Solar Cells and Their Applications, 2nd Edition (Wiley Series in Microwave and Optical Engineering)

SOLAR CELLS AND THEIR APPLICATIONS Second Edition Edited by LEWIS FRAAS LARRY PARTAIN

A JOHN WILEY & SONS, INC., PUBLICATION

SOLAR CELLS AND THEIR APPLICATIONS

WILEY SERIES IN MICROWAVE AND OPTICAL ENGINEERING KAI CHANG, Editor Texas A&M University A complete list of the titles in this series appears at the end of this volume.

SOLAR CELLS AND THEIR APPLICATIONS Second Edition Edited by LEWIS FRAAS LARRY PARTAIN

A JOHN WILEY & SONS, INC., PUBLICATION

Copyright © 2010 by John Wiley & Sons, Inc. All rights reserved Published by John Wiley & Sons, Inc., Hoboken, New Jersey Published simultaneously in Canada No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, (978) 750-8400, fax (978) 750-4470, or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748-6011, fax (201) 748-6008, or online at http://www.wiley.com/go/permission. Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Neither the publisher nor author shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages. For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at (800) 762-2974, outside the United States at (317) 572-3993 or fax (317) 572-4002. Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit our web site at www.wiley.com. Library of Congress Cataloging-in-Publication Data: Solar cells and their applications / [edited by] Lewis Fraas, Larry Partain.—2nd ed. p. cm.—(Wiley series in microwave and optical engineering) ISBN 978-0-470-44633-1 (cloth) 1. Solar cells. I. Partain, L. D. II. Fraas, Lewis M. TK2960.S652 2010 621.31'244—dc22 2010000196 Printed in Singapore 10 9 8 7 6 5 4 3 2 1

Contents Preface

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

ix

Contributors .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

xiii

INTRODUCTION TO SOLAR CELLS

.

.

.

.

.

1

.

3

.

17

Solar Cells, Single-Crystal Semiconductors, and High Efficiency . . . . . . . . . . Lewis Fraas

43

PART I

.

Chapter

1

Solar Cells: A Brief History and Introduction . Lewis Fraas and Larry Partain

Chapter

2

Solar Cell Electricity Market History, Public Policy, Projected Future, and Estimated Costs . Larry Partain and Lewis Fraas

Chapter

Chapter

3

.

.

.

67

PART II TERRESTRIAL SOLAR CELL ELECTRICITY

.

.

111

Chapter

5

Crystalline Silicon Solar Cells and Modules Leonid Rubin

.

.

113

Chapter

6

Thin-Film Solar Cells and Modules . Robert Birkmire

Chapter

7

Terrestrial Module Fabrication and Assembly Technologies . . . . Christopher Bunner

Chapter

4

8

Solar Cell Device Physics . Larry Partain

Chinese Solar Cell Status. Wang Sicheng

.

.

.

.

.

.

.

.

.

.

.

137

.

.

.

.

159

.

.

.

.

171 v

vi

Chapter

CONTENTS

Tracking the Sun for More Kilowatt Hour and Lower-Cost Solar Electricity . . . . . . Ron Corio, Michael Reed, and Lewis Fraas

.

207

Solar Cell Systems: Definition, Performance, and Reliability. . . . . . . . . . Jason Strauch, Larry Moore, and Elmer Collins

.

219

Levelized Cost of Energy for Utility-Scale Photovoltaics . . . . . . . . . Matthew Campbell

.

.

251

.

.

271

Chapter 12

Low-Concentration Crystalline Silicon Systems . Lewis Fraas

.

273

Chapter

High-Concentration, III–V Multijunction Solar Cells . . . . . . . . . . Geoffrey Kinsey

.

.

293

High-Concentration Fresnel Lens Assemblies and Systems . . . . . . . . . . Gerhard Peharz and Andreas Bett

.

313

High-Concentration Cassegrainian Solar Cell Modules and Arrays . . . . . . . . Michael Ludowise and Lewis Fraas

.

337

Concentrator Solar Cell Installations at the University of Nevada, Las Vegas . . . . Suresh Sadineni and Robert Boehm

.

.

361

Concentrator Photovoltaic Field Installations . Francisca Rubio, María Martínez, and Pedro Banda

.

377

Chapter

Chapter

9

10

11

PART III TERRESTRIAL CONCENTRATOR SOLAR CELL SYSTEMS . . . . . . . . .

Chapter

Chapter

Chapter

Chapter

PART IV Chapter

13

14

15

16

17

SOLAR CELLS IN SPACE . 18

.

.

.

.

.

.

.

395

Space Solar Cells and Applications . Sheila Bailey and Ryne Raffaelle

.

.

.

.

397

CONTENTS

vii

PART V

OTHER ASPECTS AND CONSIDERATIONS .

Chapter

19

Solar Resource for Space and Terrestrial Applications . . . . Christian A. Gueymard and Daryl Myers

.

.

.

.

425

.

.

427

.

463

Chapter

20

Solar Energy Costs: The Solar Advisor Model . Paul Gilman, Nathan Blair, and Christopher Cameron

Chapter

21

Challenges of Large-Scale Solar Cell Electricity Production . . . . David Faiman

.

.

.

.

483

.

.

.

.

497

Market Overview of Flat Panel Detectors for X-Ray Imaging . . . . . . . . . Carl LaCasce, Larry Partain, and Chuck Blouir

.

499

Amorphous Silicon Transistors and Photodiodes . . . . . Robert Street

PART VI THIN FILMS AND X-RAY IMAGER TECHNOLOGIES . . . . . . Chapter

22

Chapter

23

.

.

.

.

.

.

511

Chapter

24

Amorphous Silicon Digital X-Ray Imaging . Richard Colbeth

.

.

533

Chapter

25

Photoconductor Digital X-Ray Imaging. George Zentai

.

.

.

559

PART VII Chapter

Index .

SUMMARY 26

.

.

.

.

.

.

.

.

.

581

Summary, Conclusions, and Recommendations. . . . Lewis Fraas and Larry Partain

.

.

.

.

.

.

583

.

.

.

.

.

.

613

.

.

.

.

.

.

.

.

.

.

.

.

.

.

Preface This Second Edition is intended to be a comprehensive survey, review, and analysis of all the major factors related to the continuing technical development of solar cell electricity and its market development into a major worldwide source of electric power in response to powerful political and economic influences. It is divided into six major sections plus a Summary section including conclusions and recommendations. In contrast to the First Edition, Part I contains three initial chapters written so that nonspecialists and the more general readers and investors and policy makers can follow their contents without the need for specialized training or understanding. The goal is to allow a broad spectrum of readers to at least comprehend the market history, the influence of public policy, the likely costs of solar cell-generated electricity, and the special role that near-perfect, single-crystal semiconductor fabrication materials can have on overall performance. Chapter 4 in this part, on Solar Cell Device Physics (like the First Edition), is again aimed at advanced undergraduate and graduate college courses and other technical professionals involved in teaching, research, and commercial development. It not only covers the traditional abrupt p/n junction configuration of the First Edition but also expands into the very non-abrupt p-i-n geometries that characterize a whole new class of high-performance solar cells including interdigitated back-contact cells, point-contact cells, and heterojunction-with-intrisinsic-thin-layer (HIT) cells. It further addresses the special resistive restrictions that can limit p-i-n-type device performance as well as proposed paths to performance levels well beyond 50% efficiency levels. However, to maintain a reasonable length, this physics chapter does use the First Edition as a reference. Part II addresses the current state of terrestrial solar cell electricity technology and development programs. This includes the dominant crystalline silicon abrupt p/n junction devices and their large-scale fabrication and the emerging thin-film amorphous and polycrystalline semiconductor cells and modules. The amazing recent growth of the Chinese terrestrial solar cell program is presented in some detail. The potential advantages of tracking the sun are explored along with a detailed description of 3 years of field experience with fixed-axis crystalline silicon modules of 12% efficiency (under standard test conditions) in the Arizona desert. The emerging utility-scale installations are summarized along with their important cost-determining characteristics. Part III attempts to present a comprehensive overview of the terrestrial concentrator approach to solar cell electricity production and its special advantages ix

x

PREFACE

and challenges. This includes both low and high sunlight concentration levels with various system approaches as well as early results of small field tests at the University of Nevada and of substantial utility-scale field tests of multiple and varied concentrator systems in Spain. Part IV takes a broad look at space systems and all of the unique approaches, needs, accomplishments, plans, and future needs for space. Part V contains a chapter giving precise descriptions of the solar resource both terrestrially and in space. It also contains a chapter describing a sophisticated and detailed cost and performance model from the National Renewable Energy Laboratory (NREL). This Solar Advisor Model is reviewed and summarized. Finally, the special challenges of large-scale solar electricity production are explored. Part VI is a special four-chapter addition of the Second Edition that discusses how thin-film solar cells can be transformed into X-ray imaging devices when devices are reduced to submillimeter sizes and are aligned in columns and rows that are covered by a scintillator film that converts X-ray photons into visible light photons. If these are then attached to an array of the thin-film transistor switches, a flat-plate X-ray imager is produced. The market analysis of this whole X-ray imager field shows that its current market size of $2 billion per year should continuously evolve into a $15 billion per year wholesale market over the next 10 years or so as these devices continually improve in performance and drop in price. The final chapter summarizes the amazing growth of this solar cell electricity technology and market over the 15 years since the publication of the First Edition. It provides recommendations for how major countries and unions can play major roles from both technology and public policy perspectives and how continuing cost reduction and improved performance demands should be met under both near- and medium-term time frames. In summary, this book describes today’s baseline planar solar cell power systems as well as innovations in high-efficiency solar cells and concentrated sunlight systems that have occurred in the last 15 years, which now promise lower cost electricity competitive with other mainstream electric power sources. In addition to describing these technical breakthroughs in clear and simple terms, this book also describes the path from research breakthrough to high-volume production, emphasizing the cooperation required between government and private enterprise. Given this cooperation, solar cells can be a major contributor to the electric power production mix within the next 10 years. This book has been written for a large audience, not just a technical audience. It is hoped that any educated reader will find this book interesting, especially any reader who seeks to understand how the world’s energy supply problems can be increasingly addressed by exploiting direct solar energy resources available within a country’s borders. It further describes how most countries can start moving away from increasingly intense competition for decreasing depletable energy supplies and how they can continue moving toward a long-term, sustainable solution with inherently positive attributes.

PREFACE

xi

The thesis of this book is that solar energy can be cost competitive with other forms of electric power production and that the technical innovations required for this have already been made. Incentives for investment are needed to bring these innovations into high-volume production. It is hoped that this book will help educate the public, possible investors, as well as policy makers worldwide about the potential for a bright sunny energy future. Lewis Fraas Issaquah, WA Larry Partain Mountain View, CA

Contributors Sheila Bailey, NASA Glenn Research Center at Lewis Field, Space Environments and Experiments Branch, Cleveland, OH; email: [email protected] Pedro Banda, Instituto de Sistemas Fotovoltaicos de Concentración (ISFOC) S.A., Puertollano (Ciudad Real), Spain; email: [email protected] Andreas Bett, Fraunhofer Institut für Solare Energiesysteme (ISE), Freiburg, Germany; email: [email protected] Robert Birkmire, Institute of Energy Conversion, University of Delaware, Newark, DE; email: [email protected] Nathan Blair, National Renewable Energy Laboratory, Golden, CO; email: Nate_ [email protected] Chuck Blouir, Varian Medical Systems, Cleveland, OH; email: chuck.blouir@ varian.com Robert Boehm, Center for Energy Research, Department of Mechanical Engineering, University of Nevada, Las Vegas, NV; email: rboehm@unlv. nevada.edu Christopher Bunner, Spire Corporation, Bedford, MA; email: cbunner@spirecorp. com Christopher Cameron, Sandia National Laboratories, Albuquerque, NM; email: [email protected] Matthew Campbell, Utility Power Plant Products, SunPower Corporation, Richmond, CA; email: [email protected] Richard Colbeth, Varian Medical Systems, Mountain View, CA; email: richard. [email protected] Elmer Collins, Sandia National Laboratories, Microsystems Science & Technology, Albuquerque, NM; email: [email protected] Ron Corio, Array Technologies, Inc., Albuquerque, NM; email: rcorio@wattsun. com David Faiman, Department of Solar Energy and Environmental Physics, Jacob Blaustein Institutes for Desert Research, Ben-Gurion University of the Negev, Israel; email: [email protected] Lewis Fraas, President, JX Crystals Inc., Issaquah, WA; email: lfraas@jxcrystals. com Paul Gilman, Consultant, Chicago, IL; email: [email protected] Christian A. Gueymard, Solar Consulting Services, Colebrook, NH; email: [email protected] xiii

xiv

CONTRIBUTORS

Geoffrey Kinsey, Amonix, Inc., Seal Beach, CA; email: [email protected] Carl LaCasce, Varian Medical Systems, Salt Lake City, UT; email: carl.lacasce@ varian.com Michael Ludowise, SolFocus, Inc., Mountain View, CA; email: mike_ludowise@ solfocus.com María Martínez, Instituto de Sistemas Fotovoltaicos de Concentración (ISFOC) S.A., Puertollano (Ciudad Real), Spain; email: [email protected] Larry Moore, Sandia National Laboratories, Microsystems Science & Technology, Albuquerque, NM Daryl Myers, National Renewable Energy Laboratory, Golden, CO; email: Daryl. [email protected] Larry Partain, Varian Medical Systems, Mountain View, CA; email: larry. [email protected] Gerhard Peharz, Fraunhofer Institut für Solare Energiesysteme (ISE), Freiburg, Germany; email: [email protected] Ryne Raffaelle, U.S. Department of Energy, National Center for Photovoltaics, National Renewable Energy Laboratory, Golden, CO; email: Ryne.Raffaelle@ nrel.gov Michael Reed, Array Technologies, Inc., Albuquerque, NM; email: mreed@ arraytechinc.com Leonid Rubin, Day4 Energy Inc., Burnaby, BC, Canada; email: lrubin@day4 energy.com Francisca Rubio, Instituto de Sistemas Fotovoltaicos de Concentración (ISFOC) S.A., Puertollano (Ciudad Real), Spain; email: [email protected] Suresh Sadineni, Center for Energy Research, Department of Mechanical Engineering, University of Nevada, Las Vegas, Las Vegas, NV; email: [email protected] Wang Sicheng, Energy Research Institute, National Development and Reform Commision, Beijing, China; email: [email protected] Jason Strauch, Sandia National Laboratories, Integrated Microdevice Systems, Microsystems Science & Technology, Albuquerque, NM; email: jestrau@ sandia.gov Robert Street, Palo Alto Research Center, Palo Alto, CA; email: [email protected] George Zentai, Ginzton Technology Center, Varian Medical Systems, Mountain View, CA; email: [email protected]

PART I INTRODUCTION TO SOLAR CELLS

1 SOLAR CELLS: A BRIEF HISTORY AND INTRODUCTION LEWIS FRAAS1 AND LARRY PARTAIN2 1 JX Crystals Inc., 2Varian Medical Systems

1.1

BRIEF HISTORY

The history of the solar cell is really quite interesting [1]. In 1839, Edmond Becquerel found that two different brass plates immersed in a liquid produced a continuous current when illuminated with sunlight. We now believe that he had made a coppercuprous oxide thin-film solar cell. Later in the 1870s, Willoughby Smith, W. G. Adams, and R. E. Day discovered a PV effect in selenium. A few years later, an American named C. E. Fritts placed a sheet of amorphous selenium on a metal backing and covered the selenium with a transparent gold leaf film. He reported that this selenium array produced a current “that is continuous, constant, and of considerable force—with exposure to sunlight.” At the time, there was no quantum theory and there was considerable skepticism about his claim of converting sunlight into electricity. So he sent a sample to Werner Siemens in Germany, who was one of the most respected experts in electricity at the time. Siemens’s observation verified Fritts’s claims. However, the conversion efficiencies of both the thin-film cuprous oxide and the amorphous selenium solar cells were less than 1%. Around 75 years passed while quantum mechanics was discovered, the importance of single-crystal semiconductors was recognized, and p/n junction behavior was explained (see Chapter 3). By 1954, Chapin et al. [2] at Bell Labs had discovered, invented, and demonstrated the silicon single-crystal solar cell with 6% efficiency. Over the few following years, researchers brought the silicon solar cell efficiency up to 15%. The timing was fortunate because Sputnik was launched in 1957 and solar cells were the perfect lightweight low-maintenance remote electric power source. Today, silicon solar cells are being used to power the space station. Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

3

4

A BRIEF HISTORY AND INTRODUCTION

The solar cell industry remained small until the first Arab oil embargo in 1973. Up until that time, the solar cell industry established a firm foothold with low-level but consistent cell and array production and performance. During those first 20 years, reliability was the driver and cost was not as important. After 1973, the flat-plate silicon module was brought down to earth and modified for weather resistance. This transition also included major improvements in cell and module fabrication that brought down costs dramatically (Fig. 2.3, chapter 2). Flat-plate “champion” silicon cell efficiencies (defined in Section 2.1, Chapter 2) have improved to values as high as 25%. Production module efficiencies have improved from around 10% for early modules to as high as 19% today (SunPower Corporation). Most important, annual production quantities have grown dramatically. Worldwide production exceeded 1 GW/year in 2002 and rose to over 3.8 GW/ year by 2006 (Fig. 2.1, Chapter 2). In the late 1970s, it was discovered that good cells could be made with multicrystalline wafers as long as the crystal size is at least 20 times larger than the optical absorption length [3]. Only those carriers within an optical absorption length from the crystal boundaries are lost. This is less than 5% of the carriers. Typical production quantity multicrystalline cell efficiencies are around 14%, whereas comparable single-crystal cells have efficiencies around 15%. By 2007, modules with multicrystalline cells accounted for about 45% of sales and modules with single-crystal cells accounted for about 40% of sales. Planar silicon cell modules dominated the market in 2007 because of their early well-funded foundation years for space satellites and their huge learning curve support (Fig. 2.3, Chapter 2) from single-crystal silicon and integrated circuit technology development. While silicon-based cells still dominate the solar cell electricity market today, several other cell types have now entered the market. (Solar cells are also known as PV cells.) These newer cell types have added diversity in potential applications as well as offered alternate paths to lower-cost solar electric power. These alternate cell types include hydrogenated amorphous silicon, cadmium teluride and CIGS thin-film cells (Chapter 6), as well as concentrator cells with efficiencies as high as 41% (Chapters 13–17).

1.2

APPLICATIONS AND MARKETS

In the late 1970s and early 1980s, the traditional solar cell electricity applications [4] were at remote locations where utility power was unavailable, for example, campers and boats, temporary power needs for disaster situations, and power for remote communication station repeaters. In the late 1980s and early 1990s, solar cells began to be routinely used to provide site-specific energy for urban and suburban homes, office buildings, and a multitude of other mainstream grid-connected applications. Also, solar cell electricity systems have become very important sources of energy in the developing world. Today, for an increasing number of power needs, solar cell electricity is the cheapest and best way to generate electricity.

APPLICATIONS AND MARKETS

5

In addition to the solar power arrays on space satellites, there are now many different types of PV systems used here on Earth including 1. 2. 3. 4. 5. 6.

remote stand-alone without battery storage, remote stand-alone with battery storage, small modules for calculators and toys, residential grid connected with DC to AC inverter, commercial grid connected with inverters, and PV fields for utility power generation.

Remote solar water pumping is a nice example of stand-alone solar cell electricity where batteries are not needed. Solar water pumping is very desirable for crop irrigation, livestock watering, and clean water for remote villages. Solar water pumping systems are now installed around the world. The nice thing about this application is that underground water is pumped when the sun is shining. It can be immediately used for crop irrigation. In other areas, it can be pumped into tanks for livestock to drink. In third world countries, pumping underground water for people to drink provides cleaner water than surface water thereby limiting disease. This application is quite economical because the system is simple. Battery storage or DC to AC conversion is not necessary. Simple solar trackers are used to maximize pumping time. The electric motors driving the pumps have a threshold current that must be provided before they will operate. By tracking the sun, this power is provided from dawn to dusk, not just at around noon as would be the case without tracking. Another application where there is a good match with demand is for air conditioning in developed countries like the United States. For many remote applications, storage is needed to store electric energy for when it is needed. Examples of these applications include off-grid cabins and remote communication repeater stations. For most solar cell applications where storage is needed, secondary or storage batteries are the best alternative. Generally, batteries should be deep discharge batteries such as marine batteries or motive power batteries. Forklift trucks and golf carts use large-capacity deep discharge batteries that are designed for long life and many discharge cycles. In addition to batteries, combination systems can be used to compensate for the fact that the sun does not always shine. A solar/wind combination is particularly good since quite often, either one or the other is available. Another combination system can be a solar–thermal cell electricity system. In this case, solar cells are located on your roof for generating electricity in the summer and infrared-sensitive PV cells (also known as TPV cells) are integrated into your heating furnace to generate electricity when it is cold and dark outside and you need heat to keep warm. In a TPV cell electricity system, a ceramic element is heated in the furnace flame and its glow in the infrared is converted to electricity by infraredsensitive TPV cells [5]. Solar-powered calculators are another familiar application for solar cells. While the efficiencies of amorphous silicon solar cells are much lower than either single or multicrystalline cells, an advantage for thin-film cells is that they can be made with cell interconnections built into the process. This means that for applica-

6

A BRIEF HISTORY AND INTRODUCTION

tions like powering calculators where voltage but little current is required to run the calculator, amorphous silicon circuits are preferred to save on the cost of interconnecting multiple cells to provide voltage. Credit is due to the Japanese for recognizing this advantage and to the inventors of the amorphous silicon solar cell for making solar cells a common household item [6]. Today, more and more homes on the grid are using solar cell arrays to generate electricity to save on costs of peak electric power. The passage of the PURPA by the U.S. Congress made it possible for a small producer to install generating systems and to sell the power to the utility at a favorable price without the enormous amount of red tape usually required of a new electric power producer. Most states have now also passed net metering laws that allow the electric meter at a home to run both directions. However, at least in California, the utility charge can at most be reduced to zero and they never pay any net money to their customers who produce more electricity than they consume. This allows homeowners generating solar cell electricity to send energy to the grid if they are producing excess electricity with a credit from the utility so that they can use electric power from the grid on days without sufficient sunlight. An example of real cost savings with a solar cell electricity installation for a homeowner in San Jose, California, is shown in Figure 1.1 [7]. Current Monthly KWH 2,500 2,000

Over 300% $0.26 300% Baseline $0.24

1,500

200% Baseline $0.19

1,000

130% Baseline $0.14

500

100% Baseline $0.13

e

Ju ly Au g Se u pt st em be r O ct ob N ov er em b D ec er em be r

ay

ril

Ju n

M

Ap

Ja nu a Fe ry br ua ry M ar ch

–

Monthly KWH With PV System 2,500 2,000

Over 300% $0.26 300% Baseline $0.24

1,500

200% Baseline $0.19

1,000

130% Baseline $0.14

500

100% Baseline $0.13

e

Ju ly Au g Se u pt st em be r O ct ob N ov er em be D r ec em be r

ay

Ju n

M

Ap ril

Ja nu a Fe ry br ua ry M ar ch

–

Figure 1.1. When electric utility rates are staged, a homeowner with solar can displace electricity at the peak power rate as illustrated here. This example was originally presented by Akeena Solar on their web site in 2003 and then published in reference [7].

APPLICATIONS AND MARKETS

7

Figure 1.1 is for an actual case in 2003. Note in this figure that the utility electric rates are staged. While the homeowner pays a base rate of 13¢ per kilowatt hour that in itself is well above the national average. More importantly, the homeowner is paying twice that or 26¢ per kilowatt hour for his peak power. So his solar electric system is saving him money at the 26¢ per kilowatt hour rate. While the grid-connected solar cell electricity market started with residential customers, commercial customers are now starting to use solar arrays on their flat building rooftops. Figure 1.2 shows a photograph of two 1-kW solar cell arrays on a flat rooftop in Spokane, Washington. These arrays are mounted on carousel solar trackers (Chapter 9). People have been dreaming of the potential of solar cell electricity systems as a major electric power source for over 100 years. Now with the existence of solar power fields such as the one in China shown in Figure 1.3, this dream is becoming reality.

Figure 1.2. Two-kilowatt PV array from JX Crystals Inc on a commercial building flat rooftop.

Figure 1.3. Solar cell electricity generating field in Shanghai, China. System designed by JX Crystals Inc.

8

1.3

A BRIEF HISTORY AND INTRODUCTION

TYPES OF SOLAR CELLS AND MODULES

Unfortunately, solar cell electricity is still too expensive for widespread economical use (Section 2.4, Chapter 2). While it is hoped that traditional crystalline silicon module prices will continue to fall, there are other alternatives under development as shown in Figure 1.4. Figure 1.4 shows the three types of solar modules in use today [8]. The upper section (Figure 1.4a) of this figure shows the planar single-crystal silicon modules and fabrication procedure. This approach dominates the solar market today with over 85% of solar modules sold. As shown in Figure 1.5, retail module prices have

(a) Standard Silicon Single Crystal Module Fabrication Crystal to Ingot to Wafer to Module

(b) Concentrator Module Fabrication Smaller Single Crystal Cells With Mirrors (shown) or Lens Array

(c) Thin Film Module - Spray-on Successive Non-Crystalline Films

Figure 1.4. Alternate PV module types: (a) standard silicon single-crystal module fabrication, crystal to ingot to wafer to module; (b) concentrator module fabrication, smaller single-crystal cells with mirrors (shown) or lens array; and (c) thin-film module, spray-on successive noncrystalline films.

TYPES OF SOLAR CELLS AND MODULES

9

8

Module cost (US$/W)

7 6 5 4 3

Small Medium Large

2 1 0 1980

1985

1990

1995

2000

2005

2010

2015

2020

Year

Figure 1.5. Solar module prices for small, medium, and large volumes from 1985 through 2009. All values in then current dollars without inflation adjustments (from Photovoltaics World, September 2009).

been falling dramatically recently. Wholesale module prices are substantially lower than retail prices. The silicon cell cost accounts for about 75% of the module cost with the cost of the glass, frame, junction box, and labor accounting for the remaining approximately 25%. The lower section of Figure 1.4c shows a thin-film module. This concept is attractive because thin films require up to 100× less semiconductor material and offer a promise of lower costs per watt. Since single-crystal material is expensive, why not replace it with inexpensive thin films? The challenge is accommodating their lack of crystallinity. The latter degrades conversion efficiency, which, if too severe, limits their abilities to compete economically in the marketplace (Figs. 2.8 and 2.9 and accompanying text, Chapter 2). An appeal of multicrystalline silicon solar cells is that they offer lower manufacturing costs while still maintaining a conversion efficiency at least two-thirds that of the single-crystal ones [9] of similar Jet Propulsion Laboratory-like configurations (see Chapter 2). However, there are other useful thin-film applications, particularly for amorphous silicon, where its unique properties offer particular advantages and where high quantum efficiency but not high light conversion efficiency is a dominant factor. An example of this is use of amorphous silicon cells in medical imaging (Chapters 22–25) as shown in Figure 1.6. Here, the complete absence of crystallinity in amorphous silicon provides strong radiation damage resistance, and its higher bandgap (than crystalline silicon) gives lower dark currents. These are two strong advantages in the field of flat-plate, digital X-ray imagers that have almost totally replaced analog X-ray film. Recently, amorphous silicon imagers have also begun to displace many of the vacuum tube-based image intensifiers traditionally used in X-ray fluoroscopy. Both X-ray film and intensifier fluoroscopy replacements typically use a thin scin-

10

A BRIEF HISTORY AND INTRODUCTION Kilovoltage X-Ray Source Megavoltage X-ray Source

a-Si Kilovoltage Flat Plate X-ray Imager a-Si Megavoltage Flat Plate X-ray Imager (retracted)

Figure 1.6. Medical imaging system using amorphous silicon solar cell modules.

tillator film to convert the incident X-ray photons into visible light that the underlying amorphous silicon cells efficiently convert into electronic signals that are readily digitized. Frequently, the amorphous silicon solar cells (or pixels) measure a few hundred microns on a side, and millions of them form the rows and columns of a single X-ray imager plate. Such plates provide the digitized X-ray images at up to 30 frames/s and higher. The difficulty with module approaches (a) and (c) in Figure 1.4 is that one tries to obtain both low cost and high efficiency with the same element. In the approach shown in Figure 1.4b, one separates the two requirements of low cost and high performance into two separate elements. The single-crystal cells are the high-efficiency converters used sparingly, while mirrors or lenses are used to concentrate the sunlight onto the cells. The aluminum mirrors (or alternately glass or plastic lenses) are relativley inexpensive. For the case shown in Figure 1.4b, the cell cost is halved. The aluminum mirrors cost at least 10 times less than the singlecrystal cells. In this approach, the sunlight is concentrated onto the expensive high-efficiency single-crystal cells diluting their cost. This approach is now termed CPV. In Figure 1.4b, the sunlight intensity on the cell is doubled; that is, the concentration ratio is 2. Chapter 12 describes a configuation similar to the mirror configuration in Figure 1.4b with a concentration ratio of 3. Various concentration ratios are possible up to as high as 1000. A negative for this approach is that the modules must be aimed at the sun using solar trackers. Trackers by themselves are not a negative as the additional kilowatt per hour per installed kilowatt pays for the trackers. However, when high-concentration optical elements are used, only the direct sunlight is collected. This limits CPV to very sunny locations. However, in any case, solar cell electricity in general will be most economical first in very sunny locations such as the Southwestern United States.

ARGUMENTS FOR SOLAR CELL ELECTRIC POWER

1.4

11

ARGUMENTS FOR SOLAR CELL ELECTRIC POWER

While solar cell electricity is still expensive today, there are three strong arguments for national programs to accelerate its transition into a mainstream power source. The first argument is that there is a logical path for future lower costs for solar electricity. There are three simple steps that will lead to lower cost given development and manufacturing scale-up. These steps are based on technical breakthroughs that have now been made. In step 1, given that the cost of solar electricity today (August 2009) is about 20¢ per kilowatt hour (Solarbuzz) for commercial-sized systems for fixed flat-plate systems in the sunny Southwestern United States, by implementing solar trackers where the modules continuously point at the sun, one can gain 1.3 times more kilowatt hour per installed kilowatt, reducing the cost of solar electricity to about 16¢ per kilowatt hour. This is already being done as evidenced in Figures 1.2 and 1.3 [10]. Step 2 is then to decrease the module cost while maintaining its performance by using lower-cost optical elements as shown in Figure 1.4b. This CPV approach by itself can potentially reduce the system cost for solar electricity to under 10¢ per kilowatt hour (see Chapter 12) [10]. In step 3, one then increases the module efficiency in the CPV approach to well over 20%. As described in Chapter 3, this should reduce the cost of solar electricity still further. “Champion” CPV module efficiencies as high as 31% [11] have now been demonstrated including the one shown in Figure 1.7. While logic suggests these lower costs, this will depend on funding for manufacturing scale-up and government top-down commitment. Actually, there are multiple approaches for CPV ranging from LCPV systems using linear mirrors with silicon cells as shown in Figure 1.4b to HCPV systems

Table III: Performance Summary Packaged Projected Measure Measure Cells at STC with at Module STC 90% Operate at STC Optical Temp (April 28) Effic (April 28) DJ Cell 17.4 W 15.7 W 14.4 W 15.1 W Power DJ Cell 31.5% 28.4% 26.1% 27.3% Effic. IR Cell 3.64 W 3.28 W 2.6 W 3.1 W Power IR Cell 6.6% 5.9% 4.7% 5.6% Effic. Sum 21 W 19 W 17 W 18.7 W Power Sum 38.1% 34.3% 30.8% 32.9% Effic. NIP DNI = 0.92; Area = 600 cm2; Input Power = 55.2 W

Figure 1.7. Prototype CPV module with demonstrated outdoor module efficiency of 31%.

12

A BRIEF HISTORY AND INTRODUCTION

with newer semiconductor materials [12] such as the one shown in Figure 1.7. These LCPV and HCPV systems will be described in more detail in Chapters 12–17. Of course, while the above three steps can be implemented, this still requires investment and political commitment. This leads us to our next two arguments in favor of national programs to accelerate the penetration of solar cell electricity into the mainstream energy mix. The second reason relates to the fact that oil and natural gas resources are being depleted. Quoting from Kenneth Deffeyes’s [13] book titled Hubbert’s Peak: The Impending World Oil Shortage, “In 1956, the geologist M. King Hubbert predicted that U. S. oil production would peak in the early 1970s. Almost everyone inside and outside the oil industry rejected Hubbert’s analysis. The controversy raged until 1970 when the U.S. production of crude oil started to fall. Around 1995, several analysts began applying Hubbert’s method to world oil production, and most of them estimated that the peak year for world oil will be between 2004 and 2008. These analyses were reported in some of the most widely circulated sources: Nature, Science, and Scientific American” [14]. The 2008 peaking of world oil prices to record levels above $140 per barrel seems to support these predictions. The war in Iraq that began in 2003 was likely influenced, at least in part by the shortage of proven U.S. oil and natural gas reserves that could only last 3.0 and 7.5 years, respectively, should the United States have to depend only on its own reserves [15]. The consequence of this “impending world oil shortage” is that electricity prices are going to be rising probably abruptly within the next 5–10 years. This affects the economics of solar cell electricity as solar modules based on semiconductor devices will last for 25 years or longer. Today’s cost competition calculations for solar cell electricity usually assume a short-term payback and nonescalating energy prices. The third argument in favor of bringing solar cell electricity into the mainstream is the environmental and moral argument. It is desirable to avoid global warming as well as oil related war. When one thinks about conventional electric power production, one thinks about oil, natural gas, nuclear, and coal as fuel sources. Solar cell electricity is not on this list because it is currently too expensive. However, these conventional fuel sources have hidden unintentional costs. For example, nuclear fuels are coupled with nuclear waste management and nuclear weapons. Then nuclear waste and nuclear weapons are coupled with the cost of homeland security and our fear of weapons of mass destruction. There are hidden costs involved in attempting to guarantee that nuclear materials do not find their way into the hands of terrorists. Another example of hidden costs is the world’s dependence on oil from the Middle East that is linked unavoidably, particularly in the United States and in other developed countries, with terrorists from the Middle East. It can arguably be claimed that wars have now been fought in the Middle East to secure oil supplies.

ABOUT THIS BOOK

13

In contrast to the unintended costs just enumerated, consider solar energy. Solar energy is inevitable on the larger scale of time. Solar energy is really already a primary energy source through wind and hydroelectricity. Solar energy generated our coal, oil, and natural gas via photosynthesis a hundred million years ago. Solar cells are very much more efficient than plants at converting sunlight to useful energy. Finally, solar energy is benign and will benefit the whole world.

1.5

ABOUT THIS BOOK

The first edition of this book [16] was published in 1995 and can serve as a reference for this second edition. This second edition is divided into four main parts. Part I is an introduction to the current markets, cell and module types, and the physics of solar cell operation. The solar cell electricity market has grown appreciably over the last 14 years as described in Chapter 2. The basics of solar cell operation are presented for single-crystal cells and for thin-film cells in Chapters 3 and 4. Part II of this book focuses on the status of solar cell systems today. Single and multicrystalline silicon and thin-film cells and modules are described in Chapters 5 and 6. Over the last 3 years, silicon module automated manufacturing is coming online with the promising major cost reductions. The traditional and currently dominant silicon module manufacturing, now with automation, is described in Chapter 7. Also, over the last 3 years, China has made a major commitment to solar module manufacturing, and the status of solar electricity in China is described in Chapter 8. A major cost reduction for solar cell electricity comes through the use of solar trackers as described in Chapter 9. Large multi-MW solar cell field installations are then described in Chapters 10 and 11. Part III of this book then describes newer concentrated solar cell and system developments. Chapters 12–17 describe various concentrator solar cell electricity (also known as CPV) modules and system types and installations. Major developments have been taking place here over the last 3 years and that potentially could lead to major cost reductions over the next 5 years. While it remains to be seen if thin-film solar modules can produce electricity at rates competitive with other mainstream electricity generating technologies, nevertheless, amorphous silicon thin-film panels have found a place in other applications and in major markets like flat panel displays and medical imagers. The fourth part of this book describes successful applications of thin film technology as a spin-off from solar cell electricity. Chapters 22–25 then discuss the newest and rapidly growing applications of amorphous silicon thin films in X-ray imaging. The issue of the cost of solar cell electricity, solar modules, and solar systems is a very important subject addressed from various points of view in Chapters 2, 3, 10, 11, 20, and 26. Many believe that solar electricity prices will drop as a result of the recent investments in thin-film module manufacturing, and it is true that thin-film module prices have fallen. However, conversion efficiencies for these commercially available thin-film modules are still under 10%, and this means that

14

A BRIEF HISTORY AND INTRODUCTION

2.5–3.0 times more module area needs to be deployed relative to modules with over 19% conversion efficiencies. So, costs need to be compared at the system level, not just at the module level. Furthermore, the focus needs to be on the cost of the electricity produced in cents per kilowatt hour, not just the hardware cost. A summary and conclusions are presented in Chapter 26. Features that distinguish this second edition from the first edition are the much larger number of solar field installations and the major advances in the concentrator arena using very high-efficiency cells as well as the advances in other novel uses of thin-film modules. The current magnitude and momentum of the solar cell electricity market development (as summarized in Chapter 2) makes its eventual success both inevitable and unstoppable. This is due to the certainty of its future development path with its inherent, major advantages along with the worldwide spread of the scientific knowledge and the manufacturing and engineering know-how (covered in Chapters 3–18), plus the national commitments (alluded to in the Public Policy section of Chapter 2) that will turn promise into reality. World fossil fuel energy production rates will decline in the near to medium time frames, and the current lifestyles of the developed world cannot continue without appropriate replacements. Thus, it is no longer a question of whether this solar cell electricity power transition will occur but one of who will lead this process, who will reap the most benefits, and on what time scale it will occur. ABBREVIATIONS AC—alternating current CIGS—copper indium gallium deselenide CPV—concentrating photovoltaic DC—direct current HCPV—high-concentration photovoltaic LCPV—low-concentration photovoltaic MW—megawatt PURPA—Public Utilities Regulatory Power Act PV—photovoltaic TPV—thermophotovoltaic

REFERENCES [1] [2] [3] [4]

J. Perlin. From Space to Earth, the Story of Solar Electricity. Ann Arbor, MI, AATEC Publications (1999). D. M. Chapin, C. S. Fuller, and G. L. Pearson. A new silicon p-n junction photocell for converting solar radiation into electrical power. J. Appl. Phys. 25, 676 (1954). H. C. Card, and E. S. Yang. IEEE-TED 24, 397 (1977). R. J. Komp. Practical Photovoltaics, 3rd edition, revised. Ann Arbor, MI, AATEC Publications (2001).

REFERENCES [5] [6] [7] [8] [9] [10] [11] [12]

[13] [14] [15] [16]

15

L. M. Fraas, J. E. Avery, and H. X. Huang. Thermophotovoltaic furnace-generator for the home using low bandgap GaSb cells. Semicond. Sci. Technol. 18, S247 (2003). D. E. Carlson and C. R. Wronski. Appl. Phys. Lett. 28, 671 (1976). L. M. Fraas. Akeena Solar example cited in Chapter 2, p. 15, in Path to Affordable Solar Electric Power and the 35% Efficient Solar Cell, Issaquah, WA, JX Crystals Inc. (2004). L. M. Fraas. Path to Affordable Solar Electric Power & The 35% Efficient Solar Cell. Available at http://www.jxcrystals.com/ (2004). M. A. Green. Solar cell efficiency tables (version 29), Prog. Photovolt. Res. Appl. 15, 15–40 (2007). L. M. Fraas, J. Avery, L. Minkin, H. X. Huang, A. Gehl, and C. Maxey. Carousel trackers with 1-Sun or 3-Sun modules for commercial building rooftops. Presented at the Solar 2008 Conference, May 6, 2008, San Diego, CA (2008). L. M. Fraas, J. Avery, H. Huang, L. Minkin, and E. Shifman. Demonstration of a 33% efficient Cassegrainian solar module. Presented at 4th World Conference on PV, May 7–12, Hawaii (2006). L. M. Fraas and R. C. Knechtli. Design of high efficiency monolithic stacked multijunction solar cells. In IEEE Photovoltaic Specialists Conference, 13th, Washington, D.C., June 5–8, 1978, Conference Record (A79-40881 17-44). New York, Institute of Electrical and Electronics Engineers, Inc., pp. 886–891 (1978). K. S. Deffeyes. Hubbert’s Peak: The Impending World Oil Shortage. Princeton, NJ, Princeton University Press (2001). C. A. Campbell and J. H. Laherrere. The end of cheap oil. Sci. Am. March, 78 (1998). www.BP.com web site has a section entitled “BP Statistical Review of World Energy 2003.” This site has country and regional proven reserves and consumption data for both oil and natural gas. L. D. Partain, ed. Solar Cells and Their Applications, 1st edition. New York, John Wiley & Sons (1995).

2 SOLAR CELL ELECTRICITY MARKET HISTORY, PUBLIC POLICY, PROJECTED FUTURE, AND ESTIMATED COSTS LARRY PARTAIN1 AND LEWIS FRAAS2 1 Varian Medical Systems, 2JX Crystals Inc.

2.1

MARKET HISTORY

The worldwide solar cell (or photovoltaic) electricity generation market has grown dramatically over the past 30 years, both in terms of gigawatts per year (Fig. 2.1) (1 GW = 1 billion watts) and in billions of U.S. dollars per year sales (Fig. 2.2) [1]. Worldwide production exceeded 1 GW/year in 2002 and rose to over 3.8 GW/ year by 2006, and worldwide sales increased from US$(2007)1.5 to 9.7 billion over this same time period. The US$(2007)390 billion sales of the world’s largest oil company, Exxon Mobile, are shown in Figure 2.2 for reference [2]. Since 1975, this progress has provided a 1000-fold increase in gigawatts per year produced and almost a 100-fold increase in billions of U.S. dollars per year in sales. From 1975 to about 1985, the gigawatt per year growth rate was about a factor of 2 every 2 years. From about 1985 to 1995, it dropped to about half that rate before returning to approximately doubling every 2 years from about 1995 through 2006. Another 1000-fold increase in gigawatts per year production would place the solar cell electricity generation capacity in the range of both the total 2006 U.S. installed electricity generation capacity of 1075 GW [3] and the installed 2005 world electricity generation capacity of 3889 GW [4] even after allowing for the “capacity factor” of terrestrial sunlight being available only about 15–22% of the time [5, 6], neglecting storage issues. Since most commercial solar cells for terrestrial use

Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

17

18

SOLAR CELL ELECTRICITY

World Solar Cell Production (gigawatts)

10000

US PURPA

1000

Net Metering

JPL FSA

100

06 US Generating 05 World Capacity Generating Capacity

10 1

Rate To Double Every 2 Years

Cumulative

0.1 Annual

0.01 0.001 1970

1980

1990

2000

2010

2020

2030

2040

Year

Figure 2.1. The world solar cell production levels, on both an annual and cumulative basis, since 1975, with comparisons to the U.S. and world total electricity generating capacities in 2006 and 2005, respectively, and with support and public policy program time intervals for the JPL FSA, the U.S. PURPA, and the net metering programs. 1000 Exxon

World Solar Cell Market Size (billions 2007 US$/yr)

US PURPA

100

Net Metering

JPL FSA

Rate To Double Every 2 Years

10

1

0.1 1970

1980

1990

2000 2010 Year

2020

2030

2040

Figure 2.2. The world annual solar cell market size since 1975 with a comparison to the annual sales of the world’s largest oil company, Exxon Mobile, in 2007.

are designed for 20 years or greater lifetimes, the actual solar cell generation capacity lies between the annual and cumulative curves of Figure 2.1. The reason for the difference in the gigawatt and dollar growth rates is the constantly falling cost of terrestrial solar cells over the past 30 years as plotted in Figure 2.3. The 1975 technology was essentially slightly modified space satellite technology that cost about a hundred U.S. dollars per watt in 2007. By 2006, this had steadily dropped 25-fold to about four 2007 US$/watt. Like many high-technology and electronic products, this cost reduction roughly followed a learning

MARKET HISTORY

19

Solar Cell Module Cost (2007 US$/watt)

100 2X Cost Drop For Every 10X Increase In Cumulative Production Growth

10

8.6 gigawatts by 2006 1 0.001

0.01

0.1

1

10

100

1000

Cumulative Solar Cell Production (gigawatts)

Solar Cell Production (megawatts/yr)

Figure 2.3. The decrease in solar cell module costs since 1975 as a function of the cumulative solar cell production power through 2007. The dotted line indicates a learning curve decrease in cost by a factor of 2 with every 10× increase in cumulative production. 10000

Japan Europe China

1000 Total

Taiwan

100

Others

US

India

10

1 1994

1996

1998

2000

2002

2004

2006

2008

Year

Figure 2.4. The solar cell annual production rate from 1995 through 2006, broken down by the United States, Japan, Europe, China, Taiwan, India, and all others.

curve of dropping about a factor of 2 for every 10-fold increase in cumulative market production [7]. However, this price reduction has pretty much stalled during the last 10-fold increase in cumulative market. Many analysts attribute this cost plateau to a shortage of feedstock silicon required to manufacture the vast majority of current terrestrial solar cells [8]. The United States led in the production of solar cells from 1975 until the late 1990s. As shown in Figure 2.4, the Japanese production rate exceeded that of the United States in 1999, followed by Europe in 2003 and by China in 2006. In 2006,

20

SOLAR CELL ELECTRICITY

the United States’ fourth place position was essentially shared with Taiwan, whose rate of growth was higher. As shown in Figure 2.5 the United States held the third position in the yearly installation of solar cells in 2007, barely ahead of the rest of Europe (i.e., all but Germany) and with the rest of Asia (all but Japan) quite close behind. By 2004, Germany’s installation rate surpassed that of Japan. There are three major options to try and resume the module cost reduction process plotted in Figure 2.3. The first is to fundamentally improve crystalline cell technology and performance well beyond that of the original 1970s “modified space cell” configuration. Two potential candidates for this are the commercial SunPower interdigitated back contact cell and the SANYO HIT cell (see Chapters 3 and 4) both with prototype modules of independently confirmed “champion” efficiencies over 20% [9]. The second is thin film technologies that require much less (typically 100X) material than crystalline silicon solar cells (see Chapter 6). By 2006, the thin film annual production rate had grown to within a factor of a 1000 of the mainly silicon, total solar cell production rate, led by the United States and Japan as shown in Figure 2.6. In 2007, thin films’ growth rate was greater that that of the total market for the first time (compare with Fig. 2.1). This latest growth was mainly due to thinfilm CdTe cells that have taken the thin film lead from amorphous silicon. The decreased thin film production in Japan was mainly due to reduced activity in amorphous silicon. The third major option, for continued cost per watt reduction, is concentrator solar cell systems where the sunlight is focused down to a small area so that much smaller amounts of expensive solar cell material are also required for electricity 10000 World Solar Cell Installations (megawatts per year)

Europe Feed-In Tariffs

1000

Germany 100,000 Roofs

Germany Total

Japan

US

100

Rest of Asia

10 Rest Of Europe

1 1998

2000

2002

2004

2006

2008

Year

Figure 2.5. The annual solar cell installation rate from 2000 through 2007, broken down by the United States, Germany, Japan, the rest of Europe, and the rest of Asia, with time interval comparisons to Germany’s 100,000 Roofs Program and Europe’s Feed-In Tariff Program.

MARKET HISTORY

21

World Thin Film Solar Cell Production (gigawatts/yr)

10000 1000 100 10 Rate To Double Every 2 Years

1 0.1

Europe Japan

Total US

0.01

Rest Of World

0.001 1994

1996

1998

2000

2002

2004

2006

2008

Year

Figure 2.6. The annual world production of thin-film solar cells from 2003 through 2006, broken down by the United States, Europe, Japan, and the rest of world, and with the world total extended through 2007. The dashed line indicates a growth rate to double every 2 years.

Cumulative World Solar Cell Sales (billions 2007 US$)

1000 Rate To Double Every 2 Years

100

$36 billion by 2006

10 Germany 100,000 Roofs

1 JPL FSA

0.1 1970

1980

1990

2000

2010

2020

Year

Figure 2.7. The world cumulative solar cell sales from 1975 through 2006, with comparisons to the total expenditures of the JPL FSA Program through 1985 and Germany’s 100,000 Roofs Program through 2003.

generation (see Chapters 12–17). The latter are in their earliest, larger-scale testing phases, with most installation efforts currently centered in Spain ([10]; Banda and Rubio, Chapter 17). Figure 2.7 shows the cumulative sales totals of solar cells that have grown to US$(2007)36 billion by 2006. This is a classic learning curve-type response to a total integrated investment, which has overwhelmingly been into crystalline silicon solar cells, over this 1975–2006 time period. This total magnitude serves

22

SOLAR CELL ELECTRICITY

as a significant and growing barrier to any new technology that essentially has to become competitive at a greatly accelerated growth rate (comparatively) but with much less total integrated investment. Until recently, the thin-film growth rate had not been able to duplicate that of crystalline silicon (Figs. 2.1 and 2.6). Compounding any new solar cell approach’s problems is crystalline silicon’s constantly improving efficiency, reliability, ruggedness, acceptance, and costs as it continues down its learning curve, with ever-growing totals of integrated sales investment. The classic response strategy for this situation is a disruptive one where the new technology enters a niche market where the dominant technology does not participate and proceeds independently down its own learning curve largely unchallenged until its new concepts suddenly become competitive [11]. A potential example of the latter is the very high-efficiency, but also very high-cost, triplejunction space solar cells, based on III–V compounds (e.g., GaAs) that now dominate the space solar cell market (see Chapters 13–17).

2.2

PUBLIC POLICY

Over the past 30 years, there have been multiple significant public policy decisions, which have directly driven and strongly influenced the rate of development and the geographical location of major segments in the world total solar cell electricity marketplace. Here, five primary ones are highlighted that can be readily correlated with major market responses evident in the historical statistical market data. The earliest is the JPL FSA from 1975 through 1985 [12]. It corresponds to the initial rapid rate of market growth as indicated in Figure 2.1. It transformed delicate and expensive space silicon cells into rugged, reliable, and affordable terrestrial ones through five increasingly large (hundreds of kilowatts) and rigorously competitive block purchases of silicon cell modules whose properties were monitored and meticulously tested and reported [13]. The last two of these blocks produced the two early peaks in market growth shown on the left side of Figure 2.2. Its US$(1985)235 million (US$[2007]403 million) [14] total program size was equivalent to about 30% of the total integrated total market sales between 1975 and 1983 as is evident from Figure 2.7. Both its magnitude and sharp focus (on transforming fragile, expensive space silicon cell technology into robust and affordable terrestrial products) made it the primary driver of this early market development period. At each stage, these modules had to pass increasing stiff environmental reliability, stability, and performance standards. By the end, the project had met all of its major goals except for the price of the modules. The emerging results of singlecrystal, sawed silicon wafers, screen printed with a top metal grid contact, sandwiched between a plastic back plane and a top glass plate with polyvinyl butyral adhesive or EVA, set the standard for the next 20 years of solar cell market leadership by the United States. Following World War II, the U.S. economy was the world’s largest and it could best afford such a market generating and leading investment into this new breakthrough renewable energy technology. According to Figure 2.7, the growth

PUBLIC POLICY

23

in the world market equaled the total US$(2007)403 million JPL investment by 1988, with geometrical larger annual revenues the following years into the 1990s when U.S. companies still led in the world marketplace. For the next 10 years (about 1985–1995), the growth rate dropped by about a factor of 2, supported but only rather weakly by the PURPA that had passed in 1978. PURPA allowed “independent” electricity producers to be paid utility wholesale “avoided costs” under long-term (15–30 years), fixed-price contracts at least early on [15]. This was strongly led by California Public Utility Commission’s interpretation and requirements that ordered utilities to offer standardized contracts, most notably Standard Offer No. 4, with fixed prices. In the 1980s, PURPA was estimated to have produced up to 12 GW in non-hydro renewable energy system installations in the United States. Over half (6.1 GW) were in California, but most (over 90%) were for wind, geothermal, and biomass. Unfortunately, PURPA was not restricted just to renewables. By the early 1990s, its support of renewables stagnated with most new contracts going for nonrenewable natural gas “cogeneration” plants (producing both electric and steam) due to changing conditions including relatively low natural gas costs. Nevertheless, a significant aspect of PURPA was its first demonstration of a “feed-in”-type tariff, where renewable energy sources received “above-market” prices for their power output under longterm, fixed-price contracts, resulting in a large market response and many installations, although the majority of these were not for solar cell electricity. The next big positive stimulus for solar cell electricity was net metering, which started in the early 1990s [15]. It became prevalent enough in the United States to restore the rapid growth of solar cell production by about 1995 (as indicated on Fig. 2.1). Net metering allows independent producers, such as rooftop systems on residences and commercial buildings, to be credited by their utility for the retail avoided cost of utility-delivered electricity to their site (as opposed to wholesale avoided costs of PURPA). Typically, the utility only credits for the siteproduced power, up to the net power delivered to the site from the utility each year. Thus, any excess site electricity production is not accrued and no actual payment goes from the utility to the customer. The benefit is that the utility customer, producing local power, can offset up to 100% of the retail value of the power delivered to his or her site each year. From Figure 2.4, it can be seen that the Japanese solar cell production exceeded that of the United States by 1999 and it has maintained this world leadership production position since that time. The cost of Japanese solar cell modules decreased by over a factor of 4 from 1994 to 2003 (from ¥1733 per watt to ¥583 per watt or from US$[2007]23.62 to 4.32 per watt [8, 14, 16]) and went from being not cost competitive to being very competitive. During this same period, the efficiency of their modules increased from 12% to the 15–17% range, while the thickness of the silicon wafers used decreased from 380 to 180 microns. The performance increase came from continuous incremental improvements in light trapping (textured surfaces, double antireflection coating, reduced contact shadow). It also came from lower losses (good-quality substrates, passivation, heterojunctions) and from lower resistance (thicker electrodes, fine contact pattern). Cost reductions came

24

SOLAR CELL ELECTRICITY

from thinner substrates, larger cell sizes, simpler and lower-temperature processes, non-vacuum technology, and from making substrates directly from molten silicon. This was all largely developed directly at large commercial Japanese companies, but in close coordination with Japanese universities and with the Japanese government agencies with a clear mandate and a government resources commitment sufficient for success. The latter program was essentially that of the NEDO as part of the New Sunshine Program under the MITI [8] and its successor organizations. Approximately 8% of Japanese corporate funding (including NEDO support) goes to basic research, and this has been enough to provide the high-performance crystalline silicon HIT cell from SANYO (see Chapters 4 and 5). Since the late 1980s, the Japanese economy has been one of the world’s largest. In 2007, it was the world’s second largest economy, exceeded only by that of the United States [17]. With its comparatively small commitment to military expenditures, Japan has been particularly free to focus heavily on its internal hightechnology industrial development. “… It was in the middle of that decade [the 1990s] that the country [Japan] realized the importance of innovation and redesigned its whole strategy” [18]. “The size of NEDO’s budget illustrates this point. Japan’s corporations spend a combined 111 billion [2008] US$ on research and development every year, but more than 90% of that actually goes to incremental improvements of existing products, with 7 billion [2008] US$ going to basic research. That makes the more than 1 billion [2008] US$ that NEDO pours into its 140 research projects equal to 16% of the total corporate [Japanese] investment into basic research … Take NEDO’s long-term commitment to solar cell research. On the one hand, Japanese manufacturers like Sharp and Kyocera have won a significant share of the global market using technologies developed through NEDO projects. … With … sinister place names like ‘Devil River’, ‘Valley of Death’ and ‘Darwinian Sea’, … the illustration on the table [of the] … director general of … NEDO looks like some kind of map … It’s a vivid visual metaphor created by the U.S. National Institute of Standards and Technology to express the pitfalls that lie waiting for inventors trying to turn their research into commercial products … In Japan … [the NEDO director general] is the man charged with helping the scientific community to catapult technology and know-how past these obstacles and into the marketplace. … Since being reorganized in 2003 … [NEDO has] complete control of the national projects … [to] cancel projects that aren’t going well and direct more money into promising-looking research.” There are no details of the amounts of NEDO funding of solar cell R & D programs over the 2000–2005 time period, but when combined with Japanese corporate R & D expenditures, these combined totals are plausibly on the order of 30% of the integrated cumulative sales of solar cells of ∼10 billion (US$[2007]) by that time (see Fig. 2.7). Such NEDO-coordinated and well-supported programs in Japan, centered in large commercial operations, are reminiscent of the scale and methods of the U.S. technology development efforts termed the military–industrial complex by U.S. President Eisenhower [19]. The latter evolved dramatically from the cauldron of pressures leading up to and into World War II [20], and their implementations can be reasonably credited with keeping the U.S. military technology more than com-

PUBLIC POLICY

25

petitive with Germany’s and Japan’s early on in the World War II years and later on with the former Soviet Union’s. Its time-tested and hard-proven U.S. Department of Defense’s Science and Technology program (of US$[2007]10.8 billion) is split with 13% to basic research, 41% to applied research, and 46% to advanced technology development [21]. The latter two guide new military technology through “the Valley of Death”-type obstacles, identified by the former Advanced Technology Program of the U.S. National Institute of Standards and Technology and as highlighted and accommodated in current NEDO programs. Survivors are then ready for standard procurement or sales processes. U.S. public policy since the 1980s has largely blocked comparable programs in the United States for any technologies and industrial developments other than those focused or directed at U.S. military applications. The U.S. military–industrial complex approach was a dramatic change and improvement in the R & D process and in the rapid translation of research success into large-scale production and availability [20]. Its only major missing part, relevant to the widespread commercial introduction of solar cell-based energy, is the economic viability and market evaluations like those of the former Advanced Technology Program of the U.S. National Institute of Standards and Technology. By the mid-1990s, the EU countries and their increasingly integrated industries and markets began to build an expanding and major role for themselves in the world economy, not only for the EU in general but for Germany in particular, which was the world’s sixth largest economy by 2007 [17]. By 2007, the combined economies of the EU actually exceeded that of the United States. From 1999 until 2003, the German government instituted the 100,000 Roofs Program, targeted only at solar cell electricity, with 1.7 billion euros (US$[2007]2.2 billion) [14, 16] in favorable loans and feed-in law tariffs of up to 0.574 euros per kilowatt hour (US$[2007]∼0.75 per kilowatt hour) for up to 20 years [22]. This US$(2007)2.2 billion level investment again represents about 30% of the total world integrated sales of solar cells up to 2003 as shown in Figure 2.7. This program generated 145 GW of installed solar cell panels in Germany in 2003 to make it the leading installer since that time (see Fig. 2.5). It is estimated that the Roofs Program generated 10,000 German jobs and created 800 million euros (US$[2007]1.0 billion) in German industry revenues in 2003 alone. With this rate of revenue growth, this total loan investment was duplicated in German sales revenues within about the three following years. In comparison, the size of the California economy ranks eighth in the world [23]. Its and Germany’s (ranked sixth) have both the ability and track record of early demonstration of breakthrough technology and market trends including ones in renewable energy. Expanding on the PURPA concept, Germany introduced its version of feed-in tariffs in 1990, which was refined into its successful form by 2000 when it became a federally managed program [24, 25]. With such feed-in tariff law programs instituted in multiple European Community nations, Germany, Spain, and Denmark each provide significant portions of their countries’ electricity totaling 5%, 9%, and 20%, respectively, in 2007 from renewable sources, which were under long-term (20 year) contracts in Germany and in Spain [26]. The 2007 solar cell

26

SOLAR CELL ELECTRICITY

electricity installation totals for Germany and Spain were 47% and 23% (for a combined 70%) of the world market. These latest feed-in tariffs are justified by rationale that can include the following [24]. Long-term contracts (up to 20 years) pay solar cell electricity “tariffs” up to five times the going commercial rate (e.g., US$[2007]0.75 per kilowatt hour) [22], as opposed to representative rates (of US$[2007]0.16 per kilowatt hour typical of Italy in 2006 [27]). When 3% of a country’s electric power is produced under such agreements and the costs are spread among all the users, the 500% premium becomes a ∼15% increase in each user’s electricity bill. The integrated production that supplies these countries’ installations drives down the cost of solar cell modules, according to its learning curve (see Fig. 2.3). For a learning curve of 2X cost reduction for every 10X increase in cumulative production, the cost premium would be reduced by a factor of 2 when the cumulative production grows 10-fold, and a factor of 4 for 100-fold. After that, the cost to the customers for the next 3% of a country’s solar cell electricity installation reduces the price premium by a factor of 2–4 down to 7.5–3.75%, respectively (compared to the initial 15%). With integrated solar cell production growth doubling every 2 years (see Fig. 2.1), these 10- and 100-fold increases would occur in about 3.3 and 13.0 years, respectively. Three percent of a large user market can be quite significant particularly in the early penetration stages of solar cells into the total electricity market. For instance, 3% of the 2007 U.S. installed electricity capacity would be over 32 GW, more than eight times the total size of the world solar cell production total in 2006. The benefits that could accrue from such a program are a nation’s electricity energy supply becoming substantially less dependent on imported oil and gas, from a domestic source of sunlight that never decreases and that is delivered on-site free of charge, where the electricity conversion process itself (excluding the feedstock refining, manufacturing, installation, recycling, and disposal of solar cells) produces no pollution and generates no greenhouse gases. As a widely distributed source, it would be less subject to disruption by terrorist or military acts or by natural disasters. An orderly transition into this major new energy resource could be made steadily over many years, before world oil and gas production peak and begin to decline, which otherwise could generate sharp price fluctuations and major delivery disruptions with little or no time for orderly change. The resulting industries generated within countries could earn sales revenues equal to such subsidies within about three or four following years (if historical trends repeat) while at the same time creating large numbers of new jobs and major positive contributions to each country’s international balance of trade. Early participation in the process could well lead to world economic market leadership positions and geographic shifts of the centers of activity that historical evidence suggests can last for many years. It is the opinion of many economists [28, 29] that China is well on its way to becoming the world’s largest economy within the next few years or decades, with India likely to vie with the United States for second place perhaps 10–20 years even further out. From indications like the very high rate of expansion of its solar cell production capacity (Fig. 2.4), China may well be in the best position to invest

PUBLIC POLICY

27

the tens to hundreds of billions of 2007 U.S. dollars that historical data indicate could thrust it into the world’s leading position, at least in the area of solar cell production. And if history repeats itself, China would rapidly equal this total investment amount within 3 or 4 years of sales from its industries while also benefiting from the jobs generated and national sales revenue as well as from further increases to its international balance of payments surplus. There have been many other government-sponsored programs, some involving even larger monetary investments at their time, than the five cited above, but none with the comparably large and demonstrable effects on the world’s solar cell electricity market development history. Hence, it is not only the magnitude of a government program’s investments that provides the impetus for whole new technology applications and for relocating the geographical centers of leading activity, but it also takes wise selection of program focus and strong program execution. It is instructive to note that only the first (the JPL FSA) and the third (the Japanese NEDO development of crystalline silicon) of the five government programs covered above was directly focused on a single technology approach. Most, if not all of the other three, were essentially solar cell technology area neutral and were economic and technology stimulus programs. To a large degree, these all have mainly made use of the original JPL crystalline silicon design, with only this technology’s steady and incremental improvements, developed mostly by the manufacturers themselves (with large government subsidies at least in Japan) in response to the growing market. Over the last 30 years, the U.S. government’s (and many other countries’) renewable energy programs have had their major focus on the development of new breakthrough technology. Unfortunately, to date, the results have had little measureable impact on the solar cell electricity market and its development trajectory. However, this may be about to change; as Figure 2.3 indicates, the last 10X increase in cumulative solar cell production has produced little or no decrease in module costs, essential to the strategies covered above, for making solar cell electricity into a major alternative to depletable oil, natural gas, and coal energy supplies. A relevant public policy success example for technology implementation is the U.S. Interstate Highway System that was initiated in the Eisenhower administration beginning in 1956 [30]. It was justified as a defense-related program and was ∼90% federally funded by highway user taxes because it was considered “vitally important to national goals.” The Interstate Highway segments cost many times that of the then typical highways, and most of the highway users taxed had no direct Interstate Highway access in the early years of the program. The eventual positive impact of this system on the employment and housing opportunities of a large fraction of the U.S. population and on the U.S. economy as a whole has arguably been even greater than that of the federally funded U.S. National Aeronautics and Space Administration and its space programs. An analogue for this in solar cell electricity would be a federal tax on the users of depletable energy, namely, oil, natural gas, and coal. For example, this could pay a fraction of the initial capital costs (starting at, say, 75%) of installed solar cell electricity generating systems until approximately 30% of the then current cumulative world solar cell electricity

28

SOLAR CELL ELECTRICITY

market has been subsidized. These depletable energy users would then be directly paying for the development and deployment of renewable energy systems “vitally important to national goals” to the point that the latter becomes self-sustaining. For scale and reasonableness comparisons to the above past and potential future government programs, the following historical cost numbers may be useful. The cost of petroleum and petroleum products imported into the United States was US$327 billion in 2007, an increase of US$27 billion from the 2006 costs [31]. The U.S. 3-month balance of payment (current account) deficit increased by US$(2007)10.2 billion largely due to petroleum and petroleum product imports, which brought the total U.S. deficit to US$(2007)176 billion by the end of March 2008 [32]. Should military actions be needed at some future time to avoid major disruptions in imported gas and oil to any country, the ongoing Iraq war could provide a rough order-of-magnitude cost estimate from its 5 year costs to the United States of US$450 billion through the end of September 2007 [33]. The total final estimated cost in 1991 of the 42,795 miles of the U.S. Interstate Highway System at that point in time was $128.9 billion with $114.3 billion (or 88.7%) paid for from U.S. government funds at an average cost of $3 million per mile [34].

2.3

PROJECTED FUTURE

The historical market data above allow analysis of known quantitative results and comparatively objective evaluations of probable causes of, and contributions to, major past events and developments. Future projections require extrapolations beyond what is presently known and require a greater degree of speculation and unavoidably involve more subjective judgments and opinions. Nevertheless, important and accurate parallels can often be drawn between past history and future results, when past lessons learned are well applied, using well-accepted scientific principles and limitations combined with sound mathematical engineering and market analyses as attempted below. A continued growth rate of doubling every 2 years of production (and appropriately lower sales growth due to learning curve price reductions in Fig. 2.3) for solar cell electricity products and systems could lead to a sizeable fraction of the total U.S. and world electricity generating capacity and energy sales being provided by solar cell electricity in the decade between 2020 and 2030, as seen from Figure 2.1. One risk that could impede this progress is continuing stagnation of the Figure 2.3 solar cell module cost reduction with cumulative production. Another risk is that reasonable approaches may not be readily found to match the production of solar cell electricity to the demand schedule of electricity customers (e.g., night use of electricity when the sun is not shining). An accelerating factor for an even faster rate of growth is the rise in the real cost of energy (after correction for inflation) derived from depletable fossil fuel sources (i.e., oil, natural gas, and coal). The growing fossil fuel energy demands of the rapidly growing economies of China and India, plus the steady growth from the other leading world economies, strongly suggest that this real price increase will continue inexorably

PROJECTED FUTURE

29

for the medium term and precipitously over the longer term when the world production rates of oil and natural gas energy supplies eventually peak and begin to fall. A parallel case of a similar solid-state electronic technology growth is the approximate doubling of the calculating capacity on computer-integrated circuit chips every 2 years, which is termed Moore’s law [35]. Although all agree that this computing power growth rate cannot continue indefinitely due to fundamental and mathematical limitations, this empirical Moore’s law progress rate has essentially remained on track for more than 40 years. In comparison to solar cell electricity’s potential, 2030 is less than 25 years away. The terrestrial solar resource available is not a fundamental limitation to producing solar cell electricity each year at the level of the world generation capacity in 2005 (shown in Fig. 2.1) or even higher. The intensity of sunlight striking the outer boundary of the earth’s atmosphere is 1367 W/m2 (±4%) 24 h/day and 365 days/year as shown in the First Edition, Append B [36]. The diameter of the earth (at the equator) is 12,756 km [37] and the intensity of sunlight striking the earth’s surface (through an average atmospheric path length—i.e., AM1.5G) is 0.704 of its intensity at the atmosphere edge as given in the First Edition, Append A [36]. If one combines these with the conservative assumptions that (1) only land-based systems will be used; (2) one-third of the earth’s surface is land; (3) the availability of sunlight only matches demand 20% of the time (i.e., the capacity factor = 0.2); (4) 1% of the land is appropriate for solar cell electricity generation; and(5) the average system conversion efficiency is 10%, then the worldwide solar cell electricity resource size is 8193 GW or more than twice the 2005 world electricity generation capacity of 3889 GW shown in Figure 2.1. As the world’s dominant energy sources have evolved over time, from wood to coal and then to oil and gas, the total energy market size has increased substantially with each transition due to inherent advantages following each step. The solar input resource is large, so that the major challenge is to match solar cell electricity supply with demand. With natural hydro capacities at 121 and 923 GW for the United States and the world, respectively, in 2005 [38], this renewable storage resource can only cover demand mismatches of up to these same orders of magnitude. At Figure 2.1 extrapolated growth rates, such world levels could be reached as early as 2025. Unless battery or some other storage alternatives become practical, available and cost-effective in the meantime, solar cell electricity could be limited to supplying a fraction of the total world electric energy demand. To become the major electric energy resource, some fundamental change would likely be needed in electric energy use patterns. One of the most important and valuable U.S. exports is agricultural crops. This crop production essentially occurs only in the daytime, peaks in the summer months, with a virtual shutdown for most of the winter. Time of day and time of year utility rates could start to encourage business and personal lifestyles to take full advantage of such time availability of solar cell-generated electricity. Electric energy intense industries, such as aluminum refining and others, might well be induced to shift major fractions of their yearly

30

SOLAR CELL ELECTRICITY

production toward the daylight hours of the summer, if there were financial incentives like lower electricity rates.

2.4

COST ESTIMATES

The levelized cost of energy, L, is a detailed calculation of the effective cost of energy provided, such as 2007 U.S. dollar per kilowatt hour, considering all of the life cycle factors and expenses that contribute to such effective costs [39]. The intent of this chapter is to be as free of involved mathematical expressions and derivations as possible. Hence, Table 2.1 is used to separately summarize simple and approximate expressions that can be readily used to estimate L values and what effects them most based on parameter input values in the range of those listed in Table 2.2. This starts with baseline estimates taken from recent field experience and then proceeds to projected future cost reductions in the near- and medium-term time frames. The less mathematically inclined can thus skip the next two paragraphs and proceed directly to the results that follow. A more sophisticated and rigorous cost analysis is the complete topic of Chapter 20 (Solar Advisor Model) of this book. As listed in Table 2.1, the levelized cost of energy (L) is obtained by taking all the capital costs incurred for installing a solar cell electricity system, converting these initial capital costs into an annualized charge using fixed charge rates, and then dividing the results by the electric energy produced by such a system in a year. The result is an L value in dollar per kilowatt hour. Here, the simple approach developed by the Electric Power Research Institute (EPRI) was used [39]. The annual energy produced per unit area is just the solar energy density SEY (in kilowatt hour per square meter) that strikes the surface of the system per year, which is then adjusted for all of the relevant conversion efficiencies. The derating efficiency accounts for the mismatch and interconnect losses inevitably encountered in system assembly. A major component of fixed charge rates is the loan amortization rate, which is the per dollar cost per year of obtaining a loan at a given rate for a given period of time. This is easily obtained from loan amortization tables or from use of the Microsoft Excel PMT function. In locations where there are tax write-offs for the interest particularly on home loans, this reduces the fixed cost rate, proportionally to the tax bracket B and the fraction f of the loan payment that is interest. Additional annual financing costs are covered in the factor Δ. In a residential home loan, this Δ includes the annual taxes and insurance cost (per total loan dollar value) that adds to the amortized “interest and principal” payment (i.e., the principal, interest, taxes, and insurance [PITI] payments). The L value is conveniently expressed in terms of area-related capital costs per unit area and power-related capital cost per unit power (including the DC to AC inverter costs for the latter). One just multiplies these by the area A and the peak power P needed to generate 1 kWh in 1 year, and this gives the expression for L. For the power part, one needs to realize that module efficiency values are based on the assumption of a standard peak value of solar energy striking a system

COST ESTIMATES

31

TABLE 2.1. Levelized Cost of Energy (L) L ($ kWh ) =

TS FS + Ti Fi Annual AC energy production

Annual AC Energy Production per unit area = ηDηiηmSEY

L = (1 + r )

(Cm + Cb ) Fs A ηD ηi ηmSE Y

+ (1 + r ) Ci Fi P

AC kWh produced per year = ηDηiηmSEYA so A = 1/(ηDηiηmS) Peak output power P of inverter in kW needed for array of area A with a module efficiency ηm where the nominal peak solar input power is 1000 W/m2 = 1 kW/m2 P = ηDηiηm 1 kW/m2 A = ηDηiηm/(ηDηiηm) = 1/SEY and Cm($/m2) = Cw($/W)ηm1000(W/m2) so that (1000ηm [Cw − I ] + Cb ) FS + Ci Fi / SE Y L = (1 + r ) ηD ηi ηmSE Y where I—government incentives in $/W and η η C b FS Cw = D i {L SE Y (1 + r ) − Ci Fi } + I − 1000 1000ηm

T—total capital investment F—fixed charge rate (converts initial investment into annualized charge) Subcript S—total system except for DC to AC power inverter Subscript i—DC to AC power inverter ηi—DC to AC inverter power conversion efficiency ηm—module efficiency of converting sunlight into DC power ηD—system derating efficiency SEY—sunlight energy density striking module surface over year in kWh/m2 where FS = mS – BfSmS + Δ and Fi = mi – Bf imi + Δ r—indirect cost rate (including yearly operations and maintenance costs) Cm—module cost in $/m2 Cb—area-related balance of systems costs in $/m2 (including installation) m—loan amortization rate per year f—fraction of m that is interest B—tax bracket Δ—additional fixed charge rate A—area required to produce 1 kWh/ year in m2 Ci—DC to AC inverter cost in $/kW P—peak inverter power output rating in kW required to produce 1 kWh/year

TABLE 2.2. Parameter Values for Baseline, Near-Term, and Medium-Term Time Frames Parameter

Symbol

Value

Units

Baseline

Near Term

Medium Term

Module efficiency

ηm

0.135 (13.5%)

0.16 (16%)

0.20 (20%)

— (%)

Module cost

CW

4

2.2

1.25

$/W

Sunlight energy density/year

SEY

2435

2435

2435

kWh/m2/year (Phoenix)

Area-related BOS

Cb

307

155

150

$/m2

DC to AC inverter cost

Ci

900

690

300

$/kW

Inverter efficiency

ηi

0.95 (95%)

0.96 (96%)

0.97 (97%)

System lifetime

LTS

30

35

35

Years

Inverter lifetime

LTi

5

10

20

Years

System derating efficiency

ηD

0.95 (95%)

0.95 (95%)

0.95 (95%)

— (%)

I

2.50

1.21

0.55

$/W

0.06 (6%)

0.06 (6%)

0.06 (6%)

1/year (%/year)

Government incentive Loan rate

— (%)

Loan amortization rate system

mS

0.0726 (7.26%)

0.069 (6.90%)

0.069 (6.90%)

1/year (%/year)

Loan amortization rate inverter

mi

0.237 (23.7%)

0.136 (13.6%)

0.0872 (8.72%)

1/year (%/year)

Interest fraction of amortization system

fS

0.826 (82.6%)

0.870 (87.0%)

0.870 (87.0%)

— (%)

Interest fraction of amortization inverter

fi

0.253 (25.3%)

0.442 (44.2%)

0.688 (68.8%)

— (%)

Tax bracket

B

0.28 (28%)

0.28 (28%)

0.28 (28%)

— (%)

Additional fixed charge rate

Δ

0.0442 (4.24%)

0.0442 (4.24%)

0.0442 (4.24%)

1/year (%/year)

Fixed charge rate system

FS

0.100 (9.82%)

0.0964 (9.64%)

0.0964 (10%)

1/year (%/year)

Fixed charge rate inverter

Fi

0.265 (26.3%)

0.163 (13.7%)

0.115 (11.8%)

1/year (%/year)

Indirect cost rate

r

0.225 (22.5%)

0.225 (22.5%)

0.225 (22.5%)

— (%)

Levelized cost of energy

L

0.32

0.15

0.09

$/kWh

COST ESTIMATES

33

is 1000 W/m2. It is a standard accounting technique to multiply direct costs by an indirect rate, r, to include overhead and general and administrative costs including yearly operations and maintenance costs. For the expressions used here, it was assumed that the majority of the balance-of-systems (BOS) costs (including installation costs but excluding inverter costs) were area related. Hence, any BOS given in power-related form was converted into area-related form using the conversion relationship that is four lines from the bottom of Table 2.1. It is noted that any first-year government incentives I offered in terms of dollar per watt rebates simply subtract directly from the capital cost of the modules CW in dollar per watt. The final L expression is given three lines from the bottom of Table 2.1. The module cost CW expression on the bottom line came just from solving the L expression for CW in terms of all the other parameters. The baseline calculated costs for residential solar cell system electricity of $0.32 per kilowatt hour in Phoenix, Arizona, are given on the third column bottom line of Table 2.2. The baseline input parameter values for module efficiency, module costs, BOS costs, inverter costs, inverter efficiency, system derating efficiency, system lifetime, and inverter lifetime came from the U.S Solar Energy Technology Multiyear Program Plan 2007–2011’s [40] 2005 benchmark numbers obtained from 200 residential installations between 2000 and 2005. The loan rate and tax bracket were taken from the 2008–2012 plan [41]. The indirect cost rate and the system fixed charge rate were taken from the EPRI values [39] and its 22.5% value includes the 0.5% operations and maintenance charge used in the Solar Advisor Model. The Δ value (adding to indirect charge rate) was chosen so that the calculated FS value equaled the EPRI value of 10%. The Phoenix annual solar energy density SEY was taken from Riordan’s Chapter 20 of the First Edition [36]. As a reality check, the fixed charge rate for a 2007 home loan in Mountain View, California, was calculated as shown in Table 2.3, and its value of 9.13% (for PITI) is in this general range. The government incentive of $2.50 per watt was chosen to give the same $0.32 per kilowatt hour L value as reported in the SETP 2007–2011 plan [40]. It is in the approximate range of the most beneficial rebates offered by the state of California.

TABLE 2.3. Fixed Charge Rate F for the Purchase of a $400,000 Home with 20% Down and a $320,000 30-Year Mortgage at a 6.375% Loan Rate in Mountain View, California, in 2007 Parameter

Monthly Cost ($)

Yearly Cost ($)

Normalized Yearly Cost per Loan ($)

1995.0

23,935.0

0.0748 (7.48%)

375.0

4500.0

0.0141 (1.41%)

766.0

0.0024 (0.24%)

Principal and interest Taxes Insurance Fixed charge rate F (total)

63.85

0.0913 (9.13%)

34

SOLAR CELL ELECTRICITY

Levelized Cost of Energy L ($/kW-hr)

One of the advantages of the simple Table 2.1 expressions is the ease with which they can be used for sensitivity analyses. Figure 2.8 shows such a sensitivity plot of the levelized cost L to different module efficiency values. For the baseline case with a module efficiency of 13.5% and an L value of $0.32 per kilowatt hour (shown by a vertical line), one can see the beneficial effects of greatly increased module efficiency rapidly saturate with a sharper increase in L costs with decreased module efficiency. Similarly, Figure 2.9 shows the sensitivity of allowed module 0.5

0.4

0.3 Base Line

0.2 Near Term

0.1 Medium Term

0 0

10

20

30

40

50

60

Module Efficiency ηm (%)

Figure 2.8. The levelized cost of energy L as a function of module efficiency ηm for the baseline, near term, and the medium term obtained from the expressions of Table 2.1 using the input parameter values of Table 2.2. 6

Base Line $0.32/kW-hr

Module Cost CW ($/W)

5 4 Near Term $0.15/kW-hr

3 2

Medium Term $0.9/kW-hr

1 0 -1 -2 0

10

20

30

40

50

Module Efficiency ηm (%)

Figure 2.9. The allowed module cost CW obtained by holding all the parameters constant, except for module efficiency ηm and module cost CW, using the other parameter values of Table 2.2 for the baseline, near term, and the medium term calculated with the expression at the bottom of Table 2.1.

COST ESTIMATES

35

costs in dollar per watt (around the baseline $4 per watt value indicated by the vertical line) with module efficiency but with all the other baseline parameters held constant including L at $0.32 per kilowatt hour. Here, higher module costs can be supported with higher module efficiencies, but again this rapidly saturates. At lower module efficiencies, the module cost that can be justified drops sharply. For module efficiencies below 5%, the constant energy cost value L cannot be supported even if the module costs went to zero. This effect is due to the BOS cost described by the Cb parameter in the Cw expression at the bottom of Table 2.1. One conclusion from these sensitivity analyses is that there is no one parameter whose improvement can make the cost of solar cell electricity competitive with current electricity rates. Hence, two more cases were considered and they are the “near-term” and “medium-term” cases shown in Table 2.2 and in Figures 2.7 and 2.8 that respectively correspond to L values of $0.15 per kilowatt hour and $0.09 per kilowatt hour, module efficiency values of 16% and 20%, and module costs of $2.20 per watt and $1.25 per watt plus all the other optimistic parameter values listed in Table 2.2. These parameter values were taken from the “2011” and “2020” projections of the SETP 2007–2011 tables [40] plus similar listed adjustments to the remaining parameters. It should be noted that the SETP 2008–2012 table and plans [41] accelerated such improvements in to the “2010” and “2015” time frames. In this context, “near term” corresponds to the 2010–2011 time frame and “medium term” corresponds to the 2015–2020 time frame. Again, the corresponding curves of Figure 2.8 illustrate the positive but saturating benefits of increased module efficiencies and the sharper penalties for lower module efficiencies. Here, Figure 2.9 shows the even stronger penalties for reduced module efficiencies with the L costs held constant at increasingly competitive values. For the specific $0.09 per kilowatt hour case considered here, module efficiencies below 8% could not be justified even if their costs were zero. Further, at a module price of $2 per watt, the calculated minimum module efficiency required to be competitive is 6% for the baseline case and 14% for the near-term cases but is not achievable for the medium-term case. Similarly, at a module price of $1 per watt, the calculated minimum efficiencies required to be competitive are 5%, 6%, and 15%, respectively, for the baseline, near-term, and medium-term cases. A bottom line here is that decreased module costs are essential for achieving competitive system performance but that lower-cost modules still have to have efficiencies relatively near the more expensive modules of higher performance to provide competitive, levelized cost values. All of these emphasize the importance of resuming the Figure 2.3 module cost learning curve reductions that most recently have become quite flat. The breakdown of the three major system components and their contributions to the levelized costs L are shown in Table 2.4 for the baseline, near-term, and medium-term cases. For all three, the module costs contribute in the 70s percent range; the area-related BOS contributions are in the mid-30s to low 40s percent range, with the inverter contributions in the 30s–40s percent range. The resulting L values are only possible due to government incentives that pick up 30–40% of the total component contributions. This is the case even though the Table 2.2

36

SOLAR CELL ELECTRICITY

TABLE 2.4. Individual Parameter Contributions to the Levelized Cost of Energy L Parameter

Symbol

Contribution to L ($/kWh and %) Baseline

Near Term

Medium Term

Module cost ($/W)

CW

0.235 (73.6%)

0.117 (77.8%)

0.066 (72.7%)

Area-related BOS ($/m2)

Cb

0.134 (41.8%)

0.052 (34.2%)

0.039 (43.6%)

DC to AC inverter cost ($/kW)

CI

0.098 (30.6%)

0.046 (30.8%)

0.014 (15.6%)

Government incentive ($/W)

I

–0.147 (−46.0%)

–0.064 (−42.8%)

–0.029 (−32.0%)

Levelized cost of energy

L

0.320 (100%)

0.150 (100%)

0.090 (100%)

parameter assumptions show the actual government contributions in dollar per watt falling by more than a factor of 3 over this period. The SETP 2008–2012 Plan [41] and its Solar Advisor Model actually have the government incentive going to zero by the medium term. If these government incentives were all zero, the respective L values would be $0.467 per kilowatt hour, $0.214 per kilowatt hour, and $0.119 per kilowatt hour for the baseline, near-term, and medium-term cases (from the addition of the top three contribution rows of Table 2.4). It is such amounts that would be appropriate for “feed-in tariff ” situations. A reality check of this zero-incentive “2005” baseline value is that it is on the order of the $0.75 per kilowatt hour long-term contracts that Germany’s programs signed in the early 2000s to make their “feedin” programs financially attractive. Indeed, it is such feed-in tariff scenarios that market history indicates have been the most effective in stimulating rapid adoption and in promptly moving new geographic regions into higher world market ranking positions. The above near- and medium-term “zero-incentive” L values seem particularly reasonable for this latter type of public policy support strategy.

2.5

CONCLUSIONS

The 1000-fold growth of the world’s production of solar cell electricity in the 30 years since 1975 resulted from a growth rate that doubled output every 2 years for the first and last 10 years of this period. The middle 10 years had a growth rate reduced by about a half. Another 1000-fold growth in production would provide an electricity production level equivalent to the world’s total installed generation capacity of 3889 GW in 2005. Should a growth rate of doubling every 2 years be maintained, the latter capacity levels would be reached in the decade between 2020 and 2030.

CONCLUSIONS

37

The technical and manufacturing expertise required to support such developments is now spread throughout the world. The challenges to achieving these results are likely on the same order as those faced by the Manhattan Project that produced the first nuclear weapons or by the National Aeronautics and Space Administration’s program that placed the first men on the moon. These challenges include matching the availability of solar cell electric power with demand, continued reductions in the costs of producing electricity from solar cell systems (probably by another factor of 4 from baseline values) in ways that do not deplete basic resources needed for their construction and deployment, that does not seriously degrade the environment, and in a time frame that meets the inevitable peaking and decline of energy provided from fossil fuels. The leadership position in this technology originated in the United States but has largely shifted to the European Community and to Japan and to China over the last 10–15 years. Although the remaining challenges are daunting, the expertise to overcome them is now developing throughout the world. Changes in the rankings of dominating regions leading this technology will get increasingly expensive as the cumulative world market continues to expand. Those with the wisest and boldest strategies will likely be the major benefactors. The political entities with the highest capabilities for playing major roles in the rapidly evolving world market for solar cell electricity are those backed by the largest economies. In current order of size, these economies are those of (1) the EU, (2) the United States, (3) China, (4) Japan, and (5) India [17]. The market history-demonstrated and the best-estimated public policy approaches for promptly changing rankings among the leaders in the world market, in descending order of effectiveness, are (1) feed-in tariffs; (2) depletable energy user taxes applied to renewable energy development and deployment; (3) government-funded incentives that pay a substantial portion of the capital costs in the first year of installation; (4) long-term (30 years), low-interest rate loans often funded with tax-free bonds; (5) net metering particularly when extended to total metering so that the electricity excess delivered to a utility is paid for at least at 30–40% of the retail delivered costs; and (6) other tax incentives that include loan interest write-offs at constant interest rates over 30-year time periods. Several of these six listed approaches contain features that overlap. Of the three major options for resuming module cost reductions with market growth, key to the large-scale adoption strategy described in this chapter, is the concentrator option that has been the least explored so far and that needs the greatest assistance in making the transition from research into large-scale commercial availability and into clear feasibility demonstrations. Concentrator technology’s failure so far to transition into commercial availability is likely due in part to the “Valley of Death” syndrome described above in the Public Policy section, and that is endemic to all new technologies trying to break into large established markets. The EU is beginning to demonstrate that different countries with adjacent borders, speaking various languages, and with a spectrum of individual cultures, economic development levels, and technology competencies can join into cooperative trade, economic, consumer, and political arrangements that thrust such unions

38

SOLAR CELL ELECTRICITY

into a world leadership position, as ranked by economic metrics in general and by renewable energy implementations and deployments in particular. At its peak, the British Empire demonstrated that a dominant economic power base and trading entity could be assembled from widely spread countries whose participations were established and maintained through military “support.” A key challenge of the twenty-first century may be whether other “adjacent” and “nonadjacent” but diverse countries can cooperate through similar unions motivated primarily by economic self-interest and mutual well-being. For the latter unions to remain competitive with the large and rapidly growing economies like China’s and India’s, the involved “union” populations likely need similar sizes and mobilities and similar lack of restrictions. This latter likely includes ready access to nearly “comparable” education and training for the involved populations plus the latter’s ability to promptly relocate to job and professional opportunities unfettered by overly restrictive or artificial constraints such as those based on nationality, class, race, religion, or sex. As a nation composed essentially of diverse immigrant populations, the United States may have some advantages in the latter areas but with a remaining question of how much immigration is needed, desirable, and politically acceptable. A key to competitive solar cells and systems manufacturing in more developed countries like the EU, the United States, and Japan is automation. The fabrication equipment will likely cost about the same no matter where it is located. A Pacific Rim advantage is lower-cost labor that automation can offset to a large degree. Competitive plant construction costs including land in developed countries can likely be accommodated by appropriate tax breaks and incentives. This would offer reduced employment opportunities in manufacturing, but it would retain the product sales and factory construction revenues so that the involved country could avoid balance of payment deficits that would otherwise develop. “Champion” cells and modules are very important demonstrations of what is possible with current technology, but their performance numbers must be used with caution. When a fabricator has made and tested 10,000 devices, the very “best” by definition occurs with a yield of one in 10,000. Such a performance extreme does not necessarily provide accurate estimates of what can actually be achieved on a large scale in the near term at reasonable costs. In the past, such differences between the average achieved in large installations and in solitary “champion” devices have been as high as a factor of 4X to 5X with the differences exacerbated with attempts to provide low costs using less understood or developed materials systems (see Chapter 1 in Reference 36). When the price paid for solar cells is high using very well understood and developed materials systems, such as with space solar cells that can range in price from $300 per watt to $1800 per watt, the difference between “champion” and average cell performance can drop into the 30% or even lower range (see Chapters 4 and 23 in Reference 36). The terrestrial challenge is to simultaneously provide relatively low costs and high performance. Concentrators are an approach to using “expensive” and well-developed cell materials technologies with potentially low-cost optics to provide such a simultaneous result. However, this approach has not yet been well developed or tested

ABBREVIATIONS

39

experimentally. Early data show that the output of a two-axis tracking module of a high 17% efficiency produces 99% more kilowatt hour per square meter per year in a sunny Palo Alto, California environment in a side-by-side comparison with a near horizontal (5 ° fixed tilt) 12% efficient module (Fraas and Partain, Chapter 26). Recent tests of a 3X concentrator, two-axis tracking module of crystalline silicon interdigitated back contact cells showed that side by side, it had comparable output in kilowatt hour per square meter to a fixed axis module of similar cells in Shanghai, China, but only in measurements for a single day (L. Fraas, pers. comm.). Unfortunately, most of the current concentrator demonstration sites do not include non-concentrator modules, for direct side-by-side comparisons in terms of kilowatt hour per square meter per year outputs. Theoretical calculations of the differences are beginning to become available (Gueymard, Chapter 19). However, such calculations have not yet been carefully validated with experimental data. Their accuracy is unlikely to be trusted until such validations have been completed. Sensitivity analysis illustrates that there is no one parameter whose improvement will lead to the widespread practical implementation of solar cell electricity. However, module cost reductions (in dollar per watt) and efficiency improvements (in percent) will play dominant roles. However, major reductions in inverter costs and improvements in their lifetimes (by factors of 3X to 4X according to the Table 2.2 estimates) will also be needed. Among the other major challenges are economical means for energy storage in combination with time-of-use schemes that will be required for large-scale penetration into the traditional electricity energy market over the longer term. Most of the basic scientific understanding and principles needed to accomplish this are essentially already known. It is the engineering applications and their optimizations and scale-up that remain Herculean in scope and in magnitudes. This a situation largely shared with that of the U.S. Manhattan Project and the U.S. Man-on-the Moon Program in their early days. However, with sufficient forethought, the extremely short time frames and major diversion of resources of the Manhattan Project should not be required. While never predictable, research breakthroughs will accelerate at least some small parts of this effort. It is no longer a question whether this new energy source transition will occur. It is only a question of who will lead the process and who will reap the most benefits.

ABBREVIATIONS AC—alternating electric current CdTe—cadmium telluride DC—direct electric current EIA—U.S. Energy Information Administration EU—European Union EVA—ethylene vinyl acetate FSA—Flat-Plate Solar Array Project

40

SOLAR CELL ELECTRICITY

GaAs—gallium arsenide HIT—heterojunction with intrisic thin-layer solar cell JPL—U.S. Jet Propulsion Laboratory MITI—Ministry of International Trade and Industry (Japan) NEDO—New Energy and Industrial Technology Development Organization (Japan) PMT—name of the Microsoft Excel function that calculates mortgage payments as a function of interest rates and mortgage periods for given loan amounts PURPA—U.S. Public Utility Regulatory Policies Act R & D—research and development SETP—U.S. Solar Energy Technology Mulityear Program X—multiplier REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]

J. G. Dorn. Solar Cell Production Jumps 50 Percent in 2007, Earth Policy Institute. Available at http://www.earth-policy.org/indicatiors/solar/2007.htm. Accessed March 9, 2008 (2007). Exxon Mobile Corporation. Annual Report. Available at http://thomson.mobular. net/thomson/7/2677/3201/. Accessed September 6, 2008 (2007). Energy Information Administration, U.S. Government, Existing Generating Units in the United States. Available at http://www.eia.doe.gov/cneaf/electricity/epa/ epat2p2.html. Accessed September 6, 2008 (2007). Energy Information Administration, U.S. Government, Table H1, World Total Installed Generating Capacity. Available at http://www.eia.doe.gov/oiaf/ieo/ieoecg. html. Accessed September 6, 2008 (2008). L. Stoddord and R. Pletka. CEC Workshop on Renewable Energy. Available at http://www.energy.ca.gov/2004_policy_update/documents/2004_06_08_BLACK_ VEATCH.pdf. Accessed September 6, 2008 (2004). R. Wiser and M. Bolinger. Projecting the Impact of State Portfolio Standards on Solar Installaltions, slide 4. Available at http://www.cleanenergystates.org/library/ ca/CEC_wiser_estimates.0205.pdf. Accessed September 6, 2008 (2005). W. Wallace. Government terrestrial acceleration programs. In Solar Cells and Their Applications, L. Partain, ed., p. 495. New York, Wiley (1995). T. Tomita. Prog. Phototovolt Res. Appl. 13, 471–479 (2005). M. A. Green, K. Emery, and D. L. King. Solar cell efficiency tables (version 29). Prog. Photovolt. Res. Appl. 15, 15–40 (2007). V. Salas and E. Olias. Overview of the photovoltaic technology status and perspective in Spain. Renewable and Sustainable Energy Reviews. Elsevier, London. doi:10.1016/j.rser.2008.03.011 (2008). C. Christensen. The Innovator’s Dilemma. Boston, Harvard Business School (1997). E. Christensen. Flat plate solar array project. US Department of Energy Report, Jet Propulsion Laboratory, Pasadena, CA (1985). W. Callahan and R. McDonald. Flat-Plate Solar Array Project Final Report, JPL Publication 86-31, 5101-289, DOE/JPL 1012-125, Jet Propulsion Laboratory, Pasadena, CA (1986). Bureau of Economic Analysis, U.S. Government, Table 1.19 Implicit Price Deflators for Gross Domestic Product. Available at http://www.bea.gov/national/nlpaweb/ TableView.asp#Mld. Accessed September 8, 2008 (2008).

REFERENCES [15] [16] [17]

[18]

[19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33]

41

E. Martinot, R. Wiser, and R. Hamrin. Renewable Energy Policies and Markets in the United States. Available at http://www.resource-solutions.org/lib/librarypdfs/ IntPolicy-RE.policies.markets.US.pdf. Accessed September 13, 2008 (2005). New York Federal Reserve Bank, US$ Foreign Exchange Rates. Available at http:// www.ny.frb.org/markets/fxrates/historical/fx.cfm. Accessed September 10, 2008 (2008). Central Intelligence Agency, U.S. Government, The World Fact Book, Rank Order—GDP (purchasing power parity). Available at http://www.cia.gov/library/ publications/the-world-factbook/rankorder/2001rank.html. Accessed September 18, 2008 (2008). Fortune magazine. A catalyst for change, the Japan of the future, pp. S10-S11, Special Advertising Section, July 21. Available at http://www.timeinc.net/fortune/ services/sections/customprojects/sections/071126_Japan2.pdf. Accessed September 16, 2008 (2008). Military-Industrial Complex Speech, Public Papers of the Presidents, Dwight D. Eisenhower, 1960, pp. 1035–1040. Available at http://coursesa.matrix.msu. edu/∼hst306/documents/indust.html. Accessed September 27, 2008 (1961). D. E. Kash. Perpetual Innovation, Chapter 6. New York, Basic Books (1989). FY 2008 Budget Request for Defense S&T, FYI. AIP Bulletin of Science Policy News, February 13, 2007, No. 23, pp. 1–2 (2007). SolarBuzz. German PV Market. Available at http://www.solarbuzz.com/fastfactsgermany.htm. Accessed September 6, 2008 (2007). A brief history of the California economy. Available at http://www.dof.ca.gov/ HTML/FS_DATA/HistoryCAEconomy/index.htm. Accessed November 22, 2008 (2008). World Future Council, Feed-in Tariffs. Available at http://worldfuturecouncil.org/ fileadmin/user_upload/Maja/Feed-in_Tariffs_WFC.pdf. Accessed September 16, 2008 (2008). Wikipedia. Feed-in tariff. Available at http://en.wikipedia.org/wiki/Feed_in_Tariff. Accessed November 22, 2008 (2008). SolarBuzz. 2007 World PV Industry Report Highlights. Available at http://www. solarbuzz.com/Marketbuzz2008-Intro.htm. Accessed September 7, 2008 (2008). NEI Nuclear Notes, Italy Nuclear Update. Available at http://www.neinuclearnotes. blogspot.com/2006/02/italy-nuclear-update.html. Accessed September 15, 2008 (2006). China to become world’s largest economy by 2010, Finance Daily. Available at http://financedialy.co.uk/News/Chinatobe WorldsLargestEconomyby2010_448. html. Accessed September 27, 2008 (2008). The World’s Largest Economies in 2050, University of Phoenix, based on 2003 Goldman Sachs study. Available at http://www.everthing2.com/index.pl?node_ id=1756826. Accessed September 27, 2008 (2005). 50th Anniversary of the Interstate Highway System. Available at http://www.fhwa. dot.gov/interstate.history.htm. Accessed August 22, 2008 (2006). US oil import bill to top $400 billion this year. Available at http://www.reuters.com/ article/pressRelease/idUS236508+07-Mar-2008+BW20080307. Accessed September 7, 2008 (2008). Bureau of Economic Analysis, U.S. Government, U.S. International Transactions: First Quarter 2008. Available at http://www.bea.gov/newreleases/rels.htm. Accessed September 7, 2008 (2008). The Cost of Iraq, Afghanistan, and Other Global War. Available at http://zfacts.com/ metaPage/lib/CRS-Belasco-2006-09-Iraq-Costs-RP33110.pdf. Accessed September 7, 2008 (2008).

42 [34] [35] [36] [37] [38] [39] [40]

[41]

SOLAR CELL ELECTRICITY Interstate Cost. 50th Anniversary of the Interstate Highway System Frequently Asked Questons. Available at http://www.fhwa.dot.gov/interstate/faq.htm# question6. Accessed November 20, 2008 (2006). S. Thompson and S. Pathasarathy. Materials Today 9, 20–25 (2006). L. D. Partain, ed. Solar Cells and Their Applications. New York, Wiley (1995). Earth radius. Available at http://en.wikipedia.org/wiki/Earth-radius. Accessed November 22, 2008 (2008). Energy Information Administration, U.S. Government, Table H6, World Installed Hydroelectric and Other Renewable Generating Capacity. Available at http://www. eia.doe.gov/oiaf/ieo/ieoecg.htm. Accessed September 6, 2008 (2008). F. R. Goodman, J. C. Schaefer, and E. A. DeMeo. Terrestrial, grid connected systems. In Solar Cells and Their Applications, Chapter 16, L. D. Partain, ed. New York,Wiley (1995). US Department of Energy, Solar Energy Technologies Program, Multi Year Program Plan 2007-2011. Available at http://www.pv-era.net/doc_upload/documents/218_0 084aSolarEnergyTechnologiesProgram2007-2011_proof.pdf. Accessed October 12, 2008 (2007). US Department of Energy, Solar Energy Technologies Program, Multi Year Program Plan 2008-2012. Available at http://www1.eere.energy.gov/solar/pdfs/ solar_program_mypp_2008-2012.pdf. Accessed October 26, 2008 (2008).

3 SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS, AND HIGH EFFICIENCY LEWIS FRAAS JX Crystals Inc.

3.1

INTRODUCTION

How efficient can a solar cell be and how much power and energy can it produce and at what cost? These are the fundamental questions addressed in this chapter. In order to address these questions, it will be necessary to first describe the nature of sunlight and then the nature of semiconductors. It will be shown that singlecrystal semiconductors are necessary for high sunlight conversion efficiency and that high conversion efficiency is important for lower-cost solar electricity. In addition to silicon solar cells, a new class of semiconductors based on LED materials will also be discussed. The reader is familiar with LEDs and the fact that they come in different colors. It is noted herein that LEDs are just solar cells running in reverse. In other words, in a LED, electricity goes in and light comes out. In a solar cell, light goes in and electricity comes out. It will be shown that the LED class of semiconductors can be used to make multicolor or multijunction solar cells where the multicolor feature is critical for making solar cells with energy conversion efficiencies as high as 40%. While it is true that high-efficiency single-crystal solar cells are more expensive to make than amorphous or small grain size polycrystalline thin-film cells, fortunately, it is possible to use inexpensive mirror or lens materials to collect the sunlight and to focus the solar energy on these single-crystal converters thereby diluting their cost. This is the concentrated sunlight PV or CPV approach. The newer 40% cells are even more complex and expensive than single-crystal silicon Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

43

44

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

cells requiring higher-concentration HCPV systems, whereas the silicon singlecrystal cells can use lower concentration LCPV simpler systems. This chapter will then conclude with a discussion of the future cost potential for the various solar PV options comparing the potential future system-level cost for the conventional planar silicon cell and thin-film cell approaches with the concentrated sunlight approaches.

3.2

SUNLIGHT, RAINBOWS, AND PHOTONS

How much energy is in sunlight? The answer to this question requires a definition of some terms. The three terms are power, solar intensity, and energy. Power is in watts and solar intensity is in watts per unit area. Energy is in kilowatt hours. Homeowners pay for energy in kilowatt hours. Solar modules and arrays produce power in watts. The relevant measure of solar radiation is solar intensity. On a nice, sunny day at noon, the solar intensity is usually around 1 kW/m2. One square meter is close to 11 ft2. How does a scientist describe sunlight? The observation of rainbows proves that sunlight can be divided into different colors. Also, when a group of very fine parallel lines are scribed close to each other to make a grating, it is observed that the colors can be correlated with line spacing. This means that there is a wavelength connected to each color. So, light is an electromagnetic wave as shown in Figure 3.1, just like radio waves and microwaves with a wavelength, λ, and electric field, E. The history of the study of light is interesting. Actually, Newton thought of light as particles. However, with grating experiments in the 1800s, it was decided that light was an electromagnetic wave. Then, in about 1905, Einstein looked at the photoelectric effect and said that light comes in small energy packets called

E

Wave length

Figure 3.1. Electromagnetic wave.

SUNLIGHT, RAINBOWS, AND PHOTONS

45

quanta, which behave somewhat like particles. Einstein won the Nobel Prize for this work [1], not for his theory of relativity. Einstein noted that when the light of a certain color hits a metal, all of the electrons that escape the metal have the same peak energy and that when the light intensity is increased, the number of electrons increases, not the energy of each electron. It is like the energy in wave theory is E2, but the energy also equals nλ × eλ, where nλ is the number of photons and eλ is the energy of each photon. So, one can think of the electromagnetic wave intensity as getting smaller and smaller until one discovers that it is coming in discrete particles. It occurs in very fine steps. So now, the sun’s intensity spectrum can be divided into color or wavelength intervals as in Figure 3.2. To understand how solar cells work, first note that a photon’s energy is inversely proportional to its wavelength. This just means that shorter wavelength photons at the left in this curve have more energy than the longer wavelength photons at the right. Electric power is the product of voltage and current. In a solar cell, the power results when electrons pass through a voltage. It is the nature of semiconductors that each semiconductor material has a threshold absorption energy that then controls the voltage the electrons see. It is convenient then to talk in units called electron volts. An electron volt is the energy that an electron produces when it moves through a potential of 1 V. Photon energies and material threshold absorption energies are measured in electronvolt. For example in Figure 3.2, a red photon with an energy of 2 eV has a wavelength of about 0.6 microns, and the absorption threshold energy for silicon, called its bandgap energy, is at 1.1 eV, which equates to 1.1 microns (1 micron = 1 micrometer = 1 millionth of a meter = 1 μm). Now the efficiency limits can be seen. First, in Figure 3.2, all photons with energies less than 1.1 eV are lost; that is, all solar energy with wavelengths greater 1600

Watts/m2/micron

1400 1200 1000 800 600 400 200 0

0

0.5

1

1.5

2

2.5

Wavelength (microns)

Figure 3.2. Sun’s spectrum at AM1.5. The gray region is the usable photon energy for a silicon solar cell.

46

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

than 1.1 microns is lost because these photons are not absorbed. Next, for photons with energies large enough to be absorbed (left side of figure), each photon absorbed must create an electron that can move to generate voltage. The yield here should be one electron per photon, but the actual number is less and is called the quantum efficiency or QE. QEs of 90% occur in pure single-crystal materials. QEs in non-single-crystal materials are lower. So, single-crystal semiconductors are important for high efficiency as will be discussed in the next sections of this chapter. For now, high QEs in single-crystal materials are assumed. These moving electrons create the current. To calculate the current, the photons absorbed are simply counted. However, note that while a photon might have 2 eV of energy, it can only create an electron moving through a voltage of no more than the bandgap energy. For silicon, this is 1.1 eV. So, 2-eV photons loose at least 2.0 – 1.1 = 0.9 eV. The gray region in Figure 3.2 shows the energy that can be captured from the higher-energy photons. This now leads to a way to make higher-efficiency solar cells using multiple materials. Two stacked materials can absorb the lower-energy photons in one material generating a lower voltage and the higher-energy photons in a second material generating a higher voltage. The result is more photons better used. This concept is shown in Figure 3.3. The gray regions in Figure 3.3 show that more photon energy can now be captured. To implement this two-junction or two-color cell concept, one must choose materials wisely. The GaAs/GaSb two-color cell (2) demonstrated in 1989 uses two simple materials with bandgap energies of 1.4 eV (0.9 micron) and 0.7 eV (1.8 micron). Theoretically, the efficiency for this pair can be as high as 41%. Thirtyfive percent has been demonstrated [2].

1600

Watts/m2/micron

1400 1200 1000 800 600 400 200 0

0

0.5

1 1.5 Wavelength (microns)

2

2.5

Figure 3.3. GaAs/GaSb and solar spectrum. The GaAs absorbs higher-energy photons at the left, generating higher voltage. The GaSb absorbs lower-energy photons at the right (1 micron = 1 micrometer = 1 millionth of meter).

ELECTRONS IN ATOMS AS WAVES AND THE PERIODIC TABLE

47

For the more general reader who may not care about the chemical names, note that the top GaAs cell actually looks blue and the bottom GaSb cell actually looks red. So, one can also refer to the cells in multicolor cell stacks as blue cells glued on top of red cells. The chemical terms will be used through most of this book for the benefit of technical readers because the chemical names are more precise. More information on multijunction concentrator cells will be presented later in this chapter and in Chapter 13.

3.3 ELECTRONS IN ATOMS AS WAVES AND THE PERIODIC TABLE OF THE ELEMENTS There are several different types of solar cells made from materials ranging from single crystals to amorphous silicon. The goal here is to describe the different types of solar cells and their advantages and limitations. A fundamental description of the nature of semiconductors is presented, beginning with electrons in atoms as waves. The discussion of electrons as waves then leads to a description of semiconductors as single crystals. The theory of single-crystal semiconductors is then used to describe how diodes and solar cells work. A discussion of the effects of various defects in semiconductor materials on solar cell performance follows. The reader will see that the performances enumerated are consistent with the simple concepts presented. This chapter explains why high-efficiency cells require good single-crystal materials. In the last section, it was noted that the sun’s rays are really electromagnetic waves with varying wavelengths. Electromagnetic radiation includes radio waves, microwaves, and infrared, visible, and ultraviolet waves. When one thinks about longer wavelength radiation like radio waves, one always thinks about waves. However, for the shorter wavelengths associated with infrared and visible light, physicists start to talk about photons. A photon is like a particle or wavelet having a specific wavelength and energy. A photon is a quantum of energy or discrete packet of energy. Now, is radiation a wave or a particle? The answer is both! This is the wave–particle duality, a subject called quantum mechanics [3], a subject normally taught in graduate school physics classes along with a lot of mathematics. However, the key ideas can actually be described in simple nonmathematical terms, and these ideas are important to the understanding of solar cells. While electromagnetic radiation is normally thought of as waves, one generally thinks of electrons as particles circling an atomic nucleus just as planets circle the sun. However, an atom is really extremely small, so small that in crossing a human hair, one will pass by 200,000 atoms. Intuition based on everyday experience fails at this small size. It turns out that electrons around atomic nuclei are described by wave functions. Here is the wave–particle duality again. However, the rules that govern electrons in atoms and solids can be described in fairly simple terms. In Figure 3.4, start with the simple hydrogen atom [4, 5] with a single negatively charged electron and a single positively

48

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Nucleus

D Px, Py, Pz

4103 A 4342 A 4563 A

S

E X

6565 A

Figure 3.4. Left: potential well for an electron around the nucleus in an atom with energy level S, P, and D wave functions. Right: a spectral line sequence for hydrogen.

charged proton. The oppositely charged proton and electron attract each other and as they get closer and closer to each other, it is harder and harder to pull them apart. The electron is said to be in an energy well or a potential well as shown on the left of Figure 3.4. The question then is can the electron collapse down and sit on the proton? The answer is no. Why not? Scientists have observed the electromagnetic spectra emitted by atoms and find discrete wavelengths and energies as shown on the right in Figure 3.4. Not all energies are possible. How is this explained? Scientists hypothesize that the electron position is described by a wave function that then gives its probable position. Since one knows that the electron cannot be outside the potential well, one knows the wave functions have to be zero outside the well. Now, observe that the waves will have to have one, two, and three nodes as is shown in the wells at the left in Figure 3.4. For historical reasons, the state with one peak node is labeled S, and the states with two nodes are labeled Px, Py, and Pz. (x, y, and z are the three directions in three-dimensional space.) The next rule is that electrons can have positive and negative spins, and only one electron can occupy each state. So there will be two S states with opposite spins and two Px, two Py, and two Pz states for a total of eight state configurations possible. This wave hypothesis has proven to be very successful as it explains atomic spectra and the periodic table of the elements [6] and all of chemistry. The rule of eight including S and P orbits explains the second and third rows of the periodic table. Table 3.1 summarizes important features of the periodic table including the common commercial semiconductor materials. The D level transition metals are not shown since they are not relevant here. Compounds are formed from the elements in the periodic table. For example, table salt is a compound containing Na (sodium) and Cl (chlorine) and is written as NaCl. The two semiconductor

SEMICONDUCTORS AS CRYSTALS AND THE WAVE THEORY

49

TABLE 3.1. Periodic Table of the Elements I

II

III

IV

V

VI

VII

H Hydrogen

VIII He Helium

Li Lithium

Be Berilium

B Boron

C Carbon

N Nitrogen

O Oxygen

F Fluorine

Ne Neon

Na Sodium

Mg Magnesium

Al Aluminum

Si Silicon

P Phosphorus

S Sulfur

Cl Chlorine

Ar Argon

Ga Gallium

Ge Germanium

As Arsenic

In Indium

Sb Antimony

compounds making up the two color solar cell stack described in Figure 3.3 are gallium arsenide (GaAs) and gallium antimonide (GaSb).

3.4 SEMICONDUCTORS AS CRYSTALS AND THE WAVE THEORY Why is it important to know about electrons as waves? The answer is that waves are intrinsically periodic as are the atom locations in single crystals. It is this periodicity that makes semiconductors special. Historically, the semiconductor revolution started 80 years ago with the discovery of the importance of high-purity single crystals and the wave theory of solids. The first application of the electron wave theory was by A. Sommerfeld in 1928 [7]. As shown in Figure 3.5a, he described the motion of electrons in metals by assuming the electrons moved in a flat-bottom energy well bounded by the metal surfaces. Any wavelength would be possible, leading to a set of conduction band energy levels. The real breakthrough came in 1928 with F. Bloch [8], who then modeled a periodic single crystal with a set of periodic energy wells representing the atom positions within the larger energy well as shown in Figure 3.5b. He showed that there are then two or more energy bands separated by energy gaps. The states with wave functions centered on the atomic wells represent valence electrons, and the states with wave functions centered between the atomic wells represent conduction band electrons. A key point is that any electron can be near any atom in the crystal. A. H. Wilson [9] in 1931 then provided the elementary theory of semiconductors by noting that if the material is pure enough and if the valence band is completely full and the conduction band is completely empty, then one has a semiconductor. An intrinsic semiconductor is defined as having zero conductivity at zero temperature with increasing conductivity as the temperature increases. The increasing conductivity comes about as electrons are thermally excited into the conduction band. This is opposite from metals where the number of conduction

50

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Ec (a) Metal

Ec Ev (b) Semlconductor

Figure 3.5. Solid potential energy and energy band models for metal and semiconductor.

band electrons is fixed and collision frequency increases with temperature decreasing electrical conductivity as the temperature increases. A key point here is that there is an energy gap between the valence band and the conduction band and that this energy gap derives from the periodicity, which derives from the single crystal. This history is fascinating. However, the goal here is to explain the reasons why single crystals are important to solar cells and to probe the question of how pure and how perfect solar cell materials need to be. Most importantly, how can solar cells be integrated into systems to generate electricity on a large scale at a competitive price? Before describing semiconductors in more detail, let us return to the periodic table and contrast the semiconductors with metals and insulators to see why semiconductors are special and why they are needed to make solar cells. To preview the answer, note that in order to deliver electric power, a solar cell needs to generate both current and voltage. Generating current requires electron mobility and generating voltage requires a gap between electron energy states. Metals have electron mobility and insulators have gaps between energy states, but only semiconductors have both. The metals like sodium and magnesium are on the left in the periodic table. These atoms have only a few loosely bound electrons each, and they can be tightly packed with up to 12 nearest neighbors. Because the atoms are closely packed, the potential energy well for a metal looks like a flat-bottom well with the well extending to its surfaces. As shown in Figure 3.5a, the metal surfaces form the energy barriers confining the electrons. Because this well is so large compared to one atom, all electron wave function wavelengths and energies are possible. Electrons are then free to move around in the metal, but there are no energy gaps between energy states. Since the electrons hardly feel the metal atom core positions with the flat-bottom potential well, crystallinity is not important to metallic properties.

SEMICONDUCTORS AS CRYSTALS AND THE WAVE THEORY

51

The elements at the right of the periodic table like oxygen and chlorine have tightly bound electrons and are hungry to grab more. They readily form ionic compounds like salt (sodium chloride) and glass (silicon dioxide). The energy levels in these compounds are much like those of atoms in that the electrons only are excited between atomic energy states. There are gaps in energy, but the electrons are not mobile. Crystallinity is not very important since electrons are localized on ions. This brings us to the group IV elements like silicon. The structure of silicon in a silicon crystal is shown in Figure 3.6. Silicon has four electrons and forms four tetrahedral bonds as shown. Looking at a row of silicon atoms along the diagonal in a silicon crystal, note the alternating bonded and nonbonded spaces between silicon atoms. The energy potential well profile for this row is shown in the middle of this figure along with two wave patterns, one drawn as a solid line and one drawn as a dashed line. The peaks in the solid line wave pattern localize the electrons in the bonded regions with lower-average energy potential. Meanwhile, the peaks in the dashed line wave pattern are localized in the nonbonded regions with higher-average energy. However, both waves allow the electrons to be near any silicon pair in the crystal, implying electron mobility throughout the crystal. Because of the periodic nature of the atomic positions in a single crystal, the wave functions allowed describing the electrons in a single crystal must have a

E

Eg

c v

Eg

c v

E x Si

Si

Si

Si

Si

Si

Si

E

E x As

Ga

As

Ga

As

Ga

As

Ga

Figure 3.6. Top: tetrahedrally bonded silicon atoms in groups along cube diagonal in silicon crystal showing alternate bonded and nonbonded pairs. Middle: energy potential for top atom sequence with valence band bonding wave function as solid line and conduction band antibonding wave function as dashed line. Bottom: the potential and wave functions for a GaAs crystal.

52

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

corresponding wavelength. Thus, the two types of states with bonding and antibonding electron locations between nearest silicon pairs or farthest silicon pairs are the only states allowed. There is an energy gap between these states because no other electron wave functions are allowed. The states representing the bonding states form what is called the valence band and the states representing the antibonding states form what is called the conduction band. Figure 3.6 also shows the energy potential and wave functions for a group III–V semiconductor. The III–V semiconductors are the materials used to make LEDs. In this case, a group III element like gallium can form tetrahedral bonds with a group V element like arsenic where the result is the sharing of four electrons per atom as in silicon. Group III–V is a rich class of semiconductors. It turns out that because of the crystal periodicity, there is both energy gap and electron mobility in semiconductors. Figure 3.7 allows one to visualize this more easily. In this figure, both connected bonded regions and open channels in between can be seen. One can imagine electrons traveling in the bonded regions or separately in more energetic states in the open channels. Propagating electrons in the bonded region have energies in a valence band, and propagating electrons in the open channels have energies in a conduction band. The separation between these regions provides the energy gap. Looking at Figure 3.7, one can also imagine a large foreign atom or a crystal boundary or defect interfering with flow in the channels or a total disruption of the channels smearing the two sets of energy states into each other.

Figure 3.7. A view of a channel open for conduction electron movement in a GaAs single crystal. Small sphere: gallium atom; large sphere: arsenic atom; white cylinders: valence bonds.

JUNCTIONS AND DIODES

53

Figure 3.7 suggests intuitively that electrons will have higher mobility in single crystals than in amorphous or small crystal size thin films. This is in fact true quantitatively. Electron mobility is easily and routinely measured. The electron mobility in single-crystal silicon is typically 1500 cm2/Vs, and in single-crystal GaAs, it is 4500 cm2/Vs. However, in amorphous silicon and copper indium diselenide (CIS), two common thin-film solar cell materials, it is only 4 cm2/Vs. This is a difference by a factor of 1000 consistent with our intuitive expectations based on Figure 3.7.

3.5

JUNCTIONS AND DIODES

It has now been established that carriers are mobile, allowing current to flow in solar cells. How does one use an energy gap to create voltage? A P/N junction (P = positive, N = negative) is needed. In the above description of electron movement in semiconductors, one now needs to note that it is important to count electrons. If the semiconductor is very pure (a state called intrinsic), then all of the bonding states will be occupied by electrons and there will be no electrons to move in the conduction band. Electrons cannot move in the valence band either because there are no empty spaces to move to. Substituting a small number of phosphorus atoms for silicon atoms can rectify this problem (one in a million). Since phosphorus is from group V, it has one more electron than silicon. The resultant material is labeled N-type because the extra electrons are negatively charged. Alternately, as a complement to the N-type material, one can substitute an aluminum atom for a silicon atom leaving the bonding or valence band one electron deficient because aluminum from group III has one less electron than a silicon atom. Now instead of thinking about a million electrons in the valence band, we talk about the missing electrons in the valence band. We call this a hole. It is like watching a bubble move in water. The hole has a positive charge and we call this material P-type. Now what happens when N- and P-type materials are brought together? The result is a P/N junction diode [10–12] as shown in Figure 3.8. The band edge diagrams at the bottom of this figure describe how a diode works. When the P and N regions first come together, the electrons and holes from each side diffuse together, eliminating each other, leaving an electric field region in the junction. This happens until the valence band edge (v) in the P material almost lines up with the conduction band edge (c) in the N material as shown on the left in this figure. At this point, the free electrons and holes on both sides of the junction have the same energy as shown by the dashed horizontal line. This is the zero voltage band diagram (Fig. 3.8A). Now notice that there is an energy hill for electrons to climb in order to move from the N to the P side of the junction. An applied voltage can either decrease this hill or energy barrier for forward bias (Fig. 3.8B) or increase it in reverse bias (Fig. 3.8C). If the hill is made small enough by a forward voltage about equal to two-thirds (67%) of the bandgap energy, Eg, then current starts to

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Current (Amps)

54

Current PN y

z

x

c Eg v E

c

P v

x (a) Zero voltage

c Eg v

(A)

(C)

Voltage (V)

Forward current

c Eg v

Metal Contacts

N

(B)

P

Electron c Applied N voltage v

P

N

c v

(b) Forward voltage

(c) Reverse voltage

Figure 3.8. Upper left: P/N junction diode; upper right: current versus voltage for P/N diode; lower left (A): conduction band minimum and valence band maximum positions through P/N junction at zero applied voltage. Lower middle (B): forward voltage band diagram—reduced barrier for high current flow. Lower right (C): reverse voltage—barrier blocks current flow.

flow. This corresponds to the knee in the diode current versus voltage curve shown at the top right in this figure. In reverse bias, no current flows because the barrier just gets bigger. Thus, a diode is a rectifier allowing current flow in only one direction.

3.6

SOLAR CELL BAND DIAGRAMS AND POWER CURVES

Referring now to Figure 3.9, a solar cell is just a large P/N junction diode with a metal grid on its front side facing the sun. A solar cell converts the energy in sunrays to electric power. Now we shall refer to the sunrays as photons. In Figure 3.9, the now familiar band edge diagrams are shown at the bottom. These band edge diagrams show how a solar cell works. First, a photon is absorbed exciting an electron from the ground state or valence band in the P material to an excited conduction band state. It is mobile in the conduction band and if it lives long enough in this excited state, it can diffuse to the junction and fall down the potential barrier. Another way of thinking about this potential barrier is simply that it represents an electric field region created by the initial separation of electron and holes when the junction was formed. Anyway, when an electron enters a field region, it gains electrical energy. This can be converted to a voltage and current to do work.

HIGH-EFFICIENCY AND MULTIJUNCTION SOLAR CELLS

Sun ray Current (Amps)

Voltage (V)

P N

x

Eg

c c P

x

Max power point

Light generated current Forward dark current

c

E

Voc

Isc

Light generated current

v

55

Eg

N

Voc

v v

Short circuit current (Isc) at zero voltage

P Open circuit voltage (Voc) at zero current

Figure 3.9. Upper left: P/N junction solar cell with metal grid on top. Lower left: photon absorption excites electron into conduction band. Electron then falls through junction potential. Upper and lower right: current versus voltage curve for solar cell is diode I versus V curve moved down by light-generated current.

3.7

HIGH-EFFICIENCY AND MULTIJUNCTION SOLAR CELLS

How efficient can a solar cell be and how do we achieve these high efficiencies? Theoretically, a solar cell efficiency of 70% is possible. However, no one believes that, in practice, this can be achieved. Still, a 35% efficient solar cell has been demonstrated and 45% is probably an achievable target. What needs to be done to achieve high efficiencies is a more interesting question. In fundamental terms, three things need to be done. First, for each photon absorbed, the excited state carrier generated needs to last long enough to be collected at the junction. Second, while the sun’s spectrum contains photons of different energies, the energy available in each photon must be used as wisely as possible. And third, the voltage a cell generates should be as close as possible to the bandgap energy. We will discuss each of these requirements in succession in the following paragraphs. The first requirement of one electron collected for every photon absorbed implies single-crystal material and high-purity material. The measure of electrons collected per photon absorbed is called quantum efficiency. Figure 3.10 provides a semiquantitative answer to the semiconductor purity question. To understand Figure 3.10, let us go back to the crystal channels shown in Figure 3.7. First, how far will an electron move through one of these crystal channels? The answer is about 100 atomic spaces. This is because the atoms are not really stationary but

56

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Random walk

Light absorption

Junction

Figure 3.10. A light-generated carrier diffuses to the junction in a random walk sequence.

are vibrating small distances around their home positions because they have thermal (heat) energy. This vibration energy is small, however, so that the excited electron does not return to the valence band but just gets deflected into another channel. We think of this deflection as a step in a random walk diffusion problem. This brings us back to Figure 3.10. The next question is how far is the excited-state electron away from the junction? This depends on the photon absorption distance. This absorption distance depends on the material and on the rules for photon absorption. Now we shall divert for a minute to the rules for photon absorption. This will be important because, as we will see, silicon is fundamentally different from the III–V semiconductors in its photoelectric properties. Looking back to the hydrogen atom in Figure 3.4, a rule for photon absorption is that the wave functions involved have to have different symmetries. For example, note that the S and D wave functions are symmetric around the position of the nucleus, while the P functions are antisymmetric. Thus, absorption between S to P and P to D are allowed, but S to D is not allowed. Now let us look at the wave functions for silicon and GaAs in Figure 3.6. Note that both wave functions for silicon are symmetric around the point between two silicon atoms. This means that photon absorption in silicon is not allowed to first order. In GaAs, however, photon absorption is allowed. So the photon absorption length in GaAs is about 10,000 atomic spaces. In reality, photons are also absorbed in silicon but in about 100,000 atomic spaces. This second-order absorption in silicon results because of atomic thermal vibrations. Now, returning to the purity question and the random walk diffusion problem, remember that a step length is about 100 atomic spaces. So a carrier in GaAs will

HIGH-EFFICIENCY AND MULTIJUNCTION SOLAR CELLS

57

be about 100 steps away from the junction, and a carrier in silicon will be about 1000 steps away. However, in a random walk problem, the number of steps required to move N steps away from the start is N × N steps. So, the distance an excited electron must travel to the junction in GaAs will be 10,000 steps or one million atomic spaces. If it were to see a large impurity in a channel on this path, it could return to the valence band and be lost. So the purity requirement for GaAs is about 1 ppm. The analogous argument for silicon suggests a purity requirement of 10 ppb. This is 1/100,000,000. In fact, silicon solar cells lose performance given transition metal impurities in the range of several parts per billion. The above argument has been a little tedious, but the goal is to impress the reader with this purity requirement. By analogy, it should also be clear that good single-crystal quality without defects is as important as purity. The above purity specification is routinely met in commercial single-crystal silicon solar cells today as well as in various other single-crystal silicon-based devices that have revolutionized our lives over the last 50 years. While the reader is probably not aware of it, various single-crystal III–V devices have penetrated our everyday lives as well in the last 10 years. As the above argument about the difference in photon absorption for GaAs versus silicon suggests, III–V are often a better choice for photoelectric and optical-electronic applications. Referring to the periodic table, there are a large number of III–V materials available including GaAs, InP, InSb, and GaSb. Additionally, alloys of these materials are available including AlGaAs, GaAsP, GaInAs, InGaP, and InGaAsP. This makes a large set of bandgaps and electron mobilities available. Single-crystal III–V devices can now be found in cell phones, satellite receivers, CD music players, CD-ROMs in personal computers, taillights in cars, traffic stoplights, and military weapon systems. Single-crystal III–V devices are also key components in fiber optic phone communication and the Internet. In fact, the most efficient solar cells are made using III–V materials. This brings us back to our second requirement for making high-efficiency solar cells. We need to use the energy in the sun’s varied colored rays as efficiently as possible. A problem with sunlight is that the photons come in different colors with different associated energies. If we wanted to maximize the efficiency of a photodiode, we would illuminate it with only photons with a single energy with an energy equal to the bandgap energy Eg. Then, if the crystal quality and purity were sufficient, all of the excited carriers would be collected at the junction with 67% of the photon energy being delivered as a voltage. The energy conversion efficiency would be roughly 67%. However, referring to Figure 3.2, photons from the sun come with different energies. Some of the photons have too little energy to be absorbed, and some of the photons have energy considerable in excess of the bandgap energy. For the sun’s spectrum, this limits the single-junction solar cell efficiency to less than 30%. Fortunately, group III–V offers a solution because various materials with various bandgap energies are available. Specifically, one can stack a visible light-sensitive GaAs solar cell with metal grids on its front and back on an infrared-sensitive GaSb solar cell to arrive at the two-color or two-junction solar cell shown at the right in

58

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Figure 3.11. In this way, one absorbs the high-energy photons first in the top material generating a high voltage, while the low-energy photons pass through the top cell to be converted in the bottom cell. More photons are used and they are used more wisely. This then is the 35% efficient GaAs/GaSb two-color or two-junction solar cell demonstrated by L. Fraas et al. in 1989 [2].

Silicon Eg=1.1 eV

Ga Sb Eg=0.7 eV

Ga As Eg=1.4 eV

Figure 3.11. Left: for single-junction solar cell, sunlight contains high-energy photons with excess energy and low-energy photons with too little energy. Right: solar spectrum can be more efficiently utilized by stacking two different junctions together. Three junction cell

First Second Third

n+ n p p++ n n p+ p+ n n

Window InGaP GaInAs

Ge

p

From p. 888 (at 300 suns AM1.5) 13.3 mA/cm2 x Vop η = --------------------------------84 mW/ cm2 Vop= 2.55 V Efficiency = 40%

Active solar cell junction Shorting tunnel interconnect

Figure 3.12. Monolithic triple-junction InGaP/GaInAs/Ge CPV cell as first described by L. Fraas and R. Kinechtli in the 13th IEEE PV Specialist Conference, predicting 40% at 300 suns AM1.5 [13].

PV MODULE AND SYSTEM COST TRADES

59

Concentrating lens

Solar Cell Heat spreader back plane

Light Generated Current (Amps)

Sun light Sunlight concentrated

Sunlight not concentrated

Voltage (V)

Figure 3.13. Solar cells are more efficient with concentrated sunlight because both current and voltage increase.

As first predicted in 1978 [13], an efficiency of 40% has recently been demonstrated [14] for the monolithic three-junction InGaP/GaInAs/Ge CPV cell as shown in Figure 3.12. This brings us to the third way of increasing solar cell efficiency. For a given bandgap energy, we want to generate more voltage. Concentrating the sunlight onto the cell can do this. This is shown in Figure 3.13. Sunlight can be concentrated using a lens as is shown at the left in this figure. The resulting currents versus voltage curves with and without a lens are shown at the right. As is customary for solar cells, the diode curves here have been flipped over. Note that the higher current concentrator cell has a higher efficiency. This is because the diode is being driven harder to a higher current and voltage. In other words, if the light level goes up by 10, the current also goes up by 10 but at the same time, the voltage also goes up. In practice, the open-circuit voltage can go up from about two-thirds of Eg to about three-quarters of Eg under concentrated sunlight.

3.8

PV MODULE AND SYSTEM COST TRADES

So it is clear that single-crystal cells have higher efficiencies than amorphous or small grain size thin-film cells and that multijunction cells have still higher efficiencies. However, single-crystal cells are more expensive than thin-film cells and multijunction cells are very complex. It is also clear that CPV systems see only direct sunlight, which is less than global sunlight. So, it is necessary to look at the cost of complete PV systems. This is shown in Table 3.2. Note that the attached table has five columns comparing system-level costs measured in terms of simple payback times. The columns summarize future (ca.

60

TABLE 3.2. Comparison of Future Economics for Planar PV and Concentrated Solar PV Module Type

Thin Film

SC Silicon

LCPV

HCPV

HCPV JXC

Cell Efficiency @ temperature

9%

19%

22%

35%

44%

Annual available irradiancea

2336 kWh/m2 (fixed tilt)

2905 kWh/m2 (tracking 1-axis)

2382 kWh/m2 (1-axis; 87% global; 94% optical efficiency)

2178 kWh/m2 (2-axis; DNI = 78% global; 85% optical efficiency)

2178 kWh/m2 (2-axis; DNI = 78% global; 85% optical efficiency)

Annual kWh/m2 electricity

210 kWh/m2

552 kWh/m2

524 kWh/m2

762 kWh/m2

936 kWh/m2

Dirt penalty

–5% = >200 kWh/m2 –5% = >525 kWh/m2

–7% = >488 kWh/m2

–9% = >695 kWh/m2

–9% = >852 kWh/m2

Annual revenue at 10¢/kWh

$20/m2

$52.5/m2

$48.8/m2

$69.5/m2

$82/m2

Cell cost per m2 module

$50/m2

$300/m2

$100/m2

$100/m2

$137/m2

Module materialb

$50/m2

$60/m2

$67/m2

$67/m2

$67/m2

Optics

0

0

$57/m2

$65/m2

$65/m2

Module cost

$100/m2 ($1.11/W)

$224/m2 ($1.20/W)

$232/m2 ($0.92/W)

$360/m2 ($1.89/W)

$75/m (1-axis)

$150/m (2-axis rigid)

$150/m2 (2-axis rigid)

$25/m2

$25/m2

$25/m2

$40/m2

$40/m2

$160/m2

$460/m2

$324/m2

$422/m2

$459/m2

$35/m

Installation Total module area cost

2

2

2

$269/m2 ($0.85/W)

$75/m (1-axis)

Array support structure

2

Module system costc

$1.78/W

$2.42/W

$1.74/W

$1.67/W

$1.44/W

System with inverter at $0.3/Wd

$27/m2 $2.08/W

$57/m2 $2.72/W

$56/m2 $2.04/W

$76/m2 $1.97/W

$96/m2 $1.74/W

System payback time at $0.1/kWhd

9.4 years

9.8 years

7.8 years

7.2 years

6.8 years

a

Las Vegas. Glass, frame, Wire, J-box, etc. Without inverter cost. An explanation of the assumed peak power rating is required. For the two planar cases, this is straightforward. The efficiencies listed should be interpreted as future projected module efficiency at operating temperature. In this case, the Wp rating assumes 1 kW/m2 global illumination. So, the module peak W power ratings are 90 W/m2 for the thin-film case and 190 W/m2 for the single-crystal silicon case. For the HCPV cases, the efficiencies listed should be interpreted as future projected cell efficiency at operating temperature. However, the Wp rating now assumes 850 W/m2 of direct illumination (DNI) and 85% optical efficiency. So, for the 35% HCPV column, the module peak W power rating is 253 W/m2 and for the 44% column, the module peak W power rating is 318 W/m2. For the LCPV case, there is no standard. However, since the 3-sun sees DNI plus one-third of diffuse, the logical standard to be consistent with the other two would be to assume 90% of global illumination or 900 W/m2. In the table, a 3-sun module rating of 220 × 0.9 × 0.94 = 186 W/m2 is assumed. d Including inverter cost. From above, given that the inverter size for the system is set by the system peak power rating, one can calculate inverter cost per square meter. For the thin-film column, this is $27 per square meter; for the Si column, this is $57 per square meter; for 35%, this is $76 per square meter; for 44%, this is $96 per square meter; for LCPV, this is $56 per square meter. b c

61

62

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

2015) projected system costs for thin-film PV systems, single-crystal silicon flatplate systems, a JX Crystals 3-sun LCPV evolutionary system, an HCPV system using today’s 35% multijunction cells, and a JX Crystals HCPV Dual-Focus Cassegrainian system using a 44% efficient cell combination. These systems will be described in more technical detail in Chapters 12–17. Each column lists projected module outputs in annual kilowatt hour per square meter for Las Vegas, and each column summarizes projected system-level area-related costs in dollar per square meter. These costs include not just module cost but also support structure and installation costs. Given annual projected energy outputs and total area-related costs, it is then possible to calculate a simple system payback time given an assumed energy value of 10 ¢ per kilowatt hour. The good news is that all five columns for all PV technologies promise a simple payback time of less than 10 years. These systems should last for over 20 years. The first column is for the thin-film PV case. This case suffers from a low cell efficiency of 9%. Module costs are assumed to be those stated by First Solar and other thin-film PV companies of only $1.11 per watt. However, even given these low module level costs, because of the low cell efficiencies and lower-energy production per square meter, it takes a long time to pay off the cost of the array support structures and system installation. The projected simple payback time is 9.4 years. The second column is for the conventional single-crystal silicon planar module case. Here, the good news is the 19% cell efficiency, which then leads to 2.5 times more annual energy production per square meter of module area compared to the thin-film case. Simple one-axis tracking is one of the benefits here, giving more kilowatt hour per kilowatt installed. Trackers, unfortunately, are not affordable for the thin-film PV case. However, the cell cost is much higher than the thin-film case. Note that the cell cost assumed here is a future projection of $1.58 per watt. Given a higher-energy output, the field-related area costs are paid off faster than for the thin-film PV case, and the simple payback time works out to be 9.8 years. The third column is for an evolutionary LCPV case as will be described in Chapter 12. This case assumes the same 1-sun single-crystal cells and module manufacturing process as for the second column case. The three changes are the cut-up of the 1-sun single-crystal cells into thirds, the addition of low-cost linear mirrors, and the addition of a thin aluminum sheet heat spreader at the back of the laminated standard module. This approach has already been demonstrated in the field. The immediate benefit here is low risk, straightforward manufacturing, and a reduction of the cell cost by a factor of 3. Additional benefits are the ability to utilize more than just direct sunlight on less than perfect sunny days and higher optical throughput. The same simple one-axis trackers can be used as for the planar silicon module case. The reduction of the cell cost relative to column 2 with no additional penalties leads to a lower simple payback time of 7.8 years. The fourth column is for the HCPV case using today’s 35% multijunction cells. Here, of course, the 35% cell efficiency is spectacular. However, there are some negatives. One negative is that these HCPV systems only see direct sunlight

THE IMPORTANCE OF SINGLE CRYSTALS

63

(DNI) and not global illumination. While DNI can be as high as 90% of global illumination on the best blue-sky day, on an annual average even in Las Vegas or in Phoenix, it is only 78% of global illumination. A typical optical throughput of 85% for HCPV is also a negative. These two negatives make a 0.85 × 0.78 × 35% = 23% silicon planar cell competitive. However, the hoped-for good news for the HCPV approach is a lower module cost because of the cheaper optical materials relative to the high-efficiency single-crystal solar cells. If the low-cost modules ($0.92 per watt assumed here) and if the low-cost precision tracking ($150 per square meter assumed here) can be demonstrated in high-volume production, then this system will beat the planar system in sunny locations with a simple payback time of 7.2 years. However, this approach is more risky and will take longer to build up the manufacturing infrastructure than for the LCPV case. In the still longer term, column 5 presents an even higher-efficiency HCPV approach as will be described in Chapter 15. This approach uses a recently announced improved triple-junction monolithic cell (39% efficient) at the focus in the center of a Cassegrain optical configuration. A second infrared cell is also used behind a secondary mirror with a dichroic coating allowing the infrared to get to the second cell. The second GaSb infrared cell adds an addition 5% to give a cell combination efficiency of 44%. The GaSb cell, which is a diffused junction cell [16], is relatively inexpensive to make. Even with the cost of this second cell, its added efficiency leverages down all of the system cost to give an even more attractive payback time of 6.8 years. This approach, while all of the separate components have already been demonstrated, will require a lot of time and money to bring into high-volume production.

3.9

THE IMPORTANCE OF SINGLE CRYSTALS

From the above discussion of system-level cost trades, why have CPV systems not received more attention and why have thin film systems received so much attention? One answer is that searching for a 20% efficient low-cost thin-film solar cell is a very attractive dream. However, in this chapter, we have talked about electrons as waves and semiconductors as crystals to convey the message that this dream is not well founded on scientific principles. In fact, in graduate school solid-state physics classes, the bandgap in semiconductors is rigorously derived based on the assumption of the perfect periodic single-crystal lattice. However, the importance of single crystals to semiconductor devices is not generally conveyed in a simple understandable way. It is certainly not knowledge available to funding sources or the financial community. Figure 3.14 is an attempt to rectify this situation by making an analogy between an electron traveling in a solid and a car traveling through a forest. Organizing the atoms in single crystals is like removing the trees to make a road through a forest. Atoms out of place or atomic impurities are obstacles for the electron just like trees are obstacles for a car. Collisions with these obstacles force the electron (or the car) to lose energy. Efficiency is dramatically reduced.

64

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS

Figure 3.14. Single-crystal versus thin-film solar cells. Organizing the atoms in single crystals is like removing the trees to make a road through a forest. Atoms out of place or atomic impurities are obstacles for the electron just like trees are obstacles for a car. Collisions with these obstacles force the electron (or the car) to lose energy. If you were a car driving through the national forest, or an electron passing through a solar cell, which path would you rather take?

In any case, after 25 years of effort on thin-film solar cells, their module efficiencies are still low and they have not penetrated the mainstream electric power marketplace. Concentrator solar cells have not entered the marketplace yet in high volume either. There are several reasons for this, but it is not for lack of performance. The technology for solar concentrators is well founded on established scientific and engineering principles. One of the problems for concentrators is that a larger investment is required for hardware like lenses and trackers as well as for new solar cell manufacturing facilities. It is time for a serious top-level funding commitment. The technology is ready.

REFERENCES

65

ABBREVIATIONS AM1.5—1 and ½ Air-Mass (see Chapter 19) CPV—concentration photovoltaic DNI—direct normal incident sunlight E—electric field Eg—bandgap energy GaAs—gallium arsenide GaSb—gallium antimonide Group III–V—compounds consisting of group III and group V elements from the periodic table (LED semiconductor compounds) HCPV—high-concentration photovoltaic LCPV—low-concentration photovoltaic LED—light-emitting diode P/N junction—positive/negative junction PV—photovoltaic or solar cell QE—quantum efficiency (the number of electrons collected per incident photon) λ—wavelength

REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13]

A. Einstein. On the quantum theory of radiation. Phyikalische Zeitschrift 18, 121 (1917). L. M. Fraas, J. Avery, J. Gee, et al. Over 35% efficient GaAs/GaSb stacked concentrator cell assemblies for terrestrial applications. In 21st IEEE PV Specialist Conference, p. 190. Kissimmee, Florida, 1990. IEEE, New York (1990). R. P. Feynman, R. B. Leighton, and M. Sands. The Feynman Lectures on Physics, Volume III—Quantum Mechanics. Reading, MA, Addison Wesley (1965). N. Bohr. Radiation Spectra and the Hydrogen Atom. Philos. Mag. 25, 10 (1913). N. Bohr. The Theory of Spectra and Atomic Constitution. Fys. Tidsskr. 19, 153 (1921); 9, 1 (1922). R. P. Feynman, R. B. Leighton, and M. Sands. The Feynman Lectures on physics. In The Hydrogen Atom and the Periodic Table, Vol. 3, Chapter 19, Reading, MA, Addison Wesley (1965). A. Sommerfeld. Z. Phys. 47, 1 (1928). F. Bloch. Z. Phys. 52, 555 (1928). A. H. Wilson. Proc. R. Soc. A 133(458), 134, 277 (1931). J. M. Ziman. Principles of the theory of solids. In Electron States, pp. 72–74. Cambridge and London, Cambridge University Press (1964). C. Kittel. Introduction to Solid State Physics. 3rd Edition. New York, John Wiley & Sons (1967). S. M. Sze. Physics of Semiconductor Devices. New York, Wiley-Interscience (1969). L. M. Fraas and R. C. Knechtli. Design of high efficiency monolithic stacked multijunction solar cells. In 13th IEEE Photovoltaic Specialist Conference, Conference Record (A79-40881 17–44), Washington, D.C., June 5–8, pp. 886–891. Institute of Electrical and Electronics Engineers, Inc., New York (1978).

66 [14] [15]

SOLAR CELLS, SINGLE-CRYSTAL SEMICONDUCTORS R. R. King, D. C. Law, K. M. Edmondson, et al. 40% efficient metamorphic GaInP/ GaInAs/Ge multijunction solar cells. Appl. Phys. Lett. 90 (18), p. 3516 (2007). L. M. Fraas, H. X. Huang, and J. E. Avery. Low cost high power GaSb photovoltaic cells. Thermophotovoltaic generation of electricity. In 3rd NREL Conference, T. Coutts, C. Allman, J. Benner, eds, p. 33. Woodbury, N.Y., AIP (1997).

4 SOLAR CELL DEVICE PHYSICS LARRY PARTAIN Varian Medical Systems

4.1 DEVELOPMENT OF QUANTUM MECHANICS AND SOLID-STATE ELECTRONICS Solar cells are likely the ultimate quantum mechanical and solid-state electronic devices. Although their behavior (photovoltaics) was observed in the early 1830s, their optimization and rapid advancement awaited the development of both these fields starting at the turn of the twentieth century. Many of these discoveries are summarized by the set of equations immediately below. The person most associated with each discovery is shown in parentheses along with the year of the discovery. Asterisks identify those advances that were awarded Nobel Prizes in physics [1, 2]. (Planck, 1900)*

E N = Nh f

(4.1)

(Einstein, 1905)*

E = hf

(4.2)

(Compton,1923)*

p = hf c

(4.3)

(De Broglie, 1923)

λ=h p

(4.4)

f ( ε ) = 1 [1 + exp {( ε − ε F kT )}]

(4.5)

(Fermi and Dirac, 1926) (Schrodinger, 1928)* (Shockley et al., 1947)*

[ − (h

2

8π m ) ∇ + V ( r )] FK ( r ) = EK FK ( r ) 2

2

ε*Fe − ε*Fh = qV and J n = qnμ n dε*Fe /dx.

(4.6) (4.7)

Equation 4.1 explained the shape of the black body light emission spectrum from hot objects including the sun. The energy at any emitted light frequency f consists Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

67

68

SOLAR CELL DEVICE PHYSICS

of N discrete (quantized) energy packets each with a small energy content proportional to f with a proportionality constant (Planck’s constant) value of h. Equation 4.2 more definitively established that the wave phenomena of light is actually made up of discrete photons each with a quantized energy value hf that explained the experimentally observed photoelectric effect of electron emission from metal surfaces in a vacuum illuminated with monochromatic light. Equation 4.3 established that each such photon possesses a quantized momentum value, p, which accurately explains the trajectories of photon “collisions” with electrons. Strangely enough, Equation 4.4 indicates that rest mass particles like electrons also behave like waves with wavelengths λ that are inversely proportional to their momentum value, p. This is one description of the wave–particle duality. In 1913, Neil Bohr’s [3] Nobel Prize-winning work (when combined with the De Broglie result of Eq. 4.4) explained the discrete electron energy levels of hydrogen atoms have just the right sizes to accommodate integer multiples of each electron’s wavelength, λ. Thus, the hydrogen atom has allowed electronic energy states with unoccupied energy gaps between each allowed state or energy “band.” It is reasonable to expect that when atoms are brought together to form solids, their atomic discrete levels are perturbed and broadened into allowed electronic energy state bands. Equation 4.5 describes the probability that any of a material’s allowed electronic states is actually occupied by an electron, in terms of an energy variable expression containing Boltzmann’s constant k and referenced to a special energy value εF (the Fermi energy). The latter has ultimately been found to be the electron’s potential energy value [4, 5]. 4.2 FUNDAMENTALS OF SOLAR CELL OPEN-CIRCUIT VOLTAGE The density of available quantum states to “free” electrons in a semiconductor can be approximated by solving Schrodinger’s wave equation for the case of a cube of semiconductor material where the states’ potential energy, V, is zero everywhere inside the cube and is infinite everywhere outside this cube [6]. All of the allowed solutions to the wave equation are sine waves whose wave numbers K (i.e., in the x direction Φ(x) ∼ sin(Kx)) are integer multiples of π/a where a is the dimension of one side of the cube equal to the atomic spacing of the semiconductor. If one then integrates such quantized K values in three-dimensional space, using a differential volume that is normalized by the volume (π3/a3) occupied by one individual quantum state and applying the Fermi–Dirac expression for the probability that any available state is actually occupied, then one obtains the well-known expressions for the volume density of electrons n in the conduction band (and similarly the volume density of holes p in the valence band) with the respective conduction and valence energy band edges specified by εC and εV [7, 8]: n = N c e −(εC − ε F ) kT

(4.8)

p = N v e −( ε F − ε V ) kT .

(4.9)

FUNDAMENTALS OF SOLAR CELL OPEN-CIRCUIT VOLTAGE

69

Here, NC = 2Mc(2πmdekT/h2)3/2 and Nv = 2(2πmdhkT/h2)3/2. mde and mdh are the density-of-states effective masses for “free” electrons and holes, respectively, and Mc is the number of equivalent minima at the conduction band edge [8, 9]. Each hole denotes the absence of an electron in the valence band. Note that the product of n × p gives np = N C N V exp[− ( ε C − ε V ) kT = N C N V exp ( − ε G kT ) = ni2 ,

(4.10)

which only varies with the bandgap and temperature regardless of the doping levels to make the semiconductor either n or p type or alternatively to characterize its intrinsic or “i” properties if undoped. Then ni is the electron and hole density in such “i” materials. While the density states derivation is based on the periodic atomic spacing a, any dependence on this a disappears in the total mathematical calculation process. The results, at least to the first order, appear to accurately describe the properties of even amorphous semiconductor materials with distributions of atomic spacings and not just a well-defined and single constant a as long as there is an approximately defined bandgap. Equations 4.8 and 4.9 are equilibrium expressions for materials in the dark. With light exposure, both conduction band electron and valence band hole concentrations increase above their equilibrium values. Absorbed photons with enough energy “pump” valence band electrons into the conduction band, leaving behind the valence band holes. The standard assumption is that Equations 4.8 and 4.9 still apply but with two different quasi-Fermi energy values, ε*Fe for the electrons and ε*Fh for the holes instead of the single-equilibrium Fermi energy εF0. For the latter, subscript “0” is added to denote its equilibrium position. These expressions clearly show that for n to increase, ε*Fe must move away from εF0 toward the conduction band edge εC. Similarly, ε*Fh must also move away from ε*F0 for p to increase but toward the opposite valence band edge εV. Since quasi-Fermi energies correspond to potential energies, the light-exposed semiconductor populations of electrons and holes actually possess measurably different potential energy values that, according to the first Shockely expression in Equation 4.7, is a measureable voltage, V. The magnitude of this voltage is expressed as V = [ Δε LFe + Δε LFh ] q ,

(4.11)

where Δε LFe = ε*Fe − ε F0 and Δε LFh = ε*F0 − ε Fh with the superscript “L” added to denote the nonequilibrium light exposure. It is this splitting of quasi-Fermi and potential energy values of electrons and holes that is the first fundamental step in the quantum mechanical conversion of the microscopic photon energy (typically expressed in electron volts per photon) into a macroscopic DC voltage, V, which can be physically measured and used to do work. The challenge is to include properly configured semiconductor materials into a device that allows such a macroscopic physical V measurement with a standard voltmeter.

70

SOLAR CELL DEVICE PHYSICS

A simple explanation of the underlying physics was presented in the First Edition (pp. 2–4 [5]). It details how concentration gradients (normalized by the total concentration value) produce forces per unit charge that are just as real as the forces that electric fields exert on such charges. Moving quantized particles through such concentration gradient force fields involves work that changes potential energies. This derivation is not repeated here.

4.3

SHOCKLEY DIODE MODEL OF SOLAR CELLS

In 1947, Shockley et al. [10] produced the Nobel Prize-winning discovery and demonstration of solid-state electronic transistors fabricated from single-crystal semiconductors with energy bandgap properties. Shockley’s breakthrough devices incorporated characteristics well summarized by the two expressions of Equation 4.7 (pp. 305–310, 463 [9]) that he used extensively in his work. The second Equation 4.7 expression also clearly implies quasi-Fermi level connections to potential energy because the gradient of the former provides a force that translates into velocity (and then to a current density, J) with a proportionality constant that is the electron mobility μn, where n is the density of conduction electrons in the semiconductor material. The energy gradient and resulting velocity prescribe how much of the starting potential energy and thus the efficiency is lost in establishing a DC terminal voltage, V, and a current flow as detailed below. Interesting enough, Shockley never clearly identified quasi-Fermi levels as measures of potential energy. However, there had long been hints of such a relationship, not the least being the Einstein relation that the ratio of diffusion coefficient to mobility is a known and calculable constant [8, 9]. In their bipolar forms, transistors consist of two back-to-back Shockley diodes. All of these quantum mechanics and solid-state electronic discoveries led directly to the detailed description of the Shockley diode solar cell behaviors and their controlling parameters needed for reproducible progress in this field. These preceding advances received a significant share of the Nobel Prizes in physics in the first half of the twentieth century, and their exploitation in the second half of the twentieth century, along with one more Noble Prize-winning discovery, has resulted in major increases in solar cell performance from ∼1% efficiency level in ∼1900 to over 40% currently. One wonders what the next century of Nobel Prizewinning discoveries may provide. In 1954, Chapin et al. [11] reported a 6% efficiency for an abrupt p/n junction diode single-crystal silicon solar cell conversion device. Seven years later, the derivation and treatment of the classic Shockley model was published by Shockley and Queisser [12] for the current–voltage properties and the efficiencies of this particular configuration. The corresponding geometry and nonequilibrium energy band structure for this type of solar cell exposed to light is shown schematically in Figure 4.1. This contact metallization partially covers its emitter to allow light entrance. A macroscopic and measureable voltage, V, is clearly indicated as being equal to

SHOCKLEY DIODE MODEL OF SOLAR CELLS Emitter

71

Base Depletion region

Light n

p tE tE'

Grid metal

tB xE

tB'

xB

Back metal x

Metal fermi level V ~ 0.6 Voc

0

xn

xP

εc

Loss δεn

Light ε ε

qV

* Fe

q(V-JR)

Dark ε

* Fh

XW

Light Loss

* Fe

ε

Dark

Metal fermi level

* Fh

δεp

εv

Figure 4.1. The Shockley diode configuration (top) and its energy band diagram (bottom) in the light.

the different majority carrier Fermi levels at opposite sides of the solar cell. A significant difference in Figure 4.1 from most earlier models is its clear identification of the Fermi energy in the metal contacts and their alignment with the majority carrier Fermi levels in the p/n junction solar cell. The assumption here as that the metal contacts do not significantly respond to the incident light so that their nonequilibrium behavior is well described just by their single-equilibrium Fermi energy values. At best, the band diagrams in Shockley’s 1950 book (fig. 12-3, p. 310 [9]) only hint at such a diagram, and his classic model publication contains no energy band diagrams [12]. However, it is metal conductors that deliver electric power to loads, and any potential energy conversion in a solar cell can only be utilized if such potential energy differences are efficiency transferred to the metal contacts. Hovel’s often quoted text only weakly relates voltage to differing Fermi levels and with no connection to metal Fermi energy levels (fig. 3, p. 9 [13]). Even though Sze’s (fig. 4, p. 794 [8]) textbook does relate the open-circuit voltage to the Fermi level split, there is no connection made to Fermi levels in the metal contacts. It is such latter levels and their alignments (or misalignments) with conduction and valence band edges that form the basis for projected significant increases (or losses) in solar cell efficiencies that is the major point of this chapter.

72

SOLAR CELL DEVICE PHYSICS

The quasi-Fermi level splitting in Figure 4.1 between ε*Fe and ε*Fh gives the highest potential energy difference available from the solar cell conversion process as characterized by setting the open-circuit voltage Voc to the voltage V of Equation 4.11. For low injection conditions, where the concentrations of the light-generated carriers ΔnL = n – n0 and ΔpL = p – p0 are much greater than that of the minority carriers but are much less than that of the majority carrier concentrations, the opencircuit expression simplifies to Voc = ε G q + ( kT q ) ln [ Δ nL po ( N C N V )]

(4.12)

for the p-type side of the solar cell and to Voc = ε G q + ( kT q ) ln [ Δ pL no ( N C N V )]

(4.13)

for the n-type side of the solar cell. For high injection where ΔnL and ΔpL become larger than the majority carrier concentration, these expressions respectively become 12 Voc = ε G q + ( 2 kT q ) ln ⎡⎣ Δ nL ( N C N V ) ⎤⎦ and

(4.14)

12 Voc = ε G q + ( 2 kT q ) ln ⎡⎣ Δ pL ( N C N V ) ⎤⎦ .

(4.15)

Only significant voltages are generated if ΔnL or ΔpL >> no or po, respectively (low injection), or is greater than both (high injection). In abrupt p/n junctions with significant doping levels and low injection, this means that voltage is only really generated by minority carriers. Note the twice higher temperature coefficient of the second term in the Voc expressions for high injection versus low injection. Also note that Voc approaches the εG/q bandgap value (in all four expressions) when the temperature T approaches absolute zero [12] or when the light-generated carrier concentrations approach 1018–1019 cm−3 concentrations (depending on the density-of-states effective mass values) where the arguments in the ln functions approach values of one. It is this latter case of high light-generated carrier concentrations giving bandgap Voc values that is another major point of this chapter for significantly increasing solar cell efficiencies. Given the absorption of a photon that produces an electron–hole pair, this specific optimization minimizes the loss of energy below the bandgap. This contrasts to minimizing the losses above bandgap in the multiple bandgap devices that have recently provided efficiencies exceeding 40% (Perharz and Bett, Chapter 14; [14]). The fundamental Shockley transistor assumption implies that any minority carrier that enters the depletion region exits the other side with relatively little change in the potential energy it had on entrance. To the degree that this assumption is accurate, it provides an efficient, low-loss process for transforming minority carriers into majority ones. This conversion process from minority to majority

SHOCKLEY DIODE MODEL OF SOLAR CELLS

73

carriers is the second major step of the abrupt p/n junction solar cell energy conversion process. It is instructive to note here the most efficient step theoretically possible for converting an absorbed photon’s energy in electron volts into a DC voltage, V. This occurs when all absorbed photons’ energies hf (i.e., monochromatic photons all of a single frequency f) just exceed the bandgap energy εG, and the open-circuit voltage Voc equals the bandgap value εG/q. This would only be possible if the quasi-Fermi levels ε*Fe and ε*Fh split all the way to the band edges εC and εV. It would be a perfect quantum mechanical conversion step with no direct loss. When the energy of an absorbed photon is greater than the bandgap, its energy excess greater than the bandgap creates “hot” electrons (and/or holes) whose energies are rapidly thermalized to steady-state values near the band edge by inelastic collisions with other charged carriers and/or with semiconductor lattice vibratrions (phonons). The latter quickly convert such extra energy into heat that no longer can contribute to steady-state carrier potential energy but only to the semiconductor’s temperature increase. The third major step in the energy conversion process is the extraction of the majority carriers’ light added energy from the semiconductor and the deposition of this energy into the metal contacts whose average electron potential energy values are also specified by their respective metal Fermi energy levels indicated by the arrows in Figure 4.1. Because of semiconductor band structure, mobile electrons can only occupy electronic states whose total energy is at or above the conduction band edge εC. They thus enter the metal with offset energy δεn as a kinetic energy (hot electron) that is again quickly thermalized by scattering down to the metal’s Fermi energy level. Since only the potential energy is available at the external metal contacts, the typical way to reduce this δεn loss is to raise the electron quasi-Fermi level by moderate to heavy doping at the semiconductor’s boundary. An analogous δεp loss process also occurs for majority carrier holes at the p-type interface with its metal contact. This is likewise typically reduced by moderate to heavy doping at the semiconductor’s boundary. For the highest conversion efficiencies, both these δεn and δεp need to be simultaneously reduced to as close to zero values as possible. The second expression of Equation 4.7 indicates that a loss of potential energy is required to transport charge from one part of a device to another as described by a current density, J. The latter is a fundamental and unavoidable potential energy loss even if the conversion step to open-circuit voltage is perfect and loss free. Despite the lack of band diagram clarity, it is the one-dimensional Shockley and Queisser [12] equations, as extended by Hovel [13], and only slightly extended more in the First Edition (chapter 1 [5]), that have formed the basis for most of the successful analysis and performance improvements that have eventually and recently led to cell efficiencies exceeding 40% ([14]; Perharz and Bett, Chapter 14). These well-known derivations will not be repeated here but will only be summarized by the resulting expressions that identify key parameters. In the Shockley abrupt p/n junction model of Figure 4.1, an electrostatically charged depletion region penetrates a depth, xE, into the emitter of thickness tE and

74

SOLAR CELL DEVICE PHYSICS

also penetrates a depth, xB, into the base of thickness tB. Outside the depletion region, there is no net charge, so any potential energy change, for the mobile electrons and holes there, cannot be due to an accumulation of net electrostatic charge. The majority carrier transport is simply described by the linear Ohm’s law, current–voltage relationships, specified by a single series resistance. The more important physics is contained in the nonlinear expressions (Eqs. 4.8 and 4.9) and relationships that describe the minority carrier transport, where the effects of concentration gradients are dominant. Combining the transport equations with the one-dimensional continuity equations and the standard boundary conditions, and the assumption of current dominated by minority carrier diffusion current at each edge of the depletion region [12, 13], one obtains the standard Shockley diode equations for current density J versus voltage V in an abrupt n/p junction solar cell as J = J 0 {exp [ q (V − JR ) kT ] − 1} − J L ,

(4.16)

V = ( kT q ) ln [( J J 0 ) + ( J L J 0 ) + 1] + JR,

(4.17)

or solving for V,

where Jo = (Joh + Joe) and J oh = q ( ni2 N D ) vdh f h and J oe = q ( ni2 N A ) vde f e and the light-generated current JL is ∞

J L = ∫0 qF ( λ ) [1 − r ( λ )] QE ( λ ) dλ.

(4.18)

Here, fh and fe are hyperbolic functions of the surface or interface recombination velocities, the diffusion velocities, the diffusion lengths, and the undepleted n and p layer thicknesses specified in the First Edition [5]. The R = RA is the specific series resistance, where R is the actual series resistance in ohms describing majority carrier transport and A is the cross-sectional area of the device. The ND and NA are the respective donor and acceptor doping concentrations of the emitter and base. The F(λ) is the photon flux density of light of wavelength λ, r(λ) is the light reflectivity of the emitter surface at λ, and QE(λ) is the quantum efficiency of the solar cell at the indicated wavelength. The open-circuit voltage Voc comes from setting J = 0 in Equation 4.17 to give Voc = ( kT q ) ln [( J L J 0 ) + 1] ,

(4.19)

which is equivalent to Equations 4.12 and 4.13’s low injection expressions applied to the boundaries of the depletion region. The short-circuit current density Jsc (where V = 0) is given from Equation 4.16 by the transcendental expression J sc − J 0 [ exp ( − qRJ sc kT ) − 1] = − J L .

(4.20)

SHOCKLEY DIODE MODEL OF SOLAR CELLS

75

If R is known, the light J–V characteristics are uniquely determined when any two of the three coefficients JL(∼ –Jsc), Jo, or Voc are known. Equation 4.19 gives the third when the other two are specified. The power density is the JV product (=J{[KT/q] In[(J/Jo) + (J/JL) + 1] + JR}) from Equation 4.17 that is maximum when the d(JV)/dJ = 0, at a maximum power current density Jm and voltage Vm. This Jm is negative, with a magnitude slightly less than JL (as shown in Fig. 4.2), and it is specified by the transcendental expression ln [( J L + J m + J 0 ) J 0 ] = − J m ( J L + J m + J 0 ) − 2qRJ m kT .

(4.21)

Substituting this Jm back into Equation 4.17 then gives Vm. The efficiency of the cell η is given by 2

1

0

Voc

J (A/cm2)

–1

–2 JL –3

R = 0.3 Ω-cm2 0.22

–4 0.0091

Jsc

0.15

–5

–6 –1

–0.5

0

0

0.06

0.5

1

1.5

2

V (V)

Figure 4.2. The light and dark J–V characteristics calculated with Equation 4.16 for Jo = 9.23 (10−20) A/cm2, JL = 5.17 A/cm2. The open-circuit voltage Voc and short-circuit current density Jsc are shown as a function of the specific series resistant R in ohm-square centimeter.

76

SOLAR CELL DEVICE PHYSICS

η = − Vm J m A Pin = − Voc J sc FF ( Pin A) = − [( ε G q ) VF][ J scmax JF] FF ( Pin A) ,

(4.22)

where Pin is the optical input power to the solar cell and A is the cell’s crosssectional area. The negative signs come from Jm and Jsc being negative. The fill factor FF is conveniently defined as the ratio FF = Vm J m Voc J sc = ( J m J L )(Vm Voc )( J L J sc ) ,

(4.23)

which specifies the “squareness” of the light J–V characteristics (see Fig. 4.2), and it has a maximum value of one and accounts for the unavoidable potential energy loss in transporting the charge to the solar cell boundaries for maximum power and the unavoidable “dark” current losses of forward biased diodes as well as losses due to high series resistance values. The (qVoc/εG) = VF loss fraction is the loss when the Voc is less than the bandgap. The loss fraction JF = J sc J scmax describes when all the incident photons with energies greater than the bandgap εG do not contribute to the Jsc (i.e., the external quantum efficiency is <1). A loss factor not included in the above is the energy that absorbed photons have that exceed the bandgap value εG. The latter is decreased using multi-bandgap cells and whose loss minimizations with three differing bandgaps have recently resulted in cell efficiencies exceeding 40% ([14]; Perharz and Bett, Chapter 14). These expressions inherently include the Equation 4.7 relations because the latter were the basic underlying assumptions used to derive the Shockley abrupt p/n junction model. When sunlight is concentrated through some optical means, a concentration ratio, X, can be defined as the concentrated optical input power Pinc divided by the unconcentrated one sun value Pin1 X . If the spectral distribution is not changed too much by the concentrating optics, the light-generated J Lc is just X times its one sun value J L1X since Equation 4.18 is linear in photon flux intensity F(λ). Hence, concentrator efficiency ηc is given as ηc = − Voc J sc1 X FF ( Pin1 X A) .

(4.24)

In high-performance devices with a low R so that Jsc/JL = –1 and with Jo/JL << 1 so that In(Jo/JL) is well approximated in Equation 4.19 by qVoc/kT, the FF of Equation 4.23 is totally specified by the factor qVoc/kT independent of the value of JL [12, 13], since the transcendental Equation 4.21 then simplifies to ln [(1 − J m J sc ) exp ( qVoc kT )] = ( J m J sc ) (1 − J m J sc )

(4.25)

and Equation 4.17 similarly simplifies to Vm Voc = ( kT qVoc ) ln [( − qVoc kT ) (1 − J m J sc )] .

(4.26)

p-i-n SOLAR CELLS

77

1.0 CIGS

0.9 0.8

Ge

Si (4 data points) GaAs 1050X GaAs 240X

Fill factor

0.7 0.6

GaAs 1X

0.5

CdTe

0.4

Photochemical

0.3 0.2

Theoretical maximum

Organic polymer

a-Si (stabilized)

0.1 0.0 0.00

20.00

40.00

60.00

80.00

100.00

120.00

qVoc/kT

Figure 4.3. The variation of the fill factor FF with the qVoc/kT values. The solid line curve is the maximum values predicted by the Shockley diode model with measured data points for comparisons.

The resulting relationship is plotted as the solid line theoretical curve in Figure 4.3 for qVoc/kT values between 0 and 105. For qVoc/kT between 50 and 105, the FF rises from 0.90 to 0.95. This theoretical FF asymptotically approaches 1 for Voc >> kT/q [12] and 0 for Voc << kT/q. The measured FFs of all the Green et al. [15] champion single-junction solar cells are plotted on this figure for reference along with the high-performance Ge and 1050X GaAs cells from the First Edition [5]. Note that all these experimental values fall beneath this curve, although a few of the highest performance ones approach the theoretical limit curve.

4.4

p-i-n SOLAR CELLS

By 1969, non-abrupt p/n junctions were proposed, first for Ge cells intended for thermophotovoltaic applications with high light intensities and high short-circuit current densities [16]. This involved inserting an undoped or intrinsic (i) layer between the p- and n-doped regions. These workers noted the increased carrier lifetimes in the undoped regions and the significance of open-circuit voltage as a fraction of the bandgap. They proposed and demonstrated the advantages of interdigitated back surface contacts. By 1975, p-i-n crystalline silicon solar cells were proposed with IBCs to better meet the needs of concentrated sunlight applications including higher doping concentrations in the p and n layers for higher open-circuit voltages [17, 18]. By 1976, the first a-Si p-i-n cells were reported with thin and heavily doped n+ and p+ regions between the metal contacts and on each side of the “i” layer [19]. Some of these, at least initially, were analyzed with the Shockley-like abrupt p/n junctions expressions with various modifications to account for their non-abrupt “i” regions.

78

4.5

SOLAR CELL DEVICE PHYSICS

OHMIC CONTACTS AND HETEROJUNCTION INTERFACES

Metal contacts whose Fermi levels well align (see Fig. 4.1) with semiconductor band edges are needed to keep energy losses at these interfaces to a minimum as the majority carriers flow from the semiconductor materials and deliver power to external electrical loads. Typically, any energy losses (other than the δεn and δεp offset losses discussed above) at such interfaces are described by a contact resistance for each interface. When the voltage loss across such an interface is negligible compared to other major voltage drops in the circuit, such contacts are referred to as ohmic contacts. The latter does not necessarily mean that there is a linear relationship between the voltage drop and the current flowing through the contact but that this voltage drop increases monotonically with the current. It further means that whatever voltage drop that does develop is negligible in comparison to the major voltage drops in other parts of the solar cell circuit. Many workers have concluded that ohmic contacts correspond to interface boundaries to semiconductors with high minority carrier recombination rates, typically described by high surface recombination velocities [20–23]. The development history of high-efficiency solar cells as well as the implications of another Nobel Prize support this conclusion as discussed below. The seemingly mutually exclusive solar cell requirements (1) for ohmic contacts with low contact resistance unavoidably accompanied by high surface recombination velocities and (2) for cell interface boundaries with low surface recombination velocities were solved by the double heterostructure work pioneered by Kroemer and Alferov that began in 1957 [24]. Their work was crucial to the development of semiconductor light-emitting diodes (LEDs) and lasers that require very high current density flows between metal contact surfaces with doped semiconductors in a manner that does not limit the lifetimes of nearby excited carriers. Their breakthrough contributions were recognized with a Nobel Prize in 2000. They separated the problem into two solvable parts. Ohmic contacts were made to a higher bandgap material whose abrupt transition to a lower bandgap material keeps the latter’s minority carriers separated from the ohmic contact by the barrier height discontinuities at the conduction or valence band edges of heterojunctions. The low contact resistance is produced by a combination of high semiconductor doping of the regions adjacent to the contact metals to provide tunnel junction behavior and by microscopic roughness at the interface where pointed boundary regions provide electric field enhancement and field emission due to the differing dielectric constants of the metals and the semiconductors [23, 25–28]. Both provide higher average current densities at lower voltage drops across the contact regions. 4.6 CHAMPION 28.2% EFFICIENT GaAs AND OTHER SOLAR CELLS A “champion” cell means the highest record efficiency measured in a solar cell of some specific configuration. By definition, it is irreproducible since it is the highest value ever measured in a single device up to some specified date according to some

CHAMPION 28.2% EFFICIENT GaAs AND OTHER SOLAR CELLS

79

“validated” procedures preferably by “certified” laboratories independent of the organization that fabricated it. On a yield basis, the one-of-kind cost of a champion cell is enormous for any significantly sized development or manufacturing program. High-performance abrupt p/n junction solar cells demonstrate properties near those predicted by the ideal Shockley diode theory developed above. Analysis of the champion 28.2% efficient GaAs abrupt p/n junction concentrator cell illustrates this agreement and the relation of quasi-Fermi level splitting to fundamental performance limits. The correlations of this section help justify the predictions of solar cell performance limits developed later. The 28.2% efficiency is a corrected value calculated by multiplying the originally published 29.2% [29] by a correction factor of 0.952 to a value of 27.8%. In January 1991, Sandia National Laboratories implemented a recalibration of their primary reference cell [30] to agree with the National Renewable Energy Laboratory results, which resulted in the original 0.952 adjustment. Then recently, it was adjusted back up to 28.2% using the new reference solar spectrum ASTM G173-03 [31]. A GaAs alloy-based double-heterostructure configuration facilitated this high champion cell performance. Herbert Kroemer’s first conception of this type of structure was in 1963 for solid-state laser applications while he was at Varian in Palo Alto, California [32]. Even though Kroemer departed for academia in 1968, his remaining legacy and influence at Varian can be seen in the double heterostructure of this champion GaAs solar cell diagrammed in Figure 4.4 (and detailed in

Anti-reflection coating

3.5 μm front metallization 0.6 μm p-GaAs Cap Layer 0.04 μm p-(Al,Ga)As Window 0.5 μm p-GaAs Emitter

3.5 μm n-GaAs Base

0.2 μm n-(Al,Ga)As Mirror 0.6 μm n-GaAs Buffer

350 μm n-GaAs Substrate

Back metalization

Figure 4.4. The double heterostructure p-on-n configuration of efficient GaAs solar cells.

80

SOLAR CELL DEVICE PHYSICS

TABLE 4.1. Dimensions and Doping Levels of Optimized [33] p-on-n and n-on-p GaAs Solar Cells p-on-n Cell

n-on-p Cell

μm

cm

μm

cm−3

GaAs cap layer

0.6

p = 5(1019)

0.6

n = 2(1018)

(Al0.9Ga0.1)As window

0.04

p ∼ 1019

0.04

n ∼ 1018

GaAs emitter

0.5

p = 2(1018)

0.2

n = 1(1018)

GaAs base

3.5

n = 2(10 )

3.8

p = 5(1017)

(Al0.2Ga0.8)As barrier

0.2

n = 1(1018)

0.2

p = 2(1018)

GaAs buffer

0.6

n = 1(1018)

0.6

p = 2(1018)

−3

17

The champion 28.2% efficient solar cell was of n-on-p construction.

Table 4.1) as first reported by Varian in 1988 20 years later [29, 34]. Note that the top GaAs heterojunction was to an Al0.9Ga0.1As layer that provided a wide enough bandgap window near 3 eV for blue light to readily pass into the GaAs active light absorption region. It also confined the GaAs’s minority carriers to stay near the collecting diode junction region and away from the adjacent front surface ohmic contact with its very high surface recombination velocity for minority carriers. The lower heterojunction was to a Al0.2Ga0.8As layer for a band edge offset of only a few kT beyond that of GaAs to confine the minority carriers to also stay near the junction from the back side. A sulfuric acid–peroxide etch [35] was used to passivate its edges. Contemporary, parallel, continuing, and similar GaAs alloy solar cell development work in Russia has been well summarized recently [24]. A detailed analysis of this champion cell was presented in chapter 1 of the First Edition and is not repeated here [5]. Up to 2009, this champion cell still held the record for efficiency of a single-junction device as tabulated by Green et al. [31] even though it was fabricated and reported 21 years earlier in 1988. Here, only summary descriptions and discussions are given. Its high open-circuit voltage of 1.155 V under 240X sunlight concentration was essentially set by the minority carrier concentrations approaching the 1014 cm−3 range calculated from Equations 4.12 and 4.13. This level is much less than the majority carrier concentrations of 1018 and 5 (1017) cm−3, respectively, for the n emitter and p base in this device whose quasi-Fermi levels’ spacing from their nearest band edges were hardly changed from their equilibrium values in the dark. Assuming this is all still satisfied at minority carrier concentration values of 1017 cm−3, then such devices at 1000 times higher light concentrations would still be in the “low injection” range if the higher heat load and current densities could be managed. The reasonable Shockley diode model parameter values that well fit this device’s behavior were summarized in tables 1.3 and 1.4 of the First Edition [5].

CHAMPION 28.2% EFFICIENT GaAs AND OTHER SOLAR CELLS

81

A companion Varian p-on-n cell (listed in Table 4.1) to the champion n on p one described above had a conversion efficiency that was still increasing at a concentration ratio of 1260x where its efficiency was 26.2% (see fig. 1.8 in [5]). Note that its 0.6-μm GaAs cap layer was heavily doped at 5 (1019) cm−3. Alferov summarized Russian measurements of similar GaAs technology single-junction cells whose efficiencies were 26.2% at 1000x, 25% at 2000x and 23% at 5800x concentration values (fig. 8.4 [24]). At the 5800x concentration, the Voc had risen to 1.195 V, but the fill factor had fallen to 0.79 due to apparent series resistance problems. This open-circuit voltage indicates the minority carrier concentrations still remained in the 1014 cm−3 low injection range. For further reference at the 1680x maximum efficiency conditions for the latter device, its Voc was lower at 1.18 V but with a higher FF of 0.87 [36]. When triple junction cells were exposed to concentrations over 14,000X (fig. 8.16 [24]), rather modest increases in opencircuit voltages were observed but with kinks in their I–V curves indicative of their tunnel junctions being driven beyond low loss limits. Combined, these illustrate the difficulties of trying to make Voc approach bandgap values using high concentration ratios alone based on abrupt p/n junction low injection designs. The potential performance improvement of the low injection, abrupt n-on-p GaAs concentrator cell design can be assessed by using best reported, materials limited parameter values. The results were in 30–32% efficiency range for such single junctions without increasing the concentration ratio X beyond 240X [5]. The theoretical open-circuit voltage values Voc are specified in Equations 4.12 and 4.13 as a function of the low injection minority carrier electron and hole concentrations are shown in Figure 4.5a as the continuous curves assuming a densityof-states effective mass ratio of md/mo = 0.5. The solid line curves indicate the operational range of current champion devices with actual experimental results shown as data points, while the dashed curves indicate expected values at substantially higher and lower minority carrier concentrations. The GaAs data points appear at higher concentrations than the reported 1014 cm−3 levels reported immediately above. However, this difference is due to GaAs’s density-of-states effective masses being lower than the assumed 0.5 value used for the curves [8]. The solid lines indicate the approximate range of the best current champion cells and the upper dashed line indicates the 1019 cm−3 minority carrier concentrations that would be required to make VF approach one and the open-circuit voltages approach bandgap value (for a density-of-states effective mass ratio of 0.5). Note that this direct efficiency loss factor for voltage VF is well below 80% for data points other than the GaAs concentrator devices. Surprisingly, the champion CIGS VF is among the highest of the one sun cells with a minority carrier concentration exceeding 1013 cm−3, and CdTe is among the lowest, corresponding to a minority carrier concentration of ∼1010 cm−3. The dominant intrinsic or “i” region of p-i-n cells has such low-equilibrium carrier concentrations that almost any low light intensity provides light-generated carrier concentrations exceeding that of both the electrons and holes. This is the high injection case described by Equations 4.14 and 4.15 and plotted in Figure 4.5b. Similar contour curves are shown for constant light-generated carrier

82

SOLAR CELL DEVICE PHYSICS

Figure 4.5. The open-circuit voltage Voc as a fraction VF of the bandgap. The solid and broken line curves are theoretical values for constant light-generated concentrations from 104 to 1016 cm−3. (a) Low injection case. (b) High injection case.

concentrations (in undoped intrinsic materials, there is no well-defined “minority carrier”), again with solid curves used to emphasize the range of current champion devices indicated by data points. The low VF value of stabilized a-Si corresponds to low light-generated carrier concentrations of ∼1011 cm−3 about three orders of magnitude below that of the best GaAs concentrator cells. The IBC crystalline silicon champion cell achieved a VF value of 0.61 that corresponds to a carrier concentration of almost 1016 cm−3 if it is a high injection device but to less than 1013 cm−3 if it operates as a low injection device. Its i-like region was actually n-type and low doped to ∼1015 cm−3. Thus, it is an intermediate device whose VF = 0.61 value is plotted in both the (a) and (b) parts of Figure 4.5 due to this uncertainty. The crystalline silicon HIT cell has an

ADVANCED CONCEPT p/n AND p-i-n CRYSTALLINE SILICON CELLS

83

even higher VF =0.65 as plotted in Figure 4.5b. Its low n-type doping level has not been disclosed but is likely in the same 1015 cm−3 level range. The point contact crystalline silicon solar cell definitely has an undoped “i” layer and an even higher Voc = 0.80 V and VF = 0.73 corresponding to a light-generated concentration exceeding 1016 cm−3. This p-i-n point contact light-generated carrier concentration is over three orders of magnitude above the best GaAs concentrator cells plotted in Figure 4.5a. Direct comparison to Figure 4.5a immediately shows that same lightgenerated carrier concentrations produce much lower VF values under high injection conditions compared to low injection. This is due to the “majority” carrier Fermi level being essentially pinned close to its neighboring band edge in low injection devices due to rather heavy majority carrier doping levels (of abrupt p/n junction Shockley devices) that contribute almost half the bandgap to the observed Voc even at very low light-generated carrier concentrations. Superficially, this would seem to indicate a major advantage for low injection device designs. However, an attempt is made to argue the opposite point in Section 4.9; that is, that “high injection” devices actually provide the highest-efficiency potential when all interacting parameter trade-offs are considered.

4.7 ADVANCED CONCEPT p/n AND p-i-n CRYSTALLINE SILICON CELLS Three high-performance and innovative departures from the simple Shockley model discussed above are covered in this section, namely, the IBC cell, the point contact cell, and the HIT cell. Prior to the HIT discussion, the p/n Shockley-type PERL cell is also covered. All four are based on crystalline silicon.

4.7.1

Interdigitated and Point Contact Crystalline Silicon Cells

An innovative adaptation of the Shockley diode solar cell (of Fig. 4.1) is the IBC cell [16, 37, 38]. Its conceptual evolution can be seen by starting with a drawing in Figure 4.6 (top) of a one-dimensional symmetry cell with the light coming in from the top. All the surfaces are passivated with SiO except for narrow openings centered under each plated Cu grid line. To make ohmic contacts, a thin layer of Al is deposited, whose interface roughening reactions [26] with the underlying Si and its metallic nature provides for unimpeded majority carrier flow between the semiconductor and the exterior circuit. The Al on top of the SiO provides mirror reflection of any sunlight that strikes the sides. The small ohmic contact regions reduce the portion of the cell surface with high surface recombination velocities inherent to ohmic contacts. Heavily doped but thin p+ and n+ layers are diffused in from the right- and left-hand sides, respectively, into a low (∼1015 cm−3) n-type Czochralski silicon wafer of approximately 5-Ω-cm resistivity and ∼1-ms minority carrier lifetime. The p/n junction is at the left side at the p+–n− interface. If

84

SOLAR CELL DEVICE PHYSICS Light

p+

n+

n–

SiO Al reflector/barrier Plated Cu

εc ε*FeL q(V - JR*)

δεn ΔεFeL εFo

ΔεFhL ε*FhL

qV

δεp εv

Figure 4.6. One-dimensional conceptual representation of the interdigitated back contact cell (top) and its corresponding energy band diagram approximation (bottom).

conceptually this is folded about its center into a upside down “U” shape until the p+ and n+ regions touch, the actual IBC geometry is produced as shown schematically in Figure 4.7. As described by Mulligan et al. [39], the wafers are p+ layer diffused on both sides about 1.8 μm deep to give 16 Ω per square resistivity layers (p ∼ 5 (1018) cm−3). This is followed by a thermal oxide growth of ∼0.25-μm thickness on all surfaces. Next, a screen printed resist mask is used on the back-side oxide surface to provide ∼0.7-mm-wide strip openings (∼2-mm repetitive spacing) for etching ∼3 μm down through both the oxide and the p+-doped layer while at the same time the top side p+ layer is removed. The resist is cured thermally or with UV light. Next, an n+ diffusion is done to provide n+ layers in the 3-μm deep trenches about 0.9 μm deep with ∼40 Ω per square resistivity (n ∼ 1 (1019) cm−3). This is followed by a thermal oxide growth over all the n+ surfaces ∼0.095 μm thick. With a resist covering the back side, the textured top surface (∼1- to 10-μm-high square pyramids corresponding to the 100 wafer orientations) is etched into the wafer after the top oxide has been removed. Next, the wafers are n diffused and oxidized on both sides. The latter provides the n+ FSF under the oxide essentially for the production of a very

ADVANCED CONCEPT p/n AND p-i-n CRYSTALLINE SILICON CELLS

n+ Front surface field

85

TiO2 SiO

n– Wafer of Si

i-aSi

p+ aSi n+ aSi

Reflector/barrier

Metallization

Figure 4.7. Two-dimensional cross-sectional schematic of an interdigitated back contact solar cell.

low front surface recombination velocity [40]. The electric field of the FSF is in the direction to repel minority carrier holes from the wafer’s top or front surface. Finally, a SiN or TiO2 antireflection coating is applied to the top pyramid surface, which, in combination with the underlying SiO, provides a two-layer structure with a very low reflection coefficient over the active sunlight wavelengths of this solar cell. A screen printed resist is used to etch narrow openings in the oxide centered over the p+ and n+ strips in the back but configured as separate and disconnected “islands” to keep such exposed open areas to less than 5% the back surface. A three-layer metallization is then applied to the whole back side of ∼0.4-μm total thickness, starting with Al, followed by a titanium 10%–tungsten 90% diffusion barrier, and finally by a thin Cu layer, followed by annealing in forming gas at ∼300°C. The latter Cu outer coating provides the base for subsequent grid finger plating, but it does not exhibit strong lateral conductivity. Another screen printed resist is again used to define the grid lines over the p+ and n+ back-side regions and then Cu electroplated to ∼20-μm thickness and capped with ∼7 μm of tin plating to improve solderability. A final screen printed resist is applied so the ∼0.4-μmthick tri-layer metallization can be removed between all the grid lines that would otherwise partially short the n+ and p+ terminal contacts together. A perspective drawing of the overall structure is shown in Figure 4.8 including the 145-μm-thick Czochralski wafer thickness currently used in manufacturing [38]. The Al mirror coating of the oxides at the bottom reflects back any light at the longer wavelengths that penetrate that far, for repeat passes through the wafer. When p-type wafers were used instead of n-type, comparable performance could be achieved, but the p-type ones evidenced a few percent degradations under

86

SOLAR CELL DEVICE PHYSICS

Textured front surface n– 145 μm thick Czochralski silicon wafer SiO Passivation n+

–

p

+

+

Limited Contact Area

Cu Grid Lines

Figure 4.8. Perspective drawing of the interdigitated back contact solar cell.

Figure 4.9. Photograph of the back and front surface of an interdigitated back contact solar cell formed on a 6-in. Czochralski wafer.

sunlight due to a boron dopant and residual oxygen interaction. Fortunately, float zone silicon wafers are no longer needed to provide wafers with the desired 1-ms minority carrier lifetimes. A photo of the front and back surfaces of an interdigitated cell is shown in Figure 4.9. It shows the very black front surface and the back surface with all the alternating metal strips connected by top and bottom bus bar tabs that provide the plus and minus terminals for convenient solder attachment. Two-dimensional energy band diagrams are difficult to draw. Hence, Figure 4.6 (bottom) shows a one-dimensional approximation. The high near 1019 cm−3 doping of both the thin n+ and p+ layers makes the equilibrium (zero voltage, in the dark) Fermi level εFo penetrate both their conduction and valence band edges

ADVANCED CONCEPT p/n AND p-i-n CRYSTALLINE SILICON CELLS

87

plus provide good alignment with the metal Fermi levels with essentially negative δεn and δεp offset values for efficient conversion of the excited majority carrier concentrations into metal contact electrons with matched Fermi levels and negligible energy losses. Here, it is assumed this high doping significantly reduces the minority carrier lifetimes so the thin n+ and p+ regions do not contribute to lightgenerated current nor support high minority carrier concentration values. Very importantly, these heavily doped regions strongly define which carriers are the majority ones at the collection boundaries. Thus, when the “quasi-minority” quasi-Fermi levels collapse together to essentially a single Fermi level at the boundaries, the majority carrier Fermi levels stay aligned with the metal contact Fermi levels and at opposite band edges. The present band diagram is similar to that of Swanson (fig. 5 [37]). It is this asymmetric Fermi level merging process that coverts the quasi-Fermi level splitting in the bulk of the device into a physical, assessable, and useable voltage in the metal contacts. It plays the comparable role to the second major step of depletion-region-minority-to-majority-carrier conversion in Shockley abrupt p/n junctions. However, there are substantial and major differences in its operational mode that opens new opportunities for optimizations. In sunlight intensities low enough to produce increases in electron–hole pair concentrations below the ∼1015 cm−3 doping of the wafer, this specific resistance R is just the corresponding bulk resistances encountered in standard Shockley diode solar cells. However, at high light intensities where the photogenerated concentrations exceed the background doping level, the resistance contribution to the total R value is reduced. Once generated, the electrons (and holes) have no memory whether produced from thermally ionized doping levels or from light absorption and thus provide the same reduction in resistance regardless. The Figure 4.6 (bottom) plot is for this “high injection” case. However, one immediate disadvantage of this configuration is immediately seen. The equilibrium spacing of the Fermi level from the conduction band edge is much larger here than in standard Shockley devices. Hence, the open-circuit voltage is limited by the high injection restrictions of Equations 4.14 and 4.15 versus the low injection ones of Equations 4.12 and 4.13. One of the major cost reductions came from the replacement of all photoresist photolithography fabrication processes with screen printed resists steps. This was accommodated by widening the widths of the n+ and p+ strips shown in Figures 4.7 and 4.8 to accommodate the design rule constraints of screen printing. These wide features imposed a very flat aspect ratio for the resulting cells as illustrated schematically in Figure 4.10 (not to scale). These current trajectories are not unlike those of majority carriers in the emitters of standard Shockley diodes collected at top metal grid lines except here the “grid lines” can be very wide without any light shadowing loss. The amount of energy it takes to collect charge carriers over a given distance with a specified force is proportional to the time that force has to be applied, where that time is inversely proportional to the carrier velocity. Hence, the collection energy lost for both electrons and holes can be made similar by scaling these lateral collection lengths (approximately Wn and Wp) inversely with

88

SOLAR CELL DEVICE PHYSICS

n+

Electron trajectories p+

Hole trajectories 145 μm

+

n

p+

W Wn ~ 0.7 mm

Wp ~ 1.4 mm

Wn

Figure 4.10. The electron and hole trajectories in a flat-profile interdigitated back contact solar cell with wide 0.7- and 1.4-mm metal grid metals for the n+ and p+ regions, respectively.

their mobilities (of μn = 1350 cm2/Vs and 480 cm2/Vs for crystalline Si) in approximate agreement with the widths shown. As can be seen by the trajectories, the photoelectrons cannot be collected at the p+ region and the holes cannot be collected by the n+ region. Essentially, diffusion drives this whole process and as electrons are blocked at the p+ regions, their charge accumulation builds up to just the right distribution to generate an electrostatic field that deflects any new electrons transversely over to their proper collection region. The analogous process occurs for the photogenerated holes. The net effect is that these transverse currents can accurately be described just by doubling the diffusion coefficient value along all of the transverse paths. This is a phenomenon called the Webster effect [38], which has been well known since the early days of high-current transistor development [41]. The effect reduces the potential energy loss from the charge diversion process. Under “high injection” conditions, the 21.8% champion IBC cell with a Voc of 0.677 V, a voltage loss factor VF = 0.62, and a fill factor of 0.806 at a Jsc of 40 mA/cm2 at 25°C [15], the light-generated ΔpL and ΔnL concentrations were on the order of 4.7 (1015) cm−3. It is of interest to note that the n− doping concentration selected to reduce series resistance R problems for the interdigitated cell is in this same general concentration range. With its high measured fill factor, essentially at the limit of Shockley diode theory for this open-circuit value, there are certainly no trap-controlled SCLI limitations that are covered in the following section. However, a minority carrier concentration of 4.7 (1015) cm−3 would have produced an even higher Voc in a standard Shockley diode cell with a majority carrier doping concentration in the mid (1017) cm−3 range. For comparison, the 24.7% efficient champion Shockley diode PERL cell had a higher Voc of 0.706 V, a voltage loss factor VF = 0.64, and a higher fill factor of 0.828 at a higher Jsc of 42.2 mA/cm2 at 25°C [15]. According to

ADVANCED CONCEPT p/n AND p-i-n CRYSTALLINE SILICON CELLS

89

TABLE 4.2. Summary Properties of Exemplary Single-Junction Solar Cells Description

εG (eV)

η (%)

Voc (V)

Champion GaAs

1.41

28.2

PERL Si

1.1

IBC Si

FF

J sc1X (mA/cm2)

VF

Light-Generated Carriers (cm−3)

1.16

0.857

22.6

0.823

1.0 (1013)

24.7

0.706

0.828

42.2

0.642

4.9 (1012)

1.1

21.8

0.677

0.806

40.0

0.615

1.8 (1015)

HIT Si/a-Si

1.1

21.8

0.718

0.790

38.4

0.718

4.1 (1015)

Das Si/a-Si

1.1

18.8

0.694

0.744

36.4

0.676

2.3 (1015)

a-Si (unstabilized)

1.8

12.0

0.891

0.701

18.4

0.495

5.2 (1011)

a-Si (stabilized)

1.8

9.5

0.859

0.630

17.5

0.478

2.8 (1011)

Equations 4.12 and 4.13, it accomplished this with a lower minority carrier concentration of 4.9 (1012) cm−3. For comparisons, all these numbers are summarized in Table 4.2 as calculated from the known density-of-states effective mass values except for a-Si where md/mo was taken as 1.0 for lack of a better value. Thus, these table concentrations differ from those of Figure 4.5 calculated for md/mo = 0.5. At one sun concentrations, the interdigitated cell configurations certainly do have inherently lower open-circuit voltage potential than the champion versions of regular Shockley diode devices. As will be discussed below, the PERL fabrication requires photolithograpy fabrication steps that make it not economically competitive with alternate designs for one sun applications. Other standard Shockley diode one sun silicon cells, fabricated without photolithography, presently do not match the performance levels of the interdigitated type. Given the constraints of high production volumes at competitive costs, the IBC cells currently provide one of the best trade-offs of performance and costs.

4.7.2

Concentrator Point Contact and IBC Cells

Under concentrated sunlight, the higher costs of photolithography fabrication can be justified. For the latter, the interdigitated design uses about 10X thinner widths for the n+ and p+ regions (i.e., the Wn and Wp values) to reduce tranverse current problems. A champion version of this has demonstrated a conversion efficiency of 26.1% at a concentration value of 105X (Item 9, Table 1.5 [5]). The point contact cell has a larger fraction percentage of its back surface contacting the metal contacts than the one sun IBC one to be able to handle the higher currents encountered with concentration (see Fig. 4.11). Apparently, this design under ∼100X sun light intensities produces high enough photo carrier concentrations in the “i” region so

90

SOLAR CELL DEVICE PHYSICS n+ busbar

p+ busbar

n+ n+

n+ Undoped czochralski wafer

Oxide

Sunlight

Figure 4.11. The point contact solar cell configuration.

that little or no doping is needed in the wafers to avoid series resistance problems from the bulk of the wafer itself. 4.7.3

PERL Cell

The PERL cell, to date, is one of the closest approximations to an ultimate Shockley diode, single-junction cell design for low injection conditions, with fine light trapping features, extensive oxide passivations, reduced ohmic contact regions, and an Al reflecting back contact. The only major features it is missing are heterojunction barriers to its ohmic contact regions and possibly back surface texturing. Chapter 2 of the First Edition [5] discussed this device in detail with its picture even featured on this book’s cover fly sheet. A champion version has demonstrated the highest crystalline silicon efficiency so far, even approaching the 25.2% efficiency of the even more expensive GaAs single junction, one sun solar cell also of standard, low injection, Shockley diode design. However, the latter also had the additional Nobel Prize-winning double heterostructure barriers to it ohmic contact regions. Neither the Si point contact nor the PERL cell design (requiring photolithography) is cost competitive for one sun applications. The lower open-circuit voltages of the IBC and HIT cell compared to the PERL support the former being high injection devices described by Figure 4.5b where the PERL is a low injection device described by Figure 4.5a.

ADVANCED CONCEPT p/n AND p-i-n CRYSTALLINE SILICON CELLS

4.7.4

91

HIT Cell

The distinctly different structure of the HIT cell is shown in Figure 4.12 [42]. Its evolution from an a-Si cell design is evident by its similar configuration except that the middle section is replaced by a thin and low-doped n-type crystalline silicon wafer, and the top and bottom surfaces are almost identical, starting with a TCO under the top and bottom grid lines [43, 44]. These are followed by highly doped a-Si layers with p doping on the top and n doping on the bottom, and finally by undoped a-Si on each side of the crystalline silicon wafer. The textured nature of its top and bottom surfaces assists in reducing reflection losses and causes longer photon path lengths inside the device, particularly for the photons that exit the back of the cell and are directed back by reflection off a diffuse white bottom substrate. Since the bandgap of a-Si is 1.8 eV compared to crystalline silicon’s of 1.2 eV, this is the double heterostructure for which Alferov and Kroemer shared the 2000 Nobel Prize in physics, except that the latter was all in GaAs-related III–V compounds. For the III–Vs, the lattice constants of all the components usually need to match to give good performance. However, a-Si has no well-defined lattice constant and surprisingly appears to perform well as a heterostructure barrier that confines excited carriers to the crystalline silicon and away from ohmic contacts. This HIT cell has matched the IBC cell to have the highest “notable” module efficiency for a mass produced device of 21.8% but with a higher open-circuit voltage of 0.718 V, a voltage loss factor VF = 0.65, a slightly lower fill factor of 0.79, and a lower short-circuit current density of 36.9 mA/cm2 at 25°C compared with the IBC cell module [15]. Its lower current is due to the 1.8-eV bandgap a-Si p+ (minority carrier killer) layer’s and the top ITO layer’s absorption of blue wavelengths plus the light shading from it wide, screen printed resin contact grids on

Metal

Front TCO p+ (a-Si) i (a-Si)

n-type c-Si

n+ (a-Si) Back TCO

Metal

Figure 4.12. The heterojunction with intrinsic thin-layer (HIT) cell.

92

SOLAR CELL DEVICE PHYSICS

top of the cell [43]. Its relatively high fill factor is evidence of no strong trapcontrolled SCLI losses discussed below. The upper region current losses of HIT could be avoided by applying this a-Si/crystalline Si heterojunction to the IBC cell, all of whose contacts are on the rear of the cell where little light goes. At the same time, the protection it provides from ohmic contact high surface recombination velocities would be expected to be seen in higher minority carrier concentrations and thus higher opencircuit voltages. Encouraging preliminary data with increased Voc (compared to IBC) was found in preliminary experiments on similar structures [45] (see Table 4.2 entry Das Si/a-Si), but higher overall performance has not yet been demonstrated because it did not contain all the other efficiency-enhancing elements of the IBC described above. HIT’s one-dimensional symmetry band diagram is qualitatively the same as that shown in Figure 4.6 (but with heterojunction band edge offsets next to each doped region), and its Fermi level splitting corresponds to an excited carrier concentration of 4.1 (1015) cm−3 (from Eq. 4.15) that is about three orders of magnitude higher than the corresponding number for the champion PERL cell. However, its open-circuit voltage is only slightly higher because it operates as a high injection device (vs. a low injection one as illustrated in Fig. 4.5). This means that the HIT cell is better at absorbing, confining, and preserving photogenerated excited carriers than any other one sun, crystalline silicon-based solar cell device so far. One potential reason is the dual heterojunction that keeps the crystalline silicon photoexcited carriers from the high surface recombination velocity of ohmic contacts. Another possible reason is that the textured top and bottom is superior to single-sided texturing, particularly when it is placed onto a module’s white backing surface that reflects any light lost from the bottom back up into the cell. In reality, the high open-circuit voltage is likely due to a combination of these effects. The early versions of the HIT cells were formed on undoped crystalline silicon wafers over 200 μm thick. Recent reports have quoted thicknesses for HIT cells as low as 70 μm without wafer bowing problems but with monotonically falling short-circuit current densities with thickness below 200 μm [43]. These seem to imply that the champion-level performance occurs at the larger wafer thickness (i.e., ∼200 μm). With the same qualitative structure as an amorphous Si cell with a thick and undoped central core, the question arises as why the HIT and the IBC cells do not enter the regime of trap-controlled SCLI operation covered in the following section. The main reason is that high-quality, 1-ms-lifetime Czochralski Si wafers do not have the significant density of deep traps needed for trap-controlled behavior. However, the need to have 1015 cm−3 level doping for one sun operation free of notable resistive losses in interdigitated cells and the disappearance of this need when it and the point contact cells are under concentrated light then begs the question of whether a-Si cells might recover some of their lost fill factors if exposed to appropriate minimum levels of sunlight concentration, similar to select GaAs concentrator cells [27].

SCLI MODEL OF SOLAR CELLS

4.8

93

SCLI MODEL OF SOLAR CELLS

In the same year that Chapin et al. [11] reported a 6% efficiency for a Shockley diode single-crystal silicon solar cell, Reynolds et al. [46] reported the same 6% efficiency for what they then described as CdS solar with a Cu contact. However, the subsequent development of this type of cell showed that its higher performance configurations have a quantum efficiency that starts at ∼1-μm wavelength that corresponds to the 1.2-eV bandgap of CuxS, sharply different from the 2.4-eV bandgap (at 0.5-μm wavelength) of the CdS other part of this heterojunction device [47]. What is most distinctive about this cell is the crossover of its measured dark and light J–V curves as shown by the data points in Figure 4.13 for a device with a 6% efficiency [48]. This is in sharp contrast to the vertical only dark-light shift of high-performance Shockley solar cells as shown for the champion GaAs cell in figure 1.2 of the First Edition [5] and as specified by this model’s Equation 4.16. There are no physically realistic parameters values that one can use in Shockley diode solar cell equations to fit these measured crossover data. However, these data were well fit with a trap-controlled SCLI model as shown by the theoretical solid line curves in Figure 4.13 using reasonable parameter values that are in the range of independently measured values obtained from 20 Theory Experimental Data

15 10

Dark

J (mA/cm2)

5 0 –5 –10

Light

–15 –20 0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

V (V)

Figure 4.13. The dark and light J–V crossover properties of a 6% efficient, thin-film polycrystalline CuxS/CdS solar cell [48, 49].

94

SOLAR CELL DEVICE PHYSICS

other component materials using nonsolar cell characterization techniques [48, 49]. Like a-Si, CuxS/CdS solar cells degrade with sunlight exposure in the terrestrial atmosphere and can recover with heat treatment [47]. Unfortunately, their degradation continues unabated and is not limited to a small percentage like a-Si. Furthermore, the CuxS/CdS recovery heat treatment requires a hydrogen-reducing atmosphere (and possibly oxygen also), while a-Si requires only heat. Eventually, a CuxS/Cd(Zn)S version of this cell provided the first polycrystalline thin-film solar cell with an efficiency over 10% [50] with an open-circuit voltage of 0.84 V, a voltage loss factor VF = 0.7, a fill factor of 0.68, and a short-circuit current density of 17.8 mA/cm2. From today’s perspective, the greatest value of the CuxS/CdS solar cells is not from their ability to provide better device solutions but their validation of a theoretical model, different from Shockley’s, for a distinctly different physical mechanism for directly converting light into DC electricity. At high enough current densities, no matter how well designed and fabricated the contact and grid metalizations are in a Shockley diode type solar cell, one reaches a level where the ohmic voltage losses internal to the cell dominate over the exponential current–voltage characteristics of the abrupt p/n junction itself. This is easily seen for in-the-dark log current density versus terminal voltage plots that remain in a straight line (that is exponential) until an outward curvature appears, when the internal ohmic voltages start to dominate. Such a dark log current density versus voltage curve for an a-Si solar cell (or photodiode as termed in X-ray imagers) is shown in Figure 23.11 of Chapter 23. Here, this ohmic dominance appears at a terminal voltage of about 0.9 V and at a dark current density of about 1 mA/cm2 for this specific device. Also note that the one sun short-circuit current densities Jsc are on the order of 18 mA/cm2 in a 12% (unstablized) efficiency singlejunction a-Si device (see table 1.5 in Reference 5) and of 17.5 mA/cm2 in a 9.5% (stabilized) champion device [15]. One can work to reduce the internal series resistance by varying the thickness and properties of the ITO transparent top contact typically used for a-Si and by adding additional conducting grid lines on top of the ITO. However, as noted earlier, a-Si cells have a fundamental structural difference from the abrupt p/n junction configuration shown for the simplest Shockley diode in Figure 4.1. As shown in Figure 23.9 of Chapter 23, the a-Si solar cell structure is dominated by a ∼0.3-μm-thick intrinsic or undoped region with only thin ∼10-nm-thick n+- and p+-type doped regions, essentially the diametric opposite of an abrupt p/n junction. The most developed, trap-controlled SCLI diode model for a-Si solar cells was summarized previously [51]. A key simplifying assumption was made to keep the mathematics tractable. This is that one can find a region (the 0.3-μm-thick undoped high-resistance region) where current transport is dominated by a singlecarrier type (electrons) and where their transport is well described by drift current alone (i.e., diffusion current is negligible). This current dominance is justified in general when the mobility of conduction band electrons is quite different from that of holes, so that the higher-mobility carrier dominates. This is particularly justified in a-Si where the electron mobility is about 100 times that of its holes (Street, Chapter 23). Note that the key Shockley diode assumption, to make that mathemat-

SCLI MODEL OF SOLAR CELLS

95

ics tractable, was just the opposite. It was that a region can be found at the edges of the depletion regions where current is well described just by diffusion currents with a negligible drift current. Space charge begins to build up in this undoped amorphous layer when the velocity of electrons being injected by a negative voltage on the thin, heavily n+doped cathode exceeds the velocity with which they can be transported across the undoped region to the thin heavily p+-doped anode, as specified by its dielectric relaxation time ρε where ρ is the resistivity and epsilon ε is its dielectric constant. The threshold current density JTH where the dielectric relaxation time just equals the transit time through a resistive region of thickness L with a concentration of mobile carriers n, a mobility μ, and a dielectric constant ε is given by J TH = n2 q 2 μL ε .

(4.27)

Gauss’s law is used to describe the spatial changes in the electric field E(x) across the region. The drift-current-only expression specifies the electron concentration as n(x) = J/[qμE(x)]. In combination, these two relations define the primary differential equation that can be solved analytically for the current–voltage relationships for the case of a single trapping layer in the bandgap or solved numerically for a distribution of traps in the bandgap. To obtain rectification (and later solar cell) behavior, the n(x) boundary condition must be asymmetric with n(0) = nα (at the n-type contact), which is greater than nβ = n(L) (at the p-type contact), where L is the length of the undoped highresistance region. With a negative voltage applied to the n-type contact side of the high-resistance region, the electric field boundary condition is E(0) = J/(qμnα). The voltage injected electron concentration n(x) at each incremental position builds up so that the nonequilibrium but steady-state dielectric relaxation time through each such incremental slab exactly equals the drift velocity of the conduction electrons moving through that slab. As the current densities J (and also the terminal voltages V) continually increase, so do the trapped and mobile charge values at each and every x position under SCLI conditions. The typical transition point, for trap-controlled SCLI devices, is at a terminal voltage where pure ohmic to SCLI change begins, called the trapped-filled limit VTFL. The latter title is intended to approximately define the forward bias voltage just large enough to have the quasi-Fermi level just pass through the initial deepest trap level and to change its trapped charge probability from almost zero to almost one. This essentially is the knee where the current density begins to rise abruptly above the straight-line ohmic current density. A rough phenomenological estimate for this transition voltage is given as [52] VTFL = q L2 ptio ε ,

(4.28)

where ptio is the hole concentration in the ith trap level at equilibrium. Above this transition voltage, the resistive and trapping effects cause the current to increase nonlinearly with voltage.

96

SOLAR CELL DEVICE PHYSICS

So far, only SCLI diodes operating in the dark have been considered. For abrupt p/n junction Shockley diodes, all light absorbed in the junction “depletion” region, to produce electron–hole pairs, is assumed to be collected with 100% efficiency. Such a near-perfect collection of all photocarriers from the “i” layer of the a-Si p-i-n structure to produce a constant light-generated current apparently continues to apply relatively well. The first light exposure assumption is that if one stays at low enough voltage in general, and at zero terminal bias voltage in particular, the light-generated shortcircuit current density is well described by the Shockley diode expressions for Jsc even at high current densities. For a-Si, this means collection of all electron–hole pairs produced by light absorption in the “i” layer, but not in the n+ and p+ contact regions that are essentially dead to photocurrent generation. The second major light exposure assumption is that the charge distribution can be split into two parts, ρo(x, Φ) and ρI(x, Φ, V), where Φ is the light intensity and the “o” subscript designates the zero terminal voltage charge distribution that leads to the generation of Jsc and the “I” subscript denotes the positional change in this charge density with a voltage that controls the “injection” of charge through the device contacts. The final light exposure assumption is that the voltage-induced changes in ρI(x, Φ, V) are well approximated in the SCLI current expressions by replacing the equilibrium trap densities with their steady-state light-exposed values. Then, the dark SCLI J–V expressions become the light-exposed ones just with the addition of the constant short-circuit current density, which only varies with Φ, and the replacement of equilibrium trap densities with their steady-state light-exposed values. This SCLI model fit of the nonexponential shape of the light J–V curves for the (unstablized) 10% efficient a-Si solar cell is shown in Figure 4.14. The bandgap

0 Catalano et al. Data

J (mA/cm2)

–4

εc – εt = 0.5 eV = 0.4-0.5 eV = 0.3-0.5 eV = 0.2-0.5 eV

–8 –12 –16 –20

0

0.1

0.2

0.3

0.4 0.5 V (V)

0.6

0.7

0.8

0.9

Figure 4.14. The fit of the J–V data measured data points of a 10% efficient a-Si solar cell using the trap-controlled SCLI model and the trap densities values (round data points) of Figure 4.15 for the labeled bandgap energy ranges.

SCLI MODEL OF SOLAR CELLS

97

trap distribution densities that provide this fit are shown by the circular data points and dashed line curve in Figure 4.15. For reference, the solid line curves of the latter are the known electronic state densities inside this band or “mobility” gap from Street’s Figure 23.2 in Chapter 23. Figure 4.14 Illustrates how the lowest trap level provides the starting base to the J–V characteristic’s shape (with the highest fill factor) that then progressively fills in with each shallower trap level at higher voltage points that soften the curve and further lower the fill factor. The quasi-Fermi level essentially moves through each shallower level in approaching closer to the conduction band edge with increasing forward bias. For reference, the light J–V curve fit values for the density of states of a 9.2% efficient, polycrystalline thin-film CuxS/CdS solar cell are also shown in Figure 4.15 [53]. Both cells have a relatively high peak in the mid 1017 cm−3eV−1 trap densities 0.3–0.4 V deep below the conduction band edge that fall sharply at shallower adjacent trap energies. However, the a-Si curve intersects its band tail states and then bends back up. This softens the J–V curvature and lowers its fill factor. Apparently, the CuxS/CdS heterojuction is free of such band tails states (at least of the indicated magnitudes), and its higher fill factor of 0.71 (compared to the a-Si fill factor of 0.68) is consistent with this conclusion. The highest fill factors predicted by this SCLI model have values of approximately 0.76. They occur when there is a single deep trap in the bandgap with the equilibrium Fermi energy εFeo positioned just a few KT above it (see fig. 2.2 [52]). This implies fundamental, lower values for the achievable fill factors in solar cell devices well described by a trap-controlled SCLI model (vs. Shockley diodes ones).

Figure 4.15. The solid line curves are the density of states for amorphous silicon (from Fig. 23.2, Street, Chapter 23) as a function of energy position relative to the conduction band edge εC. The broken line curves are the trap-controlled SCLI model trap values that fit the light J–V curves of a 10% efficient a-Si cell (triangle data points) and a 9.2% CuxS/ CdS (round data points) cell.

98

SOLAR CELL DEVICE PHYSICS

For trap-controlled SCLI diodes, the theoretical values for Voc remain the same quasi-Fermi level splitting ones specified by Equations 4.12–4.15 for low or high injection and their Jsc values versus bandgap are also the same as for Shockley diodes. The only major difference is in the FF values. Fundamentally, the current density with trap-controlled SCLI cannot rise as quickly with voltage as in devices with J rising exponentially with V. This makes its shape less square and it theoretical FF lower. What is obvious for the a-Si data points in Figure 4.5b is that they have only been able to achieve rather low light-generated carrier densities up to ∼1011 cm−3 (compared to 1013 to 1015 cm−3 levels in the other champion devices). This imposes sharp relative efficiency penalties on high injection devices like a-Si despite a-Si’s higher rate of light-generated carriers in comparatively much thinner optically active layers. This low carrier concentration results from its comparatively very low minority carrier lifetimes caused by all of its bandgap states and defects and by its poorer surface passivation and its lack of heterojunction barriers to ohmic contact regions of very high surface recombination velocities. These deficiencies are present in various degrees in other defect-filled thin-film devices. One additional important implication of the SCLI solar cell model is that any bandgap material can be made to perform as a solar cell, at least to some degree, by placing it between two asymmetric contacts.

4.9 BANDGAP VOLTAGE, UNDOPED QUANTUM WELL SOLAR CELLS Shockley and Queisser [12] used a detailed balance calculation to estimate the maximum efficiency of a crystalline silicon, abrupt p/n junction solar cell described by their model equations. They took the black body radiation from the sun (appropriately represented geometrically) minus reverse black body radiation from a silicon cell surface near room temperature and used the net flux of photons above the silicon bandgap to come up with a short-circuit current density, an open-circuit voltage, and a corresponding fill factor to determine a maximum efficiency of 30% essentially for a cell located above the earth’s atmosphere. Since terrestrial cells are typically a relative 10% higher efficiency than space cells (due to less blue light photons), this corresponds to a maximum terrestrial efficiency on the order of 33%. They readily conceded that concentrated sunlight could be used to produce larger net photon fluxes from the sun and even higher efficiencies. Henry [54] extended the detailed balance modeling to the case of 1000X sunlight concentration levels. For a single junction, they found a maximum efficiency of 37% that increased to 50% with 2 junctions and to 72% with 36 junctions (and no interconnect losses) (p. 798 [8]). Henry noted that the ultimate maximum of 72% was less than the Carnot thermodynamic limit of ∼93% due to the light emission losses of high-efficiency solar cells. Such losses are readily evident in the highest efficiency direct bandgap solar cells with negligible nonradiative recombination losses but strong radiative light emission (LED behavior) when

BANDGAP VOLTAGE, UNDOPED QUANTUM WELL SOLAR CELLS

99

placed in forward bias with no sunlight exposure. This band edge emission definitely does not have the spectral distribution shape of black body radiation. The maximum low injection efficiencies were calculated and presented in figures 1.12 and 1.13 of the First Edition [5] for abrupt single p/n junction devices as a function of bandgap. For both unconcentrated and concentrated sunlight spectra, these maxima were in the 35–37% range for bandgaps between ∼1.1 and 1.4 eV in approximate agreement with Henry [54]. Such efficiency values apparently require sunlight concentration at least at 1.1-eV bandgaps. To stay in the low injection regime, the light-generated carrier concentrations would need to stay below the ∼1017 cm−3 range because the majority carrier doping in realistic devices are typically in the mid ∼1017 cm−3 levels. However, Figure 4.5a indicates that light-generated minority carrier concentrations would need to get the 1018 to 1019 cm−3 range for open-circuit voltages to approach bandgap level values. Unfortunately, low injection restrictions would then require 1019 to 1020 cm−3 majority carrier concentrations that essentially kills minority carrier lifetimes. In essence, low injection, abrupt p/n junctions are fundamentally blocked from achieving open-circuit voltages of bandgap magnitudes. In a typical 1017 to 1018 cm−3 doped abrupt p/n junction solar cell, one can attempt to shine enough light on it for high injection operation with light-generated carrier concentrations exceeding the doping levels and splitting the quasi-Fermi levels to near the band edges. However, such heavy doping decreases excited carrier lifetimes in semiconductors in general and in direct bandgap semiconductors in particular. For the latter, radiative recombination lifetime is determined by the Einstein coefficient and is inversely proportional to the doping level ([55]; Table 1.3 [5]). Hence, this radiative component of recombination lifetime can be decreased by many orders of magnitude when little or no doping is used. Combining this with excellent surface passivation and with heterojunctions that keep quantum well excited carriers away from high surface recombination velocity ohmic contacts, much higher carrier concentrations and levels of Fermi splitting to band edges become possible. In this case, the heterojunction n+ and p+ regions provide the collapse of the quasi-Fermi levels into single levels at opposite band edges so that the metal contacts provide actual open-circuit voltages at the band edge, equal to the splitting values of the quasi-Fermi levels in the quantum well (see Figs. 4.6 and 4.17b). An exemplary example is the same Figure 4.4 structure except that the 0.5-μm p emitter and the 3.5-μm n base are replaced by a 4.0-micron thick undoped GaAs layer of very high quality. Dividing the champion GaAs concentrator cell of 28.2% efficiency by its voltage loss factor VF = 0.823, the undoped quantum well design should be able to support efficiencies approaching 34% if its ohmic contacts can be made to accommodate the high current densities that would correspond to the sunlight concentration ratios that can provide band edge quasi-Fermi level splitting. An alternate quantum well structure is intrinsic amorphous Si surrounded by thin a-Si carbide (or silicon nitride) heterojuctions of higher bandgap that are heavily doped n and p types except for the 10–20 Å adjacent to the well as needed to keep the “minority” carriers in the well from tunneling into the very high

100

SOLAR CELL DEVICE PHYSICS

density of majority carriers in the barrier materials. To date, such heterojuction barriers to amorphous Si have only been applied to its top surface, and this does not provide a quantum well. Additionally, these could be capped by thin, heavily doped n and p a-Si to provide the final transition between the barrier layers and the metal contacts for better alignment of band edges in the well with the metal Fermi levels. The projected efficiencies of such high injection, undoped quantum well devices in tandem two-junction structures are plotted in Figure 4.16 as a function of the top cell bandgap with constant light-generated carrier concentration shown as contours. The bandgap pairs that provide matched junction–maximum power– current densities are (1.3 eV, 0.55 eV), (1.4 eV, 0.62 eV), (1.5 eV, 0.83 eV), (1.6 eV, 1.0 eV), and (1.7 eV, 1.2 eV) for efficiencies over 50%, slightly exceeding the values calculated by Henry [54]. The equilibrium and light-exposed energy band diagram is shown schematically in Figure 4.17 for a 1.7- and 1.2-eV tandem device interconnected via a p+/n+ tunnel junction with the actual structural layout shown as the box diagram at the bottom of Figure 4.17a for reference. For comparisons, 60 cm–3

1019 50

Efficiency η (%)

40 GaAa/GaSb 100X

1016 30

InP/GaInAs 50X

1013

20

1010 10 107

104

0 1.0

1.2

1.4

1.6

1.8

2.0

2.2

2.4

2.6

2.8

Top Junction Bandgap Energy εG (eV)

Figure 4.16. The theoretical efficiency for current matched, two-junction solar cells under high injection AM1.5D sunlight for constant light-generated carrier concentration values plus experimental data points. The solid line curve corresponds to the highest achieved concentration levels listed in Figure 4.5 and Table 4.2.

BANDGAP VOLTAGE, UNDOPED QUANTUM WELL SOLAR CELLS

101

Figure 4.17. Two-quantum well, high injection solar cell diodes of two different well bandgaps, interconnected by a tunnel junction for the production of open-circuit voltages that are a larger fraction of the bandgap values to provide higher conversion efficiencies. In (b), the bandgap voltage condition is shown with the split of the quasi-Fermi levels just beyond the band edges for an open-circuit voltage just greater than the sum of the two-well bandgaps equaling 2.9 V. Diagrams are proportioned for clear labeling of features and are not to scale.

performance data points for two champion tandem dual junction cells are shown to indicate the current state of the art (for low injection, abrupt p/n junction devices). As one indication of the practicality of such band edge quasi-Fermi splitting operation is comparison to high-efficiency LEDs with conversion efficiencies of

102

SOLAR CELL DEVICE PHYSICS

over 50% (narrow band light ouput power divided by DC input power) [56]. They achieve such bandgap splitting both with DC voltages applied to the LED contacts and also by optical pumping [57]. LEDs operate in just the opposite way from solar cells—DC electricity goes in and visible light comes out [58]. The most efficient LED conversion one can imagine is the application of a voltage just greater than the quantum well bandgap so that every electron that enters a LED quantum well combines with a valence band hole to produce a photon with the same magnitude electron volt energy as the applied voltage. This is the reverse analogue for the most efficient conversion of fixed electron volt photons into electrons flowing into the solar cell contacts at the corresponding voltage V described earlier in this chapter. Even though high-efficiency LEDs typically have many more quantum wells of much lower thickness than solar cells, all of the same bandgap, with no tunnel junction electrical interconnects, and various narrow bandwidth features like quantum dots and distributed Bragg structures [59, 60], they still provide existence proof that bandgap quasi-Fermi splitting is possible optically in quantum well structures. The 50% solar cell efficiency challenge will be to see how high the light-generated carrier concentrations can be raised as wells are made with no doping, with extremely low defect levels, and with nearly perfect boundaries to heterojunctions and to surface passivation layers. One limit will be the maximum current densities and light concentration ratios that can be tolerated by the interconnect tunnel junctions even when they are fabricated by highly precise techniques like molecular beam epitaxy. An estimate of the maximum theoretical conversion efficiencies of solar cells with many junctions all operating with open-circuit voltages of bandgap magnitudes is obtained by taking the thermodynamic limit (Carnot efficiency maximum ∼93%) and subtracting the unavoidable losses loss due to re-emmision of (LED) light (as recommended by Henry [54]). Considering undoped direct band p-i-n gap quantum well solar cells whose dark currents are dominated by radiative recombination, the worst case loss corresponds to an open-circuit voltage where all (100%) of the absorbed optical energy is lost due to radiative recombination. This “ideal” device is then operated at a maximum power voltage, Vm, which lowers its “dark current” and its corresponding radiative recombination losses by more than 95% (see such >95% reduction in dark current in the champion cell with a fill factor of 85% of figure 1.2 of the First Edition [5]). This indicates that theoretical maximum efficiencies are in the 85–90% range that falls to the ∼80–85% levels after downward adjustments for realistic material parameter values versus “ideal” ones (see [5]).

4.10

CONCLUSIONS

Fifty years of quantum mechanics theory development followed by 50 years of its application to solid-state electronic device problems have led to solar cell efficiency performance increasing from ∼1% to over 40% currently. Solar cell theory

CONCLUSIONS

103

continues to predict substantially higher efficiencies when high excited carrier concentrations are considered. As existence proofs, LEDs have achieved excited carrier concentrations at the high levels (that penetrate band edges at ∼1019 cm−3) where greatly improved performance is projected for solar cells. These high carrier concentrations correspond to diode operation where the quasi-Fermi levels split to the bandgap edges. They also correspond to terminal voltages that approximately equal the bandgaps of the photoactive regions. In retrospect, it is intuitive that such operation corresponds to the highest possible conversion efficiency. However, no current champion solar cells have come close to demonstrating open-circuit voltages at bandgap levels. The associated deficit is directly described by the voltage fraction VF shown in Figure 4.5 and identifies a promising area for substantial future device performance improvements. Prior modeling has projected 50% terrestrial efficiencies for two-junction cells at 1000X sunlight concentration that rise to 72% for 36 tandem junctions, ignoring the losses in the interconnect tunnel junctions [8, 54]. Bandgap open-circuit voltage values should assist with the challenges of exceeding 50% efficiency levels. The last 15 years has provided wide technical and market experience with solar cell configurations that depart dramatically from the traditional Shockley diode abrupt p/n junction geometry with one-dimensional symmetry. These include the a-Si, the point contact, the IBC, and the HIT solar cells. All of these except the point contact one are currently commercially successful products. And all but the first use crystalline silicon as the primary photo active component. All use textured surfaces to enhance light absorbance and trapping, and all use high doping of the regions next to the metal contacts to assure good ohmic contact formation and Fermi level alignments with the nearest band edges. Except for the HIT cell, none have applied the Nobel Prize-winning concept of a quantum well heterojunction barrier to shield excited carriers in a photoactive region from the high recombination properties of ohmic contacts. Combining the lessons learned from analyzing their behavior with the insights gained from LEDs, it appears that abrupt p/n junctions are not required for the highest efficiency performance. Indeed, under low injection limits, one is fundamentally blocked from achieving such Fermi levels bandgap splitting performance. So far, the non-abrupt configuration has not been applied to the multijunction III–V devices that currently provide the 40% efficiency levels of performance. They achieved their improvements mainly by reducing the losses of absorbed photon energy that lie above the bandgap of the absorbing material. The opportunities for improvement in improved voltage fraction VF arise from reducing the photon energy losses below the bandgap. The latter involves two components. One is the quasi-Fermi level bandgap splitting from high excited carrier concentrations. Direct bandgap materials are likely preferred due to their comparatively higher absorption coefficients to provide high light absorption in thinner layers and their absence of significant non-radiative recombination losses. The other is using heavily doped semiconductor regions next to the metal contacts to align the metal Fermi level to the associated band edge in the solar cell. The latter becomes particularly effective when implemented with the quantum well

104

SOLAR CELL DEVICE PHYSICS

heterojunction barriers. In addition, suggested options are recommended improvements to the crystalline silicon configurations for higher performance, mainly by implementing the quantum well and heterojunction barrier configuration pioneered with HIT solar cells. Summarizing the rationale for p-i-n quantum well concentrator solar cells, if the light-generated carrier concentrations can be made to equal or to exceed the standard doped semiconductor concentration levels, there is no longer any need for doping. Once generated, carriers have no memory of coming from either thermal ionization of dopants or from optical absorption of photons and thus reduce the resistive effects by equivalent amounts. However, elimination of doping can greatly increase the excited carrier lifetimes and the associated lightgenerated carrier concentrations, particularly when all cell boundaries are passivated or have heterojunction barriers shielding excited carriers from ohmic contacts’ high surface recombination. The highest possible efficiency conversion occurs by the absorption of photons with energies just above the quantum well bandgap that produce carriers that exit the solar cell with a voltage equal to the well bandgap value. Such voltage generation in the metal contact requires the metal Fermi levels to align with the well band edges, which only occurs at high doping levels and when the high light-generated carrier concentration splits the quasi-Fermi levels all the way to the band edges. Such highly doped semiconductor layers are typically optically dead and should thus be kept as thin as possible to reduce their light absorption losses. Carrier transport parameters fundamentally reduce the power delivered out of the cell by the fill factor FF ratio. However, much higher conversion efficiencies should be made possible than any current champion devices exhibit, as open-circuit voltages are made to approach quantum well bandgap values.

ABBREVIATIONS A—solar cell junction area a—atomic spacing of a semiconductor Al—aluminum AlGaAs—aluminum gallium arsenide AM1.5D—1.5 air mass direct component of sunlight a-Si—amorphous silicon ASTM—American Society of Testing and Materials c—speed of light in free space CdS—cadium sulfide CdTe—cadmium telluride CIGS—copper indium gallium diselenide Cu—copper CuxS—copper sulfide DC—direct current

ABBREVIATIONS

105

EN—black body radiation energy of N photons of frequency f E(x)—electric field as a function of position x exp—exponential f—frequency of a light photons fe—hyperbolic function describing effects of minority carrier electron surface recombination FF—fill factor fh—hyperbolic function describing effects of minority carrier hole surface recombination FSF—front surface field F(λ)—light flux intensity of wavelength λ f(ε)—probability that a quantum state of energy ε is occupied by an electron GaAs—gallium arsenide Ge—germanium h—Planck’s constant HIT—heterojunction with intrinsic thin layer i—intrinsic I—current IBC—interdigitated back contact ITO—indium tin oxide J—current density of electrons and holes J0—Shockely diode reverse saturation current density of electrons and holes J0e—electron contribution to the Shockely diode reverse saturation current density J0h—hole contribution to the Shockely diode reverse saturation current density J L1X —one sunlight-generated current density J Lc—concentrator light-generated current density JF—current density factor of J sc J scmax JL—light-generated current density from electrons and holes Jm—maximum power current density Jn—current density of electrons Jsc—short-circuit current density J scmax—light-generated current density when external quantum efficiency equals one JTH—threshold current density in SCLI devices k—Boltzmann’s constant K—quantum wave number from Schrodinger’s equation solutions L—width of the intrinsic or low-doped region ln—natural logarithm Mc—number of equivalent minima at the conduction band edge mde—electron density-of-states effective mass mdh—hole density-of-states effective mass md/mo—density-of-states effective mass ratio n—density of electrons n0—equilibrium electron carrier density ni—intrinsic carrier density nα—electron concentration boundary condition at the forward bias electron injecting contact

106

SOLAR CELL DEVICE PHYSICS

nβ—electron concentration boundary condition opposite the forward-bias electron injecting contact N—number of photons of frequency f emitted from a black body NA—hole (or acceptor) doping density NC—effective density of conduction band states ND—electron (or donor) doping density Nv—effective density of valence band states p—density of holes p—momentum of a photon p0—equilibrium hole carrier density Pin1 X —unconcentrated one sun optical input power to the solar cell Pinc —concentrated optical input power to the solar cell PERL—passive emitter, rear locally diffused p-i-n—p-type-intrinsic-n-type structure Pin—optical input power to the solar cell p/n—abrupt junction from p-type to n-type structure Pt.—point ptio—hole concentration in ith trap at equilibrium in a SCLI device QE(λ)—quantum efficiency of the solar cell at wavelength λ q—charge of an electron R—solar cell series resistance R—specific solar cell series resistance equal to series resistance R times junction area A r(λ)—the light reflectivity of the emitter surface at wavelength λ SCLI—space charge-limited current Si—silicon SiO—silicon oxide T—absolute temperature tB—solar cell base width t B′ —undepleted solar cell base width TCO—transparent conducting oxide tE—solar cell emitter width t E′ —undepleted solar cell emitter width TiO2—titanium dioxide UV—ultraviolet vde—diffusion velocity of electrons vdh—diffusion velocity of holes VF—voltage factor of qVoc/εG Vm—maximum power voltage Voc—open-circuit voltage VTFL—trapped-filled limit voltage in SCLI devices X—sunlight concentration ratio xB—base depletion region width xE—emitter depletion region width V(r)—potential energy function value at a space vector position r

REFERENCES

107

Wn—width of n contact region of IBC solar cell Wp—width of p contact region of IBC solar cell xn—x position of emitter depletion region edge xp—x position of base depletion region edge xw—x position of base boundary next to its metal contact Zn—zinc ΔnL—light-generated change in electron concentration ΔpL—light-generated change in hole concentration ε—dielectric constant εC—energy level of conduction band edge εF—Fermi energy Δε LFe—light-induced change in electron quasi-Fermi energy Δε LFh —light-induced change in hole quasi-Fermi energy δεn—electron energy loss moving from conduction band edge to metal Fermi level δεp—hole energy loss moving from valence band edge to metal Fermi level η—efficiency ηC—concentrator efficiency ε∗Fe—quasi-Fermi energy of electrons εF0—equilibrium Fermi energy ε*Fh—quasi-Fermi energy holes εG—bandgap energy εV—energy level of valence band edge λ—wavelength of a photon of frequency f ρo(x, Φ)—charge distribution as a function of position x in a SCLI device exposed to a light intensity Φ at zero bias voltage ρI(x, Φ, V)—the change in charge distribution as a function of position x in a SCLI device exposed to a light intensity Φ due to the application of a charge injecting voltage V μn—electron mobility π—the universal mathematical constant “pi” ΦK(r)—electronic wave function of quantum number K is a vector position in space r Φ—light intensity in a SCLI device ∇—mathematical operation “del”

REFERENCES [1] [2] [3] [4]

N. Herbert. Quantum Reality. New York, Doubleday (1987). Nobel Prizes. Available at http://nobelprize.org/. Accessed April 3, 2009 (2009). Bohr Model (1913). Wikipedia available at http://en.wikipedia.org/wiki/Bohr_ model. Accessed May 2, 2010. L. Partain and J. Sheldon. Quasi-Fermi level interpretation as potential energy. Lawrence Livermore Laboratory Report No. UCRL-83787, Lawrence Livermore National Laboratory, Livermore, CA, p. 16 (1980).

108 [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29]

[30] [31] [32] [33] [34] [35] [36]

SOLAR CELL DEVICE PHYSICS L. Partain, ed. Solar Cells and Their Applications. New York, Wiley (1995). D. Neamen. Semiconductor Physics and Devices, 3rd edition, pp. 25, 83–114, 327, 458. Boston, McGraw-Hill (2003). C. Kittel. Thermal Physics. New York, Wiley (1969). S. Sze. Physics of Semiconductor Devices, 2nd edition. New York, Wiley (1981). W. Shockley. Electrons and Holes in Semiconductors. Huntington, NY, Krieger (1976). J. Bardeen, Phys. Rev. 71, 717 (1947), and W. H. Brattain and W. Shockley, Phys. Rev. 72, 345. D. Chapin, C. Fuller, and G. Pearson. J. Appl. Phys. 25, 676–677 (1954). W. Shockley and J. Queisser. J. Appl. Phys. 32, 510–519 (1961). J. Hovel. Solar Cells, Semiconductors and Semimetals, Vol. 11. New York, Academic (1975). J. F. Geisz, D. J. Friedman, J. S. Ward, A. Duda, W. J. Olavarria, T. E. Moriarty, J. T. Kiehl, M. J. Romero, A. G. Norman, and K. M. Jones. Appl. Phys. Lett. 93, 123505 (2008). M. Green, K. Emery, D. King, Y. Hishikawa, and W. Warta. Prog. Photovoltaics Res. Appl. 15, 35–40 (2007). C. Kim and R. Schwartz. IEEE Trans. Electron Devices 16, 657–663 (1969). R. Schwartz and M. Lammert. Int. Electron Devices Mtg. 21, 350–352 (1975). R. Schwartz and M. Lammert. IEEE Trans. Electron Devices 24, 337–342 (1977). D. Carlson and D. Wronski. Appl. Phys. Lett. 28, 671–673 (1976). T. Ashley and C. Elliot. Semicond. Sci. Technol. 6, C99–C105 (1991). A. Prokopyev and S. Mesheryakov. Solid State Electron. 43, 1747–1753 (1999). O. Nichiporuk, A. Kaminski, M. Lemiti, A. Fave, and V. Skryshevski. Solar Energy Mater. Solar Cells 86, 517–526 (2005). A. Al-Omar. Solid State Electron. 50, 1656–1666 (2006). Z. Alferof, V. Andreev, and V. Rumyantsev. III-V solar cells and concentrator arrays. In High-Efficient Low-Cost Photovoltaics, V. Petrova-Koch, R. Hezel, and A. Goetzberger, eds., pp. 101–133. Berlin, Springer (2009). A. Rai, R. Bhattacharya, and Y. Park. Thin Solid Films 114, 379–398 (1984). L. Partain. J. Appl. Phys. 62, 4003–4005 (1987). L. Partain and D. Liu. Appl. Phys. Lett. 54, 928–930 (1989). N. Xu. The physical origin of prebreakdown electron “pin-holes.” In High Voltage Vacuum Insulation, R. Latham, ed., pp. 124–129. London, Academic (1995). N. Kaminar, D. Liu, H. MacMillan, L. Partain, M. Ladle Ristow, G. Virshup, and J. Gee. Concentrator efficiencies of 29.2% for a GaAs cell and 24.8% for a mounted cell lens assembly. In 20th IEEE Photovoltaic Specialists Conference, Las Vegas, NV, pp. 766–768. New York, IEEE (1988). D. King and B. Hansen. Validation of PDML and PTEL Primary Reference Cell Calibrations, Sandia National Laboratories Internal Memo (1991). M. Green, K. Emery, Y. Hishikawa, and W. Warta. Prog. Photovoltaics Res. Appl. 17, 85–94 (2009). H. Kroemer. Rev. Modern Phys. 73, 783–793 (2001). C. Hamaker. J. Appl. Phys. 58, 2344–2351 (1985). M. Klausmeier-Brown. Status, prospects and economics of terrestrial, single junction GaAs concentrator cells. In Solar Cells and Their Applications, L. Partain, ed., pp. 125–139. New York, Wiley (1995). Z. Fang and D. Look. J. Electron. Mater. 22, 1361–1363 (1993). V. Andreev, V. Khvostikov, V. Larionov, V. Rumyantantsev, E. Paleva, M. Shvarts, and C. Algora. 5800 suns AlGaAs/GaAs concentrator solar cells. In Proc. International

REFERENCES

[37] [38] [39] [40] [41] [42] [43] [44] [45]

[46] [47] [48] [49] [50]

[51] [52] [53] [54] [55] [56] [57]

[58] [59] [60]

109

Photovoltaic Science and Engineering Conference, pp. 147–148. Sapporo, Japan (1999). R. Swanson. Solar Cells 17, 85–118 (1986). R. Swanson. Device physics for back-side contact solar cells. 33rd IEEE Photovoltaic Specialists Conference, San Diego, CA, May 11–16, 2008. New York, IEEE (2008). W. Mulligan, M. Cudzinovic, T. Pass, D. Smith, N. Kaminar, K. McIntosh, and R. Swanson. SunPower Corporation, assignee, Solar cell and method of manufacture. U.S. Patent 7,339,110 B1, March 4 (2008). E. Van Kerschaver and G. Beaucarne. Prog. Photovoltaics Res. Appl. 14, 107–123 (2006). N. H. Fletcher. Proc. IRE 43, 551–559 (1955). M. Green. Solar Energy 74, 181–192 (2003). Y. Tsunomura, Y. Yoshimine, M. Taguchi, T. Baba, T. Kinoshita, H. Kanno, H. Sakata, E. Maruyama, and M. Tanaka. Twenty-two percent efficienct HIT solar cell, Solar Energy Materials & Solar Cells, 93, 670–673 (2009). M. Taguchi, H. Sakata, Y. Yoshimine, E. Maruyama, A. Terakawa, M. Tanaka, and S. Kiyama. An approach for the higher efficiency in the HIT solar cells. In 31st IEEE Photovoltaic Specialists Conference, Orlando, FL, pp. 866–871. U. Das, S. Bowden, M. Burrows, M. Lu, and R. Birkmire. Effect of surface passivation on Si heterojunction and interdigitated back contact cells. 17th Workshop on Crystalline Silicon Solar Cells & Modules: Materials and Processes, Vail, CO, August 5–8, 2007. D. Reynolds, G. Leies, L. Antes, and R. Marburger. Phys. Rev. 96, 533–534 (1954). L. Partain. J. Appl. Phys. 54, 5218–5226 (1983). R. Moorthy, L. Partain, D. Okubo, and D. Henderson. Space-charge-limited current and capacitance in CuxS/CdS solar cells. In IEEE International Electron Devices Meeting, Washington, DC, December 1979, pp. 442–445. New York, IEEE (1979). L. Partain, G. Armantrout, J. Leong, and P. Warter. J. Electron. Mater. 9, 467–484 (1980). A. Catalano, R. Arya, C. Fortman, J. Morris, J. Newton, and J. O’Dowd. High performance, graded bandgap a-Si : H solar cells. In 19th IEEE Photovoltaic Specialists Conference, New Orleans, LA, April 5, 1987, pp. 1506–1507. New York, IEEE, (1987). L. Partain. J. Appl. Phys. 61, 5458–5466 (1987). M. Lampert and P. Mark. Current Injection in Solids. New York, Academic (1970). L. Partain. J. Appl. Phys. 63, 1762–1767 (1988). C. Henry. J. Appl. Phys. 51, 4494–4500 (1980). R. Nelson and R. Sobers. J. Appl. Phys. 49, 6103–6108 (1978). J. Baur, F. Baumann, M. Peter, M. Engl, U. Zehnder, J. Off, V. Kuemmler, M. Kirsch, J. Strauss, R. Wirth, K. Streubel, and B. Hahn. Phys. Status Solidi C 52, S905–S908 (2009). A. Laubsch, M. Sabathil, M. Strassburg, W. Bergbauel, M. Peter, H. Laubauer, N. Linder, B. Hahn, and K. Streubel. Improving the high current efficiency of LEDs. Available at http://spie.org/documents/Newsroom/Imported/1476/1476_5485_0_ 2009-04-09.pdf. Accessed June 6, 2009 (2009). N. Holonak. Am. J. Phys. 68, 864–866 (2000). A. Erchak, D. Ripin, S. Fan, P. Rakich, J. Joannopoulos, E. Ippen, G. Petrich, and L. Kolodziejski. Appl. Phys. Lett. 78, 563–565 (2001). G. Zhang, Z. Yang, Y. Tong, Z. Qin, X. Hu, Z. Chen, X. Ding, M. Lu, Z. Li, T. Yu, L. Zhang, Z. Gan, Y. Zhao, and C. Yang. Opt. Mater. 23, 183–186 (2003).

PART II TERRESTRIAL SOLAR CELL ELECTRICITY

5 CRYSTALLINE SILICON SOLAR CELLS AND MODULES LEONID RUBIN Day4 Energy Inc.

5.1

INTRODUCTION

During the last decade, the PV industry has demonstrated an impressive 30–50% annual growth rate and, as a result, in 2008, the total production capacity has surpassed 4 GW with cumulative annual revenues reaching $20 billion. There are several reasons for this success. Initially, in Japan and later in Germany, government-supported programs for the PV industry development played extremely important roles for attracting capital investments for the PV industry. Government subsidy programs helped bridge the gap in cost between PV-generated electricity and conventional electrical grid prices, thus creating a viable marketplace for PV product manufacturers. There were also a number of other factors that have contributed to the rapid growth of the PV industry. The sudden decline in the microelectronic industry due to the Internet bubble and the availability of excess silicon feedstock production capacity provided silicon feedstock for the PV industry. The growing concern about global climate change due to greenhouse gas emission combined with oil and gas price rises suggested that the development of alternative energy sources should become an important part of government policy. Substantial progress was also made in PV cell efficiency, and PV cells were being successfully mass produced as well. One may say that it was a unique “once-in-a-life-time” opportunity for the PV industry and so, the PV companies took full advantage of it. Nevertheless, the electric energy produced with solar cells continues to be noncompetitive with traditional nonrenewable sources. The PV industry is still strongly dependent on subsidies, and PV provides a negligible contribution to the Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

113

114

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

overall energy generation market. It was expected that the economy of scale would eventually result in cost reduction sufficient to make the PV industry cost competitive with conventional sources of electrical power. There is no doubt that in recent years, some cost reduction associated with PV cells and module mass production has been achieved. But this progress has not been sufficient so far to make the PV industry cost competitive. One of the key roadblocks to further cost reduction is in the overall lack of technological innovation in PV cell and PV module manufacturing. Ironically, the very subsidy system that resulted in PV product demand explosion is also partially responsible for the lack of technological innovation during this same period. With demand exceeding supply by a wide margin, manufacturers have focused their efforts and capital on expanding production capacity, economies of scale, and profit. Crystalline silicon technology development has been largely “on the back burner.” The fact is that the majority of PV cell manufacturers are utilizing identical production technologies. Without manufacturing technology improvements, the difference between the efficiencies of mass-produced and advanced PV cells has been increasing in the last decade and is now pronounced. Even less diversity is found between PV module producers. They continue to use 35-year-old tabbing and stringing technology. In this chapter, we will present a brief overview about the design and performance of conventional mass-produced crystalline silicon PV cells and the main factors limiting their light-to-electric energy conversion efficiency. We will describe novel technologies that can lead to PV cell efficiency improvement. We will then describe some advanced PV cells with outstanding efficiency. The current status of PV module fabrication technologies and the possibilities for their improvement will be described. Finally, the optimization of PV cell and module designs to maximizing annual kilowatt hour generation capacity for PV power generation systems will be addressed.

5.2

INDUSTRIAL CRYSTALLINE SILICON PV CELL

Since the PV industry does not require the same silicon material purity as for microelectronics, it was initially possible to utilize the waste from microelectronic silicon production. This is referred to as solar-grade silicon for initially monocrystal and later for mc silicon wafer production. The most widely used technology for making monocrystal silicon is the CZ. For CZ silicon crystal growth, a silicon monocrystal seed is dipped into a crucible of molten high-purity silicon and is withdrawn slowly pulling a cylindrical monocrystal. This silicon crystal is then sliced into thin CZ wafers that are used for CZ PV cell production. Production of mc silicon typically involves a casting process in which molten silicon of slightly lower grade, referred to as solar-grade silicon, is directly cast into a typically squared mold where it is slowly crystallized into 3D ingots that are then sliced into square mc wafers. The temperature cooling gradient profile is one of the main factors that determine the impurity distribution divided between the

INDUSTRIAL CRYSTALLINE SILICON PV CELL

115

silicon microcrystal and the boundaries between them. Segregation of impurities into the grain boundaries is referred to as gettering and influences the quality of the mc wafers. It is known that the PV cells produced from CZ silicon typically have higher efficiency than the PV cells from mc material. At the same time, the throughput from the CZ process is about two times lower compared with the mc process, and the mc process consumes less energy and material with a lower cost. Therefore, the price of mc silicon wafers is lower for PV cell manufacturers, but there is a compromise leading to lower PV cell conversion efficiency. Recently, there has been substantial progress in both single-crystal and mc process technology leading to higher cell efficiencies for both. Currently, the market share of mc PV cells is slightly higher compared with CZ. Due to growing demand from the PV industry, a new solar-grade crystalline silicon production capacity has been established based not only on conventional solar grade but also on some new metallurgical silicon purification technologies. However, while the silicon material availability has increased to meet the PV industry demand, unfortunately, the PV cell efficiency produced from metallurgical silicon has been reduced. It is now evident that production of high-efficiency and low-cost CZ or mc PV cells must be based on sufficiently good quality initial feedstock silicon material. Conventional crystalline silicon PV cells are generally produced from p-type monocrystal or mc semiconductor wafers of 150- to 300-μm thickness. The front side of industrial PV cells is typically doped with phosphorus, thus creating an n-doped area of 0.1- to 2-μm thickness and a resistivity typically of 50–65 Ω per square. It is also possible to produce PV cells from monocrystalline n-type crystalline silicon with a boron-doped p-type front surface. These n-type PV cells do not suffer from light-induced efficiency degradation as often happens with monocrystalline p-type PV cells. The junction between the n-doped surface area and the p-doped bulk silicon creates a charge separation region with a strong dipole electric field. The surface area is referred to as the emitter and the bulk region is referred to as the base. When a PV cell is illuminated by light, photons produce electron–hole pairs and the dipole electric field provides for a separation of these charges. This displacement of free charges results in a voltage difference between the p and n regions of the PV cell. The emitter technical characteristics such as thickness, doping profile, and doped element concentration in the PV cell surface region are extremely important for PV cell efficiency. Conventional PV cells are typically equipped with current collecting metal contacts on the front and rear sides provided for conducting electric current when the p and n regions are connected through an external load. Figure 5.1 shows a drawing of a typical square mc silicon cell showing the front-side collection metallization pattern comprising screen printed fingers and two bus bars that collect electric current from the PV cell. The conversion efficiency of solar energy into electric energy is considered to be the main PV cell characteristic. Under illumination, PV cells generate a maximum voltage value referred to as the open-circuit voltage or Voc when the

116

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

70

Bus bar

Bus bar

Fingers

Figure 5.1. Top view drawing of a typical 156-mm2 MC silicon cell.

external circuit is open. If the external circuit is shorted, then the light-generated electric current reaches its maximum value referred to as the short-circuit current or Isc. The dependence of the current I versus the voltage V for different values of the external load is known as the I/V curve where the values of Isc and Voc are the intersection points with the current and voltage axis, respectively. Unfortunately, the maximum value of generated electric power is not the product of Isc × Voc. This maximum is never achieved due to inevitable power losses. The real value of generated power may be evaluated from the experimental I/V curve containing the so-called maximum power point referred to as Pmpp, at which I, V, and P reach their maximum values Pmpp = Impp × Vmpp. The ratio between Immp × Vmmp/ Isc × Voc represents PV cell general power losses and is referred to as the fill factor (FF). The PV cell energy conversion efficiency η is expressed as H = Voc × Isc × FF E , with FF being the fill factor and E being the light energy in watt per PV cell area.

5.3

EFFICIENCY LIMITATIONS

Although the theoretical efficiency of a crystalline silicon PV cell approaches 29% and the world record for the best silicon PV cell is 24.3%, the average efficiencies

EFFICIENCY LIMITATIONS

117

of a typical industrial monocrystalline or mc PV cells are 17% and 16%, respectively. There are several causes limiting PV cell efficiency. First of all, there are limitations based on the fundamental properties of silicon semiconductors. Photons with energies less than 1.12 eV are lost due to the silicon semiconductor band gap, and photons with energies exceeding 1.12 eV loose energy via dissipation into heat. The maximum value of the PV cell open-circuit voltage Voc is substantially lower than the silicon semiconductor band gap because it is defined by the quasi-Fermi level separation. The typical Voc values for conventional monocrystalline and mc PV cells are about 625 and 610 mV, respectively. Even high-efficiency PV cells demonstrate a Voc value not higher than 722 mV.

5.3.1

Optical Losses

The theoretical Isc maximum value for a crystalline silicon PV cell may reach 42 mA/cm2 under AM 1.5 sun irradiation. In reality, it is hard to reach this value because the Isc value of a PV cell is not determined simply by the incident solar energy intensity but instead, it depends on the fraction that is absorbed by the PV cell and converted without losses into electric energy. The main problem with crystalline silicon semiconductor material is that it absorbs light very poorly due to its high refraction index of about 3.9 and corresponding high light reflection of about 40%. The most efficient way to solve this problem is to utilize light trapping. Modern wet chemical etching and laser processing technologies allow for arranging for different types of high-efficiency textured structures like inverse pyramids or honeycomb-type structures on monocrystalline and mc PV cells. Light trapping textures can decrease light reflection to below 10%. An additional decrease in refection is achieved after antireflective dielectric coating. Sample AR coatings can consist of SiNx with a refraction index of up to 2.2. They are applied on top of the initially textured surface of the PV cell. This process can decrease the light reflection further to less then 4%. It is worth noting that most advanced technologies for achieving extremely efficient light trapping surfaces are kept as proprietary know-how. An additional optical loss is associated with the partial transmission of light in the IR spectrum. This effect is most pronounced when the PV cell thickness is less than 250 μm. Therefore, there is a need to utilize backside mirror reflective coatings, thus preventing efficiency decline due to insufficient absorption of light. The area that is occupied by the metal current collection contacts on the front side of the PV cell is referred to as the shading area. It impacts the optical losses by preventing solar radiation from reaching the surface of the PV cell and thereby generating electric current. This shading area typically occupies up to 7–10% of a PV cell’s available front surface, thus decreasing the PV cell efficiency accordingly. Later in this chapter, we will describe several promising current collecting technologies that have been developed for decreasing front-side shading.

118

5.3.2

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

Recombination Losses

The photon-generated electrical charges have to diffuse to the charge separation area and further to the current collecting contacts located on the front and rear sides of a PV cell. In general, this diffusion process encounters two main types of power losses. Charge recombination losses can occur in the semiconductor bulk and on the PV cell front and rear surfaces. Recombination in the bulk strongly depends on semiconductor impurities and crystal dislocation concentrations. These defects are responsible for energy states that function as efficient trap and recombination centers. Free electrons captured at these centers are passing to the valence band dissipating energy as heat. This type of recombination is typical for mc and monocrystalline PV cells and strongly depends on materials purity. In the case of mc material, bulk recombination may be reduced by impurity gettering to the microcrystal boundaries during casting crystallization and/or by bulk passivation, for example, with hydrogen during the SiNx AR coating application. It is obvious that PV cells produced from metallurgical silicon are particularly exposed to this type of recombination, thus demonstrating substantially lower efficiency. The recombination on the PV cell surface depends on the density of defects on the surface due to silicon crystal edge breakdown, as well as the presence of the metal current collecting contacts, and dopant concentration on the surface. Several technological methods have been developed and introduced into mass production to minimize surface recombination, which may be referred to as surface passivation. Charge recombination on the front and rear sides of a PV cell may be substantially reduced by passivation with a thin layer of a dielectric material, such as SiO2, SiNx, or SiC, by employing industrially available technologies and equipment [1, 2]. Recombination on a p-type PV cell rear surface may also be decreased by doping, for example, with boron or aluminum, thus creating a p+ layer or BSF.

5.3.3

Resistivity Losses

Conventional high-quality silicon monocrystalline PV cells of 156-mm2 area with optimized optical loss minimization typically generate Isc values of up to 8.5 A. Keeping in mind that the Voc value remains practically at the relatively low level of 625 mV, it is a challenge to collect this current with minimum power losses (I2 × R) because the PV cell resistivity starts to have an extremely high impact on the PV cell efficiency. Overall, the PV cell resistivity includes the following main components: (1) the series resistance (R) of the current collecting pattern on the front and rear sides of the PV cell, (2) the parallel or shunt PV cell resistance (Rsh), and (3) the contact resistance between the front and back-side metal patterns with the emitter (Rem) and bulk silicon (Rbl), respectively. The typical technology for making front- and back-side metal contacts is based on a conventional screen printing process in which an electrically conductive paste, containing silver and/or aluminum powder particles, is printed through a

EFFICIENCY LIMITATIONS

119

screen onto the front and back surfaces of a PV cell. The front-side screen is typically configured to produce a plurality of thin parallel line contacts referred to as “fingers” connected typically to two or three wider lines referred to as “bus bars.” The fingers collect the electrical current from the front side of the PV cell and transfer it to the bus bars. Metal ribbons are typically soldered to the bus bars to conduct electric current to the neighboring PV cell and further to the electrical circuit. Typically, the width and the height of each finger are approximately between 110 and 120 μm and between 10 and 20 μm, respectively. This corresponds to a height-to-width aspect ratio ranging between 0.1 and 0.15. It is evident that if the ratio between finger width and height is improved by making them narrower and thicker, finger conductivity will be improved along with decreased shading. It has been demonstrated that the utilization of hot melt pastes with modern heated screen printing equipment allows for printing fingers with <70-μm width and 20-μm heights and an aspect ratio of about 0.28. The back surface of a PV cell substrate is printed in two sequential steps. Initially, a conductive paste containing a composition of silver and aluminum powder particles is screen printed and dried by heating. This first pattern consists of small areas referred to as silver pads that are to act as current collecting contacts. Afterward, a partially conductive paste containing aluminum powder particles is then spread over the entire back surface of the substrate but only partially overlapping the edges of the abovementioned silver pads. This paste is then dried by heating. The PV cell is then “fired” at high temperature in an oven. During this process, the front-side silver paste enters a silver porous metallic phase, where at least part of it diffuses through the front surface of the PV cell into the emitter area and alloys with the silicon forming a silver silicide, thus creating an ohmic contact with the emitter. The resistance of this contact depends on the emitter depth, the dopant concentration, and its profile. The depth of the emitter must be controlled very carefully in order to prevent metallic particle penetration during the firing process, thus avoiding shunting. On the rear side of the PV cell, the aluminum diffuses through and alloys with the bulk silicon, thus producing a highly doped p+ layer or BSF. At the same time, the aluminum also alloys with the silver/aluminum pads in areas where it overlaps with them. These silver/aluminum pads are collecting electrical current from the rear side of the PV cell and act as back-side electrical terminals but do not provide back surface passivation. In order to achieve low contact resistance between the metal patterns and the emitter, there is a need to have an emitter with high doping concentration in the conventional PV cell’s front surface. At the same time, realization of this requirement results in higher surface recombination losses and PV cell efficiency decline in the blue spectral region. In order to increase the conversion efficiency of solar cells that employ a conventional screen printed metallization, emitter design parameters may be optimized such that under a screen printed finger, an emitter’s doping concentration and thickness is a sufficiently high, whereas in the lightilluminated areas, the emitter’s doping concentration is lower and the thickness of the emitter is thinner. An emitter with these differing doping concentration thicknesses is generally known as a selective emitter.

120

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

Selective emitters may be produced, for example, in accordance with the following process. After the highly doped emitter diffusion step, the emitter area is selectively masked and exposed to wet chemical etching for removing silicon from the nonprotecting areas, thereby making the emitter thinner at the same time keeping the highly doped emitter under the mask areas protected. The screen printer metal patterns must be precisely aligned with the former mask-protected areas, thus securing low ohmic contact between metallic contacts and emitter and avoiding shunting. Although the use of a selective emitter has proved to be effective in improving PV cell efficiency, the implementation of a selective emitter in practice is quite complicated. Another approach to reduce surface recombination and to preserve the low resistance between the screen printed metal patterns with the emitter is based on employing a deep emitter concept. After producing the p/n junction during the doping diffusion, the PV cell is exposed to high temperature, thereby driving the dopant deep into the emitter, thus lowering its surface concentration. At the same time, the increased depth of the emitter allows the metal patterns to penetrate deeper, thus increasing the contact area with the emitter, thereby lowering the contact resistance with the emitter. The negative associated with this promising approach is the necessity to use a quite expensive prolonged high-temperature process. This method also is not compatible with mc material because it does not tolerate temperatures above 900°C. The shunt resistance Rsh can also have a significant impact on the PV cell and module performance especially under low light intensity. A low shunt resistance can lead to decreased power generation during sunrise and sunset hours. A low Rsh value may also lead to PV cell irreversible damage due to a hot spot when that cell is shaded and exposed to reversed voltage. Low-quality feedstock silicon, especially metallurgical-grade materials, provides substantial risk and can lead to PV cells with low Rsh values.

5.4

NOVEL CURRENT COLLECTING TECHNOLOGIES

5.4.1

Electroless and Photoinduced Metal Plating

Electroless plating was discovered more than 50 years ago and allows for depositing metals on substrates without the application of an electric potential but requires that the surface be autocatalytic for metal plating [3, 4]. Later, it was further demonstrated that it is possible to establish low ohmic contact between electroless plated nickel with n- and p-doped silicon after an initial seed layer of nickel is sintered at about 450°C [5]. Fortunately, the contact resistance between electroless plated nickel and the solar cell emitter is more than 10 times lower when compared with conventional screen printer fingers and with the same type of emitter, thus offering the potential for PV cell efficiency improvement. The combination of electroless nickel plating with sequential electrochemical plating allows for

NOVEL CURRENT COLLECTING TECHNOLOGIES

121

the deposition of additional layers of nickel and the creation of sandwich-type structures consisting of nickel + copper + silver, thereby increasing the metal conductivity. Another technology for increasing the thickness of the initially deposited thin and narrow metallic patterns is known as light-induced plating. It has been demonstrated that when an n-type emitter of a PV cell is positioned on top of an electrolyte and the PV cell’s back side is kept out of contact with the electrolyte but connected to a negative potential, then under illumination, the PV cell generates an electric potential that is sufficient to attract metal cations from the electrolyte solution and deposits them onto the n-type emitter [6, 7]. It is obvious that the metal deposition will happen exclusively on the emitter areas that are not masked by some protective insulating material. This approach allows for the creation of fine metallization patterns on a PV cell’s front- and rear-side surfaces. After the initial publications, the light-induced silver plating technology has been further improved and has allowed for higher-efficiency PV cells [8]. It has a high potential for use in the PV industry as well. 5.4.2

Ink-Jet Printing

Ink-jet printing is a noncontact deposition technology with high line resolutions of up to 20 μm. It has proved to be more economical and less complicated when compared with the conventional photolithography approach [9–11]. This technology may be extremely useful for masking and selective etching as an alternative to the quite expensive lithography technology commonly used in microelectronics. Unfortunately, this technology has not yet been able to compete with screen printing metallization due to strict requirements for the ink to have low viscosity in order to prevent the printing head from clogging. Unfortunately, low viscosity is achieved only when the metal particle concentration is too low to secure sufficient conductivity in the printed metal patterns. 5.4.3

Aerosol Printing

A metal aerosol noncontact printing technology was originally developed by Optomec Inc. (Albuquerque, NM). In contrast to ink-jet printing, metal-containing ink initially is atomized pneumatically or ultrasonically. Afterward, aerosol is transported to the depositing head where continuous aerosol flow may be interrupted by a mechanical shutter. The metal aerosol printer operates via a graphical user interface that allows control of the main technological and operating parameters. AutoCAD drawings may be translated into mechanical codes, thus allowing the printing of any metal grid design. After the metal ink is printed onto the PV cell surface, it is dried and heated at 300–400°C for 5–10 min. In case it is needed, the cross section of the aerosol printed fingers may be further increased by electrolytic or light-induced metal plating, thus allowing thicker current collecting metal pattern lines of less than 20-μm width. Extensive investigations at ISE

122

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

Fraunhofer have demonstrated monocrystalline PV cells with 18.3% and mc PV cells with 16.1% efficiencies [12] However, the long-term stability of PV cells with aerosol direct printed metal patterns against environmental degradation has not yet been sufficiently investigated. 5.4.4

Laser-Fired Contact

An alternative technology for rear-side metallization is based on the earlier discovery that aluminum may be alloyed with the silicon semiconductor substrate after being deposited in the form of a thin film and exposed to a temperature in the range between 577 and 660°C [13]. A thin layer of aluminum, nickel, and silver may be deposited in sequence thereafter on the aluminum surface and sintered. This multilayer coating is necessary to allow the soldering of the power collecting leads to the back side of the solar cell. This technology has been further improved at ISE Fraunhofer [14] by introducing an intermediate dielectric layer between the metal layer and the rear side of the PV cell, providing for rear-side passivation. This metallic layer is further exposed to spot radiation heating, causing a localized molten mixture of the metal layer, the dielectric layer, and the semiconductor layer. Upon solidification, these regions provide an electric contact between the semiconductor layer and the metal layer. Pulse laser sources are employed. This allows both sufficient rear-side passivation and low-resistance current collection from the PV cell rear side. Although this promising technology has been licensed to several PV cell manufacturing companies, it has not yet been introduced into mass production.

5.5

EXAMPLES OF NOVEL PV CELLS

5.5.1

Laser Buried Grid PV Cell

We will start this section with the important results that have been achieved at The University of New South Wales, Australia, by implementing electroless nickel plating technology for developing a novel concept PV cell with “laser buried grids” [15]. As shown in Figure 5.2, this technology is based on the formation of laser grooves on the front side of a PV cell with further deep emitter formation and further electroless plating of a seed nickel layer inside these grooves. In these cells, the contact resistance between the sintered nickel and the emitter is at least 10 times lower when compared with the contact resistance between the standard screen printed silver pattern and the emitter. Additional plating steps with nickel, copper, and silver layers allow the laser-formed grooves to be filled with a thick layer of metal, thus providing substantial reduction of the Rs value. BP Solar successfully introduced this technology into mass production. Since the metal plated fingers are buried inside the laser-formed grooves, the metal content in the fingers is increased, improving the grid line height-to-width aspect ratio and simultane-

EXAMPLES OF NOVEL PV CELLS

123

Plated grid metal in laser groove

n+ p - silicon

p+

Back Metal

Figure 5.2. Concept for laser grooved buried grid silicon cell.

ously reducing of PV cell shading. A grid line shading as low as 4% is achieved. The buried contact technology allows substantially improving PV cell efficiency when compared to conventional screen printed solar cells. Unfortunately, this technology failed to find a widespread application in the PV industry probably due to its complexity and due to difficulties associated with its application for thin mc silicon wafers.

5.5.2

PERC PV Cell

In 1989, the same group from University of New South Wales, Australia, developed the so-called PERC with a conversion efficiency of up to 23.2% under standard terrestrial test conditions. These PERC cells were produced from high-purity silicon and have a very high Voc of 700 mV. This is achieved via very low recombination in the bulk and on the front and rear surfaces through efficient passivation with silicon oxide layers. These cells also demonstrated a very high Isc value of 41 mA/cm2, achieved mainly due to efficient light trapping via inverted pyramid texturing and optimized AR. The rear aluminum-silicon electrical contacts with the moderately doped silicon substrate were made through the openings in the rear-side silicon oxide passivating layer, thus preserving sufficient passivation.

5.5.3

PERL PV Cell

On 1990, the same group further improved the PERC cell and demonstrated the novel PERL cell with an efficiency of 24.7%. The rear-side contacts each of 10- to 30-micron width were aligned to the center of boron-diffused p+ areas of 30- to 50-micron width preserving the distance between these contacts at about 250 micron. Under standard terrestrial irradiation testing conditions, 4-cm2 PERL p-type monocrystalline silicon cells demonstrated Voc = 706 mV, Isc = 42.2 mA/cm2, FF = 82.8%, and efficiency of 24.7% performance [16]. The main differences

124

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

between the more advanced PERL cell and the PERC are (1) the introduction of novel locally highly boron-diffused areas underneath the rear-side aluminum contacts, thus decreasing contact resistance; (2) the introduction of a deep junction that more effectively insolates the high recombination area at the surface from the bulk region; and (3) the distance between the rear-side contacts was decreased from 2 mm in the case of the PERC cell to 0.25 mm, thus decreasing rear-side lateral resistance and improving FF value on the PERL cells. The PERL cell was discussed in detail in Chapter 2 of the First Edition, and its picture was featured on the book’s cover fly sheet.

5.5.4

HIT Cell

SANYO Electric Co., Ltd. started its participation in PV about 25 years ago introducing first commercial amorphous silicon PV cell and module production. Fifteen years later, SANYO introduced the novel HIT PV cell and arranged the mass production of high-efficiency PV modules based on these HIT cells. In 2007, SANYO announced that it improved its own world record for industrially produced HIT cell efficiency achieving 22%. The HIT cell of 100-mm2 size has a unique structure as shown in Figure 4.12 in the previous chapter. The HIT cell consists of n-type high-quality monocrystalline material surrounded by ultrathin amorphous silicon layers providing for high-efficiency passivation of the HIT cell front and rear sides. Proprietary, well-protected know-how technologies have been introduced into mass production for efficient texturing of the cell’s surfaces at the micron level with special surface cleaning and protection before the ultrathin amorphous silicon layer depositions, the conductive AR deposition, and the highaspect-ratio silver pattern printing on the front and rear sides. Implementation of these technologies results in Voc values in the range from 718 to 722 mV, Isc values of 38.37–38.64mA/cm2, and FF of 78.8% and efficiencies of η = 22%. It is expected that in 2010, SANYO will start mass production of 22% efficient HIT cell and modules on production capacity over 600 MW. HIT cell performance provides an excellent illustration of the importance of PV cell surface passivation as the key factor for cell efficiency improvement.

5.5.5

IBC PV Cell

On the edge of the millennium and within just 5 years, SunPower succeeded in designing and introducing into mass production both a novel substantially improved performance high-efficiency IBC PV cell and also high-efficiency PV modules [17]. The structure of this novel IBC cell was discussed and shown in the previous chapter (Section 4.7.1). Here are the main milestones of SunPower success. On May 12, 2003, NREL verified 20.4% conversion efficiency for the 125-mm semi-square, single-crystal A-300 IBC PV cell that had been introduced into mass production. On June 7,

EXAMPLES OF NOVEL PV CELLS

125

2004, SunPower announced the debut of its first mass-production line for PV modules with module conversion efficiencies approaching about 17% utilizing back the IBC solar cells. On October 16, 2006, SunPower announced the introduction into mass production of its new 22% efficient Gen-2 solar cells. On May 12, 2008, SunPower announced that it had produced a 5-in. prototype Gen-3 IBC PV cell with a world-record efficiency of 23.4%. By all means, this cell may be considered as a real masterpiece. It succeeded in incorporating in a practical way all of the valuable theoretical and technological approaches that had been discussed in the past. The main features of this cell include localization of current collecting contacts of both polarities exclusively on the rear side of the PV cell, thus minimizing front-side recombination due to elimination of front-side metallic current collecting contacts and optimizing light trapping by eliminating conventional PV cell shading and by introducing the most efficient texturing. This PV cell is produced from high-purity monocrystalline n-type silicon material with a charge carrier lifetime >1 ms that allows minority carriers to diffuse from the illuminated front surface through the entire wafer thickness to arrive at the junction and current collecting contacts of both polarities at the rear side. Front-side recombination has been additionally decreased by introducing n+-doped and SiO2 passivating layers on the front side. The back side of the IBC PV cell is processed in a very special manner. Initially, “interdigitated” n+ and p+ parallel narrow nonoverlapping strips between the PV cell’s opposite edges are produced by sequential diffusion processes. An efficient electrical insulation is built between these strips to guarantee high shunt resistance. The entire back-side surface of this PV cell is covered by a SiO2 layer providing efficient rear-side passivation. Contacting holes through this back-side SiO2 layer are produced in precision alignment with corresponding n+ and p+ strips. Metallic contacting narrow fingers are further printed in a precision-aligned manner along corresponding contacting holes, thus providing electric contact through the holes with the underlying n+ and p+ strips. Two terminal bus bars are screen printed on the opposite edges of the IBC PV cell’s rear side in such a way that one of them is connected to all n+ fingers and the other to the p+ fingers. These terminal bus bars are used for PV cell testing and interconnection in series by means of special tabbing during PV module production. These PV cells demonstrate high current (>40 mV/cm2) not only due to efficient light trapping and low shading but also due to better external quantum efficiency. The blue response is improved due to a highly n+-doped front diffusion and emitter localization on the back side, thus eliminating the front-side dead layer associated with a conventional PV cell structure with emitters localized on the front side. A long-wavelength response in the near IR spectral region has been improved as a result of a perfect rear-side passivation provided by the SiO2, which covers the entire rear-side area except for the contacting holes. Optimal front- and rear-side processing provides for an optical thickness for the PV cell about six times its actual thickness. Due to perfect frontand rear-side passivation, a Voc of about 700 mV has been achieved. An optimized screen printing process allows the FF to reach a value of >80%. In contrast to conventional crystalline silicon cells that demonstrate a Voc value decline of

126

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

2.2 mV/°C, the SunPower cell’s Voc dependence on temperature is relatively lower at about 1.87 mV/°C, thus decreasing its temperature dependence for power losses to 0.38%/°C compared to 0.5%/°C typical for conventional PV cells. This differentiation certainly will play an important role in securing annual kilowatt hour power generation. Although SunPower asserts that the production cost of these cells is low, it is evident that the cost should be substantially higher when compared with conventional PV cell fabrication due to the more expensive silicon material and the technology complexity. Several PV companies have made attempts to produce PV cells employing back contact concepts, but none of them have succeeded so far in introducing their designs into mass production mainly due to production technology complexity and high cost. The fact is that practically all crystalline silicon PV cell manufacturing companies are using screen printing technology for current collection, although it is evident that the limited conductivity of screen printed metallic patterns represents a real bottleneck for further PV cell efficiency improvement and cost reduction. In reality, the PV cell is just a small electric generator. Imagine what would happen if some gas electric generating plant were renovated and then its production capacity is doubled, and it is then placed back in service while the corresponding transmission line is kept on the same level? Isn’t it evident that a large portion of its additional electric energy will dissipate into heat due to parasitic losses in the transmission line? The same happens when conventional screen printing technology fails to collect electric power without parasitic losses from more efficient PV cells. Therefore, there is an urgent request from the PV industry for a more efficient current collecting technology.

5.5.6

Day4 PV Cell

In 2003, Day4 Energy Inc. developed a novel concept for a PV cell with improved current collection hereinafter referred to as D4 technology or D4 electrode or D4 PV cell [18, 19]. This concept is based on a proprietary current collecting electrode (Fig. 5.3) composed of copper wires coated with low melting alloys that are spaced apart and imbedded into an optically transparent adhesive layer that in turn is firmly

Film

Adhesive Wire Alloy

Figure 5.3. D4 wire-tape electrode concept.

EXAMPLES OF NOVEL PV CELLS

127 Bus bar

Wires

Wires

PV cell

Fingers

Figure 5.4. The placement of the D4 wire-tape electrode on a PV cell.

adhered to the supporting transparent polymeric film. Since the thickness of the adhesive layer is lower than the thickness of the wire it secures, then at least part of the wire protrudes above it. One may see (Fig. 5.4) that the electrode wires are soldered to the copper bus bar of about 5- to 10-mm width and 50- to 200-μm thickness. The bus bar is typically positioned outside the PV cell perimeter, thus not occupying its front surface, thereby preventing shading. The D4 electrode is placed on top of the front side of the PV cell in such a way that its wires are oriented in a transverse direction in respect to the cell’s screen printed fingers, thus allowing each wire to be placed on top of each and every grid finger. The entire composite structure is exposed to the vacuum lamination process comprising sequential vacuuming, heating, and pressing. During this process, the air is pumped out, the alloy melts, and under pressure, the electrode wires become firmly soldered to the screen printed fingers. At the same time, the electrode’s adhesive layer softens and under pressure firmly fixes the polymeric film onto the PV cell front surface. The D4 electrode secures a very low ohmic contact between the electrode wires and the screen printed aluminum on the rear side of a PV cell due to the alloy properties and strong mechanical compression during the vacuum lamination. Due to the large numbers of soldering points between the electrode wires and the screen printed fingers, this electrode secures a substantially lower risk of electrical layout failure when compared with the conventional tab spot soldering using the conventional screen printed bus bars. It is known that resistive power losses are proportional to the square of the current flow distance. Figure 5.5 shows a typical screen printed cell with screen

128

CRYSTALLINE SILICON SOLAR CELLS AND MODULES Bus bar Wires

4

70

Bus bar

Bus bar

Fingers

Figure 5.5. Comparison between the conventional and D4 PV cell.

printed bus bars on the left in contrast to a D4 cell with a wire-tape electrode on the right. Referring to Figure 5.5 (left), the typical length that the electric current must flow through the screen printed finger before reaching the screen printed bus bars is not less than 35 or 20 mm in the case of a 6-in.2 conventional PV cell that utilizes two or three screen printed bus bars. In contrast, in the case of the D4 PV cell shown in Figure 5.5 (right), with the same size PV cell with the same grid pattern, the D4 electrode allows replacing two or three screen printed bus bars with 40 D4 electrode wires, thus reducing the current flow distance along the screen printed grids to only 2 mm. One may see that these wires are soldered to the bus bar provided for series interconnection with the following PV cell. Therefore, the resistive losses for the case of the conventional current collection through bus bar ribbons are at least 20 times higher when compared with the D4 current collecting technology. Employment of the D4 technology provides several additional advantages. First, it allows eliminating conventional screen printed bus bars from the front side, thus increasing the PV cell’s Rsh value and removing silver pads from the back side of PV cells, resulting in lowered recombination losses due to improved front- and rear-side passivation. Removal of silver/aluminum pads on the rear side allows eliminating one screen printing step and one drying step. It also decreases the silver paste consumption by 40% along with lowering production cost with no compromise in the cell efficiency. In fact, the D4 technology improves the PV cell efficiency by >0.1% absolute mainly due to higher Voc and FF values. In addition, it opens the possibility to introduce into mass production PV cells with substantially narrower screen printed fingers. It has been demonstrated that after industrial conventional multicrystalline (MC) PV cells are equipped with D4 electrodes, FF values of not less than 77% with screen printed fingers of 70-μm width and less than 6-micron height provide an improved cell efficiency up to 0.5% absolute due

PV MODULE

129

to lower shading and better front and rear-side passivation. Recent results demonstrated that production costs of the PV cells with narrow fingers may be substantially reduced due to >250% lower silver paste consumption relative to conventional PV cells with two bus bars. It should be appreciated that D4 technology opens a wide range of possibilities to introduce into mass production novel low cost and more efficient PV cells based on several promising technological concepts such as extremely narrow current collection metallic patterns produced either by screen printing or electroless plating, or aerosol, or ink-jet noncontacting direct printing, or laser-fired contacts that have been already developed but failed to find commercial application so far. There is realistic possibility to employ D4 technology for electric power collection and interconnection of the back contact PV cells, thus decreasing their production cost and simplifying PV module production.

5.6

PV MODULE

In contrast to the extensive progress in PV cell technology, the development of the production technology for PV modules with crystalline silicon PV cells has remained virtually unchanged for more than 30 years.

5.6.1

Conventional PV Module Production Technology

Since a PV cell is actually is a source of low-voltage ≈0.6 V with a DC electric current of about 34 mA/cm2 in sunlight, there is a need to interconnect a large number of PV cells in series in order to achieve a necessary and useful value for the DC voltage that in turn may be converted into AC by means of an inverter. Neighbored PV cells typically are interconnected in series by means of tinned copper tabs that are spot soldered onto the front side of the bus bars of a first PV cell and then onto the silver pads on the rear side of the next sequential PV cell. A certain number of PV cells interconnected in series are known as a PV string. Strings in their turn are interconnected in series by means of tinned copper bussing, thus producing a PV module layout. PV modules with series-interconnected PV cells perform optimally only when all the series-interconnected PV cells are illuminated with an approximately similar light intensity. However, if even one PV cell within the PV module layout is shaded while all other cells are illuminated, the entire PV module is adversely affected, resulting in a substantial decrease in power output from the PV module. In addition to temporary power loss, the module may be permanently damaged as a result of cell shading because when a PV cell is shaded, it starts to act as a large resistor rather than as a power generator. In this situation, the other PV cells in the PV string expose the shaded cell to a reverse voltage that drives electric current through this large resistor. This process may result either in the breakdown of a shaded PV cell or in heating it to such a temperature that it can destroy even the

130

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

entire PV module if a very high temperature persists. In order to eliminate the risk of a PV module damage in the event of shading, practically all PV modules employ BPD connected across each of the PV string and/or an entire module depending on specific PV module design and the quality of the input PV cells. The number of PV cells in a single PV string depends on the PV cell quality, namely, the ability to withstand back-voltage breakdown that each PV cell may be exposed to if one of them within the PV string is shaded. For example, for PV cells of good quality that can withstand back-voltage breakdown of 14 V and given that each PV cell generates a Vmax = 0.56 V, then the number of PV cells in one string should not exceed 24. Given that PV cells produced from metallurgical silicon typically have lower quality and back-voltage breakdown for these cells is not higher than 7 V, using them in PV strings comprising more than 12 cells is not recommended. Although employment of the BPD allows protecting the PV cells and PV strings against damage, it also causes substantial power losses of a PV module because the shading of just one PV cell results in an entire PV string switch-off. According to the industry estimate, almost 30% of kilowatt hour annual generation may be lost in field condition due to different sources of shade. Therefore, there is a need to optimize a PV module lay out in order to secure not only sufficient shading protection of PV cells but also to minimize annual kilowatt hour losses of PV systems as well. Since PV modules are generally expected to operate outdoors for typically 25 years without degradation, their construction must withstand various weather and environmental conditions. The front side of a typical PV module construction involves the use of a transparent sheet of low-iron tempered glass. The PV cell strings are sandwiched between sheets of polymeric encapsulant material, such as ethylene vinyl acetate, or thermal plastic material, such as polyvinyl butyral. An array of PV cells is placed onto the polymeric encapsulant material in such a way that the front sides of the cells face the transparent glass sheet. The back side of the array is covered with an additional layer of encapsulant material and a backsheet layer of weather protecting material, such as Tedlar® by DuPont, or a glass sheet. The additional layer of encapsulant material and the back-sheet layer typically have openings to provide for terminal electrical conductors to be passed from the PV cell strings through the back-side encapsulant layer and back sheet of weather protecting material to the outer surface for connecting with the electrical load through a junction box. For a PV module having an array of two PV strings, typically four conductors are arranged to pass through the openings so that they are all in proximity with each other so they can be terminated in a junction box mounted on the back-sheet layer. The glass, encapsulant layers, cells, and backsheet layer are typically vacuum laminated at about 14–160°C to build a free-of-air bubbles, strong bound structure that protects the PV cells from moisture penetration from the front- and back sides and also from the edges. The electrical interconnections of the PV strings and connections to BPDs are made in the junction box. The junction box is sealed on the back side of the PV module and is equipped with electric cables for neighbored PV module interconnection.

PV MODULE

131

An aluminum frame extends around the perimeter of the PV module and protects it against damage, provides mechanical strength against wind and snow loads, and facilitates mounting of the module to a support. At the same time, it is possible to employ PV modules without the aluminum frame if external supporting structures are sufficient to provide necessary mechanical strength. The fabrication of PV modules as described above with conventional technology is quite complicated and expensive. Layout of the PV module before lamination requires a separate step of “bussing” in which PV strings are electrically interconnected typically by means of tinned copper busses that increase the area that is occupied by the PV module, thus decreasing its conversion efficiency. Due to differences in thermal expansion coefficients between the copper and silicon and glass, there is a certain risk that the soldered spots between conventional tabs and front-side bus bars and rear-side silver pads may break causing PV module irreversible damage under inevitable changes of ambient temperature. There is an additional risk especially associated with utilization of thin PV cells. Exposing cells to spot soldering may result in PV cell breakage due to local heating and pressure. Once PV cells are interconnected in series by means of conventional technology, the series resistance of the produced PV module eventually exceeds the sum of all PV cell resistances due to the additional impact of soldering points, tabs, and bussing resulting in PV module FF value decline and corresponding efficiency losses when compared to the efficiency of the PV cells utilized in the module’s fabrication. For example, when PV cells with an average FF value of ≈76% are used to produce a 48-cell PV module, the module FF value typically will be lower than 73%. In other words, excessive series resistance associated with conventional PV module production technology is responsible for about 4% power loss when compared to the power that the input PV cells were capable of generating before being interconnected in the PV module. These insufficiencies become even more pronounced when modern highefficiency PV cells are utilized for PV module production. This is a clear demonstration of a growing conflict of interest between PV cell and PV module producers. In fact, PV cell producers are motivated to build PV cells with increased power output combined with decreased production cost. The easiest way to achieve this goal is to make PV cells thinner, thus securing economy of silicon material and to increase PV cell size, thus increasing the production capacity of existing manufacturing equipment with minimal additional investments. During the last 5 years, the size of PV cells has been increased from 100 × 100 cm2 to 125 × 125 cm2 and further to 156 × 156 cm2. Even a larger cell of 210 × 210 cm2 area has been developed and tested. PV module producers have been forced to buy these cells for an increased price based on dollar per Wp, but in return, they do not have sufficient benefit because thinner cells become more fragile and experience more pronounced bowing with corresponding higher-yield losses due to higher risk of breakage. To minimize this risk, there is a need for additional investment for more sophisticated production equipment for tab soldering and general handling. At the same time, conventional PV module fabrication technology does not allow increasing PV cell

132

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

power output at PV module level if the module series resistance is not decreased accordingly. Another challenge that PV module producers have to deal with is the necessity to achieve certain voltage values that a number of interconnected PV modules must generate per occupied area in order to secure the inverter’s most efficient operation. It is evident that although increased size of more efficient PV cells results in higher power output at the PV cell level, it eventually has a negative effect on the PV module performance due to the decreased voltage value since the number of PV cells per occupied area is diminished. Yet another challenge that PV module producers have experienced is the necessity to redesign the PV module layout to allow for an increasing number of PV strings now composed of a lower number of PV cells as required for cells made with metallurgical-grade silicon. Another example of a growing conflict of interest between PV cell and PV module producers is the employment of a PV cell with a selective emitter that demonstrates a higher efficiency in the blue spectral region. This advantage allows PV cell manufacturers to sell their cells at a higher price. At the same time, PV module producers do not observe the same efficiency gain on the PV module level not only because the glass and encapsulant materials substantially cut off the blue light but also due to the inevitable accumulation of dirt on the front side of PV modules. These general considerations reflect an urgent market request for developing novel, simpler, flexible, and cost-efficient technologies that are capable of utilizing more efficient PV cells in PV module production without loosing generated power.

5.6.2

Day4™ Technology for PV Module Production

One such promising technology has been recently developed at Day4 Energy Inc. [20, 21]. It was further demonstrated [22] that PV cells may be interconnected in series by means of electrically connecting the front-side D4 electrode of the first PV cell with the back-side D4 electrode of the sequential PV cell via the tinned copper bus bar. Since the thickness of this bus bar may be ≤100 μm, it securing extremely low series resistance of this interconnection. This approach allowed replacing conventional tabbing and stringing technology along with simplifying the PV module layout. New technology provides an elegant solution to produce U-type PV strings without conventional bussing by just turning each of two sequential PV cells with front and back electrodes by 90° before placing it on top of the rear-side electrode of the previous PV cell. It further allows eliminating conventional string interconnection by means of bussing, thus not only simplifying and optimizing PV module layout but also decreasing its production cost due to material and production step reduction. At the same time, PV module efficiency is increased due to its lower series resistance and economy of space occupied by conventional bussing. It is worth to notice that flexibility of the Day4 technology allows designing and producing PV modules with an increased number of PV strings, thus improving protection against shading and preserving annual generated kilowatt hour energy.

CONCLUSION

133

D4 PV module technology and specialized production equipment were introduced into mass production in 2006 after being UL and TUV certified. To date, about 40 MW of these PV modules have been installed on European, North American, and other markets and have received positive references from customers confirming excellent and reliable performance of installation. According to neutral party test results, installations with these PV modules demonstrate higher values of annually generated kilowatt hour in comparison with neighboring installation employing PV modules from competitive producers due to sufficiently higher shunt resistance of Day4 PV cells without bus bars, resulting in higher conversion efficiency at low light intensities. Depending on the quality of the input PV cells, the average power output of a mass-produced D4 PV module containing 48 MC PV cells is not less than 175 W. The efficiency of the best produced D4 module so far is 15%, while the average is typically >14% relative to the<13% average efficiency of industrially produced 48 MC cell modules. This pronounced difference may be explained as a result of increased efficiency of input D4 PV cells and the preservation of this efficiency at the D4 PV module level. This conclusion is strongly supported by the experimental fact that the FF values of mass-produced D4 PV modules are very close to the FF value of the input PV cells. This is not the case with conventional PV module production technology. This observation has a clear explanation: D4 PV modules are characterized by about 30–40% less series resistance when compared to conventional PV modules.

5.7

CONCLUSION

In spite of the current financial and economic crisis, the global requests for new, reliable sources of energy continue to be addressed as priority number one. Compared to other renewable sources, solar electric generation is the only one that is intrinsically universally distributed. This unique feature allows establishing PV generating capacities close to end users, thus minimizing capital investments for new electric energy transmission lines. At the same time, these advantages have zero value if not supported by lower dollar per kilowatt hour cost. During the twentieth century, there were at least two moments when interest toward PV generation was extremely high and attracted substantial investments. Unfortunately, these investments ran almost dry when oil price went down. The current situation is the same. Unfortunately, the end user of electric energy does not care about the source but only about the price. The PV industry cannot continue being dependent on subsidies. The political risk associated with the government subsidies introduces a substantial amount of volatility into the PV industry demand equation. Government subsidy programs and, most importantly, political will to support them may be reduced substantially if the PV industry is unable to demonstrate its ability to reach grid-parity in a significant number of applications over the course of the next few years. Therefore, the PV industry must be ready to compete with upgraded coal, nuclear, and other so-called new clean technologies where the experienced lobbyist

134

CRYSTALLINE SILICON SOLAR CELLS AND MODULES

has the capacity to attract practically all investment capital, thus leaving the PV industry for years ahead without sufficient financial resources. With all respect to new types of PV cells and PV modules, we have to admit that just the crystalline silicon PV technology has sufficient experience in >30 years of numerous numbers of PV systems’ reliable performance in different geographical locations. That is why the 25 years warranty on the crystalline silicon PV modules is based on a real solid ground. Therefore, we believe that the crystalline silicon PV industry will continue to be attractive for the long-term investments and to be the core basis for industrial solar electric energy generation. The material presented above illustrates the high potential for cost reduction and efficiency improvement practically at all steps of the crystalline silicon PV cell and module production that may be sufficient to make PV electric energy cost competitive with conventional nonrenewable sources. The challenge is how to convert this potential into sufficient cost reduction of kilowatt hour generated in-field conditions. It might sound strange but the current financial crisis in fact has created a moment of truth that is forcing the PV industry either to change its business model or to die. During the last 8 months, there has been an amazing price reduction for the PV cells (>40%) and PV modules (>50%). Keeping in mind that last year the Si-based PV cell contributed about 70% of the PV module cost and now it is lowered to 55%, there is sufficient potential for further PV module and kilowatt hour cost reduction. This trend must be combined with a modified PV industry business model that must change its current focus from dollar per Wp toward dollar per kilowatt hour and ROI. The selling policy must also be changed from the current 25-year guarantee for PV module Wp performance toward the 25-year guarantee of a certain kilowatt hour annual generation. We believe that the success and survival of the PV industry depends on the implementation of this ambitious but feasible program.

ABBREVIATIONS AC—alternating current AR—antireflective coating BPD—bypass diode BSF—back surface field CAD—computer-assisted drawing CZ—Czochralski process D4—Day4 Energy DC—direct current FF—fill factor Gen—generation HIT—heterojunction with intrinsic thin layer I—current IBC—interdigitated back contact Impp—current at maximum power point

REFERENCES

135

IR—infrared Isc—short-circuit current ISE—Institute for Solar Energy max—maximum mc—multicrystalline n—negatively or donor-doped semiconductor NREL—National Renewable Energy Laboratory p—positively or acceptor-doped semiconductor PERC—passivated emitter and rear cell PERL—passivated emitter, rear locally diffused Pmmp—power at maximum power point PV—photovoltaic R—resistance Rbl—bulk resistance Rem—emitter resistance ROI—return on investment Rs—series resistance Rsh—shunt resistance SiC—silicon carbide SiNx—silicon nitride antireflective coating SiO2—silicon dioxide TUV—Technischer Überwachungs-Verein (Technical Inspection Association) UL—Underwriters Laboratories V—voltage Vmmp—voltage at maximum power point Voc—open-circuit voltage Wp—watt peak η—efficiency 3D—three dimensional

REFERENCES [1]

[2] [3] [4] [5]

S. W. Glunz, A. Grohe, M. Hermle, M. Hofmann, S. Janz, T. Roth, O. Schultz, M. Vetter, I. Martin, R. Ferré, S. Bermejo, W. Wolke, W. Warta, R. Preu, and G. Willeke. Comparison of different dielectric passivation layers for application in industrially feasible high-efficiency crystalline solar cells. Presented at the 20th European Solar Conference and Exhibition, June 6–10, 2005, Barcelona (2005). S. W. Glunz. High-efficiency crystalline silicon solar cells. Article ID 97370. Review article. In Advances in Optoelectronics, doi:10.1155/2007/97370 (2007). A. Brenner and E. Riddell. Nickel plating on steel by chemical reduction. Journal of Research of the National Bureau of Standards, 37, 31–34 (1946). A. Brenner. Electroless plating comes of age. Metal Finishing 37, 61–68 (1954). M. V. Sullivan and J. H. Eigler. Electroless nickel plating for making ohmic contacts. Journal of the Electrochemical Society 104, 226–230 (1957).

136 [6] [7] [8]

[9] [10] [11]

[12]

[13] [14] [15] [16] [17] [18] [19]

[20]

[21]

[22]

CRYSTALLINE SILICON SOLAR CELLS AND MODULES L. F. Durkee. Method of plating by means of light. U.S. Patent No. 4 144 139, Solarex Corporation, Rockville, MD (1979). L. A. Grenon. Electroplating method. U.S. Patent No. 4 251 327, Motorola, Schaumburg, IL (1981). S. W. Glunz, J. Knobloch, D. Biro, and W. Wettling. Optimized high-efficiency silicon solar cell with Jsc = 42 mA/cm2 and η = 23.3%. In Proceedings of the 14th European Photovoltaic Solar Energy Conference, Barcelona, Spain, pp. 392–395 (1997). K. F. Teng and R. W. Vest. Application of ink jet technology on photovoltaic metallization. IEEE Electron Device Letters 9, 591–593 (1988). T. Rivkin, C. Curtis, A. Miedaner, J. Perkins, J. Alleman, and D. Ginley. Direct write processing of photovoltaic cells. In Proceedings of the 29th IEEE Photovoltaic Specialists Conference, New Orleans, LA, pp. 1326–1329 (2002). M. F. Kaydanova, A. M. van Herst, A. Miedaner, C. Curtis, J. Alleman, M. S. Dabney, E. Garnett, S. Shaheen, L. Smith, R. Collins, J. I. Hanoka, A. M. Gabor, and D. S. Ginley. Direct write contacts for solar cell. In Proceedings of the 31st IEEE Photovoltaic Specialists Conference, Orlando, FL, pp.1305–1308 (2005). M. Hoerteis, P. L. Richter, and S. W. Glunz. Improved front side metallization by aerosol jet printing of hotmelt inks. Presented on the 23rd European Photovoltaic Solar Energy Conference and Exhibition, September 1–5, 2008, Valencia, Spain (2008). H. J. Gould. Method of manufacturing a back contact for semiconductor die. U.S. Patent 5,451,544 (1993). R. Preu, E. Schneiderlöchner, S. Glunz, and R. Loedeman. Method of producing a semiconductor-metal contact through a dielectric layer. U.S. Patent 6,982,218 (2001). M. A. Green and S. R. Wenham. Australian Patent No. 5703309, Australia (1984). M. Green, J. Zhao, A. Wang, and S. R. Wenham. Very high efficient silicon solar cell- science and technology. IEEE Transactions on Electron Devices 46(10), 1940– 1947 (1999). R. Swanson. Method of fabricating back surface point contact solar cells. U.S. Patent No. 4,927,770 (1990). L. Rubin and G. Rubin. Electrode for photovoltaic cells, photovoltaic cell and photovoltaic module. EP Patent 1 547 158 B1, 2002, August 29 (2007). A. Schneider, L. Rubin, A. Osipov, A. Smirnov, and P. Antipov. A new approach in solar cell module interconnection technique resulting in 5-10% higher PV module power output. Presented at the IEEE 4th World Conference on Photovoltaic Energy Conversion, Waikoloa, Hawaii, May 8–12 (2006). A. Schneider, L. Rubin, and G. Rubin. Solar cell efficiency improvement by new metallization techniques—The Day4™ electrode concept. Presented at the IEEE 4th World Conference on Photovoltaic Energy Conversion, Waikoloa, Hawaii, May 8–12 (2006). A. Schneider, L. Rubin, and G. Rubin. The Day4 electrode—A new metallization approach towards higher solar cell and module efficiencies. Presented at 21st European Photovoltaic Solar Energy Conference, September 4–8, 2006, Dresden, Germany (2006). L. Rubin and V. Nebusov. Photovoltaic module with edge access to PV strings, interconnection method, apparatus, and system. PCT/CA/2007/002301 (2007).

6 THIN-FILM SOLAR CELLS AND MODULES ROBERT BIRKMIRE Institute of Energy Conversion, University of Delaware

6.1

INTRODUCTION

Thin-film solar cell technologies originated in the 1960s with Cu2S/CdS, which was the first (flexible) thin-film solar cell and the first to achieve 10% efficiency in 1981 [1]. From 1970 through the 1980s, CuInSe2 [2], CdTe [3], and a-Si [4] became the solar cell materials of interest with all three achieving ∼10% efficiency. The mantra through this period was thin films will be in large-scale production and will cost ∼$1 per watt next year, which continued through the 1990s. In the 1990s, however, performance levels of the thin-film solar cells increased with CuInGaSe2 >19% [5], CdTe >16% [6], and a-Si >10% (stabilized) [7]. This, coupled with the expansion of c-Si technologies in the mid-1990s, has led to significant efforts to commercialize thin film technologies. At the turn of the century, PV began its rapid expansion with production increasing about 35%/year with most of the capacity in the c-Si module production. In 2008, worldwide solar cell production was ∼6.85 GW up from 3.44 GW in 2007 [8]. In the past few years, thin film manufacturing has undergone rapid expansion lead primarily by a-Si and CdTe technologies, and in 2008, thin film production was ∼0.89 GW up about 120% [8]. The emergence of the thin film technologies has been, in part, driven by the shortage of Si feedstock, which limited new production of c-Si and mc-Si modules. First Solar emerged as the largest thin-film PV manufacturer in the world and has expanded its manufacturing capacity of CdTe to 1 GW in early 2009. It is the first thin film company to demonstrate the “promise of thin-film PV” that high-throughput processing and large-scale facilities will significantly reduce the cost of PV modules and has achieved the lowest manufacturing cost per watt in the industry of less than $1 per Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

137

138

THIN-FILM SOLAR CELLS AND MODULES

TABLE 6.1. Best Verified Solar Cell Efficiencies for Thin-Film Solar Cells along with c-Si for Comparison under Standard Test Conditions Classification

Eff. (%)

Area (cm2)

VOC (V)

JSC (mA/cm2)

FF (%)

Test Center (Reference)

CdTe

16.5

1.03

0.845

25.9

75.5

NREL [6]

CuInGaSe2 CuInGaSe2

19.4

0.99

0.716

33.7

80.3

NREL [5]

20.0

0.419

0.692

35.7

81.0

NREL [5]

a-Si

9.5

1.07

0.859

17.5

63.0

NREL [7]

nc-Si

10.1

1.2

0.539

24.4

76.6

JQA [10]

a-Si/a-Si/a-SiGe

12.1

0.27

2.297

69.7

NREL [11]

Si (crystalline)

25.0

4.0

0.705

42.7

82.8

Sandia [12]

Si (multicrystalline)

20.3

1.002

0.664

37.7

78.9

NREL [13]

7.56

For a complete listing of the best cell efficiencies, see Green et al. [14].

watt [9]. a-Si manufacturing is rapidly increasing with Applied Materials, Oerlikon Corp, and Sharp leading the way. The production of CuInSe2-based modules has also increased substantially through efforts of Global Solar and Wurth Solar along with the entrance of Showa Shell and Honda into the market. This has spurred an infusion of venture capital for a large number of start-up companies in the United States and throughout the world in CuInSe2-, CdTe- and a-Si-based technologies as well as sparked the interest of several larger companies. The market share of thin-film PV modules is expected to increase substantially in the coming years. The performance of commercially available thin-film modules is less than crystalline Si (c-Si) modules that typically have efficiencies ranging from 12% to 19%. CuInSe2-based modules have the highest efficiency of thin-film modules in the range of 9–12% with CdTe modules from 9% to 11% and a-Si typically in the 5–8% range. The thin film technologies are at a similar development stage as c-Si was in the 1980s, and the next challenge is to increase module performance at the manufacturing level. A summary of the best cell and module performance measured at certified test laboratories is given in Tables 6.1 and 6.2 for a-Si-, CdTe-, and CuInSe2-based devices along with both single-crystal and multicrystal Si as reference points. It is interesting to note that both a-Si and CdTe, the most commercially advanced technology, have not recently increased the best performance value, mostly likely since the focus has been on increasing manufacturing capacity. In Figure 6.1, the best efficiency thin-film solar cell device is compared to the maximum obtainable single junction efficiency along with the best cell and module efficiency for single-crystal Si as a reference point. Single-crystalline Si is within about 85% of the efficiency limit, while the best CuInGaSe2 cells are ∼70% of the efficiency limit and can be directly compared

INTRODUCTION

139

TABLE 6.2. Best Verified Module Efficiencies for Thin Film along with c-Si for Comparison under Standard Test Conditions Classification

Eff. (%)

Area (cm2)

VOC (V)

ISC (A)

FF (%)

Test Center (Reference)

CdTe

10.9

4874

26.21

3.24

62.3

NREL [15]

31.2

2.18

68.9

NREL [16]

3.285

66.0

NREL [17]

CuInGa(Se,S)2

13.5

3459

a-Si/a-SiGe/a-SiGe

10.4

905

Si (large area)

20.1

16,300

66.1

6.35

78.7

Sandia [18]

Si (multicrystalline)

15.5

1017

14.6

1.37

78.6

Sandia [19]

4.353

For a complete listing of best module efficiencies, see Green et al. [14].

35 30

Cell Module Black-body limit

Efficiency (%)

c-Si 25 20

CuInGaSe2

AM0 CdTe

15 AM1.5 a-Si

10 5 0.5

1.0

1.5 Bandgap (eV)

2.0

2.5

Figure 6.1. Comparison of maximum achievable efficiency for single-junction solar cells to best-reported solar cell for different technologies. The best module performance is also included.

to the best multicrystalline Si cells since they have about the same bandgap and are polycrystalline in nature (not shown on graph). CdTe is about 55% of the efficiency limit, which indicates that advances in the laboratory are needed for the technology to reach its full potential. It is difficult to evaluate a-Si PV technology on this basis since the best devices are multijunction structures. The differential between the best cell and module can be used as a crude measure of the readiness of the technology for manufacturing. The efficiency of the CdTe and CuInGaSe2 modules is about 50–60% of the best cell, while multijunction a-Si and c-Si modules are about 80% of the best cell and mc-Si is 75%.

140

THIN-FILM SOLAR CELLS AND MODULES

The selection of PV modules will be based on annualized power output and levelized cost of energy for the specific location of the array and type of application. Figure 6.2 compares the annualized output for c-Si, mc-Si, a-Si, and CdTe at different geographic locations in the United States for residential rooftop arrays mounted at a fixed tilt for the local latitude. The graph was generated from the SAM [20], developed by DOE/NREL. Table 6.3 lists the module manufacturers and efficiencies used from SAM. The array size is ∼2.5 kW for all technologies and the SAM database is used for the modules. It is important to note that the lower-efficiency a-Si (5.7%) and CdTe (7.7%) arrays perform nearly as well as c-Si (19.3%) and better than mc-Si (13.3%) as the result of a better temperature coefficient. However, there is a large variation in temperature coefficient data for PV modules, and the effects of illu-

1500

1000

500

CdTe

0 AZ

mc-Si HI

MT Location

c-Si PA

pe

a-Si

Ty

Annual Output (kWh/kW)

2000

WA

Figure 6.2. Comparison of the annualized power output for deployment on a ∼2.5-kW roof residential rooftop array at different geographic locations for thin-film and Si modules. TABLE 6.3. Modules Used to Generate the Annualized Power from the Solar Advisor Model Database Module

Type

Eff. (%)

SunPower

c-Si

19.3

Kyocera

mc-Si

13.3

Uni-Solar

a-Si

5.7

First Solar

CdTe

7.7

THIN-FILM SOLAR CELLS AND MODULES

141

mination intensity on the module performance are not taken into account in the calculations. Thus, an accurate comparison requires a more detailed model that is validated by real data. Also, the lower-efficiency arrays require significantly more area, which increases the balance of system cost. The “figure of merit” will thus be the levelized cost of energy.

6.2

THIN-FILM SOLAR CELLS AND MODULES

The motivation for thin-film solar cells was, and still is, the potential for highspeed/high-throughput manufacturing and minimum material requirements to reduce cost. To meet this goal, large-scale manufacturing facilities are needed to obtain economies of scale, but until recently, the large up-front investment required inhibited the full-scale commercialization of thin film technologies. Thin-film modules all have several common components: (1) substrate; (2) a TCO and/or grid; (3) metal contacts; and in many cases, (4) a monolithic integration scheme. The substrate is either glass or a flexible web that is either plastic or metal foil, and the device configuration is either a superstrate or substrate design as shown in Figure 6.3. The superstrate is always glass for the thin film technologies and is coated with a TCO such as ITO, SnO2, or ZnO. Both CuInSe2- and a-Si-based modules are currently commercially being manufactured using glass and/or flexible substrates, while CdTe is always on a glass substrate. (There are several start-up companies exploring CdTe on a metal substrate.) For the substrate device configuration, glass is used for CuInSe2-based modules where Mo is used as the back contact, and flexible metal foils or polyimide substrates (web) are used for both a-Si- and CuInSe2-based modules. The attraction of the flexible substrate is the use

Figure 6.3. The two basic cell configurations for thin-film solar cells.

142

THIN-FILM SOLAR CELLS AND MODULES

of continuous roll-to-roll processing for manufacturing and was first proposed and demonstrated for Cu2S solar cells on a flexible Cu web [21]. Energy Conversion Devices and Iowa Thin Film Technologies adapted this for a-Si in the 1980s. In the 1990s, CuInSe2-based solar cells were developed on a polyimide and stainless substrate, and recently, several start-up companies are pursuing this approach on both polyimide and metal webs. A major advantage of thin film technologies is that large glass substrates or continuous webs can be coated as compared to c-Si, where individual wafer cells are fabricated and physically connected into modules using a “tab and stringing” process. For the glass and insulating webs, monolithic integration schemes are used for both superstrate and substrate modules to series connect individual cell segments, eliminating the need to assemble individual cells into the modules. Figure 6.4 illustrates the monolithic integration structure where the details of the processing steps depend on the device configuration and substrate/web. The cell segments and interconnections are defined by a series of laser and/ or mechanical scribes at different processing stages for glass substrates and are typically done as the last processing step for flexible insulating webs. For conductive webs used for both a-Si and CuInSe2, individual cells are prepared by cutting the web and applying a metal grid structure to provide the front contact to the cell and then are connected in series in a manner similar to assembling a c-Si module.

6.3

POLYCRYSTALLINE THIN FILM

Thin-film CdTe and CuInSe2 and related alloys are polycrystalline in nature with grain sizes of approximately 1–5 μm. Both CdTe- and CuInSe2-based solar cells

Figure 6.4. Typical monolithic integration scheme for thin-film modules.

POLYCRYSTALLINE THIN FILM

143

are heterojunction devices where CdTe is configured in a superstrate structure and CuInSe2 in a substrate structure. Both devices operate in a similar manner and have nearly 100% internal quantum efficiency and thus are generally not limited by Jsc. Approaches to improve device performance have primarily focused on Voc and FF, but the approach is different for each materials system. The primary mechanism limiting Voc, in both devices, is generally recombination in the space charge region, which is consistent with Shockley–Read–Hall recombination at defects within the bandgap [22, 23]. 6.3.1

CdTe Solar Cells and Modules

CdTe solar cell technology is leading the way in the development of thin-film PV in terms of cost and high-speed/high-throughput manufacturing. CdTe has a bandgap of 1.45 eV, which is a near perfect match to the AM 1.5 terrestrial spectrum for maximum solar cell efficiency (see Fig. 6.1). It is a direct bandgap semiconductor with an absorption coefficient >105 cm−1 at 700 nm and therefore only requires about a 1-μm thick film to absorb most of the incident light. All highefficiency CdTe solar cells have been made using a superstrate structure where the final processing steps, used to control the electronic properties and formation of a back ohmic contact, are carried out on a free CdTe surface. Efforts to use a substrate configuration where the CdTe is deposited on a conductive opaque substrate have not been successful, and thus CdTe solar cells are not compatible with a flexible substrate. Small area laboratory devices fabricated on a borosilicate glass substrate have achieved an efficiency of <16% [6], and a ∼0.5-m2 module with an efficiency of 10.9% has been achieved by BP Solar [15]. CdTe solar cells are p/n heterojunction devices where CdS is used as the wide bandgap, n-type window layer and a schematic of the device structure is shown in Figure 6.5.

Figure 6.5. Standard CdTe solar cell structure highlighting the individual layer in the device structure.

144

THIN-FILM SOLAR CELLS AND MODULES

The TCO is the front contact to the device, and different oxides such as SnO2, ITO, or Cd2SnO4 have been used depending on the device fabrication process and group or company fabricating the device. The properties of the TCO are chosen to minimize the sheet resistance while maintaining high optical transmission to reduce area-related losses associated with the grid structure in laboratory cells or with cell spacing in a monolithic integrated module. To reduce the possibility of shunts, a thin, high-resistance oxide film is deposited on the TCO prior to the deposition of the CdS window layer. This is needed since the thickness of the CdS is typically less than 100 nm to increase transmission and is prone to pinholes that result in shunting behavior in the final device. A variety of materials have been used for the high-resistive layer including SnO2 [24], In2O3 [25, 26], Ga2O3 [27], and Zn2SnO4 [28]. The CdS layer has been deposited by a variety of techniques such as PVD, CBD, CSS, sputtering, and electrochemical deposition. In the laboratory, the preferred method is CBD since thin, ∼50 nm, conformal CdS films can be grown at low temperature, which results in higher Jsc as a result of the increased transmission through the CdS. Also, in many cases, the film is annealed in the presence of CdCl2 at around 400°C, which increases grain size and reduces defect density [29]. For most manufacturing processes, a high-temperature process based on CSS, for example, VT, is used where Cd and S2 species are transported from a source to the substrate. A temperature difference between the source and the substrate leads to “high quality”—high-rate film growth on the lower temperature substrate typically at 500°C, and the films require no postdeposition treatments. CdTe films have been grown by a variety of techniques at substrate temperatures from near room temperature to 600°C including electrodeposition [30], CSS [31], VT [32], thermal evaporation [33], sputtering [34], chemical vapor deposition (CVD) [35], spray pyrolysis [36], and screen printing [37]. Film thicknesses have ranged from 1.5 to 15 μm. Most of the current manufacturing approaches seem to be based on a CSS-type process because of the very high CdTe growth rate, 1–10 μm/min, which can be achieved where the CdTe film thickness is about 4 μm. Figure 6.6 shows a comparison of the CSS and VT processes where the source-tosubstrate distance in CSS is small and the typical operating pressure is 10 torr.

Figure 6.6. Comparison of the CSS and VT processes that provide the basis for most of the commercial CdTe deposition systems.

POLYCRYSTALLINE THIN FILM

145

For the VT process, the source and substrate are separated and the Cd and Te2 species are entrained into a carrier gas and delivered to the substrate. The VT process can be operated over a broad range of pressures, but as the system pressure is increased, the carrier gas flow rate must be increased to obtain high CdTe film growth rates since the concentration of Cd and Te2 species entrained in the delivery gas decreases as the pressure increases [38]. Following the deposition of CdTe film, the most critical steps in fabricating a CdTe solar cell are performed, which fix the electronic properties of the device structure and the “quality” of the back ohmic contact. An overview of the processing is described below. For a detailed review of the post-processing, see McCandless and Sites [39]. The postdeposition processing consists of a CdCl2O2 annealing treatment using a solution or vapor phase delivery and heat treatment at about 400°C. The specific details of the postdeposition treatment are dictated by the CdTe film thicknesses and deposition process used to deposit the CdTe film. The treatment can result in grain growth, interdiffusion of the CdS/ CdS couple, and doping of the CdTe film. The final processing steps are a surface treatment to form the primary contact to the CdTe followed by the deposition of the current carrying metal. The surface treatment is needed to form a heavily doped or degenerate surface layer since the hole affinity of the CdTe is so high. A wet chemical etch is used to form a Te-rich surface followed a by p-type dopant, typically Cu, which is then annealed to drive the dopant in or the deposition of a degenerate semiconductor or semimetal such as ZnTe(Cu), HgTe, or PdTe. The specific details of the processing are laboratory/company specific and are not well quantified. The final step is the deposition of the secondary contact. Manufacturing of CdTe modules is dominated by First Solar, established in 1999 and utilizes the CdTe technology developed by Solar Cells Inc. (established in 1986) when the two companies were merged. The First Solar has achieved remarkable success expanding from a manufacturing capacity of 25 MW in 2005 to over a gigawatt in 2009 and reporting manufacturing costs less than $1 per watt with average module efficiencies of 10.8% [9]. They addressed the potential issues of Cd toxicity and Te availability by establishing a pre-funded module collection and recycling program. The success of First Solar has spawned a number of CdTe start-up companies where most of the new companies appear to utilize a process that is based on CSS/VT principles. There appears to be a clear approach to improve the CdTe module performance to 12–13% by increasing the transmission of glass/TCO and by further optimization of the high-resistance layer/CdS structure. However, a critical issue for CdTe manufacturers is that there is no clear pathway to increase the module performance to 15% or beyond based on current laboratory results [40]. There is a need to fundamentally improve CdTe semiconductor properties and possibly to develop a different device structure using a low-cost glass substrate and a thinner, ∼1 4-μm CdTe film.

146

6.3.2

THIN-FILM SOLAR CELLS AND MODULES

CuInSe2 and Related Alloys

Solar cells fabricated from CuInSe2 and related alloys have the potential to be the highest-efficiency modules of any of the thin film technologies. However, translation of the laboratory results to a manufacturing environment has been challenging [40]. CuInSe2 and related alloys are a class of direct bandgap materials that can be in some ways compared to the III–V semiconductors. These ternary compounds have bandgaps that range from 1.0 to 3.5 eV. Table 6.4 lists the bandgaps of the Cu-based materials. The materials, in general, form continuous solid solutions in the form of multinary compounds allowing the bandgap to be engineered. However, the electronic properties are dominated by intrinsic defects and change when forming the alloys. At any alloy, compositions yielding a bandgap of >1.3 eV, independent of the alloying materials, are less suitable for solar cells [41]. All high-efficiency CuInSe2-based devices are made using a substrate device structure where the substrate can be glass or a flexible plastic or metal foil. Efforts to use a superstrate configuration where the CuInSe2-based material is deposited on a transparent conducting substrate have not been successful [42]. The highest-efficiency solar cell, 20%, was made using a CuInGaSe2 film deposited by multisource evaporation [5], and the highest-efficiency module, 13.5%, was made using a CuInGa(Se,S)2 film grown by the reaction process [13]. CuInSe2-based solar cells are p/n heterojunction devices where a wide bandgap, n-type window layer such as CdS is used as the heterojunction partner with the p-type CuInSe2 base material. A schematic of the device structure is shown in Figure 6.7. Mo is used for the back ohmic contact and is deposited by sputtering. CuInSe2based thin films are generally deposited by the multisource elemental thermal evaporation or reaction of a precursor such as sputtered metals in a Se and/or S atmosphere. The remarkable property of CuInSe2 materials is the tolerance to

TABLE 6.4. CuInSe2 Alloy Materials Material

Eg (eV)

CuInSe2

1.0

CuGaSe2

1.68

CuAlSe2

2.72

CuInS2

1.53

CuGaS2

2.53

CuAlS2

3.5

CuInTe2

1.1

CuGaTe2

1.23

POLYCRYSTALLINE THIN FILM

147

Figure 6.7. Standard CuInSe2-based solar cell structure highlighting the individual layer in the device.

composition variations. The Cu concentration in the film can vary by 2–3% without affecting the device performance provided that the Cu-to-group III ratio is less than one. This is attributed to the broad single-phase region of CuInSe2 at the growth temperatures and may explain why, when forming a wider bandgap alloy, materials are less tolerant to composition. CuInSe2 may be the only I-III-V chalcopyrite material that has the broad single-phase regime. Another critical issue in the growth of CuInSe2-based materials is the incorporation of Na into the film that modifies the growth habit of the film and appears to reside at the grain boundaries and improves the properties of the solar cell device [43]. The soda lime glass substrate provides a source of Na, which diffuses from the glass thru the Mo contact into the growing CuInSe2-based film. However, for manufacturing, a diffusion barrier is, generally, used on the glass and the Na is controllably incorporated into the film. Two primary approaches have been used to deposit CuInSe2-based materials. Multisource evaporation has been used to deposit the CuInSe2-based thin films used for the highest-efficiency solar cells. Figure 6.8 is a schematic of the deposition system. The incident flux and substrate temperature are varied during the film growth to control the through-film composition, grain structure, and bandgap. Typically, for a CuInGaSe2 film, the through-film composition of Ga and In is controlled by varying the incident flux during film growth, creating a gradient in the bandgap through the film that depends on the Ga to (Ga + In) ratio. Variation in the Cu flux during growth can affect the structure of the film, but the Cu diffuses rapidly through the film, resulting in a uniform distribution of Cu. The best laboratory solar cell was fabricated using a three-stage process where the initial growth is just GaIn-Se followed by Cu-Se and then Ga-In-Se [5]. However, solar cells with efficiencies around 17% have been grown by simpler uniform growth and two-stage processes, which are more compatible with manufacturing [44].

148

THIN-FILM SOLAR CELLS AND MODULES Substrate

Se sparger tubes

θ Se Cu Se Ga Se

In

Se

Z Cu, Ga, In sources

Y X

Figure 6.8. Schematic of an elemental source thermal evaporation in-line system for depositing CuInGaSe2 films on a flexible web.

Figure 6.9. System for reacting precursor materials to form CuInGaSe2 films.

The second approach to growing CuInSe2-based materials is the reaction precursor films in a Se2 and/or S2 atmosphere, generally H2Se and/or H2S, and is referred to as selenization or surfurization. Figure 6.9 is a schematic of the reaction system. This approach was initially commercialized by ARCO Solar, Siemens Solar Industries, and Shell Solar Industries [45] and is the focus of many companies since the reaction process is simpler to translate to manufacturing. The simplest example of the reaction process is to sputter deposit Cu-In-Ga layers and to react these precursors in H2Se gas at atmospheric pressure. Typically, the reaction temperature is limited to less than 500°C since the H2Se will react with the Mo contact at higher temperature. Additionally, as a result of the reaction kinetics, the Ga resides near the back of the film, and thus the benefits of incorporating Ga to widen the bandgap and to increase Voc are not realized. Sulfur has been incorporated at the front surface where the junction is formed to increase the Voc, and a small area

POLYCRYSTALLINE THIN FILM

149

device of about 16% has been reported by Siemens [46]. Several companies are developing precursor processes that eliminate the metal sputtering process based on inks consisting of nanoparticles containing Cu-In-Ga-Se or electrodeposited metals. The n-type window layer for many CuInSe2-based devices is CdS and has been used for the highest-efficiency laboratory cells as well in commercial modules. The CdS is deposited by a wet chemical process, referred to as chemical bath deposition (CBD), and conformal CdS films can be deposited with a thickness of about 50 nm to minimize optical losses in the CdS layer. There has been a significant effort to replace the CdS with a non-Cd-containing window layer to make the modules more environmentally friendly. Some of the materials that have been evaluated are Zn(S,O,OH) [47], ZnSe [48], ZnIn2Se4 [49], In2S3 [50], and Inx(OH,S)y [51], which were deposited by a variety of techniques such as CBD, sputtering, MOCVD, and ALE. However, the device performance is, in general, not as good as devices using CdS. As with CdTe devices, a high-resistive TCO is deposited on the CdS, which is usually ZnO and has a resistivity from 1 to 100 Ω-cm with a thickness of ∼50 nm. Since there is a significant probability of pinholes in CdS, the highresistive TCO eliminates the possibility the conductive TCO from contacting the CuInSe2-based absorber, which can result in secondary diodes. Finally, the conductive TCO is deposited, which is usually doped ZnO or ITO having a resistivity of 10−3 to10−4 Ω-cm. Manufacturing of CuInSe2-based modules is at the early stages of commercialization and is being led by Showa Shell, Honda, Wurth Solar, and Global Solar, which have production capacity in excess of 15 MW along with numerous start-up companies throughout the world. Both multisource evaporation and reaction of precursors in a Se2 and/or S2 atmosphere are being used for the deposition of the CuInSe2-based absorber on glass, polymer, and metal foil where roll-to-roll processing is being developed using polymer and metal foil webs. Module efficiencies of close to 12% have been reported, and cell efficiencies on the steel substrate are about 10%. Recently, Ascent Solar reported monolithic integrated modules with efficiencies greater than 9.5%, verified by NREL using a roll-to-roll processing on a polyimide web. CuInSe2-based modules have the potential to achieve 15% at the commercial level based on laboratory results but lack fundamental knowledge to control properties, and the process has limited progress. The challenge with elemental multisource evaporation is the high source temperatures, required to obtain high film growth rates; typically, the Cu source operates at ∼1600°C, coupled with a Se environment [52]. Design and control of the system, particularly when depositing on a continuous web, requires in situ process control and diagnostics [53], which are being developed by several companies. For the reaction of precursors, the reaction time to form the CuInSe2-based films is limited by the reaction/diffusion rates. For a flexible web, the challenge is to react the precursors on a moving web where uniform delivery of the gas and reaction time become important issues. For modules fabricated on a glass substrate, a batch process is used where a large

150

THIN-FILM SOLAR CELLS AND MODULES

number of plates are reacted in a furnace. Developing a robust high-throughput manufacturing process needs to be demonstrated along with low cost comparable to that demonstrated for CdTe modules to accelerate the CuInSe2-based PV technologies.

6.4

HYDROGENATED a-Si, a-Si ALLOYS, AND nc-Si

Thin-film a-Si solar cells/modules have been commercially available starting with SANYO’s first commercial production facility in 1980. Today, a-Si-based modules are available from a number of companies and typically have efficiencies from 5 to 8%. The challenge for a-Si-based materials has been to improve the stabilized efficiency, which is less than their initial efficiency due to light-induced degradation, the Staebler–Wronski effect [54], which is a fundamental property of the an a-Si material. The basic device is a p-i-n structure where the n- and p-layers are 10–20 nm thick and the intrinsic i-layer is from 100 to 200 nm, depending on the device design. Engineering approaches to minimize the light-induced degradation have focused on multijunction device structures that minimize the thickness of the intrinsic layer in order to reduce the light-induced degradation. a-Si can be characterized as a direct bandgap material; however, the conduction and valence band edges are not well defined as in crystalline material as a result of band tail states. To compare the bandgap of different a-Si materials, the optical absorption coefficient as a function of energy is analyzed according to Tauc, where an optical or Tauc bandgap is rigorously defined [55]. The optical bandgap for a-Si varies from 1.7 to 1.8 eV depending on the H2 concentration in the film and has an absorption coefficient >105 m−1 at 550 nm. The bandgap can be modified from 1 to >2 eV by alloying with Ge to narrow the bandgap (a-SiGex) or with C (a-SiCx), O2 (a-SiOx), or N2 (a-SiNx) to widen the bandgap. The a-Si can be doped p-type with boron and n-type with phosphine and is used as a thin-film transistor in the display industry. The amorphous nature of the film can be transitioned to a nanocrystalline structure by controlling growth conditions and has been exploited for p-type window layers, μc-Si, to enhance optical transmission and for i-layers, nc-Si, to increase the absorption above 850 nm for use as a bottom cell in a tandem device structure [56]. The design of thin-film a-Si-based solar cells/modules is the most flexible of any of the thin film technologies. Both superstrate and substrate devices have been made on glass, foil, and plastic substrates, all based on a p-i-n structure as shown in Figure 6.10. To achieve the best performance, the device is illuminated through the p-layer, but the order of the deposition depends on the specific device design. The n-layer is deposited first for the substrate configuration, and the p-layer is deposited first for a superstrate structure. The contact to the n-layer is generally ZnO with a metal reflector, independent of the device structure, to facilitate light trapping in the device. The contact to the p-layer is a TCO where both SnO2 and ZnO have been used for modules.

HYDROGENATED a-Si, a-Si ALLOYS, AND nc-Si

151

Figure 6.10. Device structures used for a-Si-based solar cells.

Most a-Si, μc-Si, and nc-Si films used for commercial modules are grown by a PECVD process using RF (13.56 MHz) or VHF (40–100 MHz) excitation where SiH4 is the primary precursor gas and the substrate temperature is between 150 and 250°C. The growth rate for “device quality” a-Si films is from 1 to 5 A/s for the RF PECVD and is over 20 A/s for VHF PECVD [57]. However, VHF PECVD has the advantage of improved stabilized performance at the higher growth rates compared to RF PECVD, but designing the VHF electrode for uniform deposition over large areas is more challenging. Other growth processes have been used including microwave PECVD, photo- and catalytic (hot wire) CVD, and sputtering. Performance, especially stabilized, decreases as the growth rate increases regardless of the deposition method. The utilization of the SiH4 is, generally, less than 10% depending on the design equipment. Figure 6.11 shows a schematic of an RF–PECVD system. The intrinsic a-Si layer is grown using a gas mixture, SiH4 and H2, typically referred to as hydrogen dilution (ratio SiH4/H2) to control the bandgap and crystalline fraction in the film. Device quality nc-Si films have a crystalline fraction that is typically around 60% and consists of 10- to 20-nm crystallites that are encased by an a-Si:H “tissue.” The optical properties of the nc-Si:H are a mix of a-Si:H and c-Si, and thus the thickness of the intrinsic nc-Si:H layer needs to be about 1–2 μm to absorb a significant fraction of the long wavelength light. The conductivity of the a-Si is controlled by adding a dopant gas to the SiH4/H2 gas mixture, either diborane (B2H6), borntrifloride (BF3), or trimethylboron (B(CH3)3), for doped p-typing or phosphine (PH3) for n-type doping. The structure of a p-type a-Si window layer, which does not contribute to the current in the cell due to poor

152

THIN-FILM SOLAR CELLS AND MODULES Gas inlet

Substrate RF Electrode Substrate Transport RF

Plasma

Pump

Burn box

Figure 6.11. RF–PECVD system for growing a-Si films.

carrier properties, is grown using high H2 dilution to grow a μc-Si film that improves the short wavelength transmission and has a higher crystalline fraction than the nc-Si i-layer. Commercialization of a-Si-based module technology is undergoing a rapid expansion with several companies selling “turnkey” systems for single and multijunction modules based on RF or VHF PECVD. Module dimensions are increasing with Applied Materials supplying equipment to manufacture a 2.5 × 2.3 m module on a glass substrate. The Prometheus Institute projections for the thin film manufacturing capacity in 2012 is 9.6 GW with a-Si leading the way with a capacity of ∼5.2 GW. These projections are very conservative since 2008 thin-film capacity has already exceeded the projections in the report [58]. a-Si modules face several challenges to remain competitive with other thin film technologies and with c-Si wafer-based modules. The module efficiencies are low, and it is difficult to project efficiencies beyond 10%. However, a-Si-based modules have the lowest temperature coefficient of any PV technology, and thus the annualized power output can be comparable to or higher than modules with higher-efficiency ratings (see Fig. 6.2). For example, the annualized power output of a-Si modules exceeds that of c-Si by 5–15% in a large number of studies [59]. Two critical issues to improve the viability of a-Si technology are (1) improvement in stabilized performance, which is being addressed by going to a nc-S/a-Si tandem structure and (2) increased growth rate to reduce capital costs, which is particularly important for the nc-Si bottom cell where the i-layer is 1.5–2.0 μm. However, there appears to be a trade-off of performance and processing speed. a-Si technology also has several advantages: (1) flexibility in device design and structure; (2) pro-

ABBREVIATIONS

153

cessing is done at low temperature, less than 200°C; (3) material availability is not an issue even though SiH4 utilization is low; and (4) the flat panel display industry underpins the technology and scientific base.

6.5

OUTLOOK FOR THIN-FILM MODULES

The production of thin-film PV modules will continue to expand and probably, the manufacturing capacity will exceed c-Si in the next 5–10 years, primarily replacing mc-Si module market share. CuInSe2-based manufacturing challenges will be overcome and modules will reach or exceed 15% on glass and possibly flexible substrates utilizing high-speed processing technology to drive the manufacturing cost well below $1 per watt. New companies will enter the CdTe module market and module efficiencies will approach 13% with manufacturing cost less than $0.65 per watt. There will be a competition between the polycrystalline and a-Si technologies, which will be driven by the application and will levelize cost of power. As the production capacity expands, material availability of In and Te may become a critical issue, which would benefit the a-Si technology and be a driver for the organic PV technologies. The future of PV will consist of a mix of c-Si and thin film technologies where cost, application, location, and aesthetics will determine the choice of the PV technology. Thin film technologies will be the modules of choice for residential roof application and commercial build integrated applications. Medium-scale power- and utility-scale PV arrays will use both polycrystalline and high-efficiency c-Si modules. However, in hot arid locations, very high-efficiency concentrator PV arrays will also be used for utility-scale applications.

ABBREVIATIONS ALE—atomic layer epitaxy a-Si—amorphous silicon C—carbon CBD—chemical bath deposition Cd—cadmium CdS—cadmium sulfide Cd2SnO4—cadmium stannate CdTe—cadmium telluride Cl—chlorine c-Si—crystalline silicon CSS—close-spaced sublimation Cu—copper CuInGaSe2—copper indium gallium diselenide Cu2S—copper sulfide CuInSe2—copper indium diselenide

154

THIN-FILM SOLAR CELLS AND MODULES

DOE—Department of Energy Eff.—efficiency FF—fill factor Ga—gallium Ge—germanium H—hydrogen Hg—mercury I—undoped or intrinsic semiconductor III—elements in column 3 of periodic table In—indium ITO—In2O3:Sn Jsc—short-circuit current MBE—molecular beam epitaxy mc-Si—multicrystalline silicon Mo—molybdenum MOCVD—metal-organic chemical vapor deposition N—nitrogen n—negatively doped semiconductor Na—sodium nc-Si—nanocrystalline silicon NREL—National Renewable Energy Laboratory O—oxygen P—phosphorous p—positively doped semiconductor Pd—palladium PECVD—plasma-enhanced chemical vapor deposition PV—photovoltaics or solar cells PVD—physical vapor deposition RF—radio frequency S—sulfur SAM—Solar Advisor Model Se—selenium Si—silicon SiH4—silane Sn—tin SnO2—tin oxide TCO—transparent contact oxide Te—tellurium V—elements in column 5 of periodic table VHF—very high frequency Voc—open-circuit voltage VT—vapor transport Zn—zinc ZnO—zinc oxide μc-Si—microcrystalline silicon

REFERENCES

155

REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24]

R. B. Hall, R. W. Birkmire, J. E. Phillips, and J. D. Meakin. Appl. Phys. Lett. 38, 925 (1981). R. A. Mickelsen and W. S. Chen. In Conf. Rec. 15th IEEE Photovoltaic Specialist Conf., pp. 800–803 (1981). Y.-S. Tyan and E. A. Perez-Albuerne. In Conf. Rec. 16th IEEE Photovoltaic Specialist Conf., pp. 794–800 (1982). A. Catalano, R. V. D’Aiello, J. Dresner, B. Faughnan, A. Firester, J. Kane, H. Schade, Z. E. Smith, G. Swartz, and A. Triano. In Conf. Rec. 16th IEEE Photovoltaic Specialist Conf., pp. 1421–1422 (1982). I. Repins, M. A. Contreras, B. Egaas, C. DeHart, J. Scharf, C. L. Perkins, B. To, and R. Noufi. Prog. Photovoltaics 16(3), 235–239 (2008). X. Wu, J. C. Keane, R. G. Dhere, C. DeHart, A. Duda, T. A. Gessert, S. Asher, D. H. Levi, and P. Sheldon. In Proceedings of the 17th European Photovoltaic Solar Energy Conference, October 22–26, 2001, Munich, pp. 995–1000 (2001). J. Meier, J. Sitznagel, U. Kroll, C. Bucher, S. Fay, T. Moriarty, and A. Shah. Thin Solid Films 451–452, 518–524 (2004). SolarBuzz Annual World Solar PV Industry Report. Available at http://www.solarbuzz.com/Marketbuzz2009-intro.htm (2009). First Solar Quarterly Report for 2009: www.firstsolar.com. K. Yamamoto, A. Nakajima, Y. Tawada, M. Yoshimi, Y. Okamoto, and S. Igari. In Proc. of the 2nd World Conference on Photovoltaic Energy Conversion, pp. 1284– 1290 (1998). J. Yang, A. Banerjee, S. Sugiyama, and S. Guha. In Conf. Rec. 26th IEEE Photovoltaic Specialist Conf., pp. 563–568 (1997). J. Zhao, A. Wang, M. A. Green, and F. Ferrazza. Appl. Phys. Lett. 73, 1991–1993 (1998). O. Schultz, S. W. Glunz, and G. P. Willeke. Prog. Photovoltaics 12, 553–558 (2004). A. M. Green, K. Emery, Y. Hishikawa, and W. Warta. Prog. Photovoltaics 17, 85–94 (2009). D. Cunningham, K. Davies, L. Grammond, E. Mopas, N. O’Connor, M. Rubcich, M. Sadeghi, D. Skinner, and T. Trumbly. In Conf. Rec. 28th IEEE Photovoltaic Specialist Conf., pp. 13–18 (2000). Y. Tanaka, N. Akema, T. Morishita, D. Okumura, and K. Kushiya. In Proc. 17th European Photovoltaic Solar Energy Conference, pp. 989–994 (2001). J. Yang, A. Banerjee, T. Glatfelter, K. Hoffman, X. Xu, and S. Guha. In Proc. 1st World Conference and Exhibition on Photovoltaic Solar Energy Conversion, pp. 380–386 (1994). D. Rose, O. Koehler, N. Kaminar, B. Mulligan, and D. King. In Proc. 4th World Conference and Exhibition on Photovoltaic Solar Energy Conversion, pp. 2018– 2023 (2006). D. L. King, W. K. Schubert, and T. D. Hund. In Proc. 1st World Conference and Exhibition on Photovoltaic Solar Energy Conversion, pp. 1660–1662 (1994). Available at http://www.nrel.gov/analysis/sam. A. M. Barnett, B. N. Baron, J. V. Masi, and T. W. F. Russell. Method for the continuous manufacture of thin film solar cells. U.S. Patent 4,318,938, March 9, 1982. S. Hegedus and B. McCandless. Sol. Energy Mater. Sol. Cells 88, 75–95 (2005). S. Hegedus and W Shafarman. Prog. Photovoltaics 12, 155 (2004). J. Jordan and S. Albright. Photovoltaic cell with thin CdS layer. U.S. Patent 5,279,678 (1994).

156

THIN-FILM SOLAR CELLS AND MODULES

[25]

T. Takamoto, T. Agui, H. Kurita, and M. Ohmori. Sol. Energy Mater. Sol. Cells 49, 219–225 (1997). B. McCandless and R. Birkmire. In Proceedings of the 28th IEEE PVSC, ed., pp. 491–494 (2000). B. McCandless. Unpublished work. X. Wu et al. In Conf. Rec. 28th IEEE Photovoltaic Specialist Conf., pp. 470–474 (2000). B. McCandless, L. Moulton, and R. Birkmire. Prog. Photovoltaics 5, 249–260 (1997). F. J. Kruger. Electrochem. Soc. 125, 2028–2034 (1978). Y. Tyan and E. Perez-Albuerne. In Conf. Rec. 16th IEEE Photovoltaic Specialist Conf., pp. 794–800 (1982). R. Powell, G. Dorer, N. Reiter, H. McMaster, S. Cox, and T. Kahle. Apparatus and method for depositing a material on a substrate. U.S. Patent 5,945,163 (1999). B. McCandless, I. Youm, and R. Birkmire. Prog. Photovoltaics 7, 21–30 (1999). R. Wendt, A. Fischer, D. Grecu, and A. J. Compaan. Appl. Phys. 84, 2920–2925 (1998). R. Sudharsanan and A. Rohatgi. Sol. Cells 31, 143–150 (1991). J. Jordan. Photovoltaic cell and method. International Patent Application WO93/14524 (1993). S. Ikegami. In Technical Digest of the 3rd Photovoltaic Science and Engineering Conference, pp. 677–682 (1987). G. M. Hanket, B. E. McCandless, W. A. Buchanan, S. Fields, and R. W. Birkmire. J. Vac. Sci. Technol. A, 24(5), 1695 (2006). B. McCandless and J. Sites. In Handbook of Photovoltaic Science and Engineering, A. Luque and S. Hegedus, eds., pp. 633–644. John Wiley & Sons (2003). R. Birkmire. In Conf. Rec. 33rd IEEE Photovoltaic Specialist Conf., paper 370 (2008). W. Shafarman, R. Klenk, and B. McCandless. In Conf. Rec. 25th IEEE Photovoltaic Specialist Conf., pp. 763–768 (1996). T. Yoshida and R. W. Birkmire. In Proc. 11th European Photovoltaic Solar Energy Conference, pp. 811–816 (1992). M. Bodegard, K. Granath, A. Rockett, and L. Stolt. Sol. Energy Mater. Sol. Cells 58, 199–208 (1999). W. Shafarman and J. Zhu. Thin Solid Films 361–362, 473–477 (2000). R. Gay, M. Dietrich, C. Fredric, C. Jensen, and K. Knapp. In Proc. 12th European Photovoltaic Solar Energy Conference, pp. 935–940 (1994). D. Tarrant and J. Ermer. In Conf. Rec. 23rd IEEE Photovoltaic Specialist Conf., pp. 372–375 (1993). A. Ennaoui, U. Blieske, and M. Lux-Steiner. Prog. Photovoltaics 6, 447–451 (1998). Y. Othake, K. Kushiya, A. Yamada, and M. Konagai. In Proc. 1st World Conference and Exhibition on Photovoltaic Solar Energy Conversion, pp. 218–221 (1994). Y. Othake, T. Okamoto, A. Yamada, M. Konagai, and K. Saito. Sol. Energy Mater. Sol. Cells 49, 269–275 (1997). J. Guillemoles, B. Canava, E. Yousfi, P. Cowache, A. Galtayries, T. Asikainen, M. Powalla, D. Hariskos, H. Schock, and D. Lincot. Jpn. J. Appl. Phys. 40, 6065–6068 (2001). D. Hariskos, R. Herberholz, M. Ruckh, U. Ruhle, R. Schaffler, and H. W. Schock. In Proc. 13th European Photovoltaic Solar Energy Conference, pp. 1995–1998 (1995). S. T. Junker, R. W. Birkmire, and F. J. Doyle III. AIChE J. 51(3), 878 (2005).

[26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52]

REFERENCES [53] [54] [55] [56] [57] [58]

[59]

157

K. Mukati, B. A. Ogunnaike, E. Eser, S. Fields, and R. Birkmire. Ind. Eng. Chem. Res. 48, 5975–5991 (2009). D. Staebler and C. Wronski. Appl. Phys. Lett. 31, 292 (1977). J. Tauc. In Optical Properties of Solids, F. Abeles, ed., pp. 277–313. Amsterdam, North-Holland (1972). A. Shah, J. Meier, P. Torres, U. Kroll, D. Fischer, and N. Beck. In Conf. Rec. 26th IEEE Photovoltaic Specialist Conf., pp. 569–574 (1997). H. Takatsuka, M. Noda, Y. Yonekura, Y. Takeuchi, and Y. Tamauchi. Solar Energy 77, 951–960 (2004). S. Grama and T. Bradford. Thin-Film PV 2.0: Market Outlook Through 2012: Executive Summary, the Prometheus Institute. Available at http://www.greentechmedia.com/GreentechMedia/Report/Thin-FilmPV20MarketOutlookThrough2012. html. S. Hegedus. Thin film solar modules: The low cost, high throughput and versatile alternative to Si wafers. Prog. Photovoltaics 14, 393 (2006).

7 TERRESTRIAL MODULE FABRICATION AND ASSEMBLY TECHNOLOGIES CHRISTOPHER BUNNER Spire Corporation

7.1

INTRODUCTION

The global desire to produce energy from various renewable energy options increases as political, environmental, and economic landscapes transform. As expected, the solar industry benefits from this renewed attention where governments recognize and support the need for renewable energy options through grants, low tariffs, and deferred taxes to diminish the consumption of fossil fuels. Unfortunately, environmental pressure remains tied to the cost of ownership. The regional desires for renewable solutions directly correlate with governmental financial incentives. Fortunately, the solar industry has attracted notice from complementing industries as their primary customers stagnate. This attention provides new ideas and initiatives that benefit the solar module manufacturing process. Overall, the goal of terrestrial module manufacturing remains the same: to consistently produce the least expensive and most reliable module. While the basic module construction and manufacturing processes have changed little, the methods and equipment available have matured. This chapter examines the available improvements to the module fabrication process through the evolution of materials and equipment.

7.2

MATERIALS

While the solar industry continues to grow and mature, the necessary materials to construct a silicon-based photovoltaic (PV) module continue to encompass over Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

159

160

TERRESTRIAL MODULE FABRICATION Materials, 0.413, 73.1%

Labor, 0.032, 5.6% Op. & Maint., 0.020, 3.5%

Depreciation, 0.043, 7.7%

OH & G&A, 0.057, 10.1%

Figure 7.1. A breakdown is shown for the module cost elements for 220-W crystalline silicon modules produced by an automated production line at 50 MW/year excluding the cell costs. The module cost elements, excluding the cell cost, in dollar per watt, account for a total cost = $0.565 per watt.

60% of the overall module costs. This driving cost creates opportunities for material improvements and cost reduction pressure. The increasing number of module manufacturers has simultaneously consumed available material resources and have challenged material suppliers to increase their capacities. Spire Solar for many years has been doing comprehensive and exhaustive cost analysis studies for module manufacturing costs. Figure 7.1 shows a breakdown for the module cost elements for 220-W crystalline silicon modules produced by an automated production line at 50 MW/year. The module cost elements, excluding the cell cost, in dollar per watt, account for a total cost = $0.565 per watt. Silicon prices directly affect the growth of the solar industry. The silicon wafer remains the highest-priced component within the module. Silicon manufacturers have been challenged over the past decade to predict the consumption needs of the volatile semiconductor market as well as the rapidly growing solar industry. For years, the solar industry has consumed any excess higher grade and the lower grades of available silicon feedstock not desirable to the semiconductor industry. However, as capacities and innovation grow, the solar industry’s appetite for highquality silicon increases. As expected, this increasing demand, coupled with stationary supplies, leads to increasing silicon feedstock prices and higher wafer, cell, and module costs. Increasing silicon production capacities require a large commitment of capital and equipment. Over the last several years, the success of the solar industry has convinced silicon manufacturers to increase capacities. This increased capacity is just beginning to yield lower costs. Cell prices have begun to trend downward. A major driver in this cost reduction, however, has been the fabrication of thinner silicon wafers into solar cells. Where 10 years ago the “standard” solar cell thickness was 250–300 μm, today’s solar cells typically are fabricated from 180- to 200-micron wafers.

BASIC MODULE ASSEMBLY

161

These thinner cells have presented special needs for the manufacturing process to achieve the required yields necessary to take advantage of the thinner material. These special requirements include low-impact cell handling, small stacks if cells are coin stacked, more stringent heating/cooling requirements, and a myriad of machine design features to maintain manufacturing process yield. Over the last decade, traditional silicon solar cell efficiency has increased from around 12% to 17%. While the individual PV cells become more efficient and evolve into larger formats, their versatility remains. Cells may be arranged in various series and parallel configurations to meet system requirements. Three years ago, a typical module produced 70 W through the interconnection of 36 125-mm cells in series. Today, the standard module design is 220 W using 60 156-mm cells in series. A larger cell size yields various benefits. First, by increasing cell area, cell power is increased. While each cell produces about 0.6 V, regardless of size, as surface area increases, cell current increases. Second, as the cell efficiency is improved, fewer cells are required to produce the same module power. Additionally, some cell manufacturers have taken additional paths to improving cell efficiency. Reducing or eliminating the top surface contacts by use of a rear-contact cell structure has gained more absorption area. Furthermore, the use of larger and more powerful cells reduces cell handling, increases solar cell production capacities, and reduces packaging and shipping costs. Cost reductions achieved in the production of cells directly relate to reduced module costs. While these wafers and cells will require additional handling and processing requirements, the efficiency improvements will likely benefit other solar applications. Cell testing and classification is an important process step for both the cell and module fabrication processes. While cell testing quantifies cell performance, it also enables cell sorting into performance bins, which allows module fabricators to maximize power output. Essentially, the lowest-performing cell within a series string will limit the output current and the subsequent power produced by the solar module. This limiting effect also extends to the installed system. Thus, system designers expect narrow module classes to reduce loses from module mismatching (Fig. 7.2).

7.3

BASIC MODULE ASSEMBLY

While alternative materials are available for flexible solar products, low iron tempered glass remains the standard load-bearing component for today’s silicon-based solar module. The glass must be fully cleaned to ensure adequate adhesion to the subsequent module layers. After the glass is permitted to dry, the first encapsulant layer is added. EVA continues to be the primary encapsulating plastic used in PV modules. Today’s faster formulations have cut lamination cycle times in half. Additionally, EVA has demonstrated improved reliability, and UV stabilizers have been added

162

TERRESTRIAL MODULE FABRICATION

Figure 7.2. Picture of cell testing machine. In addition to machine throughput improvements now with over 1000 cells/h, today’s cell test machines must be able to handle everthinning cell thicknesses.

to protect the laminate package against the sun’s spectrum and have prevented the degradation of EVA. Other encapsulant materials have continued to play an active role in the production process. Furthermore, some thin-film processes may be incompatible with the EVA process and have returned to using PVB—a material used in the 1970s before EVA became the prominent encapsulant. Next, solar cells are interconnected into series strings. While some module manufacturers use manual labor to interconnect solar cells, most new entrants to module building are utilizing available automated tabbing and stringing equipment. There are various soldering approaches, including conduction (hot bar), convection (hot air), and radiation (infrared lamps). The standardization of cell contact printing has enabled equipment manufacturers to standardize alignment and soldering equipment. Roughly 5 years ago, four stringing machines were required to support a 12-MW line. Today, the standard throughput exceeds 500 cells/h. Coupled with larger solar cells, one stringing machine can connect soldered strings to support a 12-MW production line (Fig. 7.3). The connection of cell strings (busing) remains similar. The strings are connected using a thicker and/or wider bus wire (ribbon) that brings the module current and voltage through an egress hole or slit in the back sheet ultimately terminated in a junction box. After the cell circuit is completed, an optional fiberglass layer is added to ensure cell ribbon isolation from the rear layer of the module. Additionally, this layer aids air evacuation and minimizes cell string movement during the lamination process.

BASIC MODULE ASSEMBLY

163

Figure 7.3. Photo of automated tabbing and stringing machine. Backsheet

Fiberglass

EVA

Cells

Glass

Figure 7.4. Picture of laminate layers.

A second sheet of encapsulant and the back-sheet layers are added to encapsulate the power-producing section of the module. While module designs may vary based on system requirements, all current designs require multiple power leads to egress the rear layer for string output monitoring. While PVF is still the dominant back-sheet material, increasing module production has stressed the available supply. This has encouraged additional material suppliers to explore other material options. These newer materials are beginning to supplement PVF shortages. Additionally, glass back sheets remain a viable option for the rear surface, especially in BIPV and in thin-film modules. Modern production equipment, such as laminators, must exhibit the flexibility to handle these various module constructions (Fig. 7.4). The layered package is placed into a lamination machine. The cycle time will depend on the curing time required for the encapsulant layers. Typically, the lamination cycle is divided into to two phases. First, the air is evacuated from the heated

164

TERRESTRIAL MODULE FABRICATION

processing chamber while the encapsulant layers melt. During the second phase, pressure is applied onto the laminate package to evacuate any remaining air. The total lamination time is approximately 10–20 min, depending on materials. After the laminate is permitted to cool, excess encapsulant and back-sheet material is trimmed; module frames, a junction box, and cables are added. Within the junction box, bypass diodes are incorporated to sense reverse current. A reverse current condition is possible if one or more cells are shaded or covered during daylight hours. Essentially, the cell would not pass the module current; the cell would overheat and potentially break the module glass. The installed bypass diodes protect the module until the shading is resolved. The final testing of modules can be a labor-intensive exercise. The module must be presented to a sun simulator. The module’s cables must be connected. Next, the module is transferred to a required high-voltage tester, where the module cables and frame are connected. The test results are used to rate the module and to monitor module line performance. Ultimately, after combining all of these module materials, the resulting module is expected to last in excess of 25 years.

7.4

EQUIPMENT OPTIONS

While module manufacturers endeavor to produce the best or the cheapest module to remain competitive, the equipment manufacturers must evolve to effectively handle new materials and larger module formats. The function of the equipment is to minimize handling and stress on the solar cells whenever possible. Some module manufacturers chose to establish or move factories to economically challenged areas, to take advantage of lower labor rates. Alternatively, more automation may be incorporated to reduce labor costs. The line concept has evolved to include powered conveyors and robotics. The addition of conveyors is a natural evolution as the solar industry matures. Conveyors are available to transport the glass through the factory as it transforms into a laminate and ultimately into a solar module. By utilizing conveyors, automation may be integrated at the various process steps. Essentially, by limiting operator contact, the resulting module can be produced faster and more consistently. Robots are now engaged throughout the module assembly process. The introduction of robotics was assisted by experience from other industries. While operators feed material into the work cells, robots perform the assembly steps and advance the module through the processes. For example, an automated 50-MW module assembly line utilizes robotics to load the module glass, to complete soldered cell string layup, and to install bus ribbon prior to the lamination process (Fig. 7.5). The fundamentals of busing and laminate assembly remain similar; however, the implementation of automation has allowed module manufacturers to manage these process steps more efficiently. For example, the utilization of robots transforms the soldering and assembly processes into a timed process step eliminating the dependence on an operator’s level of experience (Figs. 7.6 and 7.7).

EQUIPMENT OPTIONS

165

Figure 7.5. Photo of glass loading. A glass-loading robot eliminates labor from the beginning of the assembly process.

Figure 7.6. After cell soldering, the robot places completed strings onto the prepared glass.

Some module equipment is developed to reduce handling and others are developed to increase throughput. Larger lamination machines address both goals. Equipped with conveyors to automatically load and unload laminates, one laminator can now laminate four 220-W laminates per cycle. Coupled with the fastest cure EVA, just two laminators are necessary for a 50-MW fully automated module line (Fig. 7.8).

166

TERRESTRIAL MODULE FABRICATION

Figure 7.7. Robots with specially equipped soldering arms complete the module circuit.

Figure 7.8. Photo of larger area laminator. Larger area laminators integrated with conveyor systems provide higher throughput with minimal to no handling.

Additionally, robots perform the edge trimming, framing, and junction box placement process steps as the laminate travels along conveyors to the sun simulator (Figs. 7.9–7.11). Larger modules are very important to reducing system costs. While modules are purchased by module power (watts), the overall system costs are lower due to fewer junction boxes, cables, and frames. Additionally, fewer modules per installation reduce the number of DC connections and simplify system mounting requirements. This has led to the development of special equipment designed to handle the increased module sizes. Today, automation can eliminate manual lifting,

EQUIPMENT OPTIONS

167

Figure 7.9. Module frames are affixed in a robotic work cell.

Figure 7.10. Photos of framing and junction box application.

loading, and connecting from the operation. Automation can also apply the power label after testing. New ideas and initiatives that benefit the solar module manufacturing process will emerge from the increasing attention enjoyed throughout the renewable energy industries. Technology improvements will likely compliment other solar segments. For example, alternative and more efficient cell designs may be incorporated into concentrator and BIPV applications.

168

TERRESTRIAL MODULE FABRICATION

Figure 7.11. The completed module is conveyed over the sun simulator for electrical performance test results. Next, the module is transported through the high-voltage tester. Lastly, the performance and safety labels are affixed to the rear side of the module.

Figure 7.12. Picture of building integrated PV (BIPV) project at Denali National Park Visitor Center in Alaska.

Through additional governmental incentives, it becomes more popular to integrate solar products into new and existing structures. Attractive BIPV projects capitalize on the versatility of silicon cells, encapsulants, and other module materials (Fig. 7.12).

7.5

FUTURE

A combination of improved cell efficiency, reduced material costs, and governmental support shall increase the amount of electricity generated from silicon-

ABBREVIATIONS

169

based modules in the coming years. Ultimately, the renewed attention and global push for renewable energy solutions will invigorate the solar industry. The collaboration of ideas and people from other industries will further challenge today’s best practices and increase productivity while lowering system costs. While this growth is beneficial, in order for the industry to evolve, lower material costs are needed. The availability and quality of bus wire for cell-to-cell and circuit connections, encapsulating plastics, and composite back-sheet materials directly affect module manufacturing. While the basic module construction and manufacturing processes have changed little, the methods and equipment available have matured. ABBREVIATIONS BIPV—building integrated photovoltaics DC—direct current EVA—ethylene vinyl acetate PVF—polyvinyl fluoride PVB—polyvinyl butyral UV—ultraviolet

8 CHINESE SOLAR CELL STATUS WANG SICHENG Energy Research Institute, National Development and Reform Commission

8.1

INTRODUCTION (BY THE EDITOR)

Driven by the pressure of global warming, rising oil prices, and a deteriorating natural environment, clean renewable energy is receiving greater attention worldwide. China’s central government has stated its support for the development of a sustainable energy system that maximizes energy efficiency and the use of renewable energy sources. A key aspect of that initiative is the application of PV technology to convert light into electricity. The Chinese PV manufacturing industry has grown dramatically in recent years due to a strong demand from overseas markets. China’s production of solar cells and modules has grown at an average annual rate of 49.5% since 2002. By 2008, the production of solar cells reached over 2 GW, which was 33% of the global production. China now contributes a large part of the worldwide solar production and is now the largest producer of solar cells in the world. Intelligent and supportive Chinese central government policy has contributed to this rapid growth in the Chinese solar cell industry. This chapter provides an overview of the coordinated efforts in China to develop a strong domestic PV industry. This chapter was written by Wang Sicheng, who has contributed to the formation of China’s government policy. It was translated into English by Huang Han Xinag at JX Crystals Inc., and editorial comments are by Lewis M. Fraas. 8.2

CHINA’S SOLAR ENERGY RESOURCES AND DATA

8.2.1

Solar Irradiation Resources in China

PV power generation largely depends on the solar energy resources of the location. Therefore, it is necessary to know the distribution of the solar energy resources in Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

171

172

CHINESE SOLAR CELL STATUS

II I III

IV

Figure 8.1. Geographic distribution of the solar irradiation of China. The territory can be divided into four resource levels shown. TABLE 8.1. Solar Irradiation Resources in China on a Horizontal Surface Level of Irradiation

Region Number

Annual Irradiation Max (MJ/m2)

Annual Irradiation Min (kWh/m2)

Daily Irradiation Average (kWh/m2)

Very high

I

≥6300

≥1750

≥4.8

High

II

5040–6300

1400–1750

3.8–4.8

Average

III

3780–5040

1050–1400

2.9–3.8

Low

IV

<3780

<1050

<2.9

the country. The distribution of solar energy resources has regional limitations, including restrictions by weather and geographic conditions. As shown in Figure 8.1, the territory of China can be divided into four solar energy resource regions in terms of the amount of solar irradiation. Annual irradiation for these four solar energy regions is shown in Table 8.1. Solar irradiation data can be acquired from county meteorological stations or from the national meteorology bureau. China has measured solar irradiation since 1953. At present, China has 2,610 meteorological observatories/stations, compared

PV R & D LEVEL AND TECHNOLOGY INNOVATIONS IN CHINA

173

with 70 in the early 1950s. Of these, 2,300 are meteorological stations (county level) and 310 are observatories (including 122 standard observatories). Before 1993, there were 66 observatories gathering solar global, scatter, and direct irradiation data. After 1993, only 17 observatories kept the global and scattering data, while the others stopped measuring scattering and direct irradiation. As summarized in Table 8.1, currently only horizontal global irradiation data can be obtained from most observatories. 8.2.2

Solar Resource Overview and Global Comparison

In China, horizontal irradiation data can be obtained from local meteorological stations. Based on these data, the irradiation on slope solar modules can be computed. This irradiation is normally about 10–15% higher than the horizontal irradiation. The additional amount depends on the site, latitude, and the ratio of direct and indirect irradiation. With this input, the output power of a PV system can be derived from the irradiation on the slope PV arrays. The appendix at the end of this chapter presents tables for the solar radiation in various provinces in China. Table 8.2 presents solar resource data for three representative locations in China with comparison data from locations in the United States and Europe. Data are presented for modules mounted horizontally and tilted to the South. Data are also presented for modules mounted on 1-axis solar trackers. Note that the solar resource is always higher with tracking regardless of location, as discussed in Chapter 9 in this book. The eastern coastal region of China, which is depicted in Figure 8.1, has average solar irradiance levels (III). Shanghai is representative of this region in Table 8.2. While the irradiance in Shanghai is lower than most of the United States, including Boston, it is well above the irradiance levels in all of Germany. In contrast, the western region of China, as represented by Beijing (II), has higher irradiance levels. As Table 8.2 indicates, the irradiance level in Beijing is similar to that of Madrid, Spain, and Medford, Oregon, in the United States, although it is not as high as the southwestern United States. China also has areas where the solar resource is well above average. For example, in Tibet (I) the resource is higher than Phoenix, Arizona, which is located in the southwestern desert region of the United States. This higher solar irradiance in western China, together with China’s goal of promoting economic development in western China, motivates the development of solar energy.

8.3 PV R & D LEVEL AND TECHNOLOGY INNOVATIONS IN CHINA 8.3.1

Solar Cell R & D Level in the World

At present, the major restrictive factor for developing a PV power system is the high cost. Solar cells accounts for over 60% of the cost of a PV power system.

174

CHINESE SOLAR CELL STATUS

TABLE 8.2. Solar Availability in China Relative to United States and Europe Annual Average Hours per Day on Horizontal Surface

Annual Average Hours per Day at Fixed Tilt

Tilt Angle

Beijing (39.6)

4.3

4.8

42

Shanghai (31)

3.6

4.0

35

Xizang (Tibet)

6.0

6.7

30

Medford, Oregon

4.4

4.9

42

6.5

Sacramento California

4.9

5.5

39

7.4

Boston, Massachusetts

3.9

4.6

42

5.7

Phoenix Arizona

5.7

6.5

33

8.6

Madrid (40), Spain

4.9

30

Lisbon (38), Portugal

5.4

30

Berlin (52), Germany

2.8

30

Region

City and Latitude

China

United States

Europe

Annual Average Hours per Day 1-Axis Tracking

Therefore, developing low-cost, high-efficiency, high-reliability, high-stability, and long-lifetime solar cells becomes the highlight of the PV technology worldwide. Currently, the R & D of the solar cells mainly focuses on the commercialized Si crystalline solar cells, a-Si solar cells, CdTe solar cells, CIGS solar cells, and concentration solar cells. Attention is also given for next-generation solar cells, which have not been commercialized yet, including crystalline Si thin-film solar cells, dye-sensitized cells, organic thin-film cells, nanometer cells, and spectral absorption cells, through large finance and research efforts (Table 8.3).

8.3.2

PV Technology Innovations in China

In China, R & D of solar cells started in 1958. In 1971, the domestic solar cells were first used successfully on the Chinese satellite DFH-II. In 1973, solar cells were used in terrestrial applications. Since 1981, the R & D of solar cells and their applications have been listed on China’s national important scientific and technology plans. During the “Sixth to Eleventh 5-Year Plan” (2006–2010), solar cell components and application technologies have achieved great progress. Since

PV R & D LEVEL AND TECHNOLOGY INNOVATIONS IN CHINA

175

TABLE 8.3. World Leading Edge Levels of Solar Cells in Lab The World Highest Efficiency of Solar Cells in Lab Type of Cells

Efficiency (%)

Researcher

Note

Monocrystalline Si solar cell

24.7 ± 0.5

University of New South Wales, Australia

4 cm2

Back-leads concentration monocrystalline Si solar cell

26.8 ± 0.8

Sun Power, USA

96 X

GaAs multijunction cell

41.1 ± 0.5

Fraunhofer Institute, Germany

Concentration

Multicrystalline Si solar cell

20.3 ± 0.5

Fraunhofer Institute, Germany

1.002 cm2

InGaP/GaAs

30.28 ± 1.2

Japan Energy Co.

a-Si solar cell

14.5 (initial) ± 0.7

4 cm2

USSC, USA

0.27 cm2

12.8 (stable) ± 0.7 CIGS

19.5 ± 0.6

NREL, USA

0.410 cm2

CdTe

16.9 ± 0.5

NREL, USA

1.032 cm2

Multicrystalline Si thin-film cell

16.6 ± 0.4

University of Stuttgart, Germany

4.017 cm2

Nanocrystalline Si solar cell

10.1 ± 0.2

Japan

2-nm film

Dye-sensitized cell

11.0 ± 0.5

EPFL

0.25 cm2

HIT

21.5

Sanyo, Japan

2000, the national “863” project and “973” project have been initiated by the Science and Technology Ministry. The industrialization technologies and basic research of PV power have been supported from these projects. The important scientific achievements and technological innovations are described in the following sections. 8.3.2.1 Polysilicon Raw Material High-purity polysilicon raw material has always been a bottleneck for the development of China’s PV industry. Before 2005, all of the semiconductor-grade and solar cell-grade polysilicon material used in China was imported. In addition, import of related technologies has been limited under an embargo to China. Therefore, the development of high-purity polysilicon manufacturing technologies was listed in the national “Tenth 5-Year” Science and

176

CHINESE SOLAR CELL STATUS

TABLE 8.4. China’s Production Output and Requirement of High-Purity Polysilicon Year

2006

Solar cell production output (MW)

2007

2008

370

1088

2000

4000

11,000

20,000

300

1100

4500

Shortage (ton)

3700

9900

15,500

Self-fulfillment percentage

7.50%

High-purity polysilicon requirement (ton) High-purity polysilicon production output (ton)

10%

22.50%

Technology R & D Plan. Several refinement processes such as the Siemens method, flowing-bed method, and metallurgy method were investigated and brainstormed. At the same time, the R & D related to solving problems caused by waste materials and gases from production, and to protecting the environment, was also granted financial support. As a result, breakthroughs have been made on cleaner manufacturing technologies for high-purity polysilicon. For example, the production technology in some major high-purity polysilicon suppliers (Luoyang China Silicon, Sichuan Xinguang Silcon, Xuzhou China Energy, Emei Semiconductor, Zhongqing Daquan, etc.) has come up to the world average level. The production output has also been growing yearly. With even further scientific and technical efforts, the indexes of unit energy consumption, quality, purity, and clean production of China’s polysilicon manufacture should achieve advanced world standards. In addition, the quantity should fulfill the requirement of the country’s solar cell production (Table 8.4). 8.3.2.2 Crystalline Si Solar Cells Since the “Sixth 5-Year” Plan, crystalline Si solar cell technology has always been a key point for science and technology R & D. The technologies for monocrystalline solar cells, multicrystalline solar cells, high-efficiency solar cells, the industrialization for crystalline solar cells, and the development of special function solar modules (BIPV) have been given ample attention and funding and, therefore, have made encouraging progress. The efficiency of lab monocrystalline solar cells is up to 20.4% and the efficiency of production of multicrystalline solar cells is normally up to 16.5% (Fig. 8.2). In addition, China can now make its own 500-kg-grade polysilicon ingot furnaces (Fig. 8.3). The capacity of polysilicon ingot furnaces was up to 40-kg grade during the “Seventh 5-Year” period and to 160-kg grade in the “Tenth 5-Year” period.

8.3.3

Thin-Film Solar Cells

Given the relative low cost of thin-film solar cells, the development of these cells has always been highly supported by the Chinese government in PV power tech-

PV R & D LEVEL AND TECHNOLOGY INNOVATIONS IN CHINA

177

Figure 8.2. Monocrystalline and multicrystalline Si solar sells.

Figure 8.3. A 500-kg-grade polysilicon ingot furnace developed by China.

nologies. During the period of the “Seventh 5-Year” Plan, the Science and Technology Committee invested 20 million yuan to set up a 100-kW annual capacity thin film production line at the Beijing Non-Ferrous Metal Research Institute. Since the beginning of the twenty-first century, under the significant support of the government through the National Science and Technology R & D Plan, “863” project, “973” project, and other innovation projects, the R & D for industrialization technologies of thin-film solar cells such as amorphous, CdTe, CuInGaSe, dye sensitization, and micro/amorphous crystalline solar cells have shown remarkable progress. Thus there is now a solid foundation for the industrialization of the thinfilm solar cells in China. 8.3.3.1 CdTe Solar Cell The CdTe thin-film solar cell is a new type of solar cell. Compared with other types of thin-film cells, it has a lower cost and higher

178

CHINESE SOLAR CELL STATUS

Figure 8.4. A 300-kW CdTe solar cell pilot production line.

efficiency. The Science and Technology Ministry began supporting the research of CdTe thin-film solar cell by setting it as a subproject under the “Ninth 5-Year” Science and Technology R & D Plan. In 2006, under the support of the national “Tenth 5-Year 863” project, “the research and pilot production line of CdTe thinfilm solar cells,” a first CdTe thin-film solar cell pilot production line having a Chinese intellectual property right was built at the Solar Energy Material Research Institute of Sichuan University. Its annual production capacity is 300 kW for 40 × 40 cm, 8.25% efficiency cells. This achievement played a positive role in promoting the PV technologies in China (Fig. 8.4). 8.3.3.2 CIGS Solar Cells The CIGS solar cell is a thin-film solar cell with a promising future due to its low cost, high efficiency, high stability, and responsiveness to weak intensity. Under the support of the Science and Technology Ministry’s R & D project in the “Eighth 5-Year” to “Ninth 5-Year” Plans and the “863 key task” in the “Tenth 5-Year” Plan, the Nankai University has developed and built a CIGS experiment platform and technology transfer center. It has developed over 14% efficiency small-area CIGS cells on glass substrate and 9.2% and 10.6% efficiency CIGS cells on polyamide and on stainless steel substrates, respectively. The university has overcome several technical obstacles and completed a process flow for large-area cell experimental line and successfully made an 804-cm2 effective area, 7% efficiency cells. With these results, the university has leaped from a small-area lab technology to a large-area prototype technology and established a solid foundation for developing its own intellectual property right production lines (Fig. 8.5).

PV R & D LEVEL AND TECHNOLOGY INNOVATIONS IN CHINA

179

Figure 8.5. CIGS thin-film solar cells developed by Nankai University.

8.3.3.3 Si Substrate Thin-Film Solar Cells The Si substrate thin-film solar cell has the advantages of low cost, abundant raw material, no poison, no pollution, short energy payback time, and ease of large-scale (LS) continuous production. Under the support of the Science and Technology Ministry’s R & D schedule in four 5-Year Plans, from the “Fifth 5-Year” to “Ninth 5-Year” and the “793” task in the “Eleventh 5-Year” Plans, the Nankai University has made a breakthrough on a-Si/μc-Si tandem cell research—the initial efficiency of a 400-cm2 integrated tandem cell assembly reached 9.2%. As a result, the company Tianjin Jinneng Cell Science and Technology Ltd. has been established with the backing of the university. Under the support of the “973” task in the “Tenth 5-Year” and “Eleventh 5-Year” Plans, Nankai University carried out research for the next generation of Si thin-film cells—a-Si/μc-Si tandem solar cells—and developed a first μc-Si thinfilm solar cell experimental platform with advanced world standards. The efficiency reached 11.8% for small-area a-Si/μc-Si tandem cells and 9.7% for 10 × 10 cm integrated a-Si/μc-Si tandem cell assemblies. Recently, with the combination of VHF and HV techniques, at 12 A/s microcrystalline film deposition rate, the efficiency of a single-junction μc-Si cell has reached 9.36%, ranked in a world advanced level (Fig. 8.6). 8.3.3.4 Dye-Sensitized Solar Cell As early as 1994, the Plasma Institute of China Science Academy started researching and developing the dye-sensitized solar cells and achieved noticeable results. Especially, under the support of the Science and Technology Ministry’s projects in the “Tenth 5-Year” and “Eleventh 5-Year” Plans, the maximum efficiency of small-area solar cells reached 9% and large-area cells (40 × 60 cm2) reached 5.7%. This achievement is one of the better research results in the world at present. In addition, they also had some encouraging research results in the areas of dye synthetic technology, nanometer semiconductor thin film, and cell encapsulation and electrodes. Moreover, the institute has designed and synthesized many new dye light sensitization materials and investigated the materi-

180

CHINESE SOLAR CELL STATUS

Figure 8.6. PECVD system developed by Nankai University.

Figure 8.7. A 500-W solar power system with dye-sensitized solar cells.

als’ absorbability and sensitization behavior in space. With these key technologies, the solar modules were assembled using 15 × 20 cm2 cells. Under 1-sun intensity, the module’s efficiency was 5.9%. By the end of 2004, a 500-W dye-sensitized cell solar power model system was successfully built and placed in operation to test its stability. It made a solid foundation for future industrialization (Fig. 8.7). 8.3.3.5 Other Types of Solar Cells Under the support of the Science and Technology Ministry, encouraging progress has been made in research on other types of solar cells in China. Table 8.5 summarizes the lab efficiency of the solar cells in China.

PV SOLAR POWER SYSTEMS AND KEY EQUIPMENT

181

TABLE 8.5. The Max Efficiency of Other Types of Solar Cells in China Solar Cell Type

Max Efficiency (%)

Researcher

Area (cm2)

GaAs cell

29.25

Tianjing Electric Power Institute

1×1

CIGS

14.3

Nankai University

0.87

CdTe

13.38

Sichuan University

0.502

7.4

Plasma Institute, Science Academy

10.2 0.253

Dye-sensitized cell a-Si/µc-Si tandem solar cells

11.8

Namkai

HIT

17.27

Graduate School, Science Academy

8.4

1.2

PV SOLAR POWER SYSTEMS AND KEY EQUIPMENT

From the “Sixth 5-Year” to “Ninth 5-Year” Plans, the National Science and Technology R & D Projects focused on independent PV power systems. In this period, individual user power supplying systems, independent PV power stations, and related controllers and inverters were developed successively. Many solar power application model systems were built, spreading into PV power user homes, rural independent power stations and into the fields of broadcast, communication, oil, meteorology, PV pumping, and school PV education. In the period of the “Tenth 5-Year” to “Eleventh 5-Year” Plans, the focus of PV power R & D shifted to the grid-connected systems and relevant devices. Encouraging results have been achieved in grid-connected inverters, PV/building integrated engineering, LS gridconnected PV power stations, and PV array tracking systems. The following are some typical innovations in PV power systems and associated equipment.

8.4.1 Independent PV Power Stations and Solar/Wind Complementary Stations and Equipment From the “Seventh 5-Year” to “Tenth 5-Year” Plans, developing the technologies for independent PV power systems has received successive support from the country’s science and technology development plans. During this period, independent PV power systems from several dozen to hundreds of kilowatts, and the associated equipment of solar/wind complementary systems, were developed by mastering the technologies of integrated system design and key equipment production. Based on this development, during the “Tenth 5-Year” Plan, which was aimed at residents in off-grid remote western provinces, China popularized independent PV power application in LS. More than 1000 individual PV power stations and wind/solar complementary power stations were built. Hundreds of thousands of PV and wind/

182

CHINESE SOLAR CELL STATUS

solar home-using systems were promoted. The difficulties of living with electricity shortages has been resolved for about four million residents. As a result, the technology’s contribution to raising people’s living standard has been practically realized (Fig. 8.8). 8.4.2 Development of Serial Equipment for Grid-Connected PV Power Systems and Model Engineering During the “Tenth 5-Year” and “Eleventh 5-Year” Plans, China supported the development of equipment and model engineering for grid-connected PV power systems as a key point in its Science and Technology R & D Plan. Through the “Tenth 5-Year” Plan, breakthrough progress was made in developing grid-connected inverters, grid-connected PV/building integrations and LS grid-connected power systems in the desert. Larger power inverters were successively developed for grid-connected inverters from 3 to 100 kVA, mainly used for PV/building

Figure 8.8. Independent PV power station and control, and inverter equipment.

Figure 8.9. Grid-connected inverters (3 and 150 kVA) developed in the “Tenth 5-Year” Plan.

PV SOLAR POWER SYSTEMS AND KEY EQUIPMENT

183

integration, and small–medium inverters to 100–500 kVA, used in large desert power stations. In PV/building integration (BIPV), various integrated styles such as PV roof, PV wall and screen, PV fence, and PV shield were developed and successfully used on Beijing Olympic Studios and sports grounds. Many types of PV array trackers were also developed for large desert PV power systems. These successes have promoted China’s PV applications to advanced international level (Figs. 8.9–8.15).

Figure 8.10. A 500-kVA large power inverter developed in the “Eleventh 5-Year” Plan.

Figure 8.11. PV/building integration engineering completed in the “Tenth 5-Year” and “Eleventh 5-Year” Plans.

Figure 8.12. PV walls of the Olympic Studios and project lights at the Olympic Center.

184

CHINESE SOLAR CELL STATUS

Figure 8.13. Large desert power stations (100 and 500 kW) at Yangbajing, Tibet, and WuWei, Gansu province.

Figure 8.14. Single-axis and dual-axis trackers developed by the Electrical Engineering Institute (EEI), Science Academy of China.

Figure 8.15. A 50-kW horizontal tracker system developed by Kenuo Weiye Inc. of EEI.

THE GROWTH OF CHINA’S PV INDUSTRY

185

8.5

THE GROWTH OF CHINA’S PV INDUSTRY

8.5.1

China’s PV Power Industry Chain and Scale

The whole PV industry chain is composed of several production links. As shown in Figure 8.16, these links include high-purity Si material preparation, Si ingot and wafer production, solar cell fabrication, PV module assembly, and PV power system building. The output capacity of each link in China is listed on Table 8.6. The following are the key features of China’s PV industry:

Metallurgic silicon

Ultra-pure silicon Mono silicon ingots/wafers Multi silicon ingots/wafers Accessory material Mono-crystalline silicon cells Multi-crystalline silicon cells PV modules (Mono silicon, Multi silicon) Subsystems

PV power systems

Figure 8.16. PV industry chain.

TABLE 8.6. China’s PV Industry Status in 2008 Item

Year 2008

Solar cell production output

2000 MW (highest in world)

Capacities of industry chain

More than 40 Si material factories, including under construction (sum of capacity: 100,000 tons/ year); 60 Si ingot/wafer factories, total capacity: 3000 MW; 62 solar cell factories, total capacity 3000 MW; 330 solar module factories, total capacity: 4000 MW

PV industry output value

200 billion yuan (US$25 billion)

Employee

200,000

186

CHINESE SOLAR CELL STATUS

• • • • • •

8.5.2

Solar cell output has experienced rapid growth, reaching 2000 MW in 2008, the highest in world. In 2008, the output of high-purity Si material was 4500 tons in China. Seventy percent of the needs still have to be imported. The shortage problem should be relieved in 2010. Chinese PV enterprises now possess enough strength to enter into capital market. For example, by the end of 2008, 11 Chinese PV power enterprises emerged in overseas stock markets and 13 enterprises entered into domestic stock markets. Reliance on imported production equipment has changed. Now, the equipment of the whole production chain have been localized (including hightech equipment such as hydrogenation furnace for making high-purity Si, PECVD reactor, polysilicon ingot puller, and multiwire saw). China’s domestic PV market still remains stagnant. While in 2008 the output of solar cell production was 2000 MW, of this amount, only 40-MW cells were used domestically. Ninety-eight percent of the cells were exported. At present, more than 200 enterprises are making PV-related products, such as solar garden lamps, solar streetlights, traffic lights, solar energy toys, watches, and calculators. Ninety percent of the products are for exportation.

Current Status of China’s PV Industrial Chain

8.5.2.1 Polysilicon Raw Material To date, Si solar cells are the main products for commercial PV power systems. Worldwide, 98% of solar cells are made of high-purity polysilicon. As an essential material of solar cell production, highpurity polysilicon becomes the most important link in the PV industrial chain. Worldwide, in the production of high-purity polysilicon, a majority of the manufacturers adopted the “improved Siemens method—trichlorosilane reduction process”. This method accounts for 80% of the total output in the world. The total energy consumption of this high-purity polysilicon production is about 125– 170 kWh/kg Si. Before 2006, the polysilicon material market was dominated by 10 companies from developed countries. Table 8.4 shows their polysilicon production quantities from 2006 to 2008. The PV market stimulated rapid growth in the polysilicon industry. In addition to existing suppliers, who are expanding their production capacity with big steps, many companies are investing actively to build new polysilicon production lines. Being promoted by overseas PV markets, China’s polysilicon production and its technology are also growing and developing fast. By introducing, digesting, and absorbing advanced technologies of the improved Siemens method from Western countries, as well as innovations in the production by several enterprises, China’s polysilicon refining technology has caught up with the international advanced level. As an example, Xuzhou Zhongneng Silicon Business has realized an entire

THE GROWTH OF CHINA’S PV INDUSTRY

187

TABLE 8.7. China’s Polysilicon Production Capacity and Output from 2006 to 2008 Manufacturer

2006 (ton)

2007 (ton)

2008 (ton)

Capacity

Output

Capacity

Output

200

100

700

200

700

500

Luoyang China Si

300

200

1000

550

3000

1000

Sichuan Xinguang

1260

0

1260

210

1260

800

40

0

40

20

40

40

300

20

300

200

1500

100

Ermei Semiconductor Factory

Shanghai Lingguang Wuxi Zhongcai Xuzhou Zhongneng

4000

1800

2000

60

Others

8700

100

20,000

4500

3300

0

Output

Zhongqing Daquan Total

1500

Capacity

300

4550

1100

closed-loop and recycle, pollution-free production by using advanced processes including chlorine hydrogenation process, high-flow and fast-deposition regeneration, and highly effective recuperation. Its refining energy consumption index has been reduced to 127 kWh/kg, which is similar to international best level. The quality of the products reached the solar grade for its first-phase polysilicon and the electronic grade for the second-phase polysilicon. The polysilicon refining technologies actually have achieved a breakthrough progress. Table 8.7 shows the polysilicon output between 2006 and 2008 in China. Total output broke through the 1000-ton target in 2007 and reached 4500 tons in 2008. 8.5.2.2 Crystalline Si Ingot/Wafer Manufacturing crystalline Si ingots/wafers is the second link in the PV industry chain. Promoted by the PV market, crystalline Si ingot/wafer manufacture is growing fast. Up to now, there are 70 Si ingot/wafer factories in China. Estimated output of Si ingots/wafers in 2008 was 20,000 tons. The technologies in monocrystalline ingot production and its process have reached advanced world level. For some leading-edge production lines, the energy consumption index is 62 kWh/kg. Recently, the output of crystalline ingots has been growing quickly in China. Table 8.8 gives the production quantities of solargrade crystalline ingots for major production factories. The sum of the solar-grade crystalline ingot output was 2216, 4550, and 8070 tons for years 2005, 2006, and 2007, respectively. In China, up to the end of 2007, there were 2400 Si crystalline pulling machines; the total capacity was about 14,400 tons/year. The crystalline

188

CHINESE SOLAR CELL STATUS

pulling machine is the key equipment in crystalline Si ingot production. The technology for making crystal pullers is very mature in China. It has about 40 years of history in the production of pullers. Recently, the crystal puller production quantity has been increasing dramatically. Sales grew from 8 million yuan in 2004 to 800 million yuan in 2007. The quality of domestic pullers fully meets the requirements of the PV industry. The price is also only one-third to one-half that of imported counterparts. Now, in China, the total number of various crystalline pullers exceeds 2400 and most of them are domestic products. There are more than 10 enterprises manufacturing crystalline pullers in China. In 2006 and 2007, they sold 400 and 800 crystalline pullers, respectively, which not only satisfied the needs of the domestic market but also allowed for exportation in large quantities. For polysilicon ingot production, the technology is more sophisticated than for crystalline Si ingots. Therefore, there are less polysilicon ingot manufacturers in China. The energy consumption in polysilicon ingot production is around 33– 55 kWh/kg in China. Table 8.9 shows China’s major polysilicon ingot manufacturers and their output. The sum of polysilicon ingot output was 300, 1120, and 3700 tons in years 2005, 2006, and 2007, respectively. Up to the end of 2007, there were 230 polysilicon casting furnaces; total capacity was 7000 tons/year. In year 2008, the output was 20,000 tons (Tables 8.8 and 8.9). Promoted by market needs, the localization speed of the polysilicon casting furnaces is increasing, and the situation of importing the casting equipment is changing. Now, several enterprises, such as the forty-eighth Institute of China Electronic Science & Technology Group and Beijing Yuntong Inc., are developing the technologies for manufacturing polysilicon casting furnaces. They have developed 240–270 kg and 400–450 kg qualified serial furnaces and placed them in the market. Wafer sawing is a key process in Si wafer production. The quality of sawing directly affects the following processes in the production chain. As an advanced technology, the multiwire wafer saw is now gradually replacing the traditional inner wheel saw and has become the major technology in the production. Now, many factories in China are using a multiwire saw. The thickness of the cut wafer can be down to 180–200 μm, and the energy consumption is about 220–300 kWh/ kWp. With the most advanced multiwire saw, the wafers can be cut as thin as 180 μm; one 400-kg Si ingot can make 13,000 wafers plus. At present, in China, most multiwire saws are imported. Now, some Chinese companies are developing special wire saws for cutting solar-grade wafers. For example, a wire saw developed by a company in Shanghai is now under trial. Another wire saw made by a company in Guansu has passed product appraisal.

8.5.2.3 Solar Cell Manufacture 8.5.2.3.1 Crystalline Si Solar Cell Manufacture Crystalline solar cells are made by means of forming a diffused junction on the monocrystalline or multicrystalline wafers and some other related processes. The process flow for manufacturing com-

THE GROWTH OF CHINA’S PV INDUSTRY

189

TABLE 8.8. Crystalline Ingot Output in 2006 and 2007 (Ton) No.

Manufacturer

2005

2006

2007

Output

Output

Output

1

Nengjin Jinglong

1126

1250

1500

2

Zhejiang Zhihui

300

750

1100

3

Jinzhou Xinri

400

750

900

4

Yangzhong Huantai

350

900

5

Changzhou Tianhe

300

680

6

Yangchou Shunda

250

500

7

Shanghai Carmdanc

30

450

8

Changzhou Yijing

200

300

9

Jiangyin Hairun

100 80

40

200

100

150

10

Beijing Longfang

11

Zhejiang Jiashan

120

12

Neimongol Hushi

100

13

Zhejiang Kaihua

70

100

14

Shanghai Hejing

30

80

15

Shanghai Songjiang

30

70

16

Xinjiang Xinnengyuan

60

17

Huzhou Xinyuantai

50

18

Shanghai Jiujing

20

50

19

Tianjin Huanou

20

40

20

Others

210

360

710

21

Total

2216

4550

8070

TABLE 8.9. Polysilicon Ingot Output of China in 2006, 2007 (Ton) Manufacturer

2005

LDK Baoding Yingli

260

2006

2007

450

2300

550

1200

Changzhou Tianhe

120

Shaoxing Jinggong

80

80

Ningbo Jingyuan

40

40

40

Total polysilicon

300

1120

3740

190

CHINESE SOLAR CELL STATUS

mercialized crystalline solar cells similarly consists of mechanical damage removal, suede-surface formation, electrode printing and baking, and so on. Under the strong push from the international PV market, China’s solar cell manufacturing grew rapidly in 2006 and 2007. In spite of the restriction from the lack of raw material, there were still new companies successively taking part in the PV industry. Up to the end of 2007, there were 50 enterprises engaging in solar cell manufacturing. Based on the calculations from the numbers of the equipment being used, the capacity of the cell reproduction reached 2900 MWp (including 100 MWp of amorphous thin-film solar cells). As an example, the solar cell outputs of Wuxi Suntech were 157.5 MWp in 2006 and 327 MWp in 2007, and accounted for 35.9% and 30.1% of the total cell outputs of China. In 2006 and 2007, the total solar cell outputs in China were 438 and 1088 MWp, respectively. Among them, 426 and 1059.7 MWp were of crystalline solar cells. In 2008, China’s solar cell output was estimated at 2000 MWp. The LS solar cell production in China has been realized basically through importing production lines from overseas. Because the latest equipment and related technologies can be acquired from the international market, the production technologies for some leading enterprises in China can keep pace with the advanced countries. The average energy consumption from wafer to cell is 220 kWh/kWp, which is similar to the rest of world. Many enterprises and institutes have been attracted by the fast growth of the Chinese PV industry and developed a lot of advanced solar cell production equipment. For example, diffusion furnaces, plasma etchers, baking ovens, and wet cleaning benches have already been localized. Now, about half of the production lines in China have used homemade wet cleaning benches. And homemade semiautomatic screen printers have been used in large quantities. Automatic screen printers and automatic sorters are also under development and expected to be in the market in 1–2 years.

8.5.2.3.2 a-Si Thin-Film Cell Manufacture Recently, because of low cost and good performance under weak light intensity, amorphous thin-film solar cells have received significant attention and a-Si cell manufacture has grown rapidly. Before 2004, in China, a-Si cells were mainly single-junction cells. After Tianjing Jingneng Inc. introduced the 2.5-MWp dual-junction amorphous cell production line in 2004, amorphous dual-junction manufacturing developed quickly. Up to the end of 2007, in China, there were 10 enterprises producing a-Si cells with a total capacity of 80 MWp. In 2006 and 2007, the output of amorphous thin-film cells was 12 and 28.5 MWp, respectively. Since 2004, China’s a-Si cell manufacture has entered a rapid growth period. Three factors have encouraged the fast development of thin-film cell manufacture—prosperous world PV market, advances in technology for thin-film cell manufacturing, and relief from the shortage of polysilicon raw material, which had largely restricted the development of crystalline Si PV manufacture before the world economic crisis. Now, many Chinese enterprises are building and introducing more sophisticated a-Si/μc-Si tandem cell production lines. For example,

THE GROWTH OF CHINA’S PV INDUSTRY

191

Suntech Power (Shanghai) and Hebei Xinao Inc. are importing 50-MWp dualjunction a-Si/μc-Si cell production lines, respectively. Meanwhile, some enterprises are setting up their own R & D centers, manufacturing equipment by themselves, and expanding their production capacities. China’s thin-film cell industry should progress significantly after these new production lines having been completed and are in production. The thin-film cells are still facing restrictions on development from issues such as low efficiency, degeneration of performance, shorter lifetime compared with crystalline cell, and low market acceptability. At the moment, in the industrialization of thin-film cells, the technologies are still improving. The production equipment has not been finalized in design yet. And the initial investment is also very high (Table 8.10).

TABLE 8.10. a-Si Solar Cell Production Capacity and Output in Year 2008 (MWp) China’s a-Si solar cell production capacity and output (ton) No.

Manufacturer

Cell

Capacity

Output

a-Si

25

11

1

Tuori New Energy, Shenzhen

2

Suosaisi New Energy, Shanghai

a-Si

20

6

3

Chuangyi Tech, Shenzhen

a-Si

6

5

4

Shihua Chuangxin Tech, Beijing

a-Si

15

4

5

Jinneng Battery, Tianjin

a-Si

6

Qingfeng E-Light, Shenzhen

7

Minghuan Solar, Shengzhen

8

Golden Solar, Fujian

a-Si

9

Fusheng Solar Energy, Zhejinag

a-Si

1.5

1

10

Hengyang E-Light, Shenzhen

a-Si

1.2

1

11

Gerui Solar Energy, Harbin

a-Si

1

1

12

Hake New Energy, Heilongjiang

a-Si

1.2

13

Jiangsheng E-Light, Nantong

a-Si

25

Start 08

14

Furi PV, Shandong

CIGS

60

Start 09

2.5

2.5

a-Si

1.5

1.5

a-Si

2

2

12

2

0.9

15

Pule, Bangbu

a-Si

12

Start 08

16

Xinao Solar Energy, Hebei

a-Si

60

Start 08

17

SunTech Thin-Film, Shanghai

a-Si

40

Start 09

18

Best Solar (LDK)

a-Si

1000

09?

1174.9

35.9

Total

192

8.5.3

CHINESE SOLAR CELL STATUS

PV Module Assembly

The crystalline solar cells are finally connected to each other and encapsulated in a PV module form by lamination procedures in order to prevent the electrodes and cells from being eroded and cracked. The quality of lamination directly relates to the lifetime of PV modules. At the moment, in China’s PV industrial chain, the PV module manufacture is the link having the most mature production technology, highest equipment localization rate, lowest operation threshold, most enterprises, fastest expanding speed, and highest output. According to incomplete statistics, now in China, there are over 200 PV module manufacturers. The module lamination capacity was 3800 MWp for 2007 and 5000 MWp for 2008. Table 8.11 shows module production output in MW in China in 2006 and 2007. Because Chinese PV power market is still under development, most of the PV modules made in China have been exported, mainly to European and American countries. In 2007, the output of PV modules in China accounted for 28% of the world and ranked first in the world. As with solar cells, the latest equipment and related technologies for modules can be bought from the international market. The PV module assembly technologies in top Chinese manufacturers are now similar to advanced international manufacturers. The energy consumption index is 44 kWh/ TABLE 8.11. Module Production Output in MW in China in 2006 and 2007 No.

Manufacturer

2006

2007

No.

Manufacturer

2006

2007

5

20

1

Suntech, Wuxi

150

364

16

Erquan, Wuxi

2

Yingli, Baoding

55

150

17

Yinjing, Changzhou

—

20

3

Jingao

30

130

18

Zhongsheng, Jiangsu

—

15

4

Jiawei, Shenzhen

45

120

19

Rixing, Wuhan

10

15

5

Lingyang, Jiangsu

45

82

20

BP Jiayang

10

15

6

Adesi, Suzhou

25

82

21

Habo, Beijing

10

12

7

Tianhe, Changzhou

30

80

22

Guofei, Wuxi

8

10

8

Zhongdian, Nanjing

30

80

23

Huayuan, Donguan

5

10

9

Solar E, Ningbo

40

70

24

Junma, Xiameng

5

10

11

Shangpin, Wusi

25

50

25

Dongying, Shandong

5

10

10

SpaceTech, Shanghai

50

50

26

Chao, Shanghai

5

10

12

Tianda, Yunnan

7

30

27

Shunda, Jiangsu

5

10

13

Top Solar

20

30

28

Junxin, Jiangyin

5

8

14

Jingci, Tianjing

10

25

29

Zhongqing, Beijing

15

Sunflower, Zhejiang

5

25

30

Shanshan Youlika

78

150

Others

Total

—

7

3

7

721

1717

PV MODULE AND SYSTEM COST AND PRICE

193

kWp and most equipment has been localized such as lamination machines, laser scribers, and module testing instruments.

8.6

PV MODULE AND SYSTEM COST AND PRICE

The cost of a solar cell module includes the high-purity Si feedstock, the ingot and wafer fabrication, the cell fabrication, the module wiring materials, and labor. The selling price then includes margin and profit added to all of the cost elements. Figure 8.17 shows these cost elements under five cases as a function of the cost of the Si feedstock (Table 8.12). The cost and price for an installed solar system is then determined by more than simply the cost of the module as illustrated in Figure 8.18 (Table 8.13). The cost of solar electricity in dollar per kilowatt hour is discussed in other chapters in this book. The results depend on various assumptions regarding economics, system lifetime, O&M cost, and system residual value, not just technical cost elements. Table 8.14 shows the results under one set of assumptions. Referring to this table, it is evident that lowering system cost will be a challenge, but lower Si cost and sunnier locations using trackers for more annual kilowatt hours will help.

4.5

Cost, Margin, & ASP (S/Wp)

4.0 3.5

$4.00/Wp

1.00

$3.44/Wp 0.79

3.0

$2.97/Wp 0.65

2.5

$2.47/Wp 0.52

2.0

$2.00/Wp 0.40

1.5 1.0

1.75 1.40

0.5

Price paid for p-Si ($/kg)

1.05 0.70

Case I $250

Case II $200

Case III Case IV $150 $100

Figure 8.17. Cost and price elements in making a solar module.

0.35 Case V $50

194

CHINESE SOLAR CELL STATUS

TABLE 8.12. Module Cost and Price as Determined by Si Feedstock Cost Si Material ($/kg)

Si Material ($/Wp)

Module Cost ($/Wp)

Gross Profit ($/Wp)

Module Price ($/Wp)

250

1.75

3.00

1.00

4.00

200

1.40

2.65

0.79

3.44

150

1.05

2.32

0.65

2.97

100

0.70

1.95

0.52

2.47

50

0.35

1.60

0.40

2.00

8 $7.20/Wp 7 1.15

Cost, Margin, & ASP (S/Wp)

6

$6.42/Wp 1.03

$5.73/Wp 0.92

$5.05/Wp

5 0.81

$4.39/Wp 0.70

4 1.00 0.79

3

0.65 0.52

2

1

0.40

1.75

1.40

1.05

0.70

– Case I

Case II

Case III

Case IV

0.35 Case V

Figure 8.18. The cost and price of a complete system can be reduced with lower-cost Si feedstock.

THE GROWTH OF CHINA’S PV POWER MARKET

195

TABLE 8.13. System Price as Influenced by Si Feedstock Price Si Material ($/kg)

Solar Module Price ($/Wp)

System Cost ($/Wp)

Gross Profit ($/Wp)

System Price ($/Wp)

250

4.00

6.05

1.15

7.20

200

3.44

5.39

1.03

6.42

150

2.97

4.81

0.92

5.73

100

2.47

4.24

0.81

5.05

50

2.00

3.69

0.70

4.39

TABLE 8.14. Projected Cost of Solar Electricity versus System Price and Annual Available Sunlight in Kilowatt Hour per Kilowatt Initial Invest. ($/W)

BIPV (1500 h)

BIPV (2000 h)

Electricity Price ($/kWh) $6

0.48

0.37

$4.5

0.36

0.28

$3

0.24

0.19

$2.25

0.18

0.14

$1.50

0.12

0.09

8.7

THE GROWTH OF CHINA’S PV POWER MARKET

Entering the twenty-first century, with the vigorous push and backing of the Chinese government, China’s PV power market achieved rapid growth. A series of state projects has been launched, including the “Tibet no-electricity counties renovation project,” “China bright project,” “Tibet Ngari PV power project,” “township electricity project,” and “the construction project in regions without electricity.” During the Ninth to Eleventh 5-Year Plan period, the Chinese government carried out a number of demonstration projects on city grid-connected PV systems and LS-PV systems in the desert. The Chinese government also missed a few opportunities to seek international assistance, greatly encouraging PV power in rural electrification. China’s renewable energy law took effect in 2006; the Chinese government departments (NDRC, Department of Science and Technology, Ministry of Construction, Ministry of Finance, Ministry of Information Industry, and Ministry of Agriculture) have actively promoted the PV power by launching

196

CHINESE SOLAR CELL STATUS

a number of projects. After successfully bidding to host the Olympic Games, various BIPV projects and 135,000 solar streetlights have already been established in Beijing, with a total power of 10 MW. By the end of 2008, the total capacity of China’s PV power was 140 MW. Table 8.15 shows the growth of China’s PV power market since 1980 (Fig. 8.19). Although China’s solar cell production has ranked highest for two consecutive years, 98% of the cells produced are for exporting, with only 2% for domestic demand. China’s PV industry, therefore, depends on the international market. China needs to expand its domestic demand.

8.7.1

The Potential Market of PV Power in China

8.7.1.1 Potential Market for Electrification in the Rural Area By the end of 2005, there were about 2.7 million households and 12 million residents without electricity in China. Among them, one million households need to be supplied with PV or wind/PV hybrid power systems before 2020. According to the target of the antipoverty program (capacity 200 Wp per household, usage 200 kWh per household per annum), the estimated installed capacity is 200 MWp; according to the outlying district power usage standard (1000 kWh per household), the potential market is 1000 MWp (1 GWp).

TABLE 8.15. The Growth of China’s PV Power Market since 1980 (kW)

Capacity Total

1980

1985

1990

1995

2000

2002

2004

2006

2007

2008

8

70

500

1550

3300

20,300

10,000

10,000

20,000

40,000

16.5

200

1780

6630

19,000

45,000

65,000

80,000

100,000

140,000

160000 140000

Capacity (kW)

Year

120000

Total

100000 80000 60000 40000

Annual

20000 0 1980 1985 1990 1995

2000 2002 2004 2006 2007 2008

Year

Figure 8.19. Annual and total capacity of China’s PV power market since 1980.

THE GROWTH OF CHINA’S PV POWER MARKET

197

8.7.1.2 BIPV Power System More than 70% of solar PV power systems are currently for on-grid power systems, primarily installed on buildings in the city (BIPV). According to long- and medium-term plans, the total installed capacity of China’s BIPV systems will be 50 MWp by 2010, and 1000 MWp by 2020. There are about 4 billion square meters of roof area in China, and about 5 billion square meters of south elevations. If 20% of this area is used to install solar systems, the total capacity would be 100 GWp. 8.7.1.3 LS-PV System in the Desert According to China’s long- and mediumterm plans for renewable energy development, by 2010, a number of 1∼10 MWp desert experimental power stations will be built in Gansu, Tibet, and Inner Mongolia (total installed capacity: 50 MWp). After further promotion during 2010–2020, by the end of 2020, the total installed capacity of LS-PV will reach 200 MWp. Twelve percent of China’s land area, or 1.2 million square kilometers, is nonfarmable desert and mudflat, mainly distributed in the rich solar radiation regions of northwest China. Its annual global solar radiation is over 1600 kWh/m2. If 1% of the desert area is used to install solar systems, the total capacity would be 1000 GWp, twice China’s national power installed capacity in 2006.

8.7.2 National Development and Reform Committee’s “Long- and Medium-Term Plans for Renewable Energy Development” The Chinese government attaches great importance to renewable energy development. On August 31, 2007, the National Development and Reform Committee released the “long- and medium-term plans for renewable energy development” (FGNY [2007] 2174); on March 3, 2008, the committee republished the Renewable Energy Development in the 11th Five-Year Plan (FGNY [2008] 610), which further defined the development objectives for renewable energy (Table 8.16).

TABLE 8.16. China’s Renewable Energy Development Plan Objectives Year

Wind power BGPG Solar power (thermal power)

2004

2010

2020

Capacity (10 MW)

126

1000

3000

Annual generation (TWh)

18.9

210

690

Capacity (10 MW)

200

550

3000

Annual generation (TWh)

51.8

212

835

Capacity (10 MW)

7.0

30 (5)

180 (20)

Annual generation (TWh)

0.84

3.9 (0.65)

23.4 (2.6)

198

CHINESE SOLAR CELL STATUS

TABLE 8.17. 2010 China’s PV Power Marketing Plan Objectives Market Classification

Grand Total Capacity (MWp)

Market Percentage (%)

Rural electrification

80

32

Communication and industry

40

16

PV products

30

12

BIPV

50

20

LS-PV

50

20

250

100

Total

TABLE 8.18. 2020 China’s PV Power Marketing Plan Objectives Market Classification

Grand Total Capacity (MWp)

Market Percentage (%)

Rural electrification

200

12.5

Communication and industry

100

6.25

PV products

100

6.25

BIPV

1000

62.5

LS-PV

200

12.5

Total

1600

100

China’s PV power development objectives for 2010 and 2020 are shown in Tables 8.17 and 8.18.

8.8

CHINA’S INCENTIVE POLICY FOR PV POWER

8.8.1

China’s Renewable Energy Law

On February 28, 2005, China’s renewable energy law was approved by the National People’s Congress and was supposed to be effective on January 1, 2006. Basically, China’s renewable energy law is similar to Germany’s “on-grid price” policy: the initial investment of the power system is borne by the developer, the developer builds the on-grid PV power system with the administrative licensing, recovers the initial cost, and makes a profit by selling the electricity produced by the PV system. Power grid utility companies offer fair prices (cost plus a reasonable profit) to buy the full output of PV electricity. The power companies will not bear the cost beyond the conventional price but will collect the electrovalence surcharge from the users of the national electricity grid systems.

CHINA’S INCENTIVE POLICY FOR PV POWER

8.8.2

199

Other Supporting Policies to Renewable Energy Law

In 2007, the State Electricity Regulatory Commission issued its twenty-fifth decree of the year, “supervisory method of power grid companies purchases the full quantity of renewable energy power” (effective on September 1, 2007), to emphasize that according to the renewable energy law, power grid companies must purchase the renewable energy power as a priority, and offer the grid integration services (the cost of grid integration services will be included in the electrovalence surcharge, apportioned by the users of the national electricity grid systems). On August 2, 2007, the General Office of the State Council transmitted the “energy-saving power dispatch regulation (draft)” (GBF [2007] 53), jointly issued by the National Development and Reform Committee, State Environmental Protection Administration, State Electricity Regulatory Commission, and State Energy Bureau, to emphasize that power grid enterprises should dispatch renewable energy power as a priority, as long as this could be done without prejudice to the adequate supply of electricity. On January 11, 2007, the National Development and Reform Committee issued the “interim provisions of deployment of renewable energy power surcharge” (FGJG [2007] 44). Renewable energy law defines that, when power grid enterprises purchase the renewable energy power, the cost difference between renewable energy power and conventional energy power will be allocated in the sales rate of electricity. The “interim provisions” clearly define the collection standard, collection scope and allocation, and balance decision of the surcharge of renewable energy power, as well as the procedure of applying for renewable energy power electrovalence subsidy, electricity price settlement, and surcharge quota allocation. To off-grid-independent power systems, the subsidies for O&M are also stipulated in “interim allocation plan for the price and cost of the renewable energy” (FGJG [2006] 7) and “interim provision of deployment of renewable energy power surcharge” (FGJG [2007] 44): the overportion of the operation cost of the independent renewable power systems, which are invested or supported by the country, to the average sale price of the electricity of the local utility will be compensated by the surcharge in the renewable energy electricity price. The application procedures and the method of compensation calculation are the same for the grid-connected power systems. From August 2006 to June 2008, a renewable energy surcharge of 0.1¢ per kilowatt hour was collected countrywide. In June 2008, this increased to 0.2¢ per kilowatt hour. This surcharge is used to subsidize renewable electricity (including wind electricity, bioelectricity, and PV electricity). The total amount of the collection is about $500 million. If $250 million of the surcharge funds are used to subsidize the grid-connected PV power and the annual generation of the PV power is 1300 kWh/kW, then 480-MW power will be supported for 4 yuan per kilowatt hour subsidy, or 960-MW power for 2 yuan per kilowatt hour subsidy, and or 1.92-GW power for 1 yuan subsidy. Unfortunately, up to now, 3 years after the renewable energy law took effect, because of the indecisiveness of the electricity price administration, PV power has not enjoyed its reasonable on-grid price.

200

CHINESE SOLAR CELL STATUS

ABBREVIATIONS a-Si—amorphous silicon BGPG—biomass power BIPV—building-integrated photovoltaics CdTe—cadmium telluride CIGS—copper–indium–gallium–selenium Co.—company CuInGaSe—copper indium gallium diselenide EPFL—Ecole Polytechnique Federale de Lausanne, Switzerland FGJG—Price Bureau of National Development and Reform Commission FGNY—Energy Bureau of National Development and Reform Commission GaAs—gallium arsenide GaP—gallium phosphide GBF—issued by the General Office of State Council GWp—gigawatt (billion watts) peak HIT—heterojunction with intrinsic thin layer HV—high voltage kWp—kilowatt peak LS—large scale max—maximum min—minimum MWp—megawatt peak NDRC—National Development and Reform Commission, China NREL—National Renewable Energy Laboratory, USA O&M—operation and maintenance PECVD—plasma-enhanced chemical vapor deposition p-Si—polycrystalline silicon PV—photovoltaic R & D—research and development Si—silicon u-Si—ultrapure silicon USSC—United Solar System Corporation VHF—very high frequency Wp—watt peak X—sunlight concentration ratio μc-Si—microcrystalline silicon

APPENDIX: CHINA SOLAR RESOURCE DATA Here, we defined the sum of irradiation on a tilted PV array as “annual usable hours.” The following tables show the average annual usable hours for each province of China.

APPENDIX: CHINA SOLAR RESOURCE DATA

201

TABLE 8A.1. Statistics of Solar Irradiation Resources of China’s 17 Southeastern Coastal Provinces and Cities Province

Annual Annual Irradiation Irradiation Max Min (MJ/m2) (MJ/m2)

Sum of Annual Horizontal Irradiation of the Capital (MJ/m2)

Horizontal Annual Usable Hours of the Capital (h)

PV Array Slope Angle (Degree)

Slope Annual Usable Hours of the Capital (h)

Heilongjiang

4683.69

4442.92

4683.69

1301.03

50

1496.179

Hebei

5008.89

5008.89

5008.89

1391.36

42

1558.323

Guangxi

4595.91

4294.11

4595.91

1276.64

25

1378.773

Jiling

5034.39

4640.64

5034.39

1398.44

45

1608.209

Guangdong

5161.46

4478.03

4478.03

1243.90

25

1343.41

Hubei

4312.92

4047.91

4312.92

1198.03

35

1377.74

Shendong

5123.01

4761.44

5123.01

1423.06

40

1636.516

Henan

5095

4764.36

4764.36

1323.43

40

1521.95

Liaoning

5068.67

4903.14

5067.41

1407.61

45

1618.757

Jiangxi

5045.26

4630.6

4832.08

1342.24

30

1449.624

Jiangsu

4855.49

4855.49

4855.49

1348.75

35

1483.621

Fujian

4410.74

4410.74

4410.74

1225.21

30

1323.221

Zhejiang

4751.82

4314.6

4314.6

1198.50

35

1318.349

Hinan

5125.1

5125.1

5125.1

1423.64

25

1494.82

Beijing

5620.01

5620.01

5620.01

1561.11

42

1748.448

Tianjin

5260.11

5260.11

5260.11

1461.14

42

1636.479

Shanghai

4729.25

4729.25

4729.25

1313.68

35

Average

4985.16

4872.12

4912.75

1364.65

1445.049 1502.04

202 TABLE 8A.2. Statistics of Solar Energy Resources of China’s Nine Western Provinces Province

Annual Irradiation Max (MJ/m2)

Annual Irradiation Min (MJ/m2)

Sum of Annual Horizontal Irradiation of the Capital (MJ/m2)

Horizontal Annual Usable Hours of the Capital (h)

PV Array Slope Angle (Degree)

Sum of Annual Irradiation on a Slope PV Array in the Capital (MJ/m2)

Slope annual Usable Hours of the Capital (h)

Xinjiang

6342.31

5304.84

5304.84

1473.57

45

6100.56

1694.601

Xizang

7910.65

6088.59

7885.99

2190.55

30

8832.31

2453.418

Inner Mongol

6195.18

5658.47

6041.35

1678.15

45

6947.56

1929.877

Qinghai

6951.76

6142.93

6142.93

1706.37

40

7064.37

1962.324

Gansu

6458.52

5442.78

5442.78

1511.88

40

6259.19

1738.665

Ningxia

5944.8

5944.8

5944.8

1651.33

42

6658.17

1849.492

Shanxi

5868.3

5513.84

5513.84

1531.62

40

6340.91

1761.365

Shaanxi

4730.51

4730.51

4730.51

1314.03

40

5440.08

1511.134

Yunnan

6156.72

4848.38

5182.78

1439.66

28

5597.4

1554.835

Average

6284.31

5519.46

5798.87

1610.80

38.89

6582.28

1828.41

TABLE 8A.3. Statistics of Solar Resources of China’s Lowest Irradiation in Four Provinces Annual Irradiation Max (MJ/m2)

Annual Irradiation Min (MJ/m2)

Sum of Annual Horizontal Irradiation of the Capital (MJ/m2)

Horizontal Annual Usable Hours of the Capital (h)

PV Array Slope Angle (Degree)

Slope Annual Usable Hours of the Capital (h)

Hunan

4212.60

4212.60

4212.60

1170.17

30

1287.19

Anhui

3792.51

3792.51

3792.51

1053.48

35

1211.50

Sichuan

4229.74

3486.96

3792.51

1053.48

35

1179.89

Guizhou

4672.40

3471.07

3797.95

1054.99

30

1139.38

Average

4226.81

3740.79

3898.89

1083.03

Province

1204.49

TABLE 8A.4. Comprehensive Efficiency for Different Power Systems PV Power System Type

Comprehensive Efficiency (%)

Individual PV power station

70

Grid-linked system on building

80

Grid-linked system on open field

85

TABLE 8A.5. Effective Annual Usable Hours for Different PV Power Systems in Each Province Region

Northwestern

Annual Usable Hours (h)

Effective Hours per Year of Individual PV System (h/year)

Effective Hours per Year on Building a Grid-Linked System (h/year)

Effective Hours per Year on Open Field GridLinked System (h/year)

Xinjiang

1694.60

1186.22

1355.68

1440.41

Xizang

2453.40

1717.38

1962.72

2085.39

Inner Mongol

1929.90

1350.93

1543.92

1640.42

Qinghai

1962.30

1373.61

1569.84

1667.96

Gansu

1738.70

1217.09

1390.96

1477.90

Ningxia

1849.50

1294.65

1479.60

1572.08

Shenxi

1761.40

1232.98

1409.12

1497.19

Shaanxi

1511.10

1057.77

1208.88

1284.44

Yunnan

1554.80

1088.36

1243.84

1321.58

Province

204

CHINESE SOLAR CELL STATUS

TABLE 8A.5. Effective Annual Usable Hours for Different PV Power Systems in Each Province (continued) Region

Eastern coastal

Poor resource region

Average

Province

Annual Usable Hours (h)

Effective Hours per Year of Individual PV System (h/year)

Effective Hours per Year on Building a Grid-Linked System (h/year)

Effective Hours per Year on Open Field GridLinked System (h/year)

Heilongjiang

1496.20

1047.34

1196.96

1271.77

Jilin

1608.20

1125.74

1286.56

1366.97

Liaoning

1618.80

1133.16

1295.04

1375.98

Hebei

1558.30

1090.81

1246.64

1324.56

Henan

1521.90

1065.33

1217.52

1293.62

Shendong

1636.50

1145.55

1309.20

1391.03

Guangdong

1343.40

940.38

1074.72

1141.89

Guangxi

1378.80

965.16

1103.04

1171.98

Hubei

1377.70

964.39

1102.16

1171.05

Jiangxi

1449.60

1014.72

1159.68

1232.16

Jiangsu

1483.60

1038.52

1186.88

1261.06

Fujian

1323.20

926.24

1058.56

1124.72

Zhejiang

1318.30

922.81

1054.64

1120.56

Hainan

1494.80

1046.36

1195.84

1270.58

Beijing

1748.40

1223.88

1398.72

1486.14

Tianjing

1459.70

1021.79

1167.76

1240.75

Shanghai

1445.00

1011.50

1156.00

1228.25

Hunan

1287.20

901.04

1029.76

1094.12

Anhui

1211.50

848.05

969.20

1029.78

Sichuan

1179.90

825.93

943.92

1002.92

Guizhou

1139.40

797.58

911.52

968.49

1551.2

1085.8

1241.0

1318.5

Note: Annual effective usable hours = annual usable hours × comprehensive efficiency of the system.

The average effective usage hours for different PV power system types: the following tables show the statistical results for China’s 26 provinces, regions, and cities, except for the four poor resources provinces.

APPENDIX: CHINA SOLAR RESOURCE DATA

205

TABLE 8A.6. The Annual Effective Usage Hours for Different System Types in China Annual Irradiation on Horizon (kWh/m2)

Annual Irradiation on Slope (kWh/m2)

Effective Usage Hours of Individual Power Systems

Effective Usage Hours of Grid-Linked on Building Systems

Effective Usage Hours of Grid-Linked on Field Systems

Northwestern

1610.80

1828.41

1250

1450

1540

Coastal

1364.65

1502.04

1000

1200

1250

Countrywide average

1487.73

1665.23

1100

1250

1350

Region

The following are the average effective usage hours for different regions: 1.

Northwestern region

TABLE 8A.7. The Average Effective Annual Usages Hours of Different Types of Systems in the Northwestern Region of China Classification

Resources Hours

Effective Usage Hours of Individual Power Systems

Effective Usage Hours of Grid Linked on Building Systems

Effective Usage Hours of Grid Linked on Field Systems

Year max

2453.4

1717.4

1962.7

2085.4

Year min

1511.1

1057.8

1208.9

1284.4

Year average

1828.4

1280.0

1462.7

1544.1

Day max

6.72

4.6

5.0

5.7

Day min

4.14

2.9

3.3

3.5

Day average

5.00

3.5

4.0

4.2

Based on the above statistics, the average annual effective usage hours in the northwestern region for different system types are the following: individual PV power system: 1200 h; on building grid-linked PV power system: 1400 h; and on field grid-linked PV power system: 1500 h. Editor’s note: sun tracking will increase these annual effective usage hours by about 30%.

206

2.

CHINESE SOLAR CELL STATUS

Eastern region

TABLE 8A.8. The Average Effective Annual Usage Hours of Different Types of Systems in the Eastern Region of China Classification

Resources Hours

Effective Usage Hours of Individual Power Systems

Effective Usage Hours of Grid Linked on Building Systems

Effective Usage Hours of Grid Linked on Field Systems

Year max

1748.4

1223.9

1398.7

1486.1

Year min

1318.3

922.8

1054.6

1120.6

Year average

1486.0

1040.2

1188.8

1263.1

Dairy max

4.79

3.4

3.8

4.1

Day min

3.61

2.5

2.7

3.1

Day average

4.07

2.8

3.3

3.5

Based on the above statistics, the average annual effective usage hours in the eastern region for different system types are the following: individual PV power system: 1000 h; on building grid-linked PV power system: 1200 h; and on field grid-linked PV power system: 1300 h. Editor’s note: sun tracking will increase these annual effective usage hours by about 30%.

9 TRACKING THE SUN FOR MORE KILOWATT HOUR AND LOWERCOST SOLAR ELECTRICITY RON CORIO,1 MICHAEL REED,1 AND LEWIS FRAAS2 1 Array Technologies, Inc., 2JX Crystals Inc.

9.1

INTRODUCTION

In the early history of terrestrial solar cells, the remote power applications required smaller expensive solar modules that were generally mounted in a fixed position. The up-front cost of the solar module in U.S. dollar per watt was the dominant economic metric, and the early government solar subsidies in California were based on this U.S. dollar per watt metric. Tracking the sun for these smaller systems simply added cost, and there was no financial incentive to do so. However, more recently, solar cell systems have gotten larger, and the economic metric has shifted to cents per kilowatt hour. By pointing the solar module at the sun all day instead of just at noon, the number of kilowatt hour produced by a given module typically increases by 25–40%. Today, the incremental cost penalty at the system level associated with tracking large arrays of 1 kW and more is now much less than this 25–40% benefit, making tracking now the preferred option for many applications. For ground-mounted solar PV systems generating electricity for peak power, solar tracked systems now account for over 50% of the new systems being installed. Solar tracking systems on commercial building flat rooftops are also now economically attractive. Fortunately, starting with the solar FIT in Germany, the government subsidies now encourage systems that maximize the kilowatt hour produced, thereby minimizing the cost of solar electricity in cents per kilowatt hour. Solar FITs are now available in Spain as well as in several states in the United States. Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

207

208

9.2

TRACKING THE SUN

THE MODERN SOLAR TRACKER ERA

It can be argued that the modern solar tracker era began as a response to the “oil and energy crisis” of the early 1970s. In response, ARCO began manufacturing PV modules in the mid-1970s. By 1983, ARCO had built two grid-connected PV power plants on the Carrizo Plains in California. The first was a 1-MW plant and the second was a 5.2-MW plant. Neither plant was economically successful. Both plants were sold to the Carrizo Solar Corporation in 1990. Both ARCO plants used very large, computer-controlled D/A trackers that had been developed for pointing heliostats for solar thermal applications. The 5.2MW plant occupied 177 ac of land. The cost of installation was high due to the module production costs and the complex D/A trackers. The electricity produced was sold at 3–4¢ per kilowatt hour. The power would have had to be sold at five times that rate for the plant to be profitable. By the end of the 1990s, Carrizo Solar had shut the plants down and sold off the assets piece by piece. This first solar PV tracker experiment was then followed with the emergence of small solar tracker companies in the late 1980s and early 1990s. Robbins Engineering, Inc. and Zomeworks Corporation manufactured and sold passive, gas-based trackers, while Array Technologies, Inc. manufactured motorized active trackers that were controlled by a closed-loop, optical-based system. Both Zomeworks and Array Technologies exhibited slow but steady growth and primarily sold their products to an emerging remote home, “off-grid,” solar market. The Robbins Engineering trackers never gained the market place share that their counterparts did. In the United States, the “on-grid” or “grid-tie” market evolved in California during the 1980s and was driven by incentive-based programs offered by the utility districts. The SMUD was the most widely known early adopter. SMUD used an innovative S/A tracker design that proved to be much more affordable for utility applications than the early D/A tracker design used by ARCO. By 1998, over 1.9 MW of grid-tied PV had been installed. By 2008, programs characterized by the California Energy Commission and the California Public Utilities Commission had promoted 279.54 MW of installed grid-tied PV. Most of the California incentives were not “performance based.” They maximized the power in watts but not the energy produced in kilowatt hour. Maximizing the delivered AC energy fed back into the grid was not the primary goal. From a global perspective during the late 1990s, the concept of national or regional energy independence fostered FITs. A FIT is a guaranteed financial ROI based on the amount of energy in kilowatt hour delivered to the grid. Because of large FITs, utility scale-sized (over 1 MW) PV plants began to flourish in Germany and South Korea. In recent years, Spain, Italy, Greece, and Australia have enacted FIT-based renewable energy programs. FITs promote growth and a stable ROI, and encourage a reduction in installed PV system energy costs. Solar tracking PV marries well to the goal of getting the most power from utility-scale power plants. In an online June 1, 2009 Renewable Energy World article entitled “Solar Trackers: Facing the Sun,” it was suggested by an industry

TRACKER GEOMETRIES

209

analyst that “between 2009 and 2012, tracking systems will be used in at least 85% of all PV installations above 1 MW.” Certainly, there is now a serious interest in solar tracking as evidenced by the participation of over 30 solar tracking companies in San Francisco during the Intersolar North America 2009 exhibition.

9.3

TRACKER GEOMETRIES

The older traditional PV mounting structures for utility-scale PV generation are typically fixed at an optimum elevation tilt angle for power production without seasonal adjustment. A typical FR system is depicted in Figure 9.1. In most utility-scale planar PV plants today, the real question is not whether to track the sun but how to track the sun. Multiple solar tracking geometries are available to track the sun through its diurnal path. These can be first segregated into two basic categories: D/A and S/A. D/A trackers follow the sun in two axes to produce the greatest power gain from the PV module by keeping the sun’s rays perpendicular to the module surface. Many point-focus CPV technologies require D/A tracking to operate. D/A trackers, however, tend to be complex and costly, and need more area to deploy, thereby reducing their GCR. S/A trackers, depending on design, are simpler, require less area to deploy, and are generally less costly to purchase and O&M than D/A trackers. The geometric variations for the D/A trackers are shown in Figure 9.2. It turns out that tracking systems are actually more economical in most of the United States. For example, Table 9.1 presents a comparison for the economics for two systems producing similar amounts of energy in kilowatt hour per year in Madison, Wisconsin. Because the tracking system produces more annual kilowatt hour per installed kilowatt, fewer kilowatt and smaller numbers of PV modules are

W Tilt N

S

Azimuth E

Figure 9.1. Fixed-tilt rack system.

210

TRACKING THE SUN Axis of rotation

Axis of rotation

W

W

N

N

Tilt Azimuth S

Axis of rotation

E

S

E

Figure 9.2. D/A tracker geometries. (a) Azimuth and elevation. (b) Tilt and roll (polar and elevation).

TABLE 9.1. PV Watts Analysis Fixed PV versus D/A Tracking PV for Madison, Wisconsin Fixed Tilt

D/A Tracking

4000 DC W

3200 DC W

Annual AC (kWh)

5089 kWh

5175 kWh

System cost ($/W)

$8.63/DC W

$9.96/DC W

$34,520

$31,872

System size

Total cost Net savings with tracker

$2648

required, producing net savings at the system level. As shown in Table 9.1, this is true even when the U.S. dollar per watt goes up in order to pay for the tracker. Specifically, a 3200-DC-W tracked system produces more kilowatt per year compared with a 4000-DC-W fixed-tilt system, and it costs $2648 less. A D/A tracker moves with the sun in two directions (Fig. 9.3). It tracks in azimuth, the E to W movement of the sun, and also tracks in elevation—the vertical rise and fall of the sun. Figure 9.4 shows the rear view of a Wattsun™ (Array Technologies, Inc., Albuquerque, NM) AZ-225 D/A tracker. A linear actuator can be seen for the elevation drive, and a motor gear assembly sits at the top of the post for the azimuth drive. The function of a D/A tracker is to keep the plane of solar array perpendicular to the sun in both azimuth and elevation. This is necessary for point-focus CPV arrays. However, planar PV arrays do not require precise pointing. Today, D/A trackers are commonly used for CPV systems as discussed in several subsequent chapters in this book.

TRACKER GEOMETRIES

211 Sunny, total solar raditaion 5.14 kWh/m2 3000

Tracked Fixed

Watts Output

2500 2000 1500 1000 500 0 4:48 7:12 9:36 12:00 14:24 16:48 19:12 Time, PST (solar time, not local)

Figure 9.3. (a) Photograph of an Array Technologies D/A post-mounted PV tracking array. (b) Comparison of the power output on a sunny day for fixed and tracked PV arrays.

Figure 9.4. Rear view of a Wattsun AZ-225 D/A solar tracker. TABLE 9.2. Increase in Annual Kilowatt Hour for Various Tracker Geometries Relative to Fixed Tilt Tracking Geometry

Annual Increase in Solar Performance

Dual axis

29–40%

Single-axis azimuth tracker with array tilt of 45°

23–34%

Single-axis linear tracker with tilt of 20°

22–32%

Single-axis linear tracker with tilt of 0°

17–25%

It is now clear that for larger PV arrays, trackers produce more kilowatt hour relative to fixed-tilt PV arrays. However, how can the cost of tracking be reduced? S/A trackers are simpler because one of the drive elements is eliminated, and it turns out that they still generate a significant increase in the annual kilowatt hour produced as shown in Table 9.2.

212

TRACKING THE SUN

9.4 GROUND-MOUNTED TRACKERS FOR UTILITY PEAK POWER A simple S/A tracker moves the plane of a PV array around one central axis of rotation. The rotational axis can be vertical (points to the zenith), or the axis can be aligned with the N/S axis of the Earth. An azimuth tracker spins the array about a vertical axis and tracks the daily passage of the sun. Figure 9.5 showed several fixed-tilt azimuth trackers. Alternately, a linear axis tracker can “roll” an array of PV modules E to W. Figure 9.6 shows the concept for the simplest horizontal beam linear axis tracker. In the case shown in Figure 9.6, there are seven beams level with the ground aligned along the N to S direction. PV modules are mounted astride these beams.

Figure 9.5. Array Technologies, Inc. fixed-tilt azimuth tracker installation—2 MW, Spain.

Figure 9.6. Building block diagram of an Array Technologies, Inc. S/A HZ Tracking System. It is a utility-scale mechanically linked system.

AZIMUTH TRACKERS FOR COMMERCIAL BUILDING FLAT ROOFS

213

Figure 9.7. Array Technologies, Inc. S/A HZ trackers. Right 6.6 MW, Alamosa, Colorado.

There is a second drive beam running E to W with a single drive motor driving this beam such that the seven beams can rotate the module array from E to W over the course of the day. Larger versions of this S/A HZ tracking system are possible such that one motor can drive 300 kW of PV modules at one time. Figure 9.7 shows photographs of two large PV fields using these tracking systems.

9.5 AZIMUTH TRACKERS FOR COMMERCIAL BUILDING FLAT ROOFS As just shown, it is now clear that solar trackers are preferred for large solar PV utility fields producing peak power. However, of the potential markets for PV, commercial customers also represent an excellent opportunity. Commercial customers can buy and use larger PV systems; they presently pay retail prices for electricity; and they need electricity during the day when the sun is shining. There is an excellent opportunity to reduce the cost of solar electricity for commercial systems when trackers are mounting on commercial building flat rooftops. However, for commercial building flat rooftops, the design constraints are different. Solar trackers are needed that are sized to fit on the roof with no roof penetration and with low profiles for low wind resistance. The JXC fixed-tilt/ azimuth carousel is a low profile tracker designed to distribute weight evenly over a large area and for low wind resistance on the roof. As shown in Figure 9.8, it is designed as a compact prefabricated unit where the module support arms fold down for easy shipping. When these arms are folded down, the carousel is only 8 in. tall. Without mounted modules, it is slightly less than 8 ft wide and slightly less than 10 ft long so that multiple carousels can be stacked and easily shipped on a flatbed truck.

214

TRACKING THE SUN

Figure 9.8. The drawing above is a top-down view of a JXC Carousel Tracker without the PV modules. The middle drawing is an edge view with the module support arms folded down, and the bottom drawing shows the carousel after the module support arms are raised and the PV modules are mounted.

A shipping and installation sequence is shown in Figure 9.9. The carousels can be installed without roof penetration by using either ballast or tether mounting. With this size determination, these carousel trackers are designed to carry four or six PV modules in two rows with two or three modules per row in front and back rows. Various module sizes can be accommodated. With panels in place at a fixed 30° tilt, and when mounted on a roof, this azimuth tracker then follows the sun from E in the morning to S at noon to W in the late afternoon. Figure 9.10 shows a 1.2-kW carousel fitted with six 200-W modules on a church rooftop in San Diego, CA.

SOLAR TRACKER ECONOMICS

215

Figure 9.9. Carousel shipping and installation sequence. Top/left: carousel stack ships prefabricated on a flatbed truck. Top/right: hoisted to the roof. Bottom/left: deployed on roof. Bottom/right: modules mounted and operational.

Figure 9.10. JXC 1.2-kW S/A Azimuth Sun Tracker in operation in San Diego, California.

9.6

SOLAR TRACKER ECONOMICS

The economic viability of solar trackers is still often debated. However, there is now enough data to prove their economic benefit for the utility field case as shown in Table 9.3. The numbers listed in this table are specifically for the baseline c-Si module case. Tracking for commercial building rooftops is newer. Several questions are often asked. In the following, some of these questions are listed and answered specifically for the 1.2-kW carousel tracker.

216

TRACKING THE SUN

TABLE 9.3. Cost/Benefit Analysis for Solar Tracking for Utility Peak Power Fields Metric

Case 1: Historical

Case 2: Current

Case 3: Future

PV module cost

$4 per watt

$2 per watt

$1.50 per watt

PV system cost

$6.50 per watt

$4.50 per watt

$3.50 per watt

12%

16%

20%

FR cost

$0.45 per watt

$0.40 per watt

$0.25 per watt

D/A tracker cost

$1.25 per watt

$1.00 per watt

$0.75 per watt

D/A––FR

$0.80 per watt

$0.60 per watt

$0.50 per watt

(12.3%)

(13.3%)

(14.3%)

38%

38%

38%

S/A HZ tracker cost

$0.75 per watt

$0.55 per watt

$0.35 per watt

S/A––FR

$0.30 per watt

$0.15 per watt

$0.10 per watt

(4.6%)

(3.3%)

(2.9%)

24%

24%

24%

Efficiency

D/A penalty D/A gain

S/A HZ penalty S/A HZ gain

Question #1: How much energy is consumed for tracking and how does that compare with the energy produced? Answer #1: A solar tracker rotates one-half revolution during the day. It does this in very small steps. It returns to the starting point overnight. Specifically, for the rooftop carousel, the energy required per day is 6 Wh, and the energy produced per day is 6 kWh. So the energy required is 0.1% relative to the energy produced. Question #2: What is the cost per watt for a solar rooftop tracker? Answer #2: The answer to this question is a function of production volume and manufacturing setup. Production volume today for the carousels is small, and the manufacturing is done in a small prototype shop. However, a cost estimate in high-volume production can be made given its weight. The weight of the carousel without modules is about 200 lb. By analogy, a Kenmore washer weighs about 250 lb. A washing machine also has a drive motor and a controller just like a solar tracker, and it can be purchased and installed for about $400. This installed price for a 1.2-kW rooftop tracker translates to $0.33 per watt. Question #3: How much does the drive system cost? Answer #3: Again by analogy, one can purchase the drive with control for a satellite dish for about $85. For a solar tracker producing 1.2 kW, this cost will translate to about $0.09 per watt.

ABBREVIATIONS

217

Question #4: What about reliability and operation and maintenance cost? Answer #4: Nearly every home today has appliances with motors. Examples are washing machines and refrigerators. These systems are very reliable, and there is an infrastructure for O&M. Furthermore, solar trackers now have over a decade of operating experience. Note that solar trackers only turn one revolution per day. This equates to 7300 revolutions in 20 years. By analogy, with a simple wristwatch, this equates to the number of revolutions the second hand on a watch will make in 121.66 h or 5 days.

9.7

CONCLUSIONS

For larger PV plants today, the question is no longer whether or not to track the sun. The relevant question now is how to track the sun. Several tracker geometries and tracker types have been developed and described here. It is now time to scale up production in order to reduce costs. ABBREVIATIONS AC—alternating current ARCO—Atlantic Richfield Oil Company CPV—concentrating photovoltaic c-Si—crystalline silicon D/A—dual-axis DC—direct current E—east FIT—feed-in tariff FR—fixed rack GCR—ground coverage ratio HZ—horizontal N/S rotation axis JXC—JX Crystals N—north N/S—north/south O&M—operate and maintain PV—photovoltaic ROI—return on investment S—south S/A—single-axis SMUD—Sacramento Municipal Utility District W—west

10 SOLAR CELL SYSTEMS: DEFINITION, PERFORMANCE, AND RELIABILITY JASON STRAUCH, LARRY MOORE, Sandia National Laboratories

10.1

AND

ELMER COLLINS

INTRODUCTION

The previous chapters have presented the development of solar cell technology and physics and the basics of terrestrial solar cells and modules. The preceding chapter described sun trackers. Trackers are an additional hardware component used to increase the energy output from solar cell systems. This chapter will begin with a description of the complete set of components that make up a solar cell system. The basic components are the same for a smaller home-based system, a large-scale terrestrial field system, and even for space-based systems that generate AC power. After the section on system definition, a detailed description of the performance of an existing 3.51-MWDC solar cell system in a field in Springerville, Arizona, is presented. The Springerville Generating Station Solar System in eastern Arizona was built and is operated by TEP. The field experience with this system has provided a treasure of information that not only establishes a baseline for today’s state-of-the-art system capabilities but also has helped guide the development of a PV system technology for the future. These are the reasons that TEP and Sandia National Laboratories entered into a collaborative effort to track, analyze, and document the cost and field performance as well as the O&M experience associated with this system. The third section of this chapter will then present a reliability study for solar cell fields based on the data available from the Springerville system. The fourth section of this chapter will then summarize the economics for this Springerville solar cell system. This chapter ends with a summary and conclusions. Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

219

220

SOLAR CELL SYSTEMS

This chapter is a condensation from papers presented by several authors from Sandia National Laboratories. The performance for the Springerville System was first described by Moore and Post [1], and the reliability study based on this system was presented by Collins et al. [2]. 10.2

SOLAR CELL SYSTEM DEFINITION

10.2.1

Basic Components

A number of components make up a PV power generating system and will be described from a system level down to the individual building blocks. At the highest level, a PV system produces electric power and, thus, is connected to an electrical load. This load can be DC batteries or equipment, or AC equipment including the electrical distribution grid. The primary focus in subsequent sections will be on grid-tied systems at larger scales, such as utility-level PV power generating fields. Many of the components and operating principles also apply to smallerscale residential systems that are grid tied, and when there is a substantial difference, it will be noted. A solar cell or PV system generates electrical power that is delivered to the load by wiring, typically transmission lines. In order to be grid connected and used by AC systems, the DC power must be converted to AC by an inverter. The inverter is a critical component to a PV system’s efficiency and reliability, as will be seen in subsequent sections. Figure 10.1 shows a schematic of a basic PV power system that is attached to the grid as depicted by the AC loads. As described previously, the most basic building block of a PV system is the P-N junction in the PV semiconductor material and cell. There are many different types of solar cells, and the focus in this section is on one of the most efficient and common, the c-Si cell. In a module, rows of cells are wired in series in what are known as strings. The strings are subsequently wired together and mounted to produce a module. These are shown in Figure 10.2. Modules are arranged into arrays, and, at larger scales, each array will typically deliver the power to an inverter. This is shown in Figure 10.3. The inverter is one of the most expensive components in the system, after the modules themselves, and is also responsible for up to 37% of system failures, as described in a subsequent section. These arrays and inverters can be grouped into larger systems that produce megawatt levels of power for the electric grid. Each array and inverter system is connected to the grid or other AC system with AC disconnects. These disconnects allow that part of the system to be isolated from the grid (or load) for servicing and repairs. There are additional components within the module that are needed to connect the modules together in an array and are used to increase the performance of the module and array. 10.2.2

Module Detail

In terms of mechanical construction, the module is physically made of a number of parts: the frame, the cell strings including cell interconnects, the junction box,

SOLAR CELL SYSTEM DEFINITION

221

Solar PV Electricity - PhotoVoltaic System

Buy from the Grid

Back-Up Battery Storage

PhotoVoltaic (PV) Flat Panel Collector

AC/DC Inverter

Import Meter Fusebox

Network / Grid

Export Meter

Lighting & Electricity - Sockets & Switches

Sell Back to the Grid

Block Diagram of PV System

Solar Irradiance

cell

Module

module array DC solar irradiance

AC AC loads

Battery

inverter

AC

+

Charge Controller

Inverter

–

engine-generator back-up (optional) battery storage (optional)

AC Loads DC Loads

Figure 10.1. Basic PV power system examples.

Figure 10.2. A single polycrystalline PV cell showing the diode grid lines, and a module cartoon showing the cell strings.

222

SOLAR CELL SYSTEMS 5

+

+

5

+

+

+

+

3 +

–

–

+

4

2

1 –

+

Inverter

Figure 10.3. PV module array tied to inverter, and module back side showing junction box and interconnects.

z o

y T2

x

Glass EVA

T1 T3

ARC Si EVA Tedlar

Figure 10.4. Module cell lamination components and Saint Gobain PV glass types.

and the bypass diodes, as shown in Figure 10.3. The frame is generally aluminum and serves the function of supporting the cell strings and other components, and provides a rigid interface to mount the module to the racking system. Although seemingly simple in design, modern module frames undergo considerable design analysis and testing to use the minimum amount of material and survive the huge loads the module may see from wind and snow. Typical minimum loading tests and specifications are defined by the UL1703 specification. Some manufacturers exceed this minimum with modules that can withstand >100-mph winds and up to 5 ft of the wettest snow. The cell string lamination consists of the Si cells themselves, and a number of layers to encapsulate and protect the cells from the elements. It typically has a glass top cover. Figure 10.4 shows a basic cell lamination. Even something as

SOLAR CELL SYSTEM DEFINITION

223

seemingly simple as the top glass is highly optimized and serves multiple functions. First, it is a cover to protect the cells from environmental structural damage. The glass is also the primary structural support component of the cell string area, just like the frame is the structural support of the entire module. The glass also needs to be highly transparent in the convertible wavelength region of the cells, so that no reduction of available solar energy occurs in the glass. Furthermore, modern glass covers have special structure patterns on both sides of the glass. On the bottom surface of the glass, the texture allows for good lamination adhesion of the cell strings. The texture on the top surface is designed to help the glass minimize soiling, which obviously can reduce the convertible light transmission. The next layer in the stack is the EVA copolymer, which serves the function of being a bond layer between the glass and cells’ ARC. The ARC helps to trap incident photons within the cells so that they may be converted to electron–hole pairs and produce electricity. Next comes a second bond layer of EVA, and finally the back-side protective coating is made of Tedlar, a polyvinyl fluoride film with high emissivity for good radiative heat rejection and low UV degradation. Also within the lamination are the electrical interconnects between the cells, generally a tin-coated copper ribbon. As stated, PV cells are wired in series to form strings (of cells), which are then wired in series within an individual module. The next component that is needed to allow the modules to be wired together into an array is the junction box. The junction box is where all of the series-connected cells that are wired together using the tin-coated copper ribbon merge into a single positive and negative electrode. The junction box also represents some expense and is responsible for about 12% of the system failures, as will also be seen in a subsequent section. Since cells and modules are series connected, there is a potential for a low-performing part of the system to drag down the entire system performance, as each series-connected component is (electrically) current limited by the weakest performing component in the series. What this means in a practical sense is that the weakest link in the chain, the lowest performing cell, string, or module, can limit the electrical current and thus the system power of the entire system. A shaded or low-performing part or region can actually go into negative bias (negative voltage), which takes power away from the module in the form of heat generation in that cell or region and is much worse than if the cell were simply disconnected and bypassed. This is exactly what a bypass diode does. A bypass diode functions by allowing current to bypass the low-performing cell so that it does not reduce the system output as much. The system power is still decreased by the relative lack of contribution of the cell or region itself, but this is less than a reverse bias cell that is actually taking power out of the system when heating up. It is probably clear to the reader at this point that every aspect of a PV system has been optimized to increase performance and reduce cost. In the next section, the performance of a state-of-the-art utility-scale PV system is described. This system is operated by the TEP Company with the system performance analysis provided by a collaboration with Sandia National Laboratories.

224

SOLAR CELL SYSTEMS

10.3

SOLAR CELL UTILITY-SCALE SYSTEM PERFORMANCE

10.3.1

Introduction

Over 12 MWDC of grid-connected PV systems are currently installed in Arizona, primarily by the state’s two largest investor-owned utilities: APS and TEP Company. The vast majority of the state’s installed generating capacity of utilityscale PV systems (100 kW and larger) utilizes flat-plate, c-Si collector technology. The APS experience has focused on one-axis, north–south-oriented, horizontal tracking arrays. The TEP systems incorporate standardized, fixed arrays. The 5 years of TEP operating experience with these systems, including performance, cost, maintenance, installation, and design, is the topic of this section.

10.3.2

TEP Company

TEP is the second largest investor-owned utility in Arizona, providing electricity to 392,000 residential, commercial, and industrial customers in Tucson and surrounding areas in southeastern Arizona. The company operates nearly 15,000 mi (24,135 km) of transmission and distribution lines throughout its service territory of 1155 mi2 (2991 km2). The utility is involved in a very active renewable energy program. The utility-scale PV generation effort is centered at the Springerville Generating Station Solar System in eastern Arizona. As shown in Figure 10.5, this facility, one of the largest PV generating plants in the world when installed, includes 4.6 MWDC of installed PV systems. Covering 44 AC (17.8 ha), this PV generating plant is grid intertied with a 34.5-kV TEP distribution line. Although the Springerville plant includes other collector technologies including amorphous

Figure 10.5. Springerville PV generating plant.

SOLAR CELL UTILITY-SCALE SYSTEM PERFORMANCE

225

Si and cadmium telluride, c-Si accounts for nearly 80% of the plant’s capacity and is the focus of this section. The field experiences with these systems provide a treasure of information that not only establishes a baseline for today’s state-of-theart system capabilities but can also help guide the development of PV system technology for the future. These are the reasons that TEP and Sandia National Laboratories entered into a collaborative effort to track, analyze, and document the cost and field performance as well as the O&M experience associated with these systems.

10.3.3

Design and Installation Experience

The 26 c-Si systems at the Springerville plant are listed in Table 10.1. Each of these systems is an identical copy of a standardized array-field configuration that utilizes the same hardware components, wiring topology, and structural mounting plan.

10.3.4

Standard System Configuration

The standard system configuration includes ASE Americas (now Schott Solar) ASE-300-DG/50 modules and a Xantrex PV150 inverter. The arrays are mounted

TABLE 10.1. List of Springerville c-Si Systems with Each System Rated at 135 kWDC System

Installation Date

System

Installation Date

SGS-135C-1

July 13, 2001

SGS-135C-14

July 15, 2003

SGS-135C-2

July 13, 2001

SGS-135C-15

July 15, 2003

SGS-135C-3

August 17, 2001

SGS-135C-16

July 30, 2003

SGS-135C-4

October 2, 2001

SGS-135C-29

October 15, 2003

SGS-135C-5

October 23, 2001

SGS-135C-30

October 30, 2003

SGS-135C-6

December 14, 2001

SGS-135C-31

August 15, 2003

SGS-135C-12

May 30, 2002

SGS-135C-32

August 30, 2003

SGS-135C-7

August 1, 2002

SGS-135C-26

June 22, 2004

SGS-135C-8

August 1, 2002

SGS-135C-27

June 22, 2004

SGS-135C-9

August 1, 2002

SGS-135C-28

June 24, 2004

SGS-135C-10

September 17, 2002

SGS-135C-23

July 20, 2004

SGS-135C-11

June 24, 2002

SGS-135C-25

July 21, 2004

SGS-135C-13

June 15, 2003

SGS-13n-24

July 23, 2004

226

SOLAR CELL SYSTEMS

at a fixed latitude tilt of 34° facing due south with 450 modules per array. Based on each system’s areal footprint of 300 ft (91 m) north–south and 140 ft (43 m) east–west, the system power density is 110.6 kWAC/AC (273.3 kWAC/ha) of ground, which allows for generous access space for construction and maintenance activities. Each array string includes nine modules with two strings per row. The power per string is 2.7 kW, and the maximum string design voltage is 595 V at −22°F (−30°C). The operating voltage of each string is 380–430 V. The Xantrex PV150 inverter converts the variable voltage DC power to a 208-V three-phase AC power. The inverters have a maximum rating of 157 kVA, at which point they will limit output or come off-line, followed by an automatic restart. The maximum inverter rating was selected by TEP to accept increased output due to cloud-edge enhancement and cold temperatures. These enhancements have at times exceeded 160 kW. In addition, the higher rating allows for normal operation to be in the optimum area of the efficiency curve and allows the inverters to run cooler extending lifetime. Each unit has a DC disconnect, 150-kVA, 208-V to 480-V step-up/ isolation high-efficiency transformer, revenue meter, and AC disconnect. Groups of four units are connected in parallel to each of eleven 500-kVA, 480-V to 34,500-V high-efficiency step-up transformers. Each transformer has a continuous rating of 500 kVA and can accommodate up to 650 kVA for brief intervals. The high-voltage sides of the transformers are connected in parallel to a 34.5-kV underground distribution line, which connects to the overhead 34.5-kV distribution line that feeds the well field pumps of the nearby 1160-MW coal-fired Springerville Generating Station. The pumps operate continuously with an average total load of about 9000 kW.

10.3.5

Instrumentation/Testing

The Springerville PV generating plant is a normally unmanned site that is continuously monitored remotely via an Internet-based communications channel. Most operational functions such as inverter reconfiguration, fault resets, IV curve tracing, diagnostic testing, and performance analyses can be performed from the remote monitor site via the Internet communications channel. Fifty points of information are taken from each of the 26 inverters and the revenue meters on 10-s scan cycles, and averaged for both 10-s and 1-min archiving. Performance information is developed from the raw data for daily review and is archived in spreadsheet format by a control operator. Near real-time performance is available on the Internet. Alarm criteria have been developed for all operational parameters, and these alarms are logged and maintenance personnel are notified in case of operation of any array or inverter that is out of specifications. Spare parts are available on-site, and localbased service personnel are dispatched to the site to perform repairs in response to alarms. Both test equipment and trained personnel are used to diagnose and repair problems in the system in addition to continued support from the inverter and module vendors. Test equipment consists of mostly traditional utility items. For these systems, a clamp-on ammeter for DC and AC with a maximum range of

SOLAR CELL UTILITY-SCALE SYSTEM PERFORMANCE

227

0–40 amp and a separate clamp-on ammeter for DC and AC with a range of 0–1000 amp are used. At least one, preferably two, voltmeter with a DC and AC range of 1000 V and an integral ohmmeter are essential. An optical temperature sensor, frequency counter, and a dual trace oscilloscope with a range to 100 kHz are needed. Harmonic meter measurements and a three-phase power factor meter are helpful as well. Specifications for the inverters included ±1% measurements of AC and DC current and voltage, ac frequency, and IGBT temperature, as well as calculated values for AC and DC power. The PV150 can acquire all AC, DC, temperature, state of operation, and power parameters at a maximum rate of one sample set per second. The first six inverters were installed, and data were taken to confirm the ±1% measurement accuracy. During the first year, these inverters were routinely checked for performance against calibrated test equipment. After the 1-year period, the PV150 power production data are validated using the revenue-grade utility meter on the output of the 480-V transformer. These kilowatt hour meters have a tested accuracy of ±0.25%. For each of the identical 26 individual systems, there are 50 strings of nine modules for a system STCs rated array nameplate power of 135 kWDC. Each string represents 2% of the rated individual system power. Individual string power degradation can be identified when a single module fails (i.e., no current in the string) or the module voltage falls to 90% or less of a good module value (i.e., low voltage cannot provide more than 50% expected current). A reduction in total array power output of 1% as compared with the output expected given solar insolation, temperature, wind speed, and wind direction can be accurately identified through the continuous performance evaluation monitors implemented in the software used for system supervision. Air temperature, wind speed, wind direction, and plane-ofarray solar insolation measurements are monitored and recorded on 10-s intervals from three locations within the plant footprint. The three temperature sensors are all NRG, model 110 s with radiation shield, with ±0.6°F (±0.33°C) accuracy. Wind speed is measured at the three locations with an NRG model #40 anemometer with ±0.2 mph (±0.32 kph) accuracy, and wind direction is measured at the three locations by NRG model #200p units, with an accuracy of ±1%. There are three plane-of-array solar radiation sensors used so that cross-comparison is possible. These instruments are MSX-01 reference cells each using a single square polycrystalline cell with a 1-Ω resistor in parallel with the cell output, resulting in a solar insolation measurement that exhibits very little temperature sensitivity.

10.3.6

Balance of System

The array configuration is designed to minimize the array-field BOS cost. The dual-stanchion array structural supports are fabricated steel with powder coating to minimize corrosion. The steel supports are staked to the ground to prevent windinduced uplift and sliding. The site preparation includes minimal surface disturbance/leveling of the natural terrain while retaining the native vegetation as much as possible to reduce surface erosion and to minimize dirt splash on the modules

228

SOLAR CELL SYSTEMS

Figure 10.6. (a) Typical 135-kWDC system. (b) Xantrex PV150 system inverter.

during rainstorms. Mounting of the arrays to the terrain may result in a slightly jagged array appearance along the row due to surface variations but the adverse PV output effects are near zero. The array electrical interconnection uses 600-V rated DC equipment and underground AC power distribution to minimize cost. Each system is installed exactly the same using a trained local labor pool. This standardized approach has resulted in a total system BOS cost of less than $1400 per watt direct current. The energy payback time for the Springerville BOS has been documented at 0.21 years, a significant improvement over previous central plant designs. A Springerville system is shown in Figure 10.6a. Note the white inverter enclosure at the back of the arrays near the center of the picture. A closeup photo of the Xantrex PV150 inverter and enclosure is shown in Figure 10.6b. 10.3.7

System Performance

To describe the system performance of the Springerville systems, the PV energy parameters that have been established by the IEA Photovoltaic Power Systems Program as described in the IEC standard 61724 are utilized. Three of the IEC standard 61724 system performance parameters—final yield, reference yield, and performance ratio—define the system field performance in terms of energy production, solar resource, and system losses. These provide an easily understood method not only to compare system performance with other system options but also to permit system owners/customers to determine if system performance is meeting expectations. This process has been proposed for widespread adoption here in the United States, and the author/editors certainly support this effort. Since all 26 Springerville systems were totally operational beginning in 2004, average performance results are presented based on data from 2004 to 2006. In addition, specific annual results are noted for each of these 3 years as well.

SOLAR CELL UTILITY-SCALE SYSTEM PERFORMANCE

229

10.3.7.1 Final Yield The final yield is the net AC energy output of the system divided by the aggregate nameplate power of the installed PV array at an STC of 1000 W/m2 solar irradiance and 25°C cell temperature: Final yield = kWhAC kWDC . It represents the number of hours that the PV array would need to operate at its rated power to provide the same energy. All UL-listed modules require a nameplate on the back of the module that identifies the STC-rated DC power. The aggregate array power can easily be determined by summing the nameplate power ratings for the array. The average monthly final yield for all 26 Springerville systems over the past 3 years is shown in Figure 10.7a. The average annual final yield is

Final yield (kWhAC/kWstc)

180 160 140 120 100 80 60 40 20 0

(a) 250

Su un-Hours

200 150 100 50 0

(b)

Figure 10.7. (a) Average monthly final yield (kWhAC/kWDC) for all systems. (b) Average monthly reference yield (sun-hours).

230

SOLAR CELL SYSTEMS 1.00

Perform mance Ratio o

0.90 0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00

Figure 10.8. Average monthly performance ratio for all systems.

1707 kWhAC/kWDC. The average final yields for the last 3 years, 2004, 2005, and 2006, are 1720, 1669, and 1731 kWhAC/kWDC, respectively. 10.3.7.2 Reference Yield The reference yield is the total in-plane solar insolation (kilowatt hour per square meter) divided by the array reference irradiance. It represents an equivalent number of hours at the reference irradiance. The reference irradiance is typically equal to 1 kW/m2; therefore, the reference yield is the number of peak sun-hours: Reference yield = total plane of array insolation per 1 kW m 2. The average monthly reference yield for the Springerville arrays over the last 3 years is shown in Figure 10.7b. The average annual reference yield over this operating period is 2138 sun-hours. The annual reference yields for the last 3 years, 2004, 2005, and 2006, are 2175, 2054, and 2185 sun-hours, respectively. 10.3.7.3 Performance Ratio The performance ratio is the final yield divided by the reference yield and is dimensionless. It represents the total losses in the system when converting from nameplate DC rating to AC output. Typical system losses include DC wiring, module mismatch, bypass diodes, module temperature effects, inverter conversion efficiency, as well as others: Performance ratio = final yield reference yield . The average monthly performance ratio for all the systems is shown in Figure 10.8. The average annual performance ratio for all systems over this operating period is 0.79. The average annual performance ratios for all systems during the last 3 years, 2004, 2005, and 2006, are 0.78, 0.81, and 0.79, respectively.

SOLAR CELL UTILITY-SCALE SYSTEM PERFORMANCE

10.3.8

231

System Maintenance Experience

For the past several years, Sandia has been working to develop a comprehensive database model to track the life cycle costs of PV systems. This database, which continues to undergo improvements, was utilized to capture, document, and track schedule and unschedule maintenance service, repairs, replacements, and labor and travel costs associated with maintenance activities for these systems. Based on Microsoft Access, the database architecture is modular to support future additions, allows associations at the component level, allows multiple components to be tracked with a system, and provides for multiple failures to be documented as a result of a maintenance visit. Failure modes (what and why), activity dates (failure and repair), and costs (labor, parts, and travel) were captured and analyzed from system maintenance activity logs covering the period of mid2001 through 2006. From these data, analyses of failure modes and O&M costs were made. The Springerville systems provide a significant database for assessing the reliability and maintenance needs for a utility-scale generating plant operating in a utility environment. Altogether, a total of 11,700 identical PV modules and 26 identical inverters have been installed since mid-2001. Over the operating history from mid-2001 through 2006, a total of 156 unscheduled maintenance events were recorded for the Springerville systems. The events are grouped by categories including DASs, inverters, junction boxes, arrays (PV), systems, and AC disconnects. Figure 10.9a presents the breakdown of these events by component as a percentage of the total number of events. The unscheduled events resulted in a loss of generating capacity that affected one or more systems and required human intervention to restore the system(s) to full operational capacity. These events could be as simple as a manual restart of a tripped inverter or considerably more complex such as the repair of damage resulting from a lightning strike (the plant experienced strikes in years 2003–2005). An examination of maintenance events by category provides some insight into just how exemplary the maintenance experience at Springerville has been. Over half of the 32 AC disconnect events were associated with high contact resistance in the 480-V outdoor rated fused disconnects. In 2006, the contacts for all Siemens brand 480-V disconnects were changed, and the factory-installed contact grease was removed. The high-resistance problem appears to be due to dirt accumulation in the grease after 5 years of operation. Ten of the 11 failures associated with the DAS were all due to a severe 100-year lightning storm at the site in July 2003. That same storm caused 14 of the 58 problems with the inverters, 12 of which required replacing the PCU card. In addition, that same storm also accounted for 8 of the 13 system events that involved replacing damaged utility meters. Many of the other inverter events were associated with manually resetting trips, a problem since avoided in 2004 when autoreset capabilities were added to the inverters. As for junction boxes, 11 of the 18 events were associated with replacing failed blocking diodes. Eleven of the 24 PV array events were associated with damage from another lightning storm in July 2004. In general, the reliability of the systems has

232

SOLAR CELL SYSTEMS

Figure 10.9. (a) Unscheduled maintenance events by component. (b) Unscheduled maintenance events by category. (c) Inverter repairs by events and category.

been excellent, and the reliability of the primary PV components, modules, and inverters, has been impressive. Figure 10.9b presents a breakdown of unscheduled events by component as a percentage of the total unscheduled repair costs. As noted, the majority of the repair costs are associated with the inverters. A more detailed examination of the unscheduled inverter events provides a real-world perspective of the maintenance expectations in a utility environment. A breakdown of inverter events by repair category is presented in Figure 10.9c. The categories refer to the inverter operation. Controller includes those functions and circuit components necessary to control the power conversion and protec-

SOLAR CELL UTILITY-SCALE SYSTEM PERFORMANCE

233

tive devices. Interlock refers to the DC disconnect and door interlock alarm circuitry. Design includes cabinet weather protection. Internal refers to fault conditions within the inverter not otherwise identified. Matrix is the power electronic inverter bridge and associated switching transistors, capacitor bank, heat sink, and cooling fans. Other captures those events such as wiring and contactor problems, inoperative switches, and unknown events that signal a fault condition. As noted earlier, 12 of the 20 controller events involved the replacement of PCU cards and the addition of enhanced lightning protection due to the July 2003 lightning strike. This enhanced protection involved the addition of lightning arrestors and associated surge-resistant components in many areas of the data collection system and on the 480 voltmeters of every inverter. The other eight events primarily involved PCU card replacements due to failures and/or intermittent problems. The interlock events were all associated with DC disconnect faults ranging from connector problems and nesting rodents to unknown causes. The five design events included faults due to one roasted spider, one roasted rodent, two cases of rain, and one case of blowing snow ingress into the cabinet. Improved gasket placement has solved those problems. In each case, the internal events involved an inoperative inverter with no obvious problem that responded to a restart. The six matrix events involved two cases of matrix failure and replacement, two cases of temperature sensor failure and replacement, one case of fan assembly and motor failure and replacement, and one unknown cause of high temperature alarm on the heat sink. The 10 other events involved loose connections, replacement of switches, a failed front panel, wiring problems, and unknown causes. Although not trouble free, the 26 inverters have provided an enviable maintenance record especially in the context of other documented inverter field performance problems. Through January 1, 2007, the 26 c-Si Springerville systems had provided 1206 system-months of continuous operation since installation. Over that same period, a total of 156 unscheduled maintenance events were recorded, which provides a mean time between unscheduled services per system of 7.7 months of operation. Scheduled maintenance was conducted on the plant each year. This included mowing the native vegetation as well as visual inspections of the arrays and power handling equipment. Table 10.2 lists the annual maintenance cost, both scheduled TABLE 10.2. Maintenance Cost as a Percentage of Capital Investment Year

Scheduled (%)

Unscheduled (%)

Total (%)

2002

0.08

0.01

0.09

2003

0.07

0.22

0.29

2004

0.06

0.04

0.10

2005

0.06

0.01

0.07

2006

0.04

0.03

0.07

234

SOLAR CELL SYSTEMS

and unscheduled, as a percentage of the cumulative capital investment by year. The average annual maintenance costs since the initial Springerville installations are 0.12% of the initial capital cost. While the above maintenance costs include unscheduled repair/service on the inverters, costs of inverter rebuild (anticipated every 10 years) are not included. Including this expense on an amortized basis is estimated to increase the annual maintenance cost by an additional 0.1 percentage point. Daily performance analysis tools pinpoint underperforming units, thus allowing for timely resolution of problems with minimal lost energy production. Consequently, overall system effective availability for years 2002, 2003, 2004, 2005, and 2006 is 99.43%, 99.78%, 99.72%, 99.81%, and 99.75%, respectively, quite high for any generating technology. (Lightning created problems in 2003 were not included as performance reductions. They were in 2004 and 2005.)

10.3.9

Capacity Factor

The average monthly capacity factor for the Springerville systems over their operating history is presented in Figure 10.10. As presented here, the capacity factor is defined as the ratio of net electrical generation for the time considered to the energy that could have been generated if the system were generating at continuous full power during the same period: Annual capacity factor = annual final yield 8760. The average annual capacity factor for all systems over the 5-year operating period was 19.5%. In addition to Sandia National Laboratories’ work with TEP Company on PV system performance and analysis, the system engineering resources of the national lab has provided the PV community with the much needed reliability analysis and

Capac city Factor %

25 20 15 10 5 0

Figure 10.10. Average monthly capacity factor for Springerville systems.

SOLAR CELL SYSTEM RELIABILITY

235

predictive PV engineering. This system reliability analysis stems from Sandia’s role in engineering high-surety, high-consequence engineering systems and will be described in the next section.

10.4 SOLAR CELL SYSTEM RELIABILITY—USING THE PAST TO PREDICT THE FUTURE This section is a condensation of an IEEE PVSC34 conference paper submitted by Sandia National Laboratories by Collins et al. [2] on PV system reliability. It describes the application of a system reliability model developed for nuclear weapons that has now been applied to the TEP Company’s PV field. It is also more generally applicable to any PV system.

10.4.1

Introduction to Solar Cell System Reliability

System level reliability and availability estimates are required to facilitate cost trade-off studies associated with competing PV systems. Estimates of reliability are necessary in developing maintenance cost projections over the system’s lifetime. Availability estimates provide an input into annual energy generation projections. The reliability and availability of large PV systems have not been thoroughly investigated. Manufacturers in the PV industry are offering warranties of 20 years and better for PV modules with incomplete knowledge of their reliability in the diverse environments in which these modules are deployed. All stakeholders, including entrepreneurs, manufacturers, regulators, utility operators, and consumers, need a useful predictive model for reliability and availability. This predictive model can facilitate trade-offs in requirements and life cycle cost; identify improvements in design, manufacturing, and in situ operation; and provide estimates of performance over the system’s lifetime. This section describes a standard methodology used to characterize a system’s reliability and availability. The elements of a reliability/availability improvement program are illustrated in Figure 10.11. The three basic activities associated with a reliability/availability program are FMEA, system reliability/availability modeling, and accelerated life tests. System reliability/availability modeling allows quantification of system reliability and availability using multiple data inputs, such as field data, test data, and accelerated life test data. A system reliability model is a diagrammatic representation of all functions, in terms of subsystem or component events, that must occur for a successful system operation. This section describes a comprehensive approach to developing reliability and availability estimates for a large PV system. System reliability and availability were defined based on the operator’s expectations, and an RBD was developed to model system behavior. The RBD developed is a hierarchical reliability model.

236

SOLAR CELL SYSTEMS System topology

Field failure data Maintenance data

System Reliability/Availability Model

Field failures

Theoretical/empirical models that predict failure

Failure effects on system

Failure Modes and Effects Analysis

Life cycle cost of system and service life prediction

Data needs for failure models

Accelerated Tests Failure mechanisms to be accelerated

Figure 10.11. Overview of reliability program for PV systems.

Larger functional elements are decomposed into smaller functional elements. The granularity of the model is determined by the level that failure data are collected. FMEA is a technique for systematically identifying, analyzing, and documenting the possible failure modes within a design and the effects of such failure modes on system performance or safety. FMEA is an inductive bottom-to-top analysis. Failure modes are identified at a basic part level, and their effects are worked upward through each level of the system to identify overall system level impact. The purpose of FMEA is to identify failure effects on the system to be included in the RBD and identify potential failure modes/mechanisms that could be accelerated to provide failure rate information. Field data, failure times, and repair times were collected and analyzed for a 5-year time period from a 4.6-MWDC PV system operated by TEP at Springerville, Arizona. Failure and repair distributions were fitted to these field data. These results were then used to populate the RBD and produce system level estimates of reliability and availability. The results of these analyses are 1. a summary of failures for each main component of the system, 2. a summary of failure distributions/rates and repair times for each main component, 3. system reliability and availability versus time projections, and 4. an estimate of the number of failures for each main component over the system’s life.

SOLAR CELL SYSTEM RELIABILITY

237

An RBD is constructed by identifying all necessary functions and their associated components that must occur for the system to provide an output. The blocks are then arranged in a diagram in order of operation. Hierarchical system block diagrams are high-level diagrams that contain one or more subsystems or components within a block. Each block may have an associated RBD defined by lowerlevel functions. Accelerated life tests are tests that are run at elevated stress levels outside the stresses expected in normal operation. The objectives of these tests are 1. 2. 3.

identification of the life distribution parameters of time to failure for the applied stress, identification of relationships (mathematical or physical/chemical) between time to failure and stress, and evidence about whether failure mechanisms stimulated by the accelerated stresses are expected to occur at operational stress levels.

Accelerated life tests allow collection of time-to-failure information in situations where the time to test at normal stress levels is inordinately long, or the sample sizes required are too large. In this program, accelerated life tests will be used to estimate time to failure for failure mechanisms that require long periods of time to manifest. 10.4.2

Case Study: Reliability Model for TEP Springerville PV System

This reliability study focused on the c-Si PV portion of TEP Springerville, Arizona, grid-connected PV system. c-Si PV modules comprise approximately 80% of the PV generating system’s capacity. (Editor’s comment: very little, if any, field performance data and reliability data are available in the literature as of this book edition for thin film PV systems.) An RBD was developed for the c-Si portion of the PV system. To develop the RBD, a definition of system success was needed. A decision was made to have a simple definition for system success. Success was defined as the PV system delivering power to the grid. A hierarchical system model was developed describing the functional elements necessary to deliver power to the grid using a commercial software tool, ReliaSoft (Tucson, AZ) BlockSim 7™. The granularity of the model, the basic blocks in the model, was based on the major components of the system and the level of identification of field failures in the reporting process. The reader is referred to the original Collins et al. [2] paper for more details. To yield useful reliability and availability metrics, the RBD must be populated with both life distributions and repair distributions within each block. 10.4.3 Life Data to Populate the Reliability/Availability Model and Results of Analysis TEP shared detailed production and outage reports for the years 2003–2007. These reports were analyzed to extract data about failures, failure causes, times to failure,

238

SOLAR CELL SYSTEMS

2 00 4

EVENTS 60

2003 2004 2005 2006 2007

2 00 5

40

2003

2006 2007

20 03

ing Ma

rsh

all

sfo ran VT 48

0V

/34

.5k

20 04 20 05

x Bo

2 0 04

rm

rm sfo an Tr 0V /48 8V 20

2006 2007

er 2 00 5

2 00 5

er

20 06 2007

2003

2 00 4

2 006 2 007

200 3

2 00 4

du Mo PV

nn co AC

dis

20 0 5

2003

le

20 07

2003

ec

20

t 2004 05

2006 2007

20 04 2 0 05

2 003

x Bo w Ro

htn Lig

Inv

ert

er

0

ing

2 00 4 2 00 5

10

2006 2007

20

2003

200 6

30

2006 2007

Number of Events

50

Coded Category By Year

Figure 10.12. Failure count for components with failure.

and repair times for elements of the system. The times to failure data for each major element of the system were analyzed using commercial software, ReliaSoft Weibull++™, to fit life distributions and estimate parameters of the distributions. A simple reliability metric is a count of failures associated with each component. Figure 10.12 provides a failure count for each major component by year. The component with the most failures is the Xantrex (Vancouver, Canada) PV Inverter. During the first 2 years, the lightning event block was second on the list of failures. This block is not a physical component of the system but represents the multiple failures caused by lightning strikes to the PV system. Lightning damaged inverters, PV modules, and the monitoring system. Severe lightning storms resulted in significant damage to the system in 2003 and 2004. 10.4.4

Life Distributions and Parameter Estimates for Blocks

The life data extracted from the production and outage reports for 2003–2007 were organized in times to failure or times to suspension for situations where no failures occurred for each component grouping. These data were then analyzed using the ReliaSoft Weibull++ and RGA 6™ (ReliaSoft, Tuscon, AZ) software tools. Components that did not fail during the 5-year data period were assumed to have a reliability of 1.

SOLAR CELL SYSTEM RELIABILITY

239

Two basic approaches to data analysis are used for the components of the PV system. For the components that are replaced when they fail, life data analysis is employed. Data are fitted to common life distributions such as lognormal, exponential, two-parameter exponential, two-parameter Weibull, and three-parameter Weibull. Once data are fitted to a distribution, the parameters of that distribution are estimated using one of several available techniques. The “best” match life distribution is then selected based on weighing three criteria: 1.

2.

3.

goodness of fit (the p value from the Kolmogorov–Smirnov test), the maximum absolute difference between the hypothesized and empirical cdfs; a likelihood ratio, the value of the log likelihood of the hypothesized distribution with the estimated parameters evaluated with the given data set; and plot fit, the mean absolute difference between the hypothesized and empirical cdfs.

Table 10.3 provides a summary of the distributions and estimated parameters used to model the components that are replaced when they fail. Scale and location parameters are in days or the applicable transformation. For components that are repaired rather than replaced, parametric recurrent data analysis is used. This approach is based on the GRP model. It is used when a component accumulates more than one failure over its service life. This model

TABLE 10.3. Summary of Life Distributions and Parameter Estimates for Replaced Components PV Component/ RBD Block

Distribution

Beta or Log SD (Shape)

Eta or Log Mean or Lambda (Scale)

AC disconnect

Weibull 3-RRX

0.35

Lightning

Exponential 1-RRX

Row box

Weibull 2-RRX

0.51

1.2E+06

PV module

Weibull 3-RRX

0.28

5.2E+12

480/34.5-kV transformer

Weibull 2-RRX

0.58

7100

208/480 transformer

Weibull 3-RRX

0.15

1.3E+10

Marshalling box

Lognormal 2-RRX

2.3

10

11,000

Gamma (Location) 3.9

0.00022

17

28

240

SOLAR CELL SYSTEMS

is particularly useful in modeling the failure behavior of a component and understanding the effects of repair on the age of the component. In order to obtain a virtual age, the exact occurrence time of failures should be available. However, the times are unknown until the corresponding event occurs. Therefore, the software uses Monte Carlo simulation to predict values for virtual time, failure number, MTBFs, and failure rate. Three coefficients are estimated by the software: two power law coefficients and a repair effectiveness coefficient. Parametric recurrent data analysis was used for the inverter only. Many inverters had multiple failures over the analysis time period 2003–2007. Life data analysis was used for all other components of the PV system. Analysis of the field data for inverters yielded a power law model with estimates of beta equal to 0.75 and lambda of 0.019 per inverter. The repair effectiveness q was estimated to be zero. A zero value means repair did not degrade the inverter’s reliability.

10.4.5

Maintenance Distributions and Parameter Estimates for Blocks

Repair times, time from failure to restoration, and other downtimes were identified for each component from the production and outage reports. For inverters, these downtime events were sorted into three categories: corrective maintenance, preventive maintenance (scheduled and unscheduled), and grid effects. Downtimes caused by grid effects are a special category. These downtimes are associated with inverters tripping and resetting due to voltage or frequency excursions in the external power grid connection that result in no physical damage to the system. A similar approach to life data analysis was taken for fitting repair distributions and estimating parameters. Data are fitted to common life distributions, such as lognormal, exponential, and Weibull. Once data are fitted to a distribution, the parameters of that distribution are estimated. Scale parameters are in days or the transformation. The results are provided in Tables 10.4 and 10.5.

TABLE 10.4. Summary of Repair Distributions and Parameter Estimates for the Inverter PV Component/RBD Block

Distribution

Corrective maintenance

Lognormal-RRX

Preventive maintenance

Exponential 1-RRX

Grid effects

Weibull 2-RRX

Beta or Log SD (Shape)

Eta or Log Mean or Lambda (Scale)

2.27

−4.25 2.62

1.07

0.16

SOLAR CELL SYSTEM RELIABILITY

241

TABLE 10.5. Summary of Repair Distributions and Parameter Estimates for Other Components PV Component/ RBD Block AC disconnect

Distribution

Beta or Log SD (Shape)

Eta or Log Mean or Lambda (Scale)

Weibull 2-RRX

0.71

1.4

Lightning

Weibull 2-RRX

0.73

10.8

Row box

Lognormal 2-RRX

2.07

−0.98

PV module

Lognormal 2-RRX

3.11

−1.37

480/34.5-kV transformer

Weibull 2-RRX

0.53

1.36

208/480 transformer

Lognormal 2-RRX

1.6

−2.33

Marshalling box

Weibull 2-RRX

0.35

3.55

10.4.6

Plots of Reliability and Availability versus Time

The life and repair distributions for all the components were imported into the BlockSim RBD for the PV System. A Monte Carlo simulation that sampled from the life distributions at discrete intervals and determined component and system state (operational, or down due to failure or maintenance) was executed. For availability, repair times are sampled for the various repair distributions associated with each block to determine component and system states. Results of these simulations at the system level were a straight-line plot indicating an availability of 100%. At this level, the model predicts at least one inverter will always supply some power to the grid. Figure 10.13 illustrates plots of reliability and availability for the inverter with PV array segment of the system. The reliability in Figure 10.13 reflects the effect of no repair or replacement of components. TEP’s rapid repair policy enables very high levels of availability with lower component reliability.

10.4.7

Expected Number of Failures in 20 Years

The expected number of failures as predicted by the model for each component for 5, 10, and 20 years are shown in Table 10.6. The number of failures predicted for the inverter assumes no additional reliability growth, a conservative assumption. The model does not currently incorporate the effects of degradation or wear out failure mechanisms. These data are not currently available for the TEP PV system. For the first 5 years, the inverter repair rate was 0.96 per inverter per year. For the PV modules, the replacement rate was approximately 5 in 10,000 PV modules per year.

242

SOLAR CELL SYSTEMS ReliaSoft BlockSim 7 - www.ReliaSoft.com

Availability and Reliability vs Time for 5 Years 1.000

Inverter With Array

Point Availability Line Point Reliability Line

A(t), R(t)

0.800

0.600

0.400

0.200

0.000 0.000

Mike Mundt Sandia Labs 5/2/2009 7:01:00 PM

365.000

730.000

1095.000

1460.000

1825.000

Time in Days, (t)

Figure 10.13. Plot of reliability and availability versus time for inverter with PV array.

TABLE 10.6. Predicted Number of Component Failures Component

Actual Number of Failures, 5 Years Cumulative

Expected Number of Failures, 5 Years Cumulative

Expected Number of Failures, 10 Years Cumulative

Expected Number of Failures, 20 Years Cumulative

125

132

231

429

PV module

29

26

31

38

AC disconnect

22

17

23

31

Lightning

16

10

20

41

PV 150 inverter (26 c-Si arrays)

208/480 transformer

4

3

3

3

34

25

35

50

Marshalling box

2

4

7

11

480/34.5-kV transformer

5

4

5

9

Row box

SYSTEM COST EXPERIENCE

243

The mean availability predicted by the model approaches 100% for the 5-year period. The effective availability reported by TEP for the system was 99.91% in 2007. Effective availability is defined as the actual power produced divided by the total power that could have been produced. Inverters are the most unreliable component in this system. Yet the availability of continuous power delivered to the grid is projected to be very high over the life of this system. However, an increase in inverter reliability can still lower corrective maintenance costs over the system’s life.

10.5

SYSTEM COST EXPERIENCE

10.5.1

Cost for TEP Springerville Solar Cell System

TEP is realizing significant cost benefits by incorporating standardized products, volume purchasing, and efficient array-field design and installation. The Springerville experience has documented some of the lowest installed system costs ever reported thereby establishing a benchmark for state-of-the-art utility-scale systems. A cost breakdown for systems installed in 2004 (the last year for system installations at Springerville) is presented in Table 10.7: 1. 2.

Modules: the module price reflects a bulk purchase from the module manufacturer. Array-field BOS: the site preparation cost includes ground leveling, fencing, and underground wiring. Structure cost includes mechanical mounting of the modules, support structure hardware, and staking. The electrical work includes module interconnect wiring, conduit, junction

TABLE 10.7. Cost Breakdown of Springerville Systems System Component

$/WDC

$/WAC

Modules

3.33

4.22

Array-field BOS

0.56

0.71

Inverter/transformers

0.40

0.51

Indirect/overhead

1.11

1.40

Total

5.40

6.84

Site preparation ($0.10/WDC) Structure ($0.15/WDC) Electrical ($0.30/WDC) AC Intertie $0.01/WDC)

244

SOLAR CELL SYSTEMS

3.

4.

10.5.2

boxes for both the string and row buses, disconnect switches, system protection and wiring on the AC side of the inverter to the 480-V transformer, and the DAS. The AC intertie cost includes the wiring and installation labor from the 480-V to the 34.5-kV transformer. Inverter/transformers: this cost includes the purchase price of the Xantrex PV150 Inverter, the 150-kVA 208/480-V transformer for each system, and one-fourth of the 480/34.5-kV transformer cost (each 34.5-kV transformer gathers four of the systems). Installation labor for these components is included. Indirect/overhead: indirect costs include system design, procurement, construction management, and project engineering. The overall project management for the Springerville installations was provided via contract by Tucson-based Global Solar Energy.

Energy Cost Figure of Merit

The true measure for comparing different PV system options is the cost of delivered kilowatt hour alternating current energy. To put the Springerville cost experience in perspective, Moore and Post [1] utilized an energy cost figure of merit defined as the average installed system cost (U.S. dollar per kilowatt direct current) divided by the energy output (kilowatt hour alternating current per kilowatt direct current) expected over a 30-year period. Although the resulting cost figure represents U.S. dollar per kilowatt hour alternating current, this figure does not include financing costs, the cost of capital, O&M costs, or any tax considerations and, thus, is not an LEC and is not portrayed as such (note that LEC for TEP is addressed in the next section). Using the 2004 system costs, this energy cost figure of merit is $0.10 per kilowatt hour alternating current for the Springerville systems. Interestingly, the Springerville energy cost figure of merit for fixed flat-plate systems is nearly identical to the energy cost figure of merit reported for one-axis, tracking horizontal flat-plate systems installed at Prescott, Arizona. It is also interesting to note the significant energy cost figure-of-merit difference between the utility-scale Springerville systems and residential systems. A total of 82 SunShare residential systems have been installed in the TEP service territory during the past few years and are being tracked by TEP and Sandia for performance, cost, and maintenance experience. These residential-size systems range from 1.2 to 5.9 kWDC. Using the average installed cost for these systems in 2004 of $7.32 per watt direct current and an average annual final yield of 1398 kWhAC/kWDC provides an energy cost of $0.175 per kilowatt hour alternating current for the residential systems, significantly higher than the utility-scale option. The energy cost of system O&M can also be described by a figure of merit defined as the annual cost of O&M divided by the annual energy output. As noted above, this is not an LEC, but it does provide a perspective on the cost impact of maintenance experience with the Springerville systems. Using the average annual

SYSTEM COST EXPERIENCE

245

maintenance cost of 0.12% of installed capital cost, the annual O&M energy cost is $0.004 per kilowatt hour alternating current. Including the expected inverter, rebuild costs increase the annual O&M energy cost to $0.007 per kilowatt hour alternating current. Considering the SunShare residential systems noted above and using the average annual maintenance cost of 1.7% of installed system cost provides a comparative annual O&M energy cost of $0.089 per kilowatt hour alternating current, an order of magnitude higher than the utility-scale systems.

10.5.3

Economic Perspective

The experience at Springerville provides a valuable utility perspective on the future use and needs of PV technology. These include actual utility-based energy generating costs, capacity factors, and operational aspects associated with solar electric generation. 10.5.3.1 Energy Cost The SEIA has provided a road map with established goals for expanding the use of solar power-generating capacity here in the United States. It is of interest to note that a photo of the Springerville systems is featured in this road map document. The road map goal over the next decade for PV systems is a selling price of $3.68 per watt alternating current in 2015 and a cumulative installed U.S. capacity of 9.6 GW. Coupling the TEP cost experience at Springerville with this SEIA cost goal provides an interesting perspective for the future of PV. Table 10.8 presents a comparison of today’s benchmark system costs for Springerville and a proposed breakdown of a 2015 utility-scale PV system meeting the road map goal in today’s dollars. Using the Springerville performance ratio of 0.79, the $3.68 per watt alternating current future system cost corresponds to an equivalent cost of $2.91 per watt direct current. The 2015 system cost components follow a proposed breakdown developed elsewhere for a c-Si system. The 2015 module cost is based on a manufacturing cost analysis for a c-Si production plant of 25 MW/ year developed by Spire Corporation. The proposed module cost is also consistent TABLE 10.8. System Costs for the Future System Component

Springerville System ($/WDC)

2015 System ($/WDC)

Modules

3.33

1.78

Array field

0.56

0.58

Inverter

0.40

0.25

Fixed

1.11

0.30

Total

5.40

6.84

246

SOLAR CELL SYSTEMS

with c-Si manufacturing cost projections developed through the U.S. DOE PVMaT program. While module costs and fixed costs require substantial cost reductions to achieve the 2015 goal, this comparison validates the creative system BOS approach developed by TEP at Springerville by already achieving the array-field BOS target projected for the next decade. As annual PV installation quantities increase in future years, it is expected that the fixed costs will be diluted over larger amounts of installed capacity and will be reduced on a U.S. dollar per watt direct current basis. The industry road map goal for 2015 is an LEC of $0.057 per kilowatt hour alternating current of PV generation. This compares to the TEP-calculated LEC in 2006 (pay as you go, no financing costs) of $0.062 per kilowatt hour alternating current for the Springerville PV generation. The TEP calculation, which includes both federal income tax credits and state property tax reductions for solar, already meets the road map baseline 2015 LEC of $0.115 per kilowatt hour alternating current. It is important to note again that the TEP strategic plan to incorporate solar generation in its service territory is focused on a pay-as-you-go funding to avoid the high costs of financing. The attractive TEP-calculated LEC for this facility is a direct result of this approach.

10.6

CONCLUSIONS

The energy data, maintenance experience, and costs with the Springerville c-Si systems provide a treasury of information that establishes a benchmark for current utility-scale fixed flat-plate PV systems technology. This operating assessment has identified a number of findings, including

• • • • • • • •

average annual AC system energy output is 1707 kWhAC/kWDC of array; average annual AC system power is 0.79 of the array DC nameplate rating; average annual O&M cost is 0.12% of initial system installed capital cost, not including rebuild/replacement cost of the inverter; the mean time between unscheduled maintenance services per system is 7.7 months of operation; innovative approaches including standardized array designs, low-cost array-field BOS, and bulk hardware purchases have resulted in an installed system cost of $5.40 per watt direct current; the average annual capacity factor for all systems was 19.5%; the LEC cost calculated by TEP (no financing costs) is $0.062 per kilowatt hour alternating current, which meets the 2015 SEIA baseline goal for PV generation; control instabilities associated with cloud passage over the PV plant require further efforts including possibly smart inverters and/or storage to smooth these generation transients and improve capacity credit for PV generation.

ABBREVIATIONS

247

ACKNOWLEDGMENTS The key contributions of Hal Post, Michael Dvorack, Jeff Mahn, Michael Mundt, and Michael Quintana are gratefully acknowledged.

ABBREVIATIONS AC—alternating current ACD—alternating current disconnect APS—Arizona Public Service ARC—antireflective coating A(t)—availability as a function of time BOS—balance of system cdfs—cumulative distribution functions c-Si—crystalline silicon DASs—data acquisition systems DC—direct current DOE—Department of Energy EVA—ethylene vinyl acetate FMEA—failure modes and effects analysis GRP—general renewal process IEA—International Energy Agency IEEE—Institute of Electrical and Electronic Engineers IGBT—insulated gate bipolar transistor IV—current voltage kWAC—kilowatt alternating current kWDC—kilowatt direct current kWhAC—kilowatt hour alternating current LEC—levelized energy cost ModJct—module junction boxes MTBFs—mean time between failures MWDC—megawatt direct current O&M—operation and maintenance PCU—power control unit PV—photovoltaic PVMaT—Photovoltaic Manufacturing Technology q—repair effectiveness RBD—reliability block diagram RGA—reliability growth analysis R(t)—reliability as a function of time SD—standard deviation SEIA—Solar Energy Industries Association Si—silicon STCs—standard test conditions

248

SOLAR CELL SYSTEMS

t—time TEP—Tucson Electric Power UL—Underwriters Laboratories, Inc. UV—ultraviolet WAC—watt alternating current WDC—watt direct current REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11]

[12] [13] [14] [15]

L. M. Moore and H. N. Post. Five years of operating experience at a large, utilityscale photovoltaic generating plant. Progress in Photovoltaics: Research and Applications 16, 249–259 (2007). E. Collins, M. Dvorack, J. Mahn, M. Mundt, and M. Quintana. Reliability and availability analysis of a fielded photovoltaic system. In 34th IEEE Photovoltaic Specialists Conference, Philadelphia, June 7–12 (2009). L. Moore, H. Post, H. Hayden, S. Canada, and D. Narang. Photovoltaic power plant experience at Arizona Public Service—a 5-year assessment. Progress in Photovoltaics: Research and Applications 13, 353–363 (2005). L. Moore and H. Post. Photovoltaic power plant experience at Tucson Electric Power. Energy Conversion and Resources 2005, 387–394 (2005). Information available at www.tucsonelectric.com. Tucson Electric Power Company. Renewables Data for Year End 2006. Annual Report to the Arizona Corporation Commission. Available at http://www.greenwatts.com/Docs/ACCAnnual2006.pdf (2006). www.GreenWatts.com/pages/solaroutput.asp T. Hansen. The systems driven approach to solar energy: A real world experience. Proceedings of Solar Energy Systems Symposium, October 15–17, Albuquerque, NM (2003). Available at www.sandia.gov/pv (2003). J. E. Mason, V. M. Fthenakis, T. Hanse, and H. C. Kim. Energy payback and lifecycle CO2 emissions of the BOS in an optimized 3.5 MW PV installation. Progress in Photovoltaics: Research and Applications 14, 179–190 (2006). IEC. Photovoltaic system performance monitoring—Guidelines for measurement, data exchange, and analysis. IEC Standard 61724, Geneva, Switzerland (1998). B. Marion, J. Adelstein, K. Boyle, H. Hayden, B. Hammond, T. Fletcher, B. Canada, D. Narang, D. Shugar, H. Wenger, A. Kimber, L. Mitchell, G. Rich, and T. Townsend. Performance parameters for grid-connected PV systems. Proceedings of 31st IEEE Photovoltaic Specialists Conference, January 3–7, Lake Buena Vista, FL (2005). M. Thomas, H. Post, and R. DeBlasio. Photovoltaic systems: an end-ofmillennium review. Progress in Photovoltaics: Research and Applications 7, 1–19 (1999). L. Moore. Sandia’s PV reliability database: helping business do business, quarterly highlights of Sandia’s solar programs. Vol. 1. Available at www.sandia.gov/pv (2001). W. Bower. Inverters—Critical photovoltaic balance-of-system components: status, issues, and new millennium opportunities. Progress in Photovoltaics: Research and Applications 8, 113–126 (2000). A. Maish. Defining requirements for improved photo-voltaic system reliability. Progress in Photovoltaics: Research and Applications 7, 165–173 (1999).

REFERENCES [16] [17] [18] [19] [20]

[21] [22] [23]

249

R. West, K. Mauch, Y. C. Qin, N. Mohan, and R. Bonn. Status and needs of power electronics for photovoltaic inverters: summary document. SAND2002-1085, Sandia National Laboratories, Albuquerque, NM, April (2002). L. Moore, H. Post, T. Hansen, and T. Mysak. Residential photo-voltaic system experience at Tucson Electric Power. Sandia National Laboratories Internal Paper (2006). SEIA. Our solar power future: The U.S. photovoltaics industry roadmap through 2030 and beyond. January. Available at www.seia.org (2005). R. Little and M. Nowlan. Crystalline silicon photovoltaics: The hurdle for thin films. Progress in Photovoltaics: Research and Applications 5, 309–315 (1997). R. Mitchell, C. Witt, R. King, and D. Ruby. PVMaT advances in the photovoltaic industry and the focus of future photo-voltaic manufacturing R&D. In Proceedings of 29th IEEE Photovoltaic Specialists Conference, N.D. Stojadinovic, ed., pp. 1444– 1447 (2002). L. H. Crow. Reliability analysis for complex repairable systems. In Reliability and Biometry SIAM, Army Material System Analysis Activity Technical Report 138, pp. 379–410. Philadelphia (1974). M. Kijima and N. Sumita. Some results for repairable systems with general repair. Journal of Applied Probability 20, 851–859 (1989). R. Mettas and W. Zhao. Modeling and analysis of repairable systems with general repair. In IEEE Reliability and Maintainability Symposium Proceedings, pp. 176– 182 (2005).

11 LEVELIZED COST OF ENERGY FOR UTILITY-SCALE PHOTOVOLTAICS MATTHEW CAMPBELL SunPower Corporation

11.1

THE DRIVERS OF THE LCOE FOR UTILITY-SCALE PVs

PV power plants have emerged in recent years as a viable means of large-scale renewable energy power generation. A critical question facing these PV plants at the utility scale is the competitiveness of their energy generation cost with that of other sources. A common means of comparing the relative cost of electricity from a generating source is through an LCOE calculation. The LCOE equation allows alternative technologies to be compared when different scales of operation, investment, or operating time periods exist. This chapter reviews the LCOE drivers for a PV power plant and the impact of a plant’s capacity factor on the system LCOE. The impact of solar tracking to a plant’s capacity factor is reviewed as well as the economic trade-offs between fixed and tracking systems. From 2004 to 2008, the market for small (<50 MW) distributed PV power plants took off around the world, particularly in Spain and Germany, where over 3 GW of power plants were installed. PV power plants have also emerged in the United States, where large plants have been constructed in recent years, or are under construction, as seen in Figure 11.1. The 25-MW (AC) FPL Desoto PV power plant built by SunPower in 2009 represents the first true utility-scale PV power plant in North America. Looking to the future, PG&E in California has announced more than 2 GW of agreements involving both solar thermal and PV power plants, including more than 750 MW of PV power—the largest utility-scale contracts for PV in the world. As part of this program, SunPower will construct a 250-MW central station, highefficiency, PV power plant in the state’s California Valley. It is expected to be the Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

251

252

LEVELIZED COST OF ENERGY

Figure 11.1. FPL Desoto (25 MW) first North American utility-scale PV power plant.

first to deliver utility-scale PV power to PG&E starting in 2010. With the LCOE falling rapidly for central station PV plants, their economic competiveness with other renewables and peaking power sources is driving adoption of the technology. 11.2

INTRODUCTION TO LCOE

The LCOE equation is one analytic tool that can be used to compare alternative technologies when different scales of operation, investment, or operating time periods exist. For example, the LCOE could be used to compare the cost of energy generated by a PV power plant with that of a fossil fuel-generating unit or another renewable technology [1]. The calculation for the LCOE is the net present value of total life cycle costs of the project divided by the quantity of energy produced over the system life: LCOE =

Total life cycle cost Total lifetime energy production

The LCOE equation can be disaggregated for solar generation as follows: N Depreciation n Initial − ∑ × ( Tax Rate ) + n Investment n =1 (1 + Discount Rate ) N Annual Costs n Residual Value × (1 − Tax Rate) − ∑ n (1+Discount Rate)n n =1 (1 + Discount Rate )

Initial kWh kWp × (1 − System Degradation Rate ) ∑ (1 + Discount Rate )n n =1 N

n

MAJOR LCOE INPUTS

253

When evaluating the LCOE and comparing other commonly known U.S. dollar per kilowatt hour benchmarks, it is important to remember that the LCOE is an evaluation of levelized life cycle energy costs. The price of energy established under PPAs or FITs may differ substantially from the LCOE of a given PV technology as they may represent different contract or incentive durations, inclusion of incentives such as tax benefits or accelerated depreciation, financing structures, and, in some cases, the value of time of day production tariffs.

11.3

MAJOR LCOE INPUTS

11.3.1

Initial Investment

The initial investment in a PV system is the total cost of the project plus the cost of construction financing. The capital cost is driven by

• • •

11.3.2

area-related costs, which scale with the physical size of the system namely the panel, mounting system, land, site preparation, field wiring, and system protection; grid interconnection costs, which scale with the peak power capacity of the system including electrical infrastructure such as inverters, switchgear, transformers, interconnection relays, and transmission upgrades; and project-related costs such as general overhead, sales and marketing, and site design, which are generally fixed for similarly sized projects.

Depreciation Tax Benefit N

Depreciation n

n =1

(1 + Discount Rate)n

∑

× ( Tax Rate)

The depreciation tax benefit is the present value of the depreciation tax benefit over the financed life of the project asset. Public policy, which enables accelerated depreciation, directly benefits the system LCOE because faster depreciation translates to faster recognition of the depreciation benefit.

11.3.3

Annual Costs N

Annual Costs n

n =1

(1 + Discount Rate)n

∑

× (1 − Tax Rate)

In the LCOE calculation, the present value of the annual system operating and maintenance costs is added to the total life cycle cost. These costs include inverter

254

LEVELIZED COST OF ENERGY

maintenance, panel cleaning, site monitoring, insurance, land leases, financial reporting, general overhead, and field repairs, among other items.

11.3.4

System Residual Value Residual Value n

(1+ Discount Rate)n The present value of the end of life asset value is deducted from the total life cycle cost in the LCOE calculation. Silicon solar panels carry performance warranties for 25 years and have a useful life that is significantly longer. Therefore, if a project is financed for a 10- or 15-year term, the project residual value can be significant.

11.3.5

System Energy Production N

∑

n =1

Initial kWh kWp × (1 − System Degradation Rate)

n

(1 + Discount Rate )n

The value of the electricity produced over the total life cycle of the system is calculated by determining the annual production over the life of the production, which is then discounted based on a derived discount rate. The first-year energy production of the system is expressed in kilowatt hours generated per rated kilowatt peak of capacity per year (kWh/kWp). The kilowatt hour per kilowatt peak is a function of

• • • • • •

the amount of sunshine the project site receives in a year; how the system is mounted and oriented (i.e., flat, fixed tilt, tracking, etc.); the spacing between PV panels as expressed in terms of system GCR; the energy harvest of the PV panel (i.e., performance sensitivity to high temperatures, sensitivity to low or diffuse light, etc.); system losses from soiling, transformers, inverters, and wiring inefficiencies; and system availability largely driven by inverter downtime.

To calculate the quantity of energy produced in future years, a system degradation rate is applied to initial system performance to reflect the wear of system components. The system degradation (largely a function of PV panel type and manufacturing quality) and its predictability is an important factor in life cycle costs as it determines the probable level of future cash flows. Finally, the system’s financing term will determine the duration of cash flows and impact the assessment of the system residual value.

LCOE VARIABLES FOR UTILITY-SCALE PV

11.4

255

THE LCOE MODEL SENSITIVITY

The LCOE is highly sensitive to small changes in input variables and underpinning assumptions. For this reason, it is important to carefully assess and validate the assumptions used for different technologies when comparing the LCOE. Figure 11.2 illustrates the model’s sensitivity to input assumptions. We provide three scenarios that all start with the same PV system price and predicted energy output using a tracker in a high insolation1 location. We then modify (1) the annual degradation rate, (2) the forecasted economic life, (3) the annual O&M expense, and (4) the discount rate. The resulting LCOE for the three scenarios range from $0.09 to $0.23 per kilowatt hour, illustrating that for the same system capital cost and initial energy output, the range of energy prices can vary by a factor of 2 or more. Comparing LCOE calculations and power plant energy pricing requires aligning assumptions across examples and calibrating against empirical data to generate a more accurate LCOE forecast. One use for LCOE calculations is to compare costs without incentives. If incentives such as the U.S. ITC are assumed in an LCOE calculation, they should be specifically referenced to make clear the basis for comparison between technologies. Given the high sensitivity of the LCOE to input variables, it is important to understand the validity of performance output over a system’s lifetime. Silicon PV systems have been operating outdoors for more than 20 years [2], and therefore, the performance and degradation mechanisms are well understood. For siliconbased PV systems, it is possible to accurately forecast future output, allowing one to populate the LCOE equation variables with a high level of confidence.

11.5

LCOE VARIABLES FOR UTILITY-SCALE PV

To understand the LCOE outlook for utility-scale PV, it is important to understand the lifetime system performance and cost. The following sections summarize key cost and performance drivers for a utility-scale PV power plant.

System Price kWh/kWp Annual Degradation System Life Annual O&M  $/kWh Discount Rate LCOE   $/kWh

Case 1 Case  2 Case  3 100% 100% 100% 100% 100% 100% 1.0% 0.5% 0.3%            15           25           40 $     0.030 $    0.010 $    0.005 9% 7% 5% $      0.23 $      0.13 $      0.09

Figure 11.2. Solar PV LCOE sensitivity to variable changes. 1

Insolation is the level of solar radiation received at a given location.

256

11.5.1

LEVELIZED COST OF ENERGY

PV Power Plant Performance

The lifetime energy generated from a PV power plant is a product of the plant location, annual performance for a given capacity, component degradation, and system lifetime. 11.5.2

System Capacity Factor

The capacity factor is a key driver of a solar project’s economics. With the majority of the expense of a PV power plant being fixed, capital cost LCOE is strongly correlated to the power plant’s utilization. The annual capacity factor for a PV power plant is calculated as Annual kilowatt hours generated for each kilowatt AC of peak capacityy ( kWh kWp ) . 8760 h in a year

Capacity Factor (AC)

A PV power plant’s capacity factor is a function of the insolation at the project location, the performance of the PV panel (primarily as it relates to hightemperature performance), the orientation of the PV panel to the sun, system electrical efficiencies, and the availability of the power plant to produce power. The economic impact of the capacity factor is substantial. Figure 11.3 illustrates a range of identical LCOE values, expressed in U.S. dollar per kilowatt hour, for a given PV power plant system price as expressed in U.S. dollar per watt peak and the associated capacity factor. As the capacity factor declines, the required installed system price must also substantially decline to maintain system economics. For example, a $2.50 per watt peak system with a 24% capacity factor (such Sample Range of Equivalent LCOE Values 40% 38% 36% 34% 32% 30% 28% 26% 24% 22% 20% $2.00 $2.50 $3.00 $3.50 $4.00 System Price - $ / Wp (DC)

LCOE Equivalence

Figure 11.3. Associated capacity factors and system prices producing an identical LCOE. (Note the listed capacity factors are based on the AC rating of the power plant at the point of grid interconnection; the DC nameplate capacity of the PV power plant will be approximately 20% higher than the AC rating depending on the PV panel type and system configuration.)

LCOE VARIABLES FOR UTILITY-SCALE PV

257

as with a fixed-tilt configuration) delivers the same LCOE as a $3.50 per watt peak system with a 34% capacity factor (such as with a tracker). The highest capacity factors are generated with trackers, which follow the sun throughout the day to keep the panel optimally oriented toward the sun. This tracking also has the benefit of generating more energy in the peak electricity demand periods of the afternoon. SunPower has developed two patented tracking systems to optimize the capacity factor of a PV power plant: the T0 Tracker—optimized for space-constrained sites—and the T20 Tracker—optimized for maximum energy production. See Figures 11.4–11.6.

Figure 11.4. SunPower’s T0 Tracker.

Figure 11.5. SunPower’s T20 Tracker.

LEVELIZED COST OF ENERGY

SunPower Production vs. Load Profile

Summer CA Peak Load 45

180

Summer total MWh produced per MW nameplate

40 160 35 140 30

120

25

100

20

80 60 40

T20 Tracker

15

Fixed Tilt

10

CAISO load 20

5

0

0

Figure 11.6. Comparison of California summer load requirements with fixed and tracking PV systems.

The LCOE model assigns an equal value to electricity generated throughout the year. However, electricity generated at peak periods is more valuable to the utility. The use of tracking with a solar system can increase the output of a plant after 4 p.m. in the summer by more than 40%, which is often a period of peak demand on the system when energy is highly valued. Figure 11.6 gives a comparison for the summer energy output of a fixed and tracking PV power plant as compared with the California ISO load. A tracker enables higher output during the peak afternoon period for a given plant capacity.

11.5.3

PV Panel Performance and Lifetime

Successful prediction of PV panel performance over time is critical to project investors. Furthermore, demonstrating the historical performance of a company’s panel technology is critical to determine financing parameters, which underpin the LCOE calculation. Silicon PV has the longest operating history of any solar cell technology. The photograph in Figure 11.7 shows a monocrystalline silicon panel after 20 years of outdoor exposure with no major visual degradation. Studies on the performance of silicon PV panels show only 4% total degradation after 23 years of outdoor exposure.2 This experience provides a high level of confidence in making future per2

Capacity factor is generally expressed as a function of the AC rating of a plant so the above kilowatt hour per kilowatt peak calculation is based on the kilowatt hour per AC watt peak as opposed to the DC watt peak.

(CAISO Load) 000s

258

LCOE VARIABLES FOR UTILITY-SCALE PV

259

Figure 11.7. Monocrystalline silicon PV panel after 20 years of outdoor exposure.

LCOE vs. System Life 100% 95% 90% 85% 80% 75% 70% 65% 60% 20

LCOE vs. System Life

25

30

35

40

Financeable System Life (Years)

Figure 11.8. LCOE sensitivity to financeable system life.

formance predictions. Note that most investors finance a solar system based on an assumed panel degradation rate of 0.5–1.0%/year, a faster rate than this historical data for silicon PV might indicate. Research on silicon PV historical performance suggests that panel life may extend much further than the 25-year design life [3]. This demonstrates that long-term performance may enable longer financeable system lives in the future. Figure 11.8 illustrates the LCOE model sensitivity to financed system life based on a 7% discount rate. As indicated in the figure, extending the financed term of the project beyond today’s 20- to 25-year values could have a material impact to the LCOE.

260

LEVELIZED COST OF ENERGY

11.5.4

Predicting System Performance

In addition to calculating PV panel output, an estimate of the system’s overall performance must be made to finance a project. The key variables in a PV power plant’s performance are plant uptime, weather-based performance (insolation, ambient temperature, soiling, etc.), inverter and power system efficiency, and system component degradation (largely from the panel). SunPower has developed an analytic model, PV grid, which accounts for the above variables and makes future performance predictions based on SunPower’s experience with more than 450 installed commercial rooftop and power plant systems. With this tool, SunPower provides project investors with a well-demonstrated means of estimating project cash flows. Figure 11.9 illustrates the actual versus expected performance for a 10-MW SunPower tracking power plant system in Germany, Bavaria Solar I. During the first 3 years of operation, the system performance has exceeded the performance estimates under which the project was financed (Fig. 11.10). This correlation between empirical data and future predictions is critical in reducing investor risk and the related cost and terms of capital investments. An important path to utility-scale LCOE reduction is to demonstrate to investors the predictable output, degradation, and system life, which would support a lower cost of project capital. As more PV data are generated and investors become more familiar with the technology, this may become possible.

Typical Year Energy Production

Expected Energy Production

Actual Energy Production

Production (In Millions of KWh)

14 12 10 8 6 4 2

2005

2006

2007

2008 (YTD)

Figure 11.9. Expected and actual energy production for 10-MW Bavaria Solar.

INITIAL PV POWER PLANT INVESTMENT

Isla Mayor Spain, 8.4 MW SunPower T0 Tracker

Jumilla, Murcia, Spain-Elecnor 23 MW SunPower T0 Tracker

261

Muehihausen, Bavaria, Germany, Trujillo, Extremadura, Spain-Elecnor 6 MW SunPower T0 Tracker 23 MW SunPower T0 Tracker

Serpa, Portugal 11 MW SunPower T0 Tracker

Las Vegas, US-Nellis AFB 14.2 MW SunPower T20 Tracker

Figure 11.10. Representative experience of SunPower PV Power Plant Technology.

11.6

INITIAL PV POWER PLANT INVESTMENT

11.6.1

PV Panel

When discussing the potential for PV solar cost reduction, the focus is understandably placed on the panel. Over the past several years, solar panel prices have represented approximately $4 per watt peak of total PV system installed prices of $6–$9 per watt peak [2], depending on the market and application type. Until 2004, PV cell and panel production costs were steadily declining following classic learning curve behavior as the solar industry grew. In 2004 and through 2008, however, the rapid growth in PV demand led to a global shortage of solar-grade polysilicon, the key raw material used in conventional silicon solar cells. The spot market price of polysilicon during this period rose from $25 per kilogram to greater than $500 per kilogram for some reported transactions. The cost of polysilicon became the driving cost of a conventional solar panel, increasing production costs to artificially high levels relative to the historical learning curve. As a secondary effect, solar cell manufacturing costs also suffered as the result of underutilized, silicon-constrained factories. In 2007, some solar manufacturers entered into new intermediate and longterm contracts that will continue through the rest of the decade, lowering feedstock costs for those that have contracted for that silicon. The polysilicon industry should also benefit from an improved cost structure—as compared with preshortage levels—due to the scale economies of the new factories being built and new silicon purification process technology. In the first half of 2008, SunPower saw its first

262

LEVELIZED COST OF ENERGY

material silicon cost reductions as we benefited from the delivery of substantial volumes of polysilicon under supply contracts from new production facilities. One benefit of the silicon shortage was that the cost and scarcity of silicon prompted a significant improvement in silicon utilization by solar cell manufacturers. In SunPower’s case, the grams of polysilicon consumed to manufacture a watt at the solar cell level declined from 13 g/W in 2004 to 6.3 g/W in 2008 and is planned to decline to an estimated 5 g/W with SunPower’s Gen 3 technology now under development. By 2011, this approximately 60% reduction in the use of silicon, coupled with an approximately 50% decline in the price of polysilicon, will independently drive large cost reductions for PV panels. Cell and panel conversion costs are driven by yield, depreciation, labor, chemical consumption, electricity cost, and materials. Conversion costs can be improved by shorter and more efficient processes, higher throughput production lines, larger plant sizes driving scale economies, and greater automation, among other factors. With a combination of lower silicon consumption and cell manufacturing costs, SunPower estimates that by the end of 2009, high-efficiency silicon panels will have a manufacturing cost below $2.00 per watt peak (DC) and that by 2014, panel costs will fall below $1.00 per watt peak (DC) with cell efficiencies of 25%. All of these costs are also leveraged by the efficiency of the solar cell. SunPower’s cost structure and cell efficiency advantage demonstrate that higherefficiency cells can absorb the increased manufacturing costs to make each cell due to the higher watts per cell. Efficiency advantages continue downstream into panel assembly, sales, marketing, and installation. For example, holding all other costs constant, an increase in cell efficiency of one percentage point will equate to approximately a 5% decrease in installed system costs. Figure 11.11 illustrates the Solar Cell Efficiency (%) 22 = Average production efficiency 20 18 16 14 12 10 8 6 Thin Films

Ribbon

Conventional

HIT

SunPower A-300

Figure 11.11. Relative solar cell conversion efficiencies.

SunPower Gen 2

INITIAL PV POWER PLANT INVESTMENT

263

solar conversion of efficiency of SunPower’s solar cells relative to conventional silicon and thin film PV technologies. 11.6.2

Area-Related Expenses

PV power plant area-related expenses include system costs, which directly scale with the area of PV panels used. These expenditures are the dominant nonpanel costs in a PV system and include steel, foundations, mounting hardware, plant installation, shipping and warehousing, field wiring, and the electrical components used to connect the panels. Area-related costs are highly correlated with the prices of steel, copper, and concrete, as well as transportation expenses. The structural materials necessary for panel installation are driven by the wind load requirements of the project. These are a function of the PV panel surface area that is exposed to the wind whether the system is tracking or fixed (similar to how the wind force on a sail is a function of the sail size). As a result, simplified tracking and fixed-tilt configurations share similar cost structures with the exception of the drive and control components. There is a common misconception that trackers significantly add to the cost of a system over fixed-tilt configurations. SunPower has developed trackers that can move up to 300 kWp of panels with a simple half-horsepower motor, which requires little maintenance. SunPower has determined that the financial benefit of the increased energy production generated by tracking the sun significantly outweighs the incremental system costs. By the end of 2008, SunPower and its partners will have deployed more than 250 MW of tracking systems on three continents. With this experience, SunPower has determined that tracking systems have delivered superior LCOE economics for its customers than fixed configurations. Area-related installation costs can vary substantially by site and by country. For example, a fence post-like support foundation might be easily driven into the ground in Bavaria, whereas a South Korean typhoon zone may require a thick steel beam placed in a hole drilled into rock and secured with reinforced concrete at four times the cost. As a result, the range of foundation costs for a fixed-tilt system or single-access tracker could vary from $30 to $200 per square meter of PV depending on the site. Additionally, differences in government electrical codes can significantly impact costs; one jurisdiction may require expensive steel wire conduit, while others allow the direct burial of cable into the ground. Once the area-related costs for a system are calculated, a simple transformation to U.S. dollar per watt peak can be accomplished by dividing U.S. dollar per square meter by the watt peak per square meter of the panel. In the case of SunPower’s high-efficiency panels, area-related U.S. dollar per watt peak costs are approximately 50% lower than thin-film PV panels. Figure 11.12 demonstrates how area-related costs are leveraged through efficiency for a sample central station PV solar power plant with 1 TWh of annual energy production. Note that although the material costs are higher for standard efficiency and thin-film panels, they are largely similar to what they would be with a fixed-tilt system, so tracking still makes economic sense provided there is available land.

264

LEVELIZED COST OF ENERGY

PV Panel Used for Project PV Panel Efficiency System Size (DC kWp) Land Required (Sq Miles) Truckloads of Concrete Required Steel Required (Tons) Cabling Required (Miles) Trenching Required (Miles) Modules to Wash (Sq Feet) Tracker Installation Labor (hours) Concrete Required (Tons)

Area Related Costs for a 1 TWh T20 Tracker Project SunPower Standard Efficiency Thin Film 20.0% 14.0% 11.0% 423,191 439,754 430,108 5.11 7.58 9.44 12,718 18,218 22,678 35,083 50,258 62,561 410 587 730 32 45 57 14,579,689 21,643,300 26,941,782 365,450 523,516 651,678 254,353 364,367 453,568

Figure 11.12. Area-related cost components for a T20 Tracker power plant with 1 TWh of annual production.

Land used for solar power plants has been readily available and inexpensive in the past, largely because the land had little economic value other than in some cases of low-yielding agricultural activities. As solar power plant developers began acquiring land in South Korea, southern Europe, and the southwest United States, prices for prime land conducive to a solar power plant rapidly increased in cost, and general land availability became an issue. Korea and southern Europe have seen solar-suitable land price increases of more than 300%, and southwest desert land has sold for prices as high as a reported $23,000 per acre for flat land [4] with high insolation located near electrical transmission lines, a roughly 15,000% increase over historical values for the same parcels. There are two fundamental drivers for the land consumed by a solar power plant: solar panel efficiency and system GCR. System GCR is the ratio of solar panel area to land area. PV panels mounted flat use the land most efficiently and have the maximum GCR but have the lowest capacity factor, meaning lower utilization of fixed plant costs. Conversely, a two-axis tracker has the maximum possible capacity factor but requires up to 10 times more land than flat configurations. To put it simply, the better the orientation to the sun (thus capacity factor), the longer the shadows created and therefore the further apart panels must be placed to avoid panel to panel shading. To deliver the best utility-scale PV LCOE, one must balance land use with the system capacity factor. SunPower addresses this optimization problem by manufacturing the world’s highest efficiency PV panels along with tracking systems that efficiently use land while increasing energy production. SunPower’s tracker offerings include the T20 Tracker, which maximizes capacity factor in an efficient land footprint, and the T0 Tracker, which optimizes land use for constrained sites while still providing a high capacity factor. Figure 11.13 illustrates the land consumption versus capacity factor for a power plant producing 1 TWh/year in a high insolation location. One can see in this example that

•

with high-efficiency PV panels, up to 75% less land is required for a given capacity factor configuration; and

PV POWER PLANT OPERATING EXPENSES

265

Land Requirements for a High Insolation PV Power Plant with 1 TWh of Annual Production 18

Square Miles Required

16 14 12 10

20% SPWR 14% xSi

8

11% CIGS/CdTe TF 6

6% aSi TF

4 2 – 34% - T20 Tracker

32% - T0 Tracker

26% - Fixed Tilt

AC Capacity Factor

Figure 11.13. Land use and capacity factors for 1-TWh production configurations.

•

11.6.3

with high-efficiency PV panels mounted on trackers, up to 30% higher capacity factors are attainable while using a similar or lower amount of land per quantity of energy produced than low- and medium-efficiency panels mounted on fixed-tilt systems. This means that lower LCOE configurations are achievable without prohibitively increasing the amount of land required. Grid Interconnection Costs

Grid interconnection costs relate to the inverter, transformer, switchgear, medium voltage substation, and electrical interconnect, the high-current electrical backbone bringing power in from the array and ultimately the transmission back to the central grid in the case of a power plant requiring a transmission upgrade for grid integration. These costs are driven by the price of the manufactured components, the skilled labor used to install, and the price of copper, which drives much of the inverter and electrical wiring costs. Power transmission costs are driven down through scale economies, more intelligent system design, and through improved plant utilization such as with solar tracking. 11.7

PV POWER PLANT OPERATING EXPENSES

The O&M of a PV power plant is relatively straightforward because there are few moving parts and no cooling systems. O&M costs generally scale with three factors: (1) system peak power dominated by inverter maintenance, (2) system annual energy production density, and (3) general site-related items.

266

LEVELIZED COST OF ENERGY

Improving the capacity factor of a system directly reduces O&M costs through higher utilization rates of fixed assets. Figure 11.14 demonstrates this as it relates to the inverter requirements to generate 1 TWh of annual energy in a PV power plant. In this example, 1 TWh of energy would require 335 inverters, each 1 MWp, with a SunPower T20 Tracker versus 442 inverters with a fixed-tilt system at the same location. The use of a tracking system would therefore significantly reduce the inverter O&M cost. Significant power-related maintenance costs also exist with respect to transformers, switchgear, and grid interconnection, and all benefit from a high capacity factor system configuration. Module cleaning, panel repair or replacement, mounting structure and wiring maintenance, and vegetation control all scale with the annual energy production density of the panels. The annual energy production density is a critically important factor for system economics (both O&M and overall LCOE). The annual energy production density is the kilowatt hour generation per unit area per year measured as Annual energy production density = kWh m 2 year . The impact of the annual energy production density can be substantial. Figure 11.15 shows the area of PV panels required in a high insolation solar power plant to generate 1 TWh of annual output.

T20 Tracker Capacity Factor 1 MWp Inverters per annual TWh Inverter O&M Cost

34.1% 335 100%

Fixed Tilt 25.8% 442 132%

Figure 11.14. Inverters required for 1 TWh of energy production in the southwest U.S. desert.

Square Miles of PV panels per 1 TWh 2.5 2.0 1.5 1.0

Sq. Miles of PV

0.5 0.0 SunPower 120 Tracker

11% CIGS/Cdte Fixed Tilt

6% aSi Fixed Tilt

Figure 11.15. PV panels required for 1 TWh of annual production.

SYSTEM RESIDUAL VALUE

267

O&M costs, which correlate with the area of PV panels used, can thus be reduced using high-efficiency PV panels mounted on tracking systems. A simple example of these O&M savings is with the cost of cleaning panels. With a high annual energy production density panel, washing costs can be reduced by up to 75%. This allows for either a direct reduction of O&M costs or allows for panels to be washed more frequently and economically, increasing system annual energy production. Although often overlooked, washing and soiling can have a material impact to a PV power plant LCOE. In a tracking system, there is the added cost of motor and controller maintenance. But in SunPower’s experience, this cost is relatively small when compared with the other O&M cost savings the tracker provides. For example, the SunPower motor requires only annual lubrication, and a single motor can control more than 300 kWp of PV. Also, the tracker bearings require no lubrication and are designed for more than 25 years of use. The O&M cost of a utility-scale tracking system would be less than $0.001 per kilowatt hour over a fixed configuration, which does not include the O&M savings from the increase in energy production. Looking to the future, opportunities for O&M cost reduction include improved inverter reliability, scale economies from larger plant sizes, automated washing and water recycling tools, and sophisticated remote monitoring.

11.8

SYSTEM RESIDUAL VALUE

Related to the previous section, solar PV financial models generally assign zero residual value to the project. The system, however, could have a useful life of 50 years or more, yielding a material residual value to the system after the 20- or 25year financed term. Additionally, the PV power plant could increase in value if fossil fuel-based energy prices continue to rise. Due to the time value of money, the LCOE impact of a system’s residual value is diluted but could still materially reduce a PV power plant’s LCOE. It is conceivable that in the future, PV systems will be treated as assets with an active secondary market. In the wind industry, secondary turbine sale and refurbishment has begun to occur.3 SunPower has seen some value being placed on the future reclamation of the structural steel used in its power plants, but placing a value, the residual energy of a PV power plant is still immature in the market.

3

Within the PV industry system, prices and sizes are often referred to in terms of the DC watt peak of the system such as here. In other instances, AC watt peak prices and sizes are published. AC watt peak prices are higher than DC values because of the losses in transforming power from DC to AC, that is, a 1-MW DC system at $7.00 per watt peak might be rated as 0.8-MW AC and $8.75 per AC watt peak.

268

11.9

LEVELIZED COST OF ENERGY

SUNPOWER’S LCOE FORECASTING TOOL

SunPower has set a company goal of reducing the LCOE of its installed system cost by at least 50% by 2012 based on 2006 costs. Through its vertical integration, SunPower has a unique window into the detailed costs of a solar system—from quartz mining for metallurgical silicon to the construction and maintenance of a PV power plant. To plan and track LCOE reductions by market and application around the world, SunPower has developed a Web-based database (Fig. 11.16) that aggregates hundreds of cost, performance, and financial inputs from its projects. The project dovetails with SunPower’s research and development work funded by the U.S. Department of Energy’s SAI. The SAI sets forth aggressive solar LCOE reductions through technological and process innovation [5]. The LCOE for an incremental PV power plant to be built in the future is influenced by a variety of external factors including exchange rates, labor prices in respective manufacturing and construction locations, scarcity of critical raw materials, the cost of capital, land prices, and many other factors. These risks are minimized by the use of a high-efficiency solar panel technology like SunPower’s since the efficiency leverages almost all non-PV plant costs. Once built, the LCOE of energy coming from a silicon PV power plant is very predictable since the LCOE is heavily influenced by capital cost, location, and systems technology choice. Based on extensive LCOE scenario analysis with a range of cost and performance structures for incumbent and emerging solar technologies, SunPower believes that utility-scale, central station solar power plants built with highefficiency silicon PV will deliver a competitive LCOE now and in the future.

Figure 11.16. SunPower’s LCOE forecasting tool.

ABBREVIATIONS

11.10

269

CONCLUSIONS

We conclude that on the many dimensions of cost and performance that underpin the LCOE for a solar power plant, high-efficiency tracking PV offers a very compelling solution. To review, the LCOE is the net present value of total life cycle costs of the project divided by the quantity of energy produced over the system life: LCOE =

Total life cycle cost . Total lifetime energy production

Key LCOE benefits for high-efficiency PV power plants include the following: 1.

2.

Lowest total life cycle cost • High-efficiency panels minimize power plant capital costs through the reduction in the number of modules and scale of the mounting system and land required to generate a given amount of energy. • Higher conversion efficiencies, more efficient use of silicon, and larger-scale manufacturing operations will drive continued high-efficiency panel cost reductions. • Life cycle O&M costs are substantially lower for high-efficiency tracking PV due to up to four times the energy production per panel per year. • A higher system residual value for a silicon PV plant drives total life cycle cost reduction. Highest total lifetime energy production • Through optimized solar tracking, SunPower PV power plants maximize the annual energy production of a system, leading to high capacity factors and a lower LCOE. • With a more than 20-year operating history, monocrystalline PV modules provide predictable energy production, which reduces investor investment risk and enables longer financeable system lives.

LCOE analysis shows how SunPower’s high-efficiency silicon PV power plants generate electricity at a price competitive with other peak power resources. Based on comparison between published cost predictions for other technologies and our internal cost reduction road map and resultant LCOE forward cost curve, we expect to maintain this competitive position into the future.

ABBREVIATIONS AC—alternating current DC—direct current

270

LEVELIZED COST OF ENERGY

FITs—feed-in tariffs FPL—Florida Power and Light GCR—ground coverage ratio ISO—Independent System Operator ITC—Investment Tax Credit kWp—kilowatt peak LCOE—levelized cost of energy MWp—megawatt peak O&M—operation and maintenance PG&E—Pacific Gas & Electric Company PPAs—Power Purchase Agreements PV—photovoltaic SAI—Solar America Initiative Wp—watt peak REFERENCES [1] [2] [3] [4] [5]

W. Short, D. Packey, and T. Holt. A manual for the economic evaluation of energy efficiency and renewable energy technologies. National Renewable Energy Laboratory, NREL/TP-462-5173, March (1995). E. Dunlop, D. Halton, and H. Ossenbrink. 20 years of life and more: Where is the end of life of a PV module? In Photovoltaic Specialists Conference, 2005. Conference Record of the Thirty-first IEEE, January 3–7, Lake Buena Vista, FL (2005). F. De Lia, S. Castello, and L. Abenante. Efficiency degradation of C-silicon photovoltaic modules after 22-year continuous field exposure. Proceedings of the 3rd World Conference on PV Energy Conversion, May 2003, Osaka, Japan (2003). A. Kimber. Long term performance of 60 MW of installed systems in the US, Europe, and Asia. Proceedings of the 22nd Annual Photovoltaic Solar Energy Conference, September 2007, Munich, Germany (2007). T. Woody. The southwest desert’s real estate boom. Fortune Magazine, July 11, 2008. http://money.cnn.com/2008/07/07/technology/woody_solar.fortune/index.htm.

PART III TERRESTRIAL CONCENTRATOR SOLAR CELL SYSTEMS

12 LOW-CONCENTRATION CRYSTALLINE SILICON SYSTEMS LEWIS FRAAS JX Crystals Inc.

12.1

INTRODUCTION—WHY CONCENTRATE SUNLIGHT?

The problem with solar electricity today is that it is too expensive. As discussed in Chapters 1 and 2, while solar module prices fell during the 1990s, traditional flat-plate module prices have stabilized recently. As discussed in Chapter 3, the problem is that there is a trade-off between solar cell cost and solar cell efficiency. The traditional single-crystal silicon cells require expensive purification limiting harmful impurities to less than 10 ppb and expensive crystal growth and cell fabrication steps. Silicon cell efficiencies as high as 23% are achieved when these steps are taken. Unfortunately, these silicon cells and modules and the resultant electricity are then expensive. Meanwhile, when thin-film cells are used with less expensive amorphous or small grain-size materials, the result is low conversion efficiency. While the thin-film modules are then less expensive, the larger installed systems and resultant electricity are still expensive. The problem is that both low-cost solar collectors and high-efficiency solar converters are needed. Concentrating the sunlight onto efficient single-crystal cells can potentially produce lower-cost solar electricity by providing a second component. Low-cost plastic or glass lenses or sheet metal mirrors can collect the sunlight and concentrate it onto the expensive high-efficiency single-crystal cells, thereby diluting their cost.

12.2

EARLY DEVELOPMENT OF CONCENTRATED PV SYSTEMS

With the first Arab oil embargo in the early 1970s, a U.S. effort toward energy independence was initiated. The solar PV community identified three approaches toward low-cost solar electricity. The first approach was simply to bring silicon Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

273

274

LOW-CONCENTRATION CRYSTALLINE SILICON SYSTEMS

solar cells down from space. This effort was sponsored by the Caltech JPL. The second approach was to attempt to make modules less expensive with thin film approaches. This effort was sponsored by NREL. The third approach was the concentrator approach sponsored by Sandia National Laboratories. Several concentrator approaches were identified as shown in Figures 12.1– 12.4. These systems can be characterized as operating with medium-, low-, or high-concentration ratios. Figure 12.1 shows an approach sponsored by Martin Marietta [1] using an array of point-focus plastic Fresnel lenses focusing sunlight 50 times only on silicon cells, and Figure 12.2 shows an ENTECH [2] linear arched plastic Fresnel lens focusing sunlight 20 times onto silicon cells. Both systems used aluminum finned extrusions for cell cooling, and both of these pioneering systems operated successfully for several years. These two systems operate in a mid-concentration range (MCPV). Their designs depart completely from the typical planar module design. However, the optical precision and tracker accuracy requirements are not extreme. There are, however, challenges in cell mounting for good voltage standoff and thermal management. For example, the cells in Figure 12.1 arrays eventually delaminated and failed. ARCO Solar made an attempt to use the typical planar silicon modules at low concentration by simply adding mirrors on either side of each module. A more recent rendition of this CPV approach is shown in Figure 12.3. The system shown in Figure 12.3 is fabricated by Abengoa and is operating [3] in Seville, Spain. A problem with this approach as illustrated is that there is no provision for additional cell cooling. This problem led to the failure of the ARCO Solar arrays at Carrisa Plains [4]. The Abengoa arrays are designed to go off-sun when the temperature gets too high. The power output from the point-focus module design shown in Figure 12.1 can be easily improved by replacing the silicon cell with a more efficient cell such

Figure 12.1. The 350-kWp SOLARAS project power plant built and deployed in Saudi Arabia by Martin Marietta using Sandia Labs technology [1].

Acrylic Fresnel Lens

US Patent # 5,498,297 Prism Covered Silicon Cell Package Extruded Aluminum Heat Sink

Figure 12.2. ENTECH MCPV module design using arched linear Fresnel lens [2].

Figure 12.3. Abengoa LCPV arrays in Seville, Spain. Fresnel Lens Panel

Cell Assembles

Sheet Aluminum Trough Housing Bulkhead

Figure 12.4. Varian high-concentration module with point-focus Fresnel lens parquet.

276

LOW-CONCENTRATION CRYSTALLINE SILICON SYSTEMS

as the MJ cells described in Chapter 3. This idea was first developed by Varian [5] in the 1980s as shown in Figure 12.4. Varian suggested increasing the concentration ratio by using a cell package with a secondary concentrator element. This HCPV approach is the subject of several of the following chapters. Unfortunately, when the oil price fell in the 1980s, solar PV funding fell dramatically. The JPL and Sandia efforts were abandoned and folded into the NREL effort, and NREL then emphasized the thin film approach. Fortunately, the air force then developed the MJ cell for satellites.

12.3

THE 3-SUN LCPV CONCEPT

There are several lessons to be learned from the pioneering work just described. A first lesson is that the MCPV modules with silicon cells will probably rapidly evolve into HCPV modules with MJ cells. However, it will then take time to develop and qualify these systems and then to develop the new optics, MJ cell, cell package, and precision tracker component manufacturing supply chain. An evolutionary design derivative from the existing silicon PV module will be much faster to market. Such a design can build off the existing silicon-module manufacturing infrastructure already in place. Since the silicon cell price today has already come down to within a factor of 3 of its economical target price, an LCPV module operating at 3-suns concentration will be sufficient. With an LCPV system in contrast to an HCPV system, very little additional complexity is added. The cost of high-purity silicon feedstock today is approximately $100 per kilogram, whereas the cost of aluminum is only about $2 per kilogram. Crystal growth adds more cost to the silicon solar cells. Therefore, substituting aluminum mirrors for a single-crystal cell area can dramatically reduce the cost of a module. This reasoning leads to the 3-sun mirror module concept [6, 7] shown in Figure 12.5. This module concept is the subject of several patents and patent applications belonging to JXC. The 3-sun LCPV mirror module concept is shown in more detail in Figures 12.6 and 12.7. In its first implementation, this concentrator module design used

Figure 12.5. JXC 3-sun mirror module concept.

THE 3-SUN LCPV CONCEPT

277

Figure 12.6. View from the back side of an A300 SunPower cell before and after being cut into third cells.

Glass Cover EVA Sheet PV Cells EVA Sheet TPT Sheet

Glass Cover EVA Sheet EVA Sheet PV Cells Voltage Stand-Off Adhesive Aluminum Sheet Heat Spreader

Figure 12.7. Top view: the standard planar silicon module laminant. Bottom view: the addition of a metal sheet heat spreader to spread the heat uniformly over the whole back plane so that the air contact area for heat removal is preserved. EVA = Ethylene Vinyl Acetate; TPT = Teflon/Polyester/Teflon.

278

LOW-CONCENTRATION CRYSTALLINE SILICON SYSTEMS

existing planar cells from SunPower. As shown in Figure 12.6, standard 125 × 125 mm SunPower A300 cells are cut into thirds. In addition, the module design used standard circuit lamination procedures and equipment. However, as shown in Figure 12.7, a thin aluminum sheet is added at the back of the laminated circuit for heat spreading. While a standard planar module contains rows of 125 × 125 mm cells, these low-concentration modules consist of rows of third cells with each row now 41.7 mm wide. Linear mirrors with triangular cross sections are located between the cell rows (Fig. 12.5). The mirror facets deflect the sun’s rays down to the cell rows. In a more generalized description of this LCPV module concept, there are three important features. The first two important features are (1) the use of rows of linear mirrors or lenses aligned with the cell rows concentrating the sunlight

A

1561 mm × 811 mm × 155mm 61.5 in × 32 in × 6.1 in Drawing dimensions in mm

Figure 12.8. Top, side, and back-side drawings of the JXC 180-W STC 3-sun PV mirror module. Note the stress relief slits in the back-side aluminum sheet heat spreader.

LCPV MIRROR MODULE DEVELOPMENT

279

onto the cell rows, and (2) the addition of a thin aluminum sheet heat spreader on the back of the circuit lamination to spread the heat away from the cell rows so that the cell operating temperature remains acceptably low. These features are highlighted in Figures 12.5–12.7. There is a third important feature shown in Figure 12.8. That feature is (3) the incorporation of slots in the back of the aluminum sheet heat spreader to accommodate the differences in thermal expansion between the silicon cells, the glass, and the aluminum so that the circuit interconnectivity is maintained over time.

12.4

LCPV MIRROR MODULE DEVELOPMENT

The development of these 3-sun PV mirror modules was enabled by a purchase order from the Shanghai city government through the Shanghai Import and Export Trading Company. This project took place in four phases over a 2-year period from 2006 to 2008. In the first phase, the 3-sun modules were designed, the first 30 were fabricated, and then tested under calibrated conditions. In the second phase, two post-mounted two-axis tracking systems were set up in Shanghai, each with 12 of the 3-sun modules. Then, in a third phase, a roof-mounted 100-kW system was built at the Shanghai Flower Park. Finally, in the fourth phase, two rooftop betasite systems were set up in the United States. In the first phase, a 3-sun module was sent to NREL for calibration measurements. The results are shown in Table 12.1. All of these measurements are consistent with a PTC 3-sun module rating of 150 W.

TABLE 12.1. NRELPV Standardized Module Performance Test Report #2K1720 JXC Low-Concentration Module (Area = 1.268 m2)

November 22, 2006

Temp (°C)

Voc (V)

Isc (A)

FF (%)

Vmax (V)

Imax (A)

Pmax (W)

22.9

24.5

9.1

67.6

19.1

7.9

150.5

32.6

23.6

10.6

65.8

17.8

9.3

LACSS Spectrolab X200 November 8, 2006 2

Direct normal outdoor 1079 W/m November 8, 2006

32.6

9.9

153.7

Normalized and spectrally corrected November 15, 2006

28.9

23.9

10.6

61.9

17.8

8.8

2

Direct normal outdoor 1081 W/m November 15, 2006

Normalized and spectrally corrected

28.9

9.8

145.6

280

LOW-CONCENTRATION CRYSTALLINE SILICON SYSTEMS Flower port flash test 300

275

250

Count

200 160

155

150 100 50 9 4 0 0 165 170 175

51 21 1

180 185 190 195 Pmax bins

0 200 205 210

Figure 12.9. STC flash-test 3-sun module power yields for 750 modules with average at 190 W and high of 205 W.

For the second, third, and fourth phases, 750 3-sun modules were then fabricated. The good yields and steadily improving performance as shown in the yield bar graph presented in Figure 12.9 was a pleasant surprise. The NREL calibrated 3-sun module was used as a reference module for these flash-test STC measurements. JXC’s 3-sun PV mirror modules have now been operating in four separate systems in the field for up to 2.5 years. Two post-mounted two-axis tracking arrays of 12 modules each were installed at the Shanghai Flower Park in April of 2006. Then 672 modules were installed in a 100-kW array on N–S horizontal beam trackers at the Shanghai Flower Port in November of 2006. Finally, two AZtracking carousels with four modules each were installed on building rooftops with one carousel at the ORNL and one carousel at the UNLV. In this section, each of these four systems will be described in succession.

12.4.1

Shanghai Two-Axis Tracking Arrays

JXC supplied twenty-four 3-sun panels for mounting on two Array Technologies post-mounted AZ225 two-axis trackers, and these were installed at the Shanghai Flower Park as shown in Figure 12.10. An SMA inverter was mounted on each post. This system is labeled as a 4-kW system based on the STC panel rating of 180 W. However, based on the PCT rating of 150 W, 12 × 150 W = 1.8 kW for each array is expected. This is consistent with the power produced from an array as shown in Figure 12.11. The peak power shown of 1.74 kW is consistent with the sun reading on July 29 of 2006 of 0.97 suns (0.97 × 1.8 kW = 1.75 kW). This system

LCPV MIRROR MODULE DEVELOPMENT

281

Figure 12.10. Two post-mounted two-axis trackers each with twelve 3-sun modules.

2kW Array 7/29/06 2 1.75

Pdc (kW)

1.5 1.25 West array 1 0.75 0.5 0.25 0 5

6

7

8

9

10

11 12 13 Time of day

14

15

16

17

18

19

Figure 12.11. Power output over a day on July 29 of 2006 for one of the 2-kW (STC) arrays.

is still operating successfully today 2.5 years after the initial installation. However, data are only available to JXC upon JXC visits to this site. On September 25 of 2008, a power output of 1.58 kW was recorded on one of these arrays with an ambient temperature of 35°C. This array has never been cleaned except by occasional rainfall.

282

12.4.2

LOW-CONCENTRATION CRYSTALLINE SILICON SYSTEMS

Shanghai Single-Axis Tracking 100-kW 3-Sun System

(1.01)

The customer’s goal from the beginning was to mount solar panels on the roof of a utility building at the Shanghai Flower Port in order to provide electric power for greenhouses to keep them cool in the summer and warm in the winter. So a low-profile horizontal beam tracking 3-sun system was designed. The building block is a 25-kW array as shown in Figure 12.12. It consists of seven beams oriented in the north–south direction driven by a motor and one central drive beam. There are twenty-four 3-sun panels astride each beam. The system that was designed and built is a nominal 100-kW (PTC) system consisting of four of these building blocks. It therefore contains 4 × 7 × 24 = 672 panels. A photograph of the completed roof-mounted 100-kW array is shown in Figure 12.13. The installation of this system was completed in the middle of November of 2006. An SMA Sunny Central inverter was used along with string monitors to read the outputs from each of the 28 strings corresponding with the 28 beams. The system performance is documented in Figure 12.14 (left) where the current for one of the typical 28 strings is shown along with the system output

(21.17)

20.76

DRIVE GEAR

9.77

1.63

12 MODULES MOUNTED IN PORTRA IT ON EACH SIDE OF DRIVE LINKAGE (168 MODULE PER TRACKER) 4 TRACKERS REQUIRED

1.56

RPC 8/20/05

ARRAY TECHNOLOGIES JX CRYSTALS FLOWER PORT SIINGLE TRACKER ASSEMBLY

nexiassy USED ON

APPICAION

Figure 12.12. Design drawing for 25-kW horizontal beam tracked array consisting of seven beams, each with twenty-four 3-sun modules, and one drive motor and drive beam.

LCPV MIRROR MODULE DEVELOPMENT

283

Figure 12.13. Photographs of the completed 100-kW 3-sun array and a series string.

voltage over a day in November of 2006. The system performance is as expected. Since the sun is at 50° off normal toward the south in the wintertime for this horizontal beam system, the max-power current reading of 5 A per string needs to be divided by cos (50) = 0.64 to predict peak summertime operation. From these data, the system peak AC power in the summer will then be (5 × 28/0.64) A × 430 V = 94 kW. Given an inverter efficiency of 94%, this then equates to 100 kW (PTC), and this is consistent with 672 × 150 W = 100.8 kW. Figure 12.14 (right) gives the more recent array power in March of 2008. Given the incidence angle i = 35.6° and cos (35.6) = 0.81, then the summer AC Pmax would be 75.8 kW/0.811 = 93.5 kW AC. This is nearly the same as for the calculated summer start-up AC power of 94 kW in November of 2006. Measurements of string performance were recently made in November of 2008 after cleaning with a projected system power again of 93 kW. There is no indication of system performance degradation.

12.4.3

Rooftop Carousel 3-Sun Systems in the United States

The Chinese 100-kW project laid the foundation for the 3-sun mirror module development to date. However, there is a significant difference between the market requirements for China versus the United States. China’s interior does not have the transmission line infrastructure already installed in the United States. So China’s solar need is for village electric power. For China, then, our modules can be used on horizontal beam trackers in fields. Meanwhile, we believe the U.S. market will be for PV integrated into commercial buildings with our target initial market in the sunny southwest. Horizontal beam trackers or post-mounted trackers are not suitable for this building-integrated market. So, JXC has designed, fabricated, and deployed a carousel tracker for use on commercial building flat rooftops [8]. This carousel tracker has been described in Chapter 9. The photograph in Figure 12.15 (left) shows a JXC carousel with 180-W 3-sun modules installed on a building flat rooftop at ORNL. This unit has been

284

100kW Array, 3/4/08, SMU 3255

100kW String Current 11/8/06 6

8 7

5

6 String Current (Amps)

Idc (amps)

4 3 2 1 0

IString 2 IString 3 IString 4 IString 5 IString 6 IString 7 IString 1

5 4 3 2 1 0

5

6

7

8

9

10 11 12 13 14 15 16 Time of Day

17 18

19

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20 Hour 1000

100

450

90

400

80

350

70

700

60

600

50

500

40

400

150

30

300

100

20

200

50

10

100

0

0

300 250 200

5

6

7

8

9

10 11 12 13 14 15 16 17 18 19 Time of Day

Pac Ppv Suns

5

6

7

8

900 800 Suns (W/m^2)

Power (kW)

Vpv (dc volts)

100kW Array, 3/4/08 500

0 9 10 11 12 13 14 15 16 17 18 19 20 Hour

Figure 12.14. Typical string currents, system voltage, and array power for a 100-kW (PTC) 3-sun system (see text for more details). Idc = DC current, Vpv = PV array DC voltage; dcc = Direct current.

FUTURE POTENTIAL MANUFACTURING AND DEPLOYMENT

285

Figure 12.15. Four CPV 3-sun modules on a carousel at ORNL (left) and UNLV (right).

successfully operating under test since September of 2007. Another carousel of this type has also been installed on a rooftop at the UNLV (Fig. 12.15, right). While 3-sun modules have been in successful operation in China, solar flux data has generally not been available for correlation with performance. The ORNL site is the first test site with good solar flux instrumentation. Data are available for the direct normal flux, diffuse, and global horizontal solar irradiance. Similar data are also available at the UNLV site. Figure 12.16 shows exemplary data collected for these two carousels over the last 18 months. These data are consistent with a 155 ± 5 W PTC 3-sun module power rating.

12.5 FUTURE POTENTIAL MANUFACTURING AND DEPLOYMENT While the goal of lower-cost solar electric power can be reached when carousels and 3-sun CPV modules reach high-volume production, unfortunately, this hardware is not yet in high-volume production. The first step has been to perform life testing and performance testing at various beta sites as has been discussed above. The performance and durability results are very promising. The next step is a commitment to manufacturing. The 3-sun LCPV approach has three advantages that promise to facilitate manufacturing. The first advantage is ready cell supply with a variety of planar silicon cell options where high-volume cell manufacturing is already in place. The second advantage is that the automated planar module manufacturing equipment is readily adaptable to 3-sun module manufacturing. And the third advantage is that a variety of one-axis trackers already in use for planar silicon module systems can be used for 3-sun systems. With regard to cell supply, a variety of planar silicon cell types can be used. In our previous work, SunPower A300 cells cut into thirds were used as shown in

286

Noon IVs

4/15/08 Pmax

12

700

8

Isc = 10.5 A

6

Voc = 94.5 V

4

FF = 0.66

500

0

20

40

Volts

60

80

400 300 200

I = 1008 W/cm2

2 0

600

Pmax = 655 W

Watts

Amps

10

100 0 10 11 12 13 14 15 16 17 18 19 20 21 22 23 0

100

1

GMT

JXC Carousel at UNLV 4/9/2008

2/5/09 Pmax

180

700

160

600

140

500

120 Watts

80 60

400 Watts

Pmax Module 1 Pmax Module 2 Pmax Module 3 Pmax Module 4

100

300 200

40 100

20 0 5

6

7

8

9

10 11 12 13 14 15 16 17 18 19 Time

0 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00 Time

Figure 12.16. 3-Sun module test data. Top/left: illuminated current versus voltage curve for ORNL array taken on March 9, 2008. Bottom/left: Pmax at UNLV on April 9, 2008. Top/right: Pmax at ORNL on April 15, 2008. Bottom/right: Pmax at ORNL on February 5, 2009. IVs = Illuminated current vs Voltage curve; I = Solar Illumination Intensity; GMT = Greenwich Mean Time.

FUTURE POTENTIAL MANUFACTURING AND DEPLOYMENT

287

Figures 12.6 and 12.17a. Based on cell efficiency, these cells are preferred. However, based on cell cost, more traditional silicon cells with front-side grids are less expensive, and there are many more potential cell suppliers. A second cell supplier has fabricated cells as per the front-side metallization design shown in Figure 12.17b. Note that in this design, the grid metallization is screen printed on the front side of the cell, and there are two edge bus bars that will fall underneath the mirror edges. In addition to cell type, there are efficiency options within each cell type. For example, in the work with SunPower A300 cells to date, cells that had higher than normal leakage currents at 1-sun were used. As a starting point, this was acceptable because 3-sun cells operate at higher light-generated currents, and therefore, higher dark level leakage currents are more acceptable. These cells had efficiencies in the 18–19% range at 3-suns. Nevertheless, SunPower has 22% efficient A300 cells that they reserve for their internal use. For the dual-bus, front-side, screen printed cell case, there are also different efficiency options. For example, SANYO has reported 22% efficient cells made with 102 × 102 mm cells. The contacts on these cells could potentially be modified to make 22% efficient dual-bus cells. Table 12.2 summarizes various future 3-sun cell options.

Figure 12.17. (a) A 125-mm A300 cell cut into thirds. (b) A 156-mm multicrystal wafer with eight screen printed patterned 3-sun dual-bus cells. TABLE 12.2. Derivative 3-Sun Cell Types and Dimensions Cell Origin

1-Sun Cell Efficiency (%)

1-Sun Cell Dimensions (mm)

3-Sun Cell Dimensions (mm)

3-Sun Active Dimensions (mm)

22

102 × 102

51 × 102

46 × 102

SunPower

23

126 × 126

42 × 126

42 × 120

China

18

156 × 156

52 × 156

47 × 156

SANYO

288

LOW-CONCENTRATION CRYSTALLINE SILICON SYSTEMS

Step 1: Step 2: Step 3: Pull Ribbons; Repeat step 1 until Solder ribbons front & back; Place Cell, Advance Cell the last cell is placed; Pull ribbons last time for the trailing end

Figure 12.18. The cells in standard planar silicon modules are electrically interconnected using automated cell stringing equipment where two metal ribbons are soldered from the back of one cell to the front of the next cell. Dual-bus 3-sun cells can be interconnected with two metal ribbons with the same automated stringing equipment. TABLE 12.3. 3-Sun Module Cell Configuration and Dimensions Cell Origin

Cell Layout

3-Sun Module Dimensions

STC Power (W)

SANYO

10 rows with eight cells per row

51.2 × 35 in. 1300 × 890 mm

200

SunPower

10 rows with seven cells per row

50 × 37 in. 1270 × 940 mm

210

China source

10 rows with six cells per row

51.2 × 38 in. 1300 × 965 mm

180

The dual-bus cell design shown in Figure 12.17b is very attractive because it can be readily adapted to standard planar module cell stringing equipment as shown in Figure 12.18. This can lead immediately to high-volume, low-cost 3-sun module manufacturing. Using the cells shown in Table 12.2, a large variety of module sizes are possible just as is the case for standard silicon planar modules. However, in addition to lower-cost modules, it is also desirable to manufacture and install lower-cost systems at the array level. In Chapter 9, a prefabricated sun-tracking carousel was described for simple and rapid field installation of arrays on commercial building flat rooftops. The 1.2-kW carousel shown in Figure 9.10 uses six SANYO 200-W modules. It may be desirable simply to begin manufacturing by designing a 3-sun module with similar dimensions to the SANYO modules used on that carousel. Table 12.3 shows various 3-sun module cell configurations leading to 3-sun modules with similar dimensions to the SANYO 200-W planar module. Each of these module designs would lead to the carousel depicted conceptually in Figure 12.19 with a power output of over 1 kW.

FUTURE POTENTIAL MANUFACTURING AND DEPLOYMENT

289

Figure 12.19. Sun-tracking 1.2-kW carousel with 3-sun modules for deployment on commercial building flat rooftops. This is a single-axis AZ tracker with modules mounted at a fixed tilt. It rotates from east to south to west over a day.

Figure 12.20. Conceptual drawing of a one-axis tripod sun tracker with array of 3-sun modules. The beam is aligned approximately with the earth’s axis, and the array rotates from east to west over the daylight hours.

The LCPV modules can also be easily deployed in fields using one-axis tripod sun trackers as depicted conceptually in Figure 12.20. These tripod trackers are similar to those deployed by SunPower Corporation in Multi MW fields as described by M. Campbell in the previous chapter.

290

12.6

LOW-CONCENTRATION CRYSTALLINE SILICON SYSTEMS

FUTURE POTENTIAL COST

The key to the 3-sun concept lies in the fact that sheet metal mirrors are at least 10 times cheaper than single-crystal cells today. As was projected in Table 3.2 in Chapter 3, this should remain true into the future even as the cost of the silicon cells comes down. Chapter 2 discussed cost goals for solar electricity system for the future. From these two inputs, it is possible to relate the cost projections for a solar electric power system based on 3-sun modules to the future cost goals enumerated in Chapter 2. This is done in Table 12.4 as an interesting comparison since one set of numbers is derived from the bottom up from detailed component cost projections, and the other set is derived as system goals from the top down. Table 12.4 suggests that the LCPV system described here provides a straightforward prescription for reaching a cost for solar electricity of less than 10¢ per kilowatt hour in sunny regions of the world as in the southwestern United States.

12.7

CONCLUSIONS

Fifty percent of the cost of a PV system today is in the PV modules, and 70% of the cost of a PV module is in the silicon cells. These cells require very high-purity silicon feedstock along with expensive ingot growth and cell fabrication processes.

TABLE 12.4. Predictions versus Goal Comparison for LCPV LCPV parameter from Table 3.2 Module efficiency

20%

Module cost

$1.20 per watt

Annual irradiance

Medium Term Goal from Table 2.2 20%

2

2382 kWh/m /year

$1.25 per watt 2435 kWh/m2/year

Area BOS

$100 per square meter

$150 per square meter

Inverter cost

$300 per kilowatt

$300 per kilowatt

Fixed charge rate

10%/year

10%/year

Indirect cost rate

22.5%

22.5%

$0.09 per kilowatt hour

$0.09 per kilowatt hour

Additional assumptions from Table 2.2

Conclusion Levelized cost of energy

ABBREVIATIONS

291

The JXC 3-sun CPV module cost is simply reduced by substituting low-cost reflecting aluminum mirrors for two-thirds of the expensive silicon cell area. The 3-sun mirror module is an evolutionary design based on planar silicon cells already in high-volume production. Its novelty is the use of mirrors and sun trackers. While the goal of lower-cost solar electric power can be reached when carousels and 3-sun LCPV modules reach high-volume production, unfortunately, this hardware is not yet in high-volume production. The first step has been to perform life testing and performance testing at various beta sites as has been reported here with promising results. Large-scale, cost-competitive solar electric power is now in sight given the two evolutionary innovations described here. The first innovation is the 3-sun PV module, and the second is a prefabricated one-axis AZ drive carousel sun tracker that can be installed on commercial building flat rooftops or over carports. Commercial customers in sunny locations are the prime users for this technology as they pay retail prices for electricity well over 10¢ per kilowatt hour. In the future, once both carousels and 3-sun CPV modules enter high-volume production, solar electric power costs are projected to fall to as low as 9¢ per kilowatt hour [9]. This should be quite affordable for commercial customers paying retail prices for electricity.

ABBREVIATIONS AC (ac) —alternating current AZ—azimuth BOS—balance of system Caltech—California Institute of Technology CPV—concentrator photovoltaic FF—fill factor HCPV—high-concentration photovoltaic Imax—current at maximum power point Isc—short-circuit current JPL—Jet Propulsion Laboratory JXC—JX Crystals Inc kWp—kilowatt peak (the peak maximum power produced by a solar module in operation in sunlight at midday or at solar noon) LCPV—low-concentration photovoltaic MCPV—mid-concentration photovoltaic MJ—multijunction (solar cell) NREL—National Renewable Energy Laboratory N-S—north–south orientation ORNL—Oak Ridge National Laboratory Pmax—power at maximum power point PTC—practical test conditions (outdoors)

292

LOW-CONCENTRATION CRYSTALLINE SILICON SYSTEMS

PV—photovoltaic STC—standard test conditions (flash test indoors) Temp—temperature UNLV—University of Nevada, Las Vegas Vmax—voltage at maximum power point Voc—open-circuit voltage REFERENCES [1] [2] [3] [4] [5] [6] [7]

[8] [9]

A. Salim and N. Eugenio. Solar Cells 29, 1 (1990). M. O’Neill. Chapter 10. In Solar Cells and Their Applications, L. D. Partain, ed., 1st edition. New York, John Wiley & Sons (1995). F. Goodman, J. Schaefer, and E. DeMeo. Chapter 16. In Solar Cells and Their Applications, L. D. Partain, ed., 1st edition. New York, John Wiley & Sons (1995). D. Silva, J. A. Muñoz, and A. Payán. 2 years of operation of the Sevilla PV 1.2 MW grid connected plant. 23rd European PV Solar Energy Conference, Valencia (2008). M. Klausmeier-Brown. Chapter 5. In Solar Cells and Their Applications, L. D. Partain, ed., 1st edition. New York, John Wiley & Sons (1995). L. Fraas, J. Avery, and H. X. Huang. Test sites and testing of 3-sun mirror modules. Presented at the IEEE 4th World Conference on Photovoltaic Energy Conversion, May 9, Waikoloa, HI (2006). L. Fraas, J. Avery, and H. X. Huang. Start-up of first 100 kW system in Shanghai with 3-sun PV mirror modules. Presented at the 4th International Conference on Solar Concentrators for the Generation of Electricity or Hydrogen (ICSC-4), March 12–16, San Lorenzo del Escorial, Madrid, Spain (2007). L. Fraas, J. Avery, and L. Minkin. Carousel trackers with 1-sun or 3-sun modules for commercial building rooftops. Presented at the Solar 2008 Conference, May 6, San Diego, CA (2008). L. Fraas, et al. Field experience with 3-sun mirror module systems. 33rd IEEE PV Specialist Conference, San Diego, CA (2008).

13 HIGH-CONCENTRATION, III–V MULTIJUNCTION SOLAR CELLS GEOFFREY KINSEY Amonix, Inc.

13.1

III–V MULTIJUNCTIONS IN HCPV SYSTEMS

Under high (>100×) optical concentration, III–V multijunction solar cells convert over 40% of sunlight into electrical power [1, 2]. The crossing of the 40% efficiency threshold in 2006 had more than just psychological value; it signaled that multijunction cells had reached a point at which their performance could more than offset their cost in high-concentration, two-axis tracking systems. Having served for more than a decade as the workhorse in space power applications, multijunctions are now being implemented for terrestrial, utility-scale solar power generation. III–V multijunction cells began displacing silicon single-junction cells in space applications in the 1990s. These structures are composed of compounds of elements in groups III and V of the periodic table. Though the cost of a III–V multijunction cell exceeds that of a silicon single-junction cell by at least an order of magnitude, this component cost is dwarfed by the cost of building and launching satellite payloads into space. In space, the higher performance of the multijunction means better economics of the overall system. For terrestrial applications, a similar situation exists with respect to high-concentration (over 100×) photovoltaic systems. High-concentration photovoltaic (HCPV) systems require relatively expensive optics and precise dual-axis tracking to maintain their focus on the sun throughout the day. The high-power densities (>10 W/cm2) incident on the cell surface require substantial investment in packaging and cooling design. Due to these higher system costs, the economics of the HCPV system favor an expensive, high-performance cell that then delivers higher energy output. Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

293

294

HIGH-CONCENTRATION, III–V MULTIJUNCTION SOLAR CELLS

A notion of the leveraging effect of cell performance on HCPV system economics is shown in Figure 13.1. The dependence of system cost on cell efficiency may be estimated for both fixed, flat-plate and concentrator system costs using reasonable assumptions for the various system component costs [3]. Component costs are taken from Swanson’s estimates for annual production volumes in the 10-kW to 10-MW range. For a 1× fixed, flat-plate system, the cell efficiency limit is assumed to be 30%. (The current record for single-crystal silicon is 25%.) Even assuming the cost of 500× cells is 300 times that for 1× cells, the low price for 1× cells is no match for the expensive, high-efficiency cells in a 500× HCPV system. The limiting case of zero-cost cells in a 500× system is also shown; it illustrates the relative insensitivity of an HCPV system to the component cell cost. Assuming modest production scale, the leveraging effect of high concentration will deliver a lower system cost in an efficiency range available with today’s III–V multijunction cells. In order to maintain high concentration on the multijunction cells, two-axis tracking is required. Though the need for this relatively expensive two-axis tracking is often cited as a disadvantage of the HCPV approach, it is worth reconsidering this viewpoint in light of the growing commitment to large-scale solar energy production. For solar energy to make a meaningful impact on the electrical grid, it needs to evolve into a large-scale, consistent energy resource. Solar panels deployed on fixed mounts or single-axis trackers are limited in this regard; they provide a peak power output that falls off when the panel surface is not directly facing the incident sunlight. By allowing panels to directly face the sun throughout the day and year, two-axis tracking provides the necessary consistency in power

Figure 13.1. 500× HCPV systems are relatively insensitive to cell cost, allowing highefficiency cells to deliver a low cost of power per system.

III–V MULTIJUNCTION DESIGN

295

output and a higher capacity factor. In this context, high-concentration photovoltaics may be viewed as the means to justify the cost of two-axis tracking. It thereby delivers a solar power source that is most steady throughout the day and year. For the vast sunny regions of the globe, this makes it most attractive for large-scale power generation.

13.2

III–V MULTIJUNCTION DESIGN

A schematic of a III–V multijunction cell is shown in Figure 13.2. The present three-junction design is similar to that used in 1× space applications. Each of the three subcell p/n junctions is composed of semiconductor material with a bandgap higher than the one below it. Longer wavelengths (with lower photon energy) pass through the upper subcells to be absorbed below. Efficient partitioning of the solar spectrum is obtained by allowing sunlight to be absorbed in the subcell with a bandgap that best matches the particular photon energy; less of the photon’s energy is lost during absorption and conversion to electron–hole pairs. The layers are grown by MOCVD on epitaxy-ready Ge substrates. Performance of the epitaxially grown layers is sensitive to the quality of the Ge growth surface; a substantial portion of the cost of Ge wafers is a result of the processing required to create this epitaxy-ready growth surface. The initial “nucleation” layers help form the bottom subcell junction in the Ge and establish an effective surface for the growth of subsequent low-defect layers above. A buffer layer may then be

Figure 13.2. Schematic of a III–V multijunction cell. Partitioning of the terrestrial solar spectrum among three subcells results in world-record conversion efficiencies.

296

HIGH-CONCENTRATION, III–V MULTIJUNCTION SOLAR CELLS

grown to contain any dislocations or other growth defects. A tunnel junction between the bottom and middle subcells passes current against the built-in field between them. The middle subcell is composed of Ga(In)As, an alloy of GaAs and a small percentage of indium to match the lattice constant of Ge. Following the second tunnel junction, the top GaInP subcell is grown. Both the top and middle subcells have BSFs to reduce carrier recombination losses at the back of the junction. Each of these subcells is also terminated with a front-side “window” layer. This consists of an alloy with a larger bandgap than the underlying emitter layer to reduce surface recombination losses in the emitter. Growth is completed above the top subcell with a thin “cap” layer of Ga(In)As that protects the underlying (aluminum-bearing) window layer from oxidation and serves as an ohmic contact for the front-side metal. During cell fabrication, the cap layer is etched away except in the vicinity of gridlines and an ARC is deposited on the window layer [4].

13.3

TUNNEL JUNCTIONS

The high efficiency of the III–V multijunction cell is due in part to its monolithic design. Since the layers are composed entirely of semiconductors, light is efficiently transmitted through intervening layers to its intended absorption layer. The monolithic design would not be possible without tunnel junctions. Due to the longer minority carrier lifetime of electrons than holes in most III–V materials, each subcell in the multijunction is doped as a (thin) n-type emitter on a (thick) p-type base. A complication in making a series connection of three such n-on-p diodes is that, at the interface of any two diodes, a p-on-n junction results. The presence of this diode of the opposite polarity would ordinarily serve to block current flow between subcells. To avoid this, a thin, degenerately doped p-on-n tunnel junction is inserted at the subcell interfaces. At relatively low current densities, the current–voltage characteristic of the tunnel junction is essentially ohmic, allowing current to flow between subcells (Fig. 13.3). Tunnel junctions make use of the quantum mechanical phenomenon that the positions of particles are not discrete but, rather, are defined by a probability distribution that falls off gradually on either side of their average positions. Such particles, placed near an energy barrier, would therefore have a nonzero probability of being on either side of the barrier: they can “tunnel” through the barrier. If the particles in question are charge carriers such as electrons or holes, and the barrier is sufficiently narrow for a large tunneling probability, large currents can be made to flow in a direction contrary to that indicated by the local electric field. The tunneling effect is strongest at voltages for which the alignment of the bandgaps at the junction is most favorable. In this region, the current–voltage response is essentially ohmic. As the current is increased, band alignment becomes less favorable and a falloff in current ensues. At higher voltages, the tunnel junction regains the characteristic of a typical p-n diode. A tunnel diode therefore has a peak tunneling current, above which the current will begin to fall off and return to that of the typical diode-like current–voltage

TUNNEL JUNCTIONS

297

Figure 13.3. Current–voltage characteristic of a tunnel junction diode compared to a typical p-n diode.

characteristic. The highest peak tunneling currents are achieved using thin, heavily doped layers of relatively low-bandgap semiconductors (such as GaAs). Tight growth control must be maintained in order to consistently form thin, highly doped tunnel junctions with correspondingly high peak tunneling currents. During the epitaxial growth process, the amount of time at high temperature that follows growth of the tunnel junction must be limited to minimize any out-diffusion of dopants that would then degrade tunnel junction performance. A trade-off also exists between maintaining high peak tunneling currents and high optical transmission through the tunnel junctions. To reduce parasitic optical absorption by the tunnel junction, materials with larger bandgaps and lower doping are favored, but these tend to deliver a lower peak tunneling current. The upper tunnel junction, which must be able to transmit light to the two subcells below it, is most sensitive in this regard. It presents a particular challenge in inverted growth structures (discussed below), where it is grown earlier in the process and therefore becomes subject to more degradation due to extended time at high temperature. HCPV systems can produce sufficient current densities to exceed the peak tunneling current of several typical tunnel junction configurations. When this occurs, the current–voltage characteristic of the tunnel junction, with its distinctive negative resistance region, becomes apparent in the solar cell characteristic and a kink in the knee of the current–voltage characteristic appears. The fill factor will drop accordingly. Such “tunnel junction failure” is reversible; once the current density is lowered, the high fill factor will return. Development of growth techniques to consistently deliver robust tunnel junctions that can handle current densities in the range of 5 A/cm2 and above has been a key factor in the progress of III–V multijunctions for use in high-concentration photovoltaics.

298

13.4

HIGH-CONCENTRATION, III–V MULTIJUNCTION SOLAR CELLS

DESIGN FOR TERRESTRIAL CONCENTRATORS

Several changes to the heritage space III–V multijunction design are made to better optimize for operation under terrestrial conditions. Space operation requires a design focus on radiation hardness. Over the lifetime of the cell, radiation-induced defects are introduced, which increase carrier recombination levels. Layer thickness and doping levels must be set accordingly. Freed from the radiation constraint, the growth layers in terrestrial concentrator cells can be adjusted to provide higher voltage and higher current under the terrestrial spectrum. The cell size, structure, and metallization pattern must be adjusted to cope with current densities that are hundreds of times greater than those encountered in space. Current densities often exceed 5 A/cm2, so the fill factor can be substantially reduced by parasitic resistance losses such as sheet resistance within the semiconductor layers, contact resistance at the metal–semiconductor interface, and metal gridline resistance (see Fig. 13.4). Cell optimization involves balancing these losses against excessive shadowing of the semiconductor active area by the metal gridlines. The size of the cell is a compromise among the competing demands of cell performance, cell yield, system optics, and assembly economics. A larger cell means fewer assemblies per watt generated, but also makes mitigation of optical, thermal, and electrical series resistance losses more challenging. Smaller cells allow for increased packing density on the expensive epitaxially grown wafers. A small cell reduces the active area affected by any one growth defect, thereby reducing the impact of isolated growth defects on overall wafer yield. In balancing these trade-offs, the resulting cell size for 500× cells is typically around 1 cm2 or smaller. The top subcell emitter requires higher doping to reduce semiconductor resistance losses. To first order, the two-dimensional nature of the built-in field profiles within the cell structure suggests that carrier transport proceeds normal to the growth layers until it reaches the top subcell emitter, at which point it curves toward the nearest metal gridline. This results in high lateral current densities in the (thin) top subcell emitter. A more highly doped top subcell emitter will reduce the voltage drop due to the lateral series resistance. If the doping is too high, however, free carrier absorption will reduce the optical transmission of the top layer and will lower the cell current. A top subcell emitter optimized for this trade-

Metal Busbar l eta RM

RContact RSemiconductor

ine

ridl

lG

ta Me

Cap Layer or uct ond c i m Se

Figure 13.4. Schematic of the parasitic resistance losses involved in current collection.

EFFICIENCY UNDER AM1.5

299

off at 500× concentration typically has an effective sheet resistance in the range of hundreds of ohms per square. In order to deliver maximum current under the terrestrial AM1.5 reference spectrum, the current balance between subcells must be readjusted with respect to the design for space (AM0). Absorption by the atmosphere, particularly in the near UV by the ozone layer, reduces the number of photons available to the top subcell. To compensate, the top subcell may be made thicker or with a lower bandgap in order to increase its current generation. When the middle subcell is GaAs or a low-indium Ga(In)As alloy, the bottom, Ge subcell will produce over 30% excess current with respect to the top and middle subcells. As a result, III–V multijunction cells are typically designed to produce equal amounts of current in the top and middle subcells to maximize the limiting current of the three-junction stack. The ARC is often designed so that this current-matched condition results when the cell is mated with a cover glass. 13.5

EFFICIENCY UNDER AM1.5

The cell efficiency measured under a particular reference spectrum is merely the first step in the prediction of the annual energy output of a III–V multijunction in terrestrial operation. However, it provides a convenient point for comparison of different technologies and provides guidance in cell optimization. Under standard test conditions, the cell is evaluated at a temperature of 25°C; illumination intensity is assumed to be 1000 W/m2; and the incident spectrum is the reference AM1.5, low AOD (G173-03 direct). To obtain a prediction of III–V multijunction efficiency, the current generated by three subcell materials over a range of bandgaps may be determined. Assuming each subcell of a given bandgap energy is able to convert available photons above this energy into an electrical current, it is possible to determine the limiting current for a set of bandgap combinations [5]. In the examples that follow, a typical external quantum efficiency of 94% is assumed for the portion of the incident spectrum with photon energy above the bandgap energy of each subcell. Wavelengths corresponding to energies below the bandgap energy are assumed to be transmitted to lower subcells or are lost. For the case where an upper subcell has available excess current with respect to the subcells below, the possibility of “leaking” excess photons from an upper subcell to a lower subcell is incorporated. In practice, this may be achieved either by thinning an upper subcell or via radiative coupling effects [6]. For the example shown in Figure 13.5, the bottom subcell is assumed to be Ge (with a bandgap energy of 0.67 eV). For ranges of top subcell and middle subcell bandgap energies, the resulting limiting current is thereby obtained. The standard diode equation may be solved for the cases where V = 0 and J = 0 to give an expression for the open-circuit voltage (VOC) as a function of shortcircuit current density (JSC): VOC =

nkT ⎛J ⎞ ⋅ ln ⎜ SC + 1⎟ , ⎝ ⎠ q J0

(13.1)

300

HIGH-CONCENTRATION, III–V MULTIJUNCTION SOLAR CELLS

Figure 13.5. Isoefficiency plot for top and middle subcell bandgap ranges on a 0.67-eV bottom subcell at 500×. The bandgap ranges for several low lattice mismatch (<1%) alloys are indicated, as well as record efficiencies (measured below 300×) for lattice-matched (circle) and metamorphic (triangle) cells on germanium [1].

where J0 is the reverse saturation current density; n is the diode ideality factor; k is the Boltzmann constant; q is the charge of an electron; and T is temperature. The increase in the short-circuit current with concentration is essentially linear, so open-circuit voltage sees a logarithmic increase with concentration [7]. Following a procedure similar to that given in Partain [8], the open-circuit voltage at each subcell bandgap energy is determined based on the position of its quasi-Fermi levels derived from the given bandgap energy and adjusted for concentration and carrier lifetimes limited by both radiative and non-radiative recombination. Assuming a linear relationship between the voltage at maximum power and the open-circuit voltage, an estimate for fill factor as a function of bandgap is obtained [9]. Finally, extrinsic losses due to series resistance and gridline obscuration (shadowing) are factored in. Using typical values for state-of-the-art processing at the target concentration, the power losses due to semiconductor sheet resistance, contact resistance, gridline sheet resistance, and gridline obscuration may be obtained (Fig. 13.4). For concentrations between 100× and 500×, the resulting (relative) power loss is in the range of 6–11%. Using these assumptions, the isoefficiency curves in Figure 13.5 indicate a maximum efficiency of about 43% at 500×. Using analytical expressions to calculate bandgap energies as a function of lattice constant [10, 11], the ranges of several alloys with lattice mismatch below 1% are indicated. Though they are by no means definitive of what is possible, significant excursions from these alloy compositions will be challenging at best.

INVERTED METAMORPHIC MULTIJUNCTION CELLS

301

For comparison, recent measured record efficiencies for both lattice-matched and lattice-mismatched (metamorphic) designs are included in Fig. 13.5. Both of these measurements were obtained below 300×, where the lower current levels allow for reduced gridline obscuration and reduced series resistance losses. Applying the modeling approach described above, the lattice-matched structure that attained 40.1% at 135× has a predicted efficiency of 40.2% at 135× and 39.7% at 500×. The metamorphic structure that attained 40.7% at 240× has a predicted efficiency of 41.1% at 240× and 40.6% at 500×. These modeling results therefore suggest that III–V multijunction structures employing a Ge bottom subcell are already within about 1% (absolute) of their maximum efficiency. For further efficiency increases, it will be necessary to look to wafer processing improvements (reduced parasitic losses) as well as more fundamental changes to the epitaxial structure.

13.6

INVERTED METAMORPHIC MULTIJUNCTION CELLS

The record for III–V multijunction cell efficiency was set in 2008 by an inverted metamorphic structure developed by NREL. This cell obtained 40.8% efficiency under the AM1.5 spectrum and 326× concentration [2]. This achievement represents a successful decoupling of III–V multijunction cell efficiency from the bandgap limitations of Ge. Though Ge provides a convenient substrate for growth of several lattice-matched III–V alloys, its bandgap of 0.67 eV delivers a lower than optimum subcell voltage of less than 400 mV at 500×. With a low bottom subcell bandgap, effective partitioning of the solar spectrum is best obtained by choosing relatively low top and middle subcell bandgaps as well. This lowers the overall cell voltage and, for the case of Ge, leads to a peak efficiency region in the lower left quadrant of Figure 13.5, beyond the reach of available lattice-matched materials. Lattice-mismatched or “metamorphic” materials are an option that, within limits, decouples bandgap energy from the constraint of lattice matching. Using graded buffer layers in the growth sequence, it is possible to grow monolithic structures with layers at different lattice constants. The buffer layers serve to transition from one lattice constant to another. Dislocations in the lattice due to the resulting strain relaxation must be contained within these transition layers; if the dislocations propagate into active cell layers, non-radiative recombination will increase dramatically. The resulting decease in cell voltage will more than offset the gains from the higher bandgap of the bottom subcell. The most straightforward approach to a metamorphic multijunction is to begin grading the lattice constant at the substrate surface. This was the method used to obtain the first 40% cell [1]. A buffer region is used to transition from the lattice constant of Ge to a larger lattice constant for the active layers above. Careful control of the growth conditions permits relaxation of the grown layers in the buffer to the new lattice constant with minimal propagation of growth defects to subsequent active subcell layers. A limitation of this approach is that all three subcells

302

HIGH-CONCENTRATION, III–V MULTIJUNCTION SOLAR CELLS

are grown subsequent to the transition in lattice constant, so all three are subject to any degradation in material quality that may result from dislocations propagating through the grown layers. As such, the degree of lattice mismatch that may be obtained is limited. In the case of the 40.7% metamorphic structure shown in [1], the lattice mismatch was 0.5%. The corresponding shift in bandgaps is not sufficient to allow for a new bottom subcell to replace the Ge. In order to attain higher levels of lattice mismatch while maintaining material quality, the NREL team successfully demonstrated an inverted metamorphic structure. By growing the structure inverted, more aggressive lattice mismatch could be pursued without compromising the growth quality of the top subcell. Buffer regions in the next two subcells allowed for increasing levels of lattice mismatch. The layers with the highest mismatch were grown last, limiting the propagation of dislocation defects. This allowed for the growth of a new bottom subcell, composed of Ga0.63In0.37As, with a lattice mismatch of 2.7% with respect to the GaAs growth substrate. The benefit of this approach is illustrated in Figure 13.6. The higher bandgap of the bottom subcell leads to a new “sweet spot” in the isoefficiency curves that is attainable by the available growth materials; the bandgap combination chosen by NREL is close to optimal. The predicted efficiency for this bandgap combina-

Figure 13.6. Isoefficiency plot assuming a bottom subcell bandgap of 0.89 eV. The recent inverted metamorphic cell record (measured at 326× [2]) is indicated by the square. The “folded” isoefficiency lines on the left indicate a region of delicate current balancing among all three subcells.

INVERTED METAMORPHIC MULTIJUNCTION CELLS

303

tion is 43.5% at 326× and 43.1% at 500×. The difference between predicted and actual efficiencies is larger in this case due both to the high level of lattice mismatch and to the difficult nature of the processing required. A schematic of the process sequence for inverted metamorphic structures is shown in Figure 13.7. In order to get the inverted structure oriented sunward, the epitaxial layers must be transferred to a “handle” substrate that provides mechanical support once the growth substrate is removed. This is a delicate process as the handle substrate must make intimate, void-free thermal and electrical contact to the thin epitaxial growth surface. Growth substrates are currently removed destructively via wet chemical etching. There are several methods of substrate removal in development that may allow for recovery of the growth substrate for economical reuse. This possibility would be more significant for GaAs substrates than for Ge, as much of the Ge epitaxial wafer cost lies in surface preparation rather than in the bulk material. The layer transfer made necessary by the inverted approach adds processing cost and will likely involve significant penalties in cell yield. The long-term reliability of metamorphic cells processed in this fashion has yet to be evaluated. The inverted metamorphic cell has pushed the limits of possibility for epitaxial growth and optimization of multijunction bandgaps for maximum efficiency. It may take considerable time, however, for this approach to find its way into commercial HCPV systems.

e trat

ubs

p-base n-emitter tunnel junction p-base n-emitter tunnel junction p-base n-emitter n+cap Growth substrate

m tto ell bo b-c su e dl l id el m b-c su p l to cel bu s

Handle substrate p-base n-emitter tunnel junction p-base n-emitter tunnel junction p-base n-emitter n+cap Growth substrate

m tto ell bo b-c su e dl l id el m b-c su p l to cel bu s

s wth Gro

n+cap n-emitter p-base tunnel junction n-emitter p-base tunnel junction n-emitter p-base Handle substrate

Figure 13.7. Inverted multijunction wafer process.

p l to cel bu s e dl ll id m b-ce su m tto ell bo b-c su

304

13.7

HIGH-CONCENTRATION, III–V MULTIJUNCTION SOLAR CELLS

PERFORMANCE IN OPERATING CONDITIONS

The operating conditions of a multijunction cell in a field-deployed HCPV system are more varied and more demanding than the standard test conditions given above. Despite their best efforts, system designers cannot maintain a cell temperature of 25°C when the ambient temperature is often over 30°C and roughly 60% of the energy reaching the cell is being converted to heat. The spectrum and intensity of light transmitted to the cell are altered by the concentrating optics and vary with time of day, day of the year, latitude, altitude, and atmospheric conditions. An understanding of how cell performance interacts with these parameters will allow for better design of III–V multijunction cells for maximizing energy generation in fielded HCPV systems.

13.7.1

Effects of Temperature and Concentration

As cell temperature is increased, the bandgaps of the subcells shrink, causing a decrease in the open-circuit voltage and fill factor, but an increase in the shortcircuit current. The decrease in voltage is the dominant effect, so a net decrease in cell efficiency results. Fortunately, the sensitivity of cell voltage to temperature decreases with concentration. The diode equation (Eq. 13.1) indicates a logarithmic increase in open-circuit voltage with concentration. The reverse saturation current density in Equation 13.1 is a function of the temperature and semiconductor bandgap (which is itself temperature dependent). Substitution of expressions for these temperature dependencies and differentiation with respect to temperature yield an expression for the VOC temperature coefficient [7, 12]: ∂VOC 1 ⎛n nkT ⎛ = − ⋅ ⎜ ⋅ Eg − VOC + ⋅⎜3 + ∂T T ⎝q q ⎝

γ ⎞ ⎞ nkT 1 ∂J SC n ∂Eg + ⋅ . ⋅ ⋅ ⎟ + 2 ⎠ ⎟⎠ q ∂T q J SC ∂T

(13.2)

The γ term is material dependent and is not well understood for the various alloy compositions involved. However, if treated as a fitting factor, good agreement is obtained with respect to measured values of the temperature coefficient (Fig. 13.8). At 500×, the multijunction VOC temperature coefficient has decreased approximately 30% with respect to its 1× value. The change in open-circuit voltage is about −1.4%/10°C at 500×; that of short-circuit current is about 0.6%/10°C [7, 13, 14]. With the effect of fill factor included, the resulting change in efficiency is about −1% (relative) per 10°C (−0.4% absolute per 10°C). The decrease in temperature sensitivity at higher concentrations provides some relief to the designer attempting to reach the economic benefits offered by HCPV systems operating at concentrations above 500×.

PERFORMANCE IN OPERATING CONDITIONS

305

Figure 13.8. The voltage temperature coefficient decreases with concentration.

13.7.2

Spectrum

Effective partitioning of the solar spectrum enables the III–V multijunction to realize the high efficiencies promised by its stacked junctions. However, there is a penalty to be paid in using spectrum partitioning to boost cell efficiency: an increase in sensitivity to variations in the spectral content of the incident sunlight. Since a monolithic multijunction cell is composed of a series connection of subcells, the overall cell current output will be limited by whichever subcell produces the lowest current. Under standard test conditions, subcell layers may be optimized to produce maximum current under a single spectral condition. For the case of the current three-junction cell under the reference AM1.5 (G173-03, direct) spectrum, there is over 30% excess current generated in the bottom subcell. Optimization is therefore achieved by equalizing the current densities in the top two subcells to obtain a “current match.” Variations in epitaxial growth and the ARC result in deviations from the intended balance of subcell current densities. The ratio of the top to middle subcell current densities (Jtop/Jmiddle) is therefore a figure of merit in tuning a given structure to a target spectrum. Though a value of Jtop/Jmiddle equal to one is desired, in practice, variations in growth and wafer process conditions lead to variations in the ratio of several percent. Proper control of growth conditions and wafer processing, particularly in the design of ARCs, can be used to adjust the ratio. If the top subcell is thin, photons within its absorption range can pass through it to generate current in the middle subcell. Radiative coupling between subcells may also occur: carriers generated in the top subcell often recombine before they are collected, emitting photons that may be absorbed by the middle subcell [6]. Complementary

306

HIGH-CONCENTRATION, III–V MULTIJUNCTION SOLAR CELLS

mechanisms for shifting current from the middle to the top subcell are not readily available, so it is generally easier to lower the current density ratio than it is to raise it. One method to increase Jtop/Jmiddle is to redesign the ARC to favor the top subcell. The loss in cell current that arises from subcell current density mismatch is partially offset by a rise in fill factor [15]. When two subcells have mismatched current outputs, the load points of the subcells move away from their individual maximum power points to points that produce equal currents. The subcell producing excess current moves to a lower-current/higher-voltage position, and the subcell producing insufficient current moves to a lower-voltage/higher-current position. The combined current–voltage characteristic therefore has a steeper “knee” in the curve and a larger fill factor. Nevertheless, the highest efficiency is obtained when the subcells have matched current densities.

13.7.3

Annual Energy Output

The high efficiencies obtained under standard test conditions are largely academic unless they can be translated into high energy output of HCPV systems in the field. In order to deliver a high annual energy output from an HCPV system, the III–V multijunction cell must be able to maintain its power output at elevated temperature, under concentrating optics, throughout the day and year. The incident spectrum is altered by diurnal and seasonal changes in air mass and atmospheric conditions such as humidity and air optical depth (turbidity). It is altered further by the transmission profile of the concentrating optics. Understanding of the effects of these changes on cell performance can inform both future cell and HCPV optical design improvements. The challenge of predicting annual cell output under varying conditions is fairly daunting, but some insight may be obtained by convolving the measured spectral response of III–V multijunction cells with predicted spectral irradiance [13]. The NREL SMARTS model [16] uses analytical expressions to generate spectral irradiance output for a given geographic location based on latitude, elevation, date, time, turbidity, and other atmospheric conditions. The application of this model to the III–V multijunction cell response provides an indication of field performance that may then be verified by direct measurement at a given site. As an example, the SMARTS model was used to generate annual spectral irradiance data for Phoenix, Arizona, assuming cloudless skies and a constant turbidity value of 0.084 (the value used for the G173-03 direct spectrum). The resulting DNI is shown in Figure 13.9. The spectral response of a III–V multijunction cell under a cover glass was used to calculate the resulting current output for each subcell. This cell had an efficiency of 37.5% under standard test conditions. For the annual output, a constant cell temperature of 65°C was assumed. The cell voltage and fill factor were adjusted for temperature, intensity variation, and resistive power losses to determine the annual cell efficiency (Fig. 13.10). Despite the regular form of the annual DNI, the multijunction efficiency contour is quite

PERFORMANCE IN OPERATING CONDITIONS

307

Figure 13.9. Predicted direct normal irradiance (DNI) assuming clear skies and an aerosol optical depth (AOD) of 0.084 in Phoenix, Arizona.

Figure 13.10. Predicted annual efficiency at 65°C of a III–V multijunction cell having an efficiency of 37.5% under standard test conditions in Phoenix, Arizona.

irregular. The highest cell efficiencies are obtained at low temperature and when the top and middle subcell current densities are most closely matched, rather than in periods of peak DNI. For example, at 13:00 on June 3, the DNI was at its annual high of 967 W/m2. The current density ratio was 1.07, so 7% excess current carriers were generated in the top subcell. Due to the current mismatch, these carriers would recombine without contributing to the overall cell current. As a result, the cell efficiency at this point was only 34.1% (16.5 W/cm2). In contrast, at 15:00 on October 3, the incident spectrum delivered an optimal current density ratio of 1.00. Despite a lower DNI of 880 W/m2, the resulting efficiency was 35.3% (15.5 W/cm2).

308

HIGH-CONCENTRATION, III–V MULTIJUNCTION SOLAR CELLS

By weighting each value by the energy available, a weighted annual current density ratio of 0.95 and an annual weighted cell efficiency of 32.8% are obtained. This suggests that higher annual cell energy output could be obtained if the cell design were modified to allow more current to be generated in the top subcell. This suggests moving away from design for AM1.5 to designs for a higher effective air mass.

13.8

RELIABILITY

To date, field experience with HCPV systems employing III–V multijunction cells is limited, so concerns remain as to whether such systems can maintain their unparalleled performance over a 25-year lifetime. Though incomplete, the heritage data for III–V multijunctions bodes well for long-term operation in HCPV systems. III–V multijunctions have proven their hardiness in the rigors of space and on the plains of Mars. Preparation for these missions involved extensive testing. Thermal cycling has been used to demonstrate tolerance to the >100°C temperature excursions that occur in satellite orbit. Due to the often prolonged periods of storage prior to a launch, III–V multijunctions have passed space qualifications that included environmental testing for terrestrial conditions as well. Comparison with other terrestrial electronic applications also provides some insight. The compound semiconductors used in III–V semiconductors are widely deployed in military electronics, wireless handsets, and in long-haul (including undersea) fiber optic communications. High-brightness LEDs provide an instance where the operating current density exceeds that required in high-concentration photovoltaics [17]. Successful qualification testing and field performance in these industries give some confidence that sufficient reliability can be demonstrated in HCPV applications as well. If failure modes do exist, they are most likely to arise in the areas that are unique to multijunctions under high concentration. The lifetime of III–V multijunctions in the field will ultimately be determined by how they are integrated into specific HCPV systems. For example, unlike III–V multijunctions in space, cells for high-concentration photovoltaics must typically be mounted using solder to provide adequate thermal and electrical conductivity for their high current densities. Care must be taken to avoid the formation of voids in the solder under the cell. A solder void can trigger a thermal runaway condition in III–V multijunctions that will rapidly induce cell failure. A hot spot in the vicinity of a void, caused by nonuniformities in the concentrating optics, for example, leads to a reduction of the bandgaps of the subcells. This causes them to generate more current, which, in turn, makes the region still hotter; thermal runaway ensues. Though such failures will usually present themselves within a few minutes of on-sun operation, aggregation of small voids or stress cracking of the solder over time could bring about long-term degradation in this mode.

SUMMARY

309

One III–V compound known to oxidize readily is AlGaAs. Effective packaging allows for its use in various optoelectronic applications, but it has been substantially eliminated from most III–V multijunction cell structures. The top “window” layer of the cell is the one most exposed to potential environmental degradation. Typically composed of AlInP, it is less susceptible to oxidation than AlGaAs, but should still be encapsulated to retard long-term degradation. The use of ARC and a cover glass helps protect both the window layer and the front-side metallization from oxidation. The low bandgap energy of the Ge subcell makes it more susceptible to reverse bias breakdown. If strings of multijunction cells are connected in parallel strings, blocking diodes must be used so that strings with a higher voltage do not induce a reverse breakdown condition in an adjacent string. The economics of lowcost assembly processes must be balanced against such potential reliability issues. Rate of failure is often accelerated at higher temperatures. Since multijunctions deliver their highest efficiencies when they are cool, HCPV systems are designed so that the cell operating temperature remains relatively low, typically below 70°C. This confers on the system an added margin in terms of reliability. Within this temperature limit, long-term degradation mechanisms, such as lateral oxidation from the edges of the cells, can be substantially mitigated. This notwithstanding, considerable effort is still needed in order to fully vet III–V multijunctions for 25-year performance in HCPV operating conditions. The HCPV industry is proceeding to retire reliability risk by adapting the old aerospace maxim: test like you fly—fly like you test. In 2007, Spectrolab completed a celllevel qualification of multijunctions adapted from the IEEE 1513 standard for CPV receivers and modules. Ongoing qualifications draw on the more recent and more extensive IEC 62108 standard for modules. In parallel, HALT is being conducted by various HCPV groups to drive III–V multijunction packages to failure and thereby identifies potential failure modes in operation. Field testing, though slower, is the ultimate test. Multijunction HCPV systems have demonstrated 2-year operation without appreciable degradation [18]. As further test cycles are completed and more systems are deployed in the field, the necessary conditions to ensure a 25-year lifetime will be established.

13.9

SUMMARY

The 40% conversion efficiency of III–V multijunction cells under high concentration makes them an economical choice for use in HCPV systems operating at 500× and above. In order to realize this promise on a large scale, the design must be tailored to accommodate the cell’s response under operating conditions. As more experience is gained in the field, the virtuous cycle of deployment and further development will permit tuning of the design for long-term energy generation. Building on their legacy in space applications, III–V multijunction cells in HCPV systems will provide the world with large-scale, long-lasting, high-performance solar energy systems.

310

HIGH-CONCENTRATION, III–V MULTIJUNCTION SOLAR CELLS

ABBREVIATIONS AlGaAs—aluminum gallium arsenide AlInP—aluminum indium phosphide AM—air mass AOD—aerosol optical depth ARC—anti-reflection coating BSF—back surface field CPV—concentrator photovoltaics DNI—direct normal irradiance Eg—semiconductor bandgap energy GaAs—gallium arsenide Ga(In)As—gallium arsenide, with trace amounts of indium GaInP—gallium indium phosphide Ge—germanium HALT—highly accelerated life testing HCPV—high-concentration photovoltaics IEC—International Electrotechnical Commission IEEE—Institute of Electrical and Electronic Engineers III–V—alloys using elements from columns III and V of the periodic table J—current density J0—reverse saturation current density Jmiddle—middle subcell current density (in a multijunction cell) Jsc—short-circuit current density Jtop—top subcell current density (in a multijunction cell) k—Boltzmann’s constant LED—light-emitting diode MOCVD—metal-organic chemical vapor deposition n—negatively doped semiconductor n—diode ideality factor NREL—National Renewable Energy Laboratory p—positively doped semiconductor q–charge of an electron SMARTS—Simple Model of the Atmospheric Radiative Transfer of Sunshine (http://www.nrel.gov/rredc/smarts/) T—absolute temperature UV—ultraviolet Voc—open-circuit voltage x—sunlight concentration ratio (direct sunlight is “1x”) γ—temperature coefficient for the ratio of semiconductor diffusion coefficients to minority carrier lifetimes

REFERENCES

311

REFERENCES [1] [2]

[3] [4] [5] [6]

[7] [8] [9] [10] [11]

[12]

[13] [14]

[15]

R. R. King, D. C. Law, K. M. Edmondson, C. M. Fetzer, G. S. Kinsey, H. Yoon, R. A. Sherif, and N. H. Karam. 40% efficient metamorphic GaInP/GaInAs/Ge multijunction solar cells. Appl. Phys. Lett. 90, 183516 (2007). J. F. Geisz, D. J. Friedman, J. S. Ward, A. Duda, W. J. Olavarria, T. E. Moriarty, J. T. Kiehl, M. J. Romero, A. G. Norman, and K. M. Jones. 40.8% efficient inverted triple-junction solar cell with two independent metamorphic junctions, Appl. Phys. Lett. 93, 123505 (2008). R. M. Swanson. The promise of concentrators. Prog. Photovolt. Res. Appl. 8, 93–111 (2000). D. Danzilio. Overview of EMCORE’s multi-junction solar cell technology and high volume manufacturing capabilities. Available at http://www.csmantech.org/ Digests/2007/2007%20Papers/01c.pdf (2007). Accessed April 3, 2009. M. W. Wanlass, K. A. Emery, T. A. Gessert, G. S. Horner, C. D. Osterwald, and T. J. Coutts. Practical considerations in tandem cell modeling. Solar Cells 27, 191–204 (1989). H. Yoon, R. R. King, G. S. Kinsey, S. Kurtz, and D. D. Krut. Radiative coupling effects in GaInP/GaAs/Ge multijunction solar cells. In Proceedings of the 3rd World Conference on Photovoltaic Energy Conversion, Vol. 1, Osaka, Japan, May 11–18, pp. 745–748 (2003). G. S. Kinsey, P. Hebert, K. E. Barbour, D. D. Krut, H. L. Cotal, and R. A. Sherif. Concentrator multijunction solar cell characteristics under variable intensity and temperature. Prog. Photovoltaics Res. Appl. 16, 503–508 (2008). L. D. Partain. Solar cell fundamentals. In Solar Cells and Their Applications, 1st edition, L. D. Partain, ed., pp. 22–33. New York, Wiley (1995). W. Shockley and H. J. Queisser. Detailed balance limit of efficiency of p-n junction solar cells. J. Appl. Phys. 32, 510–519 (1961). Ioffe Physico-Technical Institute. Available at http://www.ioffe.rssi.ru/SVA/NSM/ Semicond/. Accessed January 21, 2009. C. H. Chen, S. A. Stockman, M. J. Peanasky, and C. P. Kuo. OMVPE growth of AlGaInP for high-efficiency visible light-emitting diodes. In High Brightness Light Emitting Diodes, Semiconductors and Semimetals, Vol. 48, G. B. Stringfellow and M. G. Craford, eds., pp. 97–144. San Diego, CA, Academic (1997). K. Nishioka, T. Takamoto, T. Agui, M. Kaneiwa, Y. Uraoka, and T. Fuyuki. Annual output estimation of concentrator photovoltaic systems using high-efficieny InGaP/ InGaAs/Ge triple-junction solar cells based on experimental solar cell’s characteristics and field-test meteorological data. Sol. Energy Mater. Sol. Cells 90, 57–67 (2006). G. S. Kinsey and K. M. Edmondson. Spectral response and energy output of concentrator multijunction solar cells. Prog. Photovoltaics Res. Appl. 17, 279–299 (2009) (in press). D. Aiken, M. Stan, C. Murray, P. Sharps, J. Hills, and B. Clevenger. Temperature dependent spectral response measurements for III-V multi-junction solar cells. In Proceedings of the 29th IEEE PV Specialists Conference, New Orleans, LA, May 21–24, pp. 828–831 (2002). W. E. McMahon, K. E. Emery, D. J. Friedman, L. Ottoson, M. S. Young, J. S. Ward, C. M. Kramer, A. Duda, and S. Kurtz. Fill factor as a probe of current-matching for GaInP2/GaAs tandem cells in a concentrator system during outdoor operation. Prog. Photovoltaics Res. Appl. 16, 213–224 (2007).

312

HIGH-CONCENTRATION, III–V MULTIJUNCTION SOLAR CELLS

[16]

C. Gueymard. SMARTS, a simple model of the atmospheric radiative transfer of sunshine: Algorithms and performance assessment. Technical Report No. FSEC-PF270-95. Cocoa, FL: Florida Solar Energy Center (2005). M. Vazquez, C. Algora, I. Rey-Stolle, J. R. González. III-V concentrator solar cell reliability prediction based on quantitative LED reliability data. Prog. Photovoltaics Res. Appl. 15, 477–491. P. J. Verlinden, A. Lewandowski, H. Kendall, S. Carter, K. Cheah, I. Varfolomeev, D. Watts, M. Volk, I. Thomas, P. Wakeman, A. Neumann, P. Gizinski, D. Modra, D. Turner, and J. B. Lasich. Update on the two-year performance of 120 kWp concentrator systems using multi-junction III-V solar cells and parabolic reflective optics. Proceedings of the 33rd Photovoltaic Specialists Conference, San Diego, CA, May 12–16 (2008).

[17] [18]

14 HIGH-CONCENTRATION FRESNEL LENS ASSEMBLIES AND SYSTEMS GERHARD PEHARZ AND ANDREAS BETT Fraunhofer Institut für Solare Energiesysteme (ISE)

14.1

INTRODUCTION

One of the most important benchmarks for energy technologies is the costs per generated kilowatt hour. Primarily, the levelized costs of electricity produced by PV technologies are dominated by the area-related costs of the solar cells, the conversion efficiency, and the lifetime. In the recent decades of PV development, it has been a challenge to improve all three of these cost factors simultaneously. Nevertheless, the levelized costs of PV electricity have decreased; however, the goal of grid parity has not yet been achieved. Grid parity means that the price per PV generated kilowatt hour equals the price per kilowatt hour paid in a household. Please note that grid parity is very country specific. A concept that could accelerate the cost reduction of PV electricity is the use of PV concentrators. The basic idea of concentrator PV is to use an optical element that focuses sunlight onto a relatively small solar cell. As a result, the solar cell-related costs in a PV concentrator system are of minor impact, which allows highly efficient cells, which are usually comparatively expensive, to be used. Today, the most efficient solar cells are III–V-based triple-junction cells. Under concentrated light, they achieve efficiencies of over 40%. However, their area-related costs are close to two magnitudes higher than those of bulk silicon solar cells, which clearly dominate the solar cell market today. For terrestrial applications, the highly efficient triple-junction solar cells are ideally used in combination with cheap optics, which collect the sunlight over an area that is several hundred times larger than the solar cell area and focus it on the solar cell. One of the cheapest known optical

Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

313

314

HIGH-CONCENTRATION FRESNEL LENS ASSEMBLIES AND SYSTEMS

elements for concentrating sunlight is a Fresnel lens. Hence, Fresnel lenses are often used for concentrator applications. Fresnel lenses have been used since the beginning of PV concentrator development. In this field, the first extensive experience was made by Sandia National Laboratories. Some of the key issues for PV concentrator design are addressed in an excellent work of Burgess et al. [1]. In contrast to non-concentrating PV systems, the solar cells in a concentrator have to operate under a much higher flux density. Consequently, it is of great importance that the solar cell and its related electrical assemblies have a low series resistance. Furthermore, the generated heat flux must be removed effectively; otherwise, the solar cell or assemblies can suffer damage. Thus, an effective cooling of the solar cells is required. This can be carried out passively by using air convection or actively by conducting a cooling fluid through a heat exchanger. From a designer’s point of view, passive cooling seems to be more desirable because it is less complex than active cooling. However, passive cooling requires relatively large areas to dissipate the heat. In a Fresnel lens concentrator, the solar cell is located beneath the Fresnel lens and thus at least, an area equivalent to the lens aperture is available for heat dissipation. This option is not possible for parabolic mirror concentrators where the solar cell is located in front of the optics and causes shading. One of the fist concentrator installations at Sandia National Laboratories used acrylic Fresnel lenses with an area of 912 cm2 and silicon solar cells with an area of 15.2 cm2 [1]. The ratio of lens area and cell area defines the geometric concentration ratio, which was 60X for this system. A two-axis tracking system was used to align the arrays to the sun and to keep the focal spots on the cells. The Fresnel lens concentrator concept of Sandia National Laboratories was adopted and further developed by other research institutes in the late 1970s. At the UPM, a PV concentrator using hybrid glass silicone–glass Fresnel lenses [2, 3] was installed. A photograph of this early Fresnel lens concentrator is shown in Figure 14.1. Basically, these silicone–glass Fresnel lenses were thought to be more reliable than acrylic lenses in terms of abrasion and UV degradation. Furthermore, the costs were expected to be lower than those of acrylic Fresnel lenses. Similar to the Sandia system, the Ramón Areces concentrator used silicon solar cells. The early concentrators were equipped with solar cells having a size of about 15 cm2. Considering a concentration ratio of 50X, the cells receive a light power of about 25 W, of which less than 10% was converted into electrical energy and the rest was mostly converted into heat. This heat needed to be dissipated, and early Fresnel lens concentrators used massive metal elements of several millimeters thickness. Furthermore, special attention must be drawn to thermal matching, when using solar cells with an area of several square centimeters. Differences in the thermal expansion coefficients must be taken into account when designing the interconnection of cell and heat sink. Consequently, thermal issues were of great importance for the early concentrators due to the relatively large cell area. In the 1990s, a Fresnel lens concentrator was also developed by Fraunhofer ISE, Germany, in cooperation with the Ioffe Institute in Russia [4–6]. In contrast

INTRODUCTION

315

Figure 14.1. The Ramón Areces installation at the Universidad Politécnica de Madrid (UPM). This PV concentrator used silicone–glass Fresnel lenses, silicon solar cells, and a two-axis tracking system. The cooling was managed passively by using massive metallic elements. The person in front of the concentrator is Prof. Gabriel Sala of the UPM. Courtesy of Prof. Gabriel Sala, UPM.

to the early Fresnel lens concentrators, the dimensions were downscaled by about a factor of 100. The solar cell area was reduced to several square millimeters and was made of III–V materials. This relaxes the requirements on the passive cooling. In particular, a copper heat sink with a thickness of several 100 μm is sufficient to keep the cell cool, even when operated under a concentration ratio of up to 1000X. Already in the year 1976, James et al. demonstrated that for these high concentration ratios, solar cells made from III–V semiconductors (in particular, AlGaAs/ GaAs heteroface single junction) work very well [7]. Furthermore, work on highly efficient III–V multijunction cells has been going on since the late 1980s [8, 9]. In 2001, Fraunhofer ISE introduced monolithic multijunction solar cells in concentrator modules. A concentrator module consisting of 40 × 40 mm2 Fresnel lenses and 13 mm2 metamorphic Ga0.35In0.65P/Ga0.83In0.17As dual-junction cells reached an efficiency of 24.8% measured under natural sunlight without temperature correction [6]. At this time, it was the highest reported efficiency of a Fresnel lens concentrator module. Motivated today by the fast increase of III–V multijunction efficiencies and therefore the potential to lower the cost per kilowatt hour, many start-up companies have sprung up in the field of PV concentrators. The main application for these

316

HIGH-CONCENTRATION FRESNEL LENS ASSEMBLIES AND SYSTEMS

technologies is the production of electricity in large-sized power plants (several megawatt power scales) in sun-rich countries with a high share of direct sunlight. However, some companies are also aiming for rooftop applications. The following sections describe the most important components of a Fresnel lens concentrator and present some recent examples.

14.2

ELEMENTS OF FRESNEL LENS CONCENTRATORS

This section deals with the various elements found in every high-concentration Fresnel lens concentrator, for the most part. The Fresnel lens optics and the triplejunction cells are described in different subsections. Beside some basics, an overview is given about the manufacture as well as recent developments in the field. Please note that in particular, Fresnel lens systems with a two-dimensional concentration are considered here.

14.2.1

Fresnel Lens Optics

Basically, a Fresnel lens has optical properties that are similar to conventional lenses; however, they have lower material usages and therefore lower cost. This is due to the fact that a Fresnel lens is obtained by breaking a conventional lens into a set of concentric segments. Figure 14.2 shows a sketch of a conventional and planar Fresnel lens with the same size and focal distance. As a result of the faceted structure, a Fresnel lens offers an additional dimension in design space. The angle of each facet can be varied in order to decrease spherical aberration or the homogeneity of the light distribution in the focal spot [10]. In PV concentrators, Fresnel lenses with a flat surface are often used, similar to the one shown in Figure 14.2. They can be produced in a molding or stamping process, using acrylic or silicone materials. The relatively low manufacturing and material costs makes Fresnel lenses one of the cheapest available concentrating optics. This is the reason why they have been used from the beginning of PV

(A)

(B)

Figure 14.2. (A) A sketch of the cross section of a conventional lens and above a corresponding Fresnel lens with the same size and focal length. (B) The concentric segments of the Fresnel lens are sketched in a plan view.

ELEMENTS OF FRESNEL LENS CONCENTRATORS

317

concentrator development and are the most widely used optical system in CPV today. In comparison to a massive lens, a Fresnel lens has to pay some additional efficiency losses. A Fresnel lens consists of many prismatic facets that refract the light onto the focal spot. Ideally, each facet has a sharp peak and valley, and the groove wall is orientated parallel to the sunrays. However, the real structure of a Fresnel lens shows deviations, for example, due to the manufacturing process, which lead to losses. These losses are sketched in Figure 14.3A. 1.

2.

3.

(A)

The light impinging on rounded facet peaks or valleys is not concentrated and lost. In order to minimize these losses, accurate manufacturing technique and large facet spacing relative to the total size of the lens are mandatory. Beside the active side of a facet that refracts direct light onto the solar cell, every facet has a groove wall. In practice, this wall must be slightly tilted in order to provide a reliable de-molding during the manufacturing. This tilting causes some rays to be misguided. For simplicity and manufacturing reasons, the active side of a facet does not have the most ideal bent shape but is approximated linearly. The resulting losses are negligible when the facet spacing is small compared to the total size of the lens.

(B)

Figure 14.3. (A) Typical loss mechanisms of a flat Fresnel lens are sketched. In practice, the active side of each facet refracts incoming light rays toward the focus. Ideally, the shape of a facet is hyperbolic; however, in practice, it is often approximated linearly. Typically, light rays are misguided at peaks and valleys of the facets, which cause a decrease of the optical efficiency. Furthermore, the groove wall of each facet is typically slightly tilted in order to provide a reliable de-molding during the manufacturing process. This also has a negative impact on the optical efficiency of the Fresnel lens. (B) A dome Fresnel lens is sketched. As a result of the bowed surface structure, the light rays are refracted at two surfaces. Thus, direct light rays can be refracted in a relative steep angle at comparable low reflection and aberration losses. Furthermore, when direct light is impinging perpendicularly on the dome Fresnel lens, the grooved side of the facets and the peaks does not impact the light rays. Consequently, these parts of the domed Fresnel lens do not decrease the optical efficiency.

318

HIGH-CONCENTRATION FRESNEL LENS ASSEMBLIES AND SYSTEMS

However, when closed facet spacing is used, diffraction effects can decrease the optical efficiency of the Fresnel lens. These losses can become significant at facet diameters around 100 μm. In addition to the loss mechanisms specific to the Fresnel lens, the typical optical losses for lenses like aberration, absorption, scattering at surface imperfections, and surface reflection must be considered. Reflection losses at surfaces can be reduced with appropriate antireflection coatings; however, these coatings increase the cost of the Fresnel lens. For imaging optical elements, the maximal achievable concentration is determined by the ratio of focal length and diameter of the optics; this follows from the étendue. A flat Fresnel lens (as shown in Fig. 14.2) designed for high concentrations must refract light rays impinging the corner of the lens in a steep angle. This increases the chromatic aberration of the lens, and in extreme cases rays from the lens corners are lost due to total internal reflection. Thus, the maximal concentration of a Fresnel lens is limited by the chromatic aberration and reflection losses. The exact limit depends on a number of parameters (e.g., the refractive index). Typically, the maximum concentration limit is quoted to be about 500X for flat Fresnel lenses [11]. For geometric concentrations <500X, the optical efficiency of a Fresnel lens usually ranges between 70% and 90%, depending on the quality of the manufacturing process. An additional design parameter for a Fresnel lens is the surface curvature. In Fresnel dome lenses, the light rays are refracted at two surfaces (see Fig. 14.3B). Consequently, the required angles of refraction are smaller for a Fresnel dome lens than for a flat Fresnel lens with the same size and focal length. Using a Fresnel dome lens, short focal lengths and comparatively low chromatic aberration can be realized. This, in turn, allows higher concentration ratios to be achieved as compared with flat Fresnel lenses. The facets of the Fresnel dome lens are usually designed with undercut grooves (see Fig. 14.3B). Thus, peaks of the facets do not contribute to the concentration process. This reduces the optical losses in a Fresnel dome lens compared to a flat Fresnel lens. However, the undercut grooves complicate the manufacturing process. The molding tool must be able to collapse for the de-molding or the lens must be cast in two pieces and afterward bonded together. Anyhow, already in the early 1980s, the successful manufacture of Fresnel dome lenses for a PV concentrator module was reported. For example, a concentrator module developed for space applications used a silicone–glass hybrid Fresnel dome lens and mechanically stacked GaAs/GaSb dual-junction cells [12]. Today, Fresnel dome lenses are applied in a module manufactured by the Japanese company Daido Steel. These lenses provide a geometric concentration of about 500X and are produced with an injection molding process [13]. In 2005, Araki et al. reported a remarkably high Fresnel dome lens concentrator module efficiency of more than 28%. This module used monolithic triplejunction cells and was measured outdoors without applying a temperature correction [14].

ELEMENTS OF FRESNEL LENS CONCENTRATORS

319

Aside from cost and performance requirements, Fresnel lenses for CPV applications must also be very reliable. In particular, the lenses must withstand UV radiation, temperature cycles, humidity, wind and snow load and also extreme abrasive events such as a desert sandstorm or a hailstorm. Specifically, acrylic materials can suffer under UV light illumination. In particular, the transmittance of the lens material can decrease due to an increase of absorbance. Furthermore, the mechanical properties of the lens material can degrade due to accumulated UV irradiation. An additional degradation of optical components was reported for combined UV irradiation and an increased humidity [15]. Therefore, acrylic materials used in CPV systems need to be tested thoroughly. Thermal stress due to temperature cycles is another crucial impact factor on Fresnel lenses. In particular, injection-molded acrylic Fresnel lenses have been reported to crack at locations of high material stress [14]. This problem can be overcome by reducing the material stress during the injection molding process. In addition, the performance of Fresnel lenses can decrease due to a dust or dirt accumulation on the surface. When Fresnel lens concentrators are operated in a dry and dusty region, a continuous cleaning cycle should be considered. In contrast to reflecting surfaces, Fresnel lenses are usually not as sensitive to dirty surfaces. In experimental work performed at the UPM, it was found that the degradation due to dirt and dust was less for Fresnel lens systems than for parabolic dish systems [16]. So far, the general properties of Fresnel lenses have been described. In the following paragraph, the special requirements on Fresnel lenses as concentrators for highly efficient triple-junction solar cells are discussed. One key parameter is the concentration ratio. In order to offset the relatively higher material costs of the triple-junction cells, the concentration ratio should be at least 200X. Additionally, the Fresnel lens must provide a good optical transmission. As mentioned above, optical efficiencies greater than 80% are achievable with flat Fresnel lenses at concentration ratios of 300X–500X [17]. However, this is reached at the expense of a rather poor angular acceptance of the concentrator, which is well below 1°. In addition, the spatial distribution of light at the focal spot of the Fresnel lens is inhomogeneous. This can have a negative impact on the solar cell performance. In Figure 14.4, the measured spatial intensity distribution in the focus of a flat Fresnel lens with a lens aperture of 16 cm2 is shown. The central peak intensity decays to 10% at a distance of 0.6 mm from the center. This Fresnel lens is used together with a circular 2 mm cell, leading to a geometric concentration ratio of 500X. However, the real central peak intensity can reach concentration ratios above 2500X. The characteristics of Fresnel lenses described above can be summarized as follows: for high concentration ratios, as required for triple-junction solar cells, flat Fresnel lenses have a typically poor performance with respect to angular transmission and light homogeneity. In order to outweigh these disadvantages, secondary optical elements are often applied in CPV modules with flat Fresnel lenses.

320

HIGH-CONCENTRATION FRESNEL LENS ASSEMBLIES AND SYSTEMS

Figure 14.4. A profile of the measured intensity in the spot of a 40 × 40 mm2 Fresnel lens is shown. The measurement was carried out with monochromatic light at a setup developed at the Fraunhofer ISE [18].

14.2.2

Secondary Optics for Fresnel Lenses

Generally, a secondary optic provides a second chance to influence the path of the light rays. Usually, secondary optical elements are located close above the solar cell. The task of a secondary can be manifold. In particular, it can increase the concentration ratio and/or the angular transmission. If required, it can homogenize the spatial distribution of concentrated sunlight on the solar cell. A simple secondary approach for Fresnel lens concentrators was already used in a 200X Fresnel lens concentrator of the Sandia National Laboratories [19]. There, a highly reflective cone was located around the cell in order to catch spilled sunrays. Also in more recent developments at Fraunhofer ISE, the potential of a reflective cone secondary is investigated. In particular, these elements are manufactured from Al, which are equipped with metallic reflector layers [20]. The reflective cone secondary has been designed that only misguided light rays from the primary optics are concerned. Thus, this type of secondary is expected to have only a positive impact on the optical performance of the concentrator. Figure 14.5A presents the shortcircuit current of two FLATCON® test modules versus the angle of misalignment. Each of the two modules consists of six 4.15-mm2 triple-junction solar cells which are placed in the foci of 40 × 40 mm2 Fresnel lenses. One of the modules was, in addition, equipped with reflective secondary optics. The relative angular transmission is better for the test module with secondary optics. Obviously, the gain in optical performance increases with higher angles of misalignment. Please note that the solar disk has an angular size of about ±0.3°. The direct irradiation incoming within an opening angle between ±0.3° and ±2.5° is called

ELEMENTS OF FRESNEL LENS CONCENTRATORS

(A)

321

(B)

Figure 14.5. (A) The angular transmission of two test modules with six Fresnel lenses (40 × 40 mm2) and six solar cells (4.15 mm2) is shown. In principle, both modules have the same components, except for an additionally mounted reflective secondary optic on one of the modules. The short-circuit current of the modules was measured outdoors under a varying sun vector angle. (B) A photo of a test module equipped with reflective secondary elements is shown.

circumsolar irradiation, and the intensity is typically 5–20% (= circumsolar ratio, or CSR) of the direct sunlight [21], depending on the atmospheric conditions. As can be seen from Figure 14.5A, the CSR is poorly utilized by the Fresnel lens. Obviously, the reflective cone secondary offsets this disadvantage, and the direct sunlight and the CSR can be utilized more efficiently. For small angles of misalignment, the gain in optical transmission is relatively small and the secondary does not avoid a significant drop in optical transmission at a misalignment angle of 1°. This is a result of light reflection losses in the secondary. Alternatively to simple cone reflectors, secondary optics, which use total internal reflection, can be applied. These are dielectric optical elements that have to be placed in the concentrated beam of the Fresnel lens. At the entrance aperture, the light is refracted into the secondary and is total internally reflected at the side walls of the secondary. Finally, the light leaves the secondary at the exit aperture. Usually, the exit aperture and the solar cell are connected with an optical coupling medium in order to decrease reflection losses at the surfaces. Depending on the surface quality, total internal values up to 100% can be achieved and the angular transmission of the Fresnel lens concentrator can be increased significantly. The refraction of light at the entrance aperture provides an additional degree of freedom in optical design. However, losses occur due to surface reflection and absorption in the secondary dielectric material. Typical representatives of total internal reflection secondaries are CPCs and light rods. The latter usually has a conical shape with an entrance aperture larger than the exit aperture and can be used for homogenizing purposes. In the Fresnel

322

HIGH-CONCENTRATION FRESNEL LENS ASSEMBLIES AND SYSTEMS

dome concentrator of Daido Steel, such glass homogenizers are applied. This allows the concentrator to have a geometric concentration ratio of about 500X and a very homogeneous illumination of the solar cell and an acceptance angle of 1° [22]. A CPC is a nonimaging concentrating optical element that can allow high concentration ratio and high angular transmission values [23]. The side of a CPC is shaped parabolically and the entrance aperture is larger than the exit aperture. Beneath a Fresnel lens, it can be used to increase the concentration and the angular acceptance of the concentrator. Ray-tracing simulations show that with a Fresnel lens and a CPC secondary, a concentration of 1000X and an acceptance angle of 1° can be achieved. Typical disadvantages of all secondary optical elements that rely on total internal reflection are the high requirements for the surfaces. The roughness of the surface must be very low in order to prevent scattering and decoupling. This usually requires polishing techniques. Additionally, the shape of the total internal reflecting surface is not very tolerant to errors. Especially, the CPC shape is demanding and it is very challenging to meet these requirements with manufacturing techniques for mass production. However, already in the mid 1980s, the Varian Research Center used a conical glass element with a convex-shaped entrance aperture as secondary element. This provided a second refraction of the beam focused primarily by a Fresnel lens. In addition, a total internal reflection guided the light rays on the cell. Due to this advanced secondary concept, a Fresnel lens concentrator with a concentration factor of almost 1000X and an outstanding angular acceptance of more than 1° could be realized [24]. The modules used GaAs single-junction concentrator solar cells, achieving an efficiency of about 28% [25]. Furthermore, a module efficiency of 22.3% was reported, which was achieved with a Fresnel lens concentrator equipped with 12 GaAs solar cells and without applying a temperature correction [26]. Considering this concentrator would have been equipped with today’s triplejunction solar cells instead of the single-junction cells, one could expect a module operating efficiency of about 30%.

14.3

TRIPLE-JUNCTION SOLAR CELLS

In a PV system, the solar cells are considered to be the core element. They convert solar energy into electrical energy. However, this conversion process is limited by principal loss mechanisms. Two of these mechanisms are related to the spectral distribution of the solar energy and the bandgap of the solar cell material, respectively. In particular, photons with energies smaller than the bandgap of the semiconductor cannot be absorbed and are transmitted (transmission losses). On the other hand, photons with energies higher than the bandgap dissipate the excess energy to the semiconductor lattice (thermalization losses). In general, it is these two loss mechanisms that limit the maximal achievable efficiency of a singlejunction solar cell to approximately 30% [27].

TRIPLE-JUNCTION SOLAR CELLS

323

This limit can be exceeded by multijunction solar cells. A multijunction solar cell consists of subcells with different bandgap energies, see Figure 14.6. In particular, the subcells are placed on top of each other, sorted by bandgap energy. Thus, different subcells utilize different spectral bands, and the solar spectrum can be used more efficiently due to reduced thermalization and transmission losses. Today, the most efficient industrially produced solar cells have three subcells (triple-junction solar cell). They consist of a Ga0.50In0.50P top cell (bandgap: 1.87 eV), a Ga0.99In0.01As middle cell (bandgap: 1.44 eV), and a Ge bottom cell (bandgap: 0.67 eV). This monolithic triple-junction cell structure is grown in a MOVPE process. Figure 14.6 shows a sketch of the layer composition of a typical Ga0.50In0.50P/Ga0.99In0.01As/Ge triple-junction cell. The Ga0.50In0.50P/Ga0.99In0.01As/Ge triple-junction cell is also called latticematched triple-junction cell because all semiconductor materials used in the device have the same lattice constant. Thus, a very high material quality can be realized. Today, industrially produced lattice-matched concentrator triple-junction cells are offered with efficiencies between 37% and 39% [28–30]. Concentrator solar cells are rated at a cell temperature of 25°C and under the AM1.5 direct ASTM G173 reference spectrum. Under these conditions, the Ge bottom cell typically generates

front contact ARC cap layer n+-AlInP - window layer n-GalnP - emitter GalnP - undoped layer

Ga0.50In0.50P top cell

p-GalnP - base p+-GalnP - barrier layer p+-AlGalnP - barrier layer p++-AlGaAs n++-GaAs or GalnP n+-AlGanP/AlInAs - barrier layer n-GalnAs - emitter GalnAs - undoped layer p-GalnAs - base

tunnel diode 1 Ga0.99In0.01As middle cell

p+-GalnAs - barrier layer p+-AlGalnAs - barrier layer p++-AlGaAs n++-AlGaAs n- doped window - and nucleation layer n-Ge diffused emitter

tunnel diode 2

p-Ge substrate (100)

Ge bottom cell rear contact

Figure 14.6. A scheme of a typical Ga0.50In0.50P/Ga0.99In0.01As/Ge triple-junction cell is sketched.

324

HIGH-CONCENTRATION FRESNEL LENS ASSEMBLIES AND SYSTEMS

about 30% more current than the other two subcells. Due to the internal series connection of the triple-junction cell, the voltages are added; however, the current is limited by the subcell with the lowest current. Consequently, the excess current of the Ge bottom cell is lost. This shows the path for efficiency improvements. Ideally, every subcell in the triple-junction cell should generate the same current. Assuming a Ge bottom cell with a bandgap of 0.67 eV, a lowering of the top and middle cell bandgaps would result in a better current matching. In a triple-junction solar cell with a 1.75 eV top cell, a 1.2 eV middle cell, and a 0.67 eV bottom cell, all three cells would generate about the same current under the reference spectrum [31]. Consequently, a triple-junction solar cell with this bandgap combination can achieve a higher efficiency than a lattice-matched triple-junction cell. Figure 14.7 A shows the bandgap energies versus lattice constant for selected semiconductor materials. As described above, the materials Ga0.50In0.50P, Ga0.99In0.01As and Ge have the same lattice constant. However, when the Ga to In ratio in GaInP or GaInAs is changed, this influences the bandgap and the lattice constant. By increasing the In content in these materials, the bandgaps are lowered and the lattice constants are increased. The material Ga0.35In0.65P has a bandgap of 1.68 eV and is lattice matched to Ga0.83In0.17As, which has a bandgap of 1.2 eV. Thus, the bandgaps of these two materials are expected to be more suitable for the top and middle cells of a triple-junction cell on a Ge substrate. However, the lattice constant of Ga0.35In0.65P/Ga0.83In0.17As is about 1% greater than that of Ge. When this material combination is grown on a Ge substrate, the difference in the lattice constant introduces strain in the grown layers. If the strain is too high, it relaxes and misfit dislocations are generated (see Fig. 14.7B). Thus, a latticemismatched (an alternative term for lattice-mismatched materials is metamorphic materials) triple-junction solar cell with a good material quality can be realized when the dislocations are confined in a non-photo-active buffer structure. The success of the lattice-mismatched concept has been proven in 2009 when a new efficiency record of 41.1% was reached. Figure 14.8 shows the efficiency in dependence of the concentration level for the lattice-mismatched Ga0.35In0.65P/ Ga0.83In0.17As/Ge triple-junction solar cell. The efficiency peaks at a concentration level of 454X. For higher concentrations, the efficiency decreases due to ohmic losses; however, at about 800X, the efficiency is still higher than 40%. Spectrolab also experimented with the lattice mismatch approach and published an efficiency of 40.7% [31]. In addition to the described lattice-mismatched triple-junction cells on Ge substrates, inverted metamorphic triple-junction cells are under development. This concept was suggested by NREL [32] and uses a GaInP top cell, a GaInAs middle cell with a relatively low In content, and a GaInAs bottom cell with a high In content (bandgap: ∼1 eV). In contrast to other triple-cells, these cells are grown upside down (inverted) on a GaAs or Ge substrate. Consequently, the substrate used for the growth must be removed. This can be carried out by bonding the cell layers onto a carrier substrate (e.g., silicon or sapphire) and by removing the growth substrate afterward. In 2008, NREL reported an efficiency of 40.8% for a Ga0.51In0.49P/Ga0.96In0.04As/Ga0.63In0.37 as inverted metamorphic triple-junction cell at a concentration of 326X [33]. The results achieved

TRIPLE-JUNCTION SOLAR CELLS

325

(A) Material layers not relaxed

Material layers relaxed Material with larger lattice constant than substrate Misfit dislocations

Substrate

(B)

Figure 14.7. (A) The bandgap and the lattice constant for selected semiconductor materials are shown. The materials Ga0.50In0.50P and Ga0.99In0.01As have the same lattice constant as Ge; however, they have different bandgaps. The increase of the In versus Ga ratio decreases the bandgaps and increases the lattice constants of GaInP and GaInAs. In particular, Ga0.35In0.65P has the same lattice constant as Ga0.83In0.17A and a lattice mismatch of about 1% relative to Ge. (B) When a material is grown on a substrate with a different lattice constant, it is strained. If the strain is too high, it is relaxed by the formations of misfit dislocations.

with the lattice mismatch concept show clearly the potential of this approach, and efficiencies beyond 45% can be expected in the future.

14.3.1

Triple-Junction Solar Cells under a Varying Spectrum

As described in the previous subsection, triple-junction concentrator solar cells have achieved remarkably high efficiencies >40% under the reference spectrum. This was made possible by introducing subcell bandgaps that allow a current matching condition for the reference spectrum. However, the sun’s spectrum varies

326

HIGH-CONCENTRATION FRESNEL LENS ASSEMBLIES AND SYSTEMS

Figure 14.8. The efficiency and the fill factor of a Ga0.35In0.65P/Ga0.83In0.17As/Ge triplejunction solar cell versus concentration. This cell was manufactured at Fraunhofer ISE and measured at the CalLab of Fraunhofer ISE. One can see that the efficiency increases with increasing concentration and peaks at about 454X. For higher concentrations the series resistance of the cell and the related drop in fill factor decreases the efficiency. However, at 800X the efficiency is still remarkably high (>40%).

during the course of the day and season and is usually not equal to the reference spectrum. Due to the resulting current limitations, it is expected that concentrator systems with triple-junction cells are more sensitive to changes in the solar spectrum than CPV systems using single-junction cells. The influence of the spectral variations on the electrical parameters of a triplejunction solar cell can be investigated in the laboratory by using a multi-light source simulator [34, 35]. Here, the electrical parameters of the triple-junction solar cell are measured, while the incident spectrum is changed systematically. Thus, on the cell level, the influence of the spectral changes is well investigated. Under varying sun spectra, the power of triple-junction solar cells can be expected to vary by about 20%. Unfortunately, the influence of the solar spectrum on the energy production of concentrator systems equipped with triple-junction cells is not well investigated at present. Long-term field experience is still lacking due to the rather young triple-junction technology and the lack of information on the changing solar spectrum combined with concentrator optics. However, some investigations of Fresnel lens concentrators equipped with triple-junction cells have been already performed and show the spectral impact in the morning and evening hours [36–38]. At these high air masses, spectral distribution of the sunlight is relatively “redrich.” As a result, the top cell current is significantly lower than the current of the other subcells and the efficiency of the concentrator system decreases.

RECENT EXAMPLES AND DEVELOPMENTS

327

Obviously, the annual energy production of concentrator systems using triplejunction solar cells is more affected by the solar spectrum than systems with singlejunction cells. In a number of theoretical studies, the annual energy production of multijunction solar cells with respect to the solar spectrum has been investigated [39, 40]. The basic conclusion from these theoretical investigations is that despite the penalty of current limitation, triple-junction solar cells outperform other solar cell concepts. This is not surprising when considering that air masses <2 contribute to about two-thirds of the annual solar energy. Thus, the negative impact of the changing solar spectrum can be expected to be mitigated when a triple-junction cell is optimized for a solar spectrum at air masses around 1.5, and then the negative impact of the varying solar spectrum is expected to be less than 5% of the annual energy production.

14.4

RECENT EXAMPLES AND DEVELOPMENTS

Motivated by the quickly rising efficiency of triple-junction solar cells, there have been many new companies founded in the field of PV concentrators in recent years. An overview of the companies currently in this field can be found in Reference 41. Many current concentrator concepts rely on Fresnel lenses and on triplejunction solar cells. Although the particular construction of every Fresnel lens concentrator is unique, the basic assembly is similar for all concepts. As an example of a basic assembly concept, the FLATCON technology is described in the following section. The FLATCON concept has been under continuous development for more than 10 years [5, 42]. The optics is a silicone–glass hybrid Fresnel lens, similar to the optics used in the Ramón Areces concentrator mentioned in the introduction. In contrast to this early concentrator, the lens aperture area of a FLATCON concentrator is about 100 times smaller. This enables reduced material thickness leading to less material use and decreased material costs. From a manufacturing point of view, however, the small dimensions of the sub-elements present challenges. Due to the relatively small dimensions of the Fresnel lenses, about 100 times more sub-elements are required in order to cover the same area of one sub-element of the Ramón Areces concentrator. Thus, 100 times more cells must be tested, mounted, and electrically connected. This issue is tackled by using the manufacturing technology known from the microelectronics industry, where many small parts must be assembled precisely while maintaining a high throughput. At the Fraunhofer ISE, an automated production line for the concentrator assemblies was set up in the years 2005 and 2006 [43, 44]. At the same time, a Fraunhofer spin-off was founded—the company Concentrix Solar, to which the FLATCON technology was licensed [45]. The further development of this technology and related manufacturing techniques by Concentrix Solar in cooperation with the Fraunhofer ISE is ongoing. In the mean time, Concentrix Solar has inaugurated an automated assembly line for FLATCON modules, which is capable of producing about 25 MW annual peak power.

328

14.5

HIGH-CONCENTRATION FRESNEL LENS ASSEMBLIES AND SYSTEMS

ASSEMBLY OF FLATCON SYSTEMS

Today, triple-junction solar cells are typically delivered on 4 inch wafers. For the assembly process, the wafers are diced on a foil. The cells are picked automatically by a die bonding machine and are mounted onto a substrate. In a FLATCON module, the substrate consists basically of copper, which acts as heat sink and simultaneously as the electrical rear-side contact for the cell. This is achieved by using an electrically conducting adhesive to mount the cell on the copper plate. The front-side contact of the triple-junction solar cell is gold wire bonded to a layer that is electrically isolated to the copper substrate. Figure 14.9 shows the mounted solar cell on the cooper heat spreader. This unit is called solar cell assembly. In a pick-and-place process, the solar cell assembly is mounted with high precision onto the bottom glass plate of a FLATCON module [43]. The top plate of a module carries the quadratic Fresnel lenses, which are arranged in a matrix. The bottom

Figure 14.9. The most important assembly steps necessary for a FLATCON concentrator assembly are shown. Upper left: the photo of a 4-in. wafer with about 1000 triple-junction concentrator solar cells is shown. Upper right: the cells are mounted onto a copper heat sink. The front side of the cell is contacted with gold wire bond technology. Lower right: many of the solar cell assemblies are mounted in the focus of their corresponding Fresnel lens— the position of one solar cell assembly in the module is marked. Lower left: many FLATCON modules are mounted on a two-axis tracking system—the position of one FLATCON module is marked.

FLATCON SYSTEMS AND DEVELOPMENTS

329

glass plate with the solar cell assembly and the Fresnel lens plate are adjusted to each other. In the next step, the frames between the two plates are sealed together. A photo of the complete module is shown in Figure 14.9. In order to complete a full FLATCON system, many modules are mounted on a two-axis tracking system. Figure 14.9 shows photos of how a FLATCON system is assembled.

14.6

FLATCON SYSTEMS AND DEVELOPMENTS

A current FLATCON system from Concentrix Solar generates a peak AC power of about 5.4 kW at a direct normal irradiation of 850 W/m2 [46]. Figure 14.10 shows a photo of this CPV system. In addition, the generated AC power normalized to the reference irradiation of 850 W/m2 is shown for the years 2008 and 2009. Obviously, the generated power meets the expectations well and no degradation in power was observed during the time of measurement. The system achieved a nominal AC efficiency of more than 22%, which is a remarkably high value for a PV system. The current FLATCON design does not make use of secondary optics. In order to improve the optical efficiency of these Fresnel lens concentrators, the application of secondaries is part of an ongoing development. In particular,

(A)

(B)

Figure 14.10. (A) Photo of a state-of-the-art FLATCON system manufactured by Concentrix Solar is shown. (B) The AC power (normalized to a direct normal irradiance of 850 W/m2) of one Concentrix system located in Seville (Spain) and monitored during the years 2008 and 2009 is shown. At an irradiance of 850 W/m2, the system is expected to have a nominal power of 5.4 kW. The nominal power was reached on average and no degradation was observed during the time of monitoring. Courtesy of Dr. Andreas Gombert, Concentrix Solar.

330

HIGH-CONCENTRATION FRESNEL LENS ASSEMBLIES AND SYSTEMS

reflective secondary elements are investigated for this purpose. These elements are relatively cheap in production and can be easily introduced in the existing FLATCON design. The reflective secondary consists of a reflective cone which increases the chance of collecting light incoming from the primary Fresnel lens. Outdoor experiments with FLATCON test modules (see Fig. 14.5B) show an average power increase of about 6% by applying reflective secondary optics.

14.7

ROOFTOP APPLICATIONS

Currently, CPV is mostly considered for power plant applications in sun-rich regions. However, in recent years, some companies are trying to tackle the rooftop market with CPV systems. Large, flat rooftop areas provide enough space for several megawatts of installed peak power; thus, the rooftop application can also be suitable for PV concentrators. From a technological point of view, one has to consider the demands for low visual impact and low wind loads to meet the static requirements of the building. A standard two-axis tracking system, as shown in Figure 14.10, is not well suited for this application. Consequently, a highconcentration Fresnel lens system for rooftop applications requires a special module and tracking design. One example of such a concentrator is the GiraSol system developed by the Spanish company Sol3g [47]. This particular system is a high-concentration Fresnel lens concentrator using triple-junction solar cells combined with a glass rod secondary. The sub-modules are arranged in a line (see Figure 14.11). Consequently, the elevation of every module can be tracked apart and the azimuth angle is tracked for all modules together. Thus, the whole CPV system has a low, flat profile and is suitable for the application on flat rooftops. A photo of a Sol3g high-concentration PV GiraSol system is shown in Figure 14.11.

Figure 14.11. A photo of a photovoltaic concentrator specially developed for rooftop applications is shown. Courtesy of Ricard Pardell, Sol3g.

ABBREVIATIONS

14.8

331

SUMMARY

Fresnel lenses are considered to be a cheap and effective optic for concentrating PV applications. They have been used in concentrators since the beginning of CPV development in the 1970s. Due to the impressive recent increase in efficiency above 40% for triple-junction solar cells, PV concentrators are experiencing a renaissance. The relatively high material costs of these cells can be offset when they are operated at the focus spot of an optical system with a concentration ratio of at least 200X. This requirement can be met by using point-focus Fresnel lenses, which, however, have the nonideal optical properties in terms of angular transmission and homogeneous light distribution. Consequently, many Fresnel lens concentrators using triple-junction solar cells are additionally equipped with secondary optical elements in order to overcome these optical disadvantages. To date, the main applications of triple-junction Fresnel lens concentrator systems are power plants in sun-rich regions. Several companies have already set up power plants with several 100 kW. Under true field operation, CPV systems have demonstrated a conversion efficiency from direct sunlight to AC electricity of more than 22%. Consequently, these systems are considered to be the most efficient PV energy converters today. Moreover, further developments in the triplejunction cell technology and in the Fresnel lens optics will certainly lead to even higher efficiencies in this fast-moving field. In addition to the power plant installations, roof application is also a future market segment for concentrating PV systems. This rather nonconventional concentrator application requires special designs with respect to static issues and visual impacts for rooftops.

ABBREVIATIONS AC—alternating current Al—aluminum As—arsenic ASTM—American Society for Testing and Materials CPC—compound parabolic concentrator CPV—concentrator photovoltaics CSR—circumsolar ratio Ga—gallium Ge—germanium III–V—alloys composed of elements from columns III and V of the periodic table In—indium ISE—Institut für Solare Energiesysteme MOVPE—metal-organic vapor phase epitaxy NREL—National Renewable Energy Laboratory P—phosphorous PV—photovoltaic

332

HIGH-CONCENTRATION FRESNEL LENS ASSEMBLIES AND SYSTEMS

Sb—antimonide UPM—Universidad Politécnica de Madrid UV—ultraviolet X—geometric concentration ratio

REFERENCES [1] [2]

[3]

[4]

[5]

[6]

[7]

[8] [9]

[10] [11] [12]

E. L. Burgess and M. W. Edenburn. One kilowatt photovoltaic subsystem using Fresnel lens concentrators. In Proceedings of the 12th IEEE Photovoltaic Specialists Conference, Baton Rouge, LA, pp. 776–780. New York, IEEE (1976). E. Lorenzo and G. Sala. Hybrid silicone-glass Fresnel lens as concentrator for photovoltaic applications. In Proceedings of the International Solar Energy Society, Atlanta, GA, K. W. Boer and B. H. Glenn, eds., pp. 536–539. New York, Pergamon (1979). G. Sala and E. Lorenzo. Hybrid silicone-glass Fresnel lens as concentrator for photovoltaic applications. In Proceedings of the 2nd European Photovoltaic Solar Energy Conference, R. Van Overstraten and W. Palz, eds., pp. 1004–1010. Dordrecht, D. Reidel (1979). V. A. Grilikhes, V. D. Rumyantsev, and M. Z. Shvarts. Indoor and outdoor testing of space concentrator AlGaAs/GaAs photovoltaic modules with Fresnel lenses. In Proceedings of the 23rd IEEE Photovoltaic Specialists Conference, Washington, DC, pp. 345–348. New York, IEEE (1996). V. D. Rumyantsev, V. M. Andreev, A. W. Bett, F. Dimroth, M. Hein, G. Lange, M. Z. Shvarts, and O. V. Sulima. Progress in development of All-Glas terrestrial concentrator modules based on composite Fresnel lenses and III-V solar cells. In Proceedings of the 28th IEEE Photovoltaic Specialists Conference, Anchorage, AK, pp. 1169–1172. New York, IEEE (2000). M. Hein, M. Meusel, C. Baur, F. Dimroth, G. Lange, G. Siefer, T. N. D. Tibbits, A. W. Bett, V. M. Andreev, and V. D. Rumyantsev. Characterisation of a 25% high-efficiency Fresnel lens module with GaInP/GaInAs dual-junction concentrator solar cells. In Proceedings of the 17th European Photovoltaic Solar Energy Conference, Munich, B. McNelis, W. Palz, H. A. Ossenbrink, and P. Helm, eds., pp. 496–499. Munich, WIP–Renewable Energies, and Florence, ETA (2001). L. W. James, R. L. Moon, E. O. J. Moore, and R. L. Bell. Progress report on a 1-kW terrestrial array of AlGaAs/GaAs concentrator solar cells. In Proceedings of the 12th IEEE Photovoltaic Specialists Conference, Baton Rouge, LA, pp. 771–773. New York, IEEE (1976). J. M. Olson, A. Kibbler, and S. Kurtz. GaInP2/GaAs monolithic tandem solar cells. In Proceedings of the 19th IEEE Photovoltaic Specialists Conference, New Orleans, LA, pp. 285–288. New York, IEEE (1987). J. M. Olson, S. R. Kurtz, A. E. Kibbler, and P. Faine. Recent advances in high efficiency GaInP2/GaAs tandem solar cells. In Proceedings of the 21st IEEE Photovoltaic Specialists Conference, Kissimmee, FL, pp. 24–29. New York, IEEE (1990). L. W. James and J. K. Williams. Fresnel optics for solar concentration on photovoltaic cells. In 13th IEEE Photovoltaic Specialists Conference, Washington, DC, pp. 673–679. New York, IEEE (1978). A. Luque. Solar Cells and Optics for Photovoltaic Concentration. Bristol, England, IOP Publishing. International Standard Book Number 0-85274-106-5 (1989). M. F. Piszczor, D. J. Brinker, D. J. Flood, J. E. Avery, L. M. Fraas, E. S. Fairbanks, J. W. Yerkes, and M. J. O’Neill. A high-performance photovoltaic concentrator

REFERENCES

[13]

[14]

[15]

[16] [17]

[18]

[19] [20]

[21] [22]

[23] [24] [25]

333

array: The mini-dome Fresnel lens concentrator with 30% efficient GaAs/GaSb tandem cells. In Proceedings of the 22nd IEEE Photovoltaic Specialists Conference, Las Vegas, NV, pp. 1485–1490. New York, IEEE (1991). K. Araki, M. Kondo, H. Uozumi, Y. Kemmoku, T. Egami, M. Hiramatus, Y. Miyazaki, N. J. Ekins-Daukes, M. Yamaguchi, G. Siefer, and A. W. Bett. A 28% efficient, 400x and 200 Wp concentrator module. In Proceedings of the 19th European Photovoltaic Solar Energy Conference, Paris, W. Hoffmann, J.-L. Bal, H. Ossenbrink, W. Palz, and P. Helm, eds., pp. 2495–2498. Munich, WIP–Renewable Energies, Florence, ETA (2004). K. Araki, H. Uozumi, T. Egami, M. Hiramatsu, Y. Miyazaki, Y. Kemmoku, A. Akisawa, N. J. Ekins Daukes, H. S. Lee, and M. Yamaguchi. Development of concentrator modules with dome-shaped Fresnel lenses and triple-junction concentrator cells. Prog. Photovoltaics Res. Appl. 13, 513–527 (2005). K. Araki, H. Uozumi, T. Takamoto, M. Hiramatsu, A. Akisawa, Y. Kemmoku, and M. Yamaguchi. CPV using 3J III-V cells and Fresnel lenses—Reliability, applications and perspectives. In Proceedings of the 3rd International Conference on Solar Concentrators for the Generation of Electricity or Hydrogen, Scottsdale, AZ. Golden, CO, NREL (2005). M. Vivar, R. Herrero, R. Moretón, F. Martínez, and G. Sala. Effect of the soiling on PV concentrators: Comparison with flat modules. In Proceedings of the 33rd IEEE Photovoltaic Specialists Conference, San Diego. New York, IEEE (2008). V. D. Rumyantsev, M. Hein, V. M. Andreev, A. W. Bett, F. Dimroth, G. Lange, G. Letay, M. Z. Shvarts, and O. V. Sulima. Concentrator array based on GaAs cells and Fresnel lens concentrators. In Proceedings of the 19th European Photovoltaic Solar Energy Conference, Glasgow, W. Hoffmann, J.-L. Bal, H. Ossenbrink, W. Palz, and P. Helm, eds., pp. 2312–2315. Munich, WIP–Renewable Energies, and Florence, ETA (2000). A. Heller, P. Nitz, and W. Platzer. Indoor characterization of Fresnel-type concentrator lenses. In Proceedings of the 4th International Conference on Solar Concentrators for the Generation of Electricity or Hydrogen, El Escorial, Spain, G. Sala, R. D. McConnel, and R. R. King, eds., pp. 289–293 (2007). D. E. Arvizu. Development of the Sandia 200X experimental silicon module. In 13th IEEE Photovoltaic Specialists Conference, Orlando, FL, pp. 805–813. New York, IEEE (1984). J. Jaus, P. Nitz, G. Peharz, G. Siefer, T. Schult, O. Wolf, M. Passig, T. Gandy, and A. W. Bett. Second stage reflective and refractive optics for concentrator photovoltaics. In Proceedings of the 33rd IEEE PVSC, San Diego, CA, New York, IEEE (2008). A. Neumann, A. Witzke, S. A. Jones, and G. Schmitt. Representative terrestrial solar brightness profiles. J. Sol. Energy Eng. 124, 198–204 (2002). K. Araki, M. Kondo, H. Uozumi, N. J. Ekins Daukes, Y. Kemmoku, and M. Yamaguchi. Achievement of 27% efficient and 200 Wp concentrator module and the technological roadmap toward realization of more than 31% efficient modules. Sol. Energy Mater. Sol. Cells 90, 3312–3319 (2006). W. T. Welford and R. Winston. The Optics of Nonimaging Concentrators Light and Solar Energy. San Diego, CA, Academic (1978). N. R. Kaminar and H. C. Hamaker. 19% concentrator module efficiency using single-junction GaAs cells. In Proceedings of the 18th IEEE Photovoltaic Specialists Conference, pp. 1559–1561. New York, IEEE (1985). H. F. MacMillan, H. C. Hamaker, N. R. Kaminar, M. S. Kuryla, M. L. Ristow, D. D. Liu, G. F. Virshup, and J. M. Gee. 28% efficient GaAs concentrator solar cells. In Proceedings of the 20th IEEE Photovoltaic Specialists Conference, Las Vegas, NV, pp. 462–468. New York, IEEE (1988).

334

HIGH-CONCENTRATION FRESNEL LENS ASSEMBLIES AND SYSTEMS

[26]

M. S. Kuryla, M. L. Ristow, L. D. Partain, and J. E. Bigger. 22.3% efficient 12 cell 1000 sun GaAs concentrator module. In Proceedings of the 22nd IEEE Photovoltaic Specialists Conference, Las Vegas, NV, pp. 506–511. New York, IEEE (1991). W. Shockley and H. J. Queisser. Detailed balance limit of efficiency of p-n junction solar cells. J. Appl. Phys. 32, 510–519 (1961). AZUR SPACE Solar Power GmbH. Available at http://www.azurspace.com/. Accessed April 2, 2009. Emcore Corporation. Available at http://www.emcore.com/. Accessed April 2, 2009. Spectrolab, Inc. Available at http://www.spectrolab.com/. Accessed April 2, 2009. R. R. King, D. C. Law, K. M. Edmondson, C. M. Fetzer, G. S. Kinsey, H. Yoon, R. A. Sherif, and N. H. Karam. 40% efficient metamorphic GaInP/GaInAs/Ge multijunction solar cells. Appl. Phys. Lett. 90, 183516–183519 (2007). M. W. Wanlass, S. P. Ahrenkiel, R. K. Ahrenkiel, D. S. Albin, J. J. Carapella, A. Duda, J. F. Geisz, S. Kurtz, T. Moriarty, R. J. Wehrer, and B. Wernsman. Latticemismatched approaches for high-performance, III-V photovoltaic energy converters. In Proceedings of the 31st IEEE Photovoltaic Specialists Conference, Orlando, FL, pp. 530–535. New York, IEEE (2005). J. F. Geisz, D. J. Friedman, J. S. Ward, A. Duda, W. J. Olavarria, T. E. Moriarty, J. T. Kiehl, M. J. Romero, A. G. Norman, and K. M. Jones. 40.8% efficient inverted triple-junction solar cell with two independently metamorphic junctions. Appl. Phys. Lett. 93, 123501–123503, 123505 (2008). M. Meusel, R. Adelhelm, F. Dimroth, A. W. Bett, and W. Warta. Spectral mismatch correction and spectrometric characterization of monolithic III-V multi-junction solar cells. Prog. Photovoltaics Res. Appl. 10, 243–255 (2002). C. Baur and A. W. Bett. Measurement uncertainties of the calibration of multijunction solar cells. In Proceedings of the 31st IEEE Photovoltaic Specialists Conference, Orlando, FL, pp. 583–586. New York, IEEE (2005). K. Araki, K. Emery, G. Siefer, A. W. Bett, T. Sakakibara, Y. Kemmoku, N. J. Ekins-Daukes, H. S. Lee, and M. Yamaguchi. Comparison of efficiency measurements for a HCPV module with 3J cells in 3 sites. In Proceedings of the 31st IEEE Photovoltaic Specialists Conference, Orlando, FL, pp. 846–849. New York, IEEE (2005). A. W. Bett, J. Jaus, G. Peharz, G. Siefer, A. Hakenjos, I. Heile, H. Lerchenmüller, and J. Wüllner. Outdoor evaluation of FLATCON® modules and systems. In Proceedings of the 33rd IEEE PVSC, San Diego, CA. New York, IEEE (2008). G. Peharz, G. Siefer, K. Araki, and A. W. Bett. Spectrometric outdoor characterization of CPV modules using isotype monitor cells. In Proceedings of the 33rd IEEE PVSC, San Diego, CA. New York, IEEE (2008). P. Faine, S. R. Kurtz, C. Riordan, and J. M. Olson. The influence of spectral solar irradiance variations on the performance of selected single-junction and multijunction solar cells. Solar Cells 31, 259–278 (1991). G. S. Kinsey and K. M. Edmondson. Spectral response and energy output of concentrator multijunction solar cells. Prog. Photovoltaics Res. Appl. 17, 279–288 (2009) (early view). G. Hering. Dawn of 500 suns. Photon Magazine, November 1, pp. 146–158 (2008). A. W. Bett, C. Baur, F. Dimroth, G. Lange, M. Meusel, S. van Riesen, G. Siefer, V. M. Andreev, V. D. Rumyantsev, and N. A. Sadchikov. FLATCON™-modules: Technology and characterisation. In Proceedings of the 3rd World Conference on Photovoltaic Energy Conversion, Osaka, K. Kurokawa, L. L. Kazmerski, B. McNelis, M. Yamaguchi, and W. C. Simke, eds., pp. 634–637. Osaka, WPEC-3 (2003).

[27] [28] [29] [30] [31] [32]

[33]

[34] [35] [36]

[37] [38] [39] [40] [41] [42]

REFERENCES [43]

[44]

[45] [46] [47]

335

J. Jaus, U. Fleischfresser, G. Peharz, F. Dimroth, H. Lerchenmueller, and A. W. Bett. Heat sink substrates for automated assembly of concentrator modules. In Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden, J. Poortmans, H. Ossenbrink, E. Dunlop, and P. Helm, eds., pp. 2120–2123. Munich, WIP (2006). J. Jaus, A. W. Bett, E. Reinecke, and E. R. Weber. High precision placement of solar cell assemblies on large base plates for concentrator photovoltaics. In Proceedings of the 3rd International Conference on Multi-Material Micro Manufacture, Borovets, Bulgaria, S. Dimov, W. Menz, and Y. Toshev, eds., pp. 139–142. Borovets, Bulgaria, CRS Press (2007). A. W. Bett and H. Lerchenmueller. The FLATCON® System from Concentrix Solar, In Concentrator Photovoltaics, A. Luque and V. M. Andreev, eds., pp. 301–319. Heidelberg, Springer Verlag (2007). Concentrix Solar GmbH. Available at http://www.concentrix-solar.de/. Accessed April 2, 2009. Sol3g. Available at http://www.sol3g.com/. Accessed April 2, 2009.

15 HIGH-CONCENTRATION CASSEGRAINIAN SOLAR CELL MODULES AND ARRAYS MICHAEL LUDOWISE1 AND LEWIS FRAAS2 1 SolFocus, Inc., 2JX Crystals Inc.

15.1

INTRODUCTION

The advent of very high-efficiency multijunction III–V solar cells has opened up new possibilities in PV power conversion over the last several years. Recent laboratory demonstrations of 40% efficiency, with 38% average efficiency in commercially available quantities, indicate the growing robustness of HCPV energy generation. Equally important, efficient and manufacturable optical designs that deliver to the cells concentrated sunlight in multiples of 500 to several thousand emerged over the same time frame. Combined, the two advances have made it possible to field several hundred kilowatt-sized HCPV systems, with larger installations following in the near future. The felicitous coupling of very high-efficiency solar cells with elegant concentrating optical designs, however, is hardly coincidental. The extraordinary sophistication of high-efficiency III–V solar cells, containing upward of 60 epitaxial layers including three collecting p/n junctions and two tunnel junctions, comes at a high price. If arrayed in a flat-plate configuration, these cells would cost as much as $100 thousand per square meter, pricing the system well beyond cost-competitive levels. Instead, HCPV systems utilize inexpensive materials, such as glass and aluminum, to transfer sunlight onto a relatively small cell area, thereby reducing the unit area costs dramatically. The cost of an HCPV system will become competitive with other energy alternatives only if the optical system is designed to be efficient, inexpensive to manufacture, easy to assemble, Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

337

338

HIGH-CONCENTRATION CASSEGRAINIAN SOLAR CELL

compact for transportation, and simple to maintain. These considerations have driven the progress of Cassegrainian optical systems for HCPV in recent years, leading to a solar power receiver that satisfies all the requirements economically. Because the Cassegrainian optics (indeed, any concentrating optical design) forms, in effect, a telescope that focuses the sun onto either the solar cell or a tertiary optic, the HCPV power receiver must be mounted on a tracking system that closely follows the sun throughout the day. Tracking systems and the mounts for the power receivers come with tracking and manufacturing tolerances. Factors such as wind loading stress and flexion of the system in various orientations also increase required optical tolerances. There are design trade-offs between the precision of the tracking system and the precision of the optics. More precision in the tracker raises its cost. At the same time, if more tolerance is built into the optics, the precision of the tracker and the associated cost can be reduced. Indeed, the overall design of any HCPV system involves complex trade-offs between cell performance, optical design, tracker design, and BOS costs. A unified approach to optimizing the system design is to consider the effect of several target design points on the LCOE. The LCOE expresses the average cost of producing a kilowatt hour of electricity with the system, spanning the time from installation to decommissioning, including all the costs of components, financing, installation, repairs, and operation. This chapter discusses the design and performance of Cassegrainian HCPV systems. Comparisons to alternative concentrator designs are made. The details of the Cassegrainian optical design, application to panel design, and fabrication and tracker requirements will be discussed, and the field performance will be evaluated. Finally, a novel spectral splitting design utilizing dual solar cells is discussed.

15.2 DESIGN AND OPERATION OF HIGH-EFFICIENCY MONOLITHIC MULTIJUNCTION SOLAR CELLS To understand the design constraints placed on a concentrator system, some familiarity with the behavior of the multijunction cells themselves is necessary. Chapter 13 of this volume treats the detailed aspects of III–V multijunction solar cells. This section will discuss briefly several aspects of multijunction cells that have direct bearing on the optical concentrator and solar tracker designs. The fundamental concept behind a multijunction cell is spectral splitting (Fig. 15.1). By dividing the solar spectrum into several bands, each producing roughly the same current but in a different junction, energy losses owing to photons with energy excessively above the bandgap of the collecting p/n junctions are minimized. Kurtz et al. at NREL first demonstrated the key features of the high-efficiency cells in a 2J monolithic cell in 1990 [1]. Contemporary concentrator cells now contain three junctions at 660, 890, and 1875 nm (Ga1–xInxP, Ga1–xInxAs, and Ge, respectively). Because bandgap selection is limited by the practical choice of materials, these cells produce an overabundance of current in the germanium lower subcell. The upper two subcells, however, are closely matched in current under an

DESIGN AND OPERATION

339

Multijunction Solar Cell

Infrared

Upper Cell

Contact Grid Lines

Red

Blue

Full Solar Spectrum

Middle Cell

Lower Cell

Back Contact

Figure 15.1. Multijunction stack and spectral splitting. The multijunction cell is constructed from a multilayer stack of single-crystalline III–V semiconductor material. The blue and green light is absorbed by the uppermost subcell, red and orange by the middle subcell, and infrared light by the lower subcell.

AM1.5d spectrum. Minor changes in the spectral balance between the upper cell adsorbing band (350–660 nm) and the middle cell adsorbing band (660–890 nm) cause one of the two subcells to limit the current in the series-connected stack, reducing the cell efficiency. For example, at the low sun angles of early morning or late afternoon, the red portion of the spectrum dominates, and the upper cell becomes the current limiting subcell. Conversely, at higher sun angles, around midday, the spectrum becomes blue weighted, and the middle cell becomes the limiting element. This behavior has implications for the spectral transfer function of the optics and chromatic aberrations in the concentrating system [2]. The subcells are interconnected by tunnel junctions, which minimize the built-in diode voltage drop introduced by the two parasitic n/p junctions formed between each of the lower and middle subcell pair and the middle and upper subcell pair. Tunnel junctions exhibit negative differential conductivity at a certain critical (switching) current density (Fig. 15.2), which is determined by the bandgap of the material and by the doping densities and profiles near the junction. Normally, the junction operates at Imp, below the switching current Iswitch. Difficulties in forming the tunnel junction during the epitaxial crystal growth process may cause it to operate at a much higher voltage, Vhigh, still at Imp. Switching drops the output voltage of the cell, severely reducing the collected power [3]. At best, even a wellconstructed tunnel junction adds some undesirable series resistance.

340

HIGH-CONCENTRATION CASSEGRAINIAN SOLAR CELL

I

Iswitch Imp

0 Vtj

Vswitch

Vhigh

V

0

Figure 15.2. Tunnel junction current–voltage characteristic. The intercell connecting tunnel junctions carry current at a relatively low voltage drop up to the switching current when the voltage drop suddenly increases to a higher value. Well-constructed tunnel junctions are essential to high-efficiency cell operation.

Additional resistance sources in the cell include the resistivity of each layer, the specific contact resistance of the front and back metallization, and the distributed spreading resistances of the contact layer and the metallic grid lines. Of these, the designer has the most control over the grid resistance. Typically, solar cells are designed with as narrow a grid width as is practical and high aspect metallization in order to maximize the metallic conductivity while minimizing the shadowing of the solar cell active area. For light incident at high angles off of normal, though, the additional shadowing increases as h(tan θ), where θ is the angle from normal, and h is the height of the grid line. For high aspect ratio grid lines, shadowing loss can become a significant loss mechanism for light guided to the cell at modest angles, as in the case of fast f-number (high concentration) optical systems [4]. As the f-number decreases, a large percentage of the light is incident on the cell at high angles off normal, further eroding the efficiency at higher concentrations. The use of reflecting metal for grid lines can reduce this loss somewhat. Thus, the cell designer must face the increasing challenge of engineering grid patterns that minimize both shadowing and resistance at ever-increasing solar concentration targets. Figure 15.3 plots the typical efficiency of a concentrating solar cell as a function of solar concentration. Efficiencies rise initially primarily owing to the logarithmic dependence of Voc on the operating current. At higher concentrations, grid line resistance and spreading resistance in the semiconductor layers drive I2R losses to the point of overpowering further efficiency gains. Under actual operation, cell

DESIGN OF CONCENTRATING OPTICAL SYSTEM

341

Cell Efficiency vs. Concentration 40

Efficiency at Maximum Power (%)

38 36

25°C

34 32

90°C

30 28 26 24 10

100

1000

Concentration (suns) Figure 15.3. Typical efficiency concentration of an HCPV cell as a function of solar concentration with cell temperature as a parameter. Abstracted from Spectrolab Inc. data sheet.

temperatures also increase at higher concentrations, leading to a bandgap shrinkage of the semiconductor materials and a subsequent fall in Voc, further lowering the efficiency beyond the losses depicted by the 25°C constant temperature curve of Figure 15.3. There is an optimum solar concentration for operating a given HCPV solar cell design. Contemporary cells, however, are little challenged at the 500- to 1000-sun concentrations typically used today. This concentration range produces comparatively low current densities of 5–15 A/cm2 in the cells, a point well below the onset of nonlinear effects. Adequate cell cooling remains a primary concern in power receiver design, both to minimize efficiency loss and to extend cell lifetime. In summary, multijunction solar efficiency is sensitive to (1) solar spectral balance hour to hour and season to season, (2) the spectral transfer function of the optics, (3) the design target solar concentration, (4) the angle of the incident light on the cell, and (5) the cell operating temperature. Finally, the cells must be engineered to withstand the current densities and temperatures encountered at high concentration.

15.3

DESIGN OF CONCENTRATING OPTICAL SYSTEM

Several competing performance factors must be considered when designing any concentrator system. The final design must meet several goals:

• •

provide high-efficiency optical transmission over the critical spectral band; have a compact design;

342

HIGH-CONCENTRATION CASSEGRAINIAN SOLAR CELL

• • • • • •

allow the use of passive heat sinks; be manufacturable with proven, inexpensive technologies; allow reasonable optical tolerances; demonstrate very high reliability; achieve low maintenance costs; and achieve minimum LCOE.

Relative System Cost Effectiveness (MJ flat panel = 1)

The last point governs the majority of decisions in choosing the system design constraints. Because multijunction cell efficiencies peak near 350 suns and roll off slowly through 500 or 600 suns, cell efficiency exerts a strong influence on choosing the concentration ratio for the system. Figure 15.4 plots the relative system cost effectiveness as a function of concentration for three cases: (a, —) constant cell efficiency at constant cell temperature, (b, –䉱–) a passively cooled cell with a maximum efficiency at 400-sun concentration, and (c, ….•….) case b with tracker costs included. The curves are derived for a Cassegrainian concentrator on a caseby-case basis for several discrete concentrations. The relative system costs pass through a minimum between 800 and 1000 suns for case b. When tracker costs are added, the minimum remains near the same concentration while costs rise by more or less a uniform multiplier. Figure 15.4 points to the desirability of designing a system near 1000-sun concentration. Other practical considerations, such as the limited choice of commercially available cell sizes and constraints in manufacturing increasingly large scale optics, further narrow the choices. As emphasized on curve b of Figure 15.4, mitigating cell efficiency losses with effective heat dissipation becomes another imperative constraining the design. A 1-cm2 cell dissipates

Cost Effectiveness of Concentration 0.100

0.010

0.001 0

200

400

600

800

1000

1200

1400

Concentration (suns) Figure 15.4. Cost effectiveness of an HCPV system with (a, —) the cell at constant efficiency and temperature, (b, –䉱–) the cell with passive cooling and efficiency peak at ∼400 suns, and (c, ….•….) case b with tracker costs added.

CONCENTRATING SYSTEM CHOICES

343

roughly 50-W heat while open circuit under highest illumination. This is near the limits for passive heat dissipation. Although the fractional cost per kilowatt hour of the cells and the overall system costs diminish with increasing concentration ratio, cell operating temperature and the demands on the tracking precision also increase, pushing efficiency downward and costs upward. Balancing all of these requirements leads to choosing a 500-sun concentration ratio as the nominal design target [5].

15.4

CONCENTRATING SYSTEM CHOICES

There are four fundamental choices for concentrating optics: refractive (transparent lenses), reflective (mirrored) optical elements, dispersive (holographic, grating, and prismatic elements), and fluorescent (panels). Of these, refractive and reflective designs are the most interesting for commercial high-concentration large-scale systems. Several configurations and variations are possible for each of these. Fresnel lenses are the principal choice for refractive systems due to the weight, low material content, and low internal adsorption compared to spherical or aplanatic lenses. Plastic materials are the choice for solar concentrators due to their low cost and the ready availability of mass-production molding or stamping technologies. Silicone molded on glass substrates to form the Fresnel lens is also an emerging approach. Fresnel lenses, however, suffer from chromatic aberrations, which become more pronounced for the larger aperture lenses needed at higher concentrations. Chromatic aberrations create a local mismatch between the currents of the subcells pulling down the efficiency of multijunction cells. Molding a Fresnel lens can also be problematic. The surface tension and viscosity of molten plastic during the molding process limits the achievable sharpness at the Fresnel grooves. Also, a draft angle must be added to the groove profile to allow mold release limits. The cumulative imperfections in the plastic lenses lead to significant losses. A typical molded Fresnel lens may transmit only 83% of the incident sunlight onto the cell [6]. Plastic materials also tend to warp and droop over time, introducing beam waist enlargement, which spills light off of the target cell aperture. Necessary periodic cleaning is also problematic since plastics scratch easily. Long-term stability of many plastics suitable for Fresnel lenses has not been tested out to the 30-year time frames considered essential for HCPV energy systems. Central receiver reflectors are another popular approach to concentrator systems [7]. These systems employ a single, large reflective dish, constructed from multiple mirror segments, to focus the sunlight on a dense array of solar cells. Central receiver optical systems have high efficiencies but very narrow acceptance angles, and therefore require accurate heliostats. Cosine losses can become significant for large aperture systems with a high angle of incident rays on the solar cells. The solar cells must be actively cooled, since the back-side area, which is to say the area available to conduct heat passively to the environment, is small compared to the entrance aperture. Active cooling consumes system power and adds an

344

HIGH-CONCENTRATION CASSEGRAINIAN SOLAR CELL

additional level of complexity to system construction and maintenance. Because the light in the dense array of cells falls across the entire cell surface, the bus bars of the solar cells are also fully illuminated. The effective amount of shading is therefore larger, impacting overall system efficiency. Finally, a central tower receiver, which malfunctions and points off sun, can focus a highly concentrated region of sunlight well outside the intended system bounds, creating a safety hazard. In recent years, the optical and thermal difficulties of lens and dish concentrators have been overcome through the use of arrays of small-aperture mirrored concentrator systems. Although these systems are slightly less efficient than a single-mirror system, the Cassegrainian two-element lens system confers several advantages on a solar concentrator system: (1) the folded optical path is more compact than a single-element lens (this is true for a given level of aberration; the sole exception is a domed Fresnel lens), achieving the 1:4 theoretical minimum depth-to-width ratio; (2) higher efficiency and concentration than a Fresnel lens; and (3) absence of chromatic aberrations. With proper design, the lens can be made aplanatic. Since the focal point is positioned between the two elements and can be positioned near the apex of the primary element, the cell with its heat sink can be placed at that point. With the cell heat sink proximate to the assembly backpan, efficient thermal dissipation to the air becomes simple. In practice, choosing mirrors also enables a wider choice of materials than do lenses, and the simple shapes of the optical elements lend themselves to straightforward manufacturing techniques. Mirrors may be fabricated from any number of materials including plastics, metals, resins, and glass. Glass, however, is a nearly ideal substrate for the mirrors, owing to its rigidity, durability, and high degree of manufacturing maturity. The technologies for mirroring glass and for making protective and antireflective coatings are also highly mature. A glass lens system supported by sheet metal and glass structural members forms the basis of currently manufactured Cassegrainian concentrator systems. The following sections detail the design and construction of these systems.

15.5

OPTICAL DESIGN OF A CASSEGRAINIAN CONCENTRATOR

Several practical design goals influence the detailed configuration of a Cassegrainian optic for use in concentrating panels. Ease and economy for automated manufacturing, module compactness, passive cooling, and reasonable optical tolerances that facilitate tracker design are foremost. Minimizing secondary mirror shadowing and accurate alignment of the two mirrors during manufacturing form additional constraints. In practice, the latter constraint translates to coplanar mirror mounting. Acceptable secondary mirror shading is chosen to be <4% to retain reasonable system efficiency. Choosing <4% secondary mirror shadowing leads to optical constraints that position the focal point between the primary mirror apex and the secondary mirror. It is also advantageous to place the cell and heat sink at or below the apex of the

OPTICAL DESIGN OF A CASSEGRAINIAN CONCENTRATOR

345

primary mirror to facilitate heat transfer to the backpan. These seemingly impossible, conflicting constraints are overcome by inserting a nonimaging TIR light guide between the focal point and the cell, permitting cell placement remote from the focal point. Concurrently, tailoring the light guide design may either expand the acceptance angle or permit higher concentrations. Tapering the light guide, from a large area at the top to a smaller area at the bottom (conforming to the active cell area), forms an approximation of a field stop at the top (focal point) end. Figure 15.5 compares the focal point position at 0° and 1° pointing error at the entrance to the light guide. The diagrams are taken from ray trace models of the incoming light through the full Cassegrainian optical system. The cell active area position is also indicated in relation to the light guide entrance, with the relative intensity of the light at the focal point indicated by concentric circles. It can be seen that at higher deviation angles, light that would otherwise fall off of the cell is guided back to the active area through TIR at the guide walls. Without the guide, only a much larger cell would intercept off-track rays effectively. Alternately, the same size cell could be employed but with far more stringent demands placed upon the tracking system. The use of a light guide as a tertiary optical element results in lower cell costs, more efficient thermal management, and less onerous tracker requirements. Since the optical focal point corresponds to the top of the light guide, any unintentional defocusing protects the optics by lowering, rather than raising, the optical flux on the guide. In the event of gross pointing errors, the focal point is well above the primary mirror, thereby averting

Spot Diagrams vs. Tracking Error 0° Pointing Error

1° Pointing Error Cell Active Area

Cell Active Area

Light Guide Entrance Area = Field Stop (a)

2 5 7 Relative Light Intensity (b)

Figure 15.5. Modeling of focal point for 0° (a) and 1° (b) pointing error. At 0°, all the light strikes the cell; with no guide, a 1° pointing error moves nearly all the focused light fully off the cell; with the guide, all of the light is reflected back onto the cell active area.

346

HIGH-CONCENTRATION CASSEGRAINIAN SOLAR CELL Single Cassegrainian Concentrator Power Unit Secondary Mirror

Solar Cell

Front Glass

Primary Mirror TIR Light Guide

Figure 15.6. Cross section through the center of the optics for an aplanatic Cassegrainian optical system with tertiary element light guide. The light guide greatly expands the tolerable pointing error.

severe damage to the mirror while maintaining the concentrated flux safely within the panel. The resulting design is shown in Figure 15.6. A comprehensive discussion of the optical design may be found in Gordon [8] and in McDonald et al. [5]. While the form of the optics and the general placement of components are now defined, the working dimensions have not yet been specified. The entrance aperture is determined by starting with the available standard cell areas and considerations of the primary mirror manufacturability. Cells with a 1-cm2 active area are commonly available from commercial suppliers, while the accuracy of the vacuum-assisted slumping process limits the primary mirror size. Mirrors larger than about 30 cm are difficult to produce with the required precision. With a target concentration ratio of 500, other optical losses must be backed out of the calculation to arrive at the true geometric concentration ratio design target. Fresnel losses at the cover plate and light guide account for about 13% loss. Reflection losses in the mirrors amount to about 10%; bulk glass absorption 1%, and a guard band for soiling another 10%. There is an edge exclusion region on the cell to account for mechanical assembly tolerances of the light pipe to the cell, leaving 0.81 cm2 of illuminated active cell area. Allowing another 10% design margin, the effective collecting aperture becomes about 645 cm2, or about 28.7-cm diameter for a circular mirror, or a 25.4 cm–square mirror. Square aperture mirrors are more easily tiled into panels, so the 25.4-cm2 single power unit becomes the building block for designing panels of manageable groups of power units. The entrance dimension of the light guide is determined by considering the sun to be 0.86° (solar and near circumsolar region) and by calculating the aperture needed for a good acceptance angle. A 19-mm2 aperture light guide produces about 1.3° angle of acceptance as illustrated in Figure 15.7. This provides much greater latitude in designing a cost-effective tracker and support frame and in minimizing the solar cell size and cost.

CONSTRUCTION OF A MANUFACTURABLE CASSEGRAINIAN

347

Panel Level Power vs. Tracking Error

Normalized Power

100 80 60 40 20 0 -3.0

-2.0

-1.0

0

1.0

2.0

3.0

Acceptance Angle (deg) Figure 15.7. Measured normalized power striking the cell surface with the Cassegrainian system and light guide depicted in Figure 15.6. The pointing tolerance is ±1.3° at 90% transmittance to the cell.

15.6 CONSTRUCTION OF A MANUFACTURABLE CASSEGRAINIAN CONCENTRATOR PANEL With the basic concentrator element design clearly defined, attention shifts to constructing and fielding large numbers efficiently. To keep the manufacturing manageable, panels combine several power units into a single, modular subarray. This allows automated construction and compact shipping and controls yield loss by dividing the system into smaller pieces. Shipped to the field site for installation, arrays of panels are mounted on large, two-axis trackers capable of ∼8-kW output each. Large numbers of such trackers form the complete solar power generation site. This section will detail the manufacture of power units and panels, and a subsequent section will treat the design and configuration of trackers. The Cassegrainian design of Figure 15.6 forms the basis for SolFocus™ Inc. power receivers. The primary mirror is formed by vacuum slumping sheet glass into a precision mold, followed by wet chemical deposition of the silver reflector. Both techniques have been widely used and enjoy a high level of maturity with concurrently low manufacturing costs. The secondary mirror is formed by hot molding, a method common to making high-volume camera lenses. A highperformance coating on the secondary mirror, which operates at much higher fluxes per unit area than the primary mirror, maximizes reflectivity and minimizes heating. The tertiary optic is formed from high-index, high-transmission glass either by hot molding and polishing or by grinding and polishing. The cells are mounted on insulating ceramic substrates commonly used in high-powered LED, high-current switching circuits, and high-power microwave

348

HIGH-CONCENTRATION CASSEGRAINIAN SOLAR CELL

applications. Wire leads are attached, and the cell and ceramic assembly is bonded to a heat spreading aluminum block. The TIR light guide is attached to the front of the cell, completing the receiver assembly for the power unit, forming a separate receiver subassembly. The modular approach to receiver assembly promotes quality control and higher yield and lends itself to using modern high-speed automated semiconductor assembly processes. Groups of 20 mass-produced receivers are then ready for assembly into weather tight panels. Severe requirements targeted at reliability and ease of manufacturing govern panel design. HCPV systems work best under clear sky conditions, which generally implies hot, dry, remote areas. These locations rule out water cooling and require a short thermal path between the cell and air. Remoteness demands high reliability, so heat pipes, fans, pumps, plumbing, chilling, and heat exchangers are excluded from consideration. Simple convective heat transfer from the panel backpan to air provides unconditionally stable cooling. Enclosing the entire optical system enhances reliability. It also greatly eases cleaning by presenting a flat surface to the outside world. Further, dirt falling on the window blocks a single light path, whereas dirt falling on an exposed mirror blocks two light paths: incident and reflected. Exposed mirrors need to be cleaned at least twice as often. Arrays of small Cassegrainian systems allow for high-volume manufacturing of compact panels, minimizing logistics and shipping overhead. The SolFocus panels illustrated in Figure 15.8 meet these demands. Two major elements, the front glass panel made from a single sheet of flat glass, and the aluminum backpan, stamp drawn from a single sheet of aluminum, double as structural support and weatherproofing for the panel. Two major subassemblies are first constructed, one on the front window and the other on the backpan. The front window single glass panel serves as a mechanical mount and reference plane to hold and align the primary and secondary optical

Figure 15.8. Photograph of a SolFocus Cassegrainian concentrator panel. Central squares over each power unit are the secondary mirror backs suspended on the front glass; light guide housings show as small cones at the vertex of the primary mirrors.

TWO-AXIS CASSEGRAINIAN SOLAR TRACKING SYSTEMS

349

elements. Separately, the receivers’ heat spreader blocks are bolted to the backpan. After wiring the receivers together, the backpan and front glass are mated together. The receiver assemblies protrude through holes at the vertices of the primary mirrors. The front glass panel edges seat into gasketed edges on the backpan, joining and aligning the two subassemblies. Thus, the panel thickness is little more than the minimal thickness of the folded optical path, an important factor in minimizing transportation costs and material usage.

15.7 DESIGN OF TWO-AXIS CASSEGRAINIAN SOLAR TRACKING SYSTEMS The design of a Cassegrainian concentrator system becomes a complex exercise in system optimization. Material choice, cost and durability, weight, shipping volume, and production costs all come into consideration. At the same time, the mechanical tolerance stack up and system flexion must be engineered together with the tracking mechanism to keep the receivers in the full array within their tracking tolerance relative to the solar position. Larger apertures and concentrations demand tighter tolerances on the tracker system. HCPV systems by design have a field of view narrowly defined by the solar disk (∼0.5° diameter) and the near circumsolar region. The energy contained in this field of view is about 85% of the total full sky radiation, amounting to a 15% “concentrator tax” compared to the total energy falling on a flat-plate solar panel. This means that a concentrating system starts from an efficiency deficit compared to flat-plate systems, and must make up the difference with increased performance. The optical system must also point at the sun, a target moving at the rate of 0.25° per minute. Because of the forgiving optical acceptance angle of the Cassegrainian system, a given power unit can tolerate 1.3° of cumulative tracking error in any given direction. The allowable pointing error is significantly decreased from this by the accumulation of other mechanical tolerances. Figure 15.9a shows several angular tolerances that apply to the tracker. These include flexion of the supporting column, αp, tracking tolerance and mechanical backlash in azimuth and elevation, φ and θ, respectively, frame flexure αf, and misalignment of individual panels to the frame, αm. Not shown but equally as important are misalignments due to mechanical tolerances in mirror positioning, receiver positioning, and front glass to backpan dimension. These goals require a tracker frame that is both rigid and light. After accounting for all angular tolerances, and including the inaccuracies and imprecision inherent in the tracking software, the tracker level angular acceptance angle is reduced to 0.75°. The tracker must be rigid enough to withstand wind loading plus changes in gravitational loading as the panels are rotated throughout the day, yet must minimize the materials used to trim shipping and materials costs. The truss frame supporting the panel array is designed to be packed tightly during shipment and is engineered economically to provide high stiffness relative to pointing direction at only the four attachment points of each panel. One such solution, a truss frame, is

350

HIGH-CONCENTRATION CASSEGRAINIAN SOLAR CELL

Figure 15.9. (a) Diagram showing several sources of angular error introduced by the tracker and support frame [9]. Used with kind permission of the authors and of Springer Science & Business Media. (b) Twelve-panel truss frame support (panels not in place). The frame provides rigidity at the pane attachment points with a minimum of materials.

shown in Figure 15.9b. The panels attach to the frame quickly with four nuts, which are also used to true the panels to the frame. Each panel mounts independently, allowing easy replacement should it become necessary. LCOE costs optimize for tracker arrays of between 24 and 40 panels, although certain sites may perform better with more or fewer than this amount. The standard SF-1100S-CPV-28 model from SolFocus utilizes 28 panels. Each panel comprises 20 power units that are wired in series. In one configuration, seven panels are wired in series to produce a string of 140 series-connected cells generating about 360 V at Vmp, and three such strings are connected in parallel. Other configurations may be used depending on local codes. The output voltage is advantageously chosen to match well to cost-effective inverters. The array tracks the sun by following ephemeris equations, augmented by an adaptive calibration routine, rather than using an active sun-tracking feedback loop to avoid potentially losing tracking during cloudy conditions that often present multiple optical targets.

15.8

PANEL PERFORMANCE

The SF-1100S-CPV-28 panel tracker array has performed exceptionally well in the field. Figure 15.10 shows the normalized power output as a function of hour of the day. Nominal power output is 8.4 kW at 850 W/m2 DNI on-sun at 20°C

PANEL PERFORMANCE

351 Peak Power

100

Normalized Power (%)

90 80 Peak Demand

70 60 50 40 30 20 0 7

8

9

10 11 12 13 14 15 16 17 18 19 Time of Day

Figure 15.10. Normalized power output of a single panel throughout a clear day. From just after sunrise to just before sunset, the panel operates at >85% power and sustains output during peak demand hours (arrow) when flat-plate PV solutions have lost significant generating ability.

Average Panel Output (% of Peak)

Average Panel Power Winter 2008-2009 100 75 25 0 0

5

10

15

20

25

30

35

40

45

Day

Figure 15.11. Day-by-day average power output of a Cassegrainian 12-panel array operating in the winter of 2008–2009 near Phoenix, Arizona. Data are recorded every 5 min. Each rounded peak corresponds to a single day; low, ragged, or missing peaks are due to cloudy weather. Nighttime hours are not recorded or plotted. Nonlinearities in the time scale reflect seasonal changes in daylight hours.

ambient, or 24% efficiency under those conditions. At 40°C ambient, the array still produces slightly over 96% of the 20°C power. The arrow in Figure 15.10 indicates the period of peak power demand. Figure 15.11 plots the average panel power output of a small array of 12 panels at a test site near Phoenix, Arizona, over a 47-day period. Data are taken every 5 min during daylight hours from about 30 min after sunrise until 30 min before sunset. Periods of poor weather and clouds show

352

HIGH-CONCENTRATION CASSEGRAINIAN SOLAR CELL

as low power or intermittent output. The DNI during the period shown often exceeded 1000 W/m2 peak during the day. Power is shown normalized to 100% because of the mixed nature of the panel assemblies used in this trial.

15.9

DUAL-CELL CASSEGRAINIAN CONCENTRATORS

The HCPV Cassegrainian optical configuration offers an additional opportunity for increasing collection efficiency. The dual-focus Cassegrainian solar concentrator module concept uses a dichroic secondary mirror to split the solar spectrum into two parts and to direct the IR and near-visible portions of the spectrum to two separate cell locations. The second solar cell is located behind the dichroic secondary as shown in Figure 15.12. This is an important method for harvesting lost energy stemming from the current imbalance present in contemporary seriesconnected monolithic HCPV solar cells described earlier in this chapter, or for utilizing materials and bandgap combinations not otherwise available. The dual-focus Cassegrainian module concept was first described and developed by Fraas and Shifman et al. [10–11]. As shown in Figure 15.12, the first embodiment of this solar concentrator PV module used InGaP/GaAs DJ cells located at the near-visible focus at the center of the primary and GaSb IR solar cells located behind the secondary.

Incoming solar rays (IR and Visible)

GaSb IR cell IR Transparent cold mirror

InGaP/GaAs DJ Cell

Figure 15.12. Dual-focus Cassegrainian PV module with dichroic secondary for use with separate multijunction solar cells.

CASE STUDY

15.10

353

DUAL CASSEGRAINIAN ADVANTAGES

There are two key advantages to this concept: (1) some of the constraints on cell selection are removed, allowing for the use of higher-efficiency, more optimal cells; and (2) two separate cell locations divide the heat load so both cells run cooler at higher-efficiency points. The first advantage is demonstrated by the use of the InGaP/GaAs and GaSb 3J cell set rather than the traditional InGaP/GaAs/Ge monolithic triple-junction cell. Unfortunately, the three materials in the monolithic 3J cell do not have the ideal bandgap energies for an ideal monolithic 3J cell. The Ge bandgap energy is too low, generating excess current compared to the other two junctions. The excess current is wasted as heat. In contrast, the higher bandgap GaSb IR cell generates a higher voltage, and the separate electrical connection extracts all the power produced. Overall module efficiency is thereby increased.

15.11 CASE STUDY: OUTDOOR TEST OF FIRST DUAL-CELL CASSEGRAINIAN SYSTEM The performance improvements for the dual-focus Cassegrainian module design have already been demonstrated. Figure 15.13 shows a photograph of one of the first dual-focus Cassegrainian modules. The primary mirror in this module has dimensions of 25 cm × 25 cm. This prototype solar concentrator PV module used 0.5-cm2 InGaP/GaAs DJ cells fabricated by Spectrolab at the near-visible focus at

Figure 15.13. Photograph of a dual-focus Cassegrainian module in outdoor testing.

354

HIGH-CONCENTRATION CASSEGRAINIAN SOLAR CELL

TABLE 15.1. Performance Summary Packaged Cells at STC

Projected STC with 90% Optical Efficiency

Measure at Operating Temperature (April 28)

Measure Module at STC

DJ cell power (W)

17.4

15.7

14.4

15.1

DJ cell efficiency (%)

31.5

28.4

26.1

27.3

IR cell power (W)

3.64

3.28

2.6

3.1

IR cell efficiency (%)

6.6

5.9

4.7

5.6

Sum power (W)

21

19

17

18.7

Sum efficiency (%)

38.1

34.3

30.8

32.9

NIP DNI = 0.92 kW/m2; area = 600 cm2; input power = 55.2 W.

the center of the primary and 1-cm2 GaSb IR solar cells fabricated by JX Crystals Inc behind the secondary. The results from an outdoor test carried out in April of 2006 are summarized in Table 15.1. Referring to column 3 in Table 15.1, at actual operating temperature, the power produced by the DJ cell is 14.4 W and the power produced by the IR cell is 2.6 W for a combined electrical output power of 17 W. The direct solar intensity reading is 919 W/m2. So for a module area of 600 cm2, the input power is 55.2 W. These numbers translate to a module efficiency of 30.8%. From the Voc readings for the two cells, the individual cell temperatures are also determined. The DJ cell operates at 12.5°C above ambient and the IR cell operates at 30°C above ambient. The DJ cell temperature is really remarkable given that it was operating at a geometric concentration ratio of 1200 suns. While the IR cell operating temperature is acceptable, there is room for improvement in the IR heat sink fin design to decrease that cell’s operating temperature still further.

15.12

POSSIBLE CASSEGRAINIAN PANEL DESIGN

Given the Cassegrainian module above, the next step is to design a full-size panel. Here, a module is defined as a complete set of unique cell and optical components, and a panel is defined as an array of modules. Figure 15.14 shows a Cassegrainian

POSSIBLE CASSEGRAINIAN PANEL DESIGN

355

Figure 15.14. Dual-focus Cassegrainian panel design. This particular panel is a 3 × 6 array with dimensions of 750 mm × 1500 mm, about the same size as a silicon 180-W panel. However, extrapolating from column 2 in Table 15.1, this panel should produce 342 W.

356

HIGH-CONCENTRATION CASSEGRAINIAN SOLAR CELL

panel consisting of a 3 × 6 array of modules. As this figure shows, this panel consists of a metal back sheet with an array of holes where DJ cell packages complete with heat sinks are mounted. The panel also comprises a glass front sheet with holes where IR cell packages complete with heat sinks and the dichroic secondary mirror are mounted. An array of primary mirrors is then mounted on posts extending up from the back sheet. Finally, these three arrays are captured by sidewall aluminum extrusions.

15.13

OPTIMAL CELL SELECTION

The dual-focus Cassegrainian module design actually guarantees a panel performance greater than can be achieved with contemporary monolithic triple-junction cells alone. Higher combined cell efficiencies result by simply combining the GaSb cell as a booster cell along with a monolithic InGaP/GaAs DJ cell. However, even more optimal cell selections are possible. Figure 15.15 will aid in understanding how this works. This figure shows the accumulated current density (Jsc) or, in effect, the photon count, as a function of the cutoff wavelength for an AM1.5D terrestrial solar spectrum. Notice the four plateaus in the accumulated Jsc curve of Figure 15.15. There is a plateau at 35 mA/ cm2 and 0.9 μm corresponding to a water vapor absorption line just below the GaAs band edge. The InGaP and GaAs cells share this current equally. Notice that the fourth plateau at 1.8 μm indicates that there is really no current available in the interval between the GaSb band edge at 1.8 μm and the Ge band edge at 2 μm. This explains why the GaSb cell is a better choice than the Ge cell for a third cell.

Accumulated Jsc 80 70

mA/cm2

60 50 40 30 20 10 0 0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

Wavelength (microns)

Figure 15.15. Graph showing the current available from the terrestrial spectrum as a function of the longest wavelength that a given semiconductor can absorb.

CONCLUSION

357

If a GaSb cell is to be used as the IR low bandgap cell, what are the cell options for the near-visible, light-sensitive cell? To answer this question, first note that it will be desirable to interconnect all of the cells of both types in series. Now referring to Figure 15.15, possible 2J, 3J, and 4J options can be identified. A low-cost 2J system can be built with GaAs(0.95)P(0.05) single-junction cells in the visible receivers and GaSb cells in the IR receivers. The visible cells are series connected and joined in series with the series-connected GaSb cells. This option should allow a panel efficiency of over 30%. A 3J system can be realized with still higher panel efficiency by adding indium in both junctions of an In(0.6)Ga(0.4)P/ Ga(0.9)In(0.1)As DJ cell. A cell array of this type can then be series connected with the GaSb IR cell array for the 3J case. This option should allow a panel efficiency of over 35%. NREL and Emcore have recently announced a 40% inverted metamorphic triple-junction InGaP/GaAs/GaInAs cell [12, 13]. One key feature of this cell is that the lowest bandgap cell is no longer a Ge cell. The lowest cell bandgap can be now hypothetically tuned to 1.1 eV. A cell array using this new monolithic triple-junction cell can now be series connected with the GaSb IR cell array to make a 4J dual-focus Cassegrainian module with an efficiency potentially approaching 40%. A nice feature for both the InGaP/GaInAs-GaSb 3J cell set and for the InGaP/GaAs/GaInAs-GaSb 4J cell set is that the heat load on the GaSb cell will be further reduced relative to the InGaP/GaAs-GaSb case that has already been demonstrated. This favors lower cell operating temperatures and even higher panel performance.

15.14

CONCLUSION

Cassegrainian optics has been employed to produce a concentration PV generation system with high performance and reliability at cost points that are competitive today with other alternative energy sources. HCPV Cassegrainian panel production borrows high-volume, low-cost manufacturing techniques used in other industries, such as automotive manufacturing, architectural fabrication, and semiconductor production. As volumes increase and more efficient multijunction solar cells become available, the LCOE will be driven quickly toward a point competitive with more traditional energy sources. Further advances in configuring the Cassegrainian concentrator for more optimal pairings of dual cells may lead to yet higher conversion efficiencies. Trade-offs between added expenses involved in making a dual-cell system and the benefits derived from it will determine if the LCOE is further reduced. Combined with the rapid advances being made in the high-volume automated manufacturing of Cassegrainian systems, further gains in cell efficiency and optical configuration will make Cassegrainian concentrator systems an attractive and compelling alternative energy source.

358

HIGH-CONCENTRATION CASSEGRAINIAN SOLAR CELL

ACKNOWLEDGMENTS One of us (ML) would like to acknowledge many fruitful discussions and the helpful suggestions of Steve Horne and Mark McDonald, and the support and assistance of the entire SolFocus staff who have worked tirelessly to transform the Cassegrainian system from an innovative design to a commercial product. In particular, Prof. Jeffrey Gordon and Prof. Daniel Feuermann of Ben-Gurion University, Prof. Roland Winston of the University of California at Merced, and the staff at PARC (Xerox Palo Alto Research Center) are acknowledged for their contributions to the fundamental design, configuration, and component choices for the system throughout the early stages of development. ABBREVIATIONS 2J—two junction 3J—three junction 4J—four junction AM—air mass, a measure of the distance light travels through the atmosphere. AM1 refers to the thickness of the atmosphere. AM1.5 refers to 1.5 times the thickness of the atmosphere, corresponding to a sun angle approximately 30° from the horizon. AM1.5—see AM AM1.5d, AM1.5D—see AM; the “d” designation stands for direct, and designates only the light that reaches the solar cell directly from the sun as viewed within the solar disk. Scattered light from angles outside the solar disk is not included. BOS—balance of system DJ—dual junction, the same as 2J DNI—direct normal irradiance Ga(0.9)In(0.1)As—gallium indium arsenide; a III–V compound semiconductor alloy made from the group III elements gallium and indium and from the group V element arsenic. The subscripts indicate the group III composition is 90% gallium and 10% indium. Ga1–xInxAs—gallium indium arsenide; a III–V compound semiconductor alloy made from the group III elements gallium and indium and from the group V element arsenic Ga1–xInxP—gallium indium phosphide; a III–V compound semiconductor alloy made from the group III elements gallium and indium and from the group V element phosphorous GaAs—gallium arsenide, a III–V semiconductor formed from the group III element gallium and from the group V element arsenic GaAs(0.95)P(0.05)—gallium arsenide phosphide, a III–V semiconductor formed from the group III element gallium and from the group V elements arsenic and phosphorous. The subscripts 0.95 and 0.05 indicate the group V composition is 95% arsenic and 5% phosphorous.

REFERENCES

359

GaSb—gallium antimonide, a III–V semiconductor formed from the group III element gallium and from the group V element antimony Ge—the element germanium, a group IV semiconductor HCPV—high-concentration photovoltaic III–V—three–five compound semiconductor, any of a group of semiconductor materials formed from elements in groups III and V of the periodic table Imp—current at the maximum power point of a solar cell In(0.6)Ga(0.4)P—indium gallium phosphide, a III–V semiconductor formed from the group III element indium and from the group V element phosphorous. The subscripts 0.6 and 0.4 indicate the group III composition is 60% indium and 40% gallium. IR—infrared, light at wavelengths too long to be visible to the human eye I2R—current squared times resistance; the electrical power lost for a current of I amperes flowing through a resistor of R ohms Isc—short-circuit current; the current a solar cell sustains when run short circuited at a given solar concentration Iswitch—the current at which a tunnel junction current–voltage characteristic switches to negative differential resistance J—junction Jsc—short-circuit current density; the Isc scaled to the area of the material the current is flowing through LCOE—levelized cost of energy LED—light-emitting diode NIP—normal incident pyrheliometer, an instrument capable of measuring the total power incident from DNI solar radiation NREL—National Renewable Energy Laboratory PV—photovoltaic or solar cell TIR—total internal reflection Vmp—voltage at the maximum power point of a solar cell Voc—open-circuit voltage. The voltage a solar cell develops when run open circuit at a given solar concentration. REFERENCES [1] [2]

[3] [4]

S. R. Kurtz, P. Faine, and J. M. Olson. Modeling of two-junction, series connected tandem solar cells using top-cell thickness as an adjustable parameter. J. Appl. Phys. 68, 1890 (1990). H. Cotal and R. Sherif. The effects of chromatic aberration on the performance of GaInP/GaAs/Ge concentrator solar cells from Fresnel optics. Conference Record of the 31st IEEE Photovoltaics Specialists Conference. Lake Buena Vista, FL, January 3–7, pp. 747–750 (2005). M. Gordon, E. A. Katz, W. Tassew, and D. Feuermann. Photovoltaic hysteresis and its ramifications for concentrator solar cell design and diagnostics. Appl. Phys. Lett. 86, 073508 (2005). C. Algora, V. Diaz, and I. Rey-Stolle. Concentrator III-V solar cells: The influence of the wide-angle cone of light. Conference Record of the Twenty-Ninth IEEE

360

[5]

[6]

[7]

[8] [9] [10]

[11]

[12]

[13]

HIGH-CONCENTRATION CASSEGRAINIAN SOLAR CELL Photovoltaic Specialists Conference. New Orleans, LA, May 19–24, pp. 848–851 (2002). M. McDonald, S. Horne, and G. Conley. Concentrator design to minimize LCOE. In High and Low Concentration for Solar Electric Applications II, M. SymkoDavies, ed., pp. 66490B-01–66490B-11. Proc. SPIE, 6649, SPIE, Bellingham, WA (2007). V. Díaz, J. L. Alvarez, J. Alonso, A. Luque, and C. Mateos. Towards a technology for mass production of very high concentration flat panels. In Proceedings of the 19th European PV Solar Energy Conference, Paris, France, June 7–11, pp. 2086– 2089 (2004). G. S. Kinsey, R. A. Sherif, H. L. Cotal, P. Pien, R. R. King, R. J. Brandt, W. G. Wise, E. L. Labios, K. F. Wan, M. Haddad, J. M. Lacey, C. M. Fetzer, P. Verlinden, J. Lasich, and N. H. Karam. Multijunction solar cells for dense-array concentrators. Proc. of the IEEE 4th World Conf. on Photovoltaic Energy Conversion Vol. 1, Waikoloa, HI, May 7–12, pp. 625–627 (2006). J. M. Gordon. Concentrator optics. In Concentrator Photovoltaics, Chapter 6, A. L. Luque and V. M. Andreev, eds. New York, Springer, LLC (2007). I. Luque-Heredia, J. M. Moreno, P. H. Magalhaes, R. Cervantes, G. Quéméré, and O. Laurent. Concentrator optics. In Concentrator Photovoltaics, Chapter 11, A. L. Luque and V. M. Andreev, eds. New York, Springer-Verlag, LLC (2007). L. M. Fraas, J. E. Avery, H. X. Huang, E. Shifman, K. Edmondson, and R. R. King. Toward 40% and higher multijunction cells in a new Cassegrainian PV Module. Conference Record of the 31st IEEE Photovoltaics Specialists Conference. Lake Buena Vista, FL, January 3–7, pp. 751–753 (2005). L. Fraas, J. Avery, H. Huang, L. Minkin, and E. Shifman. Demonstration of a 33% efficient Cassegranian solar module. In Proc. of the IEEE 4th World Conf. on Photovoltaic Energy Conversion, Vol. 1, Waikoloa, HI, May 7–12, pp. 679–682 (2006). J. F. Geisz, D. J. Friedman, J. S. Ward, A. Duda, W. J. Olavarria, T. E. Moriarty, J. T. Kiehl, M. J. Romero, A. G. Norman, and K. M. Jones. 40.8% efficient inverted triple-junction solar cell with two independent metamorphic junctions. Appl. Phys. Lett. 93, 123505 (2008). R. R. King, D. C. Law, K. M. Edmondson, C. M. Fetzer, G. S. Kinsey, H. Yoon, R. A. Sherif, and N. H. Karam. 40% efficient metamorphic GaInP/GaInAs/Ge multijunction solar cells. Appl. Phys. Lett. 90, 183516 (2007).

16 CONCENTRATOR SOLAR CELL INSTALLATIONS AT THE UNIVERSITY OF NEVADA, LAS VEGAS SURESH SADINENI AND ROBERT BOEHM University of Nevada, Las Vegas

16.1

INTRODUCTION

One of the issues with solar energy is its dilute nature. In a flat-panel PV system (1 sun), the entire collector area is covered with PV cells. In a CPV system, inexpensive optical concentrators are used to focus the sunlight onto a small area, thereby reducing the amount of expensive single-crystal PV cell area required. The goal of any CPV technology is to reduce the overall initial cost of the system and hence the cost of the energy produced by the system. The use of high-efficiency PV cells such as multijunction PV cells for terrestrial power generation would not be possible without CPV technology. Though high-efficiency cells are expensive per unit area compared to 1-sun cells used in flat-panel systems, their application in CPV systems can be cost effective. The overall conversion efficiency of CPV systems can be higher than that of their flat-panel counterparts. There has been a renewed interest in CPV systems due to their cost benefits and conversion efficiencies. In a CPV system, sunlight is concentrated by reflection or refraction or by both reflection and refraction.

16.2

REFLECTIVE CONCENTRATOR

A light ray incident on a smooth mirror surface is reflected as shown in Figure 16.1. As can be seen from the figure, the light ray incident on the surface at angle Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

361

362

CONCENTRATOR SOLAR CELL INSTALLATIONS

θ

θ′

Figure 16.1. Schematic representation of reflection.

C

F

Figure 16.2. Reflective concentration from a curved surface.

θ to the surface normal reflects at angle θ′. According to the law of reflection, the angle of reflection is equal to the angle of incidence, θ′ = θ. Figure 16.2 shows the concentration of the sunlight by a curved mirror surface with radius of curvature R. As can be seen from the figure, the incident rays are concentrated at focal point F. The distance of the focal point from the center of the mirror surface can be calculated from the mirror equation. The mirror equation is given as 1 1 2 + = , p q R where p is the distance of the light source from the center of the mirror, q is the distance of the concentrated image from the center of the mirror, and R is the radius of curvature of the mirrored surface. In a concentrated solar system, the distance from the source of light (the sun) p is much greater than the radius of mirror R. It can be said that p is infinity; therefore, 1/p ≈ 0. This simplifies the mirror equation to 1 2 = . q R

REFRACTIVE CONCENTRATOR

363

R . 2 In actuality, the above equation is an approximation. The mirror surface is really a parabola described by the equation Now, the image distance or focal length f = q =

Y = x2 4 f. The equation for a circle is given by

( y − R )2 + x 2 = R 2 or 2 yR = x 2 + y 2. This reduces to y = x2/2R when y << R and x and f = R/2. Then, the sunlight incident on a curved mirror surface is concentrated at its focal point, which is equal to half of its radius of curvature R.

16.3

REFRACTIVE CONCENTRATOR

Light traveling through a transparent medium is bent at the boundary of that medium leading to another medium as shown in Figure 16.3. This is called refraction. As can be seen from the figure, the light ray traveling through the air enters the glass at angle θ1 (angle of incidence) to the surface normal and is refracted at the top surface to angle θ2 (angle of refraction). The relation between the angle of incidence and the angle of refraction is given as sin θ 2 ν2 , ≡ sin θ1 ν1 where ν1 and ν2 are the speed of light in air and glass, respectively.

θ1

θ2

θ3

θ4

Figure 16.3. Refraction of a light ray passing through a transparent material.

364

CONCENTRATOR SOLAR CELL INSTALLATIONS

Figure 16.4. Refraction of light rays passing through a (a) plano-convex lens and a (b) Fresnel lens.

Refractive index n, a property of any material, can be used to find ν for any material. Refractive index n is the ratio of speed of light in vacuum c, and the speed of light in a medium ν. In a refractive concentrator, sunlight is concentrated as shown in Figure 16.4a. As can be seen in the figure, the light traveling through a plano-convex lens is concentrated at its focal point. The focal length in this case can be calculated approximately from the basic lens equations as follows: n1 n2 n2 − n1 + = , p q R where n1 and n2 are the refractive indexes of air and lens material, respectively. As discussed earlier, p, the light source (the sun), is at infinite distance, and n1 = 1 for air. The equation simplifies to n2 n2 − n1 = . q R n Now, the image distance or focal length f = q = R, where n = n2 is the 1 − n refractive index of the lens material. As noted above, the refraction occurs only at the surface of a lens. Taking advantage of this fact, a thin lens (called a Fresnel lens) was developed and serves the same purpose as a thick lens. Fresnel lenses are being used in many applications as a less expensive and weight-reducing alternative to thick lenses. Fresnel lenses made from acrylic polymers for nonimaging applications such as CPV systems can further reduce the cost of a refractive concentrator system. Concentration of sunlight through a Fresnel lens is shown in Figure 16.4b. As can be seen from the figure, the incident sunlight is bent at the curved segment surfaces and is concentrated at the focal point. Due to the edges at each curved segment, some distor-

HCPV FRESNEL CONCENTRATOR SYSTEM

365

tion is created. But in applications such as solar concentrators, these distortions are negligible compared to the cost advantages.

16.4

CPV SYSTEMS

Various CPV systems have been evolved over the years with concentrations ranging from 2 to 1000 suns. CPV systems with concentrations in the range of 2–5 suns are generally referred to as low-concentration systems. Systems with concentrations of 100 suns or more are referred to as high-concentration systems. Currently, low-concentration systems use 1-sun silicon cells with modified grid contacts. These current 1-sun silicon cells have efficiencies lower than 20%. The high concentration PV (HCPV) systems use either high-efficiency silicon cells with efficiencies greater than 20% or multijunction cells with efficiencies close to 40%. CPV systems can also be divided as point concentration systems and line concentration systems depending on the type of concentration. In a line concentration system, the sunlight is focused onto a series of cells arranged in lines; in a point concentration system, the sunlight is focused onto a small area of cell(s). While some CPV system designs do not require sun tracking, CPV systems can better capture solar radiation by tracking the sun throughout the day. Generally, line concentrating CPV systems require one-axis tracking, making the systems less complicated and requiring less structural materials. The point concentration systems, on the other hand, require two-axis tracking, using more complicated and robust structural supports. Hence, it is a trade-off between the cost of trackers and savings in the PV cell area. Different CPV system designs take advantage of one compared to the other. Different types of CPV systems are installed in the United States. Reputable CPV system manufacturers are (1) Amonix Inc., (2) Solar Systems Australia, (3) SolFocus Inc., (4) Entech Solar Inc., and (5) JX Crystals Inc. The CPV systems from the first three manufacturers in the list are point concentration type; the latter two systems are line concentration type. Entech systems and JX Crystals are discussed in Chapter 12. The Amonix systems are discussed in the following sections. As previously discussed, concentration of light can be achieved either by reflection or by refraction. The systems developed by Amonix Inc. are point concentration systems with two-axis tracking. The important features and performance details of Amonix systems are discussed below.

16.5

HCPV FRESNEL CONCENTRATOR SYSTEM

HCPV systems are used for large-scale commercial, industrial, and utility-scale electricity generation. Amonix HCPV systems, manufactured in the United States, employ Fresnel lenses with two-axis tracking for concentrating sunlight. Amonix Inc. has made several major design improvements since the inception of their

366

CONCENTRATOR SOLAR CELL INSTALLATIONS

Figure 16.5. Sunlight concentration in an Amonix system by a Fresnel lens and a secondary optical element.

original system. The important features of fifth- and sixth-generation systems and their field performance are discussed in the following sections.

16.5.1

High-Efficiency Silicon Cell HCPV Systems

Silicon solar cells with 26% efficiency at 250-sun concentration and at 20°C are used with Fresnel concentrators. Concentration of sunlight by a Fresnel lens onto a small area is shown in Figure 16.5. For example, in an Amonix system, the sunlight is concentrated by a 7 in. × 7 in. (17.8 cm) nonimaging Fresnel lens (focal length of 21 in. or 53.3 cm) onto a PV cell with an area less than 1 cm2, achieving a 500-sun concentration ratio. A secondary optical element placed on each cell refocuses any beam dispersion. Descriptions of various components of this kind of CPV system are given in the following sections.

16.5.2

System Description

16.5.2.1 HCPV Module Four HCPV cells are attached to a patented design metal strip. These cells are connected in series (referred to as a string) through a bypass diode at each cell to isolate any malfunctioning cell. Six strips are mounted on an aluminum plate with passive heat sinks mounted on the back side of the plate. The waste heat from the cells is transferred to the ambient air by the heat sink. A set of 24 Fresnel lenses is manufactured as a single unit to match the 24 cells on each plate. A boxlike structure (shown in Fig. 16.6) separates, aligns, and secures the Fresnel lenses and cells in place. Several of these are attached to form one large module, referred to by Amonix as a MegaModule (shown in Fig. 16.6). Specifications of the module and a system with five modules are given in Table 16.1. These modules with high-efficiency silicon cells can produce 5 kW of peak power at 850 W/m2 DNI radiation and at 20°C. The modules require a sun tracking

HCPV FRESNEL CONCENTRATOR SYSTEM

367

MegamoduleTM (5 kW)

Mounting points

Fresnel lens (7” by 7”)

Fresnel lens parquet (24 lenses) Solar cells

Receiver plate (24 cells) Heat sink Receiver strip

Figure 16.6. Important components of a MegaModule™ [1].

TABLE16.1. Characteristics of a Fifth-Generation Amonix System MegaModule

System (Five MegaModules)

Rated power output

5 kW at 850 W/m2 DNI, 25°C

25 kW at 850 W/m2 DNI, 25°C

Width × length × depth

11.0 ft × 44.0 ft × 2.5 ft

55.0 ft × 44.0 ft × 2.5 ft

Weight

5528 lb

27,640 lb

Number of cells

1152

5760

Aperture area of lens

392 ft2/36.4 m2

1960 ft2/182 m2

Operating voltage

—

277/480 V AC

Maximum operating wind speed

—

30 mph

Overall system efficiency

—

18% DC or 16% AC

accuracy of less than 0.1°. Five or seven of these MegaModules are mounted on a tracker. 16.5.2.2 Tracker Structure A two-axis tracking structure is shown in Figure 16.7. It tracks the sun in azimuth and elevation. The tracker structure consists of a foundation, a pedestal, a rotating bearing head, a torque tube, and a tracker controller. A gearless hydraulic drive system is used with the tracker (not shown in the

368

CONCENTRATOR SOLAR CELL INSTALLATIONS

Figure 16.7. Installation of a two-axis tracker and mounting the modules.

pictures). The drive system applies hydraulic pressure to one side of the hydraulic actuator, positioning the torque tube such that the module surface is always normal to the sunlight. The elevation drive is a single hydraulic linear actuator, and the azimuth drive is composed of two hydraulic linear actuators and an offset cam. The two azimuth actuators apply force to a cam to achieve the commanded position. A single hydraulic pump can be used to pump the fluid from the reservoir into a pressure vessel. The pump operates only when pressure in the reservoir drops below the set lower limit and turns off as the pressure in the reservoir reaches the preset upper limit. Elevation rotation is from 10° to 95°, and the azimuth is a full 360°. The tracker is designed to survive a 90-mph wind speed and to operate up to an average wind speed of 29 mph. The arrays can move from any position to stow position in less than 15 s. 16.5.2.3 Tracker Controller A completely autonomous electronic controller that can be monitored or controlled remotely is used with these systems. It has a built-in GPS to obtain universal time. The controller calculates the sun’s position using the universal time and the array position. When the sun’s elevation reaches a set value in the morning, the controller aligns the solar array with the sun and maintains this alignment over the course of the day. The controller will continue to track the sun even when there is heavy cloud cover. When the sun’s elevation reaches the lower

HCPV FRESNEL CONCENTRATOR SYSTEM

369

preset value, the array is moved to a night stow position. The control system also has a sun sensor, which provides signals to the controller when the sun image moves beyond a threshold limit. A position error is determined by comparing the actual position to the calculated position. When this error is above a preset threshold value, the electronic controller activates the drives to move the array in an effort to reduce the error. An anemometer is mounted on these systems. When the wind speed reaches a preset value (30 mph), the controller moves the arrays into a face-up wind stow position. When the wind speed drops below the preset value, the controller moves the arrays back to sun tracking mode. 16.5.2.4 AC/DC Controller/Inverter AC/DC controller combines the DC power within the system modules, converts it to AC power, and interfaces with the AC grid. It consists of DC fuses, circuit breakers, and an inverter. When the power generation from the system reaches the designed optimal levels, the controller connects the inverter to the grid and disconnects as the power generation decreases below designed levels. 16.5.3

Field Performance

Three fifth-generation HCPV Amonix systems are installed at NV Energy’s Clark Station in Las Vegas, Nevada (NV Energy is the primary electric utility in Nevada). The three systems are grid connected and have been operating since March 2006. The peak power of these systems is 25 kW. Two of these systems are shown in Figure 16.8. The specifications of the system are given in Table 16.1. The system uses

Figure 16.8. A five-MegaModule Amonix system installed at NV Energy’s Clark Station in Las Vegas, Nevada.

370

CONCENTRATOR SOLAR CELL INSTALLATIONS

high-concentration rear-junction silicon solar cells with approximately 26% efficiency at 250-sun concentration, 850 W/m2 DNI, and 20°C. The total surface area of the system is 2475 ft2/224 m2. Several of these systems operating in Spain are shown in Figure 16.9. Field performance of the system for a winter day is shown in Figure 16.10. For this particular winter day, the system peaked at over 24-kWAC power output.

24

1000

20

800 Direct Normal Power Output

600

16 12

400 8 200 0

Power Output (kW)

Direct Normal Incidence (W/m2)

Figure 16.9. A field of Amonix systems installed in Spain.

4

8

9

10

11

12

13

14

15

16

0

Time of the Day (hr)

Figure 16.10. Variation of power output from a fifth-generation Amonix system in Las Vegas on January 15, 2007.

HCPV FRESNEL CONCENTRATOR SYSTEM

371

As can be seen from the figure, the power produced from the system increases as the solar radiation increases. This is a typical performance profile of a two-axis tracking system. During the early morning and late afternoon, when solar radiation is low (when the power generated is lower than a set value), the inverter is automatically disconnected from the grid. Las Vegas, Nevada, is located in the Southwest region of the United States. This region is well known for one of the best solar resources in the world. Additionally, climate in this region is arid. Hence, the direct beam radiation is very high in the region, making it well suited for CPV applications. A peak direct beam radiation (DNI) of 1050 W/m2 is commonly recorded in the region during summer. The averaged direct beam solar radiation available for concentrating two-axis tracking systems installed at Las Vegas is 7.1 kWh/m2/day [2]. Figure 16.11 shows the energy generated by the system each month in 2007. As can be seen from the figure, the monthly energy generation increased during summer and decreased during winter. The general trend of the system performance is as expected. Another measure of the long-term system performance of a PV system is the MEPF, with units of (kWh/m2 produced)/((kWh/m2 predicted) at the array kW rating). The MEPF is the ratio of the total monthly energy generated (kilowatt per hour) by the system, and the product of the total direct normal radiation (kilowatt per hour per square meter) and the rated power level of the system (kilowatt). As can be seen from Figure 16.12, the MEPF is high during winter and is low during summer. One of the reasons for this variation is the seasonal changes in ambient temperature. Ambient temperatures in the Las Vegas summer can be

6000

Energy Generated (kWh)

5000 4000 3000 2000 1000

Dec-07

Nov-07

Oct-07

Sep-07

Aug-07

Jul-07

Jun-07

May-07

Apr-07

Mar-07

Feb-07

Jan-07

0

Month

Figure 16.11. Monthly energy generated by a 25-kW system installed at Las Vegas, Nevada.

372

CONCENTRATOR SOLAR CELL INSTALLATIONS 1.0

MEPF kWh/(kWh/m2/kW))

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Dec-07

Nov-07

Oct-07

Sep-07

Aug-07

Jul-07

Jun-07

May-07

Apr-07

Mar-07

Feb-07

Jan-07

0.0

Month

Figure 16.12. Monthly energy performance factor (MEPF) of the system.

140000 Unit 1

Unit 2

Unit 3

Net Generated Energy (kWh)

120000 100000 80000 60000 40000 20000

Ju l-0 Se 6 p06 D ec -0 6 M ar -0 7 Ju l-0 7 Se p07 D ec -0 7 M ar -0 8 Ju l-0 8 Se p08

M

ar -0

6

0

Month

Figure 16.13. Accumulated energy generated by the three systems between March 2006 and October 2008 [3].

MULTIJUNCTION CELL HCPV SYSTEM

373

TABLE16.2. Performance of the Three Amonix Units Installed at Las Vegas, from March 2006 to October 2008 Unit 1 Total energy generated (kWh)

Unit 2

Unit 3

116,516

130,746

117,223

Total generating time (h)

7,232

7,353

7,014

Total tracking time (h)

8,029

8,173

7,765

247

261

260

Peak daily energy (kWh) Peak power (kW)

25.97

25.32

25.4

very high and have been recorded to be as high as 50°C. Since the systems are passively cooled, there is a significant influence by the ambient temperature on the system output. The efficiency of PV cells is strongly dependent on their temperatures—at higher temperatures, cell efficiency is lower; at lower temperatures, cell efficiency is higher. More discussion of temperature effects on cell efficiency can be found in Chapter 13. Although the cell efficiency is lower during summer, the total energy generated in kilowatt hour during a typical summer day is greater than the energy generated during a typical winter day (as is implied by Fig. 16.11). This is due to the longer period of solar radiation during summer days as compared with winter days. The accumulated energy output of the three systems installed at the Clark Station, Las Vegas, can be seen in Figure 16.13. The three systems have generated over 365 MWh of electrical energy since their installation in March 2006. Some performance characteristics of these systems are also given in Table 16.2.

16.6

MULTIJUNCTION CELL HCPV SYSTEM

As discussed in Chapter 10, due to the economical advantage of high concentrations, expensive high-performance PV cells can be used in HCPV systems for greater overall conversion efficiencies. While previous Amonix systems used monocrystalline silicon cells with efficiencies less than 27%, the latest generation systems utilize multijunction cells with efficiencies close to 37%. At higher cell efficiencies, the land required can be significantly reduced. It is estimated that these systems require 4–6 ac/MW of installed capacity. The systems with multijunction cells are labeled by Amonix as sixth-generation systems. In the seventh-generation systems, the concentrations are also increased to 500 suns. The physical characteristics of a five-MegaModule seventh-generation system are given in Table 16.3. Designs were also improved to reduce the use of structural materials in a module, thus reducing the weight of the modules, which allowed for higher generation capacities for each tracker. Some of these systems were also installed with seven modules on one tracker.

374

CONCENTRATOR SOLAR CELL INSTALLATIONS

TABLE16.3. Physical Characteristics of a Seventh-Generation Amonix System with Five MegaModules Attribute Physical

Environmental

Solar cells Performance

Units

Value

Dimensions

H × W (ft)

48.0 × 51.4

Maximum installed height (aboveground)

ft

50

Typical pedestal installation depth

ft

18

Tracking type

Two-axis

Tracker drive

Hydraulic

Solar tracking method

GPS and sun sensing

Zone (ASCE 07-05/IBC/ UBC/CBC)

Wind exposure D—max 90-mph fast mile

Maximum operating wind speed

mph

40

Stowage time

s

15

Maximum wind speed in stowage configuration

mph

>90

Temperature range

°C

–40 to +55

Life expectation

year

40

Solar cell type

Multijunction

Solar cell efficiency

%

37

Nominal rated power output

kWAC

38

System efficiency (AC, post-inverter)

%

25%

Annual capacity factor

%

30%

Annual electricity generation

kWh

103,744

Land usage

ac/MW

4–6

The performance of a five-MegaModule sixth-generation system installed at the University of Nevada, Las Vegas over a day is shown in Figure 16.14. As can be seen from the figure, the power produced and efficiency are relative to the solar insolation. The post-inverter power delivered by the system is over 45 kWac due to low ambient temperatures during the day. The overall system efficiency was steady at 25% except in the early morning and in the late afternoon. This means that the

SUMMARY

375 50

1200

40 DNI (W/m2) Power (kW) AC Efficiency

800 600

35 30 25 20

400

15 10

200

Power Output (kW)/Efficiency

Direct Normal Incidence (W/m2)

45 1000

5 0 7:12

8:24

0 9:36 10:48 12:00 13:12 14:24 15:36 16:48 Time of the day( hr)

Figure 16.14. Variation of power output from a seventh-generation Amonix system in Las Vegas on February 3, 2009.

system is capable of converting 25% of the incident solar radiation on the system collector area to AC electrical power, which is a significant conversion for any PV system.

16.7

SUMMARY

Although all solar radiation reaching the earth is in the form of beam radiation, as it passes through the earth’s atmospheric cover, a part of it is scattered. The part of solar radiation that directly reaches the earth is called “beam” or “direct” radiation, and the fraction of the scattered radiation that reaches the earth is called “diffuse” radiation. With an exception of some line concentrating PV systems, most of the CPV systems can use only the beam part of the solar radiation, because only the beam radiation can be concentrated. Hence, the concentrator systems are well suited for places where the beam radiation is a significant portion of the total radiation. For example, in the Desert Southwest region of the United States, the beam radiation is 80–90% of the total radiation for most of the year. The CPV systems as demonstrated and described in this chapter can be good alternatives to bring down the cost of solar power generation. Significant improvements in the CPV cells to increase the efficiency and to reduce the cost can greatly change the solar PV economics. Many CPV systems have been installed over the recent years. Experience gained through these systems can make significant improvements in the system designs to reduce the structural material use and to improve the long-term reliability of the systems. Finally, CPV technology can be an economical alternative to fossil fuel-based power generation in the near future.

376

CONCENTRATOR SOLAR CELL INSTALLATIONS

ABBREVIATIONS AC—alternating current c—speed of light in vacuum CPV—concentrated photovoltaic DC—direct current DNI—direct normal incidence F—focal point f—focal length FF—fill factor GPS—global positioning system HCPV—high-concentration photovoltaic Isc—short-circuit current MEPF—monthly energy performance factor n—refractive index p—distance of the light source from the center of the mirror PV—photovoltaic or solar cell q—distance of the concentrated image from the center of the mirror R—radius of curvature θ—angle ν—speed of light in air or glass

REFERENCES [1] [2] [3]

K. Stone, V. Garboushian, D. Roubideaux, R. Gorden, J. Turner, and D. Dutra. Design and performance of the Amonix High Concentration Solar PV System. ASES/ASME National Solar Energy Conference, Reno, NV, June 15–20 (2002). National Renewable Energy Laboratory (NREL). http://rredc.nrel.gov/solar/old_ data/nsrdb/redbook/sum2/state.html (2009). A. Sahm, R. Boehm, K. Stone, and K. Johnson. Performance of the Amonix high concentration photovoltaic systems at the NV Energy Clark Station. Proceedings of the Inaugural US-EU-China Thermophysics Conference, May 28–30, Beijing, China (2009).

17 CONCENTRATOR PHOTOVOLTAIC FIELD INSTALLATIONS FRANCISCA RUBIO, MARÍA MARTÍNEZ, AND PEDRO BANDA Instituto de Sistemas Fotovoltaicos de Concentración (ISFOC) S.A.

17.1

INTRODUCTION

The idea for decreasing the cost of PV systems using optical elements that focus sunlight onto cells to reduce cell size has been on the minds of scientists since the 1970s [1]. The key is to replace the most expensive active material area with wellunderstood and lower-cost optical elements. Unfortunately, initial markets did not encourage installation of CPV plants. Consequently, very few were constructed in the world until 2005. Module prototypes and some demonstrations of a few kilowatts were installed in the field by some universities (IES) in Spain, in research institutes (Fraunhofer in Germany and APS, Sandia National Laboratories, and NREL in the United States), and by very few research and development companies (e.g., Amonix, Entech, Solar System) [2]. For example, IES developed various projects such as Ramón Areces Miner and the EUCLIDES. A photograph of the Ramón Areces array is shown in Figure 14.1 of Chapter 14. The EUCLIDES array was installed in Madrid in 1995 with trackers, mirrors, and thermal heat sinks designed in the IES-UPM. After the success of the EUCLIDES array, a demonstration plant of 480 kWpyc was installed in Tenerife (1998). Until 2003, this plant (Fig. 17.1) was the biggest in the world using concentrated sunlight. Newer versions and optimizations of the original EUCLIDES design have been developed and put in place in the last few years [3]. Another pioneering installation was the SOLERAS project in Saudi Arabia, installed by Sandia National Laboratories in the 1980s. The system showed

Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

377

378

CONCENTRATOR PHOTOVOLTAIC FIELD INSTALLATIONS

Figure 17.1. EUCLIDES plant of 480 kW in Tenerife (Spain) (1995).

degradation of up to 20% in 6 years, mainly due to delaminating issues at high temperature. Despite these problems, this project continued in operation for 18 years. Apart from these examples, there were very few demonstration plants installed in the world during this period. However, in 2006, special feed-in tariff schemes for PV systems were established in some countries of Europe, especially in Spain, and this technology could finally find its way to profitability. At this important point in time, the ISFOC was created to support the first steps of moving CPV from prototyping to technology industrialization [4]. ISFOC was established through an agreement between the Science and Innovation Ministry of Spain and the government of Castilla-La Mancha. The main goal of ISFOC is to support the growth and to accelerate the industrialization and commercialization of the most advanced CPV technologies through the installation of various CPV pilot power plants. Since its establishment, ISFOC has executed a number of power plants (up to 3 MW in the first phase) incorporating various concentrator technologies. The objective of these pilot plants is to assist industries in setting up pilot fabrication lines and to help develop industry standards. Also, very valuable information was expected to be obtained in the process such as reliability, suitability, and producibility for each technology. As a result, ISFOC has become the leading national and international reference center for CPV.

17.2

PLANT INSTALLATION PROCEDURE

The main objective of ISFOC is to help the development of the CPV industry through the installation of demonstration plants. To carry out these first installations, the following tendering procedure was used.

PLANT INSTALLATION PROCEDURE

379

Figure 17.2. Technologies selected by ISFOC to be installed in Castilla-La Mancha (Spain).

Two international calls for tenders were launched, one in 2006 and a second one in 2007. ISFOC’s International Scientific Advisory Committee was instrumental in the selection of the best proposals. In the first call, Concentrix (Germany), SolFocus (United States), and Isofoton (Spain) were selected. In the second call for tenders, Arima ECO (Taiwan), EMCORE (United States), Sol3G (Spain), and Renovalia CPV (Spain) were chosen as summarized in Figure 17.2. Prior to installations of the systems, the CPV manufacturers had passed two technical steps. The first step was to pass the tests of the IEC 62108 standard. In the first call for tenders, the manufacturers were allowed to pass only some of the tests (as it was still in a draft status). In the second call, all of the approved and then standard tests had to be passed. In addition, the manufacturers had to assemble a whole concentrator and to pass a prototype test as an additional second call requirement. In addition to passing technical steps, there were also a number of administrative and practical procedures that both manufacturers and ISFOC needed to follow to start the installations. An engineering project document including all drawings and technical specifications was required from the manufacturers. Required permits included local licenses, government authorizations, and utility grid connection permits. The actual installations could then start once all these requirements were completed and the permits finished. Once installation was completed, the acceptance process started. ISFOC both defined and performed all the necessary acceptance tests (both in DC and AC).

380

CONCENTRATOR PHOTOVOLTAIC FIELD INSTALLATIONS

The acceptance procedure provides key information both for the manufacturers to evaluate and to further improve their design and for ISFOC to be able to properly place the power plant into production. The study of the power plants under production conditions is the key tool ISFOC uses to further evaluate the technology and to provide key information to the scientific and technical community. There are a number of project metrics determined not only for power production characteristics but also for solar resource, soiling effects, O&M operations, and so on. A description of the various procedures for both installation and acceptance of CPV power plants is presented below.

17.3

IEC 62108 TESTS

Prior to installation, CPV systems need to be qualified. The international qualification standard IEC 62108 [5] was not fully approved until December 2007. Before this date, only a few CPV manufacturers, including the partners of the ISFOC project, had carried out some of the tests of the last draft of the IEC standard. Presently, there are a number of laboratories in the process of accreditation for performing these tests. In June 2009, SolFocus announced its certification under this standard with the PTL TÜV Rheinland PTL in Arizona. The IEC 62108 consists of different types of tests:

•

•

Electrical tests: 䊊 Electrical Performance Measurement. Measurement of the I–V curve before and after the tests to identify the degradation. 䊊 Ground Path Continuity Test. Test to verify adequate electrical continuity between all exposed conductive parts and the grounding point under high-current conditions, to be measured before and after the tests. 䊊 Electrical Insulation Test. Test to determine insulation between all active parts and the frame or the outside world, to be measured before and after the tests. 䊊 Wet Insulation Test. Test to verify the insulation of the concentrator system under wet operating conditions and to verify that moisture does not enter the active parts of the sample circuitry. 䊊 Bypass/Blocking Diode Thermal Test. Test to assess the adequacy of the thermal design and relative long-term reliability of bypass/blocking diodes. 䊊 Hot-Spot Endurance Test. Test to evaluate the ability of a module to endure a hot-spot heating. This test does not need to be carried out if there is a bypass diode for the cell. Climate tests: 䊊 Thermal Cycling Test. Test to determine the ability of the receivers to withstand thermal mismatch. The receivers should follow a very aggressive thermal cycle, between −40 and 65°, 85°C, or 110°C. The

IEC 62108 TESTS

•

•

381

choice of lower temperatures will increase the number of cycles. At temperatures higher than 25°C, a 1.25 Isc should be applied into the cells. 䊊 Damp Heat Test. Test to determine the ability of the modules to withstand the effects of long-term penetration of humidity. The modules should remain 1000 h at 85% and 85°C; if the modules cannot survive at 85°C, the test could be performed at 65°C for 2000 h. 䊊 Humidity Freeze Test. Test to determine the ability of the modules to withstand the effects of thermal cycles and humidity. The modules should follow cycles of temperature between −40 and 85°C with 85% humidity, after a prethermal cycle. With 65°C, the number of cycles will be higher. Mechanical tests: 䊊 Hail Impact Test. Test to determine if the module can survive a hailstorm. Ice balls will be launched at the module. 䊊 Robustness of Terminations Test. Test to determine whether the terminations of the module will withstand stresses. 䊊 Mechanical Load Test. Applying a load to the module to verify the resistance. Ambient conditions: 䊊 Water Spray Test. Test to determine whether rain water can enter the module and if the entered water can cause a ground fault or a safety hazard. 䊊 Off-Axis Beam Damage Test. Test to evaluate that no part of the module could be damaged by concentrated solar radiation during conditions of misalignment. 䊊 UV Conditioning. Test to reveal possible premature failures due to limited UV exposure. The UV dosage is 50 kWh/m2. It can be conducted either outdoors or in a testing laboratory. 䊊 Outdoor Exposure Test. Test to make a preliminary assessment of the ability of the module to withstand outdoor conditions that may not be detected by laboratory tests. It should be a cumulative DNI 1000 kWh/m2 at a DNI higher than 600 W/m2.

ISFOC’s experience shows that as the standard procedures were being performed, improvements of design were carried out, including modifications to comply the IEC tests. The benefit of the standard is, therefore, proven. The isolation test and the damp heat test are the most difficult ones, as the modules need to show water tightness and/or electrical isolation. The problems found during the tests are typically related to

• •

Materials. Some type of plastic for the lens cannot pass the thermal and/ or humidity cycles. The seal can have similar problems. Cells. An initial control and good packaging can avoid infant mortality problems.

382

CONCENTRATOR PHOTOVOLTAIC FIELD INSTALLATIONS

•

Water tightness and isolation are related and should be improved at the same time. The way the ventilation is implemented, either active or passive, is crucial in order not to have problems in the damp heat test and in the field.

From our experience on all tests performed on our partners’ technologies, we recommend that some of the tests should be further stressed in order to identify possible material and design issues. For example, the off-axis tests should be extended in time as we have observed damage to some systems in the field operating with a minimum deviation from the tracking angle. Other tests could also be included in the procedures in the future (like salt, fog, dust deposits, and sand abrasion) in order to provide qualification for systems operating under those conditions in the field, which correspond with potential areas for CPV deployment (e.g., desert areas close to the sea).

17.4

PROTOTYPE TESTS

As part of the evaluation of the CPV technologies incorporated in the ISFOC project, a full characterization of a prototype of each is completed to demonstrate its performance. To carry out these measurements, a procedure for CPV prototype characterization has been established [6]. The main objective of the test was to verify that the DC power of the prototype, which is measured following an internal procedure further described in this chapter, is within the limits of the nominal power provided by the manufacturer. Additionally, this procedure also includes testing of the tracking system:

• • • •

17.5

manual movement function, to verify maintenance operation of the tracker system; function after power cutoff, to verify if the tracker returns to the right position after any electrical problem; function after misaiming, to verify if the tracker returns to the right position if there is a misaiming problem due to clouding, manual movement, and so on; and tracking influence on the nominal DC power; the DC power is measured during the tracking of the system to verify whether the tracker has some influence on the concentrator performance.

PLANT ACCEPTANCE

ISFOC has developed an acceptance procedure to verify the complete system of the CPV power plants, once the installation of the plants has finished [7] The acceptance procedure is organized in five steps: documental review, technical inspection, control and monitoring, and DC and AC power ratings.

RATING STANDARD

383

The documental review is carried out once the construction of the plant is finished. For the construction, according to Spanish regulations, it is necessary to have a detailed project, including health and safety studies and environmental impact studies. After the construction, the contractors have to provide the necessary documents to connect the plant to the grid: the list of components, the warranty certificates, O&M book, record of health and safety incidents, “as-built” documentation, “finished works,” and “low-voltage-form” certificates. The second step is to verify all the components of the installation: modules, trackers, fuses, LV boxes, cables, grounding systems, and so on. The installation should be inspected and all the defects repaired before any functional test is performed. After that, the control and monitoring software is evaluated; the trackers are then activated; and data communication and acquisition from all components are verified. It is important to verify the control of the plant with the real-time data but also the data storage, status, alarms, and events of all the components. After this verification, the plants should be measured with the DC and AC ratings and then should be finally accepted.

17.6

RATING STANDARD

Currently, there is no official procedure for rating the electrical performance of a CPV system. In fact, the IEC TC82 Working Group 7 (WG7) is the group of the IEC Committee now working on this task. The WG7 decided recently (June 2009) to launch a draft of the standard to the IEC Committee to perform the power rating of a CPV module. This draft is based on the draft of the standard 61853-1 for flat PV modules [8]. The purpose of this standard is to define a testing and rating procedure, which provides the PV module power (watts) at maximum power operation for a set of defined conditions. Since power depends on the radiation and temperature levels, a table with the different levels of radiation and temperature must be completed as shown in Table 17.1. This table should be adapted to CPV rating conditions, where the radiation and the temperature will be established in narrower values, as, for example, in Table 17.2. Additionally, other technical specifications are used today to measure the power of a power plant [9]. One is the American Standard Test Method for “Rating Electrical Performance of Concentrator Terrestrial Photovoltaic Modules and Systems under Natural Sunlight,” ASTM E 2527-06. This regression method has been used in the United States for several years. ISFOC’s procedure for concentrator rating [10] is used by ISFOC for the acceptance of the CPV power plants [11]. It is based on the equations of the Shockley model adapted to multijunction cells, and only one measurement of the power of the concentrator in operation conditions is needed to establish the nominal power of the CPV system in STCs.

384

CONCENTRATOR PHOTOVOLTAIC FIELD INSTALLATIONS

TABLE 17.1. Set of Parameters for IEC 61853-1 Irradiance

Air Mass

Module Temperature (°C) 15

1100

AM1.5

1000

AM1.5

800

AM1.5

600

AM1.5

400

AM1.5

200

AM1.5

25

50

75

TABLE 17.2. Array Power Output in Watts for a Set of Values of Radiation (W/m2) and Temperature (°C) Module Temperature (°C)

DNI (W/m2)

45

50

55

60

65

700

4682.32

750

—

4595.53

—

—

—

5457.54

5412.90

—

—

800

—

—

5790.69

5640.25

—

850

—

—

6133.74

5978.84

—

900

—

—

—

6441.65

6301.33

950

—

—

—

6718.48

6570.15

The STCs used by ISFOC are 850 W/m2 of the DNI and 60°C of the operating cell temperature. To obtain reliable results, the DNI has to be higher than 700 W/ m2, assuming that the sky is clear and with no clouds around the sun and the wind speed has to be lower than 3.3 m/s. This last limitation is due to the thermal stabilization of the concentrator. In this method, the operating cell temperature and the DNI is used to translate the power from the operation conditions to the STCs. To calculate the cell temperature, the backplate temperature is used together with the internal thermal resistance between the cell and the backplate. The only measurements needed for this procedure are the I–V curve of the system, the DNI, and the backplate temperature behind a cell. By applying Equation 17.1, we can obtain the operating cell temperature from the measured backplate temperature:

RATING STANDARD

385

Tcell = Tbackplate + B ⋅ Rcell-backplate , where

(17.1)

Tcell = operating cell temperature calculated, Tbackplate = temperature measured at the backplate of the module behind the cell, B = measured DNI, and Rcell-backplate = internal thermal resistance between the cell and the backplate of the module.

The current and voltage values of the I–V curve measured in the operation conditions are then translated to as-defined STCs following Equations 17.2 and 17.3, deduced from the usual Shockley equation model: I STC = I

BSTC and B

0.0257 ⋅ (TSTC − Tcell ) ⎛ ( I L1 − I ) ⋅ ( I L 2 − I ) ⋅ ( I L 3 − I ) ⎞ ⋅ ln ⎜ ⎟⎠ ⎝ 297 I L1 ⋅ I L 2 ⋅ I L 3 ⎛ T ⎞ + ( N ⋅ ( E g1 + E g 2 + E g 3 ) − VOC ) ⋅ ⎜1 − STC ⎟ , ⎝ Tcell ⎠

(17.2)

VSTC = V + N ⋅

where

I ISTC V VSTC B BSTC Tcell TSTC N IL1, IL2, IL3 Eg1, Eg2, Eg3 VOC

(17.3)

= measured current, = calculated current at STCs, = measured voltage, = calculated voltage at STCs, = measured DNI, = DNI at STCs, = calculated operating cell temperature, = cell temperature at STCs, = number of cells connected in series, = short-circuit current measured (equivalent for each junction), = bandgaps of each junction, and = open-circuit voltage measured.

This expression is valid for plants of three-junction multijunction cells. For two-junction cells or silicon cells, the terms with the 2 or 3 in the subscript, respectively, must be removed from the formula. In this formula, the temperatures must be in degree kelvin, the voltages in volts, and the energies in electronvolts. For using this equation, the photogenerated currents should be known in each of the subcells of the multijunction for a standard AM1.5D spectrum, or in agreement with the ISFOC, at any other commonly used spectrum (and any irradiance). This proportion is preserved and therefore, at the measurement’s irradiance, these values (IL1, IL2, IL3) keep the aforementioned proportion with the short-circuit current measured in operation conditions.

386

CONCENTRATOR PHOTOVOLTAIC FIELD INSTALLATIONS

Figure 17.3. I–V curve measured and translated to standard conditions.

17.7

DC RATING (PLANT)

ISFOC uses its internal rating procedure for the DC acceptance of the plant. For the first plants, ISFOC measured all the concentrators to validate the procedure and to assure the function of all the systems, so the nominal power of the plant at the STCs is the sum of the nominal power obtained for all the concentrators. The determination of the nominal power of the concentrators uses the procedure explained in the previous section. First, the ambient conditions are checked to verify if they fulfill the requirements. The performance of the system depends directly on the DNI; therefore, we select a period were the DNI is stable to obtain less dispersion. Once these periods are selected, the I–V curves measured are translated to the standard conditions (Fig. 17.3) using Equations 17.1–17.3. Currently, ISFOC has measured and evaluated 600-kW DC of CPV plants. The distribution of the nominal DC power obtained for all the concentrators measured is represented in Figure 17.4. In this graph, we can observe that 88% of the concentrators have values for the nominal DC power that are between 95% and 105% of the expected value. With these results, the plants can be accepted, as established in ISFOC’s call for tenders.

17.8

AC RATING (PLANT)

The AC rating is carried out to evaluate the real performance of the plant due to module performance in different conditions of spectrum, temperature, DNI, and tracking function over multiple days.

AC RATING (PLANT)

387

Concentrators DC Nominal Power Distribution Proportion

Acumulated Proportion

0.25

100%

0.2

80%

Proportion

70% 0.15

60% 50%

0.1

40% 30%

0.05

Acumulated Proportion

90%

20% 10% > 110

108 < <= 110

106 < <= 108

104 < <= 106

Classes

102 < <= 104

100 < <= 102

98 < <= 100

96 < <= 98

94 < <= 96

92 < <= 94

90 < <= 92

0% <= 90

0

Figure 17.4. DC nominal power distribution for a total of 600 kW in various plants.

ISFOC’s AC rating procedure consists of a control of the energy produced during a period of time. The DNI data and the heat sink temperature from several modules from different concentrators in the plant (selected in a random way) are stored. The nominal power of the plant in STCs is translated to these operation conditions (DNI and heat sink temperature) stored in order to obtain the theoretical energy production of the plant. This period of time does not include the time where the DNI is lower than 600 W/m2 or when there is shadowing between the concentrators. Comparing this theoretical energy production to the real energy generated by the plant registered with the energy meter of the plant, a correction factor (F) is calculated. This correction factor represents all the losses of the plant except the efficiency of the inverter. Finally, the AC nominal power of the plant is calculated with Equation 17.4: PAC = PDC ⋅ ηinv ⋅ F where

PAC PDC ηinv F

= AC nominal power of the plant, = DC nominal power of the plant, = inverter efficiency, and = correction factor.

(17.4)

388

CONCENTRATOR PHOTOVOLTAIC FIELD INSTALLATIONS

The nominal DC power of the plant is calculated with few measurements of the concentrators. For the AC power, a correction is made using the real energy generated by the plant, so that the ratio between these two numbers really represents all the losses of the plant. Currently, ISFOC has finished the characterization of 300-kW AC. All the obtained values for this correction factor (F) have been very good, with an average value of 95% [12]. For these cases, we have obtained an average value for this ratio of 92%, which means that the losses of that plant are only 8%. The causes of these losses are

• • • • •

the miss tracking of the system during the day, the real performance of the inverters, the mismatch and dispersion between modules and concentrators, all the wiring in the plant, and the sunlight spectral variations during the day.

So, future hard work will be required to determine the magnitudes and effects of each one of these losses. If the final AC power is 100 kW ± 10%, the plant will be accepted.

17.9

RESULTS

During the first year of operation of the plants, ISFOC has observed that the actual radiation of the site in Spain is higher than the standard value of 850 W/m2 for DNI. The most typical values lie between 900 and 940 W/m2 (see Fig. 17.5), and

Figure 17.5. Frequency (in percent) of DNI values, for which the radiation is larger than the abscise value, in Puertollano, between January and August 2009.

RESULTS

389

65% of the time the direct radiation is higher than 700 W/m2, which yields two important conclusions. The first is that the production is higher than expected, and second, the inverters are always working at the highest levels (almost 10% more than expected). In this latter case, the inverters may be limiting the production of the power for a considerable number of hours during the day. The monthly production of the ISFOC’s plants in Puertollano normalized to 100 kW during the first year of the grid connection is shown in Figure 17.6. The efficiency of the production is also very high. The efficiency is calculated dividing the energy output (from utility meters indicating kilowatt hour of production fed into the grid) by the energy input, which is the solar direct radiation (measured on site with a pyrheliometer) per optical surface. For example, Figure 17.7 shows the efficiency of one of the plants during the months of July and August 2009, once the DC and AC acceptance has been finished and the plant has worked without interruption. The losses due to shadowing, spectral response, maintenance, cleaning, special measurements, and so on are not filtered, and all the data are shown. From this graph, we can obtain an average value for the efficiency in that period of time of 20.3%. Also, we can see the influence of all the actions on the plant in the energy generation, such as maintenance, cleaning, and control failure. If we select only the data from August, we obtain a higher value for the efficiency of 21%, with maximum values of 22.1%.

Figure 17.6. Production of a normalized 100-kW plant versus monthly radiation during the first year of grid connection (from October 2008 to September 2009).

390

CONCENTRATOR PHOTOVOLTAIC FIELD INSTALLATIONS

Figure 17.7. Efficiency of the energy production during the months of July and August 2009 in one plant of 100 kW of the CPV plants installed in Puertollano (Spain).

17.10

LESSON LEARNED

During the installation of the first demonstration plants, ISFOC has learned the importance of taking into account the whole life cycle of the system, including the module’s design, manufacturing, transport, installation, assembly, and O&M [13]:

• • • • • • •

During the IEC 62108 tests, the CPV module needs to show the resistance to new challenges, like the watertightness or electrical isolation. The design of the module should stress high quality manufacturing. Cell testing, optical tuning, automation, and quality control in the manufacturing line are the key to assure product quality. The CPV modules need to be designed and dimensioned to be easily transported, avoiding special transport and breakages. The CPV plants need to follow a different procedure for the installation. The alignment of the module and the tracker’s precision are new challenges. Regulatory issues, including electrical and civil work permitting and licensing are one of the main hurdles in deploying a CPV power plant. Our experience in Spain shows it can take more than 1 year to have administrative procedures completed. The module assembly and the electrical connection is also a key point, due to the the high precision needed for small acceptance angles. The systems need to survive nonworking conditions during the off-axis tracking, wind, lack of electrical connection, and so on.

CONCLUSIONS

• •

391

Software control and data monitoring are crucial for operation and the optimal performance of the plant. The inverter should be dimensioned to work in a real operation situation.

All these points need to be solved to have a good system behavior.

17.11

CONCLUSIONS

In summary, all the work carried out for these first demonstration plants has provided critical information about the CPV systems. See Figure 17.8.

• • • • •

A new rating procedure has been written, and it is being validated together with other procedures for these first plants, to establish international technical specifications. The qualification tests of standard IEC 62108 have been carried out for the first time and will be correlated with actual degradation observed in the field. The field conditions will be totally characterized, including all solar resource and meteorological data. Initial information on the production of different technologies in the same location has been obtained with positive results. And finally, the O&M data have started to establish a validated O&M procedure and cost.

Figure 17.8. CPV 800-kW demonstration plant installed by ISFOC in Puertollano (Spain).

392

CONCENTRATOR PHOTOVOLTAIC FIELD INSTALLATIONS

17.12

FUTURE WORKS

ISFOC’s objective is to investigate its current demonstration plants to obtain improved CPV procedures, standards, and technology. Future work is needed to further develop the following: 1.

2.

3.

4.

5.

6.

7.

8.

Power and Energy Measurement. ISFOC will continue the development and validation of a new standard for the power and energy rating of CPV plants. Solar Resource. ISFOC is installing meteorological stations in the whole Castilla-La Mancha region to obtain direct radiation maps. Together with direct radiation, ISFOC is measuring global and diffuse radiation, wind speed and direction, rain, temperature, humidity, and UV radiation. The objective is to compare the real data with the theoretical data to develop a model for direct radiation. Spectrum. Together with direct radiation, wind, and temperature, the value of the spectrum is very important in understanding its influence in cell performance. ISFOC has installed a spectroradiometer to measure the spectrum and isotype cells and to determine the influence in the current of each junction of the cell due to spectrum changes. The results of this study will inform the CPV community about the necessity of having different cells or optics, depending on the spectrum [14]. Installation and BOS. With the information of these first plants and the other one, new installation procedures will be established, together with other equipments like the inverter and grid connection. Production Model. All the information from the meteorological station and production will be compiled into one big database. The data will be analyzed and processed to create a model to accurately predict production. Cleaning. The CPV systems are influenced significantly by dust and soiling; therefore, they should be cleaned frequently. The methods and frequency of cleaning will also be studied. O&M. The O&M study is one of the key projects of ISFOC. Which tasks should be carried out periodically and the cost of O&M are critical in CPV technology. Degradation and Reliability. Today, CPV systems need to demonstrate reliability. Together with the results of IEC 62108, ISFOC will study the degradation and reliability of the system in the field.

ACKNOWLEDGMENTS We would like to thank all ISFOC personnel for the daily work, which has allowed the writing of this paper.

REFERENCES

393

ABBREVIATIONS AM1.5D—air mass 1.5 APS—Arizona Public Services ASTM—American Society for Testing and Materials BOS—balance of systems CPV—concentrator photovoltaic DC/AC—direct current/alternating current DNI—direct normal incident F—correction factor for all losses of a plant except the efficiency of the inverter IEC/TC—International Electrotechnical Commission/Technical Committee IES-UPM—Instituto de Energía Solar from Universidad Politécnica de Madrid ISFOC—Instituto de Sistemas Fotovoltaicos de Concentración I–V curve—current versus voltage curve kWpyc—kilowatt peak LV boxes—low-voltage boxes NREL—National Renewable Energy Laboratory O&M—operation and maintenance PTL—Photovoltaic Testing Laboratory (based in Arizona University) PV—photovoltaic or solar cell STC—standard test condition TÜV—Technischer Überwachungs-Verein (Technical Inspection Association) UV—ultraviolet

REFERENCES [1] [2] [3] [4] [5] [6] [7]

E. L. Burgess and D. A. Pritchard. Proceedings of the 13th Photovoltaic Specialists Conference, p. 1121. New York, IEEE (1978). G. Sala and A. Luque. Past experiences and new challenges of PV concentrators. In Concentrator Photovoltaics, A. Luque and V. Andreev, eds., pp. 1–23. Berlin, Springer Series (2007). M. Vivar. Optimisation of the EUCLIDES concentrator. PhD thesis, Instituto de Energía Solar—Universidad Politécnica de Madrid (2009). F. Rubio, P. Banda, J. L. Pachón, and O. Hofmann. Establishment of the Institute of Concentration Photovoltaics Systems—ISFOC. Proceedings of the 22nd EU PVSEC, September 6, 2007, Milan (2007). IEC 62108. Concentrator Photovoltaic (CPV) Modules and Assemblies— Design Qualification and Type Approval. International Standard, International Electrotechnical Commission, Edition 1.0 2007-12, Geneva, Switzerland. F. Rubio, M. Martinez, R. Coronado, J. L. Pachón, P. Banda, G. Sala, and A. Luque. Deploying CPV power plants—ISFOC experiences. 33rd IEEE Photovoltaic Specialist Conference, May 11–16, 2008, San Diego, CA (2008). O. de la Rubia, D. Sánchez, M. L. García, M. Martínez, J. L. Pachón, and P. Banda. Acceptance procedure applied to ISFOC’s CPV plants. Proceedings of ICSC5, November 16–19, 2008, Palm Desert, CA (2008).

394 [8] [9] [10] [11] [12] [13] [14]

CONCENTRATOR PHOTOVOLTAIC FIELD INSTALLATIONS Draft of the International Standard “82/571/CDV: IEC 61853-1 Ed.1: Photovoltaic (PV) module performance testing and energy rating—Part 1: Irradiance and temperature performance measurements and power rating” (2008). F. Rubio, M. Martinez, J. Perea, D. Sánchez, and P. Banda. Comparison of the different CPV rating procedures: Real measurement in ISFOC. 34th IEEE Photovoltaic Specialist Conference, June 7–12, 2009, Philadelphia, PA (2009). M. Martínez, D. Sánchez, F. Rubio, J. L. Pachón, and P. Banda. CPV systems rating results and lesson learned at ISFOC. Proceedings of ICSC5, November 16 –19, 2008, Palm Desert, CA (2008). O. de la Rubia, D. Sánchez, M. L. García, M. Martínez, J. L. Pachón, and P. Banda. Acceptance procedure applied to ISFOC’s CPV plants. Proceedings of ICSC5, November 16–19, 2008, Palm Desert, CA (2008). M. Martínez, D. Sánchez, J. Perea, F. Rubio, and P. Banda. ISFOC demonstration plants: Rating and production data analysis. Proceedings of the 24th PVSEC, September 21–25, 2009, Hamburg (2009). F. Rubio, M. Martínez, D. Sánchez, J. L. Pachón, and P. Banda. ISFOC experiences during the CPV systems installations. Proceedings of ICSC5, November, 2008, Palm Desert, CA (2008). S. Mau, Y. Sanchez Reina, V. J. Rico, A. Martin, F. Rubio, J. Leloux, D. Pachon, T. Gómez, G. Juengst, T. Gerstmaier, and A. Hakenjos. First results of the Spanish National Project “Espectro Solar”. Proceedings of the 24th European Photovoltaic Solar Energy Conference and Exhibition, September 21–25, 2009, Hamburg (2009).

PART IV SOLAR CELLS IN SPACE

18 SPACE SOLAR CELLS AND APPLICATIONS SHEILA BAILEY1 AND RYNE RAFFAELLE2 1 NASA Glenn Research Center at Lewis Field, 2U.S. Department of Energy, National Center for Photovoltaics

18.1

INTRODUCTION TO SPACE SOLAR CELLS

From the first solar-powered satellite, Vanguard 1, launched on March 17, 1958, until today on the ISS, Si solar cells have had an important role in providing satellite power. The Vanguard cells fabricated by Hoffman Electronics for the U.S. Army Signal Research and Development Laboratory at Fort Monmouth were only ∼10% efficient, while the Si cells on ISS are ∼15% efficient. The ISS cells were fabricated by Spectrolab and were assembled into wings 107 ft × 38 ft by Lockheed Martin, producing ∼32 kW of power. A total of 32,800 cells are bonded to the flexible panels of each wing. The power assembly that was completed on March 22, 2009 has a total of 262,400 Si solar cells, 8 cm × 8 cm × 0.02 cm, which could produce ∼250 kW of photovoltaic array power(eight wings at 32 kW each—ISS electrical power system hardware was sized to provide a total of 75 kW of continuous orbital power) (see Figs. 18.1 and 18.2). Construction began on ISS in 1998. A picture of the ISS Si solar cell is shown in Figure 18.3. Data are shown in Figure 18.4 for the SAW, measured up to 10 years and predicted for 15 years. Si solar cells in the space environment, however, degrade over time (see Fig. 18.4). In space, solar cells must operate under exposure to charged particle radiation (electrons and protons primarily). In the ISS LEO, the primary exposure is to charged electrons caught in the radiation belts trapped by the earth’s magnetic field, Van Allen belts. In general, the charged particle environment is dependent on the spacecraft’s orbit (see Fig. 18.5). Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

397

398

SPACE SOLAR CELLS AND APPLICATIONS

Figure 18.1. The International Space Station.

Figure 18.2. ISS solar array wing.

It can be seen from Figure 18.5 that the MEOs are the most damaging for solar cells due to Van Allen belts. The solar wind is the source of both electrons and protons and the Van Allen belts determine the radiation environment that does vary with solar activity. Solar flares can provide intense exposure to charged particles. Less important is radiation from cosmic rays originating outside our solar system. Both the degradation of Si solar cells and their efficiency drove the transition from Si space cells to GaAs-based solar cells. These cells began to be used in spacecraft beginning in the late 1980s but did not come into predominant use until the 1990s and the advent of the MJ solar cell.

INTRODUCTION TO SPACE SOLAR CELLS

399

Figure 18.3. ISS silicon solar cell.

SAW String IV Curves

3.0

50 deg C 2.5

Current, A

2.0 1.5

BOL 3 years

1.0

10 years 15 years

0.5 0.0 0

50

NASA GRC - 9/30/08

100

150

200

250

Voltage, V

Figure 18.4. Solar array wing current versus voltage for four time periods (data courtesy of Thomas Kerslake, NASA Glenn Research Center).

SPACE SOLAR CELLS AND APPLICATIONS

EOL Specific Power (W/Kg)

400

10 year mission circular orbit

GaAs/Ge

2

10

101

*coverglass optimized for each technology at each altitude

1.2 kg/m2 panel

4

3

10

10

Altitude (km) Figure 18.5. End-of-life specific power for a single-junction GaAs cell on a germanium substrate as a function of the altitude of the orbit after 10 years on orbit.

While the production capability of terrestrial photovoltaics has risen dramatically over the past few years, the demand in the space world has been essentially stable, averaging ∼27 GEO satellites per year. The high-performance III–V cells used for space can easily be adapted to terrestrial concentrator uses, and the leading producers of these cells are looking toward a growing terrestrial market. The historic focus in the space world of increasingly higher-efficiency cells still has merit in the terrestrial market, particularly for concentrators, but cost (dollar per watt) plays a larger role in determining the future of III–V cells in that market. There are several potential things that could change the current status quo and could open the playing field to players emerging to serve new niche markets. The argument about spacecraft flying low-efficiency thin-film cells is well documented in numerous trade studies [1]. In 1994, when GaAs/Ge cells sold for ∼$644 per watt compared to $432 per watt for Si cells, it was not cost effective to consider thin-film cells unless the cells were at least 13%, AM0, efficient, and there were low-cost space-qualified array designs as well. We now have thin-film cells approaching that efficiency. A recent mass and cost comparison of lightweight and rigid array structures showed that there could certainly be an advantage in cost and mass savings particularly in high-radiation orbits for thin-film cells [2]. The Boeing HPSA concept was used in this study and was compared to a Boeing 702 rigid array structure. It should be noted that here, the cost comparison was $750 per watt at the array level for a standard 28% MJ rigid array with an array performance at 83 W/kg and $400 per watt for the baseline HPSA array up to 30 kW with a 13% CIGS cell at 180 W/kg. There has been limited investment in producing higherefficiency thin-film space cells or the infrastructure to support them. NASA ceased most research funding several years ago and AFRL has provided only limited

INTRODUCTION TO SPACE SOLAR CELLS

401

TABLE 18.1. Confirmed Terrestrial Cell Efficiencies Measured under Global AM1.5 Spectrum (1000 W/m) at 25°C [3] Cells

Efficiency (%) Global AM1.5

Efficiency (%) AM0

c-Si

25.0

23.6

Area (cm2)

Manufacturer

4.0

UNSW PERL

a

c-Si thin-film transfer

16.7

14.8

4.017

U Stuttgart

GaAs

26.1

23.0a

0.998

Radboud U Nijmegen

26.1

a

1.001

Radboud U Nijmegen

a

4.02

Spire

a

GaAs (thin film) InP

22.1

23.0

19.5

GaInP/GaAs/Ge

32.0

30.2

3.989

Spectrolab

Cu(Ga,In)Se2

19.4

16.9a

1.04

NREL

16.7

a

1.032

NREL

a

0.27

United Solar Systems Corp., stabilized

1.004

Sharp

CdTe

15.0

a-Si/a-Si/a-SiGe

12.1

10.8

Dye sensitized

10.4

9.6a

Organic polymer

5.15

1.021

Konarka

At 140 suns

0.0976

NREL, inverted monolithic

GaInP/GaAs/GaInAs

40.8 a

The efficiency for global reference conditions was translated to AM0 using ASTM E-490-2000.

support. There is a true lack of research funding resulting in very few choices to be made with regard to array structures. There are several space applications in which one could envision a lightweight, low-cost array would be a major advantage. If as a matter of national security we would want a space-based solar utility capable of supplying power to targeted terrestrial sites, then there would be considerably more interest in high-efficiency, low-cost, thin-film cells. Looking at the possibilities for the future, there are some interesting emerging technologies. A selection of the highest confirmed cells can be seen in Table 18.1 [3].

18.1.1

III–V

The III–V cells have long held the record for efficiency from a monolithic cell structure. There are currently two U.S. manufacturers of III–V space cells and terrestrial concentrator cells, Spectrolab and Emcore. Both Spectrolab and Emcore have record cells of monolithic TJ cells (GaInP/GaAs/Ge) slightly above 30% AM0. These cells, the ZTJ for Emcore and the XTJ for Spectrolab, are currently

402

SPACE SOLAR CELLS AND APPLICATIONS

undergoing AIAA S-111 testing and are part of a ManTech program for the U.S. government. Their cost has been a deterrent for terrestrial use despite the high efficiencies except for concentrator systems. There have been efforts made to reduce those costs. One of these efforts involved reducing substrate cost by reducing wafer manufacturing costs and providing more efficient use of the Ge. Others have been focused on reducing cell costs by the removal of the cell from the substrate [4]. Some of the cell configurations have also been introduced to improve cell efficiency, and the effect on cost is not yet well understood [5, 6]. These include inverted cells and mechanically stacked cells [7]. In addition, there have been efforts to enhance efficiency through new cell structures such as QD cells [8] IB cells, up and down photon conversion, hot carriers in multi-QW cells, [9] or four- or five-junction MJ cells. Last, there is an effort to produce a virtual singlecrystal MJ cell on a polycrystalline foil substrate.

18.1.2

Thin MJ Cells

If one looks at the bandgaps for the theoretical performance of MJ cells in the space environment, it is easier to understand why an inverted cell structure was chosen. In the current SOA 3J solar cell, the Ge bottom cell absorbs almost twice the low-energy photons necessary for current matching within the cell. Table 18.2 shows the bandgaps of the component cells for multiple junctions [10]. Clearly, a 1.0-eV bandgap material would improve the efficiency of a 3J cell. Unfortunately, the nitride materials that first promised the appropriate bandgap proved to have intrinsic defects limiting that application. In0.3Ga0.7As could also be used but, because of the lattice mismatch, would require graded composition buffer layers that have their own defect problems. The inverted configuration pioneered by NREL permitted the top junction to avoid the pitfalls of threading defects. The subcells are connected by tunnel junctions with the cell current density equal to that of the subcell with the lowest current density and the open-circuit voltage equal to the sum of the subcell open-circuit voltages. The epitaxial structure is grown in an inverted manner on a single-crystal substrate template. The template is removed during fabrication, leaving the cell composed of the epitaxial layer structure, n and p metal electrodes, and an antireflection coating. At the Space Power Workshop

TABLE 18.2. Bandgaps Modeled for Multijunction Series-Connected Structures Junctions

Bandgaps (ev)

Theoretical Efficiency

Jsc (mA/cm2)

3

1.9/1.33/0.92

38.8

20.2

4

2.0/1.46/1.08/0.77

41.8

17.5

5

2.13/1.64/1.28/1.0/0.75

43.9

14.2

6

2.22/1.76/1.42/1.15/0.92/0.7145.3

45.3

12.2

INTRODUCTION TO SPACE SOLAR CELLS

403

in April 2009, both Spectrolab and Emcore showed progress on an IMM TJ cell. Spectrolab measured 31.5% with 20 out of 25 cells exceeding 30%. Emcore reported a 3J cell with 33.4% efficiency and a 4J cell with 33.9%. Both companies are addressing the substantial issues of cell contact configuration, interconnect attachment, scaling up in area, handling concerns, and bonding. The low-mass (0.012 g/cm2), high-efficiency (>32.4% AM0) flexible cell is an excellent candidate for applications where high specific power (1 W/g), high areal power (400 W/m2), and cell flexibility are required. These could include unmanned aerial vehicles as well as low-mass flexible satellite solar panels. It should be noted that while the cell design is not identical, the terrestrial concentrator cells produced by both companies will benefit by all the lessons learned in achieving higher-efficiency space solar cells. It is not unreasonable to project that there will be 35% efficient space cells on the market in a few years. The IMM cells still require the use of a single-crystal substrate template that is later removed. If one wanted to preserve that substrate for further uses, thus reducing cost, there are several ways that might be accomplished. MicroLink has developed proprietary technology based on ELO to effectively peel a 4-in. wafer with epitaxial cells [4]. They fabricated DJ ELO solar cells with an efficiency of 28% at AM1.5 illumination of 1 sun. They have also succeeded in peeling an InP wafer as well. That opens the possibility for new cell configurations. There are other techniques for peeling cells using a “damaged” layer in the substrate. Years ago, we had a “cleft” process to remove cells. It should be noted that you can put these IMM cells back on a substrate of your choice. A recent study looked at a TJ cell grown on thin (100 mm) Ge and a CIGS cell on thin metal substrates [11]. Cells were tested against a set of performance indicators for an array, and the thin TJ cells proved remarkably robust. It was noted also that to fully benefit from the specific power (watt per kilogram) of these arrays, the TJ cells needed to have an alternative to the relatively thick cover glass required by some orbits. Another approach to providing thin TJ cells involves growing III–V cells on polycrystalline metal foils. First, it was demonstrated that polycrystalline GaAs cells can be grown with efficiencies greater than 20% under AM1.5 [12]. It has also been demonstrated that device-quality polycrystalline Ge suitable for OMVPE growth can be produced on metallic foils using a recrystallization process [13]. To date, this work has focused on optimizing both the virtual single-crystal Ge produced on foil and determining representative changes in cell quality in transitioning from single-crystal GaAs to polycrystalline Ge substrates. Results were presented at the 34th IEEE PVSC of single-junction (SJ) GaAs devices grown by OMVPE on single-crystal GaAs, single-crystal Ge, and poly-Ge. Trials were conducted to study the recrystallization of Ge on Mo foil to determine an optimal process to produce large (∼1 mm or greater) grains, using similar samples annealed using both a tube furnace and rapid thermal processing at varying temperatures, times, and process gases. The goal is to be able to grow a TJ cell on a flexible metal substrate. A company, Wakondo Technologies, exists to commercialize this process. It is difficult to anticipate where this will lead, but the potential promise

404

SPACE SOLAR CELLS AND APPLICATIONS

of using a low-cost process to produce higher-efficiency III–V cells is certainly worth pursuing.

18.1.3

Nanostructured Solar Cells

Nanostructured solar cells have proposed structures that will potentially exceed their conventional design counterparts. A method to achieve this involves inserting a low-dimensional structure, such as multiple QDs or QWs into a standard cell configuration, in order to form an IB within the bandgap of the host [14, 15]. Absorption of photons with energy below the bandgap of the host occurs from the valence band to the dot IB and from the dot IB to the host conduction band. Since this permits photons that would not have been absorpted to generate more current, the IB cell would gain in efficiency if the voltage of the standard cell was maintained. The theoretical prediction for this case was 45.8% compared to the standard cell limit of 31% AM0, 1 sun. There has been experimental verification of the key principles of the IB cell using InAs and GaSb dots [16, 17]. A complementary approach takes advantage of the extended absorption spectrum of the QD inserted into the current limiting junction of a TJ cell [18]. The extended absorption increases the current of the TJ and enhances the EOL efficiency. Modeling indicates that the efficiency could be increased to 47%. It is also possible the cell would have enhanced radiation tolerance as well. The problems associated with manufacturing such a cell have proven to be complex. Since the dots are initiated due to a lattice strain between the dot material and the host cell, it is necessary to compensate for that strain in growing multiple QD layers. Considerable work has gone into learning how to do strain compensation without degrading the open-circuit voltage of the cell. To date, the enhanced short-circuit current has been verified. GaAs cells with a p-i-n configuration were grown with embedded InAs QD layers. The QE measurements of the cell confirmed that the increased current was due to the photogenerated carriers with QDs [18]. A very similar story could be told about QW devices. There have been successes in enhanced short-circuit current, but many of the constraints that apply to QD cells also apply to QW cells.

18.1.4

CIGS, a-Si, CdTe

CIGS, a-Si, and CdTe are traditional thin-film cells historically produced on glass. However, we have seen these cells produced on other substrates during the last decade, and companies have been formed that produce primarily terrestrial products. While still limited in efficiency, there is a significant possibility for these cells to both improve in efficiency or form part of a hybrid cell. The major factor in favor of these cells has always been cost as they can be produced in a roll-to-roll process. Another factor for space applications has been their radiationresistant behavior. Thin film progress in small-area research cell efficiencies has

INTRODUCTION TO SPACE SOLAR CELLS

405

not translated easily to module efficiencies. While the best Si module efficiencies are on the order of 92% of cell efficiencies, CIGS modules are usually only a little over 70% and CdTe is approximately 65%. There are multiple reasons for this discrepancy in thin-film cells and arrays, and progress has been made primarily in improving the cell performance. CdTe has historically been produced on glass, and there has not been any commercialization on metal foils or polymers so CdTe has not been considered for space use. The CIGS assemblies mentioned in the previous study funded by the Italian Space Agency used CIGS cells that were 12% almost stable over a wide range of radiation fluencies [11]. The cells were acquired from an Italian university and also from the production line of Global Solar Energy. High specific power and flexible thin-film TJ a-Si:H-based solar cells are attractive for space applications. USO has championed and pioneered in thin-film a-Si alloy (a-Si:H)-based solar cells for many years and has developed a-Si:Hbased solar cells on polymer substrate made on a high-throughput basis using roll-to-roll deposition technology for space and stratospheric missions. In this conference, they will report on two new developments: (1) the monolithically laser-integrated thin-film module on flexible polymer substrate to attain an initial specific power as high as 2343 W/kg at the module level using TJ a-Si alloy solar cells and (2) a new monolithic hybrid module design that marries the advantages of our wire-bonded baseline cell with those of the advanced laser-integrated module. It uses standard size 5-MW cells as the starting material, and therefore it is roll-to-roll compatible. All layers were deposited using roll-to-roll production machines on a ∼1-mil-thick polymer substrate. With this design, we have fabricated modules having an initial AM0 aperture area efficiency of ∼9.5% and an estimated specific power of ∼1600 W/kg. Unfortunately, there is no simple answer for projecting the future of thin-film cells in space. The complete system, including stowed volume as well as the more traditional mass and power requirements, must be looked at for each mission. History has shown that it is not the cells that generally fail in an array failure but often, interconnects, deployment mechanisms, and so on are at fault. Our reluctance to utilize new array designs has slowed the progress of using thin-film cells in space.

18.1.5

Organics, Organic Hybrids, Dye

This last category has made significant gains in efficiency but is the least understood and the less well developed. More fundamental research is necessary to understand the physics and material behavior in many cell types. There are many proposed cells involving nanostructures and organic materials. One can make some estimates of the potential cost of some of these cells, but others are not developed enough to estimate. We know that for space applications, there are significant qualification costs and the behavior of the new cells under radiation and other space conditions would have to be established. This may be near term for the MJ cells but a much longer venture for the new cell types.

406

SPACE SOLAR CELLS AND APPLICATIONS

18.2

ARRAYS

Solar array designs have undergone a steady evolution since the Vanguard 1 satellite. Early satellites used Si solar cells on honeycomb panels that were body mounted to the spacecraft. Early space solar arrays only produced a few hundred watts of power. However, satellites today require low-mass solar arrays that produce several kilowatts of power. Several new solar array structures have been developed over the past 40 years to improve the array-specific power and to reduce the stowed volume during launch. The most important characteristics of solar arrays required for space applications are

• • • •

high specific power (watt per kilogram), low stowed volume (watt per cubic meter), low cost (dollar per watt), and high reliability.

In addition, several proposed space missions have put other constraints on solar arrays. Several proposed earth-orbiting missions designed to study the sun require “electrostatically clean” arrays. Inner planetary missions and missions to study the sun within a few solar radii require solar arrays capable of withstanding temperatures above 450°C and of functioning at high solar intensities (HIHT). Outer planetary missions require solar arrays that can function at low solar intensities and low temperatures (LILT). In addition to the near-sun missions, missions to Jupiter and its moons also require solar arrays that can withstand high radiation levels. The EOL power generated by an array is impacted in a variety of ways. Radiation damage will result in an 8% loss of BOL power per square meter after 5 × 1014 1-MeV electrons (i.e., typical EOL fluency in GEO orbit). The temperature correction due to the operation of the solar cell at 75°C rather than at the 25°C test conditions will reduce the power per square meter by ∼9%. Degradation due to UV exposure is around 1.7%. Loss in power over time due to micrometeors and ordinary surface contamination are each around 1% [19]. The solar arrays presently in use can be classified into six categories:

• • • • • • •

body-mounted arrays, rigid panel planar arrays, flexible panel array flexible roll-out arrays, concentrator arrays, high-temperature/high-intensity arrays, and electrostatically clean arrays.

A summary of the important typical characteristics of these arrays are given in Table 18.3 [20].

ARRAYS

407

TABLE 18.3. Space Solar Array Characteristics Technology

Specific Power (W/kg [BOL] at Cell Efficiency)

Cost ($K/W)

Area per Power (m2/kW)

High-efficiency silicon (HES) rigid panel

58.5 at 19%

0.5–1.5

4.45

HES flexible array

114 at 19%

1.0–2.0

5.12

Triple-junction (TJ) GaAs rigid

70 at 26.8%

0.5–1.5

3.12

TJ GaAs ultraflex

115 at 26.8%

1.0–2.0

3.62

CIGS thin filma

275 at 11%

0.1–0.3

7.37

Amorphous-Si MJ/thin filma

353 at 14%

0.05–0.3

5.73

a

Projected values.

Figure 18.6. Body-mounted arrays on the Mars rovers (picture courtesy of NASA JPL).

18.2.1

Body-Mounted Arrays

Body-mounted arrays are preferred for small satellites that only need a few hundred watts. Early spherical satellites and spin-stabilized cylindrical satellites used bodymounted arrays of Si solar cells on the honeycomb panels. This type of array is simple and has proven to be extremely reliable. One of the limitations of this type of array is that it puts a constraint on the direction the spacecraft must point. This type of array is still used on smaller spacecraft and on spin-stabilized spacecraft. The Mars Pathfinder Sojourner Rover and the Mars Exploration Rovers, Spirit and Opportunity, also used body-mounted solar arrays (see Fig. 18.6).

408

18.2.2

SPACE SOLAR CELLS AND APPLICATIONS

Rigid Panel Planar Arrays

Rigid panel arrays have been used on many spacecrafts requiring several hundred watts to many tens of kilowatts of power. They consist of rigid honeycomb core panels that are hinged such that they can be folded against the side of the spacecraft during launch (see Fig. 18.7). Each panel is rigid and quite strong, but can add considerably to the overall weight of the array. There has been much development recently on panels of materials other than aluminum (i.e., graphite/epoxy sheets and ribbons). Hybrid panels with aluminum honeycombs and epoxy/glass face sheets have also been developed. The folded arrays are deployed by means of pyrotechnic, paraffin, or knife blade actuators and damper-controlled springs.

Figure 18.7. Rigid panel solar array (photo credit—Boeing image).

ARRAYS

409

The BOL power density of the rigid panel array is extremely dependent on the type of solar cell used. BOL power densities range from 35 to 65 W/kg for Si cells and from 45 to 75 W/kg for GaAs/Ge cells. The panel assembly of a rigid array accounts for 75–80% of the total mass, with the stowed and deployment structure making up the balance [20]. The TRMM and Rossi XTE both employ rigid panel arrays. Available power supplied by typical rigid panel arrays range from very small to in excess of 100 kW.

18.2.3

Flexible Foldout Arrays

Flexible foldout arrays are attractive for missions that require several kilowatts of power because of their high specific power, high packaging efficiency (low stowed volume), and simple deployment system. These arrays are generally designed in two basic configurations: 1. 2.

flexible flat-panel array with linear deployment as shown in Figure 18.8 and flexible round-panel array with circular deployment as shown in Figure 18.9.

These arrays have flexible or semiflexible panels that are stowed for launch with accordion folds between each panel. On reaching an appropriate orbit, these are unfurled by means of an Astromast™, an Ablemast™, or some other similar device. The specific power of these types of arrays varies from 40 to 100 W/kg, depending on the cell type, power, mission-reliability requirements, spacecraft orientation and maneuverability capabilities, and safety requirements. Initially, they were marketed as a significant improvement in power produced per unit mass. However, even though flexible arrays have an excellent figure of merit in this

Figure 18.8. ISS array with linear deployment (figure courtesy of NASA).

410

SPACE SOLAR CELLS AND APPLICATIONS

Figure 18.9. Phoenix Lander flexible round-panel array with circular deployment (picture courtesy of ATK).

regard, the best rigid honeycomb panels have thus far matched their specific power performance. Very large flexible blanket solar arrays present complex structural and spacecraft design issues. This type of array is used on the Milstar series of spacecraft, on the Terra spacecraft, Mar’s Phoenix Lander [21], and on the ISS [22] (see Fig. 18.2). The ISS array had a BOL-specific energy of 40 W/kg due to requirements for additional maneuverability and for safety and reliability.

18.2.4

Thin-Film or Flexible Roll-Out Arrays

The flexible roll-out array is similar to the accordion-folded array mentioned earlier, except for the fact that the semiflexible or flexible substrate is rolled onto a cylinder for launch. The Hubble space telescope used such a roll-out array (see Fig. 18.10). It contained a polyimide blanket in a roll-up stowed configuration. The array was deployed by a tubular, extendable boom (BI-STEM) deployment system. The flexible roll-out array design was developed for the U.S. Air Force. After 8 years in orbit, the solar arrays on Hubble were replaced in orbit due to degradation. During the repair mission, delamination of the solar array bus bars was observed, and it was also noticed that two of the hinge pins had started to creep out. One of the arrays was returned to Earth to be studied, while the other array was jettisoned into space. The returned solar array was shipped to ESA for further study. These roll-out arrays were replaced with rigid panels that were thought to be more reliable. AFRL completed a 3-year program with two prime contractors (Boeing and Lockheed Martin) to investigate and design complete arrays uniquely tailored to thin-film solar cells. The SquareRigger™ solar array being developed by AECAble (now ATK) is a flexible blanket system composed of modular “bays.” This

ARRAYS

411

Figure 18.10. Roll-out arrays used on the Hubble space telescope (photo courtesy of NASA).

array is attempting to combine an ultrahigh-power capability (>30 kW) with a high stowed packaging efficiency. The SquareRigger solar array system is projected to achieve a specific power between 180- and 260-W/kg BOL, depending on the type of cells used. A SquareRigger system using thin-film cells is projected to offer an order-of-magnitude reduction in cost over conventional rigid panel systems. Ultraflex Solar Arrays were recently used to provide power to the Mars Phoenix Lander (see Fig. 18.9). Launched in August 2007, the Phoenix Mars Mission is the first in NASA’s Scout Program. Phoenix is designed to study the history of water and habitability potential in the martian arctic’s ice-rich soil. This was the first flight for this unique solar array technology developed by ATK’s Goleta, California facility. Each Ultraflex array unfolded like an oriental fan into a circular shape 2.1 m in diameter and will generate 770 W of power from sunlight at the distance Earth is from the sun. Since Mars is approximately 1.5 times farther from the sun, the solar arrays will produce less than half the power possible on

412

SPACE SOLAR CELLS AND APPLICATIONS

Earth. These same arrays have been projected as the solar arrays for use on Orion, the new proposed crew exploration vehicle to replace the shuttle.

18.2.5

Concentrating Arrays

Photovoltaic concentrating arrays have been proposed for missions to outer planetary missions, SEP missions, and missions that operate in high-radiation environments. These arrays are attractive for these missions because they have the potential to provide a high specific power, higher radiation tolerance, and improved performance in LILT environments. The technical issues in using concentrating arrays are precision pointing, thermal dissipation, nonuniform illumination, optical contamination, environmental interactions, and complexity of deployment. They can also decrease overall spacecraft reliability because a loss of pointing may cause significant power loss to the spacecraft. Reflective systems can have concentration ratios from 1.6X to over 1000X, with a practical limit of around 100X. Refractive designs are generally limited to the range of about 5X–100X, with a practical limit of around 20X. Solar energy may be focused on a plane, line, or point depending on the geometry of the concentrator design. These concentrators may be small and numerous if used in a distributed focus design, or they may be a single large concentrator as in a centralized focus design. AstroEdge™ array on the NRO STEX spacecraft, launched in October 1998, was the first spacecraft to use a concentrator as its main power source. This system used a reflective trough design with a nominal 1.5X concentration. The arrays were successfully deployed and cell currents were slightly higher than predicted. Thermal problems did occur on some of the panels owing to the higher concentrator operating temperature. The Deep Space 1 spacecraft launched in October 1998 used ScarletTM concentrator arrays to provide power to its ion propulsion engines [21]. Its two arrays were capable of producing 2.5 kW at 100 VDC. The ScarletTM array was developed by AEC-Able under a program sponsored by BMDO. The ScarletTM array had a refractive linear distributed focus with a 7.5X concentration ratio. The array has 720 lenses to focus sunlight onto 3600 solar cells. Deep Space 1 has two ScarletTM SAW assemblies. Each assembly is made up of a composite yoke standoff structure, four composite honeycomb panel assemblies, and four lens frame assemblies. High-efficiency TJ GaInP2/GaAs/Ge cells were used in this array. The first commercial concentrator array developed for space was the Boeing 702. It was used on the Galaxy XI spacecraft and was deployed on January 12, 2000. It had a reflective planar centralized focus concentrator design in which the sun’s rays were reflected onto a single rectangular plane of solar cells. It used thin-film reflectors and had a 1.7X concentration. It was designed for power levels of 7–17 kW over a 16+-year design life. The array deployed as expected and its initial power output was within the expected

ARRAYS

413

Figure 18.11. Flexible blanket version of the stretched lens array.

range. However, its concentrator surfaces degraded very quickly while in orbit. The specific power of this array was ∼60 W/kg using 24% efficient MJ solar cells. The similar Boeing 601 bus, which uses an ordinary planar solar array, is limited to about 15 kW of power owing to the array stowed volume limitations. Current efforts in concentrator arrays at NASA at the Department of Defense focus on the SLA (see Fig. 18.11) using a Square Rigger structure discussed above.

18.2.6

High-Temperature/High-Intensity Arrays

Missions to Mercury and other missions with close encounters to the sun (i.e., solar probe) have generated the need for cells and arrays that are capable of operating in high-light-intensity, high-radiation, and high-temperature environments. Two missions that had to contend with such an environment have already flown. Helios A, which reached 0.31 AU—the average earth-to-sun distance, was launched on December 10, 1974, and Helios B, which reached 0.29 AU, was launched on January 15, 1976. These spacecrafts used ordinary Si cells that were modified for high-intensity use and had second surface mirrors to cool the array. The remainder of their technology was very similar to what is used on standard arrays. In addition to these missions, the upcoming MESSENGER Discovery mission is planned for travel to 0.31 AU. Its solar array design is already under development. The current solar array technology can meet the needs of MESSENGER or other spacecrafts that approach the sun to about 0.3 AU, but with reduced performance and increased risk compared to other applications. Further progress is

414

SPACE SOLAR CELLS AND APPLICATIONS

required in both cell and array development for closer encounters to the sun. The common feature to the high-temperature and high-intensity solar arrays that have operated thus far is the replacement of a significant fraction of the solar cells by OSRs. These are mirrors that help control the array temperature near the sun at the cost of reduced power at larger distances. The MESSENGER design also off-points the array as the spacecraft nears the sun to keep the array below 130°C. The array is designed to tolerate pointing at the sun for a maximum of 1 h (probably much longer). However, it will be unable to function under this extreme condition (i.e., 260°C). The U.S. Air Force and BMDO also developed some high-temperature arrays in the late 1980s. The SCOPA and SUPER were designed to be capable of surviving laser attack. These were concentrator arrays that directed the incident laser light away from the solar cells. Although the laser light would not impinge directly, the arrays’ temperature would increase dramatically and thus, the arrays needed to withstand several hundred degrees Celsius. The high-temperature survivability of SCOPA and SUPER was achieved through changes to the contact metallization and through the use of diffusion barriers in the GaAs cells used. Both Tecstar and Spectrolab developed the cells in conjunction with this effort. Other smaller companies such as Astropower, Kopin, and Spire have also worked on developing high-temperature cells. GaAs cells reaching an AM0 efficiency of 18% were produced, which degraded less than 10% under 1 sun after annealing in vacuum for 15 min at 550°C. Concentrator cells that survived repeated 7-min excursions to 600°C were produced. These same cells exhibited only 10% loss with exposure to 700°C. NASA is also currently funding an effort to develop wide bandgap solar cells for high-temperature/high-intensity environments. Cells using materials such as SiC, GaN, and AlGaInP are being developed [23]. These cells may also benefit from high-emissivity selective coatings that will limit the unusable IR entering the solar cells and reduce their steadystate temperature.

18.2.7

Electrostatically Clean Arrays

There is an entire class of proposed missions designed to study the SEC. These spacecrafts typically measure the fields and particles associated with the solar wind. This requires that arrays be developed that do not distort the local environment or be electrostatically “clean.” These arrays must have their voltage separated from the space plasma and the array must be maintained at the same potential as the spacecraft. This is usually accomplished by coating the cell cover glass and arrays between the cells with a conductor. Since the coating for the cover glass must be transparent, a TCO such as indium tin oxide is used. The coatings between the cells must not short them out, so an insulating coating must first be applied to all of the interconnects before the conductive coating or “v” clips. All of this must be done within a thickness of ∼0.08 mm and within a width of about 0.8 mm.

INTEGRATED POWER SYSTEMS

415

Fabricating an electrostatically clean array presently costs three to six times as much as a typical array. This is due in large part to the hand labor involved in developing such arrays. These arrays are also less reliable due to the lack of robustness of the conductive coatings used to maintain the equipotential. In addition, these arrays are also generally body mounted, which cuts down on the available power to the spacecraft (i.e., pointing issues). The power is also limited due to the thicker cover glass that is employed owing to the high-radiation environment associated with SEC missions. Unfortunately, there is no a wide knowledge base on how to develop electrostatically clean arrays. This was demonstrated in the cost of developing the FAST solar array. The electrostatically clean body-mounted solar panels for FAST cost in excess of $7400 per test condition watt. The use of monolithic diodes on the latest generation of MJ solar cells could prove to be a tremendous advantage in developing electrostatically clean arrays. The presence of antennas, booms, and outcroppings from a body-mounted array requires that solar cells have bypass diodes to reduce the shadowing losses and potential damage to the arrays. The new built-in diodes will obviate the need and the expense involved in adding the diodes to the array circuitry. The NASA GSFC recently funded COI to study electrostatically clean arrays through the STP Program’s MMS and GEC projects. COI will be supplying the electrostatically clean solar panels for the CNOFS.

18.2.8

Mars Solar Arrays

Mars orbiters have used photovoltaic arrays that are quite similar to those used in Earth orbit with good results. However, Mars surface missions, in which the solar spectrum is depleted at short wavelengths, cause the efficiency of the cells to be lower than if the cells were operated above the atmosphere of Mars. The cell efficiency is reduced by about 8% (relative percent). In addition, the effect of dust accumulating on arrays was observed on the Mars Pathfinder mission by monitoring the JSC of cells exposed to the environment whose short-circuit current could be monitored on a routine basis. One cell indicated an increase in obscuration of about 0.3%/sol for the first 20 sols (note that a “sol” is a martian day of 24.6 h). The other cell indicated that over a longer period of ∼80 sols, the obscuration flattened out and seemed to be approaching an asymptote of around 20% obscuration [24]. Cells that are “tuned” to the martian solar spectrum and methods for mitigating dust obscuration will be necessary to produce efficient arrays for Mars surface power.

18.3

INTEGRATED POWER SYSTEMS

NASA has also been working to develop lightweight, integrated space power systems on small-area flexible substrates [24]. These systems generally consist of a high-efficiency thin-film solar cell, a high-energy density solid-state Li-ion

416

SPACE SOLAR CELLS AND APPLICATIONS

battery, and the associated control electronics in a single monolithic package. These devices can be directly integrated into microelectronic or MEMS devices and are ideal for distributed power systems on satellites or even for the main power supply on a nanosatellite. These systems have the ability to produce constant power output throughout a varying or intermittent illumination schedule as would be experienced by a rotating satellite or “spinner” and by satellites in a LEO by combining both generation and storage. An integrated thin-film power system has the potential to provide a low-mass and low-cost alternative to the current SOA power systems for small spacecrafts. Integrated thin-film power supplies simplify spacecraft bus design and reduce losses incurred through energy transfer to and from conversion and storage devices. It is hoped that this simplification will also result in improved reliability. The NASA Glenn Research Center has recently developed a microelectronic power supply for a space flight experiment in conjunction with the Project Starshine atmospheric research satellite (http://www.azinet.com/starshine/). This device integrates a seven-junction small-area GaAs MIM with an allpolymer LiNi0.8Co0.2O2. The array output is matched to provide the necessary 4.2-V charging voltage, which minimizes the associated control electronic components. The use of the matched MIM and thin-film Li-ion battery storage maximizes the specific power and minimizes the necessary area and thickness of this microelectronic device. This power supply was designed to be surface mounted to the Starshine 3 satellite, which was ejected into a LEO with a fixed rotational velocity of 5° per second. The supply is designed to provide continuous power even with the intermittent illumination due to the satellite rotation and LEO [25].

18.3.1

High Specific Power Arrays

To achieve an array-specific power of 1 kW/kg, a much higher cell specific power will be necessary. Similarly, the blanket specific power (i.e., interconnects, diodes, and wiring harnesses) must be over 1 kW/kg as well. The APSA assessment determined that the mass of the deployment mechanism and structure is essentially equal to the blanket mass for a lightweight system [26]. Therefore, a blanket specific power of approximately 2000 W/kg would be necessary to achieve a 1 kW/kg array. NASA is currently sponsoring an effort by AEC-Able Engineering to develop lightweight thin-film array deployment systems. Gains in array-specific power may be made by an increase in the operating voltage. Higher array operating voltages can be used to reduce the conductor mass. The APSA was designed for a 28-V operation at several kilowatts output, with the wiring harness comprising ∼10% of the total array mass, yielding a specific mass of ∼0.7 kg/kW. If this array was designed for a 300-V operation, it could easily allow a reduction of the harness specific mass by at least 50%. This alone would increase the APSA specific power by 5% or more without any other modification.

POWER SYSTEM FIGURE OF MERIT

417

The extremely high specific power arrays that need to be developed for SEP and SSP applications will require lightweight solar arrays that are capable of highvoltage operation in the space plasma environment. SEP missions alone will require 1000–1500 V to directly power electric propulsion spacecraft (i.e., no voltage step-up is required to operate the thrusters). NASA has proposed a thin-film stand-alone array-specific power that is 15 times the SOA III–V arrays, an area power density that is 1.5 times that of the SOA III–V arrays, and specific costs that are 15 times lower than the SOA III–V arrays [27].

18.3.2

High-Radiation Environment Solar Arrays

There are several approaches to mitigating the effects of a high radiation on a solar array. The simplest is to employ a thick cover glass (assuming that a commercial source could be developed). Thick cover glasses protect the cell from the highly damaging low-energy protons but will cause a significant decrease in the specific power of the array. However, this can be reduced if one adopts a concentrator design, assuming of course that the additional elements associated with the concentrator can withstand the high-radiation environment as well. A different approach is to try and develop cells that are more radiation resistant. Several of the materials that are being investigated for high-temperature/high-intensity missions have also shown good radiation resistance. However, these will not be suited to the high-radiation missions involving LILT. Many of the high-radiation NASA missions being considered occur at distances much greater than 1 AU. Thin films may offer a possibility since they have demonstrated some advantages with regard to radiation tolerance as previously mentioned, provided the problem of their low efficiencies are solved.

18.4

POWER SYSTEM FIGURE OF MERIT

There are many figures of merit that must be considered in developing an SSP system (i.e., specific mass, specific power, cost per watt, temperature coefficients, and anticipated radiation degradation of the solar cells used). The radiation hardness and the temperature coefficients for the III–V MJ cells are significantly better than Si cells, as previously discussed. This leads to a significantly higher EOL power level for an MJ cell as compared with a Si cell. This is shown in Table 18.4, where the BOL cell efficiencies at room temperature and the typical EOL cell efficiencies for LEO and GEO operating temperatures and radiation environments are presented. The difference in radiation degradation can have a huge impact on power system design. For example, if the area for a typical rigid panel is approximately 8 m2 and the area of a typical solar cell is 24 cm2, using a panel packing factor of 0.90 will allow the panel to have 3000 cells. Under GEO conditions, this panel populated with high-efficiency Si cells will produce 1.2 kW of EOL power. The

418

SPACE SOLAR CELLS AND APPLICATIONS

TABLE 18.4. A Comparison of the Relative Radiation Degradation of 75-μm Multijunction Cells and High-Efficiency Si under GEO and LEO Operation [24] Solar Cell Technology

BOL Efficiency at 28°C (%)

EOL Efficiency on Orbit (%)

GEO conditions (60°C)—1 MeV, 5E14 e/cm2 HES

14.1

12.5

2J III–V

20.9

20.0

3J III–V

23.9

22.6

2

LEO conditions (80°C)—1 MeV, 1E15 e/cm HES

13.4

10.6

2J III–V

19.7

18.1

3J III–V

22.6

20.3

EOL power could almost be doubled to 2.2 kW if it were populated with SOA TJ cells. Alternatively, the solar arrays populated with high-efficiency Si cells would need to be 77% larger than arrays using TJ cells in order to deliver the equivalent amount of EOL power in GEO and 92% larger in LEO. The large difference in size between solar arrays populated with Si and MJ cells is very significant in terms of stowage, deployment, and spacecraft attitudinal control. This is especially true for very high-powered GEO communications satellites in which the Si solar array area can exceed 100 m2. The comparable array with TJ cells, although by no means small, would have an area of ∼59 m2. The array size will impact the spacecraft’s weight, volume (array stowage), and system requirements on spacecraft attitude control systems (additional chemical fuel). Three important figures of merit used in power system optimization are EOL area power density (watt per square meter), specific weight (watt per kilogram), and cost (dollar per watt). Representative values for the various SOA cell technologies are listed in Table 18.5. The EOL power per unit area for an MJ cell is significantly better than a Si cell. However, the EOL specific weight for Si is almost a factor of 2 greater than an MJ cell. This results in a slightly smaller EOL cost per watt for a highefficiency Si. This demonstrates the dramatic reduction in cost of MJ cells over the past few years. If one considers the mass of the necessary array components (i.e., panel substrate, face sheet, adhesive, hinges, insulators, and wiring) along with equivalent power per area for the different cell types, and also the cost involved in having the CIC and laid on rigid panels, then the cost for developing an array using MJ cells is slightly less than that for HES cells. The EOL specific weight values at the

POWER SYSTEM FIGURE OF MERIT

419

TABLE 18.5. EOL Area Power Density (W/m2), Specific Weight (W/kg), and Normalized (to HES) Cost ($/W) for High-Efficiency Si, Dual-Junction (2J), and Triple-Junction (3J) Bare Solar Cells [24] Solar Cell Technology

W/m2

W/kg

Normalized Cell Cost ($/W)

GEO conditions (60°C)—1 MeV, 5E14 e/cm2 75-μm HES

169

676

1.00

2J III–V

271

319

1.38

3J III–V

306

360

1.22

LEO conditions (80°C)—1 MeV, 1E15 e/cm2 75-μm HES

143

574

1.00

2J III–V

245

288

1.29

3J III–V

275

323

1.15

Table 18.6. EOL Specific Weight (W/kg) at the CIC and Panel Levels for 3-mil HighEfficiency Si, Dual-Junction (2J), and Triple-Junction (3J) Cells [24] Solar Cell Technology

CIC Sp. Power (W/kg)

Panel Sp. Power (W/kg)

Normalized Panel Cost ($/W)

75

1.00

GEO conditions (60°C)—1 MeV, 5E14 e/cm2 75-μm HES

261

2J III–V

219

95

0.9

3J III–V

248

108

0.8

LEO conditions (80°C)—1 MeV, 1E15 e/cm2 75-μm HES

221

63

1.00

2J III–V

199

86

0.84

3J III–V

223

97

0.75

CIC (with 100-μm ceria-doped microsheet cover glass) and the panel levels for these cells and the normalized cost per watt for the panels are shown in Table 18.6. A similar comparison with somewhat less expensive 100-μm HES cells (at the panel level) shows a slightly smaller cost advantage for the MJ cells. The 100-μm Si cells are less radiation-hard and have lower specific power (watt per kilogram) than 75-μm Si cells, but they cost about 35% less.

420

SPACE SOLAR CELLS AND APPLICATIONS

Currently, conventional space Si cells are less expensive than MJ cells at the panel level. However, their EOL power is much lower than either the high-efficiency Si cells or the MJ cells. The increased mass and area that their usage entails would have to be considered against the cost savings and other mission considerations in any comparative study. Engineers have worked on ways to improve space solar cells and arrays in terms of all the important figures of merit since the early days of our space program. Numerous mission studies have shown that even extremely high array costs can be worth the investment when they result in a lower array mass. In general, mass saving in the power system can often be used by payload. If the revenue generated by this payload (i.e., more transmitters on a communications satellite) is greater than the cost of higher-efficiency solar cells, the choice is rather an easy one to make. However, often more instrument capabilities will require more support from the spacecraft (e.g., command and data handling, structure, and attitude control) as well as more power. These additions can negate any apparent advantage to the overall spacecraft.

ABBREVIATIONS 2J—two junction 3J—three junction AFRL—Air Force Research Lab AIAA—American Institute of Aeronautics and Astronautics AlGaInP—aluminum gallium indium phosphide AM0—air mass zero AM1.5—air mass 1.5 APSA—advanced photovoltaic solar array a-Si—amorphous silicon a-Si:H—amorphous silicon-hydrogen alloy (same as a-Si) ASTM—American Society of Testing and Materials BMDO—Ballistic Missile Defense Organization BOL—beginning of life CdTe—cadmium telluride CIC—cells interconnected and covered CIGS—copper indium gallium diselenide CNOFS—Communication/Navigation Outage Forecast System COI—Composited Optics Incorporated c-Si—crystalline silicon DJ—double junction e—charge of an electron E—exponential ELO—epitaxial liftoff EOL—end of life FAST—fast auroral snapshot

ABBREVIATIONS

421

GaAs—gallium arsenide GaInAs—gallium indium arsenide GaInP—gallium indium phosphide GaN—gallium nitride GaSb—gallium antimonide Ge—germanium GEC—geospace electrodynamic connection GEO—geosynchronous GSFC—Goddard Space Flight Center HES—high-efficiency silicon HIHT—high intensity, high temperature HPSA—high-power solar array i—undoped or intrinsic semiconductor IB—intermediate band IEEE—Institute of Electrical and Electronic Engineers III–V—semiconductor alloys composed of elements from columns III and V of the periodic table IMM—inverted metamorphic InP—indium phosphide IR—infrared ISS—International Space Station JPL—Jet Propulsion Laboratory LEO—low earth orbit LILT—low intensity, low temperature LiNi0.8Co0.2O2—lithium-ion thin-film battery MEMS—microelectromechanical systems MEO—medium earth orbit MIM—monolithically integrated photovoltaic module MJ—multijunction MMS—magnetospheric multiscale n—negatively doped semiconductor NASA—National Aeronautics and Space Administration NREL—National Renewable Energy Laboratory OMVPE—organometallic vapor phase epitaxy OSR—optical solar reflector p—positively doped semiconductor PERL—passivated emitter, rear locally diffused PVSC—Photovoltaic Specialists Conference QD—quantum dot QE—quantum efficiency QW—quantum well SAW—solar array wing SCOPA—survivable concentrating photovoltaic array SEC—sun–earth connection SEP—solar electric propulsion

422

SPACE SOLAR CELLS AND APPLICATIONS

Si—silicon SiC—silicon carbide SLA—stretched lens array SOA—state of the art sol—martian day of 24.6 h Sp.—specific SSP—space solar power STP—solar terrestrial probe SUPER—survivable power system TCO—transparent conducting oxide TJ—triple junction TRMM—Tropical Rainfall Measuring Mission U—university UNSW—University of New South Wales USO—united solar ovonic UV—ultraviolet VDC—volts direct current XTE—X-ray timing explorer

REFERENCES [1] [2] [3] [4] [5]

[6] [7]

E. L. Ralph. High efficiency solar cell arrays system trade-offs. Paper presented at 1st World Conference on Photovoltaic Energy Conversion, December 5–9, Waikoloa, Hawaii (1994). J. E. Granata and J. Merrill. Mass and cost comparison of lightweight array and rigid array structures. 33rd IEEE PVSC, May 11–16, San Diego, CA (2008). M. Green, K. Emery, Y. Hishikawa, and W. Warta. Solar cell efficiency tables (version 33). Prog. Photovoltaics 17(1), 85–93 (2009). R. Tatavarti, G. Hillier, A. Dzankovic, G. Martin, F. Tuminello, R. Navaratnarajah, G. Du, D. P. Vu, and N. Pan. Lightweight, low cost GaAs solar cells on 4” epitaxial liftoff (ELO) wafers. 33rd IEEE PVSC, May 11–16, San Diego, CA (2008). J. Geisz, S. Kurtz, M. Wanlass, J. Ward, A. Duda, D. Friedman, J. Olson, W. McMahon, T. Moriarty, and J. Kiehl. High efficiency GaInP/GaAs/InGaAs triple junction solar cells grown inverted with a metamorphic bottom junction. Appl. Phys. Lett. 91, 023502 (2007). R. King, D. Law, K. Edmondsonh, C. Fetzer, G. Kinsey, H. Yoon, R. Sherif, and N. Karam. 40% efficient metamorphic GaInP/GaInAs/Ge multijunction solar cells. Appl. Phys. Lett. 90, 183516 (2007). A. Barnett, D. Kirkpatrick, C. Honsberg, D. Moore, M. Wanlass, K. Emery, R. Schwartz, D. Carlson, S. Bowden, D. Aiken, A. Gray, S. Kurtz, L. Kazmerski, T. Moriarty, M. Steiner, J. Gray, T. Davenport, R. Buelow, L. Takacs, N. Shatz, J. Bortz, O. Jani, K. Goossen, F. Kiamilev, A. Doolittle, I. Ferguson, B. Unger, G. Schmidt, E. Christensen, and D. Salzman. Milestones toward 50% efficient solar cell modules. Paper presented at the 22nd European Photovoltaic Solar Energy Conference, September 3–7, Milan, Italy (2007).

REFERENCES [8]

[9]

[10] [11]

[12] [13] [14] [15] [16]

[17] [18] [19] [20] [21]

[22]

[23]

423

S. M. Hubbard, C. G. Bailey, C. D. Cress, S. Polly, J. Clark, D. V. Forbes, R. P. Raffaelle, S. G. Bailey, and D. M. Wilt. Short circuit current enhancement of GaAs solar cells using strain compensated InAs quantum dots. 33rd IEEE PVSC, May 11–16, San Diego, CA (2008). M. F. Führer, J. P. Connolly, M. Mazzer, I. M. Ballard, D. C. Johnson, K. W. J. Barnham, A. Bessière, J. S. Roberts, R. Airey, C. Calder, G. Hill, T. N. D. Tibbits, M. Pate, and M. Geen. Hot carriers in strain balanced quantum well solar cells. 33rd IEEE PVSC, May 11–16, San Diego, CA (2008). P. R. Sharps, B. Cho, A. Cornfeld, J. Diaz, F. Newman, M. Stan, and T. Varghese. Next generation high-efficiency space solar cells. Space Power Workshop Presentation, April 20–23, Manhattan Beach, CA (2009). R. Contini, E. Ferrando, D. Hazan, R. Romani, R. Campesato, M. C. Casale, and G. Gabetta. Comparison between CIGS and triple junction GaAs on thin Ge solar cell assemblies and related development strategies. 33rd IEEE PVSC, May 11–16, San Diego, CA (2008). R. Venkatasubramanian et al. Paper presented at the 26th IEEE PVSC, September 29–October 3, Anaheim, CA (1997). S. Bailey, D. Wilt, J. McNatt, L. Fritzenmeier, S. Hubbard, C. Bailey, and R. Raffaelle. Thin film poly III-V space solar cells. 33rd IEEE PVSC, May 11–16, San Diego, CA (2008). A. Luque and A. Marti. Increasing the efficiency of ideal solar cells by photon induced transitions at intermediate levels. Phys. Rev. Lett. 78, 5014–5017 (1997). A. Marti, N. Lopez, E. Antolin, E. Canovas, C. Stanley, C. Farmer, L. Cuadra, and A. Luque. Novel semiconducter solar cell structures: The quantum dot ingtermedial band solar cell. Thin Solid Films 511–512, 638–644 (2006). A. Marti, E. Antolin, C. Stanley, C. Farmer, N. Lopeq, P. Diaz, E. Canovas, P. Linares, and A. Luque. Production of photocurrent due to intermediate to conduction band transitions: A demonstration of a key operation principle of the intermediate band solar cell. Phys. Rev. Lett. 97, 247701 (2006). R. Laghumavarapu, A. Moscho, A. Khoshakhlagh, M. El-Emawy, L. Lester, and D. Huffaker. GaSb/GaAs type II quantum dot solar cells for enhanced intrared spectral response. Appl. Phys. Lett. 90, 173125 (2007). R. Raffaelle, S. Sinharoy, J. Anderson, D. Wilt, and S. Bailey. Multijunction solar cell spectral tuning with quantum dots. Paper presented at the 4th IEEE World Conference on Photovoltaic Energy Conversion, May 7–12, Waikoloa, HI (2006). D. Ferguson. Interactions between spacecraft and their environments. AIAA Paper #93-0705, NASA TM 106115, NASA Glenn Research Center, Cleveland, OH (1993). S. Bailey et al. Solar cell and array technology for future space science missions. Internal NASA Report to Code S (2002). M. Eskenazi, S. White, B. Spence, M. Douglas, M. Glick, A. Pavlick, D. Murphy, M. O’ Neill, A. J. McDanal, and M. Piszczo. Promising results from three NASA SBIR solar array technology development programs. In Proceedings of the Space Photovoltaic Research and Technology Conference, Cleveland, OH, September 16–18, pp. 59–94 (2003). A. Delleur and T. Kerslake. Electrical performance of the International Space Station US photovoltaic array during difacial illumination. In 37th Intersociety Energy Conversion Engineering Conference (IECEC), July 29–31, pp. 39–44, Washington, DC (2002). D. Scheiman, G. Landis, and V. Weizer. High-bandgap solar cells for near-sun missions. In Space Technology and Applications International Forum (STAIF-99),

424

[24] [25] [26] [27]

SPACE SOLAR CELLS AND APPLICATIONS January 30–February 4, Albuquerque, NM. AIP Conference Proceedings 458, 616– 620 (1999). G. Landis and P. Jenkins. Measurement of the settling rate of atmospheric dust on Mars by the MAE instrument on Mars Pathfinder. J. Geophys. Res., 105(E1), 1855– 1857 (January 25, 2000). R. Raffaelle et al. Integrated microelectronic power supply (IMPS). In Proc. 6th International Energy Conversion Engineering Conference, Vol. 1, July 29–August 2, Savannah, GA, pp. 239–242 (2001). P. Stella and R. Kurland. Latest developments in the advanced photovoltaic solar array program. In Proceedings of the 21st Photovoltaic Specialists Conference, May 21–25, Kissimimee, FL, pp. 569–574 (1990). S. Bailey, A. Hepp, and R. Raffaelle. Thin film photovoltaics for space applications. In Proceedings of the 6th International Energy Conversion Engineering Conference, July 29–August 2, Savannah, GA, pp. 235–238 (2001).

PART V OTHER ASPECTS AND CONSIDERATIONS

19 SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS CHRISTIAN A. GUEYMARD1 AND DARYL MYERS2 1 Solar Consulting Services, 2National Renewable Energy Laboratory

19.1

INTRODUCTION

Even if solar technologies, such as PV, were ideally 100% efficient, their absolute energy output would still be limited by the solar power they receive at any instant, or incident irradiance. The quantitative and qualitative aspects of the solar resource on planet Earth (or immediately around it) that are most pertinent to maximize the output and efficiency of PV applications and to determine their most appropriate location are described in this chapter. In the present context, the quantitative aspects will refer to the variations in total irradiance that is incident on a solar cell or PV collector, whereas the qualitative aspects will refer to its spectral and angular characteristics.

19.2

SOLAR RESOURCE IN SPACE

The very first application for solar cells was to provide power to satellites in space. PV panels are still the unique source of power for most, if not all, satellites in earth orbit. The principles and data provided here are essential for such applications but are also useful to expand the reach of PV applications to spacecrafts designed for more distant missions. The next section focuses on the quantitative aspects of extraterrestrial irradiance, whereas its qualitative aspects are discussed in Section 19.2.2.

Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

427

428

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

19.2.1

Solar Constant and Total Solar Irradiance

One frequently debated scientific question of the past was whether the total energy radiated by the sun was constant or not. A historical perspective on the successive estimates of the elusive SC can be found in the literature (e.g., see References 1 and 2). The modern era started in 1978 with the first regular measurements of the solar output from satellite platforms. These data series provided evidence that the solar output was not exactly constant but changed with solar activity. For this reason, the solar output over all possible wavelengths is now termed TSI. Until 1992, the satellite data series were affected by degradation problems of the various instruments and differences in their absolute calibration. All the concurrent data series were later reconciled using various correction methods. One such historical composite is the “PMOD” dataset [3], from the name of the leading institute for this effort (http://www.pmodwrc.ch). As of this writing, the latest daily TSI data use version d41 of the compositing algorithm1 and cover the “extended” period January 1976–June 2009, which spans over three complete solar cycles. (The actual TSI measurements only started in November 1978 but have been extended here with estimates derived from a calibrated proxy model for the period 1976–1978.) A plot of this data appears in Figure 19.1. A 27-day sliding average is superimposed to remove most of the short-term fluctuations. (The sun’s rotation periodicity is about 27 days.) Figure 19.1 clearly shows the rapid fluctuations in TSI related to solar activity. Low TSI values are associated with the passage of sunspots, which are dark, cold areas, whereas high TSI values are induced by the bright, hot areas called faculae [4]. In periods of intense solar activity, faculae

Total Solar Irradiance (W m-2)

1370

Total Solar Irradiance Composite Time Series 1976–2009

1368

1366

1364

1362

1360 1975

TSI, daily average TSI, 27-day average 1980

1985

1990

1995

2000

2005

2010

Year

Figure 19.1. Total solar irradiance between 1976 and 2009, according to version d41 of the PMOD extended composite dataset. 1

File filename “ext_composite_d41_ 62_0906.txt” is accessible at ftp://ftp.pmodwrc.ch/pub/ data/irradiance/composite/DataPlots/.

SOLAR RESOURCE IN SPACE

429

normally outnumber sunspots, so that the TSI is at a relative maximum. The TSI’s periodicity, which is apparent in Figure 19.1, perfectly matches the well-known 11-year periodicity in solar activity (e.g., see Reference 5). From Figure 19.1, the daily value of TSI can be estimated as the combination of two distinct components: the “long-term” TSI with a relatively smooth 11-year periodicity and amplitude of less than 2 W/m2 and a highly variable daily perturbation caused by short-term phenomena, such as sunspots. The latter component is negligible during low-activity periods but can be abruptly high during the peaking phase of a cycle. Particularly noticeable is the record deficit of ≈5 W/m2 that occurred on October 29–30, 2003 when 167 sunspots were recorded [1]. With the composite PMOD dataset covering more than 30 years of TSI measurement, is it possible to evaluate the absolute uncertainty in TSI? Between 1996 and 2003, there was a good agreement—within about ±0.1%—between the various radiometers then in service. Things changed dramatically, however, when a new type of radiometer went to orbit on the SORCE satellite in March 2003. Its data stream2 shows a systematic difference in comparison with the PMOD data (Fig. 19.2). For the period March 2003–June 2009 (2259 common data days), this difference averages 4.43 W/m2. The low standard deviation of this difference (0.04 W/ m2) indicates that the two series of data have high combined precision. Nevertheless, the causes for the important systematic bias are not resolved and are still debated. Considering this state of affair, it is safe to assume that the true TSI value is somewhere between the PMOD and SORCE evaluations. Even though TSI is variable on a daily basis, daily excursions of less than about ±0.3% around the mean TSI are presumably of little importance in engineering. This is why the SC concept is still in use. The latest value proposed for SC is

Total Solar Irradiance (W m-2)

1368

Total Solar Irradiance 2003–2009

1367 1366 1365 1364

PMOD composite PMOD composite 27-day average SORCE 27-day average

1363 1362 1361 1360 2003

2004

2005

2006

2007

2008

2009

2010

Year

Figure 19.2. TSI from the PMOD composite and SORCE measurements, 2003–2009.

2

http://lasp.colorado.edu/sorce/data/tsi_data.htm.

430

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

1366.1 W/m2 [5, 6], based on the PMOD data. Since then, the PMOD dataset has been extended at both ends and has undergone slight corrections, while the sun itself has been through a period of exceptional inactivity and low TSI since 2007. Using the revised PMOD composite and extended dataset between 1976 and June 2009 (totaling 11,245 days with available data), the current determination of SC (defined as either the average of all daily values or the mean between the minimum and maximum values of the 27-day sliding average) would be 1365.9 W/m2, that is, a decrease of 0.2 W/m2 from the previous SC value. An even lower value, 1361.5 W/ m2, would result from rather using the absolute calibration of the SORCE data. All the irradiance values indicated above refer to the average sun–earth distance, R0 = 1.496 × 1011 m, or ≈1 AU. Due to the eccentric orbit of the earth around the sun, this distance varies by up to ±1.67% during the year. This induces a maximum variation of ±3.34% in TSI due to the inverse square law of radiation propagation. Approximate daily values of this sun–earth distance correction factor, S, can be found in textbooks (e.g., see Reference 7). More precise values can be obtained from S = [1.00014 − 0.01671 cos g − 0.00014 cos ( 2 g )]

−2

(19.1)

where g is the mean anomaly in radians, calculated as a function of D, the number of days elapsed since the current astronomical epoch (January 1, 2000 noon terrestrial time, internationally referred to as J2000.0), such that g = 6.24006 + 0.0172019699 D.

(19.2)

The instantaneous solar irradiance incident on other celestial objects (planets, satellites, etc.) can be obtained by multiplying the TSI or SC values described above by the factor (R0/R)2, where R represents the actual distance between the object and the sun. Two concluding remarks need be made about the angular characteristics of solar irradiance. First, TSI and SC are defined for a receiver (instrument or collector) continuously tracking the sun to maximize the collected energy by keeping solar rays at normal incidence. If the angle of incidence on the receiver is not 0 but θ, the collected irradiance decreases by a factor cos θ, from Lambert’s law. Finally, TSI and SC refer to the spatially averaged radiative output over the solar disk. The effect known as “limb darkening” makes the sun’s periphery significantly darker than its center. This is of practical importance only if a fraction of the solar disk is observed with, for example, a telescope, which is normally not the case in energy production.

19.2.2

Extraterrestrial Spectrum

The TSI, En0, discussed in the previous section represents the integration of the solar spectrum over all wavelengths:

SOLAR RESOURCE IN SPACE

431 ∞

En 0 = ∫0 Enλ d λ

(19.3)

where Enλ is the spectral irradiance at wavelength λ. The subscript n denotes that all irradiances are defined here at normal incidence. The same correction factors S and cos θ that apply to En0 also apply to Enλ. The limb-darkening function also applies, but it varies strongly with wavelength [8]. The disk-average spectral distribution of Enλ at the average sun–earth distance and normal incidence is discussed in what follows. This quantity is often referred to as “air mass zero” (AM0) irradiance. A first approximation of the spectral distribution of Enλ can be obtained from Planck’s law, with the simplifying assumption that the sun is a blackbody radiator. Its radiative temperature, T, can then be obtained by the Stefan–Boltzmann law, E = σT 4. Using the latest determination3 of σ = 5.670 400 × 10−8 W/m2/K4 and the volumetric mean solar radius, Rs = 6.96 × 108 m, the radiant output of the sun is obtained as E = SC (R0/Rs)2; hence, T = 5775.9 K for SC = 1366 W/m2. In reality, the sun is not a perfect blackbody radiator, so that its real spectrum must be determined by either modeling or measurement. A review of such historical determinations of the spectrum can be found elsewhere [1, 5]. An accurate experimental determination of Enλ is more difficult than that of Eñ0 because spectroradiometers are more complex than broadband radiometers and cannot be calibrated with as much absolute accuracy at all wavelengths. Moreover, a single detector cannot cover the whole range of wavelengths in the solar spectrum, so that many instruments are needed in parallel, which complicates the process of merging their data into a single composite spectrum. Finally, extra care is needed at short wavelengths: at the earth’s surface, the atmosphere is completely opaque below about 300 nm, so that spaceborne instruments need to be used to monitor the solar spectrum below this wavelength, while exposure to these wavelengths in space leads to instrument degradation, which needs to be corrected a posteriori. Most useful spectra are in practice obtained by combining partial spectra of various sources and by scaling them appropriately so that they integrate to the exact value of SC per Equation 19.3, as described elsewhere [5, 6]. These spectra have moderate resolution, which is normally sufficient for engineering applications. For other applications, higher resolutions are sometimes necessary in the ultraviolet because of the intense structure in that part of the spectrum. Good results have been obtained [10] by normalizing a high-resolution spectrum with the lowerresolution Gueymard spectrum.4 The latter is shown in Figure 19.3 in both irradiance units (watt per square meter per nanometer, left axis) and photon flux units

3

All physical constants used in this chapter are from CODATA 2006 [9], available from http://physics.nist.gov/cuu/Constants/papers.html. 4 This spectrum is available at http://www.solarconsultingservices.com/Gueymard_ET_ spectrum.txt.

432

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

Figure 19.3. Planck radiation distribution and Gueymard’s extraterrestrial spectrum in irradiance units (left axis) and photon flux units (right axis).

(square centimeter per second per nanometer, right axis) for the 300- to 2500-nm spectral band, which is of most interest in PV applications. At any wavelength, the photon flux Fλ is related to Enλ, through Fλ = k Enλ λ ( h c ) ,

(19.4)

where h is the Planck constant (6.62606896 × 10−34 J s), c is the speed of light in vacuum (2.99792458 × 108 m s−1), and k is a numerical constant used to reconcile units. It equals 1 × 10−13 for λ in nanometer, Enλ in watt per square meter per nanometer and Fλ in square centimeter per second per nanometer. The Planck irradiance distribution at 5776 K is also indicated in Figure 19.3 for reference purposes. Its peak (at 502 nm) lies between those of the irradiance spectrum (451 nm) and of the photon flux (584 nm). There are differences in spectral irradiance between a quiet sun and an active sun (e.g., see Reference 1). However, most of these differences occur well below 400 m, where the sensitivity of solar cells is usually low. Therefore, despite the frequent variations in TSI noticed in Figures 19.1 and 19.2, a reference solar spectrum may be used without daily correction—except for the changing distance, per Equation 19.1.

ATMOSPHERIC EFFECTS

19.3

433

ATMOSPHERIC EFFECTS

The Earth’s atmosphere is a continuously variable filter. Broadly speaking, the changes in sun–collector geometry, weather, and local composition of the atmosphere constantly modify the spectral distribution and total magnitude of the incident solar radiation. 19.3.1

Basic Solar Geometry

The combination of the ≈23.45° tilt of the earth’s axis of rotation with respect to the plane of the earth’s orbit, in conjunction with the orbital period (≈365.25 days) of revolution around the sun, and the ≈24-hour planetary rotation period determines the varying solar position with respect to any location on the earth at a specified time. The local day length changes daily (except at the equator), and this can be predicted using standard astronomical data or simplified calculations. In order to calculate the position of the sun on the sky dome with respect to the local horizontal plane, the latitude and longitude of the place, the local standard time, solar declination and equation of time for the date are required (e.g., see Reference 7). Given these quantities, one can calculate the altitude of the sun above the horizon (solar elevation), or its complement, the angle between the zenith and the sun’s position in the sky (zenith angle) for a given time and place. The zenith angle, Z, will be used throughout this chapter. 19.3.2

Broadband Direct, Diffuse, and Global Radiation

The slender cone of radiation from the sun intercepted by the earth is called the direct beam, since it is well collimated. As the direct beam radiation passes through the earth’s atmosphere, two major extinction processes may affect the photons— absorption and scattering. Photons scattered by molecules and particles in the atmosphere that are not further absorbed contribute to uncollimated diffuse irradiance of the sky dome. Some radiation is reflected back into space by the atmosphere, clouds, or ground. Figure 19.4 shows these processes schematically. On a monthly average basis, very few areas on this planet experience a favorable solar resource representing more than 75% of the SC. In similarity with the concepts presented in Section 19.2, the collimated irradiance incident on a plane perpendicular to the sun’s direction is called DNI, denoted Ebn. DNI’s projection onto the horizontal plane is the direct horizontal irradiance, Eb, such that Eb = Ebn cos Z ,

(19.5)

where Z is also the angle of incidence of the sun rays (assumed to originate from the sun’s center) onto the horizontal plane. The sum of Eb and diffuse sky

434

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS Atomospheric Scattering

Reflected

Absorbed

Direct Diffuse

Ground-reflected Figure 19.4. Separation of solar “components” into direct beam, diffuse sky (combining to give total global hemispherical irradiance), and ground-reflected elements.

radiation Ed on a horizontal surface is the global horizontal (also called total or hemispherical) irradiance, E, which translates into E = Ebn cos Z + Ed .

(19.6)

FP collectors can use both direct and diffuse radiation. CPV collectors rely essentially on direct radiation, although devices with low concentration ratios can also make use of some diffuse radiation (see Section 19.4.8). Three frequently used indicators of the overall atmospheric transparency are Kn = Ebn/En0, Kt = E/(En0 cos Z), and K = Ed/E. The clearer the atmosphere, the higher Kn and Kt and the lower K will be. For a surface that is tilted at some angle, s, from the horizontal, and whose surface normal is not in the same vertical plane as that of the sun, the incidence angle, θ, is a function of s, Z, and of the difference in azimuth between the direction of the plane and that of the sun, ψ − γ: cos θ = cos s cos Z + sin s sin Z cos ( ψ − γ ) .

(19.7)

A planar collector in a tilted position (fixed or tracking) receives diffuse radiation from only a part of the sky hemisphere. This does not always result in a loss in incident diffuse radiation compared to a horizontal collector because the sky radiance is not uniform, and is actually much brighter around the sun—except under heavy overcast conditions (see Section 19.4.8). The ratio of the diffuse irradiance for the tilted position and the horizontal position is denoted Rd. Additionally, such a tilted collector may receive radiation that is either diffusively

ATMOSPHERIC EFFECTS

435

reflected from the ground in its field of view or specularly reflected by mirrors intentionally installed off its sides. This reflected irradiance is denoted Er. The total irradiance incident on the receiver, Es, becomes Es = Ebn cos θ + Rd Ed + Er ,

(19.8)

where Rd is the tilt factor for sky diffuse radiation. Whenever Es is not measured, Equation 19.8 must be used to evaluate it. Modeling is necessary for this task either to evaluate Es globally or each of the components in the rhs of Equation 19.8 separately: Ebn, Ed, Er, and Rd. Such models have been regularly proposed in the literature, and some of them are described in various textbooks (e.g., see References 7 and 11). If the sky radiance were ideally uniform or “isotropic,” Rd would be simply obtained from a pure geometrical expression: Rd = (1 + cos s ) 2 .

(19.9)

Many studies have shown that this simplification significantly underestimates the incident diffuse irradiance for tilted collectors facing the equator (i.e., facing south in the northern hemisphere or north in the southern hemisphere). Similar to Equation 19.9, it is frequently assumed that the ground is perfectly horizontal with an ideally isotropic reflectance, and with no shading, in which case one obtains Er = ρ E (1 − cos s ) 2 ,

(19.10)

where ρ is the ground reflectance, or “albedo,” whose value varies in practice between ≈0.1 for dark surfaces and ≈0.85 for fresh snow. Natural surfaces are generally not perfectly isotropic diffusers, so that their reflectance is enhanced in the direction of beam propagation, and therefore tends to increase with Z [12]. The addition of side mirrors or booster reflectors to a planar collector makes calculations even more intricate. This has been the topic of many investigations, but most of them either only address some specific reflector geometry or use intensive computer modeling such as ray tracing with no analytical solution. A general, simple, and accepted methodology that would encompass all possible reflector/collector geometries and various reflecting materials has yet to be proposed. The case of bifacial collectors, where reflectors redirect a fraction of the incident energy to the underside of a planar array, has also received some attention (e.g., see References 13 and 14). To optimize the performance of a PV module equipped with booster reflectors of any kind, it is important that the resulting enhanced irradiance is as homogeneous as possible over the area of the module. Since this is difficult to achieve all the time with any fixed module of large area, some compromise between module area and reflector enhancement is necessary. Only a few solutions have been proposed so far (e.g., see Reference 15).

436

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

The ideal and highly simplified cases described by Equations 19.9 and 19.10 almost never correspond to reality. This therefore introduces significant errors in the calculated POA irradiance, Es. Recent investigations have extensively discussed these sources of uncertainty using various models and methodologies and have found that current models had significant shortcomings under nonideal conditions [16]. It is therefore always preferable to measure Es with first-class instrumentation whenever the precise evaluation of the solar resource or the accurate monitoring of a solar installation is necessary. The relative magnitudes of the direct beam and sky diffuse radiation depend on the absorption and scattering properties of the atmosphere, which are discussed in what follows. 19.3.3

Absorption

The gases that constitute the earth’s atmosphere absorb solar radiation in specific spectral regions. These absorbers modify the solar spectral distribution of energy at the ground depending on their amount and path length of the beam. A convenient measure of this path length is the optical air mass (AM), generally denoted m, which reduces to ≈1/cos Z for Z less than about 75°. Thus, for a zenith sun, Z = 0° and AM = 1. For a rising or setting sun, Z = 90° and AM ≈ 38. Absorption processes remove energy from the incident irradiance and convert it to thermal energy in the atmosphere. Absorption is a strong function of air mass and of gas concentration. Most absorbing gases have a concentration that does not vary over time, and therefore their absorption can be easily predicted. Other gases are variable. It is well known that the concentrations of carbon dioxide and methane increase steadily due to human activity, but their effect in the shortwave spectrum is small. On a daily basis, the main variable gases are water vapor and ozone. The former has a significant effect on the spectrum above 700 nm, with strong absorption in the 920- to 980-nm, 1100- to 1200-nm, 1300- to 1500-nm, and 1750- to 1950-nm wave bands in particular. Ozone has a slight effect in the 550- to 750-nm wave band and has a strong effect below 350 nm. The ground-level ultraviolet irradiance is thus weak, although it can significantly affect the degradation of materials in particular. The amount of water vapor is highly variable over time and space. This amount is usually characterized in terms of precipitable water, w, which is defined as the height of water (usually expressed in centimeter) in a virtual column that would be obtained by condensing all the water vapor aloft. Typically, w varies between about 0.2 cm under cold, dry conditions and about 6 cm under hot, humid conditions. 19.3.4

Scattering

Scattering of radiation in the atmosphere is the redirection of photons by gas molecules (Rayleigh scattering) or particulates of various sizes (Mie scattering). With

ATMOSPHERIC EFFECTS

437

Rayleigh scattering, shorter wavelengths are scattered more strongly than longer wavelengths, explaining why the sky is blue. Besides gases, the atmosphere also contains larger particles, such as dust, sand, sea salt, smoke, or soot particles, which also scatter photons. The scattering process for these particles, referred to as “aerosols” because they are suspended in the atmosphere, is intricate. They also absorb radiation in small but variable proportions. The bulk measure of the total aerosol extinction effects due to both scattering and absorption is called AOD. The empirical Ångström’s law stipulates that, at any wavelength (between at least about 350 and 1000 nm), the total AOD by aerosols is generally proportional to a constant (β, the optical depth at 1 μm, also known as the Ångström turbidity coefficient) times the wavelength (expressed here in micrometer) raised to the inverse power of α, known as the wavelength exponent or the Ångström coefficient, such that τ aλ = β λ − α .

(19.11)

The value of α is an inverse function of the average size of the aerosols. Most aerosols are characterized by α between about 0.5 and 2.5. Volcanic and maritime aerosols tend to have a large size and therefore exhibit low α values, whereas the reverse is generally true for urban aerosols. Continental or rural aerosols stand in the middle, with α generally between 1 and 1.5. Accurate prediction of cloudless direct and diffuse irradiance cannot be done without a precise determination of AOD at all wavelengths, or at least of α and β. The main difficulty with aerosols is that both their type and total amount, and hence α and β, are highly variable over time and space. The actual AOD can be monitored with the help of ground-based multiwavelength sun photometers or of various spaceborne sensors. Sophisticated chemical transport models can also predict the worldwide distribution of AOD. Ground-based data are still by far the most accurate of these sources and should be preferred wherever possible. Under cloudless conditions, the amount of radiation removed from the direct beam and transformed in great part into diffuse sky radiation can be predicted accurately with appropriate radiative transfer models if AOD and some other atmospheric variables are known. The comparative performance of various broadband irradiance models has been investigated recently [17, 18] and shows the importance of using accurate AOD data. In addition to substantially affecting the magnitude of the direct and diffuse irradiance, AOD is also important with respect to changes in their spectral distribution (see Section 19.4.5). The two other most important variables that determine the magnitude of direct and diffuse irradiance are m and w. The relative effects of m, w, and β on DNI for a rural aerosol with α = 1.3 are shown in Figure 19.5. The thin vertical line indicates the turbidity conditions (β = 0.033 or τaλ = 0.084 at 500 nm) that have been chosen to develop the reference direct spectrum in standard G173 [19], which is further discussed in Section 19.4.6. The main other variables used for this standard spectrum are m = 1.5, w = 1.42 cm, and α = 1.2. The resulting DNI is Ebn ≈ 900 W/m2, which is indicated by a large gray dot on the vertical line. Figure

438

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS 1100

SMARTS Model Direct Normal Irradiance Sea level Rural aerosol, α = 1.3

1000 900

Precipitable water 0.5 cm 1.0 cm 2.0 cm 5.0 cm

Irradiance (W m–2)

800 700

m=1

600 500

m = 1.5

400 300

m=2 m=5

200 100

0

0.1

0.2 0.3 0.4 Ångström's Turbidity Coefficient β

0.5

Figure 19.5. Direct normal irradiance as a function of Ångström turbidity coefficient β, air mass m, and precipitable water w for sea-level conditions and rural aerosol. The vertical line indicates the turbidity conditions of ASTM G173. The large gray dot indicates the G173 DNI for m = 1.5 and w = 1.42 cm, that is, 900 W/m.

19.5 can be used to visually evaluate the change in DNI that would result from incremental changes in m, β, or w away from standard conditions. The aerosol amount in G173 is typical of clean sites at sea level. Extremely clean sites at high altitude often experience a 50% reduction in aerosol load (β ≈ 0.015). To the other extreme, the atmospheric turbidity may reach or may even exceed β = 0.5 under hazy conditions due to forest fires, desert dust storms, or heavy pollution. Section 19.4.5 further describes some useful sources of data for the main atmospheric variables to consider in irradiance predictions. 19.3.5

Cloud Attenuation

The solar resource is normally maximum under clear-sky conditions, but cloudy conditions are frequent at most locations. Clouds are constant “moving targets,” and studying their impact on the extinction of radiation is difficult. By far, the predominant attenuation process in clouds is scattering. Only very dense clouds absorb in a significant way. Attenuation effects depend on the type, microphysical

SOLAR RESOURCE FOR TERRESTRIAL APPLICATIONS

439

structure, thickness, and spatial distribution of clouds. Detailed microphysical properties, such as total liquid water path, water droplet size distributions, relative proportion of ice and water, and altitude and thickness of layers, have been found necessary to study cloud effects on solar radiation in detail [20]. At the other extreme, simple empirical results can be used to derive cloud attenuation factors for direct or global irradiance as a function of cloud fraction only (e.g., see Reference 21). The bulk transmittance of a cloud can also be modeled from its optical depth. These two essential quantities (cloud fraction and cloud optical depth) are now retrieved from various satellite sensors and are available on a worldwide basis in the form of gridded data, at more or less frequent time intervals. High thin clouds have low optical depths (typically 1–5), whereas dark dense low clouds have much larger optical depths (typically 50–90). Beam radiation is not transmitted by thick clouds, and only a fraction of it is transmitted by optically thin clouds. Clouds are near-neutral filters, so that their spectral transmittance is nearly flat in the visible.

19.4

SOLAR RESOURCE FOR TERRESTRIAL APPLICATIONS

The design and performance evaluation of any solar energy conversion system relies on information about the available solar resource. Measured data for a particular site are usually more accurate than model predictions. As described in Section 19.3.2, the basic radiation components at the ground are global, direct beam, and diffuse sky radiation. A PV system collects some combination of each of these components, possibly augmented by reflected radiation from the ground or booster reflectors. Solar radiation data and models for terrestrial applications are discussed below.

19.4.1

Broadband Global, Direct, and Diffuse Irradiance

Most commonly, radiation data are obtained from horizontal pyranometers, which sense global radiation. A pyrheliometer measures DNI with the help of a collimator to block most of the sky from its sensor (normal to the beam). The instrument’s field of view (about 6°) accepts some sky radiation around the ≈0.5° disk of the sun, termed circumsolar radiation (Section 19.4.8). DNI is less frequently measured than global radiation because of the need for a costly sun tracker. Diffuse radiation can be measured with a pyranometer using either a shadow band or a shading disk to mask the sun. At many sites, “direct irradiance” data are actually indirectly calculated using simultaneous measurements from a pair of pyranometers (one measuring global radiation and the other diffuse radiation) or from a rotating shadow band radiometer where global and diffuse radiation are measured in rapid succession with a single silicon cell sensor [22]. Such derived direct beam measurements are usually less accurate than pyrheliometer measurements [23, 24].

440

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

More generally, the best current practice in broadband irradiance measurement and related accuracy issues have been reviewed recently, showing that the basic accuracy of measured solar data is often not as good as one would expect [23, 25].

19.4.2

U.S. NSRDB

The NREL has established an NSRDB. The initial version of the database includes 30 years of hourly solar radiation and meteorological data for 239 sites in the United States [26]. In 2007, an update to the database was generated to include 1430 sites (including most of the original 239 sites) with hourly solar radiation and meteorological date for the period from 1991 to 2005 [24]. Both the original and updated NSRDB datasets contain more than 90% modeled solar radiation estimates from meteorological parameters, resulting in lower accuracy. The original NSRDB calculations were produced using the METSTAT model [27]. The 1991–2005 update required modifications to the original METSTAT model because of changes in the availability of certain input parameters. The update also includes hourly solar radiation estimates on a ≈10-km grid, derived from geostationary satellite data [28] for the period 1998–2005. Total global horizontal, direct beam, and diffuse irradiances for a particular cell in this latter dataset can be accessed interactively.5 Typical uncertainties for measured and modeled data in these databases is 10% for global horizontal irradiance and 15% for DNI but depends more specifically on the quality of input data and availability of some ancillary cloud cover data [24]. The objectives of the NSRDB were to provide consistent serially complete hourly time series solar data able to produce reasonably accurate (on the order of 5%) longer-term (e.g., monthly means) averaged data. Each individual hour of data could be wildly different from an actual measurement at that specific time because actual positions and geometry of clouds could not be accounted for. Statistical algorithms were developed and validated [27] to produce the correct statistical properties of frequency distributions for the long-term data. Data manuals containing computed irradiances, derived from the NSRDB basic radiation data, relative to various collector configurations such as latitude tilt, full tracking, and east–west tracking, were also prepared [29] (see http://rredc.nrel. gov/solar/ and the Appendix).

19.4.3

Solar Radiation Data for International Locations

Some national and international organizations collect and distribute radiation data from various networks for scientific research purposes. Examples include WRDC 5

http://www.nrel.gov/gis/solar.html.

SOLAR RESOURCE FOR TERRESTRIAL APPLICATIONS

441

[30], BSRN [31], ARM [32], and NOAA SURFRAD [33]. The data available from these organizations’ websites have varying resolution ranging from 1 min (ARM and SURFRAD) to daily totals and monthly totals (WRDC). Most nations have operational solar monitoring networks but have no public access to the data, although it may sometimes be purchased. The Appendix provides more details about available sources of measured and modeled data.

19.4.4 Irradiance for Fixed-Tilt, Tracking, and Concentrating Geometries When locally measured solar radiation data cannot be obtained, one falls back on solar radiation models to generate global, direct, and diffuse irradiance. Generally, one starts with clear-sky models that evaluate the radiation components from atmospheric input data, per the discussions in Sections 19.3.3 and 19.3.4. These clear-sky irradiances are then modified to take into account the effect of clouds (Section 19.3.5) to produce “all-sky” irradiance values. Various aspects of these methodologies are discussed in Reference 11, for instance. If only global horizontal irradiance is available, and one needs to estimate DNI, one of many global-to-direct conversion models of the literature needs to be used. Their performance is regularly evaluated against measured data (e.g., see References 16 and 34). Although the mean bias error of these estimates is generally reasonable on a long-term basis (depending on local climatic conditions), the hourly or sub-hourly random errors are considerable with all existing models. This important accuracy issue must be considered when using such modeled data for the design or performance simulation of CPV collectors, for instance. For fixed-tilt and tracking FP collectors, several models have been developed to solve Equation 19.8 and to produce estimates of POA irradiance, the most popular of which is the Perez model [35]. For instance, it is used in PV performance models such as PVFORM [36]. The performance of this type of models has been studied extensively, for a wide range of climatic conditions, instrumentation, local specificities, and experimental setups. Not surprisingly, these studies do not always concur in their conclusions. Recent findings [16] suggest that the accuracy of predicting the POA irradiance is seriously limited by a series of factors, including the high sensitivity of these models to uncertainties in direct irradiance, which is one critical input. Moreover, the usual complexity of local conditions (e.g., shading from obstacles and variance in ground reflection processes) is simply not considered in current models. It is therefore impossible to recommend a single model that would perform best under all possible conditions. As in the case for the globalto-direct conversion models, large random errors in the modeled irradiances (and even sometimes, significant bias errors for some POA geometries or locations) can be expected under realistic conditions. As overviewed in the Appendix, a wide diversity of modeled solar radiation products now exists. More products are actually described in the literature, mainly for climate research purposes. Although this diversity is a good thing, it also raises

442

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

the issue of which data source is preferable to obtain the needed information at any potential site in the world. This is currently an unresolved question because all models and their input data have weaknesses. Even though they might be validated in broad terms, it does not mean they are accurate under any condition, at any site, all the time. A few recent studies [37–39] have just scratched the surface of this practical problem and show significant differences from one dataset to the other. Many more in-depth studies will be necessary before it becomes possible to provide specific recommendations to all those who need highly accurate data to design or to simulate large and costly projects. Another issue of significant importance in the practice of designing PV systems is to rapidly compare the resource available for different possible collector geometries. This has received recent attention [40], based on the NSRDB modeled data discussed above. This study intercompared the annual and seasonal (summer and winter) resources for 14 different collector geometries and introduced the SREF method. A few results are outlined below and are presented in the form of questions and answers. Moreover, a numerical example is proposed at the end of this section. 1.

2.

If fixed planar collectors are to be used, what is their best tilt to maximize energy production during the summer, winter, or year? A conventional rule of thumb is that the optimal tilt angle is equal to latitude (L) for maximum annual energy production, L − 15° for maximum summer production and L + 15° for maximum winter production. Detailed calculations show that the L and L − 15° geometries both maximize the annual energy collection, therefore indicating a broad optimum. An equator-facing orientation is assumed throughout but is not too critical either [41]. At least for all the United States except Alaska, the L − 15° geometry is indeed ideal for summer, but only marginally better than a horizontal setup. For winter, the L + 15° geometry is the most favorable, but only 1–5% better than the latitude tilt setup. All these results, of course, are for ideal conditions and therefore do not consider the incidence of local microclimatic factors, partial shading, risks of snow accumulation on the collector, structural costs, aesthetics, and so on. Knowing that systems involving tracking devices cost more than a fixed design, how much more resource is available to a two-axis tracking FP than to a latitude tilt fixed collector? A two-axis tracking planar collector always benefits from the largest possible resource per collector unit area. (It is larger than what a two-axis concentrator receives at any time and location since the planar collector makes additional use of the diffuse irradiance from the sky and ground.) The annual energy enhancement of the two-axis planar geometry compared to the fixed-tilt geometry is from 25% (in cloudy climates) to 42% (in sunny climates) annually (Fig. 19.6). This range changes to 15–35% for winter operation or to 25–50% for summer operation.

SOLAR RESOURCE FOR TERRESTRIAL APPLICATIONS

443

Figure 19.6. Annual irradiation incident on a two-axis tracking flat-plate collector and a two-axis tracking concentrator, normalized by the annual irradiation incident on a latitude tilt flat-plate collector, as a function of the annual diffuse/global irradiation ratio, K. Data were processed from the 1961–1990 NSRDB (excluding Alaska).

3.

4.

Is the resource available to a two-axis concentrating collector significantly larger than to a single-axis concentrator? Remarkably, a single-axis concentrator oriented N-S and tilted at latitude receives nearly as much energy (94–96%) as a two-axis concentrator, irrespective of season and location. If the single axis is horizontal, however, the resource decreases markedly with latitude and would not be advisable to maximize winter production. Nevertheless, for summer production, this design provides 95–98% of the two-axis resource at latitudes less than 37°. Does a two-axis concentrator always benefit from a better resource than a latitude tilt planar collector? The main factor at play here is not latitude but atmospheric clearness, which can be represented by the annual diffuse-to-global irradiation ratio, K. The two-axis tracker can experience a larger annual resource only at sunny locations where K is less than 0.30–0.35. The sunniest locations in the United States are characterized by K ≈ 0.25, and the additional resource of the two-axis tracker is only about 10% there. At the other extreme, for cloudy locations and K in the 0.50–0.55 range, the two-axis tracker is harshly penalized with ≈30% less resource than the fixed latitude tilt collector (Fig. 19.6).

444

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

Example Evaluate the relative merits of several possible PV collector geometries to maximize annual collection at San Ramon, California (latitude L = 37.78°N, longitude 121.97°W), using the SREF method. Two possible sources of public domain long-term data are possible: the 1961–1990 NSRDB and the 1991–2005 NSRDB Updates. The closest location in the former dataset is Sacramento, 91-km NE. The closest locations in the latter dataset are Hayward, Oakland, and Livermore. The two first sites are directly on the Bay, whereas the latter site is in the same climatic environment as San Ramon and only 16-km SE, and is therefore preferable for this analysis. The monthly mean solar data for Sacramento are extracted from http://rredc.nrel.gov/solar/old_data/ nsrdb/1961-1990/dsf/data/23232.txt. Similarly, the data for Livermore are extracted from http://rredc.nrel.gov/solar/old_data/nsrdb/1991-2005/statistics/dsf/724927. dsf. The headers of these files are reproduced below. ID

City

State

Latitude (N)

Longitude (W)

23232

Sacramento

CA

38.52

121.50

724927

Livermore municipal

CA

37.70

121.82

Elevation (m)

Time Zone

Pressure (mb)

8

−8

1015

121

−8

999

The next step is to extract from these files the annual average daily mean values of global (H), direct normal (Hbn), diffuse (Hd), extraterrestrial on the horizontal (H0), and extraterrestrial normal (Hn0) irradiations, and report their values (in watt hour per square meter) in the next table. Sacramento, 1961–1990 H 4933

Livermore, 1991–2005

Hbn

Hd

H0

H n0

H

Hbn

Hd

5505

1556

8143

16405

4712

4892

1684

H0

Hn0

8142 16538

Using this table, the annual values of K = Hd/H, Kt = H/H0, and Kn = Hbn/Hn0 are calculated. The value of cos L is also needed to calculate the annual SREF factors (which provide the average irradiation on a collector relative to that incident on a latitude tilt planar collector or on a two-axis concentrator) using equations that are provided in the original publication [40] but are not repeated here for conciseness. The results are summarized in the following tables. Sacramento, 1961–1990 K 0.3154

Livermore, 1991–2005

Kt

Kn

cos L

K

Kt

Kn

cos L

0.3356

0.6058

0.7824

0.3574

0.2958

0.5787

0.7912

SOLAR RESOURCE FOR TERRESTRIAL APPLICATIONS

Geometry

445

SREF Sacramento

SREF Livermore

Fixed FP tilt = 0°

0.874

0.887

Fixed FP tilt = L−15°

0.985

0.987

Fixed FP tilt = L (reference)

1.000

1.000

Fixed FP tilt = L + 15°

0.960

0.959

Fixed FP tilt = 90°

0.643

0.632

Tracking FP one-axis N-S tilt = 0°

1.225

1.206

Tracking FP one-axis N-S tilt = L − 15°

1.315

1.290

Tracking FP one-axis N-S tilt = L

1.325

1.300

Tracking FP one-axis N-S tilt = L + 15°

1.293

1.267

Tracking FP two-axis

1.369

1.342

Tracking conc one-axis E-W tilt = 0°

0.656

0.624

Tracking conc one-axis N-S tilt = 0°

0.892

0.837

Tracking conc one-axis N-S tilt = L

0.974

0.914

Tracking conc two-axis

1.015

0.952

FP, flat-plate; conc, concentrator.

These results indicate that a tracking two-axis FP collector located in San Ramon would experience 34–37% more resource than the reference latitude tilt collector (indicated by italics in the above table). A two-axis concentrator (without consideration of the concentration factor, since only the outer collecting area of the receiver’s normal plane is considered here) would be nearly on par with the latter. The difference in SREF obtained at Sacramento and Livermore is not significant for FP collectors. For CPV collectors, however, the inland site (Sacramento) appears a “better” environment than Livermore. Results for San Ramon should be very close. To obtain absolute energy values, rather than SREF numbers relative to the latitude tilt FP, the annual irradiation incident on this reference collector is needed. Such data can be obtained from the 1961–1990 NSRDB6 or from an interactive 10-km resolution map7 for the period 1998–2005. For Sacramento, the first source provides a mean daily irradiation of 5.5 kWh/m2. For Livermore and San Ramon, the second source provides 5.542 and 5.483 kWh/m2, respectively. Using the Livermore data in the above table, for instance, the estimated annual resource for two-axis collectors would be 5.276 kWh/m2 for a concentrating geometry (irrespective of the concentration ratio) and 7.437 kWh/m2 for an FP geometry. Note the

6 7

http://rredc.nrel.gov/solar/old_data/nsrdb/1961-1990/redbook/sum2/state.html. http://mapserve2.nrel.gov/website/Resource_Atlas/viewer.htm.

446

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

lower concentrator resource in the former case reflects the inability to utilize diffuse radiation. Furthermore, whenever a CPV system is an option, the concentration ratio, engineering constraints, and associated risks—due, for instance, to the large interannual variability in DNI [42]—and costs significantly affect the selection of the most appropriate system.

19.4.5

Spectral Irradiance: Measurements and Modeling

Knowledge of the spectral distribution of the incident radiation on solar cells is important since their sensitivity is highly variable with wavelength. Spectral information is needed to optimize the deployment of PV technologies in different ways. For instance,

•

•

Multijunction cells can be tweaked to maximize their output under some specific spectral distributions rather than under a single reference spectrum. As a typical example, the effect of varying the incident spectrum on the performance of a tandem cell arrangement has been shown to be significant, therefore prompting adjustments to the top cell thickness so that the cell efficiency is maximized under each possible spectrum [43]. The actual performance of any solar cell under specific spectral irradiance conditions can be related to that under a reference spectrum, which is used for rating purposes. More generally, this leads to the notion of spectral mismatch, which is further discussed in Section 19.4.7.

Routine outdoor spectral measurements, using spectroradiometers, are conducted at only a few sites in the world, including NREL.8 Such measurements require sophisticated instrumentation, skilled personnel to perform regular calibrations, and delicate quality control. This explains their paucity and why most potential users prefer using modeled data. There exists a wide variety of radiative transfer models for atmospheric applications, as reviewed elsewhere [44]. However, they are complex to use and do not provide the spectral irradiance incident on tilted planes or the circumsolar contribution to direct irradiance (see Section 19.4.8). For solar engineering applications, the choice is therefore limited [45]. SMARTS has been used extensively in PV applications under cloudless conditions. Its algorithms are described thoroughly [44] and have been extensively validated against spectroradiometric measurements and advanced radiative calculations of direct and global irradiance on various tilts [46]. A comparison between predicted direct irradiance spectra and their measured counterpart at NREL (measurement uncertainty ±4% over most of the spectral range) appears in Figure 19.7 for two different air masses. The predicted spectra were smoothed here using a Gaussian filter tool in the SMARTS post-processor to 8

http://www.nrel.gov/midc/srrl_bms/.

SOLAR RESOURCE FOR TERRESTRIAL APPLICATIONS

447

Figure 19.7. Comparison between direct normal irradiance spectra predicted by SMARTS and measured with a field spectrometer at NREL under clear-sky conditions.

simulate the effect of the instrument’s bandwidth. The aerosol input data (AOD) are provided by a collocated multiwavelength sun photometer. The AOD at various wavelengths is a critical atmospheric input, which ultimately conditions the accuracy of direct and diffuse irradiance (both spectral and broadband). It is therefore essential to obtain accurate AOD data (i.e., with an uncertainty no larger than ±0.02, representing high-quality measurements). Proximity from a sun photometer site (e.g., see NASA’s Aeronet9) is the best option to obtain such data, the reduction of which is described elsewhere [44].

19.4.6

Reference and Standard Spectra

Reference spectra are needed to characterize the performance of solar cells, to compare, and to rate them against a unique standard spectral irradiance distribution. For terrestrial applications, the PV industry has relied on two distinct standard spectra since the early 1980s: a direct irradiance (DNI) spectrum and a global irradiance spectrum for a surface tilted at 37°, both for an air mass of 1.5 (hence, Z = 48.2°). A historical perspective [47] describes how these spectra were originally developed and why this specific geometry was selected. Typically, the global spectrum (“AM1.5G”) is used to evaluate the performance of solar cells for FP PV 9

http://aeronet.gsfc.nasa.gov.

448

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

Figure 19.8. Comparison between the old (thick lines) and new (thin lines) reference AM1.5D and AM1.5G reference spectra.

systems, whereas the direct spectrum (“AM1.5D”) is used for CPV applications. These measurement principles are defined in standard IEC 60904 for planar systems, for instance. The rapid development of high-efficiency multijunction cells and optical concentrator devices since the late 1990s led to the realization that the earlier direct reference spectrum (ASTM E891) was not representative of the very clear/sunny conditions where such technologies would be deployed. A temporary “low-AOD” spectrum, obtained with SMARTS version 2.8, was then circulated [48]. Based on the findings in [47], the older reference AM1.5 spectra were eventually deprecated and replaced by those in a new standard, G173 [19]. The AM1.5D reference spectrum that it contains is based on SMARTS version 2.9.2, and is close to the “lowAOD” temporary spectrum. The new AM1.5G spectrum is not too different from the older one, besides an order of magnitude better spectral resolution. A comparison of the old and new direct reference spectra is shown in Figure 19.8. More recently, similar reference spectra, but for irradiance incident on vertical and 20°-tilt planes, have been standardized [49]. They are designed to provide spectral information that can be useful in BIPV applications, for instance.

19.4.7

Spectral Mismatch Calculations

The performance of solar cells is evaluated using a set of standard conditions, which specifies a reference spectrum. For example, the 2008 revision of IEC

SOLAR RESOURCE FOR TERRESTRIAL APPLICATIONS

449

60904-3 now specifies an AM1.5G spectrum that is equivalent to the G173 AM1.5G spectrum multiplied by 0.9971. This produces an integrated irradiance of exactly 1000 W/m2 for wavelengths between 0 and infinity. When the output of a new cell is measured by comparison to that of a reference cell in the laboratory (using a solar simulator) or outdoors (using the sun), there are intricate spectral effects to take into account due to the differing spectral responses involved and the difference between the actual test spectrum and the reference STC spectrum. These effects can be condensed into a spectral mismatch factor, whose calculation procedure is defined by standards (ASTM E973 and IEC 60904-7). The same procedure can be generalized, for instance, to evaluate the impact of the pure spectral effect on the performance of PV modules of different materials and spectral responses when subjected to non-STC spectral conditions at a specific location. Since the effective incident spectrum is rapidly changing over the day, it is convenient to use a spectral radiation model such as SMARTS to generate “typical” spectra at regular intervals (e.g., hourly) during an average day at a specific location for which atmospheric data are available. Hence, a DSEF can be defined as a generalized and time-averaged spectral mismatch factor [50], with no dependence on a reference cell in this case:

∫280 Sλ d λ ⎡⎣ ∫sr Ee (t ) Eeλ (t ) dt ∫sr Ee (t ) dt ⎤⎦ 4000

DSEF =

ss

ss

4000

∫280 Erλ Sλ d λ

(19.12)

where Erλ is the reference STC spectrum; Eeλ(t) is the effective incident spectral irradiance at time t; Ee(t) is its broadband (spectrally integrated) counterpart; Sλ is the cell’s spectral response; and sr and ss are the times of sunrise and sunset, respectively (assuming no horizon shading). The weighting scheme in the numerator ensures that the large spectral effects near sunrise and sunset are not overrated. The output of a PV module will appear to be enhanced (i.e., DSEF > 1) relative to its STC rating if the PV material’s spectral response is consistently closer to the average incident spectrum than to the STC reference spectrum. Typically for instance, DSEF is lower than 1 for crystalline Si and higher than 1 for amorphous Si due to their widely different spectral response. For CPV applications, AOD is found to have the largest effect on DSEF [50].

19.4.8

Angular Characteristics and Circumsolar Effect

For non-concentrating applications, the sun can be considered as a point source. Things are different for focusing solar systems with high-concentration ratios. The limb-darkening effect described in Section 19.2.2, augmented by various optical aberrations caused by reflecting surfaces, creates an image of the sun disk with a radial Gaussian distribution of its energy flux [51]. For all CPV systems (imaging or nonimaging), a related issue is the characterization of the large diffuse radiance

450

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

in the sun’s aureole, referred to as circumsolar radiation. Although it is diffuse in nature, it behaves essentially as direct radiation and can therefore be utilized by concentrators. The variation of radiance from the sun center to the edge of the circumsolar region intercepted by a concentrating collector is often referred to as “sunshape.” Earlier studies of the sunshape [51] have been based on experimental measurements of the radiance that were carried out at 11 U.S. locations during 1976–1981 [52]. This dataset has been quality controlled and is now available online.10 More recently, similar data have been obtained with different experimental setups installed at solar furnaces in Switzerland [53] and Germany [54]. Recent studies show to what extent these different results were in agreement, while providing simplified analytical modeling that might be used in concentrating systems (e.g., see Reference 55). The main shortcoming of the proposed empirical sunshape profile functions, however, is that they do not separate the clear from the cloudobscured line-of-sight cases and are not made dependent on the sun position (Z or m), nor on the atmospheric scattering properties (AOD and/or cloud optical depth), which are all key variables to describe the circumsolar effect. It is hoped that the needed refinements will result from the deployment of a recently introduced instrument, SAM [56]. A typical result of SAM measurements appears in Figure 19.9, showing radiance profiles at 670 nm under four different sky conditions, characterized by their optical depth11 between 0.046 and 0.578 (at 670 nm). These measurements12 have been obtained in Burlington, Massachusetts, and Greenbelt, Maryland, and show remarkable differences in radiance patterns depending on the sky condition. Since only sophisticated instrumentation can be used to automatically monitor the actual sunshape, it is unlikely that such measured data will become available on the geographic scale needed for CPV applications, for instance. Various theoretical studies have provided descriptions of the spectral or broadband circumsolar radiance under specific atmospheric conditions of sun position and aerosol type and amount. Some of these studies also analyzed the sky obscuration by thin clouds, demonstrating the considerable increase in circumsolar radiation in such cases. As a general result, whenever the direct irradiance from the sun decreases, due to either increasing air mass or increasing optical depth from aerosols or clouds, the circumsolar effect increases. This additional resource recoups only a fraction of the lost direct irradiance, however. The complexity of such theoretical studies precludes their use for repetitive calculations. A simplified clear-sky aureole model has therefore been derived [57] for applications where both the solar and circumsolar irradiance are needed, and has been integrated into the SMARTS model described in Section 19.4.3. This

10

http://rredc.nrel.gov/solar/old_data/circumsolar/. This is the total optical depth for both aerosol and cloud (if present). 12 These data have been extracted from Visidyne’s public repository, ftp://ftp.visidyne.com/, and have been kindly processed by their scientists at our request. 11

SOLAR RESOURCE FOR TERRESTRIAL APPLICATIONS

451

Figure 19.9. Circumsolar radiance at 670 nm as measured by the SAM instrument at Burlington, Massachusetts, and Greenbelt (NASA’s Goddard Space Flight Center), Maryland. The vertical line indicates the edge of the solar disk.

simplified aureole model has been validated against theoretical results from the literature. It allows the prediction of the circumsolar radiance and irradiance up to 10° from the sun center, when the aerosol type is known or can be inferred from spectral measurements. Assuming that the circumsolar spectral radiance at a distance ξ from the sun center, Nc(ξ, λ), has radial symmetry, the corresponding broadband irradiance, Ec(ξ0), within the acceptance angle ξ0 is calculated numerically by SMARTS from λ2

ξ0

1

s

Ec ( ξ 0 ) = 2π ∫λ d λ ∫ξ N c ( ξ, λ ) sin ξdξ,

(19.13)

where ξs is the angular distance from the sun’s center to its outer edge (0.2665° at mean sun–earth distance), and λ1 and λ2 are the wavelength limits of the shortwave spectrum (e.g., 280 and 4000 nm as used by SMARTS). Typical results are shown in Figure 19.10 for a rural aerosol model (which should be representative of most likely cases in CPV applications), varying air mass (m = 1–3), and typical AOD for very clear (β = 0.025), average (β = 0.15), and hazy (β = 0.45) conditions. It is obvious that, for these conditions, the circumsolar contribution increases almost linearly with the acceptance angle of the receiver.

452

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

Figure 19.10. Circumsolar irradiance (in percent of the broadband direct irradiance emanating from the sun’s disk) calculated by the SMARTS model as a function of the receiver’s acceptance angle, ξ0, air mass, m, and aerosol optical depth at 1 μm, β. Each of the three groups of β results includes four curves, in ascending order of air mass from bottom to top. The acceptance angle of the most common pyrheliometer is indicated by a vertical line, showing by how much actual direct irradiance measurements can be overestimated in practice. The large gray dot on this line indicates the circumsolar contribution (about 0.25%) for the conditions of ASTM G173.

For a point-focus concentrator, the relationship between acceptance angle ξ0 and geometrical concentration C can be defined [58] by C = 1 sin 2 ξ 0 ,

(19.14)

so that C = 33.2 for ξ0 = 10°. The results in Figure 19.10 can therefore be considered as typical for focusing systems with concentration ratios above 30, which is the case for CPV power plants. These results can also be used to evaluate how the DNI measured with a regular “normal incidence pyrheliometer” should be corrected13 to obtain the actual irradiance received within the field of view of a specific concentrator for any C above 30. 13 The measured DNI should be corrected downward if ξ0 < 2.9° or upward otherwise with this type of pyrheliometer. This 2.9° limit might be slightly different from other pyrheliometer models.

SOLAR RESOURCE FOR TERRESTRIAL APPLICATIONS

453

In addition to CPV power plants, there is a new segment of the CPV market that makes use of low concentration ratios for residential and commercial applications, with the help of simple low-cost concentrators or booster reflectors. In such cases, C is typically between 1.4 and 5.0, resulting in large acceptance angles of up to 45° (assuming line-focus concentration, for which Eq. 19.14 must be replaced by C = 1/sin ξ0). The curves in Figure 19.10 cannot be extrapolated to 45° because multiple scattering processes significantly modify the sky radiance patterns beyond about 10° of the sun center. Broader measurements or calculations are therefore needed in this case. Various experiments have been conducted to measure the distribution of sky radiance across the whole sky hemisphere with sky scanners of different designs, and some empirical models have been derived from these measurements. Modeling the instantaneous radiance under partly cloudy skies is obviously most challenging, and only limited information is available [59]. So far, no comprehensive intercomparison of all the existing radiance models has been done. This is unfortunate because the limited results shown in Figure 19.11 indicate that the cloudless radiances predicted by some of these models widely disagree in the solar aureole, where accuracy matters most in CPV applications. This is a disturbing finding because cloudless conditions are supposed to be the easiest to model. Consequently,

Figure 19.11. Cloudless-sky radiance distributions predicted by various models for moderately clear summer conditions (Z = 30° and β = 0.15, left) and very clear winter conditions (Z = 60° and β = 0.025, right). These distributions are along the sun’s principal plane, that is, the vertical plane containing the zenith and the sun. The position of the sky element is defined by its zenith angle, counted here positively in the sun quadrant and negatively in the opposite direction. The vertical line indicates the sun’s position. The sky radiance is normalized by the horizontal diffuse irradiance. This relative radiance is also corrected so that its spatial integration over the sky hemisphere equals 1.

454

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

none of the existing models can be recommended for the task of predicting the radiance between 10° and 45° of the sun’s center until a detailed assessment is conducted conclusively. In the mean time, two of the possible models [60, 61] are suggested for approximate evaluations.

19.5

CONCLUDING REMARKS

The information discussed in this chapter provides the basis for what PV system designers need to quantify or qualify the solar resource available for either space or terrestrial applications. High-accuracy data can be obtained for space applications, with typical uncertainties of ±0.3% for broadband irradiance and ±5% for spectral irradiance. For terrestrial applications, the situation is more complex since it depends on the type of measurement or model used, quality of input data for models, spatial or temporal extrapolation methods used, and so on. On-site measurements with first-class thermopile instrumentation, optimal maintenance and quality control, and state-of-the-art calibration and correction methods constitute the ideal case. Uncertainties of about ±3% in DNI, ±5–7% in global irradiance, and ±5–10% in diffuse irradiance are then achievable. At the other extreme, when irradiance needs to be modeled using atmospheric data or needs to be highly extrapolated, much larger bias and random errors can be expected. On a monthly average basis, uncertainties can reach ±10–15% in global irradiance and ±15–20% in DNI, or more at cloudy sites. The latter figures will have to be improved in the future to meet the requirements of so-called bankable data needed for megawattscale solar power plant projects, for which a precise evaluation of the solar resource is essential. Solar radiation modeling is doing rapid progress, while better worldwide input data sources from, for example, spaceborne sensors become regularly available, so that “high-end” data users who need the best accuracy possible should stay regularly informed. Long-term variations in solar radiation for the next 20–50 years, in relation with climate change, should also be considered for realistic power output predictions.

APPENDIX: SOURCES OF MEASURED OR MODELED SOLAR RADIATION DATA A large number of data sources are listed below. The geographic area they cover is indicated by an acronym: United States (USA), Europe (EU), other parts of the world (PW), or worldwide (WW). A dollar sign ($) indicates data for purchase. Measured Data ARM: http://www.arm.gov/data/datastream.php?id=sirs (USA) BSRN: http://www.bsrn.awi.de/en/home (WW)

APPENDIX

455

IDMP: http://idmp.entpe.fr/ (WW) NOAA historical data: http://rredc.nrel.gov/solar/old_data/noaa/ (USA) NOAA/ISIS: http://www.srrb.noaa.gov/isis/index.html (USA) NOAA/SURFRAD: http://www.srrb.noaa.gov/surfrad/ (USA) NREL’s Renewable Resource Data Center (RredC) Solar Data: http://www.nrel.gov/rredc/solar_data.html (USA) http://rredc.nrel.gov/solar/new_data/confrrm/ (USA) http://rredc.nrel.gov/solar/old_data/hbcu/ (USA) http://rredc.nrel.gov/solar/old_data/circumsolar/ (USA) http://rredc.nrel.gov/solar/old_data/spectral/ (USA) http://rredc.nrel.gov/solar/old_data/semrts/ (USA) http://rredc.nrel.gov/solar/old_data/wa/ (USA) http://www.nrel.gov/midc/ (USA) http://rredc.nrel.gov/solar/new_data/Saudi_Arabia/ (PW) SDW: http://solardatawarehouse.com/ (USA) ($) University of Oregon: http://solardat.uoregon.edu (USA) University of Texas: http://www.me.utexas.edu/∼solarlab/tsrdb/tsrdb.html (USA) WRDC: http://wrdc-mgo.nrel.gov/ or http://wrdc.mgo.rssi.ru/ (WW) Mostly Modeled Data and Solar Resource Maps 3Tier: http://firstlook.3tier.com/solar/ (WW) ($) AWS Truepower: http://www.awstruepower.com/solutions/solar/resourceassessment/ (WW) ($) Clean Power Research: http://www.cleanpower.com (USA) ($) DAYMET: http://www.daymet.org/ (USA) DLR/SOLEMI: http://www.solemi.de/home.html (EU, PW) ($) DLR/ISIS: http://www.pa.op.dlr.de/ISIS/ (WW) EnMetSol: http://www.energy-meteorology.de/ (WW) ($) FocusSolar: http://www.focussolarusa.com/ (WW) ($) GeoModel: http://geomodel.eu/ (WW) ($) Green Power Labs: http://greenpowerlabs.com/solar_GIS.html (PW) ($) Helioclim/ESRA: http://www.helioclim.net/esra/ (EU, PW) ($) IrSOLaV: http://www.irsolav.com/ (WW) ($) MeteoControl: http://www.meteocontrol.de/ (WW) ($) Meteonorm: http://www.meteonorm.com/pages/en/meteonorm.php (WW) ($) NASA/POWER: http://power.larc.nasa.gov/ (WW) NASA/SSE: http://eosweb.larc.nasa.gov/sse or http://power.larc.nasa.gov/ (WW) NREL’s Solar Spectral Distributions: http://rredc.nrel.gov/solar/spectra/ NSRDB: http://rredc.nrel.gov/solar/old_data/nsrdb (USA) PVGIS: http://re.jrc.ec.europa.eu/pvgis/index.htm (EU) (PW) Satel-Light: http://www.satellight.com/indexgS.htm (EU) SoDa: http://www.soda-is.org/eng/index.html (WW) ($) SolarGIS: http://solargis.info/ (WW) ($) SWERA: http://swera.unep.net/ (PW)

456

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

Typical Meteorological Years (TMY) These are datasets consisting of hourly data over 365 days, normally including solar radiation, temperature, humidity, wind speed and direction, cloud cover, as well as additional pertinent information for the energy simulation of solar systems or complete buildings. Various methodologies have been proposed to develop such synthetic years, including the TMY2 and TMY3 methods used at NREL. http://rredc.nrel.gov/solar/old_data/nsrdb/tmy2/ (USA) http://rredc.nrel.gov/solar/old_data/nsrdb/1991-2005/tmy3/ (USA) http://apps1.eere.energy.gov/buildings/energyplus/cfm/weather_data.cfm (WW).

ABBREVIATIONS AM—air mass AOD—aerosol optical depth ARM—Atmospheric Radiation Measurement Program ASTM—American Society for Testing and Materials BIPV—building integrated photovoltaic BSRN—Baseline Surface Radiation Network c—speed of light (2.99792458 × 108 m/s) C—concentration ratio of a solar cell or system CPV—concentrating photovoltaic D—number of days elapsed since noon, January 1, 2000 DNI—direct normal irradiance DSEF—daily spectral enhancement factor E—global irradiance incident on a horizontal surface (W/m2) Eb—direct irradiance incident on a horizontal surface (W/m2) Ebn—DIRECT irradiance incident on a surface normal to sun rays (W/m2) Ec—circumsolar irradiance incident on a surface normal to sun rays (W/m2) Ed—diffuse sky irradiance incident on a horizontal surface (W/m2) Ee—effective incident irradiance (W/m2) Eeλ—effective incident spectral irradiance (W/m2 nm) En0—total solar irradiance at the limits of Earth’s atmosphere (W/m2) Enλ—spectral solar irradiance at the limits of Earth’s atmosphere (W/m2 nm) Er—reflected irradiance incident on a tilted receiver (W/m2) ERλ—reference STC spectral irradiance (W/m2 nm) Es—total irradiance incident on a tilted receiver (W/m2) Fλ—photon flux at the limits of Earth’s atmosphere (cm−2 /s nm) FP—flat plate g—mean sun anomaly (rad) h—Planck constant (6.62606896 × 10−34 J s) H—daily global irradiation on a horizontal surface (Wh/m2)

ABBREVIATIONS

457

Hbn—daily direct normal irradiation (Wh/m2) Hd—daily diffuse irradiation on a horizontal surface (Wh/m2) Hn0—daily extraterrestrial direct normal irradiation (Wh/m2) H0—daily extraterrestrial direct irradiation on a horizontal surface (Wh/m2) H—daily global irradiation on a horizontal surface (Wh/m2) IEC—International Electrotechnical Commission K—ratio Ed/E or Hd/H Kn—ratio Ebn/En0 or Hbn/Hn0 Kt—ratio E/(En0 cos Z) or H/H0 L—site latitude (degree) m—air mass Nc—circumsolar spectral radiance (W /m sr nm) NOAA—National Oceanic and Atmospheric Administration NREL—National Renewable Energy Laboratory NSRDB—National Solar Radiation Data Base PMOD—Physikalisch-Meteorologisches Observatorium Davos POA—plane of array PV—photovoltaic or solar cell R—actual sun–earth distance (m) R0—average sun–earth distance (1.496 × 1011 m) Rd—ratio of sky diffuse irradiance on a tilted receiver to that on the horizontal Rs—volumetric mean solar radius (6.96 × 108 m) s—receiver’s tilt angle from the horizontal S—sun–earth distance correction factor Sλ—solar cell’s spectral response SAM—(instrument for) sun and aureole measurement SC—solar constant SMARTS—Simple Model of the Atmospheric Radiative Transfer of Sunshine SORCE—Solar Radiation and Climate Experiment sr—time of sunrise SREF—solar radiation enhancement factor ss—time of sunset STC—standard test conditions SURFRAD—Surface Radiation Network t—time T—temperature (K) TMY—typical meteorological year TMY2—TMY version 2 TMY3—TMY version 3 TSI—total solar irradiance w—precipitable water (cm) WRDC—World Radiation Data Center Z—solar zenith angle α—aerosol wavelength exponent or “Ångström coefficient” β—aerosol optical depth at 1000 nm, or “Ångström turbidity coefficient”

458

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

γ—solar azimuth from North θ—angle of incidence of the sun’s rays on a surface or receiver λ—wavelength (nm) ξ—angular distance from the sun center (degree) ξ0—radial acceptance angle of a receiver (degree) ξs—angular distance from the sun’s center to its outer edge (degree) ρ—ground reflectance for shortwave radiation, or “albedo” σ—Stefan–Boltzmann constant (5.670 400 × 10−8 W/m2 K4) τaλ—aerosol optical depth at wavelength λ ψ—receiver azimuth from North (degree)

REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16]

C. A. Gueymard. Reference solar spectra: Their evolution, standardization and comparison to recent measurements. Adv. Space Res. 37, 323–340 (2006). D. Labs and H. Neckel. The solar constant. Sol. Phys. 19, 3–15 (1971). C. Fröhlich. Solar irradiance variability since 1978: Revision of the PMOD Composite during Solar Cycle 21. Space Sci. Rev. 125, 53–65 (2006). C. Fröhlich and J. Lean. The sun’s total irradiance: Cycles, trends and related climate change uncertainties since 1976. Geophys. Res. Lett. 25, 4377–4380 (1998). C. A. Gueymard. The sun’s total and spectral irradiance for solar energy applications and solar radiation models. Sol. Energy 76, 423–453 (2004). ASTM. Standard solar constant and zero air mass solar spectral irradiance tables. Standard E490-00, American Society for Testing and Materials, West Conshohocken, PA. Available at http://www.astm.org/ (2000). M. Iqbal. An Introduction to Solar Radiation. Toronto, Academic Press (1983). H. Neckel. Analytical reference functions F(λ) for the sun’s limb darkening and its absolute continuum intensities (300 to 1100 nm). Sol. Phys. 229, 13–33 (2005). P. J. Mohr, B. N. Taylor, and D. B. Newell. CODATA recommended values of the fundamental physical constants: 2006. Rev. Mod. Phys. 80, 633–730 (2008). J. J. Michalsky and P. W. Kiedron. Comparison of UV-RSS spectral measurements and TUV model runs for clear skies for the May 2003 ARM aerosol intensive observation period. Atmos. Chem. Phys. 8, 1813–1821 (2008). V. Badescu, ed. Modeling Solar Radiation at the Earth Surface. Berlin, Springer (2008). C. A. Gueymard. An anisotropic solar irradiance model for tilted surfaces and its comparison with selected engineering algorithms. Sol. Energy 38, 367–386 (1987); Erratum, Sol. Energy 40, 175 (1988). B. Robles-Ocampo, E. Ruiz-Vasquez, H. Canseco-Sanchez, et al. Photovoltaic/ thermal solar hybrid system with bifacial PV module and transparent plane collector. Sol. Energ. Mat. Sol. C. 91, 1966–1971 (2007). K. J. Weber, V. Everett, P. N. K. Deenapanray, et al. Modeling of static concentrator modules incorporating lambertian or v-groove rear reflectors. Sol. Energ. Mat. Sol. C. 90, 1741–1749 (2006). S. Hatwaambo, H. Hakansson, J. Nilsson, and B. Karlsson. Angular characterization of low concentrating PV– CPC using low-cost reflectors. Sol. Energ. Mat. Sol. C. 92, 1347–1351 (2008). C. A. Gueymard. Direct and indirect uncertainties in the prediction of tilted irradiance for solar engineering applications. Sol. Energy 83, 432–444 (2009).

REFERENCES [17] [18] [19]

[20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30]

[31] [32] [33] [34] [35] [36]

459

C. A. Gueymard. Direct solar transmittance and irradiance predictions with broadband models. Pt 2: Validation with high-quality measurements. Sol. Energy 74, 381–395 (2003); Corrigendum, Sol. Energy 376, 515 (2004). C. A. Gueymard and D. R. Myers. Validation and ranking methodologies for solar radiation models. In Modeling Solar Radiation at the Earth Surface, V. Badescu, ed., pp. 479–509. Berlin, Springer (2008). ASTM. Standard tables for reference solar spectral irradiances: Direct normal and hemispherical on 37° tilted surface. Standard G173-03, American Society for Testing and Materials, West Conshohocken, PA. Available at http://www.astm.org/ (2003). A. A. Kokhanovsky. Cloud Optics. Berlin, Springer (2006). J. A. Davies and D. C. McKay. Evaluation of selected models for estimating solar radiation on horizontal surfaces. Sol. Energy 43, 153–168 (1989). J. J. Michalsky, J. L. Berndt, and G. J. Schuster. A microprocessor-based rotating shadowband radiometer. Sol. Energy 36, 465–470 (1986). C. A. Gueymard and D. R. Myers. Evaluation of conventional and high-performance routine solar radiation measurements for improved solar resource, climatological trends, and radiative modeling. Sol. Energy 83, 171–185 (2009). S. Wilcox, M. Anderberg, W. Beckman, et al. National Solar Radiation Database 1991–2005 Update: User’s manual. Golden, CO, NREL, Rep. TP-581-41364. Available at http://www.nrel.gov/docs/fy07osti/41364.pdf (2007). C. A. Gueymard and D. R. Myers. Solar radiation measurement: Progress in radiometry for improved modeling. In Modeling Solar Radiation at the Earth Surface, V. Badescu, ed., pp. 1–27. Berlin, Springer (2008). E. L. Maxwell, W. F. Marion, D. R. Myers, et al. National Solar Radiation Data Base—Final Technical Report Vol. 2. Rep. TP-463-5784, National Renewable Energy Laboratory, Golden, CO (1995). E. L. Maxwell. METSTAT—The solar radiation model used in the production of the National Solar Radiation Data Base (NSRDB). Sol. Energy 62, 263–279 (1998). R. Perez, P. Ineichen, K. Moore, et al. A new operational model for satellite derived irradiances, description and validation. Sol. Energy 73, 307–317 (2002). W. Marion and S. Wilcox. Solar radiation data manual for flat-plate and concentrating collectors. Rep. TP-463-5607, NREL, Golden, CO (1994). A. Tsvetkov, S. Wilcox, D. Renné, and M. Pulscak. International solar resource data at the World Radiation Data Center. In Proceedings of the Solar ’95 Conference, Minneapolis, MN, R. Campbell-Howe and B. Wilkins-Crowder, eds., pp. 216–219. Boulder, CO, American Solar Energy Society (1995). A. Ohmura, E. G. Dutton, B. Forgan, et al. Baseline Surface Radiation Network (BSRN/WCRP): New precision radiometry for climate research. B. Am. Meteorol. Soc. 79, 2115–2136 (1998). T. P. Ackerman and G. Stokes. The Atmospheric Radiation Measurement Program. Phys. Today 56, 38–45 (2003). J. A. Augustine, J. DeLuisi, and C. Long. SURFRAD—A national surface radiation budget network for atmospheric research. Bull. Am. Meteorol. Soc. 81, 2341–2357 (2000). P. Ineichen. Comparison and validation of three global-to-beam irradiance models against ground measurements. Sol. Energy 82, 501–512 (2008). R. Perez, R. Seals, P. Ineichen, et al. A new simplified version of the Perez diffuse irradiance model for tilted surfaces. Sol. Energy 39, 221–231 (1987). D. F. Menicucci and J. P. Fernandez. User’s manual for PVFORM: A photovoltaic system simulation for stand-alone and grid-interactive applications. Rep. SAND850376, Sandia National Laboratories, Albuquerque, NM (1988).

460

SOLAR RESOURCE FOR SPACE AND TERRESTRIAL APPLICATIONS

[37]

D. R. Myers. General cloud cover modifier for clear sky solar radiation models. In Proceedings of the Optical Modeling and Measurements for Solar Energy Systems Conference, Vol. 6652, D. R. Myers, ed., pp. 6652–08. Bellingham, WA, SPIE, International Society for Optical Engineering (2007). D. R. Myers. Relative performance of multiple solar radiation resource assessment data sources. In Proceedings of the Solar 2007 Conference, R. Campbell-Howe, ed., paper 031A. Boulder, CO, American Solar Energy Society (2007). M. Suri, J. Remund, T. Cebecauer, et al. First steps in the cross-comparison of solar resource spatial products in Europe. In Proceedings of the Eurosun 2008 Conference Freiburg, Germany, ISES (2008). C. A. Gueymard. Fixed or tracking solar collectors? Helping the decision process with the solar resource enhancement factor. In Optical Modeling and Measurements for Solar Energy Systems II Conference, Vol. 7046, B. K. Tsai,, ed., doi: 10.1117/12.795506. Bellingham, WA, SPIE, International Society for Optical Engineering (2008). C. B. Christensen and G. M. Barker. Effects of tilt and azimuth on annual incident solar radiation for United States locations. In Proceedings of the Solar Forum 2001 Conference, S. J. Kleis and C. E. Bingham, eds., pp. 225–232. New York, NY, American Society of Mechanical Engineers (ASME) (2001). C. A. Gueymard and S. M. Wilcox. Spatial and temporal variability in the solar resource: Assessing the value of short-term measurements at potential solar power plant sites. In Proceedings of the Solar 2009 Conference, R. Campbell-Howe, ed., paper 118. Boulder, CO, American Solar Energy Society (2009). W. E. McMahon, K. Emery, D. J. Friedman, et al. Outdoor testing of GaInP2/GaAs tandem cells with top cell thickness varied. Proceedings of the International Conference on Solar Concentrators for the Generation of Electricity or Hydrogen, Golden, CO, National Renewable Energy Laboratory, Rep. No. CP-520-37819, http://www.nrel.gov/docs/fy05osti/38172.pdf (2005). C. A. Gueymard and H. D. Kambezidis. Solar spectral radiation. In Solar Radiation and Daylight Models, T. Muneer, ed., pp. 221–301, Amsterdam, Elsevier (2004). C. A. Gueymard. Advanced solar irradiance model and procedure for spectral solar heat gain calculation (RP-1143). ASHRAE Trans. 113(1), 149–164 (2007). C. A. Gueymard. Prediction and validation of cloudless shortwave solar spectra incident on horizontal, tilted, or tracking surfaces. Sol. Energy 82, 260–271 (2008). C. A. Gueymard, D. Myers, and K. Emery. Proposed reference irradiance spectra for solar energy systems testing. Sol. Energy 73, 443–467 (2002). K. Emery, D. Myers, and S. Kurtz. What is the appropriate reference spectrum for characterizing concentrator cells? In Proceedings of the 29th Photovoltaic Specialists Conference, pp. 840–843. New Orleans, LA, IEEE (2002). ASTM. Table for reference solar spectral distributions: Direct and diffuse on 20° tilted and vertical surfaces. Standard G197-08, American Society for Testing and Materials, West Conshohocken, PA. Available at http://www.astm.org/ (2008). C. A. Gueymard. Daily spectral effects on concentrating PV solar cells as affected by realistic aerosol optical depth and other atmospheric conditions. Proceedings of the Optical Modeling and Measurements for Solar Energy Systems II Conference, Vol. 7410, B. K. Tsai, ed., paper 7410–07. San Diego, CA, SPIE, International Society for Optical Engineering (2009). C. N. Vittitoe and F. Biggs. Six-Gaussian representation of the angular brightness distribution for solar radiation. Sol. Energy 27, 469–490 (1981). J. E. Noring, D. F. Grether, and A. J. Hunt. Circumsolar radiation data: The Lawrence Berkeley Laboratory reduced data base. Rep. TP-262-4429, NREL, Golden, CO. Available at http://rredc.nrel.gov/solar/pubs/circumsolar/ (1991).

[38] [39] [40]

[41]

[42]

[43]

[44] [45] [46] [47] [48] [49] [50]

[51] [52]

REFERENCES [53] [54] [55] [56] [57] [58] [59] [60] [61]

461

M. Schubnell. Influence of circumsolar radiation on aperture, operating temperature and efficiency of a solar cavity receiver. Sol. Energ. Mat. Sol. C. 27, 233–242 (1992). A. Neumann, A. Witzke, S. A. Jones, and G. Schmitt. Representative terrestrial solar brightness profiles. J. Sol. Energ.-T. ASME 124, 198–204 (2002). D. Buie and A. G. Monger. The effect of circumsolar radiation on a solar concentrating system. Sol. Energy 76, 181–185 (2004). J. G. DeVore, A. T. Stair, A. LePage, et al. Retrieving properties of thin cloud from solar aureole measurements. J. Atmos. Ocean. Technol. 26, 2531–2548 (2009). C. A. Gueymard. Parameterized transmittance model for direct beam and circumsolar spectral irradiance. Sol. Energy 71, 325–346 (2001). A. Rabl. Comparison of solar concentrators. Sol. Energy 18, 93–111 (1976). M. A. Rosen, F. C. Hooper, and A. P. Brunger. The characterization and modelling of the diffuse radiance distribution under partly cloudy skies. Sol. Energy 43, 281–290 (1989). A. P. Brunger and F. C. Hooper. Anisotropic sky radiance model based on narrow field of view measurements of shortwave radiance. Sol. Energy 51, 53–64 (1993); Erratum, Sol. Energy 51, 523 (1993). A. W. Harrison and C. A. Coombes. Angular distribution of clear sky short wavelength radiance. Sol. Energy 40, 57–63 (1988).

20 SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL PAUL GILMAN,1 NATHAN BLAIR,1 AND CHRISTOPHER CAMERON2 1 National Renewable Energy Laboratory, 2Sandia National Laboratories

20.1

INTRODUCTION

No discussion of solar cells is complete without an examination of their role as part of a complete PV system in the context of an energy project. From a technical perspective, solar cells are just one component in an electricity-generating system that consists of PV modules, inverters, and BOS components. From economic, financial, and policy perspectives, solar cells are just one element in a valuegenerating system whose worth depends on the quantity of energy the system produces, the system lifetime, installation costs, operating and maintenance costs, a range of financial options, and incentives from a variety of institutions and organizations. Computer models can help analysts understand how solar cells perform in a system and how they contribute to the cost of energy produced by the system. In this chapter, we describe and present sample analyses for one computer model designed specifically for solar energy projects: the U.S. Department of Energy’s SAM, developed by NREL, Sandia National Laboratories, and other partners. SAM calculates life cycle economic metrics such as the LCOE by integrating the results from hour-by-hour performance models with those from a cash flow cost model. SAM facilitates system-based analysis, which is an alternative to conventional approaches to valuing and comparing PV technologies. While conventional approaches focus on metrics that emphasize solar cells, such as cell efficiencies and manufacturing costs, a system-based approach focuses on energy production metrics, such as the LCOE. System-based analyses account for parameters determined by the economic context of projects in the residential, commercial, and utility electricity markets. Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

463

464

SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL

The SAM analysis examples described later in this chapter illustrate how system-based analysis can help ensure that important considerations in technology selection are included in decision making for research management, policy making, system design, and other efforts to develop the full potential of PV technologies. The system-based approach offers the following advantages:

•

•

•

Find the Low-Hanging Fruit. The location of a PV project plays an important role in both its technical feasibility and its economic attractiveness to investors. Using a systems approach on a large regional scale can help analysts discover unexpected opportunities for PV project development. For example, analysts have been quick to identify the development potential of obvious locations in sunny regions of the United States. However, they have only relatively recently shown that for some less sunny locations, factors such as the coincidence of the solar resource with demand peaks and transmission constraints limiting the availability of power from central generating plants offer opportunities for largescale implementation of distributed PV projects. A conventional analysis might focus only on annual electricity production per unit of available solar energy, and fail to capture the value of the solar resource’s seasonal or daily coincidence with high electricity prices. Example 3 described below is one example of how SAM can be used for location-specific analysis, in this case comparing system output and LCOE for identical systems operating in a hot and cold climate. Take Off the Blinders. Sometimes, analysts fall into the trap of focusing on a particular area of interest and miss the big picture. For example, by focusing exclusively on reducing PV module or inverter installation costs, an analyst may miss the effect of lower-cost component performance on the system life cycle cost. For example, a lower-cost module might degrade more rapidly than a more expensive one, or a lower-cost inverter may require more frequent repair and replacement, ultimately making the electricity generated by a system using those components more expensive in spite of the lower capital costs. Examples 4 and 5 described later in this chapter show how SAM can help with analyses of system and inverter lifetimes. Make Good Business Decisions. Looking at how components operate as part of a system and how they contribute to the system’s life cycle cost can help managers to choose the most promising direction to take research and development programs or to choose products for specific projects. Although choosing PV system equipment to minimize installation costs may seem like an obvious good business decision, in reality, other factors may come into play that cause unexpected choices to make the best business sense. Example 7 shows how the type of financing of a utility-scale PV project is just as important a consideration as the cost and efficiency of the PV modules.

LCOE

465

•

Compare Options on a Level Playing Field. Comparing projects based on different energy technologies, particularly capital-intensive ones like PV systems and operating cost-intensive ones like fuel-based systems, requires metrics that accurately express the value of the different options. Obviously, basing a technology choice between a utility-scale PV project and a natural gas plant on the installation cost alone would make the gas plant seem like the most cost-effective choice but would not capture the true cost of electricity generated by the two systems. Similarly, evaluating distributed generation projects requires a metric that is comparable to the cost per kilowatt hour of retail electricity against which the projects compete. All of the analyses in this chapter use one suitable metric for such comparisons, the LCOE, which is described next.

20.2

LCOE

The LCOE represents a project’s TLCC per unit of energy. The TLCC is the present value of the sum of the installation and project cash flows Cn over project life N, assuming a discount rate of d: TLCC =

N

Cn

n=0

(1 + d )n

∑

.

By definition, multiplying the total annual energy production Qn by the LCOE for each year in project life N and discounting the resulting stream of values to year 1 results in a value equal to the TLCC: N

Qn × LCOE

n =1

(1 + d )n

∑

= TLCC.

The LCOE is therefore mathematically equivalent to the TLCC divided by the “discounted” annual energy production of the system: LCOE =

TLCC . Qn ∑ n n =1 (1 + d ) N

The LCOE is typically expressed in cents per kilowatt hours to make it comparable to cost of energy values for projects based on different types of technologies. To calculate the LCOE of a PV system, SAM first uses an hourly simulation engine to calculate the system’s annual electric output based on a set of parameters describing the physical system, then calculates a series of annual cash flows based on a set of cost and financial assumptions, and finally calculates the LCOE using

466

SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL

the appropriate discount rates. The annual cash flows include the income and costs shown in Table 20.1. The LCOE is useful for comparing alternative options for projects based on different technologies or operating in different markets. For example, the LCOE would be useful to compare the true cost of a residential rooftop PV system with residential utility rates, or the cost of a utility-scale PV system with electricity market rates. The LCOE is also an appropriate metric for comparing different kinds of energy projects. Although this chapter focuses on the LCOE, SAM also generates other useful metrics shown in Table 20.2. The metrics depend on the type of financing.

TABLE 20.1. Annual Cash Flow Income and Cost Categories in SAM Income

Electricity sales Incentive payments Tax deductions, including for depreciation Debt

Costs

Installation Operation and maintenance Equipment replacement Debt payments and interest

TABLE 20.2. Economic Metrics Available in SAM for Each Type of Financing Type of Financing

Available Metrics

Residential cash Residential loan or mortgage Commercial cash Commercial—standard loan

LCOE Net present value (NPV) Simple payback kWh/kW—year 1 Capacity factor System performance factor Annual output—year 1

Utility and IPP Commercial third-party ownership

LCOE Internal rate of return Minimum DSCR First-year PPA PPA escalation rate Debt fraction kWh/kW—year Capacity factor System performance factor Annual output—year 1

THE SAM SOFTWARE

467

Although some metrics included in the list such as internal rate of return and simple payback do not reflect the true value of a project, they are included as available options because many analysts are familiar with them.

20.3

THE SAM SOFTWARE

SAM models a range of solar energy technologies for electricity generation, including PV arrays, solar thermal troughs, power towers, and dish–Stirling systems. SAM also includes a simple thermal model for comparisons between solar energy systems and fossil fuel-based thermal power plants. For PV systems, SAM first calculates the system’s total electricity production in kilowatt hours for the first year based on weather data for a particular location and physical specifications of the array and inverter. SAM calculates the total production for subsequent years based on an annual degradation factor, and annual cash flows based on financial and economic inputs to determine the LCOE and other economic metrics.

•

•

•

•

Weather Data. SAM’s simulation engine uses data from hourly weather files to model the solar resource at a given location. The weather file supplies radiation data for energy calculations, and wind speed, temperature, snow cover, and other data to model the effects of temperature and ground reflectance on system performance. Financial and Economic Inputs. SAM’s cash flow calculations depend on a set of input variables describing the project’s finances, such as the analysis period equivalent to the system lifetime, discount rate, inflation rate, loan amount, and loan rate. SAM provides different financing options that use different sets of input variables. Table 20.3 shows the financing options available to the different project types. Incentives. SAM models two types of incentives: tax credits and incentive payments. Tax credits can be provided by a state or federal government. Incentive payments can be provided by a state or federal government, utility, or other entity. An investment-based incentive or tax credit is a one-time payment to the project made in year 1 of the project cash flow that is either a fixed amount, a percentage of total installed costs, or a function of system size. A production-based incentive or tax credit is an annual payment based on the amount a of energy produced by the system in each year. Incentive payments may or may not be taxable by either the federal or state government. System Performance. SAM simulates the hourly performance of one or more PV modules in an array, calculating an hourly DC output value that it passes to the inverter performance model, which in turn calculates the system’s AC output (Fig. 20.1). Table 20.4 describes the three module performance models available in SAM, and Table 20.5 describes the two inverter models available in SAM.

468

SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL

TABLE 20.3. Financing Options for Different Project Types Market

Financing Options

Residential

Commercial

Central generation

•

Description

Cash

The owner pays cash in the amount of the total installed cost in year 0 of the project cash flow.

Loan or mortgage

The owner pays cash for the equity portion of the total installed cost in year 0 of the cash flow and makes a principal and interest payment in subsequent years.

Cash

The owner pays cash in the amount of the total installed cost in year 0 of the project cash flow.

Loan

The owner pays cash for the equity portion of the total installed cost in year 0 of the cash flow and makes an interest and principal payment in subsequent years.

Third-party ownership

The project earns revenues through electricity sales to cover project costs. The owner pays cash for the equity portion of the total installed cost in year 0 of the cash flow and makes an interest and principal payment in subsequent years.

Utility and independent power producer (IPP)

The project earns revenues through electricity sales to cover project costs. The owner pays cash for the equity portion of the total installed cost in year 0 of the cash flow and makes an interest and principal payment in subsequent years.

Performance Adjustments. SAM uses a set of derating factors to convert the predicted output of a single module into an array output value. A set of DC derating factors accounts for module mismatch, diodes and connections, DC wiring, shading, soiling, and, if applicable, tracking losses. A second set of derating factors accounts for post-inverter losses from AC wiring and transformers. An availability factor accounts for system outages for maintenance. To calculate the system output in years after the first year, SAM applies an annual degradation factor to account for the annual reduction in system output due to equipment aging (Fig. 20.1).

THE SAM SOFTWARE

469

TABLE 20.4. Module Performance Model Descriptions Sandia PV array performance model

An empirically based model that consists of a set of curve-fit equations for each module in a database derived from outdoor module testing conducted on a two-axis tracker

California Energy Commission (CEC) performance model

A five-parameter model developed by the University of Wisconsin’s Solar Energy Laboratory, which uses a database of standard rating condition data from the manufacturer’s module specifications or independent test laboratories and a set of semiempirical correlation equations to predict the module I–V curve. SAM assumes that modules operate at their maximum power point.

Simple efficiency model for flat-plate modules

A simple model that calculates module output by multiplying the total global solar radiation incident on the array by the module’s efficiency value and area and then by adjusting the performance through the use of a temperature coefficient, assuming a normal cell operating temperature of 25°C. A different efficiency value can be entered for up to five incident radiation values.

Simple efficiency model for concentrating modules

A simple model that calculates module output by multiplying the direct normal component of the solar radiation by the module’s efficiency value and area. A different efficiency value can be entered for up to five incident radiation values

PVWatts Solar Array

An implementation of NREL’s web-based PVWatts model that represents the complete system using a single DC-toAC derating factor. As of October 2009, this model includes preliminary battery storage and load models that will be improved and will be eventually available with the other module model options.

TABLE 20.5. Inverter Performance Model Descriptions Sandia performance model for gridconnected inverters

The model uses a set of parameters for commercially available inverters in a set of equations to calculate the AC power output based on DC power input and voltage values generated by a PV module model. The parameter sets are maintained by Sandia National Laboratories and are primarily based on a list of laboratory performance measurements maintained by the California Energy Commission.

Single-point efficiency model

A simple model that represent the inverter performance with two parameters: the AC nominal capacity rating and an efficiency value. This simple model multiplies the PV array’s DC output by the efficiency value to calculate the AC output.

470

SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL 1

4

0.9

3.5

0.8 3

2

0.5

kW

0.6

0.4

1.5

kW/m2

0.7

2.5

0.3 1

0.2

0.5 0 Total Incident Radiation (kW/m2)

0.1 0 DC Output (kW)

AC Output (kW)

Figure 20.1. Graph of hourly performance data showing the total incident solar radiation, module DC output, and system AC output for 2 days in August. TABLE 20.6. Cost Categories and Units Cost Category

Units

Module capital cost

$/WDC or $/unit

Inverter capital cost

$/WAC or $/unit

Balance-of-system capital cost

$/system

Installation labor

$/system

Miscellaneous

$/system

Engineer, procure, construct

$/system

Project, land, miscellaneous

$/system

Fixed operation and maintenance

$/year

Fixed operation and maintenance

$/kWDC-year

Variable operation and maintenance

$/MWhAC-year

•

Costs. SAM calculates the cash flow and resulting economic metrics based on two categories of costs. Capital costs account for the cost of installing modules, inverters, and BOS components. Operation and maintenance costs account for recurring costs for maintenance, repair, and replacement of equipment. For residential and commercial projects, SAM allows operation and maintenance costs to be assigned to specific years, which makes it possible to analyze projects with periodic inverter replacements or other periodic costs. Table 20.6 describes the cost inputs for PV systems.

ANALYSIS EXAMPLES

20.4

471

ANALYSIS EXAMPLES

The following examples illustrate SAM’s use in analyses based on the system approach described above. The analyses start with assumptions similar to those of the analysis described in Chapter 2, which is for a 4-kW grid-connected residential system with net metering, and using module and inverter efficiency and system cost assumptions reasonable for a representative system in the United States. Example 1 replicates the Chapter 2 analysis, and Examples 2–7 show additional analyses illustrating some of the insights that a computer model like SAM can add.

Example 1: Module and Inverter Conversion Efficiency Improvements and Cost Reductions over Time This analysis replicates the “simple approach” analysis described in Chapter 2. SAM’s cash flow method, described earlier in this chapter, replaces the fixedcharge rate method of the simple approach. In place of the solar energy density and efficiency factors used in the simple approach to predict the system’s total annual output, SAM uses the hour-by-hour performance simulation described above to calculate the system’s hourly output. The near-term and medium-term systems use assumptions reasonable for 5 and 15 years in the future with the expectation of the following improvements over time, assuming that the area remains constant:

• • • •

increases in module and inverter efficiency, increases in inverter lifetime, reduction in capital costs, and reduction in government incentives.

Table 20.7 summarizes the performance parameters that describe the physical system. The analysis uses solar radiation and position, ambient temperature, wind speed data from the TMY2 weather file for Phoenix, Arizona, and assumes a fixed array tilted at 18°. The analysis also assumes that the inverter capacity increases with the array capacity in the near and medium terms to match the array capacity, which increases with the array efficiency as the area remains constant. Table 20.8 summarizes the cost, financial, and incentive assumptions of the analysis that SAM uses to calculate the cash flow and economic metrics. In addition to the values in the table, the analysis assumes a 5.5% discount rate (to discount future values to present values) and a loan fraction of 100% for all three cases. The following adjustments were required to account for differences in measurement units used by the two methods:

•

The simple approach includes a BOS cost expressed in dollars per square meter that includes installation costs, while in SAM, the BOS cost is

472

SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL

TABLE 20.7. Summary of Performance Parameter Inputs for SAM Analysis Parameter

Baseline

PV array

—

Module efficiency Array capacity Array area Derating factor System lifetime Inverter

Near Term

Medium Term

Units

—

—

—

13.5

16.0

20.0

%

4.0

4.7

5.9

29.63

29.63

29.63

m2

0.95

0.95

0.95

—

Same as analysis period

kWDC-peak (STC)

Years

—

—

—

—

Efficiency

0.95

0.96

0.97

—

Total capacity

4.0

4.7

5.9

kWAC

Lifetime

5

•

•

•

10

20

Years

expressed in dollars, and the installation and BOS costs are in separate categories. The BOS cost in Table 20.8 is the product of the area-related BOS (Cb) cost in Table 2.2 and the array area in Table 20.7. The simple approach includes an indirect cost rate of 22.5%, of which 0.5% accounts for both operation and maintenance and inverter replacements. In SAM, the indirect cost is a fixed amount and is a component of the total capital cost. The indirect cost in Table 20.8 is 22% of the total direct capital cost. SAM calculates an annual operation and maintenance cost as a function of up to three operation and maintenance cost components: a fixed annual cost, a fixed annual cost per DC kilowatt of installed capacity, and a variable annual cost per AC megawatt hour generated per year. This analysis assumes an annual operation and maintenance calculated as 0.5% of the total installed cost for the first year, and escalated at the inflation rate of 2.5%. SAM calculates the inverter replacement cost based on the inverter capital cost, inflation rate, and replacement interval in years. SAM only applies the inverter replacement costs at the replacement intervals.

Tables 20.9 and 20.10 show a summary of the results, including LCOE values comparable to those from the simple method used in Chapter 2. The results show a trend in LCOE reductions as module and inverter efficiencies improve and as capital costs drop over time, even as government incentives decline. The magnitude of the trend shown in Table 20.10 is similar for both the SAM and simple approach methods. Note that the system output in the first year for each case increases

ANALYSIS EXAMPLES

473

TABLE 20.8. Summary of Cost, Financial, and Incentive Inputs for SAM Analysis Baseline

Near Term

Medium Term

PV module

4.00

2.20

1.25

$/WDC

Inverter (= Ci × inverter capacity)

3600

3280

1800

$/unit

Balance-of-system (BOS) cost (= Cb × array area)

9096

4592

4445

$

Installation labor (included in BOS cost)

0

0

0

$

28,696

23,965

13,652

$

Indirect (miscellaneous) (= 0.22 × A)

6313

5272

3003

$

Total installed cost (C)

8.75

4.97

2.81

$/WDC

First-year inverter replacement cost

3600

3280

1800

$/year

2.5

2.5

2.5

%/year

Total direct costs (A)

Inflation rate Inverter replacement interval

Units

5

10

20

175.00

116.80

82.90

Analysis period

30

35

35

Years

Loan rate

6.0

6.0

6.0

%/year

Loan period

30

35

35

Years

Scheduled and unscheduled maintenance (= 0.005 × C × array capacity)

Years $/kWDC-year

Capacity-based incentive

2.50

1.21

0.55

$/WDC

Federal tax rate

28.0

28.0

28.0

%

Notes: The inverter cost is based on the DC–AC inverter cost (Ci) in Table 2.2 and on the inverter capacity in Table 20.7. The BOS cost is based on the area-related BOS cost (Cb) in Table 2.2 and on the array area in Table 20.7. The table shows first-year values for scheduled and unscheduled maintenance and is based on the array capacity in Table 20.7.

because the array area is held constant as the module efficiency increases, thereby increasing the array’s peak rated capacity. Figure 20.2 shows the results generated by SAM of an analysis similar to the sensitivity analysis described in Chapter 2. The analysis varies the module efficiency for each of the three cases while holding the other assumptions constant. The inverter efficiency is held constant over the range of module

474

SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL

TABLE 20.9. Summary of Results for SAM Analysis Baseline Levelized cost of energy

34.7

First-year output

Near Term

Medium Term

15.9

7550

9040

Simple approach results

—

—

Levelized cost of energy

32.0

15.0

Units

10.1

Cents/kWh

11,300

kWh/year

—

— 9.0

Cents/kWh

TABLE 20.10. Reduction of Levelized Cost of Energy from the Baseline Baseline

Near Term

Medium Term

Units

SAM

0

54

71

%

Simple approach

0

53

72

%

Levelizec Cost of Energy (¢/kWh)

80 70 60 Base Line Near Term Medium Term

50 40 30 20 10 0

0

5

10

15 20 25 Module Efficiency (%)

30

35

Figure 20.2. Levelized cost of energy as a function of module efficiency.

efficiencies, but the number of inverters is adjusted to ensure that the inverter capacity is sufficient to handle the array’s output for each point in the sensitivity analysis. Again, the results show similar trends to the results based on the simple approach in Chapter 2. The advantage of using a computer model to perform this kind of analysis over the simple approach is that more parameters can be adjusted to explore the impact of changes in design, costs, incentives, and financial assumptions. The following examples explore some of the possibilities based on the baseline analysis presented above.

ANALYSIS EXAMPLES

475

Example 2: Module Degradation In addition to system lifetime, SAM can model the change in performance of PV modules over time. PV module output typically decreases over time as module components age. Analysts often use an annual degradation rate between 0.5% and 1.0%. Module manufacturer lifetime warranties typically guarantee that the module often will generate no less than 80% of the nameplate output at the end of the warranty period, which is usually in the 25- to 30-year range. Figure 20.3 shows the effect in SAM of changing the system degradation for the baseline case described above while holding all other assumptions constant. SAM compounds the degradation rate annually: there is no loss of output in year 1; in year 2, the system produces 99% of the year 1 output; in year 3, it produces 99% of the year 2 output, and so on. The analysis shows that each 1% change in annual degradation rate increases the system’s LCOE by about 10% (Fig. 20.3).

Example 3: Temperature Correction

Multiple of Levelized Cost of Energy (1 = No Degradation)

Module performance is affected by the cell temperature, which SAM calculates as a function of the ambient temperature, ambient wind speed, and the solar radiation incident on the array. SAM models this effect using the module’s temperature coefficient and information from the weather file. In a relatively hot climate such as Phoenix, the effect of a −0.5%/°C maximum power temperature coefficient, which is typical for multicrystalline silicon modules, on a rack-mounted module is to reduce its total annual energy output by about 10% from that of a theoretical module that does not respond to cell temperature (represented in SAM by a temperature coefficient of zero) (Fig. 20.4).

1.6 1.5 1.4 1.3 1.2 1.1 1 0

0.5

1

1.5 2 2.5 3 3.5 System Degradation (%/yr)

4

4.5

5

Figure 20.3. Effect of system degradation on levelized cost of energy.

SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL Fraction of Annual Output (1 = No Temperature Correction)

476 1 0.95 0.9 0.85 0.8

0.75 –1

–0.8 –0.6 –0.4 –0.2 Temperature Coefficient (%/°C)

0

Multiple of Levelized Cost of Energy (1 = No Temperature Correction)

Figure 20.4. Effect of module temperature coefficient on annual output for the relatively hot climate of Phoenix.

1.3 1.25 1.2 1.15 1.1 1.05 1 –1

–0.8 –0.6 –0.4 –0.2 Temperature Coefficient (%/°C)

0

Figure 20.5. Effect of module temperature coefficient on levelized cost of energy.

The temperature-related reduction in output results in a corresponding reduction in the LCOE, as shown in Figure 20.5. We used the results of SAM’s hourly simulation to further explore the relationship between module temperature and output. Figures 20.6 and 20.7 show hourly conditions for a hot day in Phoenix and for a cold day in Casper when the incident solar radiation peaks at 1 kW/m2 for both locations. At 1 p.m. on August 17 in Phoenix, the cell temperature peaks at about 62°C, while the incident solar radiation is 1 kW/m2 and the ambient temperature is about 40°C. In Casper at 1 p.m. on March 18, the total incident radiation is also 1 kW/m2, but the ambient temperature is about −5°C and the cell temperature is 13°C, well below the module’s

477

0.5

25 5 –15

°C

°C

45

kW/m2

1

65

0

65

4

45

3

25

2

5

1

–15

kW

ANALYSIS EXAMPLES

0

Noon Total Incident Radiation (kW/m2)

Noon Array Output (kW-DC)

Ambient Temperature (°C)

Cell Temperature (°C)

5 –15

0 Noon

°C

0.5

25

kW/m2

°C

45

4 3 2 1 0

65 45

1

65

25 5 –15

kW

Figure 20.6. Effect of ambient temperature on system output for a day in August of a hot climate (Phoenix, Arizona).

Noon

Total Incident Radiation (kW/m2)

Array Output (kW-DC)

Ambient Temperature (°C)

Cell Temperature (°C)

Figure 20.7. Effect of ambient temperature on system output for a day in March of a cold climate (Casper, Wyoming).

normal operating cell temperature of 25°C. The resulting output of the 4-kW array is 3.2 kW for the system in Phoenix and 4.0 kW for the system in Casper. Note that these output values are for the array of 40 modules and account for array losses. Example 4: System Lifetime PV systems are typically assumed to have a lifetime of between 25 and 30 years, often based on the expected lifetime of the modules. Module manufacturers typically provide a 25- or 30-year warranty for crystalline silicon modules. The lifetimes of modules using newer technologies are uncertain. In this analysis, we explored the relationship between the LCOE for the baseline system in Phoenix with the system lifetime. Figure 20.8 shows the LCOE for the baseline case described above for a range of system lifetimes. In order to isolate the effect of the system lifetime, we held all other assumptions, including the module efficiency and cost, constant over the range of system lifetimes.

478

SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL

Levelized Cost of Energy (¢/kWh)

44 43 42 41 40 39 38 37 36 35 34 5

10

15 20 System Lifetime (years)

25

30

Figure 20.8. Effect of system lifetime on levelized cost of energy.

Example 5: Inverter Lifetime

Levelized Cost of Energy (¢/kWh)

Inverter lifetimes are typically lower than module (and system) lifetimes, and inverters are replaced or rebuilt several times during the life of a system. In this analysis, we investigated how the inverter contributes to the LCOE of the 4-kW residential baseline system for different inverter replacement intervals over a range of inverter installation costs. Figure 20.9 shows that the sensitivity of the inverter’s fraction of the system LCOE to inverter installation cost increases as the inverter lifetime decreases. It also shows that the improvement in levelized cost energy

16 14 12 10 8 6 4 2 0 0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Inverter Cost ($/Wac) 5 years

10 years

15 years

20 years

25 years

Figure 20.9. Inverter contribution to system levelized cost of energy for different inverter lifetimes.

Multiple of Levelized Cost of Energy (1 = No O&M Cost)

ANALYSIS EXAMPLES

479

2 1.9 1.8 1.7 1.6 1.5 1.4 1.3 1.2 1.1 1 0

0.5 1 1.5 2 2.5 3 3.5 4 4.5 O&M Cost as Percent of Total Installed Cost (%)

5

Figure 20.10. Effect of operation and maintenance costs on levelized cost of energy.

decreases as the inverter lifetime approaches the system lifetime of 30 years. This type of graph also makes it possible to evaluate different inverter options when the price depends on the inverter lifetime. For example, in this analysis, the graph shows that even with an installation cost as high as $1 per watt, an expensive inverter with a 15-year lifetime is more cost-effective than a 5-year inverter with an installation cost as low as 40 ¢ per watt. Example 6: Operation and Maintenance Expenses When evaluating technology options for energy projects, investors sometimes consider only the project’s total capital cost. In this analysis, we explore the sensitivity of the LCOE to project operation and maintenance costs, which would not be accounted for an evaluation based only on capital cost. Figure 20.10 shows that for the baseline residential 4-kW system in Phoenix, each 1% increase in operating and maintenance cost increases the LCOE by just less than 10%. Example 7: Debt Fraction and Debt Rate The examples so far are based on the 4-kW residential baseline system analyzed above and in Chapter 2. The following example analyzes a 10-MW utility-scale system. The analyses for the residential system assume that the project sells electricity through a net metering agreement with an electric service provider and is financed by a loan. The electricity sales price is an input to the model and can be either a fixed rate with annual escalation or based on a time-of-use schedule. For the residential system, the LCOE represents the cost of installing and operating the system over its lifetime and includes annual loan principal and interest payments, tax payments, and benefits from incentives and electricity sales.

480

SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL

Required Revenue (¢/kWh)

Projects with utility financing, on the other hand, sell electricity through a PPA with an electricity off-taker at a fixed electricity sales price and optional annual escalation rate. For these projects, the electricity sales price is a model output that SAM calculates based on user-defined targets for the internal rate of return and minimum DSCR and on an optional constraint requiring the cash flow in all years to be positive. For utility projects, the LCOE reported by SAM is actually the revenue per unit of electricity generated that is required to recover costs and meet the targets, to cover debt- and tax-related costs, and to account for any benefits from tax credits and incentive payments. This required revenue can be considered to be the minimum power purchase price for the project to be financially feasible. Its value is very sensitive to the financial targets and assumptions, which can drive it much higher than it would be with simple loan financing. Figures 20.11 and 20.12 show how sensitive the required revenue per unit of electricity generated (reported in SAM as the LCOE) can be to both the debt fraction and loan rate, in this case with an internal rate of return target of 15%. Figure 20.11 shows the relationship when the cash flow is required to be positive for all years in the 30-year analysis period. For each loan interest rate, at a debt fraction of zero, the high capital expenditure causes the revenue requirement to be moderately high. As the debt fraction increases, the required revenue decreases because debt costs are spread out into future years. At a certain point, a debt fraction of 40% for the 2% loan interest rate case, the required revenue begins to increase again to cover higher interest and principal payments. The positive cash flow requirement forces the revenue requirement up as the loan fraction increases to cover the debt repayment expenses.

34 32 30 28 26 24 22 20 18 16 20

10 8 4 25

30

35

40 45 50 55 Debt Fraction (%)

60

65

70

6 Interest (%)

2

Figure 20.11. Required revenue versus debt fraction and loan interest rate for a project with utility financing and positive cash flow constraint.

Revenue Requirement (¢/kWh)

SUMMARY

481

22 20 18 16 14 12 10 8 6 4 2

10 8 4 20

25

30

35

40 45 50 55 Debt Fraction (%)

60

65

70

6 Interest (%)

2

Figure 20.12. Required revenue versus debt fraction and loan interest rate for a project with utility financing and no positive cash flow constraint.

Figure 20.12 shows the relationship without the positive cash flow requirement. In this case, the revenue requirement decreases as the loan fraction increases: at higher debt fraction values, SAM allows debt costs to contribute to negative project cash flows in out-years. This type of analysis can be useful in the pre-feasibility stages of a utility project to get a sense of a project’s financial requirements. The sensitivity analysis can be repeated for different system configurations and installation cost assumptions.

20.5

SUMMARY

The examples in this chapter illustrate SAM’s use for analyzing PV projects from a systems perspective, using the LCOE as a metric for evaluation. The first example shows results for an analysis replicating the study described in Chapter 2 exploring the impact of module efficiency on system LCOE. Those results showed that for module efficiencies above about 20%, efficiency improvements only result in minimal reductions in LCOE. Other examples show how array output decreases with cell temperature so an identical system installed in two different locations might have a different LCOE (Example 3) and how using modules and inverters with short lifetimes increases the LCOE (Examples 4 and 5). Examples 6 and 7 show how sensitive the LCOE can be to nontechnical assumptions such as operation and maintenance costs, project type, and loan and debt fraction for projects with debt financing.

482

SOLAR ENERGY COSTS: THE SOLAR ADVISOR MODEL

While this chapter focused on PV systems, SAM can also model concentrating solar power systems, including parabolic trough, dish–Stirling, and power tower systems. SAM also includes a simple thermal model that can be used to compare solar systems to fossil fuel-based power plants. SAM provides a useful platform for comparing different technology options using a consistent modeling framework. ABBREVIATIONS A—total direct costs AC—alternating current BOS—balance of system C—total installed cost Cb—area-related BOS costs CEC—California Energy Commission Ci—inverter cost Cn—project cash flows d—discount rate DC—direct current DSCR—debt service coverage ratio IPP—independent power producer I–V—current–voltage characteristics or curve LCOE—levelized cost of energy N—project life NPV—net present value NREL—National Renewable Energy Laboratory PPA—power purchase agreement PV—photovoltaic or solar cell Qn—total annual energy production SAM—Solar Advisor Model STC—standard test condition TLCC—total life cycle cost TMY2—typical meteorological year data

21 CHALLENGES OF LARGE-SCALE SOLAR CELL ELECTRICITY PRODUCTION DAVID FAIMAN Ben-Gurion University of the Negev

21.1

INTRODUCTION

The basic problem with large-scale solar power production stems from the diluteness of solar energy compared to other more readily available sources (such as barrels of oil). This necessitates extremely large sunlight capture areas and their associated infrastructure costs. Coupled to this is the relative remoteness of the richest terrestrial solar capture regions—the world’s deserts—from today’s population centers. In the past, the space requirement problem had been relatively muted compared to the problem of developing solar technologies that could compete economically with fossil fuel. And even the latter problem has been confused by fluctuating fuel prices and debates over a variety of geological, environmental, and political issues. However, now that the advent of CPV appears to be opening an avenue to really cost-effective solar power [1], other practical issues related to its large-scale implementation come to the forefront. Chief among these issues are storage, transmission, and financing. PV-generated electricity is less dispatchable than ST electricity, as far as utilities are concerned, because there is no intermediate thermal energy stage. Thermal energy is relatively easy to store, enabling the electricity from an ST plant to be generated when needed. In the case of PV, and particularly CPV, it is the electricity that must be stored and, as will be discussed below, this creates problems of a qualitatively different kind.

Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

483

484

LARGE-SCALE SOLAR CELL ELECTRICITY PRODUCTION

Electrical transmission is also a problem over large distances. In conventional grid networks, generating stations are typically interconnected by AC transmission lines of 100–200 km in length. For such networks, average transmission losses are found to be in the neighborhood of 7%. However, transmission lines of thousands of kilometers in length will be needed for connecting future desert-based solar power plants to distant cities. Lastly, financing takes on a new dimension when the payback times on initial investments may be tens of years. Several of these issues have been addressed in a series of volumes published by the International Energy Agency [2–4], whose Task 8 Photovoltaic Specialists Working Group coined the term VLS-PV for solar PV plants on the gigawatt scale. However, those volumes were more concerned with the problem of how to get a VLS-PV program started rather than that of implementing it on a worldwide scale, which is the concern of the present chapter. In the following sections, we shall attempt to discuss, in a structured manner, the scale of the problem and the areas in which a certain amount of R & D is still desirable. Our principal conclusion will be that technology is already sufficiently advanced and cost-effective to enable a massive VLS-PV program to go ahead.

21.2

THE SCALE OF THE PROBLEM

There are two aspects to implementing what has been termed the “grand solar plan” [5]. One is to realize the scale of investment that will be needed to replace a fossilfueled world with one whose energy needs are largely met by solar power. The other is to pinpoint the holes in our present knowledge in order to be able to create a viable solar technology in all of its details. Figure 21.1, for example, indicates that present world electricity generation is in excess of 20,000 TWh/year [6]. If we take a nominal 10% efficiency as being typical of what desert regions can provide via the employment of PV technology [7], and 2000 kWh/m2 as being a typical value for the solar irradiance in a desert region, then present world electricity production is equivalent to about 100 B square meters of solar panels and three to four times as much land area [7] for their placement. Thus, in order to meet only the present annual rise in electricity production of 610 TWh/year, seen in Figure 21.1, it would be necessary to erect 3 B square meters of PV panels each year on about 10,000 km2 of land. There are also likely to be significant political complications to a worldwide grand solar plan. In the United States, a single political system would be involved in providing solar-generated electricity to the entire population, but elsewhere, the situation is more complicated. For example, countries far removed from the world’s deserts would need to come to agreement with one another regarding the location and cost of power lines and the ensured supply of power. In addition, Europe, for example, even if treated as a single political entity, would need to reach agreement with the various Saharan states on whose territories “their” future VLS-PV plants would stand.

SPACE AND TIME REQUIREMENTS

485

22000

Annual Electricity Generation [TWh]

21000

World Electricity Production 1998-2007 (Source: BP Statistical Review 2008)

20000

y = –1.2011e+6 + 608.19x R^2 = 0.982 3 % per annum rise (base 2008)

19000 18000 17000

1% reference error bars

16000 15000 14000 1996

1998

2000

2002

2004

2006

2008

2010

Year

Figure 21.1. World electricity production during the past 10 years [6].

21.3

SPACE AND TIME REQUIREMENTS

The 10,000 km2 of additional annual land requirement that conventional PV would need in order to level off the slope of the graph in Figure 21.1 could be significantly reduced via the employment of CPV technology. This is because, under concentrated sunlight, CPV cells operate at considerably higher efficiency than 1-sun cells. Specifically, for CPV cells with a nominal (STC) efficiency of 32%, only 6 km2 of land would be needed for the generation of each 1 TWh/year [1]. By comparison, conventional fixed flat-panel PV would require 17 km2 of land [7]. Furthermore, with CPV cells already having been demonstrated with efficiencies greater than 40%, the land requirement is reduced proportionately. However, if we remain with the initial assumption of 32% cell efficiency [1], if the 610 TWh/year rise in world electricity requirements will continue unchanged, then in order for VLS-PV, in its CPV variety, to halt this rise, it will be necessary to generate an additional 610 TWh of solar electricity (from 305 GWp of CPV generating capacity) each year. In the first years, this would require 3660 km2 of new land annually, but as the technology improves (i.e., as the efficiency of massproduced CPV cells rises above the nominal 32%), the annual land requirement might eventually be reduced by perhaps 50%. Since the Sahara desert alone is approximately 10,000,000 km2 in area, this presents well over 2000 years of breathing space.

486

LARGE-SCALE SOLAR CELL ELECTRICITY PRODUCTION

On the other hand, if we were to replace all of the world’s present 20,000 TWh of electricity production by CPV technology, it would require the immediate assignment of 120,000 km2 of land (cf. 360,000 km2 for conventional stationary flat-panel PV [7]), that is, slightly more than 1% of the Sahara. Therefore, other than the political need for the various countries that share the Sahara to reach agreement on such a project, there would appear to be no serious land limitations for replacing all of the world’s electrical generating capacity by CPV and for continuing the present electrical growth rate, in like manner, for a great many years. There is also one other time requirement that should not be neglected. It stems from the fact that no PV plant on a gigawatt scale has ever been built. Raviv has estimated [1] that, taking into consideration the requirement of a certain amount of additional R & D, it would take 4 years to construct a VLS-PV manufacturing facility capable of a throughput of 1 GWp/year. Thus, this is the time that would need to elapse, after a political decision had been taken to go ahead with such a program, before construction could commence on the first VLS-PV plant. The latter would then commence operation a year later.

21.4

COST REQUIREMENTS AND FINANCING

21.4.1

The Raw Costs

Let us first address the scenario in which we were to construct VLS-PV plants at a rate that would enable the world to freeze its conventional generating capacity at its present level. As we have seen, the world would need to generate an additional 610 TWh/year, each year. If we take a 60% capacity factor as being representative of modern fossil-burning power plants, then failure to implement a renewable energy program will, in all probability, require the world to install conventional plants at a rate not less than 116 GW/year. Since the cost of conventional generating capacity is around the US$1 per watt mark, we may infer that the world will in any event be prepared to pay US$116 B per year for new generating capacity. This then is the cost goal that VLS-PV will need to reach if conventional market drivers will be prepared to finance it. For the purpose of comparison, we have already seen that 3 B square meters of conventional PV panels could perform this task. At present-day costs of US$670 per square meter of panel area, for large PV systems [7], we are talking about a global infrastructure cost of US$ 2 T per year. This is already a factor of 17 larger than the expected cost of continued conventional growth, without including the addition of storage, transmission, and (probably rising) real estate costs. By contrast, Raviv [1] has addressed the large-scale introduction of VLS-PV from the direction of CPV technology. His basic argument is that CPV cells enable units that could operate at a minimum of 25% system efficiency to be constructed, at a cost of less than $1 per Wp (excluding the storage costs, which Raviv does include in his plan). The reason that such low costs are possible is that the increased cell efficiency, coupled to the three-orders-of-magnitude reduction in PV materials

COST REQUIREMENTS AND FINANCING

487

that are brought about under very high solar concentration levels (500–1000X), reduces the expensive PV material to an almost insignificant part of the total system cost. In a series of previous publications [1, 3, 4, 8, 9], Raviv’s scheme was applied to a number of individual country case studies. For each case, it was assumed that VLS-PV plants at the GWp scale would be constructed on an annual basis, the electricity would be sold at whatever the going rate is for that country, and the steadily increasing revenues from the plants would be used to pay off the initial infrastructure costs and eventually to fully fund the continued construction of such plants. Roughly speaking, a 1-GWp VLS-PV plant would cost approximately $1 B and, at a nominal tariff of 10¢ per kilowatt hour, it would generate $200 M in annual revenues. Therefore, ignoring for a moment all of the additional costs, once five VLS-PV plants are constructed, their combined annual revenue should suffice to fund the construction of each of the subsequent plants. The fact that CPV technology can enable VLS-PV plants to be cost-effective leads to two unexpected benefits, referred to as “Type 2” and “Type 3” sustainability. The Type 2 stage is reached when all accrued costs are paid off (bank loans, interest, factory, etc.). At this stage, the electricity tariff can be dropped to something like US5¢ per kilowatt hour and the revenue from the existing VLS-PV plants is still sufficient to continue the annual plant construction program without the need for any further external financing. Type 3 sustainability is achieved when, after the construction of a sequence of 29 plants, it becomes necessary to construct two new plants per year thereafter: one to continue the ever-increasing demand for electricity and the other to replace a 30-year-old VLS-PV plant, which, by assumption, would by then have reached the end of its useful working life. Type 3 sustainability generally requires a slight adjustment of the electricity tariff in year 30 (downward in the case of California [8]) in order to allow the continued annual construction of two VLS-PV plants without the need for any external financing. These economic forecasts for CPV were based on an assumed STC cell efficiency of 32% (25% in the field) and an annual VLS-PV plant output of 2 TWh from 12 km2 of land. Many of the world’s deserts have annual DNIs higher than the 2222 kWh m2/year assumed for Raviv’s calculations [1]. This fact, together with steadily improving CPV cell efficiencies (the 41% mark having recently been exceeded [10]), means that in all probability, VLS-PV plants will occupy considerably less land than 12 km2/GWp. To generate 610 TWh/year (i.e., to freeze the continued worldwide growth of fossil fuel consumption for electricity production), it would be necessary to construct 305 GWp/year of VLS-PV plants. If we assume a nominal electricity selling price of US10¢ per kilowatt hour and 5% real interest rate, then the necessary credit line would reach a maximum value of US$1.91 T in year 12 (a figure that compares most favorably with the US$1.39 T the world would in any event have spent on new conventional generating capacity, excluding the additional fuel costs). From then onward, the credit line would decrease and would become fully paid off in year 20. At this stage, there would be 4880 GWp of installed CPV capacity, and the electricity selling price could be reduced to US4.04¢ per kilowatt hour for Type

488

LARGE-SCALE SOLAR CELL ELECTRICITY PRODUCTION

2 sustainability. These plants would be generating 9760 B kilowatt hours of electricity per year, that is, roughly 30% of the entire global projected generation at that time. At the end of year 35, the first set of VLS-PV plants would be 29 years old, and it would become necessary from then on to replace this set as well as to construct a new set each year. That is to say, it would be necessary to construct twice as many plants per year after the initial 35 years. This would require a slight upward adjustment of the electricity selling price—to US4.28¢ per kilowatt hour—in order to achieve Type 3 sustainability. Note also that at year 36, there would be 10,980 GWp of installed solar capacity, generating 21,960 TWh annually, which is about 50% of the world’s projected total electrical generation at that time. All this from infrastructure that had not cost anything and without the need for any fuel. This argument demonstrates that cost should not be an issue. Admittedly, a 20-year payback period is not very attractive for venture capital, but there are many investment portfolios, such as pension funds, that seek long-term secure investments. Clearly, electricity-generating infrastructure that does not depend on the availability or price of fuel constitutes an excellent investment for such purposes.

21.4.2

Feed-In Tariffs

A number of European countries have passed laws that require utilities to purchase PV energy from clients who wish to sell it—at rates that are greatly in excess of the amount it would cost the utility to generate the electricity itself via conventional means. The purpose of such so-called feed-in tariffs is to encourage the public to purchase PV systems that would not otherwise be a cost-effective investment. Spain is a good example, where PV electricity may be sold to utilities for approximately 50¢ per kilowatt hour. To illustrate in a quantitative way the manner in which such feed-in tariffs could help promote a more rapid introduction of a VSL-PV program, it is instructive to take the specific example of Spain and to examine two extreme case scenarios. The first is where a VLS-PV program is introduced and the electricity is sold at the minimum possible tariff that would ensure Type 3 sustainability. The second scenario is where the electricity is sold at the 50¢ per kilowatt hour “feedin” tariff. Because Spain’s electricity production has been growing at an annual rate of 10.6 TWh/year [6], it would be necessary to install 5.3 GWp/year in order to enable the country to freeze its fossil fuel consumption for electricity production. The minimum tariff that could enable this to be carried out would be 6.01¢ per kilowatt hour. Under this scenario, the entire cost of the program would be returned in year 34 (the final year in the effective life of VLS-PV plant no. 1). At that stage, the tariff could be lowered to 4.31¢ per kilowatt hour for the subsequent annual construction of 10.6 GWp (i.e., two sets of VLS-PV plants) of solar generating capacity each year—without the need for any external funding.

TECHNOLOGICAL CHALLENGES

489

Now under this first scenario, it would be necessary to establish a credit line that would reach a maximum of $66.3 B in year 20, before falling to zero in year 34. Clearly, a larger starting tariff would enable the project costs to be returned in a shorter time span. For example, a 10¢ per kilowatt hour starting tariff would result in the required credit line reaching a maximum of only $34.3 B, in year 12, before falling to zero in year 20. However, most interesting of all would be the situation of a 50¢ per kilowatt hour starting tariff (via “feed-in” legislation). Under this scenario, the maximum credit line needed for such a program would be only $12.0 B, which would be reached in year 6, and which would drop to zero by year 8 of the program. At that stage, the cost of continuing to construct 5.3 GWp of VLS-PV plants per year could be fully borne by an electricity tariff of 14.92¢ per kilowatt hour (a very reasonable tariff by today’s European standards). If the feed-in tariff were to be discontinued at this stage and the electricity sold at the new rate, then Type 2 sustainability would continue until year 35 of the program, at which point it would be possible to lower the tariff to 4.31¢ per kilowatt hour for subsequent Type 3 sustainability. One sees, therefore, that even though CPV technology when implemented on the GW per year scale, as envisaged by Raviv [1], can be fully cost-effective without the need for subsidies, the entire program could be speeded up by a factor of approximately 4 via the skillful employment of feed-in tariffs on a scale that is already in practice in Europe.

21.5

TECHNOLOGICAL CHALLENGES

21.5.1

Storage

Raviv’s original VLS-PV scheme recognized the need for storage in order to guarantee dispatchability of the CPV-generated power, and indeed, a certain amount of storage was incorporated into his program. Specifically, his suggested prescription was to incorporate 2 GWh of storage for each GWp of solar generating capacity. However, a moment’s thought suffices to enable one to appreciate that this amount of storage must be significantly increased in order truly to guarantee all-day electrical availability from a VLS-PV plant. This can be seen as follows. At a typical desert latitude of 30°, the shortest day of the year has 10 hours of daylight and 14 hours of night. If such a day (or sequence of days) were to be fully overcast, it would be necessary to charge the batteries with nighttime fossilgenerated energy. It would therefore require 14 GWh of storage capacity (at a nominal 70% round-trip efficiency) in order for “the plant” to be able to guarantee the delivery of 10 GWh of daytime energy: that is, 1 GW of power during the entire 10-hour day. On the other hand, in mid-summer there are only 10 hours available for nighttime battery charging and a daytime power requirement for 14 hours. Therefore,

490

LARGE-SCALE SOLAR CELL ELECTRICITY PRODUCTION

in the event that the following day were to be fully overcast, a 1 GWp VLS-PV plant would need a 2 GW battery in order to absorb 10 hours of nighttime charge and, assuming the same battery efficiency, to deliver 14 GWh during the following day. This argument is obviously a gross over-simplification because totally overcast days are a rarity in deserts in mid-summer. This means that a 20 GWh battery would seem to be greatly oversized for summer usage. Furthermore, a fully charged battery would be unable to accept solar input the following day. However, this simplified argument does indicate that somewhere between 14 and 20 GWh of storage per GWp of generating capacity would be needed to ensure full solar dispatchability under all weather conditions. This is an order of magnitude larger than the 2 GWh in the original Raviv model. We shall return to this matter again when storage is discussed in greater detail below. Naturally, the strategy for using such large storage would need to be configured to take into account weather forecasts and electricity tariffs that vary throughout the day. But it could provide the utility with the opportunity of using part of its nighttime base-line generation more efficiently than is generally possible without large amounts of storage. Therefore, the total cost of such a storage facility should not simply be added to the cost of its associated VLS-PV plant. Thus far, the discussion of storage obviously implies the need for low-cost, high-efficiency, long-life batteries. All of these properties will require further research and development (R&D). But there is another aspect of storage that has not hitherto received much attention and that will also require intensive R&D. It turns out that the inability of conventional electricity-generating plants to switch on and off on the same rapid time scale at which clouds obscure and permit solar irradiation limits the useful size of VLS-PV plants if large amounts of solargenerated energy are not to be wasted [11–15]. As an example, Solomon [12] and Solomon et al. [13–14] studied the compatibility of hourly solar irradiance data from various parts of the Negev Desert with the corresponding load requirements of the Israel electricity grid for the year 2006. They simulated, inter alia, the output of a fixed flat-panel PV system and found that if the Israeli grid could have had a flexibility of 100% (i.e., perfect ability to ramp up and down in synchronization with its varying PV input), then a VLS-PV system as large as 5.4 GWp could have inputted all of its generated energy that year without the need to dump any surplus, and without the need for storage. Figure 21.2 shows the hourly output of such a 5.4 GWp VLS-PV plant (dotted line), together with the corresponding output of the IEC (continuous line) during that week (March 20–26) in which the “no-dump” hour would have occurred, that is, when solar output would have precisely met the total grid requirements (on March 25) without the former having exceeded the latter at any other hour in the year. Such a VLS-PV plant would have made a substantial contribution—17.4% of the annual energy needs—given that the total, fossil-fueled, generating capacity of the grid was 10.5 GW that year. However, the same IEC data indicated that the actual grid flexibility during 2006 was only 65%. This means that only the difference between a day’s total load

Output profile (% annual peak load)

TECHNOLOGICAL CHALLENGES

491

IEC generation profile

0.7

ND size PV output profile

0.6 0.5 0.4 0.3 0.2 0.1 0

0

24

48

72

96

120

144

168

Time (hr)

Figure 21.2. Simulated hourly output of a 5.4-GWp fixed flat-panel PV plant (dotted) and that of the 10.5-GW IEC grid (continuous), assumed 100% flexible, during the week for which the precise “no-dump” condition would have occurred (on March 5, 2006) [12, 13]. Base load

0.7

Potential solarizable load

Solarized part of the load

% annual peak load

0.6 0.5 0.4 0.3 0.2 0.1 0

0

24

48

72

96

120

144

168

Time [hr]

Figure 21.3. Simulated hourly output of an 827-MWp fixed flat-panel PV plant and that of the 10.5-GW IEC grid, for grid flexibility = 65% during the week for which the precise “no-dump” condition would have occurred (April 10–16, 2006). The black band represents the unsolarizable base load; the gray area represents the solarizable portion of the turbine output; and the light gray area indicates the solar contribution [12, 13].

and an inflexible baseload part can receive a contribution from solar power. This important detail reduces the size of the largest possible no-dump VLS-PV plant to 827 MWp. Figure 21.3 shows the corresponding no-dump week of data (April 10–16, 2006), but with 35% unsolarizable base load generation in black, the remaining potentially solarizable load in gray, and the actual solar contribution in light gray. Such a plant would have made only a relatively small contribution (2.7%) to the annual electricity needs and consequently to fuel saving. It turns out that this percentage can be almost doubled by doubling the size of the VLS-PV plant and

492

LARGE-SCALE SOLAR CELL ELECTRICITY PRODUCTION

allowing approximately 10% of its total output to be dumped rather than used by the grid [12, 13]. But this is about as far as one can go without the use of storage. However, as first pointed out by Denholm and Margolis [11] and elaborated in greater detail by Solomon [12] and Solomon et al. [15], in addition to the energy capacity of storage one must also take into consideration its power capacity, namely the rate at which it can be charged and discharged. In the case of the Israeli grid, Solomon et al. [15] found that by employing appropriately sized storage with optimally chosen properties, and changing the strategy for operating the various conventional plants (including those of the base load variety!) it could be possible to increase the contribution of VLS-PV plants to approximately 60% of the annual grid requirements. Today, most utilities naturally operate their various plants in a manner that balances security of supply and profitability. With the expected advent of cost-competitive solar plants, the detailed strategy for grid management may be expected to change, particularly if conventional fuel sources begin to dwindle. The above discussion implies that special storage systems may need to be developed if VLS-PV plants of the CPV variety are to play an important role in large-scale power production. This is because of the intermittent nature of their power output. Because even fast-ramping turbines have ramp-down times of order 10–15 minutes, relatively sudden bursts of PV power would not lead to any fuel saving unless appropriately fast-acting storage were to be available. Supercapacitors could have the necessary speed for absorbing the energy, but only on a one-time basis. Depending upon the frequency of such cloud events, the supercapacitor may not have sufficient time to trickle its charge into the grid before being called upon again in order to absorb a second transient charge of energy. With appropriate meteorological data (i.e., DNI measurements on a fast time scale, and taken over a range of distances comparable to the size of a VLS-PV plant) it would be possible to assess the frequency of such events and to ignore them if the energy involved were to be a small fraction of the total annual solar availability. But such measurements have yet to be made. Furthermore, all of the simulations discussed above were performed with data on an hourly basis. Therefore, in addition to meteorological data on an appropriately fast time scale, it will be necessary to study grid output data with a time resolution that enables plant ramp-up and ramp-down properties to be quantified with greater detail.

21.5.2

Transmission

Long distance AC power transmission is conventionally carried out at high voltages in order to reduce joule losses as much as possible. On the other hand, it is not feasible to raise the voltage too far because corona discharges to the ground may dominate the joule losses. Typical voltages employed today for such purposes, depending on country, are accordingly found in the range 115–765 kV. As a simple estimate of the joule losses to be expected from the wide-scale implementation of a VLS-PV program, consider a line of 1000 km in length with an AC resistance (for 1272 AWG [16]) of 0.054 Ω/km. Suppose 1216 A (the

TECHNOLOGICAL CHALLENGES

493

maximum current in this manufacturer’s catalog) is injected at one end of the line at a voltage of 765 kV, which corresponds to an injected power of 930 MW, roughly the peak output of a 1-GWp VLS-PV plant. The joule loss would accordingly be 80 MW, or nearly 9%. On the other hand, a recently published announcement about an ultrahighvoltage 800-kV DC power line, of 2000-km length, that will deliver 6400 MW of power at a loss less than 7% [17] implies an effective line resistance less than 7 Ω. From this, one may infer that for such a line, operating at a similarly high DC voltage, the loss in power from a 930 MW VLS-PV plant over a distance of 1000 km would be less than 1%. Joule losses can, in principle, be reduced to zero via the employment of superconducting cables. However, at present, even so-called high-temperature superconductors require cryogenic cooling to liquid nitrogen temperatures. There is thus a parasitic energy requirement for liquefying the appropriate quantities of nitrogen and maintaining the liquid at approximately −200°C. Therefore, unless truly ambient-temperature superconductors are discovered, UHVDC will probably be the technology of choice for transmitting the power from VLS-PV plants over long distances. These, of course, are only the joule losses. Other losses are to be expected from the various electronics, connections, switches, and so on, needed to get the power into and out of the lines. Nevertheless, the above estimates would seem to indicate that UHVDC technology is sufficiently advanced to allow low-loss power transmission over distances comparable to those necessary to connect the Sahara to Europe, or the deserts in the SW United States to the population centers in the NE.

21.5.3

Smart Grids

The possibility of long distance, low-loss electrical transmission can, in principle, solve more problems than merely shunting power from deserts to large population centers. This is because such a grid would encompass regions having considerably different weather conditions and load requirement patterns. With an appropriately enlarged control system, load sharing could occur on a scale of complexity that was never possible in the past. To a certain extent, this would also alleviate the dispatchability problem because different cloud conditions at different locations would smooth out the combined solar inputs to such a grid from widely located VLS-PV plants. Weather forecasting would obviously also play an important role. In addition, each 15° of longitude between VLS-PV plants would extend solar availability by 1 h. Thus, the continental United States and the Sahara desert would each provide a “solar day” extended by several hours. The random on/off output of a CPV plant on cloudy days bears a certain similarity to the uncertainty in output from a wind plant. But the former is more manageable. This can be seen as follows. On a cloudless day, the output of a CPV plant depends almost entirely on geometric factors (time of year, time of day,

494

LARGE-SCALE SOLAR CELL ELECTRICITY PRODUCTION

geographic latitude). Clouds diminish the clear day pattern in a somewhat random manner. However, because, unlike the wind, clouds are visible and their velocities measurable, an appropriately automated observation system associated with each VLS-PV plant could provide advance warning to a smart grid about the expected performance of each plant during the coming hour or so [4]. Hence, the increasingly popular term “smart grid” would ultimately include input from “smart PV plants,” and probably too, “smart storage systems.”

21.6

CONCLUSIONS

In this chapter, the argument has been made that CPV technology is the key to a massive replacement of fossil fuel on a worldwide scale. High cell efficiencies make it less land intensive than the other solar technologies and, because of high levels of optical concentration (typically around 1000X), only relatively small amounts of costly PV material are needed. These two factors hold out the hope that VLS-PV plants of the CPV variety could be economically competitive with fossil-powered plants. It was shown that the world’s deserts, specifically the Sahara and the SW deserts of the United States, are more than large enough to encompass VLS-PV plants of a scale necessary to halt the present growth rate of fossil-fueled plants in Europe and in the United States, and similar results can be shown for the rapidly developing economies of China and India [9] via the Gobi and Thar deserts, respectively. Cost estimates based on the Raviv model [1] indicate that the economics of such a program are favorable even without the need for subsidies. Moreover, the timescale of investments required is of a kind that should be attractive to pension funds and other portfolios that require longer-term solidity. Storage continues to a be an area in which further R & D is needed— particularly high-speed storage for smoothing out the rapid on/off effect that passing clouds would have on a CPV plant. However, the urgent need for efficient storage is alleviated to a certain extent by the possibility of combining a VLS-PV program with a UHVDC smart grid system. Therefore, in conclusion, the technology is ripe for a cost-effective VLS-PV program to go ahead. In the United States, China, and India, such a program could commence at once, because each country enjoys a single political system. In Europe/Africa, a number of international agreements would need to precede the implementation of such a program.

ABBREVIATIONS AC—alternating current AWG—American wire gauge B—billion CPV—concentrator photovoltaics DC—direct current

REFERENCES

495

DNI—direct normal irradiance GWp—gigawatt peak IEC—Israel Electric Corporation M—million MWp—megawatt peak NE—northeast PV—photovoltaic or solar cell R & D—research and development ST—solar thermal STC—standard test condition SW—southwest T—tera (1000 billion or 1 trillion) UHVDC—ultrahigh-voltage direct current VLS-PV—very large scale photovoltaic Wp—watt peak

REFERENCES [1]

[2] [3] [4] [5] [6] [7] [8] [9] [10]

D. Raviv and R. Rosenstreich. Powering Europe from the sun: The scenario of largescale centralized CPV power stations. In 19th European Photovoltaic Solar Energy Conference, W. Hoffmann, J.-L. Bal, H. Ossenbrink, et al., eds., pp. 3745–3750. Munich: WIP and Florence: ETA (2004). K. Kurokawa, ed. Energy from the Desert: Feasibility of Very Large Scale Photovoltaic Power Generation (VLS-PV) Systems. London, James & James (2003). K. Kurokawa, K. Komoto, P. van der Vleuten, et al., eds. Energy from the Desert: Practical Proposals for Very Large Scale Photovoltaic Systems. London, Earthscan (2007). K. Komoto, M. Ito, P. van der Vleuten, et al. eds. Energy from the Desert: Very Large Scale Photovoltaic Systems for Socio-Economic Development. London: Earthscan (2009). K. Zweibel, J. Mason, and V. Fthenakis. A solar grand plan. Scientific American January: 48–57 (2008). BP Statistical Review for 2008. Available at http://www.bp.com/productlanding.do? categoryId=6929&contentId=7044622 (2008). L. M. Moore and H. N. Post. Five years of operating experience at a large, utilityscale photovoltaic generating plant. Progress in Photovoltaics: Research and Applications 16, 249–259 (2008). D. Faiman, D. Raviv, and R. Rosenstreich. Using solar energy to arrest the increasing rate of fossil-fuel consumption: The southwestern states of the USA as case studies. Energy Policy 35, 567–576 (2007). D. Faiman. Implementing VLS-PV on a worldwide scale, in a cost-effective manner. In Renewable Energy 2006 Proceedings, October 9–13, 2006, Chiba, Japan, CD-ROM, pp. 63–68. K. Schneider. World record 41.1% efficiency reached for multi-junction solar cells at Fraunhofer ISE. http://www.ise.fraunhofer.de/press-and-media/press-releases (January 14, 2009).

496

LARGE-SCALE SOLAR CELL ELECTRICITY PRODUCTION

[11]

P. Denholm and R. M. Margolis. Evaluating the limits of solar photovoltaics (PV) in traditional electric power systems. Energy Policy 35, 2852–2861 (2007). A. A. Solomon. Matching the intermittent output of renewable energy systems to the needs of an electricity grid: A mathematical study with important policy implications. PhD thesis, Ben-Gurion University of the Negev, Beersheba, Israel (2010). A. A. Solomon, D. Faiman, and G. Meron. An energy-based evaluation of the matching possibilities of very large photovoltaic plants to the electricity grid: Israel as a case study. Energy Policy (2010), doi: 10.1016/j.npol.2009.12.024. A. A. Solomon, D. Faiman, and G. Meron. The effects on grid matching and ramping requirements, of single and distributed PV systems employing various fixed and sun-tracking technologies. Energy Policy (2010), doi: 10.1016/j.npol.2010.02.056. A. A. Solomon, D. Faiman, and G. Meron. Properties and uses of storage for enhancing the grid penetration of very large photovoltaic systems. Energy Policy (2010), doi: 10.1016/j.npol.2010.05.006. Phelps Dodge Intl. Corp. On-line spec sheets available at http:www.pdic.com. T. F. Armistead. Longest, most powerful line will apply new technology. http://enr. construction.com/news/powerindus/archives/080109d.asp (January 9, 2008).

[12] [13] [14] [15] [16] [17]

PART VI THIN FILMS AND X-RAY IMAGER TECHNOLOGIES

22 MARKET OVERVIEW OF FLAT PANEL DETECTORS FOR X-RAY IMAGING CARL LACASCE, LARRY PARTAIN, AND CHUCK BLOUIR Varian Medical Systems

22.1

INTRODUCTION

Thin-film solar cells for terrestrial electric power production are of interest for their potential lower cost, but they have other characteristics that make them of value for other applications. They can be deposited on large-area glass substrates and devices can be integrally interconnected. These traits are valuable for medical imaging devices. While thin-film semiconductors have large electronic defect levels limiting their efficiencies for energy conversion applications, this trait makes them tolerant to X-ray exposures for medical X-ray imaging. This and the next three chapters address the application of advanced thin film solar cell technology to the field of flat panel digital electronic X-ray imaging. The present chapter provides an introductory overview from a market perspective that is quite positive. Chapter 23 covers the device physics of the dominant amorphous silicon semiconductor thin film technology (on glass substrates) that enables largearea flat plate electronics at much lower costs than traditional crystalline semiconductors. This includes the p+-i-n+ structures required for solar cell behavior plus the alternate n+-i-n+ configurations (plus field effect electrodes) that provide the TFTs essential for imaging. Chapter 24 describes the placement of scintillator thin films on the solar cell thin films that convert X-ray photons into the visible light photons detected by solar cells. It further details the separation of the solar cell into the millions of “photodiodes” laid out in rows and columns on glass panels that give X-ray images as each photodiode “pixel” is electronically switched Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

499

500

MARKET OVERVIEW OF FLAT PANEL DETECTORS

3,500,000

Market Size ($000)

3,000,000

Mammo

2,500,000

Dynamic 2,000,000

RAD

1,500,000 1,000,000

CR

500,000 0 2007

2008

2009

2010

2011

2012

2013

Year

Figure 22.1. The wholesale flat panel market.

by its TFT into a digital readout amplifier. It covers the critical low-noise performance essential for high-quality imaging. Finally, Chapter 25 describes the replacement of both the scintillator and the photodiodes in the X-ray imager with photoconductor films, dense and thick enough to directly absorb X-ray photons in a manner that provides potentially higher X-ray image quality at even lower costs. Medical X-ray imaging, much like the photography industry, has been undergoing a transition: a transition from analog-based technologies to solid-state digital technologies. Since the earliest days of X-ray, analog devices such as film and image intensifiers were used to convert X-ray information into a visual media. As the demands on hospital resources have increased, the need for more effective imaging devices has also increased. Hospitals require cost-effective imaging equipment capable of handling higher patient throughput, easy integration with modern picture archive solutions (PACS), and high reliability (Fig. 22.1). While human diagnostic imaging needs remain the main catalyst for digital detector design and evolution, other X-ray-based imaging markets have also been monitoring and transitioning to their use. These markets include veterinary imaging, security and inspection systems, and NDT systems. Essentially, all film-based imaging modalities are either in the transition phase or in the investigation phase to flat panel digital imaging technology. The worldwide FPD market has grown dramatically since the late 1990s. In 2009, the market was estimated to be more than $2 billion per year and is growing at more than 15% per year. Medical imaging professionals believe this trend will continue as the performance of FPDs improve, FPD costs are reduced and new applications are created.

MARKET HISTORY: FILM TO DR

22.2

501

MARKET HISTORY: FILM TO DR

The transition from film to digital imaging began in the human diagnostic imaging departments of many hospitals. The first technology was called CR and was introduced in 1983 by Fuji Photo Film Co. CR utilizes a photostimulable phosphor composed of europium-activated barium fluorohalide or europium-activated barium fluoride bromide. One of these mixtures is coated on a screen similar to a standard intensifying screen. This screen is then placed in a special cassette holder that functions like a standard film/screen cassette. The appeal of CR technology is ease of integration as CR works with standard X-ray imaging equipment. The downside to CR technology has been the lack of any significant workflow improvement in the radiology department. Cassettes still have to be manually processed similar to the conventional film cassette process. So, the patient still has to wait in the procedure room until after the technologist can verify that the proper imaging has been obtained. FPDs date back to the early 1970s when scientists at Xerox PARC began development work on amorphous silicon. Over the next 20 years, Varian Medical Systems, Xerox, GE Company, Thompson, and many other companies invested in the technologies that are utilized in modern FPDs. This technology was dubbed DR, indicating its ability to directly capture and output an image without manual operator intervention. By the late 1990s, many companies had established commercial enterprises for the manufacture of FPDs for DR applications, but the digital imaging industry remained largely dominated by CR technology until the early 2000s. Amorphous selenium was the basis for the first full-size DR image detectors offering a compact footprint, high resolution images, and fast image display. Amorphous selenium detectors were referred to as “direct” capture detectors because the X-ray photons that hit the selenium layer were directly converted to electrons for storage. Unfortunately, amorphous selenium also required very high bias voltage on the photoconductor and the selenium layer was very sensitive to temperature and humidity, all of which lead to premature failure of the detectors and poor reliability. With the exception of breast imaging, amorphous selenium detectors have been largely replaced today in most markets by the amorphous silicon detectors sold by all the major OEMs. In today’s marketplace the most common implementation of DR detector technology is based on large amorphous silicon photodiode arrays. These arrays are made up of several million TFTs coupled to photodiodes which store charge generated by light produced by X-rays absorbed in the conventional gadolinium oxysulfide or cesium iodide scintillation layers. This design is referred to as an “indirect” capture device because X-ray photons are converted to visible light photons by the scintillator before being converted to electrons for storage. Indirect capture DR products are used almost exclusively by all major diagnostic imaging OEMs and system integrators. This design provides instant imaging with excellent diagnostic quality and little operator intervention. Wait times for patients is dramatically reduced while the workflow in the radiology department is streamlined.

502

MARKET OVERVIEW OF FLAT PANEL DETECTORS

General radiography has been most impacted by the use of flat panels. Dose reduction, gains in patient throughput, and ease of use are just some of the benefits realized in radiography. The latest developments include battery-powered detectors with wireless image transmission capability. The first to announce a wireless capable detector was Trixell in 2006. The latest and most revolutionary development in this market segment was introduced at the RSNA 2008 by Carestream Health. Carestream Health’s introduction of a 14 × 17 in. (36 × 43 cm) wireless detector in a package the same size as a conventional film cassette will solve many of the problematic mechanical upgrade issues generally associated with upgrading existing film- or CR-based radiographic equipment to digital technology. In the 1990s, digital fluoroscopy systems began to appear in hospitals. Early digital systems were based on traditional image intensifiers coupled to analog then to CCD cameras and digital imaging workstations. Fluoroscopic studies as well and digital photo spot images were captured in the workstation and could be viewed on monitors, saved to other media such as videotapes, or could be used to create prints or films. This technology quickly became the industry standard, and most new systems marketed by all major OEMs included digital imaging. Following the movement of film-based general radiographic imaging to FPD technology, the next evolution for DR was to address the unique needs of fluoroscopy. The development of readout electronics capable of scanning the detector at up to 30 fps facilitated the use of these new compact imaging devices in modalities requiring live fluoroscopic imaging. While image quality was acceptable with image intensifier-based products, their huge footprint and short life span made them ripe for replacement with DR’s compact new amorphous silicon photodiode technology. 22.3

APPLICATIONS

Conventional diagnostic imaging applications such as radiography, fluoroscopy, angiography, cardiology, and urology now utilize DR technology. DR has also facilitated the development of several new imaging capabilities. Starting with NDT using slow acquisition rates and inanimate objects and rapidly progressing to realtime high-speed imaging of human anatomy, CBCT has made it possible to image a tumor during radiation treatment in order to more accurately plan and target the treatment beam. Smaller field-of-view detectors with CBCT capability have revolutionized the market for dental implants, making it possible to precisely place dental implants around critical nerves in the jaw. Ongoing research is moving these specialized DR detectors ever closer to performing actual diagnostic quality CT as well as many other imaging tasks such as cargo container inspections, bomb disposal, and food inspection, just to name a few. 22.3.1

Cardiovascular

FPDs for use in cardiovascular applications were introduced in early 2000 by GE. Doctors quickly recognized the benefits: larger field of view, high resolution, and

APPLICATIONS

503

better patient positioning as compared to image intensifier-based systems. The market quickly expanded the use of flat panels as Philips, Siemens, and Toshiba introduced flat panels into their product lines. By 2008, all the major medical OEMs had removed image intensifiers from their cardiovascular products.

22.3.2

Angiography and Other Applications

Angiography and other specialized applications followed closely behind cardiology in the transition from image intensifier-based imaging to FPDs. As the field of view, image acquisition speeds, and overall image quality capabilities increased, OEMs began adopting flat panels for use on standard C-arm-based angiography systems and on general radiographic and other diagnostic imaging systems. Many procedures such as peripheral run-off studies were dramatically simplified with the use of square or rectangular detectors, and overall patient access has dramatically improved along with a much longer life expectancy (Figs. 22.2 and 22.3). One other huge benefit of amorphous silicon flat panel image detectors over their image intensifier-based predecessors is the elimination of the need to constantly increase the radiation dose over time to maintain image quality.

Figure 22.2. Interventional R/F table system (Hitachi).

504

MARKET OVERVIEW OF FLAT PANEL DETECTORS

Figure 22.3. Multipurpose C-arm R/F table for angiography and GI (Hitachi).

Figure 22.4. Dental cone beam CT system, Imaging Sciences International.

APPLICATIONS

22.3.3

505

Dental CBCT

Flat panels have virtually created a new standard of care within the dental imaging community. Cone beam CT systems dedicated for use in dental implants are now in use throughout the world. Dentists are now able to image their patients in-office, use dedicated software customized for implant procedures, and, in some parts of the world, can bill for a CT exam (Fig. 22.4).

22.3.4

Mammography

Many modalities have attempted to address the huge volume of diagnostic images required to screen for breast cancer globally. While X-ray film-based mammography systems still dominate this procedure, flat panel-based systems are gaining market share. First introduced in early 2000 digital mammography systems have evolved rapidly, and new study results affirm that digital imaging speeds the screening process and makes it possible to use computer-aided diagnosis to reduce the tremendous workload this procedure places on radiologists. Currently, both amorphous silicon and amorphous selenium detectors are used in this modality.

22.3.5

Medical CBCT

CBCT remains in its infancy but is gaining momentum as a diagnostic tool for whole body imaging. Several systems based on mobile C-arms are showing promise along with a unique O-arm system sold by Medtronic. The O-arm system sets the standard currently for flat panel-based CBCT during surgery. Several hurdles remain to be addressed before FPDs can perform at ultrahigh speeds such as that required for diagnostic CT of the heart, but image quality is excellent for the less dynamic areas of the body (Fig. 22.5).

Figure 22.5. O-arm cone beam CT system (Medtronic).

506

MARKET OVERVIEW OF FLAT PANEL DETECTORS

Figure 22.6. Varian Oncology Systems treatment platform with IGRT.

22.3.6

Portal Imaging

Portal imaging started a revolution in oncologic imaging during radiation treatment. By providing the ability to create an image from the exit beam during treatment, oncologists were able to begin the process of seeing what they were treating during the treatment. Unfortunately, due to the megavoltage X-ray beam used in this procedure, image quality was poor. Rapid evolution, however, quickly resulted in a low-energy X-ray source and detector being mounted orthogonally to the treatment beam. This revolution, called OBI, brought near diagnostic quality imaging to the oncologist for tumor position verification during treatment. Ongoing improvements to this new technology, now termed IGRT, have made FPDs an integral part of, and the standard of care on, all new radiation treatment systems (Fig. 22.6).

22.4

FLAT PANEL MARKET EVOLUTION

The early market landscape included many companies who were either developing the basic technology or finding ways to enhance and productize core technologies

INDUSTRY LANDSCAPE

507

designed by others. Some designs attempted to tile large numbers of CMOS detectors together to form a large detector array, but major assembly and reliability problems proved too difficult to overcome. Physically large detectors incorporating one or more CCD cameras along with an elaborate lens system appeared on the market from several companies, but this design was quickly dubbed an interim technology and sales rapidly declined. The modern FPD market is dominated by amorphous silicon photodiodebased detector designs for all modalities except for mammography. The high resolution and DQE of amorphous selenium at mammography energies have become more desirable for breast imaging, but these detectors continue to be plagued by poor reliability. GE continues to successfully market an amorphous silicon photodiode-based detector for mammography with excellent image quality and reliability.

22.5

INDUSTRY LANDSCAPE

Although not a major player today, Analogic has been a long-time participant in the digital imaging system business. In June 1999, Analogic purchased the medical detector and pure metal assets of Noranda, marking its entry into the flat panel market with amorphous selenium-based radiographic detector technology. In today’s marketplace, there are some smaller companies marketing selenium-based products, but Hologic’s products designed for mammography make up the largest portion of systems based on this photoconductor technology. In April 1999, Canon formed Canon Medical Systems by combining their radiography business and a floundering PACS operation that they owned. Canon is focused on amorphous silicon panels for radiographic applications, and currently, Canon has several panel sizes on the market. Their approach has been to provide both the DR detector and the image processing software designed specifically to maximize the image quality of their panels. Canon panels use a Schottky barrier photodiode with slower speed operation than the pin photodiodes used by most other flat panel manufacturers. Thus, the Canon panels are only capable of conventional radiographic imaging. GE, along with their partner and supplier PerkinElmer, has been one of the top suppliers of DR-based diagnostic imaging products for the human medical marketplace. GE’s Revolution XQ/I dedicated chest system was first introduced at RSNA 1998. In March of 2000, GE also received FDA clearance to market its Senographe 2000D digital mammography system and their Innova 2000 flat panel-based cardiac system. GE followed these products with many other systems to date, and they remain one of the top OEMs using DR technology. Kodak entered the CR market in 1994 with the Ektascan Imagelink. Kodak sold off their healthcare imaging division, which became Carestream Health in 2006. Carestream Health is focused on digital imaging systems that include CR and DR technologies. Carestream Health recently revolutionized the general radiographic market with the introduction of a film cassette-sized wireless DR detector.

508

MARKET OVERVIEW OF FLAT PANEL DETECTORS

In March 1997, Philips Medical Systems, Siemens Medical Systems, and Thomson Tube & Electroniques (now Thales) joined forces in a consortium to develop flat panel products by creating a company called Trixell S.A.S. Trixell would be the source of DR technology for most products sold by Siemens and Philips. Varian Medical Systems was the first to market in the late 1990s with amorphous silicon photodiode-based detectors capable of real-time imaging for fluoroscopy. This was followed closely by products from GE and then by Trixell. Varian remains the largest independent supplier of FPD products in the world. Varian also pioneered CBCT technology using FPDs and remains the leader in this emerging diagnostic imaging application.

22.6

MARKET SUMMARY

Over the past 10 years, FPDs have revolutionized the radiology department in the modern hospital. The radiology department is more efficient through higher patient throughput; the dose per given X-ray is lower than with analog technology; images can be easily transmitted electronically to radiologists at remote locations; and the doctors have more imaging information available to them via advance applications. During the same period of time, the costs of FPDs have reduced and, as with all technology-based products, will continue to see price erosion as more and more players enter the market and as overall product volumes grow. Some new Schottky barrier photodiode designs allow the detector arrays to be built alongside the mega-volume LCD displays built for flat screen TVs and for computer monitors. These evolutions will drive costs down while allowing suppliers and OEMs to evolve the technology and at the same time return acceptable margins to their shareholders. The opportunity for smaller hospitals and healthcare facilities in third world countries to purchase FPD-based equipment will increase, making it very possible that all X-ray images in the future will be done with digital FPDs.

22.7

FUTURE TRENDS

The costs of digital technology, including large-area flat panel displays and flat panel X-ray imagers, continually decrease as their capabilities dramatically increase. The resulting growth with new applications is so large that the dollar size of the total world wholesale X-ray flat panel market, presently approaching $2.5 billion, continues to rise attractively at double-digit rates. This is somewhat analogous to Moore’s law like trajectory for digital computers that play an integral role in this whole imaging technology base. As with computers, the underlying imaging technology must continually evolve to keep pace on such a technology advancement treadmill. While the standard now is higher-speed pin photodiodes covered

ABBREVIATIONS

509

with scintillators, simpler and lower-cost Schottky barrier technology is beginning to replace the pin devices in the lower-performance, lower-speed, but larger-area radiology segment of this market. CR is a comparatively mature technology in terms of further cost reductions and improved performance, and its market share is expected to decrease continually. The highest-resolution performance is currently provided commercially with amorphous selenium-based photoconductor imagers. However, laboratory examples are now being reported for photoconductors of even higher sensitivity and more robust characteristics and with a potential for much lower costs (Chapter 25). In addition, low-cost, nonphotolithographic fabrication of laboratory prototypes with modest spatial resolution is emerging even on flexible plastic substrates (Chapter 23). Only time, the market, and continuous technical advancements will determine the winning technology champions of the future. ABBREVIATIONS C-Arm—support structure shaped like a “C” CBCT—cone beam computed tomography CCD—charge-coupled device CMOS—complementary metal oxide semiconductor CR—computed radiography CT—computed tomography DQE—detective quantum efficiency DR—digital radiography FDA—U.S. Federal Drug Administration FPD—flat panel detector fps—frames per second GE—General Electric GI—gastrointestinal i—undoped or intrinsic semiconductor region IGRT—image-guided radiation therapy LCD—liquid crystal display n+—heavily negatively doped semiconductor region NDT—nondestructive testing O-arm—support structure shaped like an “O” OBI—on-board imaging OEM—original equipment manufacturer p+—heavily positively doped semiconductor region PARC—Palo Alto Research Center PACS—picture archival and communications system RAD—general radiographic X-ray imaging market R/F—radiography and fluoroscopy RSNA—Radiological Society of North America TFT—thin-film transistor

23 AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES ROBERT STREET Palo Alto Research Center

23.1

INTRODUCTION

The interest in a-Si started as a basic research activity aimed at understanding how atomic disorder in solid-state materials has affected the quantum electronic properties of a semiconductor [1] and has grown into a $100 B industry making displays, solar cells, and X-ray imagers. Solar cells were the first application [2], and this industry has grown slowly but steadily since about 1980 and is perhaps now at a cusp in the growth with the current renewed interest in renewable energy. The display application of a-Si TFTs grew more rapidly to the present $100 B LCD business. Digital X-ray imagers are a smaller market than either displays or solar cells but are driving a technology revolution, replacing X-ray film [3]. The common property of these applications is the ability to fabricate TFTs and p-i-n photodiodes in a way that can be scaled to a very large area. This chapter discusses the electronic and material properties of a-Si and how these determine the structure and performance of the devices, and what issues limit their performance. 23.2

PROPERTIES OF a-Si

a-Si and the various alloys and doped layers that form the technology are deposited in a PECVD reactor [4]. Silane, SiH4, is the usual source gas, but there are a wide variety of other gases for the different compounds. The energy provided to the plasma dissociates the gases, and the resulting chemical radicals deposit on the surface and form the film. The growth surface and the resulting film are shown schematically in Figure 23.1. As a result of the hydride gas sources, the a-Si film contains about 10 atomic % of hydrogen (Fig. 23.1), which proves to be critical Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

511

512

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES a-Si growth surface H

H

H

H

H

H2 H

H

H H

H

H

H

H

H

H

H H H

H H

H H

H

H

H

H H

Log(density of states) (cm–3eV –1)

Figure 23.1. Left: illustration of the surface of an a-Si film during PECVD growth, showing the attachment of various Si-H radicals. Right: illustration of the bulk atomic structure of a-Si showing the atomic disorder and the bonded hydrogen.

valence band

mobility edge

1022 1020

extended states

1018 1016

band tail localized states

conduction band

extended states

defect states 0

1 Energy (eV)

2

Figure 23.2. The density of states of amorphous silicon showing the conduction and valence bands with extended state conduction, the localized states of the band tails, and the dangling bond defect states near the middle of the bandgap. The mobility edge divides extended from localized states.

for the electronic properties but which is also the source of the device instability problems that are discussed below. One of the main reasons that a-Si is in largescale production for TFT arrays is that the PECVD process is scalable to a very large area, and the deposition process also provides the doped contacts needed for TFTs, solar cells, and photodiodes, along with the dielectrics needed for TFT backplanes.

23.2.1

Electronic Properties of a-Si

The amorphous atomic structure has a direct effect on the electronic properties of a-Si. Figure 23.2 shows the electronic density of states, which is divided into three

PROPERTIES OF a-Si

513

distinct regions [1, 5]. The conduction and valence bands contain extended electron and hole states allowing for electronic conduction. The atomic disorder and consequent scattering reduces the carrier mobility compared to crystalline silicon by ∼100X to a value of about 10 cm2/Vs. At the edge of the bands are the band tails comprising localized states, which also originate from the atomic disorder and can easily trap mobile carriers. As a result, carrier transport is by a trapping and release process between the band tails and the extended states of the bands, and transport occurs near the mobility edge (see Fig. 23.2), which is the energy separating these two groups of states. The valence band tail turns out to be wider than the conduction band tail—a result largely related to the particular bonding properties of silicon—so that the effective electron mobility is about 1 cm2/Vs and the hole mobility is two to three orders of magnitude lower [5]. Therefore, any device that needs high mobility is restricted to electron transport. Within the bandgap are defect states identified as silicon dangling bonds— silicon atoms with three rather than four bonds. Good quality undoped a-Si has less than 1016 cm−3 defects, which is sufficiently low to allow good electronic transport and photoconductivity properties. The low defect density is the main consequence of the hydrogen that is retained in the film, without which the defect density would be at least 1000 times higher. As we shall see, the defect density is rather easily increased, either by strong illumination to the detriment of solar cells or by charge induced in the bands, which affects the transistor stability. The design of devices and circuits must take these instabilities into account. The density of states distribution in Figure 23.2 is known reasonably accurately for the valence and conduction bands, but the shape of the defect band is known only approximately. a-Si is readily doped by the addition to the plasma deposition of a gaseous hydride containing either a group III (i.e., boron) or group V (i.e., phosphorus) during the plasma growth [6]. The temperature dependence of the conductivity in heavily doped a-Si is illustrated in Figure 23.3 and indicates that there is a range of conductivity values depending on the deposition conditions and the treatment of the films. The increase in conductivity by doping is constrained by the band tails, which prevents the Fermi energy from moving close to the band edge. The wider valence band tail makes p-type doping less effective than n-type. The temperature dependence of the conductivity is thermally activated and reflects the energy separation of the Fermi energy from the mobility edge, although the activation energy decreases at lower temperature; near room temperature, the energy is 0.1–0.2 eV for electrons and ∼0.3 eV for holes. Doping also increases the defect density substantially [5] by a mechanism related to the solar cell instability (see Section 2.2). The result is that doping is effective in moving the Fermi energy, but the conductivity is modest compared to crystalline silicon, and the minority carrier lifetime of doped a-Si is poor because of recombination with the extra defects. The electronic properties of the doped film determine the design of solar cells and photodiodes, as is discussed later. The high defect density of the doped film is useful as it enables the formation of thin tunnel junctions in a-Si tandem cells.

514

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES

1

Conductivity (Ω–1cm–1)

n-type 1% PH3

10–2

p-type 1% B2H6

–4

10

10–6 2

3

4

5

Temperature 1000/T (K)

Figure 23.3. Conductivity of heavily doped n-type and p-type a-Si, corresponding to a deposition gas with 1% dopant during the PECVD deposition of material. The range of values resulting from different growth conditions and sample treatment is indicated.

While mobility is the most important parameter for a TFT, the mobilitylifetime product, μτ, determines the performance of the solar cell or photodiode, because the distance that mobile change travels before deep trapping is μτE, where E is the electric field. For a charge, Q0, induced near one side of a thin a-Si film, in a uniform electric field, the fraction of the charge that is collected by the external circuit is given by [7] Q Q0 = (μτV d 2 )[1 − exp ( − d 2 μτV )] ,

(23.1)

where d is the film thickness and V is the voltage. Figure 23.4 shows measurements of the electron and hole μτ product for doped and undoped a-Si [8]. The different μτ values for the undoped material reflect different defect densities ranging from about 3 × 1015 cm−3 for the upper right data in Figure 23.4 to 1018 cm−3 for the lower left. The value of μτ in undoped a-Si is larger for electrons than for holes, which is important for the design of photodiodes and solar cells. The μτ values for doped samples reflect the increasing defect density with doping and that only minority carriers are trapped.

23.2.2

Hydrogen and Defect Stability

One of the important limitations of a-Si is the instability of the a-Si material to the formation of additional dangling bond defects. Both solar cells/photodiodes and

PROPERTIES OF a-Si

515

Figure 23.4. Electron and hole mobility-lifetime products measured on doped and undoped a-Si. The diagonal data correspond to undoped a-Si with a range of defect densities. The doped samples have a range of doping levels [8].

1018 Saturated value

A44

NS (cm–3)

1017

HI

MI

LI

1016

1015 10–5

10–3

10–1 101 Time (hr)

103

Figure 23.5. Defect creation kinetics resulting from prolonged illumination at different intensities: high (HI), medium (MI), and low (LI), showing the power law creation kinetics leading to an eventual steady state [10].

TFTs are affected, and the defect creation imposes constraints on the design and performance of the devices. The metastable defects in solar cells [9] are created by the recombination of electron–hole pairs. Figure 23.5 shows the time dependence of the defect density at different illumination levels [10]. The defect density increases by 100X to about 1017 cm−3, where it then tends to stabilize. The increase

516

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES

in defect density is sublinear in time, roughly following a t1/3 dependence, and also sublinear in illumination intensity G. Hence, the defect density increases quickly at the start of exposure and then progressively more slowly. Stutzmann et al. [11] proposed that defects are created by the bimolecular recombination of electrons and holes in the bands, while the dominant recombination channel was by monomolecular recombination through the defect states in the gap. This model leads to a defect creation rate given by N D = const G 2 3t 1 3,

(23.2)

which fits the data well [11]. According to this model, the energy released by the electron–hole recombination breaks silicon bonds and creates the dangling bond defects. The hydrogen contained within the a-Si film is closely involved in the process—either a Si–H bond is broken or the hydrogen stabilizes the bond breaking of a Si–Si bond. Metastable defects are removed by annealing at 150–200°C, with an activation energy of ∼1.5 eV that corresponds to the hydrogen diffusion energy. A similar defect creation mechanism applies to TFTs where the source of the energy is the charge induced into the conduction band when the TFT is turned on. To summarize, undoped a-Si has a low density of deep gap states and has a carrier mobility of about 1 cm2/Vs for electrons and considerably lower for holes. n-type doped a-Si has room temperature conductivity near 0.1 S/cm, with p-type 10–100 times lower, and both have a much higher density of deep recombination centers than undoped a-Si. Additional defects are induced in undoped material by strong illumination or by charge accumulation. The TFTs and photodiodes discussed next are designed specifically to take into account these properties. 23.3

a-Si TFT

The first a-Si TFTs were reported by Spear’s group at the University of Dundee [12]. Along with the semiconductor layer, a TFT requires a low leakage gate dielectric and preferably doped ohmic source and drain contacts. The discovery of doping in a-Si provided the ohmic contact material and the PECVD deposition of silicon nitride provided the gate dielectric, so that the different layers of the TFT can be made by the same deposition process. The TFT operates in the n-channel accumulation mode. n-channel operation is used because electrons have a much higher carrier mobility than holes, and the high defect density of doped a-Si:H means that the channel must be undoped and therefore operating in accumulation mode. 23.3.1

TFT Structure and Fabrication

Figure 23.6 shows the structure of the two most common alternative a-Si TFT designs. Both have a bottom-gate structure and the same silicon nitride dielectric

a-Si TFT

517

Island etch

metal n+ a-Si

passivation a-Si channel

metal n+ a-Si

Length, L SiN gate dielectric gate substrate

Back channel etch

metal n+ a-Si

a-Si channel

metal n+ a-Si

Length, L SiN gate dielectric gate substrate

Figure 23.6. Structure of the island etch (upper) and back-channel etch (lower) bottom-gate TFTs. The difference is in the process to deposit the source and drain contacts.

and a-Si channel layer, but differ in the way that the heavily doped (n+) contacts are fabricated. In the BCE process, the n+ layer and a metal contact are deposited directly on top of the a-Si channel. These two layers are then etched to form the source and drain contacts (see Fig. 23.6), after which a passivation layer is deposited. In the alternative island etch process, also shown in Figure 23.6, the passivation layer is deposited on top of the channel, and the source and drain contact overlaps the passivation. The BCE process is widely used for displays, largely because it eliminates one lithography mask step, which is important for the manufacturing cost, and also eases the critical alignment requirements, which is important for yield particularly on large backplanes. Figure 23.6 shows that the island etch channel comprises three sections, compared to only one for the BCE device. The BCE TFTs tend to have slightly poorer performance because the channel is partially etched through, which can result in defective material, while the structure of the island etch process protects the channel. The gate dielectric is typically 300 nm thick, giving a device that can easily operate up to 40 V. The a-Si channel layer can be as thin as ∼50 nm for the island etch process but is usually thicker for the BCE device because the process requires etching into the channel. The doped contacts are about 100 nm thick to give a good

518

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES

ohmic contact but are invariably coated with a metal electrode because the n+ layer by itself has a very high sheet resistance. Both processes result in the source and drain contacts overlapping the TFT channel, which is necessary to ensure low-resistance contacts, but provides some parasitic capacitance to the gate. The parasitic capacitance has a detrimental effect on the speed and performance of TFT arrays and on the noise of image sensor arrays. Self-aligned TFTs minimize the capacitance and have been demonstrated [13] but are not currently used in general display manufacturing. Partial selfalignment of the island etch TFT can be carried out using a lithography step to pattern the passivation layer, with exposure through the substrate using the gate as the mask.

23.3.2

TFT Characteristics

The TFT transfer characteristics, shown in Figure 23.7, contain three regions— above threshold, subthreshold, and the off-current. The TFT is turned on at positive gate voltages above the threshold voltage of typically 1–2 V; the leakage current when the TFT is turned off at negative gate voltages is usually below 1 pA and can be much lower. The subthreshold region in between describes the transition between the on and off states. In the subthreshold region, the current increases exponentially due to the movement of the Fermi energy through the localized states. The subthreshold slope is typically ∼0.5 V/decade.

Source-drain current (A)

10–6 10–8 W/L = 2 Vds = 5 V

10–10 10–12

above threshold

off 10–14 10–16 –10

subthreshold

0

10

20

Gate Voltage (V)

Figure 23.7. Example of the transfer characteristics for an island etch amorphous silicon TFT.

a-Si TFT

519

The TFT ON characteristics are described by the usual MOS transistor relations [14]. In the linear regime, at a small source–drain voltage, VDS, the current, IDS, is given by I DS = CG μ FE (VG − VT − VDS 2)VDSW L ,

(23.3)

where VG is the gate voltage, VT is the threshold voltage, W is the TFT width, and L is the length. A 300-nm nitride dielectric has a gate capacitance, CG, of 5 × 10−8 F/ cm2, and a field effect mobility, μFE, of 0.8 cm2/Vs yields a current of 6 μA in a device with W/L = 10 operating at 15 V above threshold. At a high VDS, the current saturates at I SAT = CG μ FE (VG − VT ) W 2 L . 2

(23.4)

The transfer characteristics of an a-Si TFT are shown in Figure 23.7, for the linear regime. The minimum leakage current is typically when VG is about −5 V, and the small off-current gives an on/off ratio of 106–108, which is sufficient for an LCD backplane or an image sensor. The n-type contacts are essential to provide a barrier to hole conduction at negative gate bias. The leakage current may be due to bulk thermal generation current in the channel or due to injection through the barrier formed by the n-type source and drain contacts. Generally, the leakage is lower in the island etch structure compared with the BCE structure, possibly because the island protects the channel from plasma and etching damage.

23.3.3

The Bias Stress Effect in a-Si TFTs

a-Si TFTs have a threshold voltage instability that is closely related to solar cell degradation [15]. Figure 23.8 illustrates the change in TFT transfer characteristics when the device has been turned on for a period of time. The threshold voltage increases with time, with a corresponding decrease in the current, but the mobility and other parameters are essentially unchanged. The change in VT depends on gate voltage and time according to the following description [16]: ΔVT = const. t α VGγ ;

α ∼ 0.33, γ ∼ 2 − 4.

(23.5)

Measurement of the VT shift is usually done to the linear regime largely because the main application of TFTs is active matrix backplanes, which operate in the linear regime. The VT shift is smaller in saturation [17] primarily because the channel charge is lower, and it is the channel charge that determines the VT shift. There are two mechanisms contributing to the VT shift in a-Si [18]. The gate dielectric, usually silicon nitride, contains traps into which electrons can tunnel

520

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES Ids (A) 10–4 10–5

Initial

10–6

After stress e.g. Vg = 20V for 2 hours

10–7 10–8 10–9 10–10

Vds = 20V

10–11 –5

0

5

10

15

20

Vqs (V)

Figure 23.8. Threshold voltage shift of an a-Si:H TFT as a result of extended gate bias stress at a positive gate voltage. The threshold voltage shift is due to defect generated in the channel.

from the channel. The second mechanism is charge creation in the a-Si layer, and it is this process that is related to defect creation in a solar cell. The energy associated with an electron occupying a conduction band state (in the solar cell, it is the energy of an electron–hole pair) induces a bond-breaking process and creates dangling bond defects. The process is catalyzed by the presence of hydrogen, but the exact reaction has never been made completely clear. The two bias stress mechanisms both follow Equation 23.5, although with different parameters, α and γ, and so are not very easy to distinguish without careful analysis. Tunneling into the gate dielectric tends to be a stronger function of gate voltage and hence predominates at a high gate bias. In a good quality a-Si TFT of the usual structure made for displays, defect creation typically dominates up to a gate voltage of about 20 V and tunneling into the dielectric at higher gate voltage. The threshold shift is not too much of a problem for image sensors and displays because the TFTs are turned on with a low duty cycle, and it is sufficient to design the circuit so that it can accommodate a VT shift of perhaps 2–4 V over the life of the device. However, the stress effect makes it difficult to use a-Si TFTs for analog devices and when the gate bias is applied continuously for a long time.

PHOTODIODES

23.4

521

PHOTODIODES

The a-Si:H p-i-n photodiode is used as the light detector in X-ray image sensor arrays because of its high quantum efficiency and low reverse-bias leakage current. Low leakage is important because the illumination level is very weak and high signal to noise is critical. The same p-i-n structure is used for a solar cell where the illumination is many orders of magnitude larger. The photodiode structure is illustrated in Figure 23.9 and usually consists of 100 nm–1 μm of undoped a-Si between much thinner p- and n-doped layers. The doped layers provide rectifying contacts but do not contribute to the light sensitivity because doping causes a high density of charged dangling bond defects in a-Si:H as discussed earlier. The minority carrier lifetime in doped a-Si:H is therefore so small that most photogenerated carriers recombine in the doped layers before they reach the i-layer. On the other hand, the high defect density also results in a narrow depletion width, so that a doped layer thickness of 10 nm or less is sufficient to form the junction. The barrier height for the p-i-n device is larger than for a metal Schottky diode [19] even with high work function metals such as Pd or Pt, and so makes a better photodiode with a lower reverse-bias leakage current. Since charge collection of holes is less efficient than for electrons (see Fig. 23.4), the photodiode operates best when the p-doped layer faces the illumination. Electron–hole pairs are preferentially created near the illuminated surface and so holes have a smaller distance to travel for collection. The p-i-n diode must be fabricated with metal contacts on both top and bottom because the thin doped layers are not sufficiently conductive by themselves. A 10-nm-thick p-type a-Si:H layer has a sheet resistance of ∼1011 Ω per square, and the more conductive n-type layer is ∼108 Ω per square. An ITO transparent conductor is generally used for the illuminated surface, and any convenient metal serves as the reflecting back contact. The photocurrent, IPC, is given by I PC = β ηQE G e,

(23.6)

illumination

ITO transparent contact p-doped layer ~10 nm undoped a-Si ~1 μm n-doped layer ~10 nm metal contact

Figure 23.9. Structure of the p-i-n photodiode showing the thin n+ and p+ contacts, the thicker undoped layer, the transparent top electrode, and bottom metal contact. Illumination is through the top surface.

522

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES

where G is the incident photon flux, ηQE, is the internal quantum efficiency, and β is a factor containing other loss mechanisms described below. Unlike the n-i-n diode, which can have photoconductive gain, the internal quantum efficiency of the reverse-bias p-i-n diode cannot exceed unity because there is no significant charge injection. ηQE approaches close to unity when the charge collection length, μτE, is much larger than the thickness, d, as described by Equation 23.1. This condition applies in a p-i-n diode with low-defect-density a-Si and a moderate applied field. A typical external quantum efficiency spectrum is shown in Figure 23.10 and is governed by several loss factors. The decrease at long wavelength is due to the rapidly decreasing optical absorption coefficient of a-Si:H below the bandgap energy. The short wavelength decrease mostly arises from absorption in the upper doped layer, along with reflection and absorption in the transparent metal conductor. The peak quantum efficiency of 80–95% occurs in the range 500–600 nm, which is suitable for the X-ray imager, because the usual X-ray converter phosphor emits at ∼550 nm. The peak quantum efficiency can be increased by using lowabsorption doped layers and by designing the transparent metal as an antireflection layer.

23.4.1

Comparison of p-i-n Solar Cells and Photodiodes

The a-Si solar cell and the photodiode are similar devices, both being p-i-n structures. However, there are design differences that reflect the different operating conditions of the two devices. The solar cell is made substantially thinner than the

1.0 Quantum efficiency

Solar cell 0.8 0.6

Photodiode

0.4 0.2 0 400

500

600

700

800

Wavelength (nm)

Figure 23.10. Typical spectral response of a photodiode showing sensitivity across the visible region. The maximum quantum efficiency is limited by reflection at the ITO contact, absorption in the p-layer, and incomplete charge collection in the bulk.

PHOTODIODES

523

photodiode (typically 100–300 nm vs. 1 μm), primarily to allow for the high density of light-induced defects induced by the illumination, which reduces the value of μτ. The solar cell operates with a smaller built-in potential because it is in slight forward bias, also making charge collection more difficult. Hence, both the solar cell and the photodiode need a high-quality i-layer to allow high charge collection, but optimization of the charge collection is difficult in the solar cell. High quantum efficiency is important for both devices but is more critical for the solar cell. Figure 23.10 indicates that the solar cell has been engineered to have a better spectral response than the typical photodiode by texturing the film to increase optical absorption, optimizing the contacts for low absorption, and often grading the i-layer bandgap and doping for increased charge collection. The p- and n-doped contact layers are typically either microcrystalline or a wide bandgap alloy in the solar cell to provide the low optical absorption while also minimizing the series resistance when a large current flows. The X-ray imager photodiode passes a much smaller current (picoampere vs. milliampere) so the highly conductive contacts are less critical, but optimizing the p/i junction for very low reverse leakage is more important, and this is one reason why the photodiode is typically about 1 μm thick. One additional source of leakage current in the photodiode is down the sidewall of the sensor, which therefore has to be well passivated, particularly for small sensors (∼100 × 100 μm) where edge effects can dominate. The effect is not significant in a solar cell because the surface area of the device is large.

23.4.2

Diode Characteristics

The forward and reverse characteristics of the p-i-n diode are illustrated in Figure 23.11. The forward current follows the expected exponential current–voltage relation I = I 0 exp ( eV nkT ) .

(23.7)

The ideality factor n is about 1.5. The ideality factor is closer to unity in an a-Si metal Schottky diode [19], reflecting a thermionic emission process. The larger junction barrier of the p-i-n diode reduces thermionic emission and causes the generation–recombination mechanism to dominate, which gives a larger ideality factor and is also consistent with the reverse current mechanism, as discussed below. In reverse bias, a depletion layer forms with width given by WD = [ 2εε oVBI eN DI ] , 12

(23.8)

where NI is the density of ionized gap states, assumed to be constant for convenience. When the deep states are broadly distributed across the middle of the bandgap, depletion only occurs down to the middle of the bandgap because deeper

524

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES 1 10–2

Sensor Current (A/cm2)

Dark Forward Bias 10–4 Light Reverse Bias

10–6 10–8 10–10

Dark Reverse Bias

10–12 0

1

2

3

Voltage (V)

Figure 23.11. a-Si p-i-n diode characteristics showing the exponential diode region in forward bias followed by a current limited by series resistance. The dark reverse current is limited by thermal generation.

states are filled from the valence band faster than the electrons are emitted to the conduction band. This mechanism of thermal generation creates electron–hole pairs, which are the origin of the reverse leakage current [20]. Since a-Si:H is weakly n-type, the depletion layer forms from the p-type contact by the emission of electrons. Measurements find an ionized state density for undoped a-Si:H of about 5 × 1014 cm−3, which gives a depletion layer of about 1 μm for a voltage of 1 V. Hence, for a 1-μm photodiode with a more typical reverse bias of ∼5 V, the depletion width is larger than the film thickness and the internal field is approximately uniform. The reverse leakage current in the photodiode is about ∼10−11 A/cm2 (see Fig. 23.11) and is thermally activated with an energy of about 0.9 eV, consistent with thermal generation from the middle of the bandgap. Figure 23.12 shows that the leakage current decays slowly after the application of a reverse-bias voltage, over a timescale of minutes at room temperature. The transient component arises from the depletion of charge from the deep states as the depletion layer is formed, and the steady-state current, JSS, is reached when the rate of emission of electrons and holes from deep traps is equal. The thermal generation mechanism gives a steadystate current density [20], I SS = NkT ω 0 exp [ − ΔE kT ] ξ ( F ) d ,

(23.9)

PHOTODIODES

525

Figure 23.12. Transient dark current of a p-i-n photodiode showing the slow decay to a steady state over a timescale of minutes [21].

where N is the density of states near midgap, ΔE is half the bandgap energy, ω0 is ∼1012 s−1, ξ(F) represents the weak dependence on the electric field, and d is the sample thickness. Both the transient current decay and the steady-state reverse-bias thermal generation current are directly related to the defect density and are consistent with other measurements of the defect distribution of a-Si:H. Contact injection and edge leakage are other contributions to the leakage current in some photodiodes [21]. Contact injection occurs when electrons tunnel across the p–i interface leading to an increase in the leakage current, particularly at high applied bias and in samples with a high defect density. The exposed periphery of the photodiode provides a different source of leakage current, which is particularly significant in small devices where the ratio of edge to area is large. In the presence of both bulk and edge components, the total current in a square diode of edge length dS, is given by I total = I B dS2 + 4 I P dS ,

(23.10)

where IB cm−2 and IP cm−1 are the bulk and peripheral conductance. There is considerable variability in the leakage current when the bias is large, due to these mechanisms.

526

23.5

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES

TFTs AND PHOTODIODES IN SENSOR ARRAYS

The important attributes for sensors and TFTs in an X-ray detector are radiation hardness, response linearity, low image lag, and low electronic noise. a-Si turns out to be a radiation hard material, and exposures of up to 50 Mrad of gammas do not seriously degrade the photodiode or TFT performance [22, 23]. Part of the explanation is that the devices are thin and silicon has low atomic number, and so the interaction with high energy radiation is weak. Radiation does increase the defect density, but the effects are slow to appear in view of the significant intrinsic defect density in the material. Response linearity and image lag in the photodiode are closely related. An ideal p-i-n diode has unity gain and hence a perfectly linear response. However, incomplete charge transfer of the generated electrons and holes reduces the gain and can cause a departure from linearity. The trapped charge is typically released at a later time when it can create ghosts in subsequent images, the effect known as image lag. Charge trapping is governed by the mobility-lifetime product according to Equation 23.1, and when trapping is weak, the fraction of charge trapped is approximately FT = d 2 2μτV ,

(23.11)

assuming a uniform electric field. Trapping is therefore minimized in thin samples with a high bias voltage and large values of μτ. In a good quality a-Si:H photodiode of 1 μm thick with 5-V bias, 1–3% of the electrons are trapped and 5–10% holes, because of their smaller value of μτ. As noted above, charge trapping, and therefore image lag, is lowest when the sensor has light incident on the p-doped layer because holes have a shorter distance to travel to the contact. The nonlinearity in the response of an X-ray detector comes about from the way the detector operates. The bias voltage across the sensor decreases as the light flux and exposure time increase because the sensor charge is stored on the capacitance of the diode until it is read out. From Equation 23.11, the charge trapping increases as the sensor approaches voltage saturation, and hence the amount of charge trapping increases with the illumination. The collected signal charge when there is an illumination intensity, G, can be expressed as a function of voltage and time: t

QS (V , t ) = ∫ [1 − FT (V )] G dt .

(23.12)

0

Figure 23.13 shows the response of an a-Si X-ray detector under radiographic (single shot) and fluoroscopic (continuous measurement) mode [24]. The difference in the mode response is that fluoroscopic measurements are not subject to image lag because the charge trapping is exactly offset by charge release from previous frames. The radiographic mode data are subject to charge trapping and show increasing nonlinearity of response as saturation is approached. A fit to

NEW DEVELOPMENTS IN a-Si DEVICES

527

5 Fluoroscopic mode

QPIX (pC)

4

VBIAS = - 6V

Radiographic mode

3 VBIAS = - 4V

2 VBIAS = - 2V

1 0

0

200 400 600 800 1000 Incident Light Signal (# LED flashes)

Figure 23.13. The measured response of an amorphous silicon X-ray detector in radiographic and fluoroscopic modes and at different bias voltages [24].

Equation 23.12 gives a value of FT of 2–4% at 5-V bias, corresponding to an average value of μτ of ∼10−7 cm2/V. Several sources of electronic noise occur in an X-ray detector and from the associated readout electronics. Some of the noise sources are the materials and devices of the backplane, and some are external to the array. From the point of view of the pixel circuit, the most important are the resistance noise in the pixel TFT and shot noise in the photodiode current. The current noise is generally negligible in sensors at low bias voltage because of the very low leakage currents. Thermal noise of the TFT ON resistance generates the familiar kTC noise component, which is proportional to the pixel capacitance. Additional noise arises from 1/f noise in the TFT, the magnitude of which is sensitive to the device material and fabrication.

23.6

NEW DEVELOPMENTS IN a-Si DEVICES

23.6.1

TFTs and Sensors on Flexible Substrates

The ability to make a-Si TFTs and image sensors on flexible substrates is a recent development [25, 26], but a-Si solar cells on flexible substrates have been available for several years [27, 28]. The possible advantages for an X-ray detector include a more robust device, since breakage of the glass substrates is a significant practical issue. A curved sensor might be interesting for dental X-ray or for computed tomography detectors. Also, the fabrication of a sensor array on a thin X-ray transparent substrate opens the possibility to image X-rays that are incident from

528

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES

Figure 23.14. Photograph of an amorphous silicon TFT backplane fabricated on a plastic substrate.

the substrate side of the detector, which enhances the spatial resolution when used with a phosphor converter in the indirect detection mode. Either plastics or metal foils are potential flexible substrates for TFT backplanes. Plastics have the advantages of being lighter, of being insulating rather than conducting, and of generally having a better surface quality. The challenge is that many of the most suitable plastics cannot be heated above 200°C, while the standard a-Si TFT deposition occurs at up to 350°C, and p-i-n sensors up to 250°C. The effort to make low-temperature a-Si devices compatible with plastic substrates has been quite successful [25, 29]. The TFT characteristics are almost unchanged compared to the high-temperature growth. The low-temperature deposition does increase the defect density, but this effect is minimized by hydrogen dilution of the deposition gas and other changes in the deposition conditions. The increased defect density causes a small (∼1 V) increase in threshold voltage, and about a 20% increase in the bias stress threshold voltage shift. Low-temperature p-i-n photodiodes also have similar characteristics to the high-temperature devices. The increased defect density affects both the charge collection and the reverse-bias leakage current. Since both of these parameters are critical to the performance of an X-ray imager, further optimization of the growth and device configuration is probably still needed. Figure 23.14 shows an image of an a-Si TFT backplane fabricated on PEN plastic [26]. 23.6.2

Digital Lithography Processing of TFTs and Sensor Arrays

Much of the expense in the fabrication a-Si TFT detector arrays is in the photolithography processes that are used to pattern the device. Photolithography involves

NEW DEVELOPMENTS IN a-Si DEVICES

529

several steps, which are repeated for each of the masking layers that make up the device. Very large detectors with relatively low pixel resolution are of interest for some security applications—for example, screening trucks—but fabrication by conventional photolithography may make these detectors prohibitively expensive. An alternative approach to photolithography is to use printing techniques. There are several ways in which the various document printing technologies (jetprinting, offset, gravure, etc.) can be applied to the fabrication of TFTs, and one that has been successfully demonstrated is to jet-print the lithographic etch mask [30]. This approach reduces the number of steps needed in the lithography process but allows the patterning of all the conventional materials used to make an a-Si TFT backplane. Figure 23.15 illustrates the process steps used to print the etch masks and to pattern the material. The masking process gives well-defined features, although at present, the feature size is typically 40–50 microns, which is too large for conventional displays but might be suitable for large and relatively low-resolution detectors. One of the printed resist materials is a wax that has the advantage of being jetted as a hot liquid but freezing on contact with the substrate. The spread of the wax on the substrate can therefore be controlled by the substrate temperature and is largely independent of the surface energy. Wax adheres well to most surfaces and is resistant to chemical etchants.

1. Deposit film

2. Print mask

3. Etch film; strip mask

Figure 23.15. Schematic illustration of the steps in the digital lithography process that uses printed etch masks to pattern the layers of the thin film transistor backplane.

530

23.7

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES

SUMMARY

a-Si devices form the basis of three significant technologies: displays, solar cells, and X-ray image sensors. These applications share a requirement of large area fabrication, which is enabled by the PECVD method of depositing a-Si, along with the various doped layers, alloys, and dielectrics. The properties of a-Si follow from its noncrystalline atomic structure and from the small amount of hydrogen that is incorporated during deposition. The magnitude of the mobility, the defect and doping properties, and the stability of the devices can all be related back to these two factors. Once the material properties were understood, remarkably good devices could be made, which suit the applications well. Displays need the high dynamic range and low threshold voltage of the TFT. Solar cells benefit from the ability to engineer the bandgap by alloying, and the tunnel junction that is formed at the p–n interface, but suffer significantly from the instability induced by the hydrogen. Photodiodes have high gain and high dynamic range. ACKNOWLEDGMENTS Many scientists, both past and present, have contributed to the development of a-Si technology at the Palo Alto Research Center, supported by funding from PARC, Xerox, and several government agencies. The contributions of these people and organizations are gratefully appreciated. It has also been a pleasure to work with several external collaborators in the basic understanding and the various applications of a-Si. ABBREVIATIONS a-Si—amorphous silicon B—billion BCE—back-channel etch C—capacitance CG —gate capacitance of a TFT d—semiconductor film thickness dS —edge length of a square diode e—electronic charge E—electric field strength f—frequency FT —fraction of charge trapped in a-Si bandgap states G—photon illumination intensity H—hydrogen i—undoped or intrinsic semiconductor I—current IB —bulk current conductance per square centimeter IDS —source–drain current of a TFT

REFERENCES

531

III—elements in the third column of the periodic table IP—peripheral current conductance per centimeter IPC —photocurrent ISAT —saturation current of a TFT ISS —diode reverse-bias saturation current ITO—indium tin oxide k—Boltzmann’s factor L—length of a TFT LCD—liquid crystal display Mrad—mega (million) rads where rad is a unit of x-radiation dose of ionization energy MOS—metal oxide semiconductor transistor configuration n—negatively doped semiconductor n—diode ideality factor N—density of trapping states near midgap ND—light-generated defect density in the bandgap NI —density of ionized gap states in depletion region of a diode p—positively doped semiconductor PECVD—plasma-enhanced chemical vapor deposition PEN—polyethylene naphthalate Q—charge magnitude Si—silicon TFT—thin-film transistor t—time T—absolute temperature V—elements in the fifth column number of the periodic table VBI—built-in barrier height voltage of semiconductor diode VDS —source–drain voltage of a TFT VG—gate voltage of a TFT VT —threshold gate voltage of a TFT W—width of a TFT WD—width of a diode depletion layer β—photodiode loss mechanisms factor ΔE—energy of a-Si trap from the band edge ε—relative dielectric constant of a semiconductor εo—dielectric constant of free space ηQE—internal quantum efficiency of a photodiode τ—electron or hole lifetime in a semiconductor μ—electron or hole mobility in a semiconductor μFE—field effect mobility ω0—attempt to escape frequency for transitions from bandgap trap states REFERENCES [1]

N. F. Mott and E. A. Davis. Electronic Processes in Non-Crystalline Solids. Oxford, Clarendon Press (1979).

532 [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]

[15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30]

AMORPHOUS SILICON TRANSISTORS AND PHOTODIODES D. E. Carlson and C. R. Wronski. Appl. Phys. Lett. 28, 671 (1976). R. A. Street. Large area image sensor arrays. In Technology and Applications of Amorphous Silicon, R. A. Street, ed., pp. 147–221. Berlin, Springer-Verlag (2000). J. C. Knights and G. Lucovsky. CRC Crit. Rev. Solid State Mater. Sci. 21, 211 (1981). R. A. Street. Phil. Mag. B63, 1343 (1991). W. E. Spear and P. G. LeComber. Solid State Comm. 17, 1193 (1975). K. Hecht. Z. Phys. 77, 235 (1932). R. A. Street, J. Zesch, and M. J. Thompson. Appl. Phys. Lett. 43, 672 (1983). D. L. Staebler and C. R. Wronski. Appl. Phys. Lett. 31, 292 (1977). H. R. Park, J. Z. Liu, and S. Wagner. Appl. Phys. Lett. 55, 2658 (1989). M. Stutzmann, W. E. Jackson, and C. C. Tsai. Phys. Rev. B32, 23 (1985). A. J. Snell, K. D. Mackenzie, W. E. Spear, P. G. LeComber, and J. A. Hughes. Appl. Phys. 24, 357 (1981). Y. Kuo. a-Si TFT structures. In Thin Film Transistors; Materials and Processes, Volume 1, Chapter 4, Y. Kuo, ed. Dordrecht, Kluwer (2004). A. Nathan, P. Servati, K. S. Karim, D. Striakhilev, and A. Sazonov. Device physics compact modeling and circuit applications of a-Si TFTs. In Thin Film Transistors; Materials and Processes, Volume 1, Chapter 3, Y Kuo, ed. Dordrecht, Kluwer (2004). M. J. Powell. IEEE Trans. Electron Devices 36(12), 2753 (1989). T. Tsukada. Active matrix liquid crystal displays. In Technology and Applications of Amorphous Silicon, R. A. Street, ed., pp. 7–93. Berlin, Springer-Verlag (2000). K. S. Karim, A. Nathan, M. Hack, and W. I. Milne. IEEE Electron Device Lett. 24(9), 583–585 (2004). C. van Berkel and M. J. Powell. Appl. Phys. Lett. 51(14), 1094–1096 (1987). M. J. Thompson, N. M. Johnson, R. J. Nemanich, and C. C. Tsai. Appl. Phys. Lett. 39, 274 (1981). R. A. Street. Appl. Phys. Lett. 57, 1334 (1990). R. A. Street. Hydrogenated Amorphous Silicon. Cambridge, Cambridge University Press (1991). I. D. French, A. J. Snell, P. G. LeComber, and J. H. Stephan. Appl. Phys. A31, 19–22 (1983). J. M. Boudry and L. E. Antonuk. Med. Phys. 23, 743 (1996). L. E. Antonuk, Y. El-Mohri, J. H. Siewerdsen, J. Yorkston, W. Huang, V. E. Scarpine, and R. A. Street. Med. Phys. 24, 51 (1997). H. Gleskova and S. Wagner. IEEE Trans. Electron Devices 48, 1667 (2001). T. N. Ng, R. A. Lujan, S. Sambandan, R. A. Street, S. Limb, and W. S. Wong. Appl. Phys. Lett. 91, 063505 (2007). S. Guha. Multijunction solar cells and modules. In Technology and Applications of Amorphous Silicon, R. A. Street, ed., pp. 252–305. Berlin, Springer-Verlag (2000). F. R. Jeffery, D. P. Grimmer, S. Brayman, B. Scandrett, and M. Noak. PVMaT improvements in monolithic a-Si modules on continuous polymer substrates. AIP Conf. Proc. 394, 451 (1996). W. S. Wong, R. Lujan, J. H. Daniel, and S. J. Limb. J. Non-Cryst. Solids 352, 1981 (2006). W. S. Wong, S. E. Ready, J. P. Lu, and R. A. Street. Electron Device Lett. 24, 577 (2003).

24 AMORPHOUS SILICON DIGITAL X-RAY IMAGING RICHARD COLBETH Varian Medical Systems

24.1

WHY a-Si?

a-Si X-ray detectors for medical, dental, veterinary, and nondestructive test applications have seen explosive growth, with the annual number of units nearly doubling in each of the last 5 years. This growth has resulted from two decades of research and refinement by numerous academic and industrial organizations. This chapter explores the design issues unique to large-area X-ray sensors and answers the question, why a-Si? From the beginning, a-Si technology has had a number of compelling advantages that still elude competing technologies. First and foremost, a-Si is the technology of large-area electronics, from solar cells to flat panel displays to X-ray detectors. The lack of an effective lens for X-rays requires an X-ray detector to be at least the size of the object of interest. Because of image magnification, shown schematically in Figure 24.1, it is often required that the detector capture an image projection that is double the object size in medical procedures. Today, it is common to find a-Si X-ray detectors for chest radiography that have a 43 × 43 cm (17 × 17 in.) active area. When research into a-Si X-ray imagers began in the late 1980s, work on a-Si for solar cells and displays had been in progress for decades. The idea then, as now, was to build on the huge investments in flat panel display production capacity, fabrication tools, and basic research. Because of the large and growing flat panel industry, a-Si detectors are cost-effective and are becoming more so with the increasing annual volume of both detectors and displays. a-Si devices offer unique advantages not readily available in other X-ray detectors. Because there is very little long-range order, a-Si is not easily damaged Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

533

534

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

Figure 24.1. X-ray imaging geometry. SAD is the source-to-object distance. SID is the source-to-detector distance.

by radiation. This inherent radiation hardness, along with the lack of isolating field oxides and wells, makes a-Si TFTs and photodiodes resistant to more than 10,000 Gy (>1 Mrad). Because of this high radiation tolerance, one of the first applications for a-Si FPDs was in portal imaging. A portal imager (also called an EPID) is used in radiotherapy applications to verify delivery of the mega electron volt gamma radiation used to treat cancer. In this case, the EPID is positioned behind the patient, in the treatment position, to verify the relation between deposited treatment dose and the patient anatomy. The image quality of a-Si EPIDs dramatically improved the image quality available in this application, providing the information in real time. Today, nearly all medical linear accelerators are sold with a-Si imaging capability as shown in Figure 1.6 of Chapter 1 and in Figure 22.6 of Chapter 22. a-Si also has a relatively large bandgap, ∼1.8 eV, compared to 1.1 eV for crystalline Si. Hence, the photodiodes have extremely low dark current, facilitating operating temperatures up to 50°C. Unlike crystalline silicon CCDs and CMOS sensors, which often require active cooling, it is possible to build a-Si detectors that can be used over a wide ambient temperature range with minimal cooling. Furthermore, the devices are stable over large temperature and humidity variations, such that FPDs can easily survive standard storage and transportation conditions (−20°C to +70°C, 90% RH) with no special packaging considerations. The inherent stability of these devices assures a long lifetime. Perhaps most importantly, a-Si detector arrays are sensitive to visible light similar in wavelength to the light produced by X-ray scintillator screens with a long history in X-ray radiography and real-time imaging. These screens can be put in direct contact with the a-Si array, maximizing the coupling efficiency while

DETECTOR CONSTRUCTION

535

eliminating sources of image distortion. With the advent of a-Si detector technology, X-ray imaging was able to take an immediate leap forward in DQE, resolution, small area contrast and uniformity.

24.2

DETECTOR CONSTRUCTION

A picture of a 14 × 17″ portable radiography detector is shown in Figure 24.2 (Varian Medical Systems 4336R). This particular FPD has the same footprint and form factor as the traditional 14 × 17″ X-ray film cassette, and so provides a dropin upgrade to digital. The detector front cover is made of a low X-ray attenuation carbon fiber. X-rays exiting the patient are captured and converted to digital data, which are then transmitted over a high-speed link to a workstation. In this particular case, the imager link is Gigabit Ethernet, but fiber optics, LVDS, and Cameralink connections are also common. In addition to transmitting the digital image data, the Gigabit Ethernet link also delivers all the control data to the receptor. Figure 24.3 shows the internal construction typical of the indirect FPDs. Here, indirect refers to the fact that X-rays deposit energy in a scintillator screen generating visible photons, which are in turn captured and converted to electron– hole pairs in the photodiode array. In contrast, a direct FPD converts the X-ray photon energy directly to electron–hole pairs. The relative advantages of these approaches will be discussed in a later section and in Chapter 25. Typically, the

Figure 24.2. Radiographic flat panel detector (Paxscan 4336R). The active area is approximately 43 × 36 cm with 139-μm pixel pitch. The form factor matches the standard X-ray film cassette, including the thickness of 15.5 mm.

536

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

Figure 24.3. Typical internal architecture for the indirect-type FPD.

scintillator screens are made from GOS or CsI(Tl). With minor modification, both GOS and CsI were adapted from the technologies that FPDs were targeted to replace. GOS “intensifying” screens have been in use with X-ray film for a very long time, and CsI is a key component in IITs, which are the front-end capture device in traditional X-ray video imaging. The core of the detector is an a-Si photodiode array. Like the TFT arrays used in flat panel displays, the photodiode array is fabricated on a glass substrate using standard semiconductor processing. A rigid baseplate is used to support the glass and to mount the rest of the electronics. Figure 24.4 shows a Varian Medical Systems 4030CB with the back cover removed, revealing the boards and the TAB chips at the periphery. Since the mobility of electrons in an a-Si is on the order of 1 cm2/Vs, it is difficult to build quality, high-speed, low-noise amplifiers using the a-Si itself. As a consequence, the drive and charge-to-voltage amplifiers are generally external to the photodiode array. All the rows and columns of the array are brought to landing pads at the periphery of the glass in groups (e.g., 128), where they are connected to TAB packaged chips. TAB bonding technology is used extensively in flat panel displays. Because the pitch of the connections coming off the array is comparable to the pixel pitch, there is a mismatch between the connection density achievable on the array (∼100 μm) and what is readily available on the printed circuit boards (∼300 μm). To resolve this, the first level of drive and readout electronics is mounted in flexible TAB packages having glass side connections for every row (or column) and board side connections at a lower density, but sufficient to bring in power, control, and serial data signals. On the glass side, the TAB package is bonded to the TAB landing patterns by means of an ACF using a combination of heat, pressure, and ultrasonic energy. On the board side of the TAB connection, ACF, solder, or connectors are used. Figure 24.4’s inset shows a close-up view of the TAB connected readout and gate driver chips. Notice that the readout chips are connected to the board with solder, while the gate drivers are connected with ACF.

THE p-i-n/TFT PIXEL

537

Figure 24.4. Electronics of a Paxscan 4030CB real-time detector for radiography, fluoroscopy, and vascular applications. This FPD has split data lines, with readout electronics on two sides of the panel. The close-up view of the TAB bonded readout and gate driver chips are shown on the lower right.

24.3

THE p-i-n/TFT PIXEL

While there are a number of pixel architectures in use, the p-i-n/TFT architecture is found in the largest number of products and in the most varied applications. This architecture, first proposed for X-ray sensors by Antonuk and Steet in 1989 [1], is shown in Figure 24.5. Each pixel consists of just two elements, a TFT and a p-i-n photodiode. A picture of the pixel is shown in Figure 24.6. The light-sensitive p-i-n photodiode occupies the bulk of the pixel real estate. Its transparent ITO top electrode allows visible light to penetrate the diode. In addition, the top-side p-layer is made thin enough to allow most of the visible photons to be absorbed in the thick intrinsic layer of the diode. The typical quantum efficiency is shown in Figure 23.10 of Chapter 23. The peak absorption is in the green at 550 nm, which is well matched to the light output from GOS and CsI(Tl) screens. There are a few things to notice in the pixel photograph. First is that the schematic of the pixel is easily recognizable. Second is that the TFT switch and p-i-n photodiode are coplanar. The ratio of light-sensitive pixel area (p-i-n area) to the total pixel area is known as the FF. A finite amount of the pixel area must be devoted to the TFT switch, so as the pixel size shrinks beyond a certain size, the FF drops off rapidly. Broadly speaking, this type of pixel architecture is used for pixel sizes down to 100-μm pitch. Below 100 μm, there are alternative architectures that overcome the FF limitation. The third takeaway from this picture is that the p-i-n photodiodes are mesa isolated. This is quite different than the situation in crystalline CCD or CMOS detectors. The mesa isolation assures that there

538

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

Figure 24.5. Organization of the a-Si array and connected electronics.

Figure 24.6. Photograph of a p-i-n/TFT pixel.

is little cross talk or blooming of signal from one pixel to the next, even in situations where the light exposure is very high. The effect of saturating the pixel is to forward bias the photodiode, in which case the excess current is shunted on to the array bias line. For a moment, consider the signal chain. X-rays are absorbed by the scintillator producing light proportional to deposited energy. For CsI(Tl), it is typical that one visible photon will be generated for each 26 eV of energy. In diagnostic radiography, the range of energy used is roughly 10–150 keV. A single photon of 70 keV will produce approximately 2700 visible photons. This light is emitted in all directions, only half of it downward toward the photodiode array. However, it is common to use an efficient light reflector at the X-ray entrance side of the scintillator. So in the best case, all the visible photons are directed toward the array. Assuming the light is at 550 nm, 80% of these photons will be absorbed by the

THE p-i-n/TFT PIXEL

539

photodiode. For the 127-μm pixel shown, the FF is 57%. Therefore, a single 70keV X-ray photon will generate 2700 × 0.80 × 0.57 = 1230 electrons. This gain in signal from X-rays to electrons resulting from the intimate contact between scintillator and photodiode is the reason that FPDs have excellent quantum efficiency. In contrast, an X-ray detector based on CCDs may use the same CsI scintillator but collect the light through a mirror and lens system focusing the visible photons onto the CCD, which is preferably located outside of the X-ray beam. Because of the limited solid angle over which the lens collects photons, there is a quantum sink [2] or severe gain loss at this point in the signal chain. The result is that CCD-based detectors have comparatively poor quantum efficiency and also suffer from veiling glare and cross talk due to scattered light. At the pixel level, the readout timing is very straightforward. During integration, the TFT switch is open (OFF). Light absorbed by the p-i-n photodiode creates electron–hole pairs in the intrinsic layer. The internal field of the photodiode under reverse bias separates the electrons and holes forcing them to opposite electrodes, where the charge is stored on the capacitance of the p-i-n diode. Charge stored on the pixel capacitance causes the voltage on the floating node (node a in Fig. 24.7)

Figure 24.7. Equivalent circuit and timing of the front-end electronics connected to the a-Si array.

540

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

to move negatively in the direction reducing the bias on the photodiode. During readout, the TFT switch is closed (ON), connecting node a of the pixel to the data line, which is held at a virtual bias by a charge integrating amplifier. Node a is then discharged onto the data line and into the feedback capacitance of the charge integrating preamplifier. Figure 24.7 shows schematically the front-end circuitry of the readout chip, its relation to a single pixel, and the timing needed to capture the signal from one pixel. The charge integrating preamplifier is reset, and then the TFT gate is pulsed ON to discharge the pixel. The signal charge is accumulated on the feedback capacitance of the preamp, converting the signal to a voltage. Then the voltage is captured by an S&H. Chapter 23 discusses in detail the electronic properties of the TFT and p-i-n photodiode. Here we consider the key parameters in the design of the sensor, which are the following: Parameter

Significance

TFT ON/OFF resistance

Pixel discharge time/signal loss rate

p-i-n photodiode capacitance

Signal capacity Pixel discharge time kTC noise of the pixel

p-i-n diode leakage

Limits length of the integration time and dynamic range Source of shot noise

Parasitic capacitance on the data line (column)

Increases the noise gain for pixel noise and correlated line noise

Resistance of the data line

Source of pixel noise

Deep-level traps

Image lag Gain effect/ghosting

The ON resistance of the TFT in combination with the pixel capacitance sets the discharge time constant, τ. TFT τ = RON × Cpix

A typical value for τ is 3 μs. Ideally, the TFT access time would be on the order of seven time constants to assure complete discharge of the pixel. In practice, it is possible to use just a few time constants, although the shorter discharge time results in a gain loss and in an increase in the image lag due to the charge left behind on the pixel. The OFF resistance of the TFT determines how much charge will be lost from each pixel during one integration time, which can vary from 33 ms to several seconds. Also note that all the pixels hanging on a given data line will be leaking charge to the amplifier during the time in which the pixel of interest is discharged, which typically varies from 10 to 100 μs. This is a source of nonlinearity since the

ELECTRONICS ARCHITECTURE

541

amount of leakage charge will depend on the signal levels of the other pixels on the column. The problem is exaggerated by the fact that signal levels in the darkest areas of the image behind the patient can be 1000 times smaller than pixels exposed to the unattenuated beam. The data line capacitance and, to a lesser extent, the data line resistance are key parameters in the detector noise, which is discussed in more detail below. One other key photodiode parameter is the amount of deep-level traps. Traps create image lag and gain variations. A detailed description of traps in a-Si devices can be found in Chapter 23, and their effect on detector performance is explored further in a later section of this chapter.

24.4

ELECTRONICS ARCHITECTURE

The basic electronics architecture for an a-Si detector is shown in Figure 24.8. Groups of rows, typically 128 or 256, are connected to gate driver chips. The gate driver chip is a shift register with high-voltage outputs. A bit is input to the shift register and clocked at the line frequency down through one gate driver to the next until all the rows in the panel have been selected. The individual outputs of the gate driver swing between the TFT OFF voltage when deselected (∼–5 V) to the TFT ON voltage when selected (∼+15 V). On the column side of the array, the electronics is more extensive. ASICs are connected to groups of columns, typically groups of 120 or 128. In this chapter, the terms ASIC and readout chip are used interchangeably. Each column is

Pixel

128 channels

14 bit ADC

ASIC

ADC output

FPGA

Gain selection: 1,2,4,5

Off-Panel Transmission

128 channels

14 bit ADC

ASIC Gain selection: 1,2,4,5

a-Si array 16 x

Figure 24.8. Architecture of panel readout electronics.

542

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

connected to a signal acquisition channel inside the ASIC, consisting of a charge integrating preamplifier, an S&H, and an output multiplexor. The 128 voltage signals are serialized by the multiplexor into a single data stream. The data stream out of the ASIC is a series of voltage levels, output at the pixel rate, which are fed into a voltage amplifier, which in turn feeds an ADC chip. The purpose of the voltage amplification stage is primarily to match the output signal of the ASIC to the input requirement of the ADC. The output of the ADC is a digital word, typically 14 or 16 bits wide. The digital data from all the ASIC/ADC channels are connected via a data bus to an FPGA, which drives the data through the off panel transmission medium. With this background, we can visualize how a panel is read out or scanned. Each row is selected in sequence. When a row is selected, all the pixels in that row are connected to the columns (data lines) of the array. The charge signals from each pixel are converted to a voltage and are captured on the S&H capacitances, in parallel, at one moment in time. Once the voltage signals from each pixel in the row are stored on the thousands of S&H circuits, the signals are then multiplexed out to the ADCs and are converted to digital signals for transmission on the digital data bus. Very often it is possible to “pipeline” the capture process, which means that the signal acquisition at the front-end preamp can take place during the multiplexing and transmission process. A pipelined architecture can significantly speed up the frame rate by decreasing the time needed to acquire, process, and transmit each row of data. The readout method described above is called progressive scanning. Each row is read in sequence with the data from each pixel in the row coming out in sequence. One variation on this method is to use split data lines, so that the top and bottom half of the panel can be read out in parallel. In this case, data from the top half of the panel are output directly, while data from the bottom half of the panel are stored in memory until the second half of the frame time. This architecture requires double the electronics but has the advantage of faster frame rates and larger X-ray windows. Most FPDs are used in pulsed applications where the X-ray beam is pulsed ON for a fixed duration. This is very common in fluoroscopy (video X-ray) applications. And radiography is by nature a pulsed application since the objective is to acquire a single snapshot image. In pulsed applications, if the beam pulse arrives on the panel during the readout sequence, the result is a signaldependent offset shift due to capacitive coupling of the generated photocurrent signal into the preamp. In order to avoid this nonlinearity, it is best to deliver the X-ray beam pulse during a non-scanning period of the frame time, referred to as the X-ray window. With split data lines, the panel scanning is confined to half of the frame time, opening up an X-ray window at least half the frame period. The relationships between the detector readout, the X-ray beam pulse, and the digital video data output are shown in Figure 24.9. Another common variation on the readout sequence is to create super-pixels though binning. Here, multiple rows are selected simultaneously, while at the same time the signals from multiple columns are combined, either in the ASIC or digitally. For example, sending two shift-in bits to the gate driver chips and combining

ELECTRONICS ARCHITECTURE

543

Figure 24.9. Relative timing of the panel scanning (readout), X-ray beam exposure, and digital video data output stream for a progressively scanned panel (top) and a split data line panel (bottom).

the signal from every two columns results in 2 × 2 binned pixels. Using binning, the sensitivity is increased due to the larger effective pixel size, and the frame rate is increased by reducing the total number of pixels in the image matrix. The tradeoff is loss in resolution and depending on the electronics, a reduction in signal range. In fact, the signal range moves down, with the panel becoming more effective at low doses but also saturating at a lower dose. Figure 24.10 shows the SNR versus dose for the same panel in full resolution and 2 × 2 binned modes. The solid lines show the calculated X-ray quantum-limited SNR, that is, the SNR expected based on the X-ray statistics alone. Generally, FPDs are X-ray quantum limited. Notice, however, that the SNR drops off the quantum-limited curve at low doses due to the increasing influence of the panel’s electronic noise. The electronic noise effectively sets the useful lower dose limit of the detector, and so it is of critical importance in the FPD design. The sources of electronic noise are discussed extensively in the next section.

544

AMORPHOUS SILICON DIGITAL X-RAY IMAGING 1000 1×1 Binning 2×2 Binning

SNR

100

10

1 1

10

100

1000

10000

Dose (nGy/frame)

Figure 24.10. Signal-to-noise ratio versus dose for a full resolution mode (1 × 1 binning) and a high frame rate fluoroscopy mode (2 × 2 binning). The solid lines show the estimated X-ray quantum-limited SNR for the two modes.

24.5

NOISE

24.5.1

Noise from the Absorbed X-Ray Photons

In a well designed FPD, the largest noise source, over most of the dose range, is the noise of the X-ray beam itself. Conversion of the X-ray photons is described by Poisson statistics, and so both the magnitude of the noise in the X-ray beam and the SNR are equal to the square root of the total number of photons absorbed: σ X-ray = SX-ray , SNR X-ray =

SX-ray = SX-ray . σ X-ray

In addition to the inherent noise in the X-ray beam, there is noise in the X-ray conversion process (the Swank effect) and variations in the generated signal due to uncertainty in depth at which the X-ray photon’s energy is deposited (Lubbert’s effect). These are particular to the choice of scintillator and, as such, are ignored in this discussion. For the remainder of this section, the focus is on the electronic noise and its impact on the lower part of the dose range, where σelectronics is comparable to σX-ray. 24.5.2

Electronic Noise

Generally, the electronic noise sources are uncorrelated, and so the total noise is the square root of the sum of the squares of each noise source:

NOISE

545 2 σ total = σ X-ray + σ a2 + σ b2 + σ c2 + ….

The upshot is that the largest noise source tends to dominate the total noise picture, and the designer can focus on the worst offenders. However, there are two complications. First, determining the largest noise source depends on the readout mode of the panel. If the charge-to-voltage gain at the front end of the electronics is very high, then sources of noise from the preamp back through the pixel dominate. This is the situation in low-dose fluoroscopy. If the charge-to-voltage gain is low, then noise sources past the preamp, including the S&H, multiplexor, voltage gain amplifier, and the ADC, tend to be most important. Low charge-to-voltage gain settings are typically used in radiography, cine (video recording mode), and cone beam CT. The second complication is that FPDs are prone to a type of correlated line noise, which results in almost no loss of SNR but is visually very objectionable.

24.5.3

Local versus Global Noise Sources

In Figure 24.11, the front-end electronics equivalent circuit has been redrawn to emphasize the sources of noise during signal acquisition. The noise sources can be organized into two categories: local and global sources. Local noise sources are

a-Si array

readout chip

reset

TFT_gate

reset

v gate

a

pin

C fb i2

TFT

C pix

vbias

sample

C overlap

R data

C data1

vin C data 2

sample

_

Cin +

Preamp

Vref C A = data C fb

BWPreamp ≈ 1MHz

BWS&H ≈ zHk052

Figure 24.11. Redrawn front-end electronics equivalent circuit to emphasize the sources of noise during signal acquisition.

546

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

those that contribute to the noise associated with an individual pixel, that is, pixel noise, and these sources are indeed local to the data line, which the pixel is connected to. Global noise sources are those that affect all the pixels in the array. This is a convenient way to segregate the noise contributors because local and global noise sources affect the panel performance differently. The local noise sources contribute directly to pixel noise, resulting in SNR loss and in an increase in the X-ray quantum-limited dose, defined as the dose at which the statistical noise of the absorbed X-ray beam equals the electronic noise contribution. Global sources of noise, on the other hand, contribute to correlated line noise. Table 24.1 lists the major electronic noise sources, roughly in decreasing order of importance. These are groupings of noise contributors that are captured at a particular place in the signal conversion chain. In the rightmost column is the noise gain, which is the multiplier describing the base noise source’s overall impact on the panel SNR.

TABLE 24.1. Important Noise Sources in FPDs Noise Source

Symbol

Local/ Global

Origin

Noise Gain (Anoise)

Input-referred thermal noise

σinput

Local

Input-referred thermal noise of preamp Data line resistance

Cdata + Cin Cfb

Power supplies

σline

Global

Noise on the global reference voltages supplied to the a-Si array

Cdata + Cin Cfb

Back-end electronic σback noise

Local

All electronic noise generated from the ASIC S&H out through the ADC

1

kTC noise

σkTC

Local

Thermal resistance (Johnson) noise captured on a capacitor in a switched capacitor circuit kTC noise of the pixel capacitance itself is top source and it is usually captured twice.

2

Shot noise

σshot

Local

Noise associated with the TFT and photodiode leakage currents

1

Aliasing

σalias

Local

High-frequency X-ray and electronic noise aliased from above the Nyquist frequency set by the pixel pitch

1

NOISE

24.5.4

547

Input-Referred Thermal Noise

Thermal noise at the preamp inputs connected to each column of the imager is usually the largest source of electronic noise in FPDs. Thermal noise is generated at resistances in the circuit. Besides the preamp itself, the data line’s own resistance is the other major thermal noise source. The thermal noise itself is not particularly high compared to other types of detectors, but the noise gain at this point in the circuit is significant due to the large parasitic data line capacitance. Any thermal noise present at the input to the preamp will have the following noise gain: Anoise =

Cdata + Cin . Cfb

The feedback capacitance on the preamp, Cfb, is matched to the pixel capacitance or set at a lower value in order to get higher charge-to-voltage gain. A typical value for Cfb is 1 pF. On the other hand, Cdata, the data line parasitic capacitance, is on the order of 50–100 pF depending on the size of the a-Si array and on the number of pixels (resolution). The data line length can be up to 43 cm long, having more than 3000 crossover points with underlying structures and a gate–drain overlap capacitance of thousands of TFTs. Also substantial is the input capacitance to the preamp, Cin, which can be 10–15 pF. For the best noise performance, the preamp input transistors are ideally large-area PMOS devices, minimizing the 1/f noise of the amplifier, at the expense of increasing the total data line capacitance. So it is very common in FPDs that the input-referred thermal noise will be gained up by a factor 100, and unfortunately, the signal charge does not see this gain. Just 50 μV of noise at the preamp input can result in over 30,000 electrons of noise, which is roughly the average signal in low-dose fluoroscopy. Clearly, the noise performance of the ASIC preamp is critical to image quality in FPDs. In the discussion above, the noise bandwidth was ignored. It is reasonable to ask why the TFT OFF resistance, which can be mega ohms, is not a major source of thermal noise. The answer is in that the TFT OFF resistance, in combination with the data line capacitance, creates a low-pass filter with a bandwidth below the pass band of the ASIC electronics (∼1–250 kHz), so this noise is effectively rejected. In contrast, the data line resistance is on the order of 1 kΩ, so its thermal noise tends to be in the pass band of the ASIC electronics. At the higher end of the frequency spectrum, the input noise is rejected by the bandwidth limitations of the preamp and S&H circuits. The preamp bandwidth needs to be relatively high in order to hold the input node at “virtual ground,” maintaining the linearity of the charge-to-voltage conversion. However, the only speed requirement on the S&H bandwidth is that the signal settles at a rate sufficient to maintain the line rate. Reducing the bandwidth prior to the signal capture can dramatically reduce the acquired noise. For a given input noise spectral density, in units of V Hz , Qnoise = NSD × NEB1 2 × (Cdata + Cin + Cfb ) ,

548

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

where

NSD = the input-referred noise spectral density and NEB = the noise equivalent bandwidth.

24.5.5

Line Noise

Global noise sources, such as the gate driver power supplies and the array photodiode bias, create correlated line noise because all pixels in a row are sampled at the same moment in time. Noise from the power supplies coupled into the preamp front end sees the same noise gain as the input-referred thermal noise. And making the situation worse, the acceptable level of line noise is on the order of 10 times less than that of the input-referred thermal noise. To meet this condition with just power supply regulation would require power supplies with 5 μV of noise or less, which is extremely difficult, particularly in compact portable systems. Figure 24.12 shows two images of FPD electronic noise, with and without line noise rejection circuitry. In the left-hand image, the line noise variation is one-half the pixel noise, which means that the SNR degradation due to line noise is only 20%. However, it is clear from this image that the line noise dominates the visual appearance. This is a well-known phenomenon related to the human visual system’s excellent ability to distinguish horizontal lines. This problem was an issue in the early days of analog TV. At least one paper from that time proposed that the random noise in the image would need to be 15 times greater than the correlated line noise, in order for the line noise to be invisible. Here, let us define the line

Detector Electronic Noise

σ line ≈

1 × σ pixel 2

Detector Electronic Noise with Line Noise rejection

σ line ≈

1 × σ pixel 10

Figure 24.12. Two images of FPD electronic noise, with (right) and without (left) line noise rejection circuitry.

NOISE

549

noise ratio or LNR as the ratio of the total noise in the image divided by the electronic line noise: LNR =

σ total , σ line

2 2 where σ total = σ X-ray . + σ 2pixel + σ line In FPDs, the ratio of electronic pixel noise to electronic line noise should be a minimum of 10, because at the X-ray quantum-limit, where σX-ray = σelectronic ≈ σpixel, the X-ray noise contributes another factor of 1.4 to the LNR, bringing the total to approximately 15. In FPDs operating above the quantum-limited dose, noise in the X-ray beam dominates the total noise and the LNR will increase with increasing dose, making the line noise invisible at higher signals. Line noise is primarily visible in the lowest part of the signal range, that is, the darker areas of an image where the signal level has fallen below the quantum-limited dose.

24.5.6

kTC Noise

At any point in the signal acquisition process where a voltage level is set on a capacitor through a switch, the thermal noise from the resistance in the switch is captured along with the signal voltage. This noise is known as kTC noise because the magnitude of the switch resistance is irrelevant and the noise depends only on Boltzmann’s constant, the temperature in degree kelvin, and the capacitance. The reason that the resistance drops out of the noise equation, as explained in Figure 24.13, has to do with the self-filtering of the passive resistance, capacitance

A( f ) = v in

1 1 + j 2pfRC

f −3db = v2

R C

v out

1 2πRC

R

v in

C

v out

v 2 = 4kTRΔf Noise Equivalent Bandwidth (NEB) = f −3db × 2 vout = 4kTR × NEB =

π 2

=

1 4 RC

kT C

Output noise is independent of R ! Figure 24.13. The equivalent circuits explaining the resistance-independent kTC noise.

550

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

network. In our front-end equivalent circuit, we can see that kTC noise could be an issue at all the capacitors, including Cpix, Cdata, Cfb and CS&H. Using a noise reduction technique called CDS, the kTC noise associated with Cdata and Cfb can be eliminated. The kTC noise at CS&H can be reduced by using a sufficiently large capacitor. At the S&H circuit, the signal is in the form of voltage, and the noise is actually kT/C. This leaves only the kTC noise associated with the pixel capacitance. kTC noise is created at the pixel during the readout process when the TFT is switched on and node a is discharged to the virtual ground voltage set by the column preamp. When the discharge is complete and the TFT closes, the amount of signal level on the pixel capacitance is uncertain by kT C . At the next reading of the pixel, kTC noise is again introduced because the level to which the pixel discharges is uncertain by kT C . Because the kTC noise is introduced twice, at the pixel reset and again at the readout, the gain factor for the kTC noise on the pixel is 2 . CDS is a noise reduction method taken from the early days of CCDs [3]. With CDS, the reset level of the amplifier is captured and subtracted from the (signal + reset), thereby removing the offset level and the noise associated with the amplifier reset operation. CDS also has the benefit that the capture process becomes high pass, rejecting the low-frequency components of noise. This concept is represented by the AC coupling capacitor between the preamp and the S&H stage, shown in Figure 24.11. The high-frequency components of noise present on the input to the preamp are rejected by the limited bandwidth of the S&H stage, while the low-frequency noise is rejected by the CDS circuit. The low-frequency cutoff associated with CDS is important because most amplifiers have some amount of 1/f noise, noise which increases dramatically at lower frequencies. A typical CDS circuit will create a roll-off point near 1 kHz, which roughly corresponds to the knee in the amplifier 1/f curve. 24.5.7

Shot Noise

Shot noise originates with the dark current leakage processes of the photodiode and TFT. Like the X-ray beam statistics, shot noise follows Poisson statistics. During integration, charge accumulates on the pixel due to a dark current in the photodiode, while at the same time, charge leaks off the pixel due to TFT leakage. The amount of shot noise, in electrons, is equal to the square root of the total signal charge accumulated in the dark. Because the dark currents are very low and the signal capacity is very high, shot noise is usually not an issue in FPDs. There is also shot noise associated with the photogenerated signal, but because the number of electrons far exceeds the number of X-ray photons, the X-ray noise statistics dominate the SNR. 24.5.8

Aliasing

In sampled data systems, signals with frequencies greater than the Nyquist frequency, that is one over two times the sample rate, are aliased into the pass band

NOISE

551

of the system as noise. As we have seen, the electronics acquisition is in fact a sampled data system. However, in the electronics, the bandwidth is managed such that high-frequency signals above Nyquist are attenuated to avoid aliasing. This is not necessarily the case for the X-ray-to-charge conversion process. The a-Si array and scintillator form a two-dimensional sampling system. The transfer function of this system is described by the MTF of the detector. The MTF is the impulse response of the detector, that is, the panel response to a collection of sine waves over all spatial frequencies. Figure 24.14 shows the MTF for a detector with a pixel pitch of 127 μm and two CsI scintillators with different thicknesses. This is the pre-sampling MTF, where the sampling effect of the array pixel pitch is removed. The Nyquist frequency for this FPD is f Nyquist =

1 = 3.94 cycles mm . 2 × (127 μm )

Also shown in Figure 24.14 is the inherent resolution of the bare a-Si array, which follows a Sinc function or Sin ( x ) . All spatial frequencies in the image x acquired by the FPD are modulated by the MTF. For the thick 600-μm CsI, all spatial frequencies above Nyquist are effectively attenuated by the low-pass filtering effect of the scintillator. This is not true for the thin 150-μm CsI. Here, spatial

MTF of 127um FPD Sinc Function 150um CsI

1

600um CsI

MTF

0.8

Nyquist frequency

0.6

0.4

0.2

0 0

1

2

3

4

5

6

7

8

Spatial Frequency (cycles/mm)

Figure 24.14. The pre-sampled modulation transfer function for an array with a 127-μm pixel pitch and two different CsI scintillators. For reference, the array Sinc function is also shown.

552

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

frequencies above Nyquist are aliased into the pass band of the imager (0 to fNyquist) as noise. We can estimate the aliasing contribution to the NPS of the detector by dividing the integral of the MTF2 above Nyquist by the integral of the MTF2 over all frequencies. In the case of the 150-μm CsI on a 127-μm array, 17% of the signal power is aliased into the pass band as noise. Photoconductor-based detectors are discussed in Chapter 25. This type of detector can achieve unparalleled resolution approaching the theoretical limit imposed by the pixel pitch (the Sinc function). However, this resolution comes at the cost of noise performance. For general radiography and fluoroscopy, the desirable pixel pitch based on the imaging task is between 120 and 200 μm. Using smaller pixels comes at the cost of SNR (i.e., increased dose) because the number of X-ray photons per pixel shrinks and, as discussed, the SNR is the square root of the number of absorbed photons. Using photoconductor-based detectors with a 127-μm pixel pitch leads to large aliasing noise (>40%), which is one of the reasons that this technology has not succeeded in general radiography or fluoroscopy. However, in mammography where the imaging task requires a pixel pitch between 50 and 100 μm, the Nyquist frequency is sufficiently high that aliasing is reduced to an acceptable level. Today, photoconductor-type FPDs based on amorphous selenium have a large share of the full-field digital mammography market.

24.6

LAG AND GAIN EFFECT

As discussed in Chapter 23 and alluded to earlier, the a-Si in the p-i-n photodiodes has a large density of deep-level traps, which modulate the signal applied to the array. At the beginning of an X-ray pulse, some of the generated signal charge goes into filling these traps. The result is that the sensitivity of the array increases as the beam ON time progresses because less and less of the generated signal charge is lost to traps. This sensitivity modulation is known as the gain effect. When the beam shuts off, the trapped charge is emitted, contributing to the signal of later frames. This is known as image lag, since a shadow of the previous image will appear in the postexposure frames. In real-time medical imaging, there are many applications where the FPD must switch from a high-dose radiograph (highresolution snapshot) to low-dose fluoroscopy (video) in less than 1 s. Because the radiography dose can be 100–1000 times larger than the fluoroscopy dose, the radiographic image lag is often comparable to the average signal in the fluoroscopy sequence. There are several strategies used to mitigate the gain and lag effects. The first is to correct the images through an algorithmic approach [4], applied after the image acquisition. Although computation resources are required, this approach works well because, to first order, the trapping and emission processes are well behaved and consistent over large variations in signal. The other methods involve prefilling the traps prior to signal acquisition, either by using a bias light or by forward biasing the p-i-n photodiode [5]. Figures 24.15 and 24.16 show the dramatic reduction in gain and lag effect possible using these techniques.

101.00% 100.00%

Sensitivity

99.00% 98.00% Standard Mode Reference Mode Forward Bias Mode

97.00% 96.00% 95.00% 94.00% 0

20

40

60

80

100

Frame Number

Figure 24.15. The strong variation in the sensitivity or gain with frame number for the standard mode and its substantial reduction with the forward bias mode. 0.09%

2.50%

0.08% 0.07% 0.06% 1.50%

0.05%

Standard Mode Forward Bias Mode

0.04%

1.00%

0.03%

Lag with Forward Bias

Lag in Standard Mode

2.00%

0.02%

0.50%

0.01% 0.00% 0

20

40

60

80

0.00% 100

Fram e Num ber

Standard mode

Forwardbias mode

Figure 24.16. Lag effect with and without the forward bias mode (top) and the corresponding postexposure lag images (bottom).

554

24.7

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

ALTERNATIVE ARCHITECTURES

There are a number of other alternative detector architectures briefly summarized in Table 24.2. These concepts span the gamut from current products to pure research & development.

TABLE 24.2. Alternative Detector Architectures Key Features Photoconductor

X-rays converted to charge directly

Potential Advantages Highest resolution

Charge is electrostatically focused on to pixel electrode.

MIS photodiode

MIS-type photodiode No p-type a-Si is needed.

Obstacles Manufacturability of available photoconductor materials High applied bias voltage leads to noise and reliability issues

Compatible with standard LCD processing

Requires a specific reset cycle, so pixel is not continually integrating Difficult to use in real-time (video) imaging

Maxfill p-i-n/ TFT [6]

p-i-n photodiode is on top of TFT. Photodiode is still mesa isolated to avoid signal blooming in saturation.

Organic detectors

Ink-jet printed devices using organic polymers

Increased FF If a low k dielectric is used, Cdata can be reduced.

Low cost

Increased processing complexity, lower yield

Maturity of the technology

Flexible substrates Active pixel sensors

Incorporate more circuitry (TFTs) into the pixel itself.

Good low-dose performance Extended dynamic range

Increased processing complexity, lower yield Same performance may be achievable at a lower cost using standard architecture.

ABBREVIATIONS

24.8

555

SUMMARY

a-Si FPDs are enabling a digital revolution in X-ray imaging, similar to the sea change seen in commercial photography and video. The inherent radiation hardness and large area capability of a-Si detectors are clearly enabling attributes. But the excellent linearity, wide dynamic range, and edge-to-edge uniformity of a-Si FPDs provide high image quality and support advance applications like cone beam CT. Two other obvious advantages are size and weight. The IIT, used in traditional X-ray video, is roughly the size of a kitchen garbage can. An FPD with a 12″ × 16″ active area (20″ diagonal) consumes less than 25% of the volume of a 12″ IIT and less than 15% that of a 16″ IIT. In addition, the FPD takes the place of not only the IIT but also the attached image recording devices, including the CCD camera, the 35-mm cine camera, and the spot film device. The result is vastly improved access to the patient in interventional procedures. In addition to the reduction in size, the weight of a flat panel imager is less than 60% of the IIT-based imaging chain. Despite the advantages of a-Si in large area detectors, clearly their size and limited on-detector processing capability create unique design constraints. While most of the core design challenges have been overcome, developers continue to push toward lower electronic noise for better low-dose performance, improved linearity to reduce image artifacts, higher frame rates to support advance applications like dual energy imaging, and expanded dynamic range for improved cone beam CT.

ABBREVIATIONS ACF—anisotropic conductive film ADC—analog-to-digital converter, converts an input voltage to digital data Anoise—noise gain a-Si—amorphous silicon ASIC—application-specific integrated circuit; refers to custom readout chips connected to the columns of the array C—capacitance Cdata—dataline capacitance CCD—charge-coupled device CDS—correlated double sampling Cfb—preamp feedback capacitance Cin—preamp input capacitance CMOS—complementary metal oxide semiconductor; refers to a photodiode-based silicon imaging chip Cpix—pixel capacitance Cs&h—sample and hold capacitance CsI, CsI(Tl)—thallium-doped cesium iodide, a type of X-ray scintillator CT—computed tomography

556

AMORPHOUS SILICON DIGITAL X-RAY IMAGING

DQE—detective quantum efficiency EPID—electronic portal imaging detector f—frequency FF—fill factor fNyquist—Nyquist frequency FPD—flat panel detector FPGA—field-programmable gate array GOS—gadolinium oxysulfide, a type of X-ray scintillator Gy—Gray, a unit of radiation dose i—undoped or intrinsic semiconductor IIT—image intensifier tube ITO—indium tin oxide k—Boltzmann’s constant LNR—line noise ratio LVDS—low-voltage differential signaling Mrad—mega (1 million) rads, where rad is a unit of X-ray dose describing the ionization energy deposited in an irradiated object MTF—modulation transfer function n—negatively doped semiconductor NEB—noise equivalent bandwidth NPS—noise power spectrum NSD—noise spectral density p—positively doped semiconductor PMOS—p-type metal oxide semiconductor devices Qnoise—electronic noise in coulombs R—resistance RH—relative humidity S—signal intensity S&H—sample and hold circuit SNR—signal-to-noise ratio T—absolute temperature TAB—tape-automated bonding TFT—thin-film transistor σ—signal standard deviation, i.e., noise in the signal τ—pixel discharge time constant

REFERENCES [1] [2]

L. E. Antonuk and R. A. Steet. Multi-element-amorphous-silicon-detector-array for real-time imaging and dosimetry of megavoltage photons and diagnostic X rays. U.S. Patent # 5,079,426, September (1989). P. Munro, J. A. Rawlinson, and A. Fenster. Therapy imaging: A signal-to-noise analysis of a fluoroscopic imaging system for radiotherapy localization. Med. Phys. 17, 673–772 (1990).

REFERENCES [3] [4] [5]

[6]

557

M. H. White, D. McCann, I. Mack, and F. Blaha. Coherent sampled readout circuit and signal processor for a charge coupled device array. U.S. Patent 3,781,574, October (1972). L. D. Partain, I. Mollov, C. Tognina, and R. E. Colbeth. Method and apparatus for correcting excess signals in an imaging system. U.S. Patent 7,208.717, October (2002). I. Mollov, C. Tognina, and R. E. Colbeth. Photodiode forward bias to reduce temporal effects in a-Si based flat panel detectors. In Medical Imaging 2008, Proceedings of SPIE, Vol. 6913-13, J. Hsieh and E. Samei, eds., pp. 69133S-1 to 69133S-9. Bellingham, WA, SPIE (2008). R. L. Weisfield, W. Yao, T. Speaker, K. Zhou, R. E. Colbeth, and C. Proano. Performance analysis of a 127-micron pixel large-area TFT/photodiode array with boosted fill factor. Medical Imaging 2004: Physics of Medical Imaging, Vol. 536842, M. J. Yaffe and M. J. Flynn, eds., pp. 338–348. Bellingham, WA, SPIE (2004).

25 PHOTOCONDUCTOR DIGITAL X-RAY IMAGING GEORGE ZENTAI Varian Medical Systems

25.1

INTRODUCTION

Conceptually, photoconductor detectors initially seem quite different from solar cells. They consist of a highly resistive semiconductor layer placed between two metal electrodes to efficiently collect any mobile electric charge generated by incident photons based on a voltage applied to the electrodes. However, the a-Si solar cell structure shown in Figure 23.9 actually shows this exact same structure. The a-Si solar cell does have thin, highly doped p+- and n+-layers next to the metal electrodes. A standard way of producing ohmic metal contacts to semiconductors is to heavily dope the regions just adjacent to the contacts. For solar cell energy conversion, the doping needs to be of opposite types at each electrode. Solar cells are typically reverse biased when used as photon detectors and imagers so that no net positive energy conversion is produced. In essence, solar cell detectors and photoconductor detectors are fundamentally quantum converters of light, absorbing as many incident photons as possible. Ideally, every absorbed photon contributes to a net current signal that is linearly related to the intensity of the photon flux striking the detector’s surface with a minimal noise contribution. In this mode, the behavior and principles of operation of photoconductors are very nearly identical to those of solar cells that contain a very high-resistance photon active region. The latter includes point contact crystalline concentrator solar cells in addition to the a-Si ones. For solar cell energy conversion, the device bandgap needs to be in the same range as the incident photon energies. For X-ray detectors, there is no such restriction so that photoconductors are often chosen with bandgaps larger than most visible light photons because this gives the lowest dark currents important to Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

559

560

PHOTOCONDUCTOR DIGITAL X-RAY IMAGING

X-ray imaging. Furthermore, the a-Si TFTs, used in X-ray imagers, have the same basic structure with a layer of high-resistance semiconductor contacted by two metal contacts made ohmic by highly doped n-type interface layers (see Fig. 23.6). Here, the device current is controlled not by absorbed photons but by a third intervening field-effect contact separated from the a-Si by an insulating layer. Traditionally, there were two ways of X-ray imaging to capture information in medical practice: 1. radiography, a snapshot of overlapped internal organs of the imaged patients, and 2. fluoroscopy, a continuum of images or a series of snapshot images of the imaged objects. In the early 1970s, X-ray CT was invented and provided medical practitioners with an unparallel view of patient cross-sectional views without surgery. Many other X-ray imaging techniques are just slight variations of the above X-ray imaging methods. For all these medical X-ray imaging modalities, high performance at low X-ray exposure and at high spatial frequencies is demanded. Most AMFPIs currently available commercially are indirect in that they use a two-stage process, where X-ray photons are first converted into visible light in a top scintillator layer, and light then (in the second step) is transported down into an underlying layer array of photodiodes that convert part of this light into detectable electronic charge. This is in contrast to the D-AMPFI that eliminates the need for a scintillator and replaces the photodiode layer with a photoconductor, which directly converts the X-ray photons into a detectable electronic charge in a single step. In addition to simplicity and potentially lower costs, D-AMFPIs provide other major advantages of higher spatial resolution and better contrast sensitivity than the indirect configurations. This is due to the fundamental limitation of scintillators that emit light uniformly in all 4π geometric directions so that much of the light is lost, and the fraction that does transport down excites not only the pixel just below its generation position but also multiple adjacent pixels as shown in Figure 25.1. This fundamentally limits the spatial resolution, particularly in indirect devices with thick enough scintillators to efficiently absorb most incident X-ray photons. The greater the scintillator thickness, the greater image blurring and the lower associated spatial resolution (see Fig. 25.2). In direct imagers, a voltage is applied to a photoconductor layer to collect the charge, generated by the absorbed X-ray photons, so that almost all of the charge transports to the pixel underneath the generation positions due to the field focusing caused by the applied electric field as shown in Figure 25.3. This focusing effect exists even in thick photoconductor layers so the resolution does not degrade with increasing detector thickness and increasing sensitivity (Fig. 25.4). This is especially advantageous at imaging with high-energy X-rays. Such advantages are dramatically demonstrated in a-Se D-AMPFI (the single D-AMPFI currently available commercially), whose spatial and contrast resolution are unprecedented and unmatched by any indirect imager. These advantages become particularly valuable

INTRODUCTION

561

Figure 25.1. Scattered light generates charge not only in the pixel where the X-ray hits the imager but also in neighboring pixels.

Figure 25.2. Thicker scintillator—better absorption—worse resolution (MTF).

562

PHOTOCONDUCTOR DIGITAL X-RAY IMAGING

Figure 25.3. Charge generated by an X-ray beam moves perpendicular to the electrodes— no charge spreading.

Bias Bias

El. field

X-ray

Charge movement

Lower sensitivity

Good resolution

El. field

X-ray

Top electrode

Charge movement

Higher sensitivity

Pixel array

Good resolution maintained at thicker layers

Line spread function

Figure 25.4. Thicker layer—better absorption—no degradation in resolution (MTF).

PHOTOCONDUCTOR IMAGERS

563

in mammography, where the high spatial resolution requirements (all the way out to the Nyquist sampling frequency limit) are closely coupled to the additional need for resolving subtle contrast variations. The a-Se D-AMFPIs are beginning to dominate the commercial flat panel mammography market.

25.2 GENERAL REQUIREMENTS FOR PHOTOCONDUCTOR MATERIALS USED FOR X-RAY DETECTION AND IMAGING Direct detector materials preferably exhibit several attributes. These include high X-ray absorption, high charge collection, low dark current, and good uniformity. These are difficult to achieve in a single material. The most important physical parameters of the popular X-ray conversion photoconductor materials are summarized in Table 25.1 [1]. For efficient X-ray absorption at clinical exposure energies, we need high Z materials. In addition, the X-ray energy required to generate an electron-hole (e-h) pair, designated by the parameter W, should be low. The lower the W value, the higher the number of e-h pairs liberated by the interacting X-rays and the higher the X-ray sensitivity. As Table 25.1 indicates, the parameter W is much smaller for HgI2, PbI2, and CdZnTe than for a-Se. The larger the mobility-lifetime product (μτ), the greater the distance the electrical charge can move in the detector without recombination. Greater distances result in higher sensitivity due to better charge collection. The high Z number, low W, and high μτ result in a very high signal, which helps overcome noise sources in fluoroscopic (low dose) modes of operation. The ability of the imager to operate at low bias voltages allows low-voltage electronic design. The a-Se requires minimum 10 V/μm electrical field, while other photoconductor materials need only a 1 V/μm field to work at decent sensitivity levels. Another key feature of the materials is the processing temperatures they require to be fabricated into the desired form. The AMFPI panels have an a-Si switching matrix TFT layer on a glass substrate, and the highest temperature these TFTs can tolerate without damage is about 220–240°C. From this point of view, CdZnTe is not a good candidate because it requires a minimum of 600°C substrate fabrication temperature, which is way over the limit of the TFTs. There have been experiments to deposit CZT on a separate substrate, but on high-pixel-count largearea arrays, the connections to the pixellated TFT matrix readout array rapidly become unmanageable, limiting the maximum imager size only to about 9 in.2 [2]. CdZnTe, PbI2, and HgI2 are all much more stable in terms of phase transitions at elevated and at low temperatures than a-Se. a-Se imagers require temperature control not only during operation but also during shipment.

25.3

PHOTOCONDUCTOR IMAGERS

At present, only a-Se imagers are commercially available; the other imagers mentioned in this chapter are only experimental devices.

564

PHOTOCONDUCTOR DIGITAL X-RAY IMAGING

TABLE 25.1. Comparison of a Few Photoconductor Candidates for X-Ray Conversion Poly-HgI2

Poly-PbI2

a-Se

Poly-CdZnTe

80, 53

82, 53

34

48, 30, 52

Energy bandgap (Eg) (eV)

2.1

2.3

2.2

1.5–1.7

Charge pair formation energy (W) (eV)

5

5.5

50 (eff.)

4.5

Lower W—higher gain

Mobilitylifetime product (μτ) (cm2/V)

10−5

h: 1.8 × 10−6 e: 7 × 10−8

10−6–10−5

e: 8 × 10−3 h: 3 × 10−5

Higher μτ—better charge collection

Operational electric field (E) (V/mm)

0.2–1.0

0.2–1.0

10

1–2

Lower E— reduced risk of electrical breakthrough

Processing temperature (°C)

120

220

100

600

Lower temperature— simpler process

Stability (°C)

>100

>200

<60

600

Phase transition temperatures

Atomic number (Z)

25.3.1

Comments Absorption increases with Z Broader gap— lower dark current

a-Se X-Ray Imagers

Based on decades of research on a-Se xerographic drums, the first photoconductor X-ray imagers were developed using a-Se as an X-ray-sensitive material. The first company that developed an a-Se X-ray imager was Sterling Diagnostic Imaging [3]. They obtained the first FDA 510(k) approval for their a-Se imager in 1997. In 1999, Analogic purchased the digital imaging part of Sterling Diagnostic Imaging and later Anrad, the a-Se deposition company. In parallel, Hologic/LORAD also worked on a-Se imager development. One of the main issues for a-Se is the very high electrical field (bias voltage) requirement (∼10 V/μm translates 10,000 V for a 1-mm a-Se layer). The imager’s a-Si TFT switching devices have a limited breakdown voltage in the range of ≤ 80 V, but the TFT leakage current already starts to increase above ∼20 V source– drain voltage differences. Any tiny pinholes or inhomogeneities in the a-Se material can contribute to a breakdown and immediate destruction of the underlying

PHOTOCONDUCTOR IMAGERS

565

TFT transistor, plus it can damage much of the other electronic circuits. To avoid this and also to decrease the dark current, Sterling Diagnostic Imaging developed a special (patented) structure consisting of a thick insulator layer between the a-Se layer and the bias electrode [3]. An additional charge blocking layer was applied between the a-Se and the TFT array, which further decreased the dark current. The disadvantage of this method is that there is no direct path for DC current flow so that special techniques have to be used (light exposures or unique electrical pulse sequences) to reset the imager after each X-ray exposure and readout. Such reset cycles decrease the maximum imaging frame rate. Anrad developed a rectifying p-i-n diode structure in a-Se [4] that was further modified jointly with Toshiba [5]. It requires no special reset sequences. To avoid the high-voltage breakdown and to protect the electronics, two methods are available. In one, a second TFT is connected in parallel to the switching TFT, which automatically opens, when the source–drain voltage exceeds a certain level, to protect the original TFT. One of the options of this additional TFT solution is to mimic the second TFT with the dual gate method, where the second or top gate is the extension of the pixel electrode [6]. The other solution is to use “reverse” bias. This means applying a negative bias onto the top electrode instead of the usual positive bias for a-Se. It decreases the sensitivity to about two-thirds of the positive biased case but allows the TFTs to be protected. Excess negative bias on the drain electrode causes the gate–drain to become forward biased. Thus, the TFT becomes conductive and drains off the excess charge from the drain electrode, decreasing the voltage difference between the drain and the source to safe levels. Earlier, there were some experiments to develop a high-speed a-Se fluoroscopic detector [4]. However, the sensitivity of the a-Se is much lower (about half) than that of a CsI/photodiode imager; as a result, a-Se falls behind in low-dose fluoroscopic and radiographic applications. During an intensive clinical test, radiographers preferred the Trixell Pixium 4600 CsI imager over a Kodak Directview DR-9000 a-Se imager because of the better low-frequency DQE (better X-ray attenuation in the diagnostic energy range) despite much better spatial resolution of the a-Se imager [7]. These various disadvantages have greatly restricted the use of a-Se imagers in medical imaging. The main application of a-Se imagers is mammography, where they have much better spatial resolution than any other commercially available (indirect) flat panel imagers. A comparison study [8] shows that the resolution (MTF) is always much better for the LORAD a-Se imager with 70-μm pixels than for the GE Senographe imager, with 100-μm pixels (Fig. 25.5). Also, the 70-μm LORAD MTF is about equivalent to that of the highest MTF-performing Sectra slit scanning and photon counting system with 50-μm pixels. In summary, LORAD’s detectability for microcalcifications is shown to be much superior to that of the GE Senographe. The a-Se has better attenuation in the mammography energy range than CsI because its K-edge is just below this energy range (see Fig. 25.6). This helps to compensate somewhat for the lower efficiency of X-ray conversion. Figure 25.7 shows a DQE comparison of different digital imagers [8]. It compares three CR imagers: the Fuji CR Profect, the Kodak CR 850 HER-M, and

566

PHOTOCONDUCTOR DIGITAL X-RAY IMAGING 1.0 Fuji CR Profect laser scan subscan Kodak EHR-M laser scan subscan Kodak Rejius 190 laser scan subscan GE Senographe DS all directions Lorad Selenia all directions Sectra MDM slit scan array

0.9 0.8

MTF

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

1

2

3

4

5

6 7 8 9 10 Spatial frequency [mm–1]

11

12

13

14

15

Figure 25.5. Presampled MTF of a few digital mammography systems.

Figure 25.6. K-edge of the a-Se is just below the mammographic energy range so it has better attenuation than CsI for mammography. However, the attenuation of CsI is better in the radiographic (diagnostic) energy range. HgI2 and PbI2 materials have good X-ray attenuation in both energy ranges.

the Konica Regis, as well as the GE Senographe DS (indirect with a 100-μm pixel size) and the LORAD Selenia (direct 70-μm pixel a-Se) plus the photon countingslit scanning mode Sectra MDM imager. The Fuji is a dual-sided CR system, which shows a higher low-dose DQE than the other CR systems. This figure demonstrates that the DQEs are roughly equivalent for the GE Senographe DS (indirect with CsI) and the LORAD Selenia (direct a-Se) imagers in the medium to high dose regions, but at low doses, the a-Se imager falls behind.

PHOTOCONDUCTOR IMAGERS

567

Figure 25.7. Maximal low-frequency DQE of the digital mammography systems as a function of the exposure level.

The main goal is to increase the signal level over the noise of the readout electronics at low-dose fluoroscopy, to obtain quantum noise limited detection. Antonuk et al. [9] demonstrated that if the signal level for a-Se could be increased by about 10 times, then quantum noise limited detection would be achieved at low-dose fluoroscopy even with a-Se detectors. In one design, a-Se has a built-in avalanche region [10, 11]. This avalanche region multiplies the number of charges generated by the X-rays so the output electrical signal is highly increased. A slightly modified solution was proposed by Zhao et al. [12–14], in which CsI, as an indirect X-ray conversion layer, was deposited on the top of an a-Se layer. This relatively thin a-Se is used as a photo sensor for the light, generated in the CsI layer above it. It also serves as an avalanche layer to increase the number of electrons at the readout. A further solution is to apply pixel-level gain stages such as using active pixel a-Si TFT arrays [15–18]. Unfortunately, a-Se does have some further unavoidable disadvantages. In addition to low X-ray sensitivity, stability problems complicate shipment and storage and decrease the operational lifetimes of imagers. This is due to a fundamental and destructive phase transition at elevated temperatures (>60oC) and a temperature expansion–coefficient mismatch delamination at lower temperatures (<0oC). Another problem with a-Se is the ghosting effect [19–22]. This decreases the sensitivity in areas where high-intensity X-ray illumination occurs, as

568

PHOTOCONDUCTOR DIGITAL X-RAY IMAGING

Figure 25.8. Difference between image lag and ghosting.

demonstrated in Figure 25.8. This is the opposite of lag, where one can see a brighter spot on the dark image after illumination. Image lag generally disappears after a few frames, while ghosting is a long-lasting process that can take several hours or days to disappear. However, some recent investigations show that ghosting is probably not significant during a single mammographic exam. With some special methods, ghosting can be neutralized between mammography patients. This makes it possible to take several lower-dose images for breast tomography (quasi3-D imaging of the breast) for better visibility of tumors and microcalcifications. More detailed analyses on using a-Se imagers for tomosynthesis can be found in the literature [23, 24].

25.3.2

CdTe and CZT

Single-crystal CdTe and CdZnTe (CZT) devices have long been known to provide X-ray-radiation detectors with excellent energy resolution [25] and high X-ray sensitivity. CdTe detectors have very good energy resolution. Their main issue has been high dark current, preventing the application of high enough electrical fields at room temperature for good performance. One solution was to lower the temperature of these detectors [26] with TE coolers, but this also requires hermetic sealing of the detectors. Another solution was zinc addition to the CdTe. This increased the bandgap and significantly decreased the dark current at room temperature, motivating the development of CdZnTe (CZT) detectors. However, there is a technique to fabricate Schottky barrier electrodes on CdTe that helps resolve the high dark current problems and provides very low dark current and good performance [27, 28].

PHOTOCONDUCTOR IMAGERS

25.3.3

569

Large-Area CdTe and CZT X-Ray Imagers

Hamamatsu made the first attempt developing an X-ray imaging device based on CdTe [29]. It consisted of a CdS/CdTe heterojunction with two electrodes. The whole structure was mounted inside a vacuum camera tube. The imager gave excellent resolution but low sensitivity to X-rays because the layer thickness was only 2–6 μm so that the X-ray absorption was very low. The imager area was also limited by the size of the vacuum tube. The following imagers are based on an a-Si TFT readout matrix similar to the a-Se imagers. Instead of the a-Se, a polycrystalline CZT layer is used as an X-raysensitive photoconductor. The problem for direct deposition of polycrystalline CdTe or CZT onto a-Si TFT array, as shown in Table 25.1, is the very high substrate temperatures required for growing reasonable quality films. To avoid the direct deposition onto the a-Si TFT array, Shimadzu and Sharp deposited CdTe and CZT onto a substrate, separate from the TFT array and then tried to bump bond to the pixellated a-Si matrix [2, 30–32]. The maximum size of the array was 9 in.2 They got relatively high dark current, especially high for a CdTe array, reasonable resolution, and for the CZT, the DQE(0) was ∼35% at 80 kVp and 54-μR dose. The main issues were the huge pixel-to-pixel inhomogeneity and the inability to make a large array with reliable pixel-to-pixel connection (bump bonding). 25.3.4

Energy Resolution CdTe and CZT Detectors and Imagers

Briefly, the following detectors do not use a-Si TFT readout matrix with an amorphous or polycrystalline semiconductor X-ray-sensitive layer. Instead, pixellated CdTe and CZT single crystals are used. As large size CdTe and CZT crystals become cheaper along with the special, multichannel counting mode readout ASICs with energy resolution, applications of pixellated CdTe and CZT single crystal detectors for SPECT and PET as well as for other X-ray imaging fields are growing [33–35]. Experiments to develop such detectors for combined SPECT–CT [36] encountered limited count rates below the levels needed for CT scanning. Szeles et al. [37, 38] achieved 5–15 × 106 counts/s. Recently, Iwanczyk et al. [39] demonstrated a 6 × 106 maximum counts per second rate for CT scanning using a special readout ASIC. It is worth mentioning that special readout circuits were developed for increased counting rates that do not give full energy resolution. Instead, they have two or more energy bins where the X-ray photons are only sorted into a few broad energy bands. This still retains advantages over traditional CT of improved signal to noise, reduced X-ray scatter artifacts, and better detectability of tumors. 25.3.5

PbI2 and HgI2 Materials

Both PbI2 and HgI2 have excellent X-ray absorption coefficients because they are high Z photoconductor materials with a wide bandgap, preferable for low dark

570

PHOTOCONDUCTOR DIGITAL X-RAY IMAGING

currents. Both materials have low charge pair formation energy (W) values of about 5 eV/e-h generation. Theoretically, this provides much higher sensitivity than a-Se (about 10 times more e-h generated than in a-Se). Moreover, both photoconductors require much lower electrical fields of ≤1 V/μm (bias voltage) to collect most of the generated charges compared with a-Se of ≤10 V/μm. PbI2 has excellent resolution similar to the resolution of a-Se as shown in Figure 25.9 in comparison to a CsI indirect imager of the same pixel size. The resolution of HgI2 is also very similar to the other photoconductor-based imagers [40]. The excellent resolution of these materials was also demonstrated with comparative X-ray imaging in Figure 25.10 [40]. The images show the same foot phantom for direct comparisons of CsI, PbI2, and HgI2 imagers all with same pixel

Figure 25.9. Spatial resolution—MTF comparison of indirect CsI + CMOS and direct PbI2 and a-Se imagers. All of the imagers had the same 50-μm pixel pitch.

(a) CsI

(b) PbI2

(c) HgI2

Figure 25.10. Resolution comparison of foot phantom images using indirect and direct imagers. All of the imagers had the same 127-μm pixel pitch.

PHOTOCONDUCTOR IMAGERS

571

size. The details are much more visible and the bone contours are much sharper for the two photoconductor imagers. Single-crystal PbI2 has been studied by several authors [41–44] as an X-ray detector with energy resolution. However, their detection performance was not as good as that of CdTe, CZT, and cooled Ge detectors. To obtain a PbI2 X-ray imager, one has to deposit it onto a TFT array. A preferred substrate temperature for PbI2 deposition is slightly over 200°C. So, it is possible to directly deposit it onto an a-Si TFT array (whose maximum temperature limit is ∼240°C). A few authors have studied the imaging properties of PbI2 and found that it has very good X-ray absorption, a relatively low dark current, and reasonable X-ray sensitivity [45–47]. However, the main issues as confirmed in Varian studies were the slow rise time and very long image lag [48, 49] in comparison to HgI2 as seen for PbI2 #761 in Figure 25.11. Here, PbI2’s first frame lag was 33.7%, while HgI2’s was only 7.1%. Long image lag is not an inherent property of the PbI2 as it was pointed out in Reference 40, where the signal shape of singlecrystalline PbI2 and HgI2 materials was compared as seen in Figure 25.12. Rather, it is due to the imperfect polycrystalline structure of the PbI2. Hence, there is hope that long image lag could be decreased with optimized deposition processes. Improvement in radiographic image lag is shown for a new generation of PbI2 imagers, with a first frame lag of only 10.5% and a fast subsequent decay in following frames. This imager has much better crystalline morphology than the previous PbI2 imager. The initial work on HgI2 radiation detector technology began with EG&G’s single-crystalline HgI2 radiation detector research [50–52]. They investigated optimal crystal growth [53, 54], purification of starting materials [55], radiation hardness [56], and so on. Later, Constellation Technology Corporation continued the HgI2 single crystalline detector development [57–59] and Photon Imaging

Figure 25.11. First frame image lag comparison of an older PbI2 #761 and HgI2 #27 imagers to a new generation of PbI2 #8733-07H imager. The imagers run at 10 frames/s.

572

PHOTOCONDUCTOR DIGITAL X-RAY IMAGING

Figure 25.12. Signal transient comparison of single-crystalline and polycrystalline PbI2 and HgI2.

began polycrystalline imager development on TFT arrays [60, 61]. At the same time, Hebrew University and RTR, a Hebrew University’s spin-off company, also began polycrystalline depositions onto TFT arrays [62–64]. The deposition method in these cases was PVD, where HgI2 material was evaporated onto the TFT array in a vacuum system. It is known that HgI2 easily sublimates well below its melting point and deposits on colder surfaces. Thus, the deposition temperature is relative low. However, special deposition conditions are needed for good quality polycrystalline HgI2. Schieber pointed out that for low dark current, single crystalline-like growth is preferable. Some details of this technology are described by Schieber et al. [63, 65]. Street at Xerox PARC [66] and Varian also developed arrays based on RTR’s vacuum deposited HgI2 films [67]. The first arrays gave very high sensitivity and resolution, but the problem was the large dark (leakage) current. Because of the high dark current, the pixels of the imager were saturated by the dark current at longer exposure time. Hence, these experimental imagers did not work for singleshot radiography or mammography [1]. However, their low image lag (see Fig. 25.12) and high sensitivity were already very promising for low-dose fluoroscopy. Figure 25.13 compares the sensitivity of some of the best quality PVD-HgI2, PIBHgI2, PbI2, CsI, and a-Se imagers. Antonuk and his group at the University of Michigan also started to work on these new promising materials [68, 69] and compared several PbI2 samples from radiation monitoring devices (RMD) and HgI2 samples from RTR [9, 49]. Kang [70] did a detailed evaluation of several PbI2 and HgI2 samples and found that the dark currents of these samples are higher than those of good quality a-Se samples and also pointed out that both materials provide much higher sensitivity than any indirect material. Schieber at RTR also developed another special HgI2 deposition process called PIB [63], where small HgI2 crystallites are embedded in a polymer binder material. The deposition method is screen printing or a similar technology. The TFT array onto which the PIB material is deposited can be the same as those used for a-Se or for other photoconductor imagers. Su [71] compared a large number

PHOTOCONDUCTOR IMAGERS

573

Figure 25.13. Sensitivity comparison of PVD HgI2, PIB HgI2, PbI2, a-Se, and CsI imagers.

of HgI2 detectors made both by PVD and PIB methods under mammographic, radiographic, fluoroscopic, and radiotherapy irradiation conditions. He found that PIB HgI2 detectors gave lower dark currents than the PVD ones, but the signal transient properties of the PIB materials were worse. He also confirmed that the signal level given by these detectors are much higher than that of a-Se, CsI, or other indirect materials. These results have been confirmed at Varian with the sensitivity of a PIB HgI2 array also added to Figure 25.13. PIB can provide a good method for making thick detectors to improve the efficiency of the radiotherapy (portal) imaging. Du and Antonuk et al. [72] gave a summary of their HgI2 imagers’ evaluations. All of their imagers were prepared by RTR either by a PVD or PIB process. They found that the PIB imagers have lower dark current but also lower sensitivity. The image lag properties are more favorable for the PVD imagers, but both types of imagers still have problems with sensitivity inhomogeneities at the pixel-to-pixel level. PVD imagers also gave excellent resolution (MTF). They found inputquantum-limited DQE for one of the imagers at relatively low exposures. However, they did not find a single imager in the study, which would exhibit a combination of high sensitivity, low dark current, and high MTF. Another issue was the Al corrosion in the TFT arrays. It is known that HgI2 is a very corrosive material that is especially bad with Al. Unfortunately, most of the TFT arrays have some Al components, used mostly for improving the conductivity of data lines. When a pinhole exists in the TFT arrays and HgI2 finds an access, then Al corrosion starts and progresses until all of the Al line is destroyed. To avoid HgI2 contact with Al, a chemical barrier layer was developed between the TFT array and the HgI2 layer. However, this chemical barrier layer needs not to just prevent the corrosion but it also has to be conductive, but only in a vertical direction to provide a good electrical contact between the TFT array and the HgI2 material without shorting neighboring pixels that degrade resolution. It is a major

574

PHOTOCONDUCTOR DIGITAL X-RAY IMAGING

challenge to develop such a material, which fulfills all of these requirements. There is a very tight process window both for deposition of this material and for growth of a perfect crystalline structure of HgI2 layer on the top of this barrier layer. However, demonstrations came close as Zentai et al. [73] reported a PVD HgI2 imager, which had a low dark current, high sensitivity, and high DQE down to very low doses. The only remaining issues were reproducibility and yield. Zentai and Partain [74] also reported preliminary test results of using HgI2 imagers for high-energy imaging and demonstrated that the resolution of a PVD HgI2 imager did not degrade at high-energy exposure with a 6-MeV Betatron X-ray source.

25.3.6

Other Photoconductors

Some other exotic photoconductor materials have also been tested such as TlBr [75], BiI3 [76, 77], and HgBrI [78], but currently, their X-ray imaging properties do not approach those of the materials referenced above. Simon et al. measured PbO [79] and even showed some decent X-ray images, but the stability of the material was questionable and the imagers just worked in a N2 atmosphere.

25.4

NEW DIRECTIONS

Finally, some new directions in flat panel imaging warrant mentioning. One is the use of printed TFTs and also flexible substrates instead of glass substrates (as reviewed in Chapter 23). The transistors can be low-temperature deposited a-Si TFTs and even organic TFTs [80]. The main advantage of the flexible substrate is that it is more robust, does not break, and also can be bent to special shapes such as part of a cylinder for better CT imaging. At present, the charge mobility in the organic materials is about an order of magnitude lower than in the a-Si, and their radiation hardness has not been proven, but this might be a future way of getting cheap, large-area X-ray imagers. Using polycrystalline Si TFTs instead of a-Si is already well-known for LCD technology. It is also gaining popularity in the X-ray imaging field [81]. The main advantage of the polycrystalline devices over the amorphous ones is the higher charge mobility that makes it easier to provide TFT transistor amplifiers on each pixel. Moreover, multiplexer circuits and amplifiers to each pixel can also be implemented on the imager’s substrate [81]. The advantage of the pixel-level amplifiers is that the signal from each pixel is amplified so the noise is reduced and quantum noise limited detection can be achieved at very low fluoroscopic dose levels. One important remaining open issue is the radiation hardness of the polycrystalline silicon TFT. Another direction is deposition of photoconductor materials on CMOS matrices. However, only indirect-type X-ray imagers (with photodiodes) on a CMOS

SUMMARY AND CONCLUSION

575

structure are available at the present time. An advantage of CMOS technology is that the whole readout circuit can be implemented on the same wafer. The limitation of this technology is the limited wafer size, which limits the maximum imager size. However, tiling can be a solution. A further question is the radiation hardness.

25.5

SUMMARY AND CONCLUSION

It has been shown that the very high spatial resolution and simpler manufacturing (no need for photodiode) give certain advantages for photoconductor-based imagers. The a-Se had been a well-studied material for a long time and became the first photoconductor material used commercially for X-ray imaging. Today, several a-Se-based imagers exist on the market, but their main application is mammography. The major drawbacks are the low X-ray conversion efficiency and the very high electrical bias requirements. Single-crystalline CdTe and CZT are excellent X-ray detectors with energy resolution, but at present, their homogeneous deposition in polycrystalline forms on large areas is difficult, and they cannot be directly deposited onto TFT arrays because of their high substrate temperature requirements. Deposition onto a separate substrate and then bump bonding to the TFT array has not been very successful at large sizes. Pixellated single crystalline detectors are getting cheaper, but their size is limited by the crystal size as currently configured. Tiling is possible, but at present, this technology is too expensive for large-area X-ray imaging. PbI2 and HgI2 are newer materials that do not have as long an R & D history nor as extensive a learning curve development as a-Se. Both have high X-ray sensitivity and excellent resolution. At present, the polycrystalline PbI2 has longer lag, but it can be significantly decreased by optimizing the deposition parameters as shown in Figure 25.11. Another issue is that it is difficult to deposit a thicklayer PbI2 of good quality because, as the layer grows thicker, the morphology becomes porous. At present, mammography seems to be the best application area for PbI2. PVD HgI2 has already proven many good properties including low lag and high DQE. However, it still has a problem with the pixel-to-pixel homogeneity and reproducibility. If perfected, this material could be used for all X-ray imaging modalities. PIB HgI2 has further advantages that it can be easily deposited onto large areas and in thick layers. This gives opportunities to use it as a highly efficient imager in radiography and radiotherapy as well as in industrial imaging. The only drawback is the relative high image lag, which limits its application in high frame rate (fluoroscopy, CT, and tomosynthesis) imaging. New directions are X-ray imagers on flexible substrates and using polycrystalline Si and organic semiconductors for the readout instead of a-Si, but these technologies are still in their infancy.

576

PHOTOCONDUCTOR DIGITAL X-RAY IMAGING

ABBREVIATIONS 3-D—three dimensional AMFPI—active matrix flat panel imager a-Se—amorphous selenium a-Si—amorphous silicon ASIC—application-specific integrated circuit CdS—cadmium sulfide CdTe—cadmium telluride CMOS—complementary metal oxide semiconductor CR—computed radiography CsI—cesium iodide CT—computed tomography CZT—CdZnTe—cadmium zinc telluride D-AMPFI—direct active matrix flat panel imager DQE—detective quantum efficiency DR—digital radiography E—operational electric field Eg—energy bandgap e-h—electron–hole pair EG&G—Edgerton, Germeshausen, and Grier, Inc. FDA—(U.S.) Food and Drug Administration GE—General Electric HgI2—mercuric iodide kVp—kilovoltage peak LCD—liquid crystal display MTF—modulation transfer function n—negatively doped semiconductor p—positively doped semiconductor PARC—Palo Alto Research Center PbI2—lead iodide PET—positron emission tomography PIB—particle in binder PVD—physical vapor deposition R—rad, which is a unity of x-radiation dose that deposits ionization energy into an object RMD—radiation monitoring devices RTR—real time radiography SPECT—single photon emission computed tomography TE—thermoelectric TFT—thin-film transistor W—charge pair formation energy Z—atomic number μτ—mobility-lifetime product

REFERENCES

577

REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27]

G. Zentai, L. D. Partain, R. Pavlyuchkova et al. Medical Imaging 2002: Physics of Medical Imaging, p. 592. San Diego, CA, SPIE (2002). S. Tokuda, H. Kishihara, S. Adachi, et al. Medical Imaging 2002: Physics of Medical Imaging, p. 30. San Diego, CA, SPIE (2002). D. L. Y. Lee, L. K. Cheung and L. S. Jeromin. Medical Imaging 1995: Physics of Medical Imaging, p. 237. San Diego, CA, SPIE (1995). M. Choquette, H. Rougeot, J.-P. Martin, et al. Medical Imaging 2000: Physics of Medical Imaging, p. 128. San Diego, CA, SPIE (2000). O. Tousignant, M. Choquette, Y. Demers, et al. Medical Imaging 2002: Physics of Medical Imaging, p. 503. San Diego, CA, SPIE (2002). J. Lehnert and W. Zhao. Medical Imaging 2006: Physics of Medical Imaging, p. 61420S. San Diego, CA, SPIE (2006). L. L. Barski, X. Wang, J. Wandtke, et al. Medical Imaging 2006: Physics of Medical Imaging, p. 614231. San Diego, CA, SPIE (2006). P. Monnin, D. Gutierrez, S. Bulling, et al. Med. Phys. 34(3), 906 (2007). L. E. Antonuk, Q. Zhao, Z. Su, et al. Medical Imaging 2004: Physics of Medical Imaging, p. 127. San Diego, CA, SPIE (2004). D. L. Lee, G. Storti, and K. P. Golden. Penetrating Radiation Systems and Applications VI, p. 126. Denver, CO, SPIE (2004). D. L. Y. Lee. Medical Imaging 2005: Physics of Medical Imaging, p. 216. San Diego, CA, SPIE (2005). W. Zhao, D. C. Hunt, K. Tanioka, et al. Medical Imaging 2004: Physics of Medical Imaging, p. 150. San Diego, CA, SPIE (2004). W. Zhao, D. Li, A. Reznik, et al. Med. Phys. 32, 2954 (2005). W. Zhao, D. Li, J. A. Rowlands, et al. Medical Imaging 2008: Physics of Medical Imaging, p. 69130M. San Diego, CA, SPIE (2008). K. S. Karim, A. Nathan, and J. A. Rowlands. Medical Imaging 2001: Physics of Medical Imaging, p. 35. San Diego, CA, SPIE (2001). K. S. Karim, Y. K. Vygranenko, A. Avila-Munoz, et al. Medical Imaging 2003: Physics of Medical Imaging, p. 38. San Diego, CA, SPIE (2003). K. S. Karim, S. Yin, A. Nathan, et al. Medical Imaging 2004: Physics of Medical Imaging, p. 657. San Diego, CA, SPIE (2004). M. H. Izadi, K. S. Karim, A. Nathan, et al. Medical Imaging 2006: Physics of Medical Imaging, p. 61420T. San Diego, CA, SPIE (2006). W. Zhao, G. DeCrescenzo, S. O. Kasap, et al. Med. Phys. 32, 488 (2005). V. Loustauneau, M. Bissonnette, S. Cadieux, et al. Medical Imaging 2004: Physics of Medical Imaging, p. 162. San Diego, CA, SPIE (2004). M. Z. Kabir, M. Yunus, S. O. Kasap, et al. Medical Imaging 2005: Physics of Medical Imaging, p. 223. San Diego, CA, SPIE (2005). A. W. Rau, L. Bakueva, and J. A. Rowlands. Med. Phys. 32, 3160 (2005). W. Zhao, B. Zhao, P. R. Fisher, et al. Medical Imaging 2007: Physics of Medical Imaging p. 65101M. San Diego, CA, SPIE (2007). B. Zhao and W. Zhao. Med. Phys. 35, 1978 (2008). A. J. Dabrowski, J. Chwaszczewska, J. Iwanczyk, et al. IEEE Trans. Nucl. Sci. 23, 171 (1976). A. Niemela, H. Sipila, and V. I. Ivanov. IEEE Trans. Nucl. Sci. 43, 1476 (1996). C. Matsumoto, T. Takahashi, K. Takizawa, et al. IEEE Trans. Nucl. Sci. 45, 428 (1998).

578

PHOTOCONDUCTOR DIGITAL X-RAY IMAGING

[28] [29]

T. Takahashi, S. Watanabe, M. Kouda, et al. IEEE Trans. Nucl. Sci. 48, 287 (2001). Y. Tomita, Y. Hatanaka, T. Takabayashi, et al. IEEE Trans. Electron Devices 40, 315 (1993). S. Adachi, N. Hori, K. Sato, et al. Medical Imaging 2000: Physics of Medical Imaging, p. 38. San Diego, CA, SPIE (2000). S. Tokuda, S. Adachi, T. Sato, et al. Medical Imaging 2001: Physics of Medical Imaging, p. 140. San Diego, CA, SPIE (2001). S. Tokuda, S. Adachi, T. Sato, et al. Medical Imaging 2003: Physics of Medical Imaging, p. 861. San Diego, CA, SPIE (2003). H. Tsutsui, T. Ohtsuchi, K. Ohmori, et al. IEEE Trans. Nucl. Sci. 40, 95 (1993). M. Sowinska, G. Hennard, D. Feder, et al. IEEE Trans. Nucl. Sci. 49, 3306 (2002). Y. Eisen, A. Shor, and I. Mardo. IEEE Trans. Nucl. Sci. 51, 1191 (2004). W. C. Barber, K. Iwata, B. Hasegawa, et al. Nucl. Inst. Meth. Phys. Res. A 505, 595 (2003). C. Szeles, S. A. Soldner, S. Vydrin, et al. IEEE Trans. Nucl. Sci. 54, 1350 (2007). C. Szeles, S. A. Soldner, S. Vydrin, et al. IEEE Trans. Nucl. Sci. 55, 572 (2008). J. S. Iwanczyk, E. Nygard, O. Meirav, et al. Nuclear Science Symposium Conference Record, 2007, p. 2741. Honolulu, HI, IEEE (2007). G. Zentai, L. Partain, R. Pavlyuchkova, et al. IEEE Trans. Nucl. Sci. 53, 2506 (2006). J. Zhang, K. Shah, F. Olschner, et al. Nucl. Inst. Meth. Phys. Res. A 322, 499 (1992). V. Deich and M. Roth. Nucl. Inst. Meth. Phys. Res. A 380, 169 (1996). K. Shah, F. Olschner, L. Moy, et al. Nucl. Inst. Meth. Phys. Res. A 380, 266 (1996). T. Shoji, K. Hitomi, T. Tiba, et al. IEEE Trans. Nucl. Sci. 45, 581 (1998). R. A. Street, J. T. Rahn, S. E. Ready, et al. Medical Imaging 1999: Physics of Medical Imaging, p. 36. San Diego, CA, SPIE (1999). L. Fornaro, E. Saucedo, L. Mussio, et al. Nuclear Science Symposium Conference Record, 2001, p. 1476. San Diego, CA, IEEE (2001). S. Ross, G. Zentai, K. Shah, et al. Nucl. Inst. Meth. Phys. Res. A 388, 38 (1997). G. Zentai, L. D. Partain, R. Pavlyuchkova, et al. Medical Imaging 2003: Physics of Medical Imaging, p. 77. San Diego, CA, SPIE (2003). Y. Kang, L. E. Antonuk, Y. El-Mohri, et al. Nuclear Science Symposium Conference Record, 2003, p. 2119. Portland, OR, IEEE (2003). M. Schieber, I. Beinglass, A. Holzer, et al. IEEE Trans. Nucl. Sci. 24, 148 (1977). J. S. Iwanczyk, J. B. Barton, A. J. Dabrowski, et al. IEEE Trans. Nucl. Sci. 30, 363 (1983). J. S. Iwanczyk, A. J. Dabrowski, J. M. Markakis, et al. IEEE Trans. Nucl. Sci. 31, 336 (1984). M. Schieber, I. Beinglass, G. Dishon, et al. IEEE Trans. Nucl. Sci. 25, 644 (1978). M. Schieber, C. Ortale, L. Vandenberg, et al. Nucl. Inst. Meth. Phys. Res. A 283, 172 (1989). L. van den Berg. Nucl. Inst. Meth. Phys. Res. A 322, 453 (1992). B. Patt, R. Dolin, T. Devore, R. Dolin, et al. Nucl. Inst. Meth. Phys. Res. A 299, 176 (1990). L. van den Berg and R. Vigil. Nucl. Inst. Meth. Phys. Res. A 458, 148 (2001). L. van den Berg, J. S. Sandoval, R. D. Vigil, et al. X-Ray and Gamma-Ray Instrumentation for Astronomy XII, p. 100. San Diego, CA, SPIE (2002). L. van den Berg, A. Proctor, K. Pohl, et al. AIP Conference Proceedings, 632, p. 71 (2002). J. S. Iwanczyk, B. E. Patt, C. R. Tull, et al. Nuclear Science Symposium Conference Record, 2000, Vol. 1, p. 8. Lyon, France, IEEE (2000).

[30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60]

REFERENCES [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81]

579

J. S. Iwanczyk, B. E. Patt, C. R. Tull, et al. Penetrating Radiation Systems and Applications III, p. 28. San Diego, CA, SPIE (2001). M. M. Schieber, H. Hermon, R. A. Street, et al. Penetrating Radiation Systems and Applications II, p. 197. San Diego, CA, SPIE (2000). M. Schieber, H. Hermon, A. Zuck, et al. J. Crystal Growth 255, 118 (2001). M. M. Schieber, H. Hermon, R. A. Street, et al. Medical Imaging 2000: Physics of Medical Imaging, p. 48. San Diego, CA, SPIE (2000). M. Schieber, A. Zuck, H. Gilboa and G. Zentai. IEEE Trans. Nucl. Sci. 53, 2385 (2006). R. A. Street, M. Mulato, S. E. Ready, et al. Penetrating Radiation Systems and Applications II, p. 189. San Diego, CA, SPIE (2000). H. Hermon, M. M. Schieber, A. Zuck, et al. Medical Imaging 2001: Physics of Medical Imaging, p. 133. San Diego, CA, SPIE (2001). K.-W. Jee, L. E. Antonuk, Y. El-Mohri, et al. Medical Imaging 2000: Physics of Medical Imaging, p. 250. San Diego, CA, SPIE (2000). K.-W. Jee, L. E. Antonuk, Y. El-Mohri, et al. Medical Imaging 2001: Physics of Medical Imaging, p. 13. San Diego, CA, SPIE (2001). Y. Kang, L. E. Antonuk, Y. El-Mohri, et al. IEEE Trans. Nucl. Sci. 52, 38 (2005). Z. Su, L. E. Antonuk, Y. El-Mohri, et al. Phys. Med. Biol. 50, 2907 (2005). H. Du, L. E. Antonuk, Y. El-Mohri, et al. Phys. Med. Biol. 53, 1325 (2008). G. Zentai, L. Partain, and R. Pavlyuchkova. Medical Imaging 2007: Physics of Medical Imaging, p. 65100Q. San Diego, CA, SPIE (2007). G. Zentai and L. Partain. Health Monitoring of Structural and Biological Systems 2008, p. 69351Y. San Diego, CA, SPIE (2008). J. S. Iwanczyk, B. E. Patt, C. R. Tull, et al. IEEE Trans. Nucl. Sci. 49, 160 (2002). L. Fornaro, E. Saucedo, L. Mussio, et al. Nuclear Science Symposium Conference Record, 2002, p. 33. New York, NY, IEEE (2002). L. Fornaro, A. Cuna, A. Noguera, et al. Nuclear Science Symposium Conference Record, 2003, p. 3541. Portland, OR, IEEE (2003). L. Fornaro, H. Espinosa, A. Cuna, et al. IEEE Trans. Nucl. Sci. 52, 3103 (2005). M. Simon, R. A. Ford, A. R. Franklin, et al. Medical Imaging 2004: Physics of Medical Imaging, p. 188. San Diego, CA, SPIE (2004). R. A. Street, W. S. Wong, S. Ready, et al. Medical Imaging 2005: Physics of Medical Imaging, p. 7. San Diego, CA, SPIE (2005). L. E. Antonuk, Y. El-Mohri, Q. Zhao, et al. Medical Imaging 2008: Physics of Medical Imaging, p. 69130I. San Diego, CA, SPIE (2008).

PART VII SUMMARY

26 SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS LEWIS FRAAS1 AND LARRY PARTAIN2 1 JX Crystals Inc., 2Varian Medical Systems

26.1

PROGRESS

The First Edition of this book was published in 1995. This Second Edition is being published in 2010. This represents a 15-year interval in which the progress in solar cell electricity generation has been astounding. In this chapter, some of the statistics for the solar industry quoted from the concluding chapter of the First Edition are contrasted with current industry statistics. The world market for solar modules in 1992 was 65 MW. In 2008, the world market was 5.95 GW with over $10 billion in sales. This represents a market growth in gigawatts of nearly a factor of 100. In 1992, c-Si modules were dominant with 10% efficiency and mostly for off-grid industrial and residential applications. In 2008, c-Si modules with efficiencies of 19% from SunPower Corporation and 17% from Sanyo were commercially available, and the grid-tied market represented over 80% of the total. In 1992, module prices were approximately $6.5 per watt in small quantities corresponding to approximately $10 per watt in 2008 dollars. In 2008, small-quantity module prices had fallen to $4.30 per watt. Large-volume module prices in July of 2009 were below $2.80 per watt. In 1992, space cell efficiencies were at approximately 15%. In 2010, space cell efficiencies are as high as 30%, and MJ cell efficiencies for terrestrial concentrator applications have been measured at 41% with 38% cells commercially available. Most economic models indicate that solar cell module costs below $2 per watt are needed to make solar cell electricity directly competitive with the fossil fuel electricity currently produced by utilities. However, the modules are just one of the multiple elements involved in competitive solar cell electricity production as summarized further below. Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

583

584

26.2

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

CURRENT STATUS

In the following, the current status will be summarized for the technology and markets along with a discussion of predictive models at the system level for the cost of solar cell electricity in cents per kilowatt hour and the macroeconomics and timing for large-scale solar cell electricity generation.

26.2.1

Technology

The solar cell module options today are the same as they were in 1992. They are the c-Si planar modules, the TF planar module, and CMs. However, as noted above, there has been a remarkable improvement in module performance and in the volume of module sales. There has also been remarkable progress in the integration of these modules with the additional required components into complete fielded functional systems. The two planar module types dominate the commercial market with the planar silicon module market representing 85% to 90% of the total and the TF planar module market accounting for the remaining 10% to 15%. The ISE has measured the efficiencies for a large number of commercially available c-Si and TF modules as of 2007 and has reported the results as shown in Figure 26.1 [1]. As can be seen in this figure, the average efficiency for the c-Si case was 13%, with the best at 19.3%. The efficiencies for the TF case were approximately half of those for the c-Si case with an average TF module efficiency of 6.3% and the best TF module efficiency at 11.1%. The fact that TF modules have lower efficiencies relative to single-crystal materials is consistent with the arguments made in Chapter 3 for the importance of single-crystal materials for achieving high cell conversion efficiencies.

22

148 2007

1.0 TF ηm: 6.3

0.8 Si TF 0.6

CdTe

0.4

c-Si ηm: 13.0 19.3

11.1

CIS

0.2 0.0 0

2

4

6

8

10

12 14 16 ηmodule [%]

18

20

Figure 26.1. Normalized module efficiency distributions for crystalline silicon (c-Si) and thin-film (TF) modules for the year 2007 (Gerhard P. Willeke, PVSC, May 2008, ISE [1]).

CURRENT STATUS

585

The TF and c-Si module advocates represent two alternative groups. The TF module group emphasizes low module cost and the c-Si module group emphasizes module efficiency and reliability. There is a third group. The concentrator group argues that both low cost and high efficiency can be obtained by adding a low-cost concentrator element to leverage down the cost of the high-efficiency single-crystal cells. However, the problem here is that more elements are added, increasing system complexity and development cost. In addition, one now needs to point at the sun in order to focus the sunlight onto the cells. This has meant that system sizes of approximately 1 kW and larger are required in order for solar trackers to be affordable. This eliminated concentrators from the early remote off-grid market applications. However, the current solar cell market has grown and larger solar cell power systems are now commonplace. Nevertheless, new solar cell technologies evolve through stages from cell demonstration to integration into modules and then arrays. Array deployment follows for field testing followed finally by market penetration and high-volume production and then low cost. Concentrator technology has been moving through these development stages. In 1995, in the First Edition of this book, chapter 6 described the development and demonstration of the first 35% efficient MJ concentrator cells, and chapter 14 described the first integration of these cells into concentrator modules. The concentrator module described in chapter 14 was developed for a space flight demonstration, and it was successfully flown in 1994. This module is shown in Figure 26.2a.

(a)

(b)

Figure 26.2. The 200-W 28% efficient Japanese module (a) fabricated in 2004 was made to target the terrestrial solar electric power market. It resembles the Advanced Space Power Module (b) made by the Fraas team at Boeing in 1992 and described in the First Edition of this book.

586

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

Between 1995 and 2001, the focus was on improving MJ cells for space, and there was almost no work on concentrators. The focus for space was on the monolithic triple-junction InGaP/GaInAs/Ge cell because of its high power-to-weight (watt per kilogram) ratios (see Table 18.6, Chapter 18). Dramatic improvements were made in this cell with improvements in the growth technology and refinements in the required tunnel junctions. By 2000, this cell was the dominant cell for space satellites. In 2001, the Japanese government initiated the R & D project for super-high-efficiency concentrator MJ solar cells and modules for terrestrial applications. This led in 2004 to the world’s first 28% efficient solar cell module as shown in Figure 26.2b. The resemblance between the two modules in this figure is noteworthy. In order to fairly compare this 28% module efficiency with the 19% c-Si module efficiency, it should be noted that the 28% efficiency is referenced to the solar direct irradiance, whereas the 19% efficiency is referenced to the solar global irradiance. Referencing both to the global solar irradiance would reduce the 28% to approximately 24%. Nevertheless, this is a dramatic module efficiency improvement. It should also be noted that these MJ cells, developed initially for space applications, are quite expensive, and therefore, high-concentration point focus optics are required with two-axis precise trackers. This credible concentrator terrestrial module demonstration then led to the formation of several new companies planning to commercialize high-concentration solar cell modules, arrays, and field installations. Chapters 13–15 in this book provide more detailed descriptions of the latter activities by these companies. In a parallel concentrator activity, Amonix had been developing concentrators for utility applications using silicon cells. Amonix’s work pioneered concentrator field testing. In 2006, the ISFOC was formed in Spain with support from the Spanish government to support high-concentration solar cell module field testing and system qualification and certification. Chapters 16 and 17 in this book describe some of these recent concentrator field testing activities. In 2008, there were over 10 MW of c-Si-based concentrator systems installed in Spain greatly expanding on the pioneering work of Amonix. This 10 MW is small compared to the gigawatts of planar solar cell systems installed but is very significant compared to the status of this technology in 1995. Sales of high-concentration systems using the MJ cells are just beginning. The economic modeling of this Second Edition emphasizes that it is the cost per kilowatt hour that is the best single metric for evaluating the economic competitiveness for solar cell electricity, particularly for larger and utility-scale installations. For example, if two systems produce the same number of kilowatt hours of electricity per year but one has an efficiency of, say, 20% versus 10% for the other, the lower-efficiency system might only be economically competitive if its module cost is half of that of the more efficient one, everything else being equal (see Figure 2.9, Chapter 2). Of course, everything else is not equal. The two-axis trackers for high concentration cost more than that of fixed module supports for 1-sun planar module installations. However, as one- and two-axis tracking of planar modules is explored for large installations, the competitiveness and costs per kilowatt hour trade-offs are going to be determined by demonstration projects

CURRENT STATUS

587

and by the market response itself to factors beyond just the initial cost per watt of modules alone. The device physics in chapter 1 of the First Edition focused on low injection devices and their potential for substantial efficiency improvements using multiple junctions to reduce the loss of absorbed photon energy above the solar cell bandgap, particularly with sunlight concentration. The 41% efficient MJ concentrator cells are a striking example of the success of applying such concepts. The First Edition also contained descriptions of a range of solar cells from amorphous silicon to point- and back-contact crystalline cells with characteristics not well addressed by its device physics models. Low injection means that the light-generated carrier concentrations in the solar cells are higher than either the equilibrium (in the dark) conduction electron or hole concentrations but not both. High injection means these concentrations are higher than both. Such high injection treatment is a major addition to the device physics in Chapter 4 of this book. It focuses on the reduction of absorbed photon energy losses below the bandgap. This below-bandgap loss is described by a voltage factor, VF, which is the ratio of the open-circuit voltage to the bandgap. The projection is that the potential for improved efficiencies is performance multipliers between 1.18X and 1.54X beyond that of the best current devices, using high injection conditions in devices whose photoactive regions contain little or no doping. The predicted improvements come directly from making the voltage factor (VF) approach a value of 1. The challenge with high injection VF improvements will be handling the higher current densities involved with concentration so that gains in open-circuit voltage are not offset by losses in fill factor FF due to series resistance. Chapter 4 in this book also adds a detailed treatment of defect-dominated solar cells whose performance is limited by space charge accumulation. Typically, this results in lower fill factors (FF) as well as lower voltage factor (VF) values as compared with devices not limited by space charge accumulation and high defect levels.

26.2.2

Solar Cell Manufacturers

Table 26.1 summarizes the shipment by major solar cell manufacturers for the last 3 years. One cannot help but note that the leaders have been helped by the various government subsidies. For example, in 2005, four of the top 10 were Japanese companies and there were no U.S. companies in the top 10. Sharp Solar and Kyocera were numbers 1 and 2. This correlates with the Japanese government support level in 2000 with a solar cell budget of $450 million compared to the U.S. government support level of $70 million. The launch of the German government solar Feed-In Tariff Program in 2004 then led to the rise of Q-Cells from number 3 to number 1 in 2008. The Chinese government began supporting solar cell technology in 2001. Chapter 8 herein gives an excellent review of the well-coordinated development of the Chinese c-Si solar cell industry. By 2006, as a result of the Chinese government support, Suntech was number 4, and by 2008, Suntech had risen to number 2 and JA Solar appeared at number 9.

TABLE 26.1. Shipments by Major Solar Cell Manufacturers 588

Ranking

2005

2005 MWp

2006

2006 MWp

2007

2007 MWp

2008

2008 MWp

Top Ten 2008 Technology c-Si

1

Sharp Solar

375.2

Sharp Solar

434.7

Sharp Solar

363.0

U-Cells

547.2

2

Kyocera

142.0

Q-Cells

240.4

Q-Cells

344.1

Suntech

497.5

c-Si

3

Q-Cells

131.2

Kyocera

100.0

Suntach

309.0

Sharp

450.0

c-Si/a-Si

4

Scholt Solar

95.0

Suntech

152.0

Kyocera

207.0

First Solar

434.7

CdTc

5

BP Solar

85.8

Sanyo

120.5

First Solar

486.0

Kyocera

290.0

c-Si

6

Mitsubishi Electric

85.0

Mitsubishi Electric

111.0

Moloch

167.0

Moloch

263.5

c-Si

7

Sanyo

84.0

Schott Solar

96.0

Sanyo

155.0

Sanyo

220.0

c-Si/a-Si

8

Shell Solar

55.0

Molech

91.5

SolarWorld

136.1

SunPower

215.2

c-Si

9

Matech

45.0

BP Solar

78.0

Mitsubishi Electric

121.0

JA Solar

212.1

c-Si

10

Isofoten

39.3

SunPower

62.7

10a 10b

SunPower

102.0

BP Solar

148.1

c-Si

JA Solar

102.0

Mitsubishi Electric

148.0

c-Si

BP Solar

101.5

Total Year Shipments

1407.7

1984.6

3073.0

Sum top 10 above

1137.5

1557.1

2293.8

81%

79%

75%

Top 10% Total Shipments

5491.8 3434.5 63%

CURRENT STATUS

589

U.S. government support for solar cells has focused largely on the TF option. As a result, First Solar is listed as number 4. SunPower Corporation, a U.S. company, is listed as number 8 as a result of its technical innovation with highefficiency c-Si modules. It is interesting to note that the Chinese and German companies are profiting based mainly on the solar cell and module technology pioneered in the United States in the late 1970s, whereas the cell and module performance innovations are still occurring primarily in the United States. Still, the German and Chinese companies are making major contributions and advances in manufacturing. Technical innovations by SunPower Corporation in the United States and by Sanyo in Japan in the c-Si area are particularly noteworthy as described in this book’s Chapters 2, 4, and 5. Government support has also influenced customer purchasing. The German Feed-In Tariff, launched in 2004, and the Spanish Feed-In Tariff, launched in 2007, have led to the high market demands shown in Figure 26.3, where installations in Spain and Germany during 2008 reached 2.46 and 1.86 GW, respectively. It is striking to note the solar cell manufacturing ranking of the United States falling from first in 1990 to fifth by 2008. In 2008, the market leaders were Japan successively followed by Germany, China, and Spain. There is a clear correlation between the rise in these leadership positions and the support such programs received from their national governments. As was shown in chapter 22 of the First Edition, U.S. government support for solar cell development fell from its peak of $150 million in 1980 down to under $50 million in 1984, hitting a bottom of $35 million in 1988 through 1990. As noted above, the U.S. annual support level in 2000 was still only $70 million. PV Market Demand in 2008 Total: 5.95 GW

Rest of the world (row), 0.21 GW

Rest of Europe United States, (ROE), 0.31 GW 0.36 GW

Japan, 0.23 GW Italy, 0.24 GW South Korea, 0.28 GW

Germany, 1.86 GW

Spain, 2.46 GW

Figure 26.3. Solar cell (PV) market demand in 2008. Source: Solarbuzz LLC.

590

26.2.3

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

Opportunities

There are several bold opportunities, three of which are summarized below. The first is for solar modules with high efficiency and low cost simultaneously (19% and $1 per watt). The second is joint government cooperation, and the third is to combine the development of electric cars with solar-generated electricity for battery charging in employee parking lots to reduce oil imports directly. As just noted, the TF and c-Si module advocates represent two alternative groups. The TF group emphasizes low module cost and the c-Si module group emphasizes module efficiency and reliability. There is a simple evolutionary way to combine the goals of both groups. The goal of the TF group has been to reach a module cost of $1 per watt. Unfortunately, TF efficiencies are limited. Meanwhile, the c-Si technology has demonstrated cell efficiencies as high as 23% with 19% efficient modules and c-Si modules proven to last for 30 years. Unfortunately, the single-crystal cells are expensive and module prices after 30 years of development are still a factor of 2 too high with the lowest commercial c-Si module price in mid-2009 being $2.80 per watt (according to the Solarbuzz website in July of 2009 [2]). Chapter 12 herein suggests a solution by simply using three times less of the single-crystal Si in a module using mirrors for low 3X sunlight concentration (see Fig. 26.4). The simple economic argument is the following. If the difference in module cost and price for the $2.8 per watt module is 30%, the c-Si module cost is then approximately $1.96 per watt. Then, if the cost of everything other than the cells

Figure 26.4. The drawing on the left shows the standard c-Si planar module. The drawings in the center and on the right show a planar concentrator module with one-third as much expensive silicon. Since mirrors are 10 times cheaper than single-crystal cells, the module cost can be reduced dramatically while maintaining the high module efficiency.

CURRENT STATUS

591

in that module is $0.46 per watt, the single-crystal cells alone will cost approximately $1.50 per watt. These numbers are consistent with numbers presented at the 34th PVSC by Fath [3], where he projected a planar Si module cost in 2015 of $1.80 per watt with a Si cell cost of $1.35 per watt. Now note that a 3X low concentration module using one-third of the crystalline Si allows the Si cell cost to be divided by 3. It is then only $1.35/3 = $0.45 per watt. The key is that mirrors are 10 times cheaper than single-crystal cells. Noting that mirrors will cost $0.15 per watt and adding the new cell cost to the balance of module and mirror costs, one obtains a projected module cost of $0.45 + $0.46 + $0.15 = $1.06 per watt. This is the module cost goal, but now the module can use 23% efficient cells, allowing a module efficiency of 21% referenced to the direct solar irradiance. Correcting for some loss in the diffuse skylight irradiance and referencing this low concentration module efficiency to global solar irradiance, one should obtain a module efficiency of 19% rivaling that of the planar c-Si module but with approximately twice the conversion efficiency of the TF module case. Low concentration therefore represents an immediate and substantial step toward the ultimate goal of low cost and high efficiency simultaneously. The result is $1 per watt and 19% near term. The concept is both simple and straightforward. However, funding for the development of this option is required. A second opportunity can come from well-coordinated government economic development programs. The political entities that can best afford the estimated $10 billion level investments needed to assume leadership roles in the rapidly evolving world market for solar cell electricity are those backed by the largest coordinated economies. In current order of size, these are (1) the European Union, (2) the United States, (3) China, (4) Japan, and (5) India [4]. The market history-proven best public policy approaches for improved rankings in the world solar electricity market, in effectiveness order, are (1) feed-in tariffs, (2) depleted energy user tax revenues applied to renewable energy, (3) government-funded incentives that pay a substantial portion of the capital costs in the first year of installation, (4) longterm (30 years) low-interest rate loans often funded with tax-free bonds, (5) net metering particularly when extended to total metering where excess electricity delivered is paid back, and (6) other tax incentives that include loan interest writeoffs at constant interest rates over 30-year time periods. The European Union is beginning to demonstrate that different countries with adjacent borders, speaking various languages, and with a spectrum of individual cultures, economic development levels, and technology competencies can join into cooperative trade, economic, consumer, and political arrangements that thrust such “unions” into a world leadership position, as ranked by economic metrics in general and by renewable energy implementations and deployments in particular. At its peak, the British Empire demonstrated that a dominant economic power base and trading entity could be assembled from widely spread countries whose participations were established and maintained through military “support.” A key challenge of the twentyfirst century may be whether other “adjacent” and “nonadjacent” but diverse countries can cooperate through similar unions motivated primarily by economic self-interest and mutual well-being.

592

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

Over the last several decades, rising oil prices have led to major economic disruptions. Two notable examples in the United States were the supplier boycottcaused gas lines in 1973 and the recent spurt in world economic growth, led by China and India, where world demand outstripped gasoline supplies, temporarily interrupted now by a credit crisis-induced worldwide recession. Low-cost solargenerated electricity represents an opportunity to reduce oil consumption and imports and to create whole new market applications, particularly in regions with rapidly growing economies that are receptive to new technology solutions. Batterypowered cars offer major reductions in transportation’s gasoline consumption and in its carbon footprint when such cars are charged with solar cell electricity during the day in employer parking lots while their owners work. This is illustrated in Figure 26.5 where a battery-powered Tesla Roadster is parked in front of the City of Palo Alto’s Municipal Service Center in California. Tesla reports that a full Roadster charge of ∼23 kWh provides about 220 miles range, equivalent to 50 mile commutes to and from work for four employees. At a cost of 10¢ per kilowatt hour, such a solar “fill-up” would cost $2.30. Each of the seven single-crystal silicon tracking arrays at this Service Center provides a peak electrical output of about 2.4 kW (12 SunPower 205-W panels per tracking array). From their measured outputs (L. Joye, pers. comm.) of 87 kWh/day (averaged over the year), these seven trackers could provide the electric power to transport ∼17 employees to and

Figure 26.5. A battery-powered automobile parked in front of two-axis tracking arrays illustrating the opportunities for solar cell electricity to provide personal transportation when recharging occurs in employer parking lots while employees work during the day. (Photograph courtesy of Deborah Partain.)

CURRENT STATUS

593

from work each day gasoline free (assuming roundtrip commutes of 50 miles each traveling alone). Widespread use of solar daytime charging for personal transportation to and from work would need utilities that collect solar cell electricity produced from surrounding areas that is immediately delivered to employer parking lots. Progressive employers could provide this charging as a perk to those using batterypowered cars, starting with 5 or 10 charging spaces and expanding as employee demand requires. Alternatively, each charging station could be credit card activated to provide employees with very attractive and comparatively low “fuel” costs for commuting. Particularly attractive could be rooftop solar cell electricity panels (averaging about 4 kWp per installation) on home rooftops that generate most if not all of the power needed to get their occupants to and from work while reducing their home electricity bills by 90% or more due to the solar offsets currently provided by electric utilities. From the measured 15 kWh/day for a 4 kWp system (averaged over 341 consecutive days from 4 kWp of single-crystal silicon Evergreen ES-180-SL modules installed at 5° fixed tilt over the parking lot of the Palo Alto Municipal Services Center), the solar electricity produced by such a home system each day would be enough to provide ∼144 miles (varying a factor of ∼2 from December to July) of driving daily in a battery-powered car with the reported Tesla Roadster performance characteristics. An excellent match for residential rooftop generated solar electricity is its use for immediate charging of battery-powered commuter cars or for more general transportation uses like charging batterypowered commuter buses. Accelerated reduction in transportation’s gasoline consumption and carbon footprint will come from incentives that encourage employers to provide electric charging outlets in parking lots and to use onsite solar cell electricity for parking lot charging when that produces the largest reductions. Additional acceleration will come from promoting purchases of battery-powered cars and from home installation of rooftop solar cell panels that offset the electricity used for daytime car charging. At this point in time, the United States enjoys the lead in battery-powered automobiles in contrast to the leads located elsewhere for petroleum-powered automobiles including hybrids.

26.2.4

Measured Field Performance

To assess wider applicability of solar cell-produced electricity, consider the summary of actual measured field performance of installed systems not only at the above Palo Alto, California site but also at other sunny locations as listed in Table 26.2. Here, averaged field efficiency values are reported that are simply the AC kilowatt hour per year produced by a given system divided by the kilowatt hour per year of sunlight striking the system for that year as measured by on-site heliometer instrumentation. The latter on-site heliometer data were not available for the Palo Alto, California system nor for the Springerville, Arizona system described in Chapter 10. Hence, the STC efficiencies are listed for lack of better values even

594

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

TABLE 26.2. Summary of Field Performance Data for Five Sunny Locations Location

Type

Averaged Field Efficiency (%)

Averaged (kWh/ day/kWp)

Averaged (kWh/ day/m2)

Year Measured

Las Vegas, Nevada 36° latitude

Concentrator two-axis tracking

19.0a

6.89a

1.15a

2009

Puertollano, Spain 38° latitude

Concentrator two-axis tracking

19.0b

5.88b

0.969b

2009

Palo Alto, California 37° latitude

1-sun Si two-axis tracking

17.3d

5.26

0.911

2009

1-sun Si 5° fixed tilt

12.1d

3.79c

0.457c

Springerville, Arizona 34° latitude

1-sun Si 34° fixed tilt

12.3d

4.74

0.583

2004–2006

San Ramon, CA 37° latitude

Concentrator Gallium Arsenide two-axis tracking

15.2e

6.38e

0.867e

1983–1984

1-sun Si 30° fixed tilt

9.9e

5.40e

0.561e

1-sun Si 30° fixed tilt

6.2e

5.35e

0.348e

a

Calculated from 6 months January–June measured data with Wp determined under AC PVUSA conditions. b Preliminary initial 12-month start-up data (underestimates). c Average of 341 consecutive days starting March 12, 2009. d Overestimates of averaged field efficiency from use of STC efficiency values. e Converted from DC to AC assuming 95% inverter efficiency. GaAs, gallium arsenide.

though they overestimate actual field performance (see Table 15.1, Chapter 15). STC efficiencies are readily calculated from manufacturer specification sheets by taking the STC kilowatt rating of the module at a 25°C solar cell temperature and by dividing it by the exterior dimensions of the module in square meters and then by dividing again by the 1 kW/m2 incident sunlight intensity specified for this standard. The kilowatt hour per day per kWp (STC) metric is directly converted into the kilowatt hour per day per square meter metric by multiplying the former by this kWp (STC) per square meter ratio taken directly from the manufacturer’s specification sheet. The San Ramon, California field performance data are from Hester and Hoff [5].

CURRENT STATUS

595

The highest averaged field efficiencies of 19% were for the recent two-axis tracking concentrator installations in Las Vegas and in Spain. The Las Vegas system is a three-junction Amonix system described in Chapters 13 and 16. The identity of the Spain system (see Chapter 17) has not yet been released, but its high performance suggests that it is likely to also be a three-junction configuration. These 2009 performance levels are markedly higher than the 1983–1984 performance levels for both two-axis tracking concentrator and fixed-tilt 1-sun systems in San Ramon, California [5], and they confirm the strong performance improvements achieved since that time. Of strong interest is the yearly average of kilowatt hour per day of power produced per STC peak kilowatts of the systems, since purchase prices are typically in terms of dollar per STC Wp. These values tell users how much electricity they can expect for their money. The numbers for Palo Alto illustrate the 39% (=5.26/3.79) higher energy produced by the tracking system. The Palo Alto site’s biggest difference was in yearly averaged kilowatt hour per day per square meter with a 99% (=0.911/0.457) advantage for the tracking system. This additional benefit (beyond tracking) is due to the higher efficiency of the latter (17.3% vs. 12.1% STC). It is installations with significant area-related BOS costs where the latter metric is the key and where differences in efficiency provide significantly different levelized cost of energy (electricity) (LCOE) cost values as discussed in more detail below. While field system performance comparisons are only rigorously accurate for side-by-side systems at a single site, nevertheless, some important trends can be discerned. The kilowatt hour per day per kWp is within a comparatively small range for all five sunny sites particularly when the higher numbers for concentrators are adjusted for their division by a smaller solar input level (direct normal sunlight for concentrators vs. global for 1-sun systems). Also two-axis tracking plus higher efficiencies provide roughly 1.0 kWh/day/m2 performance levels versus ∼0.6 kWh/day/m2 for lower-efficiency fixed-tilt systems. This includes the 3 years of operational experience in Springerville, Arizona for edge-defined-film-fedgrowth single-crystal Si ASE-300-DG/50 modules of 12.1% STC efficiencies from ASE Americas (now Schott Solar). However, for very low 6.2% efficiency fixedtilt systems (from polycrystalline Si modules in San Ramon, California), the latter metric can fall below 0.4 kWh/day/m2 to about three times less than the best values. The latter is likely representative of the expected performance of other similarly low-efficiency technologies in similar sunny locations. The 0.911 kWh/day/m2 performance of the two-axis, 1-sun Si system in Palo Alto suggests that high-efficiency 1-sun systems are currently directly competitive with the three-junction concentrator systems particularly if their reported module efficiency improvements (from 17% to 19% STC) can be verified in large field installations. However, concentrator systems are in their very earliest stages of development in terms of total development effort magnitudes. Thus, they may have more opportunities for faster rates of improvement than the very mature 1-sun solar cell module technologies. To first order, differing solar cell technologies should produce about the same amount of energy per year per Wp STC at a given site under similar fixed-tilt or tracking conditions (see Fig. 6.2, Chapter 6) such as those listed in the “averaged

596

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

kilowatt hour per day per kWp” column of Table 26.2. This ignores second-order effects like different temperature coefficients, bandgap values, and inverter efficiencies. Similarly, first-order estimates of the energy density per unit area can be obtained by scaling with the ratios of STC efficiencies. Hence, a 10% STC device (e.g., CdTe) would be expected to have a Table 26.2 “averaged kilowatt hour per day per square meter” column value that is 58% (=10/17.1) of that listed for twoaxis tracking in Palo Alto, California. Similarly, the approximate reduction factors to listed values would be 83% (=10/12.1) and 81% (=10/12.3), respectively, for a 10% module at 5° fixed tilt in Palo Alto or at 34° fixed tilt in Springerville, Arizona. Hence, if the 10% module is substantially lower in cost per Wp, it might well provide comparable LCOE values with standard ∼12% crystalline Si modules when used for installations with significant area-related BOS costs. Such projections are useful until real field data become available and are reported for the lower dollar per Wp installations. As yet, there is no universally accepted equivalent to STC efficiency for concentrator modules, but strong recommendations for this are made in Chapter 17. As a calibration for an annual averaged 1.0 kWh/day/m2 energy output number, this corresponds to a site mean sunlight intensity of 1 kW/m2, an average of 5 h/day if the system field efficiency fraction ηF equals 0.2 (i.e., a 20% averaged field efficiency). Furthermore, if one knows the sunlight energy density SEY available at any site in kilowatt hour per square meter per year and an accurate estimate of the field efficiency fraction ηF, then the kilowatt hour per square meter of electricity that will be produced by that system at that site each year is just the product SEYηF. It is the averaged field efficiency that is needed to accurately predict installation energy output per year. Both of the annual averaged power output metrics, one normalized per peak STC kilowatt and the other per square meter, are very important for the determination of LCOE over the lifetime of the various systems. Approaches for calculating LCOE that do not readily incorporate both power and area-related BOS costs cannot provide accurate comparisons among systems of widely differing efficiencies when area-related costs are significant. An example of a program that does not yet provide convenient inclusion of area-related cost elements is the SAM program of Chapter 20.

26.2.5

Uncertainties and Unknowns

Because the solar cell electricity option has a high up-front capital cost, there has been a focus on the module capital cost expressed in dollar per watt. However, the economics of solar cells is really more complex than simply the module dollar per watt number. There is both good news and bad news beyond the dollar per watt number. The good news is that over time, once a solar cell system is installed, there will be no additional fuel cost. This saves a lot of money over the 30-year hoped-for operational lifetime of the system. The bad news is that there are additional system costs and the complete system needs to last for 30 years. The

CURRENT STATUS

597

economics is discussed further in the next section. The point here is that while we now know a lot about cells and modules, the technology has now progressed to the point where complete systems are being defined and installed. At the system level, there are still uncertainties and unknowns. As described in Chapters 5, 7, 10, and 11, there are fewer uncertainties and unknowns for the older c-Si technology. However, there are still uncertainties and unknowns for the newer TF (Chapter 6) and concentrator options. Chapters 13–17 (and Table 26.2 above) describe early field test data for the high concentration option. Unfortunately, as of this writing, there is little field data publicly available for complete TF systems. How long will they last and how much electricity will they produce under realistic field conditions? For the TF amorphous Si case, their degradation in place is well-known. Currently, First Solar is the dominant TF module provider using CdTe material as its photoactive element. As shown in Figure 26.6, there are questions about the Te supply as production volume increases [6, 7]. Note that the abundance of Te is similar to the abundance of gold (Au). A key question arises. How many megawatts per year can be produced with CdTe TF solar cells and how does that compare with the electric power needs? According to the USGS [6] and Arizona State Geologist Lee Allison [8], the world produces anywhere from 160 to 215 tons of Te a year.

Abundance, atoms of element per 106 atoms of Si

Availability 109 O

Si Al Na

106 H

Mg

C

103

10–3

10–6

Ca Fe Ti Mn

Rare earth elements Ce B Rb Nd VCr Nb Ga Pb Be Sc Co Sn La SmGdDy Er Hf Ni As Y Yb Ta Pr Cs Ge W Tl Eu Br Mo Cd SbI Tb Ho Lu Tm Ag Bi Se Hg In Major industrial Ru Te metals in Bold Au Pd Re Pt Ph Rarest “metals” Os Precious metals Ir in Italic

Li

N

1

K

P S Cl

F

Relative abundance of the chemical elements in Earth’s upper continental crust

Rock-forming elements

10

20

Cu Zn

30

Sr Zr

Ba

40 50 60 Atomic number, Z

70

80

Th U

90

P.H. Stauffer et al, Rare Earth Elements - Critical Resources for High Technology, USGS (2002)

Figure 26.6. Tellurium (Te) is as rare as gold (Au).

598

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

Te was traditionally used in metal alloys. Currently, demand from emerging new applications, like DVDs, digital camera, computer flash memory, and CPU thermoelectric cooling, among other things, has caused a severe shortage in recent years. This drove the price from below $4 a pound to over $100 in 2006, according to Lee Allison. It is plausible that investors may sense a shortage and may start to hoard Te supplies, running up prices even further. How much Te does First Solar use? They consume about 7 g of Cd and about 8 g of Te in each of their 2 ft × 4 ft CdTe solar panels. That is roughly 135 tons per each 1 GW of production. So a production rate of 1 GW/year of CdTe cells would consume all of the world’s annual production of Te. At 1 GW/year over 20 years, the CdTe installed module capacity could then grow to 20 GW, but this would be only 2% of the present U.S. electricity production capacity of 1000 GW. Is the USGS wrong on Te abundance and will there be a discovery of a large source of Te for the future of this TF solar PV option? This represents an area of uncertainty. There is no such problem for the silicon and aluminum needed for the low concentration opportunities described above.

26.2.6

Prices Today and Cost Model Predictions

The Solarbuzz website tracks low-volume solar cell module retail prices. Table 26.3 summarizes the lowest module prices available as of July 2009. The average module prices were somewhat higher. Notice that the TF module pricing is now below that of the c-Si module pricing. This reflects the fact that single-crystal cells are intrinsically the most expensive part of a c-Si module. Recently, First Solar has advertised that its module costs have now fallen below $1 per watt. However, based on its reported annual revenue for 2008 of $1.25 billion for its total sales of solar module power of 435 MW and by dividing these two numbers, one obtains a cost per watt of $2.87. This might seem inconsistent until one realizes that a solar system installed price involves much more than just the module price. In the case of this $2.87 per watt, some revenue resulted from system installation services. The fraction of revenues attributable to module sales has not been publicly disclosed. Historically, the focus of the solar industry has been on module cost expressed in dollar per watt. However, the focus is now shifting to system cost in dollar per watt. Solar electric systems cost are much higher than simply solar module costs

TABLE 26.3. Lowest solar cell module retail prices (July 2009) Lowest monocrystalline Si module price

$2.80/Wp ( 1.99/Wp)

Lowest multicrystalline Si module price

$2.48/Wp ( 1.76/Wp)

Lowest thin-film module price

$1.76/Wp ( 1.25/Wp)

CURRENT STATUS

599

Average Installed Cost (2007$/W)

and, as shown in Figure 26.7, the lower efficiency TF systems are still more expensive than c-Si systems based on the latest available data in 2007 [21]. The Solarbuzz website also provides data on complete installed solar cell system prices as shown in Table 26.4. As can be seen from this table, the installed system price is always much greater than the module price, and it depends strongly on the system size. The commercial- and utility-sized systems are less expensive than the smaller residential systems. However, the most important number for evaluating the economics of solar cell electricity is really not the capital system cost in dollar per watt but the actual delivered energy cost in cents per kilowatt hour. Solar energy marketing is going to need to shift to this LCOE metric. Various methods of calculating LCOE have been discussed in this edition’s Chapters 2, 3, 11, and 20. In chapter 16 of the First Edition, the EPRI presented an important equation for calculating the LCOE for solar cell electricity systems. This equation is repeated in Figure 26.8 below along with nine important input variables required in order to calculate a numerical value for the LCOE. These nine are important because they highlight the fact that vertical integration and cooperation is required among

12

Rack-Mounted Systems Installed in 2006 or 2007 Avg. +/– Std. Dev.

10 8 6 4 2 0

$8.1

$8.6

Crystalline Thin-Film n=13882 n=1166 64 MW 6 MW < 10 kW

$7.8

$7.9

CrystallineThin-Film n=1739 n=129 35 MW 3 MW 10 – 100 kW

$7.1

$7.7

Crystalline Thin-Film n=178 n=39 59 MW 13 MW >100 kW

Figure 26.7. Comparison of installed costs for crystalline versus thin-film systems.

TABLE 26.4. System Level (Dollar per Watt) from Solar Buzz July 2009 Solar I

Solar II

Installed Home System

Solar III

Installed Commercial System Installed Industrial System

On grid or off grid (2 kW)

On grid (50 kW)

On grid (500 kW)

Customer price

$17,394

Customer price

$327,438

Customer Price

$2,373,260

$/W

$8.70/W

$/W

$6.55/W

$/W

$4.75/W

600

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

Leveled Cost Of Electricity L = (1+r)(Cm+Cb)Fs/ηsSha + (1+r)CiFi / ha + O&M The 9 key variables are: 1. ηs = PV system conversion efficiency 2. Cm = PV module cost ($/m2) 3. Cb = area related BOS including installation ($/m2) 4. S = Site specific solar intensity (kW/m2) 5. ha = Annual solar hours for PV system (tracking) 6. Ci = Inverter cost in $/kW 7. F = Fixed charge rate (converts initial investment into annualized charge) 8. r = Indirect cost rate (permitting, NRE) 9. O&M

Figure 26.8. An LCOE expression and input parameter summary.

a large number of diverse groups in order to bring down the price of solar electricity in terms of cost per kilowatt hour. For example, TF companies emphasize low-cost modules, but this is just the Cm term in the LCOE equation. The module supports, field wiring, and installation costs can be higher when the module efficiency is low because more modules need to be installed. This is the Cb term in this equation. The sunlight, S, available at the location is certainly important. Following the sun by tracking the modules will increase the number of hours per year of operation, which is the ha term. For comparison, the product Sha equals SEY, the annual sunlight energy density of a site from Chapter 2. Increasing the annual hours of operation will reduce the impact of the inverter cost, Ci. Note also that the cost of the hardware and installation are not the only costs. The projects have to be financed by the banks, and this is the finance (F) term. Government permitting is also required, and this can cause delays increasing costs, and this is part of the project-specific overhead (r) term. Finally, the system will need some maintenance over time. In chapter 16 of the First Edition, the EPRI authors gave a table of the LCOE equation parameter values for their projections for the future TF case. These projected values are repeated here in the left number column in Table 26.5 along with our current projected values for the c-Si and low concentration cases. This allows a comparison of the projected LCOE values achievable for the cost of solar cell electricity in the fairly near term. The results are exciting with the low concentration option’s projected cost of solar electricity at 8.6¢ per kilowatt hour given volume scale-up. Notice that the BOS (BOS = Cb) number is lower for the TF case than for the c-Si or low concentration cases. This is because one-axis tracking is assumed for the c-Si and low concentration cases. Because of the lower efficiency for the TF case, tracking is not affordable. Tracking for the latter two cases then gives more annual operating hours.

CURRENT STATUS

601

TABLE 26.5. Realistic EPRI Equation Parameters for Thin-Film, c-Si, and Low Concentration Systems Variable

Module cost (Cm) BOS (Cb) (Cm + Cb) Module efficiency (η) Module power Annual hours (ha) S haη

Value

Comment

Thin Film

c-Si

$133/m2

$400/m2

$224/m2

100 MW/year

2

$100/m2

—

2

2

—

2

$29/m

Low Concentration

$100/m

2

$162/m

$500/m

$324/m

10%

19%

19%

100 W/m2

190 W/m2

190 W/m2

2321

3000 2

S at 1 kW/m2

3000 2

— 2

232 kWh/m

570 kWh/m

570 kWh/m

0.698

0.877

0.568

$234/kW

$234/kW

$234/kW

0.10

0.078

0.078

—

Fixed charge rate (F)

10.3%

10.3%

10.3%

EPRI 1995

Indirect cost (r)

0.225

0.225

0.225

EPRI 1995

F(1 + r)

0.126

0.126

0.126

—

System cost (¢/kWh)

10.0

12.0

8.1

—

O&M (¢/kWh)

0.5

0.5

0.5

—

¢/kWh

10.5

12.5

8.6

—

(Cm + Cb)/S haη Power-related BOS (Ci) Ci/ha

per year — EPRI 1995

In Table 26.5, the area-related cost variables for solar cell systems are logically given in dollar per square meter and the power converter cost is given in dollar per kilowatt. However, the solar sales and customer communities are accustomed to quotes in dollar per watt. Table 26.6 translates the key parameter assumptions given in Table 26.5 into dollar per watt to facilitate comparisons in that customary terminology. The numbers in this table are revealing. For example, the $3.52 per watt price for an installed c-Si solar cell system is within sight in the relatively near future from today’s $4.75 per watt large system price. It is also quite straightforward to see the evolutionary developments associated for the low concentration implementation to arrive at a system price of $2.39 per watt. This system capital equipment price then implies an LCOE price for solar electricity of 8.6¢ per kilowatt hour, which would be competitive for commercial customers now paying retail prices for electricity at over 10¢ per kilowatt hour.

602

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

TABLE 26.6. Translation of Key Parameter Assumptions of Table 26.5 into Customary Terminology

Module cost (Cm)

Thin Film ($/W)

c-Si ($/W)

Low Concentration ($/W)

1.33

2.10

1.18

BOS (Cb)

0.29

0.53

0.53

(Cm + Cb)

1.62

2.63

1.71

(Cm + Cb + Ci)

1.85

2.86

1.94

(1 + r) (Cm + Cb + Ci)

2.22

3.52

2.39

Note that it is the metric AC kilowatt hour per square meter per year that is required for accurate LCOE calculations regardless of whether 1 sun or concentrating systems are used in fixed tilt or with one- or two-axis tracking (see row 6 in Table 26.5). This metric is easily available for any site that records the AC kilowatt hour produced in a year. This is then divided by the area of the system, which is the external dimension of the modules multiplied by the number of modules in the installation. Site design- and system design-specific factors, like ground coverage ratios (see Chapter 11), are included in the area-related BOS cost factor. Procedures for calculating LCOE have been discussed in several chapters in this book. While the various authors’ terminology changes, the concepts and required inputs and derived results are similar to those from the EPRI equation.

26.2.7

Applications and Markets

In the summary discussion in this chapter so far, the focus has been on solar cell modules for the various electric power markets. In the last section, it was noted that this market can be segregated into the residential, commercial, and industrial or utility sectors. However, for the TF systems, interesting spin-off applications have developed, such as for medical X-ray imaging and video displays (LCDs). The X-ray applications were described in more detail here in Chapters 22 through 25. Two properties of amorphous silicon solar cells make them well suited for X-ray imaging. First is their lack of crystal structure, which makes them very resistant to X-radiation damage (up to megarad doses). The second is their relatively high bandgap and accompanying low dark current, which is particularly valuable in radiography. Devices dominated by resistive effects (photoconductors) can also offer particular advantages for such applications. A third and major advantage is their ability to be fabricated as large-area electronic imagers on glass substrates at commercially competitive costs. The 2009 digital X-ray flat panel imager market was $2 billion and growing at a 15% per yearly rate. Continued growth at the rate for the next 15 years would correspond to a $16 billion annual market size.

RECOMMENDATIONS

603

26.2.8 Government Program Investments and Solar Cell Electricity Growth The macroeconomics of solar cell electricity and its timing are becoming increasingly attractive. The historical government program sizes needed to change world solar cell market leadership positions are an estimated ∼30% of the cumulative market size, soon approaching the $30–$100 billion levels (see Fig. 2.7). Fortunately, the increased revenues produced typically equal such investment magnitudes over the following 3 or 4 years. The time required to reach the 100- to 1000-fold increase in market size and to supply substantial fractions of the world energy markets is on the order of 15–20 years (see Fig. 2.1), should current trends continue. For magnitude and timing comparisons, the costs of the Alaskan oil pipeline exceeded 23 billion (in 2007 U.S. dollar [9]), and the first oil flow through the pipe in 1977 was 9 years after the first gusher oil well demonstration in Prudhoe Bay, Alaska [10]. This project created on the order of 10,000 jobs in Alaska and produced a long-lasting increase in the Alaskan economy. Its environmental impacts were substantial, including the loss of at least 32 lives, but the disruptions and responses to the 1973 oil embargo made its construction possible, all with private financing in a very mature industry. However, the 35-year construction of the U.S. Interstate System completed in 1991 cost $129 billion with $114 billion (all in then current U.S. dollar) paid by the U.S. Federal Government from highway user taxes [11]. It proceeded because its construction was deemed “vitally important to national goals.” For further magnitude comparisons, the cost of oil and oil product imports to the United States in 2007 was $327 billion [12]. Should future oil embargos or other supply disruptions encourage direct military interventions to preserve deliveries, recent U.S. experience in Iraq indicates 5-year costs can be on the order of US$450 billion [13] with strong environmental impacts including loss of life exceeding 10,000. There are at least five world economies large enough to make these levels of investment into solar cell electricity technology and infrastructure development. Several appear ready, willing, and able to proceed with such government support.

26.3

RECOMMENDATIONS

In spite of the remarkable accomplishments of the solar industry to date, unfortunately, costs are still too high for really large-scale use on par with fossil fuels today. Government policy makers interested in bringing revolutionary innovations into the market place should understand that there are four stages that an innovation moves through on its way to large-scale commercialization. These stages are 1. 2. 3. 4.

invention and first prototype benchtop demonstration; integration of the prototype into a system and qualification testing; manufacturing technology and production; and market introduction, acceptance, scale-up, and cost reduction.

604

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

The c-Si and TF solar modules and systems have now moved to stage 4 in this process. The technical expertise and worldwide market for the manufacture, distribution, installation, development, and exploitation of these solar cells for large-scale electricity production are so widespread and so pervasive that its continued growth and prosperity is well beyond the scope, control, or even the need of participation from any one country. It is recommended that a country or economic union that wants to move up or to remain in the top rankings in this inevitable process adopt one or more of the six identified public policy programs. History has shown them to be the most effective processes for achieving shifts in rankings—with feed-in tariffs being the leading example. History also indicates that the level of investment required to accomplish market shifts is on the order of 30% of the total integrated market investment worldwide up to that point. Hence, the more time that passes, the more difficult it becomes to arrange such investments for market shifting success, plus the more restricted such opportunities become to just the largest world economies. Fortunately, history also indicates that the investment payback time for the involved economies is on the order of 3 years in terms of contributions of products sold by these economies. Furthermore, the process itself generates a large number of jobs, economic regional growth and prosperity, and even reversals in a country’s or union’s balance of international payments where all of these benefits can last for multiple decades. However, there are now opportunities for newer solar cell technologies to move through these stages with government assistance and with the tremendous promise of lower cost and the associated benefits to society. The various concentrator module options represent an example. They are now emerging out of stage 1 and are entering stage 2. There is a need for demonstration projects for these newer solar options. Certification for safety and durability can be carried out in parallel with these demonstration projects. The lack of demonstration project funding in the United States as shown in Table 26.7 is one of the reasons the United States lost its early leadership in the solar industry. Without these funded demonstration and certification projects, the newer technologies are simply stuck in the laboratory with no clear path to commercialization. In the United States and in Europe, there is also a need for government investment in automated manufacturing as required in stage 3. Without automated

TABLE 26.7. Breakdown of Annual Solar Cell Development Budgets, 2001 Japan

United States

Germany

Research and development

51.0

35.0

26.7

Demonstration

16.5

0.0

5.5

188.4

84.6

29.6

Market stimulation

Source: International Energy Agency Photovoltaic Power Systems Programme.

RECOMMENDATIONS

605

manufacturing (see Chapter 7 for the current state of the art), low-cost labor wins. This is one of the reasons for China’s recent success in the solar market. The value of forcing the cost of solar cells down toward the $1 per watt level, currently being led by TF CdTe, cannot be overestimated. However, it is not just the initial purchase price of, say, a megawatt of modules that is important. It is the total life cycle cost in cents per kilowatt hour that will ultimately determine the large-scale technology competitiveness plus a technology’s staying power as the market size grows 100-fold and then 1000-fold. It involves the whole host of parameters that determine the LCOE, which is the most comprehensive integrated performance metric for the systems. There is no single characteristic that can, for the long term, make solar cell electricity economically competitive with current utility-provided electric power. Focus on just one isolated parameter can be very misleading. One of the strongest recommendations related to that is the large-scale field testing of the various available technologies, in a single location at the same time, side by side, under identical conditions. This independent field testing should start with sunny locations and then should spread to any other less favorable localities believed conducive to solar cell electricity conversion. The key parameters that should be monitored in such installations are the kilowatt hour per square meter per year produced by the competing technologies, the integrated solar insulation energy in kilowatt hour per square meter per year striking the systems installed at this site, and thus their field efficiencies as well as the installation costs and the degradation and failure rates plus the maintenance costs per year. To the best of the authors knowledge, there are no such independent test sites collecting such comprehensive but critical data at the present time anywhere in the world. These hard data are going to be crucial for justifying the advantages and disadvantages of the new technologies of TFs (including CdTe and CIGS) and of concentrators that are struggling to make inroads into the strong and long-standing dominance of c-Si. The test conditions should include one- and two-axis tracking versus no tracking. Standard c-Si technology (without concentration) is approaching its limits for price reduction as its material costs are approaching 50% of it total manufacturing costs. This fundamentally limits future price reductions. Without the competition of new “threats” like TF CdTe and concentrators, single-crystal silicon innovations like HIT and IBC cells would not likely occur, at least in a timely fashion. Only time and the market will determine ultimate shifts in technology dominance and market leadership. Another strong recommendation is for substantial government or union support program funding for the transition of new technologies into the market place—a challenge often referred to as the “valley of death” thwarting new and novel approach introductions. The barriers to market entry grow more formidable daily for the newer technologies against the currently dominant c-Si technology. TF CdTe has apparently broken through at least initially. However, the total range of creditable competition, including concentrators, needs to be brought at least to the large-scale field testing levels to establish its true long-term potentials and pitfalls.

606

26.4

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

SUMMARY

The solar cell market has grown 100-fold in yearly power products sold since the First Edition 15 years ago with (non-concentrating) c-Si still supplying the dominant share by far. However, c-Si is having a challenging time trying to reach “grid parity,” and a niche market has developed for TF solar cells that can provide lower initial purchase prices per kilowatt relative to c-Si. An alternate life cycle analysis of equivalent cost of power produced (in cents per kilowatt hour) shows that TFs may not be economically competitive long term if their efficiencies remain low or if their fundamental compositions contain elements of comparatively low abundance or until side-by-side field evaluations can establish relevant performance, maintenance, degradation, and other cost elements versus the competition. For example, TF efficiency values are still lower by up to a factor of two. Concentrator solar cell systems represent another alternative to 1-sun c-Si and TF systems. These concentrator systems are experiencing their first large-scale demonstrations in major installations in Spain. A key metric that is being determined but is still not accurately known is the kilowatt hour per square meter per year produced by concentrator systems versus 1-sun flat plates in their various installation conditions (of fixed angle and one- and two-axis tracking; for example, see Table 26.2 Palo Alto data). However, there are early indications that concentration (at least at low concentration values) may be one of the first to really compete with the traditional fixed-tilt c-Si and maintain the pressure for reductions in the most important metric, which is the life cycle cost in cents per kilowatt hour. Unfortunately, for the United States, its market lead held until the early 1990s has successively slipped country by country to the point that four other countries, Japan, China, Germany, and Spain, now occupy the leading positions. Their positions have advanced due to one or more of the six identified effective public policy programs where feed-in tariffs have been the most dominant and effective. Multiple such programs are largely in place in most of the current market leaders. China’s success, in particular, appears to have been centered around low-cost loans for building and equipping factories, low or zero land costs, plus favored tax treatments, but most recently, it has also started including feed-in tariffs. One advantage that the United States has is the great solar resource in the Southwestern region. This is an ideal environment for the low concentration option. The latter promises a straightforward and immediate path to an LCOE of under 10¢ per kilowatt hour for customers paying retail prices for electricity. One major U.S. opportunity could be to exploit the U.S. current battery-powered cars lead by closely coupling it to solar cell power production with the corresponding major reductions in oil imports and in the carbon footprint from personal transportation based on gasoline fuels. The latest device physics analysis indicates that the steady, continuing, and substantial increases in the performance of solar cells to well above 50% and on into the over 70% efficiency range will continue for many years before approaching fundamental limits, particularly in devices intended for concentrator systems. The latter will continually demand the fabrication of solar cell structures from increas-

FINAL OVERVIEW

607

ingly near-perfect crystalline semiconductor materials. Applications of TF solar cells to digital X-ray imaging are serving a multibillion dollar per year market that could realistically grow to well beyond $10 billion per year over the next 15 years.

26.5

FINAL OVERVIEW

The solar cell or PV effect was first reported by Bequerel in 1839 (see Reference 15, chapter 1). Development of practical sunlight energy conversion devices awaited the development of quantum mechanics and solid-state electronics in the first half of the twentieth century. Applications of these were pioneered by Shockley’s development of bipolar junction transistors in the mid-twentieth century, consisting essentially of two abrupt p/n junctions formed back to back using c-Si [15]. Chapin et al. [16] adapted this technology to demonstrate an abrupt p/n junction c-Si solar cell with a 6% conversion efficiency. Shockley and Queisser’s [17] classic modeling and analysis of abrupt p/n junction solar cells followed shortly thereafter. The ultimate optimization of such an abrupt p/n junction concept is perhaps the 25% efficient PERL cell of Chapter 4. The latter utilized light trapping, surface passivation, and limited-area ohmic contact features incorporated by the use of comparatively expensive photolithography fabrication processes. The cost-performance trade-offs of the terrestrial solar cell electricity market are currently dominated by abrupt p/n junction, c-Si solar cells of about 12–14% efficiencies where the expensive photolithography process has been replaced with the much lower-cost screen printing process. Over the last 4 years or so, the continual price reductions of these types of cells have stagnated (Chapter 2). This has slowed the overall adoption rate of solar cell-generated electric power. However, polycrystalline TF cells (Chapter 6) containing up to 100 times less materials but with 9–10% conversion efficiencies have just begun to make significant market penetration to resume the learning curve price reduction trajectory (see Fig. 2.3). The Nobel Prize-winning development of heterojunctions in III–V semiconducting single crystals in the 1960s and 1970s allowed the fabrication of ohmic contacts for high-current solar cell operation simultaneously with excellent surface passivation. As this was applied to high-current level concentrator solar cells, the champion GaAs single-crystal, single abrupt p/n junction solar cell efficiency of 28% was achieved at a sunlight concentration ratio of ∼240X (Chapter 4). Such high light concentration can offset to a degree the high costs of GaAs heterojunction solar cells that do involve photolithography fabrication. The versatility of III–V semiconductor materials has allowed the extension of this heterojunction technology to solar cells of three different abrupt p/n junctions stacked inside a single device. With each junction optimally selected with different bandgaps, the broadband light from the sun is much better matched for greatly reduced “above bandgap” efficiency losses. A major result is the concentrator champion cell with a 41% efficiency as described in Chapters 13 and 14. In a parallel path, very non-abrupt p-i-n junction solar cell development began in the late 1960s with the pioneering work of Kim and Schwartz [18] in

608

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

single-crystal Ge that was first adapted to amorphous silicon in the mid-1970s by Carlson and Wronski [19]. Decades of dedicated and painstaking work ultimately lead to the p-i-n IBC c-Si cell of Swanson et al. described in Chapter 4. Replacement of photolithography with lower-cost screen printing was one of several innovations that enabled this cell to become commercially successful. Expansion of this p-i-n concept in Japan to include an amorphous silicon-c-Si heterojunction resulted in the commercially successful HIT cell also covered in Chapter 4, which also uses screen printing fabrication. A key aspect of both the IBC and HIT p-i-n c-Si cells is that both are providing commercial quantities of modules with efficiencies in the 18–19% range in contrast to the commercial abrupt p/n junction c-Si devices with efficiencies stuck down in the ∼12–14% range. Even though they are somewhat higher in price, these better performance p-i-n solar cells appear to remain commercially competitive in the market place. As evident from the single-crystal silicon PERL cell, high performance alone does not provide commercial competitiveness. As further illustrated by the diminishing market role of amorphous silicon solar cells, lowest price per watt alone also does not make a solar cell technology competitive (see Fig. 2.9). Market winners provide the best trade-offs and compromises between the total system price and performance. These total system components and cost elements vary dramatically between installations with the thin modules lying flat on a building rooftop versus the much more complex two-axis tracking systems using high sunlight concentration and somewhat thicker modules. A metric that attempts to properly adjust for these different factors is the LCOE. LCOE is the system lifetime estimate of the energy a system produces in dollars (or euros or whatever currency) per kilowatt hour. For systems with dramatically different performance levels (e.g., a factor of 2), a very important performance input parameter is the kilowatt hours produced by the system per year and per square meter of the system facing the sun. This has to be combined with a ground coverage ratio (Chapter 11). This ratio equals one for thin modules covering the flat roof of a building. However, the capacity factor for horizontal flat roof installations is low. Capacity factor is a measure of how well the time variation of the system relative output power matches the time variation of the load that the system supplies. This capacity factor is even lower for fixed mounting on vertical structures like building walls. Typically, thin solar modules are mounted tilted at an optimum fixed angle facing south to increase the capacity factor to around 20% in sunny locations for utility loads (Chapter 10) as well as to increase the energy produced per square meter per year. For this, the ground coverage ratios fall to the order of 50% to avoid excessive shading by adjacent modules. The highest possible capacity factor and energy produced per square meter per year comes with two-axis tracking where the ground coverage ratio can be on the order of 15% depending on location and system design details. The values for all of these parameter inputs for accurate LCOE estimates are just becoming widely available. One has to be careful in comparing results that were not all taken at a single site, but the Table 26.2 results for sunny locations roughly indicate up to a factor of 3 difference in the annual

ABBREVIATIONS

609

average kilowatt hours per square meter per day. For low 6% efficiency performance systems mounted at a fixed tilt, this metric can be as low as 0.35. For a ∼12% fixed tilt, it can rise to ∼0.5–0.6 levels and on up to ∼0.9–1.1 values for two-axis tracking systems with higher ∼17–19% system efficiency values. It is of particular interest that the 1-sun, IBC two-axis trackers can produce this metric’s values of almost 1, in the same general range as two-axis trackers that use the very high-performance three-junction III–V cells that involve the comparatively thicker modules required in high concentration operation. A challenge for such tracking IBC systems is to address the low prices currently approaching $1 per Wp for TF cells albeit at lower 9–10% efficiency values. One immediate and attractive trade-off opportunity is the use of the low 3X concentrator system with the IBC cells to reduce prices down toward the dollar per Wp range if the IBC’s overall high ∼17–19% efficiency performance can be maintained over a 20- to 30-year field installation lifetime. The high-concentration, three-junction, two-axis tracking systems have just begun their learning curve development process. Both their performance and prices should improve substantially with growing experience over the coming years. MJ cells with both heterojunctions and p-i-n features have not yet been reported, but theoretical models suggest such configurations should provide device efficiencies well above 50% if junction open-circuit voltage values can be made to approach bandgap values. This would be accomplished by reducing the “below-bandgap” efficiency losses and would likely require high concentration ratios and “i” regions of near-perfect single-crystal semiconductor material quality and with near-ideal passivation of all its boundaries.

ABBREVIATIONS AC—alternating current AM1.5—air mass 1.5 Au—gold BOS—balance of system Cb—area-related balance of system costs Cd—cadmium Ci—inverter cost CIGS—copper indium gallium diselenide CM—concentrating module Cm—module cost CPU—central processing unit of a computer c-Si—crystalline silicon DVD—digital video disk EPRI—Electric Power Research Institute F—fixed charge rate GaInAs—gallium indium arsenide Ge—germanium

610

SUMMARY, CONCLUSIONS, AND RECOMMENDATIONS

ha—hours per year of operation HIT—heterojunction with intrinsic thin layer i—intrinsic of undoped semiconducting material IBC—interdigitated back contact III–V—material compounds from columns 3 and 5 of the periodic table InGaP—indium gallium arsenide ISE—Fraunhofer Institute for Solar Energy Systems ISFOC—Institute for Solar Photovoltaic Concentrators kWp—kilowatt peak measured under STC LCD—liquid crystal display LCOE—levelized cost of energy (electricity) MJ—multijunction n—negatively doped semiconductor material NRE—nonrecurring engineering cost NREL—National Renewable Energy Laboratory O&M—operation and maintenance p—positively doped semiconductor material PERL—passive emitter, rear locally diffused PV—photovoltaic or solar cell PVSC—Photovoltaic Specialists Conference PVUSA—a U.S. government and utility-sponsored “Photovoltaics for Utility Scale Applications” test site where solar cell system performance was related to prevailing environmental conditions using regression analysis r—indirect cost S—sunlight intensity at a given site in kilowatt per square meter SAM—Solar Advisor Model SEY—sunlight energy density at a given site in kilowatt hour per square meter per year STC—standard test condition (25°C cell temperature, 1 kW/m2 AM1.5 global sunlight) Te—tellurium TF—thin film USGS—U.S. Geological Survey VF—voltage factor Wp—watt peak under STC η—module efficiency ηF—field efficiency fraction with values between 0 and 1

REFERENCES [1] [2]

G. P. Willeke. Progress in industrial crystalline silicon solar cell technology. In 33rd IEEE Photovoltaic Specialists Conference Record, May 11–16, 2008, San Diego, CA. New York, IEEE (2008). Solarbuzz. Available at http://www.solarbuzz.com/. Accessed July 15, 2009 (2009).

REFERENCES [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20]

611

P. Fath. The status of c-Si. Presented at the 34th IEEE Photovoltaic Specialists Conference, June 7–12, 2009, Philadelphia, Penn., N.Y. IEEE (2009). Central Intelligence Agency, U.S. Government, The World Fact Book, Rank Order— GDP (purchasing power parity). Available at www.cia.gov/library/publications/theworld-factbook/rankorder/2001rank.html. Accessed 2008 Sept. 18. S. Hester and T. Hoff. Long term PV module performance. In 22nd IEEE Photovoltaic Specialist Conference Record, pp. 198–202. New York, IEEE (1985). G. B. Haxel and M. Diggles. Rare Earth Elements—Critical Resource for High Technology. Available at http://pubs.usgs.gov/fs/2002/fs087-02/ (2002). Tellurium. Available at http://www.en.wikipedia.org/wiki/Tellurium. Accessed May 7, 2010 (2010). Scott Roberts. Reality check: First Solar and tellurium. Available at http:// seekingalpha.com/article/55845-reality-check-first-solar-and-tellurium Accessed May 7, 2010 (2010). BEA. Bureau of Economic Analysis, U.S. Government, Table 1.19, Implicit Price Deflators for Gross Domestic Product. Available at http://www.bea.gov/national/ nlpaweb/TableView.asp#Mld. Accessed September 8, 2008 (2008). Wikipedia. Trans-Alaska Pipeline System. Available at http://en.wikipedia.org/wiki/ Trans-Alaska_Pipeline_System. Accessed September 26, 2009 (2009). Interstate. 50th Anniversary of the Interstate Highway System. Available at http:// www.fhwa.dot.gov/interstate/faq.htm#question6. Accessed August 22, 2008 (2006). Reuters. US oil import bill to top $400 billion this year. Available at http:// www.reuters.com/article/pressRelease/idUS236508+07-Mar-2008+BW20080307. Accessed September 7, 2008 (2008). zfacts. The Cost of Iraq, Afghanistan, and Other Global War. Available at http:// zfacts.com/p/447.html. Accessed September 7, 2008 (2008). L. D. Partain. Solar Cells and Their Applications, 1st Edition, Chapter 1, L. D. Partain, ed. New York, John Wiley & Sons (1995). W. Shockley. Electrons and Holes in Semiconductors. 1st Edition, New York Van Nostrand Reinhold (1950). D. Chapin, C. Fuller, and G. Pearson. J. Appl. Phys. 25, 676–677 (1954). W. Shockley and J. Queisser. J. Appl. Phys. 32, 510–519 (1961). C. Kim and R. Schwartz. IEEE Trans. Electron Devices 16, 657–663. D. Carlson and D. Wronski. Appl. Phys. Lett. 28, 671–673 (1976). R. Wiser, G. Barbose, C. Peterman, and N. Darghouth. Tracking the Sun II: The Installed Cost of Photovoltaics in the U.S. from 1998–2008. Lawrence Berkeley National Laboratory Report LBNL-1516E, LBNL, October (2009).

INDEX Note: f represents a figure and t represents a table. Abengoa, 274, 275f AC (alternating current), 256, 256f, 386–88, 492–93 accelerated life tests, 237 active pixel sensors, 554 ADCs (analog-to-digital converters), 541f, 542 aerosol optical depth (AOD), 437, 447–49 aerosol printing, 121–22 aliasing, 546, 550–52, 551f aluminum, 10, 53, 118–19, 276 American Society for Testing and Materials (ASTM), 383 AMFPIs (active matrix flat panel imagers), 560, 561f, 562f, 563 Amonix Inc. current status of, 586 HCPV (high-concentration photovoltaic) Fresnel concentrator systems, 365–71, 366f, 367f, 367t, 368f, 369f, 370f, 371f, 372f, 373, 373t MegaModule, 373–75, 374t, 375f amorphous selenium (a-Se). See a-Se (amorphous selenium) amorphous silicon (a-Si). See a-Si (amorphous silicon) Analogic, 507 analog-to-digital converters (ADCs), 541f, 542 AOD (aerosol optical depth), 437, 447–49 application-specific integrated circuits (ASICs), 541–43, 541f

ARCO (Atlantic Richfield Oil Company), 208, 274 Arizona capacity factor in, 234–35, 234f instrumentation/testing in, 226–27 system configuration in, 225–26 system design, installation, 225, 225t system maintenance in, 231–34, 232f, 233t system performance in, 228–30, 229f, 230f TEP Company in, 224, 224f Array Technologies, 212, 212f, 213f arrays Abengoa, 275f body-mounted, 407, 407f concentrating, 412–13, 413f electrostatically clean, 406, 414–15 flexible foldout, 409–10, 409f, 410f flexible roll-out, 410–12 high specific power, 416–17 high-radiation environment solar, 417 high-temperature/high-intensity, 413 photodiodes, TFTs and, 526–28, 527f, 528f photodiodes and, 536, 538f, 539f, 551, 551f rigid panel planar, 408–9, 408f See also space solar arrays a-Se (amorphous selenium) history of, 3 imaging and, 501–2 X-ray imagers, 564–68, 566f, 567f, 568f

Solar Cells and Their Applications, Second Edition, Edited by Lewis Fraas and Larry Partain Copyright © 2010 John Wiley & Sons, Inc.

613

614 a-Si (amorphous silicon) about, 503, 559–60 amorphous tandem cell research, 179, 181t applications for, 5–6, 527–30, 528f, 529f, 602 detectors, 541–43, 541f, 543f, 544f efficiency of, 140, 140f, 140t, 401t HIT cell and, 91–92, 91f, 124 hydrogenated, 150–53, 151f, 152f, 514–16 manufacture of, 188–91, 191t market history of, 501 photodiode arrays and, 536, 538f, 539f, 551, 551f p-i-n photodiodes and, 521–25, 521f, 522f, 524f, 525f p-i-n solar cells and, 77, 96–98, 96f, 97f properties of, 89, 89t, 511–20, 512f, 514f, 515f, 517f, 518f, 520f, 602 SCLI (space charge-limited current) model and, 94 space solar cells and, 404–5 TFTs (thin-film transistors) backpane, 528, 528f X-ray detectors and, 526–27, 527f, 533–35, 534f X-ray imaging and, 9–10, 10f, 555 See also bandgap energy (Eg); FPDs (flat panel detectors); TF (thin-film) solar cells Asia market history, 20f ASICs (application-specific integrated circuit), 541–43, 541f See also readout chips ASTM (American Society for Testing and Materials), 383 AstroEdge arrays, 412 Atlantic Richfield Oil Company (ARCO), 208, 274 atmospheric effects, 433–39, 434f automation and assembly line, 164–68, 165f, 166f, 167f, 168f automobiles, 592–93, 592f azimuth trackers for commercial building flat roofs, 213–14, 214f, 215f geometries for, 210f, 211t structure of, 367–68, 368f See also fixed-tilt trackers; trackers

INDEX band edges, 53, 54, 68–69, 86–87, 103–4 bandgap energy (Eg) definition of, 45–46 efficiencies and, 59, 100f, 299–301 lattice constants and, 324–25, 325f multijunction cells and, 402, 402t open-circuit voltage and, 68–70, 73 See also a-Si (amorphous silicon); CuInSe2-based solar cells bandgap voltage, 98–102, 100f, 101f, 104 batteries, 5, 220, 221f, 489–90 benefit/cost analyses, 216t See also SAM (solar advisor model) BIPV (building integrated photovoltaic), 167–68, 168f, 195t, 197 body-mounted arrays, 407, 407f Boeing 702, 412–13 BOL (beginning of life) efficiencies, 417, 418t BOS (balance of system), 227–28, 228f BPDs (bypass diodes), 130 cadmium telluride (CdTe) solar cells. See CdTe (cadmium telluride) solar cells calculators, 5–6 Canon Medical Systems, 507 carousels, 213–16, 214f, 215f Carrizo Solar Corporation, 208 cars, 592–93, 592f case studies dual-cell Cassegrainian system, 353–54, 353f, 354t on raw costs, 487 reliability model, 237 Cassegrainian HCPV (high-concentration photovoltaic) systems about, 337–38, 357 cell selection for, 356–57 construction of, 347–48 design, operation of, 338–43, 339f, 340f, 341f, 342f dual-cell concentrators, 352–54, 352f, 353f, 354t optical design of, 344–46, 345f, 346f, 347f panels, 350–52, 351f, 354, 355f, 356 two-axis tracking systems, 349–50, 350f See also concentrating optical systems

INDEX CBCT (cone beam computed tomography), 505, 505f, 506f CDS (correlated double sampling), 550 CdTe (cadmium telluride) solar cells about, 142–45, 143f, 144f China and, 177–78, 178f, 181t detectors, 568 efficiencies of, 138–40, 138t, 139t, 140t market history of, 20 space solar cells and, 404–5 Te for, 597–98, 597f X-ray imagers, 569 See also TF (thin-film) solar cells CdZnTe (CZT), 563, 564t, 568, 569 cell efficiencies. See efficiencies central receiver reflectors, 343–44 champion cells, 38, 78–83, 79f, 80t champion GaAs, 88–89, 89t China average annual usable hours, 200, 201t, 202t, 203t, 204t, 205t, 206t CdTe (cadmium telluride) solar cells and, 177–78, 178f, 181t CIGS (copper-indium-gallium-selenium) solar cells and, 175t, 178, 179f, 181t equipment in, 181–83, 181t, 182f, 183f future investment by, 26–27 incentive policy, 198–99 industry chain, 185–91, 185f, 185t, 187t, 189t, 191t industry in, 171–73, 172f, 172t, 174t market growth in, 195–98, 196f, 196t, 197t, 198t market history of, 19–20, 19f module assembly, 192, 192t module prices, 193, 193f, 194f, 194t, 195t R&D, 173–80, 175t, 176t, 177f, 178f, 179f, 180f solar irradiation in, 171–73, 172f, 172t CIGS (copper-indium-gallium-selenium) solar cells, 175t, 178, 179f, 181t, 404–5 circumsolar radiation (CSR), 320–21, 449–54, 451f, 452f, 453f cloud attenuation, 438–39 CMOS (complementary metal oxide semiconductor), 574–75

615 combination systems, 5 commercial azimuth trackers, 213–14, 214f, 215f compound parabolic concentrators (CPCs), 321–22 Compton, 67–68 concentrating arrays, 412–13, 413f concentrating optical systems, 341–44, 342f, 343–44 See also Cassegrainian HCPV (highconcentration photovoltaic) systems concentrating photovoltaic (CPV) costs and, 11–12, 11f definition of, 10 efficiencies of, 340–41, 341f, 389, 390f field installations, 377–80, 379f junctions and, 58f, 59 See also rooftop CPV (concentrating photovoltaic) systems concentrating photovoltaic (CPV) systems about, 361, 365, 606 early development of, 273–74, 274f, 275f, 276 rating standards for, 383–85, 384t, 386f See also power plants concentrator point contacts, 89–90, 90f concentrators costs of, 273, 290–91, 290t, 342–43, 342f current status of, 585, 585f design of, 298–99, 298f efficiency vs., 340–41, 341f market history of, 38–39, 64 multijunctions and, 304, 305f reflective, 361–63, 362f, 433–36, 434f refractive, 363–65, 363f, 364f See also crystalline silicon (c-Si) modules; Fresnel lens concentrators; TF (thin-film) modules Concentrix Solar, 327–30, 328f, 329f conduction band junctions, diodes and, 53, 54f open-circuit voltage and, 68–70 power curves and, 54, 55f valence bands and, 512f, 513 wave theory and, 49–50, 51f, 52 cone beam computed tomography (CBCT), 505, 505f, 506f correlated double sampling (CDS), 550

616 cost/benefit analyses, 216t See also SAM (solar advisor model) costs concentration and, 273, 290–91, 290t, 342–43, 342f for CPV (concentrating photovoltaic), 11–12, 11f current status of, 590–91 economic modeling, 586–87 efficiencies and, 262–63, 262f of FITs (feed-in tariffs), 36 for large-scale solar power production, 486–89 power plants, 265–67, 266f for PV (photovoltaic) modules, 193, 193f, 194f, 194t, 195t for PV (photovoltaic) systems, 59, 60–61t, 62–63, 596 for SAM (solar advisor model), 465–67, 466t, 470, 470t of solar cell modules, 8–9, 9f, 17–20, 19f, 23–24 for solar trackers, 215–17, 216t for Springerville Generating State Solar System, 243–46, 243t, 245t See also LCOE (levelized cost of energy) CPCs (compound parabolic concentrators), 321–22 CPV (concentrating photovoltaic). See concentrating photovoltaic (CPV) CPV (concentrating photovoltaic) systems. See concentrating photovoltaic (CPV) systems CR technology, 501 crystalline silicon (c-Si) about, 102–4, 113–14, 133–34 China and, 176, 177f concentrator point contacts, 89, 90f efficiency limitations of, 116–20 HIT cell, 91–92, 91f ingots, wafers, 187–88, 189t manufacture of, 188–90 market history of, 21–22 nc-Si (nanocrystalline silicon), 150–53, 151f, 152f PERL cell, 90 point contact crystalline silicon cells, 83–89, 84f, 85f, 86f, 88f, 89t

INDEX PV (photovoltaic) cells and, 114–16, 116f technologies using, 120–29, 123f, 126f, 127f, 128f See also low-concentration crystalline silicon systems crystalline silicon (c-Si) modules about, 606 current status of, 584–85, 584f, 585f opportunities for, 590–91, 590f PV (photovoltaic) modules, 129–33 See also concentrators; TF (thin-film) modules crystals, 63–64, 64f CSR (circumsolar radiation), 320–21, 449–54, 451f, 452f, 453f CuInSe2-based solar cells, 142–43, 146–50, 146t, 147f, 148f See also bandgap energy (Eg) current collecting technologies, 120–22 CZ (Czochralski process), 114–16, 116f CZT (CdZnTe), 563, 564t, 568, 569 D/A (dual-axis) trackers, 208–11, 210f, 210t, 211f Day4, 126–29, 126f, 127f, 128f, 132–33 DC (direct current) ratings, 386, 387f deep discharge batteries, 5 demonstration plants. See ISFOC (Instituto de Sistemas Fotovoltaicos de Concentración) dental procedures, 504f, 505 Department of Defense, U.S., 25 detectors. See flat panel detectors (FPDs); photoconductors; X-ray detectors D4, 126–29, 126f, 127f, 128f, 132–33 digital imaging, 501–2 See also portal imaging digital lithography, 528–29, 529f diodes, 53–54, 54f, 130 See also p-i-n photodiodes; P/N junction diodes; Shockley diodes direct beams, 371, 433, 434f, 439–40 direct current (DC) ratings, 386, 387f DNI (direct normal irradiance) future comparisons, 60–61t, 62–63 rating standard and, 383–84, 384t

INDEX scattering and, 437–38, 438f SMARTS model and, 306–7, 307f, 437, 438f See also irradiance double heterostructures, 79–80, 79f DR technology, 501–6, 503f, 504f, 505f, 506f dual-axis (D/A) trackers, 208–11, 210f, 210t, 211f dye-sensitized solar cells, 179–80, 180f, 181t efficiencies under AM1.5, 299–301, 300f of a-Si (amorphous silicon), 140, 140f, 140t, 401t bandgap energy (Eg) and, 59, 100f, 299–301 BOL (beginning of life), 417, 418t China and, 180, 181t costs and, 262–63, 262f of CPV (concentrating photovoltaic), 340–41, 341f, 389, 390f history of, 4 increasing, 55–59, 55f, 56f, 58f, 59f inverter conversion, 471–74, 472t, 473t, 474t limitations of, 116–20 measured field performances, 593–96, 594t PV (photovoltaic) cells and, 115–20, 400–401, 401t QEs (quantum efficiencies), 46, 55, 522 space solar cells, 417–18, 418t for TF (thin-film) solar cells, 138–40, 138t, 139t, 140t of triple-junction solar cells, 324–25, 326f 28.2%, 78–83, 79f, 80t Eg (bandgap energy). See bandgap energy (Eg) Einstein, Albert, 44–45, 67–68 electricity applications for, 6–7, 6f, 7f arguments for, 11–13, 11f market history of, 17–22, 18f, 19f, 20f, 21f, 36 projected future of, 28–30

617 electroless plating, 120–21 electromagnetic radiation, 47–48 electromagnetic waves, 44–45, 44f, 45f electron volts, 44–47 electron wave theory, 49–53, 50f, 51f, 52f electronic portal imaging detectors (EPIDs), 534 electrons, 47–48, 48f, 50–51, 55 electrostatically clean arrays, 406, 414–15 Emcore, 401–3 energy, definition of, 44 ENTECH, 274, 275f EOL power, 406, 417–20, 418t, 419t EPIDs (electronic portal imaging detectors), 534 EUCLIDES, 377–78, 378f European market history, 19–20, 19f, 20f, 21f EVA (ethylene vinyl acetate), 161–62, 163f, 222f, 223 extraterrestrial irradiance, 427–32, 428f, 429f, 432f extraterrestrial spectrum, 430–32, 432f Exxon Mobile, 17, 18f failure modes and effects analysis (FMEA), 236 feed-in tariffs (FITs) about, 207, 208 costs and, 36 countries and, 25–26, 589, 589f, 606 installation rates and, 20f for large-scale solar power production, 488–89 PURPA (Public Utilities Regulatory Power Act) and, 23, 25–26 See also governments Fermi energy, 71, 513 Fermi levels ohmic contacts, heterojunction interfaces and, 78 p/n, p-i-n crystalline silicon cells and, 86–87 Shockley diode model and, 70–71, 71f, 73 FF (fill factor), 76–77, 77f, 116, 326f field performances, 370, 370f, 593–96, 594t

618 figures of merit, 417–20, 418t, 419t film imaging, 501–2 First Solar, 145, 597–98 FITs (feed-in tariffs). See feed-in tariffs (FITs) fixed-tilt trackers about, 209f, 263, 595 D/A (dual-axis) trackers vs., 210f, 210t irradiance and, 441–46, 443f linear axis trackers vs., 212, 212f, 213f See also trackers flat panel detector (FPD) market about, 499–500, 500f, 508 evolution of, 506–7 future of, 508–9 history of, 500, 501–2, 507–8 flat panel detectors (FPDs) about, 555 applications for, 502–3, 503f, 504f, 505–6, 505f, 506f construction of, 535–36, 535f, 536f, 537f future trends for, 508–9 noise from, 544–52, 545f flat panels, 361, 491f, 502, 574–75 FLATCON modules, 327–30, 328f, 329f Flat-Plate Solar Array Project (FSA), 18f, 21f, 22 flat-plates (FPs) collectors, 433–34, 434f, 441, 445 modules, 4, 460 flexible foldout arrays, 409–10, 409f, 410f flexible roll-out arrays, 410–12 flexible substrates, 527–28, 574, 575 Florida Power and Light (FPL), 251, 252f FMEA (failure modes and effects analysis), 236 FPDs (flat panel detectors). See flat panel detectors (FPDs) Fraunhofer Institute for Solar Energy, 314–15, 327 Fresnel lens concentrators about, 313–16, 315f, 331, 343–44 applications for, 330, 330f developments of, 327–30, 328f, 329f elements of, 316–22, 316f, 317f, 320f, 321f

INDEX HCPV (high-concentration photovoltaic) Fresnel concentrator systems, 365–71, 366f, 367f, 367t, 368f, 369f, 370f, 371f, 372f, 373, 373t history of, 274, 274f, 275f refraction of light rays and, 364, 364f triple-junction solar cells, 322–27, 323f, 325f, 326f, 403 FSA (Flat-Plate Solar Array Project), 18f, 21f, 22 Fuji Photo Film Co., 501, 565–66, 566f GaAs (gallium arsenide) electron movement in, 52–53, 52f InGaP and, 352, 352f, 353, 356–57 junctions and, 46–47, 46f, 607 p-on-n, n-on-p, 78–83, 79f, 80t wave functions for, 56–58 gain effect, 552, 553f GaSb (gallium antimonide) IR solar cells, 353–54, 357 junctions and, 46–47, 46f single crystals and, 63 wave functions for, 57–58 GE (General Electric), 507 Germany 100,000 Roofs Program, 21f, 25–26 market history of, 20, 20f solar cell budget of, 604t ghosting, 567–68, 568f global comparisons, 173, 174t governments economic development programs by, 591 incentive policies, 33, 35–36, 36t, 467 investments by, 587, 591, 603, 603–4 public policies and, 22–28, 37, 113, 198–99 subsidy programs and, 133 See also feed-in tariffs (FITs); specific countries grid-connected PV power systems, 182–83, 182f, 183f, 184f HCPV (high-concentration photovoltaic) systems comparison of, 60–61t, 62–63

INDEX Fresnel concentrator systems, 365–71, 366f, 367f, 367t, 368f, 369f, 370f, 371f, 372f, 373, 373t multijunctions in, 293–95, 294f, 297f, 373–75, 374t, 375f See also Cassegrainian HCPV (highconcentration photovoltaic) systems; III-V multijunctions heterojunctions, 78, 607–9 HgI2 imagers, 569–74, 570f, 571f, 572f, 573f high-concentration Cassegrainian solar cells. See Cassegrainian HCPV (high-concentration photovoltaic) systems high-concentration Fresnel lens systems. See III-V multijunctions HIT (heterojunction with intrinsic thin-layer) cells, 89t, 91–92, 91f, 124, 608 hospitals DR technology applications in, 502–3, 503f, 504f, 505–6, 505f, 506f imaging history of, 501–2 market history of, 500, 500f See also X-ray detectors Hubbert’s Peak (Deffeyes), 12 hydrogenated a-Si (amorphous silicon), 150–53, 151f, 152f, 514–16 IBC (interdigitated back contact) cells about, 83–89, 84f, 85f, 86f, 88f, 89t, 90f novel types of, 124–26 p-i-n solar cells and, 608–9 See also Shockley diodes IEC (International Electrotechnical Commission) standard 60904, 448–49 standard 61724, 228 standard 62108, 380–82, 390, 391, 392 TC82 Working Group 7, 383–85, 384t, 386f IEC (Israel Electric Corporation), 490, 491f image lag about, 552, 553f ghosting vs., 567–68, 568f of PbI2, HgI2 imagers, 571, 571f TFTs (thin-film transistors) and, 526, 540–41

619 image sensors, 527–28 imaging, 501–2 See also AMFPIs (active matrix flat panel imagers); portal imaging; X-ray imaging IMM (inverted metamorphic) multijunction cells, 301–3, 302f, 303f, 324, 402–3 incident irradiance, 427 India market history, 19–20, 19f industry about, 133–34 in China, 171–73, 172f, 172t, 174t equipment and, 164–68, 165f, 166f, 167f, 168f future of, 168–69 growth in, 113–14, 159–61 initial investment in, 261–65, 262f, 264f, 265f landscape of, 507–8 recommendations for, 603–5, 604t industry chain, 185–91, 185f, 185t, 187t, 189t, 191t See also polysilicon infrared-sensitive PV (photovoltaic) cells, 5, 63 ingots, 176, 177f, 187–88, 189t ink-jet printing, 121 Institute for Solar Energy (ISE) Fraunhofer, 121–22 Instituto de Sistemas Fotovoltaicos de Concentración (ISFOC). See ISFOC (Instituto de Sistemas Fotovoltaicos de Concentración) insulators, 50 integrated power systems, 415–17 interdigitated back contact (IBC) cells about, 83–89, 84f, 85f, 86f, 88f, 89t, 90f novel types of, 124–26 p-i-n solar cells and, 608–9 See also Shockley diodes International Electrotechnical Commission (IEC). See IEC (International Electrotechnical Commission) International Space Station (ISS) Si solar cells, 397–98, 398f, 399f, 400–401, 400f Interstate Highway System, U.S., 27–28 inverted metamorphic (IMM) multijunction cells, 301–3, 302f, 303f, 324, 402–3

620 inverters conversion efficiencies of, 471–74, 472t, 473t, 474t lifetime analysis of, 478–79 performance models for, 469t irradiance Chinese statistics, 201t, 202t, 203t, 204t, 205t, 206t extraterrestrial, 427–32, 428f, 429f, 432f for fixed-tilt, tracking, concentrating geometries, 441–46, 443f solar, 171–73, 172f, 172t spectral, 306, 446–47, 447f terrestrial applications and, 439–40 See also DNI (direct normal irradiance); TSI (total solar irradiance) ISE (Institute for Solar Energy) Fraunhofer, 121–22 ISFOC (Instituto de Sistemas Fotovoltaicos de Concentración) demonstration plants, 391f future work of, 391 IEC 62108 tests and, 381–82 lessons learned, 390–91 plant acceptance by, 382–83 plant installations, 377–80 radiation results, 388, 388f, 389f, 390f rating standards for, 383–88, 387f technologies selected by, 379f Israel Electric Corporation (IEC), 490–91, 491f ISS (International Space Station) Si solar cells, 397–98, 398f, 399f, 400–401, 400f Italian Space Agency, 405 Japan budget of, 604t market history of, 19–20, 19f, 20f, 21f production in, 23–25 JPL (U.S. Jet Propulsion Laboratory), 18f, 21f, 22–23 junctions CPV (concentrating photovoltaic) and, 57–59, 58f GaAs (gallium arsenide) and, 46–47, 46f, 607 open-circuit voltage and, 80

INDEX P/N junction diodes, 53–54, 54f See also heterojunctions; multijunction (MJ) solar cells; III-V multijunctions; tunnel junctions JX Crystals Inc., 60–61t, 62, 213, 214f, 215f See also 3-sun LCPV (low-concentration photovoltaic) concept Kodak, 507, 565, 566f Kroemer, Herbert, 79–80 kTC noise, 549–50, 549f large-scale solar power production about, 25, 483–84, 494–95 costs for, 486–89 problem with, 484, 485f space, time requirements for, 485–86 technological challenges of, 489–94, 491f Las Vegas, Nevada, 60–61t, 62, 594t, 595 See also University of Nevada, Las Vegas laser buried grid PV cells, 122–23, 123f laser-fired contact, 122 laws. See public policies LCOE (levelized cost of energy) about, 252–53, 269 drivers of, 251–52 estimation of, 30, 31t, 32t, 33–36, 33t, 34f, 36t forecasting tool, 268, 268f inputs for, 253–54 SAM (solar advisor model) and, 465–67, 466t sensitivity of, 255, 255f for solar cell systems, 599, 600f system lifetime and, 477, 478f, 479, 479f utility-scale systems and, 255–60, 256f, 258f, 259f, 260f See also costs LCPV (low-concentration photovoltaic) Abengoa arrays, 275f comparison of, 60–61t, 62 future of, 285, 287–89, 287f, 287t, 288f, 288t, 289f

INDEX mirror module development, 279–83, 279t, 280f, 281f, 282f, 283f, 284f, 285, 285f, 286f predictions vs. goal comparison, 290t See also 3-sun LCPV (low-concentration photovoltaic) concept levelized cost of energy (LCOE). See LCOE (levelized cost of energy) light history, 44–45 light rods, 321–22 light-induced plating, 120–21 line noise, 548–49, 548f linear axis trackers, 212, 212f, 213f lithography, 528–29, 529f Livermore, California, 444–46 low-concentration crystalline silicon systems about, 273, 290–91 early development of, 273–74, 274f, 275f, 276 future cost of, 290, 290t future manufacture of, 285–89, 286f, 287, 287f, 287t, 288f, 288t, 289f mirror module development for, 279–83, 279t, 280f, 281f, 282f, 283f, 284f, 285, 285f See also crystalline silicon (c-Si); nc-Si (nanocrystalline silicon); point contact crystalline silicon cells; 3-sun LCPV (low-concentration photovoltaic) concept LS-PV systems, 197, 198t mammography, 505, 507, 566–68, 566f, 568f manufacturers current status of, 587, 588t, 589 history of, 137, 140, 140t initial investment of, 261–65, 262f, 264f, 265f markets a-Si (amorphous silicon) history, 501 changes in, 583 of concentrators, 38–39, 64 demand by, 589, 589f history of, 17–22, 18f, 19f, 20f, 21f, 36 for solar cells, 4–7, 589, 589f See also China; flat panel detector (FPD) market

621 Mars, 407, 407f, 411–12, 415 See also space solar arrays mc (monocrystalline). See crystalline silicon (c-Si); crystalline silicon (c-Si) modules; low-concentration crystalline silicon systems; nc-Si (nanocrystalline silicon); point contact crystalline silicon cells measured field performances, 593–96, 594t MegaModules, 366–67, 367f, 367t, 369f, 373–75, 374t, 375f MESSENGER, 413–14 metals, 50 Middle East, 12 Mie scattering, 436 military, 24–25 models inverter performance, 469t reliability/availability, 237–38, 238f, 239t, 240t, 241t, 242f, 242t, 243 SCLI model, 93–98, 93f, 96f, 97f See also Shockley diodes; SMARTS (Simple Model of the Atmospheric Radiative Transfer of Sunshine) model module performance models, 469t modules about, 140t, 220, 222–23, 222f assembly of, 161–64, 162f, 163f, 192, 192t costs, prices of, 193, 193f, 194f, 194t, 195t Day4 technology and, 132–33 degradation of, 475, 475f, 476f, 477f equipment for, 164–68, 165f, 166f, 167f, 168f future of, 168–69 materials for, 159–61, 160f production technology for, 129–32 monolithic multijunction solar cells, 338–41, 339f, 340f, 341f Moore’s law, 29 MTF (modulation transfer function), 551–52, 551f, 565, 566f multijunction (MJ) solar cells concentrators and, 304, 305f design of, 295–96, 295f

622 multijunction (MJ) solar cells (cont’d) in HCPV (high-concentration photovoltaic) systems, 293–95, 294f, 297f, 373–75, 374t, 375f inverted metamorphic (IMM), 301–3, 302f, 303f, 324, 402–3 monolithic, 338–41, 339f, 340f, 341f operating condition performance, 304–8, 305f, 307f reliability of, 308–9 in space, 402–4, 402t See also Cassegrainian HCPV (highconcentration photovoltaic) systems; III-V multijunctions; triple-junction solar cells Nankai University, 178–79, 179f, 180f nanostructured solar cells, 404 NASA. See space solar cells National Development and Reform Committee, 197–99, 197t, 198t National Solar Radiation Data Base, U.S., 440 natural gas depletion, 12 nc-Si (nanocrystalline silicon), 150–53, 151f, 152f See also crystalline silicon (c-Si); low-concentration crystalline silicon systems; point contact crystalline silicon cells NEDO (New Energy and Industrial Technology Development Organization), 24–25 New South Wales, University of, 122–23, 123f noise, 544–52, 545f, 548f, 549f NREL (National Renewable Energy Laboratory), 124, 306, 324, 440 NSRDB (National Solar Radiation Data Base), 440 nuclear fuels, 12 ohmic contacts, 78 oil depletion, 12 100,000 Roofs Program, 21f, 25–26 open-circuit voltage, 68–70, 73–74, 75f, 80–81, 82f

INDEX optical losses, 117 See also Fresnel lens concentrators organics, 405 ORNL (Oak Ridge National Laboratory), 283–84, 285f, 286f Pacific Gas & Electric Company (PG&E), 251–52 PbI2 imagers, 569–74, 570f, 571f, 572f, 573f PERC (passivated emitter and rear cell) PV (photovoltaic) cells, 123 periodic table, 48, 49t, 50–52 PERL (passivated emitter, rear locally diffused) PV (photovoltaic) cells, 89t, 90, 123–24 See also Shockley diodes PG&E (Pacific Gas & Electric Company), 251–52 photoconductors detectors, 559, 563, 564t imagers, 563–75, 566f, 567f, 568f, 570f, 571f, 572f, 573f See also X-ray detectors photodiodes about, 521–25, 521f, 524f, 525f arrays, 536, 538f, 539f, 551, 551f TFTs (thin-film transistors) and, 526–28, 527f, 528f photoinduced metal plating, 120–21 photolithography, 528–29, 529f photons, 45–47, 45f, 46f photovoltaic (PV) solar cells. See PV (photovoltaic) solar cells Physikalisch-Meteorologisches Observatorium Davos (PMOD) dataset, 428–30, 428f, 429f p-i-n photodiodes about, 521–25, 521f, 522f, 524f, 525f lag, gain effect and, 552, 553f p-i-n solar cells about, 77, 102–4, 607–8 advanced concept, 81–83, 82f a-Si (amorphous silicon) and, 77, 96–98, 96f, 97f concentrator point contacts, 89–90, 90f HIT cell, 91–92, 91f PERL cell, 90 photodiodes vs., 522–23

INDEX point contact crystalline silicon cells, 83–89, 84f, 85f, 86f, 88f, 89t p-i-n/TFT pixel architecture, 537–41, 538f, 539f See also TFTs (thin-film transistors) planar collectors, 434, 441–46, 443f planar PV (photovoltaic), 60–61t Planck’s law, 431–32, 432f PMOD (Physikalisch-Meteorologisches Observatorium Davos) dataset, 428–30, 428f, 429f P/N junction diodes about, 53–54, 54f, 55f, 102–4 concentrator point contacts, 89–90, 90f HIT cell, 91–92, 91f PERL cell, 90 p-i-n solar cells and, 77 Shockley diodes and, 70–74, 88–89 point contact crystalline silicon cells, 83–89, 84f, 85f, 86f, 88f, 89t See also crystalline silicon (c-Si); lowconcentration crystalline silicon systems; nc-Si (nanocrystalline silicon) politics, 484 See also public policies polycrystalline CuxS/CdS, 93f, 94 Si (silicon) TFTs, 574 TJ (triple junction) cells and, 403 types of, 142–50, 143f, 144f, 146t, 147f, 148f polysilicon, 175–76, 176t, 186–87, 187t, 261–62 portable radiography detectors, 535–36, 535f, 536f, 537f portal imaging, 506, 506f, 534 See also digital imaging; X-ray imaging power, definition of, 44 power plants acceptance of, 382–83 expenses of, 265–67, 266f initial investment in, 261–65, 262f, 264f, 265f installation procedures for, 378–80, 379f power stations, 181–83, 182f, 184f power systems, 415–20, 418t, 419t power transmission, 492–93 prices current status of, 598–602, 598t, 599f, 599t, 600f, 601t, 602t

623 for PV (photovoltaic) modules, 193, 193f, 194f, 194t, 195t See also costs printing, 118–19, 121–22 progressive scanning, 542–43, 543f, 544f prototype tests, 382 public policies, 22–28, 37, 113, 198–99 PURPA (Public Utilities Regulatory Power Act), 6, 18f, 23, 25–26 PV (photovoltaic) solar cells c-Si (crystalline silicon) and, 114–16, 116f current collecting technologies for, 120–22 efficiencies of, 115–20, 400–401, 401t infrared-sensitive, 5, 63 novel types of, 122–29, 123f, 126f, 127f, 128f See also LCPV (low-concentration photovoltaic); solar cells; specific types PV (photovoltaic) systems components of, 220, 221f, 222–23, 222f costs of, 59, 60–61t, 62–63, 596 history of, 5–7, 6f, 7f, 8f lifetime analysis of, 477, 478f, 479f reliability of, 235–41, 236f, 238f, 239t, 240t, 241t, 242f, 242t, 243 residual value of, 267 See also China; concentrating photovoltaic (CPV) systems quantum efficiencies (QEs), 46, 55, 522 quantum mechanics, 47–48, 67–68 quantum well (QW) solar cells, 98–104, 100f, 101f quasi-Fermi energy, 69 quasi-Fermi levels, 70–77, 71f, 87, 97–99 Queisser, 70–77, 75f, 77f radiation electromagnetic, 47–48 ISFOC (Instituto de Sistemas Fotovoltaicos de Concentración) results for, 388, 388f, 389f, 390f solar, 433, 434f, 440–41, 454–56

624 radiography detectors, 535–36, 535f, 536f, 537f rainbows, 44–47 rating standards, 383–88, 384t, 386f, 387f Raviv model, 486–87, 489–90, 494 Rayleigh scattering, 436 RBD (reliability block diagram), 235–37 readout chips, 536, 537f, 541f See also ASICs (application-specific integrated circuit) recombination losses, 118 reference spectra, 447–48, 448f reflective concentrators, 361–63, 362f, 433–36, 434f refractive concentrators, 363–65, 363f, 364f reliability block diagram (RBD), 235–37 reliability/availability model, 237–38, 238f, 239t, 240t, 241t, 242f, 242t, 243 renewable energy development, 197–98, 197t, 198t resistivity losses, 118–20 rigid panel planar arrays, 408–9, 408f Robbins Engineering, Inc., 208 robotics and assembly line, 164–68, 165f, 166f, 167f, 168f rooftop CPV (concentrating photovoltaic) systems about, 7, 7f, 216, 593 applications for, 330, 330f on commercial buildings, 213–14, 214f, 215f comparison of, 140, 140f See also 3-sun LCPV (low-concentration photovoltaic) concept S/A (single-axis) trackers, 209–11, 211t, 212, 212f, 213f, 215f Sacramento, California, 444–46 SAM (solar advisor model) about, 463–65, 481–82 analysis examples of, 471–81, 472t, 473t, 474t, 475f, 476f, 477f, 478f, 479f, 480f, 481f costs for, 465–67, 466t, 470, 470t software for, 467, 468t, 469t, 470, 470f, 470t SAM (sun and aureole measurement), 450, 451f

INDEX Sandia National Laboratories Fresnel lenses at, 314 performance models for, 469t solar cell system reliability, 234–41, 236f, 238f, 239t, 240t, 241t, 242f, 242t, 243 SANYO Electric Company, 124, 287–88, 287t, 288t satellites, 427–32, 428f, 429f, 432f See also space solar arrays SC (solar constant), 428–30, 428f, 429f SCARLET concentrator arrays, 412 scintillators, 535–36, 560, 561f SCLI (space charge-limited current) model, 93–98, 93f, 96f, 97f SCOPAs (survivable concentrating photovoltaic arrays), 414 screen printing, 118–20 SEIA (Solar Energy Industries Association), 246–47 selective emitters, 119–20 semiconductors, 49–53, 50f sensor arrays. See arrays Shockley diodes about, 70–77, 71f, 75f, 77f PERL cells and, 90 P/N junction diodes and, 70–74, 88–89 See also IBC (interdigitated back contact) cells; PERL (passivated emitter, rear locally diffused) PV (photovoltaic) cells short-circuit current (Isc), 74, 75f, 116, 117 shot noise, 550 Siemens, Werner, 3 silicon about, 3–4, 51–52, 51f PV (photovoltaic) modules, 159–61, 160f single-crystal solar cells, 57, 63–64, 64f substrate thin-film solar cells, 179 TFTs (thin-film transistors), 574 See also a-Si (amorphous silicon) simple efficiency model for flat-plate modules, 469t single-axis (S/A) trackers, 209–11, 211t, 212, 212f, 213f, 215f single-point efficiency model, 469t smart grids, 493–94

INDEX SMARTS (Simple Model of the Atmospheric Radiative Transfer of Sunshine) model CSR (circumsolar radiation) and, 451, 452f DNI (direct normal irradiance), 306–7, 307f, 437, 438f terrestrial applications of, 446–47, 447f solar arrays. See arrays; space solar arrays solar cell efficiencies. See efficiencies solar cell electricity. See electricity solar cell systems. See PV (photovoltaic) systems solar cells applications for, 602 current status of, 584, 584f history of, 3–4, 607–9 installation rate of, 20f manufacturing companies of, 587, 588t, 589 market and, 4–7, 589, 589f modules, 8–9, 9f, 17–20, 19f, 23–24 opportunities for, 590–93, 590f photoconductor detectors vs., 559 prices for, 598–602, 598t, 599f, 599t, 600f, 601t, 602t space satellites and, 427–32, 428f, 429f, 432f types of, 8, 8f See also multijunction (MJ) solar cells; PV (photovoltaic) solar cells; space solar cells; triple-junction solar cells solar constant (SC), 428–30, 428f, 429f Solar Energy Industries Association (SEIA), 246–47 solar intensity, 44 solar irradiation, 171–73, 172f, 172t solar power production. See large-scale solar power production solar radiation, 433, 434f, 440–41, 454–56 solar trackers. See trackers solar water pumping, 5 solar/wind complementary stations, 181–82 SolFocus Cassegrainian concentrator panels, 348, 348f solid-state electronics, 29, 67–68, 70–77, 71f, 75f, 77f Sommerfeld, A., 49

625 SORCE (Solar Radiation and Climate Experiment) measurements, 429, 429f space charge-limited current (SCLI) model, 93–98, 93f, 96f, 97f space satellites, 427–32, 428f, 429f, 432f space ships, 3 space solar arrays about, 406, 407t body-mounted, 407, 407f concentrating, 412–13, 413f electrostatically clean, 414–15 flexible foldout, 409–10, 409f, 410f flexible roll-out, 410–12 high-temperature/high-intensity, 413 Mars solar arrays, 415 rigid panel planar, 408–9, 408f See also arrays space solar cells about, 397–98, 398f, 399f, 400–401, 400f CIGS, a-Si, CdTe, 404–5 current status of, 586 efficiencies of, 417–18, 418t III-V space cells, 401–2 integrated power systems and, 415–17 MJ (multijunction) cells, 402–4, 402t nanostructured solar cells, 404 organics and, 405 See also solar cells spectral irradiance, 306, 446–47, 447f See also irradiance spectral mismatch calculations, 448–49 Spectrolab, 401–2 split data lines, 542–43, 543f, 544f Springerville Generating State Solar System assessment of, 246 BOS (balance of system), 227–28, 228f capacity factor at, 234–35, 234f case study on, 237 configuration of, 225–26 costs for, 243–46, 243t, 245t design, installation of systems, 225, 225t instrumentation/testing at, 226–27 maintenance of, 231–34, 232f, 233t performance of, 228–30, 229f, 230f TEP Company and, 224, 224f Sputnik, 3 SquareRigger, 410–11 Staebler-Wronski effect, 150 STC efficiencies, 593–96, 594t

626 storage challenges, 489–92, 491f SunPower Corporation IBC (interdigitated back contact) cells, 124–26 initial investment of, 261–63, 264f LCOE (levelized cost of energy) forecasting tool, 268, 268f system performance prediction and, 260 trackers, 257, 257f, 258f, 261f sunshape, 450 SUPERs (survivable power systems), 414 survivable concentrating photovoltaic arrays (SCOPAs), 414 system-based analyses, 464–65 TAB (tape-automated bonding), 536, 537f Taiwan market history, 19–20, 19f tellurium, 597–98, 597f temperature degradation, 475–77, 476f, 477f TEP (Tucson Electric Power) Company. See Springerville Generating State Solar System terrestrial solar cells. See solar cells terrorists, 12 tests, 380–82 TF (thin-film) arrays, 410–12 TF (thin-film) modules, 584–85, 584f, 585f, 590–91, 590f, 606–7 See also concentrators; crystalline silicon (c-Si) modules TF (thin-film) solar cells China and, 176–80, 177f, 178f, 179f, 180f future of, 153 history of, 137–41, 138t, 139t, 140f, 140t single-crystal solar cells vs., 63–64, 64f types of, 141–42, 141f X-ray imaging and, 499–500 See also a-Si (amorphous silicon); CdTe (cadmium telluride) solar cells TFTs (thin-film transistors) a-Se X-ray imagers and, 564–65 a-Si (amorphous silicon), 516–20, 517f, 518f, 520f CdTe (cadmium telluride), CZT X-ray imagers, 569 future of, 574 image lag and, 526, 540–41 PbI2, HgI2 imagers and, 571–73

INDEX sensor arrays and, 526–28, 527f, 528f See also p-i-n/TFT pixel architecture thermal noise, 547–48 thermophotovoltaic (TPV) cells, 5 3-sun LCPV (low-concentration photovoltaic) concept about, 276, 276f, 277f, 278, 278f, 279 future of, 285, 287–89, 287f, 287t, 288f, 288t, 289f mirror module development, 279–83, 279t, 280f, 281f, 282f, 283f, 284f, 285, 285f, 286f III-V multijunctions design of, 295–99, 295f, 297f, 298f efficiency of, 299–301, 300f in HCPV (high-concentration photovoltaic) systems, 293–95, 294f, 297f history of, 607 performance of, 304–8, 305f, 307f reliability of, 308–9 tunnel junctions in, 296–97, 297f See also multijunction (MJ) solar cells III-V space cells, 401–2 tilted collectors, 433–36, 434f TIR (total internal reflection), 345, 356f total incident radiation, 468, 470f total solar irradiance (TSI), 428–30, 428f, 429f TPV (thermophotovoltaic) cells, 5 trackers about, 5, 207, 608–9 costs of, 215–17, 216t CPV (concentrating photovoltaic) and, 10 D/A (dual-axis), 208–11, 210f, 210t, 211f geometries for, 209–11, 209f, 210f, 210t, 211f, 211t HCPV (high-concentration photovoltaic) Fresnel concentrator systems, 367–69, 368f linear axis, 212 manufacturing companies of, 208–9 S/A (single-axis), 209–11, 211t, 212, 212f, 213f, 215f SunPower and, 257, 257f, 258f, 261f two-axis Cassegrainian systems, 349–50, 350f See also azimuth trackers; fixed-tilt trackers

INDEX transmission challenges, 492–93 triple-junction solar cells, 322–27, 323f, 325f, 326f, 403 trucks. See automobiles TSI (total solar irradiance), 428–30, 428f, 429f Tucson Electric Power (TEP) Company. See Springerville Generating State Solar System tunnel junctions, 296–97, 297f, 339, 340f See also multijunction (MJ) solar cells; III-V multijunctions Ultraflex Solar Arrays, 411–12 United States advantages of, 606 market history of, 19–20, 19f, 20f, 21f public policies in, 22–28 solar cell budget of, 604, 604t University of Nevada, Las Vegas CPV (concentrating photovoltaic) systems at, 361 MegaModule at, 374, 375f reflective concentrators at, 361–63, 362f refractive concentrators at, 363–65, 363f, 364f 3-sun modules, 285, 285f, 286f University of New South Wales, 122–23, 123f U.S. Department of Defense, 25 U.S. Interstate Highway System, 27–28 U.S. Jet Propulsion Laboratory (JPL), 18f, 21f, 22–23 U.S. National Solar Radiation Data Base, 440 utility-scale systems BOS (balance of system), 227–28, 228f capacity factor of, 234–35, 234f configuration of, 225–26 design, installation of systems, 225, 225t instrumentation/testing of, 226–27 maintenance of, 231–34, 232f, 233t performance of, 228–30, 229f, 230f TEP Company and, 224, 224f

627 valence bands, 68–70, 512f, 513 Van Allen belts, 397–98 Vanguard cells, 397–98 Varian, 275f, 276 Varian Medical Systems 4030CB, 536, 537f voltage. See bandgap voltage; open-circuit voltage wafers, 187–88, 189t Wakondo Technologies, 403–4 wars, 12 water pumping, 5 wave theory, 49–53, 50f, 51f, 52f wave-particle duality, 47–48 Webster effect, 88 Wilson, A. H., 49 wind/solar complementary stations, 181–82 X-ray detectors about, 533–35, 534f construction of, 535–36, 535f, 536f, 537f electronics architecture of, 541–43, 541f, 543f, 544f photoconductor material requirements for, 563, 564t solar cells vs., 559 TFTs (thin-film transistors), photodiodes and, 526–27, 527f See also flat panel detectors (FPDs); photoconductors X-ray image lag / gain effect, 552, 553f X-ray imaging amorphous selenium (a-Se) and, 564–68, 566f, 567f, 568f a-Si (amorphous silicon) and, 9–10, 10f, 555 CdTe (cadmium telluride) solar cells and, 569 history of, 501–2, 560, 561f, 562f, 563 TF (thin-film) solar cells and, 499–500 See also digital imaging; portal imaging X-ray photons, 544–52, 545f Zomeworks Corporation, 208