Surface Contamination and Cleaning

Surface Contamination and Cleaning, Volume 1 K.L. Mittal, Editor VSP Surface Contamination and Cleaning, Volume 1 ...

Author: N. J.) International Symposium on Surface Contamination and Cleaning (2001 : Newark

341 downloads 2462 Views 24MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Surface Contamination and Cleaning, Volume 1

K.L. Mittal, Editor

VSP

Surface Contamination and Cleaning, Volume 1

This page intentionally left blank

SURFACE CONTAMINATION AND CLEANING VOLUME 1

Editor: K.L. Mittal

UTRECHT BOSTON 2003

VSP BV P.O. Box 346 3700 AH Zeist The Netherlands

Tel: +31 30 692 5790 Fax: +31 30 693 2081 [email protected] www.vsppub.com

© VSP BV 2003 First published in 2003 ISBN 90-6764-376-9

All rights reserved. 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 or otherwise, without the prior permission of the copyright owner.

Printed in The Netherlands by Ridderprint bv, Ridderkerk

Contents

Preface Mapping of surface contaminants by tunable infrared-laser imaging D. Ottesen, S. Sickafoose, H. Johnsen, T. Kulp, K. Armstrong, S. Allendorf and T. Hoffard

vii

1

Monitoring cleanliness and defining acceptable cleanliness levels M.K. Chawla

23

Tracking surface ionic contamination by ion chromatography B. Newton

43

A new method using MESERAN technique for measuring surface contamination after solvent extraction M.G. Benkovich and J.L. Anderson

49

Methods for pharmaceutical cleaning validations H.J. Kaiser

75

Influence of cleaning on the surface of model glasses and their sensitivity to organic contamination W. Birch, S. Mechken and A. Carré

85

Decontamination of sensitive equipment R. Kaiser and K. Haraldsen

109

The fundamentals of no-chemistry process cleaning J.B. Durkee II

129

Development of a technology for generation of ice particles D.V. Shishkin, E.S. Geskin and B. Goldenberg

137

Cleaning with solid carbon dioxide pellet blasting F.C. Young

151

Development of a generic procedure for modeling of waterjet cleaning K. Babets and E.S. Geskin

159

vi

Contents

Experimental and numerical investigation of waterjet derusting technology K. Babets, E.S. Geskin and B. Goldenberg

173

Practical applications of icejet technology in surface processing D.V. Shishkin, E.S. Geskin and B. Goldenberg

193

Correlating cleanliness to electrical performance T. Munson

213

Qualifying a cleaning system for space flight printed wiring assemblies J.K. “Kirk” Bonner and A. Mehta

225

Investigation of modified SC-1 solutions for silicon wafer cleaning C. Beaudry and S. Verhaverbeke

241

Performance qualification of post-CMP cleaning equipment in a semiconductor fabrication environment M.T. Andreas

249

Spatial and temporal scales in wet processing of deep submicrometer features M. Olim

261

Microdenier fabrics for cleanroom wipers J. Skoufis and D.W. Cooper

267

Fine particle detachment studied by reflectometry and atomic force microscopy A. Feiler and J. Ralston

279

Dust removal from solar panels and spacecraft on Mars S. Trigwell, M.K. Mazumder, A.S. Biris, S. Anderson and C.U. Yurteri

293

Laser cleaning of silicon wafers: Prospects and problems M. Mosbacher, V. Dobler, M. Bertsch, H.-J. Münzer, J. Boneberg and P. Leiderer

311

Particle removal using resonant laser detachment K. Kearney and P. Hammond

335

The future of industrial cleaning and related public policy-making C. LeBlanc

345

Surface Contamination and Cleaning, Vol. 1, pp. vii–viii Ed. K.L. Mittal © VSP 2003

Preface This volume chronicles the proceedings of the International Symposium on Surfase Contamination and Cleaning held under the aegis of MST Conferences in Newark, New Jersey, May 23–25, 2001. Even a cursory look at the literature will evince that there has been tremendous interest and R&D activity in the arena of surface contamination and cleaning, so we decided to organize this symposium. Because of the importance of this topic in many technological areas, tremendous efforts have been devoted to devise novel and more efficient ways to monitor, analyse and characterize contamination on surfaces as well as ways to remove such contamination from a wide variety of surfaces. The ubiquitous nature of surface contamination causes concern to everyone dealing with surfaces, and the world of surfaces is wide and open-ended. A contaminant is defined as “unwanted matter or energy” or “material or energy in the wrong place”. Also contaminants can by broadly classified as: film-type, particulates; ionic, and biological or microbial. The technological areas where surface contamination has always been a bete noire and thus surface cleaning is of cardinal importance are too many and range from aerospace to microelectronics to biomedical. Here a few eclectic examples will suffice to underscore the importance of surface contamination and cleaning. In the world of ever-shrinking device dimensions in the microelectronics, the need to remove ever smaller particles (of nanosize dimension) is quite patent. On the other hand, film-type (organic) contamination is of crucial importance in the area of adhesive bonding, as even a very thin layer of contamination can be very detrimental in attaining good bond strength. In operation theaters, the concern about microbial contamination is all too obvious. So in light of the great concern about surface contamination, people dealing with surfaces are rightfully afflicted with molysmophobia.* The technical program for this symposium comprised 45 papers dealing with all kinds of contaminations on a host of surfaces, and many ramifications of surface contamination and cleaning were addressed. There were brisk and illuminating (not exothermic) discussions, both formally and informally, throughout the symposium. Also if comments from the participants are a barometer for the success of a symposium then this event was quite successful. Now coming to this volume, it contains a total of 24 papers (others are not included for a variety of reasons). It must be recorded that all manuscripts were rigorously peer reviewed and suitably revised (some twice or thrice) before inclusion in this volume. So this volume is not a mere collection of unreviewed papers −

viii

Preface

which is generally the case with many symposia proceedings − rather it reflects information which has passed peer scrutiny. The topics covered include: mapping of surface contaminants; various techniques for cleaning surfaces; various techniques for monitoring level of cleanliness; acceptable cleanliness levels, ionic contamination; pharmaceutical cleaning validations; cleaning of glass surfaces; decontamination of sensitive equipment; no-chemistry process cleaning; waterjet cleaning; cleaning with solid carbon dioxide pellet blasting; cleanroom wipers; dust removal from solar panels and spacecraft on Mars; laser cleaning of silicon surfaces; particle removal; implications of surface contamination and cleaning; and future of industrial cleaning and related public policy-making. I sincerely hope that this volume addressing many aspects and recent developments in the domain of surface contamination and cleaning will be of interest to a wide range of people working in many different industries. Acknowledgements It is always a pleasure to write this particular segment of a book as it offers the opportunity to thank those who helped in many ways. First, my sincere thanks are extended to my colleague and friend, Dr. Robert H. Lacombe, for taking care of the organizational aspects of this symposium. The comments from the peers are a sine qua non to maintain the highest standard of a publication, so I am most appreciative of the time and efforts of the unsung heroes (reviewers) in providing many valuable comments. I am profusely thankful to the authors for their interest, enthusiasm and contribution without which this book would not have seen the light of day. In closing, my thanks go to the staff of VSP (publisher) for giving this book a body form. K.L. Mittal P.O. Box 1280 Hopewell Jct., NY 12533

*Molysmophobia means fear of dirt or contamination, from Mrs. Byrne’s Dictionary of Unusual, Obscure, and Preposterous Words, University Books, Secaucus, NJ (1974).

Surface Contamination and Cleaning, Vol. 1, pp. 1–22 Ed. K.L. Mittal © VSP 2003

Mapping of surface contaminants by tunable infrared-laser imaging DAVID OTTESEN, SHANE SICKAFOOSE,∗ HOWARD JOHNSEN, TOM KULP, KARLA ARMSTRONG, SARAH ALLENDORF and THERESA HOFFARD1 Sandia National Laboratories, P.O. Box 969, MS 9403, Livermore, CA 94551-0969 1 Naval Facilities Engineering Service Center, 1100 23rd Avenue, Port Hueneme, CA 93043-4370

Abstract—We report the development of a new, real-time non-contacting monitor for cleanliness verification based on tunable infrared-laser methods. New analytical capabilities are required to maximize the efficiency of cleaning operations at a variety of federal (Department of Defense [DoD] and Department of Energy [DOE]) and industrial facilities. These methods will lead to a reduction in the generation of waste streams while improving the quality of subsequent processes and the long-term reliability of manufactured, repaired or refurbished parts. We have demonstrated the feasibility of tunable infrared-laser imaging for the detection of contaminant residues common to DoD and DOE components. The approach relies on the technique of infrared reflection spectroscopy for the detection of residues. An optical interface for the laser-imaging method was constructed, and a series of test surfaces was prepared with known amounts of contaminants. Independent calibration of the laser reflectance images was performed with Fourier transform infrared (FTIR) spectroscopy. The performance of both optical techniques was evaluated as a function of several variables, including the amount of contaminant, surface roughness of the panel, and the presence of possible interfering species (such as water). FTIR spectra demonstrated that a water film up to 7 µm thick would not interfere with the effectiveness of the laser-imaging instrument. The instrumental detection limit for the laser reflectance imager was determined to be on the order of a 10-20 nm thick film of a general hydrocarbon contaminant. Keywords: Infrared; tunable-laser; imaging; cleaning; surface contamination.

1. INTRODUCTION

Real-time techniques to provide both qualitative and quantitative assessments of surface cleanliness are needed for a wide variety of governmental and industrial applications. The range of potential applications include aircraft, shipboard, vehicle, and weapon component surfaces to be coated, plated, or bonded. The avail∗

To whom all correspondence should be addressed. Phone: (925) 294-3526, Fax: (925) 294-3410, E-mail: [email protected]

2

D. Ottesen et al.

ability of a convenient analysis technology for on-site, post-cleaning determination of surface contamination will allow more rapid and accurate assessments of the efficiency of chosen cleaning techniques. By developing an on-line technique, processed parts or extracted samples will not have to be sent to a separate laboratory for analysis, thereby eliminating processing delays. The information provided by the optical method will assist the process operator in distinguishing between specific contaminants and determining subsequent actions to be taken. In this paper we report the development of an infrared laser-based imaging approach that will reduce the use, emission, and handling of waste-stream materials in cleaning operations. This work is supported by the separate development of a hardened, portable Fourier transform infrared (FTIR) reflectance instrument at the Naval Facilities Engineering Service Center (NFESC), Port Hueneme, CA in cooperation with the Surface Optics Corporation. Simultaneous development of an FTIR instrument is complementary in nature to the laser-imaging technique and is described in detail elsewhere [1]. Both instruments will be used primarily for the real-time on-line or nearly on-line detection of contaminant residues on reflective surfaces. In each case, surface contamination is detected by its absorption of a grazing-incidence infrared beam reflected from the surface. The instruments differ in the nature of the information they provide. The laserbased instrument produces images that directly indicate the spatial extent and location of infrared-absorbing surface hydrocarbon contaminants. In contrast, FTIR instrumentation provides a wide-band spectral measurement of the surface reflectance averaged over a small area for nearly all organic materials, and many inorganic components. Thus, the laser-imaging system allows the rapid determination of surface cleanliness for organic residues over a large area, while the spectrallyresolved FTIR method is useful in identifying the specific molecular composition of a surface contaminant at a particular location. The imaging system under development employs a widely tunable infraredlaser illumination source in conjunction with an infrared camera. This approach provides an on-line technique for surveying contamination levels over large surface areas in a real-time imaging mode. The laser is broadly-tunable over the 1.34.5 µm wavelength range, thus allowing the detection of many hydrocarbon contaminants via absorption bands associated with CH-, OH-, and NH-stretching vibrations. Currently, the detection and identification of surface contaminants on reflective surfaces is conveniently and rapidly done by FTIR reflectance methods. These non-destructive, non-contacting optical techniques identify the chemical constituents of the contaminants, and can yield quantitative measurements with appropriate calibration. Infrared optical methods are particularly useful for cleanliness verification since the surface is probed under ambient conditions. More sensitive high-vacuum electron and ion spectroscopic techniques (X-ray photoelectron spectroscopy, Auger electron spectroscopy, and secondary-ion mass spectrometry) are not suited for on-line application.

Tunable IR-laser mapping of surface contaminants

3

Commercial instruments that employ infrared reflectance spectroscopy are available for surface analysis and provide both quantitative and qualitative information on surface coatings. These instruments are limited in their ultimate sensitivity to surface contaminants by the nature of their optical design. Infrared radiation is focused onto the surface to be analyzed at a near-normal angle of incidence, resulting in a compact hand-held apparatus. The infrared light is collected as either specularly or diffusely reflected radiation depending on the roughness and scattering properties of the surface [2, 3]. The resulting sensitivity to very thin layers of surface species is limited by poor coupling of the incident electromagnetic field with the vibrating dipoles of the surface molecular species [4-6] in layers less than 0.1 µm thick. In order to maximize the sensitivity of infrared reflectance measurements for absorption bands of thin layers of contaminants on metallic surfaces, theoretical and experimental studies [7-9] have shown that the angle of incidence of infrared radiation on the surface should be increased to at least 60° from the surface normal. This is also true for many thin-film residues on the surface of non-metals, such as dielectrics and semiconductors (although the detectability of contaminant absorption bands under these circumstances depends strongly on the optical constants of both surface and substrate, and any absorption features intrinsic to the non-metallic substrate). Additional sensitivity in the reflectance measurement is obtained by measuring only the component of the reflected infrared radiation polarized parallel to the plane of incidence. This experimental method is variously referred to as, “grazing-angle” reflectance spectroscopy or infrared reflectionabsorption spectroscopy (IRRAS). We have adapted the technique of “grazingangle” reflectance spectroscopy to utilize the newly developed tunable-laser source. 2. EXPERIMENTAL

The laser-based instrument described in this report offers the capability to rapidly survey large surface areas and to determine the location and extent of residual hydrocarbon contaminants following cleaning operations. In contrast, a spectroscopic analysis by an FTIR-based infrared reflectance instrument is able to characterize a very broad range of organic constituents and many inorganic species. However, a surface-probing FTIR instrument measures a spectrum at only a single small area on a sample, thus requiring broad area surveys to be done by sequentially probing many points. Even at a rate of ~ 10 seconds per measurement point, this can be a time-consuming process. The rate of measurement by FTIR spectroscopy is constrained by the relatively low spectral brightness (compared to a laser) of the incandescent illumination sources. This makes it necessary to use relatively long integration times to achieve an acceptable signal-to-noise ratio. The tunable-laser-based instrument overcomes these limitations by illuminating a broad surface area with a high-brightness infrared laser. This approach allows a

4

D. Ottesen et al.

single-wavelength reflectance measurement over an area of several square centimeters to be made on a timescale of less than a second. In order to acquire measurements at multiple wavelengths, the laser is tuned and an image is collected at each of the desired wavelengths. While a detailed spectral map of a surface can be generated over the laser tuning range, the primary use of the system is to provide rapid areal surveys at a few key wavelengths that are indicative of hydrocarbon contaminants. The detection sensitivity for several hydrocarbon species at various illumination wavelengths was evaluated in this work, as well as a method to suppress image noise due to laser speckle while maintaining high illumination intensity. 2.1. Quasi-phasematching tunable infrared laser The broadly-tunable infrared laser illuminator is based on a technology called quasi-phasematching (QPM) [10]. This approach has been exploited to increase the tuning range and power of the infrared light source while reducing its size. For example, continuous-wave (cw) optical parametric oscillators (OPOs) that employ the QPM material, periodically-poled lithium niobate (PPLN), are capable of tuning over the 1.3-4.5 µm spectral region while emitting more than 0.5 W of power. This technique has been used to generate tunable infrared laser light for imaging natural gas emissions, and developing laser-based spectroscopic gas sensors [10]. In this work we are extending it to the analysis of hydrocarbon residues on material surfaces. The limit of the current tuning range of the PPLN-based laser at long wavelengths is about 4.5 µm (2222 cm-1) due to the transmission characteristics of lithium niobate. This property restricts the sensitivity of the chemical imaging system to functional groups containing hydrogen atoms (C-H, N-H, O-H). Extension of the laser tuning wavelength range beyond 5 µm (2000 cm-1) is desirable to provide specific identification of hydrocarbon and some inorganic molecular species. The light source assembled for the IR imaging sensor is an OPO pumped by a continuous-wave (cw) Nd:YAG laser, as shown in Figure 1 [10]. An electric field is induced in the OPO’s PPLN crystal by the electric field of the pump laser; these fields interact to form two new laser beams whose frequencies sum to the frequency of the pump laser. The reflectivities of the mirrors in the optical cavity are selected to resonate one of the generated waves, while the other wave is simply generated and released from the cavity. The resonated wave is called the signal; the non-resonated wave is called the idler. The exact frequencies of the signal and the idler are determined by the phasematching properties of the crystal (described below), the reflectivity of the cavity, and by any spectrally-selective optics that may be added to the laser cavity (e.g. an étalon). While either the signal or the idler beam can be used for measurements, only the idler is used in the experiments reported here. As shown in Figure 1, the OPO used in the imaging sensor is of the “bowtiering” design. A diode-pumped, cw, multimode Nd:YAG laser (Lightwave Elec-

Tunable IR-laser mapping of surface contaminants

5

Figure 1. Diagram of the PPLN OPO and projection optics.

tronics, Mountain View, CA) that is capable of generating at least 6 W of output power at a wavelength of 1064 nm is used as the OPO pump source. Two flat mirrors (M3 and M4) and two curved mirrors (M1 and M2, 50-mm radius of curvature), all coated to be highly reflective at the signal and highly transmissive at the pump and idler wavelengths, form the bow-tie-shaped, single-wavelength resonant ring oscillator cavity designed to resonate the signal wave. An antireflection-coated lens, positioned between the pump laser and the OPO cavity, serves to image the Gaussian pump beam into the PPLN crystal. In this way, a beam waist (E-field radius) of 70 µm is created in the center of the crystal, which itself is centered between the two curved cavity mirrors. During normal operation, the OPO resonates on a single signal mode for minutes at a time, whereupon it hops to another cavity mode. The idler bandwidth is, however, determined by that of the pump beam, which is 10-15 GHz. The use of the QPM material, PPLN, makes cw OPO operation more tunable and efficient than it would be for a conventional birefringently phasematched crystal. Simply stated, phasematching is a condition in which all of the interacting waves (i.e., signal, pump, and idler) maintain a specified relative phase relationship as they propagate through a nonlinear medium, and is a necessary condition for efficient nonlinear generation. In birefringent materials, phasematching is

6

D. Ottesen et al.

achieved by careful selection and/or control of the crystal birefringence, temperature, and beam propagation angles. In a QPM medium, phasematching is designed into the medium during the crystal growth process. Phasematching is achieved by causing the crystal to have a periodically inverting optical axis. The engineering process used to create these crystals increases conversion efficiency by allowing the use of much stronger nonlinear coefficients of the crystal, and frees the system from reliance on birefringence thereby increasing tunability. As the light beams cross the crystal-axisinverting boundaries, any relative dephasing of the waves is corrected. For a crystal of a given periodicity, the rephasing is effective for a particular set of pump, signal, and idler frequencies. Some degree of tuning of these waves can be achieved within the crystal phasematching bandwidth (typically 10-20 cm-1). Broader tuning is achieved by accessing a portion of the same crystal having a different periodicity, or by changing the temperature of the crystal. In the present system, two 50-mm-long PPLN crystals (Crystal Technology, Palo Alto, CA) with an aperture of 11.5 mm × 0.5 mm are used as the active medium. Each crystal contains eight poled regions with different periodicities. One crystal’s periodicities range from 28.5 to 29.9 µm, and of the other crystal from 30.0 to 31.2 µm. When operating at a crystal temperature of 148°C, these periods collectively allow tuning of the idler from 2720 to 3702 cm-1. The crystals are mounted in a stacked fashion within a temperature-stabilized copper oven that is attached to a vertical translation stage. Each crystal is tuned by selecting a period using the vertical motion of the stage; horizontal motion of the oven is used to select between the two crystals. The raw output of the OPO contains the idler beam as well as portions of the signal and pump beams and some higher-order (red, green) beams created spuriously in the PPLN crystal. Spectral filtering is used to dump all but the idler beam. Prior to illumination of the sample, the idler is passed through a set of projection optics, also shown in Figure 1. The first of these is a ZnSe diffuser (mean roughness of ~ 3-4 µm) that is mounted on a motor-driven spindle. The diffuser serves to reduce the phase coherence of the idler in order to minimize laser speckle noise in the transmitted beam and viewed by the IR camera in the light reflected from the sample surface. The cone of radiation leaving the diffuser is collected by a ZnSe faceted lens (Laser Power Optics, Murrieta, CA). The faceted lens is formed to contain the equivalent of 16 6.4 mm facets and 16 partial facets around the edge of the lens on a 3.8 cm diameter with an effective f-number of 1.7. It operates as a prism array – the expanded beam is segmented into 32 different square beamlets that are subsequently overlapped at a distance of 5 cm from the surface of the lens. A ZnSe wire grid polarizer (not shown in Figure 1) is located at the overlap point, and serves to produce a p-polarized beam for the infrared reflectance measurement. The square-shaped overlap region is then imaged onto the target using an f/1.7, 8.4 cm focal-length ZnSe lens. As a unit, the system converts the Gaussian profile of the idler beam into a uniform square illumination on the sample surface.

Tunable IR-laser mapping of surface contaminants

7

The infrared laser light is incident on the sample surface at an angle of 60° from the surface normal, and the specularly reflected component is detected by an InSb focal-plane array (FPA) camera with an infrared macro-lens assembly and an array size of 256 x 256 pixels. The FPA camera is located approximately 0.3 m from the sample surface, and the resulting field of view is 20 x 35 mm. FTIR instruments at both Sandia and NFESC were used to characterize the mid-infrared spectra of contaminated surfaces via optical interfaces for grazingangle reflectance spectroscopy. The system at NFESC uses a commercially available sampling accessory that permits a variable angle of incidence from 30 to 80°, which is convenient for evaluating detection limits for contaminants on a variety of surfaces. The optical interface used by the Sandia National Laboratories FTIR instrument was constructed with a fixed 60° angle of incidence with optics external to the spectrometer. It also differs from the NFESC system in the large solidangle used both to illuminate the surface and collect reflected light. This feature is particularly useful in the examination of rougher surfaces that cause significant scattering of the infrared beam, with a consequent degradation in both signal/noise ratio and detection limits. Both systems use infrared polarizers to enhance the sensitivity of the measurements by restricting the surface illumination to p-polarization [4]. Unless otherwise noted, all reflectance spectra presented in this paper are for p-polarized measurements. 2.2. Test sample preparation for calibration In order to evaluate the usefulness of the laser-imaging technique as a cleaning verification method, we prepared a number of test surfaces with wellcharacterized levels of contamination. These were used to determine detection limits as a function of contaminant species, level of contamination, degree of surface roughness, effect of spectral interference, and instrumental parameters such as angle-of-incidence. Seven candidate materials were chosen as contaminant species for evaluation as shown in Table 1. These materials have proven to be particularly difficult to remove during cleaning operations, and are representative of many other organic contaminants encountered in government and industrial cleaning processes. Detailed measurements on the first four materials have been made in the course of this work and preliminary measurements have been made on the remaining three. A number of metals were chosen as substrates for the target contaminants, based on usage information obtained from military and contractor facilities. These were Aluminum-7075-T6, Titanium 6Al-4V, Steel Alloy 4340, Stainless Steel 304, and Magnesium AZ31B. The metals were fabricated into 3.8 x 12.7 cm flat coupons for laboratory testing and method demonstration. Six surface roughness finishes of the Aluminum 7075-T6 test coupons were obtained, ranging from 80 to 600 grit (600 grit being the smoothest). A profilometer instrument was used to examine the surface roughness profiles and provide average Ra values. A Ra value is an arithmetic average of the absolute deviations

D. Ottesen et al.

8

Table 1. Contaminant materials used for preparation of test coupon for calibration Material

Description

Usage

Drawing Agent

White soft solid – ester grease

Lubricant

Brown liquid – paraffin hydrocarbons

Silicone Mold Release 1 Mold Release 2

Silicone Green liquid – ethanol homopolymer Clear liquid – proprietary polymeric resins Yellow liquid – abietic acid or anhydride Blue liquid – castor oil base

Metal drawing, cutting, and lubricating agent Rust preventative, cleaner, lubricant, protectant for metals Lubricant Mold release agent Mold release agent

Solder Flux Hydraulic Oil MIL-H-5606A AM2

Soldering flux for electrical and electronic applications Hydraulic systems, shock and strut lubricant

from the mean surface level, in millionths of an inch; therefore, a Ra value of 1.5 = 0.00000015 inches (3.8 µm). Due to the nature of metal-shop finishing processes, surface roughness values vary considerably across a given surface area. Finishing operations also result in a directional “grain” parallel to the sample coupons’ longitudinal direction. Surface roughness measurements, therefore, exhibit large variations between measurements taken along an orientation longitudinal or transverse to the polishing axis. Two surface roughness levels, 600 and 220 grit, were obtained for the other metal alloys. Prior to contaminant application, the aluminum alloy coupons were cleaned with acetone and underwent sonication with a clean-rinsing aqueous cleaner. They were then thoroughly rinsed in distilled water and dried in an oven at 50°C. Once cooled, they were weighed on a microbalance with a precision of 0.01 mg. Two or three weighings were averaged. Both drawing agent and lubricant contaminated Al-7075 coupons were produced by two primary deposition methods – airbrushing and manual brushing. Several other techniques were attempted, including “wire-cator” drawing, coupon spinning, and “manual drop and spread.” These techniques were not used to produce test samples for calibration for these particular contaminants due to the superior results obtained from airbrushing and manual brushing. Three levels of drawing agent were applied by airbrushing to three Al test coupons for each of the six surface finishes, creating a suite of 18 panels. Varying concentrations of drawing agent in water were prepared for the airbrush solutions. Similarly, four levels of lubricant were applied to four Al test coupons for each of six surface finishes, creating a suite of 24 panels. Manual brushing was used for all but the least contaminated samples, which were airbrushed. Lubricant solutions for both tech-

Tunable IR-laser mapping of surface contaminants

9

niques were prepared using pentane as the solvent. Similar methods were used in preparing calibration samples with the mold release, solder flux, and hydraulic oil samples. All contaminated coupons were gently heated in an oven at 50°C for several days to remove both semi-volatile and volatile components. This served to stabilize the contaminants, allowing for quantification by weighing. Once the weights became stable, the coupons were cooled and weighed to determine the amount of contaminant present on the surface. When not being weighed or examined, the coupons were kept in a desiccator. 3. RESULTS AND DISCUSSION

Grazing-angle incidence reflectance spectroscopy acts to enhance the detection sensitivity for thin layers of residue predominantly through improved coupling of the electric field intensity of the incident beam with the vibrating dipoles of the surface contaminant layer perpendicular to the metallic surface. Some additional enhancement of the infrared absorption spectrum will also occur due to a lengthening of the effective path length through the absorbing thin film layer [4-6]. If the optical properties of both thin film and substrate are known (or can be determined), the reflection-absorption spectrum can be calculated as a function of film thickness and angle of incidence. This capability is particularly useful for interpreting experimental data and designing optical instrumentation. Computer codes written at Sandia [7] performed these calculations for a variety of materials. 3.1. FTIR measurements FTIR reflectance data for the full drawing-agent sample set were obtained at NFESC and Sandia using angles of incidence of 75 and 60° for average film thickness ranging from 0.1 to 1 µm, and aluminum substrates with surface finish ranging from 600 to 80 grit. Since the surface finishing operation produced a highly directional roughness, measurements were made both longitudinally and transversely with respect to the polishing grooves. Ra values were determined at NFESC using profilometer measurements, and resulted in surface roughness values of 0.3 to 1.5 µm for the longitudinal direction, and 0.5 to 6 µm for the transverse direction. The FTIR reflectance spectra were normalized using the uncoated back of a panel as a clean reference standard, and the intensity data are presented as either reflectance or –log reflectance in the following discussion. The C-H stretching vibrations near 2900 cm-1 proved to be generally useful in quantifying instrument response since these frequencies are well isolated from atmospheric interference due to water vapor and carbon dioxide. However, the baseline for these reflectance data was often non-linear. A simple single-point measurement of intensity was therefore not sufficient to determine the instrument response function.

10

D. Ottesen et al.

Optical constants (n and k) were derived for the contaminant C-H stretching vibrations using the Sandia reflectance code and a dispersion model to calculate a fit to the experimental data for one of the test coupons [7]. Reflectance-absorption spectra for the 2800-3000 cm-1 range were calculated for 1-µm thick films of a specific hydrocarbon contaminant on an aluminum surface at either 60 or 75° angle of incidence. This function was then used as a linear variable in conjunction with a second-order polynomial to produce a least-squares fit of the experimental reflectance data for the test coupons. An example is shown in Figure 2 for the longitudinal measurements of three thicknesses of drawing-agent contaminant at

Figure 2. Linear least-squares fit of experimental reflectance data for drawing-agent contaminant on 600 grit polished aluminum surfaces. Average film thickness: (Top) 0.9 µm, (Middle) 0.4 µm, (Bottom) 0.1 µm.

Tunable IR-laser mapping of surface contaminants

11

Figure 3. Integrated reflection-absorption intensity at 60° angle-of-incidence for C-H stretching bands of drawing-agent films deposited on aluminum test coupons with varying degree of surface roughness (longitudinal, top; transverse, bottom).

75° angle-of-incidence. This procedure produces extremely rapid, robust analyses of the FTIR reflectance data, even for very thin films in the presence of noise, and accounts for baseline shifts and curvature due to interference fringes. Fitting coefficients for the linear spectral function (which are proportional to the integrated intensity) are plotted against the average calculated film thickness, and these results are shown in Figures 3 and 4 for longitudinal and transverse reflectance measurements at 75° and 60° angle-of-incidence, respectively. Results for

12

D. Ottesen et al.

Figure 4. Integrated reflection-absorption intensity at 75° angle-of-incidence for C-H stretching bands of drawing-agent films deposited on aluminum test coupons with varying degree of surface roughness (longitudinal, top; transverse, bottom).

the longitudinal, 60° angle-of-incidence follow a linear relationship with film thickness except for the roughest surface finish (80 grit, Ra = 1.5 µm). The instrument response functions for transverse measurements at 60° angle-of-incidence are also reasonably linear, with the same average slope as seen in Figure 3. In contrast, analysis of the FTIR reflectance data at 75° angle-of-incidence for both longitudinal and transverse sample orientations shows a marked departure from linearity at the highest values of film thickness (Figure 4). The initial slopes

Tunable IR-laser mapping of surface contaminants

13

of the spectral response, the integrated reflection-absorption intensity, of these samples are slightly greater than the intensity of the spectral response of the same samples measured via a 60° angle of incidence data (Figure 3). This behavior is expected due to the increase in reflection-absorption sensitivity with increasing angle of incidence. Here, too, the average initial slope (and hence instrument sensitivity) is the same for both transverse and longitudinal orientations. The pronounced non-linearity in slope for the thickest films at 75° angle-ofincidence was unexpected. An increasingly non-linear response may be observed for thicker absorbing films, and this effect will become more pronounced as the angle of incidence is also increased. The interpretation of the data implying that measurement of a thicker film, sampled at a steeper angle, generated the observed non-linearity in the data is not substantiated by the calculated spectra for the present measurement conditions due to the small change from 60 to 75° in the angle of incidence. Furthermore, such a non-linear effect would be most pronounced for measurements on the smoothest substrate (Figure 4, filled circles) where the effective local orientation of the surface is most constant with respect to the illumination beam. Instead of observing such non-linear behavior the measurements made on the smoothest surface are by far the most linear sample series for the 75° data. We attribute the pronounced non-linearity of the 75° data for the thickest drawing-agent films to the morphological characteristics of the material as deposited on the aluminum test panel surface. As described above, the drawing-agent material is highly viscous and forms a visibly heterogeneous white film at 1-µm thickness. Variations in the deposition process produce relatively thick local areas of drawing-agent film and result in accretion of solid residue along the polishing grooves and ridges of the aluminum substrate. Under these circumstances, illumination of the surface with the FTIR beam at an angle of 75° may result in shadowing by contaminant material on ridge structures for all except the smoothest (600 grit polish) surface. The 12-mm diameter focal area of the infrared beam is elongated by a factor of four for this angle of incidence. In contrast, reflectance measurements at 60° result in only a factor of 2 elongation, and minimize the shadowing effect of thick films except for ridges on the roughest (80 grit polish) surfaces. This interpretation is substantiated by reflectance data for the second test set (lubricant material) as shown in Figure 5. FTIR reflectance measurements have been made at 75° angle-of-incidence for a test series similar to that of the drawing-agent set. An analysis of the C-H stretching frequencies shows a strikingly more linear dependence of instrument response with film thickness (with the exception of a single point for one of the panels with a 220 grit surface finish). We believe that this is due to the more fluid characteristic of the lubricant material, which allows the deposited film to conform much more closely to the surface topography of the test coupons. This behavior may also account for the stronger dependence of the integrated intensity slope with surface roughness, when compared to the nearly constant results for the drawing-agent contaminant examined above.

14

D. Ottesen et al.

Figure 5. Integrated reflection-absorption intensities of C-H stretching bands for lubricant films deposited on aluminum test coupons with varying degree of surface roughness for longitudinal illumination.

Even though excellent sensitivity was demonstrated for common hydrocarbon contaminants using grazing-angle infrared reflectance spectroscopy, concerns remain due to potential interference from other molecular species that may be present in the measurement environment. Chief among these is water, resulting either from cleaning operations or the local environment. Water is a very strong infrared absorber, and its presence on the surface to be measured may cause distortion or obscuration of the characteristic contaminant reflection spectrum. We performed an evaluation of this interference using lubricant-contaminated test panels with an average hydrocarbon thickness of 0.7 µm on aluminum. A water film was created on the surface of the test coupon using an airbrush, and reflection-absorption measurements were acquired at a 75° angle of incidence for several conditions. The thickness of the water film was difficult to determine due to continuous evaporation during the reflectance measurements. We estimated the thickness by measuring coupon weight gain immediately prior to and following the infrared measurements. Film thickness was calculated based on the average weight gain. Reflection-absorption spectra are presented in Figure 6 for three water films on the lubricant-contaminated test panel. These water films range in thickness from 1 µm (not visible to the eye) to 7 µm (clearly visible to the eye). Substantial interference is present in the 1700 cm-1 spectral range (not shown) due to the strong HO-H bending mode. This strong absorption obscures carbonyl absorption features that may be present in some, but not all, hydrocarbon contaminant species. The

Tunable IR-laser mapping of surface contaminants

15

Figure 6. Potential interference effects of water on C-H stretching bands of hydrocarbon lubricant film (0.7 µm) on aluminum. Three thicknesses of water film were examined (1 µm, top; 3 µm, middle; and 7 µm, bottom).

broad H-OH stretching bands centered near 3400 cm-1, however, do not obscure the C-H stretching bands near 2900 cm-1. This is particularly important for the effective and accurate use of the tunable infrared-laser imaging instrument, since images are acquired for only a small number of frequencies near 3000 cm-1, in contrast to the broad-band spectral data collected by the FTIR instrument. 3.2. Tunable infrared-laser imaging Initial images of test panel surfaces were acquired at two frequencies (2915 and 3000 cm-1) that correspond to highly absorbing and non-absorbing portions, respectively, of the hydrocarbon infrared spectrum (see above, Figures 2 and 6). We used an acquisition time of 0.5 ms per frame, and averaged a minimum of 20 frames for each frequency in order to reduce noise (shot noise and laser speckle noise). Although the InSb FPA camera is square (256 x 256 pixels), the aspect ratio of the surface area scanned by the spectrometer and the resulting images in this work are elongated by a factor of two due to the trigonometric effects of the 60° angle of incidence and reflectance. Images were acquired for illumination transverse to the polishing direction. They have been corrected for thermal background emission and normalized for system spectral response at the measurement frequencies. The normalization fac-

16

D. Ottesen et al.

Figure 7. Gray-scale on-resonance (2915 cm-1, top) and off-resonance (3000 cm-1, bottom) images for an aluminum test panel contaminated with hydrocarbon drawing agent of 0.9-µm thickness.

tor was determined by the average intensity ratio of a clean surface (the uncontaminated back surface of the test panel) for the two measurement frequencies. The ratios of successive images using the PPLN-based laser system showed a noise level of 0.44% for the entire 65,536-pixel image under our current operating conditions. This noise level corresponds to a hydrocarbon film thickness of approximately 10-20 nm for the species examined in this report, and is the primary factor in determining the present instrumental detection limit. Gray-scale images at these two frequencies for the hydrocarbon drawing-agent (thickness of 0.9 µm on aluminum) are shown in Figure 7. Structure in the images is primarily in the form of vertical lines that represent ridges in the aluminum substrate formed during surface polishing operations. A darker vertical band near the center of the image manifests the presence of an absorbing hydrocarbon in the

Tunable IR-laser mapping of surface contaminants

17

Figure 8. Reflectance images and line-intensity profiles for an aluminum test panel contaminated with a hydrocarbon drawing-agent of 0.9-µm thickness. Laser coherence noise (A) and results of Gaussian smoothing (C) are illustrated with corresponding intensity profiles (B and D, respectively) sampled along the horizontal lines superimposed on the images.

2915 cm–1 image. However, it is difficult to differentiate the absorbing organic film from the high contrast presented by the surface polishing marks in images at a single wavelength. The image created from the ratio of the two images, corrected for thermal background and normalized for the average image intensity, is a relative reflectance image, as shown in Figure 8 (A), assuming that the reflectance of the substrate remains constant at these two frequencies. Unprocessed image ratios such as these show a periodic grid pattern due to coherent interference effects that tend to obscure the hydrocarbon image, and we have investigated several image enhancement procedures to reduce noise while maintaining spatial resolution and contrast in the reflectance ratio images. Weighted Gaussian smoothing in a 7 x 7 pixel neighborhood and Fourier filtering have both been successful in suppressing this noise without significant degradation in spatial resolution, as shown in Figure 8 (B). The image ratios presented in this report have all been Gaussian smoothed. Reflectance intensity profiles along the horizontal line in each image ratio are also shown in Figure 8 (C) and (D) to demonstrate the magnitude of laser coherence noise and the effects of the smoothing procedure.

18

D. Ottesen et al.

Examples of reflectance ratio images for several test surfaces are shown in Figures 9 and 11 in false color. A calibrated color-table (“Rainbow”) for these falsecolor images is shown in Figure 10. Images for a series of 600-grit polished aluminum substrates contaminated with drawing agent are presented in Figure 9. These are the same specimens whose FTIR spectra are shown in Figure 2. Average film thicknesses for the three samples are 0.9 µm (top, left), 0.4 µm (middle, left), and 0.1 µm (bottom, left). The images are presented in false color format with identical dynamic range to help visualize the location of contaminants. Hydrocarbon material was manually deposited along the orientation of the surface polishing grooves, which is oriented vertically in these images. Heavy deposits of the hydrocarbon residue are easily

Figure 9. False-color reflectance images and thickness profiles for three aluminum test panels contaminated with a hydrocarbon drawing agent (thicknesses are: 0.9 µm, top-left; 0.4 µm, middle-left; 0.1 µm, bottom-left). Corresponding line thickness profiles are shown to the right of each false-color image.

Tunable IR-laser mapping of surface contaminants

19

Figure 10. Color bar for false-color images presented in Figures 9 and 11. Film thickness was calibrated by weight-gain measurements during sample preparation and by comparison with FTIR reflectance data.

visible in the top reflectance image (red and yellow indicating the lowest reflectance, hence the thickest deposit, locations), with a particularly thick vertical band near the center. Very few areas in this image possess high reflectance values (dark blue) characteristic of low contamination. A horizontal line across the center of the image indicates the thickness profile, shown in Figure 9 (top, right) for this sample. Reflectance values have been converted to thickness of the drawing-agent hydrocarbon contaminant using the FTIR data analysis discussed above. The data shown here indicate the thickness averaging about 0.7 µm along the profile line, with heavier deposits up to 2 µm. False color images of the test surfaces contaminated with lower amounts of hydrocarbon (Fig. 9, middle and bottom) show much less spatial variation in the distribution of hydrocarbon residue. Hydrocarbon residues are thinner and appear as predominantly green and light blue in the false-color images while the line profiles show quantitatively the thickness of lubricant in these images. The average thickness values of the three profiles presented here are consistent with the weight change and thickness values determined by FTIR. The potential value of the infrared-laser imaging method for cleanliness verification is clearly demonstrated for these test panels. For these samples distribution of the residual hydrocarbon contaminant is quite variable. In the case of the heaviest contaminated sample, a localized cleaning to effect substantial removal can be profitably applied to the most heavily contaminated areas.

20

D. Ottesen et al.

Figure 11. False-color reflectance images and thickness profiles for three aluminum test panels with a hydrocarbon drawing-agent contaminant (surface polishes are: 600-grit, top-left; 220-grit, middleleft; 80-grit, bottom-left). Corresponding line thickness profiles are shown to the right of each falsecolor image.

We also acquired reflectance ratio images for test surfaces with rougher finishes for average hydrocarbon thicknesses of 0.9 µm, again using transverse illumination. False-color images and corresponding thickness profiles for these two samples are compared to the 0.9-µm thick hydrocarbon residue deposited on the smoothest, 600-grit polished surface in Figure 11. Average thickness values from the three profiles are in reasonable agreement for all three test panels, demonstrating that large changes in surface roughness (0.5, 2.1, and 6.1 µm) do not substantially affect the measured thickness of hydrocarbon residue. We observe a qualitative change in the false-color images in Figure 11. Increasingly rough test surfaces (middle and bottom) exhibit a grainier image qual-

Tunable IR-laser mapping of surface contaminants

21

ity due to the large diversity of surface orientations relative to the infrared laser illumination beam. Distribution of the hydrocarbon residue on the 220-grit surface, however, is much more even (Fig. 11, middle, left) than for the smoothest surface (Fig. 11, top, left). The drawing-agent material shows a strong thickness gradient toward the right-hand side of the image for the roughest, 80-grit, surface (Fig. 11, bottom, left) that is clearly visible despite the grainy image appearance. 4. CONCLUSIONS

The work presented in this report has shown tunable infrared-laser imaging to be an extremely attractive method for on-line detection of hydrocarbon contaminants and determination of their spatial distribution for efficient cleaning operations. Calibrated test panels of hydrocarbon contaminants on metallic substrates were prepared and characterized with FTIR grazing-angle reflectance spectroscopy. Measurements were made over a range of film thicknesses and surface roughness, and the derived instrument sensitivity was quite robust with respect to the degree of surface roughness and the orientation of the reflectance unit to the direction of polishing grooves. Tunable infrared-laser images were acquired at both absorbing and nonabsorbing frequencies for hydrocarbon contaminants on aluminum test panels. The thickness of the contaminant layers calculated from the laser images showed good agreement with the measured film thickness determined by spatially averaged FTIR spectroscopic results. The laser images clearly reveal the heterogeneous distribution of the contaminant species on the component surfaces for a variety of film thicknesses and degree of surface roughness. Primarily, the effects of laser-coherence noise determine the current detection limits of the laser-imaging method. The noise is introduced when an image ratio is formed from images taken at absorbing and non-absorbing wavelengths. For typical hydrocarbon species, the detection limit appears to be on the order of 1020 nm for film thickness. Improvements in the system despeckling and projection optics may substantially decrease this noise level with an attendant increase in sensitivity. The configuration of a future prototype imaging system instrument will be strongly determined by system formats that employ either a pulsed or continuouswave laser, and staring focal-plane array (FPA) cameras or raster-scanned imagers. The design of an imaging system will include a consideration of the ultimate instrument cost. At the present time, it appears that a continuous-wave system with a scanned imager offers the system with the lowest cost. However, the performance of some newly developed inexpensive infrared microbolometer arrays will also be evaluated as a possible component of a low-cost pulsed imager. Future work will enlarge both the laser illumination area and image field of view in order to develop a prototype instrument capable of rapid large-area surveys during cleaning verification.

D. Ottesen et al.

22

Acknowledgments We gratefully acknowledge the financial support for these investigations by the Department of Defense through the Strategic Environmental Research and Development Program. REFERENCES 1. T.A. Hoffard, C.A. Kodres and D.R. Polly, Technical Memorandum, NFESC-TM-2335-SHR (2000).∗ 2. C.A. Kodres, D.R. Polly and T.A. Hoffard, Technical Report, NFESC-TR-2067-ENV (1997).* 3. C.A. Kodres, D.R. Polly and T.A. Hoffard, Metal Finishing 95, 48-53 (1997). 4. R.G. Greenler, J. Chem. Phys. 44, 310-315 (1966). 5. D.L. Allara, in: Characterization of Metal and Polymer Surfaces, L.H. Lee (Ed.), Vol. 2, pp. 193-206, Academic Press, New York (1977). 6. W.G. Golden, in: Fourier Transform Infrared Spectroscopy-Applications to Chemical Systems, J.R. Ferraro and L.J. Basile (Eds.), Vol. 4, pp. 315-344, Academic Press, New York (1985). 7. D.K. Ottesen, J. Electrochem. Soc. 132, 2250-2257 (1985). 8. D.K. Ottesen, L.R. Thorne and R.W. Bradshaw, Sandia Report, SAND86-8789 (1986).* 9. R.W. Bradshaw, D.K. Ottesen, L.R. Thorne, A.L. Newman and L.N. Tallerico, Sandia Report, SAND87-8241 (1987).* 10. P.E. Powers, T.J. Kulp and S.E. Bisson, Optics Letters 23, 159-169 (1998).

∗

NFESC technical reports may be ordered from the web at www.dtic.mil. Reports from Sandia National Laboratories may be ordered by contacting Sandia National Laboratories’ Technical Libraries at (505) 845-8287 or the National Technical Information Service (NTIS) at www.ntis.gov.

Surface Contamination and Cleaning, Vol. 1, pp. 23–41 Ed. K.L. Mittal © VSP 2003

Monitoring cleanliness and defining acceptable cleanliness levels MANTOSH K. CHAWLA∗ Photo Emission Tech., Inc., 3255 Grande Vista Drive, Newbury Park, CA 91320

Abstract—Defining and maintaining a “proper” level of surface cleanliness is, at best, subjective. Often the failure of surface preparation processes is not discovered until problems, such as poor adhesion, occur down stream. Surface cleanliness is critical for good surface finish or success of subsequent operations that depend on surface cleanliness. To assure consistent quality of surface cleanliness, it is important to: understand the types of contaminants that need to be monitored, most common cleanliness monitoring methods and their strengths and limitations, factors to be considered in choosing appropriate cleanliness monitoring method(s), and cost impact of various cleanliness levels. The selection of a cleanliness monitoring method should take into account several factors, such as the type of substrate and the types of contaminants to be monitored, etc. In order to define “Acceptable” level of cleanliness, a total cost approach is needed. Total cost is defined as the cost of cleaning added to the cost of non-conformance related to a particular level of surface cleanliness. An acceptable level of cleanliness is the one that minimizes or optimizes this “total cost”. Keywords: Acceptable cleanliness levels; optimum cleanliness level; total cost of cleaning; cleanliness monitoring methods.

1. INTRODUCTION

Defining and maintaining the surface preparation at “proper” levels is the key to good surface finish. However defining a “proper” level of surface cleanliness is, at best, subjective. For consistent results, it is important to define “how clean is clean”. Often the inadequacy of surface preparation processes is not discovered until problems, such as poor adhesion, occur downstream resulting in nonconformance due to poor surface cleanliness. To assure consistent quality of surface cleanliness, it is important to: understand the types of contaminants to be monitored; most common cleanliness monitoring techniques and their strengths and limitations; factors that affect the choice of cleanliness monitoring technique(s); select an appropriate cleanliness monitoring method; specify a desirable ∗

Phone: (805) 499-7667, Fax: (805) 499-6854, E-mail: [email protected]

24

M.K. Chawla

level of surface cleanliness; and monitor the surface cleanliness to an established level on an on-going basis. The selection of a cleanliness verification technique, as a minimum, should take into account the type of substrate and the types of contaminants to be monitored, desired level of cleanliness, speed of measurement, operator skill level required, and acquisition and operating costs. In addition, it is very important that the cleanliness monitoring technique be quantitative, non-destructive and readily usable. For every level of cleanliness, there is a corresponding level of product performance (i.e. failure / non-conformance rate). Each level of cleanliness has a cost associated with achieving that level, just as there is a cost associated with the failure / non-conformance rate corresponding to each level of cleanliness. These two cost components can be combined to assess “total cost” of cleaning. A minimum “total cost” can only be achieved by balancing the cost of incremental cleaning with the reduced cost of corresponding failure / non-conformance rate. The “optimum” level of cleanliness is the one that minimizes the “total cost”. Since all processes have some variation, there is bound to be some variation in the level of cleanliness achieved. An acceptable variation around the “optimum” level of cleanliness, where the total cost is minimum, would define the “Acceptable cleanliness level”. Some suggested approaches to defining acceptable levels of surface cleanliness are also discussed. 2. TYPES OF CONTAMINATION

A contamination is defined as any undesirable foreign matter that is present on a surface. Contaminations can be classified into three different categories: 1) Particulate, 2) Thin Film (Both Organic and Inorganic), and 3) Microbial or biological contamination. (1) Particulate contamination can be defined as any foreign matter present on the surface as a physical object. Some examples of particulate contaminants are dust, hair, micro-fragments and fibers. (2) Thin film contamination, also called Molecular contamination, is present on the surface in the form of a thin film covering the whole surface or some areas of the surface. This type of contamination can be organic or inorganic. Some examples of thin film contaminants are skin oil, grease, surfactant/chemical residues, oxides and other unwanted films. (3) Microbial contamination can be present on the surface in the form of particles or thin films or a combination of both and refers to generally unwanted living organisms present on the surface. Some examples of microbial contaminants are spore, bacilli and organic cultures. This type of contamination generally occurs from the environment or residues from processes.

Monitoring cleanliness and defining acceptable cleanliness levels

25

3. TYPES OF CLEANLINESS MONITORING METHODS

Cleanliness monitoring methods can also be generally classified into three different categories: 1) Indirect Methods, 2) Direct Methods, and 3) Analytical Methods. All of these methods have certain strengths and limitations, which will be discussed later; hence, it is important to select the method that will be most appropriate for a particular application. Most of these methods are appropriate for thin film or molecular contamination. (1) Indirect methods – Any technique that does not take a measurement on the actual surface or area of interest would be classified as an indirect method. See Table 1 for some of the most common indirect methods along with their features. (2) Direct methods – Any technique that takes a measurement directly from the actual surface or area of interest but does not directly identify the species of contamination present would be classified as a direct method. Some of the most common direct methods along with their features are listed in Table 1. (3) Analytical methods – Any technique that identifies the species of, and measures the amount of contamination would be classified as an analytical technique. Analytical techniques can be direct or indirect; however all of them usually determine the amount of and the species of contamination. Some of the most common analytical methods along with their features are listed in Table 2. 4. MOST COMMON VERIFICATION / MEASUREMENT METHODS

Some of the most common indirect, direct and analytical methods, with a brief discussion of their principles of operation, are presented below. 4.1. Indirect methods 4.1.1. Determination of non-volatile residue (NVR) [1] Also known as gravimetric measurement. This method requires a highly sensitive scale that can weigh parts to an accuracy of plus or minus one milligram, or better. A container is weighed before collecting fluid that flushes the part of interest. After the collected fluid has evaporated, the container is weighed again. The difference in the weight of the container before and after flushing and evaporation is the weight of the contamination removed by flushing. 4.1.2. Ultraviolet (UV) spectroscopy It involves the use of a spectrometer to analyze solvent extract from the parts of interest. Only contaminants that have an absorption wavelength in the UV region can be detected and analyzed. Calibration curves, utilizing samples with known concentration of contamination, can be developed and used to determine actual amount of contamination.

26

M.K. Chawla

Monitoring cleanliness and defining acceptable cleanliness levels 27

28

M.K. Chawla

Monitoring cleanliness and defining acceptable cleanliness levels 29

30

M.K. Chawla

4.1.3. Use of an optical particle counter (OPC) As the name implies, this method is used for detecting particulate contamination. Typically the part or surface of interest is flushed with some fluid. The fluid is then analyzed using a particle counter. OPC gives both the count and size of particles in the suspension measured. 4.2. Direct methods 4.2.1. Magnified visual inspection It is a step above visual inspection with the naked eye. Using some means of magnification, gross contamination that may not be visible to the naked eye can be observed. Due to its nature it is only effective with smaller parts that can be handled by an operator. The method also limits the surface area that can be checked. 4.2.2. Black light Using a black-light, i.e., UV light it is possible to visually detect gross level of contamination. For this technique to work, however, the contaminant of interest must fluoresce under black light. This method is somewhat similar to magnified visual inspection, except that since the contaminants fluoresce, if present, they are easier to see. Typically the level of contamination that can be detected with this method is too high for most precision cleaning applications. Experiments have shown that a skilled operator can, at best, detect 1 mg/cm2 [2]. 4.2.3. Water break test This technique utilizes the difference in surface tension of water and organic contaminants to detect contamination. This test will detect the presence of hydrophobic films on surfaces. When water is applied to the surface to be checked for contamination, water covers the areas of the surface that are clean. The presence of organic contamination on the surface prevents water from forming a film over it. This test can be used for checking small parts as well as large surfaces. It is very cost effective and will enable detection of molecular layers of hydrophobic organic contaminants. The sensitivity of the test may be questionable for rough or porous surfaces. 4.2.4. Contact angle A drop of water resting on a solid surface forms a shape that is influenced by the solid surface tension. The shape is influenced by presence of organic contaminants on the surface. If a tangent is drawn from the droplet to the solid surface, the angle formed is called “Contact Angle”. Contact angle measurements can be used to detect organic films, coatings or contaminants on the surface. “A contaminated metal part would have a high contact angle, such as 90° or more. Some parts, such as plastics, have positive contact angles even when “clean” so the method is not typically used for cleanliness analysis for these materials. While a number is obtained from this test, the test is still non-quantitative in terms of the contaminants on the part [3]”. Because of its simplicity, contact angle measure-

Monitoring cleanliness and defining acceptable cleanliness levels

31

ments have been broadly accepted for material surface analysis related to wetting, adhesion, and absorption. 4.2.5. Optically stimulated electron emission (OSEE) [4] A probe illuminates the surface to be tested with ultraviolet light of a particular wavelength. This illumination stimulates the emission of electrons from the metal surface. The emitted electrons are collected and measured as current by the instrument. Contamination reduces the electron emission and, therefore, the current measured. The equipment may be connected to a computerized scanning system that can scan a flat or cylindrical surface for cleanliness. The results can be presented as a color map or 3-D map. The user can define the level of cleanliness each color represents in the graphic presentation of the results. This feature makes it easy to compare “before” and “after” effect of a cleaning process or side-byside comparison of two pieces cleaned in alternative cleaners. OSEE is simple to operate, fast, and relatively inexpensive. In addition, it is quantitative, nondestructive, and non-contact. This technique detects both organic and inorganic contamination, such as oxides, and can be used on any shape of parts as long as the geometry of the part is presented to the sensor in a consistent manner. This system lends itself to scan small parts or large surface areas very quickly. This test can be used in the production line as well as for on-line real time measurement of surface cleanliness. The surface of interest must emit electrons for the technique to work. Nearly all materials of engineering importance emit electrons when exposed to UV light. 4.2.6. MESERAN surface analyzer – (measurement and evaluation of surfaces by evaporative rate analysis) [5] A measurement begins by depositing onto the test surface a small volume of test solution. A thin- end-window Geiger Müller detector is positioned above the droplet and a metered flow of gaseous nitrogen is passed between the detector and the test surface. To sense the volatile compound, organic compounds are used in which one or more of the carbon atoms are Carbon-14. The β-particles given off by the C-14 molecules at the surface are counted. Specifically measurements are made of how many molecules there are, how many are evaporating away, how fast they are evaporating away and, how many remain retained on the surface. Measuring molecules provides a high degree of sensitivity and the opportunity to analyze surfaces on a molecular scale with observations and results available in only a few minutes. The choice of volatile chemical compounds determines whether they react with the surface material, evaporate, or are retained by the various physical/chemical molecular forces acting at the surface. Chemical compounds can be found which tend to both volatilize (evaporate) and yet tend to be retained by the surface upon which they are placed. The balance of these tendencies determines just how long the volatile compound remains on the surface, or just how much remains. In fact, it is possible to choose a compound that reacts with specific properties of the surface, or a compound where the evaporation and/or retention are affected by certain characteristics of the surface

32

M.K. Chawla

material. By using only a monolayer equivalent of the radiochemical, the observed rate of evaporation becomes a function of the residual concentration of the non-evaporated molecules of radiochemical compound. 4.2.7. Total organic carbon (TOC) analysis [6] This method uses oxygen gas in a combustion chamber at a set temperature to combust carbon-based contaminants into carbon dioxide which is then detected by CO2 coulometer. Coulometer detection uses electricity to electrochemically measure the weight of carbon combusted in the combustion chamber. The method is very sensitive and can detect as little as one microgram of carbon. The TOC method works on a variety of materials and is surface-geometry independent. The method works only on small parts or pieces of larger parts. Due to the high temperature in the combustion chamber (more than 400°C) the method is not suitable to parts sensitive to high temperature. In addition, the TOC method detects only carbon-based contaminants, although this is generally not an issue since the majority of contaminants encountered in a manufacturing environment are carbon based. The TOC method can be used in a laboratory but is adaptable to production environment. It is a technique that works by oxidizing the sample to convert the carbon into carbon dioxide, and detecting and measuring carbon dioxide. The detection of carbon implies that there was some contamination that had carbon as its constituent. The level of TOC detected determines the level of cleanliness of a part. Since a TOC Analyzer detects only carbon, the compound of interest must contain some carbon in a detectable quantity, in order for the analysis to be carried out. 4.3. Analytical methods Any technique that identifies the species of, and measures the amount of contamination would be classified as an analytical technique. Analytical techniques can be direct or indirect; however all of them usually determine the amount of and the species of contamination. All of the analytical techniques involve “Probing the surface, near-surface region, or interior of a material with electrons, ions, or photons produced radiation that has been altered depending on the number, energy, or type of particles emitted. Changes can also occur in the frequency or absorbance of the radiation transmitted through or reflected from the material. Each type of analytical instrument looks at these emissions in a different way to provide information about certain aspects of the sample, such as structure, composition, or chemistry, and electronic or optical properties” [9]. Most of the analytical techniques test the specimen in vacuum, are expensive and require high skill level to operate and interpret the results. Testing takes time and rarely provides real-time information. Because of the cost of analytical testing, it is recommended that its use be limited to applications where identification of the species of contamination is required to enhance or improve the process. Analytical techniques can be divided into two groups; 1) Chemical/elemental surface analysis, and 2) bulk analysis techniques. There are many techniques that

Monitoring cleanliness and defining acceptable cleanliness levels

33

are capable of performing these analyses, some of the most common analytical techniques are summarized below. For a more complete list of most common analytical techniques, visit www.cea.com/table/htm, website of Charles Evans & Associates. For a more comprehensive list of analytical techniques visit the website of ESCA users group in England – www.ukesca.org/tech/list/html. 4.3.1. Chemical/elemental surface analysis techniques 4.3.1.1. Auger electron spectroscopy (AES) / scanning Auger microscopy (SAM) [7–9] They are used to obtain elemental composition information (and some chemical information) from the top two to five atomic layers of a material; identify the composition of very small features and particulates on surfaces; and provide depth composition profiles of thin films, metals, and alloys. Micro-beam AES is also used to study grain boundaries in high temperature alloys, and to examine fracture surfaces to determine composition and extent of damage. The Auger electrons, named after the discoverer of the process, are produced (among other emissions) with discrete energies, which are specific to each element, when the surface is irradiated by a finely focused electron beam. Auger electrons are collected and measured. Auger electrons have discrete kinetic energies that are characteristic of the emitting atoms, making this technique particularly useful for identifying elemental composition. The energy level of Auger electrons is specific to a species of contamination. The escape depth of Auger electrons (1–5 nm) makes this technique very surface sensitive. 4.3.1.2. Electron spectroscopy for chemical analysis (ESCA) [7–9] Also known as X-ray Photoelectron Spectroscopy, or XPS, is a surface analysis technique that provides information on both elemental identity and chemical bonding. This information can be used to identify functional groups and molecular types. This method uses special equipment to bombard the surface of interest with X-rays under vacuum conditions, causing electrons to be ejected from the surface. The actual elemental composition can be quantified by measuring the energy level of ejected electrons, since each element ejects electrons at a unique energy. Its application is limited to mostly research and development, but it can be used to calibrate and evaluate other, less sophisticated measurement methods. 4.3.1.3. Secondary ion mass spectrometry (SIMS – static) [7–9] A surface analysis technique used for identifying molecules on a surface, as well as for depth profiling for tracking very low concentrations of contaminants or ionimplanted species. SIMS technique includes static SIMS (SSIMS), dynamic SIMS, and time-of flight SIMS (TOF SIMS). SSIMS can identify organic and inorganic species. TOF SIMS is an ultra-precise and accurate technique for measuring the mass of molecules in the near-surface layers of material. A pulsed primary ion beam is used to sputter material from the surface of the sample. Secondary ions are collected and focused into a reflection time-of-flight (TOF) mass spectrometer, where they are mass analyzed. Analysis involves measuring the length

34

M.K. Chawla

of time it takes the secondary ions to reach the detector. The lighter the ion, the less time it takes to reach the detector. From the arrival time the masses of the species can be identified. High sensitivity depth profiling is a key feature. 4.3.1.4. Secondary ion mass spectrometry (SIMS – dynamic) [7–9] It uses a much higher intensity bombarding beam than Static SIMS, and is a particularly sensitive (less than part-per-billion level) method for depth profiling of dopants and trace elements in semiconductors. It can also map the X-Y distribution of atomic species with sub-micrometer spatial resolution. An energetic primary ion beam is used to sputter atoms from the sample surface. Secondary ions emitted are mass analyzed. It is inherently a profiling technique. It uses O2 or Cs ions to bombard a surface in high vacuum. High sensitivity depth profiling is a key feature. 4.3.1.5. Variable-angle spectroscopic ellipsometry (VASE) [7, 8] It is a noninvasive technique that offers information about surface composition, layer thickness, and optical properties. Its applications include examining optical surfaces and crystals, and measuring and analyzing band gaps in semiconductors, optical devices, thin films, and carbon coatings on computer hard disks. 4.3.1.6. Energy dispersive X-ray (EDX) and wavelength dispersive X-ray (WDX) analyses [7, 8] They are often combined with a scanning electron microscope or electron microprobe. EDX provides simultaneous multi-element analysis and elemental mapping capabilities for a region up to a few micrometers deep. WDX analyzes trace amounts of one element at a time and is more quantitative than EDX. An example of EDX application is identifying silicon nitride and titanium carbide inclusions in stainless steel. 4.3.2. Bulk analysis techniques The following are several analytical techniques that typically are used for chemical or elemental analysis of bulk materials, but these can also be adapted for the characterization of surfaces and thin films. Many times these techniques are used in industry for characterizing surfaces, sometimes without full knowledge of the strengths and limitations of these techniques. It is hoped that information about how these techniques work, their strengths and limitations would help the reader in determining their usefulness and limitations for their applications. 4.3.2.1. Fourier transform infrared (FTIR) spectroscopy [7, 8] It provides information about the chemical bonding and molecular structure of organics and some inorganic solids, liquids, gases and films. This technique is especially good for identifying unknowns when reference IR spectra are available. When an infrared beam impinges on a surface, the molecular constituents vibrate in the infrared regime. The identities, surrounding environments, and concentrations of these oscillating chemical bonds can be determined. FTIR is a powerful analytical tool for characterizing and identifying organic molecules. The IR spectrum of an organic compound serves as its fingerprint and provides information

Monitoring cleanliness and defining acceptable cleanliness levels

35

about chemical bonding and molecular structure. This information can be used to detect the types of organic materials present on the surface. 4.3.2.2. Raman spectroscopy (RS) [7, 8] It is used to examine the energy levels of molecules that cannot be well characterized via infrared spectroscopy. The two techniques, however, are complimentary. In the RS, a sample is irradiated with a strong monochromatic light source (usually a laser). Most of the radiation will scatter or “reflect off” the sample at the same energy as the incoming laser radiation. However, a small amount will scatter from the sample at a wavelength slightly shifted from the original wavelength. It is possible to study the molecular structure or determine the chemical identity of the sample. It is quite straightforward to identify compounds by spectral library search. Due to extensive library spectral information, the unique spectral fingerprint of every compound, and the ease with which such analyses can be performed, the RS is a very useful technique for various applications. An important application of the RS is the rapid, nondestructive characterization of diamond, diamond-like, and amorphous-carbon films. 4.3.2.3. Scanning electron microscopy (SEM) / energy dispersive X-ray analysis (EDX) [7, 8] The SEM produces detailed photographs that provide important information about the surface structure and morphology of almost any kind of sample. Image analysis is often the first and most important step in problem solving and failure analysis. With SEM, a focused beam of high-energy electrons is scanned over the surface of a material, causing a variety of signals, secondary electrons, X-rays, photons, etc. – each of which may be used to characterize the material with respect to specific properties. The signals are used to modulate the brightness on a CRT display, thereby providing a high-resolution map of the selected material property. It is a surface imaging technique, but with Energy Dispersive X-ray (EDX) it can identify elements in the near-surface region. This technique is most useful for imaging particles. 4.3.2.4. X-ray fluorescence (XRF) [7, 8] Incident X-rays are used to excite surface atoms. The atoms relax through the emission of an X-ray with energy characteristic of the parent atoms and the intensity proportional to the amount of the element present. It is a bulk or “total materials” characterization technique for rapid, simultaneous, and nondestructive analysis of elements having an atomic number higher than that of boron. Traditional bulk analysis applications include identifying metals and alloys, detecting trace elements in liquids, and identifying residues and deposits. 4.3.2.5. Total-reflection X-ray fluorescence (TXRF) [7, 8] It is a special XRF technique that provides extremely sensitive measures of the elements present in a material’s outer surface. Applications include searching for metal contamination in thin films on silicon wafers and detecting picogram-levels of arsenic, lead, mercury and cadmium on hazardous, chemical fume hoods.

36

M.K. Chawla

5. CONSIDERATIONS FOR SELECTING A CLEANLINESS MONITORING METHOD [10]

There are several factors that should be considered in selecting a method for monitoring surface cleanliness. The factors discussed here are the ones that are most important but by no means represent a complete list of factors that should be considered. There may be other factors that are pertinent to a particular application that should be considered. (1) Type of contaminant – One of the first factors that should be considered in selecting a cleanliness monitoring method is the type of contaminant that need to be monitored. Is the contaminant particulate or thin film type? If thin film contamination, is it organic or inorganic or both? Does the technique under consideration monitor the type of contaminants that need to be monitored? (2) Types of substrates – What type of substrate is going to be monitored? Are the techniques under consideration capable of monitoring this type of substrates? Are the techniques likely to damage the substrate to be monitored? (3) Level of cleanliness to be monitored – It is important that the level of contamination that is expected or tolerable can be monitored by the technique under consideration. It is recommended that samples with different levels of contamination be monitored with the technique(s) under consideration. In evaluating the technique for suitability, prepared samples should have levels of contamination spanning a range from 0% (i.e. clean surfaces) to maybe 200% of the expected level of contamination on the surface. The technique(s) should not have any problem in distinguishing between different levels of contamination. (4) Features of monitoring method – It is important to consider various features of the method under consideration. For example, is the technique non-contact and/or non-destructive? Does the technique require deposit of some medium on the surface? For example, the contact angle measurement requires that a droplet of water be placed on the surface of interest. How large an area can the technique measure? Is it sensitive to surface roughness? Can the technique check parts of different geometries? Can the technique be used on-line? Is the technique suitable for the environment it is going to be used in? Does the technique cause any permanent changes to the surface? All of these questions should be considered to determine the most appropriate monitoring technique for a particular application. (5) Measurement speed – Is the measurement speed critical for the application under consideration? If so, how fast can the technique make a measurement? Is the speed sufficient to keep up with the production flow? (6) Acquisition and operating cost – How does the acquisition cost compare among the techniques that meet other requirements for the application? Are there any expendable items that would have to be purchased for continued use of the equipment? How much does that add to the operating cost? What are the maintenance and calibration requirements and how much these require-

Monitoring cleanliness and defining acceptable cleanliness levels

37

ments will add to the operating cost? All these questions should be answered to truly compare the total cost of any cleanliness monitoring system. (7) Skill level required – The operator skill level can be a key factor in the use of some techniques, particularly the analytical techniques. Some techniques may involve interpretation of the data to determine the quality of surface cleanliness. These factors should also be considered in the selection of a cleanliness measuring technique. A high operator skill level will result in higher operating cost. In the event of personnel turnover, higher training costs may also be incurred. 6. COST OF CLEANLING [10]

For every level of cleanliness, there is a cost to achieve that level of cleanliness. There is corresponding level of failure/non-conformance for each cleanliness level, hence cost of failures/non-conformance. “Total Cost” of achieving a certain level of cleanliness is the sum of these two costs. As the achieved level of surface cleanliness increases, the cost of cleaning also increases. Eventually the incremental cost of cleaning rises exponentially. Hence the cost of surface cleaning is directly proportional to the surface cleanliness level. Intuitively, we know that the higher the cleanliness level the lower the failure/non-conformance rate, hence cost, due to surface cleanliness. The incremental drop in costs due to lower failure/non-conformance also exhibits exponential relationship. Hence the cost of failures/non-conformance is inversely proportional to the surface cleanliness level. If both of these costs were plotted on a graph, the typical result would be like the one shown in Figure 1.

Figure 1. Total cost vs. cleanliness level.

38

M.K. Chawla

An optimum level of cleanliness is the one that minimizes the total cost. Eventually one can arrive at a cleanliness level where the savings in the failure/nonconformance costs will not be offset by incremental cost of achieving cleanliness beyond the optimum level. A small range around the optimum level of cleanliness can be established as the “Acceptable Level” of cleanliness. 7. DEFINING ACCEPTABLE (“OPTIMUM”) LEVEL OF CLEANLINESS

It is expected that the non-conformance levels will increase as the level of cleanliness decreases or vice versa. It is important to understand the relationship between the level of cleanliness and non-conformance rate in order to establish the “acceptable level of cleanliness”. For example, if the failure/non-conformance rate is too high due to the surface cleanliness level, then the surface cleanliness level will have to be improved in order to reduce the failure rate. On the other hand, no failures or a very low failure rate due to the surface cleanliness level implies that the surface may be “over-cleaned.” It may be desirable to optimize the cleaning process by comparing the cost of failures/non-conformance with the cost of cleaning the surface. Generally, increasing the level of surface cleanliness will result in increased cleaning cost. An increased level of cleanliness should lower the rate of non-conformance, which, in turn, reduces the non-conformance cost. As long as the reduction in nonconformance cost more than offsets the increased cost of cleaning, it would be cost effective to increase the achieved level of surface cleanliness. When the decrease in non-conformance cost fails to offset the increase in the cleaning cost, then an optimum or “acceptable” level of cleanliness has been achieved. To establish the optimum level of surface cleanliness, two approaches are outlined here. One approach utilizes the success of the subsequent operation that depends on surface cleanliness level. The other approach is to start monitoring the cleanliness levels achieved and corresponding level of failure/non-conformance rate. Once an acceptable level of cleanliness is established using one of the two approaches, cleaning process can be monitored in production to assure ongoing product quality. 7.1. Controlled experiment This approach requires that the measure of success be defined for the subsequent operation that depends on surface cleanliness. For example, if the parts are to be bonded, then the adhesion strength of the bond will be the measure of success. If the parts are to be coated after cleaning, then the adhesion strength of the coating should be correlated to surface cleanliness. The acceptable level of surface cleanliness is the one that results in the desired level of bond/adhesion strength. One simple approach is to start monitoring and recording the cleanliness level of each part. A statistically significant sample must be monitored to assure valid

Monitoring cleanliness and defining acceptable cleanliness levels

39

Figure 2. Peel strength vs. surface cleanliness.

results. These parts then must be followed through the process to measure the level of success for each part at the subsequent operation. The level of cleanliness that results in the desired minimum level of success is the minimum level of cleanliness that must be achieved in production. This approach has its limitations. For example, the results depend on what level of cleanliness is being achieved in production. If the surface is “too clean” there may not be enough variation in the cleanliness level to identify the point where minimum success is achieved. On the other hand, if the surface is not clean enough the desirable success may not occur. A more proactive approach is to prepare parts with different levels of surface cleanliness, measure and record the cleanliness level and follow through with the subsequent operation to correlate the success level with cleanliness level. It is recommended that the range of cleanliness should be as wide as possible to help identify the minimum level of cleanliness. Once again it is important that a statistically significant sample be used. It is also recommended that, if possible, several cleanliness measurements should be taken from each part and the mean cleanliness level be correlated to the mean success level. Figure 2 [10] graphically depicts the typical result of correlating the success level to surface cleanliness level. A minimum level of cleanliness is the one that corresponds to the target minimum level of success. 7.2. “Benchmark” testing Once a cleanliness monitoring method has been selected, it can be used to establish the cleanliness level achieved by current cleaning process (“Benchmark”). The production can then be monitored to assure that benchmark cleanliness level is being achieved. In addition, the product can be followed through the manufacturing process to assure that no problems occur downstream as a result of inadequate surface cleanliness. The level of non-conformance related to the level of

40

M.K. Chawla

cleanliness achieved should be monitored. The cost associated with a given level of cleanliness and the cost of non-conformance associated with that given level of surface cleanliness should be combined to determine the “total cost”. Changes should be made to the cleaning process to minimize the “total cost”, i.e. optimize the “total cost”. The level of cleanliness associated with the “optimum total cost” should be considered the optimum or “acceptable” level of cleanliness. 8. ON-GOING, IN-PROCESS SURFACE CLEANLINESS MONITORING

Surface cleanliness monitoring system must be used to monitor the process and assure that the desired cleanliness level is being achieved on an on-going basis. Surface cleanliness monitoring system can be very useful in assessing how the surface cleanliness level is affected by making changes to the cleaning process or for evaluating alternative cleaning processes for their ability to achieve the desired cleanliness level. The required level of cleaning agent concentration in the cleaning solution can also be objectively determined and maintained by using a surface cleanliness monitoring system. Measuring the effect of varying the concentration level of the cleaning agent on surface cleanliness can help determine the “optimum” concentration level. In most industries, the chemical or cleaning agent replenishment schedule is usually time-dependent. The success of this approach relies on the level of contamination on each part and the number of parts processed in a given time interval being relatively constant. In real life, the amount of contamination can vary considerably from part to part. In addition, the number of parts cleaned during a given time frame can also vary considerably. A time-dependent replenishment schedule is not the ideal way of assuring product quality. On-going, in-process monitoring of surface cleanliness helps in replenishment of chemicals or cleaning agents only when needed, and not based on a pre-determined, somewhat arbitrary schedule. 9. SUMMARY

In order to define an acceptable level of cleanliness, it is important to minimize the total cost of cleaning. The total cost of cleaning is the sum of the cost of achieving a certain level of surface cleanliness and the cost of failure/nonconformance associated with that level of surface cleanliness. Selecting a method for monitoring cleanliness is the first step in establishing an acceptable level of cleanliness or defining “how clean is clean”. Several factors need to be considered in selecting an appropriate surface cleanliness method, which include, but are not limited to, type of contaminant to be detected, level of cleanliness to be monitored, acquisition and operating cost of the monitoring method, and the skill level

Monitoring cleanliness and defining acceptable cleanliness levels

41

required to operate the system. Surface cleanliness monitoring method may be direct, indirect or analytical. A monitoring method can be used to optimize the cleaning process by varying different parameters of the cleaning process while monitoring surface cleanliness to see how it is affected by the change. It can also help in ongoing monitoring of the cleaning process to assure that the desirable level of cleanliness is being achieved. REFERENCES 1. B. Kanegsberg and M. Chawla, “Non Volatile Residue”, A2C2 Magazine, 5, No. 3, 41 (2002) and 5, No. 4, 45 (2002). 2. R.L. Gause, “A Noncontacting Scanning Non Contact Photoelectron Emission Technique for Bonding Surface Cleanliness Inspection”, Marshall Space Flight Center, Huntsville, AL, presented at Fifth Annual NASA Workshop, Cocoa Beach, Florida (1987). 3. B. Kanegsberg and M. Chawla, “Contact Angle”, A2C2 Magazine, 4, No. 8, 41 (2001). 4. Surface Quality Monitors Brochure, Photo Emission Tech., Inc. 5. B. Kanegsberg and M. Chawla, “MESERAN”, A2C2 Magazine, 4, No. 9, 49 (2001). 6. B. Kanegsberg and M. Chawla, “Total Organic Carbon”, A2C2 Magazine, 4, No. 10, 37 (2001). 7. Charles Evans Associates Website – www.cea.com 8. R.D. Cormia, “Problem-Solving SURFACE ANALYSIS Techniques”, Surface Sciences Laboratories, Mountain View, CA: Advanced Materials & Processes, 16-23 (Dec. 1992). 9. Measurement and Characterization Website – www.nrel.gov/measurements/surface/html 10. M. Chawla, “How Clean is Clean? Measuring Surface Cleanliness and Defining Acceptable Level of Cleanliness”, in Handbook for Critical Cleaning, B. Kanegsberg and E. Kanegsberg (Eds.), pp. 415-430, CRC Press, Boca Raton, FL (2001).

This page intentionally left blank

Surface Contamination and Cleaning, Vol. 1, pp. 43–48 Ed. K.L. Mittal © VSP 2003

Tracking surface ionic contamination by ion chromatography BEVERLY NEWTON∗ Dionex Corporation, 500 Mercury Drive, Sunnyvale, CA 95032, USA

Abstract—Surface ionic contamination can cause device failures. In order to find the source of the contamination many questions must be answered first. Are the failures due to incoming materials that are not clean? Has there been a change in the process that is introducing contamination? What is the exact nature of the contaminant, ionic, particulate, metallic, etc? Is there a training issue that needs to be addressed. Can the failure be tested for or is it a long term reliability problem? These are just a few of the questions that must be answered as part of the troubleshooting process. This paper addresses how ion chromatography can be used to troubleshoot a manufacturing or cleaning process and to assure the quality and reliability of electronic devices. Topics covered include: 1. What is ion chromatography. 2. How does it differ from other cleanliness testing methods. 3. How can ion chromatography be used to troubleshoot a cleaning process. 4. Real life examples showing how the use of ion chromatography has improved cleaning processes. Keywords: Ionic contamination; ion chromatography; electronic devices.

1. INTRODUCTION

As electronic devices and assemblies become smaller and more complex, the requirements for improved quality control of product cleanliness have begun to escalate. Surface contamination from ions such as chloride, bromide, sodium, and organic acids has been shown to cause failures in electronic devices [1]. Ionic residues can cause corrosion, metal migration and electrical leakage. The failures cased by these residues may be hard or soft failures and may occur several months after the product has been manufactured and shipped to customers. Upon re-testing the returned product, the failures can be intermittent or “no trouble found” making troubleshooting the device for a root cause of the failure difficult. These residues may be on the exposed surface of an electronic device, they may be encapsulated in flux or resin deposits, they may be trapped under surface mounted devices or they may be encapsulated in polymer finishes (Figure 1). ∗

Phone: (408)4814272, Fax: (408)7372470, E-mail: [email protected]

44

B. Newton

Figure 1. Chromatogram of a board extract using IPA/water mixture.

Figure 2. Analysis of a cassette used to transport disk drive components during manufacture.

The manufacture of electronic devices typically involves a series of chemical and mechanical operations such as plating, masking, soldering, rinsing, etching, cleaning, etc. Each of these operations along with the environment in which they occur leaves some effect on the device or assembly. The processes and manufacturing environment leave chemical “fingerprints” on the device that are unique to the manufacturing process. In the same way that a forensic scientist would use fingerprints to trace a criminal, analytical techniques can be used to troubleshoot a manufacturing process or field failure to understand and correct the root cause.

Tracking surface ionic contamination by ion chromatography

45

Ionic contamination can also be found on materials that come in contact with electrical devices during manufacture, e.g. gloves, cassettes, etc. (Figure 2). These manufacturing consumables can transfer contamination to the manufactured products and need to be examined for contamination in the same way that the final product is evaluated. 2. TYPES OF IONIC CONTAMINATION

Potentially corrosive ions found on printed circuit boards and electronic devices include: ● Bromide - commonly found in solder masks ● Sulfate - comes from a variety of materials such as oils and release agents ● Chloride - commonly found in fluxes ● Organic acids such as adipic or succinic acid - found in fluxes Typically, the higher the concentration of corrosive ions on a particular assembly, the higher the risk of electrochemical failure. 3. TEST METHODS FOR IONIC CONTAMINATION

In the past, electronic component manufacturers, board manufacturers and electronic assemblers have relied on resistivity of solvent extract (ROSE) type test methods to assure ionic cleanliness. Several studies reported by Contamination Studies Laboratory (CSL, Kokomo, IN) have shown that the ROSE method is inadequate for true quantification of ionic contamination. Recently, a modified ROSE method has been proposed as an IPC (Association Connecting Electronics Industries) Standard Method IPC-TM-650 2.3.25.1. Although this new technique is an improvement for reporting overall ionic contamination, it too provides insufficient information to troubleshoot the root cause of electronic failures caused by ionic contamination. The technique of ion chromatography is uniquely qualified for troubleshooting the root cause of failures due to ionic contamination on electronic devices and printed circuit boards. Ion chromatography can provide information on the chemical nature of the residue causing the failure. The output of the ion chromatograph is called a “chromatogram” and gives the identity and quantity of each ion found in a sample of a rinse extract of the device of interest. Ion chromatography is a form of liquid chromatography. The technique is based on the use of specialized column packings for analytical separation of ions found in a chemical mixture. The main advantages of ion chromatography for residue analysis are: ● Multi-component ion analysis ● Most sensitive detection technique available for many ionic compounds ● Method versatility

46

B. Newton

Figure 3. Ion chromatography system configuration.

Ion chromatography is an analytical technique used to separate, identify and quantify ions in a sample matrix such as a water extract of a printed circuit board. The simplest ion chromatography system is composed of a sampling device, a pump, an analytical column, a suppressor and a detector (Figure 3). The analysis begins with a sample, typically a water extract containing ions of interest such as chloride, sulfate, or nitrate. A portion of the sample is injected into the ion chromatography system and combined with an eluent stream composed of sodium hydroxide or bicarbonate solution. The eluent stream carries the sample through the ion chromatography system to the analytical column. The analytical column separates the ions of interest in the sample into narrow bands within the stream of the eluent. Thus, by the time the sample leaves the analytical column, all of the chloride ions are grouped together, then all of the nitrate ions and then all of the sulfate ions. The eluent then sweeps these groups of ions into the suppressor device. This device electrolytically transforms the eluent into pure water leaving just the ions of interest in pure water to be swept along to the conductivity detector. The detector detects the ions based on their conductivity relative to the water eluent. At this point all interfering ions have been removed and the detector’s sensitivity has been maximized allowing for detection of very low (part per billion) levels of ions [2]. This is a very simplified explanation of ion chromatography but it is important to note that more complex samples and analytes can also be analyzed using this technique (for instance, cations such as sodium and magnesium, transition metals such as iron and copper and even certain biological analytes such as amines and nucleic acids).

Tracking surface ionic contamination by ion chromatography

47

4. TRACKING SURFACE IONIC CONTAMINATION IN MANUFACTURING AND ASSEMBLY OPERATIONS

There has been a growing interest in the analysis of ionic contamination on electronic components. Absolute contamination level requirements and guidelines have not been determined; however, Contamination Studies Laboratory (CSL, Kokomo, IN) recommends maximum levels of chloride ion in the range 1.0 µg/sq. in for assembled boards with sensitive components such as microBGAs. The level recommended for bare boards is less than 2.0 µg/sq. in [3]. Ion chromatography provides the unique capability of identifying the individual ions for a given contamination issue. Since the source for chloride contamination can be much different than the source for organic acid contamination it is important to know which ions the manufacturer is dealing with in order to understand and correct the root cause of the problem. This is not possible with resistivity of solvent extract (ROSE) measurements. The capability to identify and quantify individual ions makes ion chromatography a valuable troubleshooting tool for process contamination issues and process monitoring programs. In addition to being the most economical analytical technique for monitoring multiple ions, ion chromatography also provides the ability to distinguish between noncorrosive and corrosive ions, something that ROSE testing is unable to do. A number of studies have been published to show the use of ion chromatography to troubleshoot reliability issues with electronic products. One of the best sources of case study information can be found on the web site for Contamination Studies Laboratory (CSL) at www.residues.com. CSL regularly publishes case studies showing the hazards of ionic contamination to electronic device reliability on their web site and in each issue of Circuits Assembly magazine. A good explanation of how ion chromatography has been used to identify sources of CAF (conductive anodic filament) failures can be found in a study completed by Ready et al. [4]. Several studies [5-7] have been completed on the analysis of ionic contamination on failed disk drive components. As mentioned earlier, manufacturing materials and packaging can be an important source of ionic contamination. Two recent studies by Lin and Graves [8] and Bahten and McMullen [9] provide information on the use of ion chromatography for the analysis of ionic contamination on materials such as pink poly film (a common packaging material) and cleaning brushes. 5. STANDARD TEST METHODS FOR TRACKING IONIC CONTAMINATION

IPC (Association Connecting Electronics Industries) has standard test methods documented for the ROSE, Modified ROSE and Ion Chromatography analysis techniques. These are:

48

B. Newton

IPC-TM-650, TM 2.3.25 Detection and Measurement of Ionizable Surface Contaminants by Resistivity of Solvent Extract (ROSE). ● IPC-TM-650, TM 2.3.25.1 Ionic Cleanliness Testing of Bare PWBs (modified ROSE Test Method). ● IPC-TM-650, TM 2.3.28 Ionic Analysis of Circuit Boards, Ion Chromatography Method. IDEMA (International Disk Drive Equipment and Materials Association) has developed the following standard test method for ionic cleanliness testing. ● M13-99, Measurement of Extractable/Leachable Anion Contamination on Drive Components by Ion Chromatography. ●

6. CONCLUSION

The ion chromatography, ROSE and modified ROSE test methods have been developed to allow electronics manufacturers to identify and control ionic contamination before it evolves into a failed component or board. Tracking ionic contamination requires systematic troubleshooting and improved cleanliness of the product as it is manufactured. This means cleaner raw materials and processes which are controlled by systematic analysis using standard methods such as those documented by IPC, IDEMA, IEST (Institute of Environmental Sciences and Technology), and SEMI (Semiconductor Equipment and Materials International). REFERENCES 1. D. Yang, C. Lee, Y. Yang, E. Kaiser, S. Heberling and B. Newton, Precision Cleaning, 17-23 (May 1998). 2. B. Newton, Precision Cleaning, 38-39 (March 2000). 3. D. Pauls and T. Munson, Circuits Assembly, 110-112 (September 1998). 4. W.J. Ready, B.A. Smith, L.J. Turbini and S.R. Stock, Mater. Res. Soc. Symp. Proc. 515, 45-54 (1998). 5. A. Toxen, A2C2, 13-16 (September 1998). 6. P. Mee, M. Smallen and D. Vickers, IDEMA Insight, 1 (March/April 1997). 7. J. Thompson, T. Prommanuwat, A. Siriraks and S. Heberling, IDEMA Insight, 24-29 (May/June 1999). 8. S. Lin and S. Graves, Micro, 95-106 (October 1998). 9. K. Bahten and D. McMullen, Proc. Semiconductor Pure Water and Chemicals Conference, 355364 (March 1999).

Surface Contamination and Cleaning, Vol. 1, pp. 49–73 Ed. K.L. Mittal © VSP 2003

A new method using MESERAN technique for measuring surface contamination after solvent extraction MARK G. BENKOVICH∗,1 and JOHN L. ANDERSON2 1

Honeywell Federal Manufacturing & Technologies,† PO Box 419159, D/833 MS-2C43, Kansas City, MO 64141-6159 2 ERA Systems, Inc., The MESERAN Company, PO Box 3609, Chattanooga, TN 37404-0609

Abstract—The precision analytical technique known as MESERAN Analysis permits, in 2 minutes, quantitative measurement of the level of pre-existing nonvolatile organic residue (NVOR) on a substrate from <1 ng/cm2 to >100 µg/cm2. MESERAN Analysis is also applicable for determining NVOR deposited from solvents and solvent extracts. The MESERAN method is able to quantify organic contamination levels down to and below 1 ng by depositing as little as 10 µL of solvent containing a known amount of contamination on a clean substrate, allowing it to evaporate, and measuring the resultant residue. The method is described in detail. In addition, NVOR measurements determined from MESERAN data are presented for a specific project conducted at Honeywell Federal Manufacturing & Technologies (FM&T), Kansas City Plant (KCP). Keywords: MESERAN; surface contamination; solvent extraction; non-volatile organic residue.

DEFINITIONS

In this paper a number of abbreviations, special terms, and trademarks are employed: (1) µCi means microCurie, a unit of radiation which corresponds to 3.7 E 4 (37,000) disintegrations per second. (2) Carbon-14 (C-14) refers to the radioactive isotope of the element Carbon, an isotope which emits only soft or low energy beta particles; most C-14 beta particles are stopped by a sheet of paper. (3) USNRC EXEMPT means the very low level of Carbon-14 that is not regulated by the U S Nuclear Regulatory Commission. No license is required for ∗ To whom all correspondence should be addressed. Phone: (816) 997-3529, Fax: (816) 997-2049, E-mail: [email protected] † Operated for the United States Department of Energy under prime contract DE-AC04-01AL66850. ãCopyright Honeywell LLC, 2002.

50

M.G. Benkovich and J.L. Anderson

possession or use. Only USNRC licensed companies are permitted to distribute EXEMPT quantities not to exceed ten 100 µCi of C-14 (or combinations of smaller quantities that added up to 100 µCi of C-14) at one time. Multiple quantities may be stored by the user. Shipments may be made to anyone in the US. Foreign shipments under IATA regulations must conform to the regulations of the country of final destination. (4) ng means nanogram (1 E - 9 grams or 0.000000001 grams); µg means microgram (1 E - 6 grams or 0.000001 grams), mg means milligram (1 E - 3 grams or 0.001 grams). (5) µL means microliter (1 E - 6 liter or 0.000001 liter). (6) GM detector refers to a thin end-window Geiger Müller detector tube which detects the C-14 beta emissions which penetrate through the 1.4–2.0 mg/cm2 mica window. (7) One nanomole (nmole) is 1 E - 9 moles which equals approximately. 6 E 14 molecules (from Avogadro’s ~ 6 E 23 molecules per gram mole). (8) 1 square centimeter (sq cm) with a roughness factor of 3 is equal to 3 E 16 square Angstroms. (9) Monolayer refers to the number of molecules of a material which covers 1 sq cm in a conventional non-close-packed configuration. For example, on a smooth, flat surface with a roughness factor of 3, each molecule of ntridecane occupies about 50 sq. Angstroms – which equates to ~ 6 E 14 molecules per sq. cm – i.e. one nanomole. (10) NVR means non-volatile residue; NVOR means non-volatile organic residue. (11) MESERAN is an acronym for Measurement and Evaluation of Surfaces by Evaporative Rate ANalysis. (12) MESERAN, MicroSolventEvaporator (MSE), MicroOrganicResidue, and MOR are trademarks licensed to ERA Systems, Inc. (13) Ln or ln is the natural logarithm. (14) 1 mg/ft2 is equivalent to 1.0764 µg/cm2 or 1 µg/cm2 is equivalent to 0.929 mg/ft2. 1. INTRODUCTION

The principle of the MESERAN technique was discovered by one of us (JLA) in 1960. This analytical technique is used in a number of industrial and governmental facilities (within the United States and abroad) for research and development purposes as well as for quality and production control. The characterization of the surface being analyzed is carried out by depositing a chemical detector onto the test surface and observing the rate at which the chemical detector disappears from the surface. The MESERAN technique is routinely used for quantifying organic

A new method using MESERAN technique

51

contamination on surfaces and the crosslink density (or degree of cure) in polymers. In addition, the MESERAN technique can be used for quantifying chemically active sites on surfaces [1-3]. Honeywell FM&T, KCP (henceforth KCP) has been using MESERAN Analyzers for approximately 30 years to detect and quantify organic contamination on parts and evaluate various cleaning processes for removing organic contamination. KCP has used MESERAN Analyzers extensively to evaluate the ability of alternate solvents and processes for removing specific organic contaminants to eliminate the use of chlorinated and fluorinated solvents [4-10]. In recent years, KCP has been working on several projects with The MESERAN Company to improve data analysis and develop new methods for using the MESERAN technology [11-14]. 2. PRINCIPLE OF MESERAN TECHNIQUE [15]

The standard microcomputer-based MESERAN technology involves deposition, using a “clean” precision microsyringe, onto a flat or concave surface of 18 µL of a test solution consisting of a low boiling solvent or solvent combination (for these evaluations – cyclopentane) and a high-boiling-but-volatile Carbon-14 labeled compound (in a ratio of approximately 60,000:1). Figure 1 shows the application of test solution. For example, the amount of tridecane-C14 radiochemical per single test (< 0.06 µCi) corresponds to approximately 6 E 14 (6 x 1014) molecules which equates to one nanomole, the equivalent of approximately one molecular layer over one square centimeter. Metered air or nitrogen gas is permitted to flow across the surface and between the surface and a Geiger Müller detector positioned directly above the surface. The evaporation of the low boiling solvent and then the radiochemical is observed as a function of time by recording the detected emissions per second arising from the radiochemical molecules remaining on, or retained by, the surface – the vapor-phase, already-evaporated molecules having been swept out from under the detector by the metered gas (see Figure 2).

Figure 1. Application of test solution.

Figure 2. Measurement of emissions.

52

M.G. Benkovich and J.L. Anderson

Each test takes less than 3 minutes and the amount of radiochemical employed is EXEMPT from U S Nuclear Regulatory Commission and/or ‘Agreement State’ licensing regulations due to the very low level of C-14 involved. For the measurement of microorganic residues, the MESERAN method may be used: (1) Directly on a flat or concave surface and any microorganic residue thereon which is chemically compatible with the particular radiochemical employed, or (2) Indirectly using an extracting solvent followed by depositing and evaporating an aliquot amount onto a “clean” reference surface. Subsequent deposition and evaporation of the radiochemical solution permits measurement of the amount of deposited residue by comparing the results with previously obtained standards similarly deposited from volumetric dilutions. For non-polar and/or hydrocarbon type residues, tridecane-C14 in cyclopentane (designated BK) is employed. For more polar residues, tetrabromoethane-C14 in cyclopentane (designated AK) is used. In order to provide a high number of detected emissions for the minimal amount of radiochemical deposited, the tridecane-C14 has a specific activity of approximately 57 µCi/µmole (one carbon atom of tridecane is essentially pure C14 isotope) while the tetrabromoethane has both carbon atoms labeled (approximately 114 µCi/µmole). Approximately 200 ng of radiochemical are deposited in each test with similar levels of radioactivity. The MESERAN method assumes that the particular radiochemical employed is chemically compatible with the residue, that the test solution droplet covers all of the residue, and that the test solution solvent substantially dissolves the residue within the time period of the solvent evaporation. Attention to the avoidance of inadvertent contamination and the maintenance of reasonably constant temperature and pressure are required for optimal reproducibility from test to test. 2.1. Mechanism of the MESERAN technique for quantifying organic residues [1-3, 11-14] When a homogeneous chemical is permitted to evaporate, the classical mechanism of the process (normally measured by monitoring the already evaporated portion) follows first order kinetics, i.e., the plot of log concentration vs. time is a straight line. This mechanism applies to pure materials as well as to solutions of chemicals in which the components are chemically compatible and in which the second component is non-volatile under the conditions of the process. In the presence of the second component, the rate of evaporation is slowed. In the MESERAN technology, however, the amount of radiochemical retained by the surface as a function of time is measured by counting the emissions arising from the radiochemical molecules remaining on the surface. In this discussion, the temperature and pressure are assumed constant and the concentration of already evaporated molecules in the adjacent gaseous phase approaches zero due to the flowing air or nitrogen referred to above. The molecular weight of each evaporating molecule and the intermolecular forces among the near-neighbor molecules

A new method using MESERAN technique

53

are thus the primary factors in determining the tendency of each molecule to remain in solution or conversely to escape from the liquid portion of the air/liquid (or semisolid) interface. In the MESERAN technology, which employs only a monolayer equivalent of the radiochemical, the observed rate of evaporation is thus a function of the residual concentration of the non-evaporated molecules of the Carbon-14 radiochemical. Figure 3 illustrates the typical evaporation of the radiochemical solution from a clean surface. The A-B line represents the evaporation of the low boiling solvent (e.g., cyclopentane). The rationale for the initial increase in counts/second is that the C-14 soft beta emissions are partially absorbed by the solvent molecules. B represents the point at which substantially all of the low boiling solvent has evaporated and the maximal amount of residual radiation reaches the GM detector. The B-C line represents the evaporation of the radiochemical from the surface under the conditions of the test. C represents a level where the GM detector can no longer adequately differentiate the residual radiation from background. A solution of the high-boiling-but-volatile tridecane-C14 in higher boiling hydrocarbons (i.e., contamination) follows a similar but slower path than does the evaporation of the labeled tridecane itself since the non-volatile “residue” molecules occupy increasing portions of the liquid (or semi-solid) interface. The rate at which the solvent evaporates is slowed somewhat and the rate at which the radiochemical evaporates is slowed considerably with the observed rate of evaporation being a function of the amount of residue on the surface. The observed rate of evaporation of the radiochemical (the slope expressed as a positive integer) thus is an inverse measure of the amount of non-evaporating residue. The lower the slope, the more the residue and vice versa. Figure 4 illustrates typical evaporations of the radiochemical solution with increasing amounts of residue. ABC is repeated from Figure 3 and illustrates a typical evaporation of the radiochemical solution with no interactions from residue (i.e., a clean substrate). A*B*C* illus-

Figure 3. Typical evaporation of radiochemical solution from a clean surface.

Figure 4. Typical evaporations of radiochemical solution with increasing amounts of residue.

54

M.G. Benkovich and J.L. Anderson

trates a typical evaporation of the radiochemical solution with some contamination present. A**B**C** illustrates a typical evaporation of the radiochemical solution with a larger amount of contamination present. 2.2. Methods of analyzing MESERAN data [1-14] There are two general methods for analyzing the MESERAN data: (1) Total Counts (total area under each curve based on counts minus background) and (2) Slope of the evaporation of radiochemical (the post-peak portion of the curve). Based upon raw data minus background, Figure 5 illustrates three typical experimentally derived curves of natural logarithm (Ln or ln) counts per second minus background vs. time in seconds. Figure 5 is similar to Figure 4 except raw data from actual tests are shown. In Figure 5, the upper curve represents a high level of organic residue, the middle curve represents a medium level of organic residue, and the lower curve represents a low level of organic residue. The scatter, particularly at the lower values, is due to the inherent randomness of radiation (the Poisson distribution in which the square root of each count total is the best estimate of one standard deviation).

Figure 5. Plot of raw data showing low, medium, and high levels of organic contamination.

A new method using MESERAN technique

55

In an effort to reduce the effect of the randomness of radiation, the data (ln (counts – background)) are “smoothed” from six seconds through 115 seconds (for 120 second length tests) and replotted. The smoothing is accomplished by summing the ln (counts – background) at the sixth second with the previous five seconds and the subsequent five seconds and dividing this number by 10. A divisor of 10 is used instead of 11 because it is statistically sound to take the number of items being smoothed and subtract one from it because a degree of freedom is lost. This process is carried out through the 115th second and the subsequent data are replotted as smoothed (ln (counts – background)) versus time. Figure 6 represents the same data as in Figure 5 except that the data in Figure 6 are logarithmically smoothed to increase the reliability of the individual points. The plotted smoothed curve is then analyzed via linear regression to determine the slope of the post-peak line (down to near background) which best fits the data representing the evaporative process. The determined slope is multiplied by –10,000 to convert it to a positive integer; this becomes the reported MESERAN slope value with units of smoothed (ln (counts – background))/sec x (–10,0000).

Figure 6. Logarithmic plot of smoothed data showing low, medium, and high levels of organic contamination.

56

M.G. Benkovich and J.L. Anderson

The slope method is more sensitive, especially at low levels of contamination. For measuring microorganic residues, the total counts method of analysis (i.e., the area under each overall curve based on actual counts) is valid from somewhat less than 200 ng to approximately 100,000 ng (100 µg). Higher total counts are indicative of higher organic residue levels and vice versa. The slope method of data analysis, normally based on statistically smoothed data and based on the log count vs. time relationship, increases the sensitivity of the lower limit markedly (to less than 1 ng) since the total counts method (the total area under each curve) approaches statistical insignificance somewhat below 200 ng. Expressed as a positive integer, the higher (or steeper) the slope, the cleaner the surface and conversely, the lower (or more flat) the slope, the higher the residue. Both the total counts and slope methods of analysis can be used qualitatively or quantitatively. The total counts method has been used for approximately 30 years at KCP by testing a surface with the MESERAN Analyzer and comparing the results to those obtained from known clean standards for that particular surface. The total counts of the clean standard are subtracted from the total counts obtained on the surface being tested to give a net total counts representing the contamination amount. This result can be compared to previously performed calibrations of contamination to obtain a quantitative result for the contamination amount. Similarly, the slope method can be used to compare the slope obtained on the surface being tested to the slope obtained from known clean standards for that particular surface. The slope can also be compared to previously performed calibrations of contamination to obtain a quantitative result for the contamination amount. In many cases, quantitative data are not needed. For instance, if one is performing process control work to determine if the cleaning process is performing as designed, quantitative data on the actual amount of contamination may not be necessary. Often times, as long as the parts being cleaned are less than a certain level of contamination, they are clean enough. Therefore, one only has to establish the MESERAN total counts or MESERAN slopes that correspond to that level of contamination and relate the tests as being in compliance or not. KCP has used this technique for years to control cleanliness and compare the abilities of different cleaners and cleaning processes to remove various contaminants. Net total count values were established that corresponded to electrical failures and catastrophic adhesion failures. As long as the MESERAN net total counts were below these levels, no cleaning related failures occurred [4-10]. In recent years, KCP has been incorporating the use of the slope technique to give more quantifiable data for lower amounts of contamination. Calibrations of various contaminants have been performed by KCP to develop calibration curves for these contaminants on substrates of interest. MESERAN slope results obtained can now be compared to the calibration curves to determine quantitative amounts of contamination detected [11-14]. The volumetric dilution process for making calibration solutions is shown in Figure 7.

A new method using MESERAN technique

57

Figure 7. Calibration solutions formulations.

Volumetric dilutions were used to make contamination solutions for depositing known amounts of the organic contaminant on reference substrates (e.g., aluminum panels, stainless steel disks, glass cones, etc.). These calibrations were performed in the following manner. A master calibration solution was prepared in a 10-mL volumetric flask by dissolving 100 mg of the organic contaminant in 10 mL of solvent (e.g., cyclopentane, methylene chloride, or hexane that has been double distilled in an all-glass still with no grease in the joints – NVOR of these solvents are approximately 10 ppb). The master calibration solution was thoroughly mixed and 1 mL of this solution was placed in another 10-mL volumetric flask. The second volumetric flask was then diluted with the double distilled solvent until the solution level was at 10 mL and this solution was thoroughly mixed. Subsequent dilutions were carried out in a similar fashion. Ten microliters (µL) of each calibration solution were deposited on the precleaned substrates and allowed to evaporate. This resulted in the following amounts of contamination on the substrates: 1 ng, 10 ng, 100 ng, 1 µg, 10 µg, and 100 µg. Some intermediate levels were obtained by depositing 3 µL and 5 µL of the calibration solutions. The substrates that were contaminated were then tested using the MESERAN Analyzer to develop a calibration curve for the contaminant. For example, calibration curves for Dioctyl Phthalate (DOP) using radiochemical test solution AK (tetrabromoethane-C14 in cyclopentane) on aluminum panels are shown in Figures 8 and 9. Figure 8 shows the calibration curve for DOP using the total counts method of analysis. Figure 9 shows the calibration curve for DOP using the slope method of analysis. As can be seen from examining Figure 8, the total counts method of analysis loses its ability to differentiate contamination amounts (i.e., loses its statistical significance) below a few hundred nanograms of contamination (approximately a monolayer). However, the slope method of analysis shown in Figure 9 is able to differentiate contamination amounts down to 1 ng. In general practice, total counts can be used to quantify contamination amounts greater than a monolayer (a few hundred nanograms) up towards the 100 µg range. The slope method can be used to quantify contamination levels well below the monolayer (down to a nanogram) as well as up to approximately 100 µg.

58

M.G. Benkovich and J.L. Anderson

Figure 8. Calibration curve for DOP on aluminum panels using MESERAN total counts.

Figure 9. Calibration curve for DOP on aluminum panels using MESERAN low variance slopes.

A new method using MESERAN technique

59

Whenever possible it is advantageous to develop calibration curves for the contaminants of interest. KCP has developed calibration curves for numerous contaminants typically encountered in KCP operations such as oils, greases, mold releases, solder fluxes, resins, etc. However, since one does not always know all of the contaminants that may be present on a part, KCP developed a general calibration curve to use for unknown samples based upon hydrocarbon residues. Thus far, most hydrocarbon residues tested have similar calibration curves for the various amounts of residue. 3. EXPERIMENTAL

3.1. Purpose KCP conducted a cleanliness evaluation to determine the NVOR amounts on aluminum and stainless steel panels which were machined using KCP machining fluids and cleaning processes chosen for production of hardware for a particular customer. This section of the paper describes how KCP used recent advances in MESERAN technology to determine the NVOR amounts on four stainless steel panels and four aluminum panels (31 in2 each, excluding edges) by extracting the panels with methylene chloride and quantifying the extracted residues in mg/ft2. The virgin methylene chloride solvent was also evaluated so that its contribution could be subtracted from the solvent extracts. The customer specifically requested that the results be reported in mg/ft2 as opposed to µg/cm2, therefore that is how the results are reported in this paper. It is common practice in the Aerospace industry (as well as other industries) to report contamination amounts on large surfaces in mg/ft2. The conversion factors for these units are 1 mg/ft2 is equivalent to 1.0764 µg/cm2 or 1 µg/cm2 is equivalent to 0.929 mg/ft2. 3.2. Sample details For the NVOR evaluations, four samples each of the aluminum and stainless steel (10 cm x 10 cm x 0.7 cm) were machined at KCP using particular machining fluids and associated machining methods. The four KCP machining fluids evaluated were a hydrocarbon blend (mixture of 70% Pennex N 47 and 30% Hangsterfer’s Hard Cut # 511) and three aqueous-based coolants (Cimtech 200, Trimsol, and Cimstar 3700). The suppliers for these materials are: Pennex N 47 – Exxon Company, Houston, TX; Hangsterfer’s Hard Cut #511 – Hangsterfer’s Laboratories, Mantua, NJ; Cimtech 200 – Cincinnati Milacron Marketing, Cincinnati, OH; Trim Sol – Master Chemical Corporation, Perrysburg, OH; and Cimstar 3700 – Cincinnati Milacron Marketing, Cincinnati, OH. Two of the stainless steel samples were improperly labeled; therefore, the contaminant for these two panels is not known for sure. They were either contaminated with the hydrocarbon blend or Cimtech 200 and are described as such in subsequent portions of this paper (including several tables). All of the stainless steel samples were passivated by the KCP plating

60

M.G. Benkovich and J.L. Anderson

group. This evaluation was conducted on these eight small samples to allow cleanliness verifications to be performed at KCP prior to cleaning large samples (25.4 cm x 25.4 cm x 1.3 cm) that would be sent to the customer for cleanliness verification. 3.3. Cleaning process The aluminum and stainless steel panels tested in this evaluation were cleaned using the following steps: (1) ultrasonic cleaned in Dirl-Lum 603 (30 g per liter concentration) for 5 minutes at 140°F (60°C), (2) rinsed in flowing DI water for 15– 30 seconds, (3) DI water rinsed in ultrasonic cascade rinse station with 3 tanks (30 seconds in each tank) at 110–115°F (43.3–46.1°C), (4) blown dry with filtered nitrogen, and (5) baked for 30 minutes minimum at 220°F (104.4°C )in a HEPA filtered convection oven with nitrogen flowing into the oven. The panels were then packaged in nylon bags and heat sealed. Dirl-Lum 603, supplied by Blue Wave Ultrasonics, Davenport IA, is a powdered alkaline cleaner. It contains sodium metasilicate, sodium carbonate, sodium tripolyphosphate, dodecyl benzene sulfonate, polyethoxyolated phenol, and nonyl phenol. 3.4. Customer cleanliness requirements and associated problems The customer has cleanliness level requirements for this hardware which can be extremely difficult to measure. The desired cleanliness of the hardware is <0.1 mg/ft2. Cleanliness measurements performed by the customer for these parts are typically carried out using a gravimetric NVR procedure. This procedure requires that the part being measured is rinsed with a known volume of a “clean” solvent (methylene chloride) to extract contamination from the part. The solvent and extracted residues are caught in a clean receptacle and evaporated in a precleaned and preweighed dish. After all of the solvent extract has evaporated, the dish is reweighed to obtain a weight of the dish plus the residue. The dish weight is subtracted from the dish plus residue weight to determine the level of the contamination extracted from the part. Similar evaluations are performed on the virgin solvent to determine the residue in the solvent itself. This residue amount is then subtracted from the extracted residue amount to give a final result for the residue extracted. Gravimetric analysis can be a difficult technique to use consistently when measuring low levels of contamination (<1 mg) because many factors can affect these small weight measurements. Customer cleanliness criterion of <0.1 mg/ft2 further complicates matters because error is prone to being introduced during sample collection and sample processing which can be significant when trying to accurately measure to <0.1 mg/ft2. Every piece of laboratory equipment (such as glassware, weighing trays, funnels, etc.) that comes in contact with the solvent will contribute a small but variable amount of NVR to the solvent. The magnitude of this contribution is not constant due to fluctuations in contact time, surface temperature, and other variables. Although these variables are controlled as well

A new method using MESERAN technique

61

as possible, they still will contribute some level of error to the reported results, and at the low levels of contamination being discussed, this error may be significant. In addition, gravimetric analysis is typically performed after rinsing a 1 ft2 (144 in2) sample area while the samples used in this evaluation were significantly smaller (31 in2). 3.5. KCP proposed method of evaluating NVR levels Due to the problems discussed above, KCP proposed the following: (1) Extract the contamination from the panels with methylene chloride just as would be done in the gravimetric procedure, (2) use the newly developed MicroSolventEvaporator to evaporate 250-µL aliquots of the methylene chloride extract onto clean reference substrates, (3) test the evaporated residues using the MESERAN Analyzer to quantify the NVOR levels extracted from the samples, and (4) perform gravimetric analysis on the remaining extract to determine if there was enough residue to be weighed on a balance. It should be noted that gravimetric NVR methods will weigh all contamination (organic, inorganic, and particles such as metals) whereas the MESERAN method detects only NVOR (inorganic contamination and metal particles are not detected). The real thrust of these evaluations is to quantify the organic contamination and it is estimated that the majority of the contamination being extracted from the parts is organic residue. Whereas, the contribution of inorganic and metal particles is thought to be small. 3.6. KCP machining fluid information The ingredients for the machining fluids (as obtained from their Material Safety Data Sheets) are as follows: Hydrocarbon Blend Pennex N 47 – petroleum distillates (~ 78%) and proprietary additives (~ 22%). Hangsterfer’s Hard Cut #511 – petroleum oil, chlorinated paraffin, and triglycerides. Cimtech 200 Ethanolamine (10% max), caprylic acid (10% max), triethanolamine (10% max), isononanoic acid (10% max), and balance water. Trim Sol Petroleum oil (30–40%), petroleum sulfonate (20–30%), chlorinated alkene polymer (20–30%), nonionic surfactant (1–10%), aromatic alcohol (1–10%), propylene glycol ether (1–10%), propylene glycol (<1%), substituted indole (<1%), blue-green dye (<1%), silicone defoamer (<1%), and balance water. Cimstar 3700 Mineral Oil (10% max), diethanolamine (10% max), triethanolamine, (10% max), aminomethylpropanol (10% max), and balance water.

62

M.G. Benkovich and J.L. Anderson

3.7. Calibration of hydrocarbon blend Based upon the volatility evaluation and petroleum oils being the most likely contaminant from the machining fluids being evaluated, it was decided to develop a calibration curve using MESERAN Analysis for the hydrocarbon blend as a reference for future cleanliness evaluations. A master calibration solution was prepared in a 10-mL volumetric flask by dissolving 100 mg of the hydrocarbon blend in 10 mL of hexane (NVR<1 mg/L). Volumetric dilutions were used to make contamination solutions for depositing known amounts of the hydrocarbon blend on stainless steel disks. The master calibration solution was thoroughly mixed and 1 mL of this solution was placed in another 10-mL volumetric flask. The second volumetric flask was then diluted with hexane until the solution level was at 10 mL and this solution was thoroughly mixed. Subsequent dilutions were carried out in a similar fashion. Ten µL of each calibration solution were deposited on the precleaned stainless steel disks and allowed to evaporate. This resulted in the following amounts of contamination on the stainless steel disks: 1 ng, 10 ng, 100 ng, 1 µg, 10 µg, and 100 µg. Some intermediate levels were obtained by depositing 3 µL and 5 µL of the calibration solutions. The volumetric dilution process for making calibration solutions was illustrated previously in Figure 7. 3.8. Overview of extractions and NVR measurement processes One-hundred mL of methylene chloride (Optima Grade, stated residue after evaporation – 1 ppm) were used to extract both sides of the cleaned aluminum and passivated and cleaned stainless steel panels. The panels were 10 cm x 10 cm x 0.7 cm for a total of 200 cm2 (31 in2) surface area (discounting the edges). The samples were placed (one at a time) in a large cleaned stainless steel funnel. Onehundred mL of the methylene chloride were measured and poured into a cleaned Teflon squeeze bottle. Each side of the panel was rinsed with the solvent to extract contamination. The methylene chloride extract drained from the funnel into precleaned glass bottles which were then capped. The caps had Teflon inserts and were screwed onto the bottles. The funnel was ultrasonically cleaned in aqueous Dirl-Lum 603, ultrasonically rinsed in cascading DI water, blown dry with nitrogen, dried in a HEPA filtered oven, and rinsed with virgin methylene chloride after each extraction before being used for the next sample. Prior to being used, all glassware (bottles, graduated cylinders, etc.) used in the extraction process was also precleaned using the above process. The methylene chloride extracts were thoroughly mixed using an ultrasonic cleaner and vigorous shaking of the bottle prior to taking aliquots for analysis. Using the MicroSolventEvaporator and clean microsyringes, 250 µL of the extracts were deposited onto cleaned stainless steel reference substrates. The methylene chloride evaporation process using the MicroSolventEvaporator takes 10–

A new method using MESERAN technique

63

20 minutes per 250 µL sample when performed at 30°C. The evaporated residue was then tested with the MESERAN Analyzer. A minimum of 5 replicates were run for each condition tested. The data were compared to calibrations performed using the KCP hydrocarbon blend cutting oil (mixture of 70% Pennex N 47 and 30% Hangsterfer’s Hard Cut # 511) deposited on the same stainless steel substrates. Interpolations between the calibration points were performed to determine how much residue was in the extract. This amount was then multiplied by the appropriate factor to determine the quantity of residue in the entire amount of extract that was obtained from each panel. The amount of extract obtained from each sample was measured so this calculation could be performed. Recovery of 70–80% of the 100-mL of methylene chloride solvent was typical. The NVR of the methylene chloride was analyzed in a similar fashion to generate a baseline value for the solvent. Ultimately, the NVR of the methylene chloride was subtracted to determine the amount of residue actually extracted from the sample panels. This result was then converted to an amount per square foot for comparison with the customer’s specification. 3.9. KCP gravimetric analysis Only a small portion of the solvent extracts (generally about 4–5 mL out of the approximate 80 mL obtained after the extraction) was used in the evaluation conducted with the MicroSolventEvaporator and MESERAN Analyzer. Therefore, the remainder of each solvent extract (generally 70–80 mL) was sent to KCP’s Analytical Sciences Laboratory for complete evaporation in an attempt to quantify the resulting residue gravimetrically. A Mettler AE163 balance was used to make the weight measurements. The calibration label on the balance indicates the performance specification is +/– 0.005% of reading + 0.1 mg. The weights were measured to the nearest 0.01 mg. There is definitely greater error associated with weighing these small amounts. KCP Analytical Sciences personnel indicated that the residue values determined gravimetrically were at best good only to approximately +/– 0.03 mg. The virgin methylene chloride and the methylene chloride extracts were processed as follows: 1. Weighing trays were baked in an oven for 1 hour @ 105°C. 2. The weighing trays were taken from the oven and placed in a desiccator and allowed to cool. 3. The weighing trays were placed on an analytical balance and massed to the nearest 0.01 mg. 4. The methylene chloride and methylene chloride extracts were poured into individual weighing trays and allowed to evaporate. This process continued until all of the solvent evaporated for each sample. 5. The weighing trays were placed in an oven for 1 hour @ 105°C.

64

M.G. Benkovich and J.L. Anderson

6. The weighing trays were removed from the oven and placed in a desiccator to cool. 7. The weighing trays were placed on an analytical balance and massed to the nearest 0.01 mg. 8. The NVR of the methylene chloride was calculated by subtracting the result of step 3 from the result of step 7. Two samples of virgin methylene chloride were evaluated using this procedure. One sample was 100 mL of methylene chloride taken straight from the original bottle, measured in a clean graduated cylinder, poured into a precleaned glass bottle, and then sealed with caps containing Teflon inserts. The second methylene chloride NVR sample was obtained using the following procedure: (1) 100 mL of methylene chloride from the original bottle was measured in a clean graduated cylinder, (2) the methylene chloride was poured into a precleaned Teflon squeeze bottle, (3) the precleaned stainless steel funnel used to hold the samples that were extracted was rinsed with 100 mL of methylene chloride, (4) the methylene chloride drained through the funnel and was captured in a precleaned glass bottle, and (5) the glass bottle was sealed with caps containing Teflon inserts. This process captured 89 mL of methylene chloride for analysis. After aliquots from the extracts were analyzed using the MicroSolventEvaporator and MESERAN Analyzer, the remainder of each of the extracts was poured into clean graduated cylinders to measure the amount of solvent left for gravimetric analysis. The extracts were then poured back into their respective bottles and sent to be evaporated and measured gravimetrically. 3.10. Gravimetric NVR evaluations of large panels by the customer Based upon the MESERAN results obtained for the small panels (31 in2), KCP was convinced that they could clean the large panels to acceptable levels for the customer. Therefore, large panels were manufactured with the same machining fluids, cleaned, and sent to the customer for evaluation. The customer evaluated the samples using their extraction and gravimetric NVR procedure described below. The analytical procedure involves three steps: 1) rinsing the surface to be tested with solvent and collecting the rinse solvent; 2) concentrating the solvent to near dryness by evaporation with a clean gas; and 3) weighing the dry residue to determine the NVR. In the first step, the organic material is rinsed from the metal surface with methylene chloride, which must be completely captured in the sample container. The concentration stage consists of evaporating the solvent first by bubbling with clean gas (helium or nitrogen) and then by blow-down after transferring the residual solvent with the NVR into progressively smaller containers to minimize the container surface area and potential loss of NVR. All transfer steps must be quantitative and will require small aliquots of methylene chloride to rinse the containers. The final step is to transfer the remaining solvent containing the NVR into a tared weighing boat and weigh to a constant weight after all of the solvent has evaporated.

A new method using MESERAN technique

65

4. RESULTS AND DISCUSSION

4.1. KCP machining fluid volatility A quick evaluation was made to determine the volatility of the machining fluids. The fluids were weighed out in a weighing dish, allowed to air dry at room temperature for 20 hours, and then reweighed. The results are shown in Table 1. As would be expected from looking at their ingredients, the hydrocarbon blend is the least volatile and the aqueous machining fluids are the most volatile. Therefore, theoretically, there should be larger quantities of the hydrocarbon blend to be cleaned off the samples than the aqueous cutting oils. This should make it much easier to clean the aqueous machining fluids to acceptable levels since their residue amount is relatively low even if they are not removed. 4.2. Calibration results for hydrocarbon blend The stainless steel disks that were contaminated with various amounts of the hydrocarbon blend were tested using the MESERAN Analyzer to develop a calibration curve. Example plots of the MESERAN data for these calibrations at various contamination levels are shown in Figure 10. The low variance slope calibration curve for the hydrocarbon blend on these stainless steel disks is shown in Figure 11. 4.3. MESERAN analysis of virgin methylene chloride and methylene chloride extracts The newly developed MicroSolventEvaporator (a prototype designed and developed by The MESERAN Company) was used to deposit and evaporate the methylene chloride onto precleaned stainless steel disks. This system was used to evaporate 250 µL of the methylene chloride onto the stainless steel surfaces in sequential small quantities (~ 5–10 µL increments) to concentrate the residue in a small area so that it could be tested with the MESERAN Analyzer. The evaporations were carried out at 30°C to slightly speed up the process. The MicroSolventEvaporator allowed the evaporations to be carried out in a reproducible fashion to help reduce the error normally associated with manual evaporations (i.e., someone using a syringe and depositing multiple small quantities of solvent in the

Table 1. Volatility of KCP machining fluids Cutting Oil

Initial Weight (g)

Weight After 20 Hour Air Drying (g)

% Volatile

Hydrocarbon Blend Cimtech 200 Trim Sol Cimstar 3700

0.11322 0.10790 0.17500 0.11665

0.11292 0.00290 0.01100 0.00400

0.26 97.31 93.71 96.57

66

M.G. Benkovich and J.L. Anderson

Figure 10. Plot of various amounts of hydrocarbon blend (70% Pennex N 47 and Hangsterfer’s Hard Cut # 511) on stainless steel disks using smoothed Ln (counts – background). (MESERAN tests carried out using radiochemical BK).

Figure 11. MESERAN low variance slope calibration of hydrocarbon blend (70% Pennex N 47 & 30% Hangsterfer’s Hard Cut # 511) on stainless steel disks with 0.015 inch/RPM spiral grooves.

A new method using MESERAN technique

67

same area until a large volume of solvent has been evaporated). In addition, these stainless steel disks have grooves machined in them which act as a self-centering mechanism so that the solvent evaporates in a confined space and does not migrate elsewhere. After the methylene chloride was evaporated onto the stainless steel disks with the MicroSolventEvaporator, the disks were tested with the MESERAN Analyzer using radiochemical BK (best suited for detecting nonpolar residues). Table 2 shows the average low variance MESERAN slope results for the 250-µL depositions from the methylene chloride extracts. The results were compared to those obtained previously from the calibration using known amounts of the KCP hydrocarbon blend machining fluid (mixture of 70% Pennex N 47 and 30% Hangsterfer’s Hard Cut # 511) on these same substrates, which were shown previously in Figure 11. The average low variance slope obtained for the methylene chloride blanks was 2412. The MESERAN slope value has units of smoothed (ln (counts – background))/sec x (–10,0000); however, the units are generally not given when the data are shown. A linear-log Table 2. MESERAN low variance slope results for 250-µL depositions from methylene chloride extractions Description of Sample Tested

Methylene Chloride (Straight from Bottle) Hydrocarbon Blend on Aluminum Hydrocarbon Blend or Cimtech 200 on Stainless Steel Hydrocarbon Blend or Cimtech 200 on Stainless Steel Cimtech 200 on Aluminum Trim Sol on Aluminum Trim Sol on Stainless Steel Cimstar 3700 on Stainless Steel Cimstar 3700 on Aluminum

Average MESERAN Low Variance Slope (smoothed (ln (counts – background))/sec x (–10,0000))

Standard Deviation

2412

111.71

4.63

2392

186.34

7.79

2319

196.57

8.48

2324

176.14

7.58

2049

45.35

2.21

2205

351.69

15.95

2399

55.25

2.30

2208

46.33

2.10

2385

141.25

5.92

Coefficient of Variation (%)

68

M.G. Benkovich and J.L. Anderson

interpolation using the values from the calibration in Figure 11 was performed to calculate the contamination level. The slope of 2412 corresponds to a contamination level of 3 ng (or 0.000003 mg) for the 250 µL of methylene chloride evaporated on the stainless steel disks. Therefore, the NVR of the 100-mL methylene chloride blanks would be 1.2 µg (or 0.0012 mg) as calculated below: (3 ng/250 µL) x (1000 µL/1 mL) x (100 mL) x (1 µg/1000 ng) = 1.2 µg (or 0.0012 mg)

The contamination results obtained from the MESERAN Analysis for the methylene chloride extracts from the panels tested are shown in Table 3. The average low variance slopes obtained for the samples were converted into equivalent contamination amounts (in nanograms) for the 250 µL of extract evaporated by performing a linear-log interpolation using the values from the calibration curve in Figure 11. Then the amount of contamination in all of the extract collected for each sample was calculated (shown in micrograms) using similar equations as shown previously for the methylene chloride blanks. Next, the methylene chloride NVR was subtracted from the contamination amount in all of the extract collected to determine the NVR amount extracted from each sample (this result is shown as µg/31 in2). Finally, this contamination was converted to an amount per square foot to compare to the customer requirement of <0.1 mg/ft2. The results of all samples tested passed the requirement of <0.1 mg/ft2 and most of them were significantly lower. 4.4. KCP gravimetric analysis results The results obtained in the gravimetric analysis on the methylene chloride blanks and the methylene chloride extracts are shown in Table 4. Since the extract amount analyzed gravimetrically was less than the original extract amount (due to aliquots being taken for MESERAN Analysis), the gravimetrically determined NVRs were factored up to account for the lost contamination. In all cases except two, the contamination determined gravimetrically was less than the gravimetrically determined methylene chloride NVR, which by definition makes the extracted residue less than the customer limit of <0.1 mg/ft2. One of the hydrocarbon blend or Cimtech 200 samples on stainless steel had a positive result after the methylene chloride NVR was subtracted. After converting the contamination from mg/31 in2 to mg/ft2, the contamination determined is higher than the customer limit of <0.1 mg/ft2. The Cimstar 3700 on stainless steel sample also had a positive result after the methylene chloride NVR was subtracted; however, the resulting contamination was still less than the customer limit of <0.1 mg/ft2. The error associated with measuring these small amounts of contamination makes all of these gravimetric results doubtful and it is recommended that they be dismissed. 4.5. Results of gravimetric NVR evaluations of large panels by the customer The gravimetric NVR results obtained by the customer on the large aluminum and stainless steel panels (100 in2) machined, cleaned, and packaged at KCP are

A new method using MESERAN technique 69

70

M.G. Benkovich and J.L. Anderson

A new method using MESERAN technique

71

shown in Table 5. The customer determined that 100 mL of virgin methylene chloride had a NVR of 0.004 mg. They did not subtract this amount from the final results of the extracts when they reported their results. The results of every panel evaluated by the customer indicated that the contamination level was less than their requirement of <0.1 mg/ft2. Table 5. Gravimetric analysis of methylene chloride extracts from 100 in2 aluminum and stainless steel panels performed by the customer Description of Sample Tested

NVR Level (mg/ft2)

Hydrocarbon Blend on Aluminum Hydrocarbon Blend on Stainless Steel Cimtech 200 on Aluminum Cimtech 200 on Stainless Steel Trimsol on Aluminum Trimsol on Stainless Steel Cimstar 3700 on Aluminum Cimstar 3700 on Stainless Steel

0.03 0.04 0.07 0.04 0.04 0.02 0.02 0.06

5. CONCLUSIONS

Analysis of the solvent extracts with the MicroSolventEvaporator and MESERAN Analyzer indicated that all of the samples tested passed the customer requirement of <0.1 mg/ft2 with most of them significantly lower. Thus, the results indicated that the cleaning process used sufficiently cleaned the cutting oils from the panels tested. In short, the amounts of contamination being determined are too small to be accurately measured gravimetrically for the small samples extracted (31 in2) because the error associated with these measurements is extremely high at these levels. These results provide further proof that for gravimetric NVR values to be accurate, it is best to extract at least a one square foot area so there is plenty of contamination in the extract to measure properly. Therefore, when quantifying extremely small residues using solvent extracts from small parts, the only viable method currently available is the MicroSolventEvaporator/MESERAN Analysis method. The large panels (100 in2) evaluated gravimetrically by the customer indicated that all of the panels were cleaned to acceptable levels (i.e., with residual contamination less than the customer requirement of <0.1 mg/ft2). Even though panels the customer evaluated were significantly larger than ones evaluated by KCP

72

M.G. Benkovich and J.L. Anderson

and different methods were used for determining the NVR levels, the results obtained were remarkably consistent. The NVR data determined by KCP with the MicroSolventEvaporator and MESERAN technique and the NVR data determined gravimetrically by the customer follow a similar pattern. Tables 3 and 5 indicate that the results of both test methods exhibited the same cleanliness level relationship between the aluminum and stainless steel panels for each specific contaminant evaluated. Hydrocarbon blend-contaminated samples had more contamination after cleaning on the stainless steel panels than on the aluminum panels. Cimtech 200-contaminated samples had more contamination after cleaning on the aluminum panels than on the stainless steel panels. Trimsol-contaminated samples had more contamination after cleaning on the aluminum panels than on the stainless steel panels. Cimstar 3700-contaminated samples had more contamination after cleaning on the stainless steel panels than on the aluminum panels. Acknowledgements The work of Garth Christoff, Tom Hand, and Ed Fuller was instrumental in the performance of these evaluations. REFERENCES 1. J.L. Anderson, “Quantitative Detection of Surface Contaminants,” Journal of the American Association of Contamination Control, II (6),9 (1963). 2. J.L. Anderson et al., “Measurement and Evaluation of Surfaces and Surface Phenomena by Evaporative Rate Analyses,” Journal of Paint Technology, 40, No. 523, 320-327 (August 1968). 3. J.L. Anderson, “Evaporative Rate Analysis: Its First Decade”, in: Characterization of Metal and Polymer Surfaces, L.H. Lee (Ed.), Vol 2, pp. 409-427, Academic Press, New York (1977). This paper summarizes all known references prior to 1975. 4. L.C. Jackson, “Solubility Parameters and Evaporative Rate Analyses in Organic Residue Characterization” (Topical Report), UNCLASSIFIED, Bendix-Kansas City Division: BDX-6131099, March 1974. (Available from NTIS) 5. L.C. Jackson, “Contaminant Removal Based on Solubility Parameter and Evaporative Rate Analysis Technologies” (Topical Report), UNCLASSIFIED, Bendix-Kansas City Division: BDX-613-1128, August 1974. (Available from NTIS) 6. L.C. Jackson, “How to Select a Substrate Cleaning Solvent,” Adhesives Age, 22-31 (December 1974). 7. L.C. Jackson, “Solvent Cleaning Process Efficiency,” Adhesives Age, 31-34 (July 1976). 8. L.C. Jackson, “Removal of Silicone Grease and Oil Contaminants,” Adhesives Age, 29-32 (April 1977). 9. L.C. Jackson, “Contaminant Cleaning for Critical Electrical Assembly Areas” (Final Report), UNCLASSIFIED, Bendix-Kansas City Division: BDX-613-1695, February 1978. (Available from NTIS) 10. M.G. Benkovich, “Solvent Substitution for Electronic Products,” International Journal of Environmentally Conscious Manufacturing, 1, No. 1, 27-32 (1992). 11. M.G. Benkovich and J.L. Anderson, “Measurement of Organic Residues on Surfaces to a Low Fraction of a Monolayer,” Precision Cleaning, 16-28 (May 1996). This paper includes many of the more current references to MESERAN technology.

A new method using MESERAN technique

73

12. M.G. Benkovich and J.L. Anderson, “Quantification of MicroOrganic Residues to Low Nanogram Levels,” Precision Cleaning ’96 Proceedings, 115-122 (1996). 13. J.L. Anderson, R.F. Russell and M.G. Benkovich, “Quantitative Measurement of Extremely Low Levels of Non-Volatile Residues (NVR) on Surfaces and in Liquids,” Precision Cleaning ’97 Proceedings, 96-108 (1997). 14. J.L. Anderson, R.F. Russell and M.G. Benkovich, “Solvent NVR: A Problem and a Solution,” CleanTech ’98 Proceedings, 331-340 (1998). 15. MESERAN Analyzer Literature, ERA Systems Inc., The MESERAN Company, Chattanooga, TN.

This page intentionally left blank

Surface Contamination and Cleaning, Vol. 1, pp. 75–84 Ed. K.L. Mittal © VSP 2003

Methods for pharmaceutical cleaning validations HERBERT J. KAISER∗ STERIS Corporation, P.O. Box 147, St. Louis, MO 63166

Abstract—Cleaning validations are very difficult to perform. They can be made easier if an appropriate method for analyzing the samples is used. The method used should be based on the previously established residue limits of both the active and cleaning agent. There are many choices of analytical techniques that can potentially be used. This article describes various analytical techniques that are available for use, particularly for cleaning agent residues. Keywords: Pharmaceutical cleaning; cleaning validation; total organic carbon; high performance liquid chromatography; Fourier transform infrared spectroscopy; ultraviolet detection; evaporative light scattering detection; mass spectrometry; photoelectron emission; residue; method validation.

1. INTRODUCTION

In the pharmaceutical industry cleaning is an important component of the manufacturing process. Regulatory agencies both in Europe and North America require that the cleaning process be validated [1]. It must be clearly shown that product residues are removed to an acceptable level. In addition, not only do the manufacturers have to show that they have cleaned their product to an acceptable level but they have to demonstrate that they have removed the cleaning agent also to an acceptable level. Cleaning validations require tools to measure the residue levels left on surfaces. These tools are required to be validated. This poses a challenge to analytical chemists in that they must first choose the appropriate method to measure the residues and then they must validate that method. If the method cannot be validated it cannot be used to measure residue levels. There are many types of analytical techniques that can be used for these analyses. The methods can be specific or non-specific. The chemist can also utilize complementary techniques such as total organic carbon (TOC) determination or high performance liquid chromatography (HPLC). The analytical chemist must examine the pros and cons of each technique in order to choose the appropriate technique. Again, whichever method or methods are chosen they must be validatable.

∗

Phone: (314) 290-4725, Fax: (314) 725-5687, E-mail: [email protected]

76

H.J. Kaiser

2. CHOOSING THE ANALYTE

The analytical chemist must determine what is being measured. Is the residue a drug active, formulation component, or the cleaning agent? Is the residue organic or inorganic? Is the residue water soluble or water insoluble? Is the residue particulate, microbial or is it an endotoxin? The answers to these questions will narrow the choice of analytical techniques available. For example, if the residue is a drug active, HPLC or gas chromatography (GC) could be utilized. If the residue is a formulation component or a cleaning agent, TOC determination may be the method of choice. An important piece of information that the analytical chemist will require is the target residue limit. The target residue limit should always be established prior to the selection of the analytical technique. Once the residue limit has been established the analytical technique can be chosen and a method developed that can determine residue levels below the acceptance limit [2, 3]. 3. SPECIFIC VERSUS NON-SPECIFIC TECHNIQUES

Once the type of residue has been established the analytical chemist can chose between either specific or non-specific techniques. A specific technique detects a unique compound. A non-specific technique detects any compound that produces a certain response. Examples of specific techniques include HPLC, Fourier transform infrared spectroscopy (FTIR), ion chromatography (IC), atomic absorption (AA), inductively coupled plasma atomic emission (ICP), capillary electrophoresis (CE) and various protein methods. Examples of non-specific techniques include TOC determination, pH determination, acid/base titrations and conductivity. 4. TECHNIQUES FOR SAMPLING

In addition to choosing an appropriate technique a method of sampling must also be established. There are two acceptable forms of sampling. One form is a direct surface swab. This is where a swab is utilized to rub a known surface area to recover the residue. The other acceptable form of sampling is rinse water sampling. In this technique the equipment is rinsed with a known volume of water and then the water is analyzed for its residue content. The test must be a direct measure of potential contaminants and not just compendial tests for water (e.g., a U.S. Pharmacopeial test). In other words, if rinse water is used as the sample a method must be utilized that can measure the contaminants coming off the surface, not just the quality of the rinse water. The analysis cannot just be based on standard water quality parameters. Also, if rinse water is utilized it must be demonstrated that the rinse water does indeed remove the residues from the surface of the materials being cleaned. A questionable sampling technique is the use of placebos. This is a

Methods for pharmaceutical cleaning validations

77

technique where, after the manufacture of a normal drug product, the product, less the drug active, is manufactured again. A sample is then taken of this material and analyzed for the drug active in the previous batch. This is a questionable sampling technique because it assumes that the residue is distributed uniformly throughout the placebo. Another problem with placebo sampling is that the residue may be so dilute in the placebo that the analytical technique may not be able to detect it. This is an acceptable technique only if it is used along with swab or rinse water data. Therefore, since swabbing or rinse water analysis must be done in conjunction with placebo sampling, this technique is rarely utilized. Placebo sampling is important when validating the cleaning of filling equipment or small pieces of equipment where direct rinse water measurements are not practical. 5. ANALYZING THE SAMPLE

5.1. Interferences If the detection technique is non-specific a good strategy is to first identify possible sources of interference. These interferences could be either positive or negative. For example, in TOC determinations, if the person performing the sample collection accidentally coughs onto the surface being analyzed this would cause a positive interference. Since TOC measures a non-specific property, the residue amount would be calculated as if all of the measured property were due to that residue. For example, if the target residue was aspirin and the technique utilized was TOC determination, the TOC measured would be calculated as if all of the TOC came from the aspirin even though there may be excipients and/or detergents present. This provides an upper limit value and is a worst-case situation. The identification of possible interferences is important to both specific and non-specific methods. In the case of specific methods there should not be any interferences if the method was properly developed. It is important that the method is able to detect the analyte after exposure to the cleaning environment. Studies should be conducted to demonstrate that the analyte does not change after exposure to alkaline or acidic cleaners. This can be simply done by exposing the drug active to a dilution of the cleaning agent at the temperature that will be utilized during the cleaning process. This sample can then be analyzed to determine if the analyte can still be detected and quantitated. If this is not done the analyte may not be detected even though it is present in the sample. If the cleaning environment does change the nature of the analyte a new method would be required or modifications would have to be made to the existing method. When utilizing a non-specific method to measure a non-specific property, any compound that displays that property and is introduced into the sample will interfere. Possible sources of interferences could be from the environment, sampling technique, and/or people.

78

H.J. Kaiser

5.2. High performance liquid chromatography The most common analytical technique available to pharmaceutical analytical laboratories is the HPLC. There are many different detectors that are used in HPLC analyses. These include ultraviolet (UV), fluorescence, electrochemical, refractive index, conductivity, evaporative light scattering and many others. The most common detection technique is ultraviolet. However, evaporative light scattering detection (ELSD) is a very good method for the detection of surfactants. 5.2.1. UV detection There are many advantages in using a UV detector. A large number of compounds can be detected because they contain chromophores. UV detection also allows for diode array spectral possibilities. This feature is very useful when examining peaks for coeluting compounds. UV detectors are easy to use since they require no special reagents. In addition, temperature is not an important consideration since it typically does not affect molar absorptivities. This means that no special heating or cooling are required for the detector. There are disadvantages in using UV detection. UV detectors are not universal. This means that not all compounds have chromophores and, therefore, are extremely difficult, if not impossible, to detect using UV detection. Limits of detection can be higher than other detector types due to background interferences. These interferences are especially predominant when low wavelengths are used in combination with a gradient elution scheme. HPLC analysis along with UV detection is most commonly used for the analyses of drug actives. However, this technique is also sometimes utilized for the detection of surfactants that may have been left behind by the cleaning agent. There are many problems associated with surfactant analyses using chromatographic methods. High levels of detection result unless special sample preparations are taken. There are many examples in the literature where surfactants are analyzed in environmental samples, such as river water. However, this is typically accomplished by pre-concentrating the sample up to 1000-fold. Due to the higher level of detection, the quantitation levels are also higher. Again, this problem can be avoided by pre-concentrating the sample. Also, the peaks present in a chromatogram of a surfactant must be summed to obtain an appropriate level of quantitation. This is due to the fact that a surfactant material is not a single compound. Surfactants are a mixture of compounds containing various chain lengths and perhaps various degrees of ethoxylation. Another potential problem is deciding which surfactant to analyze. A cleaning agent may contain more than one surfactant. It may contain an anionic surfactant along with an amphoteric surfactant. The cleaner may contain a combination of anionic surfactants with nonionic surfactants or any number of combinations. Figure 1 shows the chromatogram of a typical ethoxylated surfactant. The different peaks represent different degrees of ethoxylation. All of the peaks must be summed in order to generate the calibration curve. The smaller peaks will disappear into the baseline as the dilution increases. This will produce less than desirable results.

Methods for pharmaceutical cleaning validations

79

Figure 1. Chromatogram of an ethoxylated surfactant using UV detection.

A significant disadvantage to the use of UV detection is that many surfactants used in formulated cleaners do not contain chromophores. ELSD can be used to overcome this. 5.2.2. Evaporative light scattering detection An ELSD is a mass detector and does not depend on the presence of chromophores. The eluent from a typical HPLC column enters the detector through a nebulizer and is carried along by a gas stream through a heated column. The mobile phase is evaporated in the column leaving small particles of the analyte. The small particles are then passed through a detector containing a laser. The light from the laser is scattered by the small particles. The detector then measures the degree of scattering. There are many advantages associated with the use of an ELSD. This type of detector is simple (in theory), versatile and rugged in use. Their manufacturers describe these detectors as “universal detectors”. This is because, in theory, they should be able to detect any non-volatile compound since they solely rely on mass. Since they do rely on mass all compounds should produce a similar response as opposed to UV detection where the extinction coefficient is important. There should also not be any baseline drifts due to mobile phase effects. This is due to the fact that the mobile phase is evaporated prior to entering the detector. There are disadvantages to the use of ELSD. There is a very limited choice of buffer salts that can be utilized. This is because the buffer has to be volatile. If the buffer is not volatile it will remain in the system and flow through the detector. The detector will then detect the buffer salts instead of the analyte. Another dis-

80

H.J. Kaiser

advantage is that the ELSD requires special care and cleaning of the nebulizer. The nebulizer must produce consistent particle sizes so that an accurate detection can occur. Therefore, careful cleaning and maintenance must be performed. 5.3. Capillary electrophoresis Capillary electrophoresis (CE) is a very good method to use for cleaning validations. CE can be used in different modes. These modes include capillary zone electrophoresis, micellar electrokinetic chromatography, and isoelectric focusing. CE can be used for both drug actives and detergent residues. The ability of CE to pre-concentrate samples makes it very useful in the detection and quantitation of surfactants. UV detection is the most common detection technique used in CE. Inverse detection can be readily used with CE. Inverse detection is a procedure where a strongly absorbing compound is utilized in the buffer solutions and produces a strong signal in the detector. When the analyte, which may not be strongly absorbing, passes through the detection window a “negative” peak is produced. Software accompanying CE units can typically invert this “negative” peak producing a normal looking “chromatogram”. 5.4. Ion chromatography Ion chromatography (IC) is another useful technique that can be utilized in residue analyses. This is because IC can be used to quantitate both organics and inorganics. The inorganics may include sulfates, phosphates, chlorides, sodium, potassium and other inorganics that may be present in a cleaning agent. IC can also be used to determine organic acids such as citric or glycolic acids and even surfactants. There are several modes of detection that can be used in IC but the most common are UV and conductivity. IC can analyze multiple analytes in the same sample in many cases. Figure 2 is a typical chromatogram showing sodium and potassium ions that are present in a cleaning agent. Low levels of detection can be achieved especially with the conductivity detector. 5.5. Total organic carbon determination TOC determination is the most commonly used analytical technique in cleaning validations. Instrument manufacturers use several different ways to determine the TOC in a sample. In general, the sample is oxidized and the carbon dioxide produced is measured. The TOC detection can occur in a variety of ways including infrared and conductivity. There are several advantages to using TOC determination as the analytical technique for cleaning validations. TOC is a non-specific measurement technique for carbon. Therefore, it can be used to quantitate any residue that contains carbon. Low limits of quantitation/detection can be achieved with TOC. Manufacturers of TOC units claim a detection limit of <1 ppb. However, quantitation and detection limits are primarily based on the quality of the water used in the analysis. There is typically a high recovery from samples in

Methods for pharmaceutical cleaning validations

81

Figure 2. Chromatogram showing sodium and potassium ions in a cleaning agent using conductivity detection.

TOC analyses. There are also minimal chemical interferences. TOC determination is easy to carry out and involves a minimal amount of method development. Typically, samples are simply diluted and analyzed. No separation procedures need to be developed unlike HPLC. As with other instrumentations, there are automated sampling accessories that can be used. These accessories include online monitors. On-line monitoring should only be used for monitoring water quality. On-line monitoring should not be used for monitoring a cleaning validation study. The instrumentation for TOC determination is also very cost effective. There are two major disadvantages to TOC determination. The first is that the sample must be water-soluble. TOC determination cannot be used as the analytical technique if the residue cannot be dissolved in water. It should be noted that solubility in the TOC determination sense means in the part per billion range rather than in the part per million range. If a compound is reported to be water insoluble it does not automatically mean that it is not soluble enough to be analyzed by TOC determination. The second disadvantage is that samples cannot be prepared in organic solvents. If the samples were prepared in organic solvents the TOC of the solvents and not of the residue would be determined. TOC determination is an excellent complimentary technique to HPLC. TOC can be used to monitor water-soluble ingredients in a formulation and HPLC can be used to monitor the water insoluble ingredients. An example of this is a formulated ibuprofen. The ibuprofen itself is not soluble even at TOC determination levels, but HPLC can be utilized to monitor the ibuprofen. The formulation components are soluble in water and TOC determination can be used to monitor their contribution to the residue.

82

H.J. Kaiser

5.6. Fourier transform infrared spectroscopy FTIR spectroscopy can be used to conduct cleaning evaluations. Fiber optics for mid-IR use is not developed far enough where it could be used in a cleaning validation situation. Therefore, FTIR spectroscopy is used in the lab to evaluate the rinsibility of various residues including detergents. The use of a grazing angle accessory allows for the semi-quantitation of residues on coupons. This can be used in a method where a coupon is spiked with a known amount of residue and allowed to dry. The coupons are then subjected to a simulated rinse procedure and the concentration on the surface of the coupon is monitored using FTIR spectroscopy. The data generated from these studies can be used in establishing rinsing procedures and for choosing appropriate cleaners for the manufacturing environment. 5.7. Photoelectron emission There is a very sensitive direct surface measurement technique that is based on photoelectron emission [4]. This relatively portable instrument has a probe that is placed onto the surface to be analyzed. The instrument uses a UV source to bombard the surface that excites the atoms on the surface. Radiation is emitted when the atoms return to the ground state. The instrument detects the radiation. Clean surfaces generate more radiation than do soiled surfaces. This is an extremely sensitive technique and any change to the surface will affect the amount of radiation detected. The normal wear and tear that occurs to equipment surfaces in the manufacturing environment will cause changes to occur which will affect the readings. 5.8. Mass spectrometry A “portable” mass spectrometer has been developed at Lawrence Livermore National Laboratories [5]. It consists of a probe connected via a cable to a vacuum pump, the electronics, and controller. The tip of the probe forms a tight seal against the surface of the substrate and a vacuum is generated. A heating element heats the surface and vaporizes the residues that may be present on the surface. This instrument not only can quantitate what is on a surface but it can also be used to identify the residues. The only drawback with this instrument is that it requires a relatively flat surface to be effective. 6. METHOD VALIDATION

It is time now to validate the methods once the residues have been identified, limits set and method(s) chosen. Worldwide regulatory agencies require that these methods be validated [6]. In order to be validated the method(s) must be shown to have linearity, precision, accuracy and robustness. Ideally, there would be at least two different analysts performing analyses on two different instruments using two

Methods for pharmaceutical cleaning validations

83

different columns (if it is a chromatographic method) on two different days with two different sets of prepared standards and samples. It is important to note that not only does the analytical method need to be validated but the sampling method needs to be validated too. For example, if a swabbing technique is utilized it must be demonstrated that the technique recovers the residues being analyzed. The percent recovery whether 90% or 50% must be established. Figure 3 shows one type of swabbing technique. It really does not matter which technique is utilized as long as it is repeated the same way each time and the entire sampling area is covered. If an analyst swabs in one manner and another analyst swabs in another manner different results will be obtained. It is very important to train whoever is doing the swabbing in the appropriate technique. A diagram of the swabbing technique should be placed in the procedure document. As was mentioned earlier, it is important to perform recovery studies and validate the sampling technique. Both swabs and rinse sampling methods must be validated. A typical procedure would be to spike a coupon with a known amount of residue and remove the residue with either a swab or a simulated rinse procedure. If a swabbing technique was used the residue from the swab should be desorbed in a suitable solvent. If a TOC analysis were being performed the solvent should be low TOC water and if an HPLC method was being preformed the solvent should most likely be the mobile phase. These studies are done below the residue acceptance limits that had been previously established. This is done to ensure that the method can at least recover the actual residue limit amount. What is an acceptable recovery? A recovery of >80% is very good. A recovery of >50% may be sufficient but a recovery of <50% is questionable. Additional work should be done to improve the <50% recovery. Factors that could cause low recoveries are adhesion of the residue to the swab or the volatility of the residue.

Figure 3. Example of one method for swabbing a surface to recover residues for subsequent analysis.

84

H.J. Kaiser

Applying a known amount of the test residue directly onto a swab can be used to perform swab recoveries. The swab is then extracted and the extract analyzed using the appropriate method. The extraction solvent spiked with the same amount should be used as the control. If the swab is retaining the residue, an alternate swab material should be sought. The recovery factor obtained should either be included in the analytical calculations or the acceptance limit calculations but not in both. If the recovery factor is included in both, too tight a limit will have been set. 7. SUMMARY

In summary, when choosing an analytical technique to determine residues, residue limits should be established first. This limit would then be used to calculate the limit in the analyzed sample. The next step would be to select, develop and validate a method to determine residue levels at and below the acceptance limit. Interferences should also be investigated and addressed. Developing and validating a sampling procedure should be the next step. The validation of the sampling procedure should include recovery studies. There are many different techniques that can be used to analyze residues. In practice, the simplest technique that can quantitate at the residue limit and below should be utilized. Less complicated techniques normally have less potential problems associated with them. REFERENCES 1. “Points to Consider for Cleaning Validation”, Technical Report No. 29, Parenteral Drug Association (1998). 2. H.J. Kaiser and M. Minowitz, J. Validation Technol., 7(3), 226-236 (2001). 3. H.J. Kaiser, J.F. Tirey and D.A. LeBlanc, J. Validation Technol., 6(1), 424-436 (1999). 4. M.K. Chawla, Precision Cleaning, 8(6), 36, 38 (2000). 5. M. Meltzer, C. Koester and C. Steffani, “Criteria Evaluation for Cleanliness Phase 0”, Lawrence Livermore National Laboratory, UCRL-CR-133199 (1999). 6. USP 23, United States Pharmacopeial Convention, Rockville, MD, 1982-1984 (1995).

Surface Contamination and Cleaning, Vol. 1, pp. 85–107 Ed. K.L. Mittal © VSP 2003

Influence of cleaning on the surface of model glasses and their sensitivity to organic contamination W. BIRCH,∗ S. MECHKEN and A. CARRÉ Corning SA, Fontainebleau Research Center, 7 bis, Avenue de Valvins, 77210 Avon, France

Abstract—Both wet and dry cleaning processes impact the chemical composition of glass surfaces. The alteration of glass surface chemistry is probed by wettability measurements as a function of pH. Wettability measurements are performed using the two-liquid method. This method consists of depositing sessile water drops on a substrate immersed in liquid octane. The measured contact angle allows an estimation of the number of hydroxyl groups exposed at the glass/water interface. Measurements with varying pH of the water drops probe the surface charge at the glass surface and indicate the isoelectric points of predominant functional groups. Variations in glass surface composition correlate with sensitivity to organic contaminant adsorption. Three typical species of glass were probed: a silica surface, the surface of sodalime glass, and the surface of a Corning aluminoborosilicate glass. The three glass species show different contamination behavior, depending on wet or dry cleaning of the glass. The surface compositions, as probed by wettability measurements, correlate to grazing incidence XPS data, indicating the alteration of surface chemical sites by leaching during the wet cleaning process. It appears that sensitivity to organic contamination is controlled by the surface chemistry and the hydration of the exposed substrate. Keywords: Glass; cleaning; contamination; surface composition; wettability; surface charge; hydroxyl group density; self-assembled monolayers; isoelectric point.

1. INTRODUCTION

The cleaning of glass surfaces [1] precedes a variety of coating processes, ranging from silane-based sol-gel coatings, to adhesion promoting films, to conductive coatings such as ITO (indium tin oxide). These cleaning processes are generally designed to remove particles from the glass surface and to render it uniformly wetting. For most applications, it is desirable that this wettability be associated with the exposure of silanol groups at the glass surface. These exposed chemical functions form part of the glass substrate and they may be used to covalently graft silane-based coatings to the glass substrate [2].

∗

To whom all correspondence should be addressed. Phone: +33-1-64 69 74 14, Fax: +33-1-64 69 74 54, E-mail: [email protected]

86

W. Birch et al.

Rendering a glass surface hydrophilic may be achieved by a variety of techniques [1]. These may be broadly classed into two categories: solution cleaning and dry cleaning. Solution cleaning is generally easy to set up for a small number of samples. It often has the disadvantage of using toxic products and generating hazardous waste, while being effective for removing particles from the glass surface [1]. Dry cleaning requires a larger initial investment. It is efficient for cleaning large numbers of samples while generating a minor amount of waste products. An advantage of dry cleaning techniques is that they generate a clean, dry surface. This should be compared to the blow-drying that follows solution cleaning, where streaks and surface inhomogeneities are almost unavoidable. This article will focus on the impact on glass surfaces of both solution cleaning and dry cleaning. Glass surfaces are not inert and their exposure to aqueous solutions generally causes a modification of their surface chemical composition. Dry cleaning processes have little or no impact on the chemical composition of the glass surface. Solution cleaning uses chromic acid, while dry cleaning uses pyrolysis. Three types of glass were analyzed. They were chosen to span a representative range of silica-based glasses, from sodalime glass to aluminoborosilicate glass, to silica. Sodalime glass is the most common and cheaply available glass substrate. It is the material used for almost all microscope slides, as well as being the standard “window glass”. Corning code 1737 aluminoborosilicate glass provides a substrate with increased chemical durability, improved optical properties, and higher thermal resistance. It is primarily used in flat panel display applications and has been used to make microscope slide substrates. Silica provides the reference substrate, bearing a surface composed of silicon oxide chemical sites. These three glass species are of increasing cost and are commercialized for different applications. Sodalime glass is a low cost substrate with a low melting point and minimal chemical durability in aqueous solution. Aluminoborosilicate glass is a more expensive glass, providing an increased thermal resistance and improved chemical durability. Its high melting point and the absence of sodium in its composition make it suitable for use in making flat panel displays. Finally, silica provides a substrate with good chemical resistance and a very high melting point. Its high cost limits its use to critical applications, where it is uniquely suited to meet the required specifications. The surface properties of glass substrates were directly probed by wettability measurements. Since clean glass surfaces are completely wetted by water, water drops deposited in air spread to give a contact angle of less than five degrees, which is below the measuring threshold of our contact angle goniometer. Thus, the glass surface wettability was measured under octane, where sessile water drops give a finite contact angle value. The wettability as a function of pH reveals features arising from the chemical functions exposed at the glass surface.

Influence of cleaning on the surface of model glasses

87

2. EXPERIMENTAL

2.1. Substrates and cleaning processes The glass materials examined were: sodalime (SDL) glass, in the form of microscope slides from Erie Scientific (Portsmouth, NH, USA); aluminoborosilicate (ABS) glass, in the form of Corning code 1737F glass (Corning, NY, USA); and silica surfaces, in the form of the native oxide layer of a silicon wafer. Silicon wafers were supplied by Unisil Corporation (Santa Clara, CA, USA) and were pdoped or n-doped four-inch wafers from standard production. All the glass surfaces were believed to have a surface roughness of 3 Å rms or less. This has been measured for silicon wafers [3, 4] and for float glass. The sodalime and aluminoborosilicate glasses bear their as-formed (“fire-polished”) surfaces, with a low surface roughness. Unfrosted microscope slides were cut from sodalime glass, made by a draw process that generates two identical surfaces. The sodalime glass made with a draw process contains no iron. This is visible by looking at the glass slide from the side, where it appears white. A lower cost process for manufacturing these slides is by floating the glass over a bath of molten tin, whereby the glass is referred to as “float glass”. The float glass has two faces: one in contact with a reducing atmosphere, referred to as the “air” side, and one in contact with the molten tin, referred to as the “tin” side, where a small amount of tin diffuses into the glass surface. This is different from the glass drawing process that generates identical glass faces. The presence of trace amounts of iron in the sodalime glass composition gives it a mild green color. This can be seen when observing a microscope slide from the side. In one experiment, frosted glass slides, made from float glass, were also used. The frosted side of these slides corresponds to the air side of the float glass. Primary chemical compositions of the three glass species used are given in Table 1. The chromic acid cleaning solution (CHR) is also known as Chromerge. It consists of a strong oxidizing agent in concentrated sulfuric acid. The solution was prepared by completely dissolving 20 grams of potassium dichromate (K2Cr2O7, from Prolabo, Fontenay sous Bois, France) in 90 grams of water. To this, 900 grams of concentrated sulfuric acid (Normapur grade, from Prolabo) was slowly added. It was then allowed to cool to room temperature. This solution could be used for cleaning while it remained yellow or brown. It was discarded when the color changed to green, indicating a change in the oxidation state of the chromium. As the concentrated sulfuric acid is hygroscopic, so the solution was stored in a closed container, avoiding dilution by ambient moisture. Dipping in chromic acid solution at room temperature for 20 minutes cleaned the glass samples. They were then rinsed in pure water and dipped for 20 minutes in a 1:1 concentrated hydrochloric acid/water solution (designed to remove chromium ions from the surface). They were then finally rinsed in pure water and dried under a stream of pure nitrogen.

88

W. Birch et al.

Table 1. Composition of glass substrates. Principal components of the glass composition are given in weight percentage. For the silica substrate, this is the composition of the native oxide layer at the surface. The bulk glass composition is given for the other two glasses. The first column recalls the process used to make the glass Substrate

Glass composition (by weight)

Silica substrate: – 1-2 nm thick silica layer – polished silicon wafer with native oxide

100% silica

Corning code 1737F glass (ABS) – aluminoborosilicate – made by fusion draw process

70% SiO2 10% Al2O3 10% B2O3

Sodalime glass (SDL) – microscope slides – made by draw process

70% SiO2 13% Na2O 10% CaO

The water used in these experiments was purified using an Elga UHP unit. It filters deionized (18 MΩ.cm resistivity) water through an activated carbon filter and a UV lamp. This water was used to rinse samples, as well as in the preparation of acidic or basic solutions to measure surface wettability. The dry cleaning process used was pyrolysis (PYR). The glass substrates were heated to 500°C, according to the following cycle: a rise in temperature from room temperature to 500°C over four hours, followed by five hours at 500°C and a return to room temperature over five hours. The samples were placed in a glass rack and the rack was enclosed in a glass container, closed with a folded aluminum foil sheet. The containers were removed from the oven after cooling to room temperature. Provided the lid was not opened, the glass samples remained clean for about one week. The samples were used within 5 minutes after removing the aluminum foil lid. The glass samples showed no visible organic contamination on their surfaces before cleaning by CHR or PYR. Neither technique is suitable for cleaning a glass surface bearing macroscopic contamination. At most, the samples had ambient contamination, in the form of a monolayer or sub-monolayer coverage of organic contaminants, expected to have a thickness of order 1 nm or less. Following cleaning, one glass sample per cleaning batch was verified as being wettable by water. To achieve this, a one-microliter drop of pure water was placed on the sample. This drop was expected to spread. When the sample was tilted, its receding edge was not expected to dewet the glass surface. The glass substrates were used if the sample passed these criteria of glass surface wettability. The sample tested was not used for further experiments.

Influence of cleaning on the surface of model glasses

89

For surface composition analyses by XPS and monolayer adsorption and deposition experiments, PYR was replaced by UV/ozone dry-cleaning (details are presented in Sections 2.3 and 2.4). Since the cleaning of silicon wafers with pyrolysis led to rapid subsequent contamination on exposure to air and the XPS equipment was far from the site where the samples were cleaned, UV/ozone cleaning was used to improve sample cleanliness during transport. The two dry cleaning techniques were expected to have a similar impact on the glass surfaces composition, which is measured by XPS to a depth of about 5 nm. 2.2. Contact angle measurements, SiOH group density and dissociation The wettability of the cleaned glass surfaces was measured under liquid octane. Octane was chosen due to its dispersive interactions with glass being almost equal to those between glass and water. This results in the measured wettability giving information on the dispersive interactions between glass and water, as described below. The clean glass substrate was immersed in liquid octane and its wettability measured with a water drop deposited on the glass surface under octane. The water drop displaces the liquid octane that was in contact with the clean glass. When the drop ceases to spread, the measured contact angle, θ, may be interpreted as shown in Figure 1. The force balance between the solid and liquid 1 (glass/octane), liquid 2 and liquid 1 (water/octane), and solid and liquid 2 (glass/water) interfacial tensions can be expressed in the form of Young’s equation:

γ glass/octane = γ water/octane cosθ + γ glass/water .

(1)

This equation can be reformulated as follows to give an expression for cosθ:

cosθ =

γ glass/octane − γ glass/water . γ water/octane

(2)

Figure 1. Schematic diagram of a sessile water drop under liquid octane. The interfacial tensions used in equation (1) are indicated and the definition of the contact angle, θ, is illustrated.

W. Birch et al.

90

Octane is a non-polar liquid. Its interaction with the glass substrate is purely dispersive giving rise to the octane/water interfacial tension. Water is a polar liquid. Its interactions with the glass substrate may be broken down into two components: dispersive and non-dispersive. The dispersive (London) interactions are those generated by instantaneous (non-permanent) dipole interactions. The nondispersive interactions include the permanent dipole interactions (Keesom and Debye interactions), hydrogen bonding and acid-base interactions. The latter two generate a surface charge on the glass substrate. The dissociation of silanol (SiOH) groups at the glass surface may be represented by the following reaction: SiO – +H 3O + ↔ SiOH + H 2O.

The surface charge generated by these acid-base interactions increases the glass/water interfacial energy. The dispersive component of the surface tension of water is almost equal to the surface tension of octane: γWD = 21.8 mJ/m2 [5] and γO = 21.6 mJ/m2 [6], where the subscript “w” stands for water and the subscript “o” stands for octane. Hence, the interaction energy between glass and octane may be considered as being equal to the dispersive interaction between glass and water: D I glass/octane = I glass/wate r

(3)

The non-dispersive interactions for the octane/glass and octane/water interfaces are considered as negligible. Using

γ glass/octane = γ glass + γ octane − I glass/octane ,

(4)

D ND γ glass/water = γ glass + γ water − I glass/wate r − I glass/water

(5)

and

in equation (1) gives ND I glass/wate r = γ water - γ octane + γ octane/water cosθ

(6)

which directly relates the non-dispersive interaction of the glass/water interface to the measured contact angle and known constants (γwater = 72.8 mJ/m2). The interfacial tension between water and octane is given by equation (7) below [7]. It can be measured using a standard procedure, such as a Wilhelmy plate set-up.

γ water/octane = γ water + γ octane − 2( γ Dwater γ octane )1 / 2.

(7)

Finally, equation (6) may be written as: ND I glass/water ≈ (1+cos θ)*51 mJ/m².

(8)

Influence of cleaning on the surface of model glasses

91

The non-dispersive (dipolar, hydrogen bond, and acid-base) interactions between a clean glass surface and water can give a simple estimate of the hydroxyl group density at the glass surface, providing a measure of the density of SiOH reactive sites at the glass surface. If we assume the non-dispersive interaction between glass and water to be dominated by hydrogen bonding and acid-base interactions, we may consider that at the point of zero charge of the glass substrate the non-dispersive interaction energy to be dominated by hydrogen bonding. At this point, the surface energy per unit area can be divided by the energy per hydrogen bond to obtain the number of hydrogen bonds per unit area. Taking the -OH bond energy between water and the glass surface as 24 kJ/mol [8] gives the following expression for the -OH group density, expressed as the number of hydroxyl bonds per square nanometer: -OH/nm² ≈ 1.3 (1+cos θ)

(9)

where θ is the contact angle of sessile water drop under octane at the point of zero charge. The above equations relating wettability measurements to the nondispersive interaction between the glass surface and water are valid if organic contamination is absent from the glass surface. The wettability measurements of water drops under liquid octane were performed in a cube-shaped glass cell (from Hellma, in Müllheim, Germany). This glass cell was first cleaned with a 2% solution of Hellmanex detergent (Hellma) in pure water for 20 minutes. It was then rinsed with pure water and blow-dried under a flow of pure nitrogen. During rinsing, the cell was verified to be uniformly wetting. Immediately after cleaning, octane liquid was poured into the cell to a depth of about two centimeters. The cleaned glass substrate was then immersed in octane. Following a waiting period, as given below, three twomicroliter sessile water drops were deposited on the sample. The contact angles on both sides of the drops were measured using a Ramé-Hart contact angle goniometer (Mountain Lakes, NJ, USA). These contact angle values were then averaged. The first measurements were made using unpurified octane. The immersion of the substrate in octane was noted and the sessile water drop contact angles were plotted as a function of the elapsed time between substrate immersion in octane and deposition of the water drop. The second set of experiments was performed using purified octane, cleaned by passing over silica and alumina chromatographic columns. Octane (purissimum grade from Fluka, Ref. 74821) was passed through a chromatography column filled with silica (silica gel, 60, Ref. 1.09385.1000, from Merck, Darmstadt, Germany), followed by passage through a chromatography column filled with alumina (basic alumina, 60, activity stage I, Ref. 1067.2000, from Merck). The silica and alumina were used as received. The chromatography columns were cleaned with Hellmanex surfactant, as described above for the glass cell. They were rinsed with pure water and dried in a clean oven with aluminum foil covering their ends. The first few milliliters of octane

92

W. Birch et al.

passed through the columns were used to rinse the original octane bottle and then discarded. The purified octane was then stored in this rinsed bottle. The reasoning behind this purification of the liquid octane is that the high active surface area of the silica and alumina particles will adsorb polar impurities in the octane. Given the similarity of the surface sites on silica and alumina particles to those found on clean glass surfaces, we expect the impurities adsorbed in the columns to be those that would have adsorbed onto a clean glass surface. To vary the pH of the water, aqueous solutions of hydrochloric acid or sodium hydroxide were used. These solutions were made using calculated volumes and verified with a pH sensitive paper. In the preliminary experiments, cleaned and dried glass substrates were immersed into octane and the wettability of water drops at different pH values was measured. These experiments showed trends in the wettability as a function of pH. To reduce the error bars on the measured contact angle values, the cleaned substrates were first immersed for twenty minutes in an aqueous solution of the same pH as that of the water drops to be measured. This was found to give the same trends in wettability as a function of pH, while decreasing the error bar on the measured contact angle values. The increased accuracy of the measurement may be due to the fact that the deposited water drop has insufficient time for it to interact with the cleaned glass surface before settling at the measured contact angle. Further, contact angle hysteresis may prevent the contact angle value from changing if the surface charge of the cleaned glass changes with its exposure the water drop. The wettability of the cleaned glass substrates was measured after their immersion in the same acidic or basic solution as used in the measuring water drops. After cleaning, each glass substrate was cut into five pieces and each piece placed in a solution of different pH. Following the 20-30 minute soaking time, the glass substrates were rinsed with water and blow-dried with pure nitrogen. Sessile drop contact angles were measured for water drops deposited five minutes after immersion. The deposited water drops came from the same solution as the one in which the substrates were immersed. For each glass species, three samples were measured at each pH and three sessile water drops were measured for each sample. The data points used for the graphs in Figures 4 and 5 represent an average of these 18 contact angle values. 2.3. Surface composition The glass surfaces were examined by normal angle x-ray photoelectron spectroscopy (XPS). This technique provides quantitative information on the atomic composition of the first 5 nm of the glass surface. The measured percentage of inorganic elements in this region provides information on changes in the glass surface composition following CHR or UV/ozone cleaning. UV/ozone cleaning was used to replace pyrolysis due to rapid adsorption of organic contaminants on the pyrolyzed silicon wafers. Placing the samples 1 cm away from the surface of a 4” x 4” low-pressure mercury lamp (BHK Inc., Claremont, CA, USA) for 30 minutes

Influence of cleaning on the surface of model glasses

93

achieved the UV/ozone cleaning. The UV lamp emits radiation at the 184.9 and 253.7 nm wavelengths, where the first generates ozone from oxygen while the second combines with ozone to oxidize hydrocarbons [10, 11]. 2.4. Monolayer adsorption and deposition Practical applications for bare glass surfaces may utilize their surface charge or surface silanol sites to adsorb or graft a layer of molecules for a thin film coating. Charge-adsorbed surfactant monolayers were deposited to probe the glass surface charge. The deposition of a grafted self-assembled monolayer of octadecylsilane was used to probe the quality of coatings grafted to the silanol groups on the glass surface. Monolayers of anionic and cationic surfactants were deposited on the cleaned glass substrates from aqueous solutions. For the anionic surfactant, sodium dodecyl sulfate (SDS, purissimum grade, from Fluka Chemie AG, Buchs, Switzerland) was used. The deposition was made by pulling the clean glass out of a solution at one half of the critical micellar concentration (cmc) for SDS, corresponding to a 4 mM solution of SDS. No deposition occurs on a negatively charged surface in solution. A monolayer of surfactant is deposited onto the glass surface from the thinning film of SDS solution as the sample is slowly pulled vertically from the surfactant solution [3]. The cationic surfactant used was hexadecyltrimethylammonium bromide (CTAB, purum grade, from Fluka). Its deposition on the negatively charged surface occurred in solution, resulting in a substrate that was pulled out “dry” (with no liquid film on its surface) from solution. The solution concentration was 0.4 cmc, corresponding to 0.4 mM of CTAB. The glass substrate was left in the surfactant solutions for at least thirty seconds and pulled out vertically at a speed of about one millimeter per second. The quality of coatings grafted to the glass via their surface silanol groups was tested by depositing a self-assembled monolayer of octadecyltriethoxysilane (OTES). For this experiment, the glass slides were cleaned using the CHR and UV/ozone cleaning processes. UV/ozone cleaning replaced the PYR dry cleaning to avoid rapid contamination of the silicon wafers following PYR cleaning. The OTES molecules were hydrolyzed before deposition for two purposes. By converting the ethoxy groups surrounding the silicon atom to hydroxyl groups, hydrolysis reduces the size of the silicon functional group at the end of the aliphatic hydrocarbon chain. This allows a dense packing of the aliphatic hydrocarbon chains into a structure perpendicular to the substrate, somewhat reminiscent of the hair on a carpet. The second purpose is an increased reactivity of the SiOH groups, as compared with the unhydrolyzed Si-ethoxy groups. This reactivity allows for cross-linking of the silanol groups and grafting of the aliphatic molecules to the silanol surface sites on the glass, resulting in a durable grafted coating on the glass surface. To form a densely-packed structure, the grafting reaction should take place after the self-assembly of the aliphatic hydrocarbon chains [12]. This results in a high quality monolayer structure with only a few defects, resulting in

94

W. Birch et al.

an optimal non-wetting behavior towards water. The OTES was pre-hydrolyzed using the recipe given in the paper by Peanasky et al. [13]. A 50-ml volumetric flask was filled three quarters with inhibitor-free THF (Ref. 24,288-8, from Aldrich). To this were added 0.25 g of a 1.31 N aqueous solution of hydrochloric acid. The hydrochloric acid solution was made by dissolving 13.27 g of 36% (specific gravity of 1.19) hydrochloric acid (ref. A 7405201, from Fisher Scientific, Chicago, IL, USA) in water, making 100 ml of aqueous solution. The mixture was stirred and 0.42 grams of OTES were slowly added (ref. SIO6642.0, from ABCR, Karlsruhe, Germany). The flask was then filled up to the 50-ml mark with THF. This solution was allowed to sit at ambient temperature for four to five hours before use. It could be refrigerated for up to one week at 4°C. To make the OTES coating solution, 18.6 grams of cyclohexane (HPLC grade, Ref. 27,062-8, from Aldrich, purified in the same manner as the octane above) were poured into a pyrolyzed glass container. While stirring, 1.11 grams of prehydrolyzed OTES solution were slowly added. Cleaned glass slides were placed in a pyrolyzed glass rack and the rack placed in the glass container for coating. The container was then closed with a sheet of folded aluminum foil taped down to its sides. The glass slides were left to incubate for 24 hours at room temperature. They were then rinsed three times with stabilized THF (Ref. 87368, from Fluka) in a clean glass container. Ultrasonic agitation was applied to each rinse by placing the glass container for five minutes in a Crest ultrasonic bath (68 kHz with modulated frequency, from Crest Ultrasonics, Trenton, NJ, USA). For optimal transmission of ultrasonic vibration energy, the ultrasonic bath contained a 3% aqueous solution of Chemcrest 14 detergent (from Crest Ultrasonics). The detergent solution was degassed and heated to 45°C to improve the uniformity and transmission efficiency of ultrasonic vibration energy in the bath. The wettability to water of three samples of each glass species and cleaning procedure was then measured. Each sample was measured using three sessile water drops and one water drop for the advancing and receding contact angles. The contact angle measurements were made as described above, using the Ramé-Hart goniometer. The sessile drop values quoted were averaged over 18 values, corresponding to both sides of three drops on each of three samples. The advancing and receding contact angle values quoted were averaged over three values, corresponding to one drop on each sample. 3. RESULTS AND DISCUSSION

3.1. Contact angle measurements: surface contamination of cleaned glass in octane In the first experiments, the contact angles of sessile water drops deposited under unpurified octane were measured. Figure 2 shows the evolution of the contact angle of the deposited sessile water drop as a function of the elapsed time between immersion of the cleaned substrate in octane and the water drop deposition. The

Influence of cleaning on the surface of model glasses

95

Figure 2. Water sessile drop contact angles measured after exposure of cleaned glass substrates to unpurified octane. The legend is the same as for Table 2. The quoted contact angles are averages of six values for three sessile drops at times of 1, 3, and 5 minutes. The contact angle values quoted for “silica PYR” at less than one minute are for one drop at each data point.

contact angle of the sessile water drops did not change after their deposition. The observed increase in contact angle suggests an evolution of the cleaned surface with time over the first five minutes after drop deposition. The increasing contact angle values also suggest a contamination of the substrate. Other possibilities, such as a hydration or dehydration of the cleaned surface were considered. Hydration of the cleaned surface is expected to lead to a decrease in the measured contact angle value. To check for dehydration, a drop of water was placed in the octane and the cleaned substrates immersed after 48 hours. The contact angle evolution was identical to that in Figure 2, suggesting that dehydration of the glass surface in octane is not what causes the observed increase in contact angle with immersion time. To verify contamination by exposure to octane, the substrates were blow-dried with pure nitrogen after immersion for at least five minutes in the liquid octane. The wettability of sessile water drops was then measured in air. The measured contact angles are given in Table 2. They range from 32 to 46 degrees, indicating a significant level of organic contamination adsorbed on the glass surfaces. Further, the large error bar of about 8 degrees indicates a nonuniform surface wettability. This behavior is typically found for random contamination deposited on a clean surface. The contact angle values measured after exposure to octane, when compared with the complete spreading of water drops on the freshly cleaned glass substrates, suggest that the liquid octane contaminates the cleaned glass substrates. The sessile drops deposited following longer immersion time of the cleaned substrate in octane showed a higher contact angle than those on a freshly immersed substrate. In fact, a water drop deposited immediately after immersion of the cleaned glass into octane spread to form a low contact angle, generally of 5°

W. Birch et al.

96

Table 2. Water sessile drop contact angles (in degrees) measured on glass substrates exposed to unpurified octane for more than five minutes. The substrates were removed from the octane and blow-dried with pure nitrogen before wettability measurements. The error bar over several sessile drops and several substrates is nominally 8°. The superscript number indicates the number of substrates over which the quoted wettability was measured. Each substrate was measured using between three and eight sessile drops. In the legend, silica refers to the native oxide layer on a silicon wafer, ABS refers to Corning code 1737 aluminoborosilicate glass, and SDL refers to sodalime glass. CHR and PYR refer to chromic acid and pyrolysis cleaning, respectively Cleaning process

silica

ABS

CHR PYR

46 1 46 5

43 35

1 7

SDL 40 32

1 2

or less. The drops, once deposited, did not change their shape, indicating that the contact angles of the sessile drops did not evolve. Young’ s equation invokes the notion of a force equilibrium at the three-phase line, where the two fluid phases, namely octane and water, meet the solid surface. The absence of any increase in the contact angle of the deposited drops, when compared to the increase in contact angle for freshly deposited drops, suggests that an equilibrium contact angle value is not reached for the sessile water drops under octane. The notion of a force balance at the three-phase line is valid. A force balance can be considered as a minimum in the energy of a system with respect to a small displacement of the system from this minimum. This concept is valid for systems that are free to move, where the influence of frictional forces is negligible. For contact angles, the concept of a contact angle hysteresis is analogous to a frictional force, preventing the contact angle from reaching a value corresponding to the overall minimum in the free energy of the system. This is the case for water drops deposited under octane, where the observed contact angle is not reached by the drop retracting to form a contact angle equal to that of sessile drops deposited at later times. Following a five minute exposure to octane, the contact angles of three or more freshly deposited sessile water drops were measured on each of the glass substrates. At least three substrates were investigated for each glass species and cleaning procedure. Table 3 gives the averaged contact angle values for each type of glass using CHR or PYR cleaning. For the silica surface, we find a slightly higher contact angle following PYR cleaning than for the CHR cleaned substrate. This is compatible with a loss of about 10% of the surface hydroxyl groups during glass pyrolysis [14]. However, for the sodalime glass, the wettability of the surface following PYR cleaning remains high, even when immersed in the liquid octane that has contaminated the other glass substrates. To obtain more information on this behavior, the advancing and receding contact angles were measured for the substrates exposed to liquid octane for more than five minutes. The measured values are shown in Table 4. If these contact angle values are interpreted in terms of surface contamination of the cleaned glass substrates by immersion in the liq-

Influence of cleaning on the surface of model glasses

97

Table 3. Water sessile drop contact angles (in degrees) measured on cleaned glass substrates under octane, following a five minute exposure to the octane. The error bar over several sessile drops and several substrates is nominally 4°. The legend is the same as for Table 2 Cleaning process

silica

ABS

SDL

CHR PYR

29 59

44 29

26 3

Table 4. Advancing and receding contact angles (in degrees) of water drops under unpurified octane. The substrates were immersed in the unpurified octane for at least five minutes before measuring the contact angle hysteresis of one water drop per substrate. The superscript number indicates the number of substrates over which the quoted wettability was measured. The legend is the same as for Table 2 Cleaning process CHR PYR

silica 5

ABS 1

SDL 1

θa

38

θr

10

θa

68

37

spreading

θr

45

0

0

9

50 15

13

35 13

3

uid octane, they suggest that the pyrolysed silica is more susceptible to adsorbing contamination than the chromic acid cleaned silica. Further, the pyrolysed sodalime glass does not strongly adsorb organic contaminants from the octane, given that the water drop spreads easily. The pyrolysed aluminoborosilicate glass shows a behavior intermediate between the silica surface and the sodalime glass, showing a finite advancing contact angle and zero receding contact angle. The absence of a finite receding contact angle value suggests that few or no organic contaminants remain adsorbed on the glass surface after passage of the water drop. On the other hand, following chromic acid cleaning, the three glass species show similar wettability, suggesting that they have almost the same affinities for adsorbing organic contaminants in liquid octane. This suggests that the sensitivity to adsorbing contaminants depends on the composition of the glass surface and the procedure used to clean the glass surface. Considering that PYR cleaned silica surface is more dehydrated than CHR cleaned silica surface, we may attribute the reduction in the affinity for adsorbing organic contaminants to the presence of a layer of water on the glass. PYR cleaned sodalime glass shows the lowest level of adsorbed contaminants. Both the soluble alkaline oxide film and the ambient moisture it adsorbs may enhance the durability of this clean glass surface. It is interesting to note that the surface behavior of the glass species is similar to that of the silica following cleaning with chromic acid.

98

W. Birch et al.

Figure 3. Schematic side-view of a drop on an inclined surface, indicating the advancing contact angle, θa, and the receding contact angle, θr.

3.2. Use of purified octane to avoid surface contamination To avoid the surface contamination described in the preceding section, subsequent measurements of water drops under octane were made in purified liquid octane, prepared as described in Section 2.2. Following exposure to the purified octane, substrates dried under a stream of pure nitrogen gave low contact angles for deposited water drops, indicating a low level of contamination from strongly adsorbing organic molecules. This suggests that the purified octane did not significantly contaminate the cleaned glass surfaces, unlike the behavior cited for the unpurified octane in Section 3.1. The contact angles for sessile water drops measured under purified octane showed a different behavior from those measured under unpurified octane. The contact angles of water under octane for silicon wafers cleaned with chromic acid were in the range 8° to 16° and showed time-dependent increases of only 2 to 6°. This result was obtained again when a freshly cleaned substrate was immersed in purified octane that had been left in the measuring cell for two days, indicating no significant increase of contaminants in the octane over two days. For Corning code 1737F glass, the contact angles of deposited water drops showed a different behavior: the water drops gave a finite contact angle immediately after deposition, later spreading to wet the glass surface within a few minutes. This behavior remains unexplained. For CHR cleaned sodalime glass, the contact angle of freshly deposited water drop increased from about 6° to 14° over the first ten minutes following immersion of the substrate. The contact angles are for freshly deposited water drops. There is no increase in the contact angle after drop deposition. For the pyrolysis-cleaned substrates, water completely wets the substrates immersed in octane, regardless of their immersion time. The spreading of water drops under octane on pyrolyzed silicon wafers was immediate. The final contact angle was below the measurement capability of the Ramé-Hart goniometer (less than about 4°). For the Corning code 1737F glass, the water drop spreads, leaving only a slightly thicker film. The contact angle is still too small to be measured. Drops of water deposited on sodalime glass under octane spread rapidly, forming dendrimers. A simple classification would rate the affinity to water of the PYR cleaned glass surfaces in the order: sodalime glass > Corning code 1737F glass > silica.

Influence of cleaning on the surface of model glasses

99

The kinetics of water drop spreading described for water drops on glass under purified octane remains unexplained. The increase in contact angle observed for glass surfaces cleaned with chromic acid is not attributable to contamination, given the high wettability of the cleaned glass following exposure to the octane. The unusual spreading behavior of water drops on acid cleaned Corning code 1737F glass also remains unexplained. Finally, the complete wetting of pyrolyzed glass substrates by water drops under octane suggests a high affinity of these glass surfaces for water. Further, the presence of some components of the glass composition that are soluble in water, such as sodium oxide and calcium oxide (see Table 1), appears to enhance the spreading of the water drops under octane. 3.3. Surface wettability as a function of pH, surface charge and SiOH group density While the contact angle values for sessile water drops at neutral pH provide a first information on the surface energy of the cleaned glass substrates, further information can be provided by measuring the surface wettability as a function of the pH. These experiments were conducted using purified octane to avoid the substrate contamination described in Section 3.1. The absence of contamination of the cleaned glass substrates and the behavior under octane of sessile water drops at neutral pH is described in Section 3.2. The pH of water drops deposited on the cleaned glass surfaces was varied in the range from 0 to 13, in order to probe the surface charge variations from acidic and basic surface groups. The surface wettability was measured at pH intervals of 1 unit for acidic pH and two units for alkaline pH values. The standard deviation of the measurements was generally 1° to 2°. About 10% of the data points showed standard deviations ranging from 3° to 5°. Figure 4 shows the non-dispersive interaction energy of glass substrates cleaned with chromic acid as a function of pH, in the range from 0 to 13. The interval of one pH unit for acidic solutions and two units for alkaline solutions was chosen arbitrarily. The contact angle values, ranging from 0 to 21°, have been converted into the interaction energy, INDglass/water, using equation (8). Figure 5 shows data from the same measurements on glass surfaces cleaned by pyrolysis. The two graphs in Figure 5 show the same data. The upper graph was plotted to show differences in the behavior of Corning code 1737 glass and sodalime glass, while the lower graph shows data for all three glasses, with silica showing a significantly lower non-dispersive interaction energy with water than the other two glass species. The hydroxyl (SiOH) surface groups on the clean glass may be positively or negatively charged, depending on the solution pH in contact with the surface. The isoelectric point, also known as the point of zero charge, or p.z.c., defines the pH at which the surface densities of positive and negative charges are equal, associated with an equal surface density of negatively charged SiOH groups (SiO–) and positively charged SiOH (SiOH2+) [9]. The isoelectric point of silica has been

100

W. Birch et al.

Figure 4. Non-dispersive interaction energy between sessile water drops and Chromerge cleaned glass substrates as a function of pH. The water drops were at the pH indicated and each point represents an average of 18 contact angle measurements. The standard deviation of the data points is typically 1-2°, with about 10% of the data points giving a standard deviation of 3-5°.

measured to be in the range of pH 2 to 3 [9]. Figure 4 shows a minimum in the interaction energy near the quoted isoelectric point of silica. While the quoted value for the isoelectric point of silica is close to pH 2, the minimum in Figure 4 is close to pH 3. At the p.z.c., we may consider the glass surface as being uncharged. If we assume the non-dispersive interactions between glass and water to be dominated by hydrogen bonding, the contact angles measured at the p.z.c. provide an estimate of the number of hydroxyl groups per unit area on the cleaned glass. Table 5 gives the measured contact angle values at pH 3 for the cleaned glass surfaces. Table 6 shows the corresponding calculated hydroxyl group densities, converted from the contact angles in Table 5 using equation (9). The values in Table 6 compare favorably to Iler’s quoted value of 2-5 -OH/nm2 (between 2 and 5 hydroxyl groups per square nanometer) for silica surfaces [14]. The measured contact angles range from 10° to 39°, while the corresponding hydroxyl group densities vary only by about 10%, indicating similar density of hydroxyl groups on the glass surfaces for all glass species and cleaning procedures. For comparison, sessile water drops measured under octane on polypropylene or polyethylene, which are not expected to show any hydrogen bonding, give contact angles of 171° [15], yielding a calculated -OH density of 0.003 (or less) per nm2.

Influence of cleaning on the surface of model glasses

101

Figure 5. Non-dispersive interaction energy between sessile water drops and pyrolysis cleaned glass substrates as a function of pH. The upper graph shows the data from sodalime and Corning code 1737 glass, comparable to that from Figure 4. The lower graph shows the full data for the silica surface, which shows an extremely different behavior, resulting in far lower interaction energy and higher measured contact angles. Data points are averaged over 18 values with standard deviations as described in the legend for Figure 4.

Table 5. Water sessile drop contact angles (in degrees) measured on cleaned glass substrates under octane. The measured water drops are aqueous hydrochloric acid at pH 3, corresponding to the point of zero charge of the data in Figures 4 and 5. The legend is the same as for Table 2 Cleaning process

silica

ABS

SDL

CHR PYR

15 38.5

21 19.5

16.2 10.2

W. Birch et al.

102

Table 6. Hydroxyl group density (-OH/nm2) for glass surfaces, calculated from the contact angles given in Table 5. These contact angles were measured on the cleaned glass surfaces, for sessile water drops under octane at the point of zero charge. The legend is the same as for Table 2 Cleaning process

silica

ABS

SDL

CHR PYR

2.56 2.32

2.51 2.53

2.55 2.58

In terms of estimated hydroxyl group density on the cleaned glass surfaces, pyrolyzed silica shows a lower hydroxyl group density than chromic acid cleaned silica. This is compatible with a minor amount of dehydroxylation during the pyrolysis cycle [14]. Unexpectedly, pyrolyzed sodalime glass appears to have a higher surface SiOH concentration than silica. This may be an artifact caused by the soluble alkali salts (see Table 1) present on the surface of the pyrolyzed sodalime glass. These salts may increase the wettability to water when compared to the same glass cleaned by chromic acid, where alkaline oxides are leached during the cleaning procedure. Returning to the graphs in Figures 4 and 5, the data show two minima in the non-dispersive interaction energy between glass and water. The first and more prominent minimum, found at pH 3, may be associated with the isoelectric point of the SiOH groups. Recalling the acid-base reaction responsible for generating the surface charge from the SiOH groups: SiO – +H 3O + ↔ SiOH + H 2O,

the hydrogen ion (H3O+) concentration at this pH is sufficient to drive the equation towards the right. At higher pH values, corresponding to a lower hydrogen ion concentration, the glass surface is negatively charged from its surface SiOH groups. At lower pH values, the surface hydroxyl groups become positively charged, according to the following equation: SiOH+H 3O + ↔ SiOH +2 + H 2O.

Thus, only at the isoelectric point does the glass surface bear no overall charge from its SiOH groups. A second, shallower, minimum in the interaction energy may be seen at pH 9. This may correspond to the sodium ions (Na+), from the sodium hydroxide used to increase the pH, being present in sufficient quantity to neutralize the SiO– ions from the surface hydroxyl groups. For all the glass samples, the measured contact angles fall sharply above pH 10 and are zero above pH 12. This is probably due to the glass surface being degraded by dissolving in the alkaline solution, leading to its being fully wettable by water.

Influence of cleaning on the surface of model glasses

103

The non-dispersive interaction energy between glass and water as a function of pH is expected to reflect the surface charge generated by the exposed chemical functions on the clean glass surface. The variations in surface charge, generated by the exposed SiOH and aluminum oxide groups, is expected to give rise to features representing the surface chemistry of the clean glass. The scatter in the data shown in Figures 4 and 5 allows only general trends to be discerned. The p.z.c.’s at pH 3 and 9 have been described in the preceding paragraphs. It is interesting to note that the chromic acid cleaned glass surfaces behave in a similar manner, showing virtually identical trends. The pyrolysis cleaned glass surfaces show differences in their behavior across the different glass compositions. These trends correlate with those observed for organic contamination of these surfaces, as described in Section 3.1, where the chromic acid cleaned glass surfaces all showed similar behavior, while the pyrolyzed glass showed significant differences in its sensitivity to contamination. In particular, the pyrolyzed silica surface shows far lower non-dispersive interaction energy with water than the pyrolyzed Corning code 1737 or sodalime glasses. This features correlates with the high degree of adsorbed contamination, described in Section 3.1, for the pyrolyzed silica surface. The datum in Figure 5 for the non-dispersive interaction energy between a pyrolyzed silica surface and water at pH 7 corresponds to a contact angle of 31°. This is significantly higher than the contact angle of water on a pyrolyzed silica surface freshly immersed into liquid octane. While the surface cleanliness was measured after cleaning, it was not measured after substrate immersion in the acidic or alkaline solutions. It is possible that the comparatively low nondispersive interaction energy observed for pyrolyzed silica is partially an artifact caused by contamination of the cleaned silica before immersion into liquid octane. Figure 4 shows similar behavior for the glass surfaces, suggesting that the aluminoborosilicate and sodalime glasses show behavior similar to that of a silica surface. This phenomenon may be due to the leaching of soluble alkaline oxides from the glass surfaces during chromic acid cleaning, leaving a surface enriched in silica that behaves essentially in the same way as a chromic acid cleaned silica surface. In Figure 5, the minimum in the non-dispersive interaction energy between glass and water at pH 9 is not present for pyrolyzed sodalime glass. This minimum was presumed to be associated with a high sodium ion concentration in solution, neutralizing the SiO– groups at the glass surface. The presence of sodium oxide (see Table 1) in the sodalime glass composition may generate a high sodium environment for the the silanol groups at the glass surface. The high sodium concentration in the glass may thus be equivalent to a high sodium concentration in solution, neutralizing the p.z.c.

W. Birch et al.

104

3.4. Glass surface composition The glass surface composition was measured using XPS following chromic acid and UV/ozone cleaning. UV/ozone cleaning replaced the PYR dry cleaning to avoid the contamination described in Section 3.1 for pyrolyzed silica surfaces. The XPS probed a 5 nm depth from the glass surface. Table 7 shows the percentage atomic composition of the glass surfaces. The silica surface shows a slight increase in the oxidation of its native surface silicon oxide film following UV/ozone cleaning. The aluminoborosilicate shows little change in its surface composition between the chromic acid or pyrolysis cleaning. The sodalime glass shows a loss of sodium following chromic acid cleaning. This reflects a leaching of the sodium oxide from the glass surface layer by exposure to acid. The residual sodium atoms may lie at a sufficient depth in the glass so as not to be removed by contact with the chromic acid. This measurement of glass surface composition indicates leaching of soluble alkaline oxides from the glass surface. However, the changes in the atomic composition of the glass surfaces appear small when compared to the observed differences in glass surface behavior. Table 7. Surface composition of cleaned glass measured by XPS. The penetration depth is estimated at 5 nm. The percentage concentrations of selected elements are given, resulting in a total of less than 100%. The carbon signal ranged from 4-7%, indicating a layer of ambient contamination adsorbed during transfer of the samples from cleaning to the measurement location. For the silicon wafer, only the signal attributed to the native oxide layer is given. Since the native oxide layer is 1-2 nm thick and the beam penetrates 5 nm into the sample, the total of the percentage compositions is much less than 100 Element %

Silica UV/O3

Silica CHR

1737 ABS UV/O3

1737 ABS CHR

SDL UV/O3

SDL CHR

Si O Al Na

13 34 0 0

12 30 0 0

27 59 5 1

29 59 6 0

26 47 1 8

29 59 0 3

3.5. Monolayer adsorption and deposition Probing the practical behavior of the cleaned glass surfaces by charge-adsorbed surfactant monolayers gave results indicating a negatively charged surface on all the cleaned glass substrates. The anionic SDS surfactant was not deposited from solution, resulting in a wettable surface when the substrate was pulled from the surfactant solution. The cationic CTAB surfactant was deposited from solution, resulting in a non-wettable surface as the glass was pulled out of the solution. A negative surface charge is predicted for silicate surfaces in water at neutral pH.

Influence of cleaning on the surface of model glasses

105

The presence of surfactant monolayers coating the glass surfaces was confirmed by measuring the wettability of non-polar tricresyl phosphate sessile drops in air. The contact angles on the SDS monolayers were in the range 61-66 degrees; those on CTAB monolayers were in the range 40-41 degrees. This wettability is compatible with a lower packing density for the CTAB surfactant monolayer, where a larger head group prevents close packing of the aliphatic hydrocarbon chains [3, 4]. Further information on the surface silanol sites can be obtained by depositing a self-assembled OTES monolayer. This monolayer was deposited on glass substrates cleaned by chromic acid or UV/ozone. A monolayer of grafted octadecylsilane molecules generates high contact angles for water, indicating a hydrophobic surface of densely-packed aliphatic hydrocarbon chains. Contact angle hysteresis probes the defect density in the self-assembled monolayers. The receding contact angle is primarily influenced by the presence of hydrophilic defects in the hydrophobic monolayer. Measured wetting properties for water drops are given in Table 8. For the UV/ozone cleaned surfaces, the data show a higher quality hydrophobic monolayer on the silica surface. A densely packed octadecylsilane monolayer gives rise to high contact angles for water. This dense packing is a result of a high degree of ordering of the octadecylsilane molecules and their grafting to each other and to the glass substrate. The aluminoborosilicate glass shows a coating quality intermediate between that on the silica and sodalime glass surface, which shows the lowest coating quality. The chromic acid cleaned glass surfaces show similar coating quality, slightly higher than that seen for the UV/ozone cleaned surfaces. This improved coating quality may be, in part, due to hydration of the chromic acid cleaned glass. The trends in coating quality for all UV/ozone cleaned glass surfaces and all chromic acid cleaned glass surfaces correlate with those seen for contamination of the glass surfaces, as described in Section 3.1. Table 8. Wettability towards water of self-assembled monolayers of OTES deposited on the cleaned glass surfaces. The sessile drop, advancing, and receding contact angles (in degrees) are given, with the sessile drop value above the advancing and receding values. The wettability measurements were made in air. For the sodalime glass (SDL), the slides used were float glass. The wettability of the deposited monolayer coatings was measured separately on the air side and the on the float side of this glass silica UV/O3 cleaned

ABS

106 108

101 97

109

108 CHR cleaned

112

SDL (air) 100 88

107

107 99

112

SDL (float) 99

90

107

107 99

112

87 108

99

112

99

106

W. Birch et al.

4. CONCLUSION

While cleaned silica-based glass surfaces have similar surface compositions, their susceptibility to strongly adsorbing organic contaminants depends strongly on the glass composition and the cleaning procedure. For the three glass species examined: silica, aluminoborosilicate, and sodalime glass, the glass surfaces behave similarly after chromic acid cleaning. They show significant differences in their properties following a dry cleaning procedure, such as pyrolysis or UV/ozone cleaning. The cleaned silica surfaces show a high susceptibility to adsorbing organic contamination following pyrolysis cleaning, while the pyrolyzed sodalime glass appears to be virtually immune to strongly adsorbing organic molecules. Pyrolyzed aluminoborosilicate glass shows an intermediate susceptibility to adsorbing organic contaminants. The chromic acid cleaned glass surfaces all show an intermediate susceptibility to contamination by adsorbed organic molecules. Thus, it may be an oversimplification to consider a clean glass surface as a high energy substrate that is bound to attract ambient organic contamination. The wettability behavior of the cleaned glass surfaces showed features associated with their exposed chemical functions. The non-dispersive interaction energy between glass and water as a function of pH showed evidence of charging of the surface silanol groups. The point of zero charge for these surface chemical functions was observed at pH 3. An estimate of the non-dispersive interaction energy between glass and water at the point of zero charge enables a reasonable estimation of the density of surface silanol groups on the cleaned glass. The trends observed for the surface charge as a function of pH correlate with the observed susceptibility for adsorbing organic contamination to the cleaned glass surfaces. Charge-adsorbed surfactant monolayers indicated a negative surface charge on the cleaned glass, as expected for silica-based glass surfaces at neutral pH. The wettability of grafted self-assembled octadecylsilane monolayers indicated high quality coatings on the cleaned glass surfaces. The coating quality was identical for all three glass species following chromic acid cleaning. The UV/ozone cleaned glass surfaces showed the highest coating quality on the silica surface, followed by the aluminoborosilicate surface and the sodalime glass surface. The trends in coating quality for all chromic acid cleaned surfaces and UV/ozone cleaned surfaces correlate with those seen for susceptibility to organic contamination of the cleaned glass surfaces exposed to unpurified liquid octane. REFERENCES 1. W.R. Birch, in: Sol Gel Handbook, M. Aegerter (Ed.), Kluwer Academic Publishers, New York (to be published) and references therein. 2. K.L. Mittal (Ed.), Silanes and Other Coupling Agents, Vol. 2, VSP, Utrecht (2000). 3. W.R. Birch, S. Garoff, M. Knewtson, R.M. Suter and S. Satija, Colloids Surfaces 89, 145 (1994). 4. W.R. Birch, S. Garoff, M. Knewtson, R.M. Suter and S. Satija, Langmuir 11, 48 (1995).

Influence of cleaning on the surface of model glasses

107

5. R.J. Good and M.K. Chaudhury, in: Fundamentals of Adhesion, L.-H. Lee (Ed.), pp. 137-151. Plenum Press, New York (1991). 6. J.J. Jasper, J. Phys. Chem. Ref. Data 1, 841 (1972). 7. F.M. Fowkes, Ind. Eng. Chem. 56 (12), 40 (1964). 8. J.N. Murrell and A.D. Jenkins, Properties of Liquids and Solutions, p. 168, John Wiley and Sons, Chichester (1994). 9. A. Carré, F. Roger and C. Varinot, J. Colloid Interface Sci. 154, 174 (1992). 10. J.R. Vig, J. Vac. Sci. Technol. A 3, 1027 (1985). 11. J.R. Vig, in: Treatise on Clean Surface Technology, Vol. 1, K.L. Mittal (Ed.), pp. 1-26, Plenum Press, New York (1987). 12. J. Davidovits, PhD Thesis, Détermination des conditions d’obtention de films monomoléculaires organisés : application aux silanes auto-assemblés sur silice, Université Paris 6, Chapter 5 (1998). 13. J. Peanasky, H.M. Schneider and S. Granick, Langmuir 11, 953 (1995). 14. R.K. Iler, The Chemistry of Silica, John Wiley & Sons, New York (1979). 15. A. Carré, S. Moll, J. Schultz and M.E.R. Shanahan, in: Adhesion 11, K.W. Allen (Ed.), pp. 8687, Elsevier Applied Science, New York (1987).

This page intentionally left blank

Surface Contamination and Cleaning, Vol. 1, pp. 109–127 Ed. K.L. Mittal © VSP 2003

Decontamination of sensitive equipment ROBERT KAISER∗ and KYLE HARALDSEN Entropic Systems, Inc. (ESI), P.O. Box 397, Winchester, MA 01890-0597

Abstract—Most of the electronic and electro-optic equipment fielded by the military is incompatible with the standard aqueous based decontamination solutions, such as 5% sodium hypochlorite solution, ESI has developed a nondestructive decontamination process for such sensitive equipment. In this process, the components to be decontaminated are immersed in an ultrasonic bath filled with an organic solvent, if they are contaminated with chemical warfare agents (CWA), or in a solution of a surfactant in this solvent, followed by a pure solvent rinse, if they are contaminated with biological agents or radioactive particles. In both cases, the contaminants are dissolved or suspended in the decontamination liquid in the bath. The contaminants are removed from the decontamination liquid by circulating it through a filtration train. In the case of CWA, the filtration train consists of an activated carbon filter, a particulate pre-filter, and a membrane filter. In the case of biological or nuclear contaminants, the circulating liquid bypasses the activated carbon beds. A prototype decontamination system has been built and operated to demonstrate the process. In this program, a wide range of sensitive equipment was contaminated with a CWA simulant. The contaminated equipment was immersed and sonicated in a flowing solvent, which recirculated around a purification loop, until the simulant could no longer be detected, and dried. The decontaminated equipment was then functionally tested. In all cases: a. no traces of simulant were found on the processed pieces, and b. the processed items were fully functional. Keywords: Decontamination; sensitive equipment; chemical warfare agents; biological warfare agents; ultrasonic cleaning; hydrofluoroethers.

1. INTRODUCTION

While much of the military equipment that is susceptible to chemical or biological threat agents can be decontaminated with aqueous decontamination agents, there are broad classes of critical equipment, including optical, electronic and communication devices, that are rendered nonfunctional by such treatment. Historically, such equipment had been decontaminated by spraying and flushing with CFC113. CFC-113 however, is an Ozone Depleting Compound (ODC) which was ∗

To whom all correspondence should be addressed. Phone: 781-938-7588, x 22, Fax: 781-9387589, E-mail: [email protected]

110

R. Kaiser and K. Haraldsen

eliminated from all DOD activities by the National Defense Authorization Act for Fiscal Year 1993. Thus, alternate methods and equipment for nondestructively decontaminating water sensitive military equipment, such as avionics, electronic, electrical and environmental systems equipment are needed. These methods need to: a. Be effective against a wide variety of threat agents, b. Be nontoxic to personnel, c. Not degrade the equipment being decontaminated, and d. Be field deployable. Decontamination system equipment should be highly mobile and selfsustaining. These methods should also be able to treat equipment that is besmirched with battlefield soils, including dirt (particulates), dried mud, oil, etc. In a broader context, the methods and equipment should also be capable of performing maintenance cleaning operations in a depot environment. These methods and equipment should also comply with environmental regulations. In terms of performance, an effective decontamination method has to be able to remove or deactivate the contaminant without affecting the part being cleaned. Because the type of equipment that would likely be decontaminated is both geometrically complex in shape and thermally sensitive, the most effective technique will likely use a flowing or agitated liquid, at ambient or modest temperatures, as a means of removing the range of threat agents from the equipment. Numerous alternate potential decontamination options have been examined, but have been determined to be of limited effectiveness [1]. Heating an article above a modest temperature may not be an option for decontaminating thermally sensitive items, which leads to problems in terms of effectively removing relatively nonvolatile contaminants. Decontamination by particle blasting methods, such as carbon dioxide snow or plastic pellets, are limited to surfaces that are in direct line of sight with the ejection nozzle. Such methods are not effective in terms of cleaning blind holes, crevices, and obstructed surfaces. These types of methods can be abrasive and destructive to the equipment being decontaminated. Capture and processing of nonvolatile contaminant-laden particles may be a problem as well. 2. APPROACH

Until recently, there were no commercially available organic (i.e. nonaqueous) liquids that would be effective cleaning/decontamination media, and that would satisfy current and projected future safety and environmental criteria. Most volatile organic liquids that exhibit good solvency for chemical threat agents are flammable, toxic or environmentally unacceptable. The 1990 Clean Air Act, designed to eliminate volatile organic compounds (VOCs), ozone depleting compounds (ODCs), and other hazardous air pollutants (HAPs), has severely limited the classes of volatile organic solvents that can be considered.

Decontamination of sensitive equipment

111

Table 1. Properties of fluorinated solvents of interest Solvent

Vertrel-XF [HFC-43-10]

HFE-7100

HFE-7200

Freon TF [CFC-113]

Chemical Formula Supplier Molecular Weight Boiling Point, °C Freezing Point, °C Heat of Vaporization, cal/g @ bp Specific Heat, cal/g @ 25°C Specific Gravity (H2O = 1) Viscosity, N·s/m2 @ 25°C Surface Tension, mN/m @ 25°C Vapor Pressure, mm Hg @ 25°C Solubility of Water in Solvent, ppm Solvent in Water, ppm Hildebrand Solubility Parameter, MPa 0.5

C5F10H2 DuPont 252 54 –80 31 0.27 1.58 0.067 14.1 226

C5F9H3O 3M Co. 250 61 –135 30 0.28 1.52 0.061 13.6 202

C6F9H5O 3M Co. 264 76 –138 30 0.29 1.43 0.061 13.6 109

C2Cl3F3 NA 163 48 –3.5 35 0.21 1.57 0.068 17.3 334

490 140 13.8

95 <12 13.1

92 20 13.5

170 110 14.7

0

0

0

1

0 1700

0 320

0 55

0.8 5000

17.1 None None 200

4.1 None None 750

0.8 None 2.4-12.4% 200

110 None None 1000

VOC, kg/kg Ozone Depletion Potential (CFC-11 = 1) Global Warming Potential [100 yr ITH (1)] Atmospheric Lifetime, yrs Flashpoint, °C Flammability Range in Air, % Exposure Guidelines, 8 hr TWA, ppm

(1) ITH: Integration time horizon Data Compiled from Published Information

Hydrofluorocarbons (HFCs), including the sub-class of hydrofluoroethers (HFEs), are a new class of organic liquids that have physical properties similar to CFC-113 (Table 1). The principal commercially available products are DuPont’s Vertrel-XF (HFC 43-10mee, 2-3 dihydrodecafluoro-pentane) and 3M’s Novec HFE-7100 (methyl nonafluorobutyl ether). The HFEs contain carbon, hydrogen, and oxygen, but no chlorine; and therefore have zero ozone depletion potential. The presence of a minority of hydrogen atoms gives HFEs many of the characteristics of a perfluoroalkane molecule, but also some characteristics of a hydrocarbon molecule.

112

R. Kaiser and K. Haraldsen

Figure 1. ESI’s process flow chart.

While retaining many of the properties and useful characteristics of CFC-113, such as wide materials compatibility, low toxicity, and lack of flammability, they do not possess the environmental limitations of CFC-113. They are not classified as VOCs, HAPs, or ODCs. HFEs also have a significantly lower global warming potential than CFC-113. Entropic Systems, Inc. (ESI) developed a process for the decontamination of sensitive equipment that meets current requirements [2]. A conceptual process flow chart for the process is outlined in Figure 1. The contaminated parts are sprayed with a fluorescent marker and immersed in a bath filled with decontamination liquid. In this bath, surface contaminants are removed from the surface of the parts and transferred to the decontamination or decon liquid, either by solution or by suspension. Contaminated decon (decontamination) liquid is withdrawn from the bath and sent to a purification module that removes the dissolved or suspended contaminants from the liquid. The purified liquid is returned to the bath through spray nozzles to further treat the contaminated parts and decontaminate the cleaning chamber. The parts remain in the bath until a prescribed cleaning regime is completed or until fluorescence sensors in the fluid circuits can no longer detect the fluorescent marker in the solvent that exits the cleaning chamber. The operator who opens the clean side door can verify that there are no longer any harmful levels of contami-

Decontamination of sensitive equipment

113

nants remaining on the treated parts by visually examining the parts for residual fluorescent marker before the parts are removed from the cleaning chamber. The principal objectives of ESI’s development program were to: a. Identify and/or develop safe and environmentally compatible cleaning agents and processes that will effectively remove biological and chemical threat agents from sensitive equipment that may be contaminated with other soils, without damaging this equipment. b. Identify and/or develop means by which the threat agents suspended in the process liquids can be subsequently deactivated, and/or safely removed from these liquids. c. Identify and/or develop methods of monitoring the presence of threat agents and/or other contaminants in the recovered process liquids. d. Demonstrate the technology developed in the above objectives on a pilot plant scale. These objectives were all met, and the major findings of this program are discussed next. 3. MAJOR FINDINGS

3.1. Identify and/or develop safe and environmentally compatible cleaning agents Ultrasonic solvent cleaning processes can effectively decontaminate sensitive equipment. Methoxyperfluorobutane (3M’s HFE-7100) is the decontamination liquid of choice because it meets the following criteria: 1. It is compatible with a wide range of sensitive equipment – the performance of electronic and optical equipment is not affected by immersion in HFE7100. In particular, it does not attack components made of poly (methyl methacrylate) or polycarbonate, as does DuPont’s Vertrel-XF. 2. The principal chemical warfare agents (CWAs) of concern are sufficiently soluble in HFE-7100, as indicated in Table 2. 3. HFE-7100 is effective in ultrasonic cleaning baths because it has very low surface tension, which allows it to penetrate small features of the surface, and because it has a low heat of evaporation, which allows the ultrasonic agitation to produce strong shear forces in order to disrupt the boundary layer and entrain the contaminants. 4. The principal CWAs of concern are quantitatively removed from solution in HFE-7100 by activated carbon. 5. When agent-contaminated HFE-7100 is passed through a bed of activated carbon, the agent adsorbs onto the activated carbon, resulting in agent-free HFE-7100 that can be recycled and reused. In comparison, there was poor adsorption of chemical agents from Vertrel MCA+ in which mustard agent (HD)

R. Kaiser and K. Haraldsen

114

exhibited a higher solubility. Vertrel MCA+ thus was not considered for two reasons – it was too aggressive a solvent to meet material compatibility requirements, and could not be reclaimed for recycling by adsorption of chemical agents. 6. It is nonflammable, nontoxic, and environmentally acceptable. Table 2. Solubility of chemical agents in solvents of interest [2]

Vertrel MCA+ Vertrel XP-10 Vertrel-XF HFE-7100 HFE-7200 CHP

GB

GD

HD

VX

M (RT) M (RT) M (RT) M (RT) M (RT) M (RT)

M (RT) M (RT) M (RT) M (RT) M (RT) M (RT)

17% (RT) 8% (40°C) 8% (40°C) 8% (40°C) 8% (40°C) M (RT)

M (RT) M (RT) M (RT) M (RT) M (RT) M (RT)

M (RT) = Miscible at room temperature

For effective decontamination to occur, sufficient shear has to be provided to result in effective mass and physical transfer of contaminants from the surfaces of the objects being decontaminated to the bulk of the decontamination liquid. In this process: 1. Ultrasonic agitation is a preferred means of providing this shear action. 2. For ultrasonic agitation to be effective, a minimum power density of 60 watts/gallon (15 watts/liter) is required. 3. The ability to generate ultrasonic power over a range of frequencies, from 40 kHz to 170 kHz, is desirable because it rapidly removes a range of particle sizes from the surface of the immersed part. 4. Oils soluble in HFE-7100, but thickened with a nonsoluble additive, are removed from exposed surfaces by high intensity ultrasonic agitation. Biological contaminants are also effectively removed or inactivated by immersion and sonication in HFE-7100 or solutions of a fluorinated surfactant, polyhexafluoropropylene oxide carboxylic acid (DuPont’s Krytox 157FS), in HFE7100, as shown in Figures 1 and 2. More specifically: 1. Vegetative cells are killed by sonication in HFE-7100. 2. Processing in HFE-7100 with up to 4 to 6% Krytox 157FS can result in the sterilization of slides initially contaminated with approximately 100 spores (i.e. > 105 spores/m2). 3. Processing in these solutions also sterilizes slides that had been initially contaminated with 104 bacteriophage particles.

Decontamination of sensitive equipment

115

Figure 2. Circuit boards in culture medium. Circuit boards with environmental contamination were processed in the Cadet and then cultured in trypticase soy broth to assess effectiveness of removal of microorganisms. The circuit board in jar A is an unprocessed control. After incubation, the culture broth is turbid, indicating multiplication of microbial contaminants (bacteria, fungi) present on the circuit board as a result of exposure to the environment. The circuit board in jar B was decontaminated using the Cadet as described in the text. The lack of turbidity in the culture indicates that the circuit board was rendered sterile as a result of processing.

4. Immersion in HFE-7100, with or without surfactant, denatures proteins. 5. The physical removal of biological species from a contaminated surface by sonication in HFE-7100 is enhanced by the presence of > 1% Krytox 157FS in the HFE-7100, and by the use of higher frequency ultrasonic (> 100 kHz) agitation. It should be noted that the mechanism for the removal of radioactive contaminants is similar to the removal of spores. The decontamination of sensitive equipment contaminated with radioactive contaminants by perfluorinated surfactant solutions was initially demonstrated by ESI under the auspices of a program sponsored by the U.S. Nuclear Regulatory Commission [3, 4], and subsequently commercialized as the Sonatol Process [5-7]. 3.2. Identify and/or develop means of removing the threat agents from process liquids The removal of CWAs and of CWA simulants dissolved in hydrofluorocarbons by adsorption on activated carbon was demonstrated in small-scale batch tests that were performed by ESI for CWA simulants, and under subcontract by a surety laboratory (Battelle Memorial Institute) for CWA.

116

R. Kaiser and K. Haraldsen

Pilot scale continuous flow adsorption tests were performed in order to establish the effect of process parameters on the rate and extent of removal of CWA simulants from a hydrofluorocarbon solution as it flowed through an activated carbon column, and on the adsorption capacity of the activated carbon in the column. Providing sufficient residence time is critical to obtaining a reasonable column loading before breakthrough of the contaminant occurs. A CWA simulant loading of 3 wt-% on granular activated carbon (Norit’s 1240 GAC) was achieved with a residence time of five minutes. The length to diameter ratio of the columns should be larger than 3 so as to minimize liquid bypassing. Activated carbon beds and filters that come into contact with contaminated liquid can be contained in commercially available housings that shield the system operator from any contained toxic contents. These sealed containers, and their contents, can be destroyed by standard methods, such as chemical deactivation or incineration. 3.3. Identify and/or develop methods of monitoring the presence of threat agents Spectrographic fluorimetry is an extremely sensitive method of detecting fluorescent materials. With this method, the detection limit of fluorescent dyes dissolved in HFE-7100 was found to be of the order of 10 parts per trillion (ppt). CWA simulants which contained small amounts (0.05 wt-% to 5 wt-%) of a fluorescent dye (Try-33 made by Day-Glo Corp., Dayton, OH) were used as contaminants in the pilot decontamination studies described in the next section. The presence of the fluorescent dye in the contaminant allowed the decontamination process to be easily monitored, in terms of being able to both: 1. Measure and detect low levels of contaminant in the process liquid leaving the ultrasonic bath and the activated carbon columns. 2. Detect traces of residual contaminant on the parts being processed by illuminating these parts with an ultraviolet lamp. 3.4. Demonstrate the technology developed in the above objectives on a pilot plant scale One of the major objectives of the program was to design and build a breadboard decontamination system in order to demonstrate: 1. The functionality of representative pieces of sensitive equipment is not affected by a process consisting of immersion and sonication in a bath of HFE7100, followed by drying in super-heated HFE-7100. 2. The chemical agent simulant is effectively removed from such pieces of sensitive equipment. 3. The chemical agent simulant dissolved in HFE-7100 is quantitatively removed in real time from solution by passing the contaminated solution through a bed of activated carbon. 4. The internal recovery of the contaminated liquid, to allow recycling of the purified process liquid.

Decontamination of sensitive equipment

117

5. The capture and removal of waste products (agent simulant, other soluble contaminants, and particulates) from the contaminated solution in fully enclosed, disposable activated carbon and filter cartridges, that are easy to install, remove, and would minimize operator exposure to the contents of the cartridges. 6. The ability to immerse parts in both pure HFE-7100 and in fluorinated surfactant or coupling agent solutions in HFE-7100. Additional objectives were to obtain operating data on specific unit operations, such as the effect of liquid circulation rate on the removal kinetics of simulant from contaminated test parts, the effects of liquid flow rate and contaminant concentration on the efficiency, and capacity of activated carbon columns. The breadboard decontamination system consisted of three modules: 1. A Poly-KleenTM Vapor Degreaser 2. An activated carbon column train 3. A circulating water chiller The Poly-KleenTM Vapor Degreaser is a manually operated vapor degreasing system designed to be used with low boiling fluorinated solvents, such as perfluorocarbons, hydrofluorocarbons and hydrofluoroethers. Figure 3 is a photograph of the inside chambers of the system. Overall dimensions are 1956 mm H x 736 mm D x 904 mm H (77" L x 29" D x 36" H). The height includes the handles on the sliding door which are 76 mm (3") high. The Poly-KleenTM Vapor Degreaser takes full advantage of the ease of fabrication of polypropylene and its compatibility with fluorinated liquids to allow a high performance system to be manufactured at a lower cost than an equivalent stainless steel system. The basic Poly-KleenTM Vapor Degreaser is a three-sump unit that requires 140 liters (38 gallons) of fluid to operate. The principal components of the system and their effective dimensions are: An immersion sump

432 mm L x 279 mm W x 254 mm H (17" L x 11" W x 10" H)

A boil sump

483 mm L x 356 mm W x 254 mm H (19" L x 14" W x 10" H)

A drying sump

356 mm L x 279 mm W x 305 mm H (14" L x 11" W x 12" H)

A vapor zone

1854 mm L x 279 mm W x 76 mm H (73" L x 11" W x 3" H)

A chilled condensate zone

1854 mm L x 279 mm W x 203 mm H (73" L x 11" W x 8" H)

A freeboard zone

1854 mm L x 279 mm W x 356 mm H (73" L x 11" W x 14" H)

A water separator

178 mm L x 102 mm W x 229 mm H (7" L x 4" W x 9" H)

A sliding cover

940 mm L x 330 mm W (37" L x 13" W)

118

R. Kaiser and K. Haraldsen

Figure 3. Interior of Poly-KleenTM system.

Figure 4. Activated carbon adsorption module.

The activated carbon module is shown in Figure 4. The modules and the chiller are interconnected as shown in the process flow diagram presented as Figure 5. Referring to this process flow diagram, the in-line process components from BV6 to CV-3 were mounted on an angle iron frame, 1829 mm L x 736 mm W x 762 mm H (72" L x 29" W x 30" H).

Decontamination of sensitive equipment

119

Figure 5. Poly-Kleenä process flow diagram.

Cleaning trials were performed with the following pieces of sensitive equipment a. Auto-Ranging LCD Digital Multimeters, Model No. 22-179A, Radio Shack, A Div. of Tandy Corp., Fort Worth, TX b. Electronic Calculator, Model No. EC-441, Radio Shack, A Div. Of Tandy Corp., Fort Worth, TX c. Global Positioning System (GPS) receiver, Model No. GlobalNav 212, Serial No.005263360, Lowrance Electronics, Inc., Tulsa, OK. d. Night Vision Binoculars, Model RO 38, 4 x 48 Nighthawk, Serial No. 982331, with Model RO45, Zoom IR Illuminator, LAN Optics International, Burlington, MA. e. 7.65 mm semi-automatic pistol, Model PP, Carl Walther GmbH Sportswaffen, Ansberg, Germany f. Inverter Circuit Boards, 38 mm (1.5 in) square, designed by Entropic Systems, Inc. Numerous tests were performed with digital multimeters, which were considered to be good prototypes for sensitive equipment. These items performed a number of electrical functions, they had a liquid crystal display covered by a clear plastic window, they contained a variety of materials that would be damaged by

120

R. Kaiser and K. Haraldsen

many solvents, and were inexpensive enough (about $12.00 each) to be considered disposable test items. In addition, some tests were performed with other items to test the effects of part geometry. These items included standard 25 mm x 75 mm microscope slides (standard flat surfaces), brass pipe nipples (easily accessible interior surfaces), and magnet assemblies (difficult to access interior surfaces). A magnet assembly consists of a circular piece of stainless screen (typically 100 mesh) that is sandwiched between two cylindrical Alnico magnets of equal diameter. The magnets were 13 mm (½") in diameter by 6.5 mm (¼" ) in height. The soil is deposited on the screen before forming a magnet assembly. This sandwich was then subjected to a cleaning trial. The changes in weight of the assembly and in the appearance of the screen were measures of the effect of the cleaning trial. The test pieces were contaminated with a variety of neat and thickened CWA simulants and other soils. CWA simulants used in these tests were diethyl phthalate (DEP), tributyl citrate (TBC), and Krytox 157 (L) and (H) fluorosurfactants. These materials are all water insoluble oils that have a low vapor pressure at ambient room temperature. They also all are miscible with HFE-7100. The CWA simulants were all doped with a fluorescent dye, Try-33, that greatly facilitated their detection on the test pieces and in the decontamination liquid. In some of the tests, a thickener was added to the simulant to mimic the behavior of thickened CWA agents. Two different types of thickeners were used: fumed silica (Cabosil LM-130, Cabot Corp.), and an acrylic polymer (Paraloid K-125, Rohm & Haas Corp.). Paraloid K-125 has been used to thicken military CWA. The consistency of the simulant depends on the amount of thickener used. At 1-2 wt-% thickener loading, the simulants flow like honey, while they become semisolid gels at thickener loading greater than 5 wt-%. One key difference between colloidal silica and an acrylic polymer is that colloidal silica is not soluble in any organic solvent, but the acrylic polymer can dissolve in a polar organic solvent. In addition to the above simulants, test pieces were also contaminated with representative soils that could be found on fielded equipment: mineral oil, SAE 30 motor oil (NAPA) thickened with Arizona road dust (Duke Scientific Co, Palo Alto, CA), multi-purpose lithium grease (Lubrimatic), and dried, 50 grit SiC water based lapping compound (Clover). The contaminant removal tests were performed in the Poly-KleenTM system according to the following general procedure: a. The equipment to be processed was weighed and photographed under visible and UV light. b. One or more tared pieces of equipment were coated with contaminant(s) or soil(s), photographed under visible and UV light, and re-weighed. c. The test piece(s) were placed into the transfer basket of the Poly-Kleenä system, which was then covered with a tight fitting screen. d. The immersion sump of the Poly-Kleenä system contained enough HFE-7100 to cover the part in the basket. Sonication for 30 minutes degassed this liquid.

Decontamination of sensitive equipment

121

e. The transfer basket containing the items to be cleaned was lowered into the immersion sump, and statically (i.e. no liquid flow) sonicated for a finite period of time, usually 15 minutes. f. After static sonication, the rinse pump was turned on and the liquid in the immersion bath was circulated through the activated carbon columns at a rate of 1,700 ml/minute for a finite period of time. The circulation time ranged from 15 minutes to 2 hours, depending on the purpose of the test. g. The rate of decontamination was monitored by following the concentration of the contaminant in the decontamination liquid (HFE-7100). h. Steps e and f were repeated until the presence of contaminant in the circulating liquid could no longer be detected. i. When the immersion sump liquid was free of contaminant, the transfer basket was moved from the immersion sump to the superheat sump and dried for 30 minutes to remove liquid drag out. j. The transfer basket was removed from the Poly-Kleenä system. The test pieces were removed from the basket, visually examined, photographed under visible and UV light, reweighed, and archived. In order to maximize ultrasonic power density, the minimum amount of liquid needed to cover the parts being cleaned was used. Typically, the sump contained from 130 to 180 mm (5 to 7 inches) of liquid, which corresponds to a liquid volume of approximately 15 liters to 30 liters (4 to 8 gallons) and a corresponding ultrasonic power density of 26 to 18 watts/liter (100 to 70 watts/gallon). In preliminary tests, it was noted that immersing and sonicating the test samples when the immersion sump was filled to the brim (about 53 liters (14 gallons)) did not result in effective cleaning. At that volume, the ultrasonic power density had dropped to a value of 8 watts/liter (30 watts/gallon). While this value would be considered marginal in a stainless steel ultrasonic bath, where the ultrasonic waves can be reflected from the walls back into the liquid, in a polypropylene bath in which the walls absorb rather than reflect the ultrasonic waves, this power density level is too low. If parts were also contaminated with biological agents, after Step h, they would be sonicated in a fluorinated surfactant/HFE-7100 solution that would be circulated through microfilters to remove suspended materials. The parts would then be rinsed in fresh HFE-7100 to remove fluorocarbon surfactant residues, and then dried as described above. Table 3 lists the sensitive equipment decontamination experiments that were carried out in the Poly-Kleenä system during the course of the program. The combination of equipment processed, contaminants used, and monitoring method(s) examined are listed in this table. The results of the various cleaning results are summarized in Table 4. This table records the weights of the items listed in Table 3, before and after contamination, as well as the post-cleaning weight and visual appearance of these items.

R. Kaiser and K. Haraldsen

122

Table 3. List of sensitive equipment decontamination experiments performed Experiment Sensitive equipment No. processed

Contaminant(s)

Monitoring method

1 2

Diethyl Phthalate Tributyl Citrate

Visual Visual, UV

Tributyl Citrate Krytox FS Tributyl Citrate Krytox FS + Mineral Oil

Visual, UV Visual, UV Visual, UV

Krytox FS + Mineral Oil

Visual

Tributyl Citrate Krytox FS + Tributyl Citrate Krytox FS

Visual, UV Visual Visual

Motor Oil, Lapping Compound, Krytox FS Krytox FS

Visual Visual

Krytox FS Krytox FS Krytox FS Krytox FS Tributyl Citrate

Fluorescence Fluorescence Fluorescence Fluorescence Fluorescence

3 4 5 6 7 8 9 10A-10E 11 12 13 14 15 16 17

Multimeter 2 Microscope Slides & 2 circuit boards 2 Multimeters Multimeter Multimeter GPS Receiver & Radio Shack Calculator Multimeter & Night Vision Goggles Multimeter & Circuit Board Walter PP Pistol Multimeter & Pipe Nipple (10 D only) Multimeter 2 Magnet Assemblies & 2 Brass nipples Multimeter - Face Down Multimeter - Face Down Multimeter - Face Down Multimeter - Face Up Multimeter - Face Down

Except for the runs where there was visible attack of the substrate by the simulant (as in run 1 in Table 3), there was an increase of less than a 0.1 gram in the weight of the object after contamination and cleaning and the original (i.e. before contamination) weight of this object. In some cases, there was a weight loss of the order of 0.1 gram (as in the calculator in run 6 and the pistol in run 9). This was attributed to the removal of other soils that were previously present on these test items. If the ratio of (weight change/contaminant weight) is used as a cleaning criterion, this value is less than 10%, except for run 1 (for the reasons cited above), and for runs 13 to 17. For these last five runs, the relatively high values of this ratio is attributable to the precision of the weight measurement. The weight measurements were performed on a balance that had an accuracy of ± 0.02 gram, which would account for most of the observed weight differences.

Table 4. Sensitive equipment cleaning results

Decontamination of sensitive equipment 123

124

R. Kaiser and K. Haraldsen

Figure 6. Decontamination of night vision goggles (NVG).

While not a quantitative measurement, visual examination under ultraviolet illumination was considered to be the most sensitive and accurate means available to ESI of assessing whether traces of fluorescent contamination remained on the processed objects. Fluorescent contamination was observed only for run 1, and runs 3 and 5, where there was no noticeable weight increase.

Decontamination of sensitive equipment

Figure 7. Decontamination of Walther PP 7.65 mm hand gun.

125

126

R. Kaiser and K. Haraldsen

Photographs of various test objects were taken under both normal and UV illuminations, a. Before the application of contaminant(s), b. After the application of contaminant(s), and c. After cleaning or decontamination. The presence or absence of contaminant on the items is clearly evident by simple examination of the photographs taken under UV illumination of the night vision goggles presented in Figure 6 and the pistol in Figure 7. Process kinetics was examined by monitoring the concentration of contaminant in the liquid in the immersion bath as a function of time. Removing the contaminant from the surfaces of the object being cleaned and transferring it into the bulk liquid took approximately 15 minutes. Once the circulation pump to the activated carbon columns was turned on, approximately three bath turnovers were required to purge the immersion sump of contaminant. In this test program, turnover time was of the order of 12 to 15 minutes. 4. POTENTIAL BENEFITS OF THE TECHNOLOGY

Potential benefits of the decontamination process discussed in this paper include: ● Demonstrated compatibility of sensitive equipment with the hydrofluoroether process liquids under processing conditions. ● Demonstrated solubility levels of CWA agents of interest in the process liquids. ● Demonstrated ability to remove CWA simulants from items of sensitive equipment. ● Demonstrated ability of HFE-7100/Krytox 157FS surfactant solutions, in conjunction with multi-frequency ultrasonic agitation to remove/neutralize a range of biological agents from solid substrates, including circuit boards. ● Demonstrated ability to quantitatively remove CWA and CWA simulants from solution in HFE-7100 by adsorption on activated carbon. ● Simple, safe, automatable decontamination process. Acknowledgements The work described in this paper was supported by the U.S. Air Force under the auspices of SBIR PROJECT AF 97-014, “DECONTAMINATION OF AIRCRAFT ELECTRONIC EQUIPMENT”, Contract Number: F41624-98-C-5061. It was facilitated by the insights and encouragement of Dr. Ngai Wong, AFRL/HEST, the Air Force Technical Monitor.

Decontamination of sensitive equipment

127

REFERENCES 1. “Joint Service Sensitive Equipment Decontamination (JSSED), Block I Systems Technology Overview”, Report PAO-01-432, October 2000 (available from DTIC, Fort Belvoir, VA). 2. R. Kaiser, “Decontamination of Aircraft Electronic Equipment”, Final Technical Report, SBIR Project AF97-014, Contract No. F41624-98-C-5061, November 2000. 3. R. Kaiser, C.S. Yam and O.K. Harling, “Enhanced Removal of Radioactive Particles by Fluorocarbon Surfactant Solutions – Process Development”, Interim Report, Prepared for US Nuclear Regulatory Commission, Washington, DC, Contract No. NRC-04-93-106, March 1995. 4. R. Kaiser, C.S. Yam, S.R. Landahl, P.A. Droof and P.H. Jones, Jr., “Enhanced Removal of Radioactive Particles by Fluorocarbon Surfactant Solutions – Process Demonstration”, Final Report, Prepared for US Nuclear Regulatory Commission, Washington, DC, Contract No. NRC04-93-106, September 1997. 5. R. Kaiser, C.Y. Yam and A.E. Desrosiers, “Decontamination of Electromechanical Parts by the Sonatol Process: II – Results”, Proc. Waste Management (WM) ’98 Conference, Tucson, AZ, March 1998. 6. A.E. Desrosiers and R. Kaiser, “Decontamination of HEPA Filters for Reuse”, Proc. EPRI International Low-Level Waste Conference, McAfee, NJ, July 1999. 7. A.E. Desrosiers, R. Kaiser and C.B. Voth, “TRU Waste Minimization During Hot Cell Decommissioning”, Proc. American Nuclear Society Annual Meeting, Boston, MA, June 1999. Note: Copies of Refs. 2, 3, 4, can be obtained from the sponsoring agencies or the author of this paper.

This page intentionally left blank

Surface Contamination and Cleaning, Vol. 1, pp. 129–136 Ed. K.L. Mittal © VSP 2003

The fundamentals of no-chemistry process cleaning JOHN B. DURKEE II∗ Creative Enterprizes, 437 Mack Hollimon Drive, Kerrville, TX 78028

Abstract—This paper describes a new method for cleaning of parts. The method converts soils to water-organic emulsions via pressure waves generated by ultrasonic transducers. Underwater photographs and symbolic representations show that the process consists of: (1) contact with ultrasonicgenerated pressure waves produces an emulsion of soil in water, and (2) rinsing the emulsion from the parts. Keywords: No-chemistry; cleaning; emulsion; ultrasonics.

1. INTRODUCTION

Cleaning without chemistry is not new; however, the information in this paper is new. This paper is not about flushing with pure water to displace particles in critical cleaning situations nor is it about slowly dissolving non-volatile residue (NVR) [1] without rinsing residues in precision cleaning situations. This paper is about cleaning large or small quantities of oils, greases, or soaps from metal (chiefly) parts using only water [2]. This paper describes a new method of solving practical cleaning problems at all levels of cleaning quality [3]. This new method is used most often in continuous processes, though it can be used easily in batch processes. In this paper, the new method is described and compared to normal aqueous cleaning technology. Photographic evidence of its use is shown. Limitations are defined. Areas for future research are discussed. 2. EXPERIMENTAL

Four photographs, Figures 1a to 1d, were taken at 32°C (90°F) in un-degassed tap water. They illustrate the experiments carried out here to describe the two-step process. In the first process step, the parts to be cleaned are bombarded with ultra∗

Phone: (830)-895-3755, Fax: (612)-677-3170, E-mail: [email protected]

130

J.B. Durkee

Figure 1a (Equivalent to Figure 3a) Initial Formation of Emulsion

Figure 1b Continued Formation of Emulsion

Figure 1c (Equivalent to Figure 3c) Formation of Gel from Dilute Emulsion

Figure 1d Further Condensation of Gel Material

Figure 1. Photographic representation of cleaning with no-chemistry process.

sonic waves from a transducer. This produces an emulsion of water and soil on the parts. In the second step, the emulsion (which contains the soil) is rinsed from the parts using any combination of water jets. The circular-shaped ultrasonic transducer produced 2000 watts over about a 30.48 cm (12 inch) length. The part was an easily recognizable form – a bullet formed of carbon steel. The part was held on a thin copper wire about 1.27 cm (0.5 inch) away from the circular transducer. The soil was 10W30 motor oil applied liberally with a paintbrush before the part was immersed in water. The four photographs in Figure 1 were taken over a period of about five seconds. The emulsion was analyzed for oil, and part cleanliness was measured.

The fundamentals of no-chemistry process cleaning

131

3. RESULTS

Emulsions similar to those seen in Figures 1a to 1d were sampled via a suction tube from beneath the liquid level. The initial emulsion (Figures 1a and 3a) was stable over several days and contained around 5% water. The final gel-emulsion was stable under boiling conditions for at least one week nonstop and contained approximately 90% water. After the gel was rinsed off via pressurized water sprays, the parts (bullets) were free of oil as ascertained by white glove tests and viewing in black light. An extraction test in hexane showed all parts to have <1.85 mg/m2 (15.6 mg/ft2) oil residue after cleaning in this manner. These parts would be very clean for industrial use, after only one step of sonication and rinsing. 4. DISCUSSION

Today’s aqueous and solvent processes both involve the same three factors: solvency, temperature (heat), and mechanical force. This is shown in Figure 2a. Naturally, solvency dominates solvent cleaning. Mechanical force dominates aqueous cleaning. The No-chemistry technology presented in this paper involves a different paradigm. This is shown simply in Figure 2b. There is no chemistry so there is no direct concern about solvency. The process is practiced from room temperature (or below) up to about 51.6°C (125°F). Consequently, the addition of heat is not a significant factor. 4.1. Another kind of mechanical force Mechanical force is the remaining of the three factors. The force is generated by ultrasonic pressure transducers which normally produce cavitation bubbles.

Figure 2a. Traditional elements of cleaning.

Figure 2b. No-chemistry cleaning.

132 J.B. Durkee

Figure 3a Pressure Waves Impact Contaminated Surface

Figure 3b Pressure Waves Produce Emulsion

Figure 3c Emulsion Collects into a Gel

Figure 3d Rinse Dislodges Gel from Surface

Figure 3e Rinsed Cleaned Surface

Figure 3. Symbolic representation of cleaning with no-chemistry process. Small blue arrows in all figures represent pressure waves produced by ultrasonic transducers. Colored shapes in Figure 3a represent individual elements of soil. Shadowed shapes in Figure 3b represent conversion of soil to water emulsions. The agglomeration of shadowed shapes in Figure 3c represents a water-bearing gel with some adhesion to the surface. Large red arrows in Figures 3d and 3e represent jets or sprays of pressurized water as flows of rinse water. The part is free of soil in Figure 3e.

The fundamentals of no-chemistry process cleaning

133

Table 1. Comparison of no-chemistry cleaning to cleaning via cavitation Item

Conventional cavitation

No-chemistry cleaning

Water degassing Optimum temperature Hydraulic conditions Standing waves Bubbles Cavitation

Essential ~ 71°C (160°F) Static flow conditions Avoided via frequency sweep Observed YES

Not done Any temperature Moving flow conditions No frequency sweep Not Observed NA

Hence, one might assume that the nature of the force is cavitation-induced implosion of tiny bubbles. This assumption is incorrect. Cavitation plays no role in this No-chemistry cleaning process. In fact, conditions are deliberately controlled to be opposite of those which normally produce cavitation bubbles. These conditions are given in Table 1, and will be later shown in Figure 3a to Figure 3d. 4.2. The no-chemistry cleaning process The No-chemistry cleaning process is simple. As stated in Section 2, only two steps are involved. Both must be performed completely, or the cleaning will not be effective. The steps are to: (1) Produce an emulsion via ultrasonic pressure waves (2) Rinse the emulsion off the parts The pair of steps are shown photographically in Figures 1a to1d, and symbolically as Figures 3a to 3e. They may be repeated as necessary to produce the desired level of cleaning in the time or space allowed. The separation in time of Figures 1a to1d and Figures 3a to 3d/3e is about five seconds. In other words, the cleaning process involving these two steps can easily be completed in twenty seconds, or less. The process is self-limiting. It is complete when there is no soil left to emulsify. 4.3. Conditions necessary to form and remove emulsion There are three necessary conditions [2]: 1. The transducers must be located close to the parts. The distance of 1.27 cm to 7.62 cm (0.5 to 3 inches) is recommended. 2. The volumetric power intensity must be high, vs applications involving ultrasonic-produced cavitation, (>26.4 watts / liter (100 watts / gal)). 3. There must be motion in the fluid (to complete the rinsing step). Both continuous flow and batch processes have been built and operated successfully. A single transducer can be located above or below the parts. A pair of transducers may be located above and below the parts. Line of sight access is not required, as with megasonic transducers.

134

J.B. Durkee

If multiple transducers above and below the parts are used, the same frequency is used for each, vs the referenced patent [4], which requires a difference in frequency. The purpose of the multiple transducers is to expose the surface which would not ordinarily be exposed to ultrasonic waves. Flat parts, which have two sides, are a common example. The rinsing nozzles can be located under water with the transducers or the parts can be raised above the water and rinsed there. Obviously, a different design of the rinse nozzle would be used in each case. The rinsing step is absolutely crucial, as is shown by the following experience. A scaled-up machine was constructed. It had two steps of emulsion production via sonication, and no intermediate step of rinsing. It was worthless as a cleaning machine, despite 2,000 watts of ultrasonic power aimed at a few non-moving parts. After a great deal of trial and error, an intermediate rinsing zone was added. It was this inadvertent experiment which showed that No-chemistry cleaning was a two-step process. Water quality is not relevant. Excellent cleaning has been obtained in deionized water, tap water, water produced by reverse osmosis, water containing tramp insoluble soil, and water containing ~ 1% emulsion. There is no practical limit to operating temperature. Issues which determine temperature [2] are soil and substrate. Lower temperatures make emulsion formation and removal more difficult because the emulsion is more viscous. Higher temperatures cause steel surfaces to rust. Typical values of operating temperature are between 26.6°C (80°F) and 51.6°C (125°F). Excellent cleaning of molybdenum-based grease from 304 stainless steel has been done at 90.6°C (195°F) [5]. This temperature was chosen to reduce the viscosity of the grease in order to increase indirectly the rate of formation and removal of emulsion. 10W50 motor oil has been removed from carbon steel at 15.6°C (60°F). 4.4. Process operation No-chemistry cleaning is not a mere curiosity, like the vortex tube. More than a dozen machines have been constructed. Each machine transports parts through zones where the parts can be sonicated to produce an emulsion and then another zone where the emulsion can be rinsed away. Some machines have operated for thousands of hours. Some machines have cleaned millions of parts. Nearly all machines have cleaned hundreds of pounds of parts per hour. No chemicals have been added to any machine. These machines differ chiefly in transport of parts, and that is the fundamental limit on the use of No-chemistry cleaning technology. Large parts are most difficult to transport and expose to ultrasonic transducers mounted close to the part’s surface. Tractor axles are the largest part successfully cleaned. Each weighs about 18.8 kilograms (forty pounds) and are about the size of a 13" television set. Two machines continuously clean oils from these parts.

The fundamentals of no-chemistry process cleaning

135

4.5. Process limitations No-chemistry cleaning technology has at least three types of limitations. 4.5.1. Technology related Proximity of the parts to the transducer is the most significant item. Shorter distances are favored: 1.27 cm (0.5 inches) to 7.62 cm (3 inches). In a recent successful application, the transducers were 12.7 cm (5 inches) from the parts. But the power intensity had to be increased to ~ 2.5 times normal. The practical effect of this restriction is to limit the dimensional size or weight of parts which can be cleaned without chemistry. 4.5.2. Inadequately designed process Design factors must be within limits determined by experiment, such as: ● Holdup time under sonication should be at least fifteen seconds. Holdup time is a tradeoff between cleaning quality and productivity [6]. ● Too few steps of sonication (formation of emulsion) and rinsing. The number of steps of sonication is a tradeoff between cleaning quality and machine size or cost. ● Operating temperature can be too low (between 26.6°C [80EF] and 51.6°C [125°F]). This can limit cleaning quality, as a more viscous emulsion is rinsed less efficiently. ● Omission of a rinse step after a sonication (formation of emulsion) step. This can limit cleaning quality if some emulsion is not removed ● Intrusion of free or tramp soil. This situation is worse than having to clean parts which are more dirty, because the soil content can be unanticipated and variable. Free soil can be fatal to any proposed application. 4.5.3. Waste management This is a problem where the soil can be removed from parts, but cannot be removed from the cleaning machine. The problem is unique to No-chemistry cleaning technology. This is because there is no chemistry to “surround” the soil, and “escort” it from the cleaning machine. Thus soil elements agglomerate or attach themselves to the cleaning machine. When No-chemistry cleaning removes the soil from the parts and allows it to be deposited on the machine, the application is a total failure. Fortunately, this happens only with some soils which do not form a stable emulsion with water. Two such soils are molybdenum disulfide and lithium-based greases. 5. FUTURE RESEARCH

Outside of developing knowhow and implementing this technology in commercial cleaning machines, future R&D will focus on achieving a more complete understanding in these areas:

136

J.B. Durkee

a. Factors which govern the rate of emulsification. b. Surface chemistry and physics associated with cleaning ferrous parts which do not rust after rinsing and drying. c. Limits of part size on cleaning rate. This will involve the following variables: weight, area per weight, part geometry, distance from transducers to parts, and deliberate reflection of ultrasonic pressure waves onto the parts. One practical example will be an attempt to clean engine blocks. d. Types of part conveyances which can or cannot be used. Currently, rotating metal drums and metal belt conveyor are the only part conveyances used. Planned experiments include hydraulic conveyance of small parts in a duct and woven spun-bonded fabric materials used as belts. e. Soil management. Current plans are to develop an efficient evaporation system which would concentrate the soil–water emulsion. Water would be recovered for reuse in cleaning; soil would be concentrated to minimize disposal costs. 6. SUMMARY

Cleaning of oil and grease soils without chemistry is technically feasible. Photographs show how these soils are emulsified in water. Cleaning data show how the emulsions are rinsed to produce clean parts. This two-step cleaning process has been implemented in continuous or batch machines. Acknowledgments I would like to thank Walter Johnson, Flo-Matic Corporation, Rockford, Illinois, USA, for his sponsorship of this work and for the freedom to experiment without regard to commercial need. I would also like to thank my colleagues Gary Kauffman and Vasilly Pekun of Flo-Matic. REFERENCES 1. “Guide to Inspections Validation of Cleaning Processes”, http;//ww.fda.gov/ora/inspect_ref/igs /valid.html 2. W.J. Johnson, US Patent 6,368,414 (2002). 3. J.Y. Baker and J.B. Durkee, A2C2 Magazine, 4-9 (October 1999), 4-6 (November 1999); and 3940 (January 2000). 4. H.B. Swainbank et al., US Patent 4,788,992 (1988). 5. W.J. Johnson, paper presented at the International Fastener Exposition, Chicago, May 27, 1999. 6. J.B. Durkee, W.J. Johnson, G. Kauffman and V. Pekun, US Patent Applications 340769/ 900016, 340769/90008, and 340769/900032 (2000).

Surface Contamination and Cleaning, Vol. 1, pp. 137–150 Ed. K.L. Mittal © VSP 2003

Development of a technology for generation of ice particles D.V. SHISHKIN, E.S. GESKIN∗ and B. GOLDENBERG Waterjet Laboratory, Department of Mechanical Engineering, New Jersey Institute of Technology, Newark, NJ 07102-1982

Abstract—The mission of this project was to develop a practical technology for formation of ice particles. Our previous works demonstrated the effectiveness of the use of ice-air mixture as a cleaning medium. However, a practical technology for fabrication of ice particles of a desired size and at a desired temperature is still under development. The physical properties of ice (tendency to agglomerate, melting, etc.) make the formation and transportation of ice particles extremely difficult. We developed a process for controllable generation of ice particles using a rotational crusher embedded into a heat exchanger. Water was supplied at the bottom of the heat exchanger and as it moved along the rotating auger ice was formed and crushed. A refrigerant or liquid nitrogen were used as cooling media. At the exit of the heat exchanger the air stream entrained the generated particles. In the course of operation the cooling conditions and the auger rotation were maintained constant while the water flow rate varied. The rate of production and the shape and size of the generated particles were monitored. We also investigated the temperature distribution along the heat exchanger and the corresponding distribution of particle sizes. As the result of this study, process phenomenology was developed and a design of the system for formation of ice particles was suggested. Keywords: Ice; particles; solidification; solid decomposition; brittle fracturing; thermal stresses.

1. INTRODUCTION – ICE AS AN ABRASIVE MEDIUM

A number of surface processing technologies based on the use of the air-ice stream have been previously suggested. A car washing machine constituted the first attempt towards utilization of the ice particles [1]. Another technology used a stream of charged ice particles directed toward surfaces [2]. Szijcs [3] proposed cleaning of sensitive surfaces by the impact of fine grade ice and air. The atomization of the liquid in the air stream and subsequent freezing of the generated fine droplets formed the blast material. The freezing was achieved by the addition of a refrigerant (N2, CO2, Freon) into the stream in the mixing chamber or by the addition of refrigerant into the jet after the mixing chamber. Another technology in∗

To whom all correspondence should be addressed. Phone: (973)596-3338, Fax: (973)642-4282, E-mail: [email protected]

138

D.V. Shishkin et al.

volved the use of ultra-clean ice particles, having a uniform grain size, for cleaning the surface and grooves of ferrite block (Tomoji [4]). An ice blasting device utilizing stored particles was suggested by Harima [5]. Vissisouk [6] proposed to use ice particles near melting temperature for surface decoating. Mesher [7] developed a nozzle for enhancement of the surface cleaning by ice blasting. Shinichi [8] suggested an inexpensive cleaning of various surfaces by mixing ice particles, cold water and air. Niechial [9] proposed an ice blasting cleaning system containing an ice crusher, a separator and a blasting gun. Settles [10] suggested producing ice particles of a size range below 100 µm within the apparatus just prior to the nozzle. Although the potential use of ice blasting has been suggested by a number of inventors, the practical use is much more limited. Herb and Vissisouk [11] reported precision cleaning of zirconium alloys in the course of production of bimetallic tubing by ice pellets. It was shown that ice blasting improved the quality of bimetal. The use of air-ice blasting for steel derusting was reported by Liu [12]. The following operational conditions were maintained: air pressure: 0.2-0.76 MPa; grain diameter: below 2.5 mm; ice temperature –50°C; traverse rate 90 mm/min; and standoff distance 50 mm. At these conditions the rate of derusting ranged from 290 mm2/min at the air pressure of 0.2 MPa to 1110 mm2/min at the air pressure of 0.76 MPa. The quality of the treated surfaces complied with ISO 8501-1 Sa 2. In the final analysis, the adoption of the ice-jet technology is determined by the effectiveness of the generation and handling of ice particles. Regular abrasives are stable at practically feasible operational conditions, while ice particles can exist only at subzero temperature. Maintaining such a temperature prior, within and outside of the nozzle is an extremely difficult task. The adhesion between the particles increases dramatically as the temperature approaches 0°C. At this condition ice tends to pack and clog the supply lines. Thus these lines must be maintained at a low temperature. This and other similar problems prevent adoption of icewaterjet (IWJ) by the industry. In order to assure the acceptance of IWJ, it is necessary to develop a practical technology for formation, transportation and acceleration of ice particles. 2. PROPERTIES OF WATER ICE

The practical application of water ice as a machining medium is determined by the physical properties of the solid water. The properties of Ice I existing at the modest pressure (below 200 MPa) have practical importance. An important feature of Ice I is the reduction of the melting temperature with an increase of the pressure. The minimum temperature of the liquid water is attained at a pressure about 200 MPa and is equal to –20°C. Another important feature determining particle behavior in the course of impact is ice elasticity. In the temperature range of –3°C to –40°C ice is an almost perfect elastic body. Hooke’s law is obeyed if the

Development of a technology for generation of ice particles

139

stresses in the ice are below a certain level and are applied for a short period of time. The dynamic elastic properties of ice at –5°C are: Young’s modulus (E) = 8.9-9.9 GPa; Rigidity modulus (G) = 3.4-3.8 GPa; Bulk modulus (K) = 8.3-11.3 GPa; Poisson’s ratio (χ) = 0.31-0.36 according to Hobbs [13]. For comparison, for an aluminum alloy 1100-H14, E = 70 GPa and G = 26 GPa. For silica glass E = 70 GPa. If the columnar ice is stressed perpendicular to the long direction of the column, the static Young’s modulus in bars is determined by the following equation: E = (5.69-0.64T)*104

(1)

where, temperature T is given in °C. The dynamic Young’s modulus of ice increases almost linearly from 7.2 GPa at –10°C to 8.5 GPa at –180°C, and is independent of the direction of loading. The data above show that the ice powder can be considered as a soft blasting material and used accordingly. One of the main issues in the use of the ice powder is sintering of the particles and their adhesion to the surface of the enclosure. The strength of adhesion depends on ice temperature. This dependence is shown in Figure 1 (a). It follows from this figure that it is necessary to maintain ice temperature below –30°C to prevent sintering of the particles. Sintering also depends on the duration of particles contact. The radius of the neck, which forms between two ice spheres,

Figure 1. (a) Strength of adhesion of ice particles, and (b) schematic of the sintering of ice particles [13].

140

D.V. Shishkin et al.

brought into contact for time t at temperature T (Figure 1 (b)) is determined by the equation n

æ x ö A(T ) ç ÷ = m t r èrø

(2)

where x is the radius of the neck, r is the radius of sphere, A(T) is a function of the temperature, n and m are constants. A(T), m and n depend on the mechanism of sintering. It follows from equation (2) that it is necessary to prevent the contact between particles in order to avoid particles sintering. The moisture contained in the atmosphere in the course of ice transportation enhances the adhesion of the ice to the walls as well as sintering of ice particles. Both phenomena result in the plug formation and clogging of the conduits. 3. EXPERIMENTAL SET UP FOR FABRICATION OF ICE PARTICLES

In our previous experiments several systems for ice powder formation were tested. One of such systems as depicted in Figure 2 was selected for further experiments. The system consists of the following functionally separated blocks: – ice making block which includes the evaporator, auger, auger driver, sealing and liquid nitrogen cooling apparatus; – ice unloading mechanism – nozzle block which includes parallel nozzles and focusing device. In our experiments water entered the heat exchanger via a special port. As it moved along the rotating nozzle it solidified and an ice plug was formed. Decomposition of this plug led to formation of ice particles. At the outlet of the heat exchanger the powder entered into the nozzle block 5 and was driven to the air gun 6. The heat exchanger and the auger of the icemaker constituted a modified commercial icemaker of the Hoshizaki Co. of America, Peachtree, GA. The design of these parts will be changed in the next generation of the device. The cooling was carried out by the refrigerant Galden HT-55 supplied by the cooling TurboJet apparatus or by liquid nitrogen stored in a tank. In both cases the supply of the cooling medium was determined by the characteristics of the source, the refrigeration system or the nitrogen tank. We replaced Hoshizaki auger driver by a more powerful device in order to prevent jamming of the ice. The rotation momentum of the auger 4 was provided via a gearbox with gear-ratio 1:100. However, the selected driver operated at a constant angular velocity of 100 rpm. At these conditions, the size and temperature of particles were determined by the rate of the water supply to the port 8, which changed gradually from 0 to 200 ml/min. Water flow rate was precisely controlled by a special valve (Figure 2 (a)). In the final analysis, we were able to generate the desired kind of particles at given cooling conditions by the proper selection of the water flow rate. The attempts to improve ice production by

Development of a technology for generation of ice particles

141

Figure 2. (a) Schematic of auger type IJ system where: 1 – evaporator, 2 – refrigerant coils, 3 – insulation, 4 – auger, 5 – ice reloading device, 6 – air gun, 7 – air supply port, 8 – water supply port, 9 – cooling medium port, 10 – gauges, A – air flow rate valve, B – water flow rate valve, C – cooling medium valve, D – data acquisition card, and (b) schematic of gauges placement inside the evaporator with their distances measured from water inlet port (water zero level).

an increase of the water pressure or a control of the supply of the cooling medium were unsuccessful. Solidified ice plug moved forward along the auger helical ways. It was determined, however, that it was absolutely necessary to eliminate any obstructions to ice flow in order to prevent jamming of the heat exchanger. It is quite obvious that the conditions of the ice production (specific cost and energy consumption, process stability, uniformity of the generated particles, output per cm2 of the outlet cross-sectional area, etc.) will be dramatically improved by the process optimization. An unloading mechanism delivered ice particles to the abrasive port of the air gun. The nozzle block consisted of two nozzles and a special focusing device. Three different sizes of the nozzles were used, however, in all cases the nozzle to focusing tube ratio was 1:2. Ice was delivered to the nozzles abrasive port through insulated flexible plastic tubes. The ice crystallization was monitored through a set of thermocouples and resistance gauges (10) imbedded into the evaporator (Figure 2(b)). Data acquisition card processed the signals generated by the gauges. The large number of control variables (rate of supply of water and coolant, rate of auger rotation, diameter and length of the heat exchanger, distribution of temperature and heat flow along the refrigerator, geometry of the auger) practically excluded empirical process optimization. At the same time, the complexity of the

142

D.V. Shishkin et al.

Figure 3. Solid ice block decomposition at the two distinct stages of ice powder formation: a) initial stage of block formation, and b) the intermediate stage of block decomposition.

phenomena that occur in the course of the powder formation (water freezing and simultaneous decomposition in the turbulent layer) made it impossible to construct a physical model of the process. Thus the only practical approach to the process design entailed experimental study of the distribution of water properties within the icemaker, construction of the process phenomenology using the acquired information and then the use of commercial packages (Pro/ENGINEER, FIDAP, etc) to evaluate the design parameters of the icemaker as well as the correlation between input variables and the ice properties. Special experiments were carried out in order to examine the state of the solid ice in the course of solidification and particles formation. In order to attain this goal the auger was stopped during a normal operation and then removed from the heat exchanger. The solid phase was visually examined. The state of the ice is depicted in Figure 3. The anatomy of the ice in the course of freezing and particles formation was investigated using the set of pictures developed. This information was supplemented by the ice temperature and electrical resistance monitored prior to the auger removal. Extensive database acquired in the course of these experiments was used to develop a hypothetical mechanism of ice particles formation The suggested hypothetical process phenomenology is described below. 4. PHENOMENOLOGY OF ICE BLOCK DECOMPOSITION

Powder formation is a complex process involving solidification, stress development in the solid phase, supercooling of the solid with cracks propagation inside the block, concentration of the structural and thermal stresses within the solid and finally decomposition of ice blocks. According to our hypothesis ice block solidification occurs in the bottom evaporator zone. The formation of a solid block and its initial decomposition are depicted in Figure 3. At this stage the formation of

Development of a technology for generation of ice particles

143

the ice structure is complete and further cooling brings about only reduction of ice temperature. Thus particles size can be controlled only at the initial stages of the process. The following phenomenology of particles formation has been suggested. Ice nucleation inside the evaporator obeys the rules of granular ice formation (Ice type I). It originates as water flow reaches the evaporator wall with a temperature of –196°C. Ice nucleation and crystal growth depend on the conditions of cooling and the auger spiral movement. The ice plug forms and propagates along the auger stem. The generated ice acquires a multilayer pattern. A supercooled ice layer directly adjacent to the evaporator and a moderately cooled layer in the vicinity of auger can be identified. An explosive character of water freezing near the evaporator wall leads to thermal ice expansion inside the volume constrained by the evaporator and the auger and formation of the ice plug. This results in the compression of the solid and nucleation of cracks. The nucleation of cracks under a compressive stress is generally due to dislocation pile-up at the grain boundaries and relief of stress concentration along the grain boundaries. The phenomenon of crack nucleation has been discussed for low to moderate loading conditions by Sanderson [14]. His studies indicate that crack nucleation is determined by the delayed elastic strain. The generated cracks as well as the cracks initially developed in the solid ice propagate within the ice body and eventually fracture the solid block. Block decomposition occurs almost instantaneously and can be considered as an explosive destruction. The ice temperature in this region ranges from –15°C to –150°C (distance from water inlet level D = 10-24 mm). The following hypothetical mechanism of ice formation and decomposition (Figure 4) was suggested. Liquid water freezes and forms ice plug in the bottom zone of the evaporator. Thermal expansion essentially stops ice advance, and develops intensive stresses within the plug. This brings about formation and development of cracks and eventually fracturing the solid ice plug into particles. The density of cracks developed in the course of freezing determines the size of particles. This density, in turn, depends on the rate of water cooling. This explains the effect of the rate of water supply on the size of particles. At a low rate of water supply, i.e. at a high rate of freezing the fine particles are generated. The periodic mode of formation and decomposition of ice blocks determines the periodical character of driver current oscillations. Indeed, the ice powder exits the icemaker periodically and the periods of ice exit exactly coincide with those of the driver operation shown in Figure 5. Before plug formation water and ice flow freely along the auger and thus loading of the driver is minimal. As long as the plug stops the ice motion, the torque of the driver increases dramatically, but drops almost instantaneously when the plug is decomposed (Figure 5). After the plug decomposition, the geometry of particles does not change and the heat exchange results in ice cooling only.

144

D.V. Shishkin et al.

Figure 4. Hypothetical schematic of ice plug formation and decomposition in an auger spiral movement: (a) bottom or solidification zone of the evaporator, (b) middle or supercooling zone of the evaporator, (c) ice plug decomposition zone, and (d) ice particles cooling zone.

Figure 5. The correlation between the driver current with time for different water flow rates. Notice distinct frequencies of maximum current value for different water flow rates.

Development of a technology for generation of ice particles

145

5. CHARACTERIZATION OF GENERATED ICE PARTICLES

Our previous work showed the possibility of using ice particles for a wide variety of surface processing operations. However, it was found that each specific technology required a narrow range of particle sizes and temperature. Derusting of steel, for example, requires particles ranging from 3 to 7 mm, while the biomedical applications require highly homogeneous ice powder in the range 0.25 mm-0.3 mm.

Figure 6. Size distribution of ice particles for two different cooling media: (a) the average diameter of ice particles as a function of a temperature at the evaporator outlet, and (b) the average diameter of ice granules as a function of water flow rate.

146

D.V. Shishkin et al.

Figure 7. Pictures show the high and low density ice-airjet streams (air pressure is 0.544 MPa) and ice particles flow rate: (a) 20 g/min, and (b) 60 g/min.

The proposed mechanism of particles formation enables us to design a technology providing a desired kind of particles. In our experiments the control strategy was developed by properly selecting the cooling medium and water flow rate. The effects of the cooling medium (refrigerant Galden HT-55 vs. liquid nitrogen) on the particles size are shown in Figure 6 (a), while Figure 6 (b) shows the effect of water flow rate on this parameter. It must be pointed out that the exit particle temperature is also an important operational parameter, because it determines the stability and hardness of particles. The actual entrainment of stable low temperature particles is shown in Figure 7. The streams containing low and high concentrations of ice particles are depicted in this figure. 6. SELECTED APPLICATIONS OF ICE PARTICLES

A series of the experiments were carried out in order to demonstrate the potential application of the ice-air jet for various surface-processing operations. 6.1. Biomedical applications of ice-airjet (IAJ) technology The experiments were conducted on two distinct types of skin, the chicken skin and the pigskin. The paint was deposited on the skin (Figs. 8 a, b left) in question by a waterproof marker. Then the IAJ was used to remove this paint. The feasibility of the paint removal without damaging the underlying layers as well as a selective removal of the epidermis layer of the skin without damaging the underneath layers was demonstrated (Figs. 8 a, b right). The removal was performed without disturbing the skin structure as well as without creating a temperature gradient in the impingement zone.

Development of a technology for generation of ice particles

147

Figure 8. (a) Waterproof marker was partially removed from the epidermis layer of pork skin. Then the epidermis layer was removed too. No damage to the underneath layers was detected, and (b) waterproof marker was removed from the highly sensitive surface of chicken skin. No damage to the epidermis was observed in course of the cleaning procedure. The pictures were taken with a Sony MVC-FD71 digital camera. Note: the marker (a) was removed without damaging the skin epidermis layer (b).

6.2. Decontamination of heavily contaminated machine parts The deposit consisted of a mixture of dry grease and dust and moderately adhered to the substrate (Fig. 9 a left). A selected area of the part surface was decontaminated (Fig. 9 a right). The visual inspection confirmed the cleanness of the generated surface.

148

D.V. Shishkin et al.

Figure 9. (a) Decontamination of a heavily contaminated machine part, and (b) removal of highly adhesive glue layer from a plastic surface. The pictures were taken with a Sony MVC-FD71 digital camera. Note: the highly adherent layer of grease (Fig. 9 (a)) was removed without damaging the painted surface of the machine part (right hand side picture). The residue of glue (Fig. 9 (b)) was removed in course of IAJ cleaning. No damage to the plastic surface was observed (right hand side picture).

6.3. Removal of an highly adherent layer Two plastic discs were glued together with a highly adherent glue (Fig. 9 b left). Then the glue remaining on the plastic disc surface was removed by the IAJ (Fig. 9b right). No surface damage was found. Notice that it was not possible to remove this deposit using mechanical means. Another example involved removal of a thin layer of fresh rust formed from the steel surface. (Figs. 10 a and b).

Development of a technology for generation of ice particles

149

Figure 10. (a) Rusted carbon steel surface. Notice that newly formed rust layer is highly adherent, and (b) carbon steel plate was partially derusted using IAJ (middle part of the plate). The pictures were taken with a Sony MVC-FD71 digital camera.

150

D.V. Shishkin et al.

7. CONCLUSION

The suggested mechanism for the formation of ice powder enables us to design an effective industrial scale device for ice jet formation. The procedure for ice particles formation was developed and a device readily available for industrial deployment was constructed. An extensive use of this device is envisioned. Acknowledgement The study was supported by NSF grant number DDM9312980. REFERENCES 1. C. Schlosser, L. Muelle and G. McDougal, US Patent 5,752,39 (1950). 2. S. Hitoshi, Japanese Patent 10137707 A (1996). 3. J. Szijcs, European Patent 0509132B1 (1991). 4. M. Tomoji, Japanese Patent 04078477 (1990). 5. I. Harima, Japanese Patent 04360766 A (1992). 6. S. Vissisouk, US Patent 5,367,838 (1994). 7. T. Mesher, US Patent 5,607,478 (1997). 8. H. Shinichi, Japanese Patent 09225830 A (1997). 9. R. Niechial, US Patent 5,820,447 (1998). 10. G. Settles, US Patent 5,785,581 (1990). 11. B. Herb and S. Vissisouk, Proc. Precision Cleaning 1996 held in Anaheim CA, pp. 172-179 (1996). 12. B. Liu, in Jetting Technology, H. Louis (Ed.), pp. 203-211, Professional Engineering Publishing Ltd., London, UK (1998). 13. P. Hobbs, Ice Physics, Clarendon Press, Oxford (1974). 14. B. Sanderson, Ice Mechanics: Risks to Offshore Structures, Graham & Trotman, London, UK (1988).

Surface Contamination and Cleaning, Vol. 1, pp. 151–158 Ed. K.L. Mittal © VSP 2003

Cleaning with solid carbon dioxide pellet blasting FRED C. YOUNG∗ Cold Jet, Inc., 455 Wards Corner Road, Loveland, Ohio 45140, USA

Abstract—Blasting with solid carbon dioxide (dry ice) pellets is a technology that is gaining wide acceptance in industry and the military for removing coatings and contaminants from surfaces. Dry ice pellet blasting is also being used to prepare surfaces prior to applying coatings. This paper explains the principles of dry ice pellet blasting, and includes a brief history of the development of the technology. A discussion follows to explain how dry ice pellet blasting works by combining kinetic energy with thermal shock. The two kinds of dry ice pellet blast systems that are commercially available, the induction (venturi) and direct acceleration types, are also discussed. The operating principles and performance characteristics of both types are explained and compared in detail. Blasting system control parameters, such as dry ice pellet flow rate, compressed air propellant flow rate, and dry ice pellet density, are defined and their effects on cleaning performance are presented. To conclude, the author provides a review of actual applications for dry ice pellet blasting technology as it is currently used in the semiconductor manufacturing industry. Keywords: Dry ice; carbon dioxide pellets; blasting; cleaning.

1. INTRODUCTION TO DRY ICE PARTICLE BLAST CLEANING

Dry ice is made from liquid carbon dioxide, a recycled byproduct of several manufacturing processes. During the blasting process the dry ice sublimates to carbon dioxide gas, just like that exhaled by humans and found naturally in our atmosphere. Using dry ice is safe for employees and the environment. Dry Ice blasting uses extruded dry ice pellets, roughly the size of a grain of rice. The pellets are made on a dry ice extrusion machine called a “pelletizer” or “nuggetizer”. The pellets can be produced, stored in sealed insulated containers, shipped, and used for blasting several days after they are produced. Dry ice blasting accelerates solid carbon dioxide pellets with compressed air in a subsonic or supersonic blast stream to remove unwanted surface contaminants. Upon impact with the surface the dry ice sublimates (turns from a solid to gas without passing through a liquid phase) into carbon dioxide. The process is dry and non-conductive, non-abrasive and non-toxic, leaving no residue on the part or equipment being cleaned. All that remains to be collected (by vacuuming and/or ∗

Phone: 513-831-3211 ext. 367, Fax: 513-831-1209, E-mail: [email protected]

152

F.C. Young

filtration) is the surface contaminant being removed. Dry ice blasting leaves no secondary waste as from sand, bead or water blasting. This allows equipment surfaces to be cleaned in-place during the manufacturing process. The absence of a secondary waste stream makes dry ice blasting a perfect non-polluting technology. 2. THE PRINCIPLES BEHIND DRY ICE PARTICLE BLAST SURFACE CLEANING

With a low temperature of –79°C, dry ice (solid carbon dioxiode) has an inherent thermal energy ready to be tapped. In addition to the kinetic energy associated with any accelerated medium blasting, dry ice blasting uses the inherent low temperature to increase shear stress in the surface coating or contaminant, enabling the particle impact to break-up the coating. Further, the thermal gradient between two dissimilar materials (the contaminant and the substrate) with different thermal expansion coefficients can serve to break the bond between the two materials. The ability of these surface mechanisms to remove the coating or contaminant varies depending on coating or contaminant. Thermal shock is most evident when blasting a thin, non-metallic coating or contaminant bonded to a metallic substrate. Thermal shock, a key element that makes dry ice blasting an effective cleaning method, does not cause thermal stress in the substrate being cleaned. The temperature decrease caused by dry ice blasting is localized at the surface where the contaminant is bonded to the substrate. (See references [1, 2].) 3. A BRIEF HISTORY OF DRY ICE PARTICLE BLASTING TECHNOLOGY DEVELPOPMENT

In the early 1930’s, the manufacture of solid phase carbon dioxide (dry ice) became possible. During this time, the creation of dry ice was nothing more than a laboratory experiment. As the procedure for making dry ice became readily available, applications for this innovative substance grew. Obviously, the first use was in refrigeration. Today, dry ice is widely used in the food industry for packaging and protecting perishable foods. In 1945, stories exist of the U.S. Navy experimenting with dry ice as a blast medium for various degreasing applications. In May 1963, Reginald Lindall received a patent for a “method of removing meat from bone” using “jetted” carbon dioxide particles. In November 1972, Edwin Rice received a patent for a “method for the removal of unwanted portions of an article by spraying with high velocity dry ice particles”. Similarly, in August 1977, Calvin Fong (then working for the Lockheed Corp.) received a patent for “sandblasting with pellets of material capable of sublimation”. The work and success of these early pioneers led to the formation of several companies in the early 1980’s that pursued the development of dry ice blasting technology. (See reference [3].)

Cleaning with solid carbon dioxide pellet blasting

153

In recent years, dry ice pellet blasting has found its most commercially successful “market niche” for in-line (in-process) cleaning of molds (rubber, plastics, aluminum foundry, food baking). Other emerging areas of successful commercial application are with specialized contract cleaning services, the wood and paper industries, the semiconductor manufacturing industry, the printing industry, and the aerospace industry. 4. HOW CARBON DIOXIDE PARTICLE BLASTING WORKS (SEE REFERENCES [4–7])

4.1. Overview Carbon dioxide pellet blasting uses compressed air to accelerate frozen carbon dioxide “dry ice” pellets to a high velocity. A compressed air supply between 415 kPa and 620 kPa pressure is required. Dry ice pellets can be made on-site or supplied. The pellets are made from liquid carbon dioxide, which is a naturally occurring compound that is non-toxic, non-flammable and chemically inert. Carbon dioxide is inexpensive and easily stored at work sites. 4.2. The principal factors contributing to cleaning performance Carbon dioxide pellet blasting works because of two factors: pellet kinetic energy (velocity) and thermal shock (temperature). The performance of solid carbon dioxide blasting for surface cleaning is optimized by combining these factors and tuning the parameters of the system specifically for the application. These parameters are compressed air pressure, type of blast nozzle, pellet size and density, and the pellet flow rate. 4.2.1. Kinetic energy High pellet kinetic energy is achieved by using high velocity supersonic nozzles that are shaped properly to aim directly at the surface of the mold or other article being cleaned. The “single-hose – direct acceleration” type of carbon dioxide pellet- blasting system provides the very high kinetic energy required to remove the most tenacious contaminants from most surfaces. Solid carbon dioxide (dry ice) possesses virtually no “hardness” when compared to sand, glass beads, or even plastic beads. It is estimated that dry ice possesses a hardness between 1.0 and 1.5 on the Mohs scale. Lack of true hardness deprives dry ice of the “chiseling” effect which is the prevailing mechanism in all other forms of particle blasting. This also explains why dry ice blasting is considered NON-ABRASIVE to most substrates. Since dry ice cannot chisel and erode away the surface contaminant or coating, it must rely on extremely high initial kinetic impact energy to create very high instantaneous shear stresses in the coating layer. Dry ice particles are like tiny “snowballs” traveling at extremely high velocity, yet possessing no coefficient of restitution, so that ALL of each individual particle’s impact energy is completely absorbed by the coating layer. Excessive shear in the

154

F.C. Young

coating layer itself, and at the interface between the coating and the substrate, causes instant fracturing and separation of coating material from the surface. In general, if a coating or contaminant is CHEMICALLY bonded to a surface, dry ice particle blasting will NOT effectively remove the coating. If the bond is PHYSICAL or MECHANICAL in nature, such as a coating of rubber residue which is “anchored” into the porous surface of an aluminum casting, then there is a good chance that dry ice blasting will work. Contaminants which are etched, or stained into the surfaces of metals, ceramics, plastics, or other materials typically cannot be removed with dry ice blasting. If the surface of the substrate is extremely porous or rough, providing strong mechanical “anchoring” for the contaminant or coating, dry ice blasting may not be able to remove all of the coating, or the rate of removal may be too slow to allow dry ice blasting to be cost effective. The classic example of a contaminant that does NOT respond to dry ice blasting is RUST. Rust is both chemically and strongly mechanically bonded to steel substrate. Advanced stages of rust must be “chiseled” away with abrasive sand blasting. Only the thin film of powderized “flash” rust on a fresh steel surface can be effectively removed with dry ice blasting. 4.2.1.1. Induction (venturi) and direct acceleration blast systems – the effect of the type of system on available kinetic energy In a two-hose induction (venturi) carbon dioxide blasting system, the medium particles are moved from the hopper to the “gun” chamber by suction, where they drop to a very low velocity before being induced into the outflow of the nozzle by a large flow volume of compressed air. Some more advanced two-hose systems employ a small positive pressure to the pellet delivery hose. In any type of twohose system, since the blast medium particles have only a short distance in which to gain momentum and accelerate to the nozzle exit (usually only 200 to 300 mm), the final particle average velocity is limited to between 60 and 120 meters per second. So, in general, two-hose systems, although not so costly, are limited in their ability to deliver contaminant removal kinetic energy to the surface to be cleaned. When more blasting energy is required, these systems must be “boosted” at the expense of much more air volume required, and higher blast pressure is required as well, with much more nozzle back thrust, and very much more blast noise generated at the nozzle exit plane. The other type of solid carbon dioxide medium blasting system is like the “pressurized pot” abrasive blasting system common in the sand blasting and Plastic Media Blasting industries. These systems use a single delivery hose from the hopper to the “nozzle” applicator in which both the medium particles and the compressed air travel. These systems are more complex and a little more costly than the inductive two-hose systems, but the advantages gained greatly outweigh the extra initial expense. In a single-hose solid carbon dioxide particle blasting system, sometimes referred to as a “direct acceleration” system, the medium is introduced from the hopper into a single, pre-pressurized blast hose through a sealed airlock feeder. The particles begin their acceleration and velocity increase

Cleaning with solid carbon dioxide pellet blasting

155

immediately, and continue to gain momentum as they travel the length of the hose. At the end of the hose, the spray nozzle “gun” actually consists of a convergent-divergent nozzle, which exchanges pressure differential across the nozzle for a huge increase in air and particle velocity. Carbon dioxide particle velocities have been measured and substantiated in excess of 215 meters per second, and up to as high as 270 meters per second at the nozzle exit plane. This is accomplished at less than one third of the flow volume (only 3000 L/min compared to 10,000 or more L/min) required by the most aggressive two-hose systems. In addition to the lighter weight and less cumbersome hand held applicator and hose of a two-hose system, the contaminant removal energy delivered to the surface is considerably higher than that provided by a two-hose inductive system. Even with solid carbon dioxide particle blasting, a significant component of the contaminant removal energy is the kinetic energy per unit of area delivered to the surface. Since kinetic energy is a function of mass and velocity of the particles, i.e., Ke=1/2 mv2, it can be seen that a two-fold increase in particle velocity, with equal particle mass and equal nozzle spray area, effectively increases the impact energy delivered to the surface by a factor of four. A three-fold particle velocity increase, from 90 to 270 meters per second, increases the blast impact energy nine times that of a two-hose system. 4.2.2. Thermal shock Unlike other blast media, the carbon dioxide particles have a very low temperature of –78°C. This inherent low temperature imparts the dry ice blasting process with unique thermodynamically induced surface mechanisms that affect the coating or contaminant to a greater or lesser degree, depending on coating type. Because of the temperature differential between the dry ice particles and the surface being cleaned, a phenomenon known as thermal stress fracturing (“fracking”) or THERMAL SHOCK can occur. As the temperature differential between the coating and the substrate increases, the thermal shear stresses in the coating increase and couple with the impact induced stresses to increase the coating removal rate. A good example is the fouling which occurs on rubber, plastic, and tire curing molds. This contaminant is a chemical compound created by the interaction of mold release products and the base polymer under high pressure and temperature. The contaminant or “fouling” resembles a very thin glass-like material which responds very readily to the thermal shock effect of dry ice pellet blasting. In fact, hot molds, at or near the cure temperature of 160°C, can be cleaned three to four times faster than the same dirty molds at room temperature. 4.3. Cleaning performance control parameters Similar to abrasive grit blasting technology, dry ice pellet blasting performance, or “cleaning power”, can be adjusted to meet the needs of the application. The changeable performance control parameters allow the user to increase cleaning aggression, as for rubber mold cleaning, or decrease the level of aggression for more delicate applications, such as cleaning soldering flux from printed circuit boards.

156

F.C. Young

4.3.1. Pellet velocity (v) It can be increased or decreased by changing the blast air pressure or the type of nozzle selected. There are as many nozzles as there are applications, and they can be designed for sub-sonic, sonic, or super-sonic air flow and corresponding lower or higher pellet velocity. 4.3.2. Pellet size (m) It can be varied by using 1 mm, 2 mm, or 3 mm diameter extruded pellets. The pellets always break up into smaller particles as they travel through the blast hose and nozzle. The larger the pellets that you start with, the larger will be the particles which exit the nozzle and impact with the surface. Pellet size can also be varied by selecting a smooth bore or rough (convoluted) bore blast hose. The rough inside surface of convoluted hose can break up the larger 3 mm diameter pellets into very fine particle sizes. 4.3.3. Thermal shock It can always be enhanced by heating the substrate surface or the entire mass of the substrate. Rubber and tire molds, and baking oven molds are good examples of starting with a hot substrate. 4.3.4. Thermal shock and kinetic energy These can be varied also by adjusting the flow rate of the pellets in the blast stream. In the single-hose system, the radial airlock feeder speed can be precisely controlled to meter out just the right amount of pellets. Sometimes too much pellet flow can cool the coating and substrate too quickly, resulting in a performance drop. Sometimes a higher pellet flow is needed if the application requires more kinetic energy than the thermal effect, like removing heavily built-up oil, grease, and grime from machinery. 5. DRY ICE PARTICLE BLAST CLEANING APPLICATIONS IN THE SEMICONDUCTOR MANUFACTURING INDUSTRY

Dry Ice particle blasting is emerging as a method of choice for critical cleaning requirements in the semiconductor industry. The areas in the semiconductor manufacturing process where dry ice particle blasting is currently being applied are: (1) Silicon wafers are polished to a high degree of surface flatness in the early stages of the diffusion process. The polishing compound is deposited randomly on the internal surfaces of the polishing machines, then dries as a hard abrasive coating. Flecks of this abrasive contaminant can fall back onto the surface of newly inserted wafers, causing deep scratches that cannot be polished out, resulting in costly scrapping of silicon wafers. Dry ice particle blast has been found to be the best method to remove the abrasive, dried-on polishing slurry from the polishing equipment, without damaging the expensive polishing equipment components.

Cleaning with solid carbon dioxide pellet blasting

157

(2) Integrated circuits (ICs, or Chips) are produced in molds where the finished silicon based integrated circuit is encased in a material called EMC (Epoxy Mold Compound). The EMC eventually builds up a light deposit on the mold surface which can cause sticking of the finished ICs in the molds, and in other forms of IC surface defects. The traditional method used to remove EMC deposition is to coat the molds with melamine, let it cure, then pull the melamine off the surface. The melamine adhesively bonds to the EMC deposit. The EMC deposit is pulled off the mold surface together with the melamine as the melamine is removed. The major drawback is that the melamine tends to also pull the chromium plating off the mold surfaces, rendering the molds useless. Carbon dioxide particle blast has been demonstrated to be very effective in cleaning the IC mold surfaces without removing the chromium plating or otherwise damaging the mold surfaces. (3) Silicon wafers are photo-etched as part of the process to cut the wafers into the individual rectangular or square ICs, and as part of the process to produce the final surface transistor circuits. A compound called photoresist is deposited on silicon wafers to mask them in areas where the photo-etching action is not desired. After the photo-etching processes, the photoresist must be removed from the wafer surface. Carbon dioxide particle blast has shown very promising results in this area, and a great deal of developmental effort is now underway to bring this process into widespread use in the semiconductor industry. (4) A contamination problem arising from the etching process is the outgassing of compounds that redeposit on the surfaces of the etching equipment (e.g., fixtures, insides of chambers). This polymer-like deposit can re-contaminate subsequent wafers being etched in this high temperature process. Also, the deposit can build up so thick that the wafer holding fixtures become unusable. Some of the etching process fixtures are made of very expensive and delicate materials, like quartz. Traditional deposit removal methods include soaking in chemical baths with toxic solutions. Carbon dioxide particle blasting has been found to be very effective for removing the etching contaminants on quartz and some other substrates. 6. CONCLUSION

Dry ice pellet blasting is a surface cleaning and preparation technology that is gaining increasing popularity and acceptance each year, primarily as an industrial mold cleaning process. Industry’s acceptance of dry ice blasting is based on random trial-and-error testing by a few companies, for the purpose of verifying acceptability for their own cleaning applications, and the great willingness of the general industry to “assume” that the mechanics and physics behind dry ice blasting were well understood, documented, and as easy to apply to any given application (like a simple cookbook recipe). In fact, there is very little scientific basis (through investigation, testing, evaluation, and reporting) to support the assumptions behind the theories of how and why dry ice blasting works. Industry has ac-

158

F.C. Young

cepted dry ice blasting for mold cleaning because it has been proven to increase profitability and improve product quality. In the harsh environment of a typical foundry or rubber parts molding shop, dry ice blasting is a relatively benign mold cleaning method compared to sandblasting and other crude forms of abrasive grit blasting. More recently, other industries, such as semiconductor manufacturing, radioactive waste decontamination, and aerospace, are beginning to view dry ice blasting as an improvement over current methods of cleaning and surface preparation, or even as a potential “breakthrough” technology for use in developing completely new manufacturing or processing methods. Many of these new potential applications require the cleaning or preparation of very delicate surfaces, such as thin metal alloys, silicon wafers, composite materials, and even populated printed circuit boards. It is for these types of applications that a much better understanding of the dry ice blast contaminant removal phenomenon is required, so that much more precise control over the process, and predictability of the outcome of using the process, can be achieved. REFERENCES 1. L.C. Archibald, “Cold Jet Thermal and Surface Cleaning Characteristics” (June 1988, The Production Engineering Research Association of Great Britain, business name PERA). PERA is located in Melton Mobray, Leicestershire, UK. Telephone 44-664-501501. The information in this report is restricted to Cold Jet, Inc., 455 Wards Corner Road, Loveland 45140, USA. Please contact Cold Jet, Inc. at 513-831-3211 to obtain a copy of this report. 2. K. Lay, “An Analysis of Mold Integrity After Carbon Dioxide Blast Cleaning”, published in the proceedings of the International Tire Exhibition and Conference (ITEC), 1996, Akron, Ohio, USA. (Available by contacting Crain Communications, Inc. 1725 Merriman Road, Akron, Ohio 44313-5251. Telephone 330-836-9180. Fax 330-836-1005) 3. J.A. Snide, “Carbon Dioxide Pellet Cleaning - A Preliminary Evaluation”, Materials & Process Associates, Inc., October 12, 1992. 4. D.R. Linger, “Fundamentals of Dry Ice Blast Cleaning Technology”, published in the proceedings of the International Tire Exhibition and Conference (ITEC), 1996, Akron, Ohio, USA. (Available by contacting Crain Communications, Inc. 1725 Merriman Road, Akron, Ohio 44313-5251. Telephone 330-836-9180. Fax 330-836-1005) 5. C. Cundiff, “Evaluation of the Cold Jet, Inc. Carbon dioxide Blast System for Paint Stripping”, Battelle, 505 King Avenue, Columbus, Ohio 43201, USA, October 18, 1989. 6. F. Young, “Blast Off” article published in Tire Technology International Magazine, December, 2000. Pages 54–58. (Available by contacting UK & International Press, Abinger House, Church Street, Dorking, Surrey RH4 1DF, UK. Telephone +44 (0) 1306 743744. Fax +44 (0) 1306 742525. E-mail [email protected]) 7. F. Young, “Tire Mold Maintenance with Solid Carbon Dioxide Pellet Blasting”, published in the proceedings of the International Tire Exhibition and Conference (ITEC), 1998, Akron, Ohio, USA. (Available by contacting Crain Communications, Inc. 1725 Merriman Road, Akron, Ohio 44313-5251. Telephone 330-836-9180. Fax 330-836-1005)

Surface Contamination and Cleaning, Vol. 1, pp. 159–172 Ed. K.L. Mittal © VSP 2003

Development of a generic procedure for modeling of waterjet cleaning K. BABETS and E.S. GESKIN∗ New Jersey Institute of Technology, Mechanical Engineering Department, Waterjet Laboratory, Newark, NJ 07102-1982

Abstract—A practical procedure for utilization of available information, both numerical and linguistic and identification of the operational conditions of the waterjet cleaning is presented. Neural Networks based prediction models were constructed using previously available information. The constructed models constituted knowledge base for the procedure. Then a single parameter, the erosion strength for cleaning, was determined experimentally. The fuzzy logic technique enabled us to determine a weighted contribution of each preliminary constructed model for the process in question. Thus, the first approximations of the operational conditions are determined. In the course of the further operation the developed model is improved. The developed procedure will assist a practitioner in the selection of a decontamination technology for an unknown surface. Keywords: Waterjet; cleaning; soft computing; process prediction; process modeling.

1. INTRODUCTION

An effective material decontamination is one of the major industrial concerns today. It is difficult to imagine a single manufacturing process where material decontamination is not involved at some level. The field of material decontamination includes such vital applications as disinfecting and wound cleaning in hospitals and extends to road deicing, maintenance of building and bridges, paint stripping from aircrafts, and so on. Currently, the most usable approach for material decontamination involves chemical cleansers. Chemical cleansers are comparatively inexpensive, and in many cases they are readily available and are extremely effective. The problem with chemical cleansers is that they are potentially hazardous to worker’s health and are environmentally unfriendly. These and other problems with chemical cleansers (such as disposal of used agents, separation of debris

∗

To whom all correspondence should be addressed. Phone: (973) 596 3338, Fax: (973) 642-42882, E-mail: [email protected]

160

K. Babets and E.S. Geskin

from cleaning agents, etc.) require alternative methods for effective material decontamination based on physical coating removal techniques. Physical coating removal techniques take advantage of differences in physical properties between the coating and the substrate to destroy the bonding and/or abrade the coating from the underlying substrate. Physical coating removal techniques use one or more of four general types of physical mechanisms [1]. ● Abrasive techniques wear the coating off with scouring action. ● Impact techniques rely on particle impact to crack the coating to remove it. ● Cryogenic techniques use extreme cold conditions to make the coating more friable and induce differential contraction to debond the coating. ● Thermal techniques use heat input to oxidize, pyrolyze, and/or vaporize the coating. These techniques include but are not limited to: plastic media blasting, wheat starch blasting, sodium bicarbonate wet blasting, high pressure water blasting and cryogenic blasting. It is clear that the water blasting constitutes the most effective technique. Water is readily available, comparatively inexpensive, and induces no damage to the environment. A complete separation of water and debris facilitates material recovery. Therefore, complete pollution prevention is feasible. Although numerous extensive studies of waterjet-based material cleaning have been implemented [2-4], and this topic is currently of interest to many researchers, there is no one universal technique that will allow practitioners to bridge the gap between the available information about the process and the current need of a practitioner to remove a specific contaminant. This difficulty renders the process unusable for most practical applications. Therefore, a goal of this research was to develop a modeling tool that would assist in practical implementation of such a technology. 2. DEVELOPMENT OF A GENERIC MODELING TOOL

The experimental studies of material decontamination enabled us to identify the range of the application of waterjet technology for surface cleaning as well as to acquire a database for development of empirical modeling and optimization techniques. The theoretical study resulted in the development of corresponding algorithms and computer codes. However, the ultimate goal of the process investigation is to provide practitioners with an effective and practical approach for processing all information available to the practitioner, regardless of the form and accuracy. 2.1. Determination of the erosion strength In our work the following approach was used to obtain a generic coefficient to characterize any combination of substrata and deposits. Following the results

Development of a generic procedure for modeling of waterjet cleaning

161

available in the literature we define an area cleaning efficiency, Ea, as the ratio of area cleaned per unit time and power delivered by the nozzle: Ea =

Area cleaned per unit time &A = Power delivered by nozzle P

(1)

where area cleaning rate &A is in m2/hour, and

P = ∆p ⋅ Q

(2)

∆p – is the pressure drop across the nozzle, Q – is the flow rate; Thus Ea has the units of [m2/kW-h], that is a unit area cleaned per unit of energy expended by the nozzle. The idea of characterizing a material’s ability to resist erosion is far from new. Thiruvengadam [5] in his studies of cavitation erosion has suggested the notion of erosion strength that was based on a strain-energy absorption concept. Heymann [6] has suggested the concept of relative erosion strength. Thus utilizing ideas of Thiruvengadam [5] and Conn [7] relates the area cleaning rate, &A , the erosion strength for cleaning Sc and the erosive intensity I, for a given waterjet nozzle and fixed set of waterjet parameters (water pressure, traverse rate, angle of impingement, standoff distance, etc.,) as: where:

&A = I Sc

or

Ea =

I P ⋅ Sc

(3)

(4)

Combining expressions (1) and (4) results in

Ea ∝ S c − 1

(5)

The relation (5) is the basic relation, used to derive the curves, representing the dependence of the area cleaning efficiency Ea and erosion strength for cleaning Sc (Figure 1). 2.2. Determination of the erosion strength based on the available cleaning examples The available experimental database reflecting material decontamination with pure waterjet was compiled, and the area cleaning efficiencies were calculated for deposit types given in Table 1. The data in Table 1 were used to derive the relationships between area cleaning efficiency Ea and erosion strength for cleaning Sc

162

K. Babets and E.S. Geskin

Figure 1. Graphical relationship between Ea and Sc.

(Figure 1). The point of departure for the line 276-310 MPa in Figure 1 was the data for Item 1 from Table 1. The calculated cleaning efficiency for removal of Epoxy #1 deposit was assigned an Erosion Strength Sc=103, relative units. This line, per relation (5) was plotted on a log-log chart. The location of the rest of the data points was based on calculated area cleaning efficiencies and working water pressure. To derive line for 138 MPa (Figure 1) items 1 and 2 were compared. Since the Erosion Strength for the deposit type Epoxy #1 is known (Sc=1000 relative units) and is constant, the first point for line 138 MPa thus could be located. Similarly to derive the line for 70-100 mPa, item #6 was located at the 138 mPa line and compared with item #7. The main result that could be inferred from the graphical relationship between the area cleaning efficiency Ea and erosion strength Sc shows that there is a definite relationship between Ea and Sc and that Erosion Strength for cleaning of similar materials is closely spaced together, and consequently the Sc parameter can used to characterize an unknown deposit-substrate combination. On the other

Development of a generic procedure for modeling of waterjet cleaning

163

Table 1. Cleaning examples

Item #

Deposit type

Water pressure (MPa)

Nozzle diameter (mm)

Flow rate m3/s

Area cleaning rate (m2/hour)

Power delivered by nozzle kW-h

Area cleaning efficiencing m2/kW-h

1

Hard Epoxy Hard Epoxy Hard Epoxy Rust Weaker1 Rust Oil Based Oil Based Weaker Epoxy1 Auto Paint1

276

0.305

4.610E-05

0.52

12.71

0.0408

138

0.305

3.260E-05

0.04

4.50

0.0085

103

0.3556

3.837E-05

0.02

3.97

0.0048

310 310

0.254 0.1778

3.391E-05 1.662E-05

0.99 0.52

10.52 5.16

0.0938 0.1040

138 69 276

0.254 0.254 0.3554

2.262E-05 1.599E-05 6.259E-05

0.69 0.17 1.04

3.12 1.10 17.26

0.2201 0.1541 0.0603

138

0.254

2.262E-05

0.14

3.12

0.0434

2 3 4 5 6 7 8 9 1

Deposits were not used in model development.

hand, it should be emphasized that the relations in Figure 1, as presented by Conn [7] and verified by our experimental studies do not constitute exact relations, at best they represent an order of magnitude comparison only. 3. DEVELOPMENT OF GENERIC PREDICTION TECHNIQUE

The problem that most of the waterjet practitioners face when dealing with an unknown surface is the lack of information about the process or, in other words, the unavailability of a generic technique that could be used as a first approximation of the process. This section is concerned with the development of such an approach. The idea behind such an approach is to combine the previous knowledge about the process in question and based on that make an informed decision as to which waterjet parameters to apply, as a first approximation. 3.1. Modeling approach In data / information processing the objective is to gain the understanding of a complex phenomenon through modeling of the system either experimentally or analytically. Then after a model of the system has been obtained, various procedures (e.g. sensitivity analysis, statistical regression, etc.) can be used to gain a better understanding of the system.

164

K. Babets and E.S. Geskin

There are, however, situations in which the phenomena involved are very complex and not well understood and for which the first principle models are not effective. Even quite often, experimental measurements are difficult and/or expensive. These difficulties led us to explore the application of Soft Computing (Artificial Intelligence) techniques as a way of obtaining models based on experimental measurements. The field of Soft Computing is comparatively new, and it includes fuzzy logic, neural networks, expert system, cellular automata, chaotic systems, wavelets, complexity theory, anticipatory systems, among others. But only fuzzy logic, neural networks and genetic algorithms have reached the stage of development where they are used for real world problems. Fuzzy logic systems address the imprecision of the input and output variables directly by defining them with fuzzy sets (fuzzy numbers) that are generally expressed in linguistic terms. Moreover, they allow for very complex and nonlinear systems to be described in very simple terms, thus making them easier to understand. Another important feature of fuzzy systems is their ability to accommodate the existing expert knowledge of a process into a model by expressing it in terms of fuzzy rules. Neural Networks, on the other hand, model a system by using sets of inputoutput data to train some generic model of the system. Neural Networks are very good at modeling very complex nonlinear relationships with large numbers of input and output variables. Models based on neural networks are also easy to optimize, since although the model itself is not given in terms of an explicitly defined function, the gradient of this function can be found numerically. The combination of the above two techniques often results in greater flexibility and/or clearer representation of the model than when they are used separately. This combination is often referred to as neuro-fuzzy model of the system. NeuroFuzzy Reasoning approach also allows overcoming some traditional problems in using fuzzy logic or neural networks, such as the problem of defining a membership function, extracting fuzzy rules, etc. We are using the notion of Erosion Strength (Sc) developed in the previous section to classify an unknown surface together with Neural Networks Fuzzy Reasoning technique, suggested by Takagi and Hayashi [8], for information processing. The prediction technique construction begins with the development of fuzzy universe for erosion strength, Sc. The experimental database allows us to construct three fuzzy sets, based on the number of experimental situations available. The hard epoxy deposit with erosion strength of 1000 represents fuzzy Class I in Figure 2. Using the notation of the Fuzzy Logic theory we state that the hard epoxy deposit has the degree of membership of one in the fuzzy set Class I, or, in simple terms, this deposit is the most representative of all deposits that might be classified as belonging to Class I. Similarly, the rust deposit, with erosion strength of 400 is assigned the degree of membership of one in the fuzzy set Class II, and, finally, the oil based paint deposit is assigned the degree of membership of one in the fuzzy set Class III. To explain the idea of “degree of membership” we refer

Development of a generic procedure for modeling of waterjet cleaning

165

Figure 2. The fuzzy universe for erosion strength, Sc.

the reader to Figure 2. Carefully inspecting this figure we notice that the classes III and II, and II and I overlap to some degree. This means that some random deposit, with erosion strength of say 100, will belong to both class III and class II. The extent to which this random deposit type is represented by either of the classes is expressed in terms of the “degree of membership”. Thus, from Figure 2 we notice that a deposit with erosion strength of 100 belongs to Class III with a degree of membership of 0.8 and at the same time belongs to Class II with a degree of membership 0.2. The higher the degree of membership in a class, the more representative this class is for a given deposit type. Clearly, if a deposit has a degree of membership of 1.0 in some class, it belongs only to that particular class, and to no other class. Thus each of the classes in Figure 2 is represented (with a degree of membership 1) by a specific deposit type available in our experimental database. In other words, these three fuzzy sets cover the ranges of all possible values for the materials with erosion strength for cleaning from 1 to 10000 relative units. Similarly, if we identify a deposit-substrate combination with some value of erosion strength, Sc, and this value happens to be inside the range [1,10000], then we can identify the degree of membership of such a deposit-substrate combination in the three fuzzy sets (Figure 2). This procedure alone can be very useful when trying to classify some unknown deposit-substrate combination.

K. Babets and E.S. Geskin

166

It should be emphasized that for each of the three basic deposits (hard epoxy paint, rust, and oil based paint) an extensive experimental study was performed. From this point on we will refer to these deposits as “base deposits” and equivalently we will refer to the three fuzzy sets (classes) they represent as “base classes”. Since we possess a required empirical knowledge for these processes, the appropriate numerical representation of each process can now be made using artificial neural networks. The procedure for the application of a neural network for process modeling and optimization was described in [9]. Thus, after each network has been created, properly trained and tested we have obtained reliable numerical models for the three base deposits, or equivalently, three fuzzy base classes of erosion strength (Sc). Therefore, it can now be stated that the process of cleaning a material with erosion strength in the range [1,10000] can now be approximated as some combination of the numerical models of the three base classes. The computational procedure is as follows. For an unknown deposit, the practitioner makes a simple experiment that allows him/her to calculate area cleaning efficiency. Then using the computed Ea we can determine the corresponding erosion strength (Sc) for this surface from Figure 1, for the given water pressure. Once a corresponding Sc coefficient has been found, the fuzzy membership in the three classes in Figure 2 can be determined. Separate experiments were conducted for the auto paint deposit removal with plain waterjet. The computed area cleaning efficiency was calculated as Ea auto paint = 0.04 m2/kW-h. From Figure 1 the corresponding coefficient Sc was found to be Sc=180 relative units. And from Figure 2, the degree of membership (µ) in the three basic classes can be calculated as µ (class I) =0, µ (class II) =0.4, µ (class III) =0.59. These degrees of membership can be interpreted as follows. The erosion strength of the auto paint deposit is approximately midway between that of the hard epoxy paint and rust deposit. Now that the surface was identified we could use the numerical models available for the three base deposits (hard epoxy, rust, and oil based paint) to obtain a first approximation to the process. We supply a set of input parameters (Water Pressure, Nozzle Traverse Rate, Nozzle Diameter, and Standoff Distance) as an input into the numerical models represented by the neural networks. The corresponding output in terms of the single strip width is obtained by each of the network. The final result is obtained by defuzzifing the output according to Equation (6). r

* i

y =

åµ s =1

AS

(x i ) ⋅ u S (x i )

r

åµ s =1

AS

, i = 1,2 ... n

(6)

(x i )

In equation (6), µ is the membership value of the deposit with erosion strength Sc in the three base classes, Us is the output of the sth neural network, and y* is the final defuzzified output.

Development of a generic procedure for modeling of waterjet cleaning

167

Figure 3. Generic modeling approach. NN1, NN2, NN3 – Artificial Neural Network models for the three basic classes. X1, X2, …, Xn – Process input variables.

The output of the model is the single clean width of the strip produced on the surface by the combination of the input parameters, which then can easily be converted into area cleaning rate. This procedure is sketched in Figure 3. 4. EXPERIMENTAL VERIFICATION OF PERFORMANCE

In order to experimentally verify the suggested modeling approach an additional experimental database was acquired. The experimental samples consisted of three types of deposits – auto paint, weaker rust, and weaker epoxy paint, (items 5, 8, 9, Table 1). Waterjet parameters varied in these additional experiments were limited to the water pressure, nozzle traverse rate, standoff distance, and nozzle diameter. The experimental setup and procedures were similar to those described in the previous paper [10]. Area cleaning efficiencies for removal of these deposits were used to identify the corresponding erosion strength coefficient from Figure 1, and the degrees of membership of these deposits in the three basic classes were identified using Figure 2. Table 2 shows the results for the test deposits along with the deposits representing the basic classes.

K. Babets and E.S. Geskin

168 Table 2. Cleaning samples

Degree of membership Sample number

Deposit

Area cleaning efficiency (m2/kW-h)

1 2

Hard Epoxy Paint Weaker Epoxy Paint Rust from Steel Weaker Rust from Steel Auto Paint Oil Based Paint

0.04 0.06

3 4 5 6

Erosion strength (relative units) 1000 665

Class I

Class II

Class III

1 0.41

0 0.58

0 0

0.0975 0.105

400 360

0 0

1 0.89

0 0.1

0.04 0.21

180 30

0 0

0.4 0

0.59 1

The model performance was tested on each of the test deposits by providing the model with a set of waterjet input parameters within the working space, obtaining the corresponding output in terms of the width of a clean strip and comparing the results with experiments. 5. DISCUSSION

The model for prediction of the results of waterjet cleaning described in the previous sections was tested on several additional test deposits. Figures 4-9 present the results of prediction. Analyzing the results, it is clear that the model prediction results are acceptable at both relative error of prediction (~ 20%), and at following the trend of the process, which is also important for any cleaning study. Nevertheless, as a first estimation of a cleaning efficiency for a given type of deposit, these results constitute a reasonable approximation. However, it should be noted that at the current stage the prediction technique was tested only in the middle of the problem space. At the outskirts of the problem space, veritable results could not be obtained. The reason for this lies in the limitations in the development of the three numerical models that represent the cleaning of the base deposits (hard epoxy, rust, and oil based paint). Since the ranges of experimental parameters used for the construction of models were different in each case, and there was no coordinated experimental setup, but rather the data were compiled at later stages, there are inconsistencies in choosing the levels of process parameters in case of a test cleaning space. For example, if for a base model development the nozzle traverse rate was in the range from 1000 mm/min to 2500 mm/min, and a test case was run at 1500~4000 mm/min, the reliable model performance will be at the intersection of these ranges. The way to

Development of a generic procedure for modeling of waterjet cleaning

169

cure the above problem is to cover all the parameter space (limited by equipment capabilities) for each process variable in all the base models. Of course, this results in quite extensive experimentation, but on the good side it needs to be done only once, when developing the base models. Also the current model does not cover the full range of all possible erosion strengths of different materials, but by extending the procedure with additional base models for lower or higher degrees of erosion strength for cleaning (Sc), this limitation can be reduced or eliminated.

Figure 4. Auto paint removal, standoff distance 13.9 cm. WP – water pressure, ND – nozzle diameter, standoff – stand off distance.

Figure 5. Auto paint removal, width of strip vs. traverse rate. Experimental vs. predicted.

170

K. Babets and E.S. Geskin

Figure 6. Removal of weaker rust. Experimental vs. predicted width of clean strip. Water pressure 241 MPa, nozzle diameter 0.1778 mm.

Figure 7. Experimental vs. predicted width of clean strip. Water pressure 172 MPa, nozzle diameter 0.1778 mm.

6. CONCLUDING REMARKS

The approach presented here for modeling of waterjet cleaning process allows a user to obtain a reliable process approximation given no or limited information about process condition. For an unknown surface a practitioner needs to determine a single coefficient, the erosion strength for cleaning (Sc), based on a simple experiment(s). The proposed approach utilizes this coefficient and approximates

Development of a generic procedure for modeling of waterjet cleaning

171

Figure 8. Removal of weaker epoxy paint. Nozzle diameter 0.356 mm, water pressure 207 MPa.

Figure 9. Removal of weaker epoxy paint. Nozzle diameter 0.3556 mm, water pressure 276 MPa.

the cleaning results in terms of area cleaning rate. It is believed that the current work will assist in practical implementation of waterjet cleaning technology, where the information deficiency on process conditions is the main reason for ineffective application of this technology.

172

K. Babets and E.S. Geskin

REFERENCES 1. United States Environmental Protection Agency, Office of Research and Development. Guide to Cleaner Technologies, EPA/625/R-93/015,Washington D.C. (1994). 2. A.F. Conn and G. Chahine, Proc. Third American Waterjet Conference, Pittsburgh, PA, Waterjet Technology Association, St. Louis, MO (1985). 3. F. Erdman-Jesnitzer, A.M. Hassan and H. Louis, Proc. Third International Symposium on Jet Cutting Technology, Chicago, IL, British Hydraulic Research Association, Cranfield, UK (1976). 4. S.T. Johnson, Proc. 7th American Water Jet Conference, Seattle, Washington, Waterjet Technology Association, St. Louis, MO (1993). 5. A. Thiruvengadam, in Erosion by Cavitation or Impingement, STP No. 408, 22-36, ASTM, Philadelphia, PA (1966). 6. F. Heymann, in Characterization and Determination of Erosion Resistance, STP No. 474, 212244, ASTM, Philadelphia, PA (1969). 7. A. Conn, Proc. Fourth American Waterjet Conference, Berkeley, CA, Waterjet Technology Association, St. Louis, MO (1987). 8. H. Takagi and I. Hayashi, Intl. J. Approximate Reasoning, 5, No. 3, 191-212 (1991). 9. K. Babets, E.S. Geskin and B. Chaudhuri, Proc. 10th American Waterjet Conference, Houston, TX, Waterjet Technology Association, St. Louis, MO (1999). 10. K. Babets, E.S. Geskin, Intl. J. Machining Sci. Technol., 4, No. 1, 81-101 (2000).

Surface Contamination and Cleaning, Vol. 1, pp. 173–191 Ed. K.L. Mittal © VSP 2003

Experimental and numerical investigation of waterjet derusting technology K. BABETS, E.S. GESKIN∗ and B. GOLDENBERG New Jersey Institute of Technology, Mechanical Engineering Department, Waterjet Laboratory, Newark, NJ 07102-1982

Abstract—The study is concerned with the development of effective technology for derusting of a steel surface. We have investigated the surface derusting by high-speed waterjet and determined the optimal operational conditions. This investigation involved topographical and metallographical studies of the substrate surfaces and subsequent classification of the substrates with respect to the degree of rust development. Then the rust was removed by a moving waterjet at various impact conditions and the generated surfaces were examined. Soft computing techniques were used to select the optimal conditions for rust removal. Due to the extremely chaotic and fuzzy nature of input information the advanced numerical procedure based on the Neural Network Driven Fuzzy Reasoning was employed. As the result, the realistic procedure for steel derusting was found and a practical technique for process design was suggested. Keywords: Derusting; fuzzy reasoning; neural network; soft computing; waterjet.

1. INTRODUCTION

The corrosion of metal structures poses a serious technological and economical problem. It shortens the life span of the steel parts and deteriorates dramatically their performance. Corrosion is a chemical or electrochemical process in which surface atoms of a solid metal either react with or dissolve in a substance that contacts the exposed surface. Corroding media are generally classified as aqueous or non-aqueous. The rate of steel corrosion in the atmosphere depends on geographical location, and can reach 1070 µm/yr. When rust depth reaches 1% of the thickness of the steel, the strength of the steel reduces by 5-10%. Throughout the world steel corrosion annually equals to 20-40% of its annual production [1]. The corrosion of all carbon steels is most devastating when the metal is subjected to an alternately wet and dry atmosphere in the presence of chloride salts. Typically this environment can be encountered on the underbodies of automobiles and trucks. The most con∗

To whom all correspondence should be addressed. Phone: (973) 596 3338, Fax: (973) 642 4282, E-mail: [email protected]

174

K. Babets et al.

ventional way of rust prevention is to coat the metal surface. But prior to the coating, a complete cleanliness of the surface must be assured to ensure that a surface is free of any rust. The conventional methods of rust removal involve acid cleaning or sandblasting. These techniques, though proven to be effective, can be environmentally hazardous. Some new derusting techniques, such as use of rust neutralizers converts rust into a chemically neutral surface, leaving the surface ready for coating application. However, the use of these neutralizers is limited to oil-based types only, and the chemicals contained in these products can be detrimental to the worker’s health. Waterjet surface derusting constitutes rather an efficient way to clean steel surfaces. The following experimental study was concerned with optimization, or at least improvement, of jet based derusting technology. 2. EXPERIMENTAL SETUP

2.1. Experimental procedure The derusting experiments were carried out at the Ingersoll-Rand waterjet system (Fig. 1). The nozzle head was mounted on a 3-axes gantry robot whose movements were guided by an Allen Bradley 8200 series CNC controller. The major obstacle in the experimental study of a derusting technology is the extreme diversity of the rusted surfaces. It is difficult to find several samples with

Figure 1. Waterjet setup.

Experimental and numerical investigation

175

a similar degree of rusting. As a result, a rather low reproducibility of the experiment is obtained. In order to at least partially overcome this problem we used the ISO developed standard for classification of the rusted steel surfaces. This standard specifies four rust grades, designated A, B, C and D [2]. The rust grades are defined by written descriptions together with representative photographic examples. The selected steel samples were sorted according to visual similarity and compared to the representative photographs. Those identified as the class C were used as experimental samples. In the above-mentioned ISO publication, C is defined as the “steel surface on which the mill scale has rusted away or from which it can be scraped, but with slight pitting visible under normal vision”. Still it should be stressed that the existent rust grades classification based on visual comparison does not provide a reliable procedure for rust identification and thus makes it quite difficult to collect uniform experimental samples. In our experiments the effects of water pressure, traverse rate and nozzle diameter on cleaning effectiveness and surface quality were investigated. The tests were run at water pressures of 310, 241, 172, 69 MPa (i.e. 45,000, 35,000, 25,000 and 10,000 psi). The water nozzles with diameters 0.127, 0.1778, 0.254, 0.3556 mm were used. In these experiments the effect of the standoff distance (the distance between the nozzle exit and the sample) as an independent process parameter was not investigated. Instead, the ratio of nozzle diameter to the nozzle standoff distance was kept constant. Thus a number of standoff distances were tested to find a near optimum value for a selected nozzle diameter. Then the obtained ratio was kept constant for the other nozzle diameters. The study was carried out at the traverse rates of 635, 2540, 7620 and 12700 mm/min. The upper bound of the nozzle traverse rate (12700 mm/min) was imposed by the equipment limitations. In order to study the effect of the waterjet parameters on the surface, a full factorial experimental design was employed. In such a design one process variable is tested at its different levels, while the other variables are held fixed at some level. The experimental procedure involved the following steps. For each cleaning situation (i.e. the combination of water pressure, nozzle traverse rate and nozzle diameter / standoff) the width of clean strip was measured with Mitutoyo Toolmakers Microscope and recorded. These values of strip width (STW) were then used to calculate process effectiveness (Rate of Area Cleaned) and specific water consumption according to the following expressions: Rate of Area Cleaned (m2 min–1) = Traverse Rate * Width of Cleaned Strip

(

C D ⋅π ⋅ D 2 ⋅ 2⋅ P

)

1

2

ρw Water Consumption m 3 / m 2 = 4⋅ Rate of Area Cleaned

(

)

(1)

(2)

where ρw is the water density, CD is the discharge coefficient of the waterjet orifice, whose diameter is D. In present work CD is taken to be 0.7 [3], P – waterjet pressure.

176

K. Babets et al.

Then several nozzle passes with 25% overlapping were made at the same operational conditions. The resulting derusted area was evaluated visually and photographed with Olympus photomicroscope with 12 x magnification in order to determine the degree of cleaning. 2.2. Surface examination In current study, the X-ray diffraction was used to evaluate the presence of oxides (rust) on the metal surface after rust removal with waterjet. The Siemens D5000 diffractometer with θ-2θ diffractometer geometry at the Stevense Institute of Technology was used for this investigation. The experimental samples consisted of rusted-cleaned pairs. The following procedure was used for sample preparation. First the metal samples were machined to a block 10.16 x 10.16 x 6.350 mm. Then the samples with similar rust were grouped in pairs. From each pair one sample was left as it was, and the other one was cleaned of rust using waterjet. The following waterjet parameters were employed: Water Pressure 200 MPa, Nozzle Diameter 0.254 mm. Two cleaning runs were made. At first the rust was removed from the metal surface at a low nozzle traverse rate, and in the second run the flash rust was removed at the high traverse rate of 3175 mm/min. Then the sample was dried in hot air. Each pair was evaluated for the presence of oxides by the diffractometer. The resulted diffraction patterns enabled us to compare the oxides content on the samples before and after the waterjet treatment. 3. EXPERIMENTAL RESULTS

3.1. Surface classification The quality of derusting by waterjet was evaluated in accordance with ISO standards (ISO 8501-1:1988). This standard defines four grades of cleanliness of the surfaces generated by jet derusting. These surfaces are termed Sa 1, Sa 2, Sa 2 ½ and Sa 3. The qualitative description of each grade along with the representative photographs of the surfaces are presented. During this experimental study it was found difficult to follow the ISO classification. Instead, the following “fuzzy” classification was suggested. According to the developed procedure we divided the derusted surfaces into two classes: “well cleaned”, and “poorly cleaned”. A surface is allowed to have a partial degree of membership in both classes. The class “well cleaned” would roughly correspond to ISO grades Sa 3 and Sa 2.5, while the class “poorly cleaned” would correspond to surface grades Sa 2.5, Sa 2 and Sa 1. Figure 2 depicts a typical well-cleaned surface. Here two shades of green can be distinguished. Light green corresponds to the derusted surface, while dark green corresponds to flash rust, which appears immediately after waterjet pass. It was found that flash rust could be removed easily by an additional application of the waterjet at a high traverse rate, or prevented by immediate drying of the surface in hot air.

Experimental and numerical investigation

177

Figure 3 shows a typical “poorly cleaned” surface. Such a surface is free from lightly adherent mill scale, rust and other contamination, but some firmly adherent rust remains on the surface. Thus the most representative surface samples were classified as either belonging to one of these classes, or, not. The “fuzzy” memberships in the two fuzzy classes for the remaining surfaces were determined using the artificial neural network assisted fuzzy classification method described in the following section.

Figure 2. Optical photograph of “well cleaned” metal surface (12 x magnification).

Figure 3. Optical photograph of “poorly cleaned” surface (12 x magnification).

178

K. Babets et al.

3.2. Surface cleanliness In order to evaluate qualitatively and quantitatively the waterjet-based derusting technology, several studies were carried out. These examinations included taking scanning electron micrographs of the surface, performing chemical analysis of the metal surface, and performing the x-ray diffraction analysis. Figure 4 shows a rusted surface at 500 x magnification. The main features of the surface are the oxidized metal grains of different sizes. Figure 5 represents a surface derusted with waterjet. No oxidized metal grains are observed; the surface is smooth and visually free of rust.

Figure 4. SEM micrograph of rust covered metal surface.

Figure 5. SEM micrograph of waterjet-derusted metal surface.

Experimental and numerical investigation

179

In order to estimate qualitatively the effectiveness of derusting, chemical analysis of the surface prior to and after treatment was carried out. The scanning electron microscopy was used for this study. The typical results of the analysis are presented in Figs. 6-7. These figures show that the oxygen content of the surface

Figure 6. Chemical composition of metal surface prior to waterjet rust removal. Oxygen content is at 900 count.

Figure 7. Chemical composition of metal surface after waterjet rust removal. Oxygen content is at 320 count.

180

K. Babets et al.

was significantly reduced after the water jet treatment. In order to evaluate the degree of derusting, the chemical analysis was supplemented by the x-ray diffraction. The corresponding diffraction patterns of the metal surfaces before and after waterjet cleaning are presented in Figs. 8 and 9, respectively. Roughly speaking each peak in these figures corresponds to a chemical compound. The intensity of

Figure 8. X-ray diffraction analysis. Diffraction pattern of rusted metal surface prior to waterjet treatment.

Figure 9. X-ray diffraction analysis. Diffraction pattern of metal surface after waterjet rust removal.

Experimental and numerical investigation

181

Table 1. Results of the metal surface chemical analysis Experimental atomic planes spacing values 0

Standard tables of atomic planes spacing values 0

d ( A)

d ( A)

3.281 3.009 2.795 2.546 2.174 2.106 2.027 1.723 1.624 1.49 1.437 1.289 1.211 1.168

3.24 2.967 2.728 2.532 2.176 2.099 2.03 1.715 1.616 1.485 1.43 1.281 1.212 1.17

Chemical compound

Fe

Fe2O3

Fe3O4

YES YES YES YES YES YES YES YES YES YES YES YES YES YES

a peak represents the relative amount of this chemical compound. The atomic spacing values shown just above these peaks allow us to determine the type of the chemical compound present on the surface. From Fig. 8 and Table 1 it is clear that a rusted surface in addition to Fe contains significant amounts of Fe2O3, and Fe3O4. After waterjet cleaning (Fig. 9) the three still remaining peaks represent Fe, with significantly increased intensity levels. Most of the oxides are no longer present in the figure, and intensity level of those still present is much lower than that of Fe content. Moreover, due to low intensity levels these peaks most probably should be attributed to the noise. The wide base peak at angles 10-25 degrees in Figs 8 and 9 is due to the presence of the holder clay used to attach the sample in the holder. Thus, Fig. 9 constitutes a compelling proof of the efficiency of the waterjet rust removal. 3.3. Effect of water pressure For each set of operational conditions there is a minimal threshold pressure below which decoating does not occur. This pressure level depends on the adhesion strength between the coating and the substrate [8]. The maximum working water pressure is defined from damage-free cleaning considerations, i.e., where the cleaning does not result in the damage to the material. In our experiments these upper and lower pressure bounds were dictated primarily by the equipment capa-

182

K. Babets et al.

bilities, since at the chosen levels of the nozzle traverse rate and nozzle diameter the removal of rust could still be performed. Between these two threshold values of water pressure the clean strip width obtained in a single nozzle pass does not vary linearly with increasing water pressure (Figure 10). The effectiveness of the process increases with increasing water pressure (Figure 11).

Figure 10. Experimental clean strip width vs. water pressure for nozzle diameter 0.3556 mm at different nozzle traverse rates.

Figure 11. Effectiveness (rate of area cleaned) vs. water pressure for nozzle diameter 0.3556 mm at different nozzle traverse rates.

Experimental and numerical investigation

183

Figure 12. Water consumption vs. water pressure for nozzle diameter 0.3556 mm at different nozzle traverse rates.

Still the relationship between the effectiveness and water pressure can be approximated as linear at lower values of the traverse rate. At high nozzle traverse rates and high nozzle diameters the relationship between the process effectiveness and water pressure can no longer be considered as linear, but rather as a polynomial. Figure 12 shows that for a large nozzle diameter there exists an extremum of the water consumption as the pressure increases. This can be attributed to the fact that the process effectiveness is not a linear function of water pressure (at least in the considered range of process variables). At small nozzle diameters and for the pressure range used in these experiments no extremum is seen, although it is reasonable to expect that the extremum will appear at higher values of water pressure. 3.4. Effect of traverse rate The effect of the traverse rate on rust removal appears to be the most significant. Figure 13 shows that, as expected, the process effectiveness increases with increasing nozzle traverse rate, while the strip width decreases (Fig. 14). The specific water consumption (Fig. 15) can be approximated as a power function of traverse rate. This actually means that there is a range of traverse rates when the increase in traverse rate results in significant drop in the specific water consumption, while the larger increase in the traverse rate insignificantly reduces water consumption. 3.5. Effect of nozzle diameter As expected the increase in nozzle diameter resulted in a higher process effectiveness (Fig. 16), but also in a higher specific water consumption (Fig. 17). It is interesting to follow the relationship between these important quantities. Let us

184

K. Babets et al.

Figure 13. Rate of area cleaned vs. traverse rate for nozzle diameter 0.3556 mm and for different water pressures.

Figure 14. Clean strip width vs. traverse rate for nozzle diameter 0.3556 mm and different water pressures (69E – 69 MPa, 172E – 172 MPa, 241E – 241 MPa, 310E – 310 MPa).

consider the experimental results for water pressure 69 MPa, traverse rate 12700 mm/min and nozzle diameters 0.1778 mm (0.007 in) and 0.254 mm (0.01 in). Here we notice that the area of the second nozzle is almost twice the area of the first nozzle. The calculated process effectiveness is 0.39 m2/hour and 0.52 m2/hour, respectively, with specific water consumption 0.064 m3/m2 and 0.094 m3/m2. If we now take two small nozzles then the total effectiveness will be 0.78 m2/hour while the water consumption will stay at 0.064 m3/m2. Thus it appears that it would be beneficial to use several small nozzles rather than a big one. This result is important from practical point of view.

Experimental and numerical investigation

185

Figure 15. Water consumption vs. traverse rate for nozzle diameter 0.3556 mm and for different water pressures.

Figure 16. Rate of area cleaned vs. nozzle diameter for water pressure 241 MPa at different nozzle traverse rates.

Figure 17. Water consumption vs. nozzle diameter for water pressure 69 MPa and at different nozzle traverse rates.

186

K. Babets et al.

4. MODEL OF THE PROCESS

4.1. Choice of modeling technique In data/information processing the objective is to gain an understanding of a complex phenomenon through “modeling” of the system either experimentally or analytically. Then after a model of the system has been obtained, various procedures (e.g. sensitivity analysis, statistical regression, etc.) can be used to gain a better understanding of the system. There are, however, situations in which the phenomena involved are very complex and not well understood and for which the first principle models are not effective. Even more often, experimental measurements are difficult and/or expensive. These difficulties led us to explore the application of Soft Computing (Artificial Intelligence) techniques as a way of developing models based on experimental measurements. The field of Soft Computing is comparatively new, and it includes fuzzy logic, neural networks, expert system, cellular automata, chaotic systems, wavelets, complexity theory, anticipatory systems, among others. But only fuzzy logic, neural networks and genetic algorithms have reached the stage of development where they are used for real world problems [4]. Fuzzy logic systems address the imprecision of the input and output variables directly by defining them with fuzzy sets (fuzzy numbers), which generally are expressed in linguistic terms. Moreover, they allow for very complex and nonlinear systems to be described in very simple terms, thus making them easier to understand. Another important feature of fuzzy systems is their ability to accommodate the existing expert knowledge of a process into a model by expressing it in terms of fuzzy rules. Neural Networks, on the other hand, model a system by using sets of inputoutput data to train some generic model of a system. Neural Networks are very good at modeling very complex nonlinear relationships with large numbers of input and output variables, and in classification problems. Models based on neural networks are also easy to optimize, since although the model itself is not given in terms of on explicitly defined function, the gradient of this function can be found numerically. The combination of the above two techniques often results in greater flexibility and/or clearer representation of a model than when they are used separately. This combination is often referred to as neuro-fuzzy model of a system. Neuro-fuzzy approach also allows overcoming some traditional problems in using fuzzy logic or neural networks, such as the problem of defining a membership function, extracting fuzzy rules, etc. Our problem at hand is a good example of a system with highly nonlinear relationship between process inputs and outputs. The problem of defining the degree of cleaning is one of the classification types. Therefore, it was found reasonable to apply an advance artificial intelligence modeling technique based on the combination of fuzzy logic and artificial neural networks. The method used is known as NN-Driven Fuzzy Reasoning [5], and was used with only slight modifications.

Experimental and numerical investigation

187

4.2. Model of the process The fuzzy classification of derusted surfaces contained two classes: “well cleaned”, and “poorly cleaned”. Some of the cleaned samples were assigned to “well cleaned” with degree of membership 1; correspondingly, they had the degree of membership 0 in “poorly cleaned”. The rest of the samples had a non-zero degree of membership in both classes. We used a special neural network to determine these degrees. The procedure involved training the neural network (NNmem), using only well-defined samples, i.e., samples having the degree of membership 1 in either class. Then, after being properly trained, such a network will not only be able to predict the binary degree of membership (either 0 or 1) for some input data, but also the dual degree of membership (fuzzy membership) for the input data points in that neighborhood. As the result of training, we obtain neural network which is able to determine the degree of membership in each of the two classes using input conditions, such as water pressure, traverse rate, etc. This procedure is described in details in Takagi and Hayashi [5] and by Ross [6]. Our actual goal, however, was to determine the process effectiveness and the resultant degree of cleanliness. In order to reach this goal we divided the available database into two data sets. The first data set contained only the data identified earlier to clearly (i.e. with degree of membership 1) belonging to the class “well cleaned”, and, similarly the second data set contained data belonging only to the class “poorly cleaned”. We then trained two separate networks, on these two data sets, and as a the result each network was able to determine effectiveness for the class it was responsible for. From this, we obtained the model of the process in terms of the three trained neural networks that were connected according to Fig. 18. According to this figure, given some input data set, the network NNmem identifies the degree of membership of a sample in each of the two classes. Network NN1 predicts productivity for class “well cleaned”, and NN2 for class “poorly cleaned”. Input information is fed to all three networks. The outputs of all three networks are then fed into special elements which process the networks outputs to determine the weighted sum and as a result predict final process effectiveness and the degree of cleanliness. Thus, we were able to obtain an accurate prediction of the process effectiveness (the average error in prediction was within 8%). Also we were able to estimate the quality of derusted surface, based on the fuzzy degrees of membership in “well cleaned’ and “poorly cleaned” classes inferred by the NNmem neural network. The results of the prediction are presented in Figs. 19-22. These figures show the process effectiveness as a function of different process parameters, without regard to the quality of resultant surface. The quality (degree of membership in two classes) for any data point in these figures is obtained using the neural network NNmem.

188

K. Babets et al.

Figure 18. Computational approach. (I1, I2, and I3 are the input parameters (water pressure, traverse rate, and nozzle diameter). NNmem is the neural network that decides the membership values (w1, w2) in each class, of the above input parameters. NN1 and NN2 determine the outputs (rate of area cleaned) y for each class.)

Figure 19. Model prediction results for water pressure 310 MPa at different nozzle traverse rates.

Experimental and numerical investigation

189

Figure 20. Model prediction results for water pressure 241 MPa at different nozzle diameters.

Figure 21. Model prediction results for nozzle diameter 0.3556 mm at different nozzle traverse rates.

190

K. Babets et al.

Figure 22. Model prediction results for nozzle diameter 0.127 mm at different nozzle traverse rates.

5. CONCLUDINGS REMARKS

We have demonstrated the feasibility and, in fact, effectiveness of steel derusting by waterjet. The effect of the operational conditions on the process was evaluated and a set of operational conditions was suggested. This set can be used as an initial state by a practitioner to search for optimum operational conditions. Because of wide variations in the states of the rusted surfaces and insufficiency of the available identification technique, an advanced soft computing procedure (neural network driven fuzzy identification) has been suggested for surface identification. The analysis of the results of derusting conditions demonstrates the effectiveness of the use of several nozzles rather than a single nozzle of the same surface area. Acknowledgements The surface examination was carried out at the Stevens Institute of Technology. The valuable advice of Josef Karagotskiy of the Electronic Microscopy Center at the Stevens Institute of Technology is gladly acknowledged. This work was partially supported by NSF grant # DDM-9312980. REFERENCES 1. B. Liu, B. Jia, D. Zhang, C. Wang, H. Li and H. Yao, Proc. 7th American Water Jet Conference, Seattle, Washington, Waterjet Technology Association, St. Louis, MO (1993). 2. International Standard ISO 8501-1:1998, Third Edition (1999).

Experimental and numerical investigation

191

3. M. Hashish, Proc. 7th American Water Jet Conference, Seattle, Washington, Waterjet Technology Association, St. Louis, MO (1993). 4. Z. Michalewicz, Genetic Algorithms + Data Structures = Evolution Programs, Third Edition, Springer (1996). 5. H. Takagi and I. Hayashi, Intl. J. Approximate Reasoning, 5, No. 3, 191–212 (1991). 6. T. Ross, Fuzzy Logic With Engineering Applications, McGraw-Hill (1995). 7. B.D. Cullity, Elements of X-ray Diffraction, Addison-Wesley Publishing Company (1978). 8. H. Jun, Proc. 7th American Water Jet Conference, Seattle, Washington, Waterjet Technology Association, St. Louis, MO (1993). 9. P. Singh, J. Munoz and W. Chen, Proc. of 11th International Symposium on Jet Cutting Technology, British Hydraulic Research Group, Dordrecht, The Netherlands (1992). 10. X. Shegxiong, H. Wangping and Z. Sheng, Proc. 7th American Water Jet Conference, Seattle, Washington, Waterjet Technology Association, St. Louis, MO (1993). 11. C. Suryanarayana and M. Norton: X-ray Diffraction, a Practical Approach, Plenum Press, New York (1998).

This page intentionally left blank

Surface Contamination and Cleaning, Vol. 1, pp. 193–212 Ed. K.L. Mittal © VSP 2003

Practical applications of icejet technology in surface processing D.V. SHISHKIN, E.S. GESKIN∗ and B. GOLDENBERG Waterjet Technology Laboratory, Department of Mechanical Engineering, New Jersey Institute of Technology, Newark, NJ 07102-1982

Abstract—The objective of this work was to acquire knowledge needed for the development and deployment of manufacturing processes utilizing the enormous technological potential of water ice. Material removal by blasting with ice media such as particles, pellets and slugs was investigated. The ice media were accelerated by entertainment in an air stream. The ice-airjet (IAJ) can replace sand blasting and the ice-waterjet (IWJ) can replace the abrasive waterjet (AWJ). The obvious advantage of the ice media is complete pollution prevention in course of materials treatment. With this technique it is possible to eliminate both contamination of the substrate as well as generation of contaminated waste streams. In addition to the obvious environmental benefits, the use of ice media will improve a number of key operational techniques, such as cleaning, decoating, polishing, deburring, drilling, cutting, etc. The “just-in-time” production of ice media at minimal environmental cost constitutes another advantage of ice-based technologies. Our previous studies have shown that the potential applications of ice abrasives range from cutting of metals to etching of photo films and precision cleaning of electronic parts. However, the rate of the cleaning and machining operations performed was insufficient. A key objective of this research was to improve ice blasting so that it was not only feasible, but also technologically and economically efficient. Keywords: Surface processing; cleaning; precision; abrasive; particle; ice.

1. INTRODUCTION

There are a number of suggested air-ice based technologies. One of the firsts of such technologies was a car washing machine, utilizing ice particles [1]. The stream of the charged frozen particles controlled by a set of coils was directed at surfaces to be cleaned [2]. Szijcs [3] proposed cleaning of sensitive surfaces by the impact of a fine grade blast material and air. The atomization of the liquid in the air stream and subsequent freezing of the generated fine droplets form the blast material. The freezing is achieved by the addition of a refrigerant (N2, CO2, Freon) into the stream in the mixing chamber or by the addition of the refrigerant ∗

To whom all correspondence should be addressed. Phone: (973)596-3338, Fax: (973)642-4282, E-mail: [email protected]

194

D.V. Shishkin et al.

into the jet after the mixing chamber. The use of ice particles, which have a uniform grain size, for cleaning the surface and grooves of ferrite block, was reported by Tomoji [4]. An ice blasting device using stored particles was suggested by Harima [5]. Vissisouk [6] proposed to use ice particles near melting temperature in order to effectively remove the coating from the substrate. Mesher [7] suggested a nozzle for enhancement of surface cleaning by ice blasting. Shinichi [8] suggested cleaning inexpensively various surfaces by mixing ice particles, cold water and air. Niechial [9] proposed an ice blasting cleaning system containing an ice crusher, a separator and a blasting gun. Settles [10] suggested producing ice particles of a size range below 100 µm within the apparatus just prior to the nozzle. Although the use of ice blasting is suggested by a number of inventors, the practical application is much more limited. Herb and Vissisouk [11] report the use of ice pellets for precision cleaning of zirconium alloys in the course of production of bimetallic tubings. It was reported that ice blasting improved the quality of the bimetal. The use of air-ice blasting for steel derusting was reported by Liu [12]. The following operational conditions were maintained during blasting: air pressure: 02-0.76 MPa, grain diameter: below 2.5 mm, ice temperature –50°C, traverse rate 90 mm/min, and standoff distance 50 mm. Under these conditions the rate of derusting ranged from 290 mm2/min at the air pressure of 0.2 MPa to 1110 mm2/min at the air pressure of 0.76 MPa. The quality of the cleaned surface complied with ISO 8501-1 Sa 2. The most important problem which actually impedes adoption of the ice-jet (IJ) technology arises from the difficulties in the generation and handling of ice abrasives. Regular abrasives are stable at all practical ranges of operational conditions, while ice particles can exist only at subzero temperature. Maintaining such a temperature both within the nozzle and the jet is an extremely difficult task. Ice particles tend to pack and clog the supply lines. The adhesion between the particles increases dramatically as the temperature approaches 0°C. Thus prior to entrance in the nozzle, ice particles should be maintained at a low temperature. These and some other problems prevent adoption of IWJ. In order to assure the acceptance of IWJ by the industry, it is necessary to develop a practical technology for formation of ice-water slurry. 2. SET UP FOR ICE-AIRJET EXPERIMENTAL PROTOTYPE

The experimental prototype depicted in Figure 1 was selected for further experiments. The system consisted of the following functionally separated blocks: – ice making block which includes the evaporator, auger, auger driver, sealing and liquid nitrogen cooling apparatus; – ice unloading mechanism – nozzle block which includes parallel nozzles and focusing device.

Practical applications of icejet technology in surface processing

195

Figure 1. (a) Schematic of auger type IJ system where: 1 – evaporator, 2 – refrigerant coils, 3 – insulation, 4 – auger, 5 – ice reloading device, 6 – air gun, 7 – air supply port, 8 – water supply port, 9 – cooling medium port, 10 – gauges, A – air flow rate valve, B – water flow rate valve, C – cooling medium valve, D – data acquisition card, and (b) picture of the ice reloading device with nozzle block.

In our experiments, water entered the heat exchanger via a special port. As it moved along the rotating auger water solidified and an ice plug was formed. Solidified ice plug moved forward along the auger helical ways. Decomposition of this plug formed ice powder. The heat exchanger and the auger of the icemaker constituted a modified commercial icemaker of Hoshizaki America Inc., Peachtree City, GA. The design of these parts will be changed in the next generation of the device. The cooling was carried out by the refrigerant Galden HT-55 supplied by the TurboJet refrigeration apparatus or by liquid nitrogen stored in a tank. We replaced Hoshizaki auger driver by a more powerful device in order to prevent jamming of the ice. The rotation momentum of the auger 4 was provided via a gearbox with gear-ratio 1:100. However, the selected driver operated at a constant speed of 100 rpm. Water flow rate was precisely controlled by a special valve (Figure 1 (a)). At the outlet of the heat exchanger the powder was entrained by the unloading mechanism which directed it to the nozzle block (5). The nozzle block consisted of two air guns (6) and a special focusing device. Three different sizes of the nozzles were used; however, in all cases the nozzle-to-focusing tube ratio was 1:2. An unloading mechanism delivered ice particles via flexible plastic tubes to the abrasive port of the air gun. In the gun the air supplied into the insulated nozzle block at the room temperature accelerated the particles. The IAJ was formed and directed to the substrate surface 3.

196

D.V. Shishkin et al.

3. EXPERIMENTAL PROCEDURE

In the course of IJ cleaning, the air pressure was maintained at 0.544 MPa (80 psi), the nozzle diameter was 2.5 mm, and the nozzle focusing tube diameter was 5 mm. The properties of ice abrasive medium were the following: ice temperature was in the range from –20°C to –70°C, granulometric composition of ice powder ranged from 0.3 mm to 7 mm, and ice flow rate was 20 g/min to 150 g/min. Iceairjet (IAJ) was used for cleaning various sensitive surfaces covered by moderately adhering deposits. The sensitive elements of the electronic boards were covered by a conductive copper paste and cleaned by the IAJ. When assembled, these components performed normally and the normal operational modes of the devices were demonstrated. The feasibility of using IAJ technology as a blasting medium for cleaning highly sensitive surfaces was shown. Another experiment involved depainting of various substrates, including mirror-like surfaces and the surfaces of soft substrates. A complete removal of the paint and the absence of surface damage were demonstrated. The generated surfaces were inspected visually. A number of experiments involved the use of the ice abrasive in waterjet (WJ) cutting applications. The experimental procedure was carried out with the following parameters: the water pressure was 306.1 MPa, the diameter of the sapphire nozzle was 0.178 mm, average standoff distance was maintained in range 7 mm– 10 mm and the traverse rate was 1.06 mm/s. Various metals and composite materials were cleaned by IWJ. The depth and cutting rate were substantially lower than in the case with conventional abrasive media. However, the IWJ produced a very narrow cutting kerf compared with AWJ and had a superior cutting ability over pure WJ. The main obstacle during ice particles entrainment in the nozzle abrasive port was their agglomeration at the port entrance and their disintegration in the mixing chamber due to intensive melting. This technology is still under development and requires further investigation. 4. EXPERIMENTAL RESULTS

A series of experiments were carried out in order to evaluate the potential of the application of IAJ for surface processing. The description of these experiments is given below. 4.1. Cleaning of electronic boards A disabled TV set was disassembled (Fig. 2a). The electronic boards were covered by a heavy dust. Then the boards were decontaminated by IAJ and reassembled. The TV set performed normally (Fig. 2b). The architecture of the boards in question was extremely complex and contained a number of very sensitive sites, like electrical contacts and conduits. Any damage to the board components would result in the TV set malfunction. It is obvious that the ice-air stream induced no damage. More difficult task, however, was a complete grease removal. Even

Practical applications of icejet technology in surface processing

197

Figure 2. (a) photograph of the electronics board of a TV set. Notice the heavy layers of dust and dirt on the electric and electronic components of board, (b) photograph of an assembled TV set. The contaminated board of TV set is shown (a). After cleaning TV set worked normally.

198

D.V. Shishkin et al.

small amount of the grease remaining at hidden pockets will disrupt the TV set performance. It is clear that the jet was able to remove soil from all the difficultto-reach pockets. Another experiment involved decontamination of computer boards. Various devices (PC, electronic watches, computer games, etc) were disassembled. The boards were covered by a mixture of lithium grease and then decontaminated by IAJ. Clean boards were reassembled and tested. All devices worked perfectly. Some of the devices above were used for several tests. No deviation in the computer operation was noticed. The boards above were populated by a large number of rather fragile components such as chips, connectors, etc. Any damage to any of these components, as well as any presence of grease on the board will disable the device. In all performed experiments the deposit was removed completely and no damage was induced to the board components. The examples of the boards decontaminated in the course of these experiments are shown in Figs. 3 (a) and 3 (b). 4.2. Decoating of sensitive surfaces The experiments involved depainting of a compact disc (CD). This involved removal of the paint as well as two layers of the coating originally deposited on the disk (Figs. 4 (a) and 4 (b)). The paint and then the emulsion layers were removed separately with no damage to the underlining surface. Another experiment involved painting and subsequent depainting of the mirror-like surface of stainless steel (Fig. 5 (a)). No change in the surface topography was noticed. Further experiments involved depainting of china (Fig. 5 (b)), egg (Fig. 6 (a)), and glass lining of a pharmaceutical reactor (Fig. 6 (b)). The most representative experiments, however, involved depainting of a LC display (Fig. 7 (a)) and degreasing of an optical glass (Fig. 7 (b)). 4.3. Decoating of soft substrates These experiments involved depainting of a soft plastic (Fig. 8 (a)) and fabric (Fig. 8 (b)). Decoating of a substrate having mechanical strength lower than that of the coating constitutes a challenging task, but IAJ was able to perform this task. 4.4. Restoration of electromechanical devices A solenoid valve (Fig. 9 (a)) and a DC motor (Fig. 9 (b)) were completely disabled by painting of all contacts. After IAJ cleaning the devices performed normally. 4.5. Removal of highly adherent surface layers An aluminum plate was covered by a thick layer of tar. Then the tar was removed mechanically from a part of the plate. However, a highly adherent thin tar layer

Practical applications of icejet technology in surface processing

199

Figure 3. (a) photograph of the board of an electronic game containing electric conduits, microchip and electronic matrix. The board was covered by a mixture of lithium grease and copper powder. Notice the cross contamination of electric conduits of the board, (b) photograph of the assembled electronic game after IAJ cleaning. The electronic game performed normally.

200

D.V. Shishkin et al.

Figure 4. (a) photograph of the CD-ROM covered by Rust-Oleum gloss protective enamel. The paint was partially removed from the CD ROM surface. No surface damage was observed in the course of IAJ cleaning, and (b) photograph of the CD-ROM partially cleaned using IAJ technique. Notice that layers of both paint and emulsion were removed. No surface damage was observed in the course of IAJ processing.

Practical applications of icejet technology in surface processing

201

Figure 5. (a) photograph of the polished steel surface. The polished steel surface was contaminated by Rust-Oleum gloss protective enamel. The paint was partially removed from the polished surface. No surface damage was observed in the course of IAJ cleaning, so the feasibility of the precision cleaning of polished surfaces was demonstrated. (b) photograph of the hand-painted china plate. The plate was covered by Rust-Oleum gloss protective enamel. Part of the deposited paint was removed by ice etching. No modification of the original surface was noticed, and thus the feasibility of IAJ etching of sensitive surfaces was demonstrated.

202

D.V. Shishkin et al.

Figure 6. (a) photograph of an egg. The egg surface was painted by Rust-Oleum gloss protective enamel. After this the egg was partially decontaminated by IAJ technique. No damage to the egg surface or penetration of the ice particles through the eggshell was noticed, so the feasibility of decontamination of highly unstable and brittle surfaces was demonstrated. (b) photograph of the cover of a pharmaceutical reactor contaminated by the lithium grease. Then the grease was partially removed from the surface of the cover by IAJ technique. No damage to the glass in the course of IAJ cleaning was noticed.

Practical applications of icejet technology in surface processing

203

Figure 7. (a) photograph of the LC display of a calculator containing electronic matrix and LCD conduits. The display was contaminated by Rust-Oleum gloss protective enamel. Then all elements of the LC display were decontaminated by IAJ technique. On assembly of the calculator the LC display performed normally. (b) photograph of a magnification lens. The lens was contaminated by lithium grease. The grease was partially removed from the lens surface, and no damage to the lens surface was observed.

204

D.V. Shishkin et al.

Figure 8. (a) photograph of a PVC tube contaminated by Rust-Oleum gloss protective enamel. The tube was partially decontaminated by IAJ technique. No damage to the tube surface in the course of IAJ cleaning was noticed, and (b) photograph of a cotton fabric. The fabric was contaminated by Rust-Oleum gloss protective enamel. Then the paint was partially removed from fabric surface, and thus the feasibility of the use of ice particles for decontamination of fabrics was demonstrated.

Practical applications of icejet technology in surface processing

205

Figure 9. (a) photograph of an electrical solenoid valve with connectors contaminated by RustOleum gloss protective enamel. The contacts of solenoid valve were cleaned by IAJ technique. After cleaning the solenoid valve was connected to an electrical supply source and performed normally. This experiment demonstrated the feasibility of using IAJ technique for decontamination and restoration of contacts of different electronic devices. (b) photograph of a DC motor. DC motor was disassembled and all elements were covered by a mixture of lithium grease and copper powder. DC motor was cleaned using IJ technique and the assembled DC motor performed normally.

206

D.V. Shishkin et al.

remained on the surface. It was not possible to remove it using mechanical means. The layer was removed completely by the IAJ (Fig. 10 (a)). A metal wall was covered by an oil paint and then was subjected to abrasive-airjet (AAJ) (the carrier medium was sodium bicarbonate). Then the same procedure was carried out using IAJ. The initial state of the graffiti covered metal surface is shown in Fig. 10 (b). The graffiti removal by conventional and ice based technologies are shown in Figs. 11 (a) and 11 (b). Another experiment involved removal of the residual highly adhesive Weldbond glue from plastic and rubber jointed surfaces (Fig. 12 (a) and 12 (b)). Average process duration in all these experiments was around two minutes. The heavily contaminated machine part with grease and dust was decontaminated by IAJ too (Fig. 13 (a) and 13 (b)). No damage to the underlying painted surface was noticed. 4.6. Etching applications The emulsion of a photo film was removed with no damage to the substrate (Fig. 14 (a)). This demonstrates the feasibility of the use of IAJ as an etching agent. 4.7. Ice-waterjet (IWJ) applications Various metals and plastic materials were subjected to IWJ cutting. The superior cutting ability of IWJ over pure WJ was seen. The cutting ability of IWJ was limited by ice abrasive disintegration in the nozzle mixing chamber. This task required further investigation. However, it was shown that the IWJ cutting kerf was thinner (Figure 14 (b) and Figure 15) and showed the potential of IWJ as an alternative cutting medium for waterjet industry. 5. CONCLUDING REMARKS

Although the ice-water jet constitutes an effective material removal tool, it is necessary to improve conditions of the jet formation in order to assure its adoption in practice. However, the ice-air jet is suitable for immediate application. It can be used for decontamination of very demanding and complex surfaces as well for such manufacturing applications as etching. Simplicity and complete absence of environmental damage constitute the main advantages of this process. A further development of IAJ surface cleaning technology will involve improvement of the control of ice particle properties and enhancement of the methods for the delivery of ice particles to the substrate. This enhancement will enable us to modify material polishing, surface cleaning, and, perhaps, grinding.

Practical applications of icejet technology in surface processing

207

Figure 10. (a) photograph of the aluminum surface contaminated by a thick layer of tar. The bulk of the tar was removed by WJ and knife scrubbing. The highly adherent thin layer was removed by ice etching. No damage to the metal surface was noticed and (b) graffiti covered painted metal surface. The oil paint is highly adherent.

208

D.V. Shishkin et al.

Figure 11. (a) Graffiti was removed with conventional AAJ. Notice discoloration occurred in the treated region and (b) surface was decontaminated by the IAJ. No damage to the underlying paint layer occurred.

Practical applications of icejet technology in surface processing

209

Figure 12. (a) The Weldbond glue was used to create a joint between plastic and rubber surfaces. Notice the highly adhesive character of the glue, and (b) the glue residue was removed by IAJ cleaning. No surface damage was noticed.

210

D.V. Shishkin et al.

Figure 13. (a) picture of the highly contaminated machine part with grease and dust, and (b) part was decontaminated by IAJ cleaning. No damage was seen on the underlying painted surface.

Practical applications of icejet technology in surface processing

211

Figure 14. (a) photograph of a strip of a photo film. The photo emulsion was partially removed from the film surface. No surface damage was observed in the course of IAJ cleaning and thus the feasibility of complete and selective emulsion removal from thin photo film was demonstrated, and (b) photographs of cutting of aluminum strip of thickness 3.1 mm (X65). Notice the reduced width of the kerf generated by IWJ cutting. Also note substrate surface erosion in the vicinity of IWJ generated kerf.

212

D.V. Shishkin et al.

Figure 15. Photographs of cutting of titanium sample of thickness 0.7 mm (X65). Notice the reduced width of kerf in the course of IWJ cutting. Also note the intensive erosion of the substrate surface in the vicinity of IWJ generated kerf.

REFERENCES 1. C. Schlosser, L. Mueller and G. McDougal, US Patent 5,752,39 (1950). 2. G. Kanno, US Patent 5,074,083 (1991). 3. J. Szijcs, European Patent 0509132B1 (1991). 4. M. Tomoji, Japanese Patent 04078477 (1990). 5. I. Harima, Japanese Patent 04360766 A (1992). 6. S. Vissisouk, European Patent 05076607 (1995). 7. T. Mesher, US Patent 5,607,478 (1997). 8. H. Shinichi, Japanese Patent 09225830 A (1997). 9. R. Niechial, US Patent 5,820,447 (1998). 10. G. Settles, US Patent 5,785,581 (1998). 11. B. Herb and S. Vissisouk, Proc. Precision Cleaning 1996 held in Anaheim, CA, pp. 172-179 (1996). 12. B. Liu, in Jetting Technology, H. Louis (Ed.), pp. 203-211, Professional Engineering Publishing Ltd., London, UK (1998).

Surface Contamination and Cleaning, Vol. 1, pp. 213–224 Ed. K.L. Mittal © VSP 2003

Correlating cleanliness to electrical performance TERRY MUNSON∗ Contamination Studies Laboratory (CSL), 201 East Defenbaugh, Kokomo, Indiana 46902

Abstract—This paper explores the correlation between the cleanliness levels on electronic assemblies and their electrical performance. It documents an experiment conducted to explore this correlation. Cleanliness was measured using Ion Chromatography (IC), and electrical performance was measured using Surface Insulation Resistance (SIR) testing under elevated humidity and temperature conditions. Furthermore, this paper discusses electronic assembly cleanliness issues, and a new cleanliness assessment approach for determining cleanliness levels required for the typical flux technology of today. We conclude – from the samples examined, and based on our past 10-years of experience analyzing similar experiments – that circuit board field performance (good or poor) is strongly correlated to the specific amount and type of invisible and visible residues between pads and holes in all areas of active circuitry. Keywords: Ionic contaminants; residues; electrochemical metal migration; ion chromatography; surface insulation resistance; electrical performance testing; cleanliness levels.

1. INTRODUCTION

Since the 1987 Clean Air Act, when Government legislation forced the electronics industry to stop using the ozone depleting chemical Freon® as a cleaner, the industry has been required to find new chemicals and processes. The industry believed at the time that Freon® solvent cleaning techniques effectively removed all surface contaminants. Subsequent research has shown that the old rosin based fluxes used in the manufacturing processes actually sealed in contaminants, whereas the new fluxes left contaminants exposed to react with humidity in the end-user environments. These inherent process residues must be removed to achieve product reliability. The change from Freon® cleaning initiated many changes in the manufacturing processes. There was no direct substitute cleaning chemistry. The new solvents could not be used for cleaning using many of the old processing chemistries. The change was an opportunity for many manufacturers to change their overall processes. As the changes were made to new processing materials and cleaning sol∗

Phone: 765-457-8095, Fax: 765-457-9033, E-mail: [email protected]

214

T. Munson

vents, product performance data indicated increasing field performance problems related to surface ionic residues. Understanding surface residues and the effects of each type became critical for product quality. Since the fluxes by themselves did not have corrosive activators and passed the coupon evaluations of the IPC (IPC TP-1042 “Phase 2 No-Clean Flux Study”) and the water-soluble process passed all the test coupon work published by the IPC (IPC TP-1043 and TP-1044 “Phase 3 Water Soluble Fluxes Study” September 30, 1992), then the focus shifted to meeting solderability performance with real product. Process cleanliness had not been an issue when the rosin flux was used. Manufactures that switched to No Clean (low solids) processes in October of 1992 would, in some cases, generate a product recall in May 1993. It has become clear that the bare board, component, temporary mask and materials and rework contaminants are critical to today’s electronic hardware performance. Before examining the correlation between cleanliness levels and electrical performance on electronic equipment, let us define cleanliness and the factors that affect it. 1.1. Definition of cleanliness In the case of electronic assemblies, finished cleanliness levels are a measure of the amount of detrimental residues remaining on completed assemblies. Electronic assembly cleanliness is a result (signature) of the assembly process, materials and secondary processes required creating the finished assembly. 1.2. Acceptable cleanliness levels The acceptable levels of cleanliness depend on how the various residue types that remain on the assembly react under electrical power in poorly controlled environmental conditions. The various residue types react differently from pad-to-pad and hole-to-hole. How they react determines the quality of electrical performance. Two of the most valuable tools for determining board and assembly cleanliness are Ion Chromatography (IC) and Surface Insulation Resistance (SIR) testing. 1.3. Why is cleanliness an important issue? Electronic assembly cleanliness is an important issue for at least four reasons: 1. The trend toward higher operating frequencies and lower operating voltages is causing circuits to be less tolerant of stray current leakage. 2. Spacing geometries have pushed traces and leads closer together and have increased the probability of power-to-ground pathways due to smaller amounts of fluids required to bridge the smaller spacing. 3. Component packages have become shorter, spaced tighter to the board surface, and smaller in relationship to the board area.

Correlating cleanliness to electrical performance

215

4. New flux chemistry residues on the newest metal composition result in new interaction issues that have effects that are inadequately understood. 1.4. Why is determining cleanliness a problem? Determining if an electronic assembly is adequately clean is difficult for the following reasons: 1. Cleanliness is not easily assessed with today’s production floor tools. Current industry-standard process cleanliness tools and test methods are not adequate gauges of product cleanliness when testing today’s low solids and watersoluble fluxes. Due to poor extraction conditions, these tools give false low levels. Also, they do not identify residues as corrosive or insulative. 2. Cleanliness is not visually assessable. 3. Adequate cleanliness depends on the circuit design, the processing materials, the process, and other factors. 4. Cleanliness is not uniform across the assembly surface, but has concentrations of residues in the critical areas such as between the component leads and the board and component interface. 5. Cleanliness is not defined in Industry Specifications as it relates to today’s field performance. 6. Cleanliness is not necessarily the same in a process from day-to-day, due to different suppliers of bare boards, operator experience, and vendor variations. 1.5. Residues sources Circuit board cleanliness is a measure of the cumulative process residues. These residues are the result of the manufacturing process steps and materials used in each step. Everything used in the manufacturing process has an impact on the types of residues that will be created. To complicate things further, boards and components that look visibly clean and dry can actually be absorbing moisture and reacting with the bare board HASL flux, creating a leakage path. These residues can come from: 1. Materials ● Bare Board Fabrication (Etching chemicals, HASL fluxes, Rinse water (tap)) ● Component Packaging (Fluxes, Mold releases) ● Component Plating or Tinning (Plating, Bath, Rinse water, Fluxes) ● Flux, Solder Paste, Cored Solder, Water Quality, Epoxies and Soaps 2. Processes ● Fluxing and Soldering (Flux amounts, Thermal effects) ● Paste and Reflow (Flux spread, and outgassing effects) ● Water Soluble Cleaning or No-Clean Processing (pressure, water, and saponifier)

T. Munson

216 ● Production

Handling (Glove and hand residues) Masking and Removal (outgassing) ● Hand Soldering and Rework (cleaning materials and procedures) ● Temporary

1.6. Residue effects Process residues are the corrosive contaminant source for corrosion cells to develop, which form when residues, a fluid medium, and a voltage differential are combined. These corrosion cells cause electrochemical migration (metal migration). Figure 1 shows dendritic growth in a corrosion cell created using a 21second Water Drop Evaluation with a high chloride level (1.29, µg/cm2 of Cl–) coupon, a 10-volt bias, and DI water, between 1-mm spacing. Corrosion cells like this cause electrical leakage and shorts, resulting in failures (or No Trouble Found (NTF) returns) under typical operating conditions (humidity levels of 50% and higher).

Figure 1. Dendrite Growth at (a) 5 seconds, (b) 10 seconds, (c) 15 seconds, (d) 20 seconds.

Correlating cleanliness to electrical performance

217

Figure 2. Umpire SIR Test Board.

2. EXPERIMENTAL DETAILS

CSL designed and conducted an experiment to explore the correlation between circuit board cleanliness levels and long-term reliability simulated through accelerated life testing under powered, high temperature and humidity conditions. 2.1. Materials To perform the experiment, 30 test coupon printed circuit boards were fabricated. The Umpire test board (shown in Figure 2) was selected because it uses a mixed technology assembly process – surface mount technology (SMT) and platedthrough-hole technology (PTH). It can also be used to assess different component areas, such as the pad-to-pad or lead-to-lead, and assess entrapment effects on a variety of components (BGA, LCC, QFP, DIP, PGA, 1206 chip, 0805 chip), and comb patterns (B24, Bellcore). 2.2. Processing conditions (three levels of bare board cleanliness with a no-clean assembly process) All 30 bare boards were fabricated using a Hot Air Solder Leveled (HASL) with three levels of cleanliness. The assembly process used a low residue solder-paste, a low residue liquid flux (alcohol-based), and a low-solids core solder (with additional liquid flux followed by alcohol and brush localized cleaning). This process also used a peelable temporary soldermask. The samples were processed as one

218

T. Munson

batch until they reached the final cleaning process. Cleanliness Level 1 (L1) used a saponifier in the cleaning solution, known to be the cleanest method. Level 2 (L2) used deionized water cleaning, known to be a marginal cleaner. Level 3 (L3) used tap water cleaning, known to be the dirtiest cleaning method (standard rinse water process at most fabricators). These two tests were performed on each sample: 1) Ion Chromatography and 2) Surface Insulation Resistance with visual inspection performed per IPC 610 protocol. 2.3. Cleanliness testing using ion chromatography Ion Chromatography is used to assess cleanliness levels of electronic equipment. Ion Chromatography is a process of separating ionic and organic residues suspended in a liquid. This separation is achieved through a finely balanced system of liquid phase eluent and resin columns. The resin has a charge opposite to the ions, causing different ions to travel through the column at different rates. As each species leaves the column (illustrated in Figure 3), a conductivity cell measures its concentration in microSiemens (µS). The IC system records this information on a chart for the duration of the analysis, and quantifies the area under the curve of each species detected. Typical species detected include: Fluoride, Chloride, Bromide, Phosphate, Sulfate, Formate, Acetate, Methane Sulfonic Acid, Weak Or-

Figure 3. Ion Chromatography System Illustration.

Correlating cleanliness to electrical performance

219

Figure 4. Surface Insulation Resistance Testing System Illustration.

ganic Acid, Sodium, Calcium, Potassium, and NH4–. Before testing, the IC system is calibrated to NIST (National Institute of Standards and Technology) traceable standards. The level of sensitivity is 0.01 part per million. CSL is able to test specific areas of boards by combining IC with an effective extraction procedure that conforms to IPC protocol TM 650 2.3.28. Each sample was IC tested in four location areas: 1) a top-side surface mount technology (SMT) area, 2) a bottom-side wave solder area, 3) a rework area, and 4) a peelable solder mask area. 2.4. Surface insulation resistance (SIR) testing SIR testing with the Umpire board allowed CSL to subject the processed samples to accelerated environmental conditions under applied power, which allowed evaluation of the electrical effects of trapped process residues. Additionally, the Umpire board allowed CSL to analyze each of the four location areas separately. Figure 4 diagrams the SIR system used. According to the IPC J-Std SIR pass-fail criterion, the patterns must maintain resistance values above 1.0e8 ohms measured at 96 and 168 hours. 3. RESULTS

The experimental results are grouped into the four test area groups: 1) top-side SMT area, 2) bottom side wave solder area, 3) rework area, and 4) solder-mask area. 3.1. Top-side SMT area results The data in Table 1 show the ionic and electrical performance mean values from each group (5 boards per condition) of samples relative to the effects of the factors on the SMT top-side.

T. Munson

220

Table 1. Top-side SMT area results for all three levels of cleanliness (L1-L3) Ionic Test using Ion Chromatography (all values are in µg/cm2) Sample #

Description

Chloride Bromide Sulfate WOA

Sodium Potassium

L1 L2 L3

SMT paste areas only SMT paste areas only SMT paste areas only

0.18 1.29 2.50

0.33 0.20 0.26

0.34 0.38 0.38

0.00 0.00 0.00

7.03 5.86 6.14

0.03 0.05 0.03

Electrical Performance using Surface Insulation Resistance Testing Sample # Initial measurement (in Ohms)

Measurement BGA (in Ohms) at 168 hours (85°C/85%RH)

LCC

LCC QFP (comb)

QFP Head1 (comb) (control)

L1 L2 L3

4.3e10 6.7e7 1.0e6

Passed Failed Failed

Passed Failed Failed

Passed Failed Failed

1.3e11 1.4e10 2.2e10

Passed Failed Failed

Passed Failed Failed

Passed Failed Failed

1 = Header pattern topside WOA = Weak Organic Acids BGA = Ball Grid Array LCC = Leadless chip carrier QFP = Quad Flat Pack RH = Relative Humidity

Based on the data in Table 1, we concluded that; 1. L1 boards performed well ionically and electrically. 2. L2 boards failed electrically due to the high amount of bare board contamination (chloride HASL flux). The BGA, LCC and QFP failed due to the chloride / flux residues from the bare board fabrication process, and not due to the assembly process since L1 boards passed. 3. L3 boards failed electrically due to the high amount of bare board contamination (chloride HASL flux). 3.2. Bottom side wave solder area results The data in Table 2 show the ionic and electrical performance of the samples relative to the effects of the factors on the bottom side wave solder areas. Based on the data in Table 2, we concluded that; 1. L1 boards performed well ionically and electrically with the exception of the bottom side B-24 comb pattern with a layer of WOA residue over the entire comb thick enough to be visibly obvious. This comb failure occurred because too much flux was applied and not all of it was complexed (all the flux carrier driven off and the crystals melted forming an insulative residue), leaving a partially dried but conductive moisture-absorbing residue between the leads.

Correlating cleanliness to electrical performance

221

Table 2. Bottom side wave solder area results for all three levels of cleanliness (L1-L3) Ionic Test using Ion Chromatography (all values are in µg/cm2) Sample # Description L1 L2 L3

Chloride

Wave soldered areas 0.66 only Wave soldered areas 1.28 only Wave soldered areas 2.61 only

Bromide

Sulfate

WOA

Sodium

Potassium

0.36

0.00

50.57

0.26

0.08

0.43

0.00

65.60

0.17

0.05

0.36

0.00

60.08

0.20

0.09

Electrical Performance using Surface Insulation Resistance Testing Sample # Initial measurement (in Ohms)

Measurement B-241 (in Ohms) at 168 hours (85°C/85%RH)

PGA

DIP1

DIP2

Head1 (control)

L1 L2 L3

3.3e8 6.7e9 1.0e6

Passed Passed Failed

Passed Passed Failed

Passed Failed Failed

Passed Failed Failed

1.2e11 1.3e10 2.6e10

Failed Failed Failed

1 = Dip pattern 2 = Dip pattern

2. L2 boards failed electrically due to the high amount of bare board contamination (chloride HASL flux) with the exception of the PGA and DIP area. The failure was due to the insulative flux effects creating a barrier between the moisture and the leads. 3. L3 boards failed electrically due to the high amount of bare board contamination (chloride HASL flux). 3.3. Rework area results The data in Table 3 show the ionic and electrical performance of the samples relative to the effects of factors on the rework areas. Based on the data in Table 3, we concluded that; L1, L2 and L3 boards failed due to the un-reacted flux and the distribution of this flux residue during the cleaning process. This failure occurred because extra flux was applied by an operator during a reworking process and not all of it was complexed (heated correctly). It then spread around leaving a conductive moisture-absorbing path, resulting in high levels of current leakage, and test failure.

T. Munson

222

Table 3. Rework area results for all three levels of cleanliness (L1-L3) Ionic Test using Ion Chromatography (all values are in µg/cm2) Sample # Description

Chloride

Bromide

Sulfate

WOA Sodium

Potassium

L1 L2 L3

0.50 1.69 3.43

0.44 0.44 0.41

0.00 0.00 0.00

59.15 0.20 65.60 0.22 60.03 0.19

0.06 0.08 0.08

Reworked areas only Reworked areas only Reworked areas only

Electrical Performance using Surface Insulation Resistance Testing Sample # Initial measurement (in Ohms)

Measurement B-243 (in Ohms) at 168 hours (85°C/85%RH)

Head2

Head1 (control)

L1 L2 L3

3.3e8 6.7e9 1.0e6

Failed Failed Failed

Passed Failed Failed

1.2e11 1.3e10 2.6e10

Failed Failed Failed

1 = Header pattern topside 2 = Header pattern bottomside 3 = B24 board bottomside stripped soldermask comb

Table 4. Temporary Soldermask area results for all three levels of cleanliness (L1-L3) Ionic Test using Ion Chromatography (all values are in µg/cm2) Sample # Description

Chloride Bromide Sulfate

WOA

Sodium

Potassium

L1

0.64

0.42

0.00

0.75

0.50

0.17

1.60

0.43

0.00

0.82

0.53

0.20

3.35

0.42

0.00

0.80

0.47

0.17

L2 L3

Temporary Masked areas only Temporary Masked areas only Temporary Masked areas only

Electrical Performance using Surface Insulation Resistance Testing Sample # Initial Measurement measurement (in Ohms) (in Ohms) at 168 hours (85°C/85%RH)

Head3

Head1 (control)

L1 L2 L3

Failed Failed Failed

Passed Failed Failed

8.e10 4.2e10 3.9e10

1.7e7 6.9e6 1.0e6

Correlating cleanliness to electrical performance

223

3.4. Temporary soldermask area results The data in Table 4 show the ionic and electrical performance of the samples relative to the effects of factors on the temporary solder-masked areas. Based on the data in Table 4, we concluded that; L1, L2 and L3 boards failed electrically due to the high amount of bare board contamination (chloride HASL flux) and contact residues from the peelable soldermask. This failure occurred because no flux residues were present to create an insulative barrier. 4. DISCUSSION

This case study shows that the low-residue assembly process used works well in the areas of SMT and Wave Solder with clean (level 1) bare boards. However, the secondary processing (rework and temporary soldermask) areas, even with clean bare boards, showed high levels of electrical failure due to excess partially or unreacted flux. The data presented here are only a small snapshot of the information gathered through this testing. This assessment is not intended to replace actual product validation or environmental testing. This case study is intended to assess the process effects as a baseline and to determine if process changes are good or bad in regards to the electrical effect. As an assessment tool, this will help establish the actual level of cleanliness required in building reliable hardware. A cleanliness assessment approach such as this will allow electronic assemblers the opportunity to document the effects of the process and materials as a baseline from which to make improvements. This case study and others we have performed over the last ten years have shown us that the residues from fabrication do have a large effect on electrical performance. In addition, the residues from secondary assembly processes have just as much effect on the field performance of the product. Although ionic and organic IC analyses of component areas on electronic assemblies detect a specific amount of flux and processing residues, determining whether the levels are good or bad is based on results from an electrical SIR evaluation of the same areas. Over the last 10 years, CSL has developed a large cleanliness level database, and has used this data to determine general levels of cleanliness, but we are constantly adjusting acceptability levels because of changing component packages, increasing operating frequency designs, lower voltage designs, and changing processing materials. The result of all the changes in the industry is electronic equipment that is more sensitive to process residues, and has less long-term reliability, especially in harsh environments, unless detrimental process residues are identified, measured, and reduced.

224

T. Munson

5. CONCLUSION

Knowing cleanliness levels is a vital part of understanding product quality. The reliability of a product can be improved with adjustments in processing that improve cleanliness. Testing at various steps of the process or locations on the product can show what processing steps are affecting the final cleanliness and reliability. It is important to be particularly cautious with incoming product cleanliness and secondary assembly processes. Acknowledgments We would like to acknowledge Dr. K.L. Mittal for constructive feedback and support. We would also like to acknowledge Diversified Systems Inc. of Indianapolis, Indiana for their manufacturing support during the experiment. REFERENCES 1. M.G. Fontana, Corrosion Engineering, 3rd edition, McGraw-Hill Book company, New York (1986). 2. H. Small, Ion Chromatography, 1st edition, Plenum Press, New York (1989). 3. IPC Committee, Post Solder No-Clean Handbook (IPC-SC-62A), IPC, Chicago, IL (1999). 4. IPC TM 650, Test Methods Handbook, “Ion Chromatography,” IPC, Chicago, IL (1995). 5. IPC TM 650, Test Methods Handbook, “Surface Insulation Resistance”, IPC, Chicago, IL (1995).

Surface Contamination and Cleaning, Vol. 1, pp. 225–239 Ed. K.L. Mittal © VSP 2003

Qualifying a cleaning system for space flight printed wiring assemblies J.K. “KIRK” BONNER∗ and ATUL MEHTA Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Pasadena, CA 91109, USA

Abstract—Cleaning critical and high reliability printed wiring assemblies (PWAs) continues to be important to ensure high reliability performance and to prevent premature failure. The necessary steps to qualifying both a cleaning system and an appropriate chemistry for cleaning such PWAs are set forth. This paper addresses a centrifugal cleaning system used in conjunction with a water-based cleaning medium to achieve optimally low levels of contaminants on PWAs. Ionograph data, ion chromatography profiling, residual rosin determination, and outgassing data are presented demonstrating the effectiveness of the centrifugal cleaning system and the aqueous cleaning agent for space flight printed wiring assemblies. It is concluded that a centrifugal cleaning system coupled with a suitable aqueous chemistry can be successfully employed to clean high reliability PWAs. Keywords: Aqueous cleaning (AC); conformal coating; ionic contamination testing (ICT); greenhouse warming potential (GWP); multilayer board (MLB); ozone depletion potential (ODP); printed wiring assembly (PWA); printed wiring board (PWB); rosin mildly activated (RMA); semi-aqueous cleaning (SAC); surface mount technology (SMT); volatile organic compound (VOC).

1. INTRODUCTION

During the last decade, the challenges of cleaning printed wiring assemblies (PWAs) have grown. Today printed wiring boards have grown more complex to meet the continuing challenges posed by the increasing uses of microdevices, fine pitch packages, and array devices, such as ball grid arrays, microball grid arrays, and flip chips. Multilayer boards with a large layer count and narrow trace widths and spaces are commonplace. The ball grid arrays, microball grid arrays, and other small devices generally have a large number of inputs/outputs (I/Os), small standoffs, and small pitches. The small standoff and small pitch, coupled with the complex circuitry needed to route such components, makes cleaning an ever more critical operation. High reliability PWAs cannot tolerate contaminants since their

∗

To whom all correspondence should be addressed. Phone: (818) 354-1320, Fax: (818) 3935456, E-mail: [email protected]

226

J.K. Bonner and A. Mehta

presence can potentially degrade the board, thus compromising the intended mission. Cleaning for high performance PWAs is normally performed as a minimum at the following stages: (1) At the bare printed wiring board (PWB) stage prior to the application of solder mask; (2) Immediately after the PWB + components are soldered to form the PWA; (3) Immediately prior to the application of conformal coating. If the PWAs are properly stored, the second and third operations are sometimes combined. A number of contaminants are potentially introduced on the PWB surface. These contaminants can be classified into three broad categories: (1) Particulates; (2) Ionic residues; (3) Non-ionic residues, chiefly organic in nature [1-4]. To ensure the reliability of a PWA, cleaning is mandatory to remove these contaminants after the soldering operation and also directly prior to the application of a conformal coating. In addition to cleaning, some sort of cleanliness verification method, such as ionic contamination testing (ICT), is normally employed. ICT can be used to ascertain that a certain level of cleanliness has been achieved. Industry-recognized devices, such as an Ionograph® or Omega-Meter®, have commonly been used for this purpose. In addition, determining the amount of residual rosin (assuming that a rosin-based flux or paste was used) is often done. Another useful technique is to remove some of the components and examine for flux residues both visually and by use of a microscope. The last decade has also seen the dramatic decrease and continuing disuse of ozone-depleting solvents. The common chlorofluorocarbon solvents, such as Freon® TMS, have been discontinued, and many PWA assemblers have switched to more environmentally-friendly cleaning agents, such as a wide variety of semiaqueous and aqueous-based materials. To enhance the performance of such materials, the proper equipment selection plays a critical role. 2. BACKGROUND

Ten years ago the Electronic Packaging and Fabrication section at the Jet Propulsion Laboratory (JPL) established a dedicated facility for producing very low volume but high performance surface mount technology (SMT) assemblies known as the SMT Laboratory. This laboratory has successfully assembled SMT PWAs for such important JPL programs as these: ● Cassini; ● ChuG Microgyro; ● Caltech-ACE;

Qualifying a cleaning system for space flight printed wiring assemblies

227

MISER; ● SEAWINDS; ● Pathfinder. and many others. Since the assemblies produced in this laboratory always fall in the high performance, high reliability category, cleaning is mandatory, not optional. With the demise of the ozone-depleting solvents that were the mainstay of the electronics industry for twenty years, it was necessary to turn to alternative chemistries and cleaning systems to ensure cleanliness and high reliability of the surface mount assemblies (SMAs). The initial cleaning system chosen for the SMT Laboratory was a two-stage batch semi-aqueous (SA) cleaning system. Although this system worked satisfactorily for a number of years, the decision was reached recently to replace it. Part of the reason was the increasing complexity of the SMT PWAs. Equipment to ensure that the cleaning solution would successfully penetrate under the small standoffs and tight spacings found under the newer components now being increasing employed was considered mandatory. Another factor in the decision was that the initial equipment manufacturer sold off this portion of the business and no longer supported the equipment. It proved increasingly more difficult to maintain it in good working condition. In addition, isopropyl alcohol (IPA), used in the original equipment, came under increasing scrutiny by the South Coast Air Quality Management District (SCAQMD). Because IPA is a volatile organic compound (VOC), its emission into the atmosphere is tightly controlled. The decision was made to investigate a new cleaning system and a chemistry that would support JPL’s need for clean PWAs to meet the newer challenges. ●

3. PERTINENT PROCESS INFORMATION

The following JPL process information is pertinent to the discussion: ● Rosin-based fluxes and pastes are used to produce all electronic hardware. Using the terminology of Mil-F-14256, the classification of these products is rosin mildly activated (RMA). ● The solder paste is applied using a semi-automated screen printer ensuring that the paste is deposited in a uniform and consistent manner. Only stainless steel stencils are used in conjunction with a stainless steel squeegee. All boards are visually inspected for proper paste deposition after the stencil operation. ● A laser-based solder paste height and width measurement system is used with a resolution of 0.0001 inch (2.5 µm). This system provides real time information on the uniformity of solder paste deposition. All boards are subjected to this measurement prior to the reflow operation. ● A batch reflow operation is used to create the solder joints of the SMT PWAs. The SMT PWAs are thermally profiled using a M.O.L.E.® – a thermocouple

J.K. Bonner and A. Mehta

228

is attached to the PWB and to the M.O.L.E. The latter is a microprocessorbased data logger attached to a computer. Thermal profiling is done to eliminate thermal shock during preheat and reflow. This operation consists of a vapor phase reflow machine using a constant boiling perfluorocarbon material (3M Perfluorocompound FC-5312®) (b.p. 216°C) for soldering the SMT PWAs. The SMT PWAs are preheated to remove paste volatiles and to initiate the activation stage of the paste. The reflow liquid, since it boils at a constant temperature, minimizes the possibility of overheating the SMT PWAs during reflow and ensures that the vapor blanket performs a uniform and consistent soldering operation. 4. CRITERIA FOR CHOOSING A NEW CLEANING SYSTEM

The key criteria in choosing a new cleaning system were: ● Safety and ease of handling; ● Performance; ● Cost. Since JPL’s need is low throughput, a batch cleaning system was acceptable. After various preliminary trials, a centrifugal cleaning system was chosen based both on performance and versatility. In addition, several new aqueous cleaning chemistries seemed very promising. One of these is based on an aqueous chemistry containing a mixture of some alkoxypropanols with one to three alkoxy units (ether linkages). The molecules are not particularly large (C2 to C4), so the hydrophobic portion is not too large. The hydrophilic part of the molecule is due to one alcohol group (-OH) and several ether groups (-O-). Overall the organic molecules exhibit excellent solubility in water. Thus, the cleaning agent in water is herein referred to as an aqueous cleaning solution. The material itself is easily biodegradable. It has zero ozone depletion potential (ODP), virtually no greenhouse warming potential (GWP), and is classified as non-flammable. The following information is supplied by the manufacturer of the aqueous solution. Although the concentrate is 91% by weight volatile organic compound (VOC), the material as used in the cleaning system is only 13.6% by weight VOC. A broader description of aqueous cleaning systems is provided in the references [5-7]. 5. NEW CLEANING SYSTEM

The new cleaning system consists of the following equipment and materials. A brief description of its operation is also given below.

Qualifying a cleaning system for space flight printed wiring assemblies

229

Figure 1. View of the centrifugal cleaning system.

5.1. Equipment The following equipment is required: ● Centrifugal cleaning system; ● Vacuum oven; ● Refractometer. 5.1.1. Equipment description The equipment consists of an enclosed stainless steel cylindrical process chamber with a series of spray nozzles located vertically. A robotic arm containing a fixture holds the PWA and moves it in and out of the chamber vertically. During the cleaning cycle, the PWA is lowered into the process chamber until it is completely sealed from ambient. (See Figure 1). 5.2. Materials The following materials are used in the centrifugal cleaning system: ● Aqueous system containing the mixture of some alkoxypropanols with one to three alkoxy units (ether linkages). – 20% by volume (see Section 4 above for further details); ● Corrosion inhibitor – 1% by volume; ● Defoamer – 0.1% by volume;

230

J.K. Bonner and A. Mehta

Deionized (D.I.) water; ● High purity nitrogen gas (N2). Note: Hereafter, the term “aqueous cleaning solution” shall refer to the entire aqueous system consisting of water, the mixture of alkoxypropanols with one to three alkoxy units (ether linkages), corrosion inhibitor, and defoamer. The pH of the working solution is 10.5 (per the vendor). ●

5.3. Principle of operation The centrifugal cleaning machine uses centrifugal energy to clean PWAs. Energy is produced when PWAs to be cleaned are rotated inside an enclosed process chamber filled with the aqueous cleaning solution (see Section 5.2). This energy causes penetration of the solution under the components, including low profile components such as ball grid arrays (BGAs), dissolving the contaminants. The contaminants are subsequently removed during the rinse operation. 5.4. Overall process description The cleaning process consists of four-stage operation. The first stage is a nitrogen purge of the process chamber. The second stage is a wash cycle with aqueous cleaning solution. The process chamber is filled with appropriate amount of aqueous cleaning solution. The PWA, while immersed in the solution, is rotated in the chamber for a predetermined duration. At the end of the cycle, the solution is cycled back to the storage tank. During the third stage, the deionized (D.I.) water rinse sprays are activated while the PWA is rotating in the chamber. During this cycle any remaining material is removed, and final cleaning is achieved. In the fourth stage, filtered hot air is pumped in the chamber as the PWA rotates and dries. During these cycles, the PWA rotates alternately, clockwise and counter clockwise, to achieve optimum cleaning and drying. 6. TESTING OF THE NEW CLEANING SYSTEM

In order to investigate the new cleaning system, a comparison was made between it and the initial cleaning system. The following objectivcs were pertinent to this investigation. 6.1. Objectives The two chief objectives were: ● Investigate the new centrifugal cleaner using the aqueous cleaning solution for flight PWAs; ● Establish the optimal cleaning cycle for the new equipment.

Qualifying a cleaning system for space flight printed wiring assemblies

231

To be able to recommend the new centrifugal cleaner using the aqueous cleaning solution, the procedure used was to compare the cleaning data of the older cleaning process using the semi-aqueous (SA) solution with the centrifugal cleaner using the aqueous cleaning solution. 6.2. Test procedure The test procedure consisted of assemblying a test PWA that would prove challenging to clean. Several alternative cleaning runs using the new centrifugal cleaning equipment were made. The data so obtained were compared with (1) the test PWA used in the semi-aqueous (SA) cleaning system using the standard SA cleaning cycle, and (2) a control PWA not cleaned at all. 6.2.1. Test PWA The test PWA was populated with ball grid arrays (BGAs), a chip scale package, quad flat packs (QFPs) (20-mil pitch and 25 mil pitch – the nearest metric sizes are 0.5 mm and 0.625 mm), a plastic leaded chip carrier (PLCC), a flat pack, a small outline integrated circuit (SOIC) and several discrete chip capacitors and resistors. Both sides of the PWA were populated. The test PWA was assembled using Sn 63 paste with rosin mildly activated (RMA) flux and soldered in a vapor phase reflow system operating at constant temperature of 216°C. (See Figures 2 and 3). 6.2.2. Test parameters The following test parameters were employed: ● The basic equipment parameters of the centrifugal cleaning machine such as the temperature of the solution, the rotational speed of PWA and the drying temperature of the air were kept constant for all the tests. ● The only parameters that were varied were the cycle times: 1. Wash cycle time; 2. Rinse cycle time. 6.2.3. Cleanliness determination methods The following methods were used to assess the achieved cleanliness levels: ● Ionic contamination levels were determined using an Ionograph® 500 ionic contamination tester. In addition, testing was performed using ion chromatography (IC) to profile the various ionic species. ● Total low volatile residue (LVR) determination consisted of an extraction with Freon® TF and isopropyl alcohol (IPA) followed by a gravimetric determination. The total LVR was considered to be equal to organic rosin residue since rosin residue predominates in flux residue. ● Residual chloride analysis (Cl–) using ion chromatography (IC) was employed. For one run, residual fluoride (F–) and bromide analyses (Br–) were also performed.

232

J.K. Bonner and A. Mehta

Figure 2. Top view of the Test PWA.

Figure 3. Bottom view of the Test PWA.

Qualifying a cleaning system for space flight printed wiring assemblies

233

Outgassing per ASTM E595, “Standard Test Method for Total Mass Loss and Collected Volatile Condensable Materials from Outgassing in a Vacuum Environment.” Either ionic contamination testing was performed using the Ionograph® 500 or total LVR was determined for a given sample, but not both, i.e., the tests are mutually exclusive of each other. This is because in the process of conducting the ionic contamination test, the PWA is cleaned, thus rendering it unfit for further cleanliness testing. This is indicated in the results (Tables 1-3) using the symbol N/A (not applicable) in one or the other column. However, the total low volatile residue (LVR) analysis and the residual chloride analysis (Cl–) are not mutually exclusive, and both examinations can be performed on the same sample. They are not mutually exclusive because first an extract is made using D.I. water to remove the very soluble anions present (Cl–, F–, Br–), and then an extract is made using the Freon TF/IPA to remove the rosin residue which is insoluble in water. The results for the new centrifugal cleaning system using the aqueous chemistry are reported in Tables 1-3. Table 4 gives ionic contamination levels using the older SA cleaning system. The outgassing test method per ASTM E595 determines the volatile content of materials when exposed to a vacuum environment. Two parameters must be measured: Total mass loss (TML) and collected volatile condensable material (CVCM). In addition, since polyimide printed wiring board material can absorb moisture, an additional parameter was determined, namely, the amount of water vapor regained (WVR). The results for the TML, CVCM and TML-WVR values are reported in Table 5. ●

6.2.4. Acceptable cleanliness levels Per JPL D-8208, “Spacecraft Design and Fabrication Requirements for Electronic Packaging and Cabling”, the ionic contamination level as determined by the Ionograph® must not exceed 10 micrograms per square inch (10 µg/in2 = 1.5 µg/cm2). If it does, the entire lot of PWAs must be recleaned and one PWA per lot retested until this ionic cleanliness level is achieved. No acceptable standard has been agreed upon for the amount of residual rosin; however, a limit of no more than 150 micrograms per square inch (150 µg/in2 = 22.5 µg/cm2) seems appropriate. In the case of ionic profiling using ion chromatography (IC), no acceptable standards have been agreed upon for the amount of individual ionic species, but one would expect that the sum of the various ionic species would be less than the limit obtained from ionic contamination testing. Based on the outgassing determination per ASTM E595, the acceptable level for the TML must be no more than 1.00%, and the CVCM must be no more than 0.10%. If the WVR is determined, then TML-WRV is also reported.

J.K. Bonner and A. Mehta

234

6.2.5. Test runs Three test runs made with the new centrifugal cleaning machine using the aqueous cleaning solution are presented in Tables 1-3. As a comparison, a test run using the older SA cleaning system is presented in Table 4. 6.2.5.1. Test run #1 – new cleaning system Centrifugal cleaning system with the aqueous cleaning solution was used. The aqueous cleaning system consisted of 20% by volume of the long chain alcohol solution, 1% by volume of the corrosion inhibitor, and 0.1% by volume of the defoamer. The wash solution temperature was 50°C; the rinse solution temperature was 50°C; the dry air temperature was 200°C; the wash cycle rotational speed = 150 RPM. The centrifugal cleaning system parameters were: Wash time = 5.0 min.; rinse time = 2.5 min.; dry time = 2.5 min. Note: The four different batches signify that the run was repeated at four different times. Ionic contamination testing using the Ionograph® was done as a cleanliness check on some of the test PWAs. In addition, total low volatile residue (LVR) analysis and the residual chloride analysis (Cl–) were performed on other PWAs. The results are presented in Table 1. Table 1. Cleanliness data from test run #1 (new cleaning system) Test PWA Serial No.

Batch No.

Ionograph® Low results volatile µg/in2* residue µg/in2*

16 14

1 1

0.40 0.23

N/A N/A

18 19

2 2

0.00 0.17

N/A N/A

8 24 9 25

3 3 3 3

0.60 0.40 0.04 0.02

N/A N/A N/A N/A

26 10 Components removed 7

4 4 4

N/A N/A N/A

3.2 6.5 0.5

Uncleaned PWA

152.2

N/A

Remarks

Batch 1 mean ionic contamination level = 0.32 (µg/in2) Batch 2 mean ionic contamination level = 0.09 (µg/in2)

Batch 3 mean ionic contamination level = 0.27 (µg/in2) Cl– residue < 0.001 (µg/in2) Cl– residue < 0.001 (µg/in2) Cl– residue < 0.005 (µg/in2)

* In the U.S., process engineering results are typically given in µg/in2. 1 µg/in2 = 0.155 µg/cm2.

Qualifying a cleaning system for space flight printed wiring assemblies

235

6.2.5.2. Test run #2 – new cleaning system Centrifugal cleaning system with the aqueous cleaning solution was used. The aqueous cleaning system consisted of 20% by volume of the long chain alcohol solution, 1% by volume of the corrosion inhibitor, and 0.1% by volume of the defoamer. The wash solution temperature was 50°C; the rinse solution temperature was 50°C; the dry air temperature was 200°C; the wash cycle rotational speed = 150 RPM. The centrifugal cleaning system parameters were: Wash time = 3.0 min.; rinse time = 2.0 min.; dry time = 2.0 min. Note: The two different batches signify that the run was repeated at two different times. Ionic contamination testing using the Ionograph® was done as a cleanliness check on some of the test PWAs. In addition, total low volatile residue (LVR) analysis was performed on several PWAs. The results are presented in Table 2. Table 2. Cleanliness data from test run #2 (new cleaning system) Test PWA Serial No.

Batch No.

Ionograph® results µg/in2

Low volatile residue µg/in2

15 17

1 1

1.35 1.47

N/A N/A

27 28

1 2

N/A 0.36

6.5 N/A

29

2

N/A

0.5

Remarks

Batch 1 mean ionic contamination level = 1.41 (µg/in2) Batch 2 mean ionic contamination level = 0.36 (µg/in2) Parts were removed from PWB first

6.2.5.3. Test run #3 – new cleaning system Centrifugal cleaning system with the aqueous cleaning solution was used. The aqueous cleaning system consisted of 20% by volume of the long chain alcohol solution, 1% by volume of the corrosion inhibitor, and 0.1% by volume of the defoamer. The wash solution temperature was 50°C; the rinse solution temperature was 50°C; the dry air temperature was 200°C; the wash cycle rotational speed = 150 RPM. The centrifugal cleaning system parameters were: Wash time = 6.0 min.; rinse time = 6.0 min.; dry time = 3.0 min. Ionic contamination testing using the Ionograph® was done as a cleanliness check on some of the test PWAs. Also, total low volatile residue (LVR) analysis, residual chloride analysis (Cl–), and in addition, residual fluoride (F–) and bromide analyses (Br–) were performed on other PWAs. In Table 3 the two aluminum plates were cleaned along with the PWAs, but they were not exposed to the solder paste. The results are presented in Table 3.

J.K. Bonner and A. Mehta

236

Table 3. Cleanliness data from test run #3 (new cleaning system) Test PWA Batch Serial No. No.

Ionograph® Low results volatile µg/in2 residue µg/in2

Remarks (Residue results are given in µg/in2)

4

1

N/A

8.4

Rework was simulated and some flux applied to this PWA

5 11 12 102

1 1 1 2

N/A N/A N/A N/A

1.9 1.6 2.3 5.8

103

2

N/A

0.7

104

2

N/A

0.7

108

2

N/A

1.4

124

2

N/A

2.6

Al plate #1 Al plate #2 Solvent (Control) 105 106 107

2

N/A

0.7

2

N/A

0.4

2

N/A

0.0

3 3 3

3.19 1.12 1.60

N/A N/A N/A

122

Uncleaned N/A PWB Uncleaned N/A PWB

123

2462 33

Cl– residue < 0.000; F– residue < 0.000; Br– residue < 0.000 (µg/in2) Cl– residue < 0.000; F– residue < 0.000; Br– residue < 0.000 (µg/in2) Cl– residue < 0.000; F– residue < 0.000; Br– residue < 0.000 (µg/in2) Cl– residue < 0.000; F– residue < 0.000; Br– residue < 0.000 (µg/in2) Cl– residue < 0.001; F– residue < 0.000; Br– residue < 0.000 (µg/in2) Cl– residue < 0.003; F– residue < 0.002; Br– residue < 0.062 (µg/in2) Cl– residue < 0.003; F– residue < 0.002; Br– residue < 0.062 (µg/in2) Cl– residue < 0.001; F– residue < 0.002; Br– residue < 0.000 (µg/in2)

Batch 3 mean ionic contamination level = 1.97 (µg/in2) Bare PWB with solder paste printed on it (bare PWB means no components). Bare PWB with solder paste printed on it and then reflowed (bare PWB means no components).

6.2.5.4. Test run #4 – old cleaning system The two-stage batch semi-aqueous (SA) cleaning system was used. The first stage consisted in placing the PWAs vertically in a suitable conventional rack followed by cleaning using a terpene-based SA material and water. The PWAs were then transferred to the second machine and rinsed using a suitable saponifier, isopropyl alcohol (IPA), and D.I. water. The conventional wash/rinse/dry cycle was used.

Qualifying a cleaning system for space flight printed wiring assemblies

237

The SA cleaning system (old cleaning system) parameters were: Wash time = 5.0 min. (with saponifier); rinse time = 10.0 min. (D.I. H2O); 5.0 min. (D.I. H2O/IPA mixture); dry time = 5.0 min. The results are presented in Table 4. Table 4. Cleanliness data from test run #4 (old cleaning system) Test PWA Serial No.

Batch No.

Ionograph® results µg/in2

Low volatile residue µg/in2

21 22 23

1 1 1

8.05 3.32 2.76

Not performed Not performed Not performed

Remarks

Batch 1 mean ionic contamination level = 4.71 (µg/in2)

6.2.5.5. Outgassing data The two samples on which the ASTM E595 outgassing test was performed were part of Test Run #3. Table 5. Outgassing data from test run #3 Test PWA Batch Serial No. No.

TML %

CVCM TML% WVR %

Remarks

121

2

0.260

0.002

0.187

101

2

0.253

0.000

0.184

This sample was a printed wiring board only. This sample was a printed wiring board assembly.

7. SUMMARY OF RESULTS

The results are summarized as follows: ● PWAs cleaned with 5.0 minutes wash and 2.5 minutes rinse had average ionic cleanliness level of 0.27 micrograms per square inch, far below the JPL maximum acceptable ionic cleanliness level of 10 micrograms per square inch. This result is much lower than that obtained by the older cleaner. (See Tables 1 and 4). ● PWAs cleaned with 5.0 minutes wash and 2.5 minutes rinse had average low volatile residue (LVR) cleanliness level of 4.84 micrograms per square inch. Although no standard exists for LVR, it is less than 6.45 micrograms per

238

●

●

●

●

●

•

J.K. Bonner and A. Mehta

square inch, which is the lowest level of the flight hardware determination standard MIL-STD-1246C Level A. PWAs cleaned with 3.0 minutes wash and 2.0 minutes rinse had average ionic cleanliness level of 0.93 micrograms per square inch, far below the JPL maximum acceptable ionic cleanliness level of 10 micrograms per square inch. This result is much lower than that obtained by the older cleaning system. (See Tables 2 and 4). The results, however, are not optimal. PWAs cleaned with 6.0 minutes wash and 6.0 minutes rinse had average ionic cleanliness level of 1.97 micrograms per square inch, still far below the JPL maximum acceptable ionic cleanliness level of 10 micrograms per square inch. These somewhat higher results may be due to the inadvertant contamination by handling of some of the boards. Since the ionic cleanliness level is still significantly lower than 10 micrograms per square inch, this result does not vitiate the overall performance of the new cleaning system. The anion profile analysis performed with ion chromatography showed exceedingly low levels of anion species, thus indicating very low levels of remaining contamination. One PWA after cleaning had its components removed to examine for flux residues. Both visual and 10x magnification were used to detect residues. Nothing was noted. The outgassing data for the boards cleaned using the new centrifugal cleaning system/aqueous chemistry indicates that the total mass loss (TML) is much less than 1.00% and the collected volatile condensable material (CVCM) is much less than 0.10%. The optimal cleaning cycle suggested by the data is: • Wash solution temperature 50°C • Rinse solution temperature 50°C • Dry air temperature 200°C • Wash cycle rotational speed 150 RPM • Wash time 5 min. • Rinse time 2.5 min. • Dry time 2.5 min.

Note on ESD There was some concern that during the hot air drying stage there might be an ESD (electrostatic discharge) problem. A medical device manufacturer that purchased a Speedline Technologies ACCEL MicrocelTM Centrifugal Cleaning System was concerned about this and performed a thorough investigation. They found no ESD problem. In addition, at JPL a normal cleaning cycle was run and an ESD meter was used to see if there was any ESD build-up. No ESD was de-

Qualifying a cleaning system for space flight printed wiring assemblies

239

tected either on the boards or on the equipment. Hence, it is concluded that no ESD problem exists. 8. CONCLUSION

The centrifugal cleaner using the new aqueous cleaning solution based on longchain alcohols shows a marked improvement in cleanliness of PWAs over the previous two-stage batch semi-aqueous (SA) cleaning system using a terpenebased SA material and water in machine #1 for cleaning and saponifier, isopropyl alcohol (IPA), and D.I. water in machine #2 for rinsing. The centrifugal cleaner using the new aqueous cleaning solution not only cleans at a higher degree of cleanliness level compared to the older SA cleaning system, but also it is cost effective to use. The total cycle time is about 50% less than the older SA cleaning system. Also, it uses single chemical (the long-chain alcohol/aqueous solution) with very small amount of additives compared to three chemicals used by the older SA cleaning system. The use of hazardous isopropyl alcohol is also eliminated. Acknowledgements The research to qualify this new cleaning system was performed at the Surface Mount Technology Laboratory at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration. The authors wish especially to thank Mr. Charles J. Bodie and Mr. Amin Mottiwala for their support and encouragement. REFERENCES 1. J.K. Bonner, in Cleaning Printed Wiring Assemblies in Today’s Environment, L. Hymes (Ed.), pp. 65-119, Van Nostrand Reinhold, New York (1991). 2. L. Hymes (Ed.), Cleaning Printed Wiring Assemblies in Today’s Environment. Van Nostrand Reinhold, New York (1991). 3. C.J. Tautscher, Contamination Effects on Electronic Products. Marcel Dekker, New York (1991). 4. C.J. Tautscher, The Contamination of Printed Wiring Boards and Assemblies. Omega Scientific Services, Bothell, WA (1976). 5. F. Cala and A.E. Winston, Handbook of Aqueous Cleaning Technology for Electronic Assemblies. Electrochemical Publications, Isle of Man (British Isles) (1996). 6. J.B. Durkee, The Parts Cleaning Handbook without CFCs: How to Manage The Change. Hanser-Gardner, Cincinnati, OH (1994). 7. M.C. McLaughlin and A.S. Zisman, The Aqueous Cleaning Handbook: A Guide to CriticalCleaning Procedures, Techniques and Validation. The Morris-Lee Publishing Group, Rosemont, NJ (1998).

This page intentionally left blank

Surface Contamination and Cleaning, Vol. 1, pp. 241–247 Ed. K.L. Mittal © VSP 2003

Investigation of modified SC-1 solutions for silicon wafer cleaning CHRISTOPHER BEAUDRY∗ and STEVEN VERHAVERBEKE Applied Materials, 974 E. Arques Ave, M/S 81307, Sunnyvale, CA 94086, USA

Abstract—The RCA clean is widely used in the semiconductor industry for many wet-chemical cleaning processes. The RCA clean consists of a particle removal step, the Standard Clean 1 or SC-1 and metallic impurity removal step, the Standard Clean 2 or SC-2 step. In this work we have investigated the addition of chelating agents in SC-1 solutions to prevent metallic deposition during the SC-1 step as well as to remove metallic contamination. We also have studied the effect of surfactants in such solutions on sub-micrometer particle removal. This leads to the development of a very fast and efficient single step RCA replacement clean. The use of a single step cleaning strategy in a single wafer mode dramatically reduces the cycle time of cleaning. Keywords: RCA clean; silicon wafer cleaning; chelating agent; modified SC-1.

1. INTRODUCTION

SC-1 cleaning is widely used in the semiconductor industry during various wetchemical cleaning processes due to its outstanding particle removal efficiency. Although SC-1, a mixture of NH4OH/H2O2/H2O, is an efficient particle removal solution, it inherently allows some metallic impurities in solution to deposit on the wafer surface [1]. For this reason a conventional SC-1 is typically followed by SC-2, a mixture of HCl/H2O2/H2O, which exhibits excellent metallic impurity removal efficiency [2]. This sequence of SC-1 and SC-2 is known as the RCA clean and has been in use for over 30 years. The most obvious advantage of adding an appropriate chelating agent to SC-1 is to prevent the deposition of metallic impurities during the particle removal step and thus to eliminate the need for a follow-up metallic impurity removal step. Not only does this reduce the number of chemical cleaning steps required, saving money and time, it also avoids the adverse effect of particle re-deposition during typical metallic impurity removal steps, such as SC-2 or an HF dip. Furthermore, an appropriately chelate enhanced SC-1 solution can potentially remove metallic contamination ∗

To whom all correspondence should be addressed. Phone: (408) 584-0957, Fax: (408) 584-1132, E-mail: [email protected]

C. Beaudry and S. Verhaverbeke

242

even more efficiently than SC-2, and its ability to bind free metal ions in solution will potentially isolate process excursions from affecting process yield. To understand the effect of adding a chelating agent to an SC-1 solution, it is important to study the interaction of metallic impurities in solution and the substrate in that solution. In this case, the substrate of interest is silicon. In aqueous solutions, such as SC-1, a silicon wafer surface is hydroxyl terminated (silanol groups: -Si-OH). The interaction between the metal ions in solution and the silanol surface groups can be described by the following equation: -Si-O-H + Mx+ ó -Si-O-M(x-1)+ + H+ x+

(1)

where M is the metallic ion. From equation (1), one can see that there are two ways to reduce metallic ions from depositing on the wafer surface. The first way is to increase the concentration of H+, shifting the reaction to the left. Unfortunately, acidifying SC-1 will degrade particle removal effectiveness of the solution (the high pH provides electrostatic repulsive forces while lowering the pH may result in attractive forces between particles and the substrate). The second way to prevent or reduce metallic ion deposition is to decrease the free metal ion concentration in solution. For many years, suppliers have supported such an approach by the development and use of ultrapure materials, chemicals, and de-ionized water. Due to the increasingly stringent requirements of wafer surface cleanliness, this approach alone cannot reach today’s required level of surface metals. In order to reduce metal deposition in SC-1 solutions to meet and even exceed the current surface metal specifications, it is necessary to not only use ultrapure components, but to also add chelating agents to bind the free metal ions present forming complexes which will remain soluble in solution. Typical chelating agents can reduce the free metal ions in solution by 6 orders of magnitude [3]. In addition to enhancing the metallic cleaning ability of SC-1 solutions, we have also investigated the use of a surfactant in our modified SC-1 solution. Although SC-1 inherently removes particles quite effectively, megasonic energy is often applied which dramatically increases particle removal efficiency. This is increasingly important as the dimensional size of semiconductor devices continues to decrease to even smaller sizes. With this mind, the addition of surfactants to SC-1 will become an important component to prevent particles removed from the wafer surface from re-deposition, thus increasing the particle removal efficiency for small particles. In liquids, the attraction or repulsion of particles to the wafer surface is dependent on the van der Waals interaction (always attractive) and the electrostatic double layer forces (usually repulsive). The combination of these interactions will determine the potential energy of interaction and thus the barrier to adhesion [4, 5]. The barrier to adhesion is related to the particle size, pH of the solution, and the respective charges on the wafer surface and particle. Cleaning down to submicrometer and smaller sizes becomes increasingly difficult as the barrier to adhesion decreases with decreasing particle size. Thus the tendency to re-deposit on the wafer surface increases as the particle size decreases. Surfactants may prevent

Investigation of modified SC-1 solutions for silicon wafer cleaning

243

deposition in two ways (i) electrostatically by increasing surface potentials and (ii) physically by steric hindrance not allowing particles to get close enough to the surface for van der Waals interaction to dominate. The focus of this work was to study the effectiveness of chelate and surfactant modified SC-1 solutions for reducing metallic ion deposition, removing metallic ions, and particle removal. In addition, we studied the potential for organic contamination residue from both the chelating agent and surfactant. 2. EXPERIMENTAL

We carried out experiments using a modified SC-1 solution with a composition of 1:2:40 (NH4OH:H2O2:H2O) to 1:2:80 (by volume). The concentration of chelating agent (carboxylic acid based) was varied, but was less than 1wt% of the solution. The concentration of the surfactant (Valtron SP2200 manufactured by Valtech Corporation, USA) was also varied, but was less than 1wt%. The measured pH value was approximately 9.6. Megasonic energy was applied during the modified SC-1 step (power density 1.13 W/cm2). The process time was 30 to 60 seconds at a temperature of 50°C or 80°C followed by a rinse at the same temperature and a spin dry. All wafers were cleaned in a single wafer mode. Sample wafers for particle removal studies were prepared with an automated aerosol particle deposition tool made by MSP Corporation, USA (Model 2300D). The particle deposition pattern was a combination of full random coverage and a spot (see Figure 1 for an example). In total, approximately 2300 Si3N4 particles

Figure 1. Particle Removal - Example of Si3N4 particle wafer maps before (left) and after (right) modified SC-1 clean (particles ≥ 0.12 µm, measured on Tencor SP-1).

244

C. Beaudry and S. Verhaverbeke

were deposited on prime 300 mm wafers. The particle measurements were performed on a Tencor SP-1 instrument. Surface metal measurements were obtained with the vapor phase decomposition-ion coupled plasma mass spectroscopy (VPD-ICPMS) technique. Time-of-flight secondary ion mass spectrometry (TOFSIMS) was used to assess if any residual chelating agent or surfactant remained on the wafer surface (after the rinsing and drying). 3. RESULTS

Figure 2 shows the particle removal efficiency for an optimized modified SC-1 solution. For one lot, consisting of 13 wafers, the average particle removal efficiency was 99.5% (1σ = 0.28; measured at ≥ 0.12 µm). Typical wafer maps illustrating the combination of full and spot Si3N4 particle deposition pattern before and after processing are shown in Figure 1. In order to determine the effectiveness of the surfactant studied, Valtron SP2200, we compared final particle counts with and without surfactant present (Figure 3). In this example we see an average of 50 less particles per 300 mm wafer. The particle cleaning performance for the modified SC-1 solution was excellent and the selected surfactant reduced the average final particle count after SC-1 cleaning. Figure 4 illustrates the effectiveness of the studied chelating agent for reducing metallic deposition. In particular, it is interesting to look at the level for Al, Fe, and Zn. These are some of the metals that readily deposit from conventional SC-1

Figure 2. Particle Removal - Average Si3N4 particle removal after modified SC-1 clean (particles ≥ 0.12 µm, measured on Tencor SP-1).

Investigation of modified SC-1 solutions for silicon wafer cleaning

245

Figure 3. Particle Removal - Final particle counts after modified SC-1 clean with and without surfactant (particles = 0.1-0.14 µm, measured on Tencor SP-1).

Figure 4. Metal Deposition - Surface trace metals levels after modified SC-1 clean as determined by VPD-ICPMS (1 sigma error bars are generally within data points).

246

C. Beaudry and S. Verhaverbeke

solutions [1]. The average surface metals levels after the modified SC-1 clean was equal to or below today’s VPD-ICPMS detection limits. For reference, typical levels for a conventional SC-1 last clean are: Al ~ 1x1011, Fe ~ 2x1010, and Zn ~ 1x1011 (atoms/cm2). The chelating agent under investigation is efficiently binding the free metal ions in solution reducing their deposition onto the wafer surface and thus can eliminate the need for an additional metal removal step. Although the modified SC-1 solution did not deposit metals from the solution we also characterized the metal removal efficiency of this solution (Figures 5 and 6). The concentration of chelating agent and exposure time were varied while the NH4OH:H2O2:H2O volume ratio and temperature were held constant at 1:2:80 and 80°C. Figure 5 shows the results for a 30 s processing time. The final concentration of all metals investigated was typically greater than 1E+10 atoms/cm2. Fe removal was found to be a function of chelating agent concentration while other

Figure 5. Metal Removal - Surface trace metals levels after a 30 s modified SC-1 clean using different chelating agent concentrations as determined by VPD-ICPMS.

Figure 6. Metal Removal - Surface trace metals levels after a 10 minute modified SC-1 clean using different chelating agent concentrations as determined by VPD-ICPMS.

Investigation of modified SC-1 solutions for silicon wafer cleaning

247

metals did not exhibit any dependence. Figure 6 shows the final concentrations of metal after 10 minutes of exposure. All metals were reduced by 2-3 orders of magnitude to close to or below 1E+10 atoms/cm2. Exposure time is obviously an important consideration for metal removal. Methods to increase the metal removal rate are now under investigation. One of the concerns with the use of chelating agents and/or surfactants in SC-1 last clean is the potential for organic contamination remaining on the surface of the wafer. However, in a typical spin cleaning equipment, the rinse process can be optimized to eliminate such concerns. The use of heated DI water rinse and high spin rates during rinsing can effectively remove all traces of the chelating agent and surfactant. TOF-SIMS measurements were carried out to confirm the absence of both organic additives. No trace of additive-specific residues was observed on the processed wafers. Thus, organic contamination through use of appropriately selected chelating agents and/or surfactants can be eliminated through process optimization. The absence of any heavy metal signature in the TOF-SIMS data also confirms our VPD-ICPMS results. 4. SUMMARY

In this paper we have shown that the addition of an appropriately selected chelating agent to SC-1 solutions can eliminate the need for an additional metal removal step, potentially saving time and money. In addition, the use of a surfactant can enhance particle removal efficiencies for very small particle sizes (<0.14 µm). This modified SC-1 solution, containing both additives, was shown to have excellent particle removal efficiency, to reduce metal deposition on the wafer surface to below VPD-ICPMS detection limits, and to remove surface metal contamination. Methods to improve metal removal efficiency are currently underway. Furthermore, rinsing can be optimized to eliminate all traces of the chelating agent and surfactant residues. REFERENCES 1. H. Hiratsuka, M. Tanaka, T. Tada, R. Yohsimura and Y. Matsushita, Ultra Clean Technol., 3, No. 3, 18-27 (1991). 2. W. Kern, in: Proceedings of the First International Symposium on Cleaning Technology in Semiconductor Device Manufacturing, J. Ruzyllo and R.E. Novak (Eds.), Vol. 90-9, pp. 3-19, Electrochemical Society, Pennington, New Jersey (1990). 3. A. Ringborn, Complexation in Analytical Chemistry, John Wiley & Sons, New York (1963). 4. R. Donovan and V. Menon, in: Handbook of Semiconductor Wafer Cleaning Technology: Science, Technology, and Applications, W. Kern (Ed.), pp. 152-197, Noyes Publications, Westwood, New Jersey (1993). 5. M. Itano and T. Kezuka, in: Ultraclean Surface Processing of Silicon Wafers: Secrets of VLSI Manufacturing, T. Hattori (Ed.), pp. 115-136, Springer-Verlag, Berlin (1995).

This page intentionally left blank

Surface Contamination and Cleaning, Vol. 1, pp. 249–260 Ed. K.L. Mittal © VSP 2003

Performance qualification of post-CMP cleaning equipment in a semiconductor fabrication environment MICHAEL T. ANDREAS∗ Micron Technology, Inc., Mail Stop 306, 8000 S. Federal Way, Boise, ID 83707-0006, USA

Abstract—An inexpensive qualification technique is described for wafer cleaning tools used after chemical-mechanical polishing (CMP). Pipette deposition of slurry onto a monitor wafer can provide the particle challenge needed to qualify the performance of a post-CMP wafer cleaning tool. In addition to gauging the performance of these tools, this pipette method is faster and less expensive than many common particle deposition techniques, including immersion, polishing and aerosol deposition. Keywords: Brush cleaning; chemical-mechanical polishing; CMP; particle deposition; PVA; slurry; wafer cleaning.

1. INTRODUCTION

Surface preparation and cleaning is one of the most critical steps in semiconductor manufacturing [1]. For all wafer cleaning tools, routine qualification is necessary to ensure that no particle contamination is introduced by the wafer cleaning equipment [2]. The broad category of wafer cleaning tools includes the poly(vinyl alcohol) (PVA) brush scrubbing tool [3]. The brush scrubber has been increasingly utilized [4] in semiconductor fabrication as a preferred technique for particle removal after CMP. Because of the high particle removal performance required of post-CMP cleaning tools, it is critical to monitor and maintain the performance of such tools [5]. The most direct measure of tool performance is inline inspection of actual product wafers [6]. While inline defect analysis is invaluable, it may require a time lag of hours or even days between wafer cleaning and discovery of high wafer defectivity. In a high-volume manufacturing environment, this delay can lead to hundreds of product wafers with possible contamination. For this reason, inline inspection of product wafers is supplemented by regular tool qualification using less expensive particle monitor (PMON) wafers. This PMON qualification should provide an accurate measure of the tool performance with the quickest possible turnaround time. For post-CMP cleaning tool qualification, it is neces∗

Phone: (208) 368-5067, Fax: (208) 368-2548, E-mail: [email protected]

250

M.T. Andreas

sary to use prepared PMON wafers with contamination analogous to that found on polished product wafers. There are several methods for preparing these contaminated monitor wafers. One way is to use polished monitor wafers [7]. In this method, test wafers can be selected which represent the surface chemistry of product wafers without the expense of underlying circuitry. The test wafer surface can be homogeneous or heterogeneous, depending on the process being qualified. These test wafers can be polished under the same CMP conditions as product wafers. This method requires the same resources as inline product inspection, therefore providing an accurate measure of post-CMP tool performance but not necessarily decreasing the turnaround time. Another common method is the slurry dip [8, 9], where monitor wafers are dipped in a wet process tank of diluted slurry. Again, monitor wafers can be selected to represent the surface chemistry of product wafers. The diluted slurry can be selected to simulate the CMP chemistry. This method has the advantage of providing contaminated monitor wafers without the time or expense of using a CMP tool. Also, it is possible to deposit dry particles using an aerosol deposition technique [10]. While it is claimed that this method is more controllable and repeatable than aqueous slurry immersion, these dry particles may not represent polishing residue as accurately as a CMP slurry. Here we describe an extremely simple contamination technique – direct pipette deposition of a small volume of undiluted slurry onto a monitor wafer. 2. EXPERIMENTAL

Bare silicon and blanket oxide wafers were used for all tests. Blanket oxide wafers were prepared by plasma enhanced chemical vapor deposition (PECVD) using tetraethoxysilane (TEOS). These TEOS derived films were deposited to 350 nm thickness on 200 mm diameter silicon substrates. All wafers were cleaned using OnTrak DSS-200 Series II brush cleaning tools. These tools were run using a dilute (<1%) basic cleaning solution. All wafers were inspected with a Tencor SurfScan 6420 laser scattering wafer inspection tool [11, 12]. Bare silicon monitor wafers were inspected for all light-scattering point defects (LPDs) >0.16 µm. Blanket oxide wafers were inspected for LPDs > 0.18 µm. 3. RESULTS

3.1. Deposition of the slurry drop Slurry drop testing was first used to investigate the scrubber response to different slurry types. The slurries investigated contained abrasive materials of alumina, ceria, fumed (furnace-grown) silica, and colloidal (solution-grown) silica. In this experiment, a large (~ 0.2 ml) drop of each slurry was deposited directly onto the center of each blanket oxide wafer. These contaminated wafers were processed through the wafer scrubber. The total brush cleaning time was varied as an ex-

Performance qualification of post-CMP cleaning equipment

251

Table 1. Initial slurry drop test results Experimental conditions

SurfScan total LPD counts > 0.16 µm

Run order

Wafer surface

Contamination

Brush time, sec

Before slurry deposition

After slurry Difference deposition and PVA scrub process

1 2 3 4 5 6 7 8-24 25 26 27 28 29 30 31 32-48 49 50 51 52 53 54 55 56 57 58 59 60 61 62

silicon CVD oxide CVD oxide CVD oxide CVD oxide CVD oxide silicon silicon silicon silicon CVD oxide CVD oxide CVD oxide CVD oxide silicon silicon silicon silicon CVD oxide CVD oxide CVD oxide CVD oxide silicon silicon CVD oxide CVD oxide CVD oxide CVD oxide CVD oxide silicon

none alumina alumina alumina alumina alumina none none none none fumed silica fumed silica fumed silica fumed silica none none none none ceria ceria ceria ceria none none colloidal silica colloidal silica colloidal silica colloidal silica colloidal silica none

80 2 20 40 60 80 80 80 80 80 2 20 40 80 80 80 80 80 2 20 40 80 80 80 2 20 40 60 80 80

4 34 36 37 33 35 12 – 6 4 32 32 65 120 5 – 2 2 62 57 72 111 4 12 29 35 40 26 52 3

256 18596 465 96 65 76 596 – 113 33 786 54 32 32 45 – 296 21 2394 40 35 29 20 52 2680 77 39 18 37 35

252 18562 429 59 32 41 584 – 107 29 754 22 -33 -88 40 – 294 19 2332 -17 -37 -82 16 40 2651 42 -1 -8 -15 32

perimental factor. Uncontaminated silicon wafers were processed immediately before and after the oxide wafers to determine any slurry particle carryover. The results from this test are summarized in Table 1. A plot of post-scrub LPD count versus brush cleaning time is shown in Figure 1. Here we determined that among all the slurry types, alumina slurry provided the highest level of wafer contamina-

252

M.T. Andreas

Figure 1. Plot of post-scrub LPD total vs. brush time.

tion (as measured by laser scatterometry) for a given drop size. Maps showing the LPD distribution after 2 second scrubbing are shown in Figures 2 and 3 for wafers contaminated with alumina and colloidal silica, respectively. In all cases, the 80 second brush cleaning time was sufficient to attain particle levels below the predefined production limit of 100 LPDs. Silicon monitor wafers run before and after each group of slurry drop test wafers did not show significant slurry carryover. Due to the simplicity of this procedure, slurry drop deposition was investigated further as a method for routine tool qualification. 3.2. Development of the slurry drop qualification method To evaluate the resolution of scrubber qualification methods (SQMs), several experimental scrubber recipes were created which simulated sub-optimal tool performance [13]. These scrubber recipes are described in Table 2. Several different SQMs were evaluated using these sub-optimal scrubber recipes. These methods, including slurry drop, slurry immersion and CMP polishing, are described in Table 3. The oxide polish method (SQM index 6) had been in use in our production line for some time prior to this experiment. Silica-containing slurries were used for all contamination methods because these slurries were the most readily available at the time. For the slurry immersion methods, 10 ml of slurry was diluted with 18 L of deionized water, and the wafers were immersed for 10 sec immediately before cleaning. For the polished oxide wafers, a 60 second pre-clean using ~ 0.5% HF was utilized between polishing and scrubbing. Many trials were repeated using new (freshly installed) and old (near end of service) PVA brushes. For each combination of qualification method and experimental scrubber recipe,

Performance qualification of post-CMP cleaning equipment

253

Figure 2. Wafer map showing LPD distribution after alumina slurry drop and 2 second scrub.

Figure 3. Wafer map showing LPD distribution after colloidal silica slurry drop and 2 second scrub.

M.T. Andreas

254 Table 2. Experimental scrubber recipes Recipe

Brush height, mm

Chemical flow, L/min

Brush rotation, 1/min

Brush time, sec/brush

Rinse time, sec

C E1 E2 E3 E4 E5 E6 E7

3.5 1.0 3.5 3.5 3.5 3.5 3.5 3.5

0.5 0.5 0.5 0.5 0.5 0.0 0.5 0.5

139 139 40 139 139 139 139 139

40 40 40 10 20 40 40 40

9 9 9 9 9 9 5 13

Table 3. Experimental scrubber qualification method details SQM index

Wafer surface

Contamination

1 2 3 4 5 6

silicon silicon silicon CVD oxide CVD oxide CVD oxide

2 drops (~ 0.1 ml) slurry dilute slurry immersion none 2 drops (~ 0.1 ml) slurry dilute slurry immersion 30 sec polish and 60 sec dilute HF clean

the difference in LPD counts (dLPD) was determined whereby dLPD = LPD(post cleaning) – LPD(pre-contamination). Comparing dLPDs for all experimental scrubber recipes for each series of SQM, slurry type and brush condition, a method dynamic range (MDR) was determined as the range between the highest and lowest dLPD results for that series. This dynamic range gives an indication of the utility for a given procedure to “catch” sub-optimal tool performance. The results from all trials are presented in Table 4. Considering dLPDs for new brush vs. old brush conditions, the most sensitive method for monitoring brush wear is clearly the bare silicon SQM (index 3). Considering the method dynamic range across scrubber recipes, the silicon wafer methods (indices 1-3) in general are more sensitive than the oxide wafer methods (indices 4-6). This may be due, in part, to the higher sensitivity of the SurfScan inspection used for bare silicon wafers. Among the silicon wafer methods, the silicon drop SQM (index 1) provides the most accurate measure of scrubber performance independent of brush age.

Performance qualification of post-CMP cleaning equipment

255

Table 4. Experimental details, dLPD results and method dynamic range (MDR) for each series Series conditions

dLPD for each scrubber recipe

MDR

SQM Silica index

Brush age

C

E1

E2

E3

E4

E5

E6

E7

1 1 2 2 3 3 3 3 4 4 4 4 5 5 6 6

old new old new old old new new old old new new old new old new

105 5 114 2 3857 357 68 10 5 39 3 5 243 -75 4 3

242 545 324 26 1107 565 126 25 21 68 55 7 13 -46 51 48

69 10 95 40 1547 149 – 15 -2 83 120 11 22 5 -1 -18

365 8 817 32 – 1020 – 10 8 – 38 – 34 1 43 -17

103 5 84 8 7495 403 33 4 9 92 -26 1 43 14 8 -50

66 30 84 12 >30000 228 72 11 0 3 26 10 133 -40 66 6

48 20 121 7 8421 204 – 24 41 – 26 2 50 -71 82 -32

1516 17 1275 8 3446 3485 – 19 1 30 99 0 2 -58 56 -3

colloidal fumed colloidal fumed – – – – colloidal fumed fumed fumed colloidal fumed colloidal colloidal

1468 540 1191 38 >30000 3336 96 21 43 89 146 10 241 89 82 98

3.3. Improvement of the slurry drop scrubber qualification method After implementation of SQM index 1 for a period of time, a new defect pattern was discovered on product wafers that was related to brush-induced residuals at tungsten CMP scrub. This defect pattern was characterized by a radial pattern of slurry residuals. Although this defect pattern was detectable on product wafers, it did not appear on scrubber qualification wafers. This detection gap led to further optimization of the scrubber qualification method. First, the full experimental space of wafer type, slurry type, and slurry amount versus scrubber performance was revisited. To simulate sub-optimal scrubber performance, two new scrubber recipes were created: A1 and A2. Both recipes feature a reduced brush pressure (adjusted by way of brush height) with shorter brush process times. Also, recipe A2 uses a slower brush rotation. The experimental conditions and inspection results are given in Table 5. Several wafer maps from this group are shown in Figures 4 and 5. The reduced efficiency of recipe A1 provided good discrimination between qualification parameters (e.g. slurry type or wafer type). In general, alumina slurry was more sensitive than silica slurry to radial defect pattern formation. As for slurry quantity, three drops of alumina appears optimal. Silicon wafers worked better than oxide wafers because the dLPD and MDR results corresponded more accurately to the expected particle removal performance of

M.T. Andreas

256

Table 5. The full factorial of scrubber qualification tests using a known good brush installation Series conditions

dLPD for each scrubber recipe

Wafer surface

Slurry

Quantity

C

A1

A2

silicon silicon silicon silicon CVD oxide CVD oxide CVD oxide CVD oxide

silica silica alumina alumina silica silica alumina alumina

1 drop 3 drops 1 drop 3 drops 1 drop 3 drops 1 drop 3 drops

25 38 69 108 -14 -4 -2 247

125 125 508 2574 2850 376 >20188 >29959

>29997 >28245 >29997 >29997 701 1535 >29932 >29933

MDR

>29972 >28207 >29928 >29889 2864 1539 >29934 >29712

Figure 4. Several wafer maps from the SQM optimization tests. These silicon wafers were scrubbed using recipe A1. The lower left note in each frame indicates drop size (1d = 1 drop, 3d = 3 drops) and slurry composition (S = silica, A = alumina).

Performance qualification of post-CMP cleaning equipment

257

Figure 5. Wafer map for a CVD oxide wafer contaminated with one drop of alumina slurry and scrubbed using recipe A1.

Table 6. Recipe details for more sub-optimal scrubber recipes used to test the sensitivity of optimized SQMs. Recipe C is the control scrubber recipe. Recipe

C A1 A2 T1 T2 T3

Brush Module 1

Brush Module 2

Rotation, 1/min

Height, mm Time, sec

Rotation, 1/min

Height, mm

Time, sec

139 139 38 139 139 139

3.5 1.5 1.5 1.5 1.5 3.5

139 139 38 139 139 139

3.5 1.5 1.5 1.5 3.5 1.5

40 30 30 40 40 40

40 30 30 40 40 40

experimental scrubber recipes A1 and A2. The improved SQM using 3 drops of alumina slurry was selected for comparison to the previous SQM using 1 drop of colloidal silica slurry. Based on results with experimental recipes A1 and A2, three more sub-optimal scrubber recipes were created. These recipes (T1, T2 and T3) were designed to perform somewhere between recipes C (control) and A1. These recipes are described in Table 6. Bare silicon wafers were used for all further tests. Wafer run order was randomized to average out any brush carryover effects. Process details and results are shown in Table 7. Wafer maps showing sensitivity to radial defect pattern formation are shown in Figure 6. The dLPD results comparing optimized alumina and silica slurry drop methods are plotted in Figure 7. The alumina slurry drop SQM shows better sensitivity to inefficient scrubber

M.T. Andreas

258

operation, especially using recipe T1. All wafers using silica and alumina SQMs met contol levels for dLPDs when using the control scrubber recipe. One of the silica drop trials showed anomalously high residue. This may have been carryover from the previously scrubbed wafer, which brought alumina slurry contamination. Further tests confirmed that carryover from 3 drops of alumina appeared when using recipe T1. All experiments confirmed the improved alumina drop SQM as more sensitive to conditions which may cause radial defect patterns. After implementing this improved SQM, no further radial defect patterns were discovered on product wafers. Table 7. Process details and inspection results for improved slurry drop SQM comparison. All tests were run using bare silicon wafers. The high dLPD result for 1dS wafer 19 (5312 adders) may be due to carryover from 3dA wafer 18 Series conditions

dLPD for each scrubber recipe

MDR

Slurry

Quantity

Run order

C

T1

T2

T3

silica

1 drop

1, 4, 9, 11 2, 5, 13, 15 3, 6, 22, 19

30 30 38

34 23 34

67 69 83

236 308 5312

206 271 5278

alumina

3 drops

7, 12, 14, 8 17, 18, 16, 10 20, 24, 21, 25

19 100 92

3711 30553 31077

80 169 128

202 580 468

3692 30453 30985

Figure 6. Representative wafer maps from the 3 drops alumina (3dA) qualification method on silicon wafers, showing sensitivity to radial defect patterns when using recipes T1 and T3.

Performance qualification of post-CMP cleaning equipment

259

Figure 7. Plot of dLPD versus scrubber recipe for optimized alumina and silica slurry drop SQM trials.

4. CONCLUSION

Manual pipette deposition of CMP slurry onto a monitor wafer is a quick and effective way to provide qualitative particle challenges to wafer cleaning equipment. This method is much faster and cheaper than other common particle deposition techniques, including polishing, aerosol deposition and immersion in dilute slurry. We have shown that this method provides enough particle loading to determine whether or not a post-CMP cleaning tool will perform within acceptable particle removal limits. This, in turn, translates to lower defects on product wafers and improved yields at a lower quality control cost. REFERENCES 1. W. Kern (Ed.), Handbook of Semiconductor Wafer Cleaning Technology, pp. 416-419, Noyes Publications, Park Ridge, NJ (1993). 2. F.W. Kern, Jr. and G.W. Gale, in: Handbook of Semiconductor Manufacturing Technology, Y. Nishi and R. Doering (Eds.), pp. 87-104, Marcel Dekker, New York (2000). 3. W. Krusell, J.M. de Larios and J. Zhang, Solid State Technol., 38, No. 6, 109-114 (1995). 4. R. DeJule, Semiconductor Intl., 56-64 (Nov. 1998). 5. J.M. de Larios, J. Zhang, E. Zhao, T. Gockel and M. Ravkin, MICRO, 15, No. 5, 61-73 (1997). 6. C. Dennison, MICRO 16, No. 2, 31-42 (1998). 7. D.W. Cooper, R.C. Linke and M.T. Andreas, MICRO 17, No. 7, 55-64 (1999). 8. A.A. Busnaina, N. Moumen, M. Guarrera and J. Piboontum, in: Semiconductor Fabtech – 9th Edition, M.J. Osborne (Ed.), pp. 279-282, ICG Publishing, London (1999). 9. S. Ramachandran, A.A. Busnaina, R. Small, C. Shang and Z. Chen, in: Semiconductor Fabtech – 13th Edition, G. Oliver (Ed.), pp. 271-277, ICG Publishing, London (2001).

260

M.T. Andreas

10. Y.H. Liu, S.H. Yoo, S.K. Chae, J.J. Sun, K. Christenson, J. Butterbaugh, J.F. Weygand and N. Narayanswami, Semiconductor Intl., 145-152 (June 2000). 11. R.S. Howland, Semiconductor Intl., 164-170 (Aug. 1994). 12. J.J. Shen and L.M. Cook, MICRO 15, No. 3, 53-66 (1997). 13. N. Moumen, M. Guarrera, J. Piboontum and A.A. Busnaina, in: Proceedings, 10th Annual IEEE/SEMI Advanced Semiconductor Manufacturing Conference, pp. 250-254 (1999).

Surface Contamination and Cleaning, Vol. 1, pp. 261–266 Ed. K.L. Mittal © VSP 2003

Spatial and temporal scales in wet processing of deep submicrometer features MOSHE OLIM∗ Seagate Technology, 7801 Computer Avenue, South Bloomington, Minnesota 55435, USA

Abstract—Liquid-phase processing is commonplace in manufacturing of thin films. Typically, the surface processed has distinct topological features such as trenches and vias. A typical liquid phase process cycle starts with a dry surface and consists of the following steps: (1) wetting of the surface, (2) dispensing a mix of chemical reagents, (3) rinsing the surface, and (4) drying the surface. Step 2 may consist of a sequence of chemical reagents either with or without a rinse in between. Each of the steps is governed by different physical processes which may have distinctly different spatial and temporal scales. These scales are addressed in the paper. A trench is used as a representative feature. Keywords: Microscale transport; hydrophilic surface; thin films; wet processing.

1. INTRODUCTION

The typical processed surface is dry as the process starts. If a chemical reagent is dispensed onto a dry surface it is likely that different parts of the surface will be subjected to the reagent for different time intervals thus resulting in a nonuniform processing result. Therefore, prior to dispensing chemical reagents on the surface, it is imperative that the surface be covered with an inert liquid. This liquid is typically deionized water which covers the whole surface and fills the trenches. The time required to fill a trench depends strongly on the characteristics of the surface processed, the surface tension of the water, and the width and depth of the trench. If the trench is hydrophilic, the capillary action of the water/air interface will ensure that the trench fills with water. The geometry of the process is shown in Fig. 1. The pressure of the gas trapped in the feature is increased due to the capillary force, and this increase in pressure enhances the diffusion of the gas into the liquid. The process continues until the gas trapped in the cavity is completely consumed by this diffusion process. The upper limit on the time required to fill the trench can be estimated (for details see [1]) as follows:

∗

Phone: 952-402-5888, E-mail: [email protected]

M. Olim

262

Figure 1. Geometry of the trench filling process. PL is the pressure in the liquid, ∆P is the capillary pressure difference across the interface, w and h are the trench width and depth, respectively, and θ is the contact angle.

Figure 2. Time required to complete the trench filling process as a function of trench width for contact angle values of 30 and 60 degrees. The surface tension of the liquid is 70 mN/m, and trench aspect ratio h/w = 10.

t fill =

k æ 1 + Ph ö 2 ç ÷h DRT è ∆P ø

(1)

where k, D, R, T, and h are, respectively, a proportionality constant, vapor/air diffusion coefficient, universal gas constant, temperature, and trench depth, and Ph and ∆P are the atmospheric pressure and capillary pressure difference, respec-

Spatial and temporal scales in wet processing of deep submicrometer features

263

tively, across the interface. The results of the calculation are shown in Fig. 2. It can be seen that the time required to fill a 0.25 µm wide trench is well below one second even for a trench whose aspect ratio is 10. 2. TRANSFER OF REAGENTS INTO AND OUT OF THE TRENCH

With the trench full of DI water, a chemical reagent is dispensed onto the substrate. In the interest of uniformity of processing along the full depth of the trench, it is important that the spatial concentration of the reagent along the trench depth be kept as uniform as possible. For analysis purposes, one may assume that the trench is full of water and the top of the trench is covered with liquid reagent. The reagent may penetrate the trench through either (a) convective or (b) diffusive mixing. In order for convective mixing to take place, the flow characteristics must allow for vortices to exist. The possibility of vortex existence may be ruled out by comparing the relevant geometric parameters to the smallest vortex diameter predicted by Kholmogorov scales (see [2]) using the following equation:

η æ ul ö ≈ç ÷ l èvø

−3/4

(2)

where η is the smallest length scale that can sustain turbulence (i.e. the smallest vortex diameter possible), l is a characteristic length of the system (in this case it is the width of a trench), and u and v are the flow velocity and kinematic viscosity, respectively. With relatively large trench width and velocity values of l = 0.5 µm and u = 1 m/s one obtains η/l ≈ 1, and with more realistic values of l = 0.25 µm and u = 0.01 m/s one obtains η/l ≈ 100. Since the smallest possible vortex diameter is noticeably larger than the trench width, it is clear that turbulent mixing cannot take place in the trench. This implies that the reagent is transferred into the trench by diffusion only. The same argument applies to transfer of reagents out of the trench when DI water is dispensed onto the surface in order to stop the reaction. Since it is desired that the results of the chemical process be uniform along the depth of the trench, it is clear that the exposure time of any point on the trench wall to the reagent should be as close as possible to that at any other point along the trench wall. The uniformity of the process may be estimated by comparing the time it takes for the reagent concentration at the bottom of the trench to equalize with that at the top of the trench. For practical purposes, let us consider the concentrations equalized when the concentration at the bottom reaches 95% of the concentration at the top. The time required for the reagent concentration at the trench bottom to reach a given concentration level can be estimated (see [1]) using the following one-dimensional diffusion equation:

264

M. Olim

Figure 3. Normalized chemical concentration at the trench bottom vs time. u is the velocity of the liquid at the top of the trench, and D is the diffusion coefficient of the chemical in the liquid.

∂C ∂ 2C + 2 =0 ∂t ∂z

(3)

where D = 1.e-9 m2/s is the diffusion coefficient and C is the reagent concentration subject to C(t,z=h) = 1, Cz(t,z=0) = 0, C(t=0,z
Typically, the de-ionized water used to rinse the substrate must be removed (in the liquid state) from the substrate and out of the features fabricated in the substrate. Water that is not removed in the liquid state would evaporate leaving originally dissolved contaminants to coagulate on the substrate and in the features thus adversely affecting the yield of the manufacturing process.

Spatial and temporal scales in wet processing of deep submicrometer features

265

Substrate rotation, since it increases the body force on the water in the features, is often utilized as a means of water removal enhancement. However, as the typical size of the features decreases, the importance of body forces compared to surface tension (manifested in the Bond number) also decreases thus reducing the efficiency of rotation as a mechanism for water removal. It has been shown [4] that the amount of water removed from a trench for a given contact angle does not change significantly below Bond number of 0.1 which may, therefore, be defined as the critical Bond number. The parameter that determines the amount of water removed when the Bond number is below the critical value is the contact angle θ, and the maximum amount of water removed from the trench of width L does not exceed πL2/4 per unit length of a long trench. Since the depth of the trench is significantly larger than its width, very little water is removed from the trench in the liquid phase and most of the water must evaporate. 4. EVAPORATION

This step is typically achieved by purging the process chamber using dry nitrogen at ambient temperature and pressure. The equations describing the time interval required to dry by evaporation a feature of a given depth are developed in [3] and they allow for any temporal variation of vapor concentration in the ambient. The rate of evaporation through a stagnant gas in one dimension is expressed in Equation (17.2-15a) in [5]. A slightly modified version of this equation is presented below: N =

PD æ P − Pv ,top ö ln ç ÷ RTz çè P − Pv ,int ÷ø

(4)

where P and T, respectively, are the pressure and temperature in the system, D is the diffusivity of the vapor in the ambient gas, R is the universal gas constant, z is the distance between the top of the feature and the liquid/gas interface, and Pv is the partial pressure of the vapor either at the top of the feature (top) or at the liquid/gas interface (int). This may be converted into an equation showing the rate of recession of the air/liquid interface which, in turn, may be integrated (for details see [3]) to yield % ∆s æ 1− P v ,top ö %z 2 = ò ln ç ÷÷ ds % ç 0 è 1− Psat ø

(5)

where 1/2

æ ρ L RT ö %z = z ç ÷ è 2 M L PD τ ø

;

% P v ,top =

Pv ,top P

;

% = Psat , P sat P

(6)

M. Olim

266

and s = t/τ is nondimensional time with τ being a characteristic time scale. To solve Equation (5) the temporal variation of the vapor pressure at the top of the feature must be known. Two potentially realistic situations in which the vapor concentration as a function of time is known are (a) Constant vapor pressure Pv,top = αPsat where 0 ≤ α ≤ 1, and (b) Exponentially decreasing vapor pressure Pv,top = Psate–Qt/V where Q and V are, respectively, the flow rate of the dry nitrogen and the volume of the chamber. Situation (b) is the slower of the two, and, for a realistic set of values Q = 1200SLPM and V = 100L, a 1 µm deep trench would require ≈ 300 ms to evaporate. 5. SUMMARY

The four main steps in wet processing of submicrometer features on hydrophilic surfaces are: (1) wetting of the surface, (2) dispensing a mix of chemical reagents, (3) rinsing the surface, and (4) drying the surface. A trench was used as a representative feature. The mechanisms driving each of these steps have been analyzed analytically in this paper. Introduction of typical dimensions and physical values into the results of the analysis yielded the time scales relevant to each of the process steps. The mechanisms and their time scales are summarized in the table below: step

driving mechanism

time scale [ms]

wetting chemical in chemical out drying

capillary action diffusion diffusion evaporation

100 10 100 100

REFERENCES 1. M. Olim, J. Electrochemical Soc., 144, 4331-4335 (1997) . 2. H. Tennekes and J.L. Lumley, A First Course in Turbulence, The MIT Press, Cambridge, MA (1990). 3. M. Olim, J. Microscale Thermophys. Eng., 3, 183-188 (1999). 4. M. Olim, J. Microscale Thermophys. Eng., 4, 223-230 (2000). 5. R.B. Bird, W.E. Stewart and E.N. Lightfoot, Transport Phenomena, John Wiley, New York (1960).

Surface Contamination and Cleaning, Vol. 1, pp. 267–277 Ed. K.L. Mittal © VSP 2003

Microdenier fabrics for cleanroom wipers JOHN SKOUFIS∗ and DOUGLAS W. COOPER† ITW Texwipe, 650 E. Crescent Ave., Upper Saddle River, NJ 07458

Abstract—As the state of technology advances in data storage and integrated circuits, the need to remove smaller particles becomes more critical in order to maintain economical yields and avoid product failures due to contamination. Present textile materials are approaching the limits of their ability to achieve particle removal. New materials are being developed and investigated for providing the high level of cleaning efficiency required. These new materials fall into the general class known as microdeniers. They are being shown to have the properties required to overcome the shortcomings of traditional textiles. Keywords: Contamination control; wipers; microdenier; cleanroom wipers; particle removal; fiber construction.

1. INTRODUCTION

The critical dimensions in data storage and integrated circuit technologies continue to get smaller and smaller, putting these high-tech products at risk from submicrometer particles. As flying heights approach 25 nm, contamination of disk media and read-write heads during their manufacture becomes even more of a concern. A half-micrometer (500 nm) particle is more than ten times the readwrite gap, perhaps leading to a read-write error or a crash. Similarly, integrated circuits continue to have ever-decreasing “line width” ground rules, requiring the control of particles of ever decreasing size limits. Particles hundredths of a micrometer in size can be “killers” [1]. The relatively large fiber diameters of standard wipers make them less efficient in picking up these small particles. Microfibers have been shown to be more efficient in this application based on their geometry and the physics of particle removal [2].

∗

To whom all correspondence should be addressed. Phone: 201-327-9100 X330, Fax: 201-3275945, E-mail: [email protected] † Now retired.

268

J. Skoufis and D.W. Cooper

2. WIPERS AND SWABS

During plant and cleanroom construction, and then throughout operations, rigorous cleaning is needed, using such consumable materials as wipers, swabs, cleaning compounds, and cleaning tapes. Such materials need to be selected carefully both with respect to contaminating potential and ability to clean. Cleanliness is measured by examining materials, extracting particulate matter in agitated liquid or in air, and by extracting chemical constituents such as ionics, hydrocarbons, non-volatile residue (NVR or residue after evaporation) etc., with appropriate solvents [3]. Cleanroom wipers (and swab heads) selected for cleanrooms of different levels of cleanliness generally follow these guidelines: A. Cleanest, Fed-Std-209E Class 1 (= Class M1.5) to Class 10 (= Class M2.5) rooms: laundered sealed-edge polyester knit (lowest contaminants) or Nylon knit (somewhat higher non-volatile residues). A portion of such a wiper is shown in Figure 1. B. Class 10 to Class 100 (= Class M3.5) rooms: laundered polyester knit or hydroentangled polyester (somewhat higher particles and fibers due to unsealed edges). C. Class 100 to Class 10,000 (= Class M5.5) rooms: hydroentangled polyester/cellulose blends.

Figure 1. Laundered sealed edge wiper representing the highest level of cleanliness for critical cleanroom applications.

Microdenier fabrics for cleanroom wipers

269

D. Classes >10,000: These consist of various composites, including polypropylene and natural fibers such as cotton and polyurethane foam. The progression of wiper offerings is logical: 1. sealed-edge products are cleaner than unsealed; 2. laundered products are cleaner than unlaundered; 3. continuous filaments give less fiber contamination than do cut (“staple”) fibers; 4. natural fibers contain contaminants harder to control than those in synthetic fibers, and natural fibers are only available as staple fibers. The efficacy of a wiper includes absorbency and ability to pick up and retain particulate matter. This will depend on the fiber or foam base material and on construction details. Inevitably, some contaminants will reach the work surface where non-contaminating materials are needed to remove them. 3. MICRODENIER FABRICS

The denier of a yarn or fiber is the linear density expressed as the number of grams in 9000 m of the yarn or fiber. Microdenier fibers (filaments) are defined as 1 denier or less, where 1 denier for polyester corresponds to a circular cylinder with a diameter D of about 11 µm [4]. For fabrics of the same fiber composition of denier d: – the total length of fibers will be proportional to 1/d – –

the radius of a fiber will be proportional to d the pores created by bundles of such fibers will have dimensions proportional to d .

the total surface area of fibers will be proportional to 1/ d the cross-sectional area of a fiber will be proportional to d. The techniques for creating such fine fibers include: 1. extruding a two-component mixture, then dissolving one component 2. extruding a two-component mixture, then fracturing the fiber with highpressure water, mechanical action, or chemical stress 3. extruding a single component, then elongating and thinning the fibers with high-temperature gas jets. The first two types, known as “islands in the sea” and “pie,” respectively, are the ones normally used for microdenier and ultra-microdenier (below 0.1 denier) materials used for cleanroom applications. They are generally made into woven or knitted goods. The third type is used for “nonwovens” only, because of the difficulties associated with weaving and knitting these fibers. – –

270

J. Skoufis and D.W. Cooper

Figure 2. Various manufacturing techniques for forming microdenier fibers. Other technologies can be used but generally will fall into one of the four shown. The first is a single component while the others are multicomponent fibers.

Mochizuki et al. [5] of Unitika Ltd. of Japan have described the formation of ultra-fine fibrous materials by splitting bi-component fibers after they have been formatted as a spunbonded fabric (molten continuous fibers are laid down and adhere where the fibers cross). They noted that three methods were conventionally used (Figure 2) to form microdenier fibers from bi-component fibers: dissolution of one of the components, separation of the components by swelling or shrinking, and separation by mechanical distortion, the last being the approach they used. A “sunflower” pattern (Figure 3) of six polyester fibers surrounding a polyethylene core was split apart by flexing, taking a 3-denier fiber and making six 0.25-denier fibers and a 1.5-denier core segment. The thin fibers contributed softness and flexibility; the core allowed convenient thermal bonding. The density of the fabric decreased, giving greater porosity. Triboelectrification of the fibers from dissimilar materials can produce charging that enhances the effectiveness of dry wiping in picking up dust particles [6]. Of course, this is not recommended for contact with semiconductors. Teijin Ltd. of Japan markets its Micro-StarTM material, made of bi-component nylon/polyester fibers, that are arranged into 16 filaments that have wedge-shaped cross sections (Figure 4). Such filaments are typically 0.16 denier, but can be made smaller. The materials are useful for absorbing both oil and water. Tests done on cleaning tapes made from these materials generally showed lower levels of ionic

Microdenier fabrics for cleanroom wipers

271

Figure 3. Sunflower pattern showing six polyester fibers surrounding a polyethylene core. Splitting results in making 3 denier fiber into six 0.25 denier fibers with a 1.5 denier core. The left figure shows an oblique view and the right figure shows a cross section of the bicomponent filaments. Courtesy of Unitika Ltd.

Figure 4. Formation of microdenier fibers by splitting bicomponent polyester/nylon fiber into 16 wedge shaped filaments of 0.16 denier. Courtesy of Teijin Ltd.

contaminants than from tapes made with conventional denier fibers (1.5 denier). This is not an inherent quality but the result of a more rigorous cleaning step. Kuraray Ltd. of Japan produces its SOLIV (R)TM fibers by splitting polyester fibers longitudinally, forming fibers that have roughly rectangular cross sections, claimed to facilitate wiping through a scraping mechanism (Figure 5) [7]. Toray Industries, Inc. of Japan markets its microdenier cleaning cloth as TorayseeTM and LuminexTM. The polyester fibers are produced by splitting thicker fibers, creating material with a 2-micrometer diameter (0.06 denier) and having sharp edges rather than being round. Figure 6 shows a comparison against materials made with larger, traditional microdenier fibers (ca. 5 micrometers in diameter). The current technology allows the formation of woven and knit goods from fibers having deniers as low as 0.06. It also allows the formation of nonwoven items, typically synthetic leathers, having deniers as low as 0.0001. While these “nanofiber” materials have not yet been utilized into cleanroom wipers, the technology is nearly available to allow weaving and knitting these fibers into cleaning materials. These fine fibers will probably be much more fragile in an unbonded state and their usefulness as wiping cloths need to be determined.

272 J. Skoufis and D.W. Cooper

Figure 5. Longitudinally split fibers showing how the rectangular shapes facilitate wiping through a scraping mechanism. The upper left shows the construction of a single filament while the upper right shows the same filaments split and knit into a wiper. Courtesy of Kuraray Ltd.

Microdenier fabrics for cleanroom wipers

273

Figure 6. Ultra microdenier polyester cleaning cloth showing split wedge shaped fibers approximately 2 micrometers in diameter (0.06 denier) is shown on the left. Comparison is made with two traditional larger microfiber cloths on the right. Courtesy of Toray Ltd.

4. FABRIC STRENGTH AND RIGIDITY

The force needed to break a fiber (in tension) is generally proportional to its cross-sectional area, a, so that the tensile strength of a textile fiber is often given as gpd, grams per denier, which should be roughly independent of fiber diameter. For polyester this is about 3-10 gpd and is commonly called the fiber “tenacity”. The yarns of interest to us are in the range of 50 to 150 denier (composed of many filaments), meaning a single yarn could suspend 150 to 1500 g without breaking, depending on the tenacity of the particular polyester material [8]. The strength of an individual filament determines whether it breaks when snagged; the strength is proportional to (denier)/(# filaments) = d/n. Bending strength is usually also proportional to cross-sectional area, so the same proportionality can be expected. To prevent pilling (formation of lint balls), most polyester in commercial use is low tenacity. Higher tenacity fibers are needed for wipers required to have abrasion resistance. The geometry of a twisted yarn is complicated. The densest plausible packing would be that of a unit cell that is a hexagon of contiguous cylinders surrounding a central cylinder, which would give less than 10% open cross-sectional area. A more plausible approximation is a square unit cell, with sides 4r in length, circumscribing four cylinders, of radius r, shown in Figure 7. This gives a porosity

274

J. Skoufis and D.W. Cooper

Figure 7. Calculation of cross-sectional porosity, e, or open space, of a twisted yarn from the densest possible packing of filaments.

e = 1-(π/4) = 0.215 = 21.5%. The spaces around these cylinders are complicated, but they are roughly cylindrical pores with radii that are about half that of the filament cylinders, r/2. r/2 will be taken here to be the characteristic dimension of these inter-filament pores in the yarn. Note that this model is quite approximate, as this geometry would in fact let liquid enter only from the ends of the yarn, and not from the periphery, which clearly is not the case in practice. The packing density is a determining factor in many of the properties of the wiper: the absorbency, ability to pick up particles, and feel are all affected. Once the filament size is determined, the science of fabric construction comes into play in order to provide optimum properties. 5. LIQUID REMOVAL EFFECTIVENESS

The liquid to be wiped up usually contains particulate and molecular contaminants that will be removed roughly in proportion to how much of the liquid is removed [9]. There is reason to believe that the residual liquid (“boundary-layer”) left by wiping with a fabric made from a yarn would be roughly the size of the interfilament spacing, r/2. The filament radius, thus the inter-filament spacing, and, by

Microdenier fabrics for cleanroom wipers

275

inference, the boundary layer will be proportional to the square root of the filament cross section and therefore to (d/n) . Some boundary-layer reduction advantage is expected for microdenier fabrics based on this analysis. The height to which a non-volatile liquid can rise in a fabric by absorption is inversely proportional to the pore size. Microdenier fabrics, because of their smaller pore sizes compared to standard denier fabrics, should raise liquids higher. The speed the liquid will travel horizontally through the fabric is proportional to the pore size. Microdenier fabrics provide less speed of absorption. This may not be critical for many applications but in cases where it is, fabric construction can overcome this disadvantage. The amount of liquid a wiper can hold is roughly proportional to its thickness. For the same knit, the thickness can be expected to be roughly proportional to the square root of the yarn denier, so wipers with larger deniers should offer greater absorption capacity. However, the weight of the wiper will be the product of the denier and the length of yarn used to knit it. To keep the basis weight the same, the length of yarn will have to be inversely proportional to the denier. This would mean fewer or shorter loops. In the limit of a very large denier, one would have almost a very loose weave, rather than a knit, which will adversely affect absorption capacity. Larger denier yarns would also feel stiffer and more difficult to get into tight spaces. It is not completely clear what the implications of denier are in absorption capacity as so much is dependent on basis weight and construction. Experience has shown that the ability to hold liquid is of less interest in critical cleaning than the ability to pick up small particles and reduce boundary layers. For different fabrics with the same basis weight (g/m2), there will be the same volume of polyester contained in a total length of filaments L'. The volume of the filaments would be πr2L'. The surface area would be 2πrL'. The surface area per unit volume (thus per unit weight) would be 2/πrL', inversely proportional to the filament radius, thus the surface area would be proportional to (n/d) . The greater the surface area, the higher the liquid absorption, but cleaning the fabric may be more difficult. 6. PARTICLE REMOVAL EFFECTIVENESS

A simple geometrical model can be used to suggest the importance of smaller wiping element size in lifting particles from a surface being wiped. Figure 8 shows two circles. One of the circles represents a particle. The other represents a fiber. The larger object (particle or fiber) has a radius = R. The smaller object (fiber or particle) has a radius = r. Arguably, to remove the particle away from the surface we need a vertical component of force. [One could also cause the particle to roll, not considered here.]

276

J. Skoufis and D.W. Cooper

Figure 8. If one circle represents a particle and the other a fiber, the force, Fv, required to lift the particle can be calculated. The smaller the fiber, the smaller the particle it can push up from the surface.

The contact (idealized) between the cylindrical fiber and the spherical particle occurs at their surfaces, along a line from one center to the other. The vertical component of the force is then Fv = F sinθ, where F is the force at the contact and θ is the angle with respect to the surface or Fv = F (r - R) / (r + R), with the smaller object being pushed toward the surface and the larger object being pushed away. The smaller the denier, the smaller the particle it can push up from the surface. Besides the size advantage microdeniers offer in picking up very small particles, microdeniers can be made in a variety of shapes, even within the same unsplit bicomponent filament. Common shapes include: shovel, wedge, star, multi-lobe, and interior hollows. The wide variety of shapes provides opportunities for investigation into their suitability for particle removal. Specific shapes may be optimal only for specific particle types.

Microdenier fabrics for cleanroom wipers

277

7. FUTURE ACTIVITIES

While microdenier materials have been in wide use throughout Asia in cleanrooms, they have been commercialized slowly in the United States. One reason is the lack of US producers; another is their higher cost compared to standard denier products. Interest in microdenier materials is gradually increasing in the U.S. More sophisticated microelectronic products demand better contamination control. The price of microdeniers is decreasing due to competition as more microdenier material manufacturers are entering the market. Several manufacturers who intend to revolutionize fiber production and significantly reduce manufacturing costs have designed new equipment in the U.S. Several companies and universities have purchased pilot-scale and production equipment. Initial trials look promising. The equipment is able to change polymer types quickly, and configurations not previously possible are now feasible. These machines can produce nanofibers of various shapes and various polymers by simple economical screen changes, rather than the time-consuming changes in expensive spinnerets previously needed. 8. CONCLUSIONS

The new microdenier fibers being used for textiles should be able to produce highly absorbent and very clean and effective wiping materials for cleanroom use. The technical developments of the industry are driving the need for fibers and fabrics which are more effective in removing ever smaller particles. Asia has led in development and use of microdenier fibers, but in the U.S. manufacturers are catching up, using economical and flexible designs and methods. REFERENCES 1. D.W. Cooper, Microcontamination, 3(8), 48–54, 73 (1985). 2. Kuraray Ltd. Publication 1241–58, “Wiping Cloth for High Class Cleanrooms” (1998). 3. “Evaluating Wiping Materials Used in Cleanrooms and Other Environments,” IES Publication RP–CE–004–2 (1992). 4. J. Skoufis, “Fabrics for Disk Media,” Internal Training Document PMG–1A (1998). Available for the author. 5. M. Mochizuki, K. Nagaoka and M. Hirai, “A Sunflower Comes into Blossom,” The Technical Progress, Unitika Publication, undated. 6. D.W. Cooper, A2C2 (October 1998). 7. Teijin Ltd. Publication 93.6.2000, “Microstar Wiping Cloth.” 8. B.P. Saville, Physical Testing of Textiles, Woodhead Publishing, Abington, Cambridge, England (1999). 9. R. Wang, Microcontamination, 14(2), 39–47 (1996).

This page intentionally left blank

Surface Contamination and Cleaning, Vol. 1, pp. 279–291 Ed. K.L. Mittal © VSP 2003

Fine particle detachment studied by reflectometry and atomic force microscopy ADAM FEILER1,∗ and JOHN RALSTON2 1

KTH, Royal Institute of Technology, Department of Chemistry, Surface Chemistry, Drottning Kristinas Väg 51, SE-100 44 Stockholm, Sweden 2 Ian Wark Research Institute, The ARC Special Research Centre for Particle and Material Interfaces, University of South Australia, Mawson Lakes, Adelaide, SA 5095, Australia

Abstract—Optical reflectometry was used to study the attachment and subsequent detachment of silica particles (diameter 25 nm) from the surfaces of titanium dioxide wafers under well-defined hydrodynamic conditions. The rate of detachment and maximum detached amount was studied as a function of both pH and added linear polyphosphate solutions. The latter have the general formula [PnO3n+1](n+2)- where n is the number of phosphorous atoms in the molecule. The maximum detached amount increased with increasing pH. The maximum detached amount also increased with n. Atomic force microscopy was used to measure the interaction between silica spheres (diameter 7 µm) and titanium dioxide wafers under the same solution conditions. The detachment force needed to separate the surfaces decreased with increasing pH as well as with n in direct agreement with the reflectometry data. It was shown that, in addition to repulsive electrical double layer forces, adsorbed polyphosphates provided a short-ranged steric layer that reduced the lateral interaction between the surfaces. The use of these two complementary techniques has given valuable insight into the processes responsible for fine particle detachment and has particular application to surface cleaning. Keywords: Fine particle detachment; reflectometry; atomic force microscopy; particle adhesion.

1. INTRODUCTION

Surface contamination due to submicrometer particulate matter is of concern in many areas including silicon wafer fabrication, mineral processing, water purification and detergency. The permanent removal of these particles from surfaces is a critical factor in these processes [1]. In solution, the combined effects of van der Waals attractive forces and electrical double layer forces govern the interaction between particles and a surface. By varying the solution conditions it is possible to alter the surface chemistry of interacting materials and change their interaction from attractive to repulsive. In this work, the attachment and detachment of ∗

To whom all correspondence should be addressed. Phone: +46 8 790 9971, Fax: +46 8 20 89 98, E-mail: [email protected]

280

A. Feiler and J. Ralston

nanosized silica particles onto titanium dioxide surfaces has been studied as a function of pH and addition of solutions of linear polyphosphates. Previous in-situ infra-red studies [2] have shown that linear polyphosphates selectively adsorb onto titanium dioxide forming strong chemical bonds. Streaming potential measurements [3] have shown that for a fixed polyphosphate concentration, the titanium dioxide becomes more negative with increasing n, which was attributed to an increased charge density with n. In addition, direct force measurements [3-5] have shown that adsorbed polyphosphate introduces a steric layer which leads to a short-ranged repulsive interaction. Reflectometry combined with a stagnation point flow cell was used to measure silica particle attachment and detachment. The stagnation point flow is ideally suited for the study of colloidal particle attachment and detachment processes in which their transport is governed only by diffusion [6]. 2. EXPERIMENTAL

2.1. Materials 2.1.1. Titanium dioxide wafers A titanium dioxide layer was deposited on the surface of silicon wafers by sputtering (prepared at Philips Research, The Netherlands) [3]. Ellipsometry measurements showed the titanium dioxide layer to be 40 nm thick and XPS analysis showed the composition of the deposited layer to be pure titanium dioxide. X-ray diffraction showed the deposited TiO2 to be amorphous. Imaging by AFM showed an rms roughness of 0.3 nm over an area of 1 µm2 with a maximum peak height of 2 nm. The wafers were cleaned by detergent washing followed by rinsing with high-purity water, ethanol, heptane and copious amounts of more high-purity water. Finally, the wafers were blown dry in a stream of nitrogen and plasma cleaned (Harrick Plasma Cleaner/Steriliser PDC-32) for 1 minute immediately prior to use. The isoelectric point of the titanium dioxide covered wafers was determined to be pH 4.2 using streaming potential measurements [3]. 2.1.2. Silica Suspensions of silica particles used for reflectometry experiments were prepared from LudoxTM AS40 (DuPont). The particles were dialyzed for 2 days in Milli Q water and then suspended in solutions of KNO3 (10-3 M) with a particle concentration of 100 mg/l. The mean particle radius was R = 12 ± 2 nm, determined by transmission electron microscopy (CESMA, Adelaide University). The silica spheres used for AFM measurements were obtained from Allied-Signal, (Chicago, Illinois). XPS analysis (CSIRO Division of Molecular Science) showed the composition of the sample to be pure silica. The typical diameter was found to be 7 µm. AFM imaging of the spheres over an area of 500 nm2 showed the rms roughness to be 0.8 nm.

Fine particle detachment studied by reflectometry and atomic force microscopy

281

2.1.3. Linear polyphosphates The linear polyphosphates were provided by Albright and Wilson (Australia) as dry sodium salts and were made up into solutions with a polyphosphate concentration of 10-5 M with background electrolyte concentration of KNO3 (10-3 M). The linear polyphosphates referred to here as P1, P2, P3 denote the monodisperse species (Na3PO4), (Na4P2O7) and (Na5P3O10) respectively, P<10> refers to a polydisperse sample with n ranging from 1-19 with an average of 10 P atoms. 2.1.4. Other reagents Analytical grade KNO3, HNO3, and KOH were obtained from BDH Chemicals (Australia) and were used as supplied. High purity water (surface tension 72.8 mN/m and resistivity 18 MΩ at 20°C) was from obtained from an Elga UHQ system. The solution pH was adjusted with drop wise addition of 0.01 M HNO3, or KOH via a micropipette. 2.2. Methods 2.2.1. Reflectometry The experimental setup for the combined reflectometry and stagnation point fluid cell is shown schematically in Figure 1. A detailed description of the experimental procedure and the theory behind the technique is given elsewhere [7, 8]. Here a

Figure 1. Schematic diagram of the combined reflectometry and stagnation point flow cell.

282

A. Feiler and J. Ralston

brief description of the essential features is given. The reflected light from a plane polarized He/Ne laser off an adsorbing surface is measured. The adsorbing surface in this work was titanium dioxide covered silicon wafers. The reflected light is split into its parallel (p) and perpendicular (s) components and the measured signal, S, is the ratio of these intensities:

S=

IP IS

(1)

This ratio depends on the refractive index profile close to the surface of the substrate. Material adsorbed at the interface, in this case silica nanoparticles, will change the refractive index profile and hence result in a change in S. Quantitative measurements of the attached particle amount (Γ) can be obtained from the change in signal via:

Γ=

1 ∆S As S 0

(2)

where S0 is the intensity ratio prior to adsorption and ∆S is the change in intensity ratio. The sensitivity factor, As, takes into account the explicit refractive index contributions from the surface, the adsorbed material and the aqueous medium. 2.2.1.1. Stagnant point flow In the stagnant point fluid cell, the collector surface is positioned at a critical distance from the inlet tube such that a stagnant point flow is generated at the point where the fluid impinges the surface. Under these conditions the hydrodynamics can be very well defined [9]. The particles arrive at the surface under the influence of surface forces and Brownian diffusion only. Solutions were gravity fed into the reflectometry cell from high-density polyethylene (HDPE) containers (250 ml) mounted on adjustable laboratory jacks. The height of the liquid above the inlet port to the fluid cell determined the flow rate. The flow rate was maintained at 1.5 cm3/min with a height of the liquid 13 cm above the inlet port. A valve was used to switch between solutions entering the cell. The cell volume was 30 cm3. A vacuum pump was used to suck excess solution from the cell. 2.2.2. AFM A sphere attached to the end of an AFM cantilever comprises a colloid probe. A silica sphere was attached to a cantilever using a heat softening resin (Epikote 1004, Shell) using a micromanipulation arm attached to a metallurgical microscope (Olympus BH2). The cantilevers were silicon nitride, tipless, 200 µm long, wide legged from Nanoprobe (Park Scientific, USA). Spring constants were determined to be 0.1 ± 0.05 N/m by measuring the resonance frequency of the cantilevers with added known masses [10]. Prior to measurement the colloid probe was rinsed with ethanol, dried in a stream of nitrogen and plasma cleaned (Harrick Plasma Cleaner/Steriliser PDC-32) for 1 minute. A Nanoscope III controller and

Fine particle detachment studied by reflectometry and atomic force microscopy

283

Atomic Force Microscope (Digital Instruments, Santa Barbara, CA, USA) equipped with a fluid cell was used to measure the forces of interaction. Electrolyte solutions at the required pH and containing the desired polyphosphate solutions were introduced into the fluid cell via Teflon tubing. The solution was allowed to equilibrate for at least 15 minutes prior to measurement. The experiments were conducted employing standard measurement procedures comprehensively described by other authors [11-13]. Measurements of the cantilever deflection against scanner (piezoelement) displacement were taken. The piezoelement was calibrated via an optical interference technique [14]. The cantilever deflection data were subsequently converted to force (F) as a function of apparent surface-surface separation (h), simply called separation hereafter. The force of interaction was normalised by the radius of the sphere, i.e. F/R, employing the Derjaguin approximation for sphere-flat interactions [15]. 3. RESULTS AND DISCUSSION

An example of typical reflectometry data is shown in Figure 2. Electrolyte solution at pH 4 was flowed into the reflectometry cell for 20 minutes before the start of the measurement. This ensured that the solution in the cell had reached thermal

Figure 2. Typical reflectometry raw data showing change in signal with time upon introduction of a particle suspension at arrow (a) followed by the introduction of a displacing solution at arrow (b).

284

A. Feiler and J. Ralston

equilibrium with the solution in the reservoirs. The baseline signal, S0 was monitored during this time to ensure that the signal was stable prior to measurement. The background electrolyte solution was permitted to flow into the cell for a further 100 seconds after the measurement began. After this time (point (a) in Figure 2), the valve was switched to allow silica particles into the cell (the electrolyte concentration remained unchanged). The reflectometer signal increases due to the attachment of the silica particles. Initially the signal increases linearly with time. The rate of attachment decreases markedly close to saturation coverage. At point (b), the valve was switched to introduce a new solution, containing either a particle-free electrolyte solution at high pH or a solution of linear polyphosphate. The decrease in the signal at point (b) is due to the detachment of the silica particles. Initially the decrease in signal is very rapid but the rate of detachment slows down as the maximum detached amount is reached. The signal reaches a “plateau detached amount” before complete detachment of the particles has been obtained. The attachment of silica particles to a titanium dioxide wafer at pH 4, followed by their detachment upon introduction of electrolyte solutions at higher pH’s, is shown in Figure 3. The rate of attachment and the maximum attached amount, Γmax, was the same for each experiment. The linear attachment regime is indicative of a rate limited only by the mass transport of the particles from solution to

Figure 3. Amount of silica particles attached at the titanium dioxide surface as a function of time after the introduction of a particle suspension in 10-3 M KNO3 at pH 4 followed (at t = 1000 s) by introduction of particle-free solutions of 10-3 M KNO3 at higher pH’s.

Fine particle detachment studied by reflectometry and atomic force microscopy

285

the surface [16, 17]. A similar observation was made in other reflectometry studies of nanometre-size particles [8, 18, 19]. No detachment was measured when electrolyte solutions below pH 6 were introduced into the cell. Upon switching to electrolyte solutions at pH 6 and higher, detachment of silica particles was detected. Both the rate of detachment and the maximum detached amount increased with increasing pH. At pH 6, only a small quantity (10%) of silica particles were detached and the detachment rate was slow compared to initial rate of attachment at pH 4. At pH 9, half of the pre-attached particles were detached and the initial rate of detachment was faster than the initial rate of attachment. In Figure 4 the detachment force measured by AFM between a silica sphere and titanium dioxide substrate as a function of pH under the same solution conditions as in Figure 3 is presented. A pH dependent adhesion force is evident. Previous studies have shown [3, 20] that the interaction force between silica and titanium dioxide surfaces is well described by the DLVO theory of colloidal stability. The adhesion results presented here may be rationalised in terms of the combined effects of an attractive van der Waals force and pH-dependent electrical double layer interactions. The isoelectric point (iep) of the silica particles and the titanium dioxide wafers have been measured to be at pH ~ 2 and pH ~ 4.5 respectively [3, 20]. At pH values below the iep of the titanium dioxide, in addition to

Figure 4. Force-distance curves for the retraction of a titanium dioxide wafer from a silica sphere (R = 3.5 µm) as a function of pH in 10-3 M KNO3. The curves correspond to, from top to bottom, data at pH 9, 8, 7, 6, and 5.6.

286

A. Feiler and J. Ralston

Figure 5. Amount of silica particles attached at the titanium dioxide surface as a function of time after the introduction of a particle suspension in 10-3 M KNO3 at pH 4 followed (at t = 1000 s) by introduction of solutions of linear polyphosphates (10-5 M) at pH 4 in 10-3 M KNO3. The subscript in Pn refers to the number of phosphorous atoms in the molecule.

the attractive van der Waals force, there will be an attractive electrical double layer interaction between the oppositely charged surfaces. At pH values above the iep of titanium dioxide both surfaces will be negatively charged and the electrical double layer interaction will be repulsive. A tensile force is needed to separate the surfaces from intimate contact at pH 5.6. This is indicated by the negative value of the normalised force at the point at which the surfaces jump out of contact. At this pH, the electrical double layer interaction will be weakly repulsive and the adhesion is due to the attractive van der Waals forces. At higher pH values the electrical double layer interactions become increasingly repulsive and the surfaces are seen to separate from contact even in the presence of a positive applied force. The separation force curves correlate well with the detachment data seen in Figure 3 and explain why no detachment of silica particles was detected below pH 6 and also why the detached amount increased with pH. The attachment of silica particles at pH 4 onto titanium dioxide substrates followed by their detachment upon switching to solutions of linear polyphosphates (10-5 M) is presented in Figure 5. As discussed in the Introduction, adsorption of polyphosphate onto titanium dioxide modifies the surface rendering it negatively charged. Note that whereas changes in pH affected both the silica and titanium

Fine particle detachment studied by reflectometry and atomic force microscopy

287

Figure 6. Force-distance curves for the retraction of a titanium dioxide wafer from a silica sphere (R = 3.5 µm) at pH 4 in 10-3 M KNO3 in the presence of linear polyphosphate solutions (10-5 M) from top downwards P<10>, P3, P2 and P1.

dioxide surface potentials, the specific adsorption of polyphosphate onto titanium dioxide leaves the silica unmodified. Thus, any detachment of silica particles can be attributed solely to chemical changes at the titanium dioxide surface. For n = 2, 3 and <10> the rate of detachment and the detached amounts of particles are similar to each other and larger than P1 on the time scale of the experiments. The discrepancy seen in the detachment profile in the presence of P1 can be understood in terms of the adhesion measurements, see below. Figure 6 shows the interaction force curves during separation between a silica colloid probe and titanium dioxide at pH 4 in the presence of solutions of polyphosphate (10-5 M) of varying n. For comparison, the interaction force between silica and titanium dioxide at pH 4 in the absence of polyphosphate is also shown. At pH 4 in the absence of polyphosphate there is a large adhesion due to combined attractive electrical double layer and van der Waals forces. The presence of polyphosphate clearly modifies the interaction force, dramatically reducing the adhesion. In the presence of P1 the detachment force is negative indicating a significant adhesion force. For n > 1 the detachment force is positive and the force curves show that the separation is dominated by a repulsive interaction. The magnitude of the repulsive interaction increases with n. The trends seen in the interaction force curves are in accord with previous AFM studies and streaming potential

288

A. Feiler and J. Ralston

Figure 7. Amount of detached particles measured by reflectometry against the normalised detachment force measured by AFM under similar solution conditions. The symbols refer to the data measured as a function of pH (♦) and in the presence of polyphosphate (•).

measurements [3] and can be directly related to chemical modification of the titanium dioxide surface due to adsorbed polyphosphate. The increase in the repulsive interaction with n is due to an increased negative surface potential on the titanium dioxide and also the presence of a steric layer, whose thickness, δ, increases with n (δ ≈ 0.4-0.6 nm for n = 1-3 and δ ≈ 1.6 nm for n = <10>) [3]. A good correlation between the detached amounts of silica particles measured by reflectometry and the normalised detachment force measured by AFM is seen in Figure 7. The detached amount of particles is plotted as a percentage of the total attached amount of particles prior to switching to the displacing solution. The detached amount of particles increases as the normalised detachment force becomes more positive (more repulsive). The detachment of the silica particles is clearly sensitive to the variation in the electrical double layer interactions brought about by changes in pH as well as due to adsorbed polyphosphate. The fact that detachment of silica particles is detected at all in the presence of P1 despite the force curves showing an adhesional interaction is evidence that the steric layer due to adsorbed polyphosphate is important in the detachment process. The slight discrepancy from a linear trend between the detached amount and detachment force seen for the P<10> data point can be understood in terms of the polydisperse

Fine particle detachment studied by reflectometry and atomic force microscopy

289

Figure 8. The effect of the sequence of addition of silica particles and P3 solution. Curve I is for the introduction of silica particles followed by the introduction of P3 (10-5 M) at t = 1500 s. Curve II is for the introduction of silica particles after pretreatment by flowing P3 solution into the cell for 1000 seconds and rinsing with electrolyte solution for 100 seconds. Curve III is for the introduction of a suspension containing a mixture of silica particles and P3 (10-5 M). All experiments were conducted in 10-3 M KNO3 at pH 4.

nature of the P<10> sample which contains a range of linear polyphosphates from n = 1-19. In the reflectometry measurements, the diffusion rate of the polyphosphate species becomes a critical factor. The diffusion rate of the polyphosphate species will decrease with n. Thus, although the larger n species will impart a more repulsive force on the silica-titanium dioxide system, the smaller n species will diffuse to the surface more quickly. The consequence is a competition between the rate of polyphosphate adsorption and the modification of the resultant polyphosphate adsorption. Finally, it is of interest from an application viewpoint as to the most efficient use of a dispersing agent such as polyphosphate in preventing particle attachment. In Figure 8 the order of addition of silica particles and P3 solution (10-5 M) to the cell is investigated. Curve I shows the attachment of particles in the absence of polyphosphate followed by their detachment upon introduction of a P3 solution (10-5 M). This sequence of addition is identical to that shown in previous figures. Curve II shows the attachment of silica particles on a titanium dioxide wafer that

290

A. Feiler and J. Ralston

was conditioned with polyphosphate. For this experiment, a solution of P3 (10-5 M) was introduced into the cell for 1000 seconds followed by rinsing with electrolyte solution for 100 seconds immediately prior to introducing the silica particles. Particle attachment was detected; however, the initial rate of attachment was less than that seen in Curve I onto the bare surface. The maximum attached amount is also reduced. The most striking effect is seen in Curve III, in which silica particles and P3 solution were introduced simultaneously into the cell. No particle attachment was detected, even after 50 minutes. The results indicate that equilibrium considerations are relevant in determining the amount of attached particles and adsorbed polyphosphate that takes place at the surface. Preconditioning the titanium dioxide surface (Curve II) reduces particle attachment by reducing the available surface sites at which the particles can attach. In Curve III the bulk solution contains both excess silica particles and polyphosphate. As a much smaller species, the polyphosphate possesses a much higher diffusion coefficient than the particles. Thus it is expected that the polyphosphate would arrive and adsorb at the titanium dioxide surface before the particles. The excess concentration of polyphosphate ensures that even if polyphosphate desorbs from a surface site, another polyphosphate molecule will quickly adsorb in its place and so prevent the attachment of particles over long periods of time. 4. CONCLUSIONS

The attachment of nanosized silica particles onto titanium dioxide surfaces and their subsequent detachment due to changes in solution pH or in the presence of linear polyphosphates was studied using reflectometry. The rate of detachment and maximum detached amount increased with both pH and n. It was seen that increasing the pH led to repulsive electrical double layer interactions, which were responsible for the detachment of the particles. In addition to electrical double layer interactions, adsorption of linear polyphosphate onto titanium dioxide provided a steric component, which facilitated the particle detachment. A good correlation was seen between the amount of detached particles and the AFM measured detachment force. Furthermore it was shown that a solution containing excess linear polyphosphate could prevent particle attachment all together. REFERENCES 1. K. L. Mittal (Ed.), Particles on Surfaces 5 & 6: Detection, Adhesion and Removal, VSP, Utrecht (1999). 2. A. P. Michelmore, W. Gong, P. Jenkins and J. Ralston, Phys. Chem. Chem. Phys., 2, 2985 (2000). 3. A. Feiler, P. Jenkins and J. Ralston, Phys. Chem. Chem. Phys., 2, 5678 (2000). 4. A. Feiler, I. Larson, P. Jenkins and P. Attard, Langmuir, 16, 10269 (2000). 5. Y. K. Leong, P. J. Scales, T. W. Healy, D. Boger and R. J. Buscall, Chem. Soc. Faraday Trans., 89, 2473 (1993).

Fine particle detachment studied by reflectometry and atomic force microscopy

291

6. Z. Adamczyk, B. Siwek, M. Zembala and P. Belouschek, Adv. Colloid Interface Sci., 48, 151 (1994). 7. J. C. Dijt, PhD thesis, Wageningen University, The Netherlands (1993). 8. M. R. Bohmer, J. Colloid Interface Sci., 197, 251 (1998). 9. T. Dabros and T. G. M. van de Ven, Colloid Polym. Sci., 261, 694 (1983). 10. J. P. Cleveland, S. Manne, D. Bocek and P. K. Hansma, Rev. Sci. Instrum., 64, 403 (1993). 11. W. A. Ducker, T. J. Senden and R. M. Pashley, Nature, 353, 239 (1991). 12. I. Larson, C. J. Drummond, D. Y. C. Chan and F. Grieser, J. Am. Chem. Soc., 115, 11885 (1993). 13. P. G. Hartley, I. Larson and P. J. Scales, Langmuir, 13, 2207 (1997). 14. M. Jaschke and H.-J. Butt, Rev. Sci. Instrum., 66, 1258 (1995). 15. J. N. Israelachvili, Intermolecular & Surface Forces, 2nd ed., Academic Press, London (1992). 16. Z. Adamczyk, L. Szyk and P. Warszynski, J. Colloid Interface Sci., 209, 350 (1999). 17. N. Kallay, M. Tomic, B. Biskup, I. Kunjasic and E. Matijevic, Colloids Surfaces, 28, 185 (1987). 18. M. R. Bohmer, E. A. van der Zeeuw and G. J. M. Koper, J. Colloid Interface Sci., 197, 242 (1998). 19. R. A. Hayes, M. R. Bohmer and L. G. Fokkink, J. Langmuir, 15, 2865 (1999). 20. I. Larson, C. J. Drummond, D. Y. C. Chan and F. Grieser, J. Phys. Chem., 99, 2114 (1995).

This page intentionally left blank

Surface Contamination and Cleaning, Vol. 1, pp. 293–310 Ed. K.L. Mittal © VSP 2003

Dust removal from solar panels and spacecraft on Mars S. TRIGWELL, M.K. MAZUMDER,∗ A.S. BIRIS, S. ANDERSON and C.U. YURTERI Department of Applied Science, Donaghey College of Information Science and Systems Engineering, University of Arkansas at Little Rock, 2801 South University Avenue, Little Rock, AR 72204-1099

Abstract—In Lunar or Martian habitat systems it is impossible to avoid contact with dust. Martian dust storms, containing submicrometer to 50 µm particles, are an environmental threat to solar cells, spacecraft, and spacesuits. Because of the high electrostatic charge of the dust and its strong adhesion properties, its deposition onto life support equipment could damage or degrade equipment, reducing the mission duration and endangering personnel. The inhalation of electrostatically charged airborne dust is also a health hazard to astronauts inside the habitat. Ways to minimize or eliminate the potential hazards caused by charged particles on space life support equipment are therefore needed. Specifically, the following topics are discussed in this paper: (1) tribocharging of insulating materials, (2) the design of a sensor to measure particle size and electrostatic charge distributions of Mars dust on a single particle basis and in real-time, (3) an experimental plan to minimize deposition of charged particles on solar cells and life support equipment, and (4) a novel method for removing deposited dust particles. Keywords: Mars dust; solar panels; electrostatic; charged particles.

1. INTRODUCTION

The atmosphere of Mars contains significant amounts of suspended dust, and in any mission to Mars it will be impossible to avoid contact with this dust. Martian dust storms containing fine particles (submicrometer to 50 µm in diameter) are a serious problem to solar cells, spacecraft, and spacesuits [1, 2]. The dust may also possess a high electrostatic charge due to tribocharging by contact with other particles or materials, or photoionization by the intense UV radiation. Because of the possibility of high charge on dust particles and resulting strong adhesion forces, deposition of dust onto support equipment could damage or hinder correct functionality of the equipment, reducing the mission lifetime.

∗

To whom all correspondence should be addressed. Phone: 501-569-8007, Fax: 501-569-8020, E-mail: [email protected]

294

S. Trigwell et al.

The settling of this dust, especially during a Martian dust storm, can have a significant effect on the efficiency of solar panels, due to the settled dust impeding the sunlight from the cells. Results from the Materials Adherence Experiment (MAE) on the Mars Pathfinder mission measured an obscuration of the solar arrays due to dust deposition at a rate of about 0.28% per day [3] with an estimate that settling dust may cause degradation in performance of a solar panel of between 22% and 89% over the course of two years [4]. Particles may also settle on the solar arrays by a process known as saltation, in which particles are lifted from the surface by the wind. These have a size range of 1–200 µm in diameter and an average trajectory of 10 to 20 cm off the surface [5]. Due to the low barometric pressure in the atmosphere, of about 10 mbar, saltation occurs at wind velocities greater than 15 m/s, which has been recorded at Viking lander sites [5]. Dust accumulation can also occur due to settling from the atmosphere. However, the real deposition rate will also depend on the geographical location and from season to season. The removal of dust settled on an array by natural wind forces on Mars has been ruled unlikely due to the low atmospheric pressure, which will necessitate high wind velocities of the order of 35 m/s [6]. The measurements of wind velocities at Viking sites showed that maximum peak wind velocity was only 25 m/s, with winds over 15 m/s occurring only 1% of the time [6]. Therefore, it was concluded that for longduration missions, prevention of deposition or periodic removal of accumulated dust must be performed to maintain the efficiency of the solar power arrays. The problem with the design of any mechanism that has to work on a Mars spacecraft is the hostile environment in which it is expected to perform. The atmosphere of Mars is quite different from that of Earth in that it is composed primarily of carbon dioxide (95.3%) with minor amounts of other gases (nitrogen – 2.7%, argon – 1.6%, oxygen – 0.13%, and trace amounts of water and neon) [2]. Although the water content of the atmosphere is about 1/1000 that of Earth, it can condense out forming clouds and even ground frost in the winter. The most significant factor is the temperature on the surface. The average recorded temperature on Mars is –63°C, with a maximum and minimum of approximately 20°C and –140°C, respectively. However, the temperature variation depends on the location. Temperatures of –133°C are observed at the winter poles, while temperatures as high as 27°C are observed on the dayside during the summer [2]. In this study, the goal was to develop an understanding of the principles of particle charging and to perform theoretical and experimental studies on the adhesion and removal of charged particles. Specifically, the following are presented: (1) a study of the effects of tribocharging of insulating materials and how it can play a role in dust accumulation on solar panels, (2) design of a sensor to measure particle size and charge distributions of Mars dust, (3) development a self-cleaning panel with electrodynamic screens to repel charged dust from settling on solar panels, and (4) development of an electrostatic wiper type brush for removing deposited particles utilizing minimum mechanical parts.

Dust removal from solar panels and spacecraft on Mars

295

2. TRIBOCHARGING OF PARTICLES

There exist two primary mechanisms of charge transfer for contact or tribocharging between two dissimilar materials. The first is electron transfer in that a linear relationship is observed between charge transferred during contact of two dissimilar materials. The second is ion transfer in that real surfaces of metals or insulators are covered by adsorbed layers, which are frequently ionic in nature, and that charge transfer is by positive or negative ion transfer between the materials. A third theory is postulated that involves material transfer that carries an associated charge. 2.1. Electron transfer In a metal at absolute zero, all the states below the Fermi level in the metal are filled, and all those above are empty. When two dissimilar metals with different work functions, A and B, are brought into contact, then electrons will flow from metal A into metal B decreasing the potential difference until equilibrium of the Fermi levels is reached. Metal B will now have a net negative charge, and metal A will have a net positive charge of equal magnitude, where the contact potential difference, Vc, is given by: VC = (φB - φA) /e

(1)

where φA and φB are the work functions of metals A and B, respectively. However, in metal/metal tribocharging, a back tunneling current exists when the two materials are separated, resulting in a net zero charge on the two metal surfaces. In a metal-insulator contact electrification, which is likely to be found on Mars as the dust comes into contact with spacecraft parts and instrumentation, electrons may pass from the metal into the empty states in the insulator, or from occupied insulator states into the metal. Insulators, specifically polymers, have been considered to have a wide forbidden band gap where very few extrinsic states exist. However, there are likely to be localized surface states, surface impurity states, bulk defect states, or bulk impurity states [7-9]. These states may emit or accept electrons in contact electrification. Bulk defect levels and surface states give rise to an “effective” work function for an insulator φI. Before contact, the surface states are filled to the equivalent Fermi level, EFP. A simple surface states theory of contact electrification of insulators is shown in Figure 1. Surface states on an insulator can be intrinsic or extrinsic. In either case, contact with a metal will cause empty states below the metal Fermi Level EF to be filled, and full states above it will be emptied. The number of electrons that transfer to the insulator will be equal to the number of surface states with energies between φI and EF. However, there is still a considerable uncertainty in this description of insulator charging [7-9], but most theories assume that the amount of back tunneling of charge when the materials are separated is negligible, and that the final charge upon separation is approximately the same as when the surfaces were in contact.

296

S. Trigwell et al.

Figure 1. (a) The insulator is uncharged with states filled below the neutral level. (b) On contact, empty states below the Fermi level are filled and the insulator charge is now proportional to φ - φI.

In covalently bonded solids, impurity atoms provide most of the additional available energy levels [1], but in molecular solids, electron traps may be associated with the ends of molecular chains or cross-links. Duke and Fabish proposed a model [10] to interpret contact electrification of pendant group polymers. This model suggests that side groups on a polymer chain can form intrinsic charge carrying sites, which may be electron donors or acceptors. The model states that the electronic states are localized and represented by double Gaussian distributions representing electron acceptor and donor states. The distribution of the states is suggested to be due to differences in the local environment for each molecule. A number of factors are involved in contact electrification under different conditions. When contacting a metal surface, it has to be considered that a metal oxide layer is always present. Similarly, the insulator surface may also be covered with an oxide layer or at least other contaminants. For this, the contact charge exchange density, σ, on the insulator is given by [11, 12]; σ = -feNS[φI - φ][1 + (fe2aNS/ε)]

(2)

where f is the fraction of area that makes intimate contact, e is the electronic charge, NS is the surface state density per unit area per unit energy (eV), φI is the insulator surface work function, φ is the metal surface work function, a is the thickness of the oxide layer, and ε is the permittivity of the oxide layer. In the case of a low surface state density, NS << ε/fe2a, σ = -feNS[φI - φ] and in the case of a high surface state density, NS >> ε/fe2a,

(3)

Dust removal from solar panels and spacecraft on Mars

σ = -fεNS[φI - φ]/ea

297

(4)

It is the concentration of the surface states, NS that determines whether the electrons occupy bulk states in addition to surface states during the time of contact [11]. If the charge transfer is completed in a very short time, then only surface states are involved. If charging is notably dependent upon time, then charge transfer into the bulk is more probable. However, the above equations (3) and (4), show that contact charging depends upon both surface oxidation and density of surface states. The physical meaning of the surface work function of an insulator, φI, and the distribution of the surface states within the forbidden gap are still not clearly established. This uncertainty in describing insulator charging as it applies to contact with metals, therefore, leads to uncertainty in the understanding of insulator contact charging with other insulators. The accepted theory that even insulators higher up in the triboelectric series will charge positive when contacted with insulators lower down has led to several published triboelectric series [7, 13, 14]. However, no two series agree absolutely, with positions on the list of some materials varying widely between different series, and only a rough agreement as to the relative positions of several polymers. There is some uniformity for certain selected polymers such as Nylon and poly(tetrafluoroethylene) (PTFE), which are consistently found on the opposing ends of the series. Similarly, polyethylene and polystyrene are usually found in consistent positions among the different series. This uniformity between different triboelectric series suggests that for insulator/insulator charging, a similar mechanism as for metal/insulator exists. Therefore, an insulator/insulator contact charging theory may be constructed. Charge exchange between insulators can be predicted from the knowledge of the charge acquired by contact with metals, and so the general conclusion is that insulator-insulator charging is caused by the same basic mechanism as metal-insulator charging. 2.2. Ion transfer Real surfaces are always covered with an adsorbed layer. This layer is frequently ionic in nature or contains a charged double layer. This covering layer can act as a significant potential barrier through which the electrons must travel. However, ion exchange between two covered contact layers can take place. In this case, the possible mechanisms for ion transfer include the difference in the affinities of the two contacting surfaces for specific ions and the abundance of a particular ion on one surface. In addition to the above, material transfer may also be considered. Fragments of one material may break off one surface and be deposited on the other. The break point is a few molecules beneath the surface, and mass transfer has been detected between combinations of certain polymers [1]. However, in this case, the amount of material transferred exceeded that considered necessary for a typical measured charge transfer. At this point it is emphasized that all theories of charge transfer for both metal-insulator and insulator-insulator contacts are still poorly understood, and a much better understanding of the nature of the surfaces

298

S. Trigwell et al.

of polymers and insulators regarding the electron energy levels and the role of impurities is needed. 2.3. Experimental In order to better understand the factors involved in how the surface properties of materials influence the charge that can be transferred to a material, the surface work function and the surface chemical composition of various metallic, ceramic, polymeric, and mineral materials were analyzed. The metal and polymer materials used were typical of those that are or may be used in a Mars mission. Pulverized quartz (SiO2), alumina (Al2O3), and pyrite (FeS2) were used as examples of minerals as spectroscopic analysis of Mars dust had shown it to be composed of silicates and iron and magnesium rich sulfates [15]. The samples were analyzed in the as-received condition with no prior cleaning except for the polymeric materials that were scraped with a clean scalpel blade to expose a fresh surface. The samples were then analyzed by X-ray photoelectron spectroscopy (XPS) to determine the surface chemistry, and by ultra-violet photoelectron spectroscopy (UPS) in air to measure the surface work function. The XPS data were obtained on a PHI Quantum 2000 ESCA Spectrometer using a focussed monochromatic Al Kα (hν = 1486.7 eV) x-ray source. The x-ray beam used was a 100 W, 100 µm diameter beam and was rastered over a 1.5 mm by 0.2 mm area. The survey scans were collected using a pass energy of 117.4 eV producing a Full Width at Half-Maximum (FWHM) of less than 1.6 eV for the Ag 3d 5/2 peak. The high energy resolution data were collected using a pass energy of 23.5 eV, producing a FWHM of less than 0.75 eV for the Ag 3d 5/2 peak. The collected data were referenced to an energy scale with binding energies for Cu 2p 3/2 at 932.67 +/– 0.05 eV, and Au at 84.0 +/– 0.05 eV. On some insulating samples, positive charging of the surface was observed due to the loss of electrons, causing the peaks to shift during data acquisition For these cases, low energy electrons were used to flood the specimen to neutralize the surface. The UPS data were obtained on a Riken Keiki AC-2 UV photoelectron spectrometer. The samples and detector were placed in open air. The UV source was a deuterium (D2) lamp with a spot diameter of 2 mm by 2 mm. For the samples with a high efficiency of photoemission (the metals and graphite) the light source power was 49.9 nW, and for those samples that have a low efficiency of photoemission (the polymers, coal, pyrite, and vitrinite) the light source power was increased to 600.2 nW. The resolution of the instrument for precision measurements is given as 0.02 eV. The samples were analyzed at a temperature of 22°C, a relative humidity of 40%, and a pressure of 1 MPa. 2.4. Results The XPS data are presented in Table 1. The relative atomic concentrations of the observed elements as reported in Table 1 were obtained by integrating the area under each peak of interest and normalizing with sensitivity factors supplied by

Dust removal from solar panels and spacecraft on Mars

299

Table 1. Surface element concentrations (atomic %) of selected materials as measured by XPS C

O

Cr

Ag Al

Mg F

Ca

Zn

Cl S

Copper

70

21 –

–

–

–

–

–

–

–

–

–

316L steel

53

35 0.8 –

0.9 –

8.7 1.7 –

–

–

–

–

–

–

–

Electro. 316L 69

24 2.1 1.8 1.3 –

0.8 0.7 –

1.1 –

0.2 –

0.3 0.3 –

Silver

40

16 –

–

–

–

–

–

40

–

–

–

–

–

2.6 –

Aluminum

23

53 –

–

–

–

–

–

–

24

1.2 –

–

–

–

–

PTFE

30

0.2 –

–

–

–

–

–

–

–

–

70

–

–

–

–

Nylon 66

79

13 7.8 –

0.2 –

–

–

–

–

–

–

–

–

–

–

Polystyrene

75

21 2.4 1.2 –

–

–

–

–

–

–

–

–

–

–

0.7

Glass

15

55 –

9.5 17

–

–

–

–

–

2.9 –

0.3 –

–

–

SiO2

9.9 64 –

0.6 26

–

–

–

–

–

–

–

0.7 –

–

–

Pyrite

51

27 0.4 0.6 –

–

5.4 –

–

–

–

–

–

–

1.4 14

Graphite

98

2.3 –

–

–

–

–

–

–

–

–

–

–

–

Polycarbonate 84

16 –

0.1 0.2 –

–

–

–

–

–

–

–

0.1 0.1 –

Acrylic

27 –

–

–

–

–

–

–

–

–

–

73

N

Na Si –

Cu Fe

1.4 7.4 –

–

–

–

–

–

–

the instrument manufacturer. The data showed that the metal specimens had particularly high levels of surface contamination (carbon and oxygen) compared to the ceramic and polymeric specimens. For example, copper showed 70 atomic % of carbon on the as-received surface, compared to the quartz (SiO2) specimen with only 9.9 atomic % of carbon. For PTFE, the XPS data showed the surface composition to be 30 atomic % carbon and 70 atomic % fluorine, which is very close to the CF2 stoichiometric composition of PTFE. This supports the fact that polymers do not pick up surface contamination in air as readily as metals. The graphite specimen shows only a minimal oxygen concentration (2.3 atomic %). The measured UPS data are shown in Table 2. The value of the work function for each material is compared with the value reported in the literature. It was observed in Table 2 that the measured work function for each material was higher than the reported values. A closer examination showed that the ratio of the measured-to-reported work function value for copper (1.17:1) was higher than for the other metals; silver (1.08:1), and aluminum (1.05:1); and these metals showed less carbon surface contamination by XPS than the copper. In contrast, the measured-to-reported work function value for PTFE was approximately unity, and the PTFE showed no discernible carbon contamination by XPS. Clearly, a correlation between the amount of surface contamination and increase in work function can be observed.

S. Trigwell et al.

300

Table 2. Measured work functions by UPS in air compared to work function values reported in the literature. Measurement errors were +/– 0.002 eV Material

Measured work function [eV]

Reported work function [eV]

Aluminum Silver Graphite Copper Stainless steel Polystyrene Pyrite Acrylic Polycarbonate Nylon 66 PTFE

4.53 4.66 5.09 5.11 5.37 5.48 5.50 5.52 5.57 5.61 5.80

4.30 4.30 4.50 4.38 Unknown 4.90 5.40 4.72 4.80 4.30 5.75

Figure 2. XPS survey scan of pyrite (FeS2).

As previously mentioned, pyrite was chosen as an example of a mineral found on Mars. For this sample, high resolution scans of the carbon, sulfur, and iron peaks were examined, and the chemical composition rather than just the elemental composition was determined. The XPS survey scan for pyrite is shown in Figure 2, and high resolution scans of the carbon, sulfur, and iron peaks are shown in Figure 3.

Dust removal from solar panels and spacecraft on Mars

301

(a)

(b)

(c)

Figure 3. High resolution XPS scans of peaks for pyrite (FeS2). a) Carbon C1s peak, b) Sulfur S2p peak, and c) Iron Fe 2p peak.

302

S. Trigwell et al.

The XPS survey spectrum shows the peaks associated with the surface composition. In addition to the peaks for carbon and oxygen, minor amounts of sodium, nitrogen, and chlorine were also detected. In the high resolution scans, the peak for each element was curve fitted with the component peaks as reported in numerous reference literature and data bases [16-18]. What can be observed from the high resolution peaks is that the carbon present on the mineral is predominantly C-C or C-H bonding indicating carbonaceous contamination, with minor amounts of carbon-oxygen species. A small amount of organic carbon-sulfur was also detected. However, the sulfur peak shows the sulfur to be mainly in the sulfate form with very little, if any, iron sulfide present. This is confirmed in the iron peaks where the iron is present predominantly as iron sulfate, and in this case no iron sulfide is detected. The data show that for the case of pyrite what is actually chemically present on the surface is very different from the bulk composition, which can affect the value of the work function. This is important in understanding how materials charge relative to each other in the triboelectric series. However, these data were taken in an Earth environment, and so it is of great interest to determine what the surface composition of minerals would be in a Martian environment. 3. DESIGN OF SENSOR

There are a number of instruments that can be used to characterize the aerodynamic size distribution of particles. Instruments such as a Faraday cup are available to estimate the net average electrostatic charge on particles samples. However, the choice of instruments for real-time simultaneous measurements of both aerodynamic diameter and electrostatic charge distributions of particles on a single particle basis is limited. The Electrical Single Particle Aerodynamic Relaxation Time (E-SPART) analyzer is used extensively for simultaneous characterization of particle size distribution (PSD) and electrostatic charge distribution [19]. The analyzer can be used in the diameter range from 0.5 to 50 µm and charges in the range from 0 to their saturation charge limit. The E-SPART analyzer, as shown in Figure 4, uses an AC electric drive to oscillate the particles in air. The resultant oscillatory motion of the particle lags behind the external AC field. The phase lag (φ) relates to the aerodynamic diameter of the particle, and the amplitude of the particle trajectory determines the particle charge and polarity, as shown in Figure 5(a). Figure 5(b) shows a still from a video image taken of charged particle tracks in the chamber. The particle tracks are analyzed by Laser Doppler Velocimetry. The details of the operation of the instrument are available [20-23]. Figures 6(a) and (b) show typical particle size and charge distribution, respectively, for a positive copier toner with a mean particle size of 8 µm. Although the data show the overall particle count, individual particle data can be extracted.

Dust removal from solar panels and spacecraft on Mars

303

Figure 4. Schematic of the E-SPART analyzer.

(a)

(b) Figure 5. (a) Principle of operation of E-SPART, and (b) video image of particle tracks. The ESPART analyzer uses an AC electric drive to oscillate the charged particles.

304

S. Trigwell et al.

(a)

(b) Figure 6. (a) Particle size distribution, and (b) charge distribution for a positive toner.

Dust removal from solar panels and spacecraft on Mars

305

4. DEVELOPMENT OF A SELF-CLEANING PANEL USING ELECTRODYNAMIC SCREENS

Dust settling out of the atmosphere onto any horizontal surface is a potential problem in the obscuration of solar arrays. A method is required to periodically remove the dust, or prevent the dust from settling in the first place. Ideally, a method that requires no moving parts and is robust in operation is most desirable, as shown in Figure 7. The static charges on the particles provide an opportunity to prevent dust deposition by using an AC voltage driven electrode screen. This type of screen creates a repelling force to the charged particles regardless of their polarity. The device consists of an electrode screen that contains a number of parallel electrodes placed equidistant from each other, embedded in a insulating coating as shown in Figure 7.

Figure 7. Solar panel with embedded electrodynamic screen.

306

S. Trigwell et al.

5. DEVELOPMENT OF A METHOD FOR REMOVING DEPOSITED DUST

At present, there are four categories of dust-removal methods, namely, natural, mechanical, electromechanical, and electrostatic [1]. It has been observed that wind velocities on Mars are insufficient to remove settled dust on its own. A possible aid to natural dust removal would be to devise a movable array so it can be turned vertical such that gravity in addition to wind may remove the dust. By designing an array with a vibration frequency that would correspond to the wind would assist in the dust removal. Electromechanical methods are similar to natural methods, and include vibrating, shocking, or using ultrasound, in combination with tilting the array to remove the dust. These methods would require sophisticated mechanisms. The problem with any mechanical device is the risk of mechanical failure of one of the components that will be catastrophic in an alien environment where it cannot be accessed to be repaired. In this section, a method is described to mechanically clean the solar panels by a mechanism using a NiTi (Nickel-Titanium, also know as nitinol) shape memory alloy. This method is unique in that it involves only a thermoelastic process (no motors or electrical components) and thus considerably reduces the probability of failure. 5.1. Proposed method for the removal of dust particles A proposal for the development of an electrostatic brush for removing particles from different surfaces was based on empirical studies using materials such as polystyrene (PS), poly(tetrafluoroethylene) (PTFE), and polyamides (Nylon). A brush made of Nylon and PTFE fibers may be effective in getting tribo-charged and in removing particles with both negative and positive polarities. The objective is, therefore, to build a device that would utilize such a brush but would have minimal moving and mechanical parts that would reduce the susceptibility to failure. The proposed device would be built using a smart material, namely NiTi alloy, which will move the wiper/brush to clean the surface of the solar array when subjected to heating by the sun. Shape memory alloys, such as nickel-titanium, are a class of unique alloys that can be deformed, but then recover their original shape when heated. This is due to the occurrence of a martensitic phase transformation and its subsequent reversal. Figure 8 [24] shows a typical plot of property changes versus temperature for a shape memory alloy. Basically, the parent phase is an austenite, and the alloy is deformed into the martensite phase. Upon heating through its transformation temperature, it reverts back to austenite and recovers its previous shape with great force. This process can be repeated millions of times. The shape recovery process occurs over a range of a few degrees. The temperature at which the alloy “remembers” its higher temperature form when heated can be adjusted by slight changes in its composition. The lowest active transformation temperature for commercially available NiTi alloy is at present 0–10°C, containing 55.8 wt% Ni (Alloy C – Shape Memory Applications, Inc., Santa Clara, CA); however, alloys with

Dust removal from solar panels and spacecraft on Mars

307

Figure 8. Property change versus temperature for martensitic transformation in NiTi alloy.

Figure 9. Daily temperature variation at Ares Vallis landing site on Mars.

lower transformation temperatures (< –50°C) are available. Data from the Pathfinder mission showed a high level of consistency in the surface temperature range at the Ares Vallis landing site [25], as shown in Figure 9. From these data, a NiTi alloy with a transformation temperature of approximately –50°C (223°K) would suffice.

308

S. Trigwell et al.

(a)

(b)

(c) Figure 10. (a) NiTi spring in tension. (b) Upon heating through the transformation temperature, and (c) after heating.

Dust removal from solar panels and spacecraft on Mars

309

NiTi is also superelastic, in that it possesses incredible amounts of flexibility and kink resistance. It has a strain recovery of about 8%, which makes it more resilient than stainless steel. Work on the corrosion resistance of NiTi in the case of biomedical implants [26] has shown it to be very resistant to corrosion in the harshest environments due to the formation of a passive TiO2 layer on the surface. The design of the cleaning device involves attaching the brush between two sets of springs, one set normal stainless steel, and the opposing set a NiTi alloy. The tension would be set so that the steel springs would be contracted, holding the brush to one side of the solar array, and the opposing NiTi spring set would, therefore, be extended against the strain. Upon heating of the device by the sun, the NiTi alloy would pass through its transformation temperature and contract, the recovery strain now being greater than that of the steel springs, and would effectively pull the brush across the array. A prototype device is shown in Figure 10, where a heat gun was used to heat the alloy spring. When the temperature drops back past the transformation temperature, the NiTi alloy will relax, allowing the steel spring to now contract again, pulling the brush back. The cleaning would be repeated several times a day as the temperature fluctuates depending upon the conditions. 6. CONCLUSIONS

The tribocharging of Mars dust can contribute to strong adhesion of dust particles to solar panels, spacecraft, and spacesuits. Both the polarity of the charged dust particles as well as the amount of charge depend upon the surface composition of the particles and the contacting materials. It has been shown that in an ambient Earth environment, surface contamination and oxidation produce significant changes to the surface composition and hence the work function of a variety of materials. An instrument has been developed that can simultaneously measure both size and charge distributions on an individual particle basis. A smaller, more robust version of which is proposed to measure size and charge distributions of Mars dust in situ. An electrodynamic screen shows promise for preventing deposition of charged dust particles, and an electrostatic brush for dust removal has been developed using NiTi shape memory alloy. The electrodynamic screen and cleaning device proposed have no mechanical parts so the probability of failure is minimized. 7. CONTINUING AND FUTURE WORK

In order to understand the tribocharging properties of Mars dust, an environmental chamber has been constructed that can effectively control the relative humidity from 0 to 98% with a +/– 1% stability. The atmospheric composition within the chamber can simulate that on Mars (predominantly CO2). The chamber can be mounted on top of an E-SPART analyzer. The charging properties of a Mars dust simulant, obtained from NASA Johnson Space Center (JSC Mars-1),

310

S. Trigwell et al.

will be investigated against stainless steel and PTFE to determine the charging characteristics in a Martian type environment. Simultaneously, an Ultra-violet Photoelectron Spectrometer is being developed to allow work function measurements to be taken as a function of relative humidity within the chamber. A compact E-SPART analyzer that is robust enough for space flight is also being developed, as well as the development of the electrodynamic screen is being continued. REFERENCES 1. G.A. Landis, Paper presented at the Intersociety Energy Conversion Engineering Conference, Honolulu, HI, July 27–August 1, 1997. 2. Mars News, www.marsnews.com/planetology, November 1999. 3. G.A. Landis and P.P. Jenkins, Proc. 26th IEEE Photovoltaic Specialists Conference, 865-869 (1997). 4. G.A. Landis, Acta Astronautica, 38, 885-891 (1996). 5. R. Greeley, N. Lancaster, S. Lee and P. Thomas, Mars, pp. 835-933, University of Arizona Press, Tuscon, AZ (1992). 6. J.R. Gaier, M.E. Perez-Davis and M. Marabito, Paper presented at the 16th AIAA/NASA/ ASTM/IES Space Simulation Conference, Albuquerque, NM, November 5–8, 1990. 7. J. Cross, Electrostatics: Principles, Problems and Applications, Adam Hilger, Bristol, England (1987). 8. Y. Murata, Jap. J. App. Phys., 18, 1-8 (1979). 9. J. Lowell and A.C. Rose-Innes, Adv. Phys., 29, 947-1023 (1980). 10. C.B. Duke and T.J. Fabish, Phys. Rev. Lett., 37, 1075-1078 (1976). 11. D.A. Hays, Proc. International Conf. on Modern Electrostatics, Ruinian Li (Ed.), Beijing, China, pp. 327-330, International Academic Publishers, New York (1988). 12. H. Bauser, Dechema Monographs, 72, 11-28 (1974). 13. W.R. Harper, Contact and Frictional Electrification, Laplacian Press, Morgan Hill, CA (1998). 14. D.M. Taylor and P.E. Secker, Industrial Electrostatics: Fundamentals and Measurements, John Wiley & Sons, New York (1994). 15. C.D. Cooper and J.F. Mustard, Paper #6164 presented at The Fifth International Mars Science Conference, Pasadena, CA (1999). 16. Ph. De Donato, C. Mustin, R. Benoit and R. Erre, Appl. Surface Sci., 68, 81-93 (1993). 17. C.D. Wagner, W.M. Riggs, L.E. Davis and J.F. Moulder, Handbook of X-Ray Photoelectron Spectroscopy, Perkin-Elmer Corp., Eden Prairie, MN (1983). 18. NIST XPS Database, http://srdata.nist.gov/xps/Bind_e_spec_query.asp 19. M.K. Mazumder and R.E. Ware, US Patent #4633714 (1987). 20. P.A. Baron, M.K. Mazumder and Y.S. Cheng, in: Aerosol Measurements: Principles, P. Baron and K. Willeke (Eds.), Chap. 17, Van Nostrand Reinhold, New York (1992). 21. M.K. Mazumder, S. Banerjee, R.E. Ware, C. Mu, N. Kay and C.C. Huang, IEEE Trans. Ind. Applications, 30, 365-369 (1994). 22. M.K. Mazumder, S. Banerjee and C. Mu, in: Dispersion and Aggregation, B.M. Moudgil and P. Somasundaran (Eds.), Engineering Foundation, New York (1994). 23. M.K. Mazumder, R.E. Ware, J.D. Wilson, R.G. Renniniger, F.C. Hiller, P.C. McLeod, R.W. Raible and M.K. Testerman, J. Aerosol Sci., 10, 561-569 (1979). 24. C.R. Wayman, MRS Bull., 49-56, April 1993. 25. Mars Pathfinder website: http://www.mars.jpl.nasa.gov (1997). 26. S. Trigwell, R.D. Hayden, K.F. Nelson and G. Selvaduray, Surface Interface Anal., 26, 483-489 (1998).

Surface Contamination and Cleaning, Vol. 1, pp. 311–334 Ed. K.L. Mittal  VSP 2003

Laser cleaning of silicon wafers: Prospects and problems M. MOSBACHER,1 V. DOBLER, M. BERTSCH, H.-J. MÜNZER, J. BONEBERG and P. LEIDERER University of Konstanz, Department of Physics, Fach M676, D-78457 Konstanz, Germany

Abstract—We report on experiments on the underlying physical mechanisms in the Dry- (DLC) and Steam Laser Cleaning (SLC) processes. Using a frequency doubled, Q-switched Nd:YAG laser (FWHM=8 ns) we removed polystyrene (PS) particles with diameters in the range of 110 nm to 2000 nm from industrial silicon wafers by the DLC process. The experiments have been carried out both in ambient conditions as well as in high vacuum (10−6 mbar) and the cleaned areas have been characterized by atomic force microscopy for damage inspection. In DLC we have determined the cleaning laser fluence thresholds for a large interval of particle sizes. Additionally we could show that particle removal was due to a combination of at least three effects: substrate thermal expansion, local substrate ablation as a consequence of field enhancement at the particle, and explosive evaporation of moisture adsorbed from the air. Which effect dominates the process depends on the boundary conditions. For our laser parameters no damage-free DLC was possible, i.e. whenever a particle was removed by DLC we damaged the substrate by local field enhancement. In our SLC experiments we determined the amount of superheating of a liquid layer adjacent to surfaces with controlled roughness. On silicon wafers the water layer could be superheated to 250 ◦ C prior to the onset of laser induced bubble nucleation. The heat transfer from the silicon substrate into the liquid was found to be limited by a thermal boundary resistance which can be characterized by a heat transfer coefficient of 3 · 107 W/(m2 K). Based on the knowledge about the particle removal mechanisms and the determined cleaning efficiency we discuss the advantages and disadvantages of DLC and SLC as possible future industrial surface cleaning procedures. Keywords: Particle removal; laser cleaning; field enhancement; cleaning mechanisms.

1. INTRODUCTION

The removal of particle contamination from surfaces is one of the crucial prerequisites for a further increase in the integration density of ICs and for the progress in nanotechnology. At all stages of the production of ICs, e.g., from the bare Si wafer to the patterned chip, particles even smaller than 100 nm in size can cause a damage to the produced structure and hence be responsible for the failure of the final device. In the late 1980s, the experts in the field of cleaning technology predicted that traditional cleaning methods such as ultrasonics and wet techniques would reach 1 To whom all correspondence should be addressed. Phone: +49-7531-882627, Fax: +49-7531-

883127, E-mail: [email protected]

312

M. Mosbacher et al.

their limit of capability [1, 2]. In addition, these traditional techniques were and still are harmful to the environment as they consume large quantities of aggressive chemicals and water. Although the traditional methods have been continuously improved [3], still particle contamination causes considerable production losses [4], and with further shrinking of line widths [5] there is a definite need to replace traditional methods by new cleaning technologies. One of these new approaches is called laser cleaning. In Dry Laser Cleaning (DLC) [6– 8] the surface to be cleaned is irradiated by a short laser pulse. In Steam Laser Cleaning (SLC) [6, 7, 9, 10] prior to the application of the laser pulse a liquid, e.g. a water-alcohol mixture, is condensed onto the surface. After the first attempts of implementing laser cleaning in prototype cleaning tools [11], this strategy was not pursued any further as there were too many open questions related to the underlying physics. In the following years several research laboratories around the world [12– 20] started to investigate the physical processes involved both in SLC and DLC, starting from the simple scenarios suggested by the authors of the first publications on the subject. These scenarios also formed the basis for certain models to describe laser cleaning and to interpret the experimental results obtained [6, 7, 10, 18, 19, 21– 33]. However, recent experiments [31, 33– 41] show that both the DLC and SLC scenarios that have been taken as common sense so far do not incorporate all the important cleaning mechanisms and hence are oversimplified. In this article we will first summarize our knowledge on the cleaning processes involved in laser cleaning and their interplay, and then present the results of systematic measurements of cleaning efficiencies in both DLC and SLC for particle sizes from 110 nm up to 2000 nm. The interpretation of these results will clearly point out the importance of the cleaning mechanisms neglected in the original SLC and DLC scenarios. Against this background we will discuss briefly the state of theoretical modeling of laser cleaning. Based on the previous sections we will finish the article with a statement of the prospects and problems of laser cleaning as an industrial cleaning process from today’s state of knowledge.

2. EXPERIMENTAL ASPECTS

2.1. Sample preparation In our quantitative studies on the cleaning efficiency we did not use irregularly shaped particle contaminants commonly used in many laser cleaning studies (Al2 O3 , Si3 N4 ,...), but spherical colloidal polystyrene (PS; Interfacial Dynamics Corporation, Portland, OR, USA) and SiO2 (Bangs Laboratories Inc., Fishers, IL, USA and Duke Scientific Corp., Palo Alto, CA, USA) particles. These particles are advantageous for investigation of the underlying physical processes involved in laser cleaning due to their narrow size distribution (standard deviation ± 5% for PS, ± 20% for SiO2 ) as compared to irregular particles. This enables studies of

Laser cleaning of silicon wafers: Prospects and problems

313

Figure 1. Typical sample as used in the laser cleaning experiments imaged in a scanning electron microscope. The displayed area is 4.8 µm × 4.8 µm and the particle size is 110 nm.

removal efficiencies for various, well-defined sizes. Their spherical shape additionally facilitates a comparison with theoretical models, as adhesion forces of particles are mostly calculated for the geometry of a sphere on a flat substrate. Some experiments were also performed using irregularly shaped Al2 O3 particles (Summit Chemicals Europe GmbH, Düsseldorf, Germany) as contaminants. As substrate we used industrial silicon (100) wafers (Wacker Siltronic, Burghausen, Germany) that were cleaned in isopropyl alcohol (IPA) in an ultrasonic bath before applying the contaminants. The particles were deposited on the silicon substrate by a spin coating process, described in detail in [17, 41]. We were able to prepare samples with more than 95% of isolated spheres at particle densities above 1000 per cm2 . A typical example can be seen in Fig. 1 where 110 nm sized PS particles were deposited onto a Si wafer. Care was taken to prevent particle agglomeration, which is important for quantitative experiments, as agglomerates exhibit a different cleaning behaviour compared to single particles [15, 35]. 2.2. Laser sources For all experiments we used a frequency doubled, Q-switched Nd:YAG-laser (λ = 532 nm, FWHM = 8 ns).

314

M. Mosbacher et al.

2.3. Evaluation of the cleaning efficiency Particle removal in the cleaned area (about 1 mm2 ) was determined under ambient conditions by a light scattering technique. A 5 mW HeNe laser (λ=633 nm) illuminated a spot with a diameter of 0.5 mm, which corresponds to several hundred particles monitored. The light scattered by the particles was detected by a photomultiplier. The monitored area was much smaller than the illuminated area, therefore in this case the laser fluence can be considered as almost homogeneous. The cleaning efficiency, defined as the fraction of particles removed, was determined by comparing the scattered light intensities before and after the cleaning with a clean reference sample [17]. In high vacuum (HV) we determined the fluences necessary for particle removal by inspecting the illuminated spot with an optical microscope prior to and after the laser pulse. In addition, we measured the cleaning laser fluence threshold relative to the melting threshold of Si by monitoring the reflected light of the HeNe laser with ns time resolution. As the laser fluence for the onset of melting of silicon is well known, this can be used for conversion of relative fluences into absolute numbers. 2.4. Determination of laser fluence The determination of laser fluences for the nanosecond pulses is described in detail in [17]. Briefly, the laser fluence was determined relative to the well-known melting threshold fluence of Si, making use of the higher reflectivity of the molten with respect to the solid phase. This was done by time-resolved monitoring of the reflected light of a HeNe laser (λ=633nm). Simultaneously the laser’s scattered light was detected in order to probe the cleaning efficiency. The techniques described above are not suitable for determining laser fluences in bulk liquids. Hence for the bubble nucleation experiments we chose another, intrinsic laser fluence calibration method. During the experiment we found a distinct change in the reflectivity of the water/silicon system. As computer simulations of the temperature dependency of this reflectivity show, this change is primarily caused by the temperature change in the water. Based on this, it is possible to derive [42] that the integral 9µs Rp (t) W := − 1 dt ∼ F (1) R0,p 4µs containing the reflectivity change in p-polarization (R0,p initial reflectivity, Rp (t) reflectivity after bubble nucleation) is directly proportional to the applied laser fluence F . The upper boundary of the integral is given by the end of the detection time, the lower boundary of 4 µs was chosen as the time where no gas bubbles were present on the surface any more. Hence, the applied local laser fluence right in the area where bubble nucleation is probed can be determined from the measured reflectivity changes of the system.

Laser cleaning of silicon wafers: Prospects and problems

315

Figure 2. Interferometric measurement of the laser-induced surface displacement. The frequency doubled Nd-YAG pulse is attenuated by glass plates (A) and guided to the silicon substrate by several prisms (P). A heterodyne interferometer (IF, B.M. industries, SH-130, bandwidth 200 kHz45 MHz) measures the surface displacement of the sample. The temporal pulse shape is captured by a photodiode (PD). Displacement and pulse shape are recorded on a digital storage oscilloscope. A lens (L) was used to increase the laser fluence at the substrate.

2.5. Surface acceleration measurement setup A detailed description of the setup shown in Figure 2 which was used for the determination of surface accelerations can be found in [43]. Briefly, we illuminated the silicon substrate by a frequency doubled Nd:YAG pulse and determined the surface displacement of the sample by a heterodyne interferometer. Both displacement and pulse shape were recorded on a digital storage oscilloscope. 2.6. Bubble nucleation experiments Figure 3 shows the setup for the bubble nucleation experiments. In order to nucleate the bubbles a Q-switched Nd:YAG laser (λ=532 nm, FWHM=8 ns) heated the silicon sample. The pulse energy was split (BS) and measured for each individual pulse by an energy meter (FM: Field Master, Coherent). Sample and liquid were placed in a fused silica cuvette that could be heated up to 360 K. Bubble growth was monitored by a cw Ar-ion laser (λ=488 nm, P=175 mW) which was focused onto the sample. Both the specularly reflected beam of the Ar-ion-laser and the light scattered by the nucleated bubbles were collected in forward direction (as shown in the diagram), as well as the scattered light perpendicular to the incident ray (similar setup, not shown in the diagram). For all the light detection we used fast photodiodes (PD: FND 100, rise time < 1 ns) covered by interference filters (IF). By a polarizing beam splitter (PBS) the reflected beam was decomposed into its p- and s-polarized constituents that were detected individually.

316

M. Mosbacher et al.

Figure 3. Optical determination of laser induced bubble nucleation (for details see text).

3. PARTICLE REMOVAL MECHANISMS

Already the authors of the first publications on laser cleaning [6, 7, 9, 10] suggested physical mechanisms to describe the particle removal process. In DLC the removal process was ascribed to the thermal expansion of the substrate due to the heating with the laser pulse. SLC was explained as a consequence of the explosive evaporation of the applied liquid. However, recent research [40, 44, 45] in the field of laser cleaning has shown that several other mechanisms are of importance for the process. In the following sections we will describe the cleaning mechanisms experimentally verified so far. 3.1. Thermal substrate expansion Most authors explain the particle removal process in DLC in the following way: during the laser pulse its energy is absorbed in the substrate. Due to the subsequent thermal expansion the surface with the adhering particle is accelerated and the particle gains kinetic energy. Depending on their elastic properties some energy is also stored in elastic deformation of both particle and surface. At the end of the pulse the expansion of the surface stops and the particle leaves the surface due to its inertia.

Laser cleaning of silicon wafers: Prospects and problems

317

From this it is clear that a measurement of the dynamics of the substrate expansion upon illumination with a laser pulse provides insight into this cleaning mechanism. Detailed discussions of our experiments can be found in [41, 43], therefore we will only highlight the results here. Using an interferometric setup as described in Section 2 we measured the expansion of a silicon substrate after illumination with a Nd:YAG laser pulse (FWHM = 10-20 ns, λ=532 nm) with ns time resolution. Typical values of this displacement were a few nm. By numerical derivation of the averaged displacement as a function of time of 600 individual experiments we obtained the deceleration of the substrate surface. This deceleration of the surface, rather than its acceleration as discussed by most authors in the description of DLC, is the relevant parameter for the particle removal [45]. For typical laser cleaning fluences of a few 100 mJ/cm2 the decelerations are in the order of about 107 m/s2 , just the order of magnitude that is thought to be necessary for removal of such small particles [7]. The experimental values are in good agreement with a simple theoretical description of the thermal expansion of the substrate. Taking into account only the 1D heat equation2 and denoting by R the reflectivity, by α the linear thermal expansion coefficient, by CP the specific heat of the substrate, by ρ its density, by f (t) the time dependent normalized intensity of the laser, and by F its fluence the surface displacement is given by α (1 − R)F d(t) = ρ CP

t

f (t ) dt

(2)

−∞

as the Grüneisen parameter α/(ρ CP ) is almost temperature independent. Expression (2) can be derived for a known pulse shape f (t), and for a gaussian pulse with FWHM τ it yields a maximum deceleration of amax = −1.71 · 106 g

ns2 F 2 τ2 mJ/cm

(3)

with a strong 1/τ 2 dependence on the pulse length. Experimentally we were able to verify the linear dependence of both displacement and maximum deceleration on the applied laser fluence. However, the 1/τ 2 dependency on the pulse length was not found, which might be explained by deviations of the actual pulse shape from an ideal Gaussian especially for long pulses. In general, the experimental data were in good agreement with the ones calculated from equation (3). 2 Thermoelastic effects [46, 47] are not taken into account, as the area where the displacement is

detected by the interferometer was at least ten times smaller than the illuminated spot and was located in its center. Thus lateral thermal gradients are neglected at this spot.

318

M. Mosbacher et al.

Figure 4. Intensity (E 2 ) of the electromagnetic field around a PS particle (diameter 1700 nm, refractive index 1.59) in vacuum when illuminated by a laser at λ=800 nm. The laser enters from the left, and the light is polarized in the image plane. At the light averted side (i.e., not directly illuminated by the laser) of the particle the intensity is enhanced by a factor of about 30.

3.2. Local substrate ablation Besides the thermal expansion of the substrate due to heating by the absorbed laser energy we experimentally identified [35, 38, 40] a second cleaning mechanism which is pronounced particularly in DLC. 3.2.1. Optics of particles with sizes comparable to the wavelength In 1908, the German physicist Gustav Mie described the scattering of light at a dielectric sphere in vacuum [48]. Two predictions of his theory are particularly important for the application of laser light to the removal of particles comparable in size with the wavelength. First, one would expect a “focusing” effect, i.e. the laser intensity underneath the particle and hence in the substrate plane should be higher than the incoming laser intensity that hits the bare part of the surface. For particles of sizes much larger than the wavelength (about one order of magnitude or more) this focusing can be explained in terms of geometrical optics as focusing by a spherical lens. However, geometrical optics fails to explain the intensity enhancement underneath particles smaller in size than the wavelength. In this case the near field distribution of the

Laser cleaning of silicon wafers: Prospects and problems

319

electromagnetic field at the particle has to be computed numerically. An example of such a computation using a program based on [49] is given in Fig. 4. But not only does the Mie theory describe an enhancement of the laser intensity in the particles’ near field, it also predicts that for certain values of the size parameter π d/λ (d denoting the particle diameter, λ the laser wavelength) the enhancement should be particularly efficient, resulting in a resonant intensity enhancement, the so-called “Mie-resonances”. 3.2.2. Near-field induced substrate damage When inspecting contaminated samples by scanning electron microscopy (SEM) or atomic force microscopy (AFM) after DLC using ns laser pulses, the consequences of the field enhancement process became obvious: all over the cleaned areas we found substrate damages localized exactly at the former particle positions [35, 37– 39]. These damages manifested as melting pools or even holes in the surface, typical examples can be seen in Fig. 5. The consequences for the laser cleaning process are obvious. The intensity enhancement reduces the maximum laser fluence that can be applied in the process. Usually in laser cleaning studies [19, 31] the laser fluence corresponding to the melting threshold of a bare surface is taken as the damage threshold fluence. Our experiments show clearly that this is an inadequate definition. Instead one must take into account the enhanced laser fluence underneath the particles, as it will be discussed in Section 4. From the obtained AFM images we were able to analyse in detail the surface profile at the damaged sites. Here we found that for high field enhancement factors the silicon substrate was not only molten, but that some material was even ablated (see Sec. 4). The momentum transfer to the particles during the ablation process significantly contributes to the cleaning process and hence local substrate ablation

Figure 5. Surface damage caused by local melting and ablation of the substrate. The melting was induced by the enhancement of laser intensity in the near field of the particles at the surface. Shown are AFM images in top view (left) and a 3D illustration (right) of the same site.

320

M. Mosbacher et al.

must be considered as a particle removal mechanism. But of course, due to the accompanying substrate damage, this removal mechanism should be suppressed in any practical application of laser cleaning. 3.2.3. Resonances in the laser cleaning threshold fluence As a second consequence, the optical resonances in the enhancement should give rise to a resonant, non-monotonous dependency of the laser cleaning threshold fluence on the size parameter. In case of a resonant enhancement, the laser fluence for particle removal should be less than in absence of resonance. In Section 4 we will demonstrate the existence of the resonances from experimental determination of threshold fluences for a variety of particle sizes. At this point it is worth to mention that the observation of the size dependent Mie-resonances in the threshold fluence was only possible due to the use of spherical colloidal particles with a small size distribution (± 5%). 3.3. Explosive evaporation of a liquid Although laser cleaning has been known for more than ten years now, still many questions related to the underlying physical processes are not answered yet. The authors of the first publications on the subject [6– 10] suggested simple physical scenarios that were accepted for the interpretation of experimental data thereafter and were taken as a basis for theoretical modeling: DLC was explained solely by the thermal expansion of the substrate and the adhering particle, SLC by explosive evaporation of the liquid condensed onto the surface [6, 7, 10, 18, 19, 21– 33]. However, it is still not clear whether these scenarios accurately reflect the actual particle removal process for all laser parameters and cleaning environments studied so far, even in the experiments that are quoted to justify the models. This is because the models suffer from major drawbacks, such as the neglect of the cleaning mechanisms field enhancement induced local substrate ablation. In addition to the thermal expansion of the substrate the early publications on laser cleaning [6, 7, 9, 10] suggested a second mechanism for the removal of dirt particles: the explosive evaporation of a liquid such as water or alcohol. Here the liquid layer is heated directly [10] (laser energy absorbed in the liquid) or indirectly [6, 7] (laser energy absorbed in the substrate and transferred to the liquid via heat conduction) by the laser pulse. Due to the short time scale of this heating the liquid is superheated to a certain extent, becomes intrinsically unstable and subsequently evaporates explosively. The momentum transfer of this explosion onto the particle is then thought to lift the particles off the surface. In the following, we will discuss this cleaning mechanism in three variations: the evaporation of liquid adsorbed at the particle-surface area as it is the case for DLC in ambient air, the evaporation of a liquid film as it is found in SLC, and the laser induced nucleation of bubbles in a bulk liquid that provides considerable information on the underlying physics.

Laser cleaning of silicon wafers: Prospects and problems

321

3.3.1. Liquid adsorbed at the particle-surface contact area Several authors report that in their DLC experiments an increase in ambient humidity led to an increase in cleaning efficiency [31, 33, 35]. This was attributed to the additional cleaning force provided by the explosive evaporation of the water adsorbed at the interstice between particle and surface. In a recent publication on the DLC of polyimide [33] the authors systematically increased the humidity from 30% to 90% and found an efficiency increase, a strong support for the above interpretation. However these experiments were not carried out on silicon and the obtained data originated only from two particle sizes. One straightforward way of minimizing the amount of water present at the interstice of particle and surface is the heating of the substrate to a temperature above the boiling point of water. We used arbitrarily shaped Al2 O3 -particles 300 nm in diameter as contaminants in order to minimize cleaning by local ablation and to prevent a deformation of the particles due to heating. For comparison we performed two different sets of DLC (Nd:YAG laser, λ=532 nm, FWHM=8 ns) experiments. In the first one we carried out DLC at ambient conditions, as commonly done in the investigations published in literature. During the second set of experiments, we heated the sample to a temperature of 120 ◦ C prior to the application of the contaminant particles and during the whole experiment. As a result, we found an increase in the cleaning threshold laser fluence for the heated sample of more than 20% versus the sample cleaned in ambient conditions. This result can be taken as a further indication for the involvement of liquid adsorbed from the atmosphere and stimulated a more systematic and detailed investigation as described in Section 4. 3.3.2. Liquid films Laser cleaning can not only be promoted by the adsorption of atmospheric humidity, but also by the condensation of a liquid onto the sample on purpose. This process, called Steam Laser Cleaning (SLC), relies very much on the explosive evaporation of the liquid film. So far the reported studies on this subject [15, 17, 36] lack a systematic investigation of the parameters of this film, i.e. its thickness and composition. However, this control on the film parameters is not easy to achieve, and thus all research on the dynamics of laser induced bubble nucleation in liquids, and hence the physical processes underlying SLC, published in the literature so far has been carried out in bulk liquids. 3.3.3. Laser induced bubble nucleation in bulk liquids The study of laser induced bubble nucleation in liquids provides the key for the understanding of liquid enhanced laser cleaning processes and their increased efficiency in comparison to dry techniques. Different methods were used for these studies: optical methods such as the change of the substrate reflectivity due to the bubble film [50– 54] and the detection of light scattered by the bubbles [50, 52, 54], the detection of the pressure wave nucleated during the growth or collapse of the bubbles [52, 54– 56] or surface plasmon spectroscopy [54].

322

M. Mosbacher et al.

Typically, these studies were carried out on rather rough surfaces such as metal films. Here quite moderate superheatings of about 20 K were found to be sufficient for the nucleation of laser induced bubbles, a value much lower than the theoretically predicted one of 200 K for water [57]. This has been attributed to the surface roughness of these substrates. But to our knowledge there are no systematic studies that confirm this roughness-dependency of the threshold. Yet this is the crucial point in the transfer of the obtained results from the water/metal film system to the water/silicon wafer system under consideration in laser cleaning studies. With respect to this several open questions arise: 1) Supposing rough metal films provide nucleation sites for gas bubbles and cause only moderate superheatings of the liquid, what will in contrast be the degree of superheating on smooth substrates like silicon wafers? 2) Is it justified to determine the temperature of the superheated liquid by measuring [58] or computing [28] the temperature of the silicon substrate and simply assuming a perfect heat transfer from the substrate into the liquid? 3.3.3.1. Bubble nucleation thresholds. We, therefore, studied [40, 42] the nucleation of bubbles at a superheated liquid/solid interface under controlled surface roughness. When the incident laser fluence reached a well defined threshold, scattered light was observed. This indicates a sharp nucleation threshold. The threshold decreases with increasing starting temperature T0 of the water as less energy is needed to reach the nucleation temperature. Figure 6 shows these temperature dependent thresholds for the examined systems. The lines are linear extrapolations to vanishing fluence and yield the nucleation temperature by their intersection with

Figure 6. Comparison of the bubble nucleation thresholds and extrapolated superheating limits of water on substrates of different roughnesses.

Laser cleaning of silicon wafers: Prospects and problems

323

the temperature axis. As a consequence of the interpolation over a large temperature interval this procedure yields a rather large error of about 20%. In a first series of experiments we determined the nucleation thresholds and superheating temperatures for water on a 50 nm silver film thermally evaporated onto glass. This system was chosen for comparison with previous experiments in [54, 56], and indeed we found a nucleation temperature of (130 ± 30) ◦ C, only slightly above the one reported in the literature. In contrast, water on conventionally polished silicon wafers (rms roughness 0.2 nm) exhibited a bubble nucleation temperature of (250 ± 30) ◦ C in close agreement with the expected value from theory and far above that one measured on the silver film. To verify the influence of the surface roughness we patterned a silicon wafer with holes (diameter approx. 500 nm, depth approx. 40 nm, hole density approx. 0.05/µm2 , for the preparation method see reference [39]). On this substrate the nucleation temperature decreased strongly to (160 ± 13) ◦ C, close to that of the rough silver film, a clear evidence for the influence of surface roughness. 3.3.3.2. Heat transfer coefficient. It is well known that there exists a discontinuity in the temperature profile at boundaries between different materials. This is due to the finite heat conductivity of the boundary region. The so-called heat transfer coefficient ˙ Q (4) ζ = AT ˙ the quantifies this thermal boundary resistance as functions of the heat flow Q, boundary area A and the temperature jump T . Clearly, in laser cleaning the thermal resistance can limit the heat flow from the substrate into the liquid and thus lower the liquid temperature considerably. None of the published computations of temperature profiles in laser cleaning incorporates this fact. Probably one reason is that although the phenomenon is well investigated for low temperatures below 50 K (Kapitza resistance, see [59]) and for technical applications at room temperature, long time scales (several seconds) and macroscopic dimensions (Nusselt-number, see [60]), there are no data at ns time scales and nm length scales, as needed for the interpretation of laser cleaning data. The data obtained in the bubble nucleation experiments allow us now for the first time to give the heat transfer coefficient for these scales. Figure 7 shows computed maximum temperatures of the water layer adjacent to the silicon surface. Our computations are based on the 1D heat equation implemented in a finite element algorithm taking into account different values for the heat transfer coefficient ζ between silicon and water. The calculations have been repeated for different starting temperatures of the water and the corresponding experimentally determined laser threshold fluence for bubble nucleation was applied. At this threshold fluence the water layer must reach a specific temperature which is the same for all starting temperatures. If the assumed value of ζ is too small or too large, however, the water layer will reach maximum temperatures that are

324

M. Mosbacher et al.

Figure 7. Computed maximum temperatures in water on a silicon surface as function of the heat transfer coefficient (HTC) ζH2 O for different starting temperatures T0 . As laser fluence we used the experimentally determined threshold fluence for bubble nucleation.

different for each starting temperature. It can be seen from these computations that for all starting temperatures of the water the graphs of the calculated maximum temperatures as a function of the assumed heat transfer coefficient intersect at one single point. This point is given by a value of ζH2 O = 3 · 107 W/(m2 K) and a maximum water temperature of 250 ◦ C – just as determined experimentally. Therefore, this ζ -value represents the heat transfer coefficient in the studied system. Interestingly, the computations exclude values of ζ < 3 · 106 W/(m2 K), as in this case the equilibrium boiling temperature of 100 ◦ C is not reached for all starting temperatures. 3.4. Further mechanisms The mechanisms described above are those that could be identified by our experiments (see Sec. 4) so far. However, this does not mean that they are the only ones involved in laser cleaning! There are at least three more mechanisms that may play a role in the process and have been suggested: particle vibrations/elastic deformations, light pressure, and surface acoustic waves. In a very recent publication [45] Arnold et al. suggested that elastic deformation of the particles might lead to particle removal if the particles were stimulated in resonance with their eigenfrequency. Vereecke et al. [61] suggested that at grazing incidence of the laser pulse the light pressure might roll the particles over the surface. Also surface acoustic waves may facilitate DLC [62]. Although in our opinion so far there is no conclusive experimental evidence for those mechanisms, they might well be found in laser cleaning as well as others unknown today.

Laser cleaning of silicon wafers: Prospects and problems

325

4. EFFICIENCY MEASUREMENTS

4.1. Steam laser cleaning For the SLC we used [17] the same experimental setup as in the DLC experiments, but supplemented it with a steam providing unit [11]: a controlled flow of filtered, pressurized air was directed through a reservoir of a water/IPA mixture (90% water) heated to 330 K. Then the steam/air mixture was directed to the sample via a nozzle at a distance of 1.5 cm from the area to be cleaned. The IPA’s role was to improve the wetting of the steam condensed onto the silicon wafer leading to the formation of a liquid film. We estimated the film thickness using ellipsometric measurements to be about 200-400 nm. The cleaning process was found [35] to be statistical in a way that the number Nr of the remaining particles after n cleaning steps is given by Nr /N0 = (1 − p)n as function of the single shot cleaning efficiency p and the original particle number N0 . Therefore, in Fig. 8 we plotted only the cleaning efficiencies after 20 cleaning steps (steam and laser pulse) for the sake of clarity. This figure shows the dependence of the cleaning efficiency on the applied laser fluence for particles of different sizes (60 nm-800 nm), materials (PS, SiO2 , Al2 O3 ) and geometries (spherical PS, SiO2 and arbitrarily shaped Al2 O3 ). Some important results can be obtained from this graph. A predominant feature is the existence of a universal cleaning threshold for all particles investigated at a laser fluence of about 110 mJ/cm2 . For slightly higher laser fluences we

Figure 8. Cleaning efficiency for various colloidal particles in SLC with λ=532 nm, FWHM=7 ns. The cleaning threshold is found to be the same for all particles.

326

M. Mosbacher et al.

observed a very steep increase in the cleaning efficiency, and values above 90% after 20 cleaning steps (laser pulse plus steam jet) are reached well below the melting threshold of a bare silicon substrate, even for particles as small as 60 nm. The threshold cleaning fluence of 110 mJ/cm2 found in SLC is larger than the threshold fluence of 80 mJ/cm2 determined for bubble nucleation in bulk water (see. Sec. 3.3). This needs further experimental investigation, however there are two possible explanations. First, the bubble nucleation process may depend on the film thickness, and hence there would be differences in the process dynamics when comparing bulk liquid to liquid films. Second, it is not clear so far whether the mere nucleation of bubbles is sufficient for an efficient cleaning or whether more laser fluence is needed to create larger or faster growing bubbles. 4.2. Dry laser cleaning in ambient conditions A first step in the investigation of DLC is the study of particle removal in ambient conditions (relative humidity 30-40%) with a ns Nd:YAG-laser. This environment represents the conditions that may be found in a possible future application of the process. A flow of pressurized, filtered air was used to blow away the removed particles and to prevent their redeposition. In Figure 9 the thresholds in applied laser fluence for the removal of PS particles are plotted as a function of the particle size. First we would like to draw the reader’s attention to two very important thresholds in the laser cleaning process. One of them is the threshold for the onset of melting of the bare substrate. As DLC is aimed for an industrial application, any change

Figure 9. Thresholds in the applied laser fluence for particle removal in DLC in ambient air. Particles smaller in diameter than 110 nm could not be removed.

Laser cleaning of silicon wafers: Prospects and problems

327

of the structure of the silicon wafer, i.e. the silicon substrate and the native oxide layer of specified thickness, as induced by melting has to be strictly avoided. From experiments [63] this melting threshold is known to be about 280 mJ/cm2 for our laser parameters, which therefore represents the upper limit of applicable laser fluence. The second threshold, also indicated in Figure 9, is the cleaning threshold of the SLC process as discussed in Section 4.1. In order to obtain information on the dependence of the cleaning threshold on the particle size we investigated many different particle diameters in the range of 110-2000 nm. It turned out that only by the investigation of this large variety of particle sizes some main characteristics of the dependence of the cleaning threshold on the particle diameter could be revealed. At first sight the shape of the curve roughly follows a 1/rk -trend, r denoting the particle radius and 1< k <2. This monotonic behavior was predicted by the first DLC models, and in fact already the first publications on DLC reported that smaller particles were harder to remove than larger ones due to the nature of the adhesion forces [6– 8]. Taking a closer look, however, one discovers an oscillating behavior of the threshold fluences which we attribute to optical resonances as discussed in Section 3.2.3. This is illustrated in the graph by the line connecting the data points. However, it should be pointed out that this line is just a guide to the eye and does not describe the field enhancement efficiency as function of the particle diameter. The number of particle sizes used in our experiments is not sufficient to resolve this dependency. In DLC using sub-ns pulses the removal of a particle is always accompanied by the formation of a hole [38], i.e. the hole formation threshold is identical with cleaning threshold. From this, one can conclude that here the dominant cleaning mechanism is local substrate ablation. Against the background of field enhancement as the origin of surface damage, it is a natural consequence also for nanosecond laser pulses to determine not only the cleaning threshold fluences in DLC, but also the local melting/ablation thresholds. The latter ones, instead of the melting threshold of the bare silicon surface, represent the true upper limit for the applicable laser fluence and are by their nature particle dependent. For its determination we made use of the Gaussian spatial beam profile of our cleaning laser. Due to this profile a spatial variation of the position in the cleaned area corresponds to a variation in the locally applied laser fluence. In a post process analysis we investigated the cleaned areas of our samples with an atomic force microscope (AFM, Digital Instruments). By imaging damage sites such as displayed in Figure 5 at different locations in the cleaned areas and especially at their borders, corresponding to the cleaning threshold fluence, we determined the damage threshold for each particle size. For all particles investigated the cleaning threshold was identical with the damage threshold. Damage-free DLC was not possible by applying the laser parameters we used! The AFM images of damage sites contain even more information on the particle removal mechanism as they reveal quantitative topographical information. All the investigated damage sites showed the same features: a “trench” surrounding a

328

M. Mosbacher et al.

Figure 10. Trench depth at the damage sites as a function of the particle size when the particles were removed in DLC applying the threshold cleaning fluence. For particles larger than 250 nm in diameter the depth increases strongly with the particle size, for smaller particles it remains almost constant.

central “hillock”. Generally speaking, at high laser fluences the hillock was lower and the trench deeper, whereas for low laser fluences a hillock was detectable but the trench almost disappeared. In Figure 10 we plotted for the investigated particles the mean trench depths for damages that occurred at the cleaning/damage threshold in a double logarithmic graph. This plot clearly shows two regimes: for particles smaller than about 250 nm in diameter the depth remains almost constant at about 1 nm. For larger particles we found a strong increase in the trench depth and the volume of the hillock was smaller than that of the trench, i.e., ablation of substrate material had taken place. From this observation we conclude that even for DLC using ns pulses local ablation of the substrate plays a role as a cleaning mechanism for “large” particles where the field enhancement is high and thus provides fluences high enough for ablation. For smaller particles field enhancement probably causes local melting, but no ablation at the threshold cleaning fluence. 4.3. Dry laser cleaning in high vacuum As already discussed in Section 3 there exists experimental evidence that “dry” laser cleaning in ambient air is facilitated by adsorbed humidity from the surrounding atmosphere. In order to minimize this effect and hence to exclude explosive evaporation of adsorbed moisture as a cleaning mechanism, we repeated the experiments reported above in high vacuum (HV, 10−6 mbar). The samples were allowed to degas approximately for 10 hours at HV before cleaning.

Laser cleaning of silicon wafers: Prospects and problems

329

Figure 11. Comparison of the cleaning thresholds in ambient air and high vacuum. In vacuum the thresholds are distinctly higher.

The results and the comparison with the ambient values are presented in Figure 11. In vacuum the laser fluences necessary for particle removal are higher for particles smaller than about 800 nm in diameter. For larger particles no difference in the threshold was detected between ambient conditions and HV. We attribute this to the predominance of different cleaning mechanisms for different particle sizes. Cleaning in ambient conditions is facilitated by explosive evaporation of adsorbed moisture from the air. When exposed to vacuum for several hours the amount of moisture at the particle-surface contact is significantly decreased, the contribution of explosive evaporation to the cleaning forces decreases, and consequently the cleaning threshold increases. This argument should also be valid for particles larger than 800 nm in diameter. However, as discussed above, for large particles local ablation induced by field enhancement appears to strongly contribute to the cleaning process. And, of course, this is the case as well in ambient air as in vacuum and hence no large difference in the cleaning thresholds for the two different environments is detected.

5. PROSPECTS AND PROBLEMS

Against the backdrop of the results presented above we now discuss the prospects as well as the problems of using laser cleaning as an industrial cleaning technique. In this context we would like to highlight three main aspects: the systematic determination of cleaning thresholds, the role of different cleaning mechanisms, and the consequences for the applicability of theoretical models proposed so far.

330

M. Mosbacher et al.

5.1. Cleaning thresholds and process efficiency A systematic determination of cleaning thresholds in both DLC and SLC should provide key information for the application of laser cleaning, as it allows to predict the minimum particle size that can be removed and to judge which of the two processes DLC or SLC is more efficient. On the basis of our measurements this comparison can be done for the first time and for a large size interval of particles. Perhaps the most striking difference in the two laser cleaning methods is the dependence of the cleaning threshold fluence on particle size. Whereas in SLC this threshold appeared to be universal, i.e. size- and material-independent for the investigated particles, in DLC we found in agreement with other authors a size dependent threshold, with smaller particles being harder to remove than larger ones. From this it is obvious that SLC is a more efficient method for small particles, i.e. for particles smaller than about 400 nm in diameter (for particles larger than 400 nm see below) which is the most interesting size regarding the cleaning of bare silicon wafers in the semiconductor industry. In addition, SLC is superior to DLC in the minimum particle size that could be cleaned from silicon wafers. Recalling that the current minimum line width in ICs is 130 nm, which means that particles of about 60-70 nm in size have to be removed, this is a key information on the quality of a cleaning method. The lower size limit of particles that could be removed by DLC was found to be 110 nm, compared to 60 nm and an efficiency above 90% in SLC. Summarizing the above, SLC is superior to DLC due to three crucial characteristics: its universal cleaning threshold, its lower threshold fluences for the relevant particle sizes, and its capability of removing sub 100 nm-particles. 5.2. Consequences of cleaning mechanisms involved Although in DLC no particles smaller than 100 nm could be removed, at a first glance it seems to be the more appropriate method for larger particles as its cleaning thresholds are distinctly lower than the universal SLC threshold. However, for a judgement of the perspectives of SLC and DLC it is not sufficient to solely determine and compare cleaning efficiency and laser cleaning threshold fluence. On the contrary, as our studies above show very clearly, this comparison must be put into perspective by taking a closer look at the cleaning mechanisms involved. The most important physical process not taken into account in traditional investigations and only recently [34, 35, 38– 40] studied is the local substrate ablation due to the enhancement of the laser intensity in the near field of the particles. The first, and most obvious, consequence of field enhancement is a locally increased laser fluence underneath the particle, and hence a decrease in the incident laser fluence necessary for particle removal. At a first sight this looks like a positive effect, but obviously a locally enhanced laser intensity drastically lowers the threshold for surface damage, and indeed we did observe surface damage caused either by melting (small particles) or local substrate ablation (large particles)

Laser cleaning of silicon wafers: Prospects and problems

331

whenever a particle was removed in DLC. This means that damage-free DLC was impossible with our laser parameters. In fact, the local substrate ablation combined with a momentum transfer to the particles was found to be the dominating cleaning mechanism with large size parameters and sub-nanosecond laser cleaning [40]. A second consequence of field enhancement also argues against a technological application of DLC. The DLC results obtained both in ambient conditions and in HV confirm a general trend already obtained by other authors [6, 7, 19, 31, 33, 64] and predicted by their models: cleaning thresholds for smaller particles tend to be higher than for larger ones. Yet this is not a strict rule. In contrast to the above cited experiments we used a large variety of particle diameters ranging from 110 nm to 4100 nm and could show that the dependence of the cleaning threshold as a function of the particle diameter was non-monotonous as a consequence of the optical resonances in the near field of the particles. This non-monotonous behaviour in DLC makes it difficult to apply the correct cleaning fluence for the removal of a specific particle size, unlike the universal threshold in SLC. It should be pointed out that field enhancement is an intrinsic physical effect that cannot be avoided in laser cleaning. However, it can be minimized by choosing suitable wavelengths to reduce the size parameter or by carrying it out in an environment with a refractive index close to that of the particles. 5.3. Models Models which describe the laser cleaning process accurately would be an important tool in the real world application, as they could predict the optimum cleaning conditions for various substrates, particles, lasers and cleaning environments. However, such models must inevitably incorporate all the knowledge gained regarding the physical processes behind both SLC and DLC. So far in DLC the particle removal has always been solely ascribed to the thermal expansion of the substrate. Field enhancement was taken into account only as increased laser fluence [19, 34], but not as the origin of an additional cleaning mechanism via local ablation. The third cleaning mechanism, evaporation of adsorbed ambient moisture, is not incorporated into any model published so far. The latter is even more important, as most of the experiments carried out to compare with laser cleaning models were performed in ambient conditions [22, 24, 26– 28, 30, 64]. Only in [19] the experiments were conducted in vacuum, and Vereecke et al. [31] varied the relative humidity and reported a higher cleaning efficiency at increased humidity levels. In addition to this, only recent models [44, 45] consider the elastic properties of the substrate/particle system which might have a great influence on the cleaning process as well. Future models should incorporate at least these three cleaning mechanisms and treat DLC as an interplay of all of them. Depending on the process parameters (laser wavelength, pulse duration, optical constants of the materials, etc.) their overall contribution to the cleaning will vary. It should also be pointed out that

332

M. Mosbacher et al.

the cleaning mechanisms are not independent of each other, e.g. adsorbed moisture may influence the field enhancement pattern. Compared to DLC the state of modeling the SLC process is at a rather initial stage. Two groups [28, 65] have suggested models to describe it. Yet these models rely on far-reaching assumptions in the description of the processes of laser induced bubble nucleation and growth as well as on the assumption of the temperature of the superheated water layer as growth medium. As our experiments on the last aspect show, it is impossible to transfer the results gained on rough metal films [50, 55, 56] to the water film/silicon system. Furthermore, it is not clear neither qualitatively nor quantitatively how the explosive evaporation differs between bulk water (as in our investigations) and water films (as in SLC) or even small water menisci as they can be found in ambient environment DLC. Therefore, a good deal of future research on the dynamics of laser induced bubble nucleation and the explosive evaporation in all these systems is necessary to accurately describe SLC.

6. SUMMARY

In this paper we have described our state of knowledge on the cleaning mechanisms responsible for particle removal in laser cleaning. Besides the well-known thermal expansion of the substrate and the explosive evaporation of a water film we identified local substrate ablation as another cleaning mechanism. Additionally we have shown the significant impact of the explosive evaporation of atmospheric moisture adsorbed at the particles for DLC. Local substrate ablation caused by field enhancement in the particles’ near field not only causes particle removal in DLC, but inevitably also causes substrate damage. Furthermore a damage-free DLC process was not possible with the laser parameters we used in our experiments. Steam laser cleaning, on the contrary, proved to be superior to the DLC process due to its higher efficiency, universal cleaning threshold and its capability to remove much smaller particles. These findings argue for the application of SLC in wafer cleaning and underline the need for further research on the physics of both DLC and SLC as only this knowledge will ensure a successful implementation of the technique in future industrial applications. Acknowledgements We thank Prof. B. Luk’yanchuk (DSI, Singapore) and Dr. Nikita Arnold (JohannesKepler-University, Linz, Austria) for useful discussions. The authors would also like to thank Dr. Bernd-Uwe Runge, Christof Bartels, Johannes Graf, Florian Lang, and Michael Olapinski (all of University of Konstanz) for constructive discussions of the findings of our experiments. Financial support by the EU TMR project “Laser Cleaning” (No. ERBFMRXCT98 0188) and the Konstanz Center for Modern

Laser cleaning of silicon wafers: Prospects and problems

333

Optics is gratefully acknowledged. Wacker Siltronic supplied the industrial silicon wafers.

REFERENCES 1. S.A. Hoenig, in: Particles on Surfaces 1: Detection, Adhesion and Removal, K.L. Mittal (Ed.), p. 3, Plenum, New York (1988). 2. T. Hattori, Solid State Technol., 8 (July 1990). 3. R. Kohli, in: Particles on Surfaces 7: Detection, Adhesion and Removal, K.L. Mittal (Ed.), VSP, Utrecht, in press. 4. R. DeJule, Semiconductor Intl., 65-68 (August 1998). 5. International Technology Roadmap for Semiconductors, Sematech International (2000). 6. W. Zapka, W. Ziemlich and A.C. Tam, Appl. Phys. Lett. 58, 2217 (1991). 7. A.C. Tam, W.P. Leung, W. Zapka and W. Ziemlich, J. Appl. Phys. 71, 3515 (1992). 8. A.C. Engelsberg, Mater. Res. Soc. Symp. Proc. 315, 255 (1993). 9. I.V. Beklemyshev, V.V. Makarov, I.I. Makhonin, Y.N. Petrov, A.M. Prokhorov and V.I. Pustovoi, JETP Lett. 46, 347-350 (1987). 10. K. Imen, J. Lee and S.D. Allen, Appl. Phys. Lett. 58, 203-205 (1991). 11. H.K. Park, C.P. Grigoropoulos, W.P. Leung and A.C. Tam, IEEE Trans. CPMT A, 17, 631 (1994). 12. J.B. Heroux, S. Boughaba, I. Ressejac, E. Sacher and M. Meunier, J. Appl. Phys. 79, 2857 (1996). 13. S. Boughaba, X. Wu, E. Sacher and M. Meunier, J. Adhesion, 61, 293-307 (1997). 14. A.C. Tam, H.K. Park and C.P. Grigoropoulos, Appl. Surf. Sci. 127-129, 721-725 (1998). 15. M. She, D. Kim and C.P. Grigoropoulos, J. Appl. Phys. 86, 6519 (1999). 16. M. Mosbacher, N. Chaoui, J. Siegel, V. Dobler, J. Solis, J. Boneberg, C.N. Afonso and P. Leiderer, Appl. Phys. A 69[Supp.], 331 (1999). 17. M. Mosbacher, V. Dobler, J. Boneberg and P. Leiderer, Appl. Phys. A 70, 669 (2000). 18. J.M. Lee, K.G. Watkins and W.M. Steen, Appl. Phys. A 71, 671-674 (2000). 19. Y.F. Lu, Y.W. Zheng and W.D. Song, J. Appl. Phys. 87, 1584 (2000). 20. D. Bäuerle, Laser Processing and Chemistry, 2nd Ed., Springer (1996). 21. J.D. Kelley and F.E. Hovis, Microelectronic Engg. 20, 159 (1993). 22. Y.F. Lu, W.D. Song, M.H. Hong, B.S. Teo, T.C. Chong and T.S. Low, J. Appl. Phys. 80, 499 (1996). 23. Y.F. Lu, W.D. Song, B.W. Ang, M.H. Hong, D.S.H. Chan and T.S. Low, Appl. Phys. A 65, 9 (1997). 24. Y.F. Lu, W.D. Song, K.D. Ye, M.H. Hong, D.M. Liu, D.S.H. Chan and T.S. Low, Appl. Surf. Sci. 120, 317 (1997). 25. Y.F. Lu, W.D. Song, Y. Zhang and T.S. Low, Proc. SPIE 3550, 7 (1998). 26. Y.F. Lu, W.D. Song and T.S. Low, Mater. Chem. Phys. 54, 181 (1998). 27. Y.F. Lu, W.D. Song, M.H. Hong, D.S.H. Chan and T.S. Low, Proc. SPIE 3097, 352 (1997). 28. Y.F. Lu, Y. Zhang, Y.H. Wan and W.D. Song, Appl. Surf. Sci. 138-139, 140 (1999). 29. Y.F. Lu, Y.W. Zheng and W.D. Song, Appl. Phys. A 68, 569 (1999). 30. W.D. Song, Y.F. Lu, K.D. Ye, C.K. Tee, M.H. Hong, D.M. Liu and T.S. Low, Proc. SPIE 3184, 158 (1997). 31. G. Vereecke, E. Röhr and M.M. Heyns, J. Appl. Phys. 85, 3837 (1999). 32. X. Wu, E. Sacher and M. Meunier, J. Appl. Phys. 87, 3618 (2000). 33. T. Fourrier, G. Schrems, T. Mühlberger, J. Heitz, N. Arnold, D. Bäuerle, M. Mosbacher, J. Boneberg and P. Leiderer, Appl. Phys. A 72, 1-6 (2001). 34. B.S. Lukyanchuk, Y.W. Zheng and Y.F. Lu, Proc. SPIE 4065, 576-587 (2000).

334

M. Mosbacher et al.

35. P. Leiderer, J. Boneberg, V. Dobler, M. Mosbacher, H.-J. Münzer, N. Chaoui, J. Siegel, J. Solis, C.N. Afonso, T. Fourrier, G. Schrems and D. Bäuerle, Proc. SPIE 4065, 249-259 (2000). 36. P. Leiderer, J. Boneberg, M. Mosbacher, A. Schilling and O. Yavas, Proc. SPIE 3274, 68 (1998). 37. M. Mosbacher, presentation at the 5th International Conference on Laser Ablation COLA 99, July 1999, Göttingen, Germany. 38. M. Mosbacher, H.-J. Münzer, J. Zimmermann, J. Solis, J. Boneberg and P. Leiderer, Appl. Phys. A 72, 41 (2001). 39. H.-J. Münzer, M. Mosbacher, M. Bertsch, J. Zimmermann, P. Leiderer and J. Boneberg, J. Microscopy 202, 129 (2001). 40. M. Mosbacher, M. Bertsch, H.-J. Münzer, V. Dobler, B.-U. Runge, D. Bäuerle, J. Boneberg and P. Leiderer, Proc. SPIE (in print). 41. M. Mosbacher, H.-J. Münzer, M. Bertsch, V. Dobler, N. Chaoui, J. Siegel, R. Oltra, D. Bäuerle, J. Boneberg and P. Leiderer, in Particles on Surfaces 7: Detection, Adhesion and Removal, K.L. Mittal (Ed.), VSP, Utrecht, in press. 42. V. Dobler, PhD thesis, University of Konstanz, Germany (2002). 43. V. Dobler, R. Oltra, J.P. Boquillon, M. Mosbacher, J. Boneberg and P. Leiderer, Appl. Phys. A 69, 335 (1999). 44. B.S. Lukyanchuk, Y.W. Zheng and Y.F. Lu, Proc. SPIE 4423, 115 (2001). 45. N. Arnold, G. Schrems, T. Mühlberger, M. Bertsch, M. Mosbacher, P. Leiderer and D. Bäuerle, Proc. SPIE (in print). 46. S.V. Buntsents, S.G. Dmitriev and O.G. Shagimuratov, Phys. Solid Stat., 38, 552-557 (1996). 47. M. Vicanek, A. Rosch, F. Piron and G. Simon, Appl. Phys. A, 59 407-412 (1994). 48. G. Mie, Ann. Physik 4, 377-445 (1908). 49. P.W. Barber and S.C. Hill, Light Scattering by Particles: Computational Methods, World Scientific Publishing, Singapore (1989). 50. O. Yavas, P. Leiderer, H.K. Park, C.P. Grigoropoulos, C.C. Poon, W.P. Leung, N. Do and A.C. Tam, Phys. Rev. Lett. 70, 1830 (1993). 51. H.K. Park, C.P. Grigoropoulos, C.C. Poon, A.C. Tam, O. Yavas and P. Leiderer, Proc. SPIE 2498, 32-40 (1994). 52. O. Yavas, P. Leiderer, H.K. Park, C.P. Grigoropoulos, C.C. Poon, W.P. Leung, N. Do and A.C. Tam, Appl. Phys. A 58, 407-415 (1994). 53. H.K. Park and C.P. Grigoropoulos, C.C. Poon and A.C. Tam, Appl. Phys. Lett. 68, 596-598 (1996). 54. O. Yavas, A. Schilling, J. Bischof, J. Boneberg and P. Leiderer, Laser Phys. 7, 343-348 (1997). 55. H.K. Park, D. Kim, C.P. Grigoropoulos, J. Appl. Phys. 80, 4072-4081 (1996). 56. O. Yavas, A. Schilling, J. Bischof, J. Boneberg and P. Leiderer, Appl. Phys. A 64, 331-339 (1997). 57. C.T. Avesidian, J. Phys. Chem. Ref. Data 14, 695-729 (1985). 58. A.C. Tam, W.P. Leung and W. Zapka, in: Particles on Surfaces: Detection, Adhesion and Removal, K.L. Mittal (Ed.), pp. 405-418, Marcel Dekker, New York (1995). 59. E.T. Swartz and R.O. Pohl, Rev. Mod. Phys. 61, 605-668 (1989). 60. G. Cerbe and H.J. Hoffmann, Einführung in die Thermodynamik, Munich (1999). 61. G. Vereecke, E. Röhr and M.M. Heyns, Appl. Surf. Sci. 157, 67 (2000). 62. A.A. Kolomenskii, H.A. Schuessler, V.G. Mikhalevich and A.A. Maznev, J. Appl. Phys. 84, 24042410 (1998). 63. J. Boneberg, Metallische Dünnfilmschmelzen von Halbleiterschichten nach ns-Laser-Annealing, PhD thesis, University of Konstanz, Germany (1996). 64. D.R. Halfpenny and D.M. Kane, J. Appl. Phys. 86, 6641 (1999). 65. X. Wu, E. Sacher and M. Meunier, J. Appl. Phys. 87, 3618 (2000).

Surface Contamination and Cleaning, Vol. 1, pp. 335–343 Ed. K.L. Mittal © VSP 2003

Particle removal using resonant laser detachment KEVIN KEARNEY and PETER HAMMOND∗ Lightforce Technology, Inc., 125 Tech Park Drive, Rochester, NY 14623

Abstract—A new photonic cleaning process that minimizes exposure of the substrate is introduced. The concepts of the Resonant Laser Detachment (RLD) are described. The RLD process uses a laser light source with intensity modulation to remove sub-micrometer contaminant particles from a substrate. Unlike other laser removal methods, which eject the particle from the surface with a single high-intensity laser pulse, RLD uses a continuous series of low-intensity laser pulses. The timing and shape of these laser pulses are tuned to exploit the kinematic properties of the particle-surface system and laser-material interactions. Theoretical analysis suggests a correlation between system resonant frequency and the particle separation mechanism. This technique results in an efficient particle removal mechanism that minimizes stress and heat loading to the underlying substrate. Keywords: Particle detachment; laser cleaning; resonance.

1. INTRODUCTION

The trends in the manufacturing of Integrated Circuits (IC’s) and Flat Panel Displays (FPD’s) and in other related manufacturing industries and the requirements for environmentally safe cleaning technologies are driving the need to investigate advanced dry photonic based cleaning methods. Contamination remaining on the substrate after a processing step may create device defects, rendering final end products useless. It is recognized that submicrometer particles can cause major defects that can result in defective components and lower production yield. With the trend towards reduced feature size in IC’s, contamination will have an even greater impact on manufacturing yield. The need is to increase yield, thus lowering production cost and increasing manufacturing competitiveness. This investigation focuses on the use of a light based cleaning technology for the purpose of developing a commercially viable critical cleaning process.

∗

To whom all correspondence should be addressed. Phone: 585-292-5610, Fax: 585-427-8422, E-mail: [email protected]

336

K. Kearney and P. Hammond

2. PARTICLE ADHESION MECHANISMS

Strong adhesion forces exist between particles and surfaces due to van der Waals, electrostatic, and capillary attraction mechanisms. Bowling [1] has discussed these forces in detail, while a concise summary can be found in Tam et al. [2]. van der Waals forces are comprised of London-dispersion force, dipole-dipole and dipole-induced dipole interactions. Capillary forces arise when atmospheric moisture condenses in the gap between a particle and a substrate, and are a function of particle radius and liquid surface tension. Coulomb electrostatic forces originate from the electrostatic double-layer formed between the particle and the substrate. Tam et al. [2] note that as particle size decreases, the relative adhesion force increases dramatically. Particle mass (m) decreases as the cube of the diameter, while adhesion forces (F) decrease directly with diameter; hence, the acceleration (a=F/m) required to detach a particle from a surface scales inversely with the square of the diameter. As will be shown later, adhesion forces on a micrometersize particle greatly exceed gravitational forces. As noted by Bowling [1], the contact area of a particle in contact with a surface is quite small, resulting in tremendous pressures at the particle-substrate contact point. For a typical 1-µm particle, force per unit area is estimated to be 10.9 N/m2 which is enough to deform the particle and increase the particle to surface contact area. Since the force of adhesion depends on the contact area, this effect will further strengthen the binding of the particle to the surface. 2.1. van der Waals forces For particles in the semiconductor-processing environment, van der Waals forces predominate. Visser’s [3] treatment of van der Waals forces between a particle of radius r and a flat surface at a distance z away from the surface shows that,

æ h öd FvdW = ç ÷ 2 , è 6π ø z

(1)

where h is the material dependent Lifshitz-van der Waals constant. Notice that for a given separation z, the van der Waals forces scale with the particle diameter d, while the mass of the particle scales as (d/2)3. Hence, the force per unit mass (acceleration) scales with particle size as 1/d2: smaller particles require greater acceleration than larger particles to remove them from the substrate. From Eq. (1), we see that the van der Waals forces on a sub-micrometer particle can greatly exceed the gravitational force on the same particle resting on the surface of the substrate. As an example, consider a 1 µm diameter particle of silicon resting on a silicon substrate. The gravitational force on the particle is approximately 1.5x10–15 N, while the van der Waals forces at a distance of 0.4 nm from the substrate (the measured equilibrium separation distance) are 7x10–8 N, a nearly 50x107 times greater.

Particle removal using resonant laser detachment

337

Figure 1. The particle contact area increases with the time the particle remains on the surface due to deformation forces, thus increasing the total particle adhesion force.

2.2. Particle deformation One consequence of the strong adhesion force is particle deformation. The adhesion forces distort or flatten the particle contact thus increasing the particle contact area. Again following Visser [3], the force due to particle deformation is a function of the particle contact radius r,

æ h Fdeformation = ç è 6π

2

ör ÷ 3 øz

(2)

Krishnan et al. [4] have shown that as a particle is allowed to rest on a substrate the increase in contact area due to deformation begins immediately upon contact and can increase by a factor of 50% within ten minutes. As shown in Figure 1, for a substrate being contaminated during the manufacturing process it is apparent that if particles are not removed shortly after being deposited they will become increasingly more difficult to remove and may ultimately be impossible to remove. In production, as the substrate is moved through the manufacturing process particles not removed shortly after being deposited may cross-contaminate subsequent processes or ultimately accumulate on the substrate, therefore contributing to potential product defects. Thus, it is important to remove particles immediately after they land on the substrate. 2.3. Particle-surface potential energy As a particle comes in close contact with the substrate, van der Waals attractive forces begin to be offset by the quantum mechanical repulsion force associated

338

K. Kearney and P. Hammond

Figure 2. A particle resting on a substrate surface is trapped in a particle-surface potential well. The potential energy of a particle at a given separation distance from a surface is shown.

with the orbital electrons of the particle and surface molecules. The two forces are estimated to be equal at a distance of 0.4 nm from the surface [1]. This distance, indicated in Figure 2, represents the minimum of the potential well in which the contaminant particle is trapped. Although the exact shape of the potential well is difficult to calculate precisely, it will have the general functional form shown in Figure 2. A particle bound to a surface may thus be considered an oscillator. As discussed below, the RLD technique exploits this kinematic behavior by exciting the resonant frequencies of the oscillator. 3. EXISTING LASER PARTICLE REMOVAL TECHNIQUES

Laser particle removal methods may be differentiated in a number of ways. The first distinction is between ablative vs. non-ablative methods. Ablation is an energetic phase transformation from a solid to a gaseous state. At high laser intensities a thin surface layer of the substrate can be removed – carrying any surface contaminants away with it. As a cleaning method, this process may be considered analogous to chemical etching, in that a thin surface layer of material is removed. Like chemical etching, laser ablation tends to “micro-roughen” the surface. In typical precision cleaning applications, micro-roughening is a very serious concern for every cleaning method – wet or dry. In fact, any uncontrolled changes in surface morphology, chemistry or physical properties are usually undesirable in these applications. Non-ablative particle removal methods use laser-generated particle-substrate interactions to break the physical and chemical bonds holding the particle to the

Particle removal using resonant laser detachment

339

Figure 3. Laser light sources utilize photothermal energy for a direct energy transfer to the particle and surface. The net effect is a rapid thermal expansion of the particle and substrate surface, dislodging the particle from the surface.

substrate. These methods may be further differentiated by the presence or absence of an energy transfer medium used to mediate the substrate-particle interaction. In direct exposure methods (no energy transfer medium), a UV laser is used to directly irradiate both the particle and an area of the surrounding substrate [5, 6]. The method acts to both photochemically break surface bonds [7, 8] and to induce a rapid thermal expansion of the particle and/or a thin substrate layer, which forcefully ejects the particle from the surface [5, 9]. The alternative method involving an energy transfer medium uses a several micrometer thick layer of a liquid on the substrate [10-12]. This liquid layer is then directly or indirectly rapidly heated by a laser, causing explosive evaporation that removes the transfer medium and trapped surface particles. 4. RADIATION FORCES

4.1. Photophoretic force The photophoretic force has two aspects: radiation force due to photon momentum coupling to matter, and radiometric force arising from the interaction of a laser-heated particle and an ambient (liquid or gaseous) medium. The radiation force operates independent of gas pressure, while the radiometric force increases from zero in vacuum to a maximum when the ambient gas molecules mean free path (mfp) is approximately equal to the particle diameter while decreasing at higher gas pressures (shorter mfp) [13]. For spherical, dielectric particles, the radiation force due to photon momentum has been found to have two orthogonal components [14, 15]. The first component

340

K. Kearney and P. Hammond

(axial) is directed along the laser beam axis. The second component (transverse) arises due to the transverse light intensity gradient (e.g., a gaussian beam profile). The combined effects of refraction and the transverse intensity gradient are such that an unequal angular distribution of light (and hence momentum) exits at the spherical surface of the particle, producing a net effective force towards the center of the beam. This effect has been well demonstrated experimentally, and can be used to form an optical “trap” for small particles [14]. In practice, the presence of a viscous medium is required to stabilize oscillations about the center of the potential well formed by the intensity profile. For typical beam diameters and light intensities, the transverse force can be made comparable to the axial force, and both can exceed the particle’s weight by many orders of magnitude. The radiometric force has been exhaustively verified, and is commonly used in aerosol studies [13]. The radiometric force is caused by uneven heating of the particle surface, and the interaction of the heated particle with the ambient background gas. Under optimum gas pressure conditions, the radiometric force can be several orders of magnitude greater than the radiation force. 4.2. Photophoresis Photophoresis has been investigated by Periasamy [16] as a mechanism to inhibit particle attachment to a substrate. The motivation in their research was to prevent particles from initial contamination. Our proposal is to use the mechanism of photophoresis to transport particles, which have been laser-detached from the substrate, away from the surface and, ultimately, to a particle collection device such as an electrostatic filter or vacuum line. 5. THE RLD TECHNIQUE

The Resonant Laser Detachment technique utilizes a low-intensity, amplitudemodulated (or repetitively pulsed) laser light directed onto the particle and substrate. Energy absorption by the particle-surface system is maximized when the laser light (energy) is applied at a pulse rate equal or near to the natural resonant frequency of the system. Thus applying the laser light at a pulse rate equivalent to the resonant frequency of the system allows for lower laser fluences and greater energy transfer to the particle-surface system. As noted earlier, laser light interacts with the system in a number of ways. First, the incident light imparts direct momentum pressure to the particle. Second, it deposits heat energy into both the particle and the substrate. This heat, in turn, generates photothermal expansion in the particle and a photothermal surface expansion in the substrate. Depending on the laser parameters, other interactions such as bond breaking and non-equilibrium carrier (electron) energy-distributions may also be present. The RLD technique uses the electrostatic field from low energy fluence laser pulses as a driving force to induce a resonant motion of the particle. The resulting

Particle removal using resonant laser detachment

341

Figure 4. A number (n) of light sources (L) of various pulse frequencies and incident angles (Q) can be applied to achieve the desired resonant response.

forces and heat loads are kept very small. Figure 4 demonstrates the application of a single light source at a given incident angle to the substrate surface. Several incident light sources may be required to achieve an effective particle detachment. Defining the particle movement to be along a path orthogonal to the surface, we can assume that the particle is bound in a potential well with an equilibrium distance of approximately 0.4 nm above the surface. By applying a small, perturbative force to the particle it is possible to cause the particle to oscillate about this equilibrium position. For very small displacements, we can represent the potential near the minimum as a parabola – i.e., a harmonic oscillator potential. This approximation will break down as the oscillations increase in amplitude. However, it is sufficient to obtain an order-of-magnitude estimate of the resonant vibrational frequency of the particle in the potential well. For a 1 µm silicon particle resting on a silicon substrate with a contact diameter of 0.03 µm we can estimate the resonant frequency by calculating the total adhesion force (van der Waals and deformation forces, F) and applying Hooke’s law to calculate the effective spring constant (k) for the particle-surface system:

F = −kx

(3)

Again, assuming the distance x at which the particle is initially separated from the substrate to be 0.4 nm gives a resonant frequency f =

1 1 ω= 2π 2π

k ≈ 94 MHz m

(4)

The actual frequency scales with the effective particle separation distance as

1 / x . For a separation distance of 5 nm the frequency decreases to approxi-

mately 1.5 MHz. For a 0.1 µm particle the initial resonant frequency can exceed 1 GHz. In any case, these order of magnitude frequencies are within the limits of

342

K. Kearney and P. Hammond

Figure 5. The calculated resonant frequencies for 0.5 µm, 1 µm, 5 µm silicon particles on a silicon substrate as a function of particle-to-surface separation distance is shown. The curves show the dependence of resonant frequency on initial separation distance, particle size, and strongly suggest that the applied driving force must be adaptive in relation to separation distance.

available laser modulation techniques. Figure 5 shows calculated resonant frequency of 0.5 µm, 1 µm and 5 µm particles at a given surface separation distance. The curves in Figure 5 suggest that to employ RLD will require a shift or “chirping” of applied pulse frequency as the particle-to-surface separation distance increases. As the particle-to-surface distance is increased the modulation frequency is reduced to match the particle-surface system resonant frequency. In practice, the incident laser light sources (suggested in Figure 4) would be spatially and temporally coordinated with each other and repetitively chirped on a continuous basis. The RLD technique used in conjunction with a method to prevent particle reattachment (such as photophoresis, as discussed earlier) can provide an efficient particle removal process. Gettering techniques can be applied to remove the particles from the surrounding environment. 6. CONCLUSION

A unique and novel photonic cleaning technique is proposed and investigated. The RLD process induces motion of the particle relative to the surface by applying a pulsed light beam as a driving force that is tuned to the kinematic characteristics of the particle-surface system. The resonant mechanism is described using a simple harmonic oscillator model. The calculated resonant frequency of a particle at rest on a substrate is dependent on the particle and substrate material, particle size, and particle-to-surface separation distance. Not accounting for substrate and

Particle removal using resonant laser detachment

343

particle surface roughness the initial resonant frequency can exceed 1 GHz for a 0.01 µm silicon particle resting on a silicon substrate. The resonant frequency of a particle on a surface decreases as the particle-to-surface distance increases requiring an adaptive, chirped light source(s) for complete particle detachment. Particle removal is accomplished by collecting particles, after resonant detachment from the surface, using electrostatic gettering and vacuum suction to prevent reattachment. REFERENCES R.A. Bowling, J. Electrochem. Soc., 132, 2208 (1985). A.C. Tam, W.P. Leung, W. Zapka and W. Ziemlich, J. Appl. Phys., 71, 3515 (1992). J. Visser, Particulate Science Technol., 13, 172 (1995). S. Krishnan, A.A. Busnaina, D.S. Rimai and D.P. DeMejo, J. Adhesion Sci. Technol, 8, 1357 (1994). 5. T.J. Magee and C.S. Leung, in: Particles on Surfaces 3: Detection, Adhesion, and Removal, K.L. Mittal (Ed.), pp. 307-316, Plenum, New York (1991). 6. T.J. Magee, J.F. Osborne, P. Gildea and C.S. Leung, U.S. Patent 4758533 (1988). 7. A.C. Engelsberg, Mater Res Soc. Symp. Proc, 315, 255 (1993). 8. A.C. Engelsberg, Proceedings Microcontamination ‘93 Conference (1993). 9. J.D. Kelley and F.E. Hovis, Microelectronic Eng, 20, 159-170 (1993). 10. S.D. Allen, S.J. Lee and K. Imen, Optics and Photonics News, 3, No. 6, 28-30 (1992). 11. S.J. Lee, K. Imen and S.D. Allen, Microelectronic Eng, 20, 145-157 (1993). 12. W. Zapka, W. Ziemlich, W.P. Leung and A.C. Tam, Microelectronic Eng, 20, 171-183 (1993). 13. O. Preining, in: Aerosol Science, C.N. Davies (Ed.), pp. 111-135, Academic Press, London (1966). 14. A. Ashkin and J.M. Dziedzic, Appl. Phys. Lett., 19, 283 (1971). 15. A. Ashkin, Science, 210, 1081-1087 (1980). 16. R. Periasamy, U.S. Patent 5472550 (1995). 1. 2. 3. 4.

This page intentionally left blank

Surface Contamination and Cleaning, Vol. 1, pp. 345–364 Ed. K.L. Mittal © VSP 2003

The future of industrial cleaning and related public policy-making CAROLE LEBLANC∗ Toxics Use Reduction Institute, University of Massachusetts Lowell, One University Avenue, Lowell, MA 01854-2866

Abstract—In this paper, the author presents some of her findings in the pursuit of safer and greener chemical solvents for hard-surface cleaning, as well as some of the new directions that the science of cleaning may take in the next five to ten years. Specifically, innovative methods of research and development into cleaning alternatives are explained, including molecular modeling, data mining, and the use of ionic liquids. A discussion on chemical risk assessment ensues, in light of the scientific concepts of hormesis and endocrine disruption. This is followed by a comparative analysis of the European approach to policy-making, known as the Precautionary Principle and recent events pertaining to cleaning issues in the U.S. Finally, conclusions are drawn based on a hypothetical case of ‘over-cleaning’. Keywords: Data mining; designer molecules; endocrine disruption; hormesis; ionic liquids; molecular modeling; precautionary principle.

1. INTRODUCTION

“Every chemical is potentially a pharmaceutical.” A. Warhurst [1] If the physical properties of liquids were any indication, successful solvent replacement with aqueous cleaners would never be possible. The components and the behavior of the water molecule are nothing like those of a typical organic solvent used for cleaning, as evidenced by the high polarity and dipole moment of water. To illustrate, some important physical properties of the chlorinated solvent trichloroethylene (TCE) and water are compared in Table 1. Chemists often try to come as close as possible to the physical characteristics of the original solvent in formulating new cleaners for the same application. This is entirely understandable. Devising test protocols in which the solvent behaves in a certain known fashion and expecting a water-based cleaner to function in the same manner, however, is neither very realistic nor even a fair analysis of its potential performance. ∗

Phone: 978-934-3249, Fax: 978-934-3050, E-mail: [email protected]

346

C. LeBlanc

Table 1. Physical properties of a chlorinated solvent (TCE) and water Properties

TCE

Water

Chemical Formula Molecular Weight Boiling Point Density

C2HCl3 131.39 87oC 1.46 g/cm3

H2O 18.02 100oC 0.99 g/cm3

For this reason, the author described the significance of choosing the right piece of mechanical equipment in process conversions involving aqueous cleaning in her thesis, “The Search for Safer and Greener Chemical Solvents in Surface Cleaning: A Proposed Tool to Support Environmental Decision-Making”.a Nevertheless, the author is familiar with at least two situations in which an overdependence on chemical properties led investigators to a much narrower field of replacement candidates than the computer program, The Aqueous Way to Go,b the tool developed during her doctoral research, would have recommended. In both cases, an organic solvent was replaced with yet another organic solvent. As a result, only incremental improvements, if any, were made to the health and safety of workers and to the protection of the environment upon implementing the alternative cleaners. The replacement cleaners shared a variety of traits with the original solvents and the inherent dangers in using any organic and/or chlorinated compound remained the same. 2. RESEARCH METHODS AND DEVELOPMENTS

2.1. Molecular modeling The advent of ‘designer molecules’ has led to the development of products without the drudgery of comparing and matching chemicals’ physical properties at every step on the bench. The term designer molecule is used by various chemical disciplines in much the same way as the term designer gene is applied in the field of genetic engineering. Designer molecules allow scientists to visualize molecular structures and how they behave under certain conditions as well as in the presence of other molecules before they are actually synthesized. Chemical modeling softa

Completed for Erasmus University’s (Rotterdam, the Netherlands) international program in cleaner production, cleaner products, industrial ecology and sustainability, in conjunction with the Toxics Use Reduction Institute at the University of Massachusetts Lowell, leading to a Ph.D. in Sustainable Development and Management (www.eur.nl/fsw/gsem/phd), the first awarded to an American woman. b A web-based, interactive matrix, i.e., a tool designed to enhance decision making in solvent substitution (www.angelfire.com/band2/greencleaners/doctoralthesis.html).

The future of industrial cleaning and related public policy-making

347

ware reveals important molecular functions, just as engineering modeling programs reveal stress and metal fatigue patterns in, for example, aircraft parts. RasMol and CHIME are two popular computer programs for the visualization of molecules. RasMol was written by Roger Sayle of Glaxo-Wellcome and CHIME, a web browser plug-in based on RasMol, is a product of MDL Information Systems (www.mdli.com). Instructions for downloading RasMol can be found at www.umass.edu/microbio/rasmol and instructions for downloading CHIME are located at www.mdli.com. Originally intended for biological systems such as proteins and nucleic acids, both programs could be used for less complicated systems such as cleaning chemicals in the development of safer and greener alternatives. An important aspect of these programs is the researcher’s ability to manipulate structures. Chemical models can be displayed as traditional stick figures, balland-stick figures or space-filled structures. They can be controlled by threedimensional rotation, size alteration and color coding. Different parts of a model (for example, the asymptotic or active site of an enzyme) can be selected and treated separately. Coupled with the information obtained from the Surfactant Virtual Library at www.surfactants.net, this ability could be very useful in creating new surface-active agents or, for that matter, new composite materials. Scientists could conceivably formulate chemicals designed to disassociate into benign forms of their components after performing certain tasks, like cleaning. Figure 1 presents a three-dimensional model, capable of rotation, of the simplest chlorinated solvent. For comparison, Figures 2 and 3 are traditional one-dimensional, line representations of more complex surfactant formulations and an aqueous metal-cleaner in action, respectively. Much work has already been done in molecular modeling and ‘virtual compounds’ can be ordered from web-based suppliers listed at www.umass.edu /microbio/rasmol/whereget.htm or ‘synthesized’ via molecular mechanics calculations with a computational chemistry package such as Chem3D. Once a model is displayed in RasMol, it can be saved in other documents as well as printed. CHIME’s program allows for the dissemination of ‘live’ molecular models on the World Wide Web. Combined with other sources of data, RasMol and CHIME are powerful mechanisms for global communication among scientists and should be helpful for improving the understanding of chemical information among all stakeholders. Computer modeling of chemical structures also advances the cause of nanotechnology, the study and control of matter at the atomic or molecular level. The manipulation of substances at the nano-level to the precise site and at the exact moment they are needed should decrease the amounts of chemicals required to achieve a certain response, thereby decreasing the generation of wastes and the likelihood of over-exposure of humans or the environment to toxic substances. Nanotechnology may make possible the bio-inspired design of enzymatic or protein-based cleaners more cost effective.

348

C. LeBlanc

Figure 1. Ball-and-stick rendition of carbon tetrachloride (CCl4).

Figure 2. Anionic (A) and nonionic (B) surfactants.

Figure 3. Saponification of a fatty oil with a strong alkali.

2.2. Data mining of cleaning performance criteria One of the first attempts to generalize chemical behavior for solvency was the Hansen method [2]. In this method, a battery of chemical reactions is conducted and monitored in test tubes. The results are ranked visually and recorded numerically. Based on the Hansen methodology, DuPont scientists developed a proprietary computer program for the selection of semi-aqueous cleaners in the 1980s. Its

The future of industrial cleaning and related public policy-making

349

application was limited to the company’s Axarel® line of products. The Aqueous Way to Go further merges the function of a computer program with Hansen-like, actual performance criteria. While application-specific testing is still required, the results of pertinent cleaning tests from the Toxics Use Reduction Institute’s Surface Solutions Laboratory (SSL) are stored in The Aqueous Way to Go program. Additional performance information from other databases is also inserted into the program and serves to (1) further decrease the time required to identify greener chemical cleaners and (2) further increase the proficiency of the final selection. Like the molecular modeling used to accelerate chemical formulating described in the previous section, a mechanism is needed, preferably computer-based for speed and accuracy, to (1) sort through a plethora of data that may, or may not, be relevant and (2) determine what chemical interactions, if any, reveal important trends for cleaning. Table 2 contains some of the newer tools available to conduct this kind of research. Recently, algorithmic programming has been applied to advance the cause of solvent substitution. In March of 2000, three simulation programs with different algorithms were reviewed for designing greener solvents by Cabezas, Harten and Green [3]. The three simulations were: (1) the U.S. EPA’s Program for Assisting the Replacement of Industrial Solvents (Paris II), (2) the Technical University of Denmark’s software, Computer Added Molecular Design (CAMD) and (3) Molecular Knowledge Systems’ chemical design software, Synapse. The Paris II algorithm (www.tds.cc) uses chemicals from the Design Institute for Physical Property Research (DIPPR) database and “looks for potential replacement solvents whose properties are as close to the required parameters as possible”. The CAMD solvent-design algorithm (www.capec.kt.dtu.dk) operates in a five-stage process using valence (i.e., molecular charge) rules. The Synapse algorithm (www.molknow.com) “generates candidate chemical structures, which are then screened as potential solvent replacements in a four-step methodology”. Unlike these programs that focus on theoretical scenarios with data that are primarily intended for the scientific community, The Aqueous Way to Go concentrates on actual performance data of existing cleaners for the end-user community, in addition to applications development. Data mining, or knowledge discovery in databases, offers the best approach for manipulating this kind of information to arrive at meaningful insights from observed tendencies (for example, the performance of certain surfactants) in would-be relational databases. Nevertheless, it would still be possible to use any, or a combination of, the remaining computer tools described in Table 2 for research and development into greener cleaners. The web site http://surfactants.net/huibers/greenchem.html lists a number of computer programs developed for property prediction, solvent replacement studies, and reaction design as well as additional solvent substitution resources on the World Wide Web, in particular, the U.S. EPA’s Envio$en$e’s (http://es.epa.gov) links to data systems.

C. LeBlanc

350

Table 2. Examples of math-based/computer-enhanced research tools Research method or principle

Algorithmic Programming

Chaos Theory

Fuzzy Logic

Visualization Software Data Mining

Description and uses

Algorithmic, or procedural, languages are designed for solving a particular type of problem. They are called high-level languages because they are largely independent of hardware. Unlike machine or symbolic languages, they vary little between computers. The first such language was FORTRAN (FORmula TRANslation), developed for scientific calculation followed by the first commercial language, COBOL (Common Business Oriented Language). ALGOL (ALGOrithmic Language), is used primarily in mathematics and science. The latest generation of languages is an outgrowth of artificial intelligence. Also known as nonlinear dynamics, chaos theory is an interdisciplinary science that attempts to reveal structure in seemingly unpredictable dynamic systems. In a linear system, a small change produces a small and easily quantifiable systematic change, but a nonlinear system exhibits a sensitive dependence on initial conditions: small or virtually immeasurable differences in initial conditions can lead to wildly differing results. (This is sometimes called the butterfly effect, in reference to a 1979 address by meteorologist E.M. Lorenz entitled, “Predictability: Does the Flap of a Butterfly’s Wings in Brazil Set Off a Tornado in Texas?)”. Uses include the study of diverse phenomena, such as dripping faucets and population growth. Whereas, classical logic holds that everything can be expressed in binary terms: 0 or 1, black or white, yes or no; fuzzy logic allows for values between 0 and 1, shades of gray, and maybe it also allows partial membership in a set. When used with an expert system, logical inferences can be drawn from imprecise relationships. Uses include automatic optimization of household appliances by sensors, automobile subsystems and smart weapons. Similar to geographic information systems (GIS) or mapping, visualization software displays sets of interconnected data, often in animation-like format. Uses include aerospace obstacle detection and landscape evaluation. Data mining (or knowledge discovery in databases, KDD), is a new research area developing methods and systems for extracting interesting and useful information from large sets of data. Uses include commercial/financial databases, telecommunication alarm sequences and epidemiological research.

2.3. Ionic liquids as solvents Recent advances in ionic liquids show promise in improving the environmental soundness of surface cleaning. Ionic liquids are salts that exist in liquid form at ambient temperature. Like all salts, they possess a positive and a negative charge. Ionic liquids do not occur naturally and must be manufactured. While not much information has been published about them yet, Song and Roh reported the use of a room temperature ionic liquid for the immobilization, recovery and recycling of a chiral catalyst [4]. The ionic liquid used was 1-butyl-3-methylimidazolium hexafluorophosphate depicted in Figure 4.

The future of industrial cleaning and related public policy-making

351

Figure 4. Structure of ionic liquid, 1-butyl-3-methylimidazolium hexafluorophosphate.

Figure 5. The combination of an ionic liquid and supercritical CO2 to separate an organic compound from solution.

Unlike water-soluble compounds that can be extracted with water, or the removal of chemicals with high vapor pressures by distillation, ionic liquids require very high temperatures to effect separation of compounds. This post-separation of chemical products from ionic liquids may be difficult to achieve since the heat needed may cause the products to degrade. Furthermore, the energy needed to drive these reactions may be too expensive. If these problems can be solved, ionic liquids may become safer, greener solvents since they do not possess any measur-

352

C. LeBlanc

able vapor pressure and so, unlike chlorinated/organic solvents, do not evaporate to be inhaled by workers or to be emitted into the atmosphere and cause air pollution. (The dermatological consequences of exposure to ionic liquids as well as their impact on water pollution are currently unknown.) To address this separation issue, Brennecke and Beckman performed experiments using a combination of an ionic liquid and supercritical carbon dioxide at room temperature [5]. Their experiment, first reported in 1999, is diagramed in Figure 5. Both the carbon dioxide and the ionic liquid are recoverable for reuse. The same system used for the separation of naphthalene could theoretically be used for the removal of organic surface contaminants. While liquid-liquid (as opposed to liquid-CO2) extractions would still be possible, they would invariably return the system to the use of organic solvents, depending on the coefficient of partition, or to the use of water, which would be almost entirely ineffectual for the separation of most hydrocarbons. The use of various polymers, surfactants or solubilizers may enhance the extraction/cleaning process. To date, no toxicological or environmental fate studies have been published on ionic liquids. This is urgently needed before much more additional application work is done. 3. DISCUSSION

3.1. Risk assessment and policy making The preceding sections dealt with the future of industrial cleaning, in terms of chemical and scientific innovation. The subsequent sections are devoted to the underpinnings of public policies that either foster or impede these advances. No other topic is as germane to the issue of chemical discovery, manufacture and use as risk assessment. And no other aspect of risk assessment has been as overlooked as hormesis. 3.1.1. The case for hormesis Hormesis may be defined as the phenomenon observed in science that the effects of chemical exposure produced at high doses are the inverse or apparent inverse of those produced at low doses in a population [6]. The study of hormesis dates back to the German physician Paracelsus (1493-1541) and father of toxicology who coined the phrase “the dose determines the poison” [7]. It is estimated that approximately 350 studies contain evidence of hormesis. These studies involve a number of different species (fungi, protozoa, bacteria, plants and animals), cover a wide range of chemical types (alcohol and its metabolites, hydrocarbons, metals and pesticides) and exhibit varying effects (alterations in growth rates, reproduction, longevity and cancer). The hormetic effect of hydrocarbons on plant growth, where growth stimulation occurred at low doses and inhibitory effects at high doses is illustrated in Figure 6.

The future of industrial cleaning and related public policy-making

353

Figure 6. Hormetic effect of organic solvents on oat seedling growth [8].

Figure 7. Identification of hormetic zone of zinc affecting cell reproduction. Source: H. Rubin, Proc. Natl. Acad. Sci. (USA), 72, 1676 (1975).

Currently, chemical risk assessments are primarily conducted by studying highlevel exposures and extrapolating to predict safe levels. Inclusion of hormesis in risk assessments would reveal hormetic zones where the chemical/biological responses may be significant. An example is given in Figure 7. Nowhere is this phenomenon more important than in the study of cancer. Approximately twenty toxicological studies have been conducted whereby hormesis occurred, followed

354

C. LeBlanc

by the onset of cancer. All three stages of the disease – initiation, promotion and proliferation – have been linked to hormetic behavior [8]. Whether or not a chemical is said to exhibit a dose-dependent beneficial or deleterious reaction depends upon the conditions defined at the time of the exposure. For example, many chemicals used in the treatment of Acquired Immune Deficiency Syndrome (AIDS) are considered toxic under almost every other nondiseased circumstance; AIDS patients themselves need to be monitored closely for toxicity levels during treatment. Problems arise when conditions are not defined prior to a chemical’s release into the general environment, turning the biosphere, if not the patient, into a laboratory. This is descriptive of the use of most of mankind’s synthesized chemicals, including the detergents and solvents used for cleaning. The point of this discussion on hormetic behavior is that exposure may be more harmful at lower, as opposed to higher, concentrations for the same chemical, toxicity notwithstanding. In fact, hormesis contains the root word hormones, which are very powerful, biologically-active compounds that function effectively at low concentrations. This refutes the principle learned by most chemists trained before 1990 that “dilution is the solution to the problem” and demonstrates the importance of identifying potential chemical hazards before they enter the biosphere, to avoid the difficulty of separating minute amounts of powerful toxins (for example, dioxin) from various waste streams. 3.1.2. Surfactants and endocrine disruption Surfactants are surface-active chemicals that are very important to the cleaning process. Their concentrations in aqueous cleaners are deceptively low (< 10%), given that they are the power horses of the cleaner’s formulation. It should, therefore, come as no surprise that some of these surface-active agents may exhibit the kinds of effects described above at very low concentrations. The proven health hazards associated with organo-chlorinated cleaning solvents were described by the author [9] while the suspected health hazards involving some surfactants in some aqueous/semi-aqueous cleaners, acting as endocrine disruptors, were only briefly mentioned. More investigative work needs to be done. The endocrine (or hormonal) system is made of glands throughout the body that synthesize and secrete hormones into the bloodstream and various receptor sites in target tissues that recognize and respond to hormones, especially the sex organs. The endocrine system controls a complex interplay between the sex hormones of the oestrogens and androgens, and other hormones, such as those of the thyroid system. The immune and nervous systems are also affected by hormonal regulation. In general, hormonal signaling is more long-lived than neural transmission. It is precisely because of these systems’ complexities that it is extremely difficult to accurately predict the behavior of a single chemical compound or its metabolites on the body’s organs.

The future of industrial cleaning and related public policy-making

355

Figure 8. The structure of oestradiol.

Oestrogens such as oestradiol, pictured in Figure 8, influence the development and maintenance of female sex characteristics, and the maturation and function of the sex organs. Chemicals that can imitate an oestrogen are known as oestrogenic chemicals. Androgens such as testosterone serve a similar purpose in males. Chemicals can disrupt the endocrine system in several ways, with the degree of disruption being influenced by timing, especially with regard to the stages of the recipient organ’s development and the age of the organism. The main mechanisms include (1) binding or activating the oestrogen receptor or other receptors, (2) modifying the production or metabolism of natural hormones and (3) modifying the number of hormone receptors. Besides endocrine disrupting, other terms used to describe this chemical behavior include: xenoestrogenic, oestrogenic (estrogenic) and hormone mimicking. Synthetic substances implicated as oestrogenic include the alklyphenols (and their derivatives) used in industrial detergents for wool washing and metal finishing, various laboratory detergents, including Triton X-100 and some liquid laundry detergents. The alkylphenols, nonylphenol and octylphenol are mainly used to make alkylphenol ethoxylate (APE) surfactants. Alkylphenols were first thought to be oestrogenic in the 1930s [10] and more evidence of such effects was published in the 1970s [11, 12]. However, it was not until 1991 that publication of the effects of nonylphenol on cultured human breast cells led to human health concerns [13]. As reported by Warhurst [1], research has shown that alkylphenols increase the growth of these cells 1000 to 10000 times greater than the natural oestradiol levels required to produce the same growth. Oestrogenic effects have also been shown on rainbow trout hepatocytes, chicken embryo fibroblasts and a mouse oestrogen receptor [14, 15]. Oestrogenic effects are present at tissue concentrations of 0.1 µM for octylphenol and 1 µM for nonylphenol [16]. A screen for recombinant yeast, using the human oestrogen receptor, has shown similar results [17]. Recent research has shown oestrogenic effects of nonylphenol at still lower concentrations and levels of 0.05 mg/L were sufficient to increase the number of eggs produced by minnows, as well as an increase in vitellogenin levels (this research also suggested that nonylphenol may lead to an increase in natural oestrogen levels) [18].

356

C. LeBlanc

Alkylphenol ethoxylate surfactants are not effectively degraded in sewage treatment plants or in the environment, tending instead to lose some of their ethoxylate groups and to also bio-accumulate up the food chain, as does dioxin (the resultant alkylphenols, alkylphenols with one or two ethoxylate groups and alkylphenoxy carboxylic acids, APEC, persist even longer). Alkylphenols accumulate where there is inadequate oxygen (for example, in sediments) and APEC persist in rivers and effluents (for example, in sewage). Human exposure to these chemicals can occur by (1) absorption through skin from shampoos, cosmetics, spermicidal lubricants and domestic and industrial detergents, (2) contaminated drinking water, (3) inhalation and ingestion from pesticide sprays and (4) contamination of food from fields treated with sewage sludge. Nonylphenol has been detected in human umbilical cords at concentrations up to 2 ppt, which may or may not be correlated to the predisposition of the infant’s sex as a consequence of exposure. 3.2. Status of related public policy The U.S. EPA formed the Endocrine Disruptor Screening and Testing Advisory Committee (EDSTAC) to develop recommendations for a screening program, which were finalized in August 1998. As a result, an “Endocrine Disruptor Screening Program”, was designed with a focus “on providing methods and procedures to detect and characterize the endocrine activity of pesticides, commercial chemicals and environmental contaminants”. By the agency’s own admission, however, “there currently is not enough scientific data available on most of the estimated 87,000 chemicals in commerce to allow us to evaluate all potential risks”, with the exception of some pesticides [19]. A number of papers from the research initiative of the National Science and Technology Council’s (NSTC) Committee on the Environment and Natural Resources (CENR) can be found at www.epa.gov/endocrine/pubs.html. 3.2.1. The precautionary principle: the European model The precautionary principle may be defined as the approach whereby lack of full scientific certainly is not used as a reason for postponing pollution prevention measures to prevent environmental degradation. It was first endorsed in 1987 by European environmental leaders concerned with toxic discharges into the North Sea. They reasoned that releases of chemicals should be reduced/eliminated if they were suspected to be harmful, even before there was clear scientific proof, hence the term precautionary. In a 1992 report, the John Snow Institute, Center for Environmental Health Studies, reported that a number of factors contributed to this scientific uncertainty [20]. These factors are listed in Table 3. That same year, the United Nations Conference on Environment and Development (UNCED) adopted Principle 15, which states that “where there are threats of serious irreversible damage, lack of full scientific certainty shall not be used as a reason for postponing cost-effective measures to prevent environmental degradation”. A version of this principle was also incorporated into the Cartagena Proto-

The future of industrial cleaning and related public policy-making

357

Table 3. Reasons for scientific uncertainty [20] I. II. III. IV. V. VI. VII.

The complexity of dose and exposure relationships The unknown cumulative effects of exposure The unknown effects of combined exposures to multiple chemicals The vast number of chemicals about which we have little or no health effects information Individual differences among humans in their receptivity and propensity for diseases Limitations of scientific knowledge Delays between exposure and occurrence of disease

col on Biosafety under the Convention of Biological Diversity. At the International Conference on Biotechnology in the Global Economy held at Harvard University in September 2000, a discussion was facilitated by the University’s Center for International Development (CID) that “supported efforts to better understand the institutions of precaution through which governments move from science to policy… highlighting the institutional differences among OECD (Organization for Economic Cooperation and Development) countries, sub-Saharan countries and international institutions”. The precautionary principle is important to industrial cleaning since its implementation in Europe has led to a ban of some surfactants while the U.S. continues to allow these chemicals in cleaners; many scientists believe that the safer, albeit more expensive, alcohol ethoxylate (Figure 2) is as effective and readily available as the suspect alkylphenol-ethoxylated surfactants. The computer program or tool, The Aqueous Way to Go can be used to ‘screen’ nonylphenol ethoxylate from potential solvent substitutes in much the same way. An overview of current policies covering chemical usage throughout the world, in particular suspect endocrine disrupters, is presented in Table 4. The author suggests that the proactive stance of the precautionary principle, rather than a variety of reactive policies, should form the basis of technical innovation paired to chemical regulation/trade. This is especially true in areas such as the production of genetically-modified organisms (GMOs) and the development of solvent alternatives, where the risks are so high for so many. More information on hormone disrupting chemicals and chemicals policy can be found at Tulane University’s web site, www.tmc.tulane.edu/ecme/eehome in the report, “Environmental Estrogens and Other Hormones”. 3.2.2. The effects of a recent policy change: the United States In a different, but related matter, a recent change in U.S. regulations has led to a loosening of the use of an important solvent chemical, the de-listing of acetone as a volatile organic compound (VOC). This may lead to an increase in the use of acetone and other hydrocarbons as cleaning agents, even though the quantity and quality of safer and greener alternatives continues to rise.

358

C. LeBlanc

Table 4. Overview of global policies affecting suspect endocrine disrupting chemicals [1]

European Union (EU)

United Kingdom (UK)

United Nations (UN) and NGOs

United States (US)

Chemical Industry

The European Commission (EC) published its strategy on endocrine disruption in Dec. 1999. Originally expected to include a list of 20-30 suspected endocrine disrupters, the list was postponed to April 2000. In March 1999 the EC’s Scientific Committee on Toxicity, Ecotoxicity and the Environment published the report, “Opinion on Human and Wildlife Health Effects of Endocrine Disrupting Chemicals, with Emphasis on Wildlife and on Ecotoxicology Test Methods”. The EU also published a communication on the precautionary principle in February 2000. Several endocrine disrupters are under review as part of Existing Substances process. The EU’s chemicals policy in under review as well, having been accepted that it is not currently effective enough. The UK Government published its new chemicals strategy in December 1999. The Environment Agency of England and Wales is currently reviewing its policy towards endocrine disrupters. The OECD has a programme on endocrine disrupters, mainly focusing on the development of testing procedures. The UN is currently negotiating a global treaty covering certain persistent organic pollutants (POPs), including PCBs, dioxin and DDT, with criteria for adding new chemicals. A similar agreement, the POPs Protocol, has already been negotiated among the UN Economic Commission for Europe. The International POPs Elimination Network is a non-governmental organization (NGO) coalition against POPs. No signs yet of any new controls on existing chemicals, even on the alkylphenols, which are already being phased out in Europe. The US National Academy of Sciences published the report, “Hormonally Active Agents in the Environment” in July 1999. Most relevant industry associations have issued statements about hormone disrupting chemicals relaying their concerns, but calling for more research before any action is taken. Industry claims that effects are not likely to be as significant as those of phytoestrogens. Some companies have stopped using suspect chemicals while others will continue to use them unless they are banned, considering endocrine disruption to be a hypothesis, rather than scientific fact.

In fact, consultants to the U.S. space agency have recommended the use of butane (lighter fluid) in some part-cleaning operations since this policy change. It would appear that as older scientists retire and/or are replaced by younger, inexperienced researchers/contractors, there is a lack of a common understanding of the past lessons learned from the use of these solvents. This may cause American society to repeat some of the same mistakes made earlier. In other words, the U.S. public may be facing a retreat to increased exposure to hydrocarbon products, and their associated health hazards, used for cleaning prior to the discovery of the destruction of the ozone layer by chlorofluorocarbons (CFCs). These developments are especially troubling in light of the United States’ active opposition to the Kyoto Protocol to decrease global-warming (i.e., carbon-based) emissions. The

The future of industrial cleaning and related public policy-making

359

Figure 9. Illustration of the three principal cleaning fields.

re-introduction of brominated cleaners, notably n-propyl bromide (nPB), is likewise a concern. UNEP’s STOC considers nPB to be ozone depleting and is not recommending it as a solvent substitute since “non-ozone-depleting solutions exist for all cleaning applications for which nPB is being promoted”. Aberrations in legal structures, especially liability issues, are no doubt at the root cause of how some societies approach environmental decision-making for cleaning applications. Consumerism, (i.e., the educated consumer) and organized labor (i.e., trade unions representing various segments of the workforce performing cleaning duties) also have roles to play. Various chemical formulators have become more adept in addressing worker safety and the environment due to these concerns. Partly because of these advances, the lines separating parts, precision and institutional (i.e., maintenance and janitorial) cleaning have blurred and are illustrated in Figure 9. As workplaces approach the safety of households in cleaning operations, overlaps among cleaning standards and performance guidelines may become more commonplace. Ironically, these same developments may also tend to increase multiple chemical sensitivities to certain, at-risk, individuals within a given population. 4. A CONCLUDING SCENARIO: PROVOCATIVE POSSIBILITIES INVOLVING SEMICONDUCTORS

In no industry is the efficacy of cleaning/rinsing cycles more essential than in the semiconductor industry. Cleanrooms, maintained at various levels of cleanliness under U.S. Standard 209E and now ISO standards according to the number and size of airborne particulates, generally require cleanliness levels many times greater than surgical fields. This is because a semiconductor, a silicon wafer with diodes and transistors, must act as a circuit at near atomic levels. Contamination is the primary cause of product failure. Moreover, the increased storage capacity of miniaturized computer chips has caused cleanliness requirements to increase ex-

C. LeBlanc

360

ponentially. Table 5 contains the water quality guidelines for chip manufacture and blood dialysis as a means to compare each system’s level of desired contamination control. The goal of this treatise is not to expose the semiconductor industry’s overreliance on energy and water resources; other researchers such as Ted Smith, founder and director of the Silicon Valley Toxics Coalition, are far more familiar with the industry and have documented this dependency. The purpose here is to reflect on the unsustainable nature of the current technology (indeed, most computers are considered obsolete within eighteen months of manufacture), more specifically, the unhealthy, unnatural conditions to which workers are exposed in cleanrooms. Table 5. Water quality guidelines for semiconductors (A) and hemodialysis (B) B

A Contaminant

Maximum concentration*

Contaminant

mg/l*

Suspended solids: residue / particulates Dissolved solids: organic (TOC) Ionic: resistivity / dissolved silica (SiO2) Aluminum (Al) cations Ammonium (NH4) cations Chromium (Cr) cations Copper (Cu) cations Iron (Fe) cations Manganese (Mn) cations Potassium (K) cations Sodium (Na) cations Zinc (Zn) cations Bromide (Br) anions Chloride (Cl) anions Nitrite (NO2) anions Nitrate (NO3) anions Phosphate (PO4) anions Sulfate (SO4) anions Bacteria

0.1 ppm / 500 counts per liter 0.020 ppm

Aluminum (Al) Arsenic (Ar) Barium (Ba) Cadmium (Cd) Calcium (Ca) Choramines Chlorine (Cl) Chromium (Cr) Copper (Cu) Fluoride (F) Lead (Pb) Magnesium (Mg) Mercury (Hg) Nitrate (AS N) Potassium (K) Selenium (Se) Silver (Ag) Sodium (Na) Sulfate (SO4) Zinc (Zn)

0.01 0.005 0.1 0.001 2.0 0.1 0.5 0.014 0.1 0.2 0.005 4.0 0.0002 2.0 8.0 0.09 0.005 70.0 100.0 0.1

18.3 megohm-cm / 3 ppb 0.2 ppb 0.3 ppb 0.02 ppb 0.002 ppb 0.02 ppb 0.05 ppb 0.1 ppb 0.05 ppb 0.02 ppb 0.1 ppb 0.05 ppb 0.05 ppb 0.1 ppb 0.2 ppb 0.05 ppb 0 counts per 100 ml

Source: P. Cartwright, Proc. of the Precision Cleaning meeting held in Rosemont, IL, May 1995. *Measurement system chosen by industry.

The future of industrial cleaning and related public policy-making

361

4.1. Hospitals, cleanrooms and cleaners: could there be a connection? “The globalization of infectious diseases is not a new phenomenon. However, increased population movements, whether through tourism or migration or as a result of disasters; growth in international trade in food and biological products; social and environmental changes linked with urbanization, deforestation and alterations in climate; and changes in methods of food processing, distribution and consumer habits have reaffirmed that infectious disease events in one country are potentially a concern for the entire world ”. So begins the World Health Organization’s (WHO) report by the Secretariat on global health security – epidemic alert and response (November 2000). Staphylococcus Aureusc has been implicated in hospital-acquired infections since the 1950s when 50% of the organism’s strains developed resistance to penicillin. It has been considered a serious bacterial pathogenic threat since that time. Known as a ‘super bug’, the organism has also become resistant to newer and more powerful antibiotics such as tertracycline and the aminoglycosides. It is common, even in the cleanest healthcare facilities, with the elderly, the seriously ill and those patients with compromised immune systems being at greatest risk [21]. The MRSA (methicillin-resistant Staphylococcus Aureus) super bug responds only to the antibiotic vancomycin, whose use is now restricted due, at least in part, to its apparent role in producing the ‘super bug’ VRE (vancomycin-resistant Enterococcid), following its application in European cattle, for which there is no known treatment. ‘Flesh-eating’ disease or Necrotizing Fasciitis is another antibiotic-resistant bacterial infection of the Streptococcus Type A variety associated with surgical or wound patients. This variant super bug is more powerful than other strains, with stronger m-protein serotypes. Benign forms of Staphylococcus Aureus are natural habitants of skin and mucus membranes of humans and can be found throughout the environment from dust to door knobs. An infectious disease expert at the American Society for Microbiology in Miami, Florida reported that even personnel who do not come into direct contact with patients can accumulate and spread bacteria, including resistant strains. Thus, the WHO’s concern is justified for the pandemic spread of these infections to the general population, including apparently young and otherwise healthy subjects, as illustrated in the ‘mad cow’ disease or Bovine Spongiform Encephalopathy (BSE), now in its fatal human form called new variant Creutzfeldt-Jakob Disease (nvCJD). Inadvertent cannibalism (that is, cow material being added to cattle feed without the ranchers’ knowledge), not an overuse c

Staphylococci are gram positive bacteria that are typically arranged in clumps or grape-like clusters. They can be distinguished from streptococci in that only the staphylococi are catalase-positive (catalase is an enzyme that liberates oxygen from hydrogen peroxide). d Enterococci are gram-positive bacteria that are widely distributed in nature and are part of the normal flora of the gastrointestinal and genital tracts.

362

C. LeBlanc

of antibiotics, appears to be the root cause of these infectious mutations. The disease involves prions. Prions are tiny biological bits that may or may not be alive and so are impossible to ‘kill’ (in the conventional sense of the term) to prevent the infection’s spread. Some of these prions can be viewed three-dimensionally with RosMol/Chime at www.mad-cow.org/prion_structure_folder/viewers.html. 4.2. The ‘cleaner’ connection Recently, Americans have been introduced to a line of household cleaners, advertised as antibacterial, for applications where no antibacterial activity is warranted. The more popular the cleaners become, the more product varieties appear on grocers’ shelves. Studies at Tufts University’s School of Medicine (Boston, Massachusetts) revealed that the antibacterial agent triclosan, used in many of these products, acts like an antibiotic to promote bacterial resistance and, potentially, the spread of untreatable infections. Furthermore, the U.S. Food and Drug Administration (FDA) reports that antibacterial soaps kill the beneficial bacteria that live on skin. Unlike their pathogenic counterparts, these bacteria apparently strengthen the immune systems of children [22]. These seemingly unconnected events or trends have one or two things in common: they are related to the things we chose to clean well or not to be able to keep clean enough. Meanwhile, microscopic forms of life have been found in Arcticlike conditions and other species such as the archaea have evolved near volcanic emissions under the sea, both environs thought to be uninhabitable by the scientific community not too long ago. Somehow, life finds a way. In hospitals, sterility is maintained (most notably for surgery), cleaning is performed and dressings (gowns, masks, gloves) are donned to protect the person from pathogens. In cleanrooms, sterility is maintained, cleaning is performed and dressings donned to protect the product from the person. Cleanrooms, whose sterile environment routinely outrivals the surgeon’s needs, may offer the next best habitat for a super bug, perhaps of prion-nature, to establish a foothold. The technical staff of these high-tech establishments may already have damaged immune systems due to the unnatural conditions in which they work on a daily basis (there is no such thing as a ‘good’ bacterium in a cleanroom). Regardless, the spread of a hypothetical cleanroom-acquired infection may not require an at-risk host, as has been previously discussed. 4.3. Closing statement Several aspects of the search for surface cleanliness are neither simple nor straightforward. The Alliance for the Prudent Use of Antibiotics located in Boston, Massachusetts, an international organization with members from more that ninety countries, has been monitoring the worldwide emergence of treatmentresistant microbial strains since 1981. This group, and others in the scientific community such as the U.S. Center for Disease Control (CDC), should be made

The future of industrial cleaning and related public policy-making

363

aware of the developments in the cleaning industry so that other studies, like those conducted at Tufts medical school, can be undertaken. In the meantime, unless and until the chemical industry provides complete chemical disclosure on a global basis, institutions such as the Toxics Use Reduction Institute’s research facility at the University of Massachusetts Lowell should assist in formulating green chemical cleaners, in addition to providing education and training programs and state-of-the-art laboratory testing of existing products. This could be accomplished through partnerships with commercial enterprises and/or other research facilities. Remaining pertinent issues, some having nothing to do with cleaning performance, could then be addressed. These include studies on the chemical additives of fragrances (over 80% of the odorants now used are synthetic in origin) and dyes or colorants (often added for worker safety in product recognition). Tighter quality control on percentages of ingredients could also be maintained, since currently the concentrations of a cleaner’s components reported on its Material Safety Data Sheet can vary by as much as 400%. Most importantly, chronic, hormetic and synergistic chemical-exposure tests need to be developed and implemented before cleaners are marketed. The development of the computer program, The Aqueous Way to Go revealed many of these trends and potential hazards in cleaning processes and chemicals. The successful technical diffusion of this tool will require an educational component, the topic of an upcoming paper. REFERENCES 1. A. Warhurst and the Friends of the Earth, London. http://website.lineone.net/~mwarhurst/policy.html (May 19, 2000). 2. C.M. Hansen, Hansen Solubility Parameters: A User’s Handbook, CRC Press, Boca Raton, FL (1999). 3. H. Cabezas, P. Harten and M. Green, Chemical Engineering, 107 (3), 109 (2000). 4. “Epoxidation Catalyst Immobilized In Ionic Liquid”, Chemical & Engineering News, 78, 21 (May 22, 2000). 5. J. Brennecke and E. Beckman, Nature, 399, 28-29 (May 6, 1999). 6. “Chemical Hormesis: Beneficial Effects at Low Exposures/Adverse Effects at High Exposures”, Brochure, Texas Institute for the Advancement of Chemical Technology, College Station, TX. 7. Paracelsus, “Seven Defensionmes: The Reply to Certain Calumniations of His Enemies,” trans. by C. Temkin in Four Treatises of Theophrastus von Hohenheim called Paracelsus, John Hopkin Press, Baltimore, MD (1994). 8. E. Calabrese and L. Baldwin, Chemical Hormesis: Scientific Foundations, Documentation and Implications for Risk Assessment, School of Public Health, University of Massachusetts Amherst (June 1997). 9. C. LeBlanc, “The Search for Safer and Greener Chemical Solvents in Surface Cleaning: A Proposed Tool to Support Environmental Decision-Making”, Ph.D. Thesis, Erasmus University, Rotterdam (2001). 10. E. Dodds and W. Lawson, Proc. Royal Soc. Lon. B. 125, 222-232 (1938). 11. R. Kociba, Toxicol. Appl. Pharmacol., 46, 279 (1978). 12. G. Mueller and U. Kim, Endocrinol., 102, 1429-1435 (1978). 13. A. Soto, H. Justicia, J. Wray and C. Sonnenschein, Environ. Health Persp., 92, 167-173 (1991).

364

C. LeBlanc

14. S. Jobling and J. Sumpter, Aquat. Toxicol., 27, 361-372 (1993). 15. R. White, S. Jobling, S. Hoare, J. Sumpter and M. Parker, Endocrinol., 135, 175-182 (1994). 16. A. Soto and C. Sonnenschein, Environ. Health Persp., 103, 113-122 (1995). 17. E. Routledge and J. Sumpter, Environ. Toxicol. & Chem., 15, 241-248 (1996). 18. “Industry Glimpses New Challenges as Endocrine Science Advances”, ENDS Report 290, 2630 (1999). Contact TURI’s library at telephone 978-934-3390 for information to obtain a copy of this report. 19. U.S. EPA Endocrine Home Page, www.epa.gov/scipoly/oscpendo (Jan. 15, 2001). 20. John Snow Institute, “Toxics Use in Massachusetts: Implications for Public Health”, Center for Environmental Health Studies, Boston, MA (Jan. 1993). 21. “Staphylococcus Aureus: An Emerging Super Bug” (Feb. 12, 2001). http://srd.yahoo.com/goo/S.+Aureus+AND+super+bug/2/*http://members.tripod.com/LouCaru/ index-9.html 22. Tufts Health Maintenance Organization, “WELL!” Magazine, Waltham, MA (Winter 2000).