Advances in Ceramic Armor V
Advances in Ceramic Armor V A Collection of Papers Presented at the 33rd International Conference on Advanced Ceramics and Composites January 18-23, 2009 Daytona Beach, Florida Edited by
Jeffrey J. Swab Volume Editors
Dileep Singh Jonathan Salem
~WILEY A John Wiley & Sons, Inc., Publication
Copyright 02010 by The American Ceramic Society. All rights reserved. Published by John Wiley & Sons, Inc., Hoboken, New Jersey. Published simultaneously in Canada. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, (978) 750-8400, fax (978) 750-4470, or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 11 1 River Street, Hoboken, NJ 07030, (201) 748-601 1, fax (201) 748-6008, or online at http:/lwww.wiley,comlgo/permission. Limit of LiabilityiDisclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Neither the publisher nor author shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages.
For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at (800) 762-2974, outside the United States at (317) 572-3993 or fax (317) 572-4002. Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic format. For information about Wiley products, visit our web site at www.wiley.com. Library of Congress Cataloging-in-Publication Data is available.
ISBN 978-0-470-45755-9 Printed in the United States of America. 10 9 8 7 6 5 4 3 2 1
Contents
Preface
ix
Introduction
xi
IMPACT, PENETRATION AND MATERIAL MODELING Fragmentation of Ceramics in the Ballistic Environment Dennis E. Grady
3
Flow Behavior of Glass at the Tip of a Penetrator
19
Rheology of Powder and Porous Media in Modeling of Penetration into Porous Ceramic
27
Computer Modeling of Shock Wave Propagation in SiC-Sample
39
Ballistic Impact Damage Observations in a Hot-Pressed Boron Carbide
45
Characterization of Microstructural Damage in Silicon Carbide Processed via Modified Chemical Vapor Deposition
57
D. A. Shockey, D. Bergmannshoff, D. R. Curran, and J. W. Simons
B.A. Galanov, V.V. Kartuzov, and S.M. lvanov
V.L. Bekenev, V.V. Kartuzov, E.V. Kartuzov, and H.V. Hachatraian
J.C. LaSalvia, R.B. Leavy, J.R. Houskamp, H.T. Miller, D.E. MacKenzie, and J. Campbell
H.T. Miller, J.C. LaSalvia, R.B. Leavy, and D.E. MacKenzie
MATERIAL CONCEPTS, PROCESSES AND CHARACTERIZATION Effects of Grain Size, Shape and Second Phases on Properties of Sintered Sic
69
P. G. Karandikar, G. Evans, S. Wong, and M. K. Aghajanian
V
Indenter Elastic Modulus and Hertzian Ring Crack Initiation
83
High Frequency Ultrasound of Alumina for High Strain-Rate Applications
91
K. T. Strong, Jr., A. A. Wereszczak, and W. L. Daloz, and 0. M. Jadaan
S. Bottiglieri and R. A. Haber
The Effect of Particle Size, Particle Loading and Thermal Processing Conditions on the Properties of Alumina Reinforced Aluminum Metal Matrix Composites
105
Pressureless Sintering of B4C-Sic Composites for Armor Applications
113
Allyn L. McCormick, Michael K. Aghajanian, Andrew L. Marshall
Rosa Maria da Rocha. Francisco C. L. de Melo
APPLICATIONS OF NDE A Portable Microwave Interference Scanning System for Nondestructive Testing of Multi-Layered Dielectric Materials
123
Destructive Testing and Nondestructive Evaluation of Alumina Structural Ceramics
135
Nondestructive Evaluation of as Fabricated and Damaged Encapsulated Ceramics
147
Microstructural Study of Sintered SIC via High Frequency Ultrasound Spectroscopy
159
Impact Damage Analysis in a Level Ill Flexible Body Armor Vest Using XCT Diagnostics
171
K. F. Schmidt, Jr. , J. R. Little, Jr., W. A. Ellingson, and W. Green
Raymond E. Brennan, William H. Green, James M. Sands
William H. Green, Raymond Brennan, and Robert H. Carter
A. R. Portune and R. A. Haber
Joe Wells, Nevin Rupert, and Murray Neal
TRANSPARENT ARMOR Impact onto Glass and Glass Ceramic Bars
183
Numerical Study of the Effect of Surface Stresses of Transparent Ceramics of Laminated Targets for Military Armor Applications
195
Analyses of Various Damage Mechanisms in Transparent Armor Subject to Projectile Impact
205
Stephan Bless, John Tolman, Scott Levinson, and Ian Polyzois
Costas G. Fountzoulas, James M. Sands, Gary Gilde, Parimal J. Patel
Kevin C. Lai, Xin Sun, and Douglas W. Templeton
vi
.
Advances in Ceramic Armor V
Pressureless Reaction Sintering of AlON Using Aluminum Orthophosphate as a Transient Liquid Phase
213
ALON@' Transparent Armor
225
Author Index
2 33
Michael Bakas, Henry Chu
Lee M. Goldman, Robyn Foti, Mark Smith, Uday Kashalikar, and Suri Sastri
Advances in Ceramic Armor V
vii
Preface
The Army’s primary goal is to provide soldiers with the equipment and resources to do their job and to do it safely. A strategic element of the future success of the US military in the on-going global conflicts is the performance of armor systems, for air and ground vehicles as well as the individual soldier, against a myriad of potential threats. Ceramic materials are currently used in many armor systems and they will be integral components of future systems. The Armor Ceramics Symposium began in 2003 following the success of the Ceramic Armor Materials by Design symposium at the Pack Rim IV International Conference on Advanced Ceramics and Glasses in November 2001. This coupled with the U.S. military actions in response to the September 11, 2001 terrorist attacks, were reasons for creating an annual meeting focused on armor ceramics. Furthermore the Pac Rim symposium showed the need for an unclassified gathering to discuss armor materials. Thus the primary objective of the Armor Ceramics Symposium is to provide an annual forum for the presentation and discussion of unclassified information and ideas pertaining to the development and incorporation of ceramic materials for armor applications. The seventh edition of this symposium focused on Impact, Penetration and Material Modeling, Material Concepts, Processes and Characterization, the Application of NDE, and Transparent Armor. On behalf of the organizing committee I would like to thank all of the presenters, authors and manuscript reviewers for their efforts in making the symposium and the associated proceedings a success. My personal thanks goes to Lisa Prokurat Franks who spent many hours on the phone with me discussing a variety of aspects and issues related to the organization of the symposium. Finally I cannot forget Marilyn Stoltz of The American Ceramic Society who was always there to answer my questions and provide the guidance and administrative support necessary to make the symposium a success. JEFFREY J. SWAB
ix
Introduction
The theme of international participation continued at the 33rd International Conference on Advanced Ceramics and Composites (ICACC), with over 1000 attendees from 39 countries. China has become a more significant participant in the program with 15 contributed papers and the presentation of the 2009 Engineering Ceramic Division's Bridge Building Award lecture. The 2009 meeting was organized in conjunction with the Electronics Division and the Nuclear and Environmental Technology Division. Energy related themes were a mainstay, with symposia on nuclear energy, solid oxide fuel cells, materials for thermal-to-electric energy conversion, and thermal barrier coatings participating along with the traditional themes of armor, mechanical properties, and porous ceramics. Newer themes included kano-structured materials, advanced manufacturing, and bioceramics. Once again the conference included topics ranging from ceramic nanomaterials to structural reliability of ceramic components, demonstrating the linkage between materials science developments at the atomic level and macro-level structural applications. Symposium on Nanostructured Materials and Nanocomposites was held in honor of Prof. Koichi Niihara and recognized the significant contributions made by him. The conference was organized into the following symposia and focused sessions: Symposium 1 Symposium 2 Symposium 3 Symposium 4 Symposium 5 Symposium 6 Symposium 7 Symposium 8
Mechanical Behavior and Performance of Ceramics and Composites Advanced Ceramic Coatings for Structural, Environmental, and Functional Applications 6th International Symposium on Solid Oxide Fuel Cells (SOFC): Materials, Science, and Technology Armor Ceramics Next Generation Bioceramics Key Materials and Technologies for Efficient Direct Thermal-to-Electrical Conversion 3rd International Symposium on Nanostructured Materials and Nanocomposites: In Honor of Professor Koichi Niihara 3rd International symposium on Advanced Processing & Manufacturing Technologies (APMT) for Structural & Multifunctional Materials and Systems xi
Symposium 9 Symposium 10 Symposium 11 Focused Session 1 Focused Session 2 Focused Session 3 Focused Session 4
Porous Ceramics: Novel Developments and Applications International Symposium on Silicon Carbide and CarbonBased Materials for Fusion and Advanced Nuclear Energy Applications Symposium on Advanced Dielectrics, Piezoelectric, Ferroelectric, and Multiferroic Materials Geopolymers and other Inorganic Polymers Materials for Solid State Lighting Advanced Sensor Technology for High-Temperature Applications Processing and Properties of Nuclear Fuels and Wastes
The conference proceedings compiles peer reviewed papers from the above symposia and focused sessions into 9 issues of the 2009 Ceramic Engineering & Science Proceedings (CESP); Volume 30, Issues 2-10, 2009 as outlined below: Mechanical Properties and Performance of Engineering Ceramics and Composites IV, CESP Volume 30, Issue 2 (includes papers from Symp. 1 and FS 1) Advanced Ceramic Coatings and Interfaces IV Volume 30, Issue 3 (includes papers from Symp. 2) Advances in Solid Oxide Fuel Cells V, CESP Volume 30, Issue 4 (includes papers from Symp. 3) Advances in Ceramic Armor V, CESP Volume 30, Issue 5 (includes papers from SYmP. 4) Advances in Bioceramics and Porous Ceramics 11, CESP Volume 30, Issue 6 (includes papers from Symp. 5 and Symp. 9) Nanostructured Materials and Nanotechnology 111, CESP Volume 30, Issue 7 (includes papers from Symp. 7) Advanced Processing and Manufacturing Technologies for Structural and Multifunctional Materials 111, CESP Volume 30, Issue 8 (includes papers from SYmP. 8) Advances in Electronic Ceramics 11, CESP Volume 30, Issue 9 (includes papers from Symp. 11, Symp. 6, FS 2 and FS 3) Ceramics in Nuclear Applications, CESP Volume 30, Issue 10 (includes papers from Symp. 10 and FS 4) The organization of the Daytona Beach meeting and the publication of these proceedings were possible thanks to the professional staff of The American Ceramic Society (ACerS) and the tireless dedication of the many members of the ACerS Engineering Ceramics, Nuclear & Environmental Technology and Electronics Divisions. We would especially like to express our sincere thanks to the symposia organizers, session chairs, presenters and conference attendees, for their efforts and enthusiastic participation in the vibrant and cutting-edge conference. DILEEP SINGH and JONATHAN SALEM Volume Editors
xii
Advances in Ceramic Armor V
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
Impact, Penetration, and Material Modeling
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
FRAGMENTATION OF CERAMICS IN THE BALLISTIC ENVIRONMENT Dennis E. Grady Applied Research Associates, Southwest Division, 4300 San Mateo Blvd NE Albuquerque, New Mexico 871 10, USA ABSTRACT The catastrophic disruption of materials in the ballistic environment commonly plays a central role in both the successful application of armor systems and the effective application of anti-armor systems. A theory of the dynamic fragmentation of solids based on continuum energy principles has provided a basis for assessing fragmentation in a wide range of ballistic applications over the past several decades’.*. Applications of the theory to the fragmentation of brittle solids, including glass and ceramic, have been problematic, however. Recently, some of the physics issues governing length scales and size distributions in the dynamic fragmentation of brittle solids have come to light. The earlier energy-based fragmentation theory has been broadened to accommodate dynamic fragmentation in brittle materials. The paper summarizes past theories and their applications to fragmentation in the ballistic environment. More recent applications to ballistic fragmentation of glass and ceramic materials are described. IhTRODUCTION Fragmentation of the component materials is a natural consequence of a terminal ballistic event. Commonly, the size distribution and trajectory of the resulting fragment debris is central to the effective application of the projectile or armor component specific to the event. The multiplicity of materials of interest in terminal ballistic applications spans a wide range of breakup and fragmentation phenomena. The catastrophic fragmentation of brittle solids under the intense dynamic loading of the ballistic event is particularly interesting, and is perhaps the least understood. The focus of this paper is on the dynamic fragmentation of brittle solids with particular emphasis on ceramic and glass. A theory of dynamic fragmentation for brittle materials is proposed. This theory is not a model of fragmentation for any specific application. Rather, the theory draws on underlying physical principles that determine the functional forms of the analytic relations that describe of the statistical fragment size distributions and the fragmentation size length scales. More broadly, the paper provides background on the topic of dynamic fragmentation of solids. The breadth and the richness of the topic are emphasized. At the same time previous theoretical work is outlined that is necessary to the present theoretical progress to brittle solids. The seminal wartime contributions of Nevill M ~ t t to~ .the ~ area of dynamic fragmentation, integral to the progress, are recognized. MOTT FRAGMENTATION In the study of the dynamic fragmentation of solids there are several historical threads that can be traced. For those concerned with the topics of terminal ballistics or exploding munitions the seminal wartime effort of Nevill M ~ t t is~ the , ~ most relevant. Of the several early efforts at achieving a theoretical understanding of dynamic fragmentation, the work of Mott has the more solid physical basis. The fragmentation theory of Mott predicts the two key elements of the dynamic fragmentation event. Namely, the statistical distribution in fragment size, and the governing distribution length scale or, equivalently, the average fragment size. A feature of Mott’s seminal theoretical study of fragmentation that has become a part of the lore in representation of terminal ballistics or exploding munitions fragmentation is the fragment distribution Mott plot. The Mott plot is a representation of the fragment distribution of a specific fragmentation
3
Fragmentation of Ceramics in the Ballistic Environment
event where the logarithm of the cumulative fragment number larger is graphed against the individual fragment mass, InN =f(m). Often the data plot linear, or near linear, in this semi-logarithmic representation of the fragment size data. Some contend that the Mott plot more specifically implies log number versus some specific power of the fragment mass. Mott himself vacillated on this point and, in one of his later wartime reports, suggested that a linear dependence on the mass m is perhaps the most appropriate. In any case, a Mott plot is here defined as a semilog plot of cumulative N versus m . The fragment size data do not always plot linear, but this representation is an appropriate space for a clear display of a broad body of dynamic fragmentation data. Mott was principally concerned with a theoretical description of fragmenting munitions and expended most of his efforts in describing the breakup of explosion-driven expanding cylindrical cases. The Mott plots for several such case fragmentation experiments performed on Aermet 100 steel cylinders’ are shown in Figure (1). 1000
AerMet 100 Steel
L
y
=
\.p
1
0
5
I
5 . 2 gm. B = 0.85
10
= 2.6
gm.p =0.67
15 20 Mass (grams)
25
I
30
Figure 1. Mott fragmentation plots for two explosive tests on AerMet100 steel cylinders. Before embarking on the principal intent of this paper, there is value in observing some selected fragmentation data within the same Mott plot theme from more diverse areas of dynamic fragmentation. An interesting counterpart to the exploding case fragmentation shown in Figure (1) occurs when a similar hollow cylindrical metal shell is instead imploded through placement of the explosive on the outside of the shell. Comparable kinetic energy imparted to the inward directed motion of the shell leads to intense shock compression and dissipation when the cylinder collapses towards the cylinder centerline. Shock dissipation is sufficiently intense that most of the metal either melts or is in a very hot solid state. The material motion is reversed when rebound at the centerline occurs. Fragmentation ensues with the hot liquid or solid metal particulating into fragments in the 10’s to 100’s of micrometer range. Ejected high-velocity fragments undergo subsequent aerothermal burning. Particle size distributions in such experiments have been determined by impeding witness plates6. A Mott fragmentation plot constructed from data from one test conveniently displays the fragment distribution as is shown in Figure (2).
4
.
Advances in Ceramic Armor V
Fragmentation of Ceramics in the Ballistic Environment
0.3
$j0.1
... \
0.03 Mass (mg)
0.2 0.3 0.5 0.7 Strain Rate (psi)
0.1
1.0
Figure 2. Mott plot for liquid fragments from one imploding metal cylinder fragmentation test is shown on the left. Fragment size data from witness plates and impact flux on one square centimeter. Mean fragment size versus expansion strain rate for three tests is displayed on the right6. A starkly different dynamic fragmentation event is illustrated by the Hubble photograph in Figure (3) of fragmentation of the Shoemaker-Levi comet. Fragmentation is caused by gravitational forces as the comet passes the Roche limit on its downward trajectory towards the planet Jupiter. Fragment luminosity provides an approximate relative measure of the fragment masses. The fragment distribution Mott plot constructed for the comet fragments is shown in Figure (3).
‘0
20 40 60 80 100 Relative Luminosity (- Mass)
Figure 3. Hubble image of the fragmented Shoemaker-Levi comet and the fragment size Mott plot assesses from relative luminosity of comet fragments.
Spiraling down many decades in length scale, the process of nuclear spallation entails the relative high energy impact of atomic nuclei, resulting in a distribution of nuclear fragments. The experimental consequences of a carbon nucleus (atomic mass 12) onto a silver nucleus (atomic mass 107) nuclear spallation event7 are illustrated in Figure (4). Sixteen electrically charged fragments are produced ranging from deuterium (two nucleons) to beryllium (nine nucleons). A Mott plot of the data provides a sensible representation of the nuclear fragment distribution. The fragment distribution is reasonably represented by a simple exponential h c t i o n * .
Advances in Ceramic Armor V
. 5
Fragmentation of Ceramics in the Ballistic Environment
(70 MeVNucleon)
1 0
5
10
Mass in Nucleons
15
Figure 4. Fragmentation recorded in a photographic emulsion caused by the 70 MeV per nucleon collision of a carbon and a silver nucleus'. Cumulative fragment number distribution Mott plot constructed from the resulting nuclear fragment data'. BRITTLE FRAGMENTATION The selected examples of dynamic fragmentation in the previous section differ markedly in properties of the materials, the nature of the fragmentation, and the vast span of length scales separating the events. Yet, they exhibit some striking commonalities, and all are sensibly displayed in the Mott plot representation of the distribution data, or, a plot of log cumulative number versus fragment mass. There are examples, however, of dynamic fragmentation where a Mott plot representation is not appropriate. Fragmentation of highly brittle materials, and perhaps other materials, results in fragment distributions that are poorly displayed on a Mott plot. Fragment fines are so numerous that counting fragments is not reasonable. Sieving techniques are used to separate fragments into size ranges. Fragment distributions are most commonly represented through plots of cumulative mass versus fragment size. Such distributions are observed to plot linear or near linear in a log cumulative mass versus log size display, or Schuhmann distribution', of the fragment size data. There are many examples of fragmentation of brittle materials that would make the salient points. An example is chosen here because of direct relevance to the terminal ballistics effects of interest. The example is that of tests in which plates of boron carbide and composite are subjected to normal impact by tungsten carbide core armor piercing projectiles. Of interest here is the shatter fragmentation of the boron carbide ceramic. The authors of this study" collected and determined size distributions for the resulting boron carbide fragment debris. Fragment distributions plotted in the Schuhmann form' are constructed from the data and are shown in Figure (5) from tests for several impact velocities. The near linear plots with power of order unity are representative of the behavior of brittle materials. Broadly, the power n ranges over about 0.5 < n < 1 . 5 , and occasionally higher, in Schuhmann plots of brittle fragment distribution data.
6
.
Advances in Ceramic Armor V
Fragmentation of Ceramics in the Ballistic Environment
100 t
5
1
I CI
'6,
865 or 926 kmir 793or814km/s
Boron Carbide
'
'
" " " ' 1.0 ' ' ' J Fragment Size (mm)
10
Figure 5. Schuhmann kagment size distribution plot constructed from data for boron carbide ceramic subjected to ballistic impact by tungsten carbide core projectiles''. EQUILIBRIUM FRAGMENTATION Before assessing the physics of fracture and fragmentation of brittle solids, it is necessary to first describe a theory of dynamic fragmentation that has demonstrated reasonably success for other materials. A representative set of test data that will be applied in describing this theory is shown in Figure (6). In these test the behind target fragment debris was captured and analyzed. The fragment size distribution Mott plots are shown for two tests from a series of experiments in which plates of stainless steel were sub'ected to impact by copper spheres over a selected range of impact velocities and angles of obliquity' 4 .
Figure 6 . Impact of a copper spheres onto a stainless steel plates". Mott plot of the behind target fragment distribution for two such tests. Impact angle from normal and velocity: Test 12, 60 degrees, 3.26 km/s; Test 13, 50 degrees, 3.09 Ms. As the fragmentation theory is described, reference will be made to equilibrium and nonequilibrium fragmentation. Briefly, equilibrium or nonequilibrium refers to the ability, or lack thereof, of the material to undergo fracture failure and fragmentation when a theoretical energy criterion is achieved. An energy criterion identifies onset of equilibrium fragmentation. Delayed failure and an additional failure criterion exemplify nonequilibrium fragmentation. At the present state of understanding equilibrium fragmentation appears to have fairly broad application across a range of material types. Nonequilibrium fragmentation applies to the brittle materials such as the competent ceramics and glasses. Some form of continuous transition from equilibrium to nonequilibrium fragmentation seems reasonable, however the boundaries of departure from one to the other are still poorly understood. The fragment size distribution data for the two tests shown in Figure ( 6 ) are sensibly described by a
Advances in Ceramic Armor V
.
7
Fragmentation of Ceramics in the Ballistic Environment
simple exponential function of the form,
The fragment size scale factor ,LI is observed to decrease as the intensity of the impact increases. In other words, higher velocity impacts resulting in more, but smaller, fragments. The theory, and sought for physical understanding, is focused on two key elements; first, prediction of the distribution of fragment size resulting from the event, second, prediction of the distribution size scale L,I and its dependence on the impact intensity and material properties.
'"1
OMott Plot
I
Figure 7 . Length L is broken into three fragments with equal likelihood of kacture placement. The fragments size distribution Mot! plot resulting from a binomial probability is shown on the right. Fragment Size Distribution I have heard a tale that if a stick of blackboard chalk (actually pressed gypsum in the later years of chalk and blackboard use) is dropped onto the floor, it breaks invariably into three pieces. (In today's increasing use of grease pencils and white boards this tale is rapidly becoming dated.) Whether true or not, it provides a theme for this first look at fragment size distributions. The chalk of length L in Figure ( 7 ) is broken at two places that are assumed to have equal likelihood of occurrence at any point within the length. The distribution of possible fragment sizes is governed by binomial statistics and yields the cumulative probability distribution for fragments of length I of the forms3'*, F(1) = 1- (1- I / L y .
(2)
The corresponding Mott distribution ( N o = 3 ) is shown in Figure (7). If the number n of fragments that the length L is partitioned into is increased, the curve in the Mott plot becomes increasingly linear and in the limit the fragment distribution approaches the exponential form, N ( 1 ) = N,e-"" (3) where /z = L i n is the average fragment length. Equation (3) is derived directly if an unbounded length is fractured along the length with frequency A, in which case the fragment size distribution is determined from Poisson probabilities. The present one-dimensional model results in an exponential distribution in fragment sizes not unlike the experimental results provided in Figure(6). One is encouraged to explore the random geometric fragmentation problem in two, and perhaps three, dimensions. Investigation of random geometric fragmentation and its possible application to real fragmentation has been undertaken by
8
Advances in Ceramic Armor V
Fragmentation of Ceramics in the Ballistic Environment
various early workers in the field. One approach was pursued by Mott and Linfoot” in their initial report on munitions fragmentation. Some of this history is provided in a recent r e ~ i e w ’ ~ . Pickup Sticks (Mott)
Figure 8. Three algorithms
H Seauential Sementation
for randomly fragmenting
a surface.
Consider random fragmentation in two dimensions. One immediately confronts a problem not evident in the random fragmentation of a line. Namely, a decision is necessary as to what form of algorithm to use in randomly partitioning the surface. Three of many possibilities are illustrated in Figure (8). Each algorithm starts through random placement of points on the surface. In the first, lines are scribed through the surface with random orientation. This fragmentation algorithm was investigated by Mott and Linfoot” in some detail. For obvious reasons the algorithm has been identified as pick-up sticks fragmentation. In the second algorithm randomly oriented lines are sequentially scribed through the points that bisect the entered fragment. Third is the well-known Voronoi tessellation. Analytic expressions for the fragment size (area) distributions have been determined for each of the fragmentation algorithms, although some intuitive license is applied in one in~tance’~. The pick-up sticks algorithm results in a Bessel function distribution of the fragment sizes. Sequential segmentation is exactly described by a simple exponential. The Voronoi tessellation provides a gamma function fragment size distribution. The Mott curves for each of the algorithms are compared in Figure (9). Notably, much of the distribution for the three algorithms is nominally the same. It is only for a reasonably small fraction of the total at the small fragment tail of the distribution that the curves differ. The Voronoi algorithm results in a dearth of small fragments whereas the pick-up sticks algorithm leads to an excess. One can speculate, based on the three examples that random geometric fragmentation of a surface, and by extrapolation a volume, will result in an exponential functional form with a single size scale parameter, although not necessarily a simple exponential. Further, the distribution is reasonable insensitive to the fragmentation algorithm (discounting the small fragment tail). Bessel (Pichp Sticks)
3
-s
Exponential (Sequential)
m 4
Figure 9. Fragment size distributions in Mott plot representation for three fragmentation algorithms. The cumulative probability distribution for each of these examples can be written,
Advances in Ceramic Armor V
.
9
Fragmentation of Ceramics in the Ballistic Environment
where the likelihood function h ( m / p ) will differ for each of the examples. The key point is that each of the distributions is constrained by a single length (size) scale p which provides the scale factor for the distribution. The preceding efforts have, of course, been an exercise in geometric fragmentation where a Poisson point process followed by a partitioning algorithm is used to randomly fragment a body. The dynamic fragmentation of many materials does not markedly differ from this geometric process, however. Dynamic spall in metals, for example, as exhibited by the images provided in the reviews of Shockey et al.” and Curran et all6, are processes in which fractures activate at random at independent points in the body, accelerate, grow, and ultimately coalesce to provided surfaces of the fragmented material. Physics-based models with increasing degrees of sophistication governing random fracture activation, growth and coalescence result in fragment distributions of the form provided in Equation (4). The classic physics-based fragmentation model of Mott4 is a one-dimensional analysis undertaken to determine the distribution in spacings resulting from the axial splitting of explosively expanding munitions. The resulting distribution is of the form of Equation (4) with a power-law likelihood function h ( m ) = ( m / p ) P . In the Mott model and analysis the resulting size scale is shown to be approximately p while the power is about p 7 / 2 . The material density and yield stress are p and Y , respectively, E is the strain rate of the expanding munitions, while y is the Mott statistical fracture parameter4’I2. These theoretical arguments, although not definitive, support the exponential or near exponential character observed in many of the dynamic fragment events where a single size scale parameter ties the distribution to the intensity of the event. Remaining then, is an assessment based on theoretical considerations of this size scale factor for a given dynamic fragmentation event.
- m,
-
Fragment Size Scale Factor Again, the impact fragmentation experiment depicted in Figure ( 6 ) Will provide the application for a theoretical assessment of the scale factor constraining the experimental fragment size distribution. Impact causes the dynamic fragmentation of a portion of the plate neighboring the point of impact. The fragment debris is ejected and expands with an intensity that is proportional to the impact velocity. If the impact velocity is increased the number of fragments produced is greater and the average fragment size is smaller. The scale factor p in the exponential expression for the fragment size distribution accounts for the dependence on intensity of the fragmentation event.
10
Advances in Ceramic Armor V
Fragmentation of Ceramics in the Ballistic Environment
Time or Correlation Distance
Figure 10. Model for visualizing the energies governing the characteristic fragment size in an equilibrium fragmentation event. The principle issues concerning the fragmentation process in this example are captured in the idealized, but conceptually more transparent, model illustrated in Figure (10). A thin circular ring, or a spherical shell, of material is provided an initial impulse leading to asymmetrically outward expansion at a velocity V . (Deceleration due to deformation dissipation of the initial kinetic energy is ignored.) A sensible measure of the expansion intensity is provided by the expansion rate B = V / R , where R is the radius of the ring or shell. If material deformation response to the imparted motion is linear elastic, then elastic strain energy increases with time according to,
u, = pc2&2t'/2,
(5)
where c is an appropriate elastic wave speed that determines the elastic modulus of the material. After onset of expansion this wave speed also determines a correlation horizon'. This correlation horizon determines at any time t a region of the material in which material points are within an elastic communication distance. Within the correlation horizon elastic stresses can adjust to modest variations in elastic modulus and concentrate stresses at points of weakness. A fundamental, and logically reasonable, precept of the theory is that, if fragmentation occurs within a time t , then fragments can be no larger than the region determined by the correlation horizon ct. A straightforward calculation provides a lower-bound fracture surface energy density over the body of,
u, = 3 r / c t ,
(6)
where r is the fracture surface energy per unit area generated in the event. For example, if the body is a metal with fracture toughness K c ,then a reasonable estimate of the fracture energy is provided by I- K;/2pc2 . A plot of both the elastic strain energy in Equation ( 5 ) that will fuel the fracture process, and the resisting fracture energy in Equation ( 6) that will be supplied by the strain energy, are plotted as a fhction of the correlation distance (or time) on the right of Figure (10). The correlation horizon as the body expands is also illustrated in the picture on the left. The strain energy is for this example an increasing quadratic function of time. The resisting fracture energy required to fragment the body is correspondingly a decreasing function of time. Fragmentation is not allowed until the energies are equal. Equality occurs at a correlation distance identified as A in the figure. Although certainly not required, an assumption of the present energy-based theory is that equilibrium fragmentation occurs when the two energies are nominally equal and the correlation length A provides the scale factor constraining the fragment size distribution. For the exponential distributions describing the fragment
-
Advances in Ceramic Armor V
.
11
Fragmentation of Ceramics in the Ballistic Environment
-
size data in Figure ( 6 ) the mass scale is p pA’. By equating the energy relations provided in Equation ( 5 ) and (6),an analytic expression,
for the governing fragment size length scale is directly calculated from both fracture properties of the material and a strain rate measure of the intensity of the fragmentation event. Application of Equation ( 7 ) to the tests provided in Figure ( 6 ) based on properties for stainless steel and kinematic properties of the fragmentation event provide size scale factors that are in remarkably good agreement with the experimental results. Similar successes of the theory have been demonstrated in numerous other applications in dynamic fragmentati~n’~. NONEQUILIBRIUM (BRITTLE) FRAGMENTATION Nonequilibrium fragmentation, in contrast to equilibrium fragmentation, occurs in those materials that, when subjected to dynamic fragmentation conditions appropriate to the plot in Figure (lo), do not fail at the juncture of the two energy curves. Instead, elastic strain energy continues to increase until some other failure criterion is achieved. In the present context this behavior applies to glass and to certain very competent ceramics such as boron carbide. As noted earlier, nonequilibrium fragmentation most probably occurs to some extent in numerous other materials. This interpretation of the energy-based fragmentation theory is less mature and has not yet been extensively e~plored’~”’. Theoretical Basis The comparable energy plot for nonequilibrium fragmentation is illustrated in Figure (1 1). When strain energy is achieved that is necessary to fuel the required fracture energy, all conditions sufficient for failure and fragmentation are not present. Strain energy, correlation length, and time continue to increase until an alternative criterion is achieved that provides both the necessary and the sufficient conditions for failure. Nonequilibrium
P
\A
I
fi
Energy Ded Exit Length Scale
Fragmentation
Time or Correlation Distance
Figure 11. Nonequilibrium conditions leading to a schism in length scales spanning the fragment size distribution in brittle solids. An application of nonequilibrium fragmentation perhaps most easily visualized is that of spall in glass. In the spall process release waves interact and carry regions of the glass into tension at a high rate of strain (of order 10s/s to 106/s). Because of the very modest energy required to create new fracture surface in glass, strain energy necessary for failure and fragmentation occurs at a tension of a few tenths of a GPa. Experimental studies show, however, that tensile stresses in the neighborhood of several GPa (one order of magnitude higher) are achieve before the spallation of glass. Reasons for the markedly high spall strengths of glass presumably relate to the very modest defect structure in the
12
.
Advances in Ceramic Armor V
Fragmentation of Ceramics in the Ballistic Environment
material, and the stress and stress correlation states that must be achieved to activate internal cavitation. Although more complex, similar nonequilibrium fragmentation governs failure in the ballistic impact tests on boron carbide ceramic target plates for which the fragment size distribution data are shown in the previous Figure (5). Namely, elastic strain energy achieves levels that far exceed the required fracture energy before failure and fragmentation proceeds. Again, the material defect structure, joined with the elastic strain energy and strain energy history, combine to determine the elevated nonequilibrium state at which failure begins. The correlation length scale A< at onset of failure determines the initial fabric of fracture. Fractures on a spacing of this characteristic length scale, however, are not sufficient to dissipate the excess elastic strain energy stored in the body preceding failure. Fracture failure proceeds through successive crack branching until available strain energy is exhausted and a fracture fabric is achieved commensurate with the fracture surface energy length scale Ae identified in Figure (1 1). In summarizing to this point, certain competent brittle solids do not exhibit equilibrium fragmentation governed by a single length scale A occurring at the juncture of the elastic strain energy and fracture resistance energy curves as shown in Figure (10). Such brittle solids instead exhibit nonequilibrium fragmentation governed by an alternative, markedly higher energy, failure criterion that leads to a schism in length scales as illustrated in Figure (1 1). The correlation length scale .1, and surface energy length scale .I, can be decades apart in highly nonequilibrium dynamic fragmentation applications. Much of the dynamic fracture and fragmentation physics is enveloped by the surface fracture energy and correlation length scales A# and ,Ic. Onset of failure and initial fracture is governed by the correlation length scale A<. Continued fracture, through successive crack branching, cascades down Over a range of length through the length scales until strain energy is exhausted at the length scale ,Ie. scales bounded by A* and A,,the physics of fracture is independent of length scale. Physical processes become self-similar within this range and physical symmetry dictates that such processes are described by power-law functions of the length scale A* < x < .1.,. The general concepts have application to other areas of physics including hydrodynamic turbulence'8 and fatigue fkacturelg. This physics has been referred to as a law ofintermediate asymptotics. In particular, within this range of length scales the resulting distribution of fragment sizes is necessarily power-law. Expressed as a cumulative fragment number distribution greater than size x , the distribution is written in the form, N(x) Y d , (8) where the exponent d is the fractal dimension of the distribution2'. The distribution is readily transformed to the Schuhmann mass distribution, M ( x ) = M,X", (9) where the Schuhmann index n is related to the fragment size fractal dimension2' through n = 3 - d . A complete fragment distribution spanning the upper and lower length scales would, of course, be functionally more complex, but must be asymptotic to power-law behavior in the intermediate range. The energy-based theory of nonequilibrium fragmentation provides a qualitative explanation for the ballistic impact generated fragment size distribution for boron carbide ceramic shown in Figure (5), as well as many other applications of dynamic fragmentation of brittle solids. The data are clearly described by a power-law function that span a finite range of fragment sizes. A more quantitative assessment of fragmentation requires both a clearer description of the impact conditions and a criterion for failure under the dynamic loading imposed.
-
Advances in Ceramic Armor V
.
13
Fragmentation of Ceramics in the Ballistic Environment
Experimental Observations The fragment distribution data from Figure ( 5 ) for ballistic impact tests on boron carbide ceramic" are repeated in Figure (12) and compared with a qualitative description of the fragment distribution inferred from the theory. The character of the experimental fragment size distribution is in sensible agreement with the predicted theoretical behavior. The measured sieve data are consistent with a power-law distribution. It can be safely inferred that the experimental distribution exhibits an upper and lower bound. A number of hrther questions, however, are raised concerning the theory. How are the correlation and energy length scales that determine the range of power-law behavior calculated? What is the appropriate dynamic failure criterion in the nonequilibrium fragmentation event? What properties of the material and/or the fragmentation event determine the exponent n in the power law? What is the span of the power-law range? Although not yet fully understood, these and related issues are discussed in the closing topics of this manuscript. 0 865or926Ms 793ar814Lm/s
o m n Carb
Id,
'
lo ' ' Fragment Size (mm) ' " " " r
, . . , L J
10
I
I Fragment Size
Figure 12. Experimental fragment size data for boron carbide ceramic" and a qualitative comparison with the functional character of the nonequilibrium fragmentation theory. Brittle Failure Criteria Energy-based fragmentation of brittle solids under equilibrium conditions provides a failure criterion firmly founded on commonly available properties of the materials. Failure occurs at the juncture of the elastic strain energy and fracture energy curves shown in Figure (lo), allowing direct calculation of the governing fragment size length scale as well as both the time and stress to failure under the loading conditions imposed. Nonequilibrium fragmentation enters a realm in which the failure properties of brittle solids are not well understood. In the brittle solids of interest elastic strain energies under the loading conditions imposed achieve dynamic levels that far exceed equilibrium values before failure and subsequent fragmentation occurs. Failure activates at internal or surface sites of weakness under stress and time conditions that are poorly understood. A fust-order assessment of the Figure ( 5 ) boron carbide experiments within the present fragmentation model is achieved by assuming that a nonequilibrium elastic strain energy commensurate with the Hugoniot elastic limit ( ok, 15 GPa ) is attained within approximately one transit of the plate thickness before failure and fragmentation proceeds. These conditions would place the Correlation length scale ,Icof the order of the plate thickness ( A c 5 mm). The lower energy length scale is calculated,
-
-
,I< =3(K
-
(10)
-
and, with a reasonable fracture toughness of K , 5 MPa m"2,provides I., 0.3 pm . The two length scales clearly span the power-law distribution displayed in Figure ( 5 ) .
14
8
Advances in Ceramic Armor V
Fragmentation of Ceramics in the Ballistic Environment
Computational simulations of the time history of elastic strain accumulation in the impact event are readily performed. The span of power-law fragmentation would be provided through the theory if a credible failure criterion can be applied within the computational model. A failure criterion that has achieved some success in the fracture of brittle solids, as well as other critical state phenomena, is that attributed to Tuler and Butcher2’. The Tuler-Butcher failure criterion can be expressed in the integral form, I, =
I(a@)- a,h)mdf5 K, .
(1 1)
The stress applicable to the applied loading conditions is a , while a,his a corresponding threshold value of the stress. Failure occurs when the time-dependent Tuler-Butcher integral achieves the material constant K, . The exponent m is a property of the material. For m = 1, Equation (1 1) provides an impulse criterion, whereas for rn = 2 the same equation is an energy or work criterion. As m becomes large Equation (1 1) approaches a constant stress failure criterion. When fit to experimental data an exponent m close to two is frequently observed. Impact breach tests on plates of soda-lime glass of Sun et a1.22, for example, are described by a Tuler-Butcher failure criterion with m = 2 for the Tuler-Butcher index23. The Tuler-Butcher criterion is appropriate for analytic models of ballistic failure where the stress within the inte a1 is a sensible average over the elastic strain energy contribution to the catastrophic fragmentation2 . The Tuler-Butcher criterion also has application in computational simulation.
F
Fragmentation Intensity Number In the theory of hydrodynamic turbulence the magnitude of the Reynolds number provides a measure of the extent of the inertial range - the span of length scales encompassing the entry level system scale powering the hydrodynamics, down to the limiting substructure scale necessary for the viscous dissipation. A similar dimensionless number characterizing the self-similar power-law range of the fragmentation event can be constructed by the ratio of the limiting correlation and energy length scale, F = Ac/Ae . (12) If the kinematic state of the body leading to failure can be characterized by a single strain rate B , then the limiting correlation length A< and the stress at failure of the body are related through, a = pc&Ac. (13) The energy limiting lower-bound length scale A< is related to the failure stress through,
where
r
is the fracture surface energy. Solving for F = .2, /Ae yields,
In a complex body (a projectile impacting a brittle plate for example) a provides a stress measure of the stored elastic strain energy at the onset of failure. For fragmentation intensity number F in the neighborhood of unity the stress can be solved for,
Advances in Ceramic Armor V
.
15
Fragmentation of Ceramics in the Ballistic Environment
This is the failure stress under conditions of equilibrium fragmentation. The intensity number increases rapidly with stress as failure criteria in excess of equilibrium are achieved. Correlation Horizon The concept of a correlation horizon’ and the meaning of this correlation within the present context of dynamic fragmentation is worthy of additional thought. Consider the thin-walled spherical shell of metal as representative of the model depicted in Figure (10). The model assumes abrupt impulsive loading up to a uniform outward velocity V . In principle this initial requirement is physically reasonable. A step pressure of magnitude P = pcV could be applied uniformly over the inner surface of the shell. A shock would propagate through the thickness d of the shell, reflect from the free surface and return a release wave. If the step pressure is maintained for a time Z d i c , the impulse conditions sought are achieved. (Elastic-plastic behavior is ignored in this discussion.) As outward motion proceeds, a uniform circumferential tension increases with time; the stress underlying the strain energy of Equation ( 5 ) . In a continuum first-order assessment this timedependent tension is independent of position throughout the body. Stress correlation plays a role in all real materials in that the modulus of elasticity is a h c t i o n of circumferential position about the shell. Any number of practical reasons can be responsible for this position dependence. The fluctuation with position would be quite subtle in a glass material for instance. It would be markedly more pronounced in other engineering materials. Thus, the correlation of circumferential stress is time dependent and determined by the elastic wave speed. A weakest point (in the sense of fracture activation) in the body within a given correlation horizon may not necessarily be the weakest point at a lower loading rate where the correlation horizon is correspondingly larger. The necessity of a correlation distance, and a corresponding correlation time, is evident in the direct imaging of impact fracture and fragmentation experiments on glass and ceramic plates of Straussberger et a ~ ‘ ~In. those tests the failure-induced stress wave is observed to propagate radially outward from the point of impact well ahead of any observed onset of fracture damage. These data are a clear example of the need for time-dependent stress correlation and concentration at preferred sites in the body to bring about brittle fracture and fragmentation. Power-Law Fragmentation Fundamental physics demands that the fragment size distribution within the range of the two bounding length scales .2, and At, and sufficiently removed from either, be a power-law function of the fragment size. This physics is, of course, based on the proposed continuum fragmentation theory and the absence of any further intervening physical length scales. The same physics does not constrain the power-law exponent, however. Assessment of the exponent requires other physics be brought to bear. The power-law nature of fragment size distributions resulting from the breakage of brittle solids has been noted by many since early in the last century, and probably earlier. Empirical evidence place the exponent n in the Schuhmann relation M ( x ) X” within the range of about 0.5 < n < 1.5 , with some applications achieving values approaching two and perhaps somewhat higher. There are studies in which values of n remarkably close to unity are observed25.A number of the earlier studies invoked the random placement of fracture flaws (a Poisson process) and some random geometric partitioning of the body. Notable are the theoretical efforts of Gaudin26,Bennett”, LienauZ8and G i l ~ a n y ~ ~ . An intriguing alternative theoretical effort within this time frame was explored by Griffith”. He
-
16
Advances in Ceramic Armor V
Fragmentation of Ceramics in the Ballistic Environment
suggested an energy argument in which a specific energy & - y l p x is associated with a fragment of size x. The fracture surface energy and specific density are y and p , respectively. He then invoked classical statistical mechanics to arrive at a Boltzmann representation of the fragment size distribution. In the large fragment limit a power-law distribution with n = 1 is achieved. Arguments are also offered for distribution in which n # 1. Turcotti2' has pursued the fragmentation of brittle solids as a fractal process leading to N x - ~ or , correspondingly to M - x" = x ~ - ~relating , the exponent n to the fractal dimension d . Turcotti explored group renonnalization methods used by others to characterize various scale invariant critical state phenomena. He identifies a parameter p , that determines the probability of fracturing of a cell of the solid body. The fractal dimension d is calculated from p m , which he in turn relates to the fragility of the brittle material.
-
CONCLUSIONS A theory of the dynamic fragmentation of solids based on continuum energy principles has provided a basis for assessing fragmentation in a wide range of ballistic applications over the past several decades. Applications of the theory to the fragmentation of brittle solids including glass and ceramic have been problematic, however. Recently, some of the physics issues governing length scales and sized distributions in the dynamic fragmentation of brittle solids have come to light. The earlier energy-based fragmentation theory is being broadened to accommodate dynamic fragmentation in brittle materials. The paper summarizes past theories and their applications to fragmentation in the ballistic environment. More recent applications to ballistic fragmentation in glass and ceramic materials are described. The principal conclusion from this study is that a previous continuum energy-based fragmentation theory has application to both equilibrium (principally ductile materials) and nonequilibrium (brittle materials) fragmentation. The latter application, however, requires an additional failure criterion which is, at present, not well understood. REFERENCES 'D.E. Grady, Local Inertial Effects in Dynamic Fragmentation, J. Appl. Phys., 53,322-325 (1982). 'D.E. Grady, The Spa11 Strength of Condensed Matter, J. Mech. Phys. Solids, 36,3,353-384 (1988). 3N.F. Mott, A Theory of the Fragmentation of Shells and Bombs, United Kingdom Ministry of Supply AC4035,May (1943). 4N.F. Mott, Fragmentation of Shell Cases, Proc. Royal Soc., A189,300-308, January (1947). 5L. Chhabildas, W. Reinhart, L.T. Wilson, D.R. Reedal, D.E. Grady, J.W. Black, Fragmentation Properties of AerMet 100 Steel in Two Material Conditions, Proceedings 19th International Symposium on Ballistics, Interlaken, Switzerland, I.R. Crewther, ed., 663-670, May 7-1 1 (2001). 6D.E. Grady, Analysis of Prompt Fragmentation in Explosively-Loaded Uranium Cylindrical Shells, Sandia National Laboratories Technical Report, SAND82-0140,February (1982). 'W. Greiner and H. Stocker, Sci. Am. 252,76 (1985). *D.E. Grady, Particle Size Statistics in Dynamic Fragmentation, J. Appl. Phys. 68, 12, 6099-6105 (1990). 9R. Schuhmann, Principles of Comminution, I., Size Distribution and Surface Calculations, AIME Tech. Publ. 1189, Mining Technologv, 1-1 1 (1940). "T.J. Moynihan, J.C. LaSalvia, M.S. Burkins, Analysis of Shatter Gap Phenomenon in a Boron CarbideiComposite Laminate Armor System, Proc. Int. Ballistics Symp., Sept. (2002). "D.E. Grady, C.A. Hall, W.D. Reinhart, Sandia National Laboratories Technical Memorandum, unpublished, (1996). "D.E. Grady, Fragmentation of Rings and Shells, Springer, (2006).
Advances in Ceramic Armor V
. 17
Fragmentation of Ceramics in the Ballistic Environment
I3N.F. Mott and E.H. Linfoot, A Theory of Fragmentation, United Kingdom Ministry of Supply AC3348, February (1943). I4D.E. Grady, Dynamic Fragmentation of Solids, in Shock Waves Science and Technology Reference Library Vol. 3, Y. Horie, ed. (2009). I5D.A. Shockey, L. Seaman, D.R. Curran, in Metallurgical Effects at High Strain Rates, R.W. Rohde et al., eds., Plenum (1973). I6D.R. Curran, L. Seaman, D.A. Shockey, Dynamic Failure of Solids, Physics Repts., 147, 253-288 (1987). "D.E. Grady, Fragment Size Distributions from the Dynamic Fragmentation of Brittle Solids, Int. J. Impact Engng., 35, 1557-1562 (2008). I8G. Falkovich and K.R. Sreenivasan, Lessons from Hydrodynamic Turbulence, Physics Today, 59, 4, 43-49 (2006). I9G.I. Barrenblatt, Scaling Phenomena in Fatigue and Fracture, Int. J. Fracture, 138, 19-35 (2006). *@D.L. Turcotte, Fractals and Fragmentation, J. Geophys. Res., 91, 1921-1926 (1986). *'F.R. Tuler and B.M. Butcher, A Criterion for the Time Dependence of Fracture, Int. J. Fracture Mech., 4,431-437 (1968). 22X.M.Sun, A. Khaleel, R.W. Davies, Modeling of Stone-Impact Resistance of Monolithic Glass Ply Using Continuum Damage Mechanics, Int. J. Damage Mech, 14, 165-178 (2006). 23D.E.Grady, Analysis of Shock and High-Rate Data for Ceramics: Failure and Fragmentation in the Shock and Ballistic Environment, Prepared for, U.S. Army TACOM-TARDEC, Applied Research Associates Tech. Rept., April (2008). 24E. Straussberg, P. Parimal, J.W. McCauley, C. Kovalchick, K.T. Ramesh, D.W. Templeton, HighSpeed Transmission Shadowgraphic and Dynamic Photoelasticity Study, Army Research Laboratory Rept., ARL-RP-203, March (2008). 25 H.C. Bergstrom, C.L. Sollenberger, W. Mitchel, Energy Aspects of Single Particle Crushing, Trans. AIME, 220, 367-372 (1961). 26 A.M. Gaudin, Investigation of Crushing Phenomena, AIME Trans., 73,253-316 (1926). "J.G. Bennett, Broken Coal, J. Inst. Fuel, 10,22-39 (1936). 28C.C.Lienau, Random Fracture of a Brittle Solid, J. Franklin Inst., 221, 485-494, 674-686, 769-787 (1936). 29 J.J. Gilvarry, Fracture of Brittle Solids. I. Distribution Function for Fragment Size in Single Fracture (Theoretical), J. Appl. Phys., 32, 391-399 (1961). 30L. Griffith , A Theory of the Size Distribution of Particles in a Comminuted System, Canadian Journal ofResearch, 21A, 6,57-64 (1943).
18
8
Advances in Ceramic Armor V
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
FLOW BEHAVIOR OF GLASS AT THE TIP OF A PENETRATOR
D. A. Shockey, D. Bergmannshoff, D. R. Curran, and J. W. Simons Center for Fracture Physics, SRI International Menlo Park, CA, USA ABSTRACT Projectiles penetrate glass armor by comminuting material at the advancing tip and forcing the fragments to flow out of the projectile path. Thus, fracture and fragment flow resistance of the glass under high pressure and shear stress control armor penetration. This paper describes a test to observe and measure the shear response of glass powder in a stress environment that simulates conditions in front of a penetrator. The data and observations provide a basis for developing physics-based models useful for computational simulations of penetration scenarios. INTRODUCTION The response of transparent armor to projectile attack cannot currently be computed with ~0nfidence.I'~ One reason is the lack of an adequate mathematical description of the failure processes occurring in the target material under high pressure and shear. Because it is difficult to generate a physics-based model, simulations typically use models from computer libraries that capture some of the important response mechanisms with values for model parameters that are selected to provide agreement with specific ballistic test data. When models and model parameters are chosen to achieve a match with the specific results they are intended to predict, they cannot be relied on to predict the outcome of arbitrary ballistic scenarios. To have general applicability, computational simulations require a model that describes the failure response of the target material that is activated by projectile penetration and that controls penetration resistance. The penetration phenomenology of frictional materials was clarified in recent re~earch.~.~ Sectioning of partially penetrated glass and ceramic targets revealed a highly comminuted region known as the Mescall zone ( M Z ) (after John Mescall, who surmised its existence from his computation^)^“ ahead of an advancing projectile and illustrated the mechanism of penetration. The pulverized material allows a projectile to advance by flowing laterally out of the projectile path and away from the penetrator tip. Resistance to penetration is thus provided by resistance to fracture and resistance to frictional flow of the fragments under high pressure. Thus an important step in achieving a model for use in simulations is to measure the flow resistance of M Z material and how flow is affected by fragment size and shape, pressure, and loading rate. To characterize the response of material in the M Z , a test is needed that (1) applies high pressure and shear to a bed of fragments in contact with a projectile surface, (2) allows pressure and shear stress to be varied independently, (3) allows the specimen to be recovered after the test to examine the change in bed density and in the number, size, and shape of fragments, and (4) allows for interruption of the shear load and unloading to observe and quantify fragment geometry at various shear strains. Such a test can reveal flow mechanisms, lead to damage evolution equations, and generate data that can be incorporated into constitutive models to achieve a failure-physics-based armor design capability. This paper describes an initial design of such a test, illustrates its use, and presents and interprets measurements of shear behavior of a bed of glass fragments as a function of pressure. EXPERIMENTAL Experiments were performed on G018-066 quartz glass powder obtained from Schott. Designated SM 3.5, the powder had a mean size distribution dso < 3.5*1 pm with 99% of the particles
19
Flow Behavior of Glass at the Tip of a Penetrator
being less than 13 pm. Material density was 2.2 g/cm3, Young’s modulus was 72 GPa, and the index of refraction was 1.46. Specimens were rings of powder about 2 mm deep with inner and outer diameters of 7.87 and 15.75 mm, respectively. The specimens were produced by pouring the powder into the annulus of the holding jig and tamping lightly to produce an even surface. Using an MTS axial-torsion machine, the specimens were loaded in combined compression and torsion by first pressing the powder with a mating steel ring to a desired normal load, then rotating the ring with respect to the specimen. The maximum normal load of 200 kN (44,000 Ib) produced a normal stress on the powder of 1.37 GPa (205 ksi). Figure 1 shows a schematic of the test.
Figure 1. Specimen, holding jig, and loading mode for compressing and shearing glass powder. Rotation angles of 10,45, and 90 degrees produced slide distances at the center of the specimen of 1.0, 4.6, and 9.3 mm, respectively. Sliding rates were varied from 0.05 m d s to 1.0 m d s . Torque was measured as a function of rotational angle. Shear stresses and strains and effective friction coefficients were computed from these data and the specimen dimensions. The shear stress reported here is the average stress over the width of the annulus; average shear strain was computed by dividing slide distance by the nominal specimen thickness, about 2 mm. Especially at higher values, what is referred to as shear strain is actually a combination of elastic shear strain in the glass particles, rearrangement and breaking up of glass particles, and sliding at the glassimetal interface.
RESULTS
Figure 2 shows shear stress as a function of shear strain under various normal forces at a displacement rate of 0.05 m d s . At all normal forces shear stress rose monotonically with strain until reaching a maximum and then remained relatively constant for higher strains. The slopes of all curves decreased with increased shear strain. At 20 kN, the curve was smooth. For higher values of normal force, large, sudden, periodic drops in shear stress occurred, likely due to slipping at the steel-glass interface. The values for peak stress at a given normal force were consistent for repeat tests; for 8 identical tests conducted at 100 kN normal load, the measured peak stress showed a standard deviation of 6%. The compressiodtorsion machine did not permit testing at ballistic rates. Stress-strain curves from tests in which the specimen was rotated 10 degrees and 90 degrees in 1 to 180 s produced shear
20
.
Advances in Ceramic Armor V
Flow Behavior of Glass at the Tip of a Penetrator
displacement rates of fiom 1 to 0.05 d s . Results from these tests differed little, and thus rate effects in this range were small. However, for tests performed at the highest rates, the effects of the stress drops overlapped and peak values were difficult to determine from the records. 0.35
0.30
0.25 A
2
2
* 0.20
v)
W
F
(I)
LI 0.15
5
ih 0 10
0 05
0 00 000
005
010
015
020
025
030
035
040
SHEAR STRAIN
Figure 2. Shear stress vs. shear strain as a fhction of pressure. Post-test specimen examination Specimens were examined post-test with optical and scanning electron microscopy 3 elucidate deformation and flow mechanisms and to seek explanations for the shapes of the stress-strain curves. Specimens tested at low normal loads and low shear strains remained in the initial loose-fragment state; fragments were generally straight-sided with sharp edges and roughly equiaxed, and thus had shapes similar to those of the original powder particles. At higher normal forces, shearing of the glass particles significantly altered the bed microstructure. Specimens tested at 200 kN, however, consisted of adherent aggregates of fragments (Figure 3a). The individual fiagments in the aggregates tended to be less angular than those in the compressed-only powder (Figure 3b).
Advances in Ceramic Armor V
.
21
Flow Behavior of Glass at the Tip of a Penetrator
(a) (b) Figure 3. (a) Quartz powder after 200 kN compression followed by shear. (b) Quartz powder after 200 kN compression only. The fragment size distribution of selected specimens was measured with a laser-based particle classifier after the tests. Figure 4 shows the effect of normal load on the flagment size distribution for tests in which the slide distance was 9.3 mm. No change from the original size distribution was observed in tests at 10 kN and 50 kN (normal stresses of 69 MPa and 342 m a ) . However, a normal load of 200 kN (1.37 GPa) narrowed and shifted the size distribution from a mean of 3.5 pm to about 2 pm, showing that larger particles were fractured and eroded.
Figure 4. Size distribution of 3.5 pm quartz glass powder after testing at normal loads of 10 kN, 50 kN, and 200 kN (normal stresses of 69 MPa, 342 MPa and 1.37 GPa). Polished cross sections through tested specimens on planes normal to a specimen radius showed cracks in two general orientations-parallel to the anvil surface, and angled at 20 to 60 degrees to the specimen axis. Left-quadrant cracks in Figure 5 formed before the right-quadrant cracks, since right-quadrant cracks terminate at the left-quadrant cracks. The cracks may have propagated
22
Advances in Ceramic Armor V
Flow Behavior of Glass at the Tip of a Penetrator
intermittently, arresting and reinitiating numerous times before reaching other cracks or specimen boundaries. It is not known if the cracks were produced upon unloading or by sectioning and polishing.
Figure 5. Polished cross section normal to the specimen radial direction showing cracks. The surfaces of the glass specimens in contact With the steel anvils were smeared and cracked (Figure 6), attesting to the sliding of the steel anvil on the glass. The directionality of the smear markings and crack profiles indicate a left-to-right movement of the anvil on the glass surface. The dried-lake-bed cracking pattern suggests a discontinuous sliding process, where adherence between glass and steel suddenly was overcome in a local area and the area was suddenly displaced.
Figure 6. Surface of glass specimen next to steel anvil showing smear markings and interface stickislip evidence. These post-test observations indicate four mechanisms of shear flow and suggest the following explanation for the shapes of the curves in Figure 2. For initial small strains the slope of the stress-strain curve probably measures the elastic shear stiffness of the bed of fragments as they deform in shear, but do not move. As the shear stress becomes higher, other mechanisms come into play that decrease the stiffness but may increase the strength of
Advances in Ceramic Armor V
.
23
Flow Behavior of Glass at the TiD of a Penetrator
the bed. One mechanism is movement of the particles, leading to shear-enhanced consolidation; shear strain is taken up by individual fragments moving into adjacent interstices. As particles move, the interstitial spaces likely become fewer and smaller and the individual fragments increasingly lock up, requiring higher stress to shear the bed. Under still higher normal loads, a second straining mechanism, particle comminution, becomes active. The larger particles that break and have their corners and edges broken off enhance the particles rearrangement mechanism, because the new smaller particles can fit into smaller interstices. A third mechanism for accommodating shear strain is slippage between specimen and anvil. When the shear strength of the bed exceeds that of the specimedanvil interface, bed deformation ceases and subsequent shearing occurs by sliding of the anvil over the fragment bed. Glass particles at the anvil interface are initially in point contact, and hence present a small effective contact area over which friction can act. As shear stress is increased, frictional strength is overcome and these contact areas slip, resulting in measurable shear strain and a flattening of the stressistrain curve, but also increasing the contact area and hence requiring higher shear stress for the next increment of displacement. Thus, even at low shear stresses and strains, the specimedanvil interface can slip as particles adjacent to the interface rearrange. The load drops that become more prevalent at high shear stresses and normal loads are predominantly due to interface stick slip. The phenomenon is clearly audible during the tests. At shear strains above 0.25, the average shear stress is nearly constant as the steel anvil slides over the compacted fragment bed. The load drops are manifestations of periodic stick-slip at the bed/anvil interface. Based on the peak shear stresses at the different values for normal stress, the measured coefficient of friction between glass and steel was 0.56 and independent of normal load for normal loads of 10 kN to 150 kN. At 200 kN the indicated friction coefficient was about 20% lower. The contribution to load drops from shear cracking in the specimens will be investigated in future tests in which specimens are subjected to progressively higher loads in the rising portion of the curves, then unloaded and examined microscopically. DISCUSSION The test described here has provided insight into the flow behavior and quantitative measurements of response for finely fragmented material in advance of a projectile penetrating a glass target. However, the test does not replicate the temperatures, rates, and pressures imposed by a penetrating projectile. Our fractographic examinations of partially penetrated glass blocks show that temperatures get high enough to melt a thin layer of glass adjacent to the penetrator and leave a residue of steel on the glass ~ u r f a c e Solidified .~ droplets and streaks of glass are observed on penetrator shafts, and previously softened and smeared glass sheet-like remnants are observed adhered to the penetrator nose. These observations suggest elevated local temperatures. A second difference is rate. The maximum sliding rate of the test, 1 d s , is several orders of magnitude less than that produced by a penetrator in MZ material. Third the maximum pressure produced in the test was 1.37 GPa, whereas pressures ahead of an advancing projectile may be several times this, depending on velocity. Finally, the tests are performed on beds of loosely poured fragments and thus the specimens differ in density and packing from the initially tight in-situ fragment beds at the tip of an advancing penetrator. However, the specimens may represent fragment packing configurations in the tunnel region alongside the projectile. The annular holding devices do not fully contain the fragments making up the specimens, allowing material to “leak out” during the test. Therefore “material properties” representative of the fragment beds are not measured. However, because the specimen geometry is identical, these edge effects are similar for all specimens and the trends in shear resistance with sliding distance, pressure, fragment size distribution, and rate are probably reliable.
24
8
Advances in Ceramic Armor V
Flow Behavior of Glass at the Tip of a Penetrator
The stresses measured at large (> 0.4) strains are indications of the glass-steel interface friction and hence may be useful to the modeler in describing the flow of glass fragments along the penetrator nose. However, observations of fragments in this location in partial penetration experiments indicate that the penetrator shaft is hot and the fragments are softened and even melted.4 Thus, friction coefficients measured here may not be characteristic of projectile penetration conditions. SUMMARY A test was developed to simulate the pressureishear load conditions experienced by Mescall zone material ahead of a penetrator advancing in a glass target. The test was applied to beds of glass fragments to observe changes in bed microstructure and fragment geometry, deduce the mechanisms of flow, and measure shear resistance as a fimction of normal load (pressure). Shear resistance of quartz glass increased monotonically with increasing normal force until it reached a constant maximum dictated by friction at the glassistee1 interface. Frequent, sudden, periodic load drops were prevalent throughout the tests, suggesting continuous flow and sticWslip of the glass on the steel anvil. The measured stressistrain relationships, their dependence on normal load, and the glass-steel coefficient of friction are data that can be used in developing physics-based models of material behavior and, hence, in computationally simulating penetration scenarios for glass armor. ACKNOWLEDGMENTS This work was performed for the Tank Automotive Command under subcontract to the Southwest Research Institute. The authors are grateful for the interest and support of Dr. Douglas Templeton and Dr. Charles Anderson and their staffs. REFERENCES 1. G. Johnson, and T. Holmquist, Some Preliminary Constitutive Models for Glass Subjected to HighVelocity Impact, Proceedings of the 32"d International Conference on Advanced Ceramics and Composites, Eds T. Ohji and A. Wereszczak, Amer. Ceram. SOC.,Wiley & Sons (2008). 2. Anderson, C. E., Presentation at the 32"d International Conference on Advanced Ceramics and Composites, (ICACC) Daytona Beach, FL (January 2008). 3. S. Chocron, C. E. Anderson, K. A. Dannemann, and A. E. Nicholls, A Preliminary Mohr-Coulomb Model for CTH to Predict Long-Rod Penetration into Borosilicate Glass, 33rd International Conference on Advanced Ceramics and Composites, Daytona Beach, FL (January 2009). 4. D. A. Shockey, A. H. Marchand, S. R. Skaggs, G. E. Cort, M. W. Burkett, and R. Parker, Failure Phenomenology of Confined Ceramic Targets and Impacting Rods, Int. J. Impact E n g 9(3), 263-275 (1990). See also Ceramic Armor Materials by Design, Ed. J. W. McCauley, et al., Ceramics Transactions 134,385-402 (2002). 5. D. A. Shockey, D. Bergmannshoff, D. R. Curran, and J. W. Simons, Failure Physics of Glass during Ballistic Penetration, Proceedings of the 32"d International Conference & Exposition on Advanced Ceramics & Composites (ICACC), Daytona Beach, FL, (January 2008). 6. J. Mescall and C. Tracy, Improved Modeling of Fracture in Ceramic Armor, Proceedings of the 1986 Army Science Conference, U.S. Military Academy, West Point (June 17-20, 1986). 7. J. Mescall and V. Weiss, Materials Behavior Under High Stress and Ultra-high Loading Rates-Part 11, Proceedings of the 29" Sagamore Army Conference, Army Materials and Mechanics Research Center, Watertown, MA (1984).
Advances in Ceramic Armor V
.
25
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
RHEOLOGY OF POWDER AND POROUS MEDIA IN MODELING OF PENETRATION INTO POROUS CERAMIC B.A. Galanov', V.V. Kartuzovl, and S.M. Ivanovl 'Institutefor Problems in Materials Science, NAS of Ukraine, Kiev, 031 42 Ukraine, Phone +380-44-424-0102, e-mail galanov@,ipms.kiev.ua
ABSTRACT In this paper, the concepts of mechanics of porous and powder media are applied for development of new analytical model of expansion of spherical cavity in porous brittle materials (ceramics). The model is based on the approach that recognizes the existence of three regions with different rheology: region of comminuted and compacted material; region of pore formation formed by radial cracks; elastically deformed region. Strain-stress state in each region is determined and analyzed. Condition of absence of pore formation region is specified. Cavity expansion pressure is determined. Energy losses on elastic deformation, fracture and compaction (non-elastic deformation) of material are calculated and compared for a number of materials. Penetration depth and target resistance depending on porosity is determined. INTRODUCTION It is well-known [l-31 that the models of expansion of spherical and cylindrical cavities in materials are useful tool for investigation of high-velocity penetration and hardness. Spherical cavity expansion model as a rule better describes high-velocity penetration. The extended review and analysis of existing models can be found, e. g., in Satapathy [ 2 ] .From discussion [ 2 ] it follows that simple and adequate analytical models of cavity expansion in porous brittle materials are still in demand. In this paper we present a new model of spherical cavity expansion in porous brittle material (ceramic) where the rheological models of porous and compressible powder materials based on approaches developed in [4-71 are used. Similar approach was used in analytical model of indentation and spherical cavity expansion in brittle non-porous materials [8,9]. Herein lies the main difference of the proposed model from existing ones. The developed model is used for an estimate of energy losses on elastic deformation, fracture and compaction of material during cavity expansion kom zero radius. The structure of these losses provides new additional information about material properties. A number of examples for some materials during cavity expansion are presented. Penetration depth and target resistance depending on porosity is determined for two ceramics. MAIN EQUATIONS OF THE MODEL Analytical model of cavity expansion in porous ceramic Figure 1 presents a geometrical scheme of spherical cavity expansion. According to model assumptions there are three zones with different stress-strain states of material: 1) r > c - elastic region (volumetric strain e = 0); 2 ) b < r < c - dilatation and pore formation (radially cracked) region (e > 0); 3) a < r < b - comminuted (pulverized) region (e < 0).
27
Rheology of Powder and Porous Media in Modeling of Penetration into Porous Ceramic
Figure 1. Model scheme of cavity expansion in spherical coordinates Or@. In elastic region r t c (see Fig. 1) solving equilibrium u(m) = 0 ; oV(r= c ) = oe(r = c) = o,[2,3,9], we obtain
equations with boundary conditions
where u - radial displacement; or,oV,oe - radial and hoop stresses respectively; stress or is a tensile strength of initial material with porosity 80; E, v - elastic constants for material with porosity 80 (initial porosity). Values oh E, v depend on 80. From (1) it follows that hydrostatic pressure and volumetric strain e are equal to 0. Cracked and dilated material of region 2 is formed at the boundary r = c from material in elastic region 1. Cracking and dilatation of material by radial cracks is accompanied by formation of pores. Radial cracks are assumed to be spatially distributed in such way that hoop stresses in region 2 are zero, i. e. oV = 0 0 = 0. Material in the region b 5 r 5 c is considered as an elastic continuum with uniaxial (radial) stress state. Fracturing takes place at the boundary r = b where stress-strain state satisfies fracture criterion and material state changes discontinuously. At the boundary r = b, stresses in the region 2 satisfy the following boundary conditions o,(r = b + 0 ) = o*< O , o V ( r =b + O)=oe(r = b+ O ) = 0,
(2)
At the boundary r = c stresses in this region are equal or(r = c - 0) =-2ofi o,(r
= c - 0) = oe(r = c -
0 ) = 0.
(3)
At the boundary r = c stress oris continuous and oV,0 0 are discontinuous. Thus, stress-strain state in the region of dilatation and cracking is defined by equations [2,3,9]
do,+*%=o, dr
r
du dr
o,=E--,
(4) (5)
I+v with boundary conditions (2) for (4)and boundary condition u =---oft for (5). Integrating (4)and ( 5 ) E we obtain
28
Advances in Ceramic Armor V
Rheology of Powder and Porous Media in Modeling of Penetration into Porous Ceramic
where u = u(r) - radial displacement of the material (u(r)- continuous function at r = c). Position of fracture front is found from (3), ( 5 ) and (7), taking into account continuity of radial stresses at r = c
1
Region 2 exists (i. e. c > b ) when 20,<(-0*) is true. If this condition is not fulfilled, then region 2 does not exist. In this case the results of work [lo] for porous plastic material can be applied for ceramics, if the fracture of porous ceramic is considered as "quasiplasticity". From this and from (7) for the displacement u of the boundary r = b and deformations in region 2 at r = b + 0 we have
E Since cracking leads to pore formation and volumetric strain rate e is connected to porosity change rate 6 by differential relation [6,7]
. 6 e=1-8'
d8 or de=--, 1-8
then porosity of material 8 in the dilatation region at r can be found from the integral of equation (10)
=b
+ 0 formed in the material after its fracture
1-8 e = l n L . 1-8
Thus, from (1 1) at e = e, for "total" porosity 9' of material at r = b + 0 we have 8' = 1 - (1 - eo)exp(-e,).
(12)
From this (at 8 0 = 0) for "induced" porosity of material 8, (i. e. formed in the region b < r < c during cracking in addition to 8,) we have
Advances in Ceramic Armor V
29
Rheology of Powder and Porous Media in Modeling of Penetration into Porous Ceramic
8, = 1 - exp(-e,).
(13)
Region 3 (a < r < b) consists of compressed fractured material which is formed from the cracked material of the region 2 at the front r = b where material of this region is comminuted (fragmented) by shearing stresses and then compacted. Therefore for the material of the pulverized region 3 we can use rheological models of compressible porous and powder materials. There are several constitutive equations of different complexity for such materials [4-71. We use the following constitutive equations for a material with porosity 8, yield stress of solid phase Y,at uniaxial compression, and porosity functions cp and y [4-71:
where T = m l o , - oq/ ;p = (o,+20T)/3.
First equation (14) is a condition of plasticity of porous body (in a space of variables lp I - T it defines flow ellipsis), second equation (14) is an associative flow law. Porosity functions cp and y~ are obtained both experimentally and theoretically [4-71. Here we use theoretical porosity functions proposed in [ 5 ] . The upper limit 8,,,=2/3 for the porosity is due to positivity requirement for the transverse deformation coefficient. Further, at the front r = b, the first equation (14) we consider as a condition o:fracturing (comminution) by shearing stresses of material of zone 2 with the porosity 0 = 8 = 1 -(1 Bo)exp(-e,), see (12). The second is considered as an associative law of fracture (compaction) of material of zone 2. In this case Y, is the strength limit for uniaxial compression of porousless material. Thus, fracture and compaction of material of zone 2 at the front r = b (i. e. formation of material of zone 3) we consider as a “quasiplasticity” of porous material. The strain state coefficient at the front r = b is s = e/y = -$@, which corresponds to compression in a spherical press mold [4-71. If we consider (14) at the fracture front r = b , take 8 = 8’ and eyclude q,,then we obtain (see (9), (12)) the following equation for determination of radial stress or= o < 0 at the front r = b
where Y,is the strength limit for uniaxial compression of porousless material. It is obvious that stress o*< 0 depends only on material properties and, therefore, is its characteristic. Non-linear equation (15) can be solved by method of successive approximations
30
.
Advances in Ceramic Armor V
Rheology of Powder and Porous Media in Modeling of Penetration into Porous Ceramic
Porosity Bk in the region 2 is assumed constant. Such material state is formed at the flow front r = b. =,(J* reach at the boundary Volumetric strain e and porosity B after compaction by the radial stress (J r = b - 0 values ek, 81 (see further). Material transforms to elastic compressible porous (porousless, if e k = 0) powder medium. From (14) at s = at the boundary r = b - 0 we have
-m
From this oq= (1- 1.58)0, or
- crlP= 1.580,~ or= 0'.At the boundary r = b hydrostatic pressure and
(J~
shearing stress intensity changes stepwise. Further, these relations (16) with B = 8k are assumed to be true for the whole region 3. Therefore equilibrium equations in the region 3 take form do, dr
3 2
-+2a3=0,
a=-B,,
r
o,(r=b)=o*,
with solution
From (18) follows, that at small Bk stress state in the region 3 is close to uniform compression or ;5 o ~ , and at large Bk - to uniaxial (radial) compression (o&rr << 1). This transition is quite natural. Let's estimate volumetric strain ek, porosity 6k of material in the region 3 and ratio a / b taking into account elastic volume compressibility of this region as an elastic porous body. Solid phase compressibility is not presupposed in (14), it is only due to porosity 14-71. Neglecting the density variation, ratio a / b is found from the mass conservation law: 4
4
4
4
-no3 =-nb3 - - - ~ ( b - u )+-nb3Ae ~ 3 3 3 3
or
a3 = b3 - ( b - u ) +~ b3Ae,
is the displacement of the boundary r = b; Ae = e, - ek is the jump of E volumetric strain at the front r = b, the volumetric strain e, at r = b + 0 is defined by formula (9) and volumetric strain ek at r = b - 0 is defined later by (21). Equality (19) with accuracy assumed earlier for the regions 1 and 2 can be represented in the form
Let's find now the volumetric strain ek at the boundary r = b - 0 caused by compression of porous material with porosity 8' (see (12)).
Advances in Ceramic Armor V
.
31
Rheology of Powder and Porous Media in Modeling of Penetration into Porous Ceramic
Since volumetric strains in the region 3 are assumed elastic, volumetric strain ek at the boundary
r = b - 0 is:
e, = p / K = o ' ( l - @ , ) / K ,
-=1
K
1 3+ 1 i 3 i Ks(l-ek) 4 ~ , ( 1 - 8 , ) ' '
where modulus of uniform compression K is found, for example, from [5-71 and K,, G, are modules of uniform compression and shearing of porousless material, respectively. The jump of volumetric strain 4 e at the boundary r = b is 4e=ec-ek
=ec-o'(l-ek)/K,
Ae=ec-o*/Ks forek<< 1,
(22)
Now, let us fmd e k . Strain e associated with compression of pores satisfies equation de = de/(i - e), e ( e = e,) = 8' ,
from which we obtain
e = 1 - (1 - B')exp(e,
- el.
From this equation, and estimating elastic compression of pores during compaction of porous material as e = ep = eke,= 8, o'(1- B,)/K, we have the following nonlinear equation for porosity €lk
8, =1-(1-8')exp(ec-8,e,),
e, = o ' ( l - e , ) l K .
Equation (23) always has a solution B k 5 8* (see Fig.2). For small e k its approximate solution is
e, = (1 + e,) 8' - e, .
Figure 2. Geometrical scheme for the solution of equation (23). Unlike porosity, e(r) in the region 3 is not constant and in accordance with (18) is:
32
.
Advances in Ceramic Armor V
(23)
Rheology of Powder and Porous Media in Modeling of Penetration into Porous Ceramic
Maximum volumetric strain is reached at r = a and its variation can be significant. Thus, from (18) for the pressure R,in the cavity we obtain
R,= -0. (b/aI2"= -o'V*-Qk ,
(26)
where G* is defined by ( 15) and v' - by (20). Now, similar to estimates obtained in the model of indentation and spherical cavity expansion in brittle non-porous materials [8,9], we give approximate formulas for analysis and estimation of work of pressure in spherical cavity spent on deformation, fracture and compaction during expansion from zero radius. For a cavity expanding from zero radius, the part of work spent on compaction from porosity 8' to porosity Ok in the region (Y 5 b ) is estimated by formula [9]:
The total work A , spent on elastic deformation during cavity expansion is [9]:
g = 3(1- 2 ~ , ) - 6 8 , (I -2v,)+
4.58;(1 - v,) , a = 1.58, ,
where E, v and Ep, vPare calculated, respectively, at 8 = 80 and 8 = & by formulas [6]: E=2(1+v)G, v=-
3K -2G G = G, (1 2(3K + G) '
-q2,
For the total work A/ spent on fracturing we have [9]
The work A of pressure R, spent on the formation of cavity with the radius a is: A=4/3xa3R, =-4/3na30*v'-Qk.
Advances in Ceramic Armor V
.
33
Rheology of Powder and Porous Media in Modeling of Penetration into Porous Ceramic
The derivation of closed approximate formula for a part of work A," spent on quasiplastic compaction of comminuted (powder, granular) material of region 3 requires knowledge of history of deformation (loading) of every elementary volume in the region 3 and subsequent integration over the whole region. Such straightforward determination of A," is rather complicated since the relation between stresses and strains in the region 3 is non-holonomic. Therefore for a part of work Ac" spent on quasiplastic compaction during cavity expansion a simple estimate of A,"/A can be made from A<'/ A = 1 - A,/ A - A, A - A; 1A .
(32)
Value A, = (A: + A,") can be considered a work spent on compaction (non-elastic deformation) powder material of region 3 with rheological equations (14). Figure 3 shows schematic stress-strain diagrams in material around the cavity. Penetration into porous ceramic Non-deformable projectile velocity u ( f )is defined by the ballistic equation (see [l 13): M Zdu = - p A ( P ) -du -
df
B(P)--C(P)R, PU2 2
, ~ ( 0=) V,,
(33)
where coefficients A , B, C depend on projectile shape and penetration depth P; rn is the projectile mass, p = p, (1 - 0,) is the target density, V, is the impact velocity. The right part of equation (33) determines total target resistance to penetration: the values
P-
B ( P) p u 2 , R, are respectively dynamic, kinematics and static components of this C ( P ) d t ' C(P) 2 resistance. Consideration of non-deformable (or elastic) projectiles is explained by that the target resistance for such projectiles depends on properties of projectile material less than for eroded ones. The case of penetration of eroded projectiles can be investigated similarly to [ 121. CALCULATIONS To illustrate the performance of the model, several key parameters were calculated for a number of materials. Table I contains mechanical properties of materials used in the calculations. Tensile strength of porous material of was determined by formulas (34)-(36) [13,14] and difference in resulting values of R, in all range of porosities (Fig. 4) proved to be very small. '5,
= 'sfs (l-O,)",
5 ',
34
-
=ofs2(1- rO'
Advances in Ceramic Armor V
m = 2,JO ; (Bal'shin)
(34)
(Skorokhod);
(35)
$i7&
Rheology of Powder and Porous Media in Modeling of Penetration into Porous Ceramic
Here q i s a tensile strength of nonporous material; Y is a radius of pores, a is a distance between them, M is an empirical parameter. The short review of such formulas is contained in [13,14]. TABLE I. MECHANICAL PROPERTIES OF SOLID PHASE N Y, ofi Target material 0 (GPa) (GPa) 1 A1203 (Coors AD995) [2] 2.62 0.262 2 AIN [2] 3.0 1.o
EX (GPa) 373 315
!JS
V,
0.232 0.25
(kg/m3) 3890.0 3260.0
TABLE 11. CALCULATED CHARACTERISTICS
It is interesting to observe that the dependence of R, on Bo is not monotonous (Fig. 4). In ref. [13] it is indicated that not all strength material dependencies on porosity are monotonous. The case in Fig. 4 should be considered in more detail. Target resistance R, calculated by formula (26) for typical ceramics A1203 and AIN as well as characteristic data calculated by the model for both materials are collected in Table I1 (see also Fig. 4). Dynamics of penetration (see (33)) of steel projectile with ogival head was investigated using the model of [ 111 for small and high porosities (see Figures 5,6). CONCLUSION A new model of spherical cavity expansion in porous plastic materials is proposed in which continuum rheological concepts of compressible powder and porous materials are used. The model assumes (on the base of experimental observations) that along the cavity expansion direction there exist three characteristic regions: comminuted and compacted material (volumetric strain e < 0); region of pore formation formed by radial cracks ( e > 0 ); elastically deformed (e = 0). From the model analysis follows that if the conditions 20/> (a') is true, then the region of pore formation is absent. In this case an appropriate adaptation of the model [lo] for porous plastic material can be used.
Advances in Ceramic Armor V
.
35
Rheology of Powder and Porous Media in Modeling of Penetration into Porous Ceramic
Figure 3. Diagrams of stress-strain state for expansion of spherical cavity. Stress-strain state was determined in each of these regions and cavity expansion pressure that plays important role in many penetration models was found. The estimate of energy losses for elastic deformation, fracture and compaction (non-elastic deformation) of material was made. The
36
Advances in Ceramic Armor V
Rheology of Powder and Porous Media in Modeling of Penetration into Porous Ceramic
investigation and evaluation of properties for two ceramics A1203 and AIN were carried out. Analysis shows strong dependence of target resistance on porosity and energy losses on compaction. Process of penetration is strongly influenced by the behaviour of target resistance.
I
10
I
I
I I
I
0.08
em 8
E
4
Y6 s' 4
0.04
2 0.0
0.2
0.4
90
0.6
0.8
1.0
0.04
0.0
Figure 4. Penetration resistance R, vs. initial porosity 90. A1203.
0.08 t, ms
0.12
0.16
Figure 5. Penetration depth P vs. time t for different initial porosities. Steel vs. A1203, V,= 1400 ds.
E
d
400
600
800
1000
v,,d
s
1200
1400
Figure 6 . Penetration depth P vs. impact velocity V,. Steel vs. A1203. ACKNOWLEDGEMENTS This work was partly supported by the TARDEC contract N62558-05-P-0330. The authors would like to acknowledge the initial impetus and continued interest in this work from William Gooch from the U S . Army Research Laboratory.
Advances in Ceramic Armor V
.
37
Rheology of Powder and Porous Media in Modeling of Penetration into Porous Ceramic
REFERENCES 'G. Ben-Dor, A. Dubinsky, and T. Elperin, Ballistic Impact: Recent Advances in Analytical Modeling of Plate Penetration Dynamics-A Review, Trans ASME, Appl Mech Reviews, 58,355-71 (2005). * S . Satapathy, Application of Cavity Expansion Analysis to Penetration Problems. IAT.RO136, Institute for Advanced Technology, University of Texas at Austin (1997). 'S. Satapathy and S.J. Bless, Cavity Expansion Resistance of Brittle Materials Obeying a Two-curve Pressure-shear Behavior, JAppr Phys, 80(7), 4004-12 (2000). 4M.B. Shtem, et a]., Phenomenological theories of powder pressing, Kiev, Naukova dumka, (1982). (in Russian) 'V.V. Skorokhod, Rheological foundations of theory of sintering, Kiev, Naukova dumka, (1972). (in Russian) %.V. Skorokhod and L.I. Tuchinsky, Plasticity criterion for porous bodies, Powder metallurgy, 11, 83-87 (1978). (in Russian) 'E.A. Olevsky, Theory of sintering: from discrete to continuum, Materials Science and Engineering, R23,41-100 (1998). 'B.A. Galanov and O.N. Grigor'ev, Analytic model of indentation of brittle materials, In: Electron microscopy and strength of materials, Proc. of the Institute of Problems in Materials Science NAS Ukraine, Kiev. Ser. "Phys. mat. science, structure and properties of materials", 13, 4-42 (2006). (in Russian) 9B.A. Galanov, V.V. Kartuzov, and S.M. Ivanov, New analytical model of expansion of spherical cavity in brittle material based on the concepts of mechanics of compressible porous and powder materials, Inf. J. Impact Engng, doi: 10.1016/j.ijimpeng. 2008.07.016 (2008). "B.A. Galanov, V.V. Kartuzov, and S.M. Ivanov, New analytical model of expansion of spherical cavity in porous plastic material and penetration into such materials, 24rd International Symposium on Ballistics, New Orleans, USA, 22-26 September 2008. p. 777-785. "B.A. Galanov, V.V. Kartuzov, and S.M. Ivanov, Numerical-analytical model of penetration of long elastically deformable projectiles into semi-infinite targets, Int. J. Impact Engng, 35(9) 1009-21 (2008). ''B.A. Galanov, S.M. Ivanov, and V.V. Kartuzov, On one new modification of Alekseevskii-Tate model for nonstationary penetration of long rods into targets, Int. J. Impact Engng, 26(1-10) 201-210 (2001). I3A.G. Kostomov, Material Science of Dispersible and Porous Metals and Alloys, v01.2, Kiev: Naukova Dumka (2003) [in Russian]. I4R.A. Andrievsky, A.G. Lanin, and G.A. Rymashevsky, Strength of refractory compounds. Moscow, Metallurgy (1974).
38
.
Advances in Ceramic Armor V
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
COMPUTER MODELING OF SHOCK WAVE PROPAGATION IN Sic - SAMPLE V.L. Bekenev, V.V. Kartuzov, E.V. Kartuzov, H.V. Hachatraian Institute for Problems of Materials Science Kyiv-142, 03680, Krzhyzhanovsky 3, Ukraine ABSTRACT. This effort is to present the results of computer modeling of shock wave propagation in cubic Sic samples with ideal and defective crystal structures. As interaction potential we used Tersoff s potential, which was well proved for this system. Sample's length was 100 unit cells and cross-section made 10 unit cells. In cross-section direction we imposed periodical boundary conditions. Shock waves with different intensity were initiated by application of constant force to atoms of the first two layers of the sample. Defects were modeled in a form of rectangular and disk-like cracks. INTRODUCTION Molecular dynamics is widely used for investigation of high-speed non-equilibrium processes appearing at high-energy affects on solids. As an example one may highlight a number of works where by means of non-equilibrium molecular dynamics these processes are investigated in metals and covalent crystals'-". Usually for metals embedded atom potentials are employed and for covalent crystals - potentials of Tersoff type". In this work we simulated propagation of shock waves in cubic silicon carbide by molecular modeling software XMD12 with the use of Tersoff s potential". A shock wave was generated in Sic sample in the crystallographic direction [IOO]. The sample had sizes of lOOxlOxl0 unit cells along x , y and z directions respectively and contained -80000 atoms. Along y and z directions the periodic boundary conditions were imposed. The sample was initially kept at constant temperature T=2 K during 1 ps to eliminate thermal fluctuations in propagation of shock wave. To obtain the shock wave an external force in the x direction was applied to two planes of atoms on one side of the sample. The force was kept constant throughout the simulation.
RESULTS
A simulation of propagation of shock wave was carried out at the followings intensities of loading: 5, 10, 20, 40, 60, 80, 100, 120, 140, 150, 160, 180, 200, 220, 260, 300, 340 and 380 GPa. Fig. 1 shows particles velocity V, along direction of propagation of shock wave depending on a distance iiom the beginning of sample for the moment of time T= 0.4 ps. Velocity profiles were analyzed at different moments of time that gave us an opportunity to calculate both the piston velocity, U p ,and the shock velocity, U , . At loading intensities more than 260 GPa is observed destruction of specimen and motion of particles opposite to shock direction.
39
Computer Modeling of Shock Wave Propagation in Sic-Sample
40
*
Advances in Ceramic Armor V
Computer Modeling of Shock Wave Propagation in Sic-Sample
On Fig. 2 three areas are selected: U p < 0.5 W s , 0.5 Ws
1.5 Ws. In the limit of a zero piston pressure shock wave velocity coincides with longitudinal sound speed'. Extrapolation of the obtained data for U p < 0.5 W Sgives a value for longitudinal sound speed 10.23 km/s that is in a good agreement with values 11.4 kmk and 12.06 km/s13. When U p > 1.5 km/s velocity of shock wave propagation stops being dependent on loading intensity. Usually a break of linear dependence U,(U,) witnesses a presence of structural phase tran~ition'~. To find out this issue a function of radial distribution of particles p ( r ) of free sample and sample after passing of shock wave with intensity of 120 GPa (Up =1.21 km/s) was built (Fig. 3). As it is seen a location of main peaks is not changed. The change of height of peaks and insignificant change of width point out on violation of near order, however the structure is saved. -1
-1
Fig. 3. Function of the radial distributing of particles. a) before passing of shock wave; b) after passing of shock wave Passing of shock wave was also investigated for samples with two types of defects, located in the middle of the sample: in a form of parallelepiped with sizes 1 0 x 2 ~ 2unit cells and in a form of disk with a diameter 10 and width 2 unit cells (Fig. 4). Deformation mechanisms of a crack in S i c under high-rate compression were investigated by molecular dynamics simulation using Tersoff potential". Two mechanisms of crack-tip response were found: a) at low tension, a disordered band is nucleated from the crack surface in the direction orthogonal to the compression; b) at a tension sufficient to allow the crack to remain open, the compression stress induces formation of disordered regions along the boundaries of the opened crack, which grow and merge into a band as the compression proceeds. Bending of the initial crack, which transforms into a curved slit, drives this process. In our case at passing of shock wave with rather high intensity (P 2200 GPa) there is an appearance of screw dislocation from a front top crack, and then from back ones. Distribution from the crack of the disordered regions is obviously asymmetric - in the front part of crack it passes more intensively, than in a back ones. A bend of crack is not observed in general.
Advances in Ceramic Armor V
41
Computer Modeling of Shock Wave Propagation in Sic-Sample
0.6 ps
0.8 ps
1.0 US
1.2 ps
Fig. 4.Shock wave passing through specimen at different times after the shock was started. Loading intensity is 200 GPa. A vertical line indicates the position of the shock wave front. In future it is supposed to continue similar investigations for silicon carbide samples oriented on crystallographic directions [I 101 and [ 1111. CONCLUSION In this work we carried out the computer modeling of shock waves propagation in cubic S i c by the method of non-equilibrium molecular dynamic with the use of Tersoff s potentials. Intensity of impact loading was varied in a broad interval from 5 to 380 GPa. Hugoniot adiabats were built in
42
a
Advances in Ceramic Armor V
Computer Modeling of Shock Wave Propagation in Sic-Sample
coordinates (piston's velocity - shock wave velocity)). Hugoniot adiabats have three areas of linear dependence with different angle of inclinations. At piston's velocities more than 1.5 kmis the velocity of shock wave propagation stops be dependent on piston's velocity. Propagation of shock wave through the samples with the defects in a form of rectangular and disk-like cracks was considered. Shock waves with intensity less than 200 GPa do not cause cracks deformation. At intensities 200 GPa and higher a propagation of disordered area from crack with formation of screw dislocation is observed. REFERENCES 'K. Kadau, T. C. Germann, P. S. Lomdahl, B. L. Holian, Microscopic view of structural phase transitions induced by shock waves, Science, 296, 1681-84,(2002). *S. V. Zybin, M. L. Elert, and C. T. White, Orientation dependence of shock-induced chemistry in diamond, Phys. Rev., B66,220102,(2002). 3E.M. Bringa, B.D. Wirth, M.J. Caturla et al, Metals far from equilibrium: From shocks to radiation damage, Nuclear Instruments and Methods in Physics Research, B202,56-63,(2003). 40.Kum, Orientation effects in shocked nickel single crystals via molecular dynamics, J. Appl. Phys., 93,3239-47,(2003). 'E. M. Bringa, J. U. Cazamias, P. Erhart et al, Atomistic shock Hugoniot simulation of single-crystal (2004). copper, J. Appl. Phys., 96,3793-99, 6D. S. Ivanov, L. V. Zhigilei, E. M. Bringa et al, Molecular dynamics simulations of shocks including electronic heat conduction and electron-phonon coupling, Shock Compression of Condensed Mutter 2003, edited by M. D. Furnish, Y. M. Gupta, and J. W. Forbes, 225-8,(2004). 'S. V. Zybin, I. I. Oleinik, M. L. Elert, C. T. White, Nanoscale Modeling of Shock-Induced Deformation of Diamond, Mat. Res. SOC.Symp. Proc., 800,AA7.7.1-AA7.7.6, (2004). '1.1. Oleynik, S.V. Zybin, M. L. Elert, and C. T. White, Nanoscale molecular dynamics simulaton of shock compression of silicon, Shock Coimpression of Condensed Mutter - 2005, edited by M. D. Furnish, M. Elert, T.P.Russel1, and C. T. White, 413-6,(2006). 9LI. Oleynik, S.V. Zybin, M. L. Elert, and C. T. White, Shear stresses in shock-compressed covalent solids, Shock Coimpression of Condensed Matter - 2005, edited by M. D. Furnish, M. Elert, T.P.Russell, and C. T. White, 417-20,(2006). ''A. Romano, J. Li, S. Yip, Atomistic simulation of rapid compression of fractured silicon carbide, Journal of Nuclear Materials, 352,224,(2006). "J. Tersoff, Modeling solid-state chemistry: Interatomic potentials for multicomponent systems, Phys. Rev., B39,5566-68,(1989).Erratum: Phys. Rev., B41,3248,(1990). "J. &kin, XMD, http://xmd.sourceforge.net, (2004). I3G. I. Kanel, S. V. Razorenov, A. V. Utkin, and V. E. Fortov, Shock Wave Phenomena in Condensed Media, Yanus-K, Moscow, (1996),[in Russian]. I4L. V. Altshuler, A. A. Papanova, Electronnaya struktura i szhimaemost metallov pri vysokih davleniyah, Soviet Physics-Uspekhi, 90, 193-215(1968)[in Russian].
Advances in Ceramic Armor V
.
43
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
BALLISTIC JMPACT DAMAGE OBSERVATIONS IN A HOT-PRESSED BORON CARBIDE J.C. LaSalvia, R.B. Leavy, J.R. Houskamp, H.T. Miller', D.E. MacKenzie, and J. Campbell US. Army Research Laboratory
AMSRD-ARL-WM-MD
Aberdeen Proving Ground, MD 21005-5069
ABSTRACT In an effort to better understand the connection between the inelastic deformation mechanisms in ceramics and their ballistic performance at a more fundamental level, observations of damage in a commercial hot-pressed boron carbide (B4C) impacted ballistically are reported. The ballistic targets consisted of B4C cylinders slip-fitted into titanium alloy (Ti6A14V) cups and welded cover plates. This target configuration provided lateral confinement for the B4C cylinders but zero-lateral pre-stress. Targets were impacted with laboratory-scale tungsten-iron-nickel (93YoW-Fe-Ni) long-rod penetrators between 800 m / s and 1600 m/s. In the limited number of ballistic experiments conducted, two apparent dwelVpenetration transition velocities (1060 m / s and 1200 d s ) were observed. The B4C cylinder impacted near the upper dwelYpenetration transition velocity was sectioned and polished to reveal sub-surface damage features. Both optical and scanning electron microscopy were utilized to examine the cross-section. The observed damage features are consistent with the low fracture toughness of this material. This suggests that the velocity range demarcated by the two apparent dweb'penetration transition velocities may actually be a wide transition zone due to stochastic variations in material and ballistic experiment parameters. The observed damage features are compared and contrasted with that previously reported for a commercial hot-pressed silicon carbide (Sic). INTRODUCTION The use of multiple flash X-rays and other high-speed imaging techniques to study the interaction between a projectile and a ceramic target can provide insight into temporal and spatial variations in fundamental ballistic performance measures and associated physics that is not provided by traditional VSOor depth-of-penetration (DOP) ballistic testing. This insight often becomes the genesis for greater hndamental understanding of the associated physics on overall ballistic performance. During the past 40 years, a number of notable fundamental ballistic studies on ceramics have occurredl-l'. Wilkens et al.' used multiple flash X-rays to capture the interaction of idealized smallann projectiles with simple ceramic targets. Using a single flash X-ray system to study the impact of long-rod penetrators on highly confined pre-stressed ceramic targets, Hauver et al.' were the first to observe the phenomenon known as "interface defeat". Burkett et al.3 also used a single flash X-ray system to study the penetration of highly confiied ceramics by long-rod penetrators. Orphal et al.4-5 used multiple flash X-rays to study the penetration of ceramics by long-rod penetrators at hypervelocities. Lundberg et al.7'82'0conducted several systematic studies of the interface defeat and dwelypenetration transition phenomena for a variety of ceramics in confined, pre-stressed targets and different penetrator characteristics using multiple flash X-rays. More recently, Behner et al.''s'3, Anderson et a ~ ' ~ , ' ~and - ' LaSalvia ~, et aI.l4 conducted studies using multiple flash X-rays to examine the effect of ceramic and target parameters on the dwelllpenetration transition phenomenon. LaSalvia et a I . l 4 examined the fimdamental ballistic performance of a commercial hot-pressed SIC known as PAD Sic-N". Figure l(a) shows the penetration rate as a function of impact velocity. Also plotted are the Sic-N results from Lundberg et aL8. The dwelVpenetration transition velocity for the LaSalvia et aI.l4 data is 1205 m / s while that for the Lundberg et aL8 data is 1507
45
Ballistic Impact Damage Observations in a Hot-Pressed Boron Carbide
ULaSalvia et al
DwelliPenetration
1000 1100 1200 1300 1400 1500 1600 1700 Impact Velocity ( d s )
(a)
(4
Figure 1. (a) Penetration rate as a function of impact velocity for SiC-N.I4 (b) Polished crosssection of Sic-N cylinder impacted at 1201 d s displaying numerous damage features.I4 (c) Damage in comminuted region. (d) Polished cross-section of Sic-B tile impacted at 1600 d s .
46
.
Advances in Ceramic Armor V
Ballistic Impact Damage Observations in a Hot-Pressed Boron Carbide
m/s. While differences in target configurations and penetrator characteristics do contribute to the overall 300 m/s difference in the observed dweltlpenetration transition velocities, it is hypothesized that the presence of a 175 MPa lateral pre-stress in the Lundberg et a1.' targets is primarily responsible for the difference. In an effort to confirm this hypothesis, LaSalvia et al.I4 examined polished cross-sections of recovered Sic-N cylinders from their experiments. While it was not possible to compare them with SIC-N cylinders tested by Lundberg et a1.' because they conducted reverse ballistic experiments (i.e. target launchei,at a stationary penetrator), it was possible to compare them to hot-pressed S i c (PAD SIC-B ) cross-sections recovered by Hauver et aL2. Figure l(b) shows the polished cross-section of a Sic-N cylinder impacted at 1201 m/s (0.5 mm from center of impact), while Figure l(d) shows the polished cross-section of a Sic-B tile impacted at 1600 m/s. The size of the comminuted region is significantly larger in the tile recovered by Hauver et al.' due to the higher impact velocity and larger penetrator diameter (6.35 mm versus 3.175 mm). The other main difference between the two cross-sections is the presence of long and steep cone cracks in the Figure l(b) that are absent in Figure l(c). This is believed to be an indication of the influence of lateral pre-stress on suppressing the growth of steep cone cracks. One might tentatively conclude that it is the formation of these steep cone cracks that leads to the lower dwell/penetration transition velocity observed by LaSalvia et aI.l4. With this limited understanding, it is natural to ask whether or not other ceramics such as B4C or sintered SIC, possess similar fundamental ballistic behaviors as Sic-N, and are governed by the same inelastic deformation mechanisms. Therefore, the purpose of this paper is to report qualitatively on the dwell/penetration transition behavior of a commercial hot-pressed B4C, as well as on its inelastic behavior as indicated by the various damage features. Comparisons to Sic-N will be presented.
EXPERIMENTAL PROCEDURES A commercial hot-pressed B4C known as PAD B4C (BAE Advanced Ceramics Division) was used in this studl. Measured physical and mechanical characteristics for PAD B4C and Sic-N are listed in Table I' 19. Figure 2 shows the microstructure for this materiaL2' The microstructure consists of equiaxed B4C grains and graphite particles of three general sizes. The largest graphite particles are on the order of the B4C grain size while smaller particles are located at grain boundaries and triple points. Chen et aL2' also showed that very fine sub-micron graphite precipitates exist within the B4C grains themselves. As in the previous study on SiC-NI4, the targets consisted of ceramic cylinders (38.1 mm diameter by 38.1 mm thick) slip-fitted (< +/- 0.05 mm) into Ti6A14V "cups" with a thin Ti6A14V cover plate (3.175 mm thick) welded to the top of the cups. The Ti6A14V cups were 76.2 mm thick by 50.8 mm in diameter, with a cavity machined into one end to accommodate both the ceramic cylinder and cover plate. Ten targets were constructed and tested. Laboratory-scale 93%W-Fe-Ni long-rod penetrators (L/D 20, 63.5 mm x 3.175 mm, p = 17.9 gkm') used previously were also used in this study. Sabot penetrator packages were launched from a powder gun at the B4C targets at velocities between 800 m/s and 1600 m/s. For additional information regarding the ballistic experiments and set-up, the reader is referred to reference 14. p Grainsize E HK4 Kc Phases (GPa) (GPa) (MPadm) (pm) (g/cm3) PADB4C 2.52 15 450 0.17 18.9 2.9 B4C,C PAD Sic-N 3.21 2 452 0.165 18.6 5.4 6H
Material
"
Advances in Ceramic Armor V
.
47
Ballistic Impact Damage Observations in a Hot-Pressed Boron Carbide
Figure 2. Microstructure of PAD B4C.” Prior to extracting the impacted B4C cylinders from the Ti6A14V cups, the recovered targets were statically x-rayed to determine the depth-of-penetration, and impact faces were examined optically. In order to keep the B4C cylinders intact and minimize the introduction of additional damage during the extraction and metallographic preparation processes, a number of epoxy resin steps were used. While still in the Ti6A14V cup, the impact face was filled with a clear epoxy resin. The rear of the cup was machined off and the exposed bottom surface of the B4C cylinder was covered with epoxy resin. The sides of the Ti6A14V cup were sectioned into a minimum of four pieces and removed from the B4C cylinder. The extracted B4C cylinder was encased in epoxy resin and sectioned to within 2 mm of the cross-section of interest. The cross-section was infiltrated with epoxy resin to keep exposed damaged ceramic intact. The cross-section was carefully ground to within 0.5 mm of the cross-section of interest. At this point, the cross-section was coated with a thin layer of epoxy resin and initial metallographic preparation steps were begun (15 pm, 6 pm, and 1 pm diamond). Final polish was performed using 0.25 pm diamond. All during this process of sectioning, grinding, and polishing, the cross-sections were examined in an effort to identify the possible introduction of preparation artifacts. The authors are confident that none were introduced. The polished cross-sections were then examined by optical and electron optical microscopy (NanoSEM 600, FEI). The low vacuum secondary electron mode on the NanoSEM 600 was used because it offered superior image quality under specimen charging conditions. RESULTS AND DISCUSSION Figure 3 shows static X-ray radiographs of B4C targets impacted at 1052 d s , 1063 d s , and 1198 d s . Radiographs for the targets impacted at 1052 d s and 1198 m/s show shallow penetration while that impacted at the intermediate velocity of 1063 d s shows significantly deeper penetration. The radiograph of B4C target impacted at 819 d s shows shallow penetration. Above 1200 d s , only deep penetration is observed. This is consistent with the behavior for Sic-N which showed penetration above 1205 4 s . Given the difference in fracture toughness between these two ceramics, it was expected that PAD B4C would possess a lower dwell/penetration transition
48
Advances in Ceramic Armor V
Ballistic Impact Damage Observations in a Hot-Pressed Boron Carbide
(a) (b) (c) Figure 3. Static X-ray radiographs of recovered B4C targets impacted at: (a) 1052 m/s, (b) 1063 m/s, and (c) 1198 m/s.
(a) (b) (c) Figure 4. Optical micrographs of the impact faces of B& targets impacted at: (a) 1052 d s , (b) 1063 m/s, and (c) 1198 m/s. velocity. Consequently, the apparent transition near 1060 m/s is not surprising; however, the transition back to shallow penetration at higher impact velocities is unexpected. Similar behavior for Sic-N is unknown since the velocity range between 1000 m/s and 1200 m/s was not carefully examined (see Figure l(a)). In an effort to better learn what is happening, the impact faces of the targets shown in Figure 3 were examined by optical microscopy. Figure 4 shows optical micrographs of the corresponding impact faces. The impact face of a B4C target that was impacted at 819 m / s (not shown) clearly shows an intact truncated impact cone formed due to Hertzian cone cracking and spallation of the surrounding material. However, the B& target impacted at 1052 m/s shown in Figure 4(a), only the lower slopes of the truncated impact cone (indicated by arrow) are evident. The upper part of the truncated impact cone appears to have been “crushed” and partially eroded. Consistent with Figure 3(b), Figure 4(b) shows a deep cavity filled with B4C fragments. In Figure 4(c), a partially-intact truncated impact cone (indicated by arrow) can be seen. This partially-intact truncated impact cone (referred to as “impact cone” in the remainder of this paper) is more clearly shown in Figure 5. The impact cone consists of top, rupture, and side surfaces. Figure 5(b) is a perspective view of the impact cone.
Advances in Ceramic Armor V
.
49
Ballistic Impact Damage Observations in a Hot-Pressed Boron Carbide
Figure 6. Division of top surface into stagnation point, core, and rim regions. Close-up of W-Ti metallic mass which defines the stagnation point region. Closer examination of the top surface of the impact cone reveals that it consists of three distinctly different regions: stagnation point, core, and rim regions as shown in Figure 6. The stagnation point region is characterized by a 300 pm diameter metallic mass determined to consist of W and Ti by X-ray energy dispersive spectroscopy. The presence of this W-Ti metallic mass supports computational simulations which show shallow penetration of both penetrator and cover plate material at the expected pressure stagnation point. The core region appears relatively smooth, while the' rim region displays wrinkled surface features. The wrinkled surface features are indicative of flow of the penetrator across the B& surface, while the smooth surface may be indicative of a stagnant layer of penetrator material at the B4C surface due to large frictional forces. A crack-like fissure is observed near the interface between these two regions. As will be shown later, this fissure appears to be a precursor for the formation of a rupture surface.
50
Advances in Ceramic Armor V
Ballistic Impact Damage Observations in a Hot-Pressed Boron Carbide
It was decided to section and polish the B4C cylinder impacted at 1198 m/ s to examine the sub-surface damage and determine possible reasons for the apparent unusual dweWpenetration transition behavior. Figure 7(a) is an optical micrograph (negative image) of the polished crosssection which clearly shows the different damage features and material that was displaced during the ballistic event. This cross-section is approximately 0.2 mm from the center of the stagnation point region shown in Figure 6 . Steep cone cracks can be clearly seen which is consistent with SiCN shown in Figure I(b). Figure 7(b) is a close-up view of the top region of the impact cone with different features indicated. These different features include an undamaged region, fissure and rupture surface, cone cracks and fragmentation, both short and long penny-shaped axial-splitting cracks, and coalesced axial-splitting cracks. SEM micrographs of these features are shown in Figure 8. The undamaged region is approximately 2 mm thick. As seen in Figure 8(a), despite being impacted at nearly 1200 m/s, the microstructure in this region looks like it does in its initial state. For comparisons, the undamaged region in Sic-N (see Figure l(b)) is less than 0.75 mm thick before the comminuted region is encountered. No comminuted region is evident in B4C. Figure 8(b) shows that the fissure observed on the top surface of the impact cone is a crack. Based on expectations of the approximate state-of-stress*' underneath the dwelling long-rod penetrator at this specific location, the fissure is a Mode I1 driven crack. Furthermore, this crack intersects with a major cone crack. If this major cone crack was an exposed surface of the impact cone due to material displacement, this crack would undoubtedly cause a rupture surface to form. Figure 8(c) shows short penny-shaped axial-splitting cracks in the region near the shot-line axis and immediately beneath the undamaged region. Figure 8(d) shows a long axial-splitting crack roughly at the same depth as the short penny-shaped cracks shown in Figure 8(c), but near the exposed surface of the impact cone. Note the fine secondary cracks that initiate off the main crack in Figure 8(d). Figure 8(e) shows the coalescence of axial-splitting cracks near the exposed surface of the impact cone. The morphology of these cracks appears to be similar to win cracks and faulting commonly observed in the deformation and failure of geological materials' 2 5 . Figure 8(f) is a close-up view of fragmentation between cracks (secondary or coalesced main cracks). It is believed that the shear deformation between main cracks causes the material isolated by secondary cracks to fragment under the action of both shear and tensile stresses. It is interesting to note that both the short and long axial-splitting cracks appear to radiate outward from the impact site suggesting their orientation to be governed by the maximum principal compressive stress. Their lengths and interaction appear to be influenced by the magnitude and sign of the lateral principal stress. Lastly, it is noted that evidence for solid-state amorphization of B4C in the form of distinct shear bands, previously observed by LaSalvia et al.'9-20in their study on the damage mechanisms in sphere-impacted B4C, was not observed in this study. The projectile characteristics, states of stress, stress magnitudes, and stress histories in both studies are very different and certainly contributed to the lack of its presence. The author's initial thoughts on the two apparent dwelUpenetration transitions was perhaps indicative of a change in the governing mechanism from microcracks to solid-state amorphization was not born out after careful examination. On the contrary, observations seem to suggest that both apparent dweWpenetration transitions are essentially governed by the same mechanism, the formation and weakening of the exposed impact cone due to the initiation, growth, and interaction of axial-splitting cracks. Since this process is expected to be sensitive to stochastic variations in experimental ballistic, target, and materials, it is suggested that the two apparent dweWpenetration transitions are simply indicative of a wide transition region. Further
P
Advances in Ceramic Armor V
.
51
Ballistic Impact Damage Observations in a Hot-Pressed Boron Carbide
Axial Cracks
(b) Figure 7. (a) Negative-image optical micrograph of the polished cross-section of the impact cone region clearly showing different damage features and displaced material. (b) Optical micrograph montage of the cone region with different features indicated.
52
.
Advances in Ceramic Armor V
Ballistic Impact Damage Observations in a Hot-Pressed Boron Carbide
(el (0 Figure 8. (a) Undamaged region. (b) Fissure. (c) Penny-shaped axial-splitting cracks near the shotline axis. (d) Lone axial-splitting crack with secondary cracks near the exposed impact cone surface. (e) Coalesced axial-splitting cracks and secondary cracks near the exposed impact cone surface. ( f ) Fragmentation of material between cracks.
Advances in Ceramic Armor V
.
53
Ballistic Impact Damage Observations in a Hot-Pressed Boron Carbide
evidence for this wide transition region in brittle materials is contained in the work by Anderson et a1.26on the interface defeat of ductile penetrators by borosilicate glass.
SUMMARY AND CONCLUSIONS Damage observations in a commercial hot-pressed B4C impacted ballistically with laboratory-scale 93YoW-Fe-Ni long-rod penetrators has been presented. Two apparent dwell/penetration transitions were observed (1060 m/s and 1200 d s ) . Examination of sub-surface damage in a B4C cylinder impacted at 1198 m/s revealed a number of interesting features. Just as in the previous work on Sic-N, long steep cone cracks near the center of impact were observed. Penetration of penetrator material into these cone cracks was also observed. Immediately beneath the impact site, a thick undamaged region with a thickness significantly greater than that in Sic-N was observed; however, no comminuted region was observed. The types and location of main cracks and the presence of fine secondary cracks indicate the relative brittleness of this ceramic compared to Sic-N. Solid-state amorphization was ruled out as the reason for the observed upper dwell/penetration transition. Based on the observed damage features, it is hypothesized that the velocity range between 1000 d s and 1200 m/s is a wide transition region controlled by stochastic variations in experimental parameters. The actual width of this transition region and which parameter variations dominate its width will be the focus of future work. FOOTNOTE Work performed with support by an appointment to the Research Participation Program at the U.S. ARL administered by the Oak Ridge Associated Universities through an interagency agreement $tween the U S . Department of Energy and U S . ARL. Armor-grade pressure-assisted densification (PAD) S i c manufactured by BAE Advanced Sframics, Vista, CA. Predecessor to PAD Sic-N. REFERENCES ‘M.L. Wilkens, C.F. Cline, and C.A. Honodel, “Light Armor,” Lawrence Radiation Laboratory, University of California, Livermore, UCRL-718 17 (1969). ’G.E. Hauver, E.J. Rapacki, Jr., P.H. Netherwood, and R.F. Benck, “Interface Defeat of Long-Rod Projectiles by Ceramic Armor”, ARL Technical Report, ARL-TR-3590, September 2005, 85 pp. 3M.W. Burkett, R.B. Parker, A D . Rollett, and G.E. Cort, “FY 90 PHERMEX and Q-Site Confined Ceramic Penetration Experiments (U),” Los Alamos Technical Report, LA- 12099-MS, August 1991,156 pp. 4D.L. Orphal and R. Franzen, “Penetration of Confined Silicon Carbide Targets by Tungsten Long Rods at Impact Velocities from 1.5 to 4.6 km/s,” Int. J. Impact Eng., 19, 1-13 (1997). ’D.L. Orphal, R.R. Franzen, A.C. Charters, T.L. Menna, and A.J. Piekutowski, “Penetration of Confined Boron Carbide Targets by Tungsten Long Rods at Impact Velocities from 1.5 to 5.0 k d s , ” Int. J. Impact Eng., 19, 15-29 (1997). 6W.A. Gooch, M.S. Burkins, G . Hauver, P. Netherwood, and R. Benck, “Dynamic X-ray Imaging of the Penetration of Boron Carbide,” J. de Physique IV, 10 [9] 583-88 (2000). ’P. Lundberg, R. Renstrom, and B. Lundberg, “Impact of Metallic Projectiles on Ceramic Targets: Transition between Interface Defeat and Penetration,” Int. J. Impact Eng., 24, 259-75 (2000). *P. Lundberg and B. Lundberg, “Transition between Interface Defeat and Penetration for Tungsten Projectiles and Four Silicon Carbide Materials,” Int. J. Impact Eng., 31,781-92 (2004). 91.M. Pickup, A.K. Barker, I.D. Elgy, G.J.J.M. Peskes, and M. van de Voorde, “The Effect of
54
Advances in Ceramic Armor V
Ballistic Impact Damage Observations in a Hot-Pressed Boron Carbide
Coverplates on the Dwell Characteristics of Silicon Carbide Subject to KE Impact,” Proceedings of the 2 1It International Symposium on Ballistics, Adelaide, Australia (2004). lop. Lundberg, R. Renstrom, and B. Lundberg, “Impact of Conical Tungsten Projectiles on Flat Silicon Carbide Targets: Transition from Interface Defeat to Penetration,” Int. J. Impact Eng., 32, 1842-56 (2005). “T. Behner, D.L. Orphal, V. Hohler, C.E. Anderson, Jr., R.L. Mason, and D.W. Templeton, “Hypervelocity Penetration of Gold Rods into SIC-N for Impact Velocities from 2.0 to 6.2 W s , ” Int. J. Impact Eng., 33,68-79 (2006). 12C.E. Anderson, T. Behner, T.J. Holmquist, M. Wickert, V. Hohler, and D.W. Templeton,” Interface Defeat of Long Rods Impacting Borosilicate Glass,” Proceedings of the 231d International Symposium on Ballistics, Tarragona, Spain, 1049-56 (2007). I3T. Behner, C.E. Anderson, T.J. Holmquist, M. Wickert, and D.W. Templeton, “Interface Defeat for Unconfined SIC Ceramics,” Proceedings of the 24thInternational Symposium on Ballistics, New Orleans, USA, 35-42 (2008). I4J.C. LaSalvia, B. Leavy, H.T. Miller, J.R. Houskamp, and R.C. McCuiston, “Recent Results on the Fundamental Performance of a Hot-Pressed Silicon Carbide Impacted by Sub-scale Long-Rod Penetrators,” in Advances in Ceramic Armor IV,Cer. Eng. Sci. Proc., 29 [6] 85-94 (2008). ”C.E. Anderson, T. Behner, D.L. Orphal, A.E. Nicholls, T.J. Holmquist, and M. Wickert, “LongRod Penetration into Intact and Pre-Damaged SIC Ceramic,” Proceedings of the 24thInternational S posium on Ballistics, New Orleans, USA, 822-29 (2008). ’ F E . Anderson, T. Behner, D.L. Orphal, A.E. Nicholls, and D.W. Templeton, “Time-resolved Penetration into Pre-damaged Hot-pressed Silicon Carbide,” Int. J. Impact Eng., 35 661-73 (2008). ”E. Strasburger, “Ballistic Testing of Transparent Armour Ceramics,” J. Euro. Cer SOC.,29 267-73 (2009). “J.C. LaSalvia, M.J. Normandia, H.T. Miller, and D.E. MacKenzie, “Sphere Impact Induced Damage in Ceramics: I. Armor-grade Sic and TiB2,” Cer. Eng. Sci. Proc., 26 [7] 171-81 (2005). ‘9J.C. LaSalvia, M.J. Normandia, H.T. Miller, and D.E. MacKenzie, “Sphere Impact Induced Damage in Ceramics: 11. Armor-grade B4C and WC,” Cer. Eng. Sci. Proc., 26 [7] 183-92 (2005). 2oJ.C. LaSalvia, R.C. McCuiston, G. Fanchini, J.W. McCauley, M. Chhowalla, H.T. Miller, and D.E. MacKenzie, “Shear Localization in a Sphere-Impacted Armor-Grade Boron Carbide,” Proceedings of the 231dInternational Symposium on Ballistics, Tarragona, Spain, 1329-37 (2007). ”M.W. Chen, J.W. McCauley, J.C. LaSalvia, and K.J. Hemker, “Microstructural Characterization of Commercial Hot-Pressed Boron Carbide Ceramics,”J. Am. Ceram. Soc., 88 [7] 1935-42 (2005). ”A.C. Fischer-Cripps, “Predicting Hertzian Fracture,” J. Mat. Sci., 32 (5), 1277-85 (1997). 23H.Horii and S. Nemat-Nasser, “Brittle Failure in Compression: Splitting, Faulting, and BrittleDuctile Transition,” Phil. Trans. R. SOC.Lond. A , 319,337-74 (1986). 24M.F.Ashby and C.G. Sammis, “The Damage Mechanics of Brittle Solids in Compression,” Pure andApplied Geophysics, 133 [3], 489-521 (1990). 25Deformation Microstructures and Mechanisms in Minerals and Rocks, Springer, Netherlands, 2000. * k E . Anderson, T. Behner, D.L. Orphal, T.J. Holmquist, V. Hohler, and M. Wickert, ”Interface Defeat of Long Rods Impacting Borosilicate Glass Experimental Results,”, SwRI Report 18.12544/009, February 2009.
Advances in Ceramic Armor V
.
55
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
CHARACTERIZATION OF MICROSTRUCTURAL DAMAGE IN PROCESSED VIA MODIFIED CHEMICAL VAPOR DEPOSITION
SILICON CARBIDE
H.T. Miller*, J.C. LaSalvia, R.B. Leavy, and D.E. MacKenzie U.S. Army Research Laboratory AMSRD-ARL-WM-MD Aberdeen Proving Ground, MD 21005-5069 ABSTRACT The role of micro and macro inelastic deformation and damage mechanisms in the high velocity penetration resistance of many technical ceramics is not M y understood. By comparing and contrasting the damage mechanisms of similar ceramics impacted at high velocities, insight can be gained, qualitatively, into the influence of crystal structure and microstructure. In this study, two commercially-available silicon carbide (Sic) variants were impacted with cemented carbide spheres (WCdwt.%Co) at velocities between 300 m / s and 1700 m/s. The two S i c variants were Sic-N (BAE Advanced Ceramics) and CVC Sic (TREX Enterprises). Ceramic tiles 76.2 mm in diameter and 19.05 mm thick were heavily confined and pre-stressed in a titanium alloy (Ti-6AI-4V) target fixture. Select impacted tiles (low and high impact velocities) were subsequently sectioned and polished. Optical and scanning electron microscopy were utilized to examine the sub-surface damage features. Radial, lateral, and cone cracking were common in both materials. In Sic-N, comminuted regions due predominately to microcracking (grain boundary) and some microcleavage (grains) were observed. In the CVC Sic, shear localization due to fragmentation (intersecting microcleavage) - , and solid-state amorphization.were observed. Grain b o u n d 6 microcracking appeared to suppress shear localization in Sic-N. INTRODUCTION Silicon carbide (SIC) is an important material in the development of lightweight ceramic In an effort to support the development of ceramic armor, a variety of ballistic testing methodologies and analyses have been used. Beyond traditional V50, depth-of-penetration (DOP), and Vs-Vr testing, other methods such as penetration resistance calculations, computational modeling, and damage observations have provided insight into the effect of system and projectile characteristics on fundamental ballistic performance. Over the years, a number of methods have been used to determine the potential of a ceramic to resist penetrati~n.~ Normandia4 conducted DOP experiments using WC 6wt.% Co spheres as the projectile, to impact commercially-available S i c variants (76.2 mm diameter by 19.05 mm thick tiles) in highly confined and pre-stressed targets. The DOP results were analyzed with respect to impact velocity, using a model based on cavity expansion theory and projectile deformation, to calculate a penetration or target resistance value for each ceramic. With these values, it was believed that ceramics could be screened against these types of projectiles without having to conduct expensive and time-consuming V50 testing. LaSalvia et al.5-6conducted sphere-impact experiments to study the mechanisms associated with the inelastic deformation response of ceramics. These experiments also proved useful in developing and validating computational models of technical ceramics. Combining cross-sectional damage observations and computational modeling should help in understanding key microstructural characteristics that contribute to penetration resistance. Making this connection is crucial in developing improved armor ceramics. The purpose of this paper is to report on the various damage features in two commerciallyavailable SIC variants resulting from sphere impact. These two Sic variants are of interest because they possess different crystal structures and microstructures. DOP data will also be presented.
-
57
Characterization of Microstructural Damage in Silicon Carbide
(a) (b) Figure 1. Representative microstructures for (a) Sic-N and (b) CVC Sic. EXPERIMENTAL PROCEDURES S i c variants manufactured by BAE Advanced Ceramics (Sic-N) and TREX Enterprises (CVC Sic) were used in this study. Sic-N is an armor-grade S i c produced via hot-pressing, while CVC (chemical vapor composite) S i c is produced by a modified chemical vapor deposition process. Material characteristics for Sic-N and CVC S i c are listed in Table I. Densities and elastic moduli were determined using the Archimedes water immersion and pulse-echo techniques (ASTM Standard E494), respectively. Knoop hardness was determined on specimens final polished with 0.05 pm colloidal silica and in accordance with ASTM Standard C1326 using a load of 40 N. Fracture toughness was determined using the single-edge pre-cracked beam (SEPB) technique in accordance with ASTM Standard C1421 and specimen (bar) dimensions of 3 mm x 4 mm x 50 mm. A standard 20 x 40 mm semi-articulating flexure four-point fixture was used with a cross-head speed of 0.5 d m i n in accordance with ASTM C1161. Phases were determined using X-ray diffraction and a commercial pattern matching program. For Sic-N, phase analysis was aided using electron backscattered diffraction. Figure 1 shows representative microstructures for SIC-N and CVC Sic. As can be seen in Figure l(a), SIC-N consists of equiaxed grains with sizes predominately below 5 pm. On the other hand, the microstructure of CVC S i c can be described as “feather-like” and consists of elongated platelets oriented parallel with the growth direction. Tiles 76.2 mm in diameter and 19.05 mm thick were inserted into Ti-6A1-4V targets that were highly confined and pre-stressed using a shrink-fit. The targets were impacted with WCdwt.%Co 6.35 mm diameter spheres at velocities between 300 d s and 1700 d s . Further detail of the ballistic experiments and set-up are given by Normandia4. Targets from two nominal impact velocities were chosen for post-mortem examination. Sic-N targets, impacted at 530 m / s and 1403 d s , and CVC S i c targets, impacted at 394 d s and 1350 d s , were selected. Ceramic tiles were extracted from the Ti-6A1-4V targets through a careful process involving infiltration with epoxy and sectioning. Following extraction, the ceramic tiles were reepoxied and sectioned within 2 mm of the impact centers. The exposed cross-sections were re-epoxied and then ground and metallographically prepared to within 0.5 mm of the impact centers. Cross-
58
’
Advances in Ceramic Armor V
Characterizationof Microstructural Damage in Silicon Carbide
h
E E
v
impact Velocity (m/s) Figure 2. DOP as a function of impact velocity for Sic-N and CVC SIC sections were fmal polished with 0.25 micron diamond. Optical and scanning electron microscopy were used to examine the various damage features. RESULTS AND DISCUSSION DOP results as a function of impact velocity are shown in Figure 2. The data shows that the penetration onset velocity for the CVC S i c is significantly higher than that for Sic-N. The penetration onset velocity for Sic-N is nominally 500 m / s while that for CVC Sic is somewhere between 850 m/s and 1050 m / s . Penetration depths are similar in the velocity range between 1050 m / s and 1200 m / s . However, above 1300 m / s , the penetration resistance of CVC SIC is much less than that of Sic-N. At 1591 m / s the DOP of CVC Sic was 14.6 mm, while DOP for Sic-N at 1597 m / s was 7.9 mm. Figure 3 shows the polished cross-sections for the Sic-N and CVC S i c tiles, with the corresponding impact velocities indicated. Figures 3(a) and 3(b) are Sic-N cross-sections, while Figures 3(c) and 3(d) are CVC S i c cross-sections. In general, regardless of the impact velocity, macroscopic damage in both materials appears to be similar. Cone cracking is clearly visible in all cross-sections. At low velocity, crater diameters and depths are small while cone cracks are few. At high velocity, the crater diameters and depths are larger and cone cracking is extensive. In the remainder of this paper, examination of select features near the impact center will be presented and discussed. A close-up view of the sub-surface damage beneath the impact site and neighboring regions for Sic-N impacted at 530 m / s is shown in Figure 4(a). The bright regions are regions of extensive damage. Immediately beneath the impact site, the comminuted region or Mescal1 zone can be seen. The comminuted region shown more clearly in Figure 4(b), consists primarily of a high-density of grain boundary microcracks and microcleavage cracks. Figure 5(a) shows the corresponding low velocity (394 d s ) impact cross-section for CVC Sic. Numerous cone cracks can be clearly seen. Directly below the impact site, a relatively undamaged region exists as shown in Figure 5(b). Unlike Sic-N, a comminuted region does not exist beneath the impact center. The damage underneath the impact center primarily consists of fine microcleavage cracks and interacting microcracks as shown in Figure 5(c).
Advances in Ceramic Armor V
.
59
Characterization of Microstructural Damage in Silicon Carbide
(4 Figure 3. Polished cross-sections showing sub-surface damage features. (a) Sic-N (530 d s ) , (b) SicN (1403 d s ) , (c) CVC S i c (394 mh), and (d) CVC Sic (1350 d s )
60
Advances in Ceramic Armor V
Characterization of Microstructural Damage in Silicon Carbide
(a) (b) Figure 4. (a) Close-up view of the impact region in Sic-N impacted at 530 m / s . (b) Comminuted region consisting of extensive grain boundary microcracking and microcleavage cracks.
(c) Figure 5. (a) Close-up view of the impact region in CVC SIC impacted at 394 m / s . (b) Undamaged region directly under impact site. (c) Microcleavage cracks and interacting microcracks.
Advances in Ceramic Armor V
.
61
Characterization of Microstructural Damage in Silicon Carbide
(b) (c) Figure 6. (a) Close-up view of the impact crater region in SIC-N impacted at 1403 d s . (b) Evidence of shear band formation. (c) Close-up view of comminution and fragmentation within the shear band. Figure 6(a) shows a close-up view of the impact crater region in Sic-N impacted at 1403 d s . As was seen in the Sic-N impacted at low velocity, comminution is seen near the impact center. However, grain boundary microcracking was seen at a distance of up to 4 mm away from the top surface, while that of the lower velocity sample was concentrated within a 2 mm radius from the center-of-impact. It is also noted that grain boundary microcracking and fragmentation was observed in regions of high shear deformation as shown in Figure 6(b). Shear bands were fust observed in this material by Shih et a].’ where they used the explosively-driven collapse of thick-walled cylinder technique to investigate the inelastic deformation behavior of ceramics under dynamic loading conditions. The shear band shown in Figure 6(b) is believed to be the fust time such behavior has been observed in this material as a result of a ballistic impact. Figure 6(c) is a close-up view of the comminution and fragmentation within the shear band. Figure 7(a) shows a close-up view of the impact crater region in CVC SIC impacted at 1350 d s . While shear deformation is evident along some cone cracks, the black ellipses outline regions of intense shear deformation. These areas are characterized by intense cracking and fragmentation as will be shown. Figure 7(b) is a close-up view of a shear band consisting of ceramic fragments of differing sizes. At low magnification, this shear band appears to be a cone crack. However, it is noted that most cone cracks were not shear bands. Closer examination of the shear band and neighboring regions
62
*
Advances in Ceramic Armor V
Characterization of Microstructural Damage in Silicon Carbide
(4 (el Figure 7. (a) Close-up view of the impact crater region in CVC S i c impacted at 1350 d s . Circled areas indicate intense shear deformation regions. (b) Shear band resulting from classic fragmentation process. (c) Region of intense shear deformation showing fragmentation due to classic process and possible solid-state amorphization. (d) Shear band (filled with epoxy) showing possible evidence of solid-state amorphization. (e) Close-up view of shear band showing possible melting and phase separation (dark spots are C-rich, grayish regions are Si-rich, lighter regions contain both Si and C).
Advances in Ceramic Armor V
.
63
Characterization of Microstructural Damage in Silicon Carbide
indicate that fragmentation occurred through the growth and interaction of finely-spaced microcleavage cracks driven by local tension and shear Stresses arising from shear deformation. Figure 7(c) shows a region corresponding to one of the circled areas in Figure 7(a). Fragmentation is clearly seen. However, the big difference between the fragmentation shown in Figures 7(b) and 7(c) is the presence of fine-scale fragmentation resulting from what appears to be possible evidence for solidstate amorphization. This fine-scale fragmentation was found as individual shear bands as shown in Figure 7(d) and within shear bands that formed predominately due to the classic fragmentation process. These features were also observed near the bottom of the impact crater. Figure 7(e) is a close-up view of the structure typical of these types of shear bands. Evidence for melting and phase separation can be clearly seen. The dark spots are evidently C-rich, the grayish regions are Si-rich, and the lighter regions contain both Si and C. Molecular dynamic simulations for nanoindentation of p-Sic have shown it to undergo solidstate amorphization rather than polymorphic transition.*-” However, it is also known that P-SiC undergoes a transition to the NaC1-type structure at a hydrostatic pressure of 100 GPa.” The application of shear in addition to hydrostatic pressure can lower the critical pressure for solid-state transformation.I2 It is not clear at this time exactly what mechanism led to the formation of the shear band shown in Figure 7(d); however, the evidence of phase separation and melting appears consistent with amorphization. CONCLUSION A fundamental ballistic study into the damage and inelastic deformation mechanisms for two commercial Sic’s, Sic-N and CVC Sic, was presented. The penetration onset velocity of CVC S i c was significantly higher than Sic-N. However, the penetration resistance of CVC S i c above 1300 m/s was significantly lower than Sic-N. Polished cross-sections for each material impacted at low velocity and high velocity were prepared and examined using optical and scanning electron microscopy. Multi-scale cracking was evident in all four cross-sectioned samples. Comminution due to grain boundary microcracking was observed in Sic-N. Fine microcleavage cracks were observed in CVC Sic. Fragmentation in Sic-N is due to intersecting macrocracks and grain boundary microcracks, while in CVC S i c fragmentation was due to intersecting macrocracks and microcleavage cleavage. Distinct shear bands were observed in CVC Sic, while diffuse shear bands were observed in Sic-N. In Sic-N, grain boundary microcracking appeared to suppress the formation of distinct shear bands. Evidence for shear-induced solid-state amorphization was observed in CVC Sic in regions of severe shear deformation. Further work is needed to gain additional insight into this phenomenon. FOOTNOTE *Work performed with support by an appointment to the Research Participation Program at the U.S. ARL administered by the Oak Ridge Associated Universities through an interagency agreement between the U.S. Department of Energy and US.ARL. REFERENCES ’C. Roberson and P.J. Hazell, “Resistance of Silicon Carbide to Penetration by a Tungsten Carbide Cored Projectile,” Cer. Trans., 151, 165-74 (2003). 2T.M. Lillo, D.W. Bailey, D.A. Laughton, H.S. Chu, and W.M. Harrison, “Development of a Pressureless Sintered Silicon Carbide Monolith and Special-shaped Silicon Carbide Whisker Reinforced Silicon Carbide Matrix Composite for Lightweight Armor Application,” Cer. Sci. Eng. Proc., 24 [3] 359-64 (2003). ’M.J. Normandia and W. Gooch, “An Overview of Ballistic Testing Methods of Ceramic Materials,” Cer. Trans., 134, 113-38, (2002).
64
.
Advances in Ceramic Armor V
Characterization of Microstructural Damage in Silicon Carbide
4M1.J.Normandia, Impact Response and Analysis of Several Silicon Carbides, Int. J. Appl. Ceram. Technol., 1 [3] 226-34 (2004). ’J.C. LaSalvia, M.J. Normandia, H.T. Miller, D.E. MacKenzie, Sphere Impact Induced Damage in Ceramics: I. Armor Grade SIC and TiB2, Cer. Eng. Sci, Proc., 26 [7] 171-79, (2005). 6J.C. LaSalvia, M.J. Normandia, H.T. Miller, D.E. MacKenzie, Sphere Impact Induced Damage in Ceramics: 11. Armor Grade B4C and WC, Cer. Eng. Sci, Proc., 26 [7] 180-88, (2005). ’C.J. Shih, V.F. Nesterenko, M.A. Meyers, “High-strain-rate deformation and comminution of silicon carbide,”J. Appl. Phys., 83 [9] 4660-71 (1998). ‘S. Yip, J. Li, M. Tang, and J. Wang , “Mechanistic aspects and atomic-level consequences of elastic instabilities in homogeneous crystals,” Mat. Sci. Eng., A317 236-40 (2001). 91. Szlufarska, R.K. Kalia, A. Nakano, and P. Vashishta, “Nanoindentation-induced amorphization in silicon carbide,” Appl. Phys. Letters, 85 [3] 378-80 (2004). ‘9.Szlufarska, R.K. Kalia, A. Nakano, and P. Vashishta, “Atomistic mechanisms of amorphization during nanoindentation of Sic: A molecular dynamics study,” Phy. Rev. B, 71 174113 (2005). “M. Yoshida, A. Onodera, M. Ueno, K. Takemura, and 0. Shimomura, “Pressure-Induced PhaseTransition in Sic,” Phy. Rev. B, 48 [14] 10587-90 (1993) ‘*K.J. Kingma, C. Meade, R.J. Hemley, H.K. Mao, and D.R. Veblen, “Microstructural Observations of Alpha-Quartz Amorphization,” Sci., 259 [5095] 666-69 (1993).
Advances in Ceramic Armor V
. 65
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
Material Concepts, Processes and Characterization
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
EFFECTS OF GRAIN SIZE, SHAPE AND SECOND PHASES ON PROPERTIES OF SMTERED SIC P. G. Karandikar, G. Evans, S. Wong, and M. K. Aghajanian M Cubed Technologies, Inc. 1 Tralee Industrial Park Newark. DE 19711 ABSTRACT Sintered Sic has high hardness, light weight, high stiffness, excellent corrosion resistance, and high temperature strength and stability. It is one of the leading candidate materials for next generation body and vehicle armors. Ideally, both high hardness and high toughness are desired for good performance. However, most literature shows that as hardness increases, toughness decreases. Thus, a key challenge in processing Sic is to obtain a microstructure that will provide the best combination of hardness and toughness. The key attributes of S i c microstructure that affect its properties include porosity, grain size, grain shape, and the type, locations and amounts of second phases. Variables that control the microstructure include starting raw materials, sintering aids, and processing (preforming and sintering) conditions. In this work, test plates were produced with controlled microstructure (e.g. coarse versus fine grains, round versus acicular grains, porosity, different second phase additions etc.) by systematically varying process parameters. Microstructure, density, elastic modulus, flexural strength, fracture toughness and h o o p hardness of these plates were characterized. Correlations are developed between the microstructure and properties. Fragments were collected after ballistic tests and their microstructure was characterized. INTRODUCTION A number of criteria must be considered when selecting materials for use in a personnel armor system for protection against ballistic threats. These include the characteristics of the specific threats to be defeated, the allowable volume and weight parameters of the system, and the system cost. Because of the range of design criteria that exist for armor systems, there is no single "best" armor material for all applications. For example, a material that provides adequate protection against specific threats in one system configuration may become inadequate if the permissible system weight is reduced. For this reason, a number of different ceramic materials have been and continue to be used in fielded armor systems. Personnel armor can be broadly classified into soft armor (e.g. textile based systems using high performance fibers) and hard armor (systems containing metallic or ceramic inserts)" '. Soft armor is adequate for certain low level threats (e.g. NIJ level I11 or less), while for higher level threats (e.g. NIJ level IV or higher) metallic or ceramic inserts are needed''4. Low density ceramics (with density 2.5-4.0 gkc) offer higher hardness and less than half the weight (at the same thickness) in comparison with armor steel (density 7.8 g/cc). A1203, B4C and SIC have been used over the past decade in personnel armor systems. Sintered and hot pressed varieties of SIC and B4C were predominant until the development of reaction-bonded Sic and B4C (RBSC and RBBC) based armor in the late 1 9 9 0 ~',~with . their use in personnel armor starting in the year 2000. B4C is the lowest density and highest hardness ceramic typically used for personnel armor. However, it has been found recently that its performance against armor piercing (AP) rounds exerting high pressure has been below
69
Effects of Grain Size, Shape and Second Phases on Properties of Sintered SIC
expectations, especially compared to less hard and denser Sicb". The recent uncertainties in the performance of B4C against more tenacious threats have been attributed to shear localization (amorphization) that occurs in this material under high pressure'. Unlike boron carbide, Sic does not undergo amorphization under high applied dynamic pressures. As a result, sintered SIC is becomin increasingly important as a ceramic material of choice against high pressure AP threats 1 0 5. SINTERED SILICON CARBIDE ProchazkaI6 first developed and patented pressureless sintered S i c with boron and carbon sintering aids. Since then, a variety of approaches have been developed for making sintered siclo-I7. These approaches can be broadly classified into solid state sintering (SSS or direct sintering) and liquid phase sintering (LPS). To make sintered ceramics, typically submicron powders are compacted (e.g. by pressing) and the compact is heated to a temperature where mass transport becomes sufficiently rapid. Fine powders have high surface area (typically 10-20 m2/g). When the powder compact is heated to a high temperature, there is a thermodynamic driving force to reduce the surface energy by growing smaller grains into larger grains. As shown in the schematic in Figure 1, the following different events occur during densification'*, 19: (1) Mass is removed from particle centers and deposited in the areas of negative curvature. This leads to shrinkage and is called sintering. Surface area reduces and results in reduction of the associated free energy. (2) New grain boundaries are formed leading to increase of grain boundary energy. (3) All pores between grains shrink, but only the ones with coordination number (number of nearest neighbor grains) smaller than critical coordination number completely disappear. (4) An equilibrium structure is formed when surface free energy reduction equals grain boundary energy increase. At this point, sintering stops. ( 5 ) Interparticle mass transport continues due to differences in the curvatures of adjacent grains. Bigger grains grow at the expense of smaller grains. This is called coarsening. Kinetics of coarsening are slower than that of sintering. Coarsening dominates after density reaches 0.77 times theoretical density. (6) Coarsening can reduce coordination number of pores and help eliminate them. ( 7 ) Coarsening also reduces curvature and re-activates mass transport or sintering. (8) Grain boundary migration can also result in grain growth and can be very rapid because just like dislocations, only nanometer level atom movements are needed.
Figure 1. Left: Mass transport during densification, Right: sintering (a+ b), coarsening ( b 9 c)
Ideally, during densification normal grain growth is desirable. Here, all grains grow uniformly increasing the scale of the microstructure, and pores remain at grain boundaries.
70
Advances in Ceramic Armor V
Effects of Grain Size, Shape and Second Phases on Properties of Sintered Sic
Unfortunately, abnormal grain growth is also seen widely where some grains can grow very big due to impurities and anisotropic grain boundary energy. In abnormal grain growth, porosity is trapped inside the abnormally grown grain. Such large grains can act as flaws and can reduce strength and ballistic performance. In the case of sintered Sic and Si3N4, phase transformations can also lead to formation of large, elongated grains. For example, if starting powders are p-Sic, the sintered Sic @redominantly a-Sic) is more prone to developing elongated grains. Thus, the size as well as phase composition of the starting powders have to be controlled very carefully to produce sintered ceramics with controlled microstructure. These events explain how pressure applied during sintering reduces the need for coarsening and achieves finer grain size in sintered parts (e.g. hot pressed Sic). Also, it underlines the importance of producing preforms for sintering with uniform porosity with low coordination number. In real life, all starting powders invariably have a grain size distribution rather than mono-sized grains. In such a situation, if some extremely large grains exist, they can show abnormal grain growth. The starting compacts for sintering are typically 45 to 70% dense. The sintered compacts are 98 to 100% dense. Thus, when densification occurs, spaces between particles have to be eliminated, resulting in change of the volume of the compact (shrinkage). Typically, shrinkage is isotropic (same % in all dimensions) - unless it is constrained. The linear shrinkage due to densification is strictly a function of starting compact density (green density) and the final sintered density -with a minor influence of weight loss during sintering. SPECIMEN FABRICATION For fabrication of sintered Sic, first a preform or a compact is made of fine (typically submicron) powders and a binder. The binder is then removed and preform is subjected to sintering temperature. Two main categories for preforming methods are (a) dry pressing, and (b) casting. The pressing category includes room temperature uniaxial pressing, isostatic pressing, etc. The casting category includes slip casting, pressure casting, tape casting, etc. Other processes such as injection molding can also be used. The preforming process used has a large impact on the microstructure and defects produced in the final sintered body. This study is focused towards fabrication of sintered Sic for defeating the next generation high pressure inducing personnel armor threat. Based on prior work, the study focused on solid state sintered Sic. Two types of specimens (100 mm x 100 mm tiles and full double curved torso plates) were made by varying the processing methods and conditions to obtain various controlled microstructures (Table I). A tile was also obtained for a commercial silicon carbide material (Hexoloy SA) for comparative evaluation. Microstructures and properties of these were thoroughly characterized. The resultant microstructure-property relations are described in this paper. Ballistic performance of these tiles and the corresponding full-scale torso plates was also characterized but is not the subject of this paper. The designation for the various specimens fabricated in this study is described in Table I. Thus, Al-1 would mean pressed preform, powder type 1 and process condition 1. Letter A - Press B - Cast
Number Powder type in pressed Additive type in cast
Number Process condition (temperaturehime)
Advances in Ceramic Armor V
.
71
Effects of Grain Size, Shape and Second Phases on Properties of Sintered Sic
As described earlier, fine powders have to be used to achieve densification. Figure 2a shows particle size distribution (PSD) for a sinterable S i c starting powder. In the case of casting approaches, this powder can be directly used. However, for pressing, the fine powder does not flow well into the tooling. Therefore, fine powders are converted to larger granules by spray drying using a binder. Sintering additives are typically added during the spray drying stage. Figure 2a also shows the PSD of the same S i c powder converted to granules by spray drying. Figure 2b shows an SEM micrograph of spray dried granule. Figure 2c shows higher magnification view of the granule showing the individual submicron powders. Figure 3a shows optical micrographs of intentionally partially sintered S i c granules. If the granules are hollow (or donut-like), a large pore is left in the center of the sintering granule, which would leave a large pore in the sintered body. The hollow center of the granule must be eliminated by controlling the spray drying parameters or by using very high pressure during pressing to flatten the pore. Figure 3b shows how a well made granule prevents formation of large voids and leaves only intergranular pores. Again, high pressing pressure can flatten out the granules to eliminate these pores. The quality of the granules, size distribution of granules as well as pressing (pressure and uniformity) must be controlled to achieve a preform with uniform pore distribution and pores with low coordination number. This allows pores to be eliminated during densification by the various mechanisms described earlier.
Figure 2. (a) particle size distributions (PSD) of as received and spray dried sinterable S i c powders, (b) SEM of spray dried SIC granules, and (c) higher magnification SEM micrograph of a granule showing individual S i c particles.
Figure 3. Microstructure of a partially sintered S i c sample made to reveal potential defects in sintered parts made with pressed spray-dried powders
72
.
Advances in Ceramic Armor V
Effects of Grain Size, Shape and Second Phases on Properties of Sintered Sic
In the casting processes, granule size and quality issues are not present. However, uniformity of packing produced during the casting process is very important. Figure 4 shows microstructures of two sintered Sic specimens made by casting. In one case, an area of lower packing (commonly called a knit line) can be observed. Knoop 2 kg hardness was measured in these samples as a function of location through the thickness. A plot of hardness versus location for the two samples is also shown in Figure 4. It is clear that in the sample with a knit line, hardness shows significant decrease in the area of knit line. Thus, it is critical to control the casting process to achieve uniform green density in the preform. 2000
e
G
1900 1800
3 1700 3 1600 5
1500
pq1400
.
1000 0
>
2000
1
4000
,
___
- - -
1100
j-
8000
8000
10000
Dlstance (micrometer)
Figure 4. Variation of hardness across the knit line in cast and sintered Sic
SPECIMEN CHARACTERIZATION From each tile, small specimens were cut, mounted, ground and polished using conventional metallographic techniques. Specimens were etched with a KOH -K3FeCN6 mixture to reveal grain structure. Grain size was measured using the concentric circle technique per ASTM El 12-96. Density was measured by the Archimedes principle per ASTM B311. Elastic modulus was measured by the ultrasonic pulse-echo technique per ASTM E494-05. Flexural strength was measured by the four point bend method per ASTM C 1161. Fracture toughness was measured by the Chevron Notch method per ASTM C1421. Minimum 5 specimens were tested for flexural strength and fracture toughness. Hardness was measured on the h o o p scale with a 2 kg load per ASTM Standard C 1236 using a Shimadzu HMV-2000 hardness tester (on un-etched specimens). Minimum 10 indents were made and an average value was reported. For ceramics, the Knoop hardness initially decreases as the applied load increases and then it reaches a stable value. The load that yields a stable hardness is determined to be 2 kg20. MICROSTRUCTURE Optical micrographs of various samples produced in this study are shown in Figures 5 and 6. As shown in the micrographs, in solid state sintered Sics, the grains can be tailored to be large (>20 micron) or small (3-5 micron) based on sintering conditions. Grains can be equiaxed (rounded) or elongated. Also, 1-3% porosity is observed based on the sintered density reached (-1% porosity for every density reduction by 0.003 g/cc from the theoretical 3.21 g/cc). Residual carbon or graphite may also be present in the microstructure.
Advances in Ceramic Armor V
.
73
Effects of Grain Size, Shape and Second Phases on Properties of Sintered Sic
Figure 5. Microstructures of various SIC samples at low magnification. Long needle like grains are seen in some samples while fine equiaxed microstructure is seen in others.
74
Advances in Ceramic Armor V
Effects of Grain Size, Shape and Second Phases on Properties of Sintered SIC
Figure 6. Microstructures of various SIC samples at high magnification. Long needle like grains are seen in some samples while fine equiaxed microstructure is seen in others.
Advances in Ceramic Armor V
75
Effects of Grain Size, Shape and Second Phases on Properties of Sintered SIC
PROPERTIES Properties of various types of Sics made in this study are summarized in Table 11. Elastic moduli and hardnesses of all the specimens are plotted as a function of density in Figure 7a. It is clear that elastic modulus reduces as density reduces. Hardness does show a decreasing tendency with density. However, the variation in the hardness data is large. Hardness and toughness of various samples are plotted as a function of grain size in Figure 7b. For the grain size ranges evaluated, no clear dependence is seen for either on grain size. For hardness, typicall a HallPetch type relation is reported by which hardness increases as the grain size decrease? ?”. Lack of such a clear trend in the current data could be due to a variety of reasons. There is significant variation in the hardness data. Since the hardness indent is only of the order of 100 micron in length, it is heavily subject to local microstructural features such as porosity (lower density samples have higher porosity). Since density and grain size both affect hardness, for an ideal study, samples with same density and differing grain sizes must be produced. Figure 8 shows several indents in the sintered Sics of different microstructures. Figure 8a shows an indent made in un-etched Sic. Figure 8b shows the same exact indent after etching. This sample has uniform, fine grain structure and the grains are much smaller than the indent. Figure 8c and 8d show instances where the indent is entirely on a single grain and on 2-3 large grains. This results in significant variation in the data. Also, grain size measurements pose special challenges when the microstructure is not equiaxed and has a bimodal mixture of long thin grain and equiaxed grains. Previous studies23 have reported significant increase in toughness with increasing grain size (with elongated grains). Such a significant increase in toughness with increasing grain size was not observed in this study. Table 11. Summary of properties of various Sics made during this study
14
/A3
bommercial Sic/ 3.13
1
4.6
1
423
1
370k 28
I
* - non ASTM slightly larger sample. These data were not used in Figure 7b.
76
.
Advances in Ceramic Armor V
2.5 i 0.04 11940k 301
Effects of Grain Size, Shape and Second Phases on Properties of Sintered Sic
Samples made by casting showed higher strength than those made by pressing. Two potential reasons for this are: (a) The porosity in the cast samples is finer and larger pores between granules that are seen in pressed parts are absent in cast parts, and (b) A grain orientation effect was observed in the cast samples where the elongated grains were oriented perpendicular to the casting direction. Samples were machined such that these elongated grains were perpendicular to the direction of stress in the bending test. 2100
4 3.8
pow
3.6
E
1.4
B1wo
3.2
3 2.8
m
2.6 2.4
Figure 7. Left: Elastic modulus and hardness Vs. density Right: Hardness and toughness Vs. Grain size.
Figure 8. Interaction of hardness measurement with microstructure (a) 100608VX-2 sample 6 indent 4(before etch); (b) same after etch; (c) 022908VX-2 indent entirely in one grain (d) 022908VX-2 indent over multiple large grains STATIC FRACTURE MODE Fracture surfaces of static flexural test bars were analyzed using a scanning electron microscope (JEOL JSM 6480). Figure 9 shows fracture surfaces for various ceramics under quasi-static (low strain rate) failure. In a ballistic event, much higher strain rates exist at the location where the projectile strikes. However, as the distance from the impact location increases, the strain rate decreases. Thus, for regions away from the impact point, quasi-static fracture mode can provide an insight into the failure mechanism. As the strain rate increases, for most materials, the failure behavior becomes increasingly more brittle. In all the cases, static fracture is found to be transgranular, consistent with the measured toughness values. ANALYSIS OF BALLISTICALLY TESTED SPECIMENS During this study, 4" x 4" flat tiles and and medium sized double curved torso tiles were made and tested against low pressure and high pressure armor piercing (AP) rounds. The actual ballistic data cannot be presented here due to export control regulations. Microstructures of
Advances in Ceramic Armor V
.
77
Effects of Grain Size, Shape and Second Phases on Properties of Sintered Sic
polished fragments of some of the tiles tested and SEM observation of the same fragment are shown in Figure 10 (low pressure threat) and Figure 11 (high pressure threat). In the static fracture, the long grains showed very flat surface while in the ballistic fragments, the long grains showed steps. Correlations were developed between these microstructures and the ballistic performance to define the best process and microstructure for scaling up the manufacturing process for sintered S i c tiles for personnel armor to defeat AP rounds.
Figure 9. Static fracture surfaces of specimens. SUMMARY For each preforming method chosen, defects that affect component microstructure were elucidated (e.g. hollow granules, uneven granule size, inadequate pressure to eliminate intergranular porosity, knit lines - regions of lower packing in cast parts). Based on the basic sintering theory, and experimental observations, it was clear that preforms with uniformly high green density are desired.
78
.
Advances in Ceramic Armor V
Effects of Grain Size, Shape and Second Phases on Properties of Sintered SIC
Figure 10.
SEM and polished cross sections of ballistic fragments (Low pressure AP threat).
Advances in Ceramic Armor V
.
79
Effects of Grain Size, Shape and Second Phases on Properties of Sintered SIC
Figure 11.
SEM and polished cross sections of ballistic fragments (high pressure AP threat)
The search to find a correlation between static mechanical properties and ballistic performance continues. Although this may be an elusive goal, individual static properties do show correlations with specific microstructural features. The grain sizes evaluated here fall in a narrow range (2-8 micron). Also, since the goal of this study was product development and not a fundamental science study, grain size and porosity (density) were not controlled independently. Thus, as grain size was reduced, density also decreased and porosity increased by a few percent. Also, the grain size increase was not due to uniformly growing equiaxed grains but due to the development of a bimodal grain structure (small equiaxed grains and elongated grains). Fracture toughness did not increase significantly with increasing grain size and elongated grains. A clear correlation between hardness and grain size was not obtained within the grain size range explored due to (a) bimodal grain structure, (b) variations in hardness due to small indent size (100 pm) and related dependence on local microstructure, and (c) compounding effect of porosity. Flexural strength did not show a systematic correlation with grain size because strength is influenced by statistical distribution of anomalous (non-uniformly distributed) defects such as large grains, large pores, and inclusions (graphite, B4C particles). Strengths of the samples made by casting were higher than strengths of the samples made by pressing due to (a) the reduced scale of porosity in the cast parts (sub-micron particle scale) versus pressed parts (granule scale), and (b) preferred orientation of grains in the cast parts. REFERENCES www.dsm.com/en-USihtmlihpf/dyneema-ud.htm
'
80
.
Advances in Ceramic Armor V
Effects of Grain Size, Shape and Second Phases on Properties of Sintered SIC
* Ballistic
Resistance of Personal Body Armor, NIJ Standard 0101.04 (http://www.ojp.usdoj .gov/nij/topics/technology/body-armor/). M. van Es, J. Beugels, and J. vanDingenen, “Development of Dyneemdceramic hybrids for SAP1 inserts,” Proceedings of PASS 2002, 163-168. P. Karandikar, G. Evans, S . Wong, and M. Aghajanian, “A review of ceramics for personnel armor applications,” Ceramic Engineering and Science Proceedings, 29 [6], 2008,50-62. M. Aghajanian, B. Morgan, J. Singh, J. Mears and B. Wolffe, “A new family of reaction bonded ceramics for armor applications,” in Ceramic Armor Materials by Design, Ceramic Transactions, Vol. 134., J. W. McCauley et a1 editors, (2002) 527-540. C. Roberson and P. J. Hazell, “Resistance of different ceramic materials to penetration by a tungsten carbide cored projectile,” Ceramic Transactions 5 1, Ceramic Armor and Armor Systems (2003) 153-163. 7C. Roberson and P. J. Hazell, “Resistance of silicon carbide to penetration by a tungsten carbide cored projectile,” Ceramic Transactions 51, Ceramic Armor and Armor Systems (2003) 165174. N. J. Woolmore, P. J. Hazell, and T. P. Stuart, “An investigation into fragmenting the 14.5 mm BS41 armor piercing round by varying a confined ceramic target set up,” Ceramic Transactions 51, Ceramic Armor and Armor Systems (2003) 175-186. 9M2.Chen, J. W. McCauley, and K. J. Hemker, “Shock induced localized amorphization in boron carbide,” Science Vol. 299 [7] 2003 1563-1566. ‘OM. Chheda, M. J. Normandia, J. Shih, “Improving ceramic armor performance,” Ceramic Industry, January (2006) 124-126. ‘IS. Elliott, “Silicon Carbide ceramic armor,” Advanced Materials & Processes, 10 (2007) 29-33. 12M. Flinders, D. Ray, R. A. Cutler, “Toughness-hardness trade off in advanced Sic armor,” Ceramic Transactions 51 Ceramic Armor and Armor Systems (2003) 37-48. 13T.M. Lillo, D. W. Bailey, D. A. Laughton, H. S . Chu, and W. M. Harrison, “Development of a pressureless sintered silicon carbide for lightweight armor application,” Ceramic Transactions 5 1, Ceramic Armor and Armor Systems (2003) 49-58. I4K. Y. Chia, W. D. G. Boecker, R. S. Storm, “Silicon carbide bodies having high toughness and fracture resistance and method of making same,” U. S. Patent 5,298,470 (1994). I5M. Flinders, D. Ray, A. Anderson, and R. A. Cutler, “High toughness silicon carbide as armor, J. Am. Ceram. SOC.88181 (2005) 2217-2226. I6S. Prochazka, “Silicon carbide sintered body,” US Patent 4,041,117 (1977). 17J.A. Coppola, N. Hailey, and C. H. McMurtry, “Sintered alpha silicon carbide ceramic body having equiaxed microstructure,” US Patent 4,179,299 (1979). “F. F. Lange, “Powder processing science and technology for increased reliability,” Journal of the American Ceramic Society, 72 [ 11 (1989) 3- 15. 19 M. N. Rahman, Ceramic Sintering and Processing, Marcel Dekker (1995). ”5. J. Swab, “Recommendations for Determining Hardness of Armor Ceramics”, Int. J. Appl. Ceram. Tech., 1 [3] 219-25 (2004). 21 A. Wereszczak, H. Lin, and G. A. Gilde, “The effect of grain growth on hardness in hot pressed silicon carbides,” J. Mater. Sci. 41 (2006) 4996-5000 22R. W. Rice C. C. Wu, and F. Borchelt, “Hardness grain size relations in ceramics,” J. Am. Ceram. SOC.77 No. 4 (1994) 2539-2553. 23 H. Miyazaki, H. Hyuga, K. Hirao, and T. Ohji, “Comparison of fracture resistance of silicon nitrides with different microstructures,” J. Euro. Cera. SOC.27 (2007) 2347-2354.
’
Advances in Ceramic Armor V
81
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
INDENTER ELASTIC MODULUS AND HERTZIAN RING CRACK INITIATION
K. T. Strong, Jr., A. A. Wereszczak, and W. L. Daloz Ceramic Science and Technology Oak Ridge National Laboratory Oak Ridge, TN 3783 1 0. M. Jadaan College of Engineering, Mathematics, and Science University of Wisconsin-Platteville Platteville, WI 53818 ABSTRACT Hertzian ring crack initiation was studied in several brittle materials using spherical indenters made from different materials. Target materials included silicon carbide, silicon nitride, and borosilicate glass. Indenter materials included glass, zirconia, steel, silicon nitride, alumina, and tungsten carbide. Decreasing the elastic modulus of the indenter against every one of the target materials resulted in the target material exhibiting ring crack initiation at lower forces. The trend is consistent and conclusive. ‘ h i s response is described in regards to the Poisson’s mismatch between indenter and target and the associated friction effects that it produces. INTRODUCTION Hertzian or spherical indentation of brittle materials is a viable way to test the contact damage that occurs in many modem ceramic applications. Unlike indentation using pyramidal-shaped indenters (e.g., Knoop, Vickers, Berkovich, cube-comer), an advantage of Hertzian indentation is the target material first responds linear elastically before permanent contact damage (e.g., twinning, dislocations) ever begins. This is an attractive characteristic because it can allow the quantification of forces that initiate fiacture and plastic-like contact damage. For this study the mechanism of interest was ring cracking. It is the first fracture mechanism that occurs during Hertzian loading of brittle materials. A “large” or blunt indenter will cause fracture prior to any deformation of the ceramic material’, and this is the situation for the present study. Ring cracking is caused by the radial tensile stress that builds on the edge of the circular contact area between the indenter and target material. During Hertzian loading the spherical indenter contacts the flat target surface causing displacements of the surfaces to occur radially inward. This causes a region of compression to build in the center of the contact circle surrounded by a region of radial tension - the latter promotes ring crack initiation if the tensile stress is sufficiently large. Considering a frictionless case between the indenter and target material (i.e., their surfaces exhibit only mutual slip), the radial stress where ring cracking initiates can be calculated (i.e. strength of the brittle material) using the classical Hertzian equation 1-2v
PF
2
m2
u,&mu =--
’
where PF is the ring crack initiation force and v is Poisson’s ratio of the target material. The parameter a is the contact radius given by
83
Indenter Elastic Modulus and Hertzian Ring Crack Initiation
where R is the sphere radius, P is the applied compressive force, ET is the target elastic modulus, and
k = 9[(1y;>+ 16
,
(1
(3)
where v is Poisson’s ratio and the subscripts T and i represent the target and indenter, respectively.2 The classical Hertzian equation assumes a frictionless contact surface exists between the indenter and target material. However, fiiction indeed is present whenever there is a mismatch in elastic properties between the indenter and target materials (i.e., whenever they are different materials). Traction will occur on the target’s surface and it will act radially outward when it is more compliant and radially inward when it is more rigid in comparison to the indenter material. The relative stiffness or compliance of the target or indenter can be represented by the Dundurs Parameter (p)
’
=
(1 - 2vT)/GT- (1 - 2v1)/G, - vT)i G, + (1 - v,)/ G,
9
(4)
where G is the shear m o d ~ l u s . ~ The mismatch effect on ring cracking was experimentally shown on glass materials by J ~ h n s o n .If~ the coefficient of friction was known then the tensile stress of ring crack initiation could be calculated. Warren proposed that, during indentation when the spherical indenter contacts the target surface, the materials initially adhere to each other causing an area of stick in the center of the contact circle. This area of stick is then surrounded by an annular area of slip. If these areas can be measured then it is theorized that the coefficient of friction can be calculated and thus so can the strength of the materiaL5 The measurement of the coeficient was not examined during this investigation but will be considered in future work. EXPERIMENTAL PROCEDURE A custom-built computer-controlled indentation system was used for the Hertzian indentation testing. An electromechanical test machine was used to apply force fiom an attached spherical indenter onto a target material. An acoustic emission (AE) sensor was used to identify the acoustic event of ring crack initiation and its associated force (RCIF) during indentation. An illustration of the test setup is shown in Fig. 1. The AE sensor’s signal was consistently the most sensitive to acoustic activity when it was mounted on the surface of the target material. In preliminary testing, a threshold of 40 dB was found to be sufficiently high to eliminate background noise yet low enough to reliably detect the initiation of a ring crack. Indents were examined in an optical microscope with differential interference contrast (DIC or Nomarski) imaging to confirm that a detected acoustic event was actually the initiation of a ring crack. Ultraviolet optical microscopy with a dye penetrant was performed when optically hard to see ring cracks were present.
84
Advances in Ceramic Armor V
Indenter Elastic Modulus and Hertzian Ring Crack Initiation
Figure 1. Schematic drawing of the test configuration. An acoustic emission sensor was used to detect an acoustic event that was then linked to a ring crack initiation force. A displacement rate of 0.1 pm/s was used to press the spherical indenter against the target specimen until initiation of a ring crack was detected. The indenter was then rapidly unloaded. Approximately 25 - 30 indents were conducted for each target - indenter material combination. Commercial statistical software was used to fit the RCIFs to a two-parameter Weibull distribution using maximum likelihood estimation. A set of indenters with a range of elastic properties was chosen to test for, and observe, the mismatch effect between the indenter and target materials. The indenter materials used, with their respective sizes and elastic properties, are displayed in Fig. 2. The specific diameters were chosen to ideally produce the,same contact areas between the indenter and target material based on their respective elastic properties. Si3N4 was designated as a reference material and a diameter of 3.00 mm was chosen as a reference size. The remaining indenter sizes were chosen with respect to the Si3N4 ball according to Eq. 5
where R is the radius of the indenter and the subscripts 1 and 2 are the Si3N4 and calculated indenters, respectively .4
Advances in Ceramic Armor V
.
85
indenter Elastic Modulus and Hertzian Ring Crack initiation
ZrO,
Glass
0
0
steel 0 = 2.38 mm E = 214 GPa v = 0.276
= 1.00 mm
E = 70 GPa v = 0.191
wc
Si3N,
= 2.38 mm E = 213 GPa v = 0.304
0
= 3.00 mm
= 4.00 mm E = 630 GPa v = 0.211
0
E = 312 GPa v = 0.276
= 3.18 mm E = 371 GPa v = 0.238
0
Figure 2. Schematic drawing showing the indenter materials and diameters used. With these, approximately the same contact area and average contact pressure is produced for a given compressive force (P). Balls are drawn to 8x scale. The target materials chosen for testing were a hot-pressed silicon carbide (SIC), a hot-pressed silicon nitride (Si3N4) and a borosilicate glass. Each material's respective elastic properties are displayed in Table I. Scanning electron microscope images show the microstructure of the polycrystalline materials displayed in Fig. 3. It should be noted that the size of the Hertzian contact areas in comparison to the grain sizes of the materials is large; therefore, these measurements sample a bulk material response. Like the ball indenter materials, the target materials were chosen for this study because their elastic moduli spanned a wide range of values.
,
86
Table I. Elastic Properties of the Target Materials. Elastic Modulus Target M;;rial
Si3N4 Borosilicate
Advances in Ceramic Armor V
$1
Shear Modulus
( G )
Poisson's Ratio ~
,
-'-
0.170 0.273 0.199
indenter Elastic Modulus and Hertzian Ring Crack Initiation
Figure 3. Scanning electron microscope images of microstructures on polished surfaces of S i c and Si3N4. RESULTS AND DISCUSSION Measured characteristic RCIF values were used rather than stress for the analysis. RCIF is a measurable value whereas stress is calculated and dependent on the friction (is., an unknown) that occurs between the indenter and target material. The results showed friction was operative in this testing because of the elastic mismatch between indenter and target materials. The characteristic RCIF are plotted as a function of the indenter elastic modulus and Dundurs Parameter f3 in Fig. 4. The drawn trendlines are fitted to the balls that remain linear elastic throughout testing. The RCIF increases as the indenter modulus and Dundurs Parameter increases. This is because of friction. The forces that initiate ring cracking should be the same for all ball materials if the classical Hertzian theory were obeyed; namely, the fitted curves in Fig. 4 should be horizontal if there were no fiiction. The data points for the steel indenters against the S i c and Si3N4 target materials do not fall on the linear trends drawn on Fig. 4. This is due to the steel ball yielding during indentation whereas the other balls remain linear elastic. However, it overlaps the Zr02 indenter data on the borosilicate glass target, which could be expected since ZrOl and steel have similar elastic properties. FEA analysis was conducted for the case of the steel indenter on all three target materials at the characteristic RCIF. A linear elastic case was considered for ease of the model and the maximum von Mises stress was determined in the steel ball. For the case of the Si3N4 and SIC, the von Mises stress in the steel ball was 2x and 3x larger than that of the yield stress of the steel ball material (2.03 GPa). The borosilicate glass caused a maximum von Mises stress in the steel ball only slightly above that of its yield stress. The Hertzian indentation with steel balls creates extra complications in the target ring crack initiation analysis because of its potential to yield (and its consequential change in radius of curvature). The Si3N4-indenter on Si3N4-target and glass-indenter on glass-target test cases are special cases because their elastic properties are matched. For only this case, the contact surfaces are believed to exhibit only slip, or the classical Hertzian case.4 From Fig. 4, when the indenter material is more rigid than the target material, higher RCIFs result. Whereas lower RCIFs result when then indenter is more compliant. The observed elastic mismatch effect on RCIFs is due to the effect of friction on the radial tensile stress on the target material's surface. The net radial stress profiles for various indenter
Advances in Ceramic Armor V
.
87
Indenter Elastic Modulus and Hertzian Ring Crack Initiation
materials in contact with Sic, Si3N4 and the borosilicate glass target under a no-slip condition are shown in Fig. 5 . Maximum radial tensile stress decreases as the indenter elastic modulus increases as Johnson and Warren For more rigid indenters, higher forces are necessary to raise the tensile stress to a value where ring cracking occurs, alternatively for a more compliant indenter, lower forces are necessary.
0
200
100
300
400
700
600
500
Indenter Elastic Modulus, E (GPa)
1600
8 g Y
s
e
+ S13N4
BS Glass
1200 1000
600
200 0 -0.100
0,000
0.100
0.200
0.300
0.400
Dundurs Parameter, 6
Figure 4. Characteristic RCIF of the materials and their respective indenters as a function of (a) indenter elastic modulus and (b) Dundurs Parameter. Drawn trendlines are fitted for balls that sustain linear elasticity throughout testing.
88
Advances in Ceramic Armor V
Indenter Elastic Modulus and Hertzian Ring Crack Initiation
Figure 5. Comparison of radial stress profiles for different indenter materials in contact against a (a) Sic, (b) SijN4, and (c) borosilicate glass target under conditions of no-slip. Shown diameters generate the same contact area and pressure for a given applied compressive force.
Advances in Ceramic Armor V
.
89
Indenter Elastic Modulus and Hertzian Ring Crack Initiation
CONCLUSIONS Characteristic RCIFs of a hot-pressed Sic, a hot-pressed SisN4, and a borosilicate glass were measured using different indenter materials and specially chosen diameters. The results showed that an elastic mismatch effect between the indenter and target material was operative, and consequential frictional forces affected the radial tensile stresses which in turn affected the forces of ring crack initiation. With respect to any target material tested, a more rigid indenter caused higher RCIF whereas a more compliant indenter lowered it. ACKNOWLEDGEMENTS Research sponsored by the Work for Others sponsor US Army Tank-Automotive Research, Development, and Engineering Center under contract DE-AC05-000R22725 with UT-Battelle, LLC. The authors would like to thank the: U.S. Army TARDEC’s D. Templeton, U S . ARL’s J. LaSalvia, J. Campell, P. Patel, J. Swab, and J. McCauley, ORNL’s H.-T. Lin, and former US.ARL’s (currently at Ceradyne) M. Normandia. Additionally, the authors thank ORNL’sJ. Hemrick and A. Shyam for their helpful comments. REFERENCES 1. R. Mouginot, “Blunt or Sharp Indenters: A Size Transition Analysis,”J. Am. Ceram. SOC.,71:658661 (1988).
2. K. L. Johnson, Contact Mechanics, Cambridge University Press, Cambridge, United Kingdom, 1985. 3. J. Dundurs, “Edge-Bonded Disimilar Orthogonal Elastic Wedges Under Normal and Shear Loading,”J. Appl. Mech., 36:650-652 (1969). 4. K. L. Johnson, J. J. O’Connor, and A. C. Woodward, “The Effect of the Indenter Elasticity on the Hertzian Fracture of Brittle Materials,” Proc. R. Soc. Lo&., 334:95-117 (1973).
5. P. D. Warren and D. A. Hills, “The Influence of Elastic Mismatch Between Indenter and Substrate on Hertzian Fracture,” J. Mar. Sci., 29:2860-2866 (1994).
90
.
Advances in Ceramic Armor V
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
HIGH FREQUENCY ULTRASOUND OF ALUMINA FOR HIGH STRAIN-RATE APPLICATIONS S. Bottiglieri and R. A. Haber Department of Materials Science and Engineering, Rutgers University Piscataway, NJ, USA
ABSTRACT Several series of high density armor-grade alumina were tested using high frequency ultrasound in order to determine any differences between multiple manufacturers and production lots. C-scans were performed at 1SMHz in order to form attenuation coefficient and elastic property maps. Qualitative and quantitative analysis enabled the separation of these samples into various categories based on their overall variability and homogeneity. Ultrasonic attenuation coefficient was chosen as the property to distinguish different regions per sample to study. The microstructure was investigated using SEM and XRD techniques to understand differences seen in the ultrasonic C-scan images. Results indicate that there is a qualitative microstructural difference when comparing sample regions initially characterized by ultrasonic NDE. INTRODUCTION There is a present motivation for research in ceramic-based composite armor systems. The low production cost of metal-based armor systems, in general, is their main advantage. The main requirement for armor ceramic systems is to decrease weight while providing the same degree of protection when compared their metal-based counterparts. Ceramics, such as alumina (A1203), can offer similar protective capabilities'. Ballistic tests have indicated certain mechanical properties, such as higher elastic properties, are generally more desirable for protective systems. However, all factors; including mechanical properties, production ability, and type of ballistic threat, must also be considered when making an armor system2. Alumina is an attractive material for armor applications as it is generally less expensive than silicon carbide (a competing armor ceramic material). Silicon carbide (Sic) and other non-oxide ceramics used as armor material are usually made by hot pressing, high temperature sintering, or another form of processing that is more expensive than how alumina can be made; slip casting, injection molding and sintering, etc3. The lower production cost of A1203 may offset the penalty paid for increase in density when compared to S i c or othernon-oxide ceramics. Many practical nondestructive testing techniques are implemented to detect large critical-size defects to ascertain the quality of macroscopic systems. However, the impetus for ultrasonic NDE techniques to go beyond flaw detection is being realized for armor ceramic systems. An armor ceramic material may be free of critical-size defects but can still be vulnerable to failure under high strain rates and/ or due to poorly distributed mechanical properties4. Ultrasonic NDE has been shown to be a useful tool in determining the homogeneity and variability of certain mechanical properties within a sample*. Knowing if and what property gradients exist in samples can give insight into how a production process may be optimized; i s . thermal fluctuations/gradients in furnaces, poorly mixed powders, mold filling directions, etc. Previous work showed that ultrasonic attenuation coefficient may correlate most strongly to ballistic performance when compared to other properties measured by ultrasonic methods'. Ultrasonic attenuation coefficient is a materials parameter that quantifies the reduction in the intensity of sound as it propagates through a material6. Acoustic attenuation is affected by the grain size of a material and the frequency of ultrasound that is being used. A preferable material would then be one in which there is less attenuation occurring thereby implying that it has a high degree of homogeneity6. Attenuation due to scattering or absorption in polycrystalline ceramics is dependent on the constituents of its microstructure (grains, pores, inclusions, second phase, etc.). This makes using the
91
High Frequency Ultrasound of Alumina for High Strain-Rate Applications
acoustic attenuation parameter of a material as a possibility of directly characterizing microstructure’. The research reported in this manuscript presents the basic theory of acoustic attenuation coefficient and a preliminary comparison of overall attenuation coefficient values and SEM micrographs. THEORY Ultrasonic attenuation in a material can be described by the Beer-Lambert law: I = I, e-ad
(1)
where a is the attenuation coefficient, d is the distance traversed by the ultrasonic wave (in units of cm), I, is the initial intensity of the ultrasonic wave, and I is the resultant ultrasonic intensity after being attenuated by the material. The experimental setup for this study uses a pulse-echo technique thereby making the ultrasound travel twice the thickness of a sample. Equation (1) must be modified to account for what is capable of being measured and reflective losses. The ultrasonic intensity terms, I and I,, correspond to the first and second bottom surface reflections of the sample. These terms equal the square of the respective measured amplitudes. Losses in ultrasound energy due to Fresnel reflection are inherent due to the acoustic impedance mismatch of two media at a surface e.g. immersion based testing will have an impedance mismatch caused by the liquid-sample interface) . For testing done in this research distilled water was used as the immersion media and A1203 as the sample causing a reflection of acoustic energy (at each surface) of 86%. The equation used to determine the reflection coefficient is shown below:
d
R = (Z, - Z2)2/ (Zi + Z2)2
(2)
where Z1 and Z2 are the acoustic impedances of the two media at a surface. After an iterative algebraic process, carefilly accounting for each reflection and how much ultrasound energy is being reflected and transmitted at each surface, the modified equation for attenuation coefficient used in this study is: a = [-8.686/ (2 . d)]
* [In{(vIo)2}- In(R2)]
(3)
where a is in units of dB/cm. As mentioned above, the attenuation of ultrasound in a material is frequency and feature size dependent. The total attenuation in an elastic solid is described by the summation of several attenuative factors (equation (4)) which are findamentally related to the bulk material and any inhomogeneities present within the sample’. UT = alnternal Absorption + aOrain Scattering + ainhornogeneity Absorption + alnhomogmeity scanning
(4)
Current methods use a deconvolving technique, such as an FFT algorithm, to understand the frequency dependent attenuative nature of studied samples at a single data point. A C-scan (many thousands of data points) image of the attenuation coefficient is representative of the overall loss of ultrasound energy over the entire frequency spectrum put out by a transducer. Understanding why a material attenuates specific frequencies can lend valuable microstructural information. This information can be known by understanding the specific attenuative factors shown in equation (4). The presented research shows initial results of possible reasons why an , 4 1 2 0 3 microstructure causes varying regions of attenuation coefficient as seen in ultrasonic C-scan images.
92
Advances in Ceramic Armor V
High Frequency Ultrasound of Alumina for High Strain-Rate Applications
EXPERIMENTAL METHOD The preliminary research shown here is a start in offering predictive nondestructive knowledge into an alumina microstructure. To understand a nondestructive technique filly one must verify that it is indeed correlated to microstructure and destructive testing techniques. The samples looked at are commercial sintered alumina tiles having dimensions of 90 x 45 x 10mm. Ultrasonic nondestructive evaluation (NDE) was performed on several series of alumina tiles. Properties were measured using a 15MHz broadband, focused, ultrasonic transducer provided by TRS technologies. The equipment and components used for this study include the following: iMoveGANTRY-500, JSR Ultrasonics DPR 500 Dual Pulser/ Receiver, A D converter card, Synapse card, and Secondwave Research Studio software, all provided by the Ultran Group. The measured properties included attenuation coefficient, longitudinal and shear velocity, and Young’s modulus. The manufacturer of these samples provided information on which series showed “good” and “bad” ballistic responses. Previous ultrasonic NDE was performed to determine any possible correlation between ballistic performance (as told by the manufacturer) and the measured properties. Results showed a strong correlation between what the manufacturer considers “good” or “bad” and the overall homogeneity, variability, and average value of attenuation coefficient’. Most alumina tiles provided by the manufacturer originally showed large variations in attenuation coefficient. There is also the presence of circular artifacts (see Figure I).
Figure I. 15MH.z attenuation coefficient C-scan of A1203 tile. Large circular artifacts present. According to the production company these artifacts are too large to be single point defects. An initial assumption is that there are clusters of unmixed sintering aids which act as a single artifact when interrogated with the ultrasound beam. The concentric circle pattern appears to be a resonant effect due to the use of a focused transducer. The sectioning of A1203 tiles into regions of low and high attenuation coefficient and regions containing the circles was done in order to investigate and draw correlations as to why these differing values and inhomogeneities in an ultrasonic C-scan image are seen.
Advances in Ceramic Armor V
.
93
High Frequency Ultrasound of Alumina for High Strain-Rate Applications
For this study a Zeiss Gemini 982 Field Emission Scanning Electron Microscope (FESEM) was used to scrutinize the microstructure of the differing regions (low/ high attenuation coefficient and circle). Two alumina samples were chosen, which showed high variability and many acoustic circles in their attenuation coefficient C-scan maps. These samples are approximately 90 x 45 x lOmm and have densities (Archimedes) of 3.85dcc. The regions which were sectioned are labeled as L1, L2, H1, H2, C1, and C2 to represent low and high attenuation coefficient and a circular area. The number next to L, H, or C indicates which sample it is (arbitrarily chosen). The 15MHz ultrasonic attenuation coefficient C-scan images are below in Figure 2 and Figure 3. Please note the slightly different scales used in order to draw attention to the variability of each tile.
Figure 3. 15MHz attenuation coefficient C-scan image of All03 tile number 2. The regions of high and low attenuation coefficient vary by roughly 1.5 - 2.0 dB/cm while C1 and C2 have attenuation coefficient values of approximately 8.6 and 9.6dB/cm, respectively. The differences in attenuation coefficient values must be caused by differences in microstructure, all other things being equal (such as frequency and testing NDE testing conditions). Alumina tile number 1 was originally sectioned around C1 in order to ascertain the feasibility of continuing with microscopy on the rest of the sections. This was due to a concern that the differences in attenuation coefficient and the acoustic circles detected by ultrasound may be too small or indistinguishable when looking for them with the FESEM. The sections were sliced vertically through the areas of interest and the vertical orientation is the same for each of the resulting micrographs. Below in Figure 4 is an image from the preliminary look at section C1, done to verify that possible features would be observed.
94
.
Advances in Ceramic Armor V
High Frequency Ultrasound of Alumina for High Strain-Rate Applications
Figure 4. SEM image of section C1. Figure 4 shows an SEM image of a representative area of section C1. In order to perform EDS on the feature shown in Figure 4, the accelerating voltage was increased from 2.89kV to 1OkV. This increase in voltage caused the sample to charge, which is why the image in Figure 4 shows the white sections. This feature (-15pm) appeared in a cluster of similar looking inclusions (not shown here). To understand what this inclusion was, energy dispersive spectroscopy (EDS) was performed. The results, Figure 5 and Figure 6 , indicate that these inclusions are most likely CaO, a known sintering aid used in the production of these samples. The presence of these CaO clusters adds credence to the initial idea that the acoustic circles are possibly caused by a cluster of inhomogeneities acting as a single resonator.
Figure 5. EDS analysis of inclusion seen in Figure 4.
Advances in Ceramic Armor V
95
High Frequency Ultrasound of Alumina for High Strain-Rate Applications
Figure 6 . EDS map analysis of inclusion seen in Figure 4. The EDS spectra and map show that the inclusions are most likely CaO. If allowed to run longer and collect more data, it seems likely that there would be more oxygen coinciding where calcium is located in the point map. According to the manufacturer, CaO is more likely to appear than pure metallic calcium. The presence of Si may be an artifact of the detector used to record this data; a silicon scintillator material is used in the detection of X-rays". However, Si02, another known additive which may be used and the detected CaO appear to be the result of the sintering additives used. With regards to what may be causing the acoustic circles, there is evidence that they may be due to unmixed CaO, or other sintering aids used during the production of these alumina tiles. A preliminary look as to what may be causing the acoustic circles showed results compatible with how the manufacturer processed these alumina samples. To further understand the microstructure, and what possible causes it could be having on the differing values of attenuation coefficient, all of the sectioned samples mentioned above were polished and then etched in order to reveal grains and fine microstructure. An organized discussion of each section follows below. RESULTS AND DISCUSSION SEM images of each section were taken at several magnifications for comparison. Any features that were present and prominent were examined more closely. The images below (Figures 7 17) show the microstructure of alumina samples which were polished and then thermally etched to reveal grains. Table I shows the densities of the entire tiles and the sectioned samples. It is clear that there is a large density variation in each sample which is correlated to the differences seen in attenuation coefficient. The total density for each sample is a simple linear rule of mixing of every point in the sample. There are three points for each sample (L, H, and C) which have been independently tested for density (shown in Table I). Hence, for Tile 2, the low total density implies there are other regions of the sample which have lower density than the values shown. Table I. Density values (gicc).
I
Tile1 Tile2
96
.
1
I
Advances in Ceramic Armor V
Total
1
1
1
H
I
C
3.85 1 3.93 I 3.80 I 3.88 3.86 1 3.93 1 3.89 I 3.89
I
High Frequency Ultrasound of Alumina for High Strain-Rate Applications
Regions of Low Attenuation Coefficient: L l and L2 Figure 7 shows images of the general microstructures of samples L1 and L2 at 500x magnification. According to ultrasound theory, an area of lower attenuation coefficient should be more homogeneous and have fewer attenuators (a feature which may scatter or absorb ultrasound energy). Samples L1 and L2 have average attenuation coefficient values of 8.6 and 9.3dB/cm, respectively. As the ultrasound measurements are volumetric, in which the beam diameter is much larger than the cross sectional area of each sample section, there is the likelihood that the surfaces shown in the SEM micrographs do not show large variations for such a small difference in attenuation coefficient between samples. At this magnification there appears to be more porosity and pullout for sample L1 than there is in sample L2.
Figure 7. SEM images of sections L1 (left) and L2 (right) at 500x magnification. At higher magnifications, as seen in Figures 8 and 9, it is clear that there is an increase in what appears to be glassy features distributed throughout the bulk when looking from sample L1 to sample L2.
Figure 8. SEM images of A1203 microstructure of sample L1 at 5000x.
Advances in Ceramic Armor V
.
97
High Frequency Ultrasound of Alumina for High Strain-Rate Applications
Figure 9. SEM images of A1203 microstructure of sample L2 at 5000x. Circled areas show representations of glassy features. The lighter features seen between the grains in Figure 9 are qualitatively distinguishable as a possible glassy phase due to the contrast differences caused by atomic number. These were c o n f i i e d to be glassy features by EDS and the manufacturer through SEM analysis as well. One concern is that the glassy features did not evolve and redeposit as they did in the circular region samples. This may be that these smaller glassy features are not the exact composition as the ones seen in the circular regions, which appear to be much larger. In the low attenuation coefficient regions (as seen above) they are approximately on the same size order of a grain; about 1 - 7pm. Overall, at higher magnifications, there does appear to be qualitative reasons as to why there is a slightly higher attenuation coefficient when comparing L2 to L1. Regions of High Attenuation Coefficient: HI and H2 Sample HI has an overall attenuation coefficient value of 9.2dBicm while sample H2 has a value of IO.2dB/cm. There should be an increase in the amount of inhomogeneities in sample H2 when compared to other sections of lower attenuation coefficient. However, this may not necessarily be seen for the same reasons as mentioned above. Shown below in Figure 10 are general SEM images of samples H1 and H2 at 500x magnification.
Figure 10. SEM images of sections HI (left) and H2 (right) at 500x magnification.
98
Advances in Ceramic Armor V
High Frequency Ultrasound of Alumina for High Strain-Rate Applications
When compared to each other, the microstructure of HI appears to have a slightly higher degree of homogeneity than that of H2. This is determined on a qualitative basis. When comparing each sample, HI and H2, to the respective low attenuation coefficient samples, L1 and L2, there is a noticeable increase the amount of porosity and pullout for the regions of high attenuation coefficient.
Figure 11. SEM images of sections HI (left) and H2 (right) at 5000x magnification. Figure 11 shows high magnification images (5000~)comparing samples H1 and H2. The alumina samples seen in Figure 11, H1 and H2, show signs of abnormal grain growth, more porosity/ pullout, and re-deposits of a glassy phase (H1 (left)). These features are all causes as to why sections H1 an H2 have overall higher attenuation coefficient when compared to sections L1 and L2. Sections L1 and L2 show much less variability in the distribution of size and shape of the A1203 grains present when compared to HI and H2. The increase in attenuation coefficient (by about 1 dBicm) seen from sample H1 to sample H2 can be caused by differences in porosity, inhomogeneities, or grain size (as mentioned above). Due to the differing density values from H1 to H2 it seems likely that the increase in attenuation coefficient is related to and increase in poorly dispersed sintering additives. Hence, the increase in gaps between grains seen in sample H2 may be caused by pullout of a glassy phase present in that region. Regions of Acoustic Circles: C 1 and C2 The acoustic circle regions in each sample can be described as a high attenuation coefficient circle with a center of lower attenuation coefficient. The average values of the entire circular regions for each section, C1 and C2, are 9.0 and lO.OdB/cm. When comparing these attenuation coefficient values and the recorded densities for each section it does seem likely that these acoustic circles would be caused by clusters of inhomogeneities, such as sintering aid inclusions. The theoretical density of pure A1203 is 3.97 glcc while common alumina sintering additives such as CaO and NalO have theoretical densities of 3.35glcc and 2.27glcc, respectively. The relative high density of these circular regions imply that there are less lower density sintering additives when compared to a high attenuation coefficient region. However, the qualitative results showing the acoustic circle imply that these low density sintering additives are clustered together. This offers a possible reason as to why these regions of inhomogeneity, shown in the attenuation coefficient C-scan images, do not show a much higher value of attenuation coefficient. The overall microstructure should be more like that of a low attenuation coefficient region, but have the same attenuative inhomogeneities present as a region of high attenuation coefficient region in the form of clusters. Our current speculation is that the clusters
Advances in Ceramic Armor V
. 99
High Frequency Ultrasound of Alumina for High Strain-Rate Applications
interact with the ultrasound as a single resonator. These clusters are then shown as the acoustic circles seen in the attenuation coefficient C-scan images. This concept is taken from ultrasound work done with colloidal systems and is not yet readily understood in elastic solids”. Figures 12 and 13 show regions of sample C1 at 500x and 875x magnifications. These regions show that there are more localized regions of porosity and pullout. These two images are representative of what was seen throughout the entire section.
Figure 12. SEM image of section C1 at 500x magnification.
Figure 13. SEM image of section C1 at 875x magnification. When compared to other sections there are instances of larger porosity and regions of porosity clusters. An interesting feature found (Figure 14) appears to be the remnant of a large inclusion (approximately 90pm). The crater would seem to be left by a large glassy agglomerate. The act of thermally etching the section caused what was in the crater to melt out and redeposit.
100
Advances in Ceramic Armor V
High Frequency Ultrasound of Alumina for High Strain-RateApplications
Figure 14. SEM image of section C1 at lOOOx magnification. Shows large crater and porosity. The porosity contained within the crater in section C1 may be caused by the partial evolution of the inclusion during sintering. Sample C2 showed a similar crater with the same internal porosity most likely caused by the same processing conditions and additive. This is seen below in Figure 15.
Figure 15. SEM image of section C2 at lOOOx magnification. Shows large crater and porosity. These craters are only seen in the acoustic circle regions which make them be the probable cause of the circles seen in the ultrasonic C-scan maps. A qualitative answer as to why the cavity was left by an inclusion which melted and redeposited during the etching process is shown in Figure 16. These images show the inside of the crater (left image) and the immediate outside (right image).
Advances in Ceramic Armor V
.
101
High Frequency Ultrasound of Alumina for High Strain-Rate Applications
Figure 16. SEM images of inside (left) and outside (right) of crater in section C2 at 5000x magnification. The right image of Figure 16 show ‘pimples’ or bumps on the grains which are indicative of the glassy inclusion melting and redepositing on the surface beside it. The contrast difference between the two images implies that the surface directly outside the crater is covered by a thin layer of the redeposited inclusion. The bumps appear to be areas which started to accumulate more of the glassy inclusion. The same bumps were seen next to the crater seen in C1 (as seen in Figure 17).
Figure 17. SEM image of surface outside of the crater in C1. Magnification of 1 0 0 0 0 ~ . Energy dispersive spectroscopy was performed in order to determine the composition of the bumps on the grains. However, the penetration depth of the electron probe used for this analysis is greater than the size of the bumps. Hence this analysis did not (and could not) provide information on exactly what the bumps are; but only what the underlying material is, alumina. Interrogation of the microstructure of sections C1 and C2 reveal that the acoustic circles are most likely caused by large glassy inclusions which are byproducts of partially/ unmixed sintering additives. CONCLUSIONS The initial study of the correlation of ultrasonic attenuation coefficient to microstructure for selected alumina samples was performed. Analysis utilized ultrasonic testing methods and qualitative
102
Advances in Ceramic Armor V
High Frequency Ultrasound of Alumina for High Strain-Rate Applications
SEM information. Preliminary results show that there is an obvious direct correlation between the quality of an alumina microstructure and the overall loss of ultrasound energy measured at 15hfHz. The correlation for these samples appears to be strongly governed by the distribution and type of glass sintering additive used in production. The acoustic circles are caused by clusters of large unmixed additives and porosity possibly caused by the outgassing of these additives. Even without the compositional analysis of the bumps seen outside the craters it is clear that they are artifacts left over from glassy inclusions. Out of the possible inclusions which may have been artifacts or inhomogeneities left in during production, it seems likely that CaO was the large inclusion preceding the crater left by the thermal etching process. A larger sampling and more quantitative testing must be done in order to be able to conclusively draw predictions on an alumina microstructure based off of ultrasonic NDE techniques. ACKNOWLEDGEMENTS The authors would like to thank the financial support of NSF WCRC Ceramic and Composite Materials Center and the Army Research Laboratory Materials Center of Excellence for lightweight vehicular armor. REFERENCES ‘E. Medvedovski, “Alumina Ceramics for Ballistic Protection Part 2”, American Ceramic Society Bulletin, Vol. 81, No. 4,45-50 (2002). 2R. Brennan, “Ultrasonic Nondestructive Evaluation of Armor Ceramics”, Ph.D. Thesis, Rutgers University (2007). 3E. Medvedovski, “Alumina Ceramics for Ballistic Protection Part l”, American Ceramic Society Bulletin, Vol. 81, No. 3,27 - 32 (2002). 4H. Karagulle, “Analysis of signals generated by multiply reflected ultrasonic waves in plates”, Ultrasonic Methods in Evaluation of Inhomogeneous Materials, Series E: No. 126, (1987). ’S. Bottiglieri, R.A. Haber “High frequency ultrasound of armor-grade alumina ceramics”, In ublication of QNDE proceedings, June 22, Chicago, IL, 2008. %TIT Resource Center: Collaboration for NDT Education, http://www.ndt-ed.org ’A. Evans, “Ultrasonic Attenuation in Ceramics”, Journal ofApplied Physics, Vol. 49, No. 5, (1978). ‘L. Lynworth, Ultrasonic Measurements for Process Control, Publisher: Academic Press Inc., San Diego, 1989. 9A. Portune, R.A. Haber, “Role of microstructure in ultrasound response for armor ceramics”, In Ytblication of QNDE proceedings, June 22, Chicago, IL, 2008. J. Goldstein, Scanning Electron Microscopy and X-Ray Microanalysis, Publisher: Kluwer Academic Publishers, 2002. ”A. Sdukhin and P. Goetz, “Ultrasound for Characterizing Colloids”, Elsevier Science B.V. (2002).
Advances in Ceramic Armor V
.
103
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
THE EFFECT OF PARTICLE SIZE, PARTICLE LOADING AND THERMAL PROCESSING CONDITIONS ON THE PROPERTIES OF ALUMINA REINFORCED ALUMINUM METAL MATRIX COMPOSITES Allyn L. McCormick, Michael K. Aghajanian, Andrew L. Marshall M Cubed Technologies, Inc. 1 Tralee Industrial Park Newark, DE 197 11 ABSTRACT Particle reinforced metal matrix composites (MMCs) have shown viability in many applications, including thermal management, precision equipment, automotive components, wear and armor. The composite materials, as compared to their matrix metals, exhibit higher strength to density ratios, higher stiffness to density ratios, better fatigue resistance, better elevated temperature properties (higher strength, lower creep rates), and better wear resistance. The vast majority of work to date has examined AliSiC particle reinforced composites. The present work examines AI/A1203 composites. Relative to SIC, A1203 offers many advantages as a reinforcement in aluminum alloys, including lower cost, no reactivity with aluminum (allows a wide range of alloy chemistries), better CTE match to aluminum (reduces residual stress), and composites with improved corrosion resistance. Mechanical and physical properties of these materials can easily be manipulated via reinforcement size, shape, volume loading and thermal processing conditions. The present study examines the effects of A1203 size and content on microstructure, physical properties, and mechanical properties of AIIA1203 composites. INTRODUCTION Particulate reinforced metal-matrix composites are of interest for various applications due to their improved mechanical properties relative to unreinforced alloys. The discontinuous reinforcement provides isotropic properties as compared to that of continuous-fiber-reinforced composites. Moreover, these materials are versatile and can be tailored for specific uses by altering processing conditions and/or raw materials. For instance, factors such as the alloy chemistry, the reinforcement shape (particulate, platelet, whisker, etc.), the reinforcement chemistry (A1203, Sic, etc.), the reinforcement loading, the processing method, post heat treatment, and cold work can have a significant impact on the structural behavior of the resultant composite. Due to the tailorability of these materials it is important to know the associated impact that manipulating any of these variables has on the resultant properties and thus aid in the design aspect of the composite. To this end, the present paper examines the effect of two variables, namely reinforcement size and loading, on the microstructure, physical and mechanical properties of the composite. Discontinuously reinforced metal matrix composites are fabricated using a variety of methods, including the powder metallurgy technique (mixing of ceramic and metal powders followed by solid state hot pressing) and compocasting (the mixing of ceramic particles into molten metal, followed by casting). Systematic studies have been performed and data exist characterizing the effect of particle loading on the mechanical behavior for both the powder metallurgy technique’ and compocasting’. The present work utilized a pressureless liquid metal infiltration technique3’ for fabrication of the test composites, which ranged in A1203 loading
105
Effects of Particle Size, Particle Loading and Thermal Processing Conditions
from 32 to 46 volume percent. This technique allows the fabrication of composites with different reinforcement contents without affecting other microstructural features. EXPERIMENTAL PROCEDURES Four composite billets were fabricated to examine the variables of reinforcement size and reinforcement loading. In each case a binary AI-Mg alloy was infiltrated into a loose compact of reinforcement material. In all cases white fused A1203 was chosen as the reinforcement material. As detailed in Table I, the A1203 reinforcement ranged in size from 12 to 154 pm, with particulate loadings varying from 32 to 46 volume percent. The billets were fabricated by compacting the A1203 reinforcement particles into a graphite container after which a binary Al8Mg alloy was placed atop the compressed powder. The alloyireinforcement pairs were then placed into a controlled atmosphere furnace at room temperature with a flowing nitrogen atmosphere. The furnace was ramped to an ultimate temperature of 825°C and held for an appropriate time to fully infiltrate the reinforcement. The furnace was then cooled to 700°C at which time the graphite boats were removed from the furnace and placed on a cold plate to allow directional solidification. This controlled solidification process yielded composites free of solidification porosity. Table I: Reinforcement Characteristics of AIiA1203 Composites Evaluated
TEST PROCEDURES Physical and mechanical properties were measured with the test methods shown in Table 11. Density was measured only once per material on the bulk billet of composite. Young’s modulus was measured at three locations on each billet. Flexural strength, tensile strength and fracture toughness were each measured using five to ten samples. The mechanical tests utilized a Sintech universal test frame in conjunction with Test Works materials testing software. Microstructures were evaluated using a Leica D 2500 M optical microscope and the Clemex Vision PE imaging software. Table ZI: Test Methods
106
Advances in Ceramic Armor V
Effects of Particle Size, Particle Loading and Thermal Processing Conditions
The flexural strength and fracture toughness test methods employed during this study are primarily suited for brittle ceramic-based materials. The present AIIAIIO~ composites characterized, however, display some plasticity due to the presence of the ductile A1 alloy phase. Therefore, the results of these tests are not fully quantitative. Nonetheless, they are useful for comparison purposes. It is recognized that more focus needs to be placed on tensile-based properties as this work proceeds. RESULTS AND DISCUSSION Microstructures of representative samples are provided in Figure 1. Constant magnification was used for the coarser (larger particle size) composites. For the finest composite (12 pm particle size), higher magnification was needed to image the particles. Complete wetting with no porosity is seen in all four composite systems.
C
D
Figure I : Optical photomicrographs of AI2O3/AIcomposites: A: 154 pm A1203 B: 58 pm A1203 C: 44 pm A1203(58112 pm blend) D: 12 pm AI2O3(higher magnification)
Advances in Ceramic Armor V
.
107
Effects of Particle Size, Particle Loading and Thermal Processing Conditions
A summary of all property data is provided in Table 111. The results of density and Young’s modulus follow the reinforcement content of the composite, as expected, due to the higher density (4.0 vs. 2.7 gicc) and higher Young’s modulus (380 vs. 68 GPa) of the AlzO3 The strength and toughness data show much reinforcement relative to the A1 alloy matrix5’ less of a reinforcement content effect. Instead, the trends are strongly associated with reinforcement size. In particular, the composite containing the smallest (1 2 pm) reinforcement particle has far superior structural properties, whereas the use of the coarsest particles (154 pm) led to significantly lower properties than the other composites. The trend even held with the relatively small change from 44 to 58 pm reinforcement size, with the finer size leading to small increases in flexural strength, tensile strength and fracture toughness.
‘.
Table ZZP Summary of Property Data
Figure 2 plots the results of Young’s modulus versus reinforcement content. In addition, the Hashin-Shtrikman bounds’ are provided. These models, which account for stress contributions, provide a better prediction of Young’s modulus for the case of a binary composite than the more straight forward Voigt and Ress rules of mixtures. Specifically, the calculations were conducted using Equation 1 (below), which was provided by Zhang et a1.8. Zhang’s work studied the bounds for the specific case of a binary particle-reinforced metal matrix composite. Handbook data5’ for Young’s moduli of the A1203 reinforcement (380 GPa) and AI-8Mg matrix alloy (68 GPa) were utilized in the calculations. All of the data fall well within the predictive bounds, following other studies in the literature’, lo. This result suggests good load transfer (Le., good bonding) between the A1 and A1203 phases. Equation 1 can be written as:
where E,, Emand E, are the calculated, matrix and particle Young’s moduli, respectively; and V, and V, are the matrix and particle volume fractions, respectively. The upper and lower bounds are obtained by reversing the m and p subscripts in Equation 1.
108
.
Advances in Ceramic Armor V
Effects of Particle Size, Particle Loading and Thermal Processing Conditions
Volume Percent A120,
Figure 2: Young's modulus of A1203/A1 composites as a function of A1203 content with calculated Hashin-Shtrikman bounds plotted.
Over the small range of reinforcement contents studied in the present work, little effect of reinforcement content on strength and toughness is observed. In other studies where larger ranges of reinforcement content were studiedg, a trend of increasing strength with increasing reinforcement content has been shown. A very strong relationship between reinforcement size and strength is shown in the data. Figure 3 provides flexural strength as a function of reinforcement size. The tensile strength data show a similar trend. In all cases, properties are improved as reinforcement size is decreased. For instance, the flexural strength with a 12 pm A1203 particle reinforcement is 535 MPa, as compared to only 189 MPa with a 154 pm reinforcement size. Strength is expected to increase with decreasing Fain size assuming the Orowan relation where the strength is proportional to the grain diameter (- '2) ". The flexural strength results in Figure 3 show this trend where the data are fit to a power curve. As with strength, a strong effect of particle size on fracture toughness is seen. As shown in Figure 4, the results indicate a strong trend of increasing fracture toughness with decreasing particle size in the Al/A1203 system. These data suggest that the reinforcement particle size is related to the critical flaw in the composites - i.e., as the reinforcement size is decreased, the critical flaw in the composites is decreased, thus leading to increased strength and toughness.
Advances in Ceramic Armor V
.
109
Effects of Particle Size, Particle Loading and Thermal Processing Conditions
Mean Particle Diameter (micron)
Figure 3: Flexural Strength of A1203IA1 Composites as a Function of A1203 Particle Size
Mean Particle Diameter (micron)
Figure 4: Fiacture Toughness as a Function of Reinforcement Size
110
Advances in Ceramic Armor V
Effects of Particle Size, Particle Loading and Thermal Processing Conditions
SUMMARY Particulate-reinforced A120jIAl metal matrix composites with various particulate sizes and loadings were fabricated by a pressureless-liquid-metal-infiltration process. The elastic, flexural, tensile and toughness properties were evaluated, and the results related to reinforcement size and content. From this work the following conclusions were made. 1. By increasing the reinforcement content of the composite, systematic increases in density and Young’s modulus were obtained. The composite system with the highest reinforcement content (46%) had a stiffness of 1.8 times that of the base alloy. The HashinShtrikman bounds provided an excellent prediction of Young’s modulus. No effect of particle size on Young’s modulus was observed 2. Flexural and tensile strength were strongly affected by reinforcement size. Flexural strengths over 500 MPa were achieved with a 12 pm A1203 reinforcement, as compared to less than 200 MPa with a 154 pm reinforcement. Little effect of reinforcement content on strength was observed across the range studied. 3. As with strength, fracture toughness increased as particle size decreased. In this material system, there is a clear relationship between the reinforcement particle size and the critical flaw size. REFERENCES 1 . D.L. McDANELS, “Analysis of Stress-Strain, Fracture, and Ductility Behavior of Aluminum Matrix Composites Containing Discontinuous Silicon Carbide Reinforcement,” Metall. Trans. A16 1105-15 (1985). 2. J.W. McCOY, C. JONES and F.E. WARNER, “Preparation and Properties of Cast CeramiciAluminum Composites,” SAMPE Q,19(2) 37-50 (1988). 3. M.K. AGHAJANIAN, J.T. BURKE, D.R.WHITE and A S . NAGELBERG, “A New Infiltration Process for the Fabrication of Metal-Matrix Composites,” SAMPE Q,20(4) 43-47 (1989). 4. M.K. AGHAJANIAN, M.A. ROCAZELLA, J.T. BURKE, and S.D. KECK, The Fabrication of Metal Matrix Composites by a Pressureless Infiltration Technique,” J. Mater. Sci. 26 447-54 (1991). 5 . Metals Handbook: Desk Edition (ASM International, Metals Park, OH, 1985). 6. Engineered. Materials Handbook, Vol. 4, Ceramics and Glasses (ASM International, Metals Park, OH, 1991). 7. Z. HASHIN, S. SHTRIKMAN, “A Variational Approach to the Theory of the Elastic Behavior of Multiphase Materials,” J. Mech. Phys. Solids, 11 127-140 (1963). 8. L. ZHANG, X.H. QU, X.B. HE, B.H. DUAN, S.B. REN and M.L. QIN, “ThermoPhysical and Mechanical Properties of High Volume Fraction SiCp/Cu Composites Prepared by Pressureless Infiltration,” Mat. Sci. & Eng. A489 285-93 (2008). 9. M. K. AGHAJANIAN, R. A. LANGENSIEPEN, M. A. ROCAZELLA, J. T. LEIGHTON, C. A. ANDERSON, “The Effect of Particulate Loading on the Mechanical Behaviour of A1203/A1 Metal-Matrix Composites,” J. Mat. Sci.,28 6683-6690 (1993). 10. P.K. ROHATGI, S. RAMAN, B.S. MAJUMDAR and A. BANERJEE, “Ultrasonic Techniques in Evaluation of Metal Matrix Particulate Composites,” Mater. Sci. Eng. A123 89-96 (1990). 11. W. D. KINGERY, H.K. BROWN, W.R UHLMANN, Introduction to Ceramics Second Edition (Wiley-Interscience, 1960).
Advances in Ceramic Armor V
.
111
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
PRESSURELESS SINTERING OF B4C-SiC COMPOSITES FOR ARMOR APPLICATIONS Rosa Maria da Rocha, Francisco C. L. de Melo CTA-IAE Comando-Geral de Tecnologia Aeroespacial, Instituto de Aeronautica e Espaeo Slo Jose dos Campos, Sb Paulo, Brazil ABSTRACT The development of lightweight and inexpensive ceramic armor is under ongoing consideration by both ceramic armor manufacturers and armor users. Boron carbide (B4C) exhibits attractive properties such as low density, high hardness and high wear resistance, which makes it very useful as lightweight armor. However, poor sinterability as well as relatively low strength and fracture toughness place restrictions on its wide application. One suitable material candidate to improve B4C mechanical properties is silicon carbide (Sic). B4C-Sic ceramic composites are very promising armor materials because B4C and S i c are intrinsically very hard. In this work a pressureless sintering study of B4C-SiC ceramics was made. B4C-Sic mixtures were prepared with S i c concentration from 10 to 50 wt%. Without the external applied pressure during sintering it was necessary to add sintering aids. Two groups of additives, namely A1203-Y203and AlN-Y20, were investigated. Samples were densified by pressureless sintering at 2000 "Ci30 min in an argon atmosphere. B4C-SiC composites were analyzed by XRD and SEM. Bulk density, total linear shrinkage and total weight loss were also measured. Density higher than 93.0 YOof the theoretical value was determined and microhardness of 30.3 GPa was achieved for composite with 10 wt% of SIC sintered with AIN-Y203 additive. INTRODUCTION Boron Carbide, also known as black diamond, is the third hardest material at ambient conditions (next to diamond and cubic boron nitride, c-BN) and is the hardest material at temperature above 1100 "C. This property combined with a low density (2.52 g . ~ m -and ~ ) high Young's modulus (457 MPa), makes it attractive for application such as lightweight armor and wear-resistant parts I. However, despite the technological interest in B4C based ceramics, the sintering of pure and dense B4C components has proved quite difficult, maid due to the strong covalent bonding, which is responsible for the intrinsically low diffision mobility Suitable sintering aids have therefore been sought and various elements and compounds were proposed to aid the consolidation of B4C 3-5. Single-phase B4C materials prepared by pressureless sintering as well as by hot pressing have poor cracking resistance. That is because boron carbide has strictly directional interatomic bonds, high deformation resistance, and low atomic mobility of its atoms right up to temperatures near the melting point. Addition of viscous, brittle, or fiber form components (secondary phase) to the matrix tends to increase the strength and crack resistance of the composite. One suitable material candidate to improve B4C mechanical properties is silicon carbide 6-9. S i c as well as B4C has been recognized as an important structural ceramic material because of its unique combination of properties. S i c ceramics have been used in industry because of their favorable properties such as high elastic modulus and hardness, good thermal and chemical stability, high thermal conductivity, and relatively low thermal expansion coefficient lo. S i c as well as B4C is a highly covalent bonded material which makes densification very difficult. Presently, the production of high density sintered S i c products cannot be achieved without using sintering additives. Pressureless sintering of p-Sic with B and C, probably in the solid state, was first announced by Prochazka in 1975 'I. Since then, many developments and
?
113
Pressureless Sintering of B4C-Sic Composites for Armor Applications
advances in understanding pressureless sintering of S i c have been made. Other additive systems based on the B-C system have been investigated: B-C-A1 system being the most prominent I2,l3. Liquid phase sintering with oxide additives, like A1203 and Y203, has also been extensively studied, mainly addressed to lowering S i c sintering temperature I4,l5. The objective of this study was to investigate ceramic composites based on B4C and S i c obtained by pressureless sintering. The effects of S i c concentration and the sintering additive compositions on the composite densification were investigated. Three different concentrations of S i c were used: 10, 30 and 50 wt% of S i c and two compositions of sintering additive, AIN:Y203 and A1203:Y203. EXPERIMENTAL PROCEDURE Commercially available high-purity fine powders of B4C (H.C.Starck-Germany), P-SiC (BF-12 H.C.Starck-Germany), AIN (fine grade-H.C.Starck-Germany), Y2O3 (fine grade- H.C.Starck-Germany) and A1203 (A-16 SG- Alcoa) were used to prepare the mixtures. The powder mixtures were prepared by planetary milling in isopropanol using a polyethylene container. Three different composites were prepared with S i c concentration of 10, 30 and 50 wt%, while maintaining a constant amount of 10 vol% of each additive system. Two different additive systems, namely A1203-Y203 and AIN-Y203 were investigated. The molar ratio of AIN to Y203 and A1203 to Y203 was 3:2 and 5:3, respectively. For comparison purpose, samples without additive were also prepared. After drying and sieving, the powders were uniaxially pressed in disc shape samples of 14 mm diameter and 5 mm thickness and followed by a cold isostatic pressing at 300 MPa. A graphite resistance furnace was used for pressureless sintering which was comprised of graphite heating elements and fibrous insulation. Samples were placed inside the furnace without bed powder protection. The temperature was monitored using an optical pyrometer sighted on the graphite casing near the samples. Thermal cycle was characterized by heating and cooling rate of 20"-30" Cimin and dwell time of 0.5 h at sintering temperature of 2000 "C. The final densities of the sintered samples were determined by liquid (distilled water) immersion method (Archimedes' Principle). The theoretical densities were calculated from the densities of the starting phases by rule of mixtures. Total weight loss and total linear shrinkage were also determined. Microhardness of the samples with highest densities values was assessed by Vickers microindentation using load of 9.8 N. X-ray diffractometry (XRD- Model Phillips PW 18/30) using monochromated CuKa-radiation was used to identify the as-sintered phase composition of both the S i c polytypes and the crystalline grain-boundary phases. Diffraction analysis was performed directly from a machined surface of the sintered samples and qualitative analysis was done by comparison with JCPDS standards. The microstructure of the sintered samples was examined by scanning electron microscopy (SEM - Model LEO 435i-Zeiss) on the fracture surface. RESULTS AND DISCUSSION The results of relative density of the sintered composites as a function of S i c concentration are presented in Fig. 1. As it can be seen, the relative density decreased with increasing in S i c concentration. This effect was observed in composites without additive and those with additives in both systems. Very little densification was reached for B4C-SiC samples without sintering additives. Composites sintered with AIN-Y203 addition showed the highest densities for the three S i c concentrations. The use of AIN-Y203 as an additive system has been demonstrated to be effective in the pressureless LPS sintering of S i c without the use of a powder bed 16. Because of the high level of porosity in B4C-30SiC and B4C-5OSiC samples, microhardness was measured only in the samples with higher densification, B4C-1OSiC. Samples with A1203-Y203 and AlN-Y203 additions showed microhardness of 29.5 1.2 and 30.3 f 1.2 GPa, respectively.
*
114
Advances in Ceramic Armor V
Pressureless Sintering of B,C-SIC Composites for Armor Applications
92-
P
-m-0-
-A-
88:
v,iW additive AI2O;Y2O3 NN:Y203
&):
s ": 76-
72684
,
10
.
, 20
,
,
, 30 SIC (Mh)
,
40
.
,
50
Figure 1. Relative density as a function of S i c concentration for samples sintered at 2000 'Ci30min. Composites prepared without additive, with 10 vol% of AIN-Y203 and A1203-Yz03 additives. Figure 2 shows the total weight loss of the sintered composites as a h c t i o n of S i c amount. It can be observed that the weight loss is almost constant regardless of the S i c concentration. However the additive presence has significantly increased the composites weight loss. Among the most evident result it should be noted that A1203-Y203 containing compositions exhibit the highest weight loss, of about l6%, while those with AlN-Yz03 have a weight loss of about 12%. This result is related to the interaction between oxides and S i c with massive gaseous products formation leading to high weight loss. These reactions occur more actively with the increase of alumina content and the temperature of sintering lo.
10
20
30
sic (M%)
40
50
Figure 2. Total weight loss versus SIC concentration for samples sintered at 2000 "Ci30min. Composites prepared without additive, with 10 vol% of AlN-Y203 and A1203-Y203additives.
Advances in Ceramic Armor V
.
1 15
Pressureless Sintering of B,C-Sic Composites for Armor Applications
16
,
8 C (Wh)
Figure 3. Total linear shrinkage as a function of S i c concentration. Composites prepared without additive, with 10 vol% of AlN-Y203 and A1203-Y203 additives. Figure 3 shows the total linear shrinkage of the samples as a function of S i c content. Sintered samples with additives showed a significant decrease in shrinkage with the increase in S i c content. Moreover, composites with A1203-Yz03 present a lower shrinkage than samples with AIN-YzO,. These results suggest that sintering additives enhance densification for low S i c concentration (B&- IOSiC), but as S i c amount increases the densification is hindered. In the case of A1203-Y203 containing composites the densification reduction was rather more pronounced. SEM micrographs of the composites sintered without additive are shown in Fig. 4. Microstructure is fine with a homogeneous distribution of S i c grains (white phase) in the B4C matrix for all S i c concentrations. XRD analysis of these samples (Fig. 5 ) show besides B4C peaks, the corresponding peaks of the 6H a - S i c polytype. Overlapping of the Bragg reflections from the different S i c polytypes makes a qualitative X-ray diffraction phase analysis extremely difficult. Because of that it is not possible to verify if all 3C original p-Sic was transformed to a-Sic.
(a)
(b)
(4
Figure 4.SEM micrographs of the fracture surface of the samples sintered without additive: (a) 10 wt% Sic; (b) 30 wt% Sic; (c) 50 wt% Sic.
116
Advances in Ceramic Armor V
Pressureless Sintering of B4C-Sic Composites for Armor Applications
i r
c 10 sic
30
35
4 0 4 5
50
55
€
0
6
5
2@(") Figure 5. X-ray diffraction patterns of the sintered composites B4C-SiC without additive.
Figure 6 presents the microstructure of the sample B4C-lOSiC, B4C-30SiC and BdC-SOSiC sintered with AIN-Y203 additives. It can be noted from Fig. 6c the high amount of platelet shape S i c grains (grey phase). The difference in the composite microstructures is directly related to S i c phase transformation and its concentration in the composite. It is well k n o w that at higher temperatures cubic p-Sic transforms to hexagonal a-Sic, especially 4H polytqpe, stimulating grain growth and a change in grain shape from equiaxed to high aspect ratio plate-like grain 15,'7. S i c phase transformation with the associated platelet like grains morphology can explain the lower density results of composites with higher S i c concentrations. The formation of such morphology could result in early impingement of large grains, forming a skeleton that effectively arrested densification in pressureless sintering of the B4C-Sic composite.
(a)
(b)
(c)
Figure 6. SEM back scattered images of the composites sintered with AIN-Y203 : (a) B4C-1OSiC; (b) B4C-30SiC; (c) B4C-SOSiC.
Advances in Ceramic Armor V
.
1 17
Pressureless Sintering of B,C-SIC Composites for Armor Applications
The results of XRD analysis of the composite B4C-SOSiC sintered without additive and with AlN:Y203 and Ai203:Y203 are shown in Fig. 7. It was identified the polytype 4H instead of the 6H in both systems with additive. The polytype transformation is very sensitive to composition and it has been reported that SIC materials sintered with A1 favors the formation of 4H polytype with more elongated and interlocking structure '3218.
2@(") Figure 7. X-ray diffraction patterns of the composite B4C-SOSiC sintered without additive and with AlN:Y203 and A1203:Y203. Considering the results of relative density and linear shrinkage of the two additive systems (Fig. 1 and Fig. 3), it can be suggested that the amount of aluminum in the composite is directly related to extent of 4H transformation and its morphology. Since phase transformation plays an important role for the microstructural development and the resulting properties, detailed investigations of the transformation kinetics must be conducted further. For composites B4C-SiC an accurate investigation by means of manipulating /3 to a transformation, including starting powder compositions, sintering aids, and sintering temperature must be accomplished to obtain dense and self-reinforced composites. In this study the optimal S i c concentration of 10 wt% and the addition of AlN-Y203 as sintering aids present the best densification results. This composite shows the highest density (93.4 %TD) and Vickers hardness of 30.3 f 1.2 GPa. This is a very promising result, once high hardness is one of the most important requisite for a ballistic material application. Moreover, the SIC platelet grains can act as a factor to increase composite fracture toughness, which play an equally important role in obtaining light weight ballistic material. CONCLUSION B4C-SiC composites were fabricated by pressureless sintering of B4C and SIC powder mixtures. Three different concentrations of SIC, 10, 30 and 50 wt% and two different systems of sintering aid were analyzed in samples sintered at 2000 "C/30 min. The relative density of the composites decreased with the S i c concentration. B4C-SOSiC composites showed the lowest densities, even for sample with
118
Advances in Ceramic Armor V
Pressureless Sintering of B4C-SiCComposites for Armor Applications
AIN:Y203 addition. The low densification is probable related to the 3C to 4H S i c polytype transformation with abnormal grain growth and the a-Sic grains platelet morphology. B4C-1OSiC composite sintered by LPS using AIN:Y203 showed the highest density (93.4 %TD) and high microhardness (30.3 1.2 GPa). Further research at the optimization of both composition and processing of the composite will be necessary for applying this material as ballistic armor.
*
REFERENCES IF. Thevenot, Boron Carbide-A Comprehensive Review, J. Eur. Cerum. Soc., 6,205-25 (1990). 'G. L. Kalandadze, and S. 0. Shalamberidze, Sintering of Boron and Boron Carbide, J. Solid. Stute Cbem.,
154, 194-98 (2000). 'J. E. Zorzi, C. A. Perottoni, and J. A. H. da Jornada, Hardness and Wear Resistance of B4C Ceramics Prepared with Several Additives, Muter. Lett., 59,2932-35 (2005). 4S. Yamada, K. Hirao, and Y. Yamauchi, Sintering Behavior of BdC-CrBz Ceramics, J. Muter. Sci. Lett.,21, 1445-47 (2002). 5L.S. Sigl, Processing and Mechanical Properties of Boron Carbide Sintered with Tic, J. Eur. Cerum. SOC.,18, 1521-29 (1998). 6Y. G. Tkachenko, V. F. Britum, E. V. Prilutskii, D. Z. Yurchenko, and G. A. Bovkun, Structure and Properties of B4C-Sic Composites, Powder Mefull. Met. Cerum., 44, 196-201 (2005). 'M. Uehara, R. Shiraishi, A. Nogami, N. Enomoto, and J. Hojo, SiC-B4C Composites for Synergistic Enhancement of Thermoelectric Property, J. Eur. Cerum. Soc., 24,409- (2004). * S. Tariolle, C. Reynaud, F. Thevenot, T. Chartier, and J. L. Besson, Preparation, Microstructure and Mechanical Properties of SIC-SIC and B4C-B4CLaminates, J. Sol. Star. Chem., 177,487-92, (2004). G. Mapani, G. Beltrani, G. L. Minoccari, and L. Pilotti, Pressureless Sintering and Properties of aSiCB4C Composite, J. Eur. Cerum. Soc., 21,633-38 (2001). lo V. A. Izhevskyi, L. A. Genova, J. C. Bressiani, and A. H. A. Bressiani, Liquid Phase Sintered Sic: Processing and Transformation Controlled Microstructure Tailoring, Muter. Res., 3,13 1-38 (2000). I ' S. Prochazka, The Role of Boron and Carbon in the Sintering of Silicon Carbide, Special Ceramics, 6, 171-8, edited by P. Popper, British Ceram. Res. Assoc., Stoke-on-Trent, 1975). "H.N. Yoshimura, A. C. da Cruz, Y. Zhou, and H. Tanaka, Sintering of 6h(u)-Sic and 3C(p)-SiC Powders with B4C and C Additives, J. Muter. Sci., 37, 1541-46 (2002). l 3 X.F. Zhang, Q. Yang, and L. C. De Longhe, Microstructure development in hot-pressed silicon carbide: effects of aluminum, boron and carbon additives, Acfu. Mut., 51, 3849-60 (2003). I4M. Omori, and H. Takei, Preparation of Pressureless-Sintered SiC-Yz03-Al209. J. Muter. Sci., 23, 374449 (1988). Is V. A. Izhevskyi, J. C. Bressiani, and A. H. A. Bressiani, Effect of Liquid Phase Sintering on Microstructure and Mechanical Properties of Yb203-A1NContaining Sic-Based Ceramics, J. Am. Cerum. SOC.,88, 1115-21 (2004). I 6 J. Schneider, K. Biswas, G. Rixecker, and F. Aldinger, Microstructural Changes in Liquid-Phase-Sintered Silicon Carbide during Creep in an Oxidizing Enviroment, J. Am. Cerum. Soc., 86, 501-7 (2003). I'M, Nader, F. Aldinger, and M. J. Hoffmann, Influence of the a-p Sic Phase Transformation on Microstructural Development and Mechanical Properties of Liquid Phase Sintered Silicon Carbide, J. Muter. Sci.,34, 1197-1204 (1999). Y . Zhou, H. Tanaka, S. Otani, and Y. Bando, Low-Temperature Pressureless Sintering of alpha-Sic with AI4C3-B4C-CAdditions, J. Am. Cerum. Soc., 82, 1959-64 (1999).
'*
Advances in Ceramic Armor V
.
1 19
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
Applications of NDE
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
A PORTABLE MICROWAVE INTERFERENCE SCANNING SYSTEM FOR NONDESTRUCTIVE TESTING OF MULTI-LAYERED DIELECTRIC MATERIALS
K. F. Schmidt, Jr., J. R. Little, Jr. Evisive, Inc. Baton Rouge, Louisiana USA W. A. Ellingson Argonne National Laboratory Argonne, Illinois USA W. Green US Army Research Laboratory Aberdeen Proving Ground, Maryland, USA ABSTRACT
A portable nondestructive technique that can be used for in-situ monitoring of the condition of multilayer composite material ceramic armor has been developed and demonstrated using the Evisive Scan microwave interference scanning method. The portable microwave interference scanning system has been demonstrated to detect damage on rest specimens and engineered features in surrogates. Work has been conducted using specially fabricated surrogates and as-produced and non-ballistic impact damaged specimens. The microwave interference scanning technique detects and images cracks, laminar features and variations in material properties such as density. It requires access to only one surface, and no coupling medium. Other methods, including through-transmission x-ray, and destructive examination are used to establish quantitative performance. It appears that the method is suitable for in-theater health monitoring of composite ceramic armor and that damage level data are not affected by separation of outer lamination layers. Test panels used in this work were provided by the US Army Research Laboratory, Ballistics Testing Station through Argonne National Laboratory. This paper will describe the system and present current results. INTRODUCTION High-performance technical ceramics are widely used in body armor, aircraft armor and ground vehicle armor. Ceramic armor is employed in the form of plate inserts in garments and seats; in panels in vehicles, aircraft and vessels; and as an applique in armored vehicles. Ceramic armor provides effective and efficient erosion of and defeat of ballistic threats. Effectiveness of ceramic armor can be degraded by defects present from production and by operational damage resulting from handling or impact with objects in the environment, other than projectiles. In normal use, ceramic armor is routinely exposed to the possibility of such damage"'. A means to detect damage from ballistic projectiles, non-projectile impacts, handling and manufacturing defects is needed to determine the integrity of the ceramic armor so that appropriate replacement can be made. Recently, (reported here in 2008'2', US and international patents issued earlier'3') a microwave-based method has been developed and demonstrated that is applicable to ceramic armor systems. The process will be referred to here as Evisive Scan TM. It has been configured and tested as a field useable system. The system has been demonstrated on several armor panels provided
123
Portable Microwave Interference Scanning System for Nondestructive Testing
by a variety of independent sources. The Evisive Scan method permits real time evaluation by inspection from one surface only, through non-contacting encapsulation, with panels hung in place.
DESCRIPTION OF THE METHOD The Evisive microwave interference scanning technique requires access to only one side of a part. The microwave interference pattern is created by bathing the part in microwave energy as illustrated in Figure 1. The probe (transmitter and receiver antenna) is moved over the part and the signal at the receivers is sampled. The voltage values for both receivers are saved with the associated X-Y position on the object. The saved voltages are displayed as a function of X-Y position on a computer monitor. The displayed “image” is presented as a surface; with each X-Y position being shown as a gray scale, a false-color or as a 3D Z value.
The displayed image is surface created from a point cloud of measured voltages at (x,y) positions. The voltage at each point is an integrated signal from the entire thickness. Some energy is reflected and transmitted at every interface of changing dielectric constant in the field of the transmitter. This includes the front and back surfaces of the part, and every “feature” in the part that has a discontinuity in dielectric properties. A microwave interference pattern is created when the reflected energy is combined at each of the detectors with the transmitted signal to create the measured detector voltage. The dielectric constants of the engineered ceramic materials in this report are about 4 and the microwave frequency used in the tests is about 24 GHz. Thus the wavelength in the material is about 8 mm (0.33 inches). The magnitude of the phase difference between the emitted signal and reflected signal determines the voltage of the signal. The distance or depth to a reflection defmes the phase relationship of the reflected signal to the emitted signal and the amplitude of the reflected signal, as illustrated in Figure 2. The phase to depth relationship is repeated at a cycle of one wave length. The relative magnitudes of dielectric constants at a detected discontinuity also defines the magnitude of the reflected energy. The combined emitted signal and the reflected energy are measured at the Channel A and Channel B receivers. This variable DC voltage is sampled at points over the part surface. Hardware Channels A and B are separated by a quarter wavelength (hi4) in the wave propagation dimension, Z. Channel C data is the value of the voltage of Channel A minus the voltage of Channel B for each (x,y) location. In any “image” data, the rate of change of the detected signal value impacts the “clarity” of that image. This is true for detected Z axis features as well. Thus the “image” data of a feature is optimized visually at a Z dimension associated with maximum rate of change of the signal in the Z dimension. This is achieved for each channel by moving the
124
.
Advances in Ceramic Armor V
Portable Microwave Interference Scanning System for Nondestructive Testing
emitter (and receiver) within a quarter wave length in the Z direction. This position is referred to as the “Stand-Off’ distance.
Every change in dielectric constant in the material returns a reflection, which, with the emitted signal creates an interference pattern. It is the sum at the measurement plane, of these patterns from all changes in dielectric constant in the inspection volume that creates the detector voltage. This voltage is sampled and saved as the Evisive Scan data file. The interference pattern from a point reflector is illustrated theoretically in Figure 3. Depth information may be inferred from the number and frequency of interference “ripples”.
The Z position of a detected feature may be inferred by observing the difference in the indications of the features in Channels A, B or C, the optimized Stand-off, and the number and frequency of
Advances in Ceramic Armor V
.
125
Portable Microwave Interference Scanning System for Nondestructive Testing
interference rings. X and Y positions of features are estimated from rate of change of interference signal in the measurement plane. PORTABLE EQUIPMENT CONFIGURATION The portable version of the Evisive Scan equipment is shown in Figure 4a. This shows the laptop-computer with the driving electronics and the scan head. The same computer and electronics can be coupled in the laboratory to an XY Positioning Table as shown in Figure 4b. Operating, interface and display software resides on the interface and display computer.
(3)
(b)
Figure 4 Photographs of portable scanning microwave system a-computer, electronics and bead b-X-Y positioning table with head attached
Data are collected via an X-Y raster scan over the surface of the part. The data rate is sufficiently high that mechanical positioning or position feedback for manual positioning is the only limitation in scan speed. The scan data are available in near real time. This scanning technology has been applied in the laboratory, with X-Y planar, X-Y cylindrical and r-8 positioning, and in the field with surface X-Y and multi-degree of fieedom positioning devices. With the exception of the infrared tracking position system, the images presented here were acquired on the X-Y positioning table. Datum spacing in the scan direction was 0.003 inches and raster increment was 0.05 inches unless otherwise stated. Scan speed on the X-Y positioning table reach 3 inches per second and ramp to start and stop. Scan speed with the hand held infrared tracking system varies and may exceed 10 inches per second. PORTABLE FIELD CONFIGURATION: A number of portable configurations have been applied to field use: The instrument and control system has been interfaced to mechanized pipe scanners and with a multi-axis position system for free-form manual positioning, and with a variety of manual position encoders.
126
*
Advances in Ceramic Armor V
Portable Microwave Interference Scanning System for Nondestructive Testing
A portable system has been interfaced to an infrared camera for correlation in other related studies. The equipment is shown in the field in Figure 5 . The probe is manipulated manually, position tracked ad presented in real-time (the position tracking display is shown in Figure 6 ) . Evisive has also successfully demonstrated wireless interface of the probe and control computer, and is currently optimizing that technology.
DETECTION OF CRACKED ARMOR TILE The Evisive Scan system has been demonstrated to detect cracked ceramic armor tile in a typical ceramic armor layered configuration. Figure 7 shows a corelation between a throughtransmission x-ray image, Figure 7a, and the resulting microwave scan Figure 7b, of the same cracked
Advances in Ceramic Armor V
.
127
Portable Microwave Interference Scanning System for Nondestructive Testing
tile. The wide, light gray patterns shown in the microwave scan follow the crack centerlines. The Evisive Scan data has sufficient dynamic range, (greater than 12 bit resoultion), to identify the centerlines and edges of features within the data position precision. The scans presented in Figure 7 have spatial precision of 75 mm (0.003 inches) by 1.25 mm (0.050 inches).
(a) (b) Figure 7 Correlation between through-transmission x-ray image and microwave scan of same 8-inch square cracked layered ceramic armor tile a)-x-ray image b)-microwave image
ESTIMATION OF CHARACTERISTICS OF VARIOUS DETECTED FEATURES The scanning microwave data can be used to estimate characteristics of detected features including depth below a surface, size, and whether or not the feature lies in a certain plane. Data were acquired from the front surface only and in these cases used the XY Positioning Table. Scan speed was about 15 cm (6 inches) per second, with programmed start and stop ramps. Sample spacing on the scan axis was 75 mm (0.003 inches), triggered by optical encoder. Raster spacing was 1.25 mm (0.050 inches). Standoff was 0.0 mm, unless otherwise specified. All data were acquired using 24 GHz. The gray scale assignments in the “image” data are optimized to the voltage range within a region of interest ESTIMATION AND COMPENSATION FOR PLANAR OR NON-PLANAR FEATURES Planar orientation of features can be obtained by observing the voltage differences in any scan. This can be seen in Figure 8 which shows the non-planar detection of the surface of a tile. The tilt plane of the tile, less than 3 degrees, is compensated for in the presentation software. The consistent background voltage across the surface of the tile confirms the tilted planar orientation.
128
Advances in Ceramic Armor V
Portable Microwave interference Scanning System for Nondestructive Testing
ESTIMATING DEPTH OF FEATURES Depth of a feature below the surface can be estimated by determining the he phase differences of the microwave signal at the front and back surface and from features within the part . The phase difference of Channels A and B present this difference for every scan. Changes in the Standoff, which adjusts the phase at the front surface of the material under test, changes the phase relationship at all depths in the material and both Channels. The differential relative voltage changes in the image of specific features at different standoff values indicate differences in the feature attributes at different depths in the material. This process allows segregation of stacked features and determination of extent of features at different depths which appear at a common location. Effect of Stand Off on depth of feature detection is illustrated in Figure 9. In Figure 9, the features with distinct depth dependent characteristics are shown in boxes 0. These features appear to have similar relative values at all three standoffs. A/4 is about 3 mm (0.12 inches) in air. Each 0.025 increment of standoff moves the phase about 3 mm (0.012) inches in the material. The shape difference in the circled feature 0 may be due to the phase translation through the encapsulating fabric. Selection of Stand-off value for scanning or optimization of Channel phase separation can optimize detection and resolution of specific depths of interest.
Figure 9 Detection of features at different depths
Advances in Ceramic Armor V
. 129
Portable Microwave Interference Scanning System for Nondestructive Testing
FEATURE PRESENTATION DIFFERENCE BETWEEN CHANNELS AND STAND-OFF The phase difference between Channel A and Channel B (U4)is evident in Figure 10 and Figure 11 which present Channel A and Channel B images of a tile at Standoff of 0.01875 inches. Indications of features at different depths in the material are noted. The Figures illustrate the difference of the response of features of interest and of the surrounding dielectric material at the Channel A and B phases. Ch A
- Evisive NOT file - 05-28-2008 11:37:21
i-2808llB.S2 top [email protected] SimplifVirmor Sampler Delamtumeur Sample 6.2808H8-12 mp wcpo3s00or876.a~
nand onO.oW76
X Inch
Figure 10 Features in Channel A at Standoff0.01875
The phase difference of ;V4 (about 0.08 inches in the material (based on an estimated dielectric constant of about 4) illustrates a depth-based aspect of the circled
130
Advances in Ceramic Armor V
0 features..
Portable Microwave interference Scanning System for Nondestructive Testing
-
Ch B - Evisive NDT file 05-28-2008 11:37:21
ample 6.2808\1B.S2 top w ~ 3 s o 0 1 8 7 5 . e v d o rSampledArmor Sampler DelamrL4rmour Sample 6.2808HB-SZ Stand off 0.01876
X Inch
Figure 11 Features in Channel B a t Standoff 0.01875
In Figure 12 the separation of the features at (1.5,2) is more clear at Standoff of 0.0375 inches (about 0.01 inches depth in the part). C0N CLUS I 0N S
A portable microwave-based system has been developed and demonstrated on layered ceramic armor to detect cracks and delaminations within ceramic armor systems. Examination requires access from one side only and is effective in applications with metal backing. Coupling of the system to both laboratory X-Y scanning systems as well as multi-axis scanning systems has been demonstrated. The capability of the method allows determination of size, depth and orientation of features within the dielectric solid.
Advances in Ceramic Armor V
. 131
Portable Microwave Interference Scanning System for Nondestructive Testing
Ch A
- Evisive NDT file - 05-28-2008 1 2 4 6 5 3
so0375.evdmor SamplesWrmor Samples DelamsWrmour Sample 5-28-081 Stand off 0.0375
X Inch Figure 12 Features in Channel A at Standoff 0.0375
Further laboratory testing including destructive analysis of samples will establish scan and data interpretation protocols and qualify the technique for field nondestructive testing applications. The equipment needs further optimization and hardening for field use. While wireless communications with the system have been demonstrated, wireless communication needs improving. ACKNOWLEDGEMENT Evisive, Inc. expresses its sincere appreciation to the US Army Small Business Innovative Research Program, and US Army Research Laboratory and US Army Tank Automotive Engineering Research and Development Command.
132
Advances in Ceramic Armor V
Portable Microwave Interference Scanning System for Nondestructive Testing
REFERENCES 1 “ADVANCES IN CERAMIC ARMOR III”, Jonathan Salem, Dongming Zhu, Edited by Lisa Prokurat Franks, Ceramic Engineering and Science Proceedings, Vol. 28, Issue 5,2007 2 “A PORTABLE MICROWAVE SCANNING TECHNIQUE FOR NONDESTRUCTIVE TESTING OF MLnTILAYERED DIELECTRIC MATERIALS”, Karl Schmidt, Jack Little, William Ellingson, , Proceedings of the 32”dInternational Conference &Exposition on Advanced Ceramics and Composites, 2008 3 United States and International Patents United States Patent 6,359,446, “APPARATUS AND METHOD FOR NONDESTRUCTIVE TESTING OF DIELECTRIC MATERIALS”, MAR. 19.2002 United States Patent 6,653,847, “INTERFEROMETRIC LOCALIZATION OF IRREGULARITIES, Nov. 25,2003 International Patent PCTNS2005/026974, “HIGH-RESOLUTION, NONDESTRUCTIVE IMAGING OF DIELECTRIC MATERIALS”, International Filing Date 1 August, 2005 Canadian Patent 2,304,782, “NONDESTRUCTIVE TESTING OF DIELECTRIC MATERIALS”, Mar. 27,2007 New Zealand Patent 503733, “NONDESTRUCTIVE TESTING OF DIELECTRIC MATERIALS”, PCTNS20051026974, International Filing Date 1 August, 2005 Australian Patent 746997, “NONDESTRUCTIVE TESTING OF DIELECTRIC MATERIALS”, PCTNS20051026974, International Filing Date 1 August, 2005
Advances in Ceramic Armor V
133
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
DESTRUCTIVE TESTING AND NONDESTRUCTIVE EVALUATION OF ALUMINA STRUCTURAL CERAMICS Raymond E. Brennan, William H. Green, James M. Sands US Army Research Laboratory Aberdeen Proving Ground, MD, 21005-5066 ABSTRACT A combination of destructive and nondestructive testing methods was utilized to evaluate the impact velocity and energy conditions that caused fracture in alumina structural ceramics. Incremental damage was produced in aluminum-backed A1203 tiles using drop tower testing for low velocity impact with a high mass indenter. Under these test conditions, a damage threshold range was determined below which there was no impact damage and above which fracture occurred. The majority of damage in the fractured samples consisted of radial and cone cracking. The samples were nondestructively evaluated using digital radiography, x-ray computed tomography, and ultrasound Cscan imaging. Bulk damage detected by these techniques was compared to surface damage observed by visual inspection. Qualitative evaluation of surface and bulk cracks and quantitative percent damage assessment were used to compare the integrity of the nondestructive techniques. While all three methods were able to detect surface cracks, C-scan imaging was more effective at detecting internal damage in the alumina samples while x-ray computed tomography was more effective at producing three-dimensional images of the crack patterns. INTRODUCTION It is a common practice in ballistic testing of armor materials to overmatch the test samples in order to determine the limits of their performance’-I2. The main goals of this work were to produce incremental levels of damage in armor ceramic materials and to characterize and compare the resulting damage using nondestructive means. Fracture of alumina structural armor ceramics was studied to determine the velocity and energy conditions under which different levels of damage were produced. The purpose of determining damage data, both qualitative and quantitative, was to gain a better understanding of the impact behavior of structural armor ceramics, which could also be used to provide input to impact damage models. Impact damage was created using drop tower testing on alumina test specimens. In this technique, a high mass indenter was used for low impact velocity testingI3. By altering the drop height and, therefore, the velocity and energy conditions, parameters that caused fracture were determined13. The presence of damage was indicated by visual inspection of surface damage as well as nondestructive evaluation methods including ultrasound testing (UT), digital radiography (DR), and xray computed tomography (XCT). Quantitative damage assessment comparing visual and nondestructive inspection was used to distinguish surface and bulk damage. The NDE methods were also compared to determine effectiveness of bulk damage detection. X-ray digital radiography is a 2-D rojection radiology technique in the same category as film radiography and real-time radiographyleP6. Compared to other radiology techniques, DR has a significantly higher dynamic range and is able to produce digital images of the test specimen^'^-'^. While a digital radiograph projects the entire volume onto a single plane, x-ray computed tomography produces a series of individual densitometric slices that show the true dimensions and spatial relationship of features withim the sample”. These high resolution XCT slices can be viewed individually or in sequence at any desired orientation through computer volumetric multi-planar reconstruction techniques. Using commercial image processing software, it is also possible to reconstruct 3-D digital solid images that can reveal virtual sections on arbitrary planes. Selected features within single slices or entire 3-D solid volumes can be isolated from the surrounding bulk
135
Destructive Testing and Nondestructive Evaluation of Alumina Structural Ceramics
material and visualized as 3-D point cloud images, where the points in space define the feature boundaries. Ultrasound NDE uses the transmission of acoustic waves to nondestructively characterize a test specimen”-19. Reflection of these waves is caused by acoustic impedance mismatch, which occurs at material boundaries”-’9. This can aid in the detection of defects such as pores and inclusions. For Cscan imaging, an ultrasonic transducer is rastered over the desired sample area, and the collected signals from the A-scans, or amplitude scans, are assigned to x and y coordinates. The changes in gated, or selected, reflected amplitude signals are evaluated. In this study, the bottom surface reflected signal was gated to study variations and look for distributed effects of inhomogeneities in the bulk of each sample. The reflected signal amplitude data, given in mV, represented signal gain while the attenuation, given in decibels (dB), represented signal loss. These values were assigned to a cblor scale or gray scale which represented amplitude changes over the gated regions. The data were mapped according to the assigned scales and the x and y coordinates to produce bulk image maps of the samples. SAMPLE PREPARATION AND EXPERIMENTAL SETUP T h e samples used in this study were (12.7 mm thick, 101.6 mm by 101.6 mm) CoorsTek AD995 A1203 tiles adhesively bonded to (3.17 mm thick, 101.6 mm by 101.6 mm) aluminum backing plates, A1 5083, which were used for structural support. They were made using a manual process in which a 2.5 mm thick layer of GE Silicone SCSlOOO Sealant was applied to the backing layer and bonded to the ceramic tile. The bonding procedure utilized consistent application of the adhesive at a constant pressure (weight) to bond the backing plate to the ceramic during the 24-hour cure time. Drop tower testing was conducted using an Instron Dynatup Model 8200 Impact Test Instrument, as shown in Figure 1. The instrument was utilized for high mass drop weight impact testing with a 12.7 mm diameter high strength steel hemispherical indenter13. A 1077.0 g lead weight was added to the drop weight assembly for a total drop weight of 3735.0 g. A total of 30 samples were fabricated and tested. The initial drop heights covered a wide range, from 127.0 mm up to 1016.0 mm, where 1016.0 mm was the maximum height of the test instrument. The impact energies ranged from 4.57 J at a drop height of 127.0 mm to 37.19 J at a drop height of 1016.0 mm. Subsequent drop tests focused on a narrow transition height range between 419.1 to 469.9 mm, and a broader range between 508.0 and 932.2 mm that was used to test repeatability. Five samples were selected for fiuther nondestructive evaluation.
Figure 1. Drop tower testing setup with A1203 sample.
136
Advances in Ceramic Armor V
Destructive Testing and Nondestructive Evaluation of Alumina Structural Ceramics
Table I. Drop tower testing conditions and percent damage calculations.
The drop height, impact velocity, impact energy, and fracture results for these five samples are listed in Table I. After completing the drop tower testing, visual inspection was conducted by acquiring digital images of the sample surfaces. Projection digital radiography and XCT were performed through the thickness of selected samples using a computed tomography system with a 420 kW x-ray source and a 512 element linear detector array. The tube energy and current used were 400 keV and 2.0 mA, respectively, and the focal spot was 0.80-mm. The source-to-image-distance and source-to-objectdistance were 940.00-mm and 750.00-mm, respectively. Ultrasound testing was conducted using a 15 MHz longitudinal transducer with a water immersion testing setup. The reflected signal amplitude values from the bottom surfaces of each A1203 ceramic were measured to produce C-scan image maps through the bulk of selected samples. DROP TOWER TESTING RESULTS After completing drop tower testing, the data indicated that a transition from non-fractured to fractured samples occurred at an approximate range of 15.24 to 16.19 J, which corresponded to drop heights between 431.8 and 450.8 mm. This is graphically depicted in Figure 2, where the system energies were plotted against the impact velocities.
Impact Velocity (mls) Figure 2. Drop tower plot of energy vs. velocity indicating sample fracture results and conditions.
Advances in Ceramic Armor V
.
137
Destructive Testing and Nondestructive Evaluation of Alumina Structural Ceramics
In this graph, the squares represented non-fractured samples and the circles represent fractured samples. The circled threshold region in the figure, containing both fractured and non-fractured samples, indicated that no samples tested at impact velocities below 2.86 d s e c were fractured and all samples tested at impact velocities above 2.94 d s e c were fractured. In the fractured samples, the majority of surface cracks produced by indenter impact were visible. The radial cracks typically extended out to the edges. The indenter also produced a cratered impact region in the fractured samples. With increasing impact energy, the number of surface cracks and the size of the impact craters appeared to increase. A more detailed description of drop tower testing for samples A, B, C, D, and E (Table I), which were selected for further nondestructive evaluation, is provided. While there was a small mark on sample A, tested at 355.6 mm, where the high strength steel indenter impacted the sample surface, there were no signs of cracking. As in the general sample trends, drop tower testing at heights above 431.8 mm caused fracture for samples D and E, which were tested at 591.8 and 762.0 mm, respectively. However, the test conditions for samples B and C, which were both tested at a drop height of 431.8 mm, fell within the threshold region as shown in Figure 2. Within this threshold region, approximately half of the samples fractured while the other half did not. This trend was consistent for samples B and C, as sample C fractured while sample B did not. The impact velocities and impact energies varied slightly for sample B at 2.88 d s e c and 15.50 J when compared to sample C at 2.87 d s e c and 15.40 J. Digital surface images of samples A through E were acquired for visual inspection, as shorn in Figures 3-7. Samples A and B showed small marks where impact occurred but did not cause fracture while samples C, D, and E showed fracture origins and surface crack patterns in the impact regions.
-
10mm
Figure 4. Comparison of visual inspection (left), DR (center), and UT (right) for sample B.
138
.
Advances in Ceramic Armor V
Destructive Testing and Nondestructive Evaluation of Alumina Structural Ceramics
Figure 7. Comparison of visual inspection (left), DR (center), and UT (right) for sample E. DIGITAL RADIOGRAPHY IMAGING X-ray digital radiography was conducted on samples A through E, as shown in Figures 3-7, for damage assessment and comparison to visual inspection. The 2-D bulk imaging data collected through the thickness of each sample showed crack patterns that were consistent with visual inspection. Several additional features were also detected that were not picked up by visual inspection of the sample surfaces. Darker features in the DR images corresponded to a lower degree of x-ray attenuation and more pronounced damage. The largest cracks showed up as the darkest features in contrast to lighter undamaged regions.
Advances in Ceramic Armor V
.
139
Destructive Testing and Nondestructive Evaluation of Alumina Structural Ceramics
The DR images for samples A and B (Figures 3-4), which were tested at the lowest heights of 355.6 and 431.8 mm, respectively, and did not fracture, appeared to be homogeneous, with no distinct features that could be detected. The point impact regions in which the indenter contacted the alumina samples did not have enough additional x-ray transmission beyond that through the areas surrounding these regions. Sample C (Figure 5), which was dropped from 431.8 mm and did fracture, showed the six radial cracks that originated from the impact area and extended to the sample edges. Some of the cracks, such as the left crack extending to the top surface and the right crack extending to the bottom surface appeared to be wider since they were imaged through the full thickness of the sample. The DR image of sample D (Figure 6), which was dropped from 591.8 mm, showed eleven radial cracks that extended from the impact area to the edge of the sample. Two additional cracks with slight attenuation differences compared to the undamaged material were also found, one extending from the major crack in the bottom left comer, and another connecting two cracks in the upper left comer. Sample E (Figure 7), which underwent the highest indenter drop height of 762.0 nun, also showed eleven radial cracks in the DR image. In addition, several other features exhibiting lower x-ray attenuation were noted including a circular feature in contact with one of the major cracks, a small crack extending to the lower right comer of the sample, and an isolated edge crack near the bottom right comer. While DR imaging was able to detect all of the major cracks that were visible on the sample surface, only minor variations attributed to bulk specimen damage were found. For quantitative damage comparison, the images were inverted and the percent area from darker regions of the DR images was used to estimate the degree of damage in each of the five samples. The calculated values were 0.00% for A, 0.00% for B, 14.06% for C, 14.22% for D, and 23.37% for E. While no damage was found in samples A and B, the percent damage values for samples C, D, and E, increased with drop height, impact velocity, and impact energy. ULTRASOUND C-SCAN IMAGING Ultrasound C-scan imaging was conducted on samples A through E for damage assessment and comparison to visual inspection. Selection of the alumina bottom surface signal for amplitude evaluation enabled bulk characterization of the tile without interference from the aluminum backing plate. The scale was set up in terms of attenuation, or signal loss, in decibels (dB), with darker features representing the highest degree and lighter features representing the lowest degree of signal loss. Ultrasound C-scan images are shown in Figures 3-7. The image for sample A showed a circular region in the center of the sample with a significant level of signal loss compared to the surrounding areas, as shown in Figure 3. This region was at the point of impact and was considered to be the only source of damage to the sample, as no cracking was detected. Material variations in the sintered alumina that were not a result of low velocity impact were also detected by the longitudinal 15 MHz transducer. A region along the left side of the sample with a higher level of attenuation indicated a slightly lower density cornpared to the rest of the sample. The C-scan image for sample B also showed a circular region in the center of the sample with significant signal loss, as shown in Figure 4. This region was slightly larger in area and higher in attenuation as compared to the region in sample A. Other features that were detected in the scan were three surface defects in the upper half of the sample and some minor density variations in the upper right hand comer of the sample. While there was no crack damage detected for samples A and B, ultrasound C-scan images of samples C, D, and E demonstrated significant impact damage and crack patterns. The regions with the highest degree of attenuation were most evident for the largest cracks and at the point of impact. Sample C (Figure 5 ) , which was tested at a drop height of 431.8 111111, showed eight radial cracks that extended from the impact region to the edge of the sample. Three additional cracks were also identified including one that originated from a radial crack and extended to the upper edge, one that started from the impact area and extended halfway toward the bottom left comer, and one that extended straight down to the bottom edge. All of these cracks exhibited a lower degree of attenuation
140
.
Advances in Ceramic Armor V
Destructive Testing and Nondestructive Evaluation of Alumina Structural Ceramics
compared to the initial eight radial cracks. Regional material variations were also present, as the bottom left comer showed lower signal loss than the rest of the sample. Sample D (Figure 6), which was tested at a drop height of 591.8 mm, appeared to have an impact region similar in area to sample C, but showed more cracking throughout the sample. Fourteen radial cracks extended from the impact region to the edge of the sample. Five cracks originated at the impact region without extending to the edge. There were also four additional cracks that extended from the initial radial cracks. Surface damage and density variations not caused by impact were also detected in sample D. Sample E (Figure 7), which was dropped from a height of 762.0 mm, had the largest impact area, which was roughly twice as large as those in samples C and D. There were twelve radial cracks that extended from the impact area to the sample edge, two cracks that originated from initial radial cracks, and five that did not reach the edge. Qualitatively, while the number of cracks did not directly correspond to the drop height, the highest velocity, highest energy impact from 762.0 mm produced the largest impact area. For quantitative damage comparison, the images were inverted and converted to grayscale, with the dark regions representing areas in which damage occurred. The values for the samples were 0.1 7% for A, 0.18% for B, 23.73% for C, 28.85% for D, and 36.33% for E. The degree of quantitative damage increased with drop height for the five samples. As compared to the DR images, the damage percentages were higher in the ultrasound images due to additional damage detected in the bulk. The percent damage also increased incrementally for the three fractured samples. X-RAY COMPUTED TOMOGRAPHY IMAGING X-ray computed tomography was conducted on fractured samples C, D, and E, for threedimensional imaging and analysis of the radial and cone cracks that propagated through the bulk of each sample. Just as in the DR images, darker features in the XCT images corresponded to a lower degree of x-ray attenuation and more pronounced damage. However, unlike the DR images, which were projected onto a single plane, the XCT images consisted of multiple slices through the thickness of each sample. These high resolution 2-D XCT slices, 25 to 30 collected for each fractured sample, were used to study the intricate crack patterns as they traveled from the impact surface to the bottom of each sample. Figures 8, 9, and 10 show selected image slices through the axial (x-y) plane as well as cross-sectional slices through the sagittal (y-z) and frontal (x-z) planes for samples C, D, and E. Volume Graphics Studio” software was utilized to take the sets of 2-D XCT slices and reconstruct them as 3-D volumetric scans. The 3-D images, as shown in Figure 11, were rotated to show isometric views of the bottom surface of each sample. The software was also used to cut through the images at various planes to simultaneously show the surface and bulk damage, as demonstrated in Figure 12.
Figure 8. Axial (top), sagittal (left), and frontal (right) 2-D XCT images of sample C.
Advances in Ceramic Armor V
. 141
Destructive Testing and Nondestructive Evaluation of Alumina Structural Ceramics
Figure 9. Axial (top), sagittal (left), and frontal (right) 2-D XCT images of sample D.
Figure 10. Axial (top), sagittal (left), and frontal (right) 2-D XCT images of sample E.
Figure 12. 3-D XCT cut plane images of samples C (left), D (center), and E (right).
The three-dimensional XCT images characterized the damage in a way that the 2-D DR and UT images could not. While the other methods imaged the damage as a collective volume through the entire thickness, the ability of XCT to account for progressive damage through ordered slices enabled
142
.
Advances in Ceramic Armor V
Destructive Testing and Nondestructive Evaluation of Alumina Structural Ceramics
more accurate imaging of the crack patterns as they propagated through each sample. In this way, a definitive distinction was made between the radial cracks and cone cracks. For sample C, the selected individual slices in Figure 8 showed changes in damage from the top surface to the bottom surface of the A1203 tile. While the first slice showed a small crater in the region where the indenter impacted the tile, the cracks originating from the point of impact became progressively wider from slice to slice, forming a partial cone crack. The cross-sectional images representing the sagittal and frontal slices of the sample also showed the angle of the cracks widening from the top to the bottom of the sample. Another noticeable change was the formation of a crack in the bottom right corner of the sample as the slices continued through the sample thickness. While the six radial cracks from the DR images were also found in the XCT images, the small internal cracks detected by UT were not detected in the XCT images. For sample D, the top surface XCT slice showed the central impact crater as well as the initial formation of a partial cone crack extending toward the bottom left corner of the sample (Figure 9). While this cone crack continued to move away from the center in consecutive scans, another partial cone crack started to open up in the third slice, extending toward the top right corner of the sample. While the cone cracks were not as well defined as they were in sample C, they were still detectable in the images. The same eleven radial cracks from the DR images were also detected in the XCT images, but the internal UT cracks could still not be resolved. For sample E, the impact crater was visible in the first slice, and the formation of three partial cone cracks was evident starting in the second slice (Figure 10). Just as in samples C and D, the cone cracks widened as the slices continued to the bottom surface of the sample. Some of the cross-sectional XCT slices through the center of the sample also showed the wide angle of crack propagation. As in sample D, sample E also showed a total of eleven radial cracks, but no additional internal damage detected by UT imaging. Quantitative data was extracted fiom the XCT images by examining the crack propagation of both radial and cone cracks. The radial crack lengths and angles of samples C, D, and E are shown in Tables 11, 111, and IV. The radial crack lengths were estimated by fitting planes through the 3-D XCT images, and the angles between neighboring radial cracks were estimated by determining the angles between the planes. The crack lengths ranged from 40 mm to 68 mm, but the average crack length for each of the three samples was identical at 54 mm. The angles between neighboring cracks ranged from 14 degrees to 112 degrees, with average angles of 61 degrees for sample C, 33 degrees for sample D, and 32 degrees for sample E. The smaller average angles between neighboring cracks for samples D and E reflected the higher number of radial cracks in those samples. The cone crack angles were also estimated for the three samples by comparing the dimensions of the impact crater at the surface to the approximate dimensions of the cone crack at the bottom of each sample. The estimated cone crack angles were 59.8 degrees for sample C, 57.2 degrees for sample D, and 54.1 degrees for sample E. While the cone crack angles were similar for the three fractured samples, they decreased slightly at higher velocities and impact energies. Table 11. Crack length and neighboring crack angle measurements for sample C. Crack #
Crack Length [mm]
1
46 53 46 59 59 63
2
1
3 1 1
4
1
5 6
1
Angle with Neighboring Crack
I
53 79 47 39 34
["I
I
112
Advances in Ceramic Armor V
.
143
Destructive Testing and Nondestructive Evaluation of Alumina Structural Ceramics
Table 111. Crack length and neighboring crack angle measurements for sample D. Crack #
Crack Length [mm]
I
68
26
2
50
38
3 1
Angle with Neighboring Crack
49
4
1
62
[“I
Table IV. Crack length and neighboring crack angle measurements for sample E.
I
59 29
41
I
10
40
Angle with Neighboring Crack
1
52
23
55
61
[mm]
1
5
1
56
1
6
1
52
1
7
1
58
43
I
25 34
[“I
31
I I I
t
I
Crack Length
25
I
24
I
Crack #
I I I I
I
I I II
23 14 18
8
1
52
41
9
1
68
21
10
I 1
48
11
56
I I
18
77
I
After completing quantitative analysis of the XCT images, an additional study was performed on sample C to look at the damage in more depth. Surfacer*’ software was utilized to perform a more detailed analysis of the radial and cone cracks. By selecting thresholds to separate damaged and undamaged regions, 3-D point clouds of the damaged sample areas were created. In contrast to the other XCT images in which the damaged regions were characterized by the absence of data, the 3-D point clouds served as a volumetric representation of the actual damage in sample C. By studying these damage patterns in greater detail, the radial and cone cracks were distinguished and separated from each other. In order to achieve this, the cone crack patterns were curve fit at each 2-D XCT slice. The reconstruction of these slices resulted in a 3-D representation of the cone crack, which was isolated from the rest of the sample and separated from the remaining radial cracks. Axial and isometric views of the cone crack reconstructions were overlaid onto the radial cracks, as shown in Figure 13. Axial and isometric views of the cone crack reconstructions were extracted and removed from the full damage volume, as shown in Figure 14. The 3-D representation of the cone crack demonstrated the ability to separate specific features and damage types from a given matrix.
Figure 13.Top view (left) and isometric view (right) of 3-D damage images of sample C.
144
.
Advances in Ceramic Armor V
Destructive Testing and Nondestructive Evaluation of Alumina Structural Ceramics
Figure 14. Side view (top), top view (left) and isometric view (right) of cone crack from sample C.
CONCLUSIONS A combination of destructive drop tower testing methods and nondestructive imaging methods was described. Drop tower testing was effective for creating damage in the form of through thickness radial and cone cracks. A well defined transition range between the behavior of non-fractured and fractured A1203 samples was established. A threshold region between 2.87 and 2.90 d s e c was found below which no fracture occurred and above which fracture always occurred. Comparison of visual inspection to ultrasound, digital radiography, and x-ray computed tomography demonstrated the capabilities of NDE methods for detecting internal bulk damage in the A1203 samples. All three techniques successfully identified the major radial cracks, which added up to six for sample C, eleven for sample D, and eleven for sample E. The DR images also detected some details that were not evident by visually observing the sample exterior, but these were relatively minor. The XCT images were most effective at using 3-D capabilities to characterize radial and cone crack propagation through the thickness of the fractured samples. By using Volume Graphics and Surfacer software, crack lengths and angles were calculated, and a more in-depth analysis of the crack types enabled a separation of the cone cracks from the radial cracks. A 3-D representation of the cone crack pattern is sample C was successfully reconstructed from the XCT images. While the XCT images held the advantage of imaging the detectable damage, it was not until ultrasound evaluation was conducted that the full extent of bulk damage was found. In addition to detection of all the major cracks found by visual inspection, DR, and XCT, a large degree of bulk damage was detected during C-scan imaging. This damage was the result of internal cracks that were present in the bulk of the samples. The number of significant internal cracks found by C-scan imaging of the drop tower tested samples was five in sample C, twelve in sample D, and eight in sample E. By identifying the extended bulk damage in the samples, more accurate quantitative damage values were obtained. Besides the characterization of internal damage in the samples, additional material and density variations were also detected. While UT held the advantage of small internal crack detection, the images were limited to showing volumetric damage on a 2-D image map. The UT scans showed large regions of damage at the impact points, but XCT images determined that the damage did not behave the same way through the entire thickness. Rather, the impact crater propagated as a cone crack, getting wider in angle, as opposed to remaining uniform. This was something that could be indicated in the 3-D XCT images but not the UT scans. The detection differences between NDE techniques were due to the nature of inspection. While the DR and XCT methods detected bulk features based on x-ray attenuation, the ultrasound method detected bulk features based on acoustic impedance mismatch and acoustic signal attenuation. For the DR images, the internal cracks that could not be detected had widths that were smaller than the x-ray
Advances in Ceramic Armor V
.
145
Destructive Testing and Nondestructive Evaluation of Alumina Structural Ceramics
spatial resolution. These were hairline fractures with narrow gap sizes that were averaged out during collection of the x-ray data. On the other hand, the C-scan images were very sensitive to acoustic impedance mismatch between the alumina tile and the immersion medium. These variations were identified by the 15 MHz longitudinal transducer during C-scan imaging of the samples. This inherent detection characteristic proved to be the difference during detection of internal damage. REFERENCES ID. Sherman, Impact Failure Mechanisms in Alumina Tiles on Finite Thickness Support and the Effect of Confinement, International Journal o f h p a c t Engineering, 24,3 13-328 (2000). 2N. Bourne, J. Millett, Z. Rosenberg, and N. Murray, On the Shock Induced Failure of Brittle Solids, Journal of the Mechanics and Physics of Solids, 46, 1887-1908 (1998). 3D. Sherman and T. Ben-Shushan, Quasi-static Impact Damage in Confined Ceramic Tiles, International Journal oflmpact Engineering, 21,245-265 (1998). 4B.A. Latella, B. O’Conner, N.P. Padture and B.R. Lawn, Hertzian Contact Damage in Porous Alumina Ceramics, Journal of the American Ceramic Society, 80, 1027-103 1 (1997). 5C.E. Anderson, J.D. Walker and J. Lankford, Investigations of the Ballistic Response of Brittle Materials”, US Army Research Ofice Technical Report - SwRI Project 06-51 17/002, 1-112 (1995). 6J.N. Singletary, R.A. Coffelt, J.W. Gillespie and B.A. Gama, Impact and High Strain Rate Response of 3-D Woven Systems, US Army Research Ofice Technical Report- AR041472.1-EG-ST1,1-71 (2001). 7J.G. Chacon-Nava, F.H. Stott, S.D. de la Torre and A. Martinez-Villafane, Erosion of Alumina and Silicon Carbide at Low-Impact Velocities”, Materials Letters, 55,269-273 (2002). ‘S. Bueno, L. Micele, C. Baudin and de G. Portu, Reduced Strength Degradation of Alumina-Alumina Titanate Composite Subjected to Low-Velocity Impact Loading”, Journal of the European Ceramic Society, 28, 2923-2931 (2008). 9J.H. Park, S.K. Ha, K.W. Kang, C.W. Kim and H.S. Kim, Impact Damage Resistance of Sandwich Structure Subjected to Low Velocity Impact, Journal OfMaterials Processing Technologv, 201,425430 (2008). ”Z.Y. Zhang and S.O.W. Richardson, Low Velocity Impact Induced Damage Evaluation and its Effect on the Residual Flexural Properties of Pultruded GRP Composites”, Composite Structures, 81, 195201 (2007). “P.M. Schubel, J. Luo and I.M. Daniel, Low Velocity Impact Behavior of Composite Sandwich Panels, Composites: Part A, 36, 1389-1396 (2005). ’*J. Gustin, A. Joneson, M. Mahinfalah and J. Stone, Low Velocity Impact of Combination KevlariCarbon Fiber Sandwich Composites, Composite Structures, 69,396-406 (2005). I3S. McMichael and S. Fischer, Understanding Materials with Instrumented Impact”, ME, 47-50 (1989). 14M.J.Dennis, ASM International Handbook Vol. 17 on Nondestructive Evaluation and Quality Control: Industrial Computed Tomography (1989). I5J.H.Stanley, Physical and Mathematical Basis of CT Imaging: ASTM Tutorial Section 3, ASTM CT Standardization Committee (1986). I6T.H. Newton and D.G. Potts, Technical Aspects of Computed Tomography Vo1.5,The C.V. Mosby Company (198 1). I7P.E.Mix, Introduction to Nondestructive Testing, John Wiley & Sons, pp. 104-153 (1987). “5. Krautkramer. and H. Krautkramer, Ultrasonic Testing of Materials, Springer-Verlag (1990). I9M.C. Bhardwaj, Ceramic Monographs - Handbook of Ceramics, 41, (1992). 2owww.volumegraphics.com 2’D. Koker, A. Clemons, M. Wiesner and K. Prior, SDRChageware Basic Reverse Engineering with Surfacer Training Guide, 9, 1999.
146
.
Advances in Ceramic Armor V
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
NONDESTRUCTIVE EVALUATION OF AS FABRICATED AND DAMAGED ENCAPSULATED CERAMICS William H. Green, Raymond Brennan, and Robert H. Carter U.S. Army Research Laboratory Weapons and Materials Research Directorate ATTN: AMSRD-ARL-WM-MD Aberdeen Proving Ground, MD, USA ABSTRACT X-ray computed tomography (XCT) and ultrasonic testing (UT) are two major nondestructive evaluation (NDE) methods used to inspect a wide variety of materials and components. Current work in the area of NDE of materials at the Army Research Laboratory includes inspection and analysis of both individual ceramic targets and ceramic panels. A number of samples have been evaluated using both XCT and UT, including encapsulated panels. XCT and UT results from NDE studies of as fabricated and damaged encapsulated ceramic panels will be shown and discussed. The results from the two NDE methods will be compared and contrasted. INTRODUCTION Nondestructive evaluation (NDE) or nondestructive testing (NDT) is a discipline of materials science that encompasses a wide variety of inspection modalities. NDE is applicable to an extremely wide variety of materials, components, and systems and is utilized to inspect objects at the surface, subsurface, and in the interior. Two methods used for the evaluation and analysis of internal geometrical and physical characteristics of materials are x-ray computed tomography (XCT) and ultrasonic testing (UT) or scanning. Both of these methods have been used to characterize armor ceramics, including ballistically damaged ceramics [l-51, and XCT has been used to characterize and evaluate ballistically damaged encapsulated ceramic panels [6]. Ceramic materials are currently typically combined with other materials in armor panel structures in order to decrease weight without losing ballistic performance. Encapsulated panels in which the ceramic material is enclosed and backed by a supporting material are an example of this approach. Two encapsulated ceramic panel specimens were characterized and evaluated using XCT and, in the case of the second specimen, also characterized using UT. DESCRIPTION OF SPECIMENS AND DIGITAL RADIOGRAPHY SCANS The first encapsulated specimen was an approximately 207 mm (8.1”) by 361 mm (14.2”) rectangular section from a larger fully penetrated impacted test panel. The specimen included a complete penetration and the surrounding area as well as undamaged material farther away. The backing material was not present on the specimen as received. Figure 1 shows photographs of the first specimen with its detached backing material. The second specimen was an approximately 180 mm (7.1”) by 210 mm (8.25”) rectangular section from a larger non-impacted test panel. It is believed that the physical sectioning process may have resulted in cracking in the ceramic material of the specimen. Cracking in the ceramic material is visible in the exposed sectioned side of the specimen. Digital radiographs (DRs) of each specimen were taken through their thickness using the 420 keV x-ray tube and linear detector array (LDA) setup in centered rotate-only (RO) mode. The x-ray technique (parameters) of the DRs of the first and second specimen were (405 keV, 2.0 mA) and (400 keV, 2.0 mA), respectively, and geometries of source-to-object-distance (SOD) = 750.00 mm and source-toimage-distance (SID) = 940.00 mm. Figure 2 shows digital radiographs of each specimen. The two DRs of the first specimen (2a and 2b) have been processed or “windowed” to accentuate one or some features over others. In the first the damage itself immediately around and farther away from what is
147
Nondestructive Evaluation of as Fabricated and Damaged Encapsulated Ceramics
left of the penetration cavity is emphasized, as is the damage in the ceramic tiles below the penetrated tile. It also shows the tile layup very clearly. In the second image the actual perimeter of the entrance hole of the penetration cavity itself is emphasized, showing only relatively thin material left intact at the edges making the rest of the image significantly lighter. The diameter of the entrance of the penetration cavity is approximately 60 mm. The cracking in the second specimen is not apparent in Figure 2c, which is at a larger scale. The sectioned side of the specimen is on the left and the two approximately horizontal features are visible on the surface of the specimen.
(a) (b) FIGURE 1. (a) Front and (b) Back photographs of the ballistically damaged specimen with the detached titanium backing next to it (on the right in both images). The upper left section in (a) has fractured away from the rest of the panel, and is not included in the x-ray evaluation.
(a) (b) (c) FIGURE 2. Digital radiographs of both specimens. (a) Image emphasizing damage and tile layup, and (b) Penetration cavity in ballistically damaged specimen. (c) As-fabricated and sectioned specimen (larger scale). XCT AND UT SCANNING PROCEDURES A preliminary series of XCT scans incrementally spaced by 20.00 mm were taken at the bottom to the top of the first specimen. The specimen was held with its faces in a vertical orientation by a portable vice for scanning. Thus, the specimen faces were perpendicular to the horizontal x-ray (collimated) fan beam resulting in through thickness cross-sectional CT images. The vice was suitably stabilized on the scanning turntable. The middle of the penetration cavity was at a vertical position of approximately 288 mm. The vertical scan positions were based on multiples of 20.00 mm below,
148
Advances in Ceramic Armor V
Nondestructive Evaluation of as Fabricated and Damaged Encapsulated Ceramics
above, and including the 288.00 position. The undamaged bottom (edge) of the specimen was at a vertical position of about 20 mm and the top position of the remaining upper portion was at about 381 mm. The specimen was scanned using the 420 keV x-ray tube and LDA set up in offset RO mode. The slice thickness was 0.500 mm and each slice was reconstructed to a 1024 by 1024 image matrix. The field of reconstruction (FOR) diameter was 230.00 mm. The tube energy and current used were 405 keV and 2.0 mA, respectively, and the focal spot was 0.80 mm. The SOD and SID were 750.00 mm and 940.00 mm, respectively. The first series of scans were done to get a good understanding of the overall changes in damage features throughout the specimen. A second set of scans was done starting below the penetration cavity and ending within the penetration cavity. These scans were vertically overlapping with a slice thickness and increment of 0.500 mm and 0.450 mm, respectively. This was the majority of the slice (image) data analyzed using 3-D solid and point cloud visualization. Several sets of XCT scans were taken of the second specimen. This resulted in an overall collection of scans incrementally spaced by 5.00 mm starting at the vertical position of the specimen just above the top of the vice. The second specimen was also fixed and held with its faces vertical relative to the scanning turntable like the first specimen. Three sets of overlapping scans were taken in particular sections of interest between vertical positions of 97.00 mm and 122.20 mm, 125.00 mm and 148.40 mm, and 182.00 mm and 206.30 mm. The slice thickness of all the scans was 1.000 mm and the slice increment of the overlapping scans was 0.900 mm. The bottom (edge) of the specimen was at a vertical position of about 54 mm and the position of the top was at about 264 mm. Each slice was reconstructed to a 1024 by 1024 image matrix and the FOR diameter was 195.00 mm. The tube energy and current used were 400 keV and 2.0 mA, respectively, and the focal spot was 0.80 mm. The SOD and SID were 750.00 mm and 940.00 mm, respectively. The second specimen was also ultrasonically scanned using a pulse-echo immersion (water) setup. Frequencies of 1 M H z , 5 MHz, 10 MHz, and 15 M H z were used in the scans. Amplitude difference C-scans gated to show the bulk characteristics of the specimen through its thickness were taken at 5 MHz and 10 MHz. The 5 M H z and 10 M H z transducers used for this study were a Panametrics V307 broadband type with a 25.4 mm (1.00”) diameter element size and a 215.9 mm (8.50”) spherical point focus and a Panametrics V315 broadband type with a 19.05 mm (0.75”) diameter element size and a 163.8 mm (6.45”) spherical point focus, respectively. The acoustic signals reflected from the top surface and bottom surface of the specimen were selected, or gated. The amplitude of these individual gated signals was measured in addition to measuring the amplitude difference between the gates. The time-of-flight (TOF) was also measured by calculating the difference in milliseconds between the top and bottom surface gated signals. Using a step size of approximately 0.25 mm (0.01’3, the amplitude and TOF variations were collected and mapped over the area of the specimen (210 mm x 180 mm) to look for acoustic variations in the bulk of each sample that represented defects or inhomogeneities [7, 81. EVALUATION OF BALLISTICALLY DAMAGED SPECIMEN (SPECIMEN 1) Computed Tomography Scans Figure 3 shows a series of CT scans (images) of the specimen starting at a vertical position of 148.10 mm (3a). The scans in Figures 3b to 3j were taken at vertical positions of 167.90 mm (3b), 188.15 mm, 207.95 mm, 228.20 mm, 248.00 mm, 251.15 mm, 262.85 mm, 288.00 mm, and 308.00 mm, respectively. The darker vertical bands in these images are indications of the area between adjacent tiles. The slightly darker oblong feature with a somewhat crosshatched appearance in the middle of some of the images (lower heights) is an image artifact. It is not an indication of a real physical feature in the specimen. Damage is clearly evident in the top, or back, (exit) side of the specimen at 148.10 mm. There is very slight cracking as far away from the center of the penetration cavity as 160 mm at a scan height of 128.00 mm. The damage is significantly more severe at 167.90
Advances in Ceramic Armor V
.
149
Nondestructive Evaluation of as Fabricated and Damaged Encapsulated Ceramics
mm with connected cracking over most of the width of the specimen towards the back side. Cracks are also present in the middle thickness region of the specimen. Multiple cracks are near and adjacent to the bottom, or front, (impact) side of the specimen at 188.15 mm (3c), about 100 mm from the center of the penetration cavity. Braking up of some of the ceramic into rubble as well as bulge in the back face is also present at this height. The amount of rubble is higher with relatively large pieces and the bulging is more severe at 207.95 mm (3d). The ceramic material is clearly cracked through from the front to the back and both the front and back side parts of the encapsulant (case) are also cracked with multiple cracks in the front. The distance between the outer cracks in the front part of the case is about 81 mm. At 228.20 mm (3e), about 60 mm from the center of the penetration cavity, the rear part of the case is blown open and peeled back and the distance between the two cracks in the front part of the case is about 25 mm. Secondly, at least half of the thickness of the ceramic material no longer has any structural integrity. At 248.00 mm ( 3 0 the rear of the case is peeled back more and there is no ceramic material in the center area between the front and the back of the specimen. The front of the case is also on the verge of being penetrated with a major crack. At 251.15 mm (3g) the case and ceramic material have been completely penetrated, with some residual penetrator material (white) just above and to the right of the hole in the front of the case. The hole in the front of the case and the penetration cavity are larger at 262.85 mm (3h), about 25 mm from the center of the penetration cavity. There is not very much material left at 288.00 mm (39 being the scan height of the approximate center of impact. Essentially very little is still intact at this height, including the part of the specimen in the left hand side of the image which was missing. Physically this was the right hand side of the specimen as viewed looking at the exit side from the x-ray source perspective (see Figure 2). At 288.00 mm and higher the right hand side of the specimen is missing having been blown off by the impact and the image at 308.00 mm (3j) is just showing the remaining side.
150
Advances in Ceramic Armor V
Nondestructive Evaluation of as Fabricated and Damaged Encapsulated Ceramics
(9 (r) FIGURE 3. A series of cross-sectional CT scans (images). Scans (a-j) were taken at vertical positions of148.10mm, 167.90mm, 188.15mm,207.95mm,228.20mm,248.00mm,251.15mm,262.85mm, 288.00 mm, and 308.00 mm, respectively. Three-Dimensional Solid Visualization The excellent dimensional accuracy and the digital nature of XCT images allow the accurate volume reconstruction of multiple adjacent or overlapping slices. A virtual three-dimensional (3-D) solid image is created by electronically stacking the XCT images, which have thickness over their cross sections (i.e., voxels), one on top of the previous from the bottom to the top of the specimen or scanned height to generate its virtual volume. The 3-D solid images of the specimen were created using the second set of overlapping scans from a height of 140.00 mm to a height of 265.55 mm. Figure 4 shows a series of 3-D solid images of the scanned volume with sections virtually removed in (c-f). The method of virtual sectioning, which is essentially only showing a portion of each scan, allows viewing of generated surfaces anywhere in the scanned volume in 3-D space. The view in Figure 4 is looking at the front of the specimen with it tilted forward, except for 4b which is looking at the back of the specimen with that side tilted forward. Figures 4a and 4b show the entire scanned volume with surfaces and no sections virtually removed. Figure 4b shows the break up and rubble of the ceramic material at the edges of the penetration cavity. The virtual sectioned surfaces in Figures 4c-4f are approximately 2.2 mm, 6.7 mm, 11.2 mm, and 15.7 mm from the front face of the specimen, respectively. The increasing amount of damage around the penetration cavity with increasing distance from the front face is readily apparent. Damage is visible relatively far from the penetration cavity at a height of about 140 mm at the through thickness distance of 15.7 mm (4f). These 3-D solid images and the CT scans they were created from are indicative of the failure of the ceramic material and encapsulation at the back side ‘of the specimen relatively far from the penetration cavity before the specimen was completely penetrated. The view in Figure 5 is looking at the side of the specimen rotated to the right such that the front of it is on the left. The beginning of the entrance hole of the penetration cavity is near the top of the images in the front face. The virtual sectioned surfaces in Figures 5a-5h are approximately 9.9 mm, 19.7 mm, 29.6 mm, 39.5 mm, 59.7 mm, 69.6 mm, 90.7 mm, and 125.3 mm from the side of the specimen, respectively. It can be clearly seen how ceramic material cracked and broke away with the case in the back right area (impact view) around the penetration cavity. The surface at 90.7 mm (5g) is about at the center of the penetration cavity and the surface at 125.3 mm (5h) is at the other side of the penetration cavity. Secondly, damage is visible relatively far from the penetration cavity towards the back of the specimen (right side) at a height of about 140 mm, which is at the bottom of the images, in Figures 5e-5h. Three-Dimensional Point Cloud and Surface Visualization A 3-D point cloud is a set of points in space that define geometrical characteristics (i.e., shape, size, location) of a specimen or scanned volume and features within it. Location of the points is determined by appropriate (image) segmentation of the feature or features of interest. Figure 6 is a point cloud of the hole in the front of the panel as defined by the DR (Figure 2b) with a circle fit to it.
Advances in Ceramic Armor V
151
Nondestructive Evaluation of as Fabricated and Damaged Encapsulated Ceramics
(4 (el (9 FIGURE 4. A series of 3-D solid visualization images with material removed from the front face towards the back face. (a) Front (impact) side. (b) Back (exit) side. (c-f) Front view with 2.2 mm, 6.7 mm, 11.2 mm, and 15.7 m of material removed, respectively.
(9 (g) (h) (el FIGURE 5 . A series of 3-D solid visualization images with material removed uemendicular to the faces from right side as viewed from the front. (a-h) ciew looking at front and side with 9.9 mm, 19.7 mm, 29.6 mm, 39.5 mm, 59.7 mm, 69.6 mm, 90.7 mm, and 125.3 mm of material removed, respectively. The diameter of the entrance hole is 58.40 mm with the center at a height of 282.00 mm. Figure 7 is a point cloud of the overall damage and back face of the specimen in the second overlapping set of
152
Advances in Ceramic Armor V
Nondestructive Evaluation of as Fabricated and Damaged Encapsulated Ceramics
scans. The view is looking at the front of the specimen with it tilted forward 50 degrees from a perpendicular line of sight. The back face is the flat looking areas of points to the left and right of the middle “bulge”, which is at the front of the specimen, and peeled back case. The points in the bulge include the bowing of the front of the specimen, damage outside of the penetration cavity, and the penetration cavity itself. The presence of damage relatively far from the penetration cavity at the top of the image is shown by the isolated points at the bottom of the image. Figure 8 is a point cloud of the bulge and the damaged case behind it, in which the back face and isolated points have been removed (from Figure 7), and is also tilted at 50 degrees. Figure 9 is a point cloud created using only the set of overlapping scans that went through a section of the penetration cavity itself. The top image is looking at the top of the specimen in the negative z (-z) direction. The bottom image is only the walls of the cavity in which they are tilted into the page relative to the top down view in the top image. Free form 3-D surfaces (Non Uniform Rational Bezier/Basis Spline, or NURBS, method [9, lo]) were fit to the cavity wall point clouds. These are shown by the wavy surfaces passing through the points on either side of the section of penetration cavity in Figures 10 and 11, which are isometric views from rear side and impact side perspectives, respectively, including all of the points to show the specimen edges. Although these surfaces follow the section of penetration cavity quite well, it is useful to take a simpler and more directly informative approach of fitting planes to these point clouds. The fit planes (NURBS) are shown in Figures 12 and 13, which are also isometric views from rear side and impact side perspectives including all of the points. The tilt of the two planes defining the section of penetration cavity can be seen more easily than the general behavior of the free form surfaces. The internal angle between the two planes in the physical x-y plane (CT scan plane) is about 125 degrees.
FIGURE 6 . Point cloud of entrance hole with fit circle for diameter.
FIGURE 7. 3-D point cloud of overall damage and back face.
FIGURE 8. 3-D point cloud of front bulge, penetration cavity, and damaged case.
FIGURE 9. 3-D point cloud of section of penetration cavity.
Advances in Ceramic Armor V
153
Nondestructive Evaluation of as Fabricated and Damaged Encapsulated Ceramics
FIGURE 10. Point cloud of cavity section w/ fit FIGURE 11. Point cloud of cavity section wi fit free form surfaces (NLTRBS). Rear perspective. free form surfaces (NURBS).Impact perspective.
FIGURE 12. Point cloud of cavity section w/ fit FIGURE 13. Point cloud of cavity section w/ fit planar surfaces (NURBS). Rear perspective. planar surfaces (NURBS). Impact perspective. The equations [mm] of the left and right planar surfaces in Figure 13 (impact side perspective)
are
.437x + .674y +.595z -136 = 0 (Left) - . 4 7 ~+ , 6 0 8 +~. 6 3 9 ~-147 = 0 (Right)
. which can be written as = -1.54~ 1.362 + 31 1 ~ , = 1 . 2 9 y +1.362-313
XL
Table I gives the data set of {z, yn, x,~, xnR} for five values of z (vertical position) within the penetration cavity starting near the bottom of the cavity, where three values of y (n=3) were chosen , z) spanning the through thickness depth (+y direction) of the cavity. The {xd, yn, z} and { x ~ R yn, points give the approximate location of the walls of the section of the cavity and provide the fitting Table I. Data Points {z, yn, XL, xa}on Fit Planar Surfaces [mm].
154
.
Advances in Ceramic Armor V
Nondestructive Evaluation of as Fabricated and Damaged Encapsulated Ceramics
data to generate a representative penetration cone surface. Figure 14 shows isometric views from an impact side perspective of the fit penetration cone relative to segmented point cloud representations of selected CT scans, which show the boundaries of the damaged specimen. The penetration cone has an internal angle of 121.9 degrees and an upward tilt out of the x-y plane of 34 degrees. The surface of the cone is mesh shaded in Figure 14a in order to maximize the visibility of the points in the vicinity of the penetration cavity. The cone is shaded with an opaque surface in Figure 14b in order to emphasize the location, angle, and tilt of the cavity within the damaged structure of the specimen.
(a) (b) FIGURE 14. Isometric impact side views of fit cone characterizing entire penetration cavity and selective bottom-to-top CT scans of specimen that have been (gray scale) segmented and converted to point cloud representations to show the outer boundaries of the specimen. The penetration cone has an internal angle of 121.9 degrees and an upwards tilt out of the x-y plane (CT scan plane) of 34 degrees. EVALUATION OF AS-FABRICATED AND SECTIONED SPECIMEN (SPECIMEN 2) Ultrasonic Scans Figure 15 shows two through thickness amplitude difference C-scans, both gated to show the bulk characteristics of the specimen, taken using 5 MHz (15a) and 10 MHz (15b) columnar immersion transducers. The sectioned side of the specimen is on the right and the long vertical dimension (210 mm) is labeled. The up direction from the bottom of the images towards 210 mm corresponds to increasing z in the CT scans of the specimen. Darkest and lightest gray represent the least and most attenuation (signal loss), respectively. The two toned lighter band around the perimeter of the specimen is caused by edge effect attenuation. The pattern of the tile layout can be clearly seen as well as two spacers in the left hand side of the specimen, especially in Figure 15b. Figure 15a shows that there is some significant attenuation over most of the specimen and Figure 15b shows a number of individual features, including an approximately double lobe shaped area of varying attenuation down the center of the specimen starting at about 28 mm from the bottom of it. Figure 15b also shows two horizontal bands of higher attenuation vertically centered at about 74 mm and 143 mm from the bottom of the specimen. The lower band extends from the left hand side of the middle tile in its bottom area to the right hand side of the specimen. The upper band is mostly in the top area of the middle tile with a small portion of it in the lower area of the top middle (partial) tile. The entire area of the top right (partial) tile is relatively highly attenuated.
Advances in Ceramic Armor V
. 155
Nondestructive Evaluation of as Fabricated and Damaged Encapsulated Ceramics
(a) (b) FIGURE 15. Through thickness amplitude difference C-scan images of the specimen, both gated to show bulk characteristics, taken using, (a) 5 MHz and (b) 10 MHz, immersion transducers. The scale is given in order to provide the location of CT scans above the bottom of the specimen for comparison to the areal C-scans. Computed Tomography Scans Figure 16 shows a series of CT scans (images) of the specimen at vertical positions of 80.00 mm, 105.00 mm, 11 1.40 mm, 126.80 mm, 144.80 mm, 148.40 mm, 165.00 mm, 194.60 mm, 195.50 mm, 204.5 mm, 225.00 mm, and 240.00 mm (a - I). The front of the specimen is at the top of the images. The darker vertical bands in the images are indications of the area between adjacent tiles. The very faint concentric rings feature in the center of the images is an image artifact due to using relatively narrow image windowing to increase contrast. It is not an indication of a real physical feature in the specimen. From the lowest to the highest scan they were approximately 26 mm, 5 1 mm, 57mm,73mm,91 mm,94mm,111mm,141mm,142mm,150mm,171mm,and186mmabovethe bottom of the specimen, respectively, which is at the bottom of the scans in Figure 15. The sectioned side of the specimen is on the right as it is for the ultrasonic scans. Cracking is evident in the sectioned side of the specimen as well as in the left side of the specimen. The scan at 26 mm (16a) shows the cracking in both sides of the specimen and a faint indication of cracking in the bottom right area of the center tile, which has fair correlation with the Cscan image in Figure 15b. The scan at 51 mm (16b) shows a horizontal band of lower density material below the middle tile. This is a few millimeters below the top of the bottom middle tile in the C-scans, which is in the area of the double lobe shaped feature. The entire middle tile area has a lower density in the scan at 57 mm (16c), which corresponds to the horizontal region between the middle and bottom middle tiles in the C-scans. Lower density horizontal bands are also apparent below the middle and right tiles of the scan at 73 mm (16d), which correlates well with the higher attenuation region centered at about 74 mm in the C-scans. The scans at 91 mm (16e) and 94 mm (169 both show low density bands below portions of the middle and right tiles and the cracking in the sectioned side of the specimen extending into the middle tile. The scan at 94 mm, which appears to have a slightly lower density band than the scan at 91 mm, is near the top of the bottom right tile in the C-scans. The scan at 111 mm (16g) shows lower density bands below the middle and right tiles with similar thickness and a somewhat faint short linear feature starting at the bottom right comer of the middle tile just above the low density band, which probably indicates that the cracking in the sectioned side of the specimen extends into the middle tile in this scan also. The crack in the left side of the specimen also appears to
156
a
Advances in Ceramic Armor V
Nondestructive Evaluation of as Fabricated and Damaged Encapsulated Ceramics
reach the middle tile. This correlates well with the higher attenuation region just above the center of the middle tile and the region to its right in the C-scans. There is also some indication of the extended crack in the left side of the specimen near the bottom of the top left tile in the C-scans. The scans at 141 mm (16h) and 142 mm (16i) show wavy and non-uniform regions of low density, respectively, near the front (top) of the middle tile that extend into the right tile. This correlates well with the higher attenuation region centered at about 143 mm in the C-scans. The low density feature below the middle tile in the scan at 150 mm (16j) appears to extend into the tiles on each side of it. This is in the vicinity of the top of the middle tile in the C-scans, which is in fair correlation. The scans at 171 mm (16k) and 186 mm (161) both show low density bands below the left and right tiles with the wider band below the right tile, which is the sectioned side of the specimen. Both scans also show a low density band above the middle tile that appears to be wider in these scans than in the other scans. These scans seem to have a better correlation with the C-scan in Figure 15a than the one in Figure 15b. The scan at 186 mm also shows a crack in the bottom right comer of the middle tile. This particular feature correlates well with the higher attenuation region that forms a protrusion out of the upper portion of the top right tile into the top middle tile in the C-scan in Figure 15b.
FIGURE 16. A series of cross-sectional CT scans (images). Scans (a-1) were approximately 26 mm, 51 mm, 57 mm, 73 mm, 91 mm, 94 mm, 111 mm, 141 mm, 142 mm, 150 mm, 171 mm, and 186 mm above the bottom of the specimen, respectively.
Advances in Ceramic Armor V
.
157
Nondestructive Evaluation of as Fabricated and Damaged Encapsulated Ceramics
CONCLUSIONS A wide range of ballistic damage in a sectioned encapsulated ceramic panel specimen including complete penetration was scanned and extensively characterized using XCT 2-D cross-sectional (planar) and 3-D volumetric analysis. Several damage features including low severity ceramic cracking relatively far from the penetration cavity, ceramic cracking, fragmentation, and rubble, encapsulation cracking and exit (rear) side peel back, impact (front) face bulging, and penetration cavity size and geometry were captured and discussed. Successive application of XCT 2-D evaluation, volumetric solid visualization and analysis, and volumetric point cloud visualization and derived feature surface analysis provided extensive and important qualitative and quantitative data about damage features. Characteristics of captured damage features provided better understanding of the physical processes of damage initiation and growth. Several features in a different sectioned non-impacted encapsulated ceramic panel specimen were characterized using UT and XCT, including cracking that may have been caused by the physical sectioning process. Ultrasonic amplitude difference C-scans through the thickness of the specimen showed a number of attenuation features, including wide bands and regions over large areas of tiles of higher attenuation. XCT scans showed a number of spatial features located in 3-D space, including cracks in the edges of tiles extending hrther into the interior and low density regions. The ultrasonic bulk C-scans and through thickness spatial XCT scans had a fairly good correlation with a number of features matching up. The XCT and UT methods can synergistically provide comprehensive and detailed information about the internal geometrical and physical characteristics of materials. REFERENCES 'W. Green, H. Miller, J. LaSalvia, D. Dandekar, and D. Casem, Evaluation of Ballistically-Induced Damage in Ceramic Targets by X-ray Computed Tomography, Proceedings of 32"d International Conference on Advanced Ceramics and Composites - Topics in Ceramic Armor, (2008). 2W. Green, N. Rupert, and J. Wells, Inroads in the Non-invasive Diagnostics of Ballistic Impact Damage, Proceedings of 2jrhArmy Science Conference, (2006). 3H. Miller, W. Green, and J. LaSalvia, Ballistically-Induced Damage in Ceramic Targets as Revealed by X-ray Computed Tomography, Proceedings of 31" International Conference on Advanced Ceramics and Composites - Topics in Ceramic Armor, (2007). 4N. Bourne, W. Green, and D. Dandekar, On the One-dimensional Recovery and Microstructural Evaluation of Shocked Alumina, Proceedings of the Royal Sociep A: Mathematical, Physical, and Engineering Sciences, published online: doi: 10.1098/rspa.2006.1713,(2006). 'J. Wells, W. Green, N. Rupert, and D. MacKenzie, Capturing Ballistic Damage as a Function of Impact Velocity in Sic-N Ceramic Targets, Proceedings of 3dhInternational Conference on Advanced Ceramics and Composites -Advances in Ceramic Armor, (2006). 6W. Green and R. Carter, Evaluation of Ballistic Damage in an Encapsulated Ceramic Panel via X-ray Computed Tomography, Proceedings of Review ofprogress in Quantitative NDE, (2008, to be published). 'P. E. Mix, Introduction to Nondestructive Testing, John Wiley & Sons, pp. 104-153, (1987). *J. Krautkramer and H. Krautkramer, Ultrasonic Testing of Materials, Springer-Verlag, (1990). 'Wikipedia, Nonuniform rational B-spline, http://en wikipedia.ora/wiki/NURBS, (January 2009). ''SDRChagewareO, Basic Reverse Engineering With Surfacer: Training Guide, pp. (244-245, 326-327), (March 1999).
158 . Advances in Ceramic Armor V
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
MICROSTRUCTURAL STUDY OF SINTERED SIC VIA HIGH FREQUENCY ULTRASOUND SPECTROSCOPY A. R. Portune and R. A. Haber Rutgers University Materials Science and Engineering Piscataway, New Jersey, USA ABSTRACT In order to form better ballistic models of real armor ceramic materials, it is necessary to understand the distributions of size and quantity of inhomogeneities present within the material. Ultrasound NDE is well suited for this task as it interacts with all of these features elastically throughout the volume. In this study, a new method has been developed to characterize Sic microstructure by looking at the degree to which each acoustic frequency is attenuated by the material. Frequencies from S to 3SMHz were used to understand how the microstructure affects the propagation of ultrasound at a broad range of frequencies. Acoustic spectroscopy performed on defect engineered Sic samples illustrates how different acoustic responses are measured from ‘good’ and ‘defective’ regions for multiple samples. Comparisons between the attenuation behavior of defect engineered Sic provides evidence that acoustic spectroscopy can quantitatively measure differences in bulk microstructure between similar samples. INTRODUCTION Ceramic materials are very attractive for armor systems because they exhibit many desirable properties, including low weight, high hardness, and high strength’. However, these materials have been known to show a high degree of property variability both within an individual sample and throughout sample sets. Heterogeneities within the bulk material can act to amplify stresses incurred during ballistic events, causing the premature failure of the armor system. Historically, the role of nondestructive evaluation (NDE) has been to locate any large anomalies which may reduce the quality and performance of the material. NDE has also been applied to form predictions of a material’s behavior without any damaging tests’. Recent work by Ramesh et al. has shown that during high strain rate events, defects of all sizes are activated simultaneously, resulting in catastrophic failure’ This contrasts directly with classic static or semi-static tests, in which only the most damaging flaw need be considered. The locating of single anomalous defects is insufficient to provide performance predictions during high strain rate events, as it does not provide data regarding the distribution of stress concentrators throughout the sample volume. New techniques must be developed which address this problem by seeking to quantify the distribution in size and quantity of all heterogeneities present throughout a sample. This information could allow for the creation of extremely accurate modeling based on real experimental data. Acoustic waves are well equipped for this kind of characterization because they interact elastically throughout the entire sample volume. The high frequency ultrasound regime (
159
Microstructural Study of Sintered Sic via High Frequency Ultrasound Spectroscopy
used in colloidal science with a high degree of accuracy to determine particle size distributions in aqueous mediums4. THEORY Attenuation of Ultrasonic Energy The attenuation of ultrasound energy is most simply described by the Beer-Lambert Law:
I = I, emax where I is the intensity of the ultrasound energy, I, is the intensity before propagation in the medium, a is the attenuation coefficient, and x is the path length traveled by the ultrasound energy. For the testing performed in this study, I and I, correspond to the amplitude of each frequency in the first and second bottom surface reflections, illustrated in Figure 1. The path length used in this case was twice the sample thickness, as the sample was examined using a pulse-echo configuration. The attenuation coefficient is comprised of multiple mechanisms which interact additively. The two fundamental mechanisms are absorption and scattering, but these can arise from multiple sources. In a sintered S i c system, the attenuation will come from two sources: the bulk material, which will incorporate internal absorption and grain boundary scattering, and the inhomogeneities, which will either primarily absorb and scatter the ultrasound energy depending on the size of the inhomogeneity and the size of the wavelength. The following equation shows how these four factors are summed up. aT = aInternal Absorption + agrain scattering + %homogeneity Absorption -k ahhomogeneity Scattering (214
Understanding how these four mechanisms work in sintered silicon carbide is critical for determining microstructural features from acoustic spectroscopy.
160
Advances in Ceramic Armor V
Microstructural Study of Sintered SIC via High Frequency Ultrasound Spectroscopy
Top Surface First btforn Surface Peak
500
I
0
I
603
\ Sbcond Bottom Surface Peak
.loo0 12
11
13
14
Tlme (us)
Figure 1: A-Scan of a sintered S i c sample with labeled peaks. Attenuation from the Bulk Material The sintered S i c bulk material will attenuate the ultrasound through frictive absorption and scattering at the grain boundaries. The absorption in this case is due to internal friction as the stress wave propagates, converting ordered particle oscillations to disordered movement, or heat. This mechanism has not been studied extensively for structural ceramics, but it well understood for tissue materials in the medical field’. The following equation describes the absorption of ultrasound in the bulk material: a h t e r n a l Absorption
=
a A k f2
where A is a constant based upon the medium in question, kT is thermal conductivity, f is the frequency of ultrasound, p is density, c is the speed of sound, and C, is specific heat at constant pressure. For sintered SIC, the attenuation due to internal absorption is less than O.ldB/cm at frequencies below 35h4Hz. This mechanism only dominates attenuation in perfect single crystals. Attenuation due to scattering from grain boundaries has been studied extensively in solid materials, but mostly in metal systems such as steel6 Using the long wavelength approximation, the attenuation caused by scattering from the grains is described as: agrain scattering
=A
D~ P
where A is a constant based upon the medium in question, D is the longest dimension of the grain, and f is the frequency of ultrasound. The long wavelength approximation is valid for cases where kD << 1, where k is the wavenumber, and therefore inversely related to the ultrasound wavelength. Research in steel has shown the material constant A to be on the order of 10-29.6For cases in which the ultrasound wavelength is less than one hundred times the grain size, the contribution from grain scattering is extremely small and in many cases negligible. For sintered S i c with a mean grain size of 4-8pm7, this
Advances in Ceramic Armor V
. 161
Microstructural Study of Sintered SIC via High Frequency Ultrasound Spectroscopy
is true for frequencies less than roughly 20MHz. The long wavelength approximation can be used extensively for grain scattering as the acoustic mismatch is relatively minor. Attenuation due to Inhomogeneities Ultrasound energy will be lost whenever the sound wave hits an interface where the acoustic properties of both mediums are not identical. The mechanism responsible as well as the severity of this loss will be determined by the size of the inhomogeneity, its composition, and the size of the ultrasound wavelength. Work in colloidal systems has shown that different mechanisms dominate in different frequency regimes4. Absorption tend to dominate at small ka values (usually caused by low frequencies), while scattering dominates at high ka values (usually caused by high frequencies), where k is the wavenumber (2nii;)and a is the inhomogeneity's effective radius. Specific absorption mechanisms caused by inhomogeneities in a dense elastic solid have not been thoroughly examined by the scientific community. Work in colloidal systems, where the inclusions have high density relative to the bulk medium, shows an inverse parabolic relationship between the attenuation coefficient and frequency4. This is defined approximately as: ahhomogeneity Absorption
=A
f 1 a*
where A is a constant based on the composition and morphology of the inhomogeneity, f is the frequency of ultrasound used, and a is the size of the inhomogeneity. It is likely that low frequency attenuation of ultrasound in structural ceramics will be dominated by this mechanism. Much more attention has been given to attenuation from scattering by inhomogeneities in ceramic systems. The Rayleigh approximation is only appropriate for low ka values, and as such may become inappropriate at higher frequencies. This is because the mismatch in speed of sound for many inclusions within sintered S i c is very severe. An air pore in a silicon carbide matrix will have a speed of sound ratio of 1:35. Because of this, the Mie scattering solution will provide a more accurate analysis, at the expense of requiring more rigorous mathematics. Physical modeling has shown that the cross sectional scattering area is inversely related to the feature radius raised to the sixth power', This study by Tittman et al. concluded that scattering due to inhomogeneities should be orders of magnitude stron er than scattering due to grain boundaries, and should dominate overall attenuation at high frequenciesf The Mie scattering equation for the determination of the cross sectional scattering area follows: aInhomogeneity Scattering
=A
f 1 a6
where A is a constant defined by the composition and morphology of the inhomogeneity. The constant may also be influenced by the exact acoustic impedance difference between the inhomogeneity and the bulk material. EXPERIMENTAL METHOD Twenty lOOmm x lOOmm x 6.5mm defect engineered sintered S i c samples were acquired and tested using ultrasound nondestructive evaluation in order to determine the location and severity of engineered inclusions and defects. Each sample was made to have a distinct type of inclusion which was placed at approximately the center of the sample at half the thickness. A list of the sample numbers and their defect types is displayed in Table I. Each defect was between O.5mm-l.Omm in diameter. Single transducer testing was performed using a 20MHz Olympus transducer in a water filled immersion tank. The transducer used pulses a range of frequencies from roughly 5-4OMHz. Cscans were performed by rastering the transducer over the sample in pulse-echo configuration,
162
Advances in Ceramic Armor V
Microstructural Study of Sintered SIC via High Frequency Ultrasound Spectroscopy
meaning that the same transducer both emitted the ultrasound energy and received the reflections from the sample. Table I: A li
1
ristic defect
103 112 132
150
1 Large non-spherical pits I Large non-spherical pits
I
1 Machining defects Silicon inclusions
Silicon inclusions Pittin
Three different tests were performed on these samples in order to see their acoustic behavior: overall signal attenuation coefficient, the longitudinal acoustic velocity, and the acoustic frequency spectrum. The first of these displays the loss of ultrasound energy in dB/cm caused by interactions with the sample bulk. This is a measure of the total attenuation of ultrasound energy, and is not specific to any one individual frequency. The second property tested was the longitudinal speed of sound in the material. This was achieved by measuring the time of flight (TOF) from the top surface reflection to the first bottom surface reflection, after which the following equation was applied: CL = 2
d / TOF
where d is the sample thickness. The sample thickness was assumed to be constant when creating Cscan images of longitudinal velocity. This assumption introduces a degree of error into the data, but as sample thickness deviations were less than 0.04mm when measured by calipers, the amount of error in velocity calculations is minimal. The third measurement made was the acoustic frequency spectra. For this measurement, a fast Fourier transform (FFT) is applied to the first and bottom surface peaks in order to see how strong they are at each individual frequency. From here, the Beer-Lambert law is applied to measure how each frequency was attenuated. The corresponding graph allows for an analysis of how different frequencies are attenuated by the sample’s microstructure. Point measurements was made using this technique for each sample over the ‘good’ part of the sample and over the ‘defective’ part. In order to more easily see how the frequencies are attenuated through time, a discrete time Fourier transform
Advances in Ceramic Armor V
.
163
Microstructural Study of Sintered Sic via High Frequency Ultrasound Spectroscopy
(DTFT) was applied. The DTFT allows for the A-scan measurement to be directly shown with the frequency strength distribution, enabling a better visualization of how different frequencies are attenuated by the sample bulk. RESULTS AND DISCUSSION For each sample, the following images were constructed for each sample and are presented along with this report: C-scan image of overall sample attenuation coefficient, C-scan image of longitudinal velocity, C-scan image of bottom surface reflection amplitude, A-scan of ‘good’ and ‘defect’ regions, DTFT of A-scans for each region shown at various angles, and the acoustic frequency spectra for both ‘good’ and ‘defect’ regions. Because of space considerations, these images have been included in this file for only one of the samples in order to illustrate the meanings of the graphs and images. This sample was randomly selected as M11, which included a laminar flaw. Comparison of the attenuation spectra of multiple samples will be discussed at the end of this section. The C-scan image of overall signal attenuation coefficient can be seen in Figure 2. The white area seen towards the center of the sample corresponds to the region where the laminar flaw was located. Because of the scale used, the overall signal attenuation here appears to be much lower than it actually is. The attenuation coefficient in the ‘defective’ region is much higher than what is shown in this scale, on the order of 7-8dBicm. A few smaller regions of above average attenuation can also be seen. These are likely indicative of increased concentrations of heterogeneities within the bulk. On average, the overall signal attenuation coefficient was 1.55dB/cm*O. 17dBicm.
Figure 2: Overall signal attenuation coefficient for sample M11 - Laminar flaw Figure 3 shows the longitudinal velocity variations for sample M1 1. Not much variation is seen in this sample, although somewhat higher values are shown towards the center of the sample as compared to the edges. The longitudinal velocity is related to the Young’s modulus, as E is proportional to .:c The dark area seen in the image is where the laminar flaw was located. The speed of sound in this region was markedly slower than for the rest of the sample.. The longitudinal velocity does not seem to be very sensitive to areas around the laminar flaw, as there is no sudden increase or
164
Advances in Ceramic Armor V
Microstructural Study of Sintered SIC via High Frequency Ultrasound Spectroscopy
decrease in velocity values in this portion of the sample. This indicates that changes in elastic properties in areas around the flaw are at most relatively minute.
Figure 3: Longitudinal velocity measurements for sample M11 - Laminar flaw The bottom surface reflection signal amplitude is shown in Figure 4. While not directly relevant to this study, this image is included for completeness. The bottom surface amplitude is dependant on many factors, only one of which is bulk property variations, and as such it is not a reliable source of information regarding sample homogeneity. The bottom surface amplitude image has been included in this report to show how signal strength rapidly decreases in the region around the included defect. For sample M11, the signal dropped fiom approximately 1400mV to approximately 500mV over the ‘defect’ region.
Figure 4: Bottom surface reflected signal amplitude for sample M11 - Laminar flaw. Scale is in mV.
Advances in Ceramic Armor V
.
165
Microstructural Study of Sintered Sic via High Frequency Ultrasound Spectroscopy
Point measurements were taken to compare the attenuating behavior of the ‘good’ area of the sample with that of the ‘defect’ area. The A-scan saved from each region can be seen in Figure 5 . Looking from the ‘good’ A-scan to the ‘defect’ A-scan, two important characteristics are seen. First is that the ‘defect’ region includes additional peaks, corresponding to reflections from the interface between the bulk material and the laminar flaw. Second is that the original sample peaks have been drastically reduced in amplitude. This occurs because a significant portion of the ultrasound energy has been reflected by the flaw, increasing the overall signal attenuation through the sample.
Figure 5 : A-scans of sample M11 -Laminar flaw from a) ‘good’ and b) ‘defect’ regions using a 12dB receiver gain.
In order to characterize the specific frequencies attenuated over the ‘good’ and ‘bad’ regions, a DTFT was used on the A-scans seen in Figure 5 . The DTFT images are shown for both regions at various angles in Figures 6 - 9. Viewing the DTFT image at different angles allows for an easier understanding of how individual frequencies are being attenuated in the sample. It also provides an understanding of which frequencies are reflected by the laminar flaw in the ‘defect’ region. When viewing these DTFT images, it is important to recognize that the frequency distribution of the top surface reflection is not necessarily true. This is because the top surface reflection peak in the A-scan was actually higher than it was possible to display on the oscilloscope, causing the apparent frequency distribution to be different from the true frequency distribution. As the frequency distribution of the top surface peak was not used in sample analysis, this was determined to be inconsequential.
a) Figure 6 : DTFT images of sample M11 - Laminar flaw from a) ‘good’ and b) ‘defect’ regions at a 333” viewing angle.
166
Advances in Ceramic Armor V
Microstructural Study of Sintered Sic via High Frequency Ultrasound Spectroscopy
In order to quantify how the sample attenuates each at each frequency it was necessary to look at the spectra of the first and second bottom surface reflections. By applying the Beer-Lambert law ‘at each frequency a graph of the frequency dependant attenuation coefficient was constructed. This is shown for the ‘good’ and ‘defect’ regions of sample M11 in Figure 10. This graph clearly displays the difference in attenuating behavior for these two regions of the sample. The ‘defect’ region shows a much more erratic attenuation coefficient curve which has generally higher values than the ‘good’ region. It is interesting to note that the ‘defect’ curve displays two local minima at -15MHz and 25MHz and a local maximum at -2OMHz. The ‘defect’ curve shows behavior akin to broad Mie resonances. This is seen because the laminar flaw is on the order of the ultrasound wavelength, and as
Advances in Ceramic Armor V
.
167
Microstructural Study of Sintered Sic via High Frequency Ultrasound Spectroscopy
such the Rayleigh approximation no longer holds. Both the ‘good’ and ‘defect’ curves tend towards zero around 35MHz due to a self-focusing effect seen at this frequency.
-5 a
E * .$ E
rzz
10
+ ‘Good’
-00 0
8 -
0
0
6
-
0 O””0,
A comparison of the attenuation spectra of the ‘good’ points of multiple samples is shown below in Figure 11. The same basic curvature is seen for each sample, but with significant variations indicative of the different microstructures present in each sample bulk. The most significant difference occurs at either end of the frequency range analyzed in this test. In the <20MHz and >30MHz ranges, each sample diverges significantly from the others. This change in behavior indicates the regimes where the specific distribution of heterogeneities present begins to dominate the attenuation spectra. In the frequency range between 20MHz and 30MHz, each sample shows near identical behavior in regard to its attenuation. Techniques for quantizing the different heterogeneity distributions based on the differences seen are being developed. A comparison of the ‘defect’ points for these five samples is shown below in Figure 12. Unlike the spectra for the ‘good’ points, each inclusion type shows significantly different behavior. The attenuative behavior seen is specific to the inclusions’ size, geometry, and composition. The laminar flaw, pitting, and carbonized defect all show resonant behavior to varying degrees. Frequency specific minima can be seen in the low density inclusion spectra as well. The silicon inclusion spectra (M141) shows remarkably similar to the spectra from the ‘good’ spectra seen in Figure 11. This indicates that silicon inclusions attenuate ultrasound energy with less severity than other inclusions. This result is not surprising as silicon’s acoustic properties are more similar to silicon carbide than the other defect types.
168
.
Advances in Ceramic Armor V
Microstructural Study of Sintered SIC via High Frequency Ultrasound Spectroscopy
Advances in Ceramic Armor V
. 169
Microstructural Study of Sintered SIC via High Frequency Ultrasound Spectroscopy
CONCLUSION A method has been developed for analysis of the frequency dependant attenuation of ultrasound in armor-grade silicon carbide. This method is capable of quantizing the contributions of various acoustic loss mechanisms due to both the bulk material and any heterogeneities present. The acoustic spectroscopy methods have been applied to a series of defect engineered sintered S i c samples. Spectra of ‘good’ and ‘defective’ regions indicate measurable differences between different samples and different defect types. Closer analysis of the specific behavior of frequency dependant attenuation by these materials should be able to provide insight into the inhomogeneity distributions and microstructural architecture causally responsible for the acoustic losses measured. REFERENCES 1. M.J. Normandia, J.C. LaSalvia, W.A. Gooch, J.W. McCauley, A.M. Rajendran, AMPTIAC Quarterly, 8(4), pp 21-27 (2004). 2. R. Brennan, Ph.D. Thesis Dissertation “UltrasonicNondestructive Evaluation ofArmor Ceramics,” Rutgers University (2007) 3. B.Paliwal, K.T. Ramesh, Journal of the Mechanics and Physics ofsolids, 56, pp. 896-923, (2008). 4. A. S. Dukhin, P. J. Goetz, Ultrasound for Characterizing Colloids, Elsevier Science B.V. (2002). 5. L. Luo, J. Molnar, H. Ding, X. Lv, G. Spengler, Diagnostic Puthologv, 1:35 (2006) 6. D. W. Nicoletti, N. Bilgutay, B. Onaral, 1990 Ultrasonics Symposium Proceedings, pp. 1119 - 1122 ( 1990). 7. Presentation on sintered S i c - www.grc.nasa.goviWWW/TurbineSeal/papers/2002/Owens.pdf 8. B. R. Tittmann, L. Ahlberg, A. G. Evans, R. K. Elsley and P . T. Khuri- Yakub, Ultrasonics Symposium Proceedings, IEEE Cat. #76 CHI 120-5SU, pp. 653-658 (1976).
170
.
Advances in Ceramic Armor V
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
IMPACT DAMAGE ANALYSIS IN A LEVEL I11 FLEXIBLE BODY ARMOR VEST USING XCT DIAGNOSTICS I
2
Joe Wells , Nevin Rupert , Murray Neal
3
1. Consultant, Mashpee, MA, USA, 774-836-0904,jwells24@jhuedu 2. NLR Technologies LLC, Kittanning, PA, USA. 3. Pinnacle Armor Inc., Fresno, CA, USA. ABSTRACT A comprehensive interrogation and ballistic impact damage analysis was recently conducted on a level I11 flexible body armor vest using non-invasive x-ray computed tomography, XCT, diagnostic techniques. The archival vest was manufactured 1999 by Pinnacle Armor, Inc. and was twice impacted with a level I11 M-80 ball round at the same impact location. The vest incorporates an imbricated network of zirconia toughened alumina, ZTA, ceramic disks 50mm in diameter in stitched, multi-layered ballistic fabric architecture. Full vest XCT scans were conducted sequentially prior to impact, after the first impact, and again after the second impact. Three dimensional solid object renderings of the vest in each impact condition were reconstructed using the Volume Graphics StudioMax 3-D voxel analysis and visualization software package. Results of the damage analysis discussed for this flexible body armor vest include: the morphology and metrology of the localized impact cavity and the limited damage confined to the ceramic disks immediately adjacent to the impact location; the absence of damage in the ceramic disk array at distances removed from the impact cavity; the location, size, and morphology of the major residual projectile fragments residing in front of the ceramic striking face; and the complete lack of perforation of the vest back face. These results are briefly contrasted with a rigid level I11 ceramic plate technology from a similar XCT damage diagnostic study reported previously. The Pinnacle Dragon Skin @ flexible body armor vest, while structurally more complex than the monolithic ceramic and composite backing plates of the rigid interceptor body armor, demonstrated smaller and more localized damage and appears quite amenable to non-invasive XCT interrogation and damage diagnostics.
-
BACKGROUND The non-invasive in situ XCT diagnostics, characterization, and 3-D visualization of ballistic impact damage details in laboratory armor ceramic terminal ballistic targets have been successfully demonstrated over the past decade. Until recently, no known application of this technology had been demonstrated and published on actual ballistically impacted body armor protective vests. Our recent work reported on the successful application of the XCT damage diagnostic modality for the ballistic impact damage characterization in a level-I11 rigid body armor vest [l]. Subsequently it was desired to evaluate the applicability of this technology toward the characterization of impact damage in the considerably more complex architecture of a level-I11 Pinnacle Dragon Skin @3 flexible body armor vest.
171
Impact Damage Analysis in a Level 111 Flexible Body Armor Vest Using XCT Diagnostics
FLEXIBLE BODY ARMOR VEST The subject personal protection vest was provided by Pinnacle Armor, Inc. of Fresno, CA. It was identified as a Dragon Skin@ Level-I11 flexible body armor vest produced circa 1999. The unique architectural design of the vest includes an imbricated array of ceramic disks in 10 horizontal rows with the number of disks per row varying from 3 in the top row to a maximum of 9 disks in the widest row to 6 disks in the bottom row as shown in Table I. These ceramic disks were made of ZTA (zirconia toughened alumina) and measure S0mm in diameter. Each disk is encased within a ballistic fabric sheath and is stitched in between layers of ballistic fabric so as to allow local movement with respect to the adjacent overlapping neighboring disks in the overall array. The disk array is then inserted into an external outer tactical vest. OTV. Table I. Details of the ZTA ceramic disk arrangement in the subject Level-111 vest
BALLISTIC TESTING AND XCT SCANNING APPROACH The subject vest was ballistically impacted at the same location with two sequential impacts by a level-I11 NATO 7.62 x 51-mm M-80 ball projectile at a velocity (VO)ranging from 2,850 to 2,890 fps. Full vest XCT scans were conducted sequentially prior to impact, after the first impact, and again after the second impact. The XCT scans were conducted post-impact with a 450 kV x-ray fan beam orientation perpendicular to the normal projectile direction. It is not presently possible to conduct these XCT scans in real time during impact. Three dimensional solid object renderings of the vest in each impact condition were reconstructed using the Volume Graphics StudioMax 3-D voxel analysis and visualization software package [2]. Subsequently, virtual sectioning, in situ metrology, grey level filtering, virtual transparency, variable magnification and rotational adjustments, and feature segmentation and isolation image processing steps were all applied to distinguish and discriminate both the architectural as well as the damage features of interest. UN-IMPACTED VEST RESULTS The original 3-D solid object images reconstructed from the XCT scans of the un-impacted vest condition are shown in Figures 1 and 2. Note that the Volume Graphics StudioMax@software presents the front or strike face view with the vest top facing upward in these images and the rear or backside view with the vest top facing downward. This orientation protocol is used here throughout this study and makes it relatively easier to distinguish the front versus rear face views. In the reconstructed x-ray images that are shown throughout this report, the lighter the feature, the greater the grey level and hence, the greater is the density of that feature. Thus the darker grey levels represent the various low density ballistic fabrics in Figure 1 and are virtually eliminated by raising the grey level filtering level as shown in Figure 2. It also must be noted that all of the images presented in this report are NOT conventional photographs but rather are high resolution images reconstructed from the x-ray computed tomography, XCT, x-ray scan data.
172
*
Advances in Ceramic Armor V
Impact Damage Analysis in a Level Ill Flexible Body Armor Vest Using XCT Diagnostics
Figure 1. Vest condition - S-0 (un-impacted) at a low grey level, GL, (density), segmentation showing vest fabric outlines and individual disks from both the front (left) and rear (right) normal views.
Figure 2. Vest condition - S-0 (un-impacted) at an increased grey level, GL, (density), filtering showing the virtual absence of vest fabrics revealing only the individual imbricated disks from both the front (left) and rear (right) normal views. SINGLE-IMPACTEDVEST RESULTS The uniformly imbricated ceramic disk arrangement, the higher density projectile fragments, and the surrounding ballistic fabrics with multiple-stitched fabric “bar tack” peripheral segments on the outside vest edges are clearly apparent in the single impacted condition as shown in Figure 3. Of main interest in this figure, however, is the single irregular shaped ballistic cavity just off center in the sixth and seventh disk row from the top of the vest. This is the sole impact damage area observed.
Advances in Ceramic Armor V
.
173
Impact Damage Analysis in a Level 111 Flexible Body Armor Vest Using XCT Diagnostics
Figure 3. Vest condition S=l (single impact) at relatively low grey level, GL, density segmentation showing the sole irregularly shaped impact crater with a substantial sized residual projectile fragment visible from both the front (left) and rear (right) normal views. Close up views of the ballistic cavity are shown from a front normal, front oblique, and then from a completely segmented front oblique view of the fragments alone, respectively, in Figure 4. It is observed in the front normal view that the major residual projectile fragment is actually split into two separate large segments, both residing in fronf of the ceramic disk array, as clearly seen in the front oblique view.
Figure 4. Vest condition-S-1 (single impact) close up views - front normal (left), front oblique (middle), and fragment only segmented front oblique views showing the split main fragment morphology located in front of the front strike face. DOUBLE-IMPACTED VEST RESULTS A comparable view of the vest following the second impact is shown in Figure 5 . Again, only a single ballistic cavity is observed in the same location as in the S-1 condition. No other ceramic disk array damage is observed apart from this location.
174
Advances in Ceramic Armor V
Impact Damage Analysis in a Level 111 Flexible Body Armor Vest Using XCT Diagnostics
Figure 5 . Vest condition-S-2 (double impact) at relatively low grey level, GL, (density) segmentation showing the sole irregularly shaped impact crater slightly enlarged with the substantial sized residual projectile fragment remaining visible from both the front (left) and rear (right) normal views. A close up front oblique view and a front transverse view of this impact cavity location in the S-2 condition is shown in Figure 6 . The location of the large residual projectile fragment(s) is again shown to be in front of the ceramic array strike face as before. Ceramic damage is observed only in those disks immediately adjacent to the ballistic cavity. Initial attempts to obtain higher resolution images of the damaged ceramic fracture faces in situ within the large overall vest reconstructed XCT scanning file were unsuccessful.
Figure 6 . Vest condition C (S-2) - Oblique front side view (left) and frontal transverse virtually sectioned view (right) with ceramic disk fracture surfaces and residual projectile fragments suspended within the ballistic fabric (non-visible) in front of ZTA ceramic disks. BALLISTIC IMPACT CAVITY DETAILS A close up comparison of magnified normal images of the ballistic impact cavity from both the S-1 and the S-2 impacted vest conditions are shown in Figure 7. Two in situ linear measurements made on the arbitrary L and W axes of the impact cavity in each impact condition are indicated in this figure. While close, all images in this figure are not at the identical magnification.
Advances in Ceramic Armor V
.
175
Impact Damage Analysis in a Level Ill Flexible Body Armor Vest Using XCT Diagnostics
Figure 8. Close up view of the ballistic cavity in vest condition S-2 (double impact) It is readily apparent that some enlargement of the cavity occurred with the 2nd impact, particularly, along the W-axis (from 50.1 to 65mm). Also seen in Figure 8 is shown the encircled damaged regions of three of the six ceramic disks immediately surrounding the ballistic cavity as well as the major and minor residual projectile fragments front view (left) and some in situ positional and dimensional metrology from the rear view (right). All ceramic crater damage is found to be quite localized to a few of the disk edges immediately surrounding the crater as indicated by the three circled irregular disk edges in Figure 8. A second substantial, albeit somewhat smaller, residual projectile fragment also becomes observable at the cavity location in the S-2 vest condition. Selective in situ positional and dimensional metrology results also shown in Figure 8 reveal a separation distance between the two fragments as -19.7mm. The approximate dimensions of the larger fragment are shown a -19.5mm by -18.2mm.
176
.
Advances in Ceramic Armor V
Impact Damage Analysis in a Level 111 Flexible Body Armor Vest Using XCT Diagnostics
TRANSVERSE AND LONGITUDINAL SECTIONS IN VEST CONDITION S-2 Longitudinal and transverse virtual sections showing the fmal rest position of the larger residual projectile fragments in front of the ceramic array are shown in Figure 9.
Figure 9. Longitudinal section (above) and transverse (frontal - middle) and longitudinal (sagittal botton) cross-sections of vest condition C at GLseg values of 1207 (left) and 2239 (right). BRIEF COMPARISON OF RESULTS BETWEEN THE FLEXIBLE VS RIGID VESTS A brief summary comparison is made between the XCT damage diagnostics results found in the previous paper [ 1 J on the level I11 interceptor rigid body armor vest with those results found for this Pinnacle flexible body armor vest and is presented in Table II. Both vests were impacted by the identical type NATO 7.62 x 51 mtn M-80 Ball round at velocities ranging from 2,850 to 2890 fps.
Advances in Ceramic Armor V
.
177
Impact Damage Analysis in a Level 111 Flexible Body Armor Vest Using XCT Diagnostics
TABLE I1 CC dF'ARISON OF RESULTS FOR Body Armor/ Interceptor (Rigid)- Level 111 Observations Basic Architecture Curved solid B4C ceramic strike plate adhered to a rigid organic composite backing plate. Three impacts at widely Hit Locations separate locations. Three large individual through Impact Cavity cavities in B4C ceramic strike plate. Less wide cavities behind in the composite backing plate. Size distribution of multiple Projectile projectile fragments captured in Fragments the backing plate material behind the cavity location. Interface Separation Extensive (- 240 degrees) interface separation between the B4C strike plate and composite backing plate. Cracking
Extensive fragmentation plus extended and interconnected radial cracking within B4C strike face about each impact cavity.
1 Perforation Status
Through perforations of the backing plate rear surface were observed at all 3 hit locations.
'HE FLEXIBLE VS RIGID VEST! Pinnacle (Flexible) Level 111
-
Multiple overlapping ZTA ceramic disks embedded within multiple layers of inter-stitched ballistic fabrics. Two impacts at essentially the same location. Single through cavity in the ZTA ceramic diskarray, slightly enlarged with the 2nd impact. One large and a few smaller projectile fragments captured by ballistic fabric layers in front of the cavity location. Intermittent interlaver seoaration of ballistic fabrics enveloping major projectile fragments in front of the ZTA ceramic array cavity. Cracking and fracture damage localized within only 3 damaged ZTA ceramic disks adjacent to cavity. No damage observed in disks away from the impact cavity. No perforations of the rear ballistic fabric face were observed. I
1
SUMMARY & CONCLUSIONS: The conclusions reached from this study are as follows: Only a single we11 delineated impact damage cavity in the regular overlapping ceramic disk architecture is observed, clearly indicating that the second impact in vest condition S-2 was very close to the location of the first impact in vest condition S-1. No other significant damage in the ceramic disk arrangement was observed other than at, or immediately adjacent to, the impact location in vest conditions S-1 and S-2. This common ceramic disk damage cavity consists essentially of fragmented and displaced ceramic with one disk completely obliterated and some partial fragmentation of immediately adjacent ceramic disks. The ceramic cavity in vest condition S-2 is somewhat larger than in condition S-1. A common large projectile fragment is observed in the cavity area following both impacts. This large residual projectile fragment was found to unequivocally reside in ballistic fabric directly in front ofthe cavity in the ceramic array. This terminal fkagment location is quite distinct from the projectile fragment capture location typically observed behind the ceramic striking face of solid rigid armor
178
*
Advances in Ceramic Armor V
Impact Damage Analysis in a Level 111 Flexible Body Armor Vest Using XCT Diagnostics
ceramic systems. Additional smaller irregular fragments of both the ceramic disks and the projectile are observed surrounding the impact cavity on both the front and the rear sides of impacted vest conditions. Several smaller projectile fragments are also located at the bottom of the flexible vest. It is assumed that these latter projectile fragments originally resided near the impact cavity and then fell to the vest bottom during the inevitable handling in between the successive ballistic testing impacts and the XCT scanning stages. In situ metrology provided measurements of the cavity linear and area dimensions, as well as measurements of the local positioning, diameter, surface area and volume of the major residual projectile fragments. Cracking damage and impact-induced fracture surfaces were highly localized as observed only in three of the six ceramic disks immediately adjacent to the ballistic impact cavity. The ceramic disks were f o y d to have a deviation range in grey level values indicating a density gradient and/or a non-homogeneous composition. This GL range is attributed to the lower volume fraction, but higher density, of the zirconia material mixed within the alumina ceramic disk composition. Observations in the oblique view and transverse sections show the larger fragments are positioned in front of the ceramic array and within the highly plied inter-stitched layers of ballistic fabric which remain between all residual projectile fragments and the vest rear surface. No perforation by the projectile fragments was observed on the rear surface in either vest impact condition. While structurally more complex than the monolithic ceramic and composite backing plates of the rigid interceptor body armor, the flexible vest revealed appreciably smaller and more localized damage following two sequential impacts in the same location as compared to that found previously at the location of any one of three widely separated individual impacts in the prior rigid vest examination. The Pinnacle Dragon Skin@ flexible body armor vest appears quite amenable to non-invasive XCT interrogation and damage diagnostics. Most of the diagnostic impact damage details observed and presented in this report would not have been achievable with the traditional 2D digital radiographic NDE modality.
REFERENCES:
1. J.M. Wells and N.L. Rupert, “On the XCT Diagnostics of Ballistic Impact damage in a B4C Ceramic Body Armor Plate,: Proc. 24” International Symposium on ballistics, NDIA, (2008), 430438. 2. http://www.volurnenraphics.corn/
Advances in Ceramic Armor V
.
179
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
Transparent Armor
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
IMPACT ONTO GLASS AND GLASS CERAMIC BARS Stephan Bless, John Tolman, Scott Levinson, and Ian Polyzois Institute for Advanced Technology, The University of Texas at Austin, Austin, TX, USA ABSTRACT Glass and glass ceramic bars were struck with steel projectiles, causing compressive failure. The bars were observed with a high-speed camera, and the rear free-surface velocity history was measured with a photonic Doppler velocimeter. Propagating compressive fractures were observed at the impact ends, and explosive tensile failures were observed at the distal ends. Compressive strength and tensile strength were computed from the free-surface velocity records. There were shot-to-shot variations in the free-surface velocity histories. The compressive strength of the glass was in the range of 1.6 5 0.1 GPa at 104/sstrain rate. The compressive strength of glass ceramic was about 2.6 GPa. The tensile strength in glass bars varies between 0.6 and 1.2 GPa and is greater than glass ceramic; however, in this experiment, tensile failure is probably an extrinsic-as opposed to intrinsic-process, due to surface flaws. INTRODUCTION The strength of materials under one-dimensional (1-D) stress loading is one of the most frequently used parameters for characterizing behavior. There are only a few ways to perform dynamic measurements of this strength, commonly referred to as the yield stress (Y). Most frequently used is the Hopkinson bar. However, use of the Hopkinson bar technique is difficult for brittle materials because failure can occur before stress equilibrium has been obtained or at stress concentrations associated with specimen design. Another problem with Hopkinson bar measurements is the small sample size, which can result in large test-to-test variability. Furthermore, for many materials, there is a very large apparent change in strength between Hopkinson bar results and plate impact results, and questions remain whether this is due to a change in loading rate or a change in test geometry. For these reasons and others, direct bar impact is a useful complement to the Hopkinson bar device. Figure 1 is a generic sketch of the bar impact geometry. A high-speed projectile strikes one end of a bar and creates a stress exceeding the strength of the bar. Variations of the impact geometry include symmetric impact of two bars of the same material and using a Hopkinson bar as a driver coupled to a target bar. The consequence of the high impact stress is that an elastic wave is created in the bar, and the amplitude of the wave is the largest elastic stress that the bar can support, e.g., Y. Initially, the wave propagates in 1-D strain, and the speed is the conventional longitudinal sound velocity, CL = dM/p, where M is the constrained modulus and p is density. However, after a few bar diameters, the deformation transitions to 1-D stress, and the wave speed becomes equal to c~=dE/p, where E is the Young's modulus. Quantitative measurements of strength with the bar impact geometry seem to have begun with Rosenberg et al.'. Following their work, the bar impact has been used to study many ductile materials. In order to assure I-D strain, almost all experimenters use bars of 10 diameters or greater. Bar impact on brittle materials is more difficult but has become relatively common. Instrumentation has included embedded manganin gauges2, strain gages, and VISAR interferometry to measure the velocity of the distal end, which may be free2 or backed with a transparent reference material3. High-speed photography is often employed for failure diagnostics 2, 3, 4, 6
183
ImDact onto Glass and Glass Ceramic Bars
Figure 1 . Basic bar impact geometry. The bar geometry shown in Figure 1 has been used to measure compressive strength of alumina and other polycrystalline opaque ceramics5,6 3 7 3 8 .Reflection of the compressive wave from the fiee end of the bar enables measurement of tensile strength. This has also been applied to alumina9, PMMA’O and concrete”. However, Chhabildas7(1997) and Simha’* have advised caution in the interpretation of bar impact stress because of the competing tensile and compressive failure modes that may exist in the impact region. Chhabildas concluded that better results are obtained with a graded impedance impactor to slow down the pressure rise time at the impact surface. There have been a limited number of studies on transparent brittle materials, including H~rnalite’~, AIONl4, PMMA’O, and fused silicaI5. And there have been several studies of silica glasses. Bless et al. observed that a failure wave phenomena fxst discovered in 1-D strain experimentsI6 also appeared to occur in glass bars6. There have been several subsequent studies of failure waves in glass bars235.All investigators report compressive failure waves whose speed depends on stress but at impact stresses of several GPa, reach a limiting speed equal approximately to 42 times the shear wave speed. It is also reported that the radial expansion associated with the failure wave is heterogeneous, and the extent of jetting associated with compressive failure waves is greater in soda lime glass than in borosilicate glassI5. Examination of high-speed photographs and recovered articles has led to the conclusion that the distal end of the bar had experienced an explosive failure’ qualitatively similar to what is seen in Prince Rupert’s drops’*. Embedded stress gau es recorded levels of 2.5 GPa compressive stress in borosilicate glass and 1.9 GPa in soda lime glass . These levels may represent the compressive strength of the glasses. However, use of embedded gauges in glass is problematic, since Kanel” showed that failure waves are generated at interfaces, so the gauge is always surrounded by failed material, and heterogeneity of deformation on the gauge plane can lead to over estimates of stress. Based on the previous results, we hypothesize that the phenomena occurring in glass bars are as sketched in Figure 2. The impact produces a compressive wave which initially propagates at CL, but after a few diameters transitions to Q. A compressive failure is generated at the impact zone, in which the stress relaxes to near zero. Thus, unlike the case of a ductile material, in which the flow stress is maintained, impact onto a brittle material is expected to produce a stress pulse whose duration is the time to failure. Failure results in a network of cracks that propagate into the bar from the impact end. After failure, the stress relaxes to the unconfined compressive strength of the failed material, which is probably close to zero. We think that this is not quite analogous to the failure waves observed in plate impact, because the failure wave in a bar cannot draw significant energy from the elastic compression of the material into which it propagates. The compressive pulse reflects from the rear surface of the bar. Here there is the possibility of a tensile failure wave that propagates back into the bar.
P
f
184
.
Advances in Ceramic Armor V
ImDact onto Glass and Glass Ceramic Bars
'f
Imp.',
/
'
Figure 2 . x,t diagram for bar impact onto a brittle material. If tensile failure occurs upon reflection of the compressive pulse, then a spall signal will result, as sketched in Figure 3 . The relationship between stress and particle velocity for the waves traveling in the bar is the conventional acoustic equation (1)
0 = PCEU
where u is the particle velocity. Particle velocity doubles at a free surface, which has been verified even for inelastic waves in bars", so the compressive strength of the bar can be computed from
where ufsmaX is the maximum free-surface velocity. If tensile failure occurs, the most tensile stress experienced by the target ot is computed from the conventional spall equation = PAUCEI~
(3)
( ~ t
where Au is defined in Figure 3 .
+ PRESENT EXPERIMENTS The present experiments have sought to expand the bar impact technique and apply it to borosilicate glass and glass ceramic. The impactor was a steel anvil, hardened in ail but the first few tests. The impact end of the bar was either bare or covered with a graded-impedance wafer supplied by Nguyen2'. The borosilicate sample bars were round in cross section, 152.4 mm long, 12.7 mm in diameter. The glass ceramic samples were square in cross section, 150 mm long, and 12.5 mm on a side.
Advances in Ceramic Armor V
. 185
IrnDact onto Glass and Glass Ceramic Bars
The bars were held in place by two slender plastic mounts. Self-shorting pins were placed parallel to and up-range of the target in order to trigger instrumentation. The impact was observed with a Cooke high-speed framing camera. Unlike most previous studies, the sample was front-lit instead of back-lit, so failed material appears bright instead of dark. The velocity of the free end was measured with either the camera or with a photonic Doppler velocimeter (PDV)2'. Two surface conditions were used. In one, a piece of retro-reflective tape was affixed to the surface. This created a very intense signal that was detected without the expense of imaging probes but presented some problems with interpretation. Alternatively, we observed the rough-polished end of the bar directly with a focusing collimator. In a PDV, an incoming, single-mode laser light at wavelength 10 is electronically mixed or heterodyned with the Doppler-shifted reflected light from the experiment, producing an electronic beat signal. The instantaneous frequency of this signal is the minute frequency shift Af= 2vll0 of the reflected light, which is directly proportional to the velocity of the reflector. If multiple reflectors are present, the beat signal is described by multiple frequency components. Thus, short-time Fourier analysis of the beat signal can be used to provide highly resolved andor highly accurate velocity profiles of the reflector(s). The lengths of the time subintervals over which short-time Fourier transforms are conducted are inversely proportional to the smallest frequencies (velocities) that are resolved. Any section of data can be analyzed to provide maximum time resolution or maximum velocity resolution. An example is provided in Figure 4, in which the same data (in this case the initial motion of the distal end of a glass bar) have been analyzed with two different sampling windows.
Figure 4. Signal from shot A1027. Left: high temporal resolution obtained with fast Fourier transforms (FFTs) over At = 82 ns intervals (Av = 9.5 d s ) . Right: High velocity resolution (Av = 0.59 d s ) obtained with FFTs over At = 1.3 ps intervals. The signals from the projectile (seen through the glass bar) and the bar distal end are indicated. MATERIALS Two materials were used in these experiments, borosilicate glass (BSG) and Coming glass ceramic (GC). Densities of these materials are 2.21 and 2.78 g/cm3, respectively. Young's modulus for BSG is 61 GPa. The GC is 25% crystalline spinel, and Coming reports its modulus to be 93 GPa22. EXPERIMENTS CONDUCTED Table I is a compendium of all of the data-producing tests conducted on glass and glass ceramic bars. All velocities were obtained from the PDV record for the tests that have PDV data. The velocities of the tests without PDV records were determined from photography. The use of a gradient density impactor is indicated. For the tests that have PDV data, the use of retroreflective tape is indicated. The
186
Advances in Ceramic Armor V
Impact onto Glass and Glass Ceramic Bars
impact velocity is ui, and the maximum and final free surface velocities are ufsmaXand ufs,respectively. The strain rate is calculated from the acceleration, measured at the half-max point, divided by the free wave velocity.
Shot#
Material
u, (mk)
GDI
PDV
Retroreflective Tape
u.,
Uhmax
1003
BSG
266
247
1004
GC
260
287
245 264
297
232
160.8 255
210
156
31 0
371
307
222
303
236
FAILURE PHENOMENOLOGY The failure process was similar in the glass and glass ceramics. Figure 5 illustrates frames from the three records that were obtained with BSG bars.
Figure 5 . Early, mid, and late time images of fracture of BSG bars. Top shows shot 1008; bottom shows shot 1005.
Advances in Ceramic Armor V
. 187
Impact onto Glass and Glass Ceramic Bars
The impact produces comminuted glass, which spreads laterally on the face of the anvil. The spreading zone is manifested about five diameters into the bar. At early times, the damage shows up as a sharp front. This front moves at about ce, and we think it is due to activation of surface flaws in the glass because there is little to no initial radial expansion. Within the spreading zone, which certainly is pulverized, the expansion is relatively uniform, with little jetting. These observations are contrary to some previous reports of well-developed failure waves in BSG, which are followed by a rather heterogeneous expansion5. The mid-time images show a region of tensile failure, which is presumably due to the tensile reflection of the incoming wave. We interpret the relatively quiescent damage as meaning that the light reflection is due to activation of surface flaws. It appears that there is a central portion of the bar that does not fragment, even at relatively late times. At late times, the distal end of the bar undergoes a radial explosion. Here, the expansion becomes fast and heterogeneous, with numerous jets. We believe this phenomenon is a consequence of a failure that began at surface flaws and propagated toward the axis, resulting in complete comminution of the bar. The free end of the bar expands. The expansion velocity is different than the projectile velocity; thus, the free end must be decoupled from the front that is in contact with the projectile. The behavior of the distal end of the bar probably indicates that this is an extrinsic, not intrinsic, behavior. The spall strength of intact glass is many gigapascalesZ3,and the stresses associated with these waves should not be enough to spall the glass. Further evidence for this hypothesis is the contrasting behavior observed in alumina bars. Alumina is a polycrystalline ceramic in which the tensile strength of the bulk is not different than the surface. In alumina, distinct spall planes develop, and there is little radial expansion at the distal end4324. Figure 6 illustrates damage in shots on glass ceramic bars. Initially, the damage proceeds as in BSG samples. The compressive wave ignites surface flaws. The radial expansion adjacent to the anvil is less violent than in glass targets. However, in the glass ceramic bars, the entire bar shows damage and surface jetting-unlike glass, where the center of the bar resisted damage. The explosive radial failure is similar to the phenomenon in glass. Jetting is more severe in the glass ceramic than in the BSG. Thus, it seems that in most respects the failure phenomenology of the glass ceramic is similar to BSG-e.g., compressive and tensile failure associated with surface flaws. The coarser jetting observed in glass ceramic indicates that that there is a larger flaw size distribution. Failure initiation only takes place at a relatively few number of large flaws.
Figure 6 . Early, mid, and late time images from impacts onto glass ceramic bars. Top, shot 1004. Bottom, shot 1007.
188
.
Advances in Ceramic Armor V
Impact onto Glass and Glass Ceramic Bars
FREE SURFACE VELOCITY DATA Free surface velocity data are available from three sources: high-speed photography analysis and PDV records with and without retroreflective tape. From the data, the parameters (defined in Figure 3) ufnmax, Au, and UF have been determined for each experiment where available. In many of the camera-only experiments, a thin disk was affixed to the bar rear surface to allow us to distinguish qsmaX from UF. Separation of the flyer from the glass was never resolvable, and we believe that the velocity measured from camera records corresponds to UF. The retroreflective tape records may not give reliable Au, both because the spectral peaks are much less crisp than where the surface was viewed directly and because the tape can separate from the glass when the glass slows down. Figures 7 and 8 illustrate the free surface velocity records from two experiments, 1017 and 1027. The impact velocity and bar end surface were different in the two shots. Test 1017 had a smooth, flame-polished free surface and was impacted at 3 10 d s . Test 1027 had a frosted, rough-polished free surface and was impacted at 222 d s . The spectral broadening of the signal during spa11 indicates that the surface was broken up, and many different particle velocities were being recorded at the same time.
Figure 7. Free surface velocity vs. time for test 1017 (BSG with a smooth finish struck at 3 10 d s ) .
Figure 8. Free surface velocity vs. time for test 1027 (BSG with a frosted finish struck at 222 d s ) . The records in Figures 7 and 8 show some similar features. A precursor has traveled at velocity what one would expect from previous publications, this wave is clearly still present ten diameters from the impact site. There follows a wave traveling at CE, which has a rise time of typically 5 ps. This is a more than twice the wave transit time across the bar, so the rise time is longer than
CL. Contrary to
Advances in Ceramic Armor V
.
189
Impact onto Glass and Glass Ceramic Bars
would be predicted by elastic theory. There is a peak stress whose duration is a microsecond or less. The velocity then relaxes; in test 1017, it drops by about 12%, and then falls slowly, until a clear but noisy spall signal occurs. In test 1027, it falls only 3% to a plateau, followed by a spall signal. Whether these variations in peak behavior are due to the polished versus frosted surface, test-to-test variations, or are a consequence of the higher impact velocity in test 1017 is yet to be determined. The spall signal is very sharp in test 1017 (in which the bar end was flame-polished and smooth), but in test 1027 (in which the bar end had a frosted texture), spall occurs over about a microsecond. After the spall signal, the velocity recovers to a constant value, U F . The rebound from the minimum value of velocity associated with spall is normally due to the reflection of a tensile wave generated at the free surface that is then reflected back to the rear surface from the spall crack. Thus, the rebound after spall is an indication that the failure initially occurred away from the free surface. However, the absence of significant reverberation must mean that we are observing a cloud of comminuted material, as opposed to an intact spall plate. The damage that originated some distance away from the free surface quickly percolated to the end of the bar. The bar velocity persists at U F for hundreds of microseconds. This is a time during which the front and rear of the bars are decoupled. The sharp spall signal and reverberations in the record for 1017 suggest that the failure occurred away from the free surface. The slower spall signal and lack of reverberations in the record for 1027 suggest that the failure occurred at the surface and that there was no spall plate. The effect of the bar lateral surface finish has not been experimentally investigated. It has been observed that when glass bars are backed with aluminum, the tensile failure excavates a crater in the backup bar, whereas this effect is absent in polycrystalline brittle material^'^. This is consistent with the notion that at the distal end of the bar failure waves converge to the center axis from the lateral surfaces. It seems clear that the compressive strength can be computed from PCEUfsmaX/2. It is not clear what physical meaning should be associated with U F ,other than pcEziF/2 is a lower bound on strength. Stresses have been computed from the observed free surface velocities, and the trends are plotted in Figure 9 for final velocity U F and in Figure 10 for peak velocity. The abscissa in these graphs is impact velocity. Strain rates, defined as um,,/2c~,were 5.6 to 7.8 x 103/s. The ordinate is the stress computed from the final or the maximum particle velocity. Experiments in which the final free surface velocity was measured with the PDV and camera are in excellent agreement. Thus, a relatively simple imaging device is indeed adequate to measure U F . However, it is clear the U F is typically 20% less than the maximum free surface velocity. Thus, if the strength is measured solely from the final free surface velocity, it will be underestimated by about 20%. The data in Figures 9 and 10 indicate that addition of the graded density impactor to the bar front surface reduced the peak stress in the bar. This is contrary to previous calculations of the effect of a graded density impactor in alumina bars’. We speculate that, in glass, longer duration of impact loading results in failure from surface flaws or stress concentrations at comers before peak stress can be achieved. This is contrary to alumina, where failure is a bulk process. In plate impact experiments, the spall signal Au can be used to compute spall stress. In these tests, it still has the significance that pAuc~i2is probably equal to the highest tensile stress that exists in the bar, but since failure initiates at the surface, this value represents a competition between the longitudinal stress waves that are driving the interior tensile stress and cracks that propagate in from the lateral surfaces. We have computed the tensile stress using what appears to be the dominant velocity in the spall region. For the glass rods, the spall strength increases monotonically with impact velocity, from 0.6 to 1 . 1 GPa. For the glass ceramics, the spall strength is 0.58 GPa.
190
*
Advances in Ceramic Armor V
Impact onto Glass and Glass Ceramic Bars
Figure 9. Stress computed from final bar velocity vs. impact velocity. Symbols indicate which material, presence or not of a graded density impactor, and instrumentation.
Figure 10. Stress computed from maximum bar velocity vs. impact velocity. Symbols are the same as in Figure 9, with the addition that shots with retroreflective tape are so indicated, since the values of peak velocity may be underestimated in those shots. SIMULATIONS Impacts on glass bars were modeled using a program called EMU, created and compiled by Sandia National Laboratories. EMU is a Unix-based mesh-free fmite element program that uses lattice points or nodes held together by bonds. It predicts the deformation and failure of structures and bodies under dynamic loading and failure due to fracture and is the fust code based on the peridynamic theory of solid mechanics. The following specifications were used: a hardened steel anvil 100 mm across with a radius of 29 mm and an initial velocity of 2 19 m/s axially impacting a square cross-sectioned borosilicate glass
Advances in Ceramic Armor V
.
191
ImDact onto Glass and Glass Ceramic Bars
bar 150 mm long and 12.5 mm on the sides. The simulation was run for approximately 100 ps after impact. For brittle materials, EMU uses what it calls the micro elastic linear flat model. Nodes within the material interact purely elastically, and the bonds behave like elastic (linear or nonlinear) springs. Input parameters include the bulk sound speed of the material, the yield strength, and the bond failure stretch or failure strain for that material. The material’s density is a separate parameter. The bond failure model was elastic-brittle in tension and elastic-plastic in compression. The glass model is 1 GPa compressive yield stress, 0.002 maximum tensile bond stress, and bulk sound speed of 3.776 mtn/p~*~. Figure 11 shows the damage that results in the EMU calculation. The impact face of the bar particulates and spreads on the impactor surface. The bar develops a concentrated damage zone at about 85 ps at a distance of about 65 mm from the (original) impact face. After that time the front and back half of the bar appear to be decoupled. The compressive damage of the front half of the bar and subsequent decoupling of the impactor and distal end agree with observation.
Figure 11. Damage levels in the EMU calculation, at 100 ps after impact. The rear portion of the bar fails in discrete spall zones. This is not realistic. The explosive failure of the distal end is not captured because the model does not recognize the surface weakness of the glass. The free surface velocity in the EMU calculation jumps to 1 4 0 d s when the stress wave arrives. When the bar breaks, the velocity jumps to approximately the impact velocity. These features of the calculations do not comport well with the PDV calculation. After the bar breaks, the front damage half is moving slower than the impactor and is being crushed. The jump to impact velocity in the code is apparently due to reflection of the tensile wave generated at the free surface from the surface of the impactor.
CONCLUSIONS The failure phenomena in borosilicate glass and glass ceramic bars are qualitatively and quantitatively similar. The front portion of the bar fails in compression, after which the rear portion of the bar is decoupled from the impactor. The rear portion of the bar fails in tension, apparently from the surface, and this phenomenon results in a radial explosion. The compressive strengths at rates of 104/s are about 1.7 GPa for glass and 2.6 GPa for glass ceramic. Addition of a graded density plate on the bars reduces the measured strength of the bars. A distinct spall signal occurs in the free surface velocity histories, but the meaning of the spall stress does not correspond to bulk tensile strength. The spall signal in glass is greater than in glass ceramic and increases with velocity.
192
*
Advances in Ceramic Armor V
IrnDact onto Glass and Glass Ceramic Bars
ACKNO WLEDGMXNT The research reported in this document was performed in connection with Award Number N00014-06-1-475 from the Office of Naval Research. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the Office of Naval Research. REFERENCES I Z.Rosenberg, M.Mayseless, Y. Partom, The Use of Manganin Stress Transducers in Impulsively Loaded Long Rod Experiments, Trans ASME 202-207, March 1984 * N.H.Murray, N.K.Bourne, J.E.Field, Z.Rosenberg, Symmetrical Taylor Impact of Glass Rods, Shock Compression of Condensed Matter - 1997, ed. Schmidt, 533-536, 1998 A.Cosculluela, J.Cognoux, F.Collombet, Univaxial compression of alumina: structure, microstrucre andstrainrate, J.PhysC3, 109-119, 1991. F. Galvez Diaz-Rubio, J.Rodriguez Perez, V. Shchez Galvez, The spalling of long bars as a reliable method of measuring the dynamic tensile strength of ceramics, 1nt.l J. Impact Eng’ng, 27, 161-177, 2002 W.G.Willmott, D.D.Radford, Taylor impact of glass rods, J. Appl Phys 97,2005 N.S. Brar, S.J. Bless, Z. Rosenberg, “Brittle Failure of Ceramic Rods Under Dynamic Compression, J.Phvsiaue c3.607-612, September 1988 L.C.Chhabildas, M.D.Furnish, G.E.Grady, Impact of Alumina Rods - A Computational and Experimental Study, J. Phys IV, C3-137-143, 1997 H. Simha, “High Rate Loading of a High Purity Ceramic: One Dimensional Stress Experiments and Constitutive Modeling,” Ph.D. Dissertation, The University of Texas at Austin, December 1998 J.Najar, Dynamic tensile fracture phenomena at wave propagation in ceramic bars, J. Phys. C3-647652, 1994 l o N.A.Zlatin, GXPugachev, E.N.Bellendir, E.L.Zil’berbrand, Tensile strength of polymethylmethacrylate rods subjected to lO-’-sUniaxial dilation pulses, Sov Phys. Tech. Phys. 29,469-472, 1984 ‘ I P.Forquin, B.Erzar, J.R.Klepaczko, Experimental Approach and Modeling of the Dynamic Tensile Behavior o f a Micro-Concrete, International Center for Applied Computational Mechanics, Symp on Materials under Extreme Loading, Rocamadour, France, May 28-30,2008 l2 C.H.M. Simha, S.J.Bless, A.Bedford, What is the Peak Stress in the Ceramic Bar Impact Experiment?, Shock Compression of Condensed Matter - 1999, ed. M.D.Fumish, L.C.Chhabildas, R.S.Hixson, Am Institute of Physics, 615-618,2000 R.Russel1, %Bless, T.Beno, Impact Induced Failure Phenomenology in Homalite Bars, Shock Compression of Condensed Matter-200 1, American Institute of Physics, 2002
’
’’
l 4 J.U.Cazamias, S.J.Bless, Bar Impact Tests on Transparent Materials, 18th Int’l Symp. Ballistics, San Antonio, TX, November 1999 D.D.Radford, G.R. Willmott, J.E. Field, J.E., The effect of structure on failure front velocities in lass rods, AIP Conference Proceedings, 2004,no.706, pt.1, pp. 755-8 G.LKane1, S.V.Rasorenov, V.E.Fortov, The Failure Waves and Spallations in Homogenous Brittle Materials, 45 1-454, Shock Compression of Condensed Matter 1991, Elsevier, 1992 l 7 P.Zeinert, S.Bless, T.Beno, Comminuted Particles Originating from Catastrophic Failure of a Glass Bar, Proc 2005 A P S Topical Conference on Shock Compression of Condensed Matter, 2006 I’ S.Chandrasekar, M.M.Chaudri, The explosive disintegration of Prince Rupert’s drops, Phil Mag. B, V0170, NO 6, 1195-1218, 1994 l9 G.Kane1, Failure waves in shock-compressed glasses, Shock Compression of Condensed Matter 2005, American Physical Society, 2006
’‘
Advances in Ceramic Armor V
.
193
ImDact onto Glass and Glass Ceramic Bars
L.P.Martin, J.R.Patterson, D.Orlikowski, J.H.Nguyen, Application of tape-casse graded impedance impactors for light-gas gun experiments, J. Appl. Phys, 102,23507,2007 O.T.Strand, D.R.Goosman, C.Martinez, T.L.Whitworth, W.W.Kuhlow, Compact system for ighs eed velocimetryusing heterodyne techniques, Rev. Sci Intru, 77,No. 6, p83108ff, 2006. 2PJ.Zhang, private communication, 2008 23 %Bless, N.S.Brar, Failure Waves and Their Effects on Penetration Mechanics in Glass and Ceramics, in High Pressure Shock Compression of Solids, Spring Verlag, 2007, 24 T. Beno, S.Bless, S.Nichols, New Phenomena Observed in Plate Impacts onto Alumina Bars, Proc 2005 APS Topical Conference on Shock Compression of Condensed Matter, 2006. 2 5 S.Silling, private communication, 2008. 2o
194
.
Advances in Ceramic Armor V
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
NLTMERICAL STUDY OF THE EFFECT OF SURFACE STRESSES OF TRANSPARENT CERAMICS OF LAMINATED TARGETS FOR MILITARY ARMOR APPLICATIONS Costas G. Fountzoulas, James M. Sands, Gary Gilde, Parimal J. Patel U.S. A m y Research Laboratory, Survivability Materials Branch, 4600 Deer Creek Loop, AMSRDARL-W-MD, APG, MD, USA ABSTRACT With the rapid advancement of the computer power and the recent advances in the numerical techniques and materials model have allowed the accurate simulation of the ballistic impact into muitilayer transparent armor configurations. The development of next generation ceramics is a key to providing the enhanced protection and extended service performance for future U S . military systems. A significant development program is underway to improve processing and reduce costs associated with magnesium aluminate spinel. Pre-stressing of the ceramic materials has been shown to improve its ballistic characteristics. The purpose of this effort is to apply the existing modeling tools in advancing ceramic transparent armor materials to fielded applications. This paper reports on the effect of various compressive stress levels on the depth of penetration (DOP) in the polycarbonate of a spinel/polycarbonate laminate impacted by a projectile. The ballistic behavior of the pre-stressed spinel was studied by ANSYSiAUTODYN commercial software and the results will be discussed in detail. Keywords: armor, modeling, spinel, failure analysis, defects. INTRODUCTION The damage resulting from standard projectile impact on a brittle target has been studied extensively over the past decades and it corresponds to the three stages [ 1-51: Crushing occurs in the high pressure compressive zone in the region of contact, which is associated with the onset of micro-cracking, - the formation of interconnected cracks and fragmentation (loss of ceramic continuity). The dynamic erosion at the interface between the ceramic and the projectile proceeds by a process of high pressure grinding and the mixture of target and impactor particles is accelerated and ejected from the front surface of the target. The size distribution of the ceramic fragments is determined by the network of radial, transverse and spa11 cracks spreading from the contact zone. The constraint of the target is expected to affect mostly these failure processes. The propagation of cracks in brittle solids, such as ceramics, results in their failure. Various defects in the ceramics such as, holes, inclusions, microcracks and surface scratches facilitate the nucleation and propagation of cracks [6].Ashby and Sammis [6]reported that “the difference between compressive and tensile fracture is that in tension a single crack grows unstably (once started, it accelerates across the sample to cause failure) while in compression a population of small cracks extends stably, each growing longer as the stress is raised, until they interact in some cooperative way to give final failure. Because of this, the strength of a brittle solid in compression is usually greater, by a factor of ten or more, than that in tension”. Transparent armor systems using ceramics as the striking face have been explored since the early 1970’s because they potentially provide superior ballistic protection to conventional glass based transparent armor systems [7]. However, commercial manufacturers have not experienced a demand for large ceramic transparent plates for commercial markets other than point-of-sale scanners and
195
Numerical Study of the Effect of Surface Stresses of Transparent Ceramics
fluorescent lighting. Therefore, the development of large (> 645 cm2) and thick (>12.7 cm) ceramic plates for transparent annor applications has not reacted to commercial maturity. The U.S. Army has invested heavily in the development of next generation materials, including ceramics, for military systems [8]. The result of the on-going investments is a critical understanding of ceramics strengths and weaknesses for military platforms. As large transparent ceramic materials are available commercially in sizes up to 900 cm’, progress in ballistic designs has offered substantial increases in performance in transparent armor systems. Among the potential ceramic materials considered for armor - sapphire, edge-form-growth sapphire, magnesium aluminate spinel, aluminium oxynitride - one was selected for the current pursuit, magnesium aluminate spinel (MgA1204). Although individually, single impact testing has shown small variations in ballistic efficiencies ( 4 0 % ) in multi-hit performance, all of the ceramics are effectively equivalent. Technology Assessment and Transfer (Annapolis MD, USA) is providing ceramic spinel plates produced via hot-pressing in sizes up to 28 cm x 36 cm x 1.5 cm for this report [91. Finite element modeling has progressed substantially in the ability to predict failure of materials under extreme dynamic loading conditions. One of the limitations of predictive models is lack of a complete dynamic materials properties database which is needed for each of the materials in the simulations. In order to compensate for parameters whose dynamic values were extrapolated from their static or quasi-static properties, baseline experiments are often used to calibrate the models. (10, 11) However, the recalibration method of modeling lacks many of the physical properties and failure mechanisms associated with real-world materials. Therefore, recalibrated models often lack the ability to predict within statistical error hture failures over any substantial ranges due to the existence of defects, and materials substitutions often lead to new calibration requirements. The desired approach is to validate a fully characterized materials database with one calibration model, and subsequently apply the model to modified designs. However, despite its apparent problems, recalibration of existing materials models has been proven to be an effective tool in the hands of the modeler by minimizing the number of simulation iterations and resulting in more successful predictions. Regardless of methodology, finite element tools can be applied effectively to reduce the variability between impact tests and can be used to parameterize designs with fewer experimental failures, when robust models are created [lo]. MODELING The ballistic behavior of a target, which consisted of spinel and polycarbonate, and impacted by a surrogate projectile, was simulated using the non-linear ANSYWAUTODYN commercial package [ 121. The modeled laminate consisted of panels of spinel and polycarbonate of 232 cm2 cross sectional area. The thickness of the spinel ranged from 3, 6, 9 and 12 mm, and the thickness of the polycarbonate was set at 100 mm. The projectile applied in the models was a 30 mm long, 1095 steel projectile, of conical frustum geometry, (6-mm large base, 1-mm small base), using two dimensional axisymmetric models. Lagrange solver was used for the spinel and polycarbonate, and smooth particle hydrodynamics (SPH) for the steel projectile. The element size used for the 2D Lagrange and the particle size were set to 0.5 mm. Compressive stresses ranging from 10, 20,40, 80, 100 and 200 MPa were applied on the spinel free surface as shown in Figure 1 as a boundary condition. The polycarbonate was simply supported at the comers by applying zero velocity along the x-direction as a boundary condition. Results were obtained by simulating projectile impacting the targets at constant velocity of 878 m/s. The material models used for the polycarbonate and steel projectile were obtained from the AUTODYN material library [12]. The polycarbonates (PC) was modeled using a shock equation of state (EOS), piecewise Johnson-Cook (JC) strength model, and a plastic strain failure criterion. The
196
Advances in Ceramic Armor V
Numerical Study of the Effect of Surface Stresses of Transparent Ceramics
projectile steel was modeled using a shock EOS and a JC strength model (12). The spinel was modeled using a polynomial EOS and Johnson-Holmquist (JH2) strength and failure models (12).
Figure 1. For all loading cases (a-e) the polycarbonate is supported at each corner. P indicates stress applied as a boundary condition. Solid lines on the striking surface of spinel, d and e, are boundary conditions restricting the spinel from moving along the x-direction (simply supported).
Advances in Ceramic Armor V
.
197
Numerical Study of the Effect of Surface Stresses of Transparent Ceramics
RESULTS A comparison base for all loading cases was created by conducting a series of simulations for each of the 3, 6 , 9 and 12-mm thick spinel targets without applying any compressive stress on it. The depth of penetration (DOP) of the projectile in the simply supporting polycarbonate was measured and these results were compared with the results of the analysis of all the loading cases (Figure 2). As it is shown in Figure 2, all data points lie on a second order equation curve. As it is expected, the DOP decreases with increasing spinel thickness.
Depth of Penetration in Polycarbonate (No Compressive Stresses) 10 1
I
Spinel Thickness (mm) I
Figure 2. Depth of penetration of the projectile in the polycarbonate vs. spinel thickness. The solid line represents the best fitted line among the data points, whose equation is shown in the chart. LOADING CASE 1 In this case, compressive stresses were introduced at the back surface of the spinel hard face (exit surface of the projectile) (Figure la). Figure 3 shows the results of this analysis for all spinel thicknesses and magnitude of the introduced compressive stresses. The analysis indicated that the applied compressive stress had nosignificant effect on the DOP into the polycarbonate. The DOP is essentially the same for 9- and 12-mm thick spinel targets. LOADING CASE 2 In this case, compressive stresses were introduced at the front surface of the spinel hard face (entrance surface of the projectile) (Figure Ib). Figure 4 shows the results of this analysis for all spinel thicknesses and magnitude of the introduced compressive stresses. The results of this analysis are similar to the results of Loading Case 1. The introduced compressive stresses in the striking surface has no effect on the DOP into the polycarbonate. LOADING CASE 3 In this case, compressive stresses were introduced at the sides of the spinel hard face (Figure lc). Figure 5 shows the results of this analysis for all spinel thicknesses and magnitude of the introduced compressive stresses. The results of this analysis are similar to the results of Loading Case
198
.
Advances in Ceramic Armor V
Numerical Study of the Effect of Surface Stresses of Transparent Ceramics
1 and 2. The introduced compressive stresses on the entire thickness had no effect on the DOP in the polycarbonate.
LOADING CASE 4 In this case, compressive stresses were introduced at the interface of the spinel hard face with the polycarbonate (entrance surface of the projectile), while the spinel front face was confined along certain parts of the fiont face (Figure Id). Figure 6 shows the results of this analysis for all spinel thicknesses and magnitude of the introduced compressive stresses. The analysis indicated that as the magnitude of the applied compressive stress increased the DOP decreased. For 3-mm and 6-mm thick spinel with 40 MPa compressive stress, the DOP decreased by 18% and 68% respectively. However, for 9-mm thick spinel as the compressive stress increased the DOP decreased to zero Table I. Compressive stress level introduced into the spinel. The numbers in the first column are related to Figures 3-6 (Abscissa) I l l OMPa 1
I
J
Figure 3. Depth of penetration vs. compressive stresses introduced at the back surface of the hard face (Fig.la).
Advances in Ceramic Armor V
.
199
Numerical Study of the Effect of Surface Stresses of Transparent Ceramics
Figure 4. Depth of penetration vs. compressive stresses introduced at the front surface of the hard face (Fig. lb)
Figure 5 . Depth of penetration vs. compressive stresses introduced at the sides of the spinel (Fig lc).
200
.
Advances in Ceramic Armor V
Numerical Study of the Effect of Surface Stresses of Transparent Ceramics
Depth of Penetration in Polycarbonate (Loading 4) 1
1
20 10 16 14 p 1 2 E 10
1 l i
8n °6
‘
4 2
i
I
I
+3-rnm ‘b-6-rnm 1 +Q-mm Ip-12-mm
o
-2 1
2
3
4
5
6
7
I
Compressive Stress (MPa)
Figure 6 . Depth of penetration vs. compressive stresses introduced at the back surface of the hard face, while the front face is simply supported at certain parts (Fig. Id). LOADING CASE 5 In this case, like in Loading Case 4, while compressive stresses also were introduced at the interface of the spinel with the polycarbonate (entrance surface of the projectile) the striking face of spinel was confined along its entire length (Figure le). For all spinel thickness from 3 to 12-mm, the numerical analysis showed no penetration of the projectile into the polycarbonate. EFFECT OF STRESS ON DAMAGE Figures 7 and 8, show the effect of 200 MPa compressive stress for all loading cases on the 9-mm thick spinel target. The damage appeared to be less only for the Loading Case 5, where the spinel was was compressed in a direction parallel to the impact direction. The effect of the compressive stresses on the damage of the spinel was similar to the effect shown on Figures 1 and 8 for all spinel thicknesses.
(a) (b) Figure 7. (a) Damage on 9-mm spinel (a) no stress applied; (b) Loading 1, 200 MPa compressive stresses. The notation for “Spinel”, “Interface” and “Polycarbonate” applies also in Figure 8.
Advances in Ceramic Armor V
.
201
Numerical Study of the Effect of Surface Stresses of Transparent Ceramics
Figure 8. (a) Damage on 9-mm spinel (c) Loading 2, 100 MPa compressive stress; (d) Loading 3,200 MPa compressive stress; (e) Loading 4,200 m a ; (0 no stress applied CONCLUSIONS While the need for advanced materials solutions to protect of vehicles from ballistic threats continues to grow, the ability to predict materials performance using advanced modeling tools increases. The current efforts underway in the U S . Army include the use of ballistic modeling, ballistic testing, and historic knowledge of ballistic design to create structural armors for the transparent armor needs of the U.S. military. In the current paper, the results of an initial study of the effect of compressive stresses introduced on the surface of a spinel on polycarbonate composite were reported. The ballistic behavior of the spinel was studied numerically through 2D simulations by using the commercial software ANSYS-AUTODYN. Compressive stresses were introduced in the striking and in the back surface of spinel, at the edges of the spinel target, and at the interface of spinel with the polycarbonate parallel to the impact direction (Figures la-e). In addition, when a compressive stress
202
.
Advances in Ceramic Armor V
Numerical Study of the Effect of Surface Stresses of Transparent Ceramics
was applied at the interface of spinelipolycarbonate, the impact striking face of the spinel was simply supported, thus placing the spinel in compression in a direction parallel to the impacting impactor. The thickness of the spinel varied from 3, 6, 9 and 12 mm and the compressive stresses varied from 10, 20, 40, 80, 100 and 200 MPa. The analysis showed that the thickness of the spinel was the main parameter for decreased DOP in polycarbonate. As the thickness of the spinel increased the DOP decreased for all cases, with or without compressive stresses. However, it was also shown, that prestressing the spinel with stresses above 10 MPa throughout its thickness contributed to a decrease of the DOP for all spinel thicknesses. The numerical analysis also showed that for spinel thicknesses above 6-mm the compressive stresses had no effect on the DOP in the polycarbonate but only on the extent of the spinel damage. This work is the fust effort of modeling the adhesive stresses between the hard surface and adhesives. At the moment this paper was written they were not any experimental data available to validate our models predictions. Future efforts will use 3D simulations to model these stresses and understand their physics in depth. ACKNOWLEDGEMENT The authors wish to thank the reviewers wholeheartedly for their successful suggestions and corrections which made this paper better and provoked the authors for further investigation. REFERENCES D.G. Brandon, L. Baum and D. Sherman, “Ballistic failure processes in alumina”, Colloque C8, supplement au Journal de Physique 111, Volume 4, septembre 1994 Brandon D.G. “Ceramic Armor”, Encyclopedia of Adv.Cer.Mat. (Pergamon 1991) pp.22 Laible R.C., ‘Ballistic materials and penetration mechanics’ ed. Laib1e.R.C. (Elsevier Sci.Publ.1980) pp. 133 Wilkins M.L., Cline C.F. and Hondol C.A., Tech.Rep. UCRL-50694 (Lawrence Rad.Lab. 1969) Wilkins M.L., ‘Ballistic materials and penetration mechanics’ ed. Laib1e.R.C. (Elsevier Sci.Publ.1980) pp.225 Ashby, M.F. and Sammis, C.G., “The damage mechanics of brittle solids in compression”, Pure Appl. Geophy. Pp. 489-521,133(3) (1990) Gatti,A & Noone, M J, Feasibility Study for Producing Transparent Spinel (MgA1204), AMMRCFebruary 1970. * CR-70-8. 2006 &y Modemization Plan, “Building, Equipping, and Supporting the Modular Force,” Annex D. March2006. http://www.techassess.com/tech/spinel/spineljrop.htm,20 September 2007. lo C.G.Fountzoulas, B.A. Cheeseman, P.G.Dehmer and J.M.Sands, “A Computational Study of Laminate Transparent Armor Impacted by FSP“, Proceedings of 231d Inter, Ballistic Symp., Tarragona, Spain, 14-19 April 2007 ‘I C. G. Fountzoulas, J.C.LaSalvia, B.A.Cheeseman, “Simulation of Ballistic Impact of a Tungsten Carbide Sphere on a Confined Silicon Carbide Target“, Proceedings of 23rdInter, Ballistic Symp., Tarragona, Spain, 14-19 April 2007 ANSYSiAUTODYN Vol 11.0, Manual, Century Dynamics Inc., Concord, CA
’
’
Advances in Ceramic Armor V
. 203
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
ANALYSES OF VARIOUS DAMAGE MECHANISMS IN TRANSPARENT ARMOR SUBJECT TO PROJECTILE IMPACT Kevin C. Lai and Xin Sun Pacific Northwest National Laboratory K6-08 PO BOX 999 Richland, WA 99352 Douglas W. Templeton USARMY RDECOM-TARDEC Warren, Michigan 48397 ABSTRACT Design and manufacturing of transparent armor have been historically carried out using experimental approaches. In this study, we use advanced computational modeling tools to study the various stress components during the impact event and to identify the different cracvdamage driving mechanisms for the different cracking patterns. Experimentally observed damage patterns for a thick glass laminate under fragmentation simulation projectile (FSP) impact are used to compare with the modeling results. AHPCRC developed modeling software EPIC’06 [ l ] is used in predicting the penetration resistance of transparent armor systems. It is found that a 1-parameter single state model can be used to predict the impact penetration depth with relatively good accuracy. In addition, regions with high radial stress and high hoop stress are found to correlate well with locations of ring cracks and radial cracks respectively. It is also found that a region of high radial stress and circumferential stress is produced by the impact at the back side of the target plate, possibly causing tensile damage leading to higher penetration depth for thinner laminates or higher velocity impact. INTRODUCTION The response of brittle solids to the impact loading of different projectiles has been a subject of considerable interest because of its significance to strength degradation and structural integrity of this important class of materials. It becomes particularly important in designing transparent armor materials because of the conflicting goals of providing ballistic protection and maintaining good transparency for visibility and survivability. The performance requirements for transparent armor are: 1.) Single and multi-shot protection against various projectiles; 2.) Transparency and visibility for vehicle/structure occupants. Various kinds of glasses are attracting more and more attention as transparent armors against ballistic threats for their unique combination of properties such as outstanding optical transparency, light weight, and high failure strength. Glass strength, as most other materials in the ceramics family, is known for its dependency on surface flaws. In order to examine penetration resistance of various glasses under ballistic impact condition, many researchers have conducted controlled impact experiments. For example, Behner et al. [2] measured failure front dynamics in borosilicate glass under gold rod impact and penetration. The dynamic failure wave front and rod penetration depth are measured simultaneously using high speed photography and flash X-rays. Nie et al. [3] reported the dynamic strength of borosilicate glass under combined compression and shear loading using Split-Hopkinson Pressure Bar (SHPB) technique. More recently, Bless and Chen [4] reported the experimental ballistic data for layered soda lime glass under the impact loading of a fragmentation simulated projectile (FSP). Even though
205
Analyses of Various Damage Mechanisms in Transparent Armor Subject to Projectile Impact
penetration is limited in the top two glass layers, various damage patterns extend throughout every plate, severely degrading the transparency of the target. On the modeling front, with more experimental data on penetration depth becoming available, many activities are focused on finding the appropriate material constitutive model and the associated material parameters in predicting the penetration depth for transparent armor subject to various projectile impacts. Under higher strain rate and higher Confinement pressure, Johnson and Holmquist [S] examined the applicability of various constitutive models in EPIC’06 for the gold rod penetration experiments carried out by Behner et al. [2]. It was found that a 1-parameter single state model can be used to predict the penetration velocity as well as the hole profile and penetration depth with reasonably good accuracy. Sun et al. [6] examined the applicability of the simple 1-paremeter single state model in EPIC’06 [ l ] in modeling the FSP penetration experiment reported by Bless and Chen [4] on layered soda lime transparent annor systems. It was found that with the simple, constant flow stress model, EPIC’06 can predict the penetration depth of the layered glass armor with relatively good accuracy. This is true for both the gold rod penetration prediction on monolithic borosilicate [S] and the FSP impact simulation on laminated soda lime glass target. The authors then suggested that the comminuted glass particles at the tip of the projectile flow like a viscous fluid under high confinement pressures, and that the strength of the fractured glass follows the material behavior of a perfectly plastic flow without any strain hardening, regardless of glass type. Even though the single state model in EPIC’06 can predict the depth of penetration with relatively good accuracy, it does not predict the multiple cracking mechanisms as shown and discussed by Bless and Chen [4]. The damage away from the projectile impact zone is due to the stress waves generated by the impact shock, and this is the fbndamental difference between the impact behaviors of brittle armor and ductile armor: Ductile armor has a single penetration point surrounded by material that undergoes considerable amount of plastic deformation; For brittle armor, in addition to the penetration crater and the finely comminuted zone immediately adjacent to it, there are large regions of damagedfractured zones in the target with various crack morphologies. These different, discrete damaged zones will influence the glass transparency as well as the multi-hit capability of the brittle armor systems in different degrees. Therefore a comprehensive understanding of the driving forces for the different cracking mechanisms in the transparent annor system has to be established. In this study, we attempt at correlating the various cracking mechanisms for the transparent armor plate reported by Bless and Chen [4] to the various stress components predicted during the FSP impact. The single state 1-parameter model in EPIC’06 is used to predict the evolution of various stress components during the impact event. In addition to depth of penetration, ring cracks, radial cracks, needle cracks are correlated with principal tensile stresses in different directions. It is shown that the surface flaw driven tensile failure of glass has a much lower critical strength level in comparison to the failure strength under high confinement, i.e., 570h4Pa. The results of this study can be used in developing a more comprehensive constitutive model and the associated material parameters in modeling the overall damage and cracking behaviors for transparent armor under projectile impact. GLASS CONSTITUTIVE MODEL VALIDATION WITH FSP SINGLE SHOT EXPERIMENT Figure 1 shows the FSP penetration experiment results reported by Bless and Chen [4] on layered soda lime glass. The target used in the experiment was made of seven layers of 30cm x 30cm soda lime glass of 7.6cm total thickness and two layers of polycarbonate of 12.7mm total thickness. The individual layers were assembled with polyurethane adhesive. A standard 12.7(SO-caliber)
206
Advances in Ceramic Armor V
Analyses of Various Damage Mechanisms in Transparent Armor Subject to Projectile Impact
fragment simulating projectile was shot into the glass layers at 11 I l d s . The projectiles weigh 13.4g and are made of HRC30 steel [4]. The projectile was stopped in the second glass layer, with the total depth of penetration of 18.9mm. Figure 1 is a replot from Ref. 4 showing the through thickness different crack morphologies observed for the entire target. The detailed definitions of the various crack morphologies listed in Figure 4 are explained in details in Ref. 4 and will not be covered here due to the brevity of this paper.
Figure 1. Damage map of complete target for the single shot FSP experiment (re-plotted from Ref. 4) Figure 2 is a replot from Ref. 4 showing the various damage regions on the strike face. Starting from the inner circle, the different damagekrack regions can be characterized as: the impact crater, the bundled radial cracks, the outward fan cracks, the coarse radial cracks, the bow-tie region, dicing cracks, and the inward fan cracks [4].
Figure 2. Damageicrack patterns on the strike face (Replot from Ref. 4)
Advances in Ceramic Armor V
. 207
Analyses of Various Damage Mechanisms in Transparent Armor Subject to Projectile Impact
MODELING OF THE IMPACT EVENT In EPIC’06 calculations, the following physical properties for soda lime glass are used: density = 2530kgim3, Young’s modulus = 72GPa, Poisson’s ratio = 0.25. The single state 1-parameter model is used and the constant strength for the soda lime glass is assumed to be 570MPa. Note that the strength definition here is based on equivalent stress, and that the single state model does not differentiate damage state and intact state. The constitutive relation used here is essentially assuming that the strength of glass under high rate impact follows the material behavior of an elastic-perfectly plastic flow without any strain hardening. The polyurethane adhesive is modeled with EPIC library material #142 for polyvinyl butyral (PVB) phenolic resin [ l ] with the Johnson-Cook model [7]:
o =[CI + ~ 2c N,][I + c3,1n.i0][1- T * ~ ]~ +4P . (4) where E is the equivalent plastic strain, 6’ is the dimensionless total strain rate, T* is the homologous temperature, and P is the hydrostatic pressure (compression is positive). The default constants for material #I42 in EPIC’06 are: C1= 75.8MPa, C2=69MPa, N=1.0, C3=0, M=1.85 and C4=0.
Figure 3(a) shows the predicted target geometry at the end of the impact event with the depth of penetration to be 18.97mm, which is very close to the experimental measurement. Figure 3(b) shows the kinetic energy evolution during the impact event for the projectile and the target: It can be seen that the penetration event is finished at around 40microsecond. Since infinite boundary (nonreflecting) condition is considered at the edge of the target, interaction between the outgoing and rebounding stresses can not be modeled in EPIC’06. These stress wave interactions are potentially responsible for the formation of inward fan cracks and the bow tie cracks reported by Bless and Chen [4]. Next, we examine the evolution of various stress components during the FSP impact event to correlate them with the various damagehacking patterns.
E
-
s
. - t
(.,
.**: ~
“ A
*u
u
XUI
(1 I.
>
7
/.
(a) (b) Figure 3. Predicted penetration depth and kinetic energy evolution at 1OOmicrosecond Evolution of Pressure Figure 4 shows the evolution of hydrostatic pressure during the impact event. In addition to the outward expanding of the compressive wave in every layer, it is also clear that there is a column of material right underneath the projectile impact crater that experiences high contact pressure. This
208
.
Advances in Ceramic Armor V
Analyses of Various Damage Mechanisms in Transparent Armor Subject to Projectile Impact
column of high contact pressure is labeled as ‘compress disk’ by Bless and Chen [4]. The compact disk is characterized as damaged but hlly consolidated glass separated by regions where the glass was heavily micro-fissured. These results indicate that the depth of penetration should be calculated with the damaged strength of glass rather than the intact strength of glass.
Figure 4. Predicted pressure contour-during impact Ring Crack and Radial Crack - ox and cr: Figure 5 shows the evolution of or and uzduring the entire impact event with the dimension of the compressed disk predicted in Figure 4 also illustrated. Note that the maximum contour value of the tensile stress plots is lOOMPa, which is much lower than the 570MPa glass strength under hydrostatic compression. This value is based on the measured bi-axial tensile strength of soda lime glass under ring-on-ring bending tests [8]. High radial stress ox in the x-direction leads to a circumferential crack that can be characterized as a ring crack or a ripple crack [9]. Results in Figure 5(a) show that most ripple cracks are located right outside the compressed disk. This is consistent with the experimental observations reported by Bless and Chen [4]. The predicted width of the ripple crack zone, which is roughly the same as the compressed disk, is similar to the experimental measurement shown in Figure 1. Several discrete ring cracks are also predicted toward the edge of the target.
Advances in Ceramic Armor V
.
209
Analyses of Various Damage Mechanisms in Transparent Armor Subject to Projectile Impact
Figure 5. Evolution of radial stress
6,and hoop
stress o2during the entire impact event
Since the kinetic energy of the projectile reaches its asymptotic value at 40microsecond, see Figure 3(b), the impact event can be considered over at that time. Consequently, the contours for ox and o: both reach their most severe states at t=50microsecond as shown in Figure 5. It is interesting to note that, underneath the compressed disk, the last two layers of glass plates are also predicted to experience a combination of high ox and c z .This indicates a possible tensile damage region in the last two glass layers upon impact. This is potentially the floret region observed by Bless and Chen [4], which is characterized by very small asterisk-like features located in the center region of the back of the target. Should the projectile have more kinetic energy to reach deeper into the target, this predamaged glass with lower confinement pressure will potentially lead to lower glass strength, which can yield higher penetration depth than predicted values since no damage is considered in the single parameter model here. Predicted contours of oz in Figure 5(b) indicate that there is a region in the second and third layers of the target that experiences very high hoop stress. This area corresponds well to the experimentally observed area with needle cracks and radial cracks. Furthermore, the predicted hoop
210
Advances in Ceramic Armor V
Analyses of Various Damage Mechanisms in Transparent Armor Subject to Projectile Impact
stress evolution pattern in Figure 5(b) for the loading cycle does not suggest a high radial crack region as reported by Chen and Bless [4]. This is consistent with the observations [lo] and predictions [9] for a semi-infiiite glass block under static indentation in which radial cracks only occur during the unloading cycles. DISCUSSIONS AND CONCLUSIONS In this study, we attempt to correlate the different glass crackingdamage mechanisms with the various stress components experienced by the glass target during an FSP impact event. With the simple, constant flow stress model, EPIC'06 can predict the penetration depth of the layered glass armor with relatively good accuracy. However, the various damage morphologies cracking patterns can not be directly predicted by the single parameter model. By studying the stress evolution in the glass target, we are able to identify the stress components responsible for the various cracking mechanisms: ox for ripple cracks, 0: for needle crack and high ux and oz at the back plate for floret cracks. We also validated that the strength of glass under bi-axial tension is much lower than its strength under hydrostatic compression. This is consistent with the surface flaw-driven nature of the glass strength under tensile loading. Detailed examination of the modeling results also explains why the single-state 1-parameter model can well predict the depth of penetration for very thick targets and for impact events with relatively low impact energy. For thinner laminated targets or for higher velocity impact, tensile stress induced damage at the back side of the target can cause pre-damage of the target prior to the arrival of the projectile head, leading to higher penetration depth than the single state model can predict. Our next steps include incorporating the various cracking mechanisms and the associated material parameters into a constitutive model to accurately predict the penetration depth as well as the various damage patterns for transparent armor systems. ACKNOWLEDGMENTS Pacific Northwest National Laboratory is operated by Battelle for the U.S. Department of Energy under contract DE-AC06-76RL01830. Technical discussions with Mr. Tim Holmquist of Southwest Research Institute are also gratefully acknowledged. FOOTNOTES *Corresponding author, phone: 509 372 6489, fax 509 375 2604, email: [email protected] REFERENCES 'G.R. Johnson, S.R. Beissel, C.A. Gerlach, R.A. Stryk, T.J. Holmquist, A.A., Johnson, S.E. Ray and J.J. Arata, User Instructions for the 2006 Version of the EPIC Code, 2006. 'T. Behner and V. Hohler, Failure Kinetics of DEDF and Borosilicate Glass against Rod Impact, Presented at Progress Meeting EMI-SwRI-IRA-RDECOM at SwRI, San Antonio, 05-06, Oct. 2005. 'X. Nie, W. Chen, X. Sun and D.W. Templeton, Dynamic Failure of Borosilicate glass under CompressiodShear Loading: Experiments, Accepted for publication by Journal of American Ceramics Society, 2007. 4S. Bless and T. Chen, Impact Damage in Layered Transparent Armor, to appear International Journal of Fracture, 2009. 5G.R. Johnson and T.J. Holmquist, Determination of Simple Constitutive Models for Borosilicate Glass using Penetration-Velocity Data from Ballistic Experiments, Presented at the 15th APS Topical Conference on Shock Compression of Condensed Matter, June 24-29,2007, Hawaii.
Advances in Ceramic Armor V
.
21 1
Analyses of Various Damage Mechanisms in Transparent Armor Subject to Projectile Impact
6X. Sun, K.C. Lai, T. Gorsich and D.W. Templeton, Optimizing Transparent Armor Design Subject to Projectile Impact Conditions, Proceedings of the 32"d International Conference on Advanced Ceramics and Composites, Daytona Beach, 2008. 'G.R. Johnson and W.H. Cook, A Constitutive Model and Data for Metals Subjected to Large Strains, High Strain Rates, and High Temperatures. In Seventh International Symposium on Ballistics, The Hague, The Netherlands, April 1983. *S.T. Gulati, J.D. Helfinstine, T.A. Roe, M.A. Khaleel, R.W. Davies and J.L. Woods. Biaxial Strength of Automotive Windshields, Technical Report for CRADA "Structural Reliability of Lightweight Glazing Alternatives CRADA Project", 1999. 9X. Sun and M.A. Khaleel, Modeling of Glass Fracture Damage Using Continuum Damage Mechanics- Static Spherical Indentation, International Journal of Damage Mechanics, 2004, 13[3] 263-285.
''C.G. Knight, M.V. Swain and M.M. Chaudhri. Impact of Small Steel Spheres on Glass Surfaces, J. Matls. Sci., 1977, 12 1573-1586.
212
Advances in Ceramic Armor V
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
PRESSURELESS REACTION SINTERING OF AlON USING ALUMINUM ORTHOPHOSPHATE AS A TRANSIENT‘LIQUID PHASE Michael Bakas, Henry Chu Idaho National Laboratory Idaho Falls, ID USA ABSTRACT Use of aluminum oxynitride (AION) in transparent armor systems has been difficult due to the expense and limitations of the processing methods currently necessary to achieve transparency. Development of a pressureless processing method based on direct reaction sintering of alumina and aluminum nitride powders would reduce costs and provide a more flexible and practical manufacturing method. It may be possible to develop such a processing method using liquid phase sintering; as long as the liquid phase does not remain in the final sample. Alp04 forms a liquid phase with A1203 and AIN at the temperatures required to sinter AION, and slowly decomposes into P205 and alumina. TI-erefore, it was investigated as a possible transient liquid phase for reaction-sintered AION. Small compacts of alumina and aluminum nitride with up to of 15.0 wt% Alp04 additive were pressed and sintered. It was found that Alp04 formed the requisite transient liquid phase, and it was possible to adjust the process to produce AlON samples with good transmission and densities of 3.66-3.67 g/cc. XRD confvmed the samples formed were AION, with no trace of any remaining phosphate phases or excess alumina or aluminum nitride. Based on the results, it was concluded that Alp04 could be utilized as a transient liquid phase to improve the density and optical transmission of AlON produced by pressureless reaction sintering. INTRODUCTION Aluminum oxynitride (AION) is a transparent polycrystalline material. Reviews of the history of the material and its development have been presented by Corbin (1989)l and McCauley et a1 (2009)*. The most complete A1203/AIN phase diagram was published by McCauley and C ~ r b i n . ~ Interest in AlON is driven by its potential for use in armor and infixed detector applications. Currently the cost of AlON is high due to the hot pressing or HIPing necessary for a pore-free microstructure and the expense of pre-synthesized AlON powder. The goal of this research was to develop a method for producing transparent AlON at reduced cost by eliminating the need for pressure-assisted densification and utilizing cheaper starting powders. An alternative to using AlON powder is reaction sintering of A1203 and AlN.1,2,4Reaction sintering was used by McCauley and Corbin to produce translucent AlON in their initial work4 forming y-AION as a dense polycrystalline material. When fired in a pressureless process, AlON has intragranular porosity that is difficult to r e m ~ v e . ” ~ A liquid phase sintering process could possibly eliminate this porosity prior to it becoming intragranular. It is necessary for any liquid phase sintering of AlON to be transient as any remaining liquid phase would impair optical transmission. The sensitivity of transient liquid phase processes to processing conditions makes them difficult and time consuming to develop.’ Not only must the correct additive be identified, but numerous processing variables such as heating rate, maximum temperature, and hold times must be defined if porosity or swelling is to be a v ~ i d e d . ” ~ Previous research has discovered that certain additives can be beneficial to the sintering of AION. Hartnett, Gentilemen, and Macquire have reported that additives such as yttrium and lanthanum improve the sintering and microstructure of AION, and subsequent researchers have also reported on the beneficial nature of these additive^.^^^"^ However, these additives can only be added in limited amounts; otherwise their solubility in AlON is exceeded and second phases form.5310 An additive that
213
Pressureless Reaction Sintering of AlON Using Aluminum Orthophosphate
can be added in amounts capable of eliminating nearly all porosity is needed if transparency is to be achieved. LiAl50~was investigated as such an additive by Cutler, Clay, Poslusny, Flinders, and Jacobs.” In this work, Alp04 was investigated as an additive that possibly meets this criteria. Alp04 was chosen as a possible solution because the AIP04/A120j phase diagram’‘ indicates the formation of a liquid hase at -185OOC. At temperatures greater than 16OO0C, Alp04 decomposes into P205 gas and AI;Oj!As the A1203 will react to form AION, it should be possible to add as much Alp04 as desired without the formation of second phases. EXPERIMENTAL PROCEDURE The alumina powder used alternated between Ceralox SPA-0.5 and Alcoa calcined alumina. The aluminum nitride powder was high purity Tokuyama powder grade E, and the aluminum orthophosphate was produced by Acros Organics. The alumina and aluminum nitride were batched according to the desired AIN mol% relative to the alumina. Subsequent discussions that reference the AIN mol% of a sample or settling powder refer to this batched value, not the A N mol% including other additives or the final AIN mol% of the sample. The Alp04 was added as a wt% to the batched alumina and aluminum nitride mixture, with the amount varying from 0.5 to 20.0 wt%. Batches of 20-35g of the desired mixture were dispersed in ethanol with polyvinyl pyrrolidone (PVP) added as a dispersant and Rhoplex B60A@ as a binder. Mixing was performed with a high speed shear mixer for a minimum of ten minutes at 6000 rpm. The slip was subsequently poured into a drying tray and dried at 4 5 T for 24 hours. After drying, the powder cake was removed from the tray and ground with an alumina pestle and mortar. This mixture was then sieved through a 100 mesh and measured out into -5g samples that were pressed with a ram pressure of 10,000 Ibs/in2. Samples were subsequently CIPed at a pressure of 60 ksi. After CIPing, samples were fired in air at 6OOOC for 15 hours to remove any binder and organics. Samples were stored in a desiccation chamber that was first put under vacuum and subsequently flooded with nitrogen to minimize any reaction of AIN with moisture. Samples were fired in BN crucibles machined from Saint Gobain AX05 BN. The furnace used was an Astro Industries lOOOA series furnace with carbon elements. The atmospPere was dry nitrogen at a pressure of 2 psig. Temperature was controlled with an optical pyrometer. As carbon has been established to be detrimental to A10N,I4 samples were packed in a settling powder mixture of alumina, aluminum nitride, and occasionally other additives, such as boron nitride or yttria, for protection. The settling powder mixtures were mixed dry for 1 hour in a Turbla mixer in a Nalgene bottle. Prior to firing, a small bed of settling powder was laid out on the bottom of the crucible. The sample was placed on this powder bed and covered with additional settling powder. After firing, sample densities were measured using Archimedes’s method with water with a Mettler Toledo AG204 DeltaRange@ scale with an error of *0.0004g. A Bruker-AXS D8Advance Xray diffraction system was used to identify the primary phases present. For microscopy, samples were polished to a 1 micron finish and etched with hot phosphoric acid for ten minutes. Optical microscopy was performed using a Keyence DHX digital microscope, and scanning electron microscopy was performed with a Phillips XL 300 ESEM with an attached EDAX EDS system for compositional analysis. When trace compositional analysis was needed, samples were analyzed with a JEOL 8900 electron microprobe. RESULTS First Series of Experiments The goal of the first series of experiments was to determine if the Alp04 could be used for reaction sintering of AlON without any phosphate phases remaining in the final sample. Samples fired with small amounts (0.5 wt% to 1.0 wt%) of Alp04 do not have phosphate phases measurable with XRD. Figure 1 is an XRD of a sample batched to 35.0 mol% AIN with 1.0 wt% Alp04 that was fired
214
.
Advances in Ceramic Armor V
Pressureless Reaction Sintering of AlON Using Aluminum Orthophosphate
100
10
M
60
b!
2-Theta -Scale
a t 8 1 1 0 w% AIP04 - WL1 1 5406 - General Operations Back (in Sqrt) 0 977 1 000 I Im
* A1506N - Cubic 00-048-0686(*) AIN - Hexagonal - 00-025-1133 (*) ~
Figure 1 . XRD of sample batched to 35.0 mol% AIN with 1.0 wt% Alp04 added, fired at 195OOC for 18 hours. for 18 hours at 195OOC. It indicates that the primary phase present is y-AION, with some AIN peaks from incomplete reaction. The appearance of the samples is white with limited optical translucency. Having demonstrated that Alp04 could be added without adverse effect on the reaction sintering of AION, the amount of Alp04 was increased in subsequent experiments. Sample appearance becomes more translucent with increasing AIP04 content. These early samples were fired in a settling powder mixture of A1203 and AIN batched to 35.0 mol% AM. A problem occurs with the settling powder as Alp04 content increases. While the settling powder is easily brushed off samples with little or no AlP04, additions of 10.0 wt% of Alp04 cause the settling powder to bond to the surface of the samples in such a manner that mechanical grinding is necessary to remove it. In an attempt to prevent this bonding, 10.0 wt% of BN was added to the settling powder. BN powder did not prevent the settling powder from bonding to the sample. A subsequent XRD found non-AION phases in the sample. Even accounting for the A1203 added by the decomposition of the phosphate, the composition of these samples should have been within the single phase AlON region of the A1203/AIN phase diagram. A subsequent sample filed without any settling powder had an extremely low density. Based on these results, it was suspected that the bonding of the settling powder to the sample was indicative of a more significant interaction than first suspected.
Advances in Ceramic Armor V
.
215
Pressureless Reaction Sintering of AION Using Aluminum Orthophosphate
Figure 2. XRD of 28.0 mol% AIN sample with 15.0 wt% Alp04 fired at 1950°C for 18 hours in a 28.0 mol% AIN settling powder with 0.5 wt% Y203. Second Series of Experiments The goal of the second series of experiments was to further investigate the interaction between AIP04 containing samples and the settling powder and determine if this interaction could be of a beneficial nature. The results of this series indicate that during the course of the interaction, material is introduced from the settling powder into the sample. This additional material can shift the AIN mol% of the sample from its nominal batched value and introduce additives from the settling powder into the sample. The phases observed in the final sample depend on the initial compositions of sample and settling powder and the severity of the interaction. Results indicate that settling powder composition, temperature, and sample composition all influence the degree of interaction. Different phases observed in various samples included the desired y-AION, a phase tentatively identified as cp’ phase AION,” and corundum. Figures 2 and 3 show some example XRD patterns of samples created in this series. Temperatures of greater than 1900°C are necessary for any significant densification to occur. Increasing the temperature seemed to increase the degree of interaction, causing the compositional shift to be of greater magnitude and sample density to increase. Increases in the Alp04 content up to 15.0 wt% assist the interaction, while amounts greater than 15.0 wt% appear to be detrimental to densification. Y203 additions in the settling powder, even in small amounts (0.3-0.5 wt”/.), greatly increase the degree of interaction. SEM performed on samples fired in settling powders with Y203 additions found yttrium-rich phases on the grain boundaries.
216
. Advances in Ceramic Armor V
Pressureless Reaction Sintering of AlON Using Aluminum Orthophosphate
Figure 3. XRD of 32.0 mol% AIN sample with 15.0 wt% Alp04 fired at 1950OC for 18 hours in 28.0 mol% AIN settling powder with 0.5 wt% Y203. Samples containing primary phases of y-AION and cp’ phase AlON appeared to have good optical translucency compared to other samples produced in the second series. Microscopy confirmed the presence of a second phase on the grain boundaries, and EDS indicated the composition of these phases was similar to that published for cp’ phase A10N.3 These samples can be produced reliably using a settling powder mixture of 20.0 mol% AIN, and 0.5 wt% Y20,. These samples of y-AION and cp’ phase AlON have a translucent appearance, but when polished appear to have small areas of optical transparency. Subsequent anneals of these samples do not eliminate the cp’ phase AION. Third Series of Experiments This series of experiments focused on improving the appearance of the samples by eliminating the cp’ phase AION. It was hypothesized that the interaction with the settling powder was driving the AIN content of the sample out of the y-AION region and into a region of y-AION and liquid (see reference 3), with the liquid phase forming cp’ phase AlON upon cooling. An increase of the AIN mol% could compensate for this shift and produce samples of pure y-AION. A series of samples containing 15.0 wt% Alp04 and increasing amounts of AIN were fired with all other factors kept constant. The settling powder composition was a 20.0 mol% AIN mixture, with a 0.5 wt% Y203 addition. Hold time was 18 hours at 195OoC,with 10°C/min heating and cooling rates. As AIN mol% in the sample increases, the XRD peaks identified with cp’ phase AlON are less prominent relative to those of y-AION. At 44.0 mol%, the XRD patterns appear to be pure y-AION, but
Advances in Ceramic Armor V
. 217
Pressureless Reaction Sintering of AlON Using Aluminum Orthophosphate
Figure 5. 44.0 mol% AIN sample with cp’ Figure 4. 500X Optical Micrograph of 44.0 mol% phase AlON on grain boundaries. A1N sample. cp’ phase AlON is outlined. microscopy finds cp’ phase AlON present. Figure 4 shows an optical micrograph of the cp’ phase AlON and Figure 5 shows the appearance of this sample. While somewhat transparent, the cp’ phase AlON is clearly visible in the sample and impairing transmission. Increasing AIN mol% to 46.0 mol% eliminates the cp’ phase AION, but sample density decreases and appearance becomes poor. Subsequent experiments indicate that samples of a nominal 44.25 mol% AIN or greater have poor density under the firing conditions in this series. As can be seen from Table I, the drop in density is quite abrupt for samples of greater than 44.0 mol% AIN. Fourth Series of Experiments The fourth series of experiments focused on finding methods to improve the density of samples containing greater than 44.0 mol% AIN. Increases in hold times, peak temperature, and Y203 content were all explored as possible solutions. Small increases of Y203in the settling powder improved the density of the samples. Additions of roughly 0.7 wt% to 0.9 wt% YzO, seem to consistently produce samples with high density. Some examples of the densities achieved under different conditions can be viewed in Table 11. Additions of Y203 greater than 0.9 wt% cause the settling powder to fuse to the sample in such a manner that the distinction between the settling powder and sample is lost. Instead of a layer distinct from the sample that can be removed with minimal grinding, the settling powder in the “fused” samples forms a hard and dense layer requiring extensive grinding. Longer hold times improve density to a limited extent. Samples eventually achieve a maximum density that additional time does not improve. This peak density is achieved faster with increased Yz03 content. Prolonged holds of 30 hours or greater often cause sampIe/powder fusion similar to that Table I Densities of Selected Series 3 Samples
I
218
.
Sample AIN mol% 40.0% 41.0% 44.0% 44.25% 46.0%
I I
I
Advances in Ceramic Armor V
Density 3.632 gJcc 3.619 g/cc 3.637gJcc 3.545 d c c 3.563gJcc
-
1
I
1
AlON Phases Present cp’ and y cp’ and y cp’ and y V
Y
Pressureless Reaction Sintering of AlON Using Aluminum Orthophosphate
Figure 7.500X Optical micrograph Figure 6. lOOOX SEM micrograph of yttrium-rich phase. of yttrium-rich phases. caused by high Y203 additions. Higher temperature improves densification slightly, but significant problems emerge at these temperatures. Very high temperatures (>198OoC) often cause the entire sample to melt. While high, these temperatures should not completely melt AlON. It is unknown if this is caused by the phosphate, poor temperature control, the yttria additions, or an unexpected compositional shift. Temperatures in the range of 1965°C-19750Cimprove densities, but speed the fusion of sample and powder. The results of the fourth series indicate that while adjustments can be made to improve the densities of >44.0 mol% samples, too great an increase in any direction can result in an undesirable fusion of the sample and settling powder. Y203 additions in a range 0.7 wt% to 0.9 wt% combined with hold times of 18-30hours at 1930 to 1950"C,(less time is needed with greater Y203 content or temperature) can consistently avoid sampleipowder fusion while producing dense samples of y-AION. However, the appearance of these samples is not improved over the 44.0mol% AIN sample shown in Figure 5. The increase in Y203 necessary to achieve high density also.increases the yttrium rich phases on the grain boundaries. While smaller in size than the cp' phase AlON regions in previous samples, these phases are still sufficient to cloud the samples and often give them a yellow tint. Elimination of these phases is necessary to attain an optically transparent sample. Fifth Series of Experiments Yttrium-rich phases had been observed throughout the development of the process, and after the completion of the fourth series of experiments they appear to be the primary defects that prevent transparency. The polishing process frequently pulled out or damaged these phases. Figures 6 and 7 contain micrographs showing examples in which the yttrium-rich phases are readily apparent. Elimination of these phases is the focus of the fifth series of experiments. This series of experiments is the current stage of development. Compositional analysis using EDS finds yttrium, aluminum, oxygen, and nitrogen present in Table 11- Densities of Selected Samples Fired in Series 4.Firing Temp 1950°C
Advances in Ceramic Armor V
.
219
Pressureless Reaction Sintering of AlON Using Aluminum Orthophosphate
Figure 8. XRD of 44.25 mol% AIN sample fired for 30 hours at 195OOC in a settling powder with 0.7 wt% Y203 cooled at IOCimin to 1725OC. these phases, although the atomic percentages of the elements, especially yttrium, varies between and even within phases. A range of from 1.5 to 9.2 at% yttrium has been observed thus far. Small amounts of calcium ranging from 0.02 to 0.75 at% are sometimes observed in these phases. This calcium content is likely from trace impurities in the starting powders. The yttrium content in the AlON grains measured with the electron microprobe ranged from 0-250 ppm with up to 450 ppm detected in material very close to an yttrium-rich phase. These measurements are below the solubility of Y in yAION, defined as -1700 ppm at 187OOC by Miller and Kaplan." Some samples produced in the fourth series were annealed at various temperatures in an attempt to eliminate these phases. Sample densities dropped after annealing and sample appearance does not improve. Examples of typical results achieved with these anneals are presented in Table 111. Low temperature anneals (165OoC-1775"C) diffuse yttrium out of the phases, but SEM reveals that non-AION phases containing Al, N, and 0 remain. Subsequent high temperature anneals of these low temperature annealed samples reduce density and do not improve the optical appearance. In an attempt to more uniformly disperse the phases, the cooling rate was slowed to l"C/min down to 1725OC for one sample. Sample composition was 44.25 mol% AIN, and was fired for 30 hours at 1950°C in a settling powder mixture of 20 mol% AIN, 0.7wt% Y203.Below 1725T, a 10°C/min rate was resumed to room temperature. Yttrium-rich phases are still present in this sample, but are reduced in number and size. The density of this sample improved from 3.663 g/cc (see Table 11) when cooled at 10°C/min to a value of 3.676g/cc. While small, this difference in density exceeds the margin of error for the density measurements; estimated at a maximum of *0.004 g/cc based on the scale's margin of error and inaccuracies in the temperature measurement of the water. When viewed with the
220
.
Advances in Ceramic Armor V
Pressureless Reaction Sintering of AlON Using Aluminum Orthophosphate
Table 111- Densities of Selected Series 4 Samples Subjected to 60 hour Anneal at 1870°C
3.670 gkc
3.617 gicc
Figure 9.200X optical micrograph of 44.25 mol% A N sample fired for 30 hours at 1950°C in a settling powder with 0.7 wt% Y203 and cooled at IoC/min to 1725°C. settling powder. Adjustments and changes to the firing cycle are planned to allow slower cooling. DISCUSSION The exact mechanics that govern this interaction are currently unknown. It is possible that the large amount of liquid phase formed by the ALP04 additions causes the sample to incorporate some of the surrounding powder. However, the interaction seems to proceed long after the liquid from the ALP04 additions should have decomposed. Infiltration of an A1203N203 based liquid phase into a SIC compact has been reported by Taguchi, Ribeiro, and Balestra." It is possible that similar mechanisms are applicable in this process. The decomposition of the phosphate should cause significant porosity in the sample. A liquid phase formed in the settling powder would be in the position to infiltrate the sample via the porosity formed by the decomposition of the AlP04. McCauley and Corbin's AlN/A1203 phase diagram3 indicates a liquid forms at 20.0 mol% AIN at the temperatures used in this process. Ternary phase diagrams of the AIN-A1203-Y203 system were plotted by Medraj, Hammond, Thompson, and Drew.I6 These diagrams indicate eutectics at 1776OC and 1783OCI6 and the presence of a liquid phase for the settling powder compositions at the temperatures used in this process. Based on these different phase diagrams, the settling powder mixtures will form liquid phases at the temperatures used in this process. Higher temperatures andor greater yttria content increases the amount of liquid phase. This suggests a possible explanation as to why increases in either of these factors can cause an increase in the speed of the interaction. However, infiltration is simply one possible explanation for the observed behavior. It is possible that other mechanisms govern or influence the interaction, and further work is necessary to better define the nature of this interaction. The yttrium-rich phases appear to be the remains of an unstable liquid phase. In their modeling of transient liquid phase sintering kinetics, Turriff and Corbin" indicate the removal of a transient liquid phase can become slower at higher temperatures, and in extreme cases, the liquid may become a persistent liquid phase. This offers an explanation as to why the yttrium-rich phases occur when the yttrium content is below the solubility limit for y-AION. While elimination of these phases may be possible at lower temperatures, the cooling rate of lO"C/min used in the majority of experiments does not allow the liquid phase sufficient time to dissolve. The drop in density observed for annealed samples may be caused by the removal of these phases by either dissolution or a reformed liquid penetrating the grain boundaries. The solution to the problem of the yttrium-rich phases may be to
Advances in Ceramic Armor V
.
221
Pressureless Reaction Sintering of AlON Using Aluminum Orthophosphate
prevent their formation by slower cooling rates combined with holds at lower temperatures to allow time for the complete dissolution of the liquid phase. The promising results achieved with a slower cooling rate of 1"Cimin seem to indicate that slower cooling rates improve the process. The problem of sampleipowder fusion encountered at slower cooling rates may be prevented by further adjustments to the process, and should not yet be considered an insurmountable obstacle. The optical or mechanical properties of the AlON produced by this process have yet to be measured. Optical transmission measurements will be performed if the yttrium-rich phases are eliminated. CONCLUSIONS Reaction sintering of AlON can be accomplished using additions of Ap04. Sufficient addition of Alp04 causes interaction with settling powders surrounding the sample. The exact nature of this interaction is unknown, but is possibly caused by infiltration of the sample by a liquid phase formed in the settling powder. Under the correct conditions, this interaction can be used to produce dense reaction sintered AION. Optical transparency of this AlON is currently impaired by second phases caused by yttria additions in the settling powder. Further development of the process is needed to eliminate these phases and to better understand the underlying mechanisms governing the process. ACKNOWLEDGEMENTS The submitted manuscript has been authored by a contractor of the U.S. Government under DOE Contract DE-AC07-05ID14517. Accordingly, the U.S. Government retains a nonexclusive, royaltyfree license to publish or reproduce the published form of this contribution, or allow others to do so, for U.S. Government purposes. This information was prepared as an account of work sponsored by an agency of the U.S. Government. Neither the US.Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. References herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise, does not necessarily constitute or imply its endorsement, recommendation, or favoring by the U.S. Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the U S . Government or any agency thereof. FOOTNOTES * Unfortunately, it was discovered later in this work that there were inaccuracies in the temperatures measured by the pyrometer, and some uncertainty as to the uniformity of the hot zone due to an aging heating element. The degree of inaccuracy was measured by comparing pyrometer measurements with temperatures measured by a type C thermocouple, and the values reported have been compensated from the initial pyrometer reading based on this comparison. However, there is still some uncertainty as to the actual temperatures due to the non-uniform hot zone. Based on results obtained after a new element was installed, all temperatures reported in this paper should be treated as possibly exceeding $e actual temperature by a margin of up to 20OC. The diffraction pattern available for identification of cp' phase AlON was considered of questionable quality. Based on the composition of these phases measured with EDS, and for simplicity, these observed phases will be referred to as cp' phase AlON in discussion, but some uncertainty exists that it is genuine cp' phase AION. REFERENCES IN. Corbin, Aluminum Oxynitride Spinel: A Review, J. Eur. Cerarn. Soc., 5, 143-154 (1989).
222
*
Advances in Ceramic Armor V
Pressureless Reaction Sintering of AlON Using Aluminum Orthophosphate
'J. McCauley, P. Patel, M. Chen, G. Gilde, E. Strassburger, B. Paliwal, K.T. Ramesh, and D. Dandekar, AION: A Brief History of its Emergence and Evolution, J. Eur. Ceram. Soc., 29, 223-236 (2009). 3J. McCauley, and N. Corbin, High Temperature Reactions and Microstructures in the A1203-AIN System, Progress in Nitrogen Ceramics, ed. F.L. Rley, Martinus Nijhoff Pub., The Netherlands, 1 1 1118 0 1983 4J. McCauley, and N. Corbin, Phase Relations and Reaction Sintering of Transparent Cubic Aluminum Oxynitride Spinel (AION), J. Am. Ceram. Soc., 62,476-479 (1979) 'C. Martin, and B. Cales, Synthesis and Hot Pressing of Transparent Aluminum Oxynitride, Proc. of SPIE, V. 1112,20-24 (1989) 6J. Zhou, Z. Liao, J. Qi, W. Pang, Y. Wen, J. Wen, H. Wang, J. He, D. Wu, and T. Lu, A New Method for Preparation of Transparent AION Ceramic, Key Engineering Materials, Vols. 368-372, 44 1-443 52008) R. German, Liquid Phase Sintering, 164-172, Plenum Press, New York, NY, 01985 *H. Danninger, Homogenization and Pore Formation During Sintering with Transient Liquid Phase, Powder Metallurgy International, 20,21-25, (1988) 9R. Gentilman, E. Macquire, L. Dolhert, Transparent aluminum oxynitride and method of manufacture, U s Patent 4,720,362 (1988) 'OL. Miller, W. Kaplan, Solubility Limits of La and Y in Aluminum Oxynitride at 187OoC, J. Am. Ceram. Soc., 91, 1693-1696 (2008) I'D. Clay, D. Postlusny, M. Flinders, S. Jacobs, and R. Cutler, Effect of L i A l s 0 ~additions on the sintering and optical Transparency of LiAION, J. Eur. Ceram. SOC.,26, 1351-1362 (2006) "I. Tananaev, E. Maksimchuk, Y. Bushuev, and S . Shestov, Inorg. Muter., 14, 562-565 (1978) I3K. Allaf, and A. Rouanet, Decomposition Thermique des Phosphates 11. Le phosphate d'aluminum AIP04, Rev. Int. Hautes Temper. Repact., 19,89-99 (1982) I%. Corbin, The Influence of Carbon, Nitrogen, and Argon on Aluminum Oxynitride Spinel Formation, Masters Thesis, Department of Materials Science and Engineering, Massachusetts Institute of Technology, Cambridge, MA (1982) " S . Taguchi, S. Ribeiro, and R. Balestra, Infiltration of A1203N203 mix into S i c Ceramic Preforms, Ceramics International, 34, 625-629 (2008) 16M.Medraj, R. Hammond, W. Thompson, and R. Drew, Phase Equilibria, Thermodynamic Modelling and Neutron Diffraction of the AIN-A1203-Y203 System, Canad. Metall. Quart., 42,495-507 (2003) I'D. Turriff, and S . Corbin, Modelling the Influences of Solid-state Interdiffusion and Dissolution on Transient Liquid Phase Sintering Kinetics in a Binary Isomorphous System, Metall. And Muter. Tram., 37A, 1645-1655 (2006)
Advances in Ceramic Armor V
. 223
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
ALON" TRANSPARENT ARMOR Lee M. Goldman, Robyn Foti, Mark Smith, Uday Kashalikar and Suri Sastri Surmet Corporation 3 1 B Street, Burlington, MA USA ABSTRACT Aluminum Oxynitride ( A L O P Optical Ceramic) is a transparent ceramic material which combines transparency fiom the UV to the MWIR and excellent mechanical properties with the ability to be fabricated by conventional powder processing techniques. ALON has isotropic optical and mechanical properties by virtue of its cubic crystalline structure. Consequently, ALON is transparent even in polycrystalline form and thus can be produced by conventional powder processing techniques. This combination of properties and manufacturability make ALON suitable for a range of applications from IR windows, domes and lenses; to transparent armor. ALON" transparent armor laminates represent the state of the art protection against armor piercing threats. ALON" transparent armor offers a factor of two weight and thickness savings over conventional glass laminates against 30 cal AP threats, and more than a factor of two savings against 50 cal AP threats. Single and multi-hit results will be presented against both 30 and 50 cal AP threats. Surmet has developed tiled ALON@armor windows as a means of affordably producing large flat armor windows. Tiled windows as large as 2'x3' have been successfully produced and tested. Recent results will be presented. NTRODUCTION Transparent ceramic materials such as A L O P Optical Ceramic, spinel and sapphire offer a quantum leap in ballistic performance over conventional glass laminates. Each of these materials has been shown to provide protection against armor piercing rounds at about one half the weight and thickness of conventional glass laminates. However, the relatively high cost (compared to glass) and availability of these materials are currently the largest obstacles to their wide spread use. While these materials will always be more expensive than glass, the ability to substantially decrease the cost and increase the production volume will determine if any of these materials will be viable towards transparent armor market. It is precisely in these areas that ALON@Optical Ceramic has an advantage over spinel and sapphire.
ALOP OPTICAL CERAMIC
Originally developed for IR windows and domes, A L O P Optical Ceramic is now available in several materials grades and is used for a wide range of military and commercial applications. The grades are based upon optical requirements, ranging from our armor grade A L O g (AG-ALOP) which is used for transparent armor and various commercial applications, to high purity A L O P (HP-ALON") which is used for IR windows and domes, to the highest grade Low Scatter ALON" (LS-ALON") which is used for reconnaissance windows. All grades of material have equivalent mechanical and chemical properties. Surmet performs all steps of the ALON" manufacturing process in house, beginning with the synthesis of our own A L O P powder, continuing on through forming of the green bodies, the subsequent heat treatment and culminating with the final optical fabrication of the finished component.
225
ALON" Transparent Armor
Surmet is producing three grades of ALON" optical ceramic material: AG-ALON" is a grade of ALON" that has been specifically designed for transparent armor applications. The material has a light tint, and cuts out the UV completely below 0.3 pm. While Sumet can produce AG-ALON" panels in sizes up to 16.5 in x 30 in, we have developed a tilin approach to produce large transparent armor windows using 12x12-in. to 14x20-in ALON0 tiles. Tiling is a method of producing larger andor more cost-effective transparent armor panels. HP-ALON@is designed for infrared window, dome and lens applications. It has higher transmission in visible than AG-ALON". L S - A L O e is designed for the most demanding optical applications such as reconnaissance (Recce) windows that require high homogeneity and isotropy in optical properties in addition to excellent transmission in visible and infrared regions. Comparison of transmission of HP-ALON@and AG-ALON@is shown below in Figure 1. It should be noted that the measured transmission for HF-ALON" and L S - A L O e are equivalent.
+AG -CAG -A- HP +HP +HP
02
04
O6
Mkrono
Oa
sample 1 sample2 sample 1 sample2 sample3
1
1 1
12
Figure I : Comparison of transmission scans for IR Grade ( H P A L O P ) andArmor Grade (AGALOP).
COMPARISON OF ALON AND SPINEL While Surmet produces both ALON and spinel, we market only A L O e for transparent armor applications based upon cost and producibility advantages. While both ALON and spinel are produced by conventional powder processing techniques (Figure 2), and the processes for producing these two materials are very similar, ALON has several advantages over spinel in terms of producibility: The process for ALON is more robust and mature, and less susceptible to lot to lot variations. Surmet produces its own ALON powder, at a significantly lower cost than that of spinel powder The process yields for ALON are currently higher than for spinel
226
Advances in Ceramic Armor V
ALONm Transparent Armor
For these reasons we are able to produce larger lots of A L O P material at a consistently high quality, than are currently possible for spinel. These same factors make ALON a more affordable option than spinel.
Figure 2: Schematic of Process for Producing ALON by Powder Processing
COMPARISON OF ALON TO SAPPHIRE While ALON is made by conventional powder processing techniques (Figure 2), sapphire is grown by melt based single crystal growth techniques. This limits the size (particularly thickness and width) that can be grown in reasonable cycle time and at affordable price. Furthermore, there is little economy of scale for single crystal growth techniques. If you want to double your capacity, you must double the number of crystal growth equipment. By comparison, there is a considerable economy of scale for powder processing based production equipment. Dramatic improvements in throughput, and cost savings, can be attained through the use of larger furnaces. ALON can easily and affordably be produced in thicknesses well above 0.3 in. (the currently supplied standard sapphire layer thickness). ALON" transparent armor thus holds an advantage particularly against armor piercing threats larger than 30 caliber round (12.7 mm, 14.5 mm, etc.) and against improvised explosive devices (EDs). For threats larger than 30calAP, this sapphire layer thickness falls below the commonly used thumb rule for hard faced armor. Thus, A L O e transparent armor is expected to provide a more efficient solution against larger AP and E D threats. While thicker sapphire layers can be produced, it is not likely that it can be done cost effectively, as thicker layers require much slower crystal growth rates.
Advances in Ceramic Armor V
.
227
ALON@Transparent Armor
BALLISTIC RESULTS Surmet has developed and demonstrated technology to produce larger windows through tiling. Typical laminates consist of an outer layer of ALON over a layer glass on top of a polycarbonate layer. The glass and polycarbonate layers are monolithic over the entire window. The ALON tiles are face bonded onto the glassipolycarbonate laminate. The first concern for a tiled window is ballistic performance. The seams in a tiled design can be weak points, and the overall laminate design must account for them. We have conducted ballistic testing at the seams on tiled windows and obtained excellent results. Figure 3 shows a 5-panel,2h2ftFt, tiled window before and after ballistic testing against 30 cal M2AP rounds. The seams employed here were simple butt joints. The panel was shot only on the seams, multiple times until failure. The window shown was shot 13 times, and stopped 11 of the rounds. An equivalent window was shot 13 times and stopped 12 shots. Obviously, the ballistic performance of the tiled window is excellent. Figure 4 shows a similar %panel 2ftx2ft, tiled window before and after ballistic testing against 50 cal M2AP rounds. This tile array successfidly withstood 5 shots, one in the center of each ALON tile. The tiled ALON" transparent armor thus offers multi-hit capability.
Figure 3: 2ftu2ft tiled ALON Armor Window Before and After 30 CalM2AP Ballistic Testing of Seams
Figure 4: 50 Cal Multi-Shot Ballistic Testing The next consideration for tiled transparent armor is aesthetics. Vehicle operators do not like to see seams in their windows. Furthermore, while butt joint seams appear to be pencil thin lines at normal
228
Advances in Ceramic Armor V
ALON@Transparent Armor
incidence, they become more noticeable when the window is viewed at angular incidence. Surmet has developed engineered seams which are less visible, particularly when viewed at an angle. Photographs of the butt joint and engineered seams are shown in Figure 5.
Figure 5: Photograph of 4x441ALON tiles with simple bunjoint (lefa and refinedjoint (right).
RECENT RESULTS ON 50 CAL AP TESTING Over the past year we have concentrated much of our testing on 50 cal threats. This is in large part due to the interests of our particular customer base, and in part to the recognition that it is easier to justify the relatively higher cost of ALON transparent armor for higher level threats where the benefits are more substantial. While much of the testing that we have performed in the recent past is customer specific and consequently quite sensitive, one set of testing was recently performed for, and shown on the Discovery Channel show ‘How Stuff Works’ (January Sthepisode). In this testing the performance of ALON transparent armor laminates was compared to the performance of conventional glass laminates. In particular, an ALON laminate nominally 1.6411thick, with an areal density of 19 psf was compared to a glass laminate nominally 3.6-in thick and with an areal density of approximately 43 psf. Both samples were shot with 50 cal M2 AP rounds at muzzle velocity, at the University of Dayton Research Institutes (UDRI) ballistic test range. High speed video (-5000 framesisec) was taken of the impacts on these two types of laminates. A sequence of images of the high speed impact of the 50 cal round on the all-glass laminate, comsponding to the initial impact itself, is shown in Figure 6. A photograph of the backside of the laminate, after impact, is shown in Figure 7, clearly showing that round passed entirely through the allglass laminate.
Advances in Ceramic Armor V
.
229
ALONa Transparent Armor
Figure 6. High Speed Impact on All-Glass Laminate
Figure 7. Photograph of backside of all-glass laminate, clearly showing an a i t hole The equivalent set of images for the ALON@ laminate is shown below. The high speed images corresponding to the initial impact of the 50 cal AP round on the A L O e laminate are shown in Figure 8. A photograph of the backside of the laminate, after impact, is shown in Figure 9, clearly showing that the ALON* laminate stopped the round.
230
Advances in Ceramic Armor V
ALON@Transparent Armor
Figure 8. High Speed Impact on A L O P Laminate
Figure 9. No Penetration in A L O P Laminate
Advances in Ceramic Armor V
.
231
ALON@TransDarent Armor
SUMMARY Surmet produces three grades of A L O P Optical Ceramic for various applications: AG-ALON' for transparent armor applications HP-ALON' for IR windows and dome applications L S - A L O e for reconnaissance and visible applications Each grade is specifically tailored for its particular application. Armor grade ALON" (AG-ALON@) has been specifically formulated for transparent armor applications in terms of performance and producibility. Individual window panels as large as 16.5~30-in have been produced for certain applications, although for armor applications, panel size is limited to 12x12-in. to 14x20-in. Large windows are produced by tiling smaller panels. 30 cal AP and 50 cal AP testing has been performed on ALON' windows, including tiled window configurations, showing that ALON@transparent armor provides state of the art protection against armor piercing threats. Recent tests against 50 M2AP rounds show that A L O e laminates stop these threats at less than half the weight and thickness of standard glass based transparent armor laminates.
232
9
Advances in Ceramic Armor V
Advances in Ceramic Armor V Edited by Jeffrey J. Swab Copyright 02010 The American Ceramic Society
Author Index
Aghajanian, M. K., 69, 105 Bakas, M., 213 Bekenev, V. L., 39 Bergmannshoff, D., 19 Bless, S., 183 Bottiglieri, S., 91 Brennan, R. E., 135, 147 Campbell, J., 45 Carter, R. H., 147 Chu, H., 213 Curran, D. R., 19 Daloz, W. L., 83 da Rocha, R. M., 113 de Melo, C. L., 113 Ellingson, W. A., 123 Evans, G., 69 Foti, R. 225 Fountzoulas, C. G., 195 Galanov, B. A., 27 Gilde, G., 195 Goldman, L. M., 225 Grady. D. E., 3 Green, W. H., 123, 135, 147
Haber, R. A., 91, 159 Hachatraian, H. V., 39 Houskamp, J. R., 45 Ivanov, S. M., 27 Jadaan, 0. M., 83 Karandikar, P. G., 69 Kartuzov, E. V., 39 Kartuzov, V. V., 27, 39 Kashalikar, U., 225 Lai, K. C., 205 LaSalvia, J. C., 45, 57 Leavy, R. B., 45, 57 Levinson, S., 183 Little, Jr., J. R., 123 MacKenzie, D. E., 45, 57 Marshall, A. L., 105 McCormick, A. L., 105 Miller, H. T., 45, 57 Neal, M., 171 Patel, P. J., 195 Polyzois, I., 183 Portune, A. R., 159
233
Author Index
Rupert, N.,171 Sands, J. M., 135,195 Sastri, S., 225 Schmidt, Jr., K. F.,123 Shockey, D. A., 19 Simons, J. W., 19 Smith, M., 225 Strong, Jr., K. T., 83 Sun, X., 205
234
.
Advances in Ceramic Armor V
Templeton, D. W., 205 Tolman, J., 183 Wells, J., 171 Wereszczak, A. A., 83 Wong, S., 69