Professional Documents
Culture Documents
ASM Principles of Soldering
ASM Principles of Soldering
www.asminternational.org
Principles of Soldering
www.asminternational.org
Copyright 2004 by ASM International All rights reserved No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the written permission of the copyright owner. First printing, April 2004
Great care is taken in the compilation and production of this book, but it should be made clear that NO WARRANTIES, EXPRESS OR IMPLIED, INCLUDING, WITHOUT LIMITATION, WARRANTIES OF MERCHANTABILITYOR FITNESS FORAPARTICULAR PURPOSE,ARE GIVEN IN CONNECTIONWITHTHIS PUBLICATION.Although this information is believed to be accurate byASM,ASM cannot guarantee that favorable results will be obtained from the use of this publication alone. This publication is intended for use by persons having technical skill, at their sole discretion and risk. Since the conditions of product or material use are outside of ASMs control, ASM assumes no liability or obligation in connection with any use of this information. No claim of any kind, whether as to products or information in this publication, and whether or not based on negligence, shall be greater in amount than the purchase price of this product or publication in respect of which damages are claimed. THE REMEDYHEREBYPROVIDED SHALLBE THE EXCLUSIVEAND SOLE REMEDYOF BUYER,AND IN NO EVENT SHALL EITHER PARTY BE LIABLE FOR SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES WHETHER OR NOT CAUSED BY OR RESULTING FROM THE NEGLIGENCE OF SUCH PARTY.As with any material, evaluation of the material under end-use conditions prior to specication is essential. Therefore, specic testing under actual conditions is recommended. Nothing contained in this book shall be construed as a grant of any right of manufacture, sale, use, or reproduction, in connection with any method, process, apparatus, product, composition, or system, whether or not covered by letters patent, copyright, or trademark, and nothing contained in this book shall be construed as a defense against any alleged infringement of letters patent, copyright, or trademark, or as a defense against liability for such infringement. Comments, criticisms, and suggestions are invited, and should be forwarded to ASM International. Prepared under the direction of the ASM International Technical Book Committee (20032004), Charles A. Parker, Chair. ASM International staff who worked on this project include Charles Moosbrugger, Acquisitions Editor; Bonnie Sanders, Manager of Production; Kathy Dragolich and Nancy Hrivnak, Production Editors; Kathryn Muldoon, Production Assistant; and Scott Henry, Assistant Director of Reference Publications. Library of Congress Cataloging-in-Publication Data Humpston, Giles. Principles of soldering / Giles Humpston, David M. Jacobson. p. cm. Includes bibliographical references and index. ISBN 0-87170-792-6 1. Solder and soldering. 2. Brazing. I. Jacobson, David M. II. Title. TS610.H84 2004 671.56dc22 2003058379 SAN: 204-7586 ASM International Materials Park, OH 44073-0002 www.asminternational.org Printed in the United States of America
www.asminternational.org
Contents
Preface .............................................................................................................................................. vii About the Authors............................................................................................................................ ix History ................................................................................................................................................ x Chapter 1: Introduction................................................................................................................... 1 1.1 Joining Methods .................................................................................................................... 1 1.1.1 Mechanical Fastening ................................................................................................. 1 1.1.2 Adhesive Bonding ...................................................................................................... 2 1.1.3 Soldering and Brazing ................................................................................................ 3 1.1.4 Welding ....................................................................................................................... 4 1.1.5 Solid-State Joining ..................................................................................................... 4 1.1.6 Comparison between Solders and Brazes ................................................................. 5 1.1.7 Pressure Welding and Diffusion Bonding ................................................................. 8 1.1.7.1 Pressure Welding .................................................................................................. 9 1.1.7.2 Diffusion Bonding ................................................................................................ 9 1.2 Key Parameters of Soldering ............................................................................................. 12 1.2.1 Surface Energy and Surface Tension ....................................................................... 12 1.2.2 Wetting and Contact Angle ...................................................................................... 13 1.2.3 Fluid Flow ................................................................................................................ 18 1.2.4 Filler Spreading Characteristics ............................................................................... 19 1.2.5 Surface Roughness of Components ......................................................................... 22 1.2.6 Dissolution of Parent Materials and Intermetallic Growth ..................................... 24 1.2.7 Signicance of the Joint Gap ................................................................................... 25 1.2.8 The Strength of Metals ............................................................................................ 27 1.3 The Design and Application of Soldering Processes ........................................................ 28 1.3.1 Functional Requirements and Design Criteria ........................................................ 28 1.3.1.1 Metallurgical Stability ........................................................................................ 29 1.3.1.2 Mechanical Integrity ........................................................................................... 29 1.3.1.3 Environmental Durability ................................................................................... 29 1.3.1.4 Electrical and Thermal Conductivity ................................................................. 30 1.3.2 Processing Aspects ................................................................................................... 30 1.3.2.1 Jigging of the Components ................................................................................ 30 1.3.2.2 Form of the Filler Metal .................................................................................... 31 1.3.2.3 Heating Methods ................................................................................................ 33 1.3.2.4 Temperature Measurement ................................................................................. 34 1.3.2.5 Joining Atmosphere ............................................................................................ 35 1.3.2.6 Coatings Applied to Surfaces of Components .................................................. 37 1.3.2.7 Cleaning Treatments ........................................................................................... 37 1.3.2.8 Heat Treatments Prior to Joining ....................................................................... 37 1.3.2.9 Heating Cycle of the Joining Operation ............................................................ 38 1.3.2.10 Postjoining Treatments ....................................................................................... 39 1.3.2.11 Postjoining Cleaning .......................................................................................... 39 1.3.2.12 Statistical Process Control ................................................................................. 42 1.3.3 Health, Safety, and Environmental Aspects of Soldering ....................................... 42 Chapter 1: Appendices ..................................................................................................................43 A1.1 Solid-State Joining with Gold, Indium, and Solder Constituents ........................... 43 A1.2 Relationship among Spread Ratio, Spread Factor, and Contact Angle of Droplets ................................................................................................................. 44
iii
www.asminternational.org
Chapter 2: Solders and Their Metallurgy................................................................................... 49 2.1 Survey of Solder Alloy Systems ........................................................................................ 51 2.1.1 Lead-Tin Solders ...................................................................................................... 56 2.1.2 Other Tin-Base Solders ............................................................................................ 58 2.1.3 Zinc-Bearing Solders ................................................................................................ 60 2.1.4 Gold-Bearing Solders ............................................................................................... 64 2.1.5 High-Lead Solders .................................................................................................... 72 2.1.6 Indium Solders ......................................................................................................... 73 2.2 Effect of Metallic Impurities .............................................................................................. 75 2.3 Application of Phase Diagrams to Soldering .................................................................... 77 2.3.1 Examples Drawn from Binary Alloy Systems ........................................................ 79 2.3.2 Examples Drawn from Ternary Alloy Systems ....................................................... 83 2.3.3 Complexities Presented by Higher-Order and Nonmetallic Systems ................................................................................................................. 92 2.4 Depressing the Melting Point of Solders by Eutectic Alloying ........................................ 93 2.4.1 Liquid Alloys Based on Gallium ............................................................................. 93 2.4.2 Cadmium-Base Solders ............................................................................................ 93 2.4.3 General Features ....................................................................................................... 93 2.4.4 Implications for Lead-Free Solders ......................................................................... 95 Chapter 2: Appendices ..................................................................................................................96 A2.1 Conversion between Weight and Atomic Fraction of Constituents of Alloys .......................................................................................... 96 A2.2 Theoretical Modeling of Eutectic Alloying ............................................................. 97 Chapter 3: The Joining Environment........................................................................................ 103 3.1 Joining Atmospheres ......................................................................................................... 103 3.1.1 Atmospheres and Reduction of Oxide Films ........................................................ 105 3.1.2 Thermodynamic Aspects of Oxide Reduction ....................................................... 106 3.1.3 Practical Application of the Ellingham Diagram .................................................. 107 3.1.3.1 Soldering in Inert Atmospheres and Vacuum .................................................. 107 3.1.3.2 Soldering in Reducing Atmospheres ............................................................... 109 3.1.3.3 Alternative Atmospheres for Oxide Reduction ................................................ 111 3.1.4 Forming Gas as an Atmosphere for Soldering ...................................................... 111 3.2 Chemical Fluxes for Soldering ......................................................................................... 111 3.2.1 Fluxes for Tin-Base Solders ................................................................................... 116 3.2.1.1 Soldering Fluxes That Require Cleaning ......................................................... 116 3.2.1.2 No-Clean Soldering Fluxes .............................................................................. 118 3.2.1.3 Measure of Cleaning Effectiveness: The Surface Insulation Resistance (SIR) Test ................................................................................... 119 3.2.2 Fluxes for Unsolderable Metals ......................................................................... 120 3.2.2.1 Aluminum Soldering Fluxes ............................................................................ 121 3.2.2.2 Stainless Steel Soldering Fluxes ...................................................................... 122 3.2.2.3 Magnesium Soldering Flux .............................................................................. 122 3.2.3 High-Temperature Fluxes ....................................................................................... 122 3.3 Fluxless Soldering ............................................................................................................ 123 3.3.1 Oxide Formation and Removal .............................................................................. 124 3.3.2 Self-Dissolution of Solder Oxides ......................................................................... 125 3.3.3 Reduction of Solder Oxides by Hydrogen ............................................................ 126 3.3.4 Reduction of Solder Oxides by Atomic Hydrogen ............................................... 127 3.3.5 Mechanical Removal of Oxides (Ultrasonic Soldering) ....................................... 128 3.3.6 Reactive Gas Atmospheres for Reduction of Oxides ........................................... 130 3.3.7 Surface Conditioning Processes ............................................................................. 131 3.3.8 Fluxless Soldering Processes Considerations ........................................................ 132 3.3.8.1 Solderable Component Surfaces ...................................................................... 133 3.3.8.2 Preform Geometry ............................................................................................ 133
iv
www.asminternational.org
3.3.8.3 Mechanically Enhanced Solder Flow .............................................................. 134 3.3.8.4 Metallurgically Enhanced Solder Flow ........................................................... 134 3.3.9 Example of a Fluxless Soldering Process Using In-48Sn Solder .................................................................................................................. 135 3.3.10 Fluxless Soldering of Aluminum ........................................................................... 136 Chapter 3: Appendix ...................................................................................................................137 A3.1 Thermodynamic Equilibrium and the Boundary Conditions for Spontaneous Chemical Reaction ........................................................................ 137 Chapter 4: The Role of Materials in Dening Process Constraints ...................................... 145 4.1 Metallurgical Constraints and Solutions .......................................................................... 147 4.1.1 Wetting of Metals by Solders ................................................................................ 147 4.1.2 Wetting of Nonmetals by Solders .......................................................................... 149 4.1.2.1 Solderable Coatings on Nonmetals .................................................................. 149 4.1.2.2 Active Solders .................................................................................................. 152 4.1.3 Erosion of Parent Materials ................................................................................... 153 4.1.4 Phase Formation ..................................................................................................... 154 4.1.5 Filler-Metal Partitioning ......................................................................................... 155 4.2 Mechanical Constraints and Solutions ............................................................................. 157 4.2.1 Controlled Expansion Materials ............................................................................. 159 4.2.1.1 Iron-Nickel Alloys ............................................................................................ 160 4.2.1.2 Copper-Molybdenum and Copper-Tungsten Alloys ........................................ 161 4.2.1.3 Copper-Surface Laminates ............................................................................... 162 4.2.1.4 Composite Materials ......................................................................................... 163 4.2.2 Interlayers ............................................................................................................... 164 4.2.3 Compliant Structures .............................................................................................. 165 4.2.4 The Role of Fillets ................................................................................................. 167 4.3 Constraints Imposed by the Components and Solutions ................................................. 168 4.3.1 Joint Area ................................................................................................................ 169 4.3.1.1 Trapped Gas ...................................................................................................... 169 4.3.1.2 Solidication Shrinkage ................................................................................... 173 4.3.2 Void-Free Soldering ............................................................................................... 173 4.3.3 Joints to Strong Materials ...................................................................................... 175 4.3.3.1 Joint Design to Minimize Concentration of Stresses ...................................... 175 4.3.3.2 Strengthened Solders to Enhance Joint Strength ............................................ 178 4.3.4 Thick- and Thin-Joint Gap Soldering .................................................................... 178 Chapter 4: Appendices ................................................................................................................180 A4.1 A Brief Survey of the Main Metallization Techniques ......................................... 180 A4.2 Critique of Void-Free Soldering Standards ........................................................... 183 A4.3 Dryness and Hermeticity of Sealed Enclosures .................................................... 184 Chapter 5: Advances in Soldering Technology ......................................................................... 189 5.1 Lead-Free Solders ............................................................................................................. 189 5.1.1 The Drive for Lead-Free Soldering ....................................................................... 190 5.1.2 Compatibility with Lead-Tin Solder ...................................................................... 191 5.1.3 Alternatives to Lead-Tin Solder ............................................................................ 191 5.1.4 Silver-Copper-Tin Ternary Phase Equilibria ......................................................... 193 5.1.5 Metallurgical, Physical, and Chemical Properties of Lead-Free Solders ............................................................................................... 193 5.1.5.1 Surface Tension ................................................................................................ 193 5.1.5.2 Other Physical Properties ................................................................................. 194 5.1.5.3 Mechanical Properties ...................................................................................... 194 5.1.5.4 Corrosion Resistance ........................................................................................ 195 5.1.5.5 Susceptibility to Tin Pest and Tin Whiskers ................................................... 195
v
www.asminternational.org
Process Window for Lead-Free Solders ................................................................ 196 Wetting and Spreading Characteristics of Lead-Free Solders ................................................................................................................ 197 5.1.8 High-Melting-Point Lead-Free Solders ................................................................. 197 5.2 Flip-Chip Interconnection ................................................................................................. 199 5.2.1 The Flip-Chip Process ............................................................................................ 199 5.2.2 Characteristics of Flip-Chip Technology ............................................................... 202 5.2.3 Underll .................................................................................................................. 203 5.2.4 Inspection ................................................................................................................ 203 5.2.5 Rework .................................................................................................................... 204 5.2.6 Self-Alignment of Flip-Chip Structures ................................................................ 204 5.2.7 Surface Topography ................................................................................................ 206 5.2.8 Step-Soldered Flip-Chip Interconnects .................................................................. 206 5.3 Solderability Test Methods and Calibration Standards ................................................... 207 5.3.1 Assessment of Wetting ........................................................................................... 207 5.3.2 Assessment of Spreading ....................................................................................... 210 5.3.3 Solderability Calibration Standards ....................................................................... 212 5.4 Amalgams as Solders ....................................................................................................... 214 5.4.1 Amalgams Based on Mercury ............................................................................... 215 5.4.2 Amalgams Based on Gallium ................................................................................ 216 5.4.3 Amalgams Based on Indium .................................................................................. 217 5.5 Strengthening of Solders .................................................................................................. 217 5.5.1 Grain Renement ................................................................................................... 218 5.5.2 Oxide-Dispersion-Strengthened Solders ................................................................ 218 5.5.3 Composite Solders .................................................................................................. 219 5.6 Reinforced Solders (Solder Composites) ......................................................................... 222 5.7 Mechanical Properties and Numerical Modeling of Joints ............................................. 223 5.7.1 Measurement of Mechanical Properties ................................................................ 223 5.7.2 Numerical Modeling of Joints ............................................................................... 224 5.7.2.1 Dimensional Stability of Soldered Joints ........................................................ 224 5.7.2.2 Prediction of Joint Lifetime ............................................................................. 226 5.8 Solders Doped with Rare Earth Elements ....................................................................... 227 5.8.1 Effect of Rare Earth Additions on Solder Properties ............................................ 227 5.8.2 Implications for Soldering Technology ................................................................. 229 5.9 Diffusion Soldering ........................................................................................................... 230 5.9.1 Process Principles ................................................................................................... 230 5.9.2 Diffusion Soldering of Silver ................................................................................. 231 5.9.3 Diffusion Soldering of Gold .................................................................................. 233 5.9.4 Diffusion Soldering of Copper ............................................................................... 234 5.9.5 Practical Aspects ..................................................................................................... 234 5.9.6 Modeling of Diffusion-Soldering Processes .......................................................... 235 5.10 Advances in Joint Characterization Techniques .............................................................. 235 5.10.1 Ultrasonic Inspection (Scanning Acoustic Microscopy) ....................................... 235 5.10.2 X-Radiography ....................................................................................................... 236 5.10.3 Optical Inspection ................................................................................................... 237 Abbreviations and Symbols.......................................................................................................... 243 Index................................................................................................................................................ 245
5.1.6 5.1.7
vi
www.asminternational.org
Preface
Since the rst edition of Principles of Soldering and Brazing, published in 1993, the authors have received valuable feedback from readers representing a wide range of technical interests. This has prompted the decision to expand the text and organize it into two companion books, one covering soldering and the other brazing. This rst book primarily aims at providing information about soldering in a form that is hopefully readily accessible and as easy to assimilate as possible. Priority is given to the fundamental principles that underlie this eld of technology rather than recipes for making joints. The largely articial distinctions between soldering and brazing are preserved because, despite their many commonalities, it has been found that practicing engineers are either concerned with soldering or brazing and seldom are involved with both simultaneously. The planned companion book, Principles of Brazing, addresses this complementary need. A large proportion of the literature on soldering and brazing may be charged with being heavy on description and light on critical analysis. We have endeavored to redress the balance, while striving to avoid being unduly simplistic or overly mathematical in our approach. Admittedly we may not always have succeeded in this aim. As in Principles of Soldering and Brazing, we have striven to maintain the focus on the fundamental aspects of soldering and have deliberately avoided entering into specic joining technologies in detail. At the same time, we recognize that the range and extent of the knowledge base of metal joining is not immediately obvious, and it requires a fairly deep understanding of materials. To cite a single example, nichrome (an alloy of nickel and chromium), which is a perfectly satisfactory and widely used metallization for soldering, is rendered useless if the solder contains bismuth. If there is an evident bias towards electronic and photonics applications, this reects the recent professional orientation of the authors. Some topics are inevitably not accorded due consideration, although it is hoped that sufficient references are provided to enable the reader to pursue these further. No attempt has been made to gather a comprehensive list of published papers. Those that are included have been selected because they are useful basic texts, cover important subject matter, or relate to exemplary pieces of work, whether in respect of methodology, technique, or other noteworthy features. It was felt that if the value of the book depended on its bibliography, it would rapidly become dated. The advent of computer search facilities and databases of scientic journal and conference abstracts should enable the reader who wishes to nd references on a specic topic to obtain further information without too much difficulty. The search term lead-free solder will yield an astounding 25,000 publications in the public domain, virtually none of which are more than 10 years old. The reader should note that all compositions given in this book are expressed in weight percentage in accordance with the standard industrial practice. These have, for the most part, been rounded to the nearest integer. The ratio of elements in intermetallic compounds, again by convention, refers to the atomic weight of the respective constituents. The general convention used for specifying alloy compositions is that adopted by the alloy phase diagram community, namely in the alphabetical order of the elements, by chemical symbol. We have not been entirely rigorous in this regard as it is sometimes helpful to group alloys by the dominant constituents. Minor additions to bulk compositions are given in order of concentration; for example, Pb-62Sn-0.5Lu-0.02Ce. Specic references are given with each chapter. For those wishing to read more generally on particular topics, the authors would recommend the texts listed as Selected References at the end of this preface. Many phase diagrams are subject to ongoing research, resulting in continued improvement in the accuracy and detail of the information. The most recent version of a diagram may be identied by consulting the latest cumulative index of phase diagrams, published in the Cumulative Index of the periodical Journal of Phase Equilibria (ASM International). This will refer to the source of the thermodynamically assessed diagram of interest. The reader is advised that the four compendia of binary phase diagrams published in the 1960s, 70s and 80s (colloquially referred to as Hansen, Elliott, and Shunk) are now known to contain many errors and omissions.
vii
www.asminternational.org
Information on new developments in soldering and brazing is scattered throughout a wide range of periodicals, as reected in the sources cited in the references appended to the individual chapters. To keep abreast of the literature, the authors have found especially useful the following abstract publications: Metals Abstracts and Science Abstracts. Technical libraries can provide automated searches against specied key words as a monthly service. We wish to thank our many colleagues and ex-colleagues for their helpful advice and encouragement, particularly James Vincent, for insights into lead free soldering. Giles Humpston David M. Jacobson
SELECTED REFERENCES Soldering
Brandon, D. G., and Kaplan, W. D., 1997. Joining Processes: An Introduction, John Wiley & Sons Frear, D.R., Jones, W.B., and Kinsman, K.R., 1990. Solder Mechanics: A State of the Art Assessment, TMS
Hwang, J.S., 1996. Modern Solder Technology for Competitive Electronics Manufacturing,
McGraw-Hill
International Organization for Standardization (IOS), 1990. Welding, Brazing, and Soldering
Processes: Vocabulary, (ISO/DIS 857-2), ISO (currently under revision) Klein Wassink, R.J., 1989. Soldering in Electronics, 2nd ed., Electrochemical Publications Liebermann, E,, 1988. Modern Soldering and Brazing Techniques, Business News Manko, H.H., 2002. Solders and Soldering, 4th ed., McGraw-Hill Nicholas, M.G., 1998. Joining Processes: Introduction to Brazing and Diffusion Bonding, Kluwer Academic Strauss, R., 1998. SMT Soldering Handbook, 2nd ed., Butterworth-Heinemann Thwaites, C.J., 1983. Capillary Joining: Brazing and SoftSoldering, Books Demand UMI Woodgate, R.W., 1996. The Handbook of Machine Soldering: SMT and TH, John Wiley & Sons
Alloy Constitution
John, V.B., 1974. Understanding Phase Diagrams, Macmillan Prince, A., 1966. Alloy Phase Equilibria, Elsevier West, D.R.F., 1982. Ternary Equilibrium Diagrams, Chapman and Hall
viii
www.asminternational.org
ix
www.asminternational.org
History
Origins of Solders and Soldering
The word solder derives from the Old French, soudure, which in turn stems from the Latin solidare, which means to fasten together. Its earliest use in a completely English context as a noun meaning a fusible metallic alloy used for uniting less fusible metal surfaces or parts dates to about 1350. It is interesting to note that in 19th century English, just as in modern French, the l would have been omitted and the word pronounced sod-der, a form that still persists in the United States of America today. Although the origin of solders and soldering is lost to antiquity, it is possible to speculate on how the invention arose. Lead was rst obtained as a by-product of silver production. Silver extraction from ores involved cupellation of lead, and the base metal was then recovered from the litharge [Tylecote 1976]. The softness and malleability of lead were clearly recognized, and there exist examples of lead being used as a setting agent to x posts in the ground and lock morticed stones. It was observed that in this instance the lead ller could give a stronger joint than a simple friction grip. Lead was used by the Mesopotamians (3000 B.C.) to join pieces of copper together, although perhaps more by luck than design since pure lead does not wet copper at all readily. The Romans are known to have produced lead separately from silver, taking advantage of the fact that this metal can be easily extracted from its sulde ore, galena, simply by roasting the mineral in air [Tylecote 1976]. The earliest examples of tin are Egyptian and date from 2000 B.C. What might be construed as a manufactured solder alloy has been found in King Tutankhamuns tomb (1350 B.C.), although there is some debate among scholars about the deliberateness of the metallurgy of this joint. Solders comprising alloys of lead and tin were almost certainly used during the Iron Age [Tylecote 1962]. By the Roman Imperial period there is evidence, both from literary sources and from surviving artifacts, that lead-tin solders were in regular use. Pliny the Elder (1st century A.D.) speaks of tertiarum, an alloy of two parts of (black) lead and one part of white lead (tin) being used for joining metal pipes [Pliny, Natural History xxxiv 161 (Rackham 1952)]. Pliny also remarks that the price of this alloy is 20 denarii per pound. With 25 denarii (silver pieces weighing approximately 4 gm, or 0.14 oz, each) to 1 gold aureaus of close to 8 gm (0.28 oz), the price of Roman solder works out at $70 per kilogram, assuming that gold has maintained its purchasing power since Plinys day. The current price for the same alloy (Pb-33Sn) is lower by an order of magnitude, which indicates how much more precious solder was in antiquity. An analysis of soldered joints in Roman artifacts has shown that both tin-rich and lead-rich alloys were used. The solder in a force-pump from Roman Silchester contains lead to tin in a weight ratio of close to 3 to 1, which is similar to the composition of plumbers solder [Tylecote 1962]. Elsewhere, solders containing mainly tin (80 to 100% Sn), have been encountered in nds from 4th and 5th century sites in Britain [Lang and Hughes 1991]. Soldering, unlike many Roman crafts, either did not die out during the Dark Ages or enjoyed an early revival. The soldering iron, not mentioned at all in Classical times, was well known and in widespread use by the early Middle Ages. Soldering was used for joining the lead strips in stained glass windows, with the oldest complete examples being the Five Prophets windows in Augsburg Cathedral that date from the late 11th century. From 1700 onwards it is clear that soldering was well established with the appearance of tinsmiths and white-iron men as trades. Newcomens discovery of the effectiveness of the internally condensing steam engine in 1708 is attributed to the faulty repair, by soldering, of a blowhole in the cast bronze cylinder. This permitted a spray of external condenser water into the cylinder and the development of the internal condenser; a design that was not superseded until Watt developed the separate condenser nearly 70 years later. Modern soldering practice dates to the early 20th century when improved extraction techniques, which enabled exotic metals to be available at affordable cost, coupled with the appearance of alloy phase diagrams, gave rise to the diversity of alloys now available.
x
www.asminternational.org
REFERENCES
Tylecote, R.F., 1976. A History of Metallurgy, The Metals Society Tylecote, R.F., 1962. Metallurgy in Archaeology: A Prehistory of Metallurgy in the
British Isles, Edward Arnold
Rackham, H., 1952. Natural History, Vol 10, Cambridge, MA, Translation of Pliny
1. Historia Naturalis, Vol 34 (No. 161)
Lang, J. and Hughes, M.J., 1991. Joining Techniques in Aspects of Early Metallurgy, British Museum Occasional Papers, No. 17, British Museum, p 169177
xi
engineers and scientists, a worldwide network dedicated to advancing industry, technology, and applications of metals and materials. ASM International, Materials Park, Ohio, USA www.asminternational.org This publication is copyright ASM International. All rights reserved. Publication title
Principles of Soldering To order products from ASM International: Online Visit www.asminternational.org/bookstore Telephone 1-800-336-5152 (US) or 1-440-338-5151 (Outside US) Fax 1-440-338-4634 Mail
Product code
#06244G
Customer Service, ASM International 9639 Kinsman Rd, Materials Park, Ohio 44073-0002, USA
Email CustomerService@asminternational.org
American Technical Publishers Ltd. 27-29 Knowl Piece, Wilbury Way, Hitchin Hertfordshire SG4 0SX, In Europe United Kingdom
Telephone: 01462 437933 (account holders), 01462 431525 (credit card)
www.ameritech.co.uk Neutrino Inc. In Japan Takahashi Bldg., 44-3 Fuda 1-chome, Chofu-Shi, Tokyo 182 Japan Telephone: 81 (0) 424 84 5550
Terms of Use. This publication is being made available in PDF format as a benefit to members and
customers of ASM International. You may download and print a copy of this publication for your personal use only. Other use and distribution is prohibited without the express written permission of ASM International. No warranties, express or implied, including, without limitation, warranties of merchantability or fitness for a particular purpose, are given in connection with this publication. Although this information is believed to be accurate by ASM, ASM cannot guarantee that favorable results will be obtained from the use of this publication alone. This publication is intended for use by persons having technical skill, at their sole discretion and risk. Since the conditions of product or material use are outside of ASM's control, ASM assumes no liability or obligation in connection with any use of this information. As with any material, evaluation of the material under end-use conditions prior to specification is essential. Therefore, specific testing under actual conditions is recommended. Nothing contained in this publication shall be construed as a grant of any right of manufacture, sale, use, or reproduction, in connection with any method, process, apparatus, product, composition, or system, whether or not covered by letters patent, copyright, or trademark, and nothing contained in this publication shall be construed as a defense against any alleged infringement of letters patent, copyright, or trademark, or as a defense against liability for such infringement.
CHAPTER 1
Introduction
1.1 Joining Methods
surfaces. This method often, but not always, relies on the use of clamping members such as screws and rivets. In crimping, the components are keyed together by mechanical deformation. Characteristic features of mechanical fastening include:
SOLDERING AND BRAZING represent one of several types of methods for joining solid materials. These methods may be classied as:
Mechanical fastening Adhesive bonding Soldering and brazing Welding Solid-state joining
Other methods, such as glass/metal sealing, electrostatic welding, and so forth, are dealt with elsewhere [Bever 1986]. Schematics of these joining methods are given in Fig. 1.1. These different methods have a number of features in common but also certain signicant differences. For example, soldering and brazing are the only joining methods that can produce smooth and rounded llets at the periphery of the joints. The joining methods are listed in the rst paragraph of this chapter in the order in which they lead to fusion of the joint surfaces and tend toward a seamless joint. Because soldering and brazing lie in the middle of this sequence, they share several features with the other methods. For example, soldered and brazed joints can be endowed with the advantageous mechanical properties of welded and diffusion-bonded joints; at the same time they can be readily disassembled, without detriment to the components, like mechanically fastened joints. These features make soldering and brazing highly versatile. The principal characteristics of the various joining methods are summarized in the paragraphs that follow.
1.1.1
Mechanical Fastening
Mechanical fastening involves the clamping together of components without fusing the joint
2 / Principles of Soldering
parameters, such as the clamping force and the materials used, and in service the conductance is unlikely to be constant.
1.1.2
Adhesive Bonding
between the cured adhesive and the respective components. The original interfaces of the joint are preserved in this type of bonding process. Characteristic features of adhesive bonding include:
Adhesive bonding involves the use of a polymeric material, often containing various additives, to stick the components together. The process involves a chemical reaction, which may simply comprise exposure of the adhesive to air, leading to the formation of a hydrogen-type bond
Fig. 1.1
Chapter 1: Introduction / 3
ductive adhesives, and anisotropically conductive adhesives. Table 1.1 provides an indication of the product range of thermally conductive adhesives available from one manufacturer. These advanced materials are being augmented by the development of polymers with high impermeability to moisture and low thermal expansion coefficients for use as electronic packaging materials. Other polymers have been developed that function as both ux and underll material for ip-chip applications. More exciting advances will no doubt continue to become available.
1.1.3
Polymer chemistry is a rapidly evolving science. As a result, some very advanced adhesives have appeared on the market in the last few years with properties highly tailored for particular function in the electronics industry. These include thermally conductive adhesives, electrically con-
Soldering and brazing involve using a molten ller metal to wet the mating surfaces of a joint, with or without the aid of a uxing agent, leading to the formation of metallurgical bonds between the ller and the respective components. In these processes, the original surfaces of the components are eroded by virtue of the reaction occurring between the molten ller metal and the solid components, but the extent of this erosion is usually at the microscopic level (<100 m, or 4000 in.). Joining processes of this type, by convention, are dened as soldering if the
Table 1.1 Selection of commercially available conductive adhesives, used in place of solder for some applications
4030SD 4030LD 4030Hk 4030SR 4130HT 5030P 6030Hk(a)
Paste Properties Viscosity at 10 rpm, kcP(b) 25 C (77 F) shelf life, months Paste density, g/cm3 Processed Properties Die shear(c)(d), kgf/cm2 (psi) Bulk resistivity(d), cm Thermal conductivity, W/m K Youngs modulus, GPa (ksi) Thermal expansion, 106/C (106/F) Rework temperature(e), C (F) Designed product use 56 (800) 40 15 0.9 (125) 15 (28) 100 (212) Highly conductive adhesive 56 (800) 40 15 0.9 (125) 15 (28) 100 (212) Large area device attach 42 (600) 13 3540 0.9 (125) ... 100 (212) 98 (1400) 160 5 ... ... 150 (300) 105 (1500) 40 20 1.8 (250) 17 (30) 150 (300) Highertemperature applications 105 (1500) 25 2025 2.5 (350) 13 (23) 105 (1500) 610 3060 ... ... 3045 6 3.17 2540 6 3.42 2545 6 4.35 3550 6 3.07 3040 6 3.33 3040 6 3.7 30 (typical) 6 3.84.5
Very high Solder thermal and replacement electrical for SMT conductivity adhesive(g)
200250 (f) (390480) Withstands Extremely wire bonding high at 250 C thermal/ (480 F) electrical conductivity
Note in particular the cited thermal conductivity values are comparable to those of many solders. (a) A novel thermoset development material capable of room-temperature storage. (b) kcP is 1000 centipoise (1 Pa s). (c) 6.48 mm (0.255 in.) die, ceramic. (d) 175 C/15 min prole. (e) 70 kPa (10 psi) force. (f) Rework as for typical epoxies. (g) Maximum die size 7 7 mm (0.28 0.28 in.). Source: Multicore Solders Ltd.
4 / Principles of Soldering
ller melts below 450 C (840 F) and as brazing if it melts above this temperature. Characteristic features of soldering and brazing include:
used heating sources are plasma arcs, electron beams, lasers, and electrical current through the components and across the joints (electrical resistance). Filler metals may be used to supplement the fusion process for components of similar composition, as for example when the joint gap is wide and of variable width. In that situation, the ller is often chosen to have a marginally lower melting point than the components in order to help ensure that it completely melts. Characteristic features of welding include:
Soldering and brazing can be applied to a wide variety of materials, including metals, ceramics, plastics, and composite materials. For many materials, and plastics in particular, it is necessary to apply a surface metallization prior to joining.
1.1.5
Solid-State Joining
1.1.4
Welding
Welding involves the fusion of the joint surfaces by controlled melting through heat being specically directed toward the joint. Commonly
The term solid-state joining covers a very wide range of joining processes. The two extremes are pressure welding and diffusion bonding. Pressure welding, at its simplest, involves physically deforming two abutting, faying surfaces to disrupt any intervening surface lms and enable direct metal-to-metal contact. Diffusion bonding in its purest form merely requires placing two faying surfaces in contact and heating the assembly until the voids at the interface have been
Chapter 1: Introduction / 5
removed by diffusion. Further details of process parameters for diffusion bonding of gold and indium are given in Appendix A1.1. Pressure welding generally works better if the components are heated (e.g., friction welding), and diffusion bonding is usually greatly accelerated by the application of pressure or mechanical agitation (e.g., thermosonic ball and wedge bonding, see Fig. 1.2) to force a greater area of the faying surfaces into contact. Solid-state joining constitutes a subject in its own right, quite separate from soldering and brazing, which rely on liquid-state metal joining. However with the development of the diffusionsoldering and diffusion-brazing processes, which are a hybrid of the two, some consideration of solid-state joining, in particular diffusion bonding, is merited. Pressure welding is sometimes used to prepare ller metals in various geometries and to tack preforms in position. Occasionally, conventional ller metals are used to produce a pressure weld or diffusion bond between dissimilar metals in a solid-state joining process. For this reason, further information on both pressure welding and diffusion bonding is included in section 1.1.7 in this chapter. Characteristic features of solid-state joining:
extreme being explosive welding, while diffusion bonding is much slower (>10 min). The joints have no llets. The service temperature of joined assemblies can be higher than the joining temperature and tend toward the melting point of the components. Solid-state joining is limited in application to specic combinations of materials that provide specic combinations of mechanical or diffusion characteristics. Of all the joining methods, they are the least tolerant of poor mating of the joint surfaces. Joint surfaces need to be scrupulously clean because solid-state joining is a uxless process. The properties of solid-state joints can approach those of the parent materials.
Further details on pressure welding are given in section 1.1.7.1 and on diffusion bonding in section 1.1.7.2 in this chapter.
1.1.6
An electronic module in which the semiconductor dies have been interconnected using ne wire attached by thermocompression bonding
Fig. 1.2
In many respects it is fruitful to consider solders together with brazes. This integrated treatment can be justied on metallurgical grounds. These two classes of ller cannot be demarcated by a temperature boundary as is habitually done: conventionally, solders are dened as ller metals with melting points below 450 C (840 F) and brazes as having melting points above this value. This distinction has a historical origin: the earliest solders were based on alloys of tin, while brazes were based on copper-zinc alloys (see History of Soldering in this volume and History of Brazing in the planned companion volume Principles of Brazing for a brief historical background of solders and brazes, respectively). The type of metallurgical reaction between a ller and parent metal is sometimes used to differentiate soldering from brazing. Solders usually react to form intermetallic phases, that is, compounds of the constituent elements that have different atomic arrangements from the elements in solid form. By contrast, most brazes form solid solutions, which are mixtures of the constituents on an atomic scale. However, this distinction does not have universal validity. For example, Ag-Cu-P brazes react with steels to form the interfacial phase of Fe3P in a similar manner to the reaction of tin-base solders with iron and steels to form FeSn2. On the other hand, solid
6 / Principles of Soldering
solutions form between silver-lead solders and copper just as they do between the common silver-base brazes and copper. Also, there exist brazes for aluminum that melt below 450 C (840 F). Soldering and brazing involve essentially the same bonding mechanism: that is, reaction with the parent material, usually alloying, to form metallic bonds at the interface. In both situations, good wetting promotes the formation of llets that serve to enhance the strength of the joints. Similar processing conditions are required, and the physical properties are comparable, provided the same homologous temperature (the temperature at which the properties are measured as a fraction of the melting temperature expressed in degrees Kelvin) is used for the comparison. The perpetuation of the distinction of solders from brazes on the basis of the 450 C (840 F) boundary has arisen from the signicant gap that exists between the melting points of available solder alloys, the highest being Au-3Si, which melts at 363 C (685 F), and the lowest temperature standard braze, the Al-4Cu-10Si alloy, which melts at 524 C (975 F) but, being a noneutectic alloy, is fully liquid only above 585 C (1085 F). Eutectic alloys are dened in Chapter 2, section 2.3; for the present, it shall suffice to state that eutectic alloys are akin to pure metals in melting and freezing at the same tempera-
ture. The temperature ranges of the principal solder and braze alloy families are shown in Fig. 1.3 and 1.4. For most purposes, the temperature gap between solders and brazes is substantially wider than 160 C (290 F). This is because the goldbase solders are very expensive and are largely limited in use to the high added-value manufacturing of the electronics industry. Removing the high-gold-content alloys from consideration, the highest-melting-point solders are the lead-rich alloys, which melt at about 300 C (570 F). The lowest-melting-point brazes that are used commercially in signicant quantities are the reasonably ductile aluminum-silicon alloys, which melt at 577 C (1070 F). A selection of eutectic alloys with melting points in the temperature interval 300 to 550 C (570 to 1020 F) that at some time or other have been promoted as solders and brazes are listed in Table 1.2. They are, without exception, brittle and often contain one or more volatile constituents, notably magnesium, cadmium, or zinc. Some multicomponent alloys that have been developed and are designed to ll the temperature gap are described in Chapter 2. However, none of them are readily available from commercial sources. The dearth of ller metals with melting points in the range 300 to 550 C (570 to 1020 F) is not necessarily a handicap; techniques are avail-
Fig. 1.3
Chapter 1: Introduction / 7
able for making joints using molten ller metal with effective melting points in this temperature interval. Transient-liquid-phase diffusion bonding is one such example and is discussed in Chapter 5, section 5.9. From the maps of solders and brazes in Fig. 1.3 and 1.4, it might appear that there are many more solders than brazes. In fact, the contrary is true. The alloys that are specically indicated in
these gures are the mostly eutectic compositions or those characterized by minimum melting ranges. Most commercially used solders are included because these are almost all of eutectic composition. However, whole families of brazes have been omitted because there is no eutectic in the alloy system; instead they exhibit complete intersolubility. Examples are the copper-nickel, silver-gold, silver-palladium, and silver-gold-
Fig. 1.4
Table 1.2 Selected eutectic alloys that are offered as high-melting point solders and low-melting point brazes
Solder composition(a), wt% Melting point C F Problems
5Ag-95Cd 75Au-25Sb 88Au-12Ge 97Au-3Si 6Al-94Zn 48Al-52Ge 36Al-37Mg 75Pb-25Pd 56Ag-44Sb 58Au-42In 68Al-27Cu-5Si 23Ag-53Cd-24Cu 24Cu-76Sb 62Cd-38Cu
(a) All compositions given are in weight percent.
340 356 361 363 381 424 450 454 485 495 524 525 526 549
644 673 682 685 718 795 842 849 905 923 975 977 979 1020
Toxic fumes, volatile High cost, brittle High cost, brittle High cost, brittle Volatile, brittle High dross, brittle Volatile, brittle Poor fatigue resistance, brittle Volatile, brittle High cost, brittle Difficult to clean, brittle Toxic fumes, volatile, brittle Volatile, brittle Toxic fumes, volatile, brittle
8 / Principles of Soldering
palladium alloys. Alloys in such systems melt over a temperature range that varies with the composition. The higher process temperatures needed to make a brazed joint have important consequences because more thermal activation energy is present. These are:
Several general features distinguish the majority of solders from common brazes, namely:
Most commercial solders are of eutectic composition because there is usually a need to minimize the processing temperature while maintaining reasonable uidity of the molten ller. Also, solders are intrinsically soft and must be conferred with optimal mechanical properties; generally these are achieved by having a ne-grained microstructure, which is a characteristic feature of a true eutectic alloy. Most brazes, by comparison, possess mutual solid solubility between their constituents and are therefore offered with a wide range of compositions and melting ranges. The low degree of intersolubility and the propensity to form intermetallic compounds possessed by solder alloys are related to their constituent elements having a noncubic crystal symmetry. Solders nd application at temperatures at a fraction between 50 and 90% of their melting point in degrees Kelvin, under strain levels that often exceed 10%. At these relatively high temperatures, the alloys are not metallurgically stable and the joint microstructure tends to change with time. Brazes tend to be used at temperatures that are relatively much lower and usually below half their melting point in degrees Kelvin. These points are discussed in further detail in Chapters 2 and 3, and reference should also be made to the planned companion volume Principles of Brazing. Notwithstanding the differences, solders and brazes operate on similar principles, and hence the frequent use of the collective term ller throughout this book has some justication.
More extensive metallurgical reaction between the ller metal and the substrate. Solders typically do not dissolve more than a few microns of the component surfaces, whereas brazes often dissolve tens of microns. Larger changes in the composition of the ller metal therefore occur during brazing, which in turn signicantly affects the uidity of and wetting by the molten ller as well as the properties of the joint. Greater reactivity with the atmosphere surrounding the workpiece. All other factors being equal, brazes are less tolerant of oxidizing atmospheres than solders, but, for the same reasons, are also better suited to cleaning by reducing atmospheres. When joints are made in air with the aid of a ux, the greater reactivity of brazes means that a higher proportion of ux to ller metal is generally required. In consequence, uxcored solders are adequate for use in air, while brazing rods intended for use in ambient atmosphere must be provided with a thick external coating of ux. Fluxes are discussed in Chapter 3, section 3.2. Most, but not all, soldering and brazing processes are performed at small excess temperatures above the melting point of the ller metal, commonly referred to as the superheat. Much higher process temperatures are occasionally used where it is desirable to exploit thermal activation. For example, tin-containing solders can wet and join nonmetallized ceramics provided the solder incorporates an active ingredient, such as titanium, and the alloy is heated above about 900 C (1650 F) [Kapoor and Eagar 1989]. Although the freezing point of the solder is unchanged at about 250 C (480 F), such activated solders have several of the characteristics of brazes at the process temperature. Further information on these alloys can be found in Chapter 4, section 4.1.2.2. On the other hand, the aluminum-germanium eutectic alloy melting at 424 C (795 F) behaves like a typical braze on aluminum and copper surfaces, although by conventional denition, it is classed as a solder.
1.1.7
Solid-state joining methods are not new, and examples of gold-base artifacts fabricated using pressure welds have been dated to 1000 B.C. [Tylecote 1968], while a cup and chalice decorated by diffusion bonding have been dated to 3200 B.C. [Tylecote 1967]. Although more recent interest in welding has been almost totally dominated by fusion welding processes, both pressure welding and diffusion bonding continue to satisfy niche applications because of the unique combination of process and joint param-
Chapter 1: Introduction / 9
eters they offer. Some solid-state joining procedures are a combination of pressure welding and diffusion bonding, as is evident from the fundamental characteristics of each. 1.1.7.1 Pressure Welding
Pressure welding utilizes pressure to rupture surface lms at the joint interface and also to extrude virgin parent metal between islands of surface contamination so that metallic bonding can take place. Thus, the process is characterized by high pressures, applied for short periods of time, on metals that may be either cold or hot. By necessity, bulk plastic deformation of the metals will occur. Possibly the most common examples of pressure welding that are pertinent to solders and brazes are butt welding to join lengths of wires, roll-bonding, and indentation welding. In pressure welding, it is generally accepted that bond formation is controlled by the extent of deformation of the faying surfaces. The term threshold deformation is used extensively in the literature on this subject and is dened as the minimum deformation needed to achieve any bonding, although the strength of a bond at this level of deformation is generally much less than that of the parent metal (see Fig. 1.5). The bonding process can be described as four consecutive stages: 1. Removal of surface contamination and breakup of brittle surface layers, in particular oxides. This is frequently accomplished
by mechanically abrading the surface immediately prior to bonding. Adsorbed water is believed to be the main surface contaminant and responsible for preventing bonding if the deformation is less than 8%. Typically, 40% deformation is required to affect a sound joint when bonding metals in atmospheres other than vacuum. 2. Establishment of physical contact between regions of uncontaminated metal as virgin metal extrudes between gaps in the ruptured surface lms 3. Activation of contacting atoms to form a metallic bond. The contact area determines the extent of bonding. 4. Subsequent atom rearrangement as a consequence of postweld heat treatment and/or stress relaxation Pressure welding is particularly effective when joining dissimilar metals. For good weldability, the softer metal should have the more brittle and stronger oxide lm and vice versa. The hard oxide layer can then promote and assist in the breakup of the surface layers on the harder metal, but is itself easily ruptured by the yielding metal supporting it. For example, the oxide on aluminum fullls the requirements of strength and brittleness compared to the oxides of most other metals, while the metal is relatively soft, and therefore pressure welding of aluminum to other metals occursatlowerdeformationsthanwhenautogenously welded. Also, the different deformation characteristics of dissimilar metals may result in interfacial movement that will enhance bonding compared to autogenous welding.The use of pressure welding to fabricate ductile preforms of brittle alloys using partitioned constituents is discussed further in Chapter 4, section 4.1.5. 1.1.7.2 Diffusion Bonding
The strength of pressure-welded joints as a function of the deformation induced during the bonding process. Below the threshold deformation level, no joining occurs. With increasing deformation the joint strength also increases eventually up to that of the parent materials. Note that the joining process modies the properties of the parent material as it will work-harden when mechanically deformed.
Fig. 1.5
Diffusion bonding relies on a combination of temperature and time to remove voids from the free interfaces between two abutting metal parts. Fundamentally, the process is dened as one in which no plastic deformation of the components being joined takes place, although it is usual to apply some pressure to ensure that the nominally at faying surfaces are indeed in intimate contact. Typical process conditions are durations of up to several hours at temperatures that may be as high as two-thirds of the melting point of the least thermally stable metal in the bonded couple. The use of long times at relatively high tem-
10 / Principles of Soldering
peratures necessitates some form of atmosphere control to preserve surface cleanliness. Soft (roughing) vacuum and controlled atmospheres are equally suitable. Since diffusion processes are the main mechanisms for bonding, with no means for the physical displacement of any intervening nonmetallic surface lms, there are two major considerations in diffusion bonding. The rst is the necessity to ensure that these lms do not constitute a barrier to atom migration. Secondly, in bimetallic systems the formation of intermetallic compounds and porosity arising from inequality of diffusion rate by different species (Kirkendall porosity) must be controlled. Table 1.3 presents some of the better-known direct diffusion-bonding combinations of metals and metalloids. Diffusion bonding does not take place by one dominant mechanism, but is a consequence of one or more possible mechanisms that often operate in parallel. Each mechanism results in material (or void) transport so that the surface energy associated with the interface is progressively reduced as joining proceeds. Some possible mechanisms include:
Plastic yielding of surface asperities Creep of the surface asperities Surface and volume diffusion altering the
shape of the voids
Table 1.3
Ag
Al
Au
Be
Ag Al Au Be Cr Co Cu Fe Nb Mg Mo Ni Pd Pt Ta Ti U V W Stainless steel Cast iron Carbides Graphite ... ... ... ... ... ... ... ... ... X ... ... ... X ... ... ... ... ... ... ... X X ... ... ... ... ... ... ... ... ... ... ... ... X ... ... ... ... ... ...
X X ... X ... ... X X ... ... ... ... ... ... ... X ... ... ... X
X X ... ... ... ... X ... ... X ... X ... ... ... X ... X ... X
... ... X X ... ... X ... ... ... ... X ... ... ... ... ... ... ... ...
X X X X ... ... X ... ... ... ... ... ... ... ... ... ... ... ... X
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... X ...
... ... ... ... ... X ... ... ... ... ... X ... ... ... ... ... ... ... ...
... X ... ... ... ... X ... ... ... ... X X ... ... X ... ... X X
... ... ... ... ... ... X ... X ... X X ... ... X X ... X X ...
... X ... ... ... ... ... ... ... X ... X ... ... ... ... ... ... ... ...
... ... ... ... ... X ... X ... ... X X X ... ... ... ... ... X X
... ... ... ... ... ... ... X ... ... X X ... X ... ... ... ... X ...
... ... ... ... ... ... ... ... X ... X ... ... ... X X ... X ... ...
... ... ... ... ... ... ... ... ... ... ... X ... ... ... X X ... ... ...
... X ... ... ... ... X ... X ... ... ... ... ... X ... ... X ... ...
... ... ... ... X ... ... X X ... X X X X ... ... ... ... X ... X ... ...
X X X ... ... ... X ... X ... ... X X ... ... X ... ... ... X ... ... X
... ... ... ... ... ... ... ... ... ... ... X ... ... ... ... X ... X ... X ... ...
... ... ... ... ... ... ... ... X ... ... X ... ... ... ... ... ... ... ... ... X ...
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... X ... ... ... X ... ... ...
Chapter 1: Introduction / 11
12 / Principles of Soldering
1.2
The quality of soldered joints depends strongly on the combination of ller and component materials, including surface coatings that may be applied to the components, and also on the processing conditions that are used. It is precisely for this reason that a sound understanding of the metallurgical changes accompanying the sequence of events that occur in making soldered joints is so vital for developing reliable joining processes. Soldering technology has generally evolved in an empirical manner, largely by trial and error. Theoretical principles have helped to furnish insights, guidelines, and qualitative explanations for this technology, but have rarely provided reliable data for use in the design of joining processes. The basic difficulty is that the real situation is highly complex, as it brings into play a large number of variables, some of which may not be easy to recognize. Among the relevant factors are the condition of the solid surfaces (i.e., the nature of any oxides or other coatings, surface roughness, etc.) and the temperature gradients that develop during the joining operation, as well as the metallurgical reactions involving the ller and parent materials and also the chemical reactions with uxes, where these are used. Another key aspect of joining with llers is the manner and extent of ow of the molten ller into the joint. These are inuenced by:
be negligible. It is also assumed that the composition and other characteristics of the solid and liquid components, likewise, do not change with time. This assumption is not generally valid, as is shown in this section. Modeling of self-alignment and other wetting and spreading processes can be done with a program entitled Surface Evolver. A search of the World Wide Web using this as the keyword should identify a site from which the software can be obtained.
1.2.1
The concepts of surface energy and surface tension are briey reviewed in this section. Figure 1.6 provides a simplied representation of the atomic structure of a solid close to one of its free surfaces. The atom at position A, in the bulk of the solid, has a balanced array of neighboring atoms, whereas atom B at the surface of the solid is lacking in neighbors above it, apart from the occasional vapor molecule and, therefore, it has some unsaturated bonds. The potential energy of atoms at the free surface, such as B, is higher than the energy of atoms within the bulk of the solid, such as A, by the energy of the unsaturated bonds. The aggregate of this excess energy that is possessed by atoms in the vicinity of the free surface constitutes the surface energy of the solid. In a similar manner, a liquid also possesses a surface energy, which is directly manifested in the tendency to draw up into drops. If small, the droplets are
Dimensions of the joint Spread characteristics of the ller metal Surface condition of the components
The limitation of theory in accounting for observed behavior is well illustrated by the classical model of wetting and spreading. This model nevertheless does provide useful concepts and insights. It is given a detailed treatment by Harkins [1952] and is not repeated here. For the purposes of the present discussion, it suffices to outline the main features of this model. In the classical model of wetting, the surface of the solid is taken to be invariant as a liquid droplet spreads over it. That is to say, the reaction between the liquid and the solid components across their common interface is considered to
Simplied diagram of surface energies. Atom B, at the surface, has unsaturated bonds and thus a higher energy than atom A. This difference in energy is the origin of surface energy SV.
Fig. 1.6
Chapter 1: Introduction / 13
perfect spheres. Because a sphere has the smallest surface-to-volume ratio of any shape, it is clear that the surface energy of a liquid is greater than its volume energy. In the classical model, when a liquid spreads over a surface, the volume remains constant, because evaporation and reaction with the substrate are excluded. Therefore, only surface-energy changes must be considered. A surface of a liquid acts like an elastic skin covering the volume; in other words, the surface is in a state of tension. The tensile force (F), known as surface tension (), is dened as the force acting at right angles to a line of unit length (L) drawn in the surface. The relationship between surface tension and surface energy under specic conditions can be seen as follows. Consider a liquid lm of length L and width W. Apply a force F at a barrier AB, as shown in Fig. 1.7, parallel to one surface of the lm, so as to extend the liquid lm a distance x. The increase in area of the lm is x L. The work done in obtaining this increase is the mathematical product of the force applied times the distance moved, or F x. The work done by the liquid lm in opposing the increase in area, under isothermal conditions (i.e., constant temperature), is 2 x L, where is the surface tension force acting on each surface at the prescribed temperature. At a xed temperature (under isothermal conditions):
Fx 2xL
Thus, surface energy is equivalent to surface tension under isothermal conditions. In the modern metric or International System of Units (SI), the unit of surface energy is joule per square meter (J/m2) and that of surface tension is newton per meter (N/m). Because these parameters are properties of an interface (e.g., between liquid and air), surface energy and tension must be dened with reference to the appropriate pair of materials that meet at the interface, and the test conditions, such as temperature and atmosphere, also must be specied.
1.2.2
According to the classical model of wetting, the liquid will spread over a solid surface until the three surface tensionsbetween the liquid droplet and the solid substrate, the liquid droplet and the atmosphere, and the substrate and the atmosphereare in balance as shown in Fig. 1.8. According to the balance of forces:
SL SV LV cos (Eq 1.1)
where SL is the surface tension between the solid and liquid, LV is the surface tension between the liquid and vapor, SV is the surface tension between solid and vapor, and is the contact angle of the liquid droplet on the solid surface. Equation 1.1, known as the wetting or Youngs equation, shows that < 90 corresponds to the condition SV > SL.This imbalance in surface tension (i.e., surface energy) provides the driving force for the spreading of liquid over the solid surface and diminution of the unwetted surface area. The contact angle provides a measure of the quality of wetting. Thus, if 90 < < 180, some wetting is said to occur, but a liquid droplet will not spread on the surface with which it is in contact. On the other hand, if < 90, a liquid droplet will wet the substrate and also spread
Fig. 1.7
Diagram used to explain the relationship between surface energy and surface tension
Fig. 1.8
Surface tension forces acting when a liquid droplet wets a solid surface, according to the classical model
14 / Principles of Soldering
over an area dened by the contact angle . Clearly, the area of spreading will increase with decreasing contact angle. For further details of the interrelationship between these two parameters, refer to the Appendix A1.2. Rewriting Eq 1.1 in terms of cos :
cos SV SL LV
Thus, wetting is improved by decreasing as cos increases; that is, as approaches 0. Cos may be maximized by:
highly temperature dependent and usually decreases rapidly with increasing temperature [Schwartz 1987, Table 1.4], thereby providing a ready means of controlling spreading. The term LV is a constant at a xed temperature and pressure for a particular liquid-vapor combination, but can be varied by altering the composition and pressure of the atmosphere. Although the composition of the atmosphere used for the joining operation is known to affect the contact angle, in practice it is often easier to promote spreading by reducing the pressure of the atmosphere. This is one of the reasons for the popularity of vacuum-based joining processes. In general, the relative magnitudes of the surface energies are SV > SL > LV. For water wetting on mica, subjected to an atmosphere of water vapor, the following values have been measured [Tabor 1969]:
SV 0.183 N/m SL 0.107 N/m LV 0.073 N/m
Thus, cos (0.183 0.107)/0.073 1 (within the limits of experimental error) and the contact angle, 0. The surface energies of pure metals correlate quite well with their melting points. This is to be expected because the refractoriness of metals reects the strength of the bonds between adjacent atoms in the lattice, and the asymmetry between this and a free atom is responsible for surface energy and tension.
Table 1.4 Surface roughness (Ra) of cold-rolled copper after sanding with wet silicon carbide paper or polishing with a colloidal suspension of alumina in water
Abrasive Nominal particle size, m Ra obtained on cold-rolled copper, m
Wetting angle of lead-tin solder on copper at 10 C above the melting point, 1 min after reow using rosin mildly activated (RMA) ux, as a function of lead concentration. Adapted from Liu and Tu [1998]
Fig. 1.9
200 63 23 5 0.05
For comparison, copper surfaces on electronic component leads usually have an oriented Ra of approximately 0.1 m (4 in.).
Chapter 1: Introduction / 15
It is possible to calculate the surface tension of solder alloys from thermodynamic principles using data for the pure metals. It mostly varies as an essentially linear relationship between the values for the two pure metals. Solders that exhibit poor spreading even at large superheats above the liquidus temperature, such as Sn-40In and Sn-65Bi, have similar surface tension to other solders, but the wetting is controlled by the reaction kinetics at the solder/ substrate interface, which are less favorable [Park et al. 2000]. The wetting equation (Eq 1.1) applies when the liquid is practically insoluble in the solid over which it spreads (i.e., the solubility is less than 0.1%). For binary metal systems where this condition is satised (e.g., tin-chromium), it has been shown that the wetting equation can be reduced to:
to an equilibrium height h where the capillary force balances the hydrostatic force (as shown in Fig. 1.10), such that:
2LV cos gD
(Eq 1.2)
cos 1 k
S Tm L Tm
where k is a constant equal to approximately 0.3, L TS m is the melting point of the solid metal, and T m is the melting point of the liquid metal. This expression has been veried experimentally [Eustathopoulos and Coudurier 1979]. Higher-order metal systems (ternary, quaternary, etc.) are considerably more complex, and the wetting equation cannot be truncated to such a simple form.Amore sophisticated analysis of wetting that takes into account the inuence of certain microscopic features, including the inuence of local defects and van der Waals forces, is provided by de Gennes [1985]. However, this is still a continuum analysis and does not consider the local atomic environment. Indeed, it has further been suggested that Youngs equation is only valid under certain special cases and there are some difficulties with the theoretical denition of solid-surface tension [Xian 2000]. Further academic endeavor will hopefully resolve these issues. So far, this chapter has idealized ller spread over a single surface. In a joint there are always two facing surfaces. If both contact angles are less than 90, the surface energies will give rise to a positive capillary force that will act to ll the joint. For a pair of vertical parallel plates D mm apart and partly immersed in a liquid, the capillary force per mm length of joint is equal to 2LV cos . Under this force, the liquid will rise
where is the density of the liquid and g is the acceleration due to gravity. As might be expected, experimental assessment of capillary rise of solders reveals that capillary rise is less than predicted by theory, although the general principles of Eq 1.2 are substantiated. Meniscus rise is usually greatest for solders that exhibited the lowest contact angle and surface tension and in the narrowest gaps. However, for many solders the correlation with gap width is weak. This has been studied and ultimately was attributed to be due to a rapid increase in voids in the joint as the gap width was decreased progressively (see Chapter 4, section 4.3.1). Indeed, in practical terms capillary ow in narrow gaps is largely dictated by the efficacy of the ux/ller combination used [Vianco and Rejent 1997]. The actual situation in soldering is much more complex than that represented by Eq 1.2 and the classical wetting model. The irreversible nature of spreading and the time dependence of contact angle that is commonly observed are at variance with this simple model. These and other departures from the classical model occur because the
Fig. 1.10
16 / Principles of Soldering
joining process almost invariably involves a degree of chemical reaction between the ller metal and the solid surface, which is neglected in the conventional model. This is clearly demonstrated in a study [Schwartz 1987] that showed that the contact angle for various liquid metals on freshly cleaned beryllium generally decreased with time, over a timescale of several minutes, at a xed temperature. Predictably, perhaps, it was also found that the contact angle decreased with increasing temperature and the atmosphere in which the test was conducted also made a difference. Reactions between a ller metal and the substrate often result in dissolution of the surface of the substrate; this process usually leads to the growth of new phases. Frequently, these phases are intermetallic compounds that either appear distributed throughout the joint or form as layers adjacent to the surface of the solid substrate. The energy of formation of an intermetallic layer by reaction between a molten ller and a solid substrate has been calculated by Yost and Romig [1988] and Wang and Conrad [1995]. The energy of formation considered is the thermodynamic function known as the Gibbs free energy. This function and its properties are briey explained in the Appendix A3.1. In order to simplify the analysis, Yost and Romig limited their consideration to the clean surfaces of pure metals, wetted by liquids of elemental metals, in the absence of uxes, to form binary interfacial phases. It was demonstrated that the free energy of formation of intermetallic phases by reaction of liquid antimony, cadmium, and tin with solid copper was approximately two orders of magnitude larger than the energy release created by the surfaceenergy imbalance during the advance of a spreading solder droplet, which is exclusively considered in the classical model. Therefore, in these cases, and probably more generally in soldering processes, the Gibbs free energy change that occurs on reaction by a ller with the substrate is demonstrably the dominant driving force for wetting. Empirical evidence for this is provided, for example, by the fact that the measured contact angle of molten germanium on silicon carbide at 1430 C (2600 F) is approximately 120, whereas that of molten silicon on this ceramic at the same temperature is 38 [Li and Hausner 1991]. The substantial difference in the two contact angles cannot be accounted for by the difference in
LV in the wetting equation (Eq 1.1). It can only be due to the greater intersolubility of silicon with silicon carbide. This example clearly demonstrates that the simple classical wetting equation cannot be relied on for a quantitative description of wetting, contact angle, or spreading. A more direct example is provided by Fig. 1.11. In this simple system of copper-silicon braze wetted onto graphite, the nal contact angle is insensitive to alloy composition, but the rate of attainment of equilibrium wetting is directly related to the concentration of silicon, which is the active ingredient in the braze. Modications have been proposed to incorporate the Gibbs free energy change accompanying metallurgical reaction into the classical wetting equation by adding additional terms. In particular, the following equation has been developed for the contact angle in reactive wetting [Kritsalis, Coudurier, and Eustathopoulos 1991; Laurent, Chatain, and Eustathopoulos 1991]:
cos cos 0
0 SL SL
LV
Gr LV
(Eq 1.3)
where SL is the solid-liquid interfacial energy 0 after reaction, SL is the interfacial energy before reaction, 0 is the contact angle before reaction, and Gr is the Gibbs free energy of the reaction. Equation 1.3 is probably more of theoretical interest than practical value because its use presupposes knowledge not only of the Gibbs free energy of reaction, but also values of the before and after contact angle or interfacial energy. The effect of metallurgical interaction between ller and the component material in pro-
Fig. 1.11
Contact angle of copper-silicon brazes of different composition on vitreous carbon substrates demonstrating the effect of driving force of alloying on wetting rate and the dependence of the equilibrium wetting angle on the reaction product. Adapted from Landry, Rado, and Eustathopoulos [1996]
Chapter 1: Introduction / 17
moting wetting is exploited in active ller metals: the addition of a small fraction of a reactive metal such as titanium, hafnium, or zirconium to llers enables them to wet and spread over ceramic materials. In this instance, wetting of and reaction with the ceramic are inextricably linked. Activated ller alloys are discussed further in Chapter 4, section 4.1.2.2 and the planned companion volume Principles of Brazing. Although a low contact angle is used as an index for judging the quality of wetting, there are situations where higher contact angles are preferred. This is illustrated in Fig. 1.12, which shows two joints, one between two component surfaces of unequal area and the other between component surfaces that entirely correspond. In the rst case, a low contact angle serves to form a gentle concave llet, which enhances the mechanical properties of the joint. In the other conguration, a low contact angle encourages the formation of a neck in the joint, which can be a source of weakness. A contact angle close to 90 will eliminate this problem.
It is usually presumed that a substrate surface is perfectly wettable, or at least will be when the ux has had sufficient time to perform its cleaning action. Sometimes, however, the majority of the surface area will be wettable and the remainder covered with an array of nonwettable patches. This can be due to inadequate surface preparation or an incorrect choice of ux or process conditions, or it is simply an inherent feature of the substrate material. An example of the latter is some of the new generation of metal-matrix composites that comprise a nely dispersed mixture of a metal and powder of refractory compounds such as Be/BeO and Au/TiN. The net effect of these nonwettable patches is to cause local impediments to wetting and spreading and an increase in the effective contact angle, as shown in Fig. 1.13. A further point to be aware of in connection with wetting is that a situation can arise where the molten ller is physically prevented from achieving its equilibrium contact angle, as, for example, when a solder droplet is conned to a
Fig. 1.12
Effect of contact angle on llet formation and joint lling. Low contact angles tend to be preferred when external llets can form. In other geometries, higher contact angles result in lower stress concentrations.
18 / Principles of Soldering
Effect of nonwettable surface features on the contact angle of solder on copper. Data of Yost, Hosking, and Frear [1993] augmented by the authors. Lead-tin solder wetted onto a copper surface containing embedded nonwettable particles 1020 m (400800 in.) in diameter (RMA ux, 180 C, or 356 F).
Fig. 1.13
small metallized area. This is commonly encountered on electronic circuit boards, where solder droplets are constrained to individual metal pads. In this situation, the pad is often too narrow to accommodate the spherical metal cap that would form if this restriction did not apply. The enforced wetting angle imposes a pressure on the droplet that is often adequate to cause the solder to ow along the length of the conductor that leads away from the pad and, more seriously, to lift and ow under the solder resist that surrounds the pad. An analysis of the pressure arising from a nonequilibrium contact angle, using the classical wetting model, is given by Klein Wassink [1989].
liquid ow, dV/dt, between a pair of horizontal parallel plates, length l, separated a distance D, under a pressure P per unit area transverse to the plates. The viscosity of the liquid is .
dV dt PD3 12l
It is assumed that the liquid front will advance at a rate (dl/dt) equal to the mean velocity of ow, that is:
dl dt
1 dV dt D
PD2 12l
1.2.3
Fluid Flow
The wetting equation determines the degree of wetting for a given liquid-solid combination, but will not provide information on the rate of wetting. Knowledge of the contact angle(s) enables the surface energy to be determined and hence the force that acts to ll the joint gap with liquid. The liquid will ow into the joint under this force at a rate that is governed by its viscosity. Simple uid-ow theory assumes that:
From the wetting equation (Eq 1.1), under isothermal conditions the change in surface energy as a unit area of a surface becomes wetted by the liquid is:
SV SL LV cos
Therefore, the change in surface energy when the pair of parallel plates becomes wetted is:
2l (SV SL ) 2l LV cos
It follows that the force acting on the liquid to cause it to wet the plates is:
F 2l LV cos l
Chapter 1: Introduction / 19
and the velocity of ow of the liquid into the space between two parallel surfaces, of separation D, according to this simple model is given by:
LVD cos 6l
dl dt
(Eq 1.4)
Equation 1.4 shows that the rate of liquid ow increases when: The liquid-vapor surface tension, LV, increases. The joint gap, D, increases. The contact angle decreases. Filler metal viscosity is low. Andrade [1952] derived an empirical formula relating viscosity, when molten, to the molecular weight of metals (in SI units):
0.5 m 1.65 107 T 0.5 m A
for determining solderability. It should be noted that, although the rate of lling is proportional to the joint gap D, the driving force for lling, according to the classical model, is inversely proportional to D; that is, these two aspects of lling act in opposition. This simple model needs to be modied in situations where interfacial reaction occurs while liquid spreading is proceeding. Models that have been tentatively proposed for this situation have been reviewed by Meier, Javernick, and Edwards [1999]. Currently, the lack of relevant data on reaction-rate kinetics, interfacial energy before and after reaction, and diffusion hampers a more complete understanding of spreading, and also wetting, of molten llers, especially where interfacial reaction with solid components is signicant. However, much can be learned from empirical observations, as shown in the following section.
1.2.4
V 2/3
where m is the viscosity at the melting point of the metal, Tm is the absolute melting point, V is the molar volume, and A is the atomic weight of the metal. By assuming limited solubility between the constituents in an alloy and applying the rule of mixtures, it is thereby possible to provide an estimate of the theoretical viscosity of solder. Rates of ow calculated from Eq 1.4 for molten solders in joints 50 m (2000 in.) wide are typically 0.3 to 0.7 m/s (1 to 2.3 ft/s). In other words, a joint 5 mm (0.2 in.) long will be lled in a time of the order of 0.01 s. This implies that joint lling by the molten solder occurs virtually instantaneously and that transient effects associated with uid ow can generally be neglected in joining processes. De Gennes [1985] offers a more developed model of the dynamics of liquid spreading, in which the surface-energy driving force is opposed by viscous drag and surface irregularities. Joint lling times of the order of 0.1 s are routinely measured on instruments used
Molten ller metals do not all have the same spreading characteristics, although, with few exceptions, the degree of spread over an ideal substrate increases as the temperature is raised and the environment is made more reducing. In this context, an ideal substrate, suitable for reference purposes, needs to be dened. This is understood to possess a perfectly clean metal surface that is highly wettable by the ller metal under consideration, but with which it does not signicantly alloy. Any alloying reactions will be highly specic to the combination of materials in question, so that the substrate will lose its ideal characteristics. An example of a substrate that approximates the ideal, and that has been used by the authors in comparative soldering assessments, comprises a at glass plate sputter coated with 0.1 m (4 in.) of chromium and overlaid with 0.1 m (4 in.) of gold. The chromium represents a metal that is essentially insoluble in most solders, and the gold layer provides this reactive metal with protection against oxidation. The gold layer is sufficiently thin to not signicantly alter the composition of a solder pellet as it spreads over the substrate [Humpston and Jacobson 1990]. Eutectic composition alloys are often regarded as having the best spreading characteristics, and this is frequently one of the reasons cited for their selection in preference to hypoeutectic and hypereutectic compositions. The superior spreading of alloys of eutectic composi-
20 / Principles of Soldering
tion in comparison with off-eutectic alloys of the same system, which is often observed, can be explained by the different melting characteristics in the two cases. An alloy of eutectic composition melts instantly. Spreading of the molten alloy is then driven by interaction with the substrate [Ambrose, Nicholas, and Stoneham 1992]. In the case of a noneutectic ller metal, melting, wetting, and spreading commences before the alloy is entirely molten and it will tend to be somewhat viscous. Under these conditions, movement of the ller will be relatively sluggish. By the time the alloy is completely molten, the ller will have partly alloyed with the substrate, and the driving force for spreading will have been diminished. Eutectic composition alloys also have lower viscosity than adjacent compositions when completely molten; further details are given in Chapter 2, section 2.3.1. Whether or not the ller alloy is of eutectic composition is of much less importance to the phenomenon of spreading than the composition per se. The spreading of a ller metal depends greatly on the elemental constituents present and their relative proportions. The authors have compared the spreading characteristics, as a function of excess temperature above the melting point (superheat), of all combinations of the elements bismuth, indium, lead, silver, and tin when used as eutectic solders on ideal substrates [Humpston and Jacobson 1990]. The results presented in Fig. 1.14 show that the area of spreading increases at an accelerating rate as a function of the excess temperature above the melting point of the solder. Furthermore, this study has demonstrated that there is a consistent ranking order for these elements in their ability to promote spreading namely, tin > lead > silver > indium > bismuth. This ranking order is maintained even for ternary and quaternary solders and when applied to a range of substrates, in air, using mild uxes. Although high uidity of a ller metal is a desirable property when it is required to ow into the joint gap of a heated assembly by capillary action, it is not quite so important when the preferred method of applying the ller is to sandwich a thin foil preform between the components, which are then joined together in an appropriate heating cycle. For this type of conguration, a high degree of spreading is detrimental to joint lling, as the ller tends to ow out of the joint. Placement of the ller metal and its inuence on joint lling is discussed in Chapter 4, section 4.3.1.1. In a vacuum or neutral protective atmosphere, the spreading of a ller metal will tend to be in-
ferior to that obtained in air in the presence of a chemical ux. This is to be expected in view of the limited effectiveness of these environments: neither a vacuum nor a protective atmosphere is usually capable of removing oxides that form on the surface of components or the ller while exposed to air before the joining operation. In both cases, the spreading is inferior to that achieved in the presence of an active ux that can remove the surface oxide [Humpston and Jacobson 1991]. Detailed investigation reveals that even an ostensibly simple parameter such as contact angle exhibits somewhat complex behavior. Figure 1.15 shows the contact angle of Pb-60Sn solder on copper at 194 C (381 F), protected by a chemically inert ux, as a function of wetting time. The results indicate that there are at least four distinct stages of wetting and spreading. During the rst 10 s of melting, the solder forms a spherical cap and there is subsequent rapid spreading with a corresponding decrease in the contact angle. The contact angle then temporarily stabilizes, as a dynamic balance is struck among the growth rate of interfacial intermetallic compounds, the diffusion rate in the molten pool of liquid, and the efficacy of the ux in cleaning the substrate surface. This situation persists for about 500 s. Thereafter, a further reduction in contact angle occurs. This is thought to be associated with a progressive change in the composition of the solder resulting in a sudden change in the intermetallic compounds formed in the halo at the edge of the solder pool and hence a change, in this case a further decrease, in the wetting angle. Finally, after many minutes, the contact angle reaches a settled value as the solder pool becomes saturated with the substrate. The molten liquid then commences isothermal freezing as the solidus temperature progressively increases, owing to the alloying with the substrate [Wang and Conrad 1995]. Some attempt has been made to undertake a theoretical analysis of the kinetics of spreading of a molten metal over a wettable solid surface. The current theoretical approach considers the spreading of an inert sessile drop on a smooth and perfectly wetted substrate as a balance between surface-energy drive and viscosity impedance [de Gennes 1985]. However, comparison of this theoretical model with practical experience reveals a number of aws with the model, not the least is that measured ow rates are in the region of four orders of magnitude slower than predicted by theory (see Fig. 1.16). These discrepancies are mostly due to the added metallurgical and
Spread characteristics of binary solder alloys on an ideal substrate as a function of excess temperature above the melting point. The substrate is a at, microscope slide, sputter metallized with 0.1 m (4 in.) of chromium overlaid with 0.1 m (4 in.) of gold. Spread ratio is dened in Appendix A1.2.
Fig. 1.14
22 / Principles of Soldering
ing behavior of a ller. Surface roughness reduces the effective contact angle *, where * is related to , the contact angle for a perfectly at surface through the relation:
cos * r cos
where:
Actual area of rough surface Plan area
r
Contact angle of lead-tin solder on copper as a function of wetting time, using an inert ux and low superheat. There are four distinct stages of wetting, the last being the equilibrium contact angle that is obtained using more typical process parameters.
Fig. 1.15
physical complexity in ller metal wetting and spreading as discussed earlier and are among the simplifying assumptions of this model. Nevertheless, the de Gennes model does predict some interesting dependencies of spreading. First, the initial spreading of molten ller metal is described by the imbalance between Youngs forces and viscous damping. This model also predicts a relative insensitivity of spreading to excess temperature in ller/substrate combinations that wet well. Continued research in this area may achieve a complete mathematical description of wetting and spreading by ller metals that takes into account isothermal solidication and the physical and chemical state of the surface [Ambrose, Nicholas, and Stoneham 1993].
1.2.5
The roughness of joint surfaces can have a signicant effect on both the wetting and spread-
At the same time, by producing a network of ne channels, the texturing may increase the capillary force acting between the ller and the component surfaces. Both phenomena will tend to aid spreading. A directionally oriented surface texture promotes preferential ow parallel to the channeling [Nicholas and Crispin 1986]. From surface-energy calculations it is possible to show that if the instantaneous contact angle of the molten ller is less than the surface angle (i.e., the root angle of V-shaped valleys), then profuse wetting tends to occur along the valleys. This is a frequent observation and, indeed, represents a problem when soldering to rough machined surfaces in that the ller does not spread uniformly in all directions. Exactly the same situation pertains to microscopically rough surfaces. An example of a microscopically rough surface is thick electroplated copper that has been deposited without brighteners (surface leveling agents) or a steel surface that has been etched so as to furrow out the grain boundaries. The resulting surface comprises roughly spherical nodules, which therefore have a continuous network of valleys
Fig. 1.16
Comparison between measured and predicted rates of spreading by molten solder. The large discrepancy arises because the models are based on uid ow and do not take into account the metallurgical driving force for spreading.
Chapter 1: Introduction / 23
between the nodule peaks. Prolometer measurements indicate that the many valleys have contact angles in the region of 15 and are therefore capable of enhancing spreading. The difference here is that rather than a contact line that undulates over a rough surface, the wetting front has a fractal, or lacelike character. The solder wets the channels between the peaks and gives the wetted solder an extensive halo ahead of the main molten pool [Yost, Michael, and Eisenmann 1995]. Where capillary enhancement of spreading is required, the surface texture should be as jagged as possible. A surface prepared by grit blasting or abrading with silicon carbide impregnated paper is therefore preferable to a shot-peened surface. The reason for this is as follows. Sharp reentrant angles that exist on jagged surfaces coincide with sudden changes in the crystallographic orientation of the exposed parent metal. The adhesion of native oxides at these microstructural discontinuities will tend to be relatively weak and provide sites at which the oxide layer can be more readily undermined or penetrated. There is a limit to the roughness of surface that can be used to promote spreading by a molten ller. If the texturing is too deep, then capillary dams can be formed and these will impede the spreading of the ller metal [Funk and Udin 1952].Another factor that should be considered in connection with texturing is the extent of alloying between the ller and the parent material. For ex-
ample, when lead-tin eutectic solder is used to join gold-coated components, there is a limit of about 4% to the concentration of gold that can be accommodated before the solder is embrittled (see Chapter 2, section 2.3.2). If the solder spreads over a gold-coated surface, the critical thickness of the gold coating will be reduced by a factor related to the surface roughness, r, and which must be considered when calculating the total volume of gold to be applied corresponding to a given plan area of spread. Table 1.4 indicates values of surface roughness that can be obtained by abrasion of copper by various means. Attempts have been made to improve joint lling by introducing capillarity enhancers to the joint gap. Such enhancers include nely divided powders and ne meshes that are wetted by the ller but that are effectively inert. This type of approach has been explored by the authors and has not been found to radically improve joint lling. A volume fraction of powder that was calculated to signicantly increase capillary forces had an adverse effect on the uidity of the ller metal. On the other hand, meshes provided stable traps for air and evolved gases in the joint. This led to the formation of an array of voids corresponding to each aperture in the mesh, as revealed by a high-resolution radiograph of a joint with a No. 400 gauze that was soldered using a foil of Ag-96Sn alloy (Fig. 1.17). The importance of surface roughness is exemplied by a recent study of the fracture tough-
Radiograph of a 50 mm (2 in.) diam component soldered using two 50 m (2000 in.) thick foils of Ag-96Sn solder, in high vacuum and incorporating a No. 400 gauze. (b) High-resolution radiograph that reveals the true nature of the joint lling in (a), with a void present at the center of each aperture in the mesh. Magnication: 640
Fig. 1.17
24 / Principles of Soldering
ness of joints made to copper components using Ag-96Sn solder. It was found that the only signicant variable was surface roughness. Joint toughness was found to be largely independent of joint thickness and soldering time for limited ranges of these variables (60200 s and 150400 m, or 616 mils, respectively). The response to variation in surface roughness is reproduced in Fig. 1.18. The low fracture toughness at intermediate roughness occurs when adjacent sites of intermetallic compound growth meet at unfavorable crystallographic orientations [Stromswold, Pratt, and Quesnel 1994].
lling, because the reactions that are occurring transverse to the ow directions will accelerate [Tunca, Delamore, and Smith 1990]. Dissolution of the substrate and resulting growth of intermetallic compounds both follow Arrhenius-type rate relationships, represented by:
Rate exp
Q kT
1.2.6
It is frequently observed that a ller metal will continue to spread beyond an initially wetted surface area over an extended period of time (>10 s), which would not be expected from classical uid-ow theory. Clearly, classical expressions for uid ow, exemplied by Eq 1.4, do not strictly apply in such cases. Indeed, this type of ow can usually be associated with solid-liquid interfacial reactions, which are neglected in the model described in Milners paper [1958]. Where joint lling is sluggish because of reactions occurring between the ller and the solid surface, increasing the temperature to reduce the viscosity of the molten ller is unlikely to enhance
where Q is an activation energy that characterizes the reaction taking place at temperature T (in degrees Kelvin) and k is the Boltzmann constant. The alternative of increasing the joint gap is not usually an option because this is likely to lead to a reduction in joint lling and/or joint strength, as discussed in Chapter 4, section 4.3. The solution then is to change the materials system; several means by which this can be achieved without changing the parent materials are described in Chapter 4, section 4.1. Interfacial reactions are important, not only in determining the ow characteristics of the ller and its wetting behavior, but also the properties of the resulting joints. When a molten ller wets the parent materials, there is normally some intersolubility between them. It is usually manifested as dissolution of the surfaces of the parent materials in the joint region and the formation of new phases at either interface between the parent materials and the molten ller or within the ller itself when it solidies. The effects of dissolution of the parent materials and compound formation on joints are discussed in detail in Chapter 2, section 2.3. The rate of dissolution of a solid metal in a molten metal is described by Weeks and Gurinsky [1958, p 106161] and Tunca, Delamore, and Smith [1990]:
dC dt
K A (Cs C ) V
(Eq 1.5)
Effect of the surface roughness of copper substrates on the fracture toughness of joints made with silvertin eutectic solder. It is worth noting that the joints under which the test joints were made are relatively extreme in terms of joint thickness (i.e., quantity of solder present) and soldering time, so that the thickness of the copper-tin intermetallic layers formed is likely to be substantially thicker than encountered in normal soldering practice.
Fig. 1.18
where C is the instantaneous concentration of the dissolved metal in the melt, Cs is the concentration limit of the dissolved metal in the melt at the temperature of interest, t is the time, K is the dissolution rate constant, A is the wetted surface area, and V is the volume of the melt. This equation is known as both the Nernst-Shchukarev and
Chapter 1: Introduction / 25
C Cs
1 exp KAt V
(Eq 1.6)
assuming initial conditions of C 0, t 0. Equation 1.6 reects the fact that, in general, the concentration of dissolved metal in the molten ller increases in an inverse exponential manner with respect to time. That is, the dissolution rate is initially very fast but then slows as the concentration of the dissolved parent material tends toward its saturation limit (i.e., equilibrium), as shown in Fig. 1.19. Substituting measured values into Eq 1.6 shows that, for a soldered joint of typical geometry, the equilibrium condition is reached within seconds at the process temperature. Thus, it is possible to use an equilibrium phase diagram to predict the change in the composition of the ller metal that will occur in typical joining operations and the associated depth of erosion of the joint surfaces. Equilibrium phase diagrams and their use in soldering and brazing are considered more fully in Chapter 2, section 2.3. In some materials systems, the product of reaction between molten ller metal and the parent materials is a continuous layer of an intermetallic compound over the joint interface. Once formed,
the rate of erosion greatly decreases because it is then governed by the rate at which atoms of the parent material can diffuse through the solid intermetallic compound. As a rough guide, solidstate diffusion processes are two orders of magnitude slower than solid-liquid reactions, and thus continued dissolution of the parent materials effectively ceases, within the timescales of typical joining processes. Intermetallic growth will, however, continue throughout the life of the product, the practical implications of which are discussed in Chapter 4, section 4.1.4.
1.2.7
The joint gap at the process temperature inuences both the joint lling and the mechanical properties of the resulting joint. The relationship between joint dimensions and mechanical properties is discussed in Chapter 4, section 4.3 and in the planned companion volume Principles of Brazing. In summary, the thinner a joint is, the greater its load-bearing capability tends to be, until a limiting condition is reached. Contact angle, surface tension, and viscosity all reduce with increasing temperature, making good joint lling in narrow joints more readily achievable as the joining temperature is raised above the melting point of the ller metal. A lower practical limit to the joint gap is imposed by three factors:
The concentration of a solid metal in a liquid metal wetted by it changes in an inverse exponential manner with respect to time and is limited by the saturation concentration of the solid constituent in the liquid at that temperature.
Fig. 1.19
26 / Principles of Soldering
further dissolution, or alloying with the ller, which will change its composition and hence its melting point. On the whole, solders tend to form interfacial compounds with parent materials, while brazes usually exhibit more extensive alloying between the materials. This can be partly explained by the fact that most solder alloys are based on elements with crystal structures that differ from those of most common parent metals. Consequently, intermetallic compounds tend to form in preference to solid solutions. A reaction that depresses the melting point of the ller metal is desirable for narrow joints, because its uidity will be enhanced by such a reaction at a constant temperature. A reaction that raises the melting point of the ller metal will tend to increase its viscosity and can cause the ller to solidify at the process temperature before it has lled the entire joint. Wider joints mitigate this effect because the alloying will tend to be diluted. b. Dissolution of the ller in the parent metal. In this situation, the volume of ller will shrink as the reaction progresses; therefore, a larger volume of ller metal accommodated in a wider joint gap is again preferred and for similar reasons. However, absorption of the ller is generally undesirable, because its constituents will tend to penetrate into the parent materials preferentially along grain boundaries, generally to the detriment of the mechanical properties of the assembly and sometimes resulting in embrittlement and/or hot shortness.
Polymers Polymers, rubbers Polymers, semicrystalline Polymers, amorphous Metals Zinc alloys Aluminum alloys Copper alloys Stainless steels Iron alloys Nickel alloys Cast irons Titanium alloys Tungsten/molybdenum alloys Low-expansion alloys (Fe-Ni-base) Graphite Ceramics Ceramics, glass Ceramics, oxide Ceramics, porcelain/clay Ceramics, nitride/carbide Diamond/silica/carbon ber 610 48 37 26 1 to 1 2530 2023 1619 1517 1315 1215 1013 810 47 15 79 150300 100200 50100
The values given are representative of the most widely used materials, rather than provide absolute limits for the different classes listed. The thermal expansivity will depend not only on elemental composition but also on microstructure and temper. Composite materials can have expansivities that effectively range between those of the constituents and depend on the relative proportions of the matrix and reinforcement phases. To convert to customary units of 106/F, multiply given values by 0.55556.
Chapter 1: Introduction / 27
calculations of capillary force and viscous drag of liquid ow; see Fig. 1.20. Generally, when components rest freely on one another and the assembly is heated until the ller is molten, the joint will tend to self-regulate to widths around 50 m (2 mils). Indeed, it has been demonstrated that for a xed combination of ller metal, component materials, and process conditions the joint gap will tend to a xed value specic to the combination. This value must be determined by experiment. If there is insufficient ller metal to ll this gap then the joint will contain voids, or if too much ller is applied the excess will spill out [Bakulin, Shorshorov, and Shapiro 1992]. Where thinner or wider joints are required, it is necessary to insert spacers (such as wires) of the desired width between the components and, for thin joints, to apply pressure during the bonding cycle to overcome the hydrostatic forces that will act to levitate the upper component.
1.2.8
strong is the interface between the parent material and the ller metal in an ideal situation? The cohesive strength of metals results from attractive forces between the constituent atoms. Normally, each atom will occupy a physical location where the net force on it is zero. When the solid metal is strained by the application of an external load, the atoms move from their equilibrium positions, and an opposing stress is set up in the metallic crystal. The attractive force between atoms that share the same electron cloud increases with the distance between them up to a maximum and thereafter decreases abruptly, when failure occurs. A perfect metal lattice will fail at this point by cleavage across the crystallographic plane because this is the region where the interatomic forces are weakest. To a rst approximation, the interatomic force per unit area varies with interatomic separation, x, according to a sine wave with wavelength , as shown in Fig. 1.21. The interatomic force per unit area may then be represented by a sine wave:
o sin 2 x
The purpose of making soldered joints is usually to form a metallic bond between components. A fundamental question, therefore, is how
(Eq 1.7a)
where o is the maximum theoretical strength. The work done per unit area in completely separated neighboring planes of atoms, which are an equilibrium distance apart (i.e., a /4), is then:
dx
0 0
/2
/2
sin 2
dx
Fig. 1.20
Calculated time for molten tin and copper to ow up a perfectly wetted capillary [Nicholas 1989]
This work corresponds to the total surface energy of the two new surfaces created in the fracture, that is, 2SV, where SV is the surface energy per unit area of the solid. Accordingly:
2SV
(Eq 1.7b)
Within the elastic range of strain, Hookes law applies, that is: Fig. 1.21
Variation of interatomic force, per unit area, with distance
x a
28 / Principles of Soldering
cos 2
d dx
o
x0
2 a
Hence:
o E 2a
(Eq 1.8)
its theoretical strength. Only in special materials such as carbon ber are the two values remotely comparable. In ductile metals, application of stress results in the movement of dislocations and other defects through the lattice of individual grains. The interfaces between grains are another region where physical material transport and plastic ow takes place. Failure occurs when the rate of increase in strength of the material due to work hardening falls below the rate of decrease in the load-bearing cross section resulting from the plastic ow. The preceding discussion pertained to bulk materials, that is, the components and the ller metal, when considered in isolation. In reality, the joint interfaces will often be a source of voids, microcracks, local interfacial mismatch stresses, and brittle intermetallic layers. These features tend to be a common source of joint weakness, and they should be minimized through judicious choice of ller/parent material and joining conditions.
so that:
1.3
ESV a
1/2
The theoretical fracture stress is about o /10 for metals, although in practice strengths of metals tend to be only one-tenth of this value (i.e., o /100), owing to the presence of lattice defects and other discontinuities. Possibly somewhat surprisingly, soldered joints subject to simple mechanical stress will often fail in a brittle manner. The reasons for this are elaborated in Chapter 4, section 4.3.3. In brittle materials, failure takes place by the extension of cracks that either preexist in the structure or nucleate at lattice imperfections. The stress to cause fracture can be deduced from Eq 1.9 by replacing the denominator with c, where c is the crack length, thus:
b (ESV c)1/2
A soldered joint is usually required to satisfy a specic set of requirements. Most frequently, it must achieve a certain mechanical strength, which it must retain to the highest service temperature in the intended application. The joint must also endure a particular service environment, which may be corrosive, and it may have to provide good electrical and thermal conductance. In addition, the joint must be capable of being formed in a cost-effective manner without detriment to other parts of the assembly. The principal aspects that need to be addressed can be divided:
The functional requirements of the application and the means of satisfying these through appropriate structural design The achievement of the specied assembly through successful processing Each of these stages is examined in this section.
1.3.1
Since c is very much larger than a, the mechanical strength of a brittle material is low relative to
All soldered joints used in manufactured products must remain solid in service and retain the
Chapter 1: Introduction / 29
associated components in xed positions when subjected to stress. These requirements are usually satised by suitable design of the geometry and the metallurgy of the joint, but there are also other aspects to consider. Not the least among these is the fact that solders, in particular, are often operating under conditions that are relatively at least as severe as those encountered in jet engines [Plumbridge 1995]. Factors that have a bearing on the functional integrity of soldered joints are discussed below.
1.3.1.1
Metallurgical Stability
For a joint to remain solid, in most cases the melting point (solidus temperature) of the ller metal needs to exceed the peak temperature that the component is ever likely to experience. There are exceptions to this rule, which are discussed in Chapter 5, section 5.9. Because the strength of all metals decreases rapidly as the melting point is approached, the peak operating temperature should not exceed about 70% of the melting point of the ller, in degrees Kelvin, if the joint is required to sustain a load. The phases that form on solidication in a soldered joint are frequently unstable at elevated service temperatures. Instability of phases present in the joint at the service temperature may be undesirable, because it can affect its integrity. The effects of continued reactions between the ller and the components also must be considered, as explained in Chapters 2 and 3. Because most solders are softer than the materials that are commonly joined, the mechanical properties of a joint are generally limited by those of the ller metal. An exception arises when the joints are very thin and constrained between parent materials of high modulus, as described in Chapter 4, section 4.3.3. 1.3.1.2 Mechanical Integrity
1.3.1.3
Environmental Durability
The durability of engineering and consumer products often depends on joints maintaining their mechanical integrity for the duration of their expected service life. The mechanical integrity of a soldered joint depends on a number of factors, including:
Joints are normally expected to be robust in relation to the service environment. This most commonly involves exposure to corrosive gases, including sulfur dioxide and other constituents of a polluted atmosphere; to moisture, perhaps laden with salt; and to variable temperature. The corrosion and stress-corrosion characteristics of the joint are then relevant. Corrosion mechanisms are generally very complex and specic to a given combination of materials, chemical environment, and joint geometry. Therefore, each situation should be determined empirically. The temperature of a joint can be shifted well beyond a normal ambient range, especially in aerospace applications and in situations where heat is generated within the assembly itself. Then, thermal fatigue and other changes to the metallurgical condition of the joint, such as the growth of phases, can occur, and these invariably affect the properties of the joint. In other words, there is interdependence between environmental stability and microstructural stability. An appropriate choice of the materials combination used should enable these changes to be constrained within predictable and acceptable limits. In this regard, solders tend to be used at service temperatures that are proportionately very close to their melting points, with respect to the thermodynamic reference point of absolute zero temperature (273 C, or 459 F). Hence, they are metallurgically unstable and microstructural changes take place readily.
30 / Principles of Soldering
1.3.1.4
In certain applications, soldered joints are required to provide electrical and/or thermal contact between components. Generally, thin, welllled soldered joints amply satisfy this requirement. Only in a few extreme situations are the thermal and electrical properties of such joints close to the allowed limits. A case in point is high-power silicon device assemblies, where the joint between the silicon device and the metal backing plate is required to conduct away >1 W/mm2 of thermal power. Here, it is crucial to ensure that the joints are kept thin (<30 m, or 1200 in.) and essentially void-free (<5% by volume) [Humpston et al. 1992].
1.3.2
Processing Aspects
An important aspect that must be considered when designing a joint is the practicality of the process involved. Among the relevant issues are:
Jigging of the components The form of the ller metal Heating method Temperature measurement Joining atmosphere Coatings applied to surfaces of components (as necessary) Cleaning treatments Heat treatments prior to joining Heating cycle of the joining operation Postjoining treatments Jigging of the Components
1.3.2.1
The components being joined normally must be held in the required conguration until the ller metal has solidied. Even if the components can be preplaced without a xture, the use of some form of jig is still frequently benecial to ensure that the components are not disturbed by capillary forces originating from the molten ller metal. On the other hand, it is also possible to exploit the capillary forces to make the assembly self-align during the joining process; this is widely practiced in the fabrication of electronic circuits using lightweight surface-mounted components, for example, style 0201 and ip-chip assembly (see Chapter 5, section 5.2). The jig can be used to fulll more than a holding function. For example, it can serve as a heat
spreader, as a heat sink, or as a heat source. A jig should be constructed from a nonporous material to prevent contamination of the atmosphere surrounding the workpiece. Moreover, the jig material should not be wettable by the molten ller metal, in case they come into contact through accidental spillage. Materials such as brass should be avoided as zinc readily volatilizes. Care should be taken in designing jigs so as not to stress the constrained components through thermal expansion mismatch. A note of caution needs to be sounded when joining to stressed components, as this can lead to brittle failure of the components through a mechanism known as liquid metal embrittlement. The degradation in the mechanical properties of an engineering steel, when stressed in tension and wetted by Pb-4Sn solder, has been studied by Watkins, Johnson, and Breyer [1975]. In practice, this failure mechanism is seldom encountered. Graphite is often favored as a jig material. It is inexpensive, easy to machine, a good thermal conductor (making it an effective heat spreader), is an absorber of radiant heat, and is not wetted by the majority of molten ller metals. Graphite also has the merit of mopping-up oxygen in an oxidizing atmosphere to form CO and CO2. However, if the oxygen partial pressure is already low, then its role as a reducing agent is negligible. Care should be taken to ensure that the graphite used for jigs is of a dense grade so that it will be mechanically robust and have a low porosity to minimize outgassing. The desorption of water vapor, in particular, frequently determines the quality of a gas atmosphere in the vicinity of the workpiece. Jigs are sometimes used to apply a controlled pressure to a joint in an assembly. One component can then be deliberately and elastically distorted to bring it into close and uniform contact with its mating part. This is an advantage when very narrow joints are required and when solidstate diffusion constitutes an important part of the joining process. Compressive loading on the joint also aids expulsion of air and vapor from the joints, which are otherwise trapped in pockets and produce voids. Another application is outlined in Fig. 1.22. Because this is a uxless soldering process, the applied load serves a dual function. First, it helps puncture the oxide lms on the surfaces of the ller. Second, when theller metal melts, the applied force acts against any dewetting capillary force of the liquid to ensure ow. The combination of these two factors leads to improved joint lling. Additional
Chapter 1: Introduction / 31
considerations about uxless joining processes can be found in Chapter 3, section 3.3.
1.3.2.2
Filler metals are available in many different forms. These include congurations that normally can be produced from an ingot by mechanical workingfor example, wire, rings, and foil. Such geometries are not restricted to ductile alloys. If the constituents are individually ductile, then the preform can be partitioned. This method is discussed further in Chapter 4, section 4.1.5. The development of rapid-solidication processes has led to the availability of foils and wire of joining alloys that are inherently brittle. These foils are produced directly from the melt: the process involves forcing molten metal through a hole or slot onto a rapidly spinning, water-cooled, metal wheel. Figure 1.23 shows such a strip casting process in operation, and Fig. 1.24 gives some typical foils produced by this route. The high rate of heat extraction that occurs in this process causes the molten metal to solidify almost immediately on striking the wheel, resulting in the formation of a strip of the alloy with a ne crystalline, or occasionally amorphous, microstructure. The dimensions of the cast material can be controlled by varying the nozzle dimensions, the ejection pressure, the speed of rotation of the wheel, and other parameters of the casting process [Jones 1982]. The rened microstructure of the rapidly solidied alloys generally improves strength and ductility compared
with the same alloys produced by conventional casting and mechanical working (see Fig. 1.25). A good example is the Bi-43Sn eutectic solder, which is normally brittle. However, when prepared as a foil by rapid solidication, the ductility of this alloy is comparable to that of other solders. Solders are also available as nely divided powders that can be mixed with a binder to form a paste capable of being screen printed onto a substrate or applied to the workpiece, via a dispenser, in an automated production line. However, powders and pastes containing powders have an extremely high ratio of surface area to volume of ller, which generally produces high oxide fractions and in the absence of suitable precautions would be detrimental to the quality of the resulting joints. Paste manufacturers go to great lengths to produce solder spheres with polished surfaces and thin oxide skins specically for this very reason. Some of the more common ller metals are available as wire or rod that incorporates a ux. Most readers are probably familiar with uxcored solder wire. In this form, the ller metals can be readily used in air without any additional precautions. In more specialized joining processes, the solder can be deposited as a coating on the components by screen printing, electroplating, and by vapor deposition techniques such as evaporation. Where it is not possible to deposit the actual alloy, sequential layers of the constituent elements can be applied. The former is generally preferred because the melting point of an alloy
(a) Sensor comprising three piezoelectric ceramic elements. These are metallized and soldered onto a metallized substrate. The component is then encapsulated in a polymer to provide protection against the environment. (b) To ensure the designed degree of acoustic coupling between the piezoelectric elements and the substrate, the soldered joints must be of a specied thickness. This was achieved using tungsten spacer wires in each joint and a spring-loaded jig to apply a compressive stress to each element during the process cycle. Also shown is the mask used to apply the metallization pattern to the substrate.
Fig. 1.22
32 / Principles of Soldering
is well dened, whereas there is no guarantee that melting will take place at the desired temperature in the case of the composite layers, unless signicant solid-state diffusion has occurred rst to form the appropriate low-melting-point phases. The use of some form of preplaced ller metal has a number of advantages. Most particularly, because the thickness and area of ller metal are predetermined, the volume of molten ller may be carefully controlled. Also, the number of free surfaces is reduced from four (corresponding to a foil preform sandwiched in the joint) to just two, thereby considerably reducing the proportion of oxides and other impurities deriving from exposed surfaces. Tin-lead solder is the only solder that can be readily applied to faying surfaces as an alloy directly by electroplating. More recently, and
Fig. 1.23
many
Production of foil directly from a molten charge by strip casting. Source: Vacuumschelze GmbH, Ger-
aided by the development of advanced pulseplating methods, coelectroplating of gold-tin solder has been developed. Prior to this, the solder had to be realized as a deposit of gold, overlaid with tin (the respective electropotential of these elements makes it difficult to reverse this order). This order is undesirable because the tin is able to oxidize during storage and on heating to the process temperature. Also, it is difficult to realize thick solder deposits using this strategy because of the need to heat above 420 C (788 F) to destabilize the layer of AuSn intermetallic that will otherwise form as a barrier layer at the interface between the two metals. The new method uses pulse plating to extract a controlled ratio of gold to tin atoms from a cyanide-free, weakly acidic solution and offers the prospect of highly reproducible deposits over large areas to any desired thickness [Sun and Ivey 2001]. Some ingenuity has been applied to other, lessdirect methods of selective application of applying ller metals. One of these exploits a chemical reaction to selectively apply tin-lead solder to ne-pitch pads on printed circuit boards without the normal complexity of precision stencil printing. The so-called Super Solder system uses tin powder combined with a lead salt of an organic acid to make a paste. On heating, the organic acid decomposes, thereby providing a partial uxing action, and the lead so liberated combines with the tin to form the solder alloy in situ. The paste is applied over the entire surface of the circuit board. On heating above 183 C (360 F), any tin-lead formed where there is an exposed copper land will wet to it. The excess paste can be simply washed off as there are no
Fig. 1.24
al. [1988]
Dendrite arm spacing decreases with increasing cooling rate and hence ne-grained microstructures have improved mechanical properties. The data pertain to hypereutectic cast iron. Adapted from Seah, Hmanth, and Sharma [1998]
Fig. 1.25
Chapter 1: Introduction / 33
ux residues to promote adhesion. By introducing the lead content of the solder in essentially liquid form, the diameter of the tin particles can also be decreased. By this means, it is claimed that this process is suitable for component pitches as ne as 80 m (3 mils). The solder deposit is typically 100 m (4 mils) thick [Fuse, Obara, and Irie 1992]. In the electronics industry there exists a very different method of applying the solder to the workpiece. It is wave soldering, which is a process used to manufacture many millions of printed circuit boards (PCBs) each year. It is fundamentally different in that molten solder is applied to the faying surfaces. There are many variations of the process, but, in essence, molten solder is pumped continuously over a weir so that there exists a stationary wave whose height is precisely controlled. The circuit board, already populated with electronic components, is uxed, preheated, and then passed through the very top of the solder wave. The transit time is typically 1 to 2 s, making it a very rapid assembly method. Because the entire area (volume) of the joint is immersed in molten solder during transit through the wave, this removes from the process the need to achieve any spreading, so the solder need only wet the PCB lands and component terminations. 1.3.2.3 Heating Methods
Heat must be supplied to the joint to raise the temperature of the ller metal and joint surfaces above the melting point of the ller. The joint surfaces need to be heated; otherwise the ller metal will be incapable of wetting them and therefore will ball up. To prevent this situation, it is good practice always to heat the ller metal via the components to be joined and never vice versa. The available methods of heating are: local heating, in which only that part of the components in the immediate vicinity of the joint is heated to the desired temperature, and diffuse heating, where the temperature of the entire assembly is raised. Common local heat sources include soldering irons, gas torches, and resistance heating using the assembly as the resistive element. More sophisticated heating techniques, such as induction heating and laser heating, also fall within this category. Although some methods of local heating are applicable to joining in a controlled atmosphere, this is not usually the case with a soldering iron or torch, and a ux must then be used.
In local heating, the rate of heat energy input must be high to overcome the heat conducted away by the components and jigging. A high rate of heat input can achieve the desirable characteristic of fast heating and cooling of the joint. Fast heating coupled with short heating cycles minimizes erosion of substrate surfaces and therefore restricts the formation of undesirable phases, while rapid cooling ensures a ne grain size to the solidied ller and thereby superior mechanical properties. However, these potential benets can be offset by the thermal distortion that might be produced in the components being joined. Local heating can be used to create specied temperature gradients that will restrict the ow of the molten ller metal to the immediate vicinity of the joint. Diffuse heating sources include furnaces (both resistance and optical), hot plates, and induction coils. The features of diffuse heating methods are the opposite of those of the local heating methods. For example, the total energy requirement is higher, as the temperature of the entire assembly has to be raised, which also signicantly increases the process cycle time. On the other hand, there is less risk of thermal distortion and accurate control of temperature is easier to achieve. Diffuse heating methods tend to impose fewer constraints on the atmosphere surrounding the workpiece, because the source of heat is relatively remote from the components. For example, a torch is not generally compatible with a special atmosphere. If diffuse heating is to be used in the fabrication of complex assemblies, the designer must ensure that all of the component parts are able to withstand the peak process temperature. With local heating, heat sinks can be used to protect sensitive areas from excessive thermal excursions. A related consideration when using diffuse heating in a situation where several joints must be made is that the melting point of the ller metal used for the preceding joining operations must be higher than the peak process temperature used in the succeeding cycle. Several different ller metals will therefore be required to fabricate a multijointed product in a step-joining process. A diffuse heating method often used in the electronics industry is reow soldering. A printed circuit board is prepared by populating with components and at each joint location a controlled volume of solder and ux is applied, often in the form of a paste. Heating is carried out in the saturated vapor of a very precisely formulated organic uid that has a condensation temperature
34 / Principles of Soldering
some tens of degrees above the liquidus temperature of the solder. Heaters are used to vaporize the uid, and the circuit board is placed in the vapor stream produced. When this intersects with this cold body of the printed circuit board, the vapor condenses and liberates its heat of vaporization, thereby heating the board and components. The process is particularly effective when the thermal mass of adjacent parts differs greatly as the coldest components condense more vapor and therefore receive the greatest heat input. Temperature gradients are therefore automatically minimized. Through the correct choice of uid, it is also possible to design the working uid so that ux residues are removed and the nished assembly emerges from the process chamber dry and clean. The principal drawbacks are the cost of the working uids and their attendant health and safety issues. The oldest method of heating joints is by naked ame. The gases predominantly used now are acetylene and propane, burnt in oxygen. These gases are inexpensive, widely available, easy to use, and can be made oxidizing, reducing, or neutral by adjustment of the oxygen-to-gas ratio. These three combustion conditions are also readily discernible by eye, allowing a skilled operator to adjust the torch to satisfy the requirements of the job at hand. The thermal characteristics of some common fuel gases burnt in oxygen are given in Table 1.6. A relatively new development in ame technology is the microame (Fig. 1.26). This is essentially a conventional gas torch that uses a hypodermic needle as the gas tip. By this means the ame diameter can be reduced to submillimeter dimensions. This tool is particularly useful for precision hand (or robotic) soldering, brazing, and indeed welding of three-dimensional components, such as jewelry items, musical instruments, and heavy electronic parts, examples of which include coils and connector pins. The ame has an equivalent heat output of somewhere in the region of 500 to 2000 W (370 to
1475 ft lbf/s), so it is a considerably more intense heat source than a soldering iron. A further convenience of these miniature gas torches is their compatibility with a gas source derived from the electrolysis of water, for example, using an iron cell containing a concentrated sodium hydroxide solution. Both the energy source (electricity) and fuel (deionized water) are readily available; the gas supply is effectively instantaneous and does not require pressurized cylinders. This type of equipment is available commercially from a number of European manufacturers and possibly several others, worldwide. 1.3.2.4 Temperature Measurement
The liquid-solid metallurgical reactions that occur during soldering operations are highly temperature dependent. Therefore, reliable measurement of temperature is essential. Thermocouples and pyroelectric elements are the most common types of temperature sensor. A number of precautions should be taken when employing thermocouples. Regular calibration checks should be made to determine if the thermoelectric characteristics of the thermocouple materials have altered and to test for electrical interference affecting the display system. Correct temperature measurement requires good thermal contact between the thermocouple and the object being monitored. This tends to present a problem in vacuum joining processes where thermal contact by mechanical meansnamely,
Table 1.6 Thermal characteristics of common fuel gases burnt in oxygen. In each case, the ame temperature is in the region of 3000 C (5430 F)
Fuel gas Thermal output, kJ/cm3/s (kW/cm3)
15 7 6 9
Fig. 1.26
Ernest Spirig
Microame torch to solder small objects requiring a small, intense heat source. Source: Dipl. Ing.
Chapter 1: Introduction / 35
resting the thermocouple against a surface tends to be inadequate. The thermal mass of the thermocouple and its protective sheath impedes the thermocouple junction from sensing the true temperature of the component surface. These effects can be minimized by embedding the thermocouple within the workpiece to improve thermal transfer. Even when thermocouples are used for temperature measurement in gas atmospheres, where the thermal coupling is better than it is in a vacuum, a change in the measured temperature will lag behind that actually occurring. This delay, which can be of the order of seconds, is difficult to measure accurately, but it must be taken into account if a thermocouple is being used to monitor the temperature of assemblies exposed to high heating and cooling rates. Pyrometers have one important advantage over thermocouples: they are noncontacting sensors of temperature. Measurements may be made remotely from the workpiece, and the response time of the instrument can be accurately determined. Traditional pyrometers are primarily designed for operation above about 750 C (1380 F). However, recent advances in pyroelectric technology have led to the commercial development of thermal imaging bolometers that are capable of measuring radiated energy down to and even below ambient temperature with a fast response time. Bolometers are now worth considering not only for high-temperature brazing, but also for general application to soldering and brazing processes. Apart from their exibility, bolometers are capable of rapidly measuring temperature over a large surface area. By contrast, a thermocouple provides only a highly localized measurement of temperature at its tip; this might not be representative of the entire joint region. 1.3.2.5 Joining Atmosphere
types of atmosphere, dened according to the reaction that occurs between the atmosphere and the constituent materials:
Essentially inert (e.g., nitrogen, vacuum) Reducing (e.g., carbon monoxide, halogen
containing) The implications associated with using each of these atmospheres are considered below. Oxidizing Atmospheres. Air is the most common oxidizing atmosphere. The principal advantages of joining in air are that no special gas-handling measures are required and that there are no difficulties associated with access to the workpiece during the joining process. However, because most component surfaces and those of the ller metal are likely to form oxide scale when heated in air, normally uxes must be applied to the joint region. An active ux is capable of chemically and/or physically removing an oxide lm. The ux may be applied either as a separate agent or as an integral constituent of the joining alloy. The subject of uxes is discussed in detail in Chapter 3, section 3.2. Gold and some of the platinum-group metals do not oxidize when heated in air. Although silver will oxidize at air ambient temperature, the oxide dissociates on heating to about 190 C (375 F). These precious metals are therefore sometimes applied as metallizations to the surfaces of the components being joined in uxless processes. The use of wettable metallizations is discussed in Chapter 4, section 4.1.2.1. Solders that contain signicant proportions of the precious metals are generally less susceptible to oxidation, enabling mild uxes to be used. Inert Atmospheres. From a practical view point, an atmosphere is either oxidizing or reducing. This is because it is not possible to remove and then totally exclude oxygen from the workpiece, except perhaps under rigorous laboratory conditions. Thus, when dening an atmosphere as inert it must be taken as meaning that the residual level of oxygen present is not sufcient to adversely affect the joining process under consideration. An atmosphere that might be suitable for soldering to silver is likely to be inadequate for nickel. Because the inertness of an atmosphere is judged relative to the specic application, it is necessary to dene a quantitative measure of the oxygen present. This parameter is the oxygen
For a molten ller metal to wet and bond to a metal surface, the latter must be free from nonmetallic surface lms. Although it is possible to ensure that this condition is met at the beginning of the heating cycle, by prescribed cleaning treatments, signicant oxidation will generally occur if the components are heated in air. Steps must therefore be taken to either prevent oxidation or remove the oxide lm as fast as it forms. The approach adopted will depend largely on the atmosphere surrounding the workpiece. Soldering processes are conducted in one of three
36 / Principles of Soldering
partial pressure. Partial pressure provides a measure of the concentration of one gas in an atmosphere containing several gases. The partial pressure of a gas in a mixture of gases is dened as the pressure it would exert if it alone occupied the available volume. Thus, dry air at atmospheric pressure (0.1 MPa, or 14.5 psi) contains approximately 20 vol% O2, so that the oxygen partial pressure in air is 0.02 MPa (2.9 psi). Typical inert atmospheres among the common gases include nitrogen, argon, and hydrogen. Hydrogen is included here because it is not capable of reducing the oxides present on the majority of metals at normal soldering and brazing temperatures. The oxygen partial pressure in standard commercial-grade bottled gases is of the order of 10 mPa (1.5 106 psi). Higherquality grades are available, but their cost is usually too prohibitive to permit their use in most industrial applications. Vacuum is frequently used as a protective environment for ller metal joining processes. Vacuum offers several advantages compared with a gas atmosphere, particularly the ability to measure and control the oxygen partial pressure more readily. In a substantially leak-free system, the oxygen partial pressure is one-fth of the vacuum pressure, which is relatively easy to determine, as compared with direct measurement of oxygen partial pressure. Although a roughing vacuum of 100 mPa (15 106 psi) will provide an atmosphere with the same oxygen partial pressure as a standard inert gas, it is possible to improve on this value, by several orders of magnitude, using a high-vacuum pumping system. Alternatively, a low oxygen partial pressure may be achieved by obtaining a roughing vacuum, backlling with an inert gas and then roughing out again. The effect of the second pumping cycle will be to reduce the oxygen partial pressure to less than typically one-thousandth of that in the inert gas, that is, approximately 10 Pa (1.5 109 psi). This estimate assumes that the furnace chamber is completely leak tight and does not outgas from interior surfaces, nor does any oxygen or water vapor backstream through the pump. The disadvantages of using a vacuum system for carrying out a joining process are, principally, restricted access to the workpiece and the inadvisability of using either uxes or ller metals with volatile constituents, such as cadmium, as the vapors can corrode the vacuum chamber, degrade its seals, and contaminate the pumping oils. A frequently overlooked consideration in reduced-pressure atmospheres is adsorbed water
that naturally exists on surfaces that are exposed to ambient atmospheres. The continuous streaming of water vapor that desorbs from surfaces and ows past the workpiece as the pressure in a vacuum chamber is reduced is a source of oxidation; this is discussed in quantitative detail in Chapter 3, section 3.1. In a vacuum system operating at 10 mPa (1.5 106 psi), the desorbing water vapor constitutes the major proportion of the residual atmosphere. An adsorbed monolayer of water vapor of just 100 mm2 (0.16 in.2) in area desorbs to a gas pressure of 4 mPa (6 107 psi) per liter of chamber volume. The surfaces of the chamber should therefore be smooth to minimize the surface area and also dry. In order to reduce this problem further, the walls of the vacuum chamber should be heated and the system should be vented to a dry atmosphere. To effectively desorb water vapor, the bakeout temperature should be at least 250 C (480 F), which may be difficult to achieve in practice owing to design constraints and the employment of rubber and other organic seals. Another source of oxidizing contamination in a vacuum system is oil vapor mixed with air and water vapor, backstreaming from a rotary pump. This can occur whenever the pressure inside the vacuum chamber drops below 1 Pa (1.5 104 psi), but can be largely eliminated by employing a foreline trap or by isolating the pump from the chamber once the required pressure reduction has been obtained. The widespread practice of relying on an open gas shroud to provide an inert atmosphere is often unsatisfactory because it is extremely difcult to control such an atmosphere reliably. For example, turbulence in the inert gas shroud can result in a supply of air actually being directed at the workpiece. Recent advances in furnace technology now permit open furnaces, often belt furnaces intended for reow soldering of PCBs, to achieve very high specication atmospheres in the working zone through careful design of the gas ow at the open portals. A reducing atmosphere is one that is capable of chemically removing surface contamination from metals. Gases that provide reducing conditions are generally proprietary mixtures that liberate halogen radicals. Specic gas-handling systems are usually needed for these in order to satisfy health and safety legislation. For a few metals, hydrogen is satisfactory as a reducing atmosphere. No less important for meeting its functional requirement than the oxygen partial pressure of the gas is its water con-
Chapter 1: Introduction / 37
tent. Hydrogen is a relatively difficult gas to dry, and the water vapor present can present a serious problem. A frost point of 70 C (95 F) is equivalent to a water content of 0.0002% by volumethat is an oxygen partial pressure of about 10 mPa (1.5 106 psi). There is also the risk of explosion when dealing with hydrogen at high temperatures, and hydrogen can embrittle some materials. A more detailed treatment of reducing atmospheres and their use is given in Chapter 3, sections 3.1 and 3.3. 1.3.2.6 Coatings Applied to Surfaces of Components
Only occasionally is the desired joining alloy (chosen on the basis of melting temperature and physical properties) metallurgically compatible with the substrate in the sense that the ller will wet the substrate uniformly, without the consequential formation of embrittling phases by reaction. A solution is to apply a surface coating that will promote wetting by the ller and react with it in a benign manner. Coatings can be applied by a variety of techniques and to thicknesses that suit the particular application. On metals, coatings are usually applied by wet plating methods, which are quick, economical, and exible with regard to the coating thickness. Electroplating cannot often be used directly for most nonmetals, and it is more common instead to rely on vapor-deposited coatings. If the substrate is refractory in character, adhesion of metal coatings tends to be poor unless the metal is itself sufficiently refractory so that it will form a strong reactive bond to the substrate. Widely used metallizations are chromium or titanium as the reactive layer, overlaid by gold or a platinum group metal to provide protection from the atmosphere. These and other metallizations and the principles on which they are designed are described in Chapter 4, section 4.1.2.1. 1.3.2.7 Cleaning Treatments
derlying materials. Thick oxides and other nonmetallic surface layers can be removed chemically. However, mechanical cleaning is generally preferable because chemicals tend to leave residues, which then also have to be removed. A procedure that has proved effective with solder wire is to wipe the latter with a brous tissue soaked in solvent. The rst few wipes leave a black mark on the tissue, which is a combination of native oxides and manufacturing residues (including lubricating oils). As shown by the data presented in Fig. 1.27, the solder wire should be wiped multiple times and, ideally, used within 1 h (see Chapter 3, section 3.3). Mechanical abrasion exposes a fresh metal surface. The roughness of the abraded surface can be readily controlled, and this can be used to advantage in promoting the spreading of the molten ller metal. The rougher the surface, the better the wetting and spreading of the molten ller tend to be, for the reasons given in Chapter 1, section 1.3.5. However, rough surfaces create problems when it is required to cover them with thin metal coatings. For example, thickness uniformity of thin metallic lms is difficult if not impossible to achieve on a rough-textured substrate. Soft components, such as solder foils, can be difficult to mechanically clean using abrasive particles because these tend to get embedded. These and thin vapor-deposited and electroplated metallizations can be protected against atmospheric corrosion by the application of a noble metal overcoat. If correctly stored and handled, such components will not require cleaning prior to bonding and, for obvious reasons, must never be abraded. 1.3.2.8 Heat Treatments Prior to Joining
Prejoining heat treatments are occasionally useful in providing stress relief and thereby pre-
The surfaces of the components to be joined and the ller metal preforms must be free from any nonmetallic lms, such as organic residues and metal oxides, to enable the molten ller metal to wet and alloy with the underlying metal. Fluxes are often capable of removing surface oxides, provided they are reasonably thin. Organic lms can be removed with solvents, which obviously should not react with the un-
Improvement in shear strength of a uxless joint made with In-48Sn wire as a function of the number of wipes made with a paper tissue soaked in ethanol, immediately prior to melting of the solder
Fig. 1.27
38 / Principles of Soldering
venting unpredictable distortion during heating of the components to the bonding temperature. Other situations where prejoining heat treatments can be benecial include those involving components with nonmetallic surface lms that are thermally unstable. In the case of silver, for example, the oxide will readily dissociate when heated above 190 C (375 F) in an ambient atmosphere. Likewise, silver sulde dissociates on heating above 842 C (1548 F). Heating cycles may be used to produce solder alloys from layers of the constituent elements applied to surfaces by screen printing, electroplating, or vapor deposition. By heating the substrate above the melting point of the constituent with the lowest melting point, alloying will occur by solid-liquid interaction. The joint can then be formed in a subsequent heating cycle that is usually referred to as the reow stage. 1.3.2.9 Heating Cycle of the Joining Operation
The prepared components and ller metal, possibly mounted in jigs, are joined by applying heat. The heating cycle involves four important processing parameters: the heating rate, the peak bonding temperature, the holding time above the melting point of the ller, and the cooling rate. In general, it is desirable to use a fast heating rate to limit reactions that can occur below the prescribed bonding temperature. However, the maximum heating rate is normally constrained by adverse temperature gradients developing in
the assembly. These can produce distortions in the components and give rise to nonuniform reactions between the ller and the two joint surfaces. Also, temperatures are difficult to measure reliably during fast heating schedules. A better practice, when joining in a vacuum or special atmosphere furnace, is to heat the assembly rapidly to a preset temperature that is just below the melting point of the ller metal and then hold at this temperature for sufficient time (which can range from a few seconds to over one hour, depending on the size of the assembly and the heating method) to allow the assembly to thermally equilibrate and for water vapor to ush out of the joint. Following this dwell, the assembly may then be rapidly heated to the bonding temperature. This method is used to make joints to PCBs using lead-free solders where only a few degrees of superheat are permitted. Proles of typical temperature cycles are shown in Fig. 1.28. The bonding temperature should be such that the ller is guaranteed to melt, but at the same time should not be so high that the ller degrades through the loss of constituents or by reaction with the furnace atmosphere. The optimal temperature is normally determined by metallurgical criteria, most importantly the nature and extent of the ller-substrate interaction. The peak process temperature is frequently set at about 50 and 100 C (90 and 180 F) above the melting point, because accurate temperature measurement and control is not always readily achievable, especially in reduced-pressure atmospheres, where bulky jigging is used or where conduction be-
Fig. 1.28
Proles of typical temperature cycles. (a) Heating cycle with a controlled prole incorporating dwell stages to reduce thermal gradients. (b) Heating cycle dened solely by attainment of a peak temperature
Chapter 1: Introduction / 39
tween the heat source and the workpiece is poor. Moreover, the reported melting temperatures of some llers are not based on accurate measurements, and it is prudent to make some allowance for this uncertainty. The minimum time that the assembly is held above the melting point must be sufficient to ensure that the ller has melted over the entire area of the joint and the maximum time is usually a compromise based on practical and metallurgical considerations. Extended dwell times tend to result in excessive spreading by the molten ller, possible oxidation gradually taking place, and deterioration of the properties of the parent materials. The cooling stage of the cycle is seldom controlled by the operator, but tends to be governed by the thermal mass of the assembly and jig. Forced cooling can lead to problems, such as exacerbating mismatch stresses. Occasionally one or more dwell stages are required, either to provide stress relief to the bonded assembly, or to induce some requisite microstructural change. An example of the latter would be the aging treatment of a precipitation-hardening alloy. In this case, the solution treatment and quenching stages are carried out in tandem with the actual reow operation. A heat treatment temperature of about 75% of the freezing point of the ller metal in Kelvin usually provides the optimal relief of residual stresses. An example of an assembly that required a stress-relief treatment in order to avoid catastrophic cracking is shown in Fig. 1.29. 1.3.2.10 Postjoining Treatments
remove ux residues and the tarnishing that are produced when joining in air. Flux residues must be removed, as they are usually corrosive, especially when moist and can affect the long-term reliability of the component in service. Both chemical and mechanical means of ux removal are employed. Tarnishing tends to be removed chemically, often with acids, followed by thorough rinsing. If a heat treatment is not integrated into the cooling stage of the bonding cycle, a separate heat treatment may be carried out subsequently. 1.3.2.11 Postjoining Cleaning
Various types of postjoining treatments can be applied. A cleaning schedule is generally used to
Large-area silicon chip soldered into a metallized ceramic package using an alloy based on the Au3Si composition. The expansion mismatch between silicon and the ceramic makes it necessary to cool the assembly at a controlled rate in order to prevent fracture of the components.
Fig. 1.29
For many years it has been standard practice to clean assemblies after soldering. This need arose from the corrosive nature of traditional rosinbased uxes and the subsequent reliability and aesthetic degradation that can occur if the ux residues are not removed. Particularly in the electronics industry, the preferred cleaning agent has been a blend based on chlorouorocarbons (CFCs). These liquids were used in vapor cleaning equipment of low complexity and therefore low cost. In addition to being relatively efficient solvents, CFCs had good compatibility with most engineering materials including solders, no ammability risk, and low toxicity. The process had the additional benet that the product emerged dry from the cleaning operation. The ease of cleaning in this manner was largely responsible for it to be the norm to clean all parts and assembliesit was less expensive and easier to clean than to risk subsequent problems in service. The discovery in the mid-1970s that CFCs were contributing to stratospheric ozone depletion led to elimination of CFCs from solder cleaning and ultimately to a ban on their use and manufacture. The response by industry and particularly those engaged in electronics manufacture was threefold. First, alternative cleaning chemicals and methods were devised. Many of these are now commercially available systems. As a result of this substitution endeavor it is perhaps ironic that it has been proven that cleaning with CFC-based solvents is actually fairly ineffective, especially with modern surface-mount and chip-on-board technologies. Second, new and improved ux chemistries were formulated, including water soluble and no-clean uxes (see Chapter 3, section 3.2). Finally, many manufacturers undertook a thorough reappraisal of the need for cleaning, par-
40 / Principles of Soldering
ticularly given the short life of some products owing to technical obsolescence or the transience of fashion; examples include wristwatches, portable computer games consoles, and mobile telephones. Each of these options is considered further in the paragraphs that follow. Cleaning of Electronic Assemblies. The current scene of cleaning methods is very analogous to the situation regarding lead-free solders as a replacement for lead-tin eutectic; namely, an almost universal cleaning route has now been replaced with a plethora of alternatives. This is partly due to commercial pressures as the various manufacturers compete for market share. The available cleaning methods can be broadly grouped into four categories:
Solvents Semiaqueous formulations Aqueous formulations All of the above with ultrasonic or highpressure jet assistance
stand-off heights of 50 m (0.002 in.) are much more difficult to clean under than plastic leaded chip carriers (PLCCs), which have standoff heights around 100 m (0.005 in.). Different substrates and components have different levels of inherent cleanliness. This is due to the relative chemical affinity and degree of mechanical adhesion between the ux/ux residues and the substrate. Thus, for example, ceramic circuit boards have only half the contamination of ber-reinforced laminate (FR4) boards as reowed, specied in terms of g NaCl equivalence. After cleaning, the results are not greatly different. Best cleaning results are obtained by optimizing the thermal prole of the soldering cycle for the cleaning system used. Some cleaners are far more effective at removing ux residues than unspent ux, and vice versa for others. Usually, manufacturers sell a package comprising a ux and recommended cleaning system and will be able to advise on the most appropriate thermal cycle as well. Comparative studies reveal that the solvent, ux, and equipment manufacturers have a clear understanding of the capability of their products to clean (see Fig. 1.30). There is negligible technical merit in deviating from their recipes. Water-Soluble and No-Clean Fluxes. Water-soluble uxes are designed such that residues and unspent ux are miscible with water. Thus, the assembly can be cleaned in water. To avoid water stains on the product, it is usual to complete the cleaning process with an organic rinse, usually some form of alcohol. Another family of uxes has been developed that carries the designation no-clean. The chemistry of these materials has been carefully formulated to ensure that unspent ux and ux residues are chemically bound or otherwise buffered and remain relatively inert in the presence of moisture. Modern versions of these uxes have been further tailored so that the residue can provide additional functionality. A prime example is provided by no-clean uxes developed for ipchip underll applications, where the ux residue is designed to have a particular modulus and expansion coefficient that helps to extend the fatigue life of the solder joints. As a general observation, the soldering process window for water-soluble and no-clean uxes tends to be somewhat narrower than for a process utilizing a more conventional ux formulation.
Comparative studies have been made of the relative effectiveness of each of these approaches [Richards et al. 1993], from which the following conclusions can be made:
Chapter 1: Introduction / 41
Alternatives to Cleaning. There are three alternatives to cleaningsimply do not clean at all, use uxless processes, or switch to conductive adhesives [Lea 1998a; Lea 1998b]. Electrically conductive adhesives have made considerable technological progress and are widely used to interconnect to LCD displays. However, they have not yet reached the point of maturity where they can be used reliably for interconnects that carry more than a few microamperes of current or high frequencies. There are also issues concerning their stability in service: deterioration of their adhesion and the slow evolution of volatile species, which is of particular concern when conductive adhesives are used to interconnect bare die or are in the vicinity of naked optical facets. No doubt, improved stability will be achieved as better-formulated products become available. At face value, uxless processes are highly attractive because there is no ux and therefore no cleaning required. However, uxless processes are extremely difficult to implement and are incompatible with low-cost volume manufacture. To obtain satisfactory wetting and spreading, the process atmosphere needs to be denuded of oxygen and water vapor to levels that can only be obtained in closed vessels. More details on uxless soldering are given in Chapter 3, section 3.3. Some care needs to taken to ensure that a uxless process does leave a clean product after soldering. The authors encountered one process in
which the solder perform was etched in 10% hydrochloric acid and then rinsed in deionized water prior to use. The process was claimed to be uxless because no ux was used, and all the acid was supposed to be rinsed off the solder. However, as the solubility of hydrochloric acid in water decreases faster than does the dilution ratio, in practice the dried solder preform and nal product were both proved to be severely contaminated with chlorides. It transpired that better-quality joints with lower levels of ionic contamination of the product were obtained by using a commercial no-clean ux without cleaning! The ultimate alternative to cleaning is simply not to do it. Cleaning adds to the capital and consumable process cost and results in an assembly yield loss. Provided the ux residues do not adversely impact the reliability of the product, then the merits of cleaning are truly questionable. This is particularly true for short-life consumer products, for example, musical greeting cards. Because the principal risk to leaving ux residues on the product is corrosion arising after the residues react with moisture, another frequent approach is to coat the entire product in lacquer. This provides a sufficient barrier to moisture ingression over the recommended life of the product. Obviously for life- or mission-critical products, cleaning will probably always be undertaken, but in many instances acting in recognition of a nite life is a technically and economically sound approach.
The relative effectiveness of different cleaning agents on glass-reinforced epoxy laminate FR4 printed circuit boards (PCBs) as measured by residual sodium chloride contamination. The reowed bar references as-made and uncleaned PCBs. Adapted from Richards et al. [1993]
Fig. 1.30
42 / Principles of Soldering
1.3.2.12
All processes are subject to variation, and achieving stability of processes is an important step in any quality-improvement program. The application of statistics to monitor and control the variability is termed statistical process control (SPC) [Ledolter and Burrill 1999]. Many industrial soldering processes are subject to SPC. A modern PCB assembly line can achieve joint defect rates of a few ppm. This means, quite literally, that only one or two soldered joints in every million made would fail a quality inspection. This degree of manufacturing quality is only achieved through SPC. A fundamental tool in SPC is a graphical display, known as a control chart. This chart provides the basis for deciding whether the variation in the output of a process is due to common, randomly occurring variations or to unusual causes, which would require investigation and action. The control chart is a chronological plot of particular characteristics, such as joint strength or peak reow temperature, sampled at periodic intervals. This information furnishes data on the process stability and provides an understanding of improvements, where made. Whenever a signicant deviation from the norm is detected, a decision can be made to adjust a process variable in order to bring the output back to the required quality level. Obviously, to accomplish this there must be a proper understanding of the relationship between the process variables and the output. There are different types of control charts, designed for different situations, which are classied by the type of data they contain. Control
charts designed to monitor the proportion of defective items are referred to as p-charts, while charts that track the number of defects on the product are known as c-charts. Both are used to describe attribute data, that is, a record of the presence and absence of certain characteristics. Quantitative data are monitored using a mean or x-bar chart, while process variability is measured using range charts (r-charts) and standard deviation charts (s-charts). The basis of SPC is, for each parameter tracked, to select upper and lower control limits (see Fig. 1.31). These are set at a multiple number of standard deviations from the mean such that there will be a high probability that the data will fall between these limits when the process is working as desired. Only when the process metrics drift further from the mean is intervention required. It is possible to apply SPC to virtually any process or machine output. In many instances, it is a highly effective basis for controlling manufacturing processes. As with any tool it is necessary to use some discretion and critical thought to ensure that SPC is appropriate and the cost of implementing and sustaining it is justied.
1.3.3
Soldering encompasses the use of a large number of different materials, covering metallic and nonmetallic elements for the llers and the parent materials, and organic and inorganic chemi-
Fig. 1.31
Statistical process control chart for peak reow temperature (indicated by solid squares) measured at a test point on a printed circuit board, showing the upper and lower control (i.e., intervention) points for this process
Chapter 1: Introduction / 43
cals used in uxes, controlled atmospheres, and for removing ux residues. Several of these materials are hazardous in varying degrees to the operators or to the environment [Sax and Lewis 1989]. Accordingly, they must be handled, used, and disposed of according to national codes of practices or regulations governing hazardous substances. Official listings produced by national health and safety authorities classify materials according to their toxicity level, for example, the exposure limits for hazardous vapors and dusts. The main problem with solders and soldering uxes arises when they are heated to make a joint. The fume contains a cocktail of gases that can cause eye and nose irritation, dermatitis, asthma, and respiratory problems. The fume contains ne particles, in the range 0.1 to 1 m (4 to 40 in.), which is the most dangerous size distribution for causing long-term lung damage. The recommended solution is to ensure that the workplace or work chamber is ventilated using an appropriately designed extraction system that is able to exhaust the gases and trap the particulates [Jakeway 1994]. Preventing exposure to the hazard by appropriate measures should always be given higher priority than protective measures. All materials that are likely to be encountered in a joining context will have an assigned value of maximum exposure limit, usually in weight per unit volume (normally in mg/m3). The form of the material is also relevant. Powders and dusts are more hazardous than nonvolatile liquids and monolithic solids: they are ranked according to the maximum inhalable quantity in mg/m3, time weighted over a period of time, either short term, meaning minutes, or over a longer period of many hours. For correct interpretation of the rules, regulations, and audits, reference should always be made to a qualied safety practitioner, as there are often also legal aspects to consider. Care must also be taken in the storage of materials both prior to use and in the procedures for the subsequent disposal of residues, exhaust emissions, and other associated effluent, such as solutions containing rinsed uxes. These are usually subject to statutory controls. For many organic chemicals and gases, in this context solder uxes, binders used in pastes, and halogenated gases, there may also be re risks to consider. The ammability is rated according to ashpoint, which is the lowest temperature at which the substance can be spontaneously ignited when it is in a saturated condition.
Appendix A1.1: Solid-State Joining with Gold, Indium, and Solder Constituents
Gold has three desirable characteristics that render it a most suitable metal for making diffusion-bonded joints. These are its low modulus, rapid self-diffusion, and absence of an oxide skin when heated in air. As a result, diffusion bonding of gold can be achieved at room temperature with plastic deformation of as little as 20%. The temperature/pressure curve for a process time of 1 h is given in Fig. 1.32 [Humpston and Baker 1999]. Extrapolation of the graph predicts that a successful gold-gold bond can be achieved without pressure above approximately 450 C (840 F). Certainly it is the authors experience that gold rods merely placed in contact bond readily at 500 C (932 F). Indium can also be used to make pressurewelded joints. Because indium tends to be covered by a relatively thick layer of oxide, the process usually relies on extensive physical deformation to ensure adequate virgin metal contact. Heating is not normally necessary because of the extensive deformation and, at room temperature, the metal is already very close to its melting point. A typical temperature/pressure process curve for indium is given in Fig. 1.33 [Plotner et al. 1991]. Diffusion bonding and pressure welding can, perhaps surprisingly, be achieved using standard solders. The authors have witnessed one highvolume manufacturing line where a hermetic seal was made between two gold metallized surfaces using a wire of ordinary (uxless) solder. Two
Temperature/pressure curve for diffusion bonding of gold, for a process time of 1 h. The line on the graph differentiates between joints of acceptable (above) and unsatisfactory (below) tensile joint strength after fabrication.
Fig. 1.32
44 / Principles of Soldering
Fig. 1.33
1 V h (h2 3A2) 6
solders used in this manner include Sn-40Pb and In-48Sn. In both instances, the trick to obtaining a satisfactory joint lies in ensuring that the faying surfaces of both the components and the solder are as free from nonmetallic contamination as possible. The joint was made by forming a length of freshly cleaned solder wire into a ring, butt welding the ends, and simply pressing it between the two component surfaces, at room temperature. Sufficient load was applied to cause ow of the solder out to the joint edge and a substantial narrowing of the joint gap; so the process was clearly pressure welding. The resulting joint provided a key seal on a hermetic cavity, and the product carried a 20-year guarantee against hermeticity failure by the manufacturer. The merit of using solder preforms for this application is that they are readily available at low cost compared with specially deposited coatings of gold or indium.
A and a are related by the conservation of the volume of the droplet, that is:
4 1 V a3 h (h2 3A2) 3 6
Therefore:
a 1 2
h(h2 3A2)1/3
Appendix A1.2: Relationship among Spread Ratio, Spread Factor, and Contact Angle of Droplets
Expressions describing the spread of a molten metal droplet are derived under the following set of idealized conditions:
Chapter 1: Introduction / 45
and
4A
2
R Sr
A2 h2 2h
Therefore:
sin REFERENCES 2
(A/h) (h/A)
(1 3A /h )
2
4 cot2 /2
Sf
Dh D
A (R h R
2
)2
ham, A.M., 1992. Kinetics of Braze Spreading, Proc. Conf. British Association for Brazing and Soldering, 1992 Autumn Conference, Coventry, U.K. Ambrose, J.C., Nicholas, M.G., and Stoneham, A.M., 1993. Dynamics of Liquid Drop Spreading in Metal-Metal Systems, Acta Metall. Mater., Vol 41 (No. 8), p 23952401 Andrade, E.N.C, 1952. The Viscosity of Liquids, Proc. Royal Soc., A, Vol 215, 36/356 Bakulin, S.S., Shorshorov, M.Kh., and Shapiro, A.E., 1992. A Thermodynamic Approach to Optimising the Width of the Brazing Gap and the Amount of Brazing Alloy, Weld. Int., Vol 6 (No. 6), p 473475 Bever, M.B., Ed., 1986. Encyclopedia of Materials Science and Engineering, Pergamon Press, p 24582463 Brandon, D.G. and Kaplan, W.D., 1997. Joining Processes: An Introduction, John Wiley & Sons de Gennes, P.G., 1985. Wetting: Statistics and Dynamics, Rev. Mod. Phys., Vol 57 (No. 3), p 827863 Eustathopoulos, N. and Coudurier, L., 1979. Wettability and Thermodynamic Properties of Interfaces in Metallic Systems, Proc. Conf. British Association for Brazing and Soldering, Third International Conference, London, paper 5 Feature, 1976. Diffusion BondingTomorrows Low-Cost Fabrication Tool, Met. Mater., Feb, p 3739 Fleetwood, M.J., et al., 1988. Control of Thin Strip Casting, Proc. Second International Conf. Rapidly Solidied Materials, 79 March (San Diego, CA) Frear, D.R., Jones, W.B., and Kinsman, K.R., 1990. Solder MechanicsA State of the Art Assessment, TMS Funk, E.R. and Udin, H., 1952. Brazing Hydromechanics, Weld. Res. Suppl., Vol 6, p 310s316s
46 / Principles of Soldering
Chapter 1: Introduction / 47
Strength of Cast Iron, J. Mater. Sci., Vol 33, p 2328 Stromswold, E.I., Pratt, R.E., and Quesnel, D.J., 1994. The Effect of Substrate Roughness on the Fracture Toughness of Cu/96.5Sn3.5Ag Solder Joints, J. Electron. Mater., Vol 23 (No. 10), p 10471053 Sun, W. and Ivey, D.G., 2001. Microstructural Study of Co-Electroplated Au/Sn Alloys, J. Mater. Sci., Vol 36, p 757766 Tabor, D., 1969. Gases, Liquids and Solids, Penguin Books Tunca, N., Delamore, G.W., and Smith, R.W., 1990. Corrosion of Mo, Nb, Cr, and Y in Molten Aluminium, Metall. Trans. A, Vol 21A (No. 11), p 29192928 Tylecote, R.F., 1967. Diffusion Bonding, Part l, Weld. Met. Fabr., Vol 35 (No. 12), p 483 489 Tylecote, R.F., 1968. The Solid Phase Welding of Metals, Edward Arnold Vianco, P.T. and Rejent, J.A., 1997. Capillary Flow Solder Wettability Test, Solder. Surf. Mt. Technol., Vol 25 (No. 2), p 47 Wang, X.H. and Conrad, H., 1995. Kinetics of Wetting Ag and Cu Substrates by Molten
60Sn40Pb, Metall. Mater. Trans., Vol 26A, p 459469 Watkins, M., Johnson, K.L., and Breyer, N.N., 1975. Effect of Cold Work on LiquidMetal Embrittlement by Pb Alloys on 4145 Steel, Proc. Conf. Fourth Inter-American Conference on Materials Technologies (Caracas), p 3138 Weeks, J.R. and Gurinsky, D.H., 1958. Liquid Metals and Solidication, American Society for Metals Xian, A., 2000. Thermodynamic Discussion on Youngs Equation in Wetting, Z. Metallkde., Vol 91 (No. 4), p 316322 Yost, F.G., Hosking, F.M., and Frear, D.R., 1993. The Mechanics of Solder Alloy Wetting and Spreading, Van Nostrand Reinhold Yost, F.G., Michael, J.R., and Eisenmann, E.T., 1995. Extensive Wetting due to Roughness, Acta Metall. Mater., Vol 43 (No. 3), p 299305 Yost, F.G. and Romig, A.D., 1988. Thermodynamics of Wetting by Liquid Metals, Materials Research Society Symposium Proceedings, Vol 108, p 385390
CHAPTER 2
50 / Principles of Soldering
must be mutually suited. For example, solders containing zinc, lead, or other volatile constituents are not usually appropriate for furnace joining at elevated temperature, especially when these involve reduced pressures. The degree of temperature uniformity that can be achieved over the joint area will have an inuence in determining the minimum temperature difference that can be tolerated between the melting temperature of the ller and the melting or degradation temperature of the parent material. This consideration is particularly relevant to the joining of aluminum and stainless steel components, for contrary reasons. Stainless steel possesses relatively poor thermal conductivity, whereas aluminum has a low thermal heat capacity, both of which make it difficult to obtain uniform heating throughout a bulk assembly using local heat sources that do not encompass the workpiece. The condition of the surface of the parent material may affect its compatibility with the solder, especially when uxes are not used. As an obvious example, ller metals will less readily wet an oxidized surface than a freshly cleaned metal surface. This consideration often determines the acceptable shelf life of components prior to joining. The term atomically clean should not generally be used to describe cleaned surfaces. It is an abstract ideal, not normally relevant in practical situations! In order to establish whether a particular parent metal (or nonmetal with a surface metallic coating) is compatible with a given solder, it is necessary for the appraisal to be carried out under conditions that are likely to be representative of those used in any practical implementation of the joining process. Parameters such as process time and temperature can be critical in this regard. Storage shelf life of the components (and the ller when in the form of paste) is another relevant factor that needs to be taken into account but is often neglected during transfer of a process from the laboratory to the factory. This is particularly true where the parent metal has a multilayer metallic coating, a situation that is commonly met with in soldering. Although the outermost layer, often gold, will have indenite shelf life, if it is not fully dense or sufficiently thick, oxygen will be able to percolate through it and oxidize less noble metals beneath the gold layer. The problem then only becomes apparent when the gold dissolves in the solder, exposing the underlying metal that has been rendered nonwettable during storage.
The properties of the solder and the resulting joint must also be compatible with the service requirements. These are likely to involve a combination of at least some of the following considerations:
The simultaneous attainment of several of the desired characteristics is frequently achievable with common ller metals, provided the basic
design guidelines and process conditions are satised. Lead-tin eutectic solder, which is taken to include the Pb-60Sn alloy, accounts for the largest proportion of joints involving ller metals that are encountered. It is the more exceptional and demanding service requirements that have given rise to the development of the hundreds of additional ller metal compositions. However, in general, soldering is not a difficult process to master. The commercially available llers and matching uxes have been designed so that, when used in conjunction with the common engineering parent materials, they meet many if not all of the requirements listed previously.
2.1
The survey begins with consideration of leadtin solder on account of its preeminent position among the lower-melting-point ller metals. The review then extends to consider the less common solders, arranged in rough order of usage and decreasing melting point. To readers familiar with brazing technology, one important difference of solders will soon be apparent. That is, solders are largely based on unique eutectic compositions. The addition of other alloying elements usually destroys the eutectiferous character of these alloys and the desirable soldering properties that stem from it. Thus, the alloy compositions are normally exact rather than encompassing a broad range, an important exception being the indium-lead series of alloys, which are not eutectiferous. In any case, there is little incentive to introduce such modications because there is a ready choice of alternative eutectic solders having similar melting points, as illustrated by Fig. 1.3. For detailed coverage of the available solder alloys, the reader should consult reference publications [e.g., Klein Wassink 1989, Manko 2002, Schwartz 2003] as well as suppliers data sheets and manuals. All solder alloys are based on combinations of two or more constituents, chosen from nine common elements, namely antimony, bismuth, cadmium, gold, indium, lead, silver, tin, and zinc. Of these, the cadmium-containing alloys have largely been removed from manufacturers catalogs because they offer no clear advantages over other solders, and their use is subject to restrictions arising from the toxicity of cadmium fumes. As with beryllium, it transpires that cadmium metal is relatively safe, and it is the oxide that is
hazardous to health. In recent years, the use of lead-containing solders for joining domestic water pipes has been largely banned. This has led to the development of lead-free solders for plumbing applications [Irving 1992]. These solders comprise typically 95% tin with the balance being one or more of silver, antimony, bismuth copper, nickel, and zinc. Solders that do not contain lead are also actively being used in electronics manufacture as part of a portfolio of measures by companies toward environmental responsibility. Further information on this initiative is given in Chapter 5, section 5.1. High-melting-point solders contain either gold or lead as the principal constituents. The majority of lower-melting-point solders, that is, those with melting points below 150 C (300 F), fall into two groups depending on whether the active constituent of the alloy is tin or indium. This classication is determined on the basis of the likely reaction of the solder with the substrate material, which, in the temperature range covered by solders, is necessarily a metal because there is insufficient thermal activation available for wetting nonmetals. In most cases, wetting of a component by a solder results in the formation of intermetallic compounds, either within the ller or at the interface between the solder and the parent material. These intermetallic phases have a pronounced effect on the mechanical properties of the joint. As constituents of solders, indium and tin have a dominant role in determining the intermetallic compounds that form by reaction with the components: the compounds invariably contain either indium or tin. When both indium and tin are present, the composition of the parent materials (i.e., the material on the surface of the components) determines which of these elements is the predominant constituent of the resulting intermetallic compounds. For solder alloys that contain neither element, intermetallic compounds are not always formed and those that do necessarily depend on the composition of both the solder and of the parent materials. The mechanism by which these intermetallics form and some of the implications of their presence in joints are discussed in section 2.3 in this chapter. Each of the major elements in the lowest melting point soldersbismuth, indium, lead, and tinconfers different properties to the ller metals, often in a manner that is not obvious. For example, the fact that solders tend to be based on eutectic composition alloys means that their melting points are determined by the eutectic reaction
52 / Principles of Soldering
rather than by the melting points of the individual elements present. There is no simple relationship between the melting temperatures of the constituent elements and the eutectic temperature. Therefore, while the elements fall into the following sequence, lead > bismuth > tin > indium, when ranked in descending order of melting point, the solder alloys do not fall into this pattern; those containing bismuth have the lowest melting points of all. Despite such complexities, it is possible to make some generalizations about the role of each element in a solder. To this list is also appended antimony, not that there are common solders based on this element, but because it is sometimes a minor constituent in other solder alloy families. Tin is a preferred constituent of many solder alloys because it confers uidity, benets wetting, enhances mechanical and physical properties, and possesses exceptionally low vapor pressure.After lead, cadmium, and zinc, tin is the least expensive ingredient of solder alloys. Nevertheless, the price of tin is still some 30 times that of lead, which accounts for the popularity of lead-rich lead-tin solders. By way of comparison, bismuth and antimonyareaboutthesamepriceastin,withindium and silver being over an order of magnitude more expensive still. However, tin-bearing solders tend
to form brittle compounds on reaction with many parent materials and metallizations used in engineering, particularly copper and gold. Silver is one of the few exceptions, with silver-tin intermetallic phases being comparatively ductile. Therefore, considerable care must be taken in designing the joining process and specifying the service environment so as to restrict the formation of intermetallic compounds to concentrations below those that would otherwise weaken and embrittle the joints. This point is discussed in connection with the phase diagrams of the relevant alloy systems in section 2.3 in this chapter. Indium and lead are the two softest and most ductile constituents of solder alloys. Despite their inferior mechanical properties, solders with high lead concentrations nd wide application because they are the least expensive and the easiest to use of the high-melting-point solder alloys available. The indium-bearing solders are particularly attractive for use with gold metallizations because these are not readily dissolved and the interfacial phases that form are comparatively ductile, so that joints are not embrittled by their presence. The low level of gold erosion stems from a combination of the steep slope of the liquidus line on the phase diagram between indium and gold (see Fig. 2.1) and the formation
Fig. 2.1
of a thin, continuous intermetallic compound AuIn2 between the molten solder and the gold metallization. This layer of compound then acts as a barrier against signicant further gold dissolution from taking place and results in the prole of the erosion curves shown in Fig. 2.2. Thus, indium-containing solders can be reliably used in conjunction with very thin gold metallizations. Silver is a constituent of several solder alloys, but only as a minor proportion, not simply on account of its price premium, but because higher concentrations (more than about 5%) result in a sudden increase in the liquidus temperature toward those of the silver-bearing brazes. Small additions of silver are used primarily to enhance mechanical properties of solders and joints and to promote uidity by destabilizing native surface oxides on the molten solder. Silver oxide is not stable in air above 190 C (374 F). Unfortunately, however, owing to industrial pollution silver tarnish often contains sulde, which is far more stable against thermal degradation. The Ag96Sn solder has among the best mechanical and physical properties of any low-melting-point solder alloy [Harada and Satoh 1990]. Silvercontaining solders tend to be preferred when joining silver-coated components because the presence of silver in the solder reduces the rate and extent of scavenging from the metallization, as shown by the data in Fig. 2.3. Bismuth is the most brittle constituent of the common solders and, for this reason, few solders contain more than 50% of this element. Bismuthbearing alloys comprise the majority of the lowest melting point solders, as can be seen from Fig. 1.3. Bismuth exhibits the unusual characteristic of expanding on freezing, enabling solders to be tailored to have essentially zero liquid-
to-solid volume contraction by appropriately adjusting the bismuth concentration. It has been claimed that this property can confer benets in making hermetic soldered joints [Dogra 1985]. Although alloys such as the bismuth-tin eutectic solder can be prepared in a manner that renders it soft and ductile, by rapid solidication, it will subsequently embrittle, even at room temperature, owing to changes in the atomic lattice spacing of the bismuth phase that is responsible for the solid-state expansion [Hare, Corwin, and Reimer 1985]. The mechanical properties of bismuth-tin solders may be improved considerably by the addition of 0.5% Ag. This low concentration of silver does not appreciably affect the melting range of the solder, but acts as a very efficient grain rener that improves the tensile ductility of the solder by a factor of three and reduces the susceptibility of the solder to strainrate-dependent deformation behavior [McCormack et al. 1997]. The low melting point and inferior uidity of the bismuthbearing solders impose constraints on the joint design and processing conditions. For example, their relatively low melting temperature means that aggressive inorganic uxes are needed to chemically clean the surfaces of the parent materials. Antimony is often found as a deliberate minor addition (<10%) in many solders, particularly lead-tin alloys. This is because the addition improves some key mechanical properties. In leadtin solder, antimony causes solid-solution
Erosion of a gold metallization by molten indium as a function of reaction time and temperature. Similar results are obtained for indium-base solders, including goldindium, silver-indium, indium-lead, and indium-tin.
Fig. 2.2
Substantial reduction of the dissolution rate of silver in lead-tin eutectic composition solder obtained by small additions of silver to the alloy. Adapted from Bulwith and Mackay [1985]
Fig. 2.3
54 / Principles of Soldering
strengthening, up to a concentration of about 3% (see Fig. 2.4). As the proportion of antimony increases, SbSn cuboids form in the solder and, while this produces further improvements in properties such as creep, these benets are offset by a tendency toward brittleness that is responsible for high and undesirable scatter on reliability data [Tomlinson and Bryan 1986]. Gold is the most expensive major constituent of solders and, for that reason, the applications of gold solders tend to be limited to high-value electronics, photonics, and jewelry manufacture. One of the chief attractions of these solders is their melting point, which falls within the 300 to 500 C (570 to 930 F) gap that separates the upper limit of the lead-base solders from the lower limit of the available aluminum-bearing brazes. Zinc forms a eutectic alloy with aluminum at the composition, Al-94Zn (melting point 381 C, or 718 F). It is the only low-melting-point metal that does not grossly enhance the corrosion of aluminum alloys because its electrode potential is close to that of aluminum, as indicated by the data in Table 2.1. Accordingly, zinc-aluminum solders nd use in the joining of aluminum alloys. However, the refractory nature of aluminum oxide requires the use of aggressive uxes to promote wetting by zinc-base solders, and this creates other problems, namely highly corrosive ux residues. This outline of solder families is brief and fairly generalized. Many solder alloys have unique properties, conrmed in numerous studies, that show that there is often no regular pattern among solder composition, melting point, and joint properties. Indeed, the ranking order for even simple mechanical tests can be radically changed merely by altering the test conditions used [Tomlinson and Fullylove 1992].
A list of some of the more common binary, ternary, and quaternary composition solder alloys is given in Table 2.2, and the associated phase diagrams of those in widest use are shown in Fig. 2.5 to 2.15. Most of these alloys are ductile and can be mechanically worked to produce preforms of virtually any desired geometry. The majority of solders are based on eutectic compositions, for reasons that are explained in section 2.3 in this chapter. Many solder manufacturers also offer off-eutectic compositions in these and other alloy systems. They are included in the product range because they are either less expensive, easier to fabricate as wire and foil, or have a melting range, which is sometimes a technical advantage. It is usually the case that uidity and mechanical properties suffer on moving to an off-eutectic composition, but sometimes these characteristics are desirable. For example, the high-lead solders are expressly used below their liquidus temperature for the purpose of bridging wide gaps. On the other hand, the indium-lead alloys offer a continuum of solder compositions with intermediate melting temperatures. Although not included in Table 2.2, reference has been noted in the literature to one other highmelting-point solder. Details are sparse, but it appears to be a quaternary alloy based on the copper-silver eutectic braze with substantial additions of indium and tin to depress the liquidus and solidus temperatures. The melting point of the solder is claimed to be 450 C (842 F), and, therefore, it technically qualies as a solder [Hirakawa, Tanahashi, and Terasawa 1995]. Further details on conventional brazes based on copper-silver and other alloys can be found in the planned companion publication Principles of Brazing.
Fig. 2.4
Shearstrength of -brass joints made with lead-tinsolder containing varying concentrations of antimony
Gold Silver Copper Hydrogen Lead Tin Nickel Cadmium Iron Zinc Silicon Aluminum Magnesium
1.50 0.80 0.34 0.00 0.13 0.14 0.25 0.40 0.44 0.74 1.30 1.66 2.37
Table 2.2
Ag
... ... ... ... ... ... ... 3.0 ... 5.0 ... ... 1.5 ... ... ... ... 3.5 ... ... ... 25.0 ... ... 2.5 2.5 1.5 ...
49.0 33.7 52.0 67.0 ... 55.5 57.0 ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
21.0 66.3 ... 33.0 51.0 ... ... 97.0 96.0 80.0 99.5 75.0 ... ... 50.0 ... ... ... ... ... ... ... ... 25.0 5.0 ... ... ...
18.0 ... 32.0 ... ... 44.5 ... ... ... 15.0 ... 25.0 36.0 38.0 50.0 ... 85.0 ... ... ... 85.0 ... 88.9 75.0 92.5 97.5 97.5 99.5
12.0 ... 16.0 ... 49.0 ... 43.0 ... ... ... ... ... 62.5 62.0 ... 91.0 ... 96.5 99.3 95.0 3.5 65.0 ... ... ... ... 1.0 ...
... ... ... ... ... ... ... ... 4.0Zn ... 0.5Au ... ... ... ... 9.0Zn 15.0Au ... 0.7Cu 5.0Sb 11.5Sb 10.0Sb 11.1Sb ... ... ... ... 0.5Zn
57 72 95 109 120 123 139 144 144 142149(a) 156 165170(a) 179 183 178210(a) 198 215 221 227 235245(a) 240 232234(a) 251 255265(a) 300310(a) 304 309310(a) 318
135 162 203 228 248 253 282 291 291 288300(a) 313 330340(a) 354 361 352410(a) 388 419 430 441 455475(a) 464 450453(a) 484 490510(a) 572590(a) 579 588590(a) 604
(a) These alloys are not eutectic compositions, but have been included on account of their narrow melting range and their industrial exploitation.
Fig. 2.5
56 / Principles of Soldering
2.1.1
Lead-Tin Solders
This survey necessarily starts with lead-tin solders because they account for about 95% of all soldered joints. Soldered joints are made on a vast, global scale. It is estimated that more than 60,000 tonnes per annum (6 107 kg/yr or 54,000 tons/yr) of lead-tin solder are used in the electronics assembly industry alone. The benecial characteristics of lead-tin alloys have been appreciated since at least Roman times, and the elder Pliny, writing in the rst century A.D., in his Historia Naturalis [Rackham 1952] specically mentions an alloy containing two parts of black lead (modern lead) and one part of white lead (i.e., tin) used for soldering pipes. He also remarks that the price of this alloy is 20 denarii per pound. It is interesting to note that this works out at roughly $100/kg, assuming that gold has maintained its purchasing power since Plinys day. Todays price for the same solder is lower by an order of magnitude. Lead-tin alloys offer the following advantages compared with other solders:
Superior wetting and spreading characteristics, compared with most other solders, especially those containing bismuth, antimony, and indium. The difference is quantied in Fig. 1.14.
Fig. 2.6
In most other respects, such as corrosion resistance and electrical and thermal conductivity, there is little to distinguish lead-tin alloys from the majority of other low-melting-point solders (see Table 5.19 in Chapter 5, section 5.7). Lead-tin solders are seldom pure, but often contain other elements, either as incidental impurities or as deliberate additions designed to modify specic properties. Three metal additions are routinely encountered; antimony, bismuth, and silver. Antimony is frequently present to a maximum concentration of about 1%. It has a benecial role in improving the mechanical properties of joints though a solid-solution-strengthening mechanism.Antimony can substitute for double the proportion of tin, without greatly widening the melting range of the solder and, therefore, is often favored by solder manufacturers as it decreases materials costs. At higher concentrations of antimony, there are solder alloys based on the ternary system Pb-Sb-Sn. The liquidus projection of this alloy system is given in Fig. 2.13, from which it can be seen that there is a ternary eutectic at 240 C containing 11.5% Sb. Bismuth is added for similar functional reasons as antimony. The maximum concentration is usually limited to about 3%, as higher levels
result in a signicant widening of the melting range and a noticeable impairment to wetting and spreading behavior. In addition, there is a range of low-melting-point ternary solders based on Bi-Pb-Sn alloys in which bismuth is present in high concentrations. The liquidus surface of this alloy system is given in Fig. 2.16. Small quantities of bismuth and also antimony are added to lead-tin solders to prevent degradation at low temperatures through a mechanism known as tin pest. Further details of this failure mechanism are given in section 2.2 in this chapter. Silver. Lead-silver-tin solders nd application for soldering silver-coated surfaces. The addition of 2% Ag to the lead-tin eutectic alloy has a marked effect in reducing the erosion of silver coatings, as shown by the data in Fig. 2.3. The presence of silver in lead-tin solders results in the formation of a ne dispersion of the relatively ductile intermetallic compound Ag3Sn, on solidication, which boosts the mechanical properties of joints. Small quantities of copper and gold have a similar benecial effect, as described in section 2.2 in this chapter. However, these elements are not usually present as deliberate alloying additions since they are incorporated automatically when soldering copper- and goldcoated components, respectively.
Fig. 2.7
58 / Principles of Soldering
All of the aforementioned additions to lead-tin solder, in small quantities, enhance the strength of joints to copper because they alloy preferen-
tially with some of the tin and thereby serve to reduce the thickness of the copper-tin intermetallic phases. These form at copper-solder inter-
Fig. 2.8
Fig. 2.9
faces and can be a source of mechanical weakness (see section 2.3 in this chapter) [Quan et al. 1987].
Attempts have been made to improve the mechanical properties and resistance to failure of lead-tin solders through creep and fatigue by a
Fig. 2.10
Fig. 2.11
60 / Principles of Soldering
number of strengthening mechanisms. These novel alloys are discussed in Chapter 5, sections 5.5, 5.6, and 5.8.
2.1.2
Two tin-base solders nd wide industrial application. These are silver-tin eutectic and a AgSb-Sn solder often referred to as Alloy J. Silver-Tin. Silver-tin is a eutectic alloy that has a xed point generally accepted to be 221 C. The composition of the eutectic is close to Ag96.5Sn. Alloys offered by manufacturers will deviate from this by up to 1.5% on account of cost, but this results in a widening of the melting range. The silver-tin eutectic forms the basis of virtually all of the lead-free alloys offered as replacements for lead-tin solder (see Chapter 5, section 5.1). In many respects, this solder is the closest binary alloy to lead-tin solder. The melting points are not greatly different, and their wetting and spreading characteristics are comparable to those of lead-tin. So are their mechanical properties, although silver-tin alloys are generally slightly superior in this respect for service temperatures below 150 C (300 F). Both solders form closely related intermetallic compounds on wetting of the common engineering parent materials and
metallizations. The hue and luster of solder llets are visually indistinguishable, but the surfaces of lead-tin solder llets are often smoother as this solder contains a more even ratio of the two constituents and hence a ner grain size. Alloy J. The majority of solders are referred to in manufacturers catalogs by a name or terminology that gives a clue to the alloy composition, although part of the product range of one well-known solder manufacturer is denoted by a numbering system that can only be decoded with aid of a reference card. More recently, it has been noted that, with the advent of lead-free solders, marketing departments have had some role in the naming of new products. These names are designed to convey an impression of green and environmentally friendly products rather than technical information about the product. One of the cryptic solder names in the literature is Alloy J, and particular reference is made to this relatively prominent alloy here to elucidate its key features. Alloy J has the composition 25Ag-65Sn10Sb. It has a narrow melting range, usually cited simply as 233 C (451 F) [Olsen and Spanjer 1981, Mackay and Levine 1986]. Containing 65% Sn, Alloy J qualies as a tin-base solder. The alloy was originally developed as an alter-
Fig. 2.12
native solder for silicon semiconductor dieattach with ostensibly more favorable mechanical properties, melting range, and price than the gold-tin and gold-silicon solders. The name derives from the systematic search of the Ag-Sb-Sn system that was undertaken for suitable candidatesAlloy J was the tenth alloy composition evaluated and the rst one found largely to meet the target requirements, that of substantially superior performance compared to lead-tin eutectic under conditions of power cycling of semiconductor die, as shown in Fig. 2.17 [Mackay and Levine 1986]. Reference to the ternary phase diagram for this alloy system (Fig. 2.15) shows that the 25Ag65Sn-10Sb solder is a three-phase mixture of tin, Ag3Sn, and SbSn. The two intermetallic compound constituents are relatively hard and brittle. When produced and used conventionally, the presence of these intermetallics makes the fabrication into convenient forms such as wire and preforms extremely difficult and is detrimental to the joint properties. This problem has been resolved through the production of the alloy using rapid-solidication technology (see Chapter 1, section 1.3.2.2). The rapid-solidication process yields an alloy with a substantially ner micro-
structure with a grain size below 0.3 m (12 in.) compared with 10 m (400 in.) when produced by conventional ingot casting. The rapidsolidication process also extends the solid solubilities of both silver and antimony in tin so that the volumetric proportion of intermetallic phase in the alloy is correspondingly decreased. These two features result in improved ductility and a reduction in hardness. Provided the soldering cycle is short and the subsequent service temperature is not excessive, the benets of a rened grain size persists in joints made using solder prepared by rapidsolidication processing. The reasons for this are not fully understood, but the net effect is that silicon die soldered to ceramic packages using rapidly solidied Alloy J have improved resistance to fatigue failure on power cycling over solder produced by conventional casting, as indicated by the data presented in Fig. 2.18 [Pinamaneni and Solomon 1986].
2.1.3
Zinc-Bearing Solders
One of the major applications of solders containing zinc is for joining aluminum and its alloys [Finch 1985]. This metal plays an additional
Fig. 2.13
62 / Principles of Soldering
role in helping to disrupt alumina surface layers and so helps to promote wetting by the solder. A ux-cored, zinc-base solder that is intended for joining aluminum is available commercially [Rubin 1982]. Such joints are obviously susceptible to corrosion and tend to be used for making electrical connections to functional aluminum parts rather than structural assemblies. The key to making a successful soldered joint to aluminum is the formulation of the ux used. This topic is discussed in Chapter 3, section 3.2.2.1. When soldering to aluminum parts, a frequently overlooked consideration is the combination of high thermal conductivity and expansivity of this metal, coupled with low heat capacity. Therefore, particular attention needs to be paid to the heating method to minimize thermal gradients and thermally induced distortion. One of the zinc-bearing solders that is widely used for joining aluminum is the eutectic composition alloy Al-94Zn (melting point is 381 C,
or 718 F). The aluminum-zinc phase diagram is given in Fig. 2.19. Often a few ppm of gallium is added to the solder to aid wetting, molten gallium being one of the few metals that will not ball up on alumina, which is prevalent on the surface of aluminum alloys. Spreading of the solder is promoted by a reaction between gallium oxide and alumina, rather than between the respective metals. The relatively small depression in the melting temperature of aluminum achieved by alloying with a large proportion of zinc has one negative consequence, namely the occurrence of signicant erosion of the aluminum components in the region of the joint. This can be a problem in situations where it is required that component geometry is not perceptibly changed by the joining operation. Zinc-bearing alloys are not suitable for joining processes that entail reduced pressure atmospheres, owing to the inherent volatility of this element.
Fig. 2.14
Liquidus surface of the Ag-Cu-Sn system. Adapted from Petzow and Effenberg [1988]
All of the zinc-bearing solders contain high percentages of zinc, generally in the range 75 to 95%, and their solidus temperatures cover the temperature band 197 to 419 C (387 to 786 F). The alloying elements present in signicant percentages are aluminum, cadmium, copper, and tin. A representative list of these solders is given in Table 2.3. Pure zinc is not used as a ller metal because it tends to ball up on heating when used in normal atmospheres, whereas when small amounts of other elements are added, wetting of surfaces is more readily achieved. The cadmiumcontaining alloy listed in Table 2.3 is now generally avoided because cadmium fume is classed as a health hazard. The choice of alloying additions is made on grounds of reducing the erosion of the substrate metals and, as far as possible, of improving the spreading and owing characteristics. In general, the additions cannot be entertained as melting-point depressants because they mostly do not reduce the liquidus temperature, only the solidus temperature. Over the years, considerable research has been devoted to developing improved new solders
based on zinc [Harrison and Knights 1984] because these alloys have the following potential attractions:
They can be used at sufficiently low temperatures so as not to destroy the workhardened strength of the copper alloys. Yet, at the same time, the joints that are produced are typically two to four times stronger than those obtained with the common solders. The zinc solders are compatible with galvanized steel components. They are inexpensive in relation to most solders, being approximately one-quarter the cost of the lead-tin alloys, for a given volume of ller metal. They are lightweight, having approximately half the density of lead-tin eutectic solder. They possess high thermal conductivity, exceeding 100 W/m K (60 Btu/h ft F). They are not hazardous: unlike cadmium and lead, the effects of small amounts of zinc absorbed in the human body are noncumulative and temporary. Zinc is one of the vital trace elements in the human diet (as are many
Fig. 2.15
Liquidus surface of the Ag-Sb-Sn system. Adapted from Petzow and Effenberg [1988]
64 / Principles of Soldering
elements that are toxic in larger concentrations). The factors that have greatly limited the adoption of the zinc solders include:
The potential strength of joints to copper alloys are compromised by the presence of embrittling copper-zinc compounds, which form as interfacial phases. For this reason, the zincbase solders that have been formulated for use with copper alloys often contain numerous minor alloying additions to modify the growth of these phases.
Fig. 2.16
Fig. 2.17
Degradation of joint quality, as measured by through-thickness electrical resistivity for silicon semiconductor die attached joints using three different solders and subject to power cycling
zinc, which is of the order of 2 to 3 m/min (80 to 120 in./min) at 400 C (750 F), in air, coupled with high surface tension and viscosity (both approximately double those of lead-tin alloys), are the main factors. The high vapor pressure of zinc means that these solders cannot be used in reduced pressure atmospheres because the volatility of
this element exacerbates the formation of voids in joints and contaminates furnace equipment. The high affinity of zinc for oxygen requires the use of aggressive uxes when soldering in ambient atmospheres which, in turn, leads to cleaning and corrosion problems. The ux efficacy is highly specic to the particular
Cumulative failure data for TO-220 silicon semiconductor die, subject to power cycling. Joints were fabricated using preforms prepared using conventional casting (dark bars in chart) and mechanical working or rapid solidication technology (white bars on chart). Adapted from Pinamaneni and Solomon [1986]
Fig. 2.18
Fig. 2.19
66 / Principles of Soldering
parent metal/solder combination on which it is used. This makes it extremely difficult to achieve tolerant soldering processes using zinc-base ller metals, although consistent results do appear to have been achieved with nickel and copper substrates with uxes based on L-glutamic acid and dimethylammonium chloride using process temperatures as low as 250 C (480 F). For many situations, the disadvantages of zincbearing solders outweigh their prospective benets, which accounts for the fact that they have not found much favor in industry. Nevertheless, further development continues. One example is the development of uxes based on organic tin compounds. These chemicals are designed to decompose at zinc soldering process temperatures and coat the substrate and ller metal with a thin lm of tin. By this means, contact angles as low as 20 can be obtained on copper substrates when soldered in air using tin-zinc eutectic solder [Vaynman and Fine 2000]. Another example is the development of zinc-base solders as leadfree substitutes for the high-lead family of ller metals. In particular, an alloy of composition Al-3Ga-3Mg-90Zn, which has a melting range of 309 to 347 C (588 to 657 F), has been proposed as a hard solder for die-attach applications in the electronics industry [Shimizu et al. 1999]. Flux pastes of zinc-base solders have also been developed [Noguchi 1999].
2.1.4
Gold-Bearing Solders
Gold is unusual in that it is the only element on which both brazes and solders are based; that is, this element is the major constituent of both solders and brazes. Further details about goldbase brazes are given in the planned companion volume Principles of Brazing; also discussed is the use of gold solders for jewelry applications. The gold-bearing solders are all gold-rich alloys of eutectic composition and have melting points between 278 and 363 C (532 and 685 F). They are listed in Table 2.4 and their associated phase diagrams are given in Fig. 2.20 to 2.23. In view of their high cost, the applications of these alloys tend to be limited and specialized. One of the chief attractions of these solders is their melting point, which falls within the 300 to 500 C (570 to 930 F) gap that separates the upper limit of the lead-base solders from the lower limit of the available aluminum-bearing brazes.
Gold-bearing solders have the advantage of being suitable for joining to gold-metallized components. Their compatibility with gold metallizations is explained more fully later in this section. Three of these solders (the gold-antimony alloys being the exception) are used widely in the electronics industry as high-temperature solders for attaching semiconductor devices into packages and building hermetic enclosures for sensitive compound semiconductors and optical devices. Gold-antimony alloys are brittle, even when prepared using rapid-solidication technology and, as far as the authors are aware, are not much used as ller metals. The principal gold-base solders are considered briey in the paragraphs that follow. Gold-Silicon and Gold-Germanium. On account of their closely similar characteristics as solders, these alloys may be considered together. Gold-silicon alloys are primarily used in the form of a foil preform for bonding silicon semiconductor chips to gold-metallized pads in ceramic packages. The alloy compositions used as solders are slightly gold-rich with respect to the eutectic composition, generally at or close to Au-2wt%Si. This is deliberate because the eutectic alloy is too hard and brittle even to hot roll to foil. By making the alloy gold-rich, the proportion of the ductile gold phase in the microstructure is sufficient to improve the mechanical properties to tolerable limits. Rapid solidication is unable to produce ductile foil because rapid cooling of molten gold-silicon alloys results in the formation of a large volume fraction of (metastable) gold-silicon intermetallic compounds, principally Au3Si, which render the foil too brittle to handle [Johnson and Johnson 1983]. An alternative method of applying the solder to silicon components is simply to coat the back surface of the silicon die with a thin layer of gold, applied by a vapor-phase technique. On heating the gold-metallized silicon to above 363 C (685 F), the resulting interdiffusion between the gold and silicon generates liquid ller metal in situ. The Au-2Si solder is an exception to the rule of eutectic composition alloys, possessing favorable characteristics as ller metals. In particular, it suffers from high viscosity when molten. This characteristic is a direct consequence of the low temperature of the eutectic transformation, relative to the high melting points of both constituent phases, together with the silaceous dross that tends to form on the surface of the alloy and the absence of suitable uxes that might enhance wetting. This can be seen by comparing
the spreading behavior of the Au-2Si alloy, as a function of temperature above its melting point, shown in Fig. 2.24 with similar data for other
Table 2.3
Solder composition
Zn-6Al Zn-5Al-2Ag-1Ni Zn-7Al-4Cu Zn-10Cd Zn-25Cd Zn-1Cu Zn-30Cu-3Sb-1Ag Zn-10Sn-1Pb Zn-70Sn Zn-2Ni
381 381387 379390 266399 266370 418424 416424 197385 199311 419560
718 718729 714735 511750 511700 784795 781795 387725 390592 7861040
Table 2.4
Gold-bearing solders
Eutectic temperature C F
Composition, wt%
common solder alloys given in Fig. 1.14. Poor uidity of the solder increases the risk of inadequately lled joints, which mar the performance and reliability of the product. It is generally recommended that gold-silicon foil is given a light etch in hydrouoric acid and then used within 30 minutes. This process step removes the surface silica and silicon and greatly improves the wetting and spreading behavior. While technically successful, this approach does require handling of this rather aggressive acid and special operator training is usually required to satisfy health and safety regulations. Furthermore, when using gold-silicon solders, precautions must be taken to ensure that the initial cooling rate of solidied joints does not exceed 5 C/s (9 F/s). If this condition is not satised, Au3Si is formed, and its subsequent decomposition with time to gold and silicon can produce cracks within the joint, due to the associated volume contraction [Johnson and Johnson 1984]. A number of alloying additions are known to be capable of promoting solder spreading [Humpston and Jacobson 1990]. For the Au-2Si solder, one of the most effective promoters of spreading is tin. Results of spreading tests conducted under identical conditions, but with different concentrations of tin in the solder showed
Fig. 2.20
68 / Principles of Soldering
that increasing the level of tin gave a progressive improvement in solder spread, as illustrated in Fig. 2.25. Hardness measurements revealed that
this was accompanied by a softening of the alloy, with its hardness decreasing by more than 150 HV, a welcome feature because it makes the al-
Fig. 2.21
Fig. 2.22
loy more amenable to mechanical working into foil and wire for solder preforms. The simplied phase diagram of the Au-Si-Sn alloy system, reproduced in Fig. 2.26, indicates that, by restricting the concentration of tin to below 8%, the melting temperature of the alloy is determined by the reaction among gold, silicon, and the Au5Sn intermetallic compound, which occurs over a narrow melting range of 356.5 to 358.0 C (673.7 to 676.4 F). This is closely similar to the solidus temperature of the Au-2Si alloy (363 C, or 685 F), and switching of this latter alloy for one additionally containing tin will therefore not upset subsequent manufacturing steps. Semiconductor die attach represents one of the main areas of application of gold-silicon and gold-germanium solders. Neither will directly wet bare silicon. Silicon is a relatively refractory element that, when exposed to air, becomes covered with a stable, continuous, and extremely adherent layer of the native oxide (silica). For this reason, all conventional solder alloys are unable to wet silicon even with uxes. In order to achieve wetting, it is necessary to either coat the back of precleaned die with a solderable metallization, such as gold, or to manually scrub the molten ller between the die and the package and thereby physically
displace the silica layer. The rst option is not always compatible with other stages of semiconductor fabrication and so scrubbing is resorted to, using a protective shroud of inert gas. With the advent of automatic machines to perform this operation, it has evolved to be a rapid joining method that delivers consistent quality joints. A successful method of promoting wetting of oxidized metal surfaces is to incorporate elements into the ller that are capable of reacting with the oxide to form an adherent bond [Crispin and Nicholas 1984, Xian and Si 1991]. Research has shown that addition of just 1% Ti to the Au-2Si and Au-20Sn solders is highly effective in promoting wetting of bare silicon at soldering temperatures of around 400 C. Concentrations of titanium of up to 2% were found to have little effect on the melting ranges of the alloys and are actually benecial in reducing their hardness. Figure 1.29 shows a silicon die soldered into a gold-metallized package using a gold solder with the composition Au-2Si-8Sn-1Ti; the joint was made by heating in a nitrogen atmosphere without scrubbing. Wetting of the ceramic package and the silicon die by the solder was found to be highly uniform. The gold-germanium eutectic composition alloy is sometimes recommended in place of the
Fig. 2.23
70 / Principles of Soldering
Fig. 2.24
Spreading behavior of Au-2Si solder as a function of the excess temperature above its melting point
Fig. 2.25
Spreading behavior of Au-2Si solder containing tin. Left to right: 0%, 4%, 8% Sn
gold-silicon solder where cost margins are particularly tight. As can be seen from the phase diagram given in Fig. 2.22, it offers the benet of lower gold concentration with little change in melting point (note in some older texts the melting point of the gold-germanium eutectic is erroneously given as 256 C). The gold-germanium binary eutectic contains 12.5 wt% Ge compared with 3.2 wt% Si in the case of the gold-silicon eutectic alloy. However, the price of germanium is much higher than that of silicon (because this element has to be produced as a minor by-product of zinc), but is considerably below that of gold. Historically, the price of germanium is about 5% that of gold, although in 2002, it stood at 10% of the gold price, so that the saving in materials cost with respect to goldsilicon is fairly signicant. As a solder, goldgermanium is no easier to use than its silicon equivalent. The Au-12Ge solder reportedly exhibits excellent wetting characteristics on copper, silver, and nickel surfaces [Hosking, Stephens, and Rejent 1999]. Where this solder is to be used in conjunction with steel parts, the parts should be plated with nickel prior to soldering to avoid the formation of iron germides. Hosking, Stephens, and Rejent [1999] have also shown that although Au-12Ge solder is brittle at temperatures up to 170 C (340 F), above about 220 C (430 F), this alloy is reasonably ductile
Fig. 2.26
Gold-rich portion of the Au-Si-Sn phase diagram. The system can be divided into three regions by melting point. The circle indicates the alloy composition selected as the solder. Adapted from Prince, Raynor, and Evans [1990]
and it is possible to substantially remove expansion mismatch stresses by a postsoldering heat treatment at this temperature. The same will presumably be true of the Au-2Si solder, but the softening temperature is not known for this alloy. Gold-Tin. The Au-20Sn eutectic solder is hard and moderately brittle. These properties arise from the fact that its constituent phases are two gold-tin intermetallic compounds, namely, AuSn () and Au5Sn (). The Au5Sn phase is stable over a range of compositions and, consequently, has some limited ductility that is imparted to the solder alloy (approximately 2% at room temperature). Although difficult, this solder can be hot rolled to foil and preforms stamped from it. By using rapid-solidication casting technology, it is possible to produce thin ductile foil, up to about 75 m (3 mil) thick and having an amorphous microstructure. However, this state is somewhat unstable, and within about 30 min at room temperature the rapidly solidied strip is indistinguishable in its mechanical properties from foil prepared by conventional fabrication methods [Mattern 1989]. Nevertheless, this crystallization can be suppressed for about a year if the quench-cooled material is stored under liquid nitrogen (196 C, or 320 F), and for close to one month at 20 C (4 F), so that it is possible to manufacture shaped foil preforms while the strip is still ductile, which can then be either immediately placed in a jig or returned to cold storage. The alloy can also be readily gas atomized to form spherical powder and, because it is relatively inert, will survive for long periods in an organic binder medium (optionally containing a ux) without degradation. For this reason, the solder alloy is often used in the form of paste, provided the components being joined are compatible with the chemicals involved. An alternative method of introducing the Au20Sn solder into an assembly is to selectively coat the joint area with a thick layer of gold, overlaid with a thinner layer of tin in the thickness ratio of 3Au to 2Sn. If electroplated, the layers must be deposited in this order owing to their electrochemical potentials. By subjecting the tin to a light acid etch and then immediately applying an outer layer of gold by evaporation or immersion plating, the predeposited solder is suitable for use without ux and can be endowed with a shelf life of several months. Other methods of producing ductile preforms of gold-tin solders are described in Chapter 4, section 4.1.5. The principal applications of the Au-20Sn solder are die attach of gold-metallized chips
and the hermetic sealing of ceramic semiconductor packages. The die-attach process that is used when the sealing operation is performed with the Au-20Sn solder is usually a highermelting-point gold-silicon solder. Thus, the gold-silicon and gold-tin solders are used in the initial joining operations of a step-soldering sequence, which ends with the soldering of the packages containing the chips onto printed circuit boards. This nal soldering operation is most frequently performed using a lead-tin or lead-free solder at a process temperature in the region of 240 C (464 F). Of the gold-bearing solders, only the Au-20Sn alloy has a signicant degree of uidity when molten. In their major eld of application, that of semiconductor manufacture, the use of uxes to promote spreading is not usually permitted. Instead, the techniques and methods normally applied to uxless joining are invoked to effect reliable joints. The options available and underlying principles of uxless joining are described in detail in Chapter 3, section 3.3. The gold-tin eutectic solder tends mostly to be used for soldering to thick gold metallizations. There are some difficulties with this, notably the dissolution of gold increases the freezing point and prevents good spreading. Identication of wettable, but insoluble barrier metallizations for this solder is therefore desirable. Copper, nickel, chromium, and nichrome are wholly unsuitable as they all dissolve readily in the Au-20Sn alloy. Palladium is a good wettable metallization for the molten solder, but, on extended aging in the solid state, Kirkendall voids form at the interface between the residual palladium and the Pd3Sn2 intermetallic compound [Anhock et al. 1998]. The solubility of platinum in Au-20Sn solder is strongly temperature dependent. At normal soldering temperatures with short process cycles, a thin layer of platinum (200 nm, or 8 in.) provides a readily wettable and stable barrier layer [Wada and Kumai 1991]. However, if large superheats or prolonged thermal excursions are necessary, then published work suggests that cobalt is a better choice. The dissolution rate of platinum in gold-tin eutectic at 330 C (626 F) is 10 nm/s (0.4 in./s), but nearly an order of magnitude slower for cobalt. Both gold and tin have very low solubility for cobalt. The mechanical properties of cobalt-tin intermetallics are not known, but, as the layers formed during normal soldering cycles will be extremely thin, they are unlikely to inuence joint properties signicantly [Park et al. 2002].
72 / Principles of Soldering
A high-melting-point quaternary ller metal based on the gold-tin eutectic became available relatively recently, intended for joining to thicklm metallizations [DuPont Electronics 1998]. The design strategy appears to have been to substitute some of the gold by silver and copper, both of which form solid solutions with gold. The exact composition and melting range have not been disclosed, but the recommended process temperature is 400 C (750 F), compared with 350 C (660 F) for the Au-20Sn binary composition from the same supplier. This implies the melting range is probably somewhere around 300 to 350 C (570 to 660 F). Among the claimed benets of the substitution are higher melting point, greater ductility, and improved wettability. A noneutectic gold-indium alloy of composition 82Au-18In is offered commercially by a number of manufacturers. The alloy is gold-rich to the nearest eutectic and deliberately so because of the poor mechanical properties of the eutectic composition alloy. The demand for this solder and its inclusion in catalogs stems from its melting range of 451 to 485 C (844 to 905 F), which makes it the highest melting point solder available. However this gold-indium alloy is a difficult solder to work with, as it is inferior to gold-silicon and gold-tin alloys with regard to its wetting and ow characteristics, as indicated in Table 2.5. Also, there is no developed ux for this solder and it must be used uxless, with all the attendant problems and limitations, although the process temperature is sufficiently high that this is one of the rare examples where a hydrogencontaining atmosphere is chemically active toward surface oxides on the solder (see Chapter 3, section 3.1.3).
2.1.5
High-Lead Solders
solidus temperatures above 300 C (570 F). A representative list of these solders is given in Table 2.6. They possess small melting ranges, usually, about 5 to 10 C (9 to 18 F), and their usage arises largely on account of their melting point. They are, in effect, inexpensive alternatives to the Au-20Sn eutectic alloy or, with care, can be used in a step-soldering sequence in which the next ller metal is the gold-rich solder. These high-lead solders differ in one important respect to the Au-20Sn alloy in that they are extremely soft and creep readily. This can be either an advantage or drawback, depending on the application and the function that the solder joint is required to provide. The high-lead alloys possess other advantages as solders. They are suitable for use with goldplated components without the risk of catastrophic embrittlement through the formation of continuous AuSn4 intermetallic phase at the joint interface. Although there is a small percentage of tin in some of the high-lead solders, the alloy constitution allows for appreciable gold dissolution before the formation of undesirable gold-tin phases because the small amount of tin present is mostly incorporated in the lead solid solution. With small adjustments of composition, some high lead-base solders have wide melting range and consequently have excellent gap-bridging characteristics. In the development of electronic packaging fabrication technology, the Ag92.5Pb-5Sn alloy applied in the form of a proprietary solder paste was found to produce consistent hermetic seals for feed-throughs into hybrid microwave packages despite joint clearances varying from below 20 m (0.8 mil) to more than 150 m (60 mil), as can be seen in Fig. 2.27 [Jacobson and Sangha 1997]. An obvious omission from any list of highmelting-point lead-rich solders is pure lead. The reason for this is pure lead has atrocious wetting
Table 2.6 Selected high-melting-point lead-rich solders
Melting range Solder composition C F
High-lead solders can be loosely dened as alloys containing 95% or more Pb and having
Table 2.5 Contact angle of Au-12Ge and Au-18In solders on common metal substrates
Contact angle, degrees Substrate Au-12Ge(a) Au-18In(a)
4 4 5
35 58 43
(a) The process temperature for gold-germanium was 430 C (806 F) and gold-indium 550 C (1022 F), with 5 min hold at peak temperature in a vacuum of 7 mPa ( 5 105 torr). Adapted from Hosking, Stephens, and Rejent [1999].
98.5Pb-1.5Sb 95Pb-5Sn 92.5Pb-5In-2.5Ag 97.5Pb-1.5Ag-1Sn 93Pb-3Sn-2In-2Ag 97.5Pb-2.5Ag 95Pb-5In 93.5Pb-5Sn-1.5Ag 90Pb-5In-5Ag 92.5Pb-5Sn-2.5Ag
310322 308312 300310 309310 304305 304 300313 296301 290310 287296
590612 586594 572590 588590 579581 581 572595 565574 554590 549565
and spreading behavior. This is a consequence of the fact that lead does not alloy readily with most engineering metals and metallizations, so that there is only marginal thermodynamic driving force for wetting and spreading. Indeed, this is the sole reason for lead being unsuitable on its own as a solder and requiring alloying additions. The list of high-lead alloys given in Table 2.6 is by no means exhaustive. Many other variations exist, but the basis of their formulationnamely, small additions of silver, tin, indium, antimony, and other elementsis similar. Data sources cite slightly different solidus and liquidus temperatures for many of these alloys, but the discrepancies are mostly less than 5 C (9 F). Although the mechanical properties of leadrich solders are often maligned, they will actually deliver a joint with respectable mechanical properties. Chadwick peel-test results with a 98Pb-2Sn solder can have peel-fracture initiation strengths of more than 35 kN/m (200 lb/in.) when optimally made [Beal 1984]. Lead-rich solders containing 98% Pb appear to be a good initial choice for applications where mechanical properties are important, as indicated in Table 2.7. At lower lead contents, the deformation characteristics are slip bands that are balanced in distribution between the grains and the grain boundaries. At 98% Pb, the grain boundaries appear to be more robust, leading to more heavily localized strain effects and superior creep resistance. Above this concentration of lead, the solder progressively behaves more like a homogeneous single-phase material.
2.1.6
Indium Solders
Indium-base solders share the common characteristics of low melting point and being extremely soft and ductile. The mechanical properties are mostly a reection of the fact that at room-temperature indium solders are operating at a very high homologous temperature. That is, 25 C (77 F) is close to their melting point when expressed in Kelvin. Indium remains ductile even at cryogenic temperatures. At high homologous temperatures, the rate of solid-state diffusion in metals with simple crystal structures is sufficiently fast that microstructural changes can occur in timescales that are comparable to changes in the service environment of joints in components. This is exemplied by the stress-strain curve of a thick indium soldered joint and the continuum between stress-strain and creep data given in Fig. 2.28 and 2.29, respectively [Freer Goldstein and Morris Jr. 1994; Darveaux and Linga Murty 1993]. This means that recovery and recrystallization occur as fast as work hardening is induced, and mechanical failure of joints made using indium-base solders tends to be stress overload or unidirectional creep. The ready creep of indium solders implies that joints are unlikely to fail on thermal cycling unless the load-displacement curve is asymmetric. It follows that indium solders are well suited for making joints between dissimilar materials that will be subject to thermal cycling. In fact, creep in joints made with indium solders can usually occur sufficiently fast to ensure that the stress is always close to zero, with roughly 80% stress relaxation occurring within seconds of a step change in strain. Then, the solder will not work harden. If joints are suitably designed to take advantage of these benecial thermomechanical characteristics, indium solders can provide superior life (see Fig. 2.30, also Humpston et al. 2001). When indium soldered joints do fail, the mechanism is predominantly by classical creep rupture, having its origins in the nucleation
Table 2.7 Mechanical properties and corrosion resistance of high-lead solders containing tin
Corrosion Stress-rupture weight loss time, 142 N (32 ASTM D 1384 lbf) load, days 2 weeks, mg
Microsection through a joint of varying width made using the 1.5Ag-92.5Pb-5Sn solder in the form of a paste to two electronic packaging components in lightweight Al-70Si (Osprey CE7 alloy), plated with nickel and gold. The blackish region constituting the joint is the solder, which completely lls the gap, which varies in width from less than 20 to more than 150 m (0.79 to 5.9 mil).
Fig. 2.27
100 99 98 97 96
74 / Principles of Soldering
and coalescence of cavities, arising from the material redistribution associated with stress relaxation. In contrast, tin-base solders, when subject to conditions of low-cycle (low-strain rate) fatigue, fail because the plastic deformation in the solder leads to microstructural changes in the region of maximum strain. Internal cavities then develop, which coalesce to form voids and then grow into cracks.
Shear stress strain curve for a 500 m thick indiumtin (In-48Sn) joint held at 40 C (104 F) ambient and strained at a rate of 5 104/s. Adapted from Freer Goldstein and Morris Jr. [1994]
Fig. 2.28
Continuum between stress-strain and creep data for indium-tin eutectic solder (In-48Sn) at room and elevated temperature. Adapted from Darveaux and Murty [1993]
Fig. 2.29
Resistance of a ip-chip daisy chain between silicon and alumina, versus the number of cycles (N) of a thermal shock test 25 to 196 C, 30 s dwell for the solders indicated. Adapted from Shimizu et al. [1995]
Fig. 2.30
Indium-base solders used in optoelectronic and photonics applications can also fail by a process known as phase segregation. This occurs when there is an extreme and unidirectional electrical and/or thermal gradient sustained across a joint. The result is migration of indium toward one joint interface and the accumulation of voids at the other until the joint fails. This mechanism is generally only observed in applications such as die attach of microwave power ampliers and laser diodes where, although the absolute power levels are modest, the small physical size of the parts results in very high energy ux. Pure indium is not often used as a solder because the wetting and spreading characteristics are mediocre as are the mechanical properties of joints made using it. One exception stems from exploitation of the complex oxide that forms on indium. Very-high-purity indium is readily available because this metal is chemically extracted from zinc residues as a minor by-product. Provided the indium is of purity better than 99.99995%, it will wet and spread over unmetallized oxide ceramics and glass, in air, without ux. The resulting joints do not have the same strength (5 to 10 MPa, or 725 to 1500 psi) and fracture toughness as conventional soldered joints, but are nevertheless hermetic and usable in a limited range of applications [Indium Corporation of America 1994]. The low melting point of indium solders stems from the low melting point of indium itself, which is 157 C (315 F). It is interesting to note that the cited melting point of indium has increased by nearly 5 C (9 F) over the last 40 years with the development of improved rening methods; the melting point of the metal being very sensitive to low levels of metallic impurities. Albeit largely as a point of metallurgical curiosity, the melting point of indium is also unusually sensitive to pressure; the application of 4000 MPa (580 ksi) will cause the melting point to roughly double to 300 C (570 F; see Fig. 2.31). Alloying additions made to indium to form lower-melting-point eutectics confer a number of practical benets. Table 2.8 lists some of the more common indium solders and their melting points. The merits of alloying are principally the generation of multiphase microstructures, which improve the mechanical properties of the solder, and the formation of mixed-composition oxides, which are generally easier to chemically remove than pure indium oxide. Indium oxide forms readily as a sticky and tenacious lm that requires specially formulated uxes to effect its
Fig. 2.31
Melting point of indium as a function of pressure. Adapted from Kennedy and Newton [1963]
removal. Even so, as can be seen from Fig. 1.14, indium solders are appreciably less uid than their tin-base counterparts and require greater superheats to achieve comparable ow. Indium-lead alloys are a useful class of solders with readily adjustable melting ranges. Consideration of the phase diagram for this alloy system, which is given in Fig. 2.12, reveals that indium and lead form an essentially miscible mixture that extends between the melting points of the two parent materials, that is, 157 to 328 C (315 to 622 F). Thus, by appropriate selection of the composition it is possible to obtain a solder, albeit one with a melting range and mediocre mechanical properties, that has any desired solidus temperature between the two limits. Although indium-lead alloys solidify by peritectic reaction, the solidication rate of most soldered joints is usually sufficiently slow to prevent problems associated with microstructural coring, but this must be considered if a fast heating method, such as laser soldering, is being employed. Indium-containing solders are often recommended for joining to gold-coated components because gold is less soluble in these alloys than
Table 2.8 Composition and melting ranges of some common indium-base solders
Melting range Composition, % C F
in lead-tin solders. The restricted solubility of gold in a joint containing indium is largely associated with the formation of a continuous layer of the intermetallic compound AuIn2 at the solder/gold interface, which effectively suppresses further reaction between these metals (see Fig. 2.2). In the presence of lead, the interfacial layer takes the form of separate grains of the AuIn2 compound embedded entirely in primary lead. The lead provides an easy diffusion path between the solder and the gold coating and so permits the reaction to continue, albeit at a modest rate over extended timescales [Yost, Ganyard, and Karnowsky 1976; Jacobson and Humpston 1989]. Intermediate melting temperature indiumbase solders, such as the In-15Pb-5Ag composition, are signicantly more ductile than lead-tin eutectic and can yield joints with correspondingly superior thermal fatigue performance [Edwards, Nixon, and Lakes 2000]. Although the melting range of this alloy is close to the eutectic temperature of lead-tin solder, its price differential is a factor of about 80, which restricts its use to high-added-value and more specialist applications.
2.2
As might be expected, metallic impurities present in ller metals can be either highly benecial, benign, or totally undesirable, depending on the solder, the impurity, and the application. The effects of impurities have been most widely studied for lead-tin solder. Lead-tin solder is generally prepared to highpurity specications because the presence of
76 / Principles of Soldering
many other elements can have a deleterious effect on joint formation, even in concentrations as low as 5 ppm [Schmitt-Thomas and Becker 1988; Becker 1991]. With some elements, joint formation may be inhibited or even prevented. The detrimental characteristics of the elemental additions and the critical concentrations that produce these adverse effects are summarized in Table 2.9. There was no minimum level of aluminum that did not adversely affect solderability, except on brass substrates (see below): the lowest concentration investigated was 0.0005%. A detailed study of the effect of different elements on the wetting properties of eutectic leadtin solder was made by Ackroyd, Mackay, and Thwaites [1975]. They assessed the solderability on copper, brass, and steel coupons using areaof-spread tests, rotary dip tests, and wetting balance measurements. High-purity Pb-60Sn solder was deliberately adulterated with low concentrations (<1%) of aluminum, arsenic, bismuth, cadmium, copper, phosphorus, sulfur, antimony, and zinc and used under an ambient atmosphere with selected active and nonactivated uxes. Of the substrate materials, brass exhibited a consistent enhancement in spread area with increasing impurity addition, irrespective of the element in the above selection. By contrast, the solderability of copper and steel almost invariably deteriorated in the presence of subpercentage levels of the impurity and then continued to decline or tended to a steady value at higher concentrations. The effect of major ternary additions on the wetting angle of Pb-80Sn solder on copper, brass, and mild steel substrates is given in Table 2.10 [Raman 1977]. Bismuth was found to be generally benecial; other additions such as antimony and indium exhibited a dependence on the substrate. The ternary additions consistently imTable 2.9 Lowest impurity concentrations producing detrimental effects, in terms of wetting, in Pb-60Sn solder
Impurity element Impurity concentration, % Detrimental effect
proved the strength of butt joints to copper and brass substrates, but markedly decreased the joint strength for the mild steel substrates. One important area where impurity elements have a positive role in solders is in suppressing tin-pest in lead-tin solders. Pure tin can exist in two allotropic forms: white tin, which is the common metallic form and gray tin, to which white tin can transform below 13 C (55 F). Owing to the 26% increase in volume that accompanies the phase change from white to gray tin, the solid metal disintegrates into a crumbly mass having essentially no strength. The transformation from white to gray tin is clearly disastrous for the mechanical properties of soldered joints. Tin-pest can affect electrical systems used in subzero environments, for example, in high-ying aircraft and also soldered containers and conduits of refrigerants. The transformation does not occur spontaneously, but is always preceded by an incubation period, which may be as much as several years if the tin is exceptionally pure. The maximum rate of conversion occurs at about 40 C (40 F). Figure 2.32 shows the rate of transformation as a function of time at various temperatures. The incubation period is much shorter if white tin is mechanically worked at low temperatures or is inoculated with either gray tin or other elements and compounds having similar diamondlike crystal structures, such as silicon and ZnSb [Macintosh 1968]. Copper, too, accelerates the process [Rogers and Fydell 1953], whereas lead at higher concentrations than the eutectic retards it slightly [Williams 1956]. Relatively small additions of certain elements have been shown to suppress the allotropic transformation of tin. The addition of 0.15% antimony will prevent it in tin-base solders. Bismuth is even more effective and a 0.05% addition is all that is necessary [Bornemann 1956]. Existing solder specications accommodate sufficient quantities of impurity elements to prevent tinpest. However, in recent years there has been a
Table 2.10 The effect of major alloying additions on the wetting angle of Pb-80Sn solder
Wetting angle, degrees Solder/addition Copper Brass 0.15% C steel
Oxidation Reduced spreading Reduced spreading Oxidation Reduced spreading High melting range Dewetting High melting range Oxidation
26 38 22 30 34
50 43 36 38 37
95 82 70 102 98
trend toward high-purity solders in order to improve certain characteristics in wave soldering machines. In some solder specications, the maximum level of antimony permitted has been reduced from 0.5 to 0.12%, so that tin-pest could occur in soldered assemblies unless other impurities, notably bismuth, remain at appropriate levels [EP&P 1992]. Metallurgical reactions of solders with the parent materials can also have an effect on the properties and performance of joints made with ller metals. The most common examples are modication of melting range and physical properties. This subject is dealt with further in section 2.3 in this chapter. Other examples of favorable effects of impurities on different solders include:
Tin also improves the spreading characteristics and reduces the hardness. However, to achieve these benets, the concentration of tin ideally needs to be several percent (see section 2.1.4 in this chapter). The addition of germanium to gold-silicon solder also reduces the alloy hardness with consequential benets for joint ductility. The softening of the multicomponent alloy arises as a result of grain rening of the silicon phase, the grain size being decreased by roughly half for comparable cooling conditions. Similarly, a 0.5% silver addition provides grain renement of bismuth-tin solder. Silver, gold, and antimony, at moderate concentrations (<4%), act as strengthening elements in lead-tin solder. The addition of 0.5 to 0.6% Bi to the Au-20Sn solder is said to improve the wettability when the alloy is used for joining gold-plated products [Rapson and Groenewald 1978]. Although the presence of aluminum is undesirable in conventional solders, as it greatly exacerbates oxide formation, the same is not true of all solders. A case in point is the leadfree alloy Sn-8.5Zn-1Ag. Minor additions of aluminum to this solder, up to a maximum of 1 wt%, actually improves wetting of copper. In this instance the increased oxidation is outweighed by modication of the metallurgy such that the formation of copper-zinc intermetallic compounds is suppressed in favor of copper-tin phases because of the ready association between aluminum and zinc in the molten state [Cheng and Lin 2002].
Small quantities of the transition metals cobalt and iron (<0.5%) increase the mechanical strength of Ag-1.7Cu-93.6Sn solder by about 25% at room temperature. Cobalt renes the Ag3Sn phase, while iron promotes renement of primary tin dendrites. The benets are lost at elevated temperature (>150 C, or 300 F) owing to rapid solid-state diffusion [Anderson et al. 2002]. Rare-earth elements have a similar effect, as discussed in Chapter 5, section 5.8. Zinc improves the spread and ow of Au-2Si solder, but only at concentrations above 0.5%. Zinc has the additional benet of softening the alloy, reducing the hardness from 320 to 128 HV for a 0.75 wt% addition.
This list is by no means exhaustive, and some additional examples are given in Chapter 3, section 3.3.8.4. In general, most of these effects have been discovered by accident rather than by design. Manufacturers supply solders of suitable purity for most applications, but the process designer must always be aware of the possibility of unfavorable composition changes to the solder as a result of alloying with the components being joined. This topic is examined in the next section.
2.3
Allotropic transformation of white tin into gray tin as a function of time and temperature. Adapted from Bornemann [1956]
Fig. 2.32
The selection of a solder for a particular application is often based exclusively on the melt-
78 / Principles of Soldering
ing point and mechanical properties of the solder and its ability to wet the parent materials. The solder is regarded as a uniform layer of metal that simply bridges the gap between the components and binds them together. If only life were that simple! In reality, the formation of the desired metallic bond between the solder and a component requires a degree of alloying. The ensuing metallurgical reactions usually lead to a heterogeneity of phases in the joint. To further complicate matters, kinetic factors tend to accentuate the development of this nonuniformity. Such inhomogeneities often determine the quality and overall characteristics of joints, such as their mechanical properties, the ease and extent of solder spreading, the nature of any llets formed, and so on. Metallurgical reactions do not cease once the joint has been made, but continue to proceed, to a greater or lesser extent, during the service life of the assembly. The rate-controlling step for reaction between two solid metals is the diffusion of atoms between the reacting phases. The relative position of the product of the reaction and the reacting phases will be governed largely by the diffusion coefficients of the participating metals. For individual metals, it has been established empirically that the rate of diffusion R, increases rapidly with absolute temperature, T, following an exponential relationship:
R A expQ kT
where k and A are constants; k is the Boltzmann constant, and A is an experimentally determined factor for each combination of reacting phases that may vary with concentration. Q is the activation energy for diffusion which, to a rst approximation, is proportional to the melting point, Tm, of the particular metal [Birchenall 1959]. The rate of reaction will therefore be dependent on the homologous temperature dened as the ratio of T/Tm and will be more pronounced for low-melting-point solders that see service at or close to normal ambient temperatures. The reactions produce perceptible metallurgical changes in the constitution and microstructure of soldered joints [Frear, Jones, and Kinsman 1990, Chapter 2]. For a proper understanding of metallurgical reactions between solders and parent materials, it is essential to have some grasp of the subject of alloy constitution. The constitution of an
alloy refers to features such as its composition, melting range, range of phase stability, solubility limits, and related parameters that can be deduced from the phase diagram of the system in which the alloy appears. In the following sections, some attention is given to highlighting the value of phase diagrams and suggesting how this valuable source of information might be tapped. Frequently, the appropriate phase diagram for elucidating specic solder/substrate reactions and joint microstructures is not available in the literature. Experimental techniques and a methodology for elucidating gaps in phase diagrams are described in the literature [Humpston and Jacobson 1993, Chapter 3]. Note that all of the phase diagrams in this book are dened in weight percentages of the constituent elements as this is more appropriate to soldering technology than the atomic percentage scale. Also, the relative proportion of the elements in intermetallic compounds, such as Cu3Sn, refer to atomic weights. General equations for converting atomic to weight percent of constituents in alloys, and vice versa, are given in Appendix A3.1. The fundamentals of alloy phase diagrams are covered in many metallurgical textbooks and will not be repeated here. Readers without a background in this eld are referred to the publications listed in the Preface under the heading Alloy Constitution, which provide an excellent introduction to the subject. Here it will suffice to state that a phase diagram is a representation of the thermodynamic stability of phases as a function of composition with respect to particular thermodynamic variables such as temperature or, less commonly, pressure. What is important to remember is that the information given by the diagrams relates to essentially equilibrium conditions. The phase diagram tells us about the ultimate balance of phases within the joint and those that are likely to be encountered during the progression toward equilibrium. A joined assembly in which the solder and abutting components are different materials are never in true compositional equilibrium, as long as the joint remains distinct. In most practical contexts, the composition of a joint will be tending toward equilibrium over most of its width, and therefore phase diagrams are applicable to an assessment of its constitution. However, at the edges of the joint, marked compositional gradients will exist, causing a signicant deviation from equilibrium. These will be exacerbated by any temperature gradients that develop during the process cycle and are manifested as the appearance of different
phases in those regions. Even here, phase diagrams can assist in the elucidation of the metallurgical reactions and the resulting phases, as shown in the following section. Phase diagrams can provide the following practical information:
The melting temperatures of the virgin solder and of the abutting components
The rate of reactions that might occur between the solder and the components and their variation with time and temperature. This applies both when the solder is molten and when it is solid during service. The spatial distribution and morphology of phases in the joint, although frequently it is possible to deduce whether intermetallic phases are likely to form as interfacial layers or will be dispersed throughout the solidied solder. This is explained in section 2.3.1 in this chapter. The wetting characteristics of a particular solder/parent materials combination, of even perfectly clean surfaces. In practice, wetting is likely to be heavily inuenced by the oxides, impurities, and residues that are inevitably present on component and solder surfaces, but that are extraneous to the alloy phase diagram. Physical properties of joints, in particular the mechanical and corrosion characteristics. However, it is often possible to predict the likely range of certain physical properties by comparison with other known alloy systems. The simplest diagrams that are encountered in a joining context are those relating to binary alloys where, for example, the solder is a single metal being used to join components of another metal. This situation is represented by the use of pure tin to solder copper pipes. Although the available literature on phase diagrams may appear to be reasonably comprehensive, it is worth bearing in mind that reliable diagrams exist for, roughly, only 50% of binary
combinations, 5% of ternary systems, and 0.5% of quaternary mixtures. A compendium of authoritative alloy phase diagrams is being prepared under the auspices of the International Programme for Alloy Phase Diagram Data (IPAPD) and the rst volumes have already appeared. This work is ongoing, and updates are to be found in the Bulletin of Alloy Phase Diagrams and Journal of Phase Equilibria. On a periodic basis, these publications include a cumulative index that lists all phase diagram evaluations published by members of IPAPD. It is worthwhile consulting the most recent phase diagram available, because older diagrams can contain signicant errors or omissions. High-order alloy systems are naturally more complex and are less well documented, as noted earlier. However, for a given joining process only a very limited portion of the phase diagram is required, and if one is unavailable it is often possible to experimentally determine the necessary data (for example, [Humpston and Jacobson 1993, Chapter 3]). Recently, an exciting breakthrough appears to have been achieved in a method that is able to predict, with claimed accuracy exceeding 99%, whether compound formation will occur for any binary, ternary, or quaternary system. This will be of great assistance in reducing the time and effort to establish the phase relationships in alloy systems, particularly those of higher order that are often necessary to encompass soldering and brazing processes. This work also proved that materials properties are quantitatively contained in the elemental property parameters of the constituent elements, so that once additional information retrieval methods are automated, the selection of materials for specic applications will be greatly facilitated, and it is hoped many new materials with exciting property combinations will be discovered [Villars et al. 2001]. The value of alloy phase diagrams for understanding and optimizing soldering processes can best be appreciated by describing a few specic examples in the following sections.
2.3.1
Example 1: A Binary Eutectic Composition Solder Used with Components of One of the Constituent Metals, with No Intermetallic Compound Formation. A representative example of this type of reaction is a silver-lead solder used to join components of pure silver.
80 / Principles of Soldering
The silver-lead phase diagram represented in Fig. 2.11 shows that in this binary alloy system there is minimum melting point, that of the eutectic composition (Ag-97.6Pb). Alloys with this unique composition transform between a liquid (L) and two solid phases (S1 and S2) at a xed temperature, 304 C (579 F), according to the reaction:
L S1 S2
For all other compositions, except those of the pure metals, there is a separation between the liquidus and solidus temperatures. On either side of the eutectic the liquidus and solidus separate, with the alloys melting over a range of temperatures. At temperatures within the melting range, the alloy is partly liquid and partly solid. On cooling below the solidus temperature, all alloys in this system exist as duplex of S1 and S2, in direct proportion to the alloy composition. It can be seen from the silver-lead phase diagram (Fig. 2.11) that silver is soluble in the Ag97Pb solder when molten. Thus, at the joining process temperature (say, 400 C, or 750 F), the solder will dissolve silver from the components until the equilibrium concentration of silver is attained, determined by the intersection of a line drawn on the phase diagram at the process temperature with the liquidus curve. Thus, at 400 C (750 F) the dissolution of silver by the Ag-97Pb solder changes its composition to approximately Ag-93Pb. That is, the dissolution of silver increases the liquidus temperature of the solder in the joint, but not its solidus temperature, because eutectic transformations are isothermal. Because soldering cycles are short, the solder will not normally dissolve more silver. (In theory, if left at temperature for a sufficiently long time, the lead would diffuse through the entire volume of the silver components so that the assembly had a uniform composition determined by the total quantities of silver and lead in contact. This is actually the basis of diffusion-soldering processes (referred to in Chapter 5, section 5.9)). At the commencement of the cooling stage of the process cycle, the molten solder no longer corresponds to the eutectic composition, but is rich in silver and, in consequence, now possesses a freezing (i.e., melting) range. On cooling below the liquidus temperature, the excess silver will solidify rst, as indicated by the phase diagram. This precipitation tends to occur preferentially at
the interface between the components and the solder because this interface is usually slightly cooler than the volume of the molten solder. Precipitation continues until the temperature and composition of the remaining liquid reach the eutectic point so that nal solidication by the molten solder results in the formation of a small volume fraction of nely divided eutectic. The alloy microstructure will therefore comprise primary dendrites of silver with the interdendritic spaces lled with the duplex eutectic mixture. The primary silver phase will contain a local concentration gradient as the amount of lead it incorporates varies with temperature. The silver phase is then said to be cored. It is possible to quantitatively follow the change in composition of the solder as it cools by application of the lever rule. Referring to the enlarged and schematic portion of the phase diagram given in Fig. 2.33, at 350 C (662 F) the weight fraction of the solder that is solid, under equilibrium conditions, is:
% solid XO XY 100
where X is the composition of the liquid phase, Y is the composition of the solid phase, and O is the composition of the alloy.
Fig. 2.33
scale)
Application of the lever rule to an enlarged portion of the silver-lead system diagram (not drawn to
Alloys of eutectic composition are preferred as solders on account of the following characteristics:
stage, minimizes the chance of such an interruption. However, this assumes that alloying of the solder with the component materials does not greatly shift the composition of the solder from its eutectic point. A disturbed joint generally has inferior mechanical properties, and the llets will acquire a rough surface with a frosty appearance. For these reasons, most solders are of eutectic composition. Example 2: A Binary Eutectic Composition Solder Used with Components of One of the Constituent Metals, with Intermetallic Compound Formation. The silver-tin phase diagram is given in Fig. 2.9. The eutectic composition solder (Ag-96.5Sn) is widely used to join silvercoated components. One of the main reasons for this choice is that the erosion of the metallization on such components is low and highly predictable. The restricted erosion is a consequence of the formation of the Ag3Sn intermetallic compound () as a layer that separates the molten solder from the remaining silver, because it has a melting point above normal soldering temperatures. This interfacial layer restricts further reaction to a degree determined by temperature and the duration of the soldering process, as indicated by the data given in Fig. 2.35. Using such data, it is possible to determine the minimum thickness of silver that is required for a given volume of solder. This feature can be used to advantage in soldering to nonmetals by ensuring that the application of a silver metallization of dened thickness will prevent erosion through to the nonmetallic base material in a prescribed soldering cycle and thereby avoid catastrophic dewetting. While the phase diagram can provide guidance about whether a new phase will form, it cannot be used to determine the ultimate distribution within the joint as this is greatly inuenced by a combination of factors. Another piece of important information that cannot be ascertained from equilibrium phase diagrams is the rate of growth of phases. The different rates at which solders can react with substrate metals is vividly illustrated by comparing the rate of erosion of gold by molten indium in Fig. 2.2 with that by molten tin in Fig. 2.36. Supercially, the phase diagrams of these binary systems are identical. Although it would appear from Fig. 2.2 that the indium-gold reaction is self-limiting, this is only true in relation to the short timescales of
Tensile strength of cast bars of lead-tin alloys. Optimal mechanical properties are coincident with the eutectic composition (Pb-62Sn). Adapted from Inoue, Kurihara, and Hachino [1986]
Fig. 2.34
82 / Principles of Soldering
typical soldering operations (seconds). Over longer periods of time, the reaction will proceed to a signicant extent. Even when the solder is solid, the extent of the reaction can be appreciable as shown in Fig. 2.37, despite the fact that reaction rate may be one or two orders of mag-
nitude slower than when the solder is molten, the exact ratio depending on the respective temperatures. This point must be borne in mind when considering the stability of a joint over the service life of the associated assembly. The design life may be as long as 25 years, which is longer
Fig. 2.35
Erosion of silver by molten tin as a function of reaction time and temperature. Adapted from Evans and Denner [1978], augmented by authors own data
Fig. 2.36
than the duration of the joining cycle by a factor of six orders of magnitude. Example 3: A Binary Peritectic Solder, Illustrating Problems Associated with Using a Solder in This Category. The second common type of phase transformation is the peritectic reaction where a liquid, L, on cooling, partly solidies to form a solid phase S1 and at the peritectic temperature the remaining liquid reacts with S1 to form a new solid phase, S2. This may be written as:
L S1 S2
Indium-lead solders are frequently employed in situations where the standard tin-base solders cannot be used, for example, for soldering at intermediate temperatures between the available eutectic composition solders. The indium-lead phase diagram, which is given in Fig. 2.12, contains two peritectic reactions. These are:
L Pb at 171.6 C (340.9 F) L In at 158.9 C (318.0 F)
a peritectic reaction, it is not possible to maintain equilibrium conditions. This is due to the fact that diffusion rates in solids are about two orders of magnitude slower than in liquids, so that a nonequilibrium microstructure develops consisting of islands of the primary solid phase, S1, completely surrounded by a rim of the second solid phase, S2. A quaternary aluminum alloy microstructure exhibiting a peritectic transformation is shown in Fig. 2.38. In such an alloy, liquid that is rich in the lower-melting-point elements will be retained below the peritectic transformation temperature. Thus, the melting and freezing range of the alloy is widened and the remelt temperature cannot be reliably predicted. Furthermore, its microstructure will be grossly inhomogeneous and relatively coarse, to the detriment of the mechanical properties of joints made with this alloy. In higher-order alloys, a number of other types of phase transformation can occur and these are generally referred to as transition reactions. The majority of these possess features akin to a peritectic transformation, and they tend to be avoided when selecting alloy compositions as solders on the same grounds.
2.3.2
Alloys exhibiting this type of transformation are generally undesirable as solders because, during
It is rare for a soldered joint to be limited to a combination of just two elements, forming a binary alloy system. Usually the solder is an alloy of at least two metals, while engineering alloy substrates are frequently multicomponent. Com-
Continued growth of gold-indium intermetallic phases at the interface between a gold metallization and indium-lead solder at elevated temperature but below the solidus temperature. Adapted from Frear, Jones, and Kinsman [1990]
Fig. 2.37
An alloy microstructure characteristic of a peritectic transformation. The alloy contains four constituents: aluminum, copper, nickel, and silicon. The primary phase is totally surrounded by a rim of a second phase as a result of the peritectic reaction failing to maintain equilibrium conditions during solidication.
Fig. 2.38
84 / Principles of Soldering
monly, intermetallic compounds form between the constituents. The volume, distribution, and morphology of these intermetallic phases in a joint can have a pronounced effect on mechanical properties, in particular. From the relevant phase diagram, it is possible to predict whether the intermetallic compound will tend to form as a continuous interfacial layer against the parent materials or is more likely to be dispersed throughout the joint. If the intermetallic compound has poor mechanical properties, then a dispersion of it is preferred because this not only avoids the source of weakness represented by the interfacial intermetallic phase, but actually works to advantage by strengthening the solder. Examples of these different situations are described below. Some care needs to be taken when referring to the intermetallic phases that form in a ternary system. An example is provided by binary copper-tin solder wetted on nickel substrates to form a mixture of copper-tin and nickel-tin intermetallic phases at the interface. The apparent confusion in the literature over the composition of these phases can be explained by reference to the Cu-Ni-Sn phase diagram that reveals that the binary compounds Cu6Sn5, Ni3Sn2, and Ni3Sn4 not only have very extensive ternary solubility (up to 30%), but there exists a continuous solid solution between Cu3Sn and Ni3Sn. Hence, there is considerable scope for composition variation within the stable phase elds, and the convenient labels adopted from binary alloy practice do not adequately describe the real situation [Lin, Chen, and Wang 2002]. A ternary system is most usually represented by an equilateral triangle, with each of the vertices corresponding to the three elements. A grid is normally drawn on the triangle to provide a linear scale of composition. Temperatures are then represented by a series of isotherms, so that the position of the liquidus on the diagram is mapped as a topographical surface viewed in plan. Phase stability as a function of temperature is commonly represented by a diagram resembling a binary alloy phase diagram, where either one of the constituents or the ratio of two constituents is held xed. A single diagram cannot be used to track the solidication sequence because the ensuing composition changes can extend outside the plane of the diagram. For a similar reason, the lever rule cannot be applied to this representation in order to calculate the proportions of phases that exist in equilibrium. However, the lever rule can be used in conjunction with a series of isothermal sections.
Example 4: Interfacial Compound Formation between a Eutectic Solder and the Component Metals. As an example of interfacial compound formation, it is well known that leadtin solders, when used with copper components, form several copper-tin intermetallic compounds, predominantly at the copper/solder interface. Despite the many billions of soldered joints that have been made to copper using leadtin solder since antiquity, a complete and credible ternary phase diagram for this alloy system only become available in 1994. The binary copper-tin phase diagram, a liquidus projection, and the key isothermal sections and isopleths relevant to soldering to copper are reproduced in Fig. 2.39 to 2.42 [Yost, Hosking, and Frear 1993]. The features to note are:
Copper is soluble in molten lead-tin solders. The dissolution of copper by the solder marginally depresses the melting point of the lead-tin eutectic alloy. Liquid lead-tin eutectic solder that has dissolved a small quantity of copper will solidify via a ternary eutectic at 182 C (360 F). A reduction in melting point of a solder on dissolution of substrate constituents is a key driving force for spreading. The ternary eutectic point is actually at 61.75Sn-38.05Pb0.2Cu (wt%). The third constituent in the majority of reactions between lead-tin solder and copper, in addition to the tin and lead phases, is the intermetallic compound Cu6Sn5. It is only for very lead-rich solder compositions or when other molten solder compositions have had the opportunity to dissolve appreciable concentrations of copper that the intermetallic compound Cu3Sn is present as a primary phase. The rate of dissolution of copper in molten Pb-60Sn solder is initially rapid, as is the erosion of most other engineering metals and metallizations by this alloy, as can be seen from Fig. 2.43. Consequently, despite the short process cycle times normally associated with soldering practice, the molten solder will dissolve sufficient copper to reach the saturation concentration, dictated by the process temperature and the composition of the liquidus surface in the Cu-Pb-Sn phase diagram. On cooling this ternary solder, it would be expected from consideration of the phase diagram that copper-tin intermetallic phases will precipitate and become distributed within the
bulk of joint. If the peak process temperature is kept below 375 C (705 F), the phase diagram shows that the primary phase that precipitates on
cooling will be Cu6Sn5. This has been conrmed by experiment. At normal solder process cooling rates, the precipitates are typically of the order of
Fig. 2.39
Fig. 2.40
Liquidus surface of the Cu-Pb-Sn system. Adapted from Yost, Hosking, and Frear [1993]
86 / Principles of Soldering
Fig. 2.41
Expansion of the solder-rich, copper-poor, region of the Cu-Pb-Sn system. Regions are identied by the primary or rst phase that solidies on cooling in that region. Adapted from Yost, Hosking, and Frear [1993]
Fig. 2.42
Isothermal section of Cu-Pb-Sn system at 150 C. Adapted from Yost, Hosking, and Frear [1993]
5 to 20 nm (0.2 to 0.8 in.) in size and therefore can only be resolved using high-resolution imaging techniques such as transmission electron microscopy [Felton et al. 1991]. However, if the peak process temperature exceeds 375 C (705 F) and remains above this level long enough for more than 1.6% of copper to dissolve, then the rst phase to precipitate will be Cu3Sn. It is also apparent from the ternary phase diagram that copper-tin intermetallic compounds will form at the interface between the solid copper and molten solder. Owing also to the prevailing concentration gradient, the copper-rich Cu3Sn phase forms adjacent to the copper component with the more tinrich Cu6Sn5 phase between it and the solder, as shown in Fig. 2.44. The rate of formation of interfacial copper-tin intermetallic phases is governed by the rate of interdiffusion of copper and tin through them and
is well characterized. A distillation of the data that are available in the literature is presented in Fig. 2.45. The Cu3Sn phase layer does not grow to any signicant thickness during typical industrial soldering processes because the inward diffusion of tin through the Cu6Sn5 phase does not normally reach sufficient concentration [Dirnfeld and Ramon 1990]. The formation of continuous layers of intermetallic compounds between the copper and the molten solder greatly restricts the further dissolution of copper in the solder [Felton et al. 1991]. Copper-tin intermetallic compounds do not possess particularly desirable mechanical or physical properties in bulk form (see Table 2.11). These phases will continue to increase in thickness during elevated-temperature service of the component, albeit much more slowly than when the solder is molten as can be seen
Rate of dissolution of a range of engineering parent metals and metallizations in lead-tin solder as a function of temperature. Adapted from Klein Wassink [1989]
Fig. 2.43
Growth of copper-tin intermetallic compounds on a copper substrate wetted by lead-tin eutectic solder as a function of reaction time and temperature
Fig. 2.45
Table 2.11 Mechanical and physical properties measured for three common intermetallic compounds, on bulk specimens
The phases are generally hard with low fracture toughness. However, their thermal expansivities lie between the substrate and solder, which probably plays a role in decreasing the stress concentration in joints.
Property Cu6Sn5 Cu3Sn Ni3Sn4
Fig. 2.44
heat treatment
Hardness, HV Toughness, MPa m1/2 Electrical resistivity, cm Thermal conductivity, W/m K Thermal expansivity, 106/K Phases formed by reaction between a lead-tin solder and a copper substrate, following extended
88 / Principles of Soldering
from Fig. 2.46. The presence of thick layers of coppertin intermetallic phases is widely reported as being detrimental to the mechanical properties of the joints, and this has prompted many studies of the effect of their presence, for example Dirnfeld and Ramon [1990]. Figure 2.47 shows the relationship between the tensile strength of joints and the thickness of the copper-tin intermetallic layer as determined in these studies, and Fig. 2.48 shows the relationship between intermetallic thickness and creep rupture life [Hongyuan, Ying, and Yiyu 1994]. Notwithstanding this general concern, there are
very few reports of joint failures that can denitely be pinpointed to the copper-tin intermetallic phases. This unclear picture is understandable because tin forms hard intermetallic phases, of comparable thickness, by solid-state diffusion at elevated temperature with virtually all the metals commonly used in engineering [Kay and Mackay 1976]. Some relevant data are reproduced in Fig. 2.49. Not surprisingly then, joints made to copper testpieces using lead-tin solder are found not to be signicantly weaker than joints to other common substrate materialsin terms of their strength, ductility, and fatigue resistancewhen formed using conventional process cycles such that the intermetallic layers remain relatively thin. Not all the binary lead-tin solders form two intermetallic phases on reaction with copper. If
Growth of copper-tin intermetallic compounds on a copper substrate in contact with lead-tin solder for 100 s at different process temperatures
Fig. 2.46
Relationship between thickness of copper-tin intermetallic and creep rupture life of joints made to copper components using lead-tin eutectic solder
Fig. 2.48
Effect of thickness of copper-tin intermetallic compounds in soldered joints on tensile strength at room temperature
Fig. 2.47
Growth of tin-base intermetallic phases by solidstate diffusion at 170 C (340 F) on various substrate materials. Adapted from Kay and Mackay [1976]
Fig. 2.49
the tin level is less than 25%, the phase diagram in Fig. 2.41 indicates that only the Cu3Sn intermetallic will precipitate at the solder/copper interface under near-equilibrium conditions [Grivas et al. 1986]. Example 5: Distributed Compound Formation between a Eutectic Solder and the Component Metals. Another industrially important alloy system comprises gold, lead, and tin, on account of the widespread use of gold as a solderable metallization for uxless joining in the electronics industry. The liquidus surface of Au-Pb-Sn ternary phase diagram is given in Fig. 2.50, and a section from the eutectic Pb-62Sn composition toward gold is shown in Fig. 2.51. The principal features of these diagrams are closely akin to those of the Cu-Pb-Sn system discussed previously, albeit with the difference that in the gold-base system at least three embrittling gold-tin intermetallic compounds are able to form directly from the liquid. Therefore, it might be expected that, when lead-tin eutectic solder is used to join solid gold components, the resulting microstructure of joints will be similar. However, there are two important differences.
Gold metallizations are seldom more than a few microns (sub-mils) thick, largely on account of the cost of this metal, so that there is a limit to the concentration of gold that will accumulate in the molten solder. The precise value is determined by both the volume of solder alloy and the thickness of the metallization, but is usually well below the saturation concentration. Also, due to the high rate of dissolution of gold in molten lead-tin solders, even thick metallizations will be completely dissolved during the heating cycle. On cooling a Au-Pb-Sn ternary alloy formed by reaction of lead-tin eutectic solder with a gold metallization, the phases that precipitate are indicated by the phase diagram. Because of their unfavorable mechanical properties, it is gold-tin phases that are of interest. If the gold concentration is below 1%, then the gold remains dissolved as a solid solution in the lead and tin phases. Between 1 and 5% Au, solidication terminates by the ternary eutectic reaction so that the AuSn4 phase present will be well-dispersed throughout the volume of the solder and nely divided. Between 5 and 8% Au, AuSn4 precipitates as a secondary phase and is consequently present as large conglomerates in addition to be-
Liquidus projection of the Au-Pb-Sn ternary system. The rst phase to form on solidication is labeled for each phase eld. The 4% Au isoconcentration line is marked on the gure as is the tie line between lead-tin eutectic solder and pure gold. A concentration of 4% gold in the eutectic is taken as a safe limit because AuSn4 cannot form as a primary phase. Adapted from Humpston and Davies [1984, 1985]; Humpston and Evans [1987]
Fig. 2.50
90 / Principles of Soldering
ing a constituent of the ternary eutectic. Above 8% Au, AuSn4 is the primary phase formed on solidication and becomes the dominant constituent of the solder microstructure. At even higher concentrations of gold (above about 13% Au), the solder constitution changes abruptly with gold-tin phases being the major components of primary, secondary, and tertiary solidication. When gold-tin phases form as the primary or major phase, they embrittle the soldered joints on account of their intrinsically low fracture toughness and the weak interface between them and the lead-phase in lead-tin eutectic [Harding and Pressly 1963]. A maximum gold concentration of 4% is usually taken as a safe working limit in industry. As can be seen from Fig. 2.50, highlead and high-tin content in lead-tin solders can accommodate slightly higher levels of gold before gold-tin intermetallic compounds become dominant (i.e., constitute the primary phase) and cause joint embrittlement. Likewise, from the appropriate sections of the phase diagrams of other tin-base solder alloys in combination with gold, it is possible to derive safe working limits for them in a similar manner.
Low concentrations of the AuSn4 phase actually enhance the mechanical properties of many tincontaining solders, including lead-tin [Wild 1968]. This strengthening is illustrated in Fig. 2.52 for the Ag-96Sn solder. From a concentration of about 3.5 to 8% Au, the solder microstructure is characterized by a ne dispersion of the AuSn4 phase, present as a secondary phase. However, when the level of gold in the solder rises toward 10%, there is a sudden change in properties and ingots of the solder become completely unworkable. This change corresponds to the appearance of AuSn4 as the primary phase; that is, AuSn4 is the rst solid phase to form on solidication and consequently adopts a massive form. Above 19% Au the primary phase is AuSn2, and at even higher gold concentrations (greater than 38%) it is AuSn. The safe working limit for this alloy system is usually taken to be 8% Au. The ternary Ag-Au-Sn phase diagram and relevant pseudobinary section between eutectic silver-tin solder (Ag-96.5Sn) and gold are shown in Fig. 2.53 and 2.54. It is noteworthy that although the Ag-96Sn alloy contains more tin than lead-tin eutectic solder, it is tolerant to approximately twice the volume fraction of gold before the al-
Vertical section through the Au-Pb-Sn ternary system between eutectic lead-tin (Pb-62Sn) composition and gold, with the 8% Au concentration marked by a dashed line. The plan view of this section is marked by a dashed line on the liquidus projection of the Au-Pb-Sn ternary system shown in Fig. 2.50.
Fig. 2.51
loy is embrittled by AuSn4. This is thought to be due to a better match between the atomic lattices of Ag3Sn and AuSn4 than between lead and AuSn4. The critical levels of gold that give rise to AuSn4 as the primary phase are listed in Table 4.3 for a selection of solders. Soldering processes that involve gold metallizations are discussed in Chapter 4, section 4.1.3 and Chapter 3, section 3.3.8.1. The mechanical properties of joints containing intermetallic phases can be inferred from a phase diagram according to whether they are compounds of exact stoichiometric composition (i.e., they are in integral atomic ratios of their constituents) or exist over a range of compositions. Exact stoichiometric compounds tend to form when one of the two elements is strongly metallic in character and the other signicantly less so, in terms of the density of free electrons that bind the atoms of the metal together. AuSn4, FeSn2, and Cu6Sn5 are typical examples of this type of compound. These compounds tend to be hard and brittle. Moreover, because their crystal structures are frequently of low symmetrythat is, they deviate from simple cubic or hexagonal structuresthe interfaces of these compounds with other phases tend to be weak. These characteristics are transferred to the joint unless the compounds form as a ne dispersion within the solder.Therefore, their occurrence should be minimized or, even better, avoided wherever possible.
Compounds that are stable over a range of compositions tend to be ductile and have crystal structures exhibiting high symmetry, as do most elemental metals. Therefore, they tend to have a benign effect on joint properties. An example of such a compound is Ag3Sn, which is stable over
Fig. 2.53
phase eld.
Liquidus surface of the Ag-Au-Sn system. The rst phase to form on solidication is labeled for each
Fig. 2.52
Vertical section through the Ag-Au-Sn ternary system between eutectic silver-tin solder (Ag-96.5Sn) and gold. The plan view of this section is marked by the dashed line in Fig. 2.53.
Fig. 2.54
92 / Principles of Soldering
the composition range from 13 to 20%Ag at room temperature and has a hexagonal close-packed crystal structure. This compound forms as an interfacial layer when silver-tin solder is used to join silver-coated components in a manner analogous to the reaction between leadtin eutectic and copper. An example is shown in Fig. 2.55, and the associated binary phase diagram in Fig. 2.9.The rate of growth of the Ag3Sn layer decreases exponentially with time at a xed temperature as shown in Fig. 2.56. The growth of this intermetallic layer correlates with the measured rate of erosion of silver by tin (Fig. 2.35). The growth of the Ag3Sn layer progressively reduces the dissolution of silver. For this reason, the silver-tin eutectic solder, used in conjunction with silver-coated substrates,
forms the basis of a soldering process that is highly tolerant to the processing time and temperature, inasmuch as the risk of totally dissolving a fairly thin (10 m, or 0.4 mil) silver metallization can be minimized.
2.3.3
Micrograph revealing a continuous interfacial layer of the intermetallic compound Ag3Sn formed on reaction between tin-silver solder and silver. Magnication: 20
Fig. 2.55
More often than not, higher-order alloy systems are encountered in joining, because both the solder and the parent materials usually each have a minimum of two constituents. Combinations involving ve or even larger numbers of elements are not uncommon. The denition of the plethora of phases that can exist in these higher-order systems represents a daunting task. In order to make the problem more tractable, a reductive approach is often employed. This method usually involves partitioning the multicomponent system into a series of quasi-binary or quasi-ternary alloy systems, each containing different but xed proportions of the other components and ascertaining these sections of the relevant phase diagrams, in turn. Much care should be exercised in extracting quantitative information from these partial phase diagrams because the tie lines, triangles, quadrilaterals, and so forth that are used with the lever rule to determine the relative proportions of
Fig. 2.56
Thickness of the Ag3Sn intermetallic layer formed by reaction between Ag-96Sn solder and silver as a function of reaction time and temperature. After Evans and Denner [1978], with authors own data
phases present often do not lie in the plane of the selected sections.
low-melting-point aluminum brazing alloys, are given in the planned companion volume Principles of Brazing.
2.4
2.4.1
The number of commercially available solders is nite, and it is not uncommon to have an application where it would be desirable either to extend the lower temperature limit on the use of a particular family of solders or to identify ller compositions that melt within a specied range. An obvious case in point is the search for replacements for lead-tin solder that do not contain lead, but that can be used at comparable processing temperatures. For the reasons elaborated in section 2.3 in this chapter, eutectic alloys possess several key characteristics that make them a natural choice for llers, namely superior spreading behavior when molten, with complete melting occurring at a single temperature that is lower than, and usually well below, that of either constituent metal. This property of instantaneous melting enables joining operations to be carried out at temperatures only slightly in excess of the solidication temperature, and molten eutectic alloys generally possess a high degree of uidity. Eutectic alloys also possess favorable mechanical properties arising from their well-distributed duplex or higher-order microstructure. As shown for the particular example of the Bi-Pb-Sn solder in Fig. 2.16, a eutectic alloy formed of three constituent metals (that is, a ternary eutectic) always has a melting point that is lower than the three constituent binary eutectics. What then happens to the melting point if further constituents are added? In particular, can it be lowered at will and, if not, how might one determine the limits? Answers to these questions have been provided by analyzing the pattern of behavior observed in two simple examples, namely the development of cadmium solders and gallium alloy systems intended as substitutes for mercury. From their general features it is possible to devise certain empirical ground rules of general applicability that can be applied to other cases, such as the quest for lead-free solders with a similar melting point to lead-tin eutectic (183 C, or 361 F). Other examples, pertaining to
There are requirements for metals that are liquid at normal ground temperatures. Mercury is the only metallic element that is molten down to subzero temperatures (freezing at 39 C, or 38 F). It has therefore enjoyed a near monopoly in thermometry for more than two centuries as well as in other applications employing liquid metals such as mirrors, high-current switches, and slip rings. With the growing awareness of health and safety issues in recent years, there has been a move away from the use of mercury in these applications, mostly to functional substitutes, rather than toward nding direct replacements. Some direct mercury-free alloy substitutes have been developed, based on gallium eutectics, gallium being the secondlowest-melting-point metal next to mercury. Thus, for example, Ga-In-Sn, melting at 12 C (54 F), and Ga-In-Sn-Zn, melting at 9 C (48 F), have been used in high-current switches [Walkden, Kowalczyk, and Cooke 1978]. The progressive reduction of the melting point of gallium alloys, as the number of constituents is increased, is shown in Tables 2.12 and 2.13.
2.4.2
Cadmium-Base Solders
Cadmium forms a series of binary eutectic alloys with other common solder ingredients that melt at temperatures down to 123 C (253 F), as shown in Table 2.14. The alloying elements selected in this example are indium, tin, and zinc, which all melt at relatively low temperatures and form binary eutectics with one another, as well as with cadmium. In Table 2.14, it can be seen that ternary alloys of cadmium with these elements span a melting range from 163 to 93 C (325 to 199 F) and on down to 90 C (194 F) for the quaternary eutectic in the Cd-In-Sn-Zn system.
2.4.3
General Features
The melting point behavior follows the following pattern for the three alloy systems considered here:
94 / Principles of Soldering
ing element that enters into the eutectic reaction becomes progressively smaller, so that the melting point tends to an asymptotic minimum. These features are consistent with elementary thermodynamic and statistical models, as ex-
Table 2.12 Temperature of eutectiferous phase transformations in which one of the participating phases is gallium
Order Alloy system Melting point, C Temperature depression, T
1 2
4 5
(a) Authors own measurements
Ga Ga-Ag Ga-In Ga-Sn Ga-Zn Ga-Ag-In Ga-Ag-Sn Ga-Ag-Zn Ga-In-Sn Ga-In-Zn Ga-Sn-Zn Ga-Ag-In-Sn Ga-Ag-In-Zn Ga-In-Sn-Zn Ga-Ag-In-Sn-Zn
0 5 14 9 4 16 9 10 20 17 13 21 21 21 24
Table 2.13
Order
3 4
0 3 3 2 2 2 2 3
Alloy system
1 2
Cd Cd-In Cd-Sn Cd-Zn In-Sn In-Zn Sn-Zn Cd-In-Sn Cd-In-Zn Cd-Sn-Zn In-Sn-Zn Cd-In-Sn-Zn
321 123 177 266 117 142 199 93 116 163 108 90
0 198 144 95 204 179 122 228 205 158 213 231
plained in Appendix A2.2 to this chapter, and are generic to eutectic alloying. From a practical point of view, the implications are clear. When seeking a reduction in the melting point of a pure metal or of an existing eutectic alloy, for use in soldering or other applications, there is a trade-off between keeping the number of constituents low and judiciously chosen to optimize the melting-point depression and increasing the number of components, which is likely to produce only a relatively small further reduction in the melting point. Multicomponent solder are beset by disadvantages, chief among which is a considerably more complex phase diagram, if it is indeed known, and often, a harder and less workable alloy. On account of the cumulative complications associated with increasing the number of constituents in the ller, it is advantageous, where possible, to limit the choice to alloying additions that are most effective at depressing the melting point. Generally, these are elements that have low melting points and very limited solid solubility in the host metal. The implications from the foregoing discussion are not encouraging with regard to the search for drop-in replacements for eutectic lead-tin solders, as explained below. Further details on lead-free solders are given in Chapter 5, section 5.1.
drawn about the solidication of tin-rich alloys and are described below. Those tin-rich multicomponent alloys that solidify eutectiferously do so at a temperature that is either above 200 C (392 F) or signicantly below 150 C (302 F). The available range of suitable alloying additions will either only lower
Table 2.15 Alloying sequences that show that drop-in replacements for lead-tin solders, based on tin, are unobtainable
A suitable replacement for lead-tin eutectic solder would need to meet the following criteria: Contain only inexpensive and nonvolatile and nonhazardous constituents Be a eutectic with a melting point between 150 and 200 C (302 and 392 F) 1. Candidate constituent binary systems and their eutectic temperatures are: Ag-Sn (221 C) Bi-Sn (139 C) Cu-Sn (227 C) In-Sn (120 C) Sn-Zn (199 C) Another binary alloy that might provide the basis for a multicomponent solder is the noneutectic, namely: Sn-6Sb, melting range: 232250 C (450482 F). 2. Candidate ternary systems for replacing eutectic lead-tin solder: Ag-Cu-Sn, ternary eutectic at 217 C [Miller, Anderson, and Smith, 1994] Ag-Sb-Sn(a), ternary transition reaction terminating on the Ag-Sn binary eutectic (221 C) Ag-Sn-Zn(a), ternary transition reaction terminating on the Sn-Zn binary eutectic (199 C) Cu-Sb-Sn, ternary transition reaction terminating on the Cu-Sn binary eutectic (227 C) Cu-Sn-Zn(a), ternary transition reaction terminating on the Sn-Zn binary eutectic (199 C) All other combinations result in ternary eutectic alloys melting below 150 C, thus: Ag-In-Sn(a) (113 C) Bi-Sn-Zn(a) (130 C) In-Sn-Zn(a) (108 C) or, in the case of the Ag-Bi-Sn and Bi-Sb-Sn alloys, terminate on the Bi-Sn binary eutectic composition (139 C) [see Kattner and Boettinger, 1994 for the Ag-Bi-Sn system and Ohtani and Ishida 1994 and Ghosh, Loomans, and Fine, 1994 for the Bi-Sb-Sn system] 3. Quaternary alloys based on the ternary alloys that melt above 150 C Freezing of these alloys concludes at the binary eutectic melting point. Thus, for example: Ag-Cu-Sn-Zn(a) undergoes a ternary transition reaction terminating on the Sn-Zn binary eutectic.
(a) Authors own measurements
2.4.4
The wealth of published studies of alternative solders has generally arrived at a consensus view that alloys that provide the closest alternatives to lead-tin eutectic solder, which can meet the acceptance criteria for printed circuit board (PCB) manufacture are tin-rich with the minor constituents being one or more of the following: antimony, bismuth, copper, silver, and zinc [Hampshire 1993]. Ideally, the replacement solder should have a melting point close to that of leadtin, lying in the range between 150 and 200 C (302 and 392 F), and be of eutectic composition. Here, the objective is not to minimize the melting temperature, but to restrict it to within a prescribed temperature window. Because none of the known binary eutectic alloys of tin satisfy the above criteria, a study was made of higher-order systems to establish whether these might contain suitable eutectic compositions. The results are given in Table 2.15. From the experimental data, conclusions can be
96 / Principles of Soldering
the melting point slightly below that of tin, melting point 232 C (450 F), as in the case of the 3.5Ag-0.9Cu-95.6Sn ternary eutectic, melting point 217 C (423 F) [Loomans and Fine 1999], or will effect a radical reduction in melting point, for example, alloying with indium and/or bismuth [Yoon et al. 1999; Hassam, Dichi, and Legendre 1998]. A eutectic alloy of silver-tin with gold, of composition 3.6Ag-3.6Au-92.8Sn, has a melting point of 206 C (403 F), but the cost premium associated with the gold content would be unacceptable for the manufacturing industry. The Sn-9Zn eutectic alloy, melting at 199 C (390 F), may be judged unsuitable for most soldering applications because of its propensity to corrode with the generation of a conductive white zinc chloride surface lm when subjected to a chlorine-containing atmosphere. Solidication of all other tin-rich multicomponent alloys occurs via one or more transition reactions in a sequence that terminates on one of the tin binary eutectics, that is, at a temperature close to that of tin and above 200 C (392 F). An example is provided by the Bi-Sb-Sn system [Ohtani and Ishida 1994; Ghosh, Loomans, and Fine 1994]. It has been shown that some of the multicomponent alloys that are being put forward as dropin replacements for lead-tin are handicapped by the presence of a low-melting-point eutectic fraction that can severely compromise the reliability of soldered joints if segregation occurs toward the low-melting-point alloy composition, either continuously at joint interfaces or along grain boundaries [Vincent and Humpston 1994]. One of these alloys is the Ag-20In-77Sn alloy being offered as a lead-free solder, which substantially melts at 177 C (351 F). However, a signicant fraction of this alloy melts at a much lower temperature of 113 C (239 F) [Korhonen and Kivilahti 1998], associated with a low-meltingpoint Ag-In-Sn composition, close to the In48Sn binary eutectic [Artaki, Jackson, and Vianco 1994]. Likewise, the In-86Sn-9Zn solder largely melts between 188 and 198 C (370 and 388 F), but it contains a proportion of the In46Sn-2Zn ternary eutectic melting at 108 C (226 F) [McCormack, Jin, and Chen 1994]. It may reasonably be concluded that there are no higher-order tin-base eutectics, without gold, that solidify in the target temperature range. Hence, the multicomponent alloys offer no major advantages over the binary eutectic alloys of tin, with regard to melting point. It follows that leadfree solders that are developed in the future are
likely to be based on the silver-tin binary alloy system, which has a eutectic temperature of 221 C (430 F). Economic and functional benets are obtained by substituting some of the silver for copper and exploiting the Ag-Cu-Sn ternary alloy system, which has a eutectic temperature at 217 C (423 F), as mentioned above [Miller, Anderson, and Smith 1994; Vincent and Humpston 1994]. More information on the rationale for this choice is given in Chapter 5, section 5.1.
Appendix A2.1: Conversion between Weight and Atomic Fraction of Constituents of Alloys
In an alloy containing N constituents, conversion from weight to atomic fraction of constituent, n, may be made using:
Pn /An
at.% n
i1
/ Ai
P'i
100
where P is the weight percentage of the constituent denoted by the subscript, A is the atomic weight of the constituent denoted by the subscript, subscript n refers to constituent n, and subscript i refers to each constituent in turn. Similarly, in an alloy containing N constituents, conversion from atomic to weight fraction of constituent, n, may be made using:
P'n An
wt% n
i1
Pi Ai
100
where P' is the atomic percentage of the constituent denoted by the subscript, A is the atomic weight of the constituent denoted by the subscript, subscript n refers to constituent n, and subscript i refers to each constituent in turn.
tion, Eq A2.1 can be extended stepwise to multicomponent alloys, where, for example, the binary alloy A B may be considered as the solution matrix AB of a pseudobinary alloy AB C. Then Eq A2.1 may be rewritten in the form:
H(2) m R
1 T (3) m
1 T (2) m
) ( )
ln x2 x* 2 x2 x* 2 x1 x* 1
1 T (2) m
1 T (1) m
) ( )
ln x1 x* 1
where x1 is the mole fraction of component A of , and the liquidus composition at temperature T (2) m * xi is this mole fraction at the limit of solid solubility at the same temperature. T (1) is the melting m (1) temperature of pure A, and H m is its latent heat of fusion. R is the universal gas constant. The latent heat Hm (or enthalpy of fusion) of a metal is proportional to its melting point. This is because entropies of fusion (Sm) have similar values for all metals (at roughly 10 J/K mol), and to a rst approximation, in the absence of any phase changes:
Hm Tm Sm 0 (Eq A2.1)
The relationship x2x * x1x * implies that 2 1 for the third component, C, the attainable liquidus depression T (3) T (2) is in most cases sigm m nicantly smaller than for the second component B, equal to T (2) T (1) . If this sequential prom m cedure is applied to additional elements, the general formula for the ith constituent and the corresponding liquidus temperature, T im is obtained:
i1) H( m
i) T (m
1
i1) T (m
) ()
ln xil xi*1
(Eq A2.2)
that is:
Hm Tm
Equation A2.1 therefore implies that the attainable melting-point depression is determined by the melting point of the addition; the lower the melting point of the addition, the lower will be the melting point of the resulting eutectic (assuming that one exists). The effect on melting point of multiple alloying additions, all with similar melting points, can be deduced as follows. As a simple approxima-
This expression is a member of a series, of decreasing size for increasing value of i. The melting-point depression is maximized for a large difference between the concentration of the ma. trix xi1 and its solid solubility limit x* i1 Despite the fact that this model considerably oversimplies reality, it accounts for the two principal features in common with the experimental results on eutectic alloys, namely:
The progressive reduction in melting temperature as the number of alloying constituents is increased The asymptotic narrowing of the meltingpoint depression of the alloy with the introduction of each additional constituent
98 / Principles of Soldering
Composition-specic thermochemical data on multicomponent alloys (Hm, xi) are needed in order to apply Eq A2.2 in calculating the liquidus temperature depression through progressive alloying. This information is mostly unavailable for ternary and higher-order systems. However, from knowledge of thermochemical data for pure metals and binary alloys, it can be inferred that the respective values will differ widely from one metal to another, and therefore large variations in temperature drop are to be expected between different alloying additions. The physical picture of this behavior may be more clearly understood in terms of the entropy changes accompanying progressive alloying. At the microscopic level, a material system may be viewed as an ensemble of atoms or molecules and, on this basis, entropy provides a measure of the degree of atomic or molecular disorder in the system, according to the relationship:
S k ln
there is a tendency for the entropy to increase as the number of constituents in the material is increased, as follows: If Ni 10 for all values of i, then Stirlings Formula can be applied, whereby:
ln Ni! Ni ln Ni Ni
and the change in entropy in increasing the number of constituents from two (binary system) to three (ternary system) is:
S2,3 kN [ln 3 ln 2] R [ln 3 ln 2]
in the particular case where N1 N2 12 N for the binary system and N1 N2 N3 13 N for the ternary system. Then, on increasing the number of constituents from three to four (quaternary system), the entropy rises further by an increment:
S3,4 R [ln 4 ln 3] S3,4 S2,3
where S is entropy, k is the Boltzmann constant, and is the degree of disorder, as measured by the number of different distributions available to the atoms or molecules in the system. In the simplest case, where the volume of the material is shared by i different species of atom, representing different constituents of an alloy, each present in amounts N1, N2, N3, , N i, such that:
N1 N2 N3 N i N
where N1 N2 N3 Ni1 Ni (1/i) N and i 2. The pattern is established where each successive addition increases the entropy of the system overall, but by progressively smaller amounts, as shown by the data in Tables 2.12 to 2.14. In other words, each additional constituent has a relatively smaller effect on the degree of disorder of the system, as one might intuitively expect. Entropy, S, is related to the Gibbs free energy, G, by the relationship:
then assuming that the different species of atom are equally interchangeable, the number of ways, , in which all the atoms may be arranged among the N available sites is:
N! N1! N2! N3! Ni!
and
S k ln N! N1! N2! N3! Ni!
()
dG dT
As the number of constituents increases, while keeping the total number of atoms, N, constant,
Therefore, as the entropy increases, the depression of the Gibbs free energy of a system as a
function of temperature increases. This in turn will tend to depress its melting point, although the actual relationship will be governed by the specic free energies of the constituents in the molten and solid states and of the solution that they form. In general terms, the picture provided by this elementary expression is consistent with that furnished by Raoults law and the ClausiusClapeyron equation. At rst sight, this model may appear inappropriate for a eutectic alloy. However, the special case of a eutectic alloy with a low degree of intersolubility of the pure metal constituents in the solid approximates reasonably well, insofar as each phase is tantamount to a pure constituent and is well dispersed throughout the alloy. This model therefore serves as a crude, but nevertheless graphical illustration of the physical effect of progressive eutectic alloying.
REFERENCES
Solder, 1920 June (Atlantic City, NJ), American Society for Testing Materials, p 129149 Bulwith, R.A. and Mackay, C.A., 1985. Silver Scavenging Inhibition of Some Silver Loaded Solders, Weld. J., Vol 64 (No. 3), p 86s90s Cheng, S. and Lin, K., 2002. The Thermal Property of Lead-Free Sn-8.05Zn-1Ag-xAl Solder Alloys and Their Wetting Interaction with Copper, J. Electron. Mater., Vol 31 (No. 9), p 940945 Crispin, R.M. and Nicholas, M.G., 1984. Reactive Brazing of Alumina to Metals, Brazing Soldering, Vol 6 (No. 1), p 3739 Darveaux, R. and Murty, K.L., 1992. Effect of Deformation Behaviour on Solder Joint Reliability Prediction, Proc. Conf. Microstructures and Mechanical Properties of Aging Materials, 25 Nov (Chicago), p 399 408 Dirnfeld, S.F. and Ramon, J.J., 1990. Microstructural Investigations of Copper-Tin Intermetallics and the Inuence of Layer Thickness on Shear Strength, Weld. J., Vol 69 (No. 10), p 373s377s Dogra, K.S., 1985. A Bismuth Tin Alloy for Hermetic Seals, Brazing Soldering, Vol 9 (No. 3), p 2830 DuPont Electronics, 1998. Low Temperature Braze System 5087D Braze Alloy Composition, Technical Information Sheet, Sept Edwards, L.K., Nixon, W.A., and Lakes, R.S., 2000. Development of a Low-Cycle Fatigue Life Curve for 80In15Pb5Ag, J. Electron. Mater., Vol 29 (No. 9), p 10841089 EP&P 1992. Reduction in Antimony Puts Solder Joints at Risk, Electron Packag. Prod., Vol 32 (No. 2), p 20 Evans, D.S. and Denner, S.G., 1978. An Apparatus for the Determination of Solid/Liquid Metal Interactions under Controlled Conditions, Pract. Metallogr., Vol 15, p 486493 Felton, L.E., et al., 1991. n-Cu6Sn5 Precipitates in Cu/Pb-Sn Solder Joints, Scr. Metall. Mater., Vol 25 (No. 10), p 23292333 Fields, R.J., Low III, S.R., and Lucey Jr., G.K., 1991. Physical and Mechanical Properties of Intermetallic Compounds Commonly Found in Solder Joints, Proc. TMS Fall Meeting, 2024 Oct (Cincinnati, OH), p 165173 Finch, N.J., 1985. Soft Soldering Aluminium and Its Alloys, Brazing Soldering, Vol 9 (No. 59), p 4649
ing of Dissimilar Metals, Weld. J., April, p 127s136s Humpston, G. and Davies, B.L., 1984. Thermal Analysis of the AuSn-Pb Quasibinary Section, Met. Sci., Vol 18 (No. 6), p 329 331 Humpston, G. and Davies, B.L., 1985. Constitution of the AuSnPbSn Partial Ternary System, Mater. Sci. Technol., Vol 1 (No. 6), p 433441 Humpston, G. and Evans, D.S., 1987. Constitution of the Au-AuSn-Pb Partial Ternary System, Mater. Sci. Technol., Vol 3 (No. 8), p 621627 Humpston, G., et al., 2001. Compliant and Hermetic Solder Seal, U.S. Patent 10,020,018, Oct Humpston, G. and Jacobson, D.M., 1993. Principles of Soldering and Brazing, p 9195 Humpston, G. and Jacobson, D.M., 1995. Do 18 Carat Gold Solders Exist?, Gold Bull., Vol 27, p 110116 Indium Corporation of America, 1994. Research Solder Kits brochure Inoue, H., Kurihara, Y., and Hachino, H., 1986. Pb-Sn Solder for Die Bonding of Silicon Chips, IEEE Trans. Components, Hybrids Manuf. Technol., Vol 9 (No. 2), p 190 194 Irving, B., 1992. Host of New Lead-Free Solders Introduced, Weld. J., Vol 71 (No. 10), p 4749 Jacobson, D.M. and Humpston, G., 1989. Gold Coatings for Fluxless Soldering, Gold Bull., Vol 22 (No. 2), p 918 Jacobson, D.M. and Sangha, S.P.S., 1997. Novel Low Expansion Packages for Electronics, GEC J. Technol., Vol 14, p 4855 Johnson, A.A. and Johnson D.N., 1983. The Room Temperature Dissociation of Au3Si in Hypoeutectic Au-Si Alloys, Mater. Sci. Eng., Vol 61, p 231235 Johnson, D.N. and Johnson, A.A., 1984. Surface Cracking in Gold-Silicon Alloys, Solid State Electron., Vol 27 (No. 12), p 11071109 Kattner, U.R. and Boettinger, W.J., 1994. On the Sn-Bi-Ag Ternary Phase Diagram, J. Electron. Mater., Vol 23, p 603610 Kay, P.J. and Mackay, C.A., 1976. The Growth of Intermetallic Compounds on Common Base Materials Coated with Tin and TinLead Alloys, Trans. Inst. Met. Finish., Vol 54, p 6874
(Atlantic City, NJ), American Society for Testing Materials, p 149157 Xian, A. and Si, Z., 1991. Wetting of TinBased Active Solder on Sialon Ceramic, J. Mater. Sci. Lett., Vol 10 (No. 22), p 1315 1317 Yoon, S.W., et al. 1999. Investigation of the Phase Equilibria in the Sn-Bi-In Alloy System, Metall. Mater. Trans. A, Vol 30A, p 15031515 Yost, F.G., Ganyard, F.P., and Karnowsky, M.M., 1976. Layer Growth in Au-Pb/In Solder Joints, Metall. Trans., Vol 7A (No. 8), p 11411148 Yost, F.G., Hosking, F.M., and Frear, D.R., 1993. The Mechanics of Solder Alloy Wetting and Spreading, Van Nostrand Reinhold, p 130134
CHAPTER 3
3.1
Joining Atmospheres
There are many types of assembly that demand soldering under a protective atmosphere. These include assemblies intended for service in a vacuum environment, which must be free from volatile contaminants and parent metal components that are disgured by oxide scale. The categories of joining atmospheres that are available and their interrelationships are shown in Fig. 3.1. Generally, uxes are needed only when carrying out the joining operation in air or other oxidizing environments.
Chemically inert gas atmospheres (e.g., argon, nitrogen, helium, vacuum). These function by excluding oxygen and other gaseous elements that might react with the components to form surface lms and inhibit owing of and wetting by the solder. Chemically active atmospheres, both gases and uxes, which are designed to react with surface lms present on the components and/ or the ller metal during the joining cycle and remove them in the process. These atmospheres may either decompose surface lms (as does hydrogen when acting on certain oxide or sulde layers, for example) or react with the lms to produce compounds that can be displaced by the molten ller metal. Traditional rosin uxes predominantly function in this manner. Controlled gas atmospheres require a conning vessel, and this invariably means a furnace of some type. Furnace joining also offers other advantages:
suitable for parts that are likely to distort if heated locally. The atmospheric protection afforded leads to economies on the use of ux and on nishing operations, such as cleaning and the removal of ux residues. Against this must be considered the following potential disadvantages:
The process may be easily automated for either batch or continuous production, because the heating conditions can be accurately controlled and reproduced without the need for much operator skill. Furnace joining offers uniform heating of the components of almost any geometry and is
Fig. 3.1
Compound semiconductors (e.g., gallium arsenide) can have their electrical functionality poisoned by hydrogen, while some brasses are intolerant of ammonia. Thus, the requirements of each component in an assembly must be individually assessed and the atmosphere chosen to suit.
dG VdP SdT
3.1.1
A principal process requirement for successful soldering is to ensure that the joint surfaces are free from oxides and other lms that can inhibit wetting by the molten solder and the formation of strong metallic bonds. The ability to remove a layer of oxide from a given metal depends on the ease of either physically detaching the lm from the underlying metal or of chemically separating the oxygen ions from the metallic ions present in the oxide, that is, the strength of the relevant molecular bonds. Chemical reduction of metal oxide by atmospheres is considered rst. Chemical thermodynamics can be used to determine the propensity for a metal to spontaneously oxidize or, conversely for an oxide to disassociate. An understanding of oxide reduction by gas atmospheres is very relevant to brazing, and for that reason an introduction to the subject is provided in the planned companion volume Principles of Brazing. At typical soldering temperatures there is insufficient driving force available to reduce or prevent oxidation of almost all metals, save those classied as noble and under limited circumstances, possibly also extending to silver and copper. For this reason, only a very supercial treatment of chemical thermodynamics is provided here, sufficient only to grasp the most important fact that gaseous atmospheres, including pure hydrogen, are generally inert toward oxides of solders and common substrate materials at normal soldering temperatures. There are, of course, exceptions, and some of these are elaborated on later in this chapter. A measure of the strength of a metal-tooxygen chemical bond is given by the change in the Gibbs free energy that occurs when that metal reacts to form the oxide, as detailed in Appendix A3.1. Here, it is noted that the Gibbs free energy, G, is an important thermodynamic function in chemistry because incremental changes in its value only involve incremental changes in pressure, P, and temperature, T, for reversible reactions:
V is the volume and S is the entropy, which is explained further in Appendix A3.1 Chemical reactions, such as oxidation and reduction, which are reversible, can take place at constant pressure and temperature, so that the Gibbs free energy of the material system does not then change in the course of the reaction. Table 3.1 shows the Gibbs free energy of formation of oxides for a selection of metals at room temperature. This formation energy is sometimes referred to reciprocally as the dissociation potential of the oxide. The least stable metal oxides are those of the noble metals, gold, silver, and members of the platinum group. These metals are therefore the most readily soldered, while the refractory metals and the light metalsnotably aluminum, beryllium, and magnesiumhave particularly stable oxides so that these metals are the most difficult to join. It is precisely because gold will not form a stable oxide in air that it is widely used as a surface coating for uxless joining processes (see Chapter 3, section 3.3.8.1). Other factors need to be considered in connection with oxide reduction. In particular, many metals form different oxides of varying stabilityfor example, cuprous oxide (Cu2O) and cupric oxide (CuO). Furthermore, oxides formed on alloy surfaces are not generally pure metal
Table 3.1 Comparative values for free energies of formation of metal oxides of common solder constituents and selected metals at room temperature (25 C or 77 F)
The more negative the value, the more stable the oxide.
Free energy of formation at 25 C (77 F), kJ/mol (rounded values)
Element
Common oxide
Gold Silver Copper Bismuth Lead Tin Zinc Antimony Indium Chromium Titanium Aluminum
Au2O3 Ag2O CuO Cu2O BiO Bi2O3 PbO PbO2 Pb3O4 SnO SnO2 ZnO Sb2O3 Sb2O4 Sb2O5 In2O3 Cr2O3 TiO2 Al2O3
50 10 130 150 170 460 190 210 570 260 490 300 580 740 780 620 700 840 1580
oxides but rather compound or other forms of mixed oxide. Often these are of nonuniform composition and structure, adding further complexity to the subject. This is particularly true of solders that are almost inevitably multicomponent alloys. In its present state of development, chemical thermodynamics is not able to predict accurately conditions under which dissociation of oxides will occur, but can only provide a semiquantitative indication, particularly when the kinetics of reaction are taken into account. Owing to such complexities, the thermodynamic principles for analyzing oxide reduction are considered only for pure metals.
3.1.2
All chemical reactions are reversible, including oxidation reactions. In general, the oxidation of any metal can be described by an equation of the form:
nM m 2 O2 MnOm
The reaction will proceed spontaneously in either directionnamely, oxidation of the metal, or, conversely, reduction of the oxide, if it is energetically favorable to do so. A condensed treatment of relevant thermodynamic functions and their relationships, which has by necessity required a degree of oversimplication, is given in Appendix A3.1 at the end of this chapter. A more rigorous treatment is given in standard textbooks on thermodynamics, such as those listed in the Selected References given in the Preface. The free-energy change (G) for oxidation reactions involving a series of metals can be charted on a diagram as a function of temperature, as shown in Fig. 3.2. This representation is known as an Ellingham diagram or a Richardson-Jeffes diagram. The diagram can be used to determine whether, in principle, an atmosphere is capable of reducing surface oxides, although it does not provide any indication of the kinetics of the reactions. The use of the Ellingham diagram in soldering practice is described in the next section. At any given temperature, the smaller the equilibrium partial pressure of oxygen in the metal oxide, the stronger the bond between the oxide
Simplied Ellingham diagram showing the free-energy change for oxidation of several metals. Oxide stability is reduced by elevated temperature and decreased oxygen partial pressure. Each dashed line corresponds to the Gibbs free-energy change as a function of temperature, relating to a particular oxygen partial pressure. mpt, melting point
Fig. 3.2
and the parent metal, that is, the greater is the stability of the oxide. The partial pressure of a gas in an atmosphere is dened in Chapter 1, section 1.3.2.5. Thus, the tendency for the oxide to decompose will be greater the lower the oxygen content of the atmosphere and the higher the temperature. The further down the diagram a particular metal-oxygen reaction curve lies, the more inherently stable is the oxide, and it is correspondingly more difficult to reduce; in other words, higher temperatures and atmospheres of lower oxygen content are required to effect reduction.
3.1.3
3.1.3.1
For many metals, heating alone in air is not adequate to reduce the oxide, because the components are degraded or even melt before the critical temperature, Tc, is reached at which the oxide will spontaneously decompose. Moreover, the rate of oxidation roughly doubles with each 25 C (45 F) rise in temperature. Thus, stable oxides become progressively thicker and more tenacious, and consequently more difficult to remove, over the time interval that the component is being heated to the critical temperature. Excessive oxidation can damage component surfaces, particularly if the lm spalls off locally, because the rate of oxidation will be nonuniform over the surface, producing an unsightly nish. For these reasons, where uxes are not employed, it is usual practice to heat the components in a protective atmosphere or vacuum, which will both protect the surfaces from further oxidation and reduce the partial oxygen pressure, and hence the critical temperature. The conditions of temperature and oxygen partial pressure required to spontaneously reduce a metal oxide can be deduced from the Ellingham diagram. Reduction will occur when the free-energy curve for metal-oxide formation lies above the oxygen partial pressure curve at the temperature of interest; that is, the oxygen pressure in the atmosphere is less than that which will cause the metal under consideration to oxidize. Thus, the critical temperature for the reduction of PdO decreases from 920 C (1688 F) in pure oxygen at atmospheric pressure to 380 C (715 F) if the oxygen partial pressure is decreased to 1010 atm (102 mPa). It can be seen from the more detailed
Ellingham diagram reproduced in Fig. 3.3, that oxide reduction in vacuum is practicable only for palladium and silver at normal soldering temperatures (<450 C, or 840 F). The minimum partial oxygen pressure that can be achieved using highquality industrial equipment is of the order of 1010 atm (102 mPa). Note that it is convenient to use the atmosphere as the unit of pressure in thermodynamic calculations, and this convention is also applied to Ellingham diagrams. For metals having oxidizing reaction curves that are located below the 1010 atm (102 mPa) oxygen partial pressure curve (that is, a line joining the point O on the T 273 C axis on the left, to the 1010 atm value on the partial oxygen pressure (PO2 ) scale on the right of the Ellingham diagram, as shown in Fig. 3.2), it will be energetically favorable for the metal to oxidize by reaction with the residual oxygen and any water vapor present in the furnace atmosphere. This includes most common solder constituents, including tin and indium. Thus, industrial-quality vacuum and inert atmospheres are incapable of preventing degradation of solders during normal heating cycles. Obviously, an atmosphere that is largely free of oxygen and water vapor will greatly slow further oxidation, but cannot prevent or reverse it. As mentioned previously, care must be taken to select an atmosphere that is inert toward all of the metals in the assembly being joined. Vacuum can degrade certain materials, notably brass, even at soldering temperatures, due to the loss of zinc through volatilizationa consequence of the high vapor pressure of this element. Likewise, lead-containing solders are unstable in high vacuum at temperatures much above 300 C (570 F) and are not recommended for use under these conditions. Table 3.2 lists the boiling/sublimation temperatures of selected elements at 1010 atm (102 mPa). For metals to be joined under reduced pressure, the process temperature must be considerably less than the boiling/sublimation temperature (by a factor of < in K/K), if volatilization is not to be signicant. The oxygen partial pressure in a vacuum furnace can be reduced substantially below the gas pressure in the vacuum by repeatedly pumping out and backlling the chamber with a dry, oxygen-free gas (see Chapter 1, section 1.3.2.5). Care must be taken to ensure that the inlet system is completely leak-tight. Otherwise, some oxygen will be bled into the furnace, and this will impair or even nullify the benet of the inert atmosphere. A periodic ushing of the chamber with an inert gas will also serve to minimize any
buildup of oxygen released in the dissociation of oxides during the heating cycle. The effectiveness of using the process temperature and oxygen partial pressure to control oxide reduction, or at least prevent oxidation, is
limited by the presence of adsorbed water vapor on the walls of the vacuum chamber and on other free surfaces. The desorption of water vapor effectively increases the oxygen partial pressure in the chamber, and this has a deleterious effect on
Fig. 3.3
Ellingham diagram for selected oxides. M, melting point of metal; B, boiling point of metal; M', melting point of oxide
the oxide-removal process. Therefore, it is good practice to heat the walls of the chamber to promote desorption, while simultaneously removing the vapor from the chamber by alternately pumping out and/or ushing with dry, inert gas before commencing the heating cycle. Additional background information on moisture in closed spaces in given in Appendix A4.2. Nevertheless, as can be seen from the nomogram in Fig. 3.2, the quality of vacuum required to prevent adsorption on to the faying surfaces needs to exceed that which can usually be economically achieved in an industrial process. Large-scale industrial processes often rely on liquid nitrogen for several reasons. Not least of these is the ease of convenience of delivery and storage. Furthermore, nitrogen boiled off from a cryogenic tank containing the liqueed gas possesses lower levels of oxygen and water vapor (typically <2 ppm combined) than all but the purest grades of bottled nitrogen. It is also relatively inexpensive, being comparable in price per liter to bottled mineral water. Owing to increasingly stringent environmental legislation, joining in inert atmospheres is gaining in popularity. This is particularly true in the electronics industry, where the trend is toward nitrogen atmosphere furnaces for both wave and reow soldering processes. In commercial systems, the nitrogen ambient contains less typically than 10 ppm of other species. The running costs associTable 3.2 Boiling/sublimation temperature of selected elements at a pressure of 10!10 atm (10!2 mPa)
Boiling/sublimation temperature Element C F
ated with the large volumes of nitrogen that are required to achieve this quality of atmosphere are offset by the ability to dispense with postjoining treatments because reduced quantities of uxes and cleaning uids are required, with their associated health and environmental problems. For certain applications, inert gases other than nitrogen may be more appropriate. Of these less common inert gases, argon and carbon dioxide are probably the most widely used. Both can be purchased in high-purity form. Carbon dioxide is often recommended in applications where the atmosphere is conned, but open to air at various portals, because the greater molecular weight of carbon dioxide enables it to displace air more effectively than does nitrogen [Esquivel and Chavez 1992]. Argon is more expensive than the other two gases, and its use is therefore largely conned to joining in closed volumes. 3.1.3.2 Soldering in Reducing Atmospheres
If the partial oxygen pressure surrounding the workpiece cannot be sufficiently lowered to effect oxide removal by introducing a vacuum or inert gas environment, then a reducing atmosphere might be able to remove the oxide. The three most widely used reducing gases are hydrogen, carbon monoxide, and cracked ammonia (that is, ammonia dissociated into nitrogen and hydrogen). The basic chemical reduction processes for hydrogen and carbon monoxide are:
yH2 MxOy xM yH2O
Cd Zn Mg Sb Bi In Mn Ag Al Sn Cu Cr Au Pd Fe Co Ni Ti Mo W
100 150 210 300 350 525 550 630 725 730 780 800 880 905 950 1020 1025 1130 1680 2230
212 302 410 572 662 977 1022 1166 1337 1346 1436 1472 1616 1661 1742 1868 1877 2066 3056 4046
and
yCO MxOy xM yCO2
Values are rounded. Note the high position of tin and the low position of manganese and zinc in the table in relation to their melting points.
For convenience to the user, the Ellingham diagram is provided with a series of side scales giving the partial oxygen pressure corresponding to ratios of H2/H2O and CO/CO2, as shown in Fig. 3.3. On the left-hand side is shown an axis for T 273 C (0 K) with points marked at values of the free energies, G, at this temperature for the hydrogen/oxygen (point H), the carbon monoxide/oxygen (point C) and other reactions. Each of these points is associated with one of the side scales shown in Fig. 3.3.
As an example of the use of the Ellingham diagram, consider the conditions for reduction of chromium oxide. The free energy of formation of Cr2O3 as a function of temperature is represented by curve AB in Fig. 3.4. Values of the free energy of formation of water vapor from the reaction of hydrogen with oxygen are represented by a family of curves diverging from the point H, each curve corresponding to a different molar ratio H2/H2O in the atmosphere. When curve AB crosses a particular curve belonging to the family of water vapor curves representing different water vapor/hydrogen partial pressure ratios, the chromium/oxygen and hydrogen/oxygen reactions are in equilibrium because their respective free energies are the same. This means that the oxygen potentials for the two reactions are identical. When curve AB lies above the water vapor curve at a particular temperature, chromium oxide will be spontaneously reduced to form chromium and water vapor by the hydrogen because the latter combination is more stable than Cr2O3. The reverse is true when curve AB lies below the
water vapor curve. For the equilibrium between chromium/oxygen and water vapor/hydrogen reactions to be achieved at a soldering temperature of 400 C (750 F), the H2O/H2 ratio must be lower than 1010indicated by lines AB and HQ. This condition cannot be achieved in practice, so that hydrogen is ineffective in removing chromium oxide in soldering operations. It is evident from the Ellingham diagram that the stability of metal oxides decreases as the temperature is increased and the oxygen partial pressure is reduced. Commercial supplies of hydrogen, nitrogen, and other gases will inevitably contain some oxygen and water vapor, which are oxidizing agents, and component and furnace surfaces will usually contribute desorbed water to the process atmosphere. In addition, atmospheres with a high concentration of hydrogen present an explosion risk, which usually precludes their use. Although not given in Fig. 3.3, the reaction curve for tin and its common oxide lies in close proximity to that for carbon monoxide/dioxide. Thus, at 400 C, it is theoretically possible to re-
Simplied Ellingham diagram illustrating the graphical method for determining the temperature and H2O/H2 ratio that will spontaneously reduce a metal oxide to metal (here, Cr2O3 to Cr). The set of dashed lines corresponds to the Gibbs free-energy change as a function of temperature for the reaction of hydrogen with oxygen to produce water vapor for different H2O/H2 ratios.
Fig. 3.4
duce tin oxide to tin with a H2O/H2 ratio as low as 102. However, the reaction rate is so slow that, in practice, the gases hydrogen and carbon monoxide are not effective for reducing the oxides of most common industrial metals, including tin, below about 500 C (1380 F) at readily obtainable oxygen partial pressures. The issue of reaction rate is discussed further in section 3.3.3 in this chapter. 3.1.3.3 Alternative Atmospheres for Oxide Reduction
Other gases such as chlorine and uorine are more effective than hydrogen and carbon monoxide at removing surface oxides of particular metals, as is clearly indicated on the relevant Ellingham diagrams [Wicks and Block 1963]. The power of chlorides and uorides as cleaning agents accounts for the effectiveness of chlorouorocarbons (CFCs) and the difficulty experienced in replacing them by more environmentally friendly chemicals, following the implementation of the Montreal Protocol on Substances That Deplete the Ozone Layer: 1991 [McLaughlin et al. 1998; Lea 1991]. Such gases partly operate by converting the oxide to a halide that is volatile at the joining temperature and that vaporizes during the heating cycle. These halide atmospheres also chemically attack the underlying metal and physically undermine the oxide, as occurs in the uxing of aluminum. This point is discussed in section 3.2.2.1 in this chapter. A different option, particularly for tin-base solders, is to use a process atmosphere that contains a chemical ux in its vapor state. Formic acid is one such example. Owing to the absence of a liquid ux, joining under gaseous uxes is generally misdescribed as a uxless process. Further information on chemically active atmospheres is to be found in section 3.3.6 in this chapter.
3% hydrogen, and nitrogen-40% hydrogen. From this trial it can be concluded that there is no particular advantage in using forming gas at small excess temperatures above the solder melting point. In fact, the European specication for commercial grade forming gas allows it to contain a higher level of oxygen and water vapor than is found in bottled nitrogen. In the United States, dry nitrogen and forming gas have roughly similar specications, but forming gas carries a price premium as a specialty gas mixture. Experimental investigators are advised that 40% hydrogen is well above the explosive limit for this gas in air, so its use requires many safety features to comply with health and safety legislation or makes for something of a white knuckle ride for the process operator! However, as explained in section 3.3, if a process using higher than normal soldering temperatures can be adopted, then a dry hydrogen atmosphere may facilitate uxless soldering, particularly for the highermelting-point gold-base solders, on account of the greater thermal activation available and the higher intrinsic nobility of the alloys. It is sometimes claimed that better results in terms of wetting and spreading are obtained when using forming gas compared with pure nitrogen. Frequently, this has more to do with the fact that the quality (i.e., leak tightness) of the furnace and gas control and conveyance system used with forming gas are of a higher standard than for a nitrogen furnace, as required by safety protocols and hydrogen has a thermal conductivity seven times that of nitrogen (see Table 3.3). Thus, for the same process temperature setting, the parts are likely experiencing additional superheat when reowed in forming gas.
3.2
3.1.4
Despite the scientic knowledge embodied in the preceding discussion, it is a common misconception that forming gas (nitrogen/hydrogen mix) is a better choice than nitrogen as a process gas for soldering because it will reduce tin and indium oxide. Figure 3.5 shows the result of melting controlled amounts of four common solders in atmospheres of pure nitrogen, nitrogen-
Successful soldering is largely dependent on the ability of the solder to wet and spread on component surfaces. A major barrier to wetting is presented by stable nonmetallic lms and coatings on the surfaces, in particular oxides and carbonaceous residues. Oxide lms on the faying surfaces present more than a physical barrier to wetting and spreading. Oxides are typically poor thermal conductors, compared to metals, and act as barriers to heat transfer, thereby exacerbating temperature gradients present and delaying fusion of the solder with the parent metal. The thermal contact resistance of electroplated and
then reowed lead-tin eutectic solder on copper is known to increase with the logarithm of the
oxide thickness [Di Giacomo 1986]. Surfaces with a coarse-grained microstructure lose sol-
Spread tests of four common solders, melted on NiCr/Au substrates at 10 C (18 F) superheat, in controlled atmospheres. There is negligible benet from a hydrogen-rich atmosphere. Although somewhat subjective, the solders melted in the 40% hydrogen atmosphere do appear to be cleaner with a brighter metallic luster. The holes in the substrates are of identical size and spacing and act as scale markers. The solder disks measured 6 mm (0.24 in.) diameter by 25 m (1 mil) thick and were only supercially cleaned before use. (a) Ag-80In-15Pb solder melted in nitrogen. (b) Same alloy melted in nitrogen-3% hydrogen. (c) Same alloy melted in nitrogen-40% hydrogen. (d) Ag-96.5Sn solder melted in nitrogen. (e) Same alloy melted in nitrogen-3% hydrogen. (f) Same alloy melted in nitrogen-40% hydrogen. (g) Au-20Sn solder melted in nitrogen. (h) Same alloy melted in nitrogen-3% hydrogen. (i) Same alloy melted in nitrogen-40% hydrogen. (j) Au-12Ge solder melted in nitrogen. (k) Same alloy melted in nitrogen-3% hydrogen. (l) Same alloy melted in nitrogen-40% hydrogen. Source: BAE Systems
Fig. 3.5
derability much more readily than those with ne-grained microstructures because the oxide lms grown on the former are more continuous, with fewer defects. Fluxes are chemical agents that are used to remove these layers and thereby promote wetting by the molten ller.
In order to be effective in exposing a bare metal surfaces, a ux must be capable of fullling the following functions:
(g) Au-20Sn solder melted in nitrogen. (h) Same alloy melted in nitrogen-3% hydrogen. (i) Same alloy melted in nitrogen-40% hydrogen. (j) Au-12Ge solder melted in nitrogen. (k) Same alloy melted in nitrogen-3% hydrogen. (l) Same alloy melted in nitrogen-40% hydrogen. Source: BAE Systems
physical means, often involving reaction of the ux with surface oxides to form metal salts, which are then dissolved by the ux Protection of the cleaned joint from oxidation during the joining cycle Wetting the joint surfaces, but being displaced by the molten solder as the latter spreads While molten, uxes form a thermal blanket around the joint that helps to spread the heat evenly during the heating cycle. The ux also tends to reduce the surface tension between the solder and the joint surfaces, thereby enhancing spreading. Ideally, the ux should leave no residues or produce residues that are easily removed by, for example, being soluble in water. It should also be compatible with the ller and substrate materials. For example, ammonia-containing uxes are not suitable for brass components, because intergranular corrosion can result through chemical reaction. Chemical uxes always function while in a gaseous or liquid form, although the active constituents are frequently solid at room temperature. Fluxes can be introduced to the joint in a number of ways, the most common of which are discussed here. A ux can be applied in the form of a powder, paste, or liquid immediately prior to the heating cycle. The joint is then heated to the required bonding temperature, by which point solid uxes have become molten, ideally just before the ller metal melts. In wave soldering, a liquid ux is applied to a circuit board using a foam or spray dispenser only seconds, or less, before the part enters the solder wave (see Chapter 1, section 1.3.2.2). A ux can also be placed within or adjacent to the joint together with the ller metal as a preform and the assembly heated to the bonding temperature. As a properly chosen ux will melt at a temperature below the melting point of the ller, the molten ux is able to spread over the joint surfaces and clean them before the ller metal melts and displaces the ux.
Table 3.3 Thermal conductivities of soldering atmospheres, relative to air
Soldering atmosphere Relative thermal conductivity
Another method involves introducing the ux together with the ller into a joint already held at the bonding temperature, in the form of uxcored solder wire. Although this technique is widely practiced because it is fast and convenient, it is not recommended because the heated component surfaces are unprotected until the ller is applied. More aggressive uxes are then required, which in turn tends to accentuate corrosion and cleaning problems. Alternatively uxes can be applied together with the ller, prior to the heating cycle, in the form of pastes and creams, which are normally proprietary formulations. They comprise mixtures of the ller metal, which is present as a powder of a prescribed shape and size distribution together with a ux and an organic binder that is selected to produce the desired viscosity and to burn off without leaving contaminating residues. These pastes and creams are particularly useful in automated reow soldering operations because they can be screen printed or dispensed using syringes. The large surface area of the powdered ller metal in contact with the ux means that corrosion is inevitable during storage; therefore, these products have a nite shelf life and storage requirements, which are best strictly observed. The mechanisms of ux action are almost as diverse as are the ux formulations that are commercially available. In many cases, the mechanisms have not been fully elucidated, which is in no short measure due to the commercial secrecy surrounding ux compositions. While solder compositions are mature, ux chemistry is still radically evolving. Todays soldering uxes are very sophisticated chemicals compared with even a few years ago, and new products continue to be launched almost monthly. Several different uxing mechanisms cover the majority of soldering operations that are encountered. Even these are sufficiently complex not to be understood in detail at the present time. However, uxing mechanisms can be classied according to whether they remove the nonmetallic surface coating by physical or chemical means. A ux can chemically remove a surface oxide coating by:
the underlying metal. In this mechanism, the ux does not react with the surface coating itself, but is able to percolate through it and react with the underlying metal, thereby causing detachment of the coating. In addition, physical uxing action can be achieved by applying mechanical agitation, without the need for a material uxing agent, as discussed in section 3.3.5. Many uxes function by a combination of mechanisms, and for this reason uxing action is best illustrated with reference to specic examples. Environmental considerations have impinged heavily on the use of certain uxes, particularly those incorporating organic materials (volatile organic components, or VOCs), because these have tended to rely on CFCs for the removal of their residues, which attack the earths ozone layer when they are discharged to the atmosphere. These older VOC formulations have been largely replaced in manufacturing industry by uxes whose residues are soluble in water or cleaning agents that do not contain substances that are environmentally harmful [Lea 1991; Ellis 1991] (see Chapter 1, section 1.3.2.11). Even when soldering using ux, the employment of an inert atmosphere, such as nitrogen, is benecial. This is particularly true for the AgCu-Sn lead-free solders, which are used at much lower superheats than lead-tin solder. The nitrogen atmosphere helps prevent oxidation of the solder and substrate before the ux becomes active and reduces the work that the ux has to do in maintaining oxide-free interfaces ahead of the advancing solder front. This means that less ux is required, making it environmentally and economically advantageous. If large superheats can be applied, the benets of a protective atmosphere are reduced because the solder spreads much faster, as can be seen from Fig. 3.6 [Buckley 2000]. Further information on lead-free solders is given in Chapter 2, section 2.4.4 and Chapter 5, section 5.1. The effect of different concentrations of oxygen on the results obtained when fabricating uxed printed circuit boards (PCBs) using leadtin solder has been studied. In general, if the oxygen level can be decreased to below 10 ppm then the soldering process will generally be insensitive to many process variables, because the molten solder has a low surface tension and shorter wetting time. Soldering processes can be realized in an inert atmosphere where the oxygen concentration is up to 1% (10,000 ppm). However, to be successful, most process parameters such as ux type, heating rate, peak temperature,
and time at temperature all need to be carefully optimized to achieve low-defect soldering. At elevated oxygen levels, the rate of dross formation on the solder becomes signicant and will interfere with the solder wetting and spreading, as can be seen in Fig. 3.7 [Nowotarski and De Wilde 1996]. Soldering in very low oxygen level atmospheres does not necessarily translate to signicantly improved production costs through maximizing yield. Indeed, just from the standpoint of the cost of the increased gas consumption necessary to reduce the oxygen level below 20 ppm, at least one study showed that it was economically unfavorable (see Table 3.4) [Verite, Verbockhaven, and Alleaume 1997]. Obviously this analysis will vary with the value of the product, in this instance the PCB for a mobile telephone, which is produced in high volumes in a highly competitive market.
Wetting of copper by Pb-63Sn solder using rosin ux. Soldering with ux generally benets from a protective atmosphere (unless the atmosphere detrimentally affects the chemistry of the uxing action), because the ux has to work less to protect the substrate and ller from oxidation.
Fig. 3.6
Effect of oxygen concentration in the atmosphere on the rate of dross (scum) formation on an exposed solder reservoir
Fig. 3.7
3.2.1
The overwhelming majority of soldered joints made are to interconnect electronic components. The principal material that is joined in these applications is copper, due to its high electrical conductivity; the solder is almost invariably a lead-tin (or lead-free tin-base) alloy. Because electronic components tend to be manufactured and stored under reasonably cool, clean, and dry conditions, they are likely to have only a thin layer of copper oxide as a barrier to wetting. This, coupled with the fact that soldering is usually performed rapidly and mostly below 300 C (570 F), means that the uxes required tend not to be highly aggressive chemicals. Major considerations pertaining to uxes intended for electrical/electronic applications, other than their cleaning ability, are the nature of the residues and the ease of their removal. These factors are of concern owing to the need to avoid subsequent deterioration of the joints and ultimately failure of the circuitry through corrosion [Turbini et al. 1991]. Indium and zinc solders require uxes with a chemistry different from tin-base solders, because of the composition of the oxides and the different process temperatures involved. As such, they have an even more specialized formulation compared with uxes for tin-solders and are best procured from specialist suppliers and used in accordance with manufacturers guidelines. 3.2.1.1 Soldering Fluxes That Require Cleaning
Conventional soldering uxes contain at least four basic ingredients, each of which has an identied role [Klein Wassink 1989; Manko 2002]:
100 0 100
37 22 59
30 32.7 62.7
where R is an organic radical. Copper chloride is soluble in water, while copper abiet [Cu(R3COO)2] is miscible with rosin. Hence, the compounds that now contain the copper oxide dissolve in the excess liquid ux to leave a clean metal surface when the ux is displaced by the molten ller metal. Like most active ux constituents, these acids become progressively more active with increas-
ing temperature until a point is reached when their effectiveness ceases, because they either thermally decompose or boil off. The corrosive properties of acids at room temperature can present handling problems and application difculties especially when the ux must either be introduced directly to the components or mixed with solder to form pastes sometime before the joining operation. However, there are salts that liberate acid only when heated, thus avoiding this type of problem. One of these is zinc chloride, which produces hydrochloric acid by reaction with moisture at elevated temperatures, according to the reaction:
ZnCl2 H2O Zn(OH)Cl HCl
tant needs to be inert toward the other constituents of the ux and also to the clean metal surface. For this reason, ion-free organic complexes, similar to those widely used in the electroplating industry, tend to be favored. The rheological agent is provided to impart the correct degree of body to the ux to suit the application method, whether syringe dispensing, screen printing, or stenciling. A common commercial designation of uxes used for soldering is:
By using a mixture of similar salts, the activation temperature of the ux and its corrosiveness can be adjusted over a wide range. In all cases, the residues are highly corrosive. Other halide uxes operate in a very similar manner to zinc chloride. For example, the amine hydrohalides, such as hydrazine hydrochloride (R2NH2Cl), thermally decompose with the liberation of hydrochloric acid:
R2NH2Cl R2NH HCl
R: rosin RMA: rosin mildly activated RA: rosin activated OA: organic acid IA: inorganic acid WS: water soluble SA: synthetically activated
The ux constituent that protects the clean metal surface from reoxidation usually also serves as the carrier for the other ingredients. It need only be effective for a few seconds in many soldering processes. Alcohols, oils, esters, glycol, and even water are capable of fullling this function at the relatively low temperatures used for most soldering operations (<250 C, or 480 F). As the ux carrier constitutes by far the largest volume fraction of the formulation, it has a strong inuence on the aggregate properties even though it is not intended to be chemically active. The ux carrier is usually a mixture of chemicals formulated partly to prevent the ux from boiling violently when the workpiece reaches a specic temperature. As mentioned previously, it may also play a role in mopping up the products of reaction. Surfactants are added to lower the surface tension of the liquid ux. The effect of minute additions of detergents, soaps, and soluble oils to water is well known, and these are often added to water-based uxes to ensure satisfactory wetting when either oil or grease are likely to be present on the component surfaces. The surfac-
This classication is somewhat loose and can be more misleading than helpful. It does not strictly indicate the chemical characteristics of the ux, particularly its aggressiveness to specic oxides. However, because this classication is widely used, it warrants consideration here. An alternative classication of soldering ux types, devised by the International Organization for Standardization, is given in Table 3.5. Although this classication is scientically more precise, it is less descriptive; therefore, the traditional terms tend to be preferred. The different categories can be grouped as shown below. R, RMA, RA. Fluxes with these designations contain rosin as the principal active chemical ingredient, the difference between them being a progressive increase in level of chemical activity from R through to RA. In modern uxes, the rosin is a synthesized chemical and no longer the natural product obtained from the sap of pine trees. OA. OA stands for organic acid, which provides the cleaning action in this type of ux. The organic acid used will tend to be one of those described previously, such as a carboxylic acid. These uxes are generally more aggressive than RA uxes. IA. IA uxes are based on inorganic acids, usually hydrochloric acid and are among the most aggressive of the available uxes. SA. Synthetically activated uxes are formulated to have residues that are soluble in chlorouorocarbon solvents. With concern growing about atmospheric pollution, and CFCs in particular, the SA uxes have been superceded by
newer formulations that have water-soluble (WS) residues. WS. Water-soluble uxes containing soluble halides have been available for several decades. These are among the most commercially developed uxes available with chemical activity tailored over a wide range, roughly in proportion to the total halide content. Particular care needs to be taken to clean off ionic residues, which can cause corrosion and generate electrical malfunction. Water-soluble uxes tend to have shorter printing life than rosin uxes, have less tackiness and are therefore more difficult to use for reow soldering. 3.2.1.2 No-Clean Soldering Fluxes
In response to manufacturers concern over the cost of cleaning, a new generation of noclean uxes has come into use that are said to leave no deleterious residues (in PCB assembly applications). These uxes have been formulated with a wide range of chemical ingredients, but all share the characteristics of leaving residues that are judged to be benign and so can be left on the assemblies. No-clean uxes mostly contain an alcohol solvent carrier and a small percentage of active ingredients, which may comprise resins or simply organic acids. Some of the newer no-clean uxes are water-based and are totally free of VOCs, including alcohols. Fortuitously, using water as the solvent in place of a VOC, such as an alcohol, has been found to enhance no-clean uxing. In particular, water-based no-clean uxes appear to cope better with circuit boards that enter the soldering operation with surface oxide. Their enhanced activity with respect to metal oxides has
been ascribed to a more complete ionization of the activators in a water base than in alcohol [Hyland and Rao 1996]. The need to clean is obviated by diluting the activators to low levels. Therefore, these uxes contain a much lower content of rosin, or halidesreferred to simply as solidsthan do traditional uxes. The solids content of no-clean uxes is normally in the range 2 to 3 wt%, as compared with 25 to 35 wt% for normal rosinbased uxes. These low-solid uxes are more difficult to work with than the rosin uxes precisely because they are less active. An alternative approach used for formulating no-clean uxes is to use activators of normal strength or thereabouts in conjunction with other constituents that undergo polymerization when subjected to soldering temperatures. Gelatin is one such compound. The resulting polymer fraction encapsulates residual active species, thereby incapacitating it. However, this mechanism can be impaired if not all the ux is fully heated, leaving areas of the joint exposed to active residues. It is also important that no-clean ux residues are not cleaned because, unless the cleaning is scrupulous it leaves behind destabilized encapsulant from which corrosive agents can subsequently escape. From the foregoing outline, it is clear that no-clean uxes are less forgiving, being more intolerant to variability in the soldering process conditions than conventional alternatives. Sufcient ux must be provided to achieve the required solderability and, as noted previously, the ux must be heated uniformly and in the correct manner to react fully and leave innocuous byproducts. These considerations are consistent with the experience of industrial users, who nd
Table 3.5 Classication of soldering uxes using the method adopted by the International Organization for Standardization
Flux type Flux basis Flux activation Flux form
1 Resin
1 Rosin 2 Resin 1 Water soluble 2 Not water soluble 1 Salts 2 Acids 3 Alkalis
A Liquid
2 Organic
3 Not halogen activated B Solid 1 With NH4Cl 2 Without NH4C1 1 Phosphoric acid 2 Other acids 1 Ammonia and/or amines C Paste
3 Inorganic
that no-clean uxes work best in assembly-line soldering operations when the process conditions are optimally tuned and strictly maintained from one assembly operation to another. While this situation is readily achieved in a factory using modern equipment, it does mean that noclean uxes are not generally suitable for handsoldering processes. When correctly instigated, the residues left on a PCB are not detrimental in most applications and accelerated life-test environments [Anson et al. 1996].
3.2.1.3
Because ux residues, like other contaminants, can easily affect the performance and service life of electronic circuits, effective cleaning procedures assume a critical importance. A semiquantitative test has been devised to assess the effectiveness of a cleaning process for removing ux residues from circuit boards and to provide assurance that boards soldered using no-clean ux will not fail in service due to unbound residues. This test measures the surface insulation resistance
(SIR) to current ow between pairs of conducting tracks in interdigitized comb patterns on boards as a function of time at a prescribed temperature and level of humidity. It is capable of catching electrochemical failure mechanisms associated with vestigial ionic residues, namely unacceptable current leakage under humid conditions, corrosion, metal migration, and dendritic growth. An electric eld is applied between the two intermeshed sets of combs, as shown in Fig. 3.8. The widely accepted SIR test standard (ISO 9455, part 17) calls for a test pattern with 400 m (16 mil) wide tracks, separated by 500 m (20 mil) wide gaps, and a test eld of 100V/mm (2.5V/mil) (i.e., a test voltage of 50 V). The test environment is a constant temperature of 85 C (185 F) and a relative humidity of 85%. Surface insulation resistance measurements are made twice a day over a test duration of 168 h. If the surface resistance is maintained above 108 throughout the test, the surface is judged to be clean. The presence of ionic contamination on the surface of the test board can result in the formation of conducting dendrites, as shown in Fig. 3.9. These dendrites can grow rapidly (in minutes), starting from residues deposited on a cath-
Fig. 3.8
International Electrotechnical Commission (IEC) test coupon for the evaluation of board cleanliness using SIR. Schematic of typical detail of an interdigitated comb is illustrated. Courtesy of Concoat Systems
ode and progressing toward a neighboring anode. When a dendrite has bridged the gap, the SIR value will rapidly decrease. As a dendrite grows, it will progressively carry a larger proportion of the available current until at some point it burns out and the SIR will return to a high value. This sequence of events repeats until carbon deposits build up sufficiently to permanently lower the resistance below the acceptance threshold. Examples of SIR test plots (log SIR versus time) are shown in Fig. 3.10. A recent assessment of the SIR test has highlighted the fact that the test procedure, as currently followed, can miss detection of failure events [Hunt 2000]. It is possible that resistance drops, due to dendrite formation, might escape observation if the electrical measurements are widely spaced in time: half a day is clearly too long. Moreover, the bridging of a dendrite across neighboring conductors, which lasts for even only a few seconds before burnout, can represent a large number of operations in equipment operating at a clock frequency of MHz and above. For such equipment, intermittent dendrite bridging would cause a corresponding incidence of failures, which the test might not pick up. This limitation can be overcome by using modern high-speed multichannel measuring equipment in the SIR test. Other recent recommendations to improve the test include using a lower voltage stress (5 V/mm) and ner test structures (100 m/100 m, or 4 mil/4 mil track/gap), which would be more appropriate of current circuitry and operating conditions. It is also proposed to change the test environment from 85 C (185 F)/85% RH to a more demanding 40 C (104 F)/93% RH.
As a result of such reviews of the SIR test procedure, the following compromise set of conditions is being proposed for a new Standard specication (draft IEC 61189-5):
Example of conducting dendrites growing across the gaps of a test pattern on a circuit board during a SIR test. Courtesy of Concoat Systems
Fig. 3.9
Example of an SIR test plot (log SIR vs. time), showing features usually associated with dendritic growth. Courtesy of Concoat Systems
Fig. 3.10
able uxes exist for both of these unsolderable metals. 3.2.2.1 Aluminum Soldering Fluxes
Aluminum forms a natural refractory oxide that is remarkably stable and tenacious. It is mechanically durable, with a hardness that is only inferior to that of diamond and its high melting point (2050 C, or 3722 F) reects its high degree of physical stability. Alumina is also chemically stable to the extent that it cannot be directly reduced to the metal by aqueous reagents. On exposure to air, a layer of alumina will form almost instantaneously on the surface of aluminum, and this will grow to an equilibrium thickness of between 2 and 5 nm (0.08 and 0.2 in.) at ambient temperature. On heating to 500 to 600 C (930 to 1110 F), the thickness of this surface coating will increase to about 1 m (40 in.). Therefore, special uxes have been formulated for use with aluminum alloys. These have to be particularly effective in protecting the metal surface from oxidation before the solder melts and spreads. Aluminum uxes all contain halide compounds. These are highly corrosive especially in the presence of moisture, including humid atmospheres. Therefore, all ux residues must be removed as completely as possible, as residues left behind after the cleaning procedures may cause corrosion in the vicinity of the joint. The uxes used for soldering of aluminum and its alloys are of two types [The Aluminium As-
sociation 1990]: organic uxes and chloridebased uxes. Organic uxes contain amines, uoborates, and a heavy metal compound in an organic carrier. They come in the form of viscous liquids or powders. A typical example of this type of ux has the composition: 83% triethanolamine, 10% uoboric acid, and 7% cadmium uoroborate (a viscous liquid). Its operating range is 180 to 280 C (355 to 535 F). The uxing action relies on disrupting the oxide, which cracks and crazes during the heating operation due to the differential thermal expansion between the metal and oxide. This enables the ux to come into direct contact with the aluminum and deposit a lm of the metal ion in the ux (in this instance, cadmium) onto the aluminum surface via an exchange reaction. Organic uxes must not be exposed to a torch or ame; otherwise they will char, and this will impede solder ow. Aluminum alloys containing more than about 1% Mg cannot be satisfactorily soldered using these uxes, because magnesia is more refractory than alumina and the ux is correspondingly less effective. Chloride-based uxes contain zinc or tin chlorides with ammonium chloride and uoride and are generally applied as a water-based slurry or paste to precleaned component surfaces. An example of such a ux has the formulation: 88% tin chloride, 10% ammonium chloride, 2% sodium uoride (powder) with a working range of 300 to 400 C (570 to 750 F). By substituting the tin chloride with zinc chloride, the tempera-
Table 3.6
Parent material
Aluminum Aluminum alloys Beryllium Brass Chromium Copper Copper-nickel Gold Invar/Kovar Lead Magnesium Nichrome Nickel Palladium Platinum Silver Stainless steel Steel Tin Titanium Zinc
... ... ... X ... X ... X ... X ... ... ... X X X ... ... X ... ...
... ... X ... ... ... X ... ... ... ... ... X ... ... ... ... X ... ... X
X ... ... ... ... ... ... ... X ... ... X ... ... ... ... X ... ... ... ...
... X ... ... X ... ... ... ... ... X ... ... ... ... ... ... ... ... X ...
ture of operation can be raised to 380 to 450 C (715 to 840 F). The uxing mechanism is essentially the same as that for the organic uxesnamely, one involving an exchange reaction whereby aluminum on the surface is replaced by zinc or tin. The effectiveness of these uxes is reduced by the presence of silicon in the parent material, because silicon is not as amenable to the exchange reaction as is aluminum. Because both of these two uxes operate by substituting for aluminum a metal that has reasonable oxidation resistance and that is more readily wetted by the ller metal, they are fundamentally different from conventional soldering uxes. The latter act simply by cleaning and protecting the original surfaces of the components. Because the replacement metal has a higher density than aluminum, this type of uxing process is commonly referred to as heavy metal deposition. The quantity of ux that needs to be applied is a function of the humidity of the ambient atmosphere. In moist atmospheres, a proportion of the ux is rendered ineffective through hydrolization by reaction with water vapor. It has been found that the quantities of ux that need to be applied can be reduced considerably by carrying out the soldering operation in a completely dry environment. 3.2.2.2 Stainless Steel Soldering Fluxes
ux [Wakelin 1993]. The solder possesses sufcient ductility to survive bending several times by hand and can be cold worked by rolling. Details of the solder composition have not been disclosed, and it has not yet been developed as a commercial product. 3.2.2.3 Magnesium Soldering Flux
Magnesium is also deemed to be unsolderable by conventional means. However, an interesting development is the discovery that molten acetamid will directly dissolve the surface oxides on both aluminum and magnesium [Golubtchik 1984]. Because this organic compound melts at 83 C (181 F), it offers a route whereby indium-tin solder (melting point 120 C, or 248 F) could be used to wet the exposed metal on the substrate. An experimental trial conducted by one of the authors under very rudimentary conditions conrmed that the approach does have some potential and possibly merits further investigation, particularly as a method for tinning magnesium prior to more conventional soldering.
3.2.3
High-Temperature Fluxes
Stainless steel is not an easy material to solder for two reasons. Firstly, it is covered with an extremely stable (self-repairing) oxide layer, which indeed gives it its stainless characteristic. Secondly, for a metal it has an unusually low thermal conductivity, one-thirtieth of that of copper, which causes complications when endeavoring to heat joints quickly and to a uniform temperature. Aggressive halogen-based uxes can attack the oxide layer on stainless steel but, unless the residues are fully removed, they are prone to cause pitting corrosion. A solution of zinc chloride in hydrochloric acid constitutes the chemically active ingredients of uxes for stainless steel [Mei and Morris 1992]. Phosphoric acid is also extremely effective as a ux on stainless steel, but unfortunately it polymerizes at temperatures above 200 C (392 F), which precludes its use with the common tinbase solders. Multicomponent, lead-free, solders based on bismuth-tin have been devised for soldering to stainless steel using phosphoric acid
Fluxes for soldering at high temperatures (in the range 250 to 450 C, or 480 to 840 F) are formulated differently to uxes designed to work with lead-tin eutectic solder and alloys of similar melting point. There are three reasons for this. First, the higher process temperature means that there is more thermal activation available so that the active constituents need not be as chemically aggressive. Second, these uxes are usually used with high-lead solders or gold-tin eutectic solder. Thus, the oxide that impedes wetting and spreading on the surface of the solder is either lead oxide or tin oxide diluted by the noble metal so, again, decreased chemical activity is sufficient. The nal consideration is that the ux needs to provide signicantly more protection to the ller metal and component surfaces against oxidation, because higher process temperatures usually mean that the duration of the process cycle is longer. For these reasons, high-temperature uxes tend only to be mildly active and are based predominately on high-molecular-weight hydrocarbons. As might be expected, the long-chain molecules are very viscous at room temperature, and these uxes are very difficult to dispense, especially on a cold morning. To overcome this prob-
lem, the ux is usually thinned with a lowermolecular-weight species. On heating, the lighter fraction of the mixture progressively volatilizes as the viscosity of the high-molecular-weight hydrocarbon declines, so ensuring that the uidity of the ux remains essentially constant throughout the working temperature range. A point to be aware of when using these uxes, especially in conned spaces such as ip-chip assembly operations, is that if the heating rate is too fast, then on reaching about 150 C (300 F), the ux can boil and physically displace components. This is simply rapid volatilization of the lighter hydrocarbon fraction. Incorporating a dwell in the heating cycle or a slow ramp through the critical temperature range are appropriate remedies. In air, high-temperature uxes produce a plume of smoke when heated to the soldering temperature. This fume is ammable, so that soldering cannot be undertaken with a naked ame. In enclosed spaces the fume tends to recondense on cooler surfaces, which requires that cleaning is undertaken on a planned basis. Highertemperature uxes are available in all of the modern designations including water soluble and no-clean.
3.3
Fluxless Soldering
The vast majority of commercial soldering operations involves the use of chemical uxes. The use of ux makes the soldering operation tolerant to an ambient air environment and maintains the solderability of surfaces through the heating cycle, as explained in section 3.2. However, there is a penalty to be paid when uxes are used in the residues that they leave behind. Flux residues are never completely removed in normal cleaning procedures, and the contamination can impair the product function, performance, and life. High-performance electronics systems, particularly those enclosed in hermetic packages, usually proscribe the use of all organic materials in their assembly, including uxes, as a means of guaranteeing service life. A particularly critical application is hybrid assembly for space electronics, where even small traces of organic substances present an unacceptable risk, owing to the tendency of hydrocarbons to outgas. Because aerospace electronics are often contained within hermetic packages, the presence of moisture in the package environment can activate a litany of semiconductor failure mechanisms and is therefore undesirable.
The problem is no less acute in relation to optoelectronic and photonic modules. A typical example is a laser diode soldered to a heat sink (for example, a thermoelectric cooler) in order to stabilize the junction temperature. The joining operation must leave the adjacent light-emitting facet in a pristine state. Contamination of this surface with virtually any species will either degrade the service life of the device or even have catastrophic consequences for the operational characteristics. Similar considerations pertain to the assembly of tiny microelectromechanical systems (MEMS) [Kuhmann et al. 1998]. For these and other comparable applications, there is the need to use uxless soldering processes, which will also furnish thin and welllled joints to satisfy the thermal and mechanical requirements of the assembly. Fluxless soldering offers other attractive features: it eliminates a processing material and more signicantly in terms of cost saving, processing steps including cleaning after the soldering operation. Fluxless soldering also obviates a major source of voids in joints, namely, that arising from trapped ux products. The crux of a successful uxless soldering process is to eliminate all surface contamination from the faying surfaces and to provide sufficient protection of exposed surfaces from oxidation through the heating cycle. At the same time, deleterious contaminants must be excluded from the joining environment. A ux is designed to achieve these conditions, and if it is dispensed with then other means have to be found to satisfy them. As might be expected, there is direct correlation between oxide thickness and solderability. Similarly, there is a clear relationship between the solderability of electronic component terminations and defect levels in volume manufacturing of electronic circuit boards. Thus, it is possible to establish a relationship between defect levels and oxide thickness on a given surface nish. Figure 3.11 presents results from board trials that were conducted in air using a nonactivated ux, involving copper lands. Clearly, the rst 5 nm (0.2 in) of oxide layer thickness has a very signicant impact on solderability and the number of defective joints that result. While it is fairly straightforward to achieve a nonoxidizing atmosphere of sufficient quality to largely maintain the solderability of joint surfaces during a rapid heating cycle, initial removal of surface contamination, and particularly the native oxides, is more complex. Tackling
this problem rst requires knowledge of the nature and thickness of the oxides involved.
3.3.1
During uxless soldering the prime impediment to wetting and spreading is the presence of oxide lms on the surfaces of the parent and ller metals. All base metals are covered with a thin lm of oxide through contact with air. In this respect, the noble metals, gold and platinum, are exceptional. Surface coatings of these metals are exploited for their perpetual solderability, stemming from their inertness to oxygen. The thickness of the native oxide on a base metal surface depends very much on the history of the particular component. The principal factors governing the thickness of the oxide layer are time of exposure to air and temperature. The oxidation of some common metals and metallizations at room temperature, after mechanical cleaning of the surface, are indicated in Fig. 3.12.
Fig. 3.11
Effect of oxide thickness on copper lands on the defect level of joints incurred during PCB assembly
This graph shows that the surface of these base metals will be covered with more than 5 nm (0.2 in.) of oxide within 5 min of the cleaning operation, although the oxide layer could be substantially thicker if the same metal were exposed to certain chemicals or heated in air. Obviously, this surface contamination must be removed or displaced by some means before wetting by the solder can proceed. The curves represented in Fig. 3.12 should be taken as being indicative only, because in practice the growth rate is affected by a multitude of factors, including surface roughness, residual stresses, and humidity. For elevated temperature, an approximate rule of thumb is that the native oxide thickness on base metals eventually doubles with every 200 C (360 F) increase in temperature. For this reason, it is desirable to heat the joint rapidly and as soon as possible after cleaning of the surfaces, even when carrying out a soldering operation with the benet of ux and especially so in its absence. A detailed theoretical treatment of native oxide growth on base metals is given by Martin and Fromm [1977]. Because oxide lms grow so rapidly on most base metals, as shown in Fig. 3.12 and 3.13, in practice clean and solderable component surfaces can only be achieved if these are of gold or platinum. Accordingly, for uxless soldering, a gold coating is generally applied to the component surfaces, which is of sufficient thickness and adequately pore-free to ensure good solderability over its specied storage life. Further information on this subject is given in Chapter 4, section 4.1.2.1. An oxide lm will likewise grow rapidly on any solder preform exposed to air. To a rst approximation, metallic oxide growth has a parabolic relationship with time, following a standard diffusion-type equation:
X2 Dt Dot e(Q/RT)
Fig. 3.12
where X is the layer thickness, D is the diffusivity, t is the time in seconds, Do is the diffusion coefficient (m2/s), T is the temperature in Kelvin, Q is the activation energy (J/mol), and R is the universal gas constant (8.314 J/mol K). For pure tin, Q 33 kJ/mol and Do3.7 1018 m2/s, while for eutectic lead-tin solder the respective values are 40 kJ/mol and 2.5 1017 m2/s. Thus, the native oxide on solid solder pre-
forms will be typically a few nm thick shortly after cleaning, but will rapidly increase in thickness, as can be seen in Fig. 3.14, and especially as the heating cycle progresses, unless the atmosphere is either inert or actively prevents reoxidation. The dominant species of oxide that forms on solder is usually consistent with that predicted based on classical thermodynamics, namely the stable oxide of the constituent element that has the lowest free energy of oxide formation. This metallic constituent will tend to oxidize preferentially at the surface of the alloy preform. Thus, even on the high lead 95Pb-5Sn solder, the oxide skin contains somewhere in the region of 20 times more tin than lead. The free energies of oxide formation of typical elements used in solders are listed inTable 3.1. Metal hydroxides are thermally unstable with respect to the oxides, so native surface lms are usually not hydrated to any appreciable extent. From Table 3.1 it can be seen that only the stability of indium and zinc oxides ex-
ceeds that of tin. Hence the majority of solder alloys are predominantly covered with a layer of tin oxide and only indium- and zinc-containing alloys have different surface oxides. This is also the reason why indium- and zinc-base solders each have their own unique ux formulations. The tin oxide growth on molten lead-tin solder follows a square-root time dependence, whereas the combination of oxides that form on In-48Sn solder follows a parabolic growth law, as shown in Fig. 3.15. Indium oxide initially thickens much more rapidly than tin oxide, particularly when the solder is molten and accounts for the relative difficulty of obtaining satisfactory wetting and spreading using indium solders without ux. Indium-tin solders oxidize to form a mixed oxide of In2O3 and SnO and so are somewhat easier to use than other indium solders, but still less so than tin-base solders. Likewise, on gold-tin alloys, the tin oxide is diluted through the incorporation of gold. As gold does not oxidize, except in very unusual environments, for gold-tin solder the rate of oxide growth is slightly slower than on pure tin.
3.3.2
Oxide growth on three base metals as a function of temperature. The time at temperature is of the order of a few minutes and has some correspondence with typical soldering cycle times once component mass and heating rates are taken into consideration
Fig. 3.13
Published work indicates that if a molten metal or alloy is heated above some critical temperature then dissolution of thin lms of surface oxide into the bulk of the melt can occur, the extent depending on the solubility of the oxide. If the heating cycle is conducted in a controlled atmosphere that prevents reoxidation of the free surface, then a uxless soldering process may be possible under these conditions.
Fig. 3.14
Oxide thickness versus oxidation time for a range of solders held in air 140 C (284 F) above their melting point. Adapted from Dong, Schwarz, and Roth [1977]
The superheats (temperatures above the melting point) required to dissolve the surface oxides on seven common solder alloys, in nitrogen, have been determined experimentally and are given in Table 3.7 [Dong, Schwarz, and Roth 1977]. Interestingly, the high-tin-containing alloys perform well in this regard, compared with lead-tin eutectic solder, and this phenomenon might partly explain the success being achieved in running leadfree soldering process at very low superheats, using Ag-Cu-Sn alloys with only mildly active uxes. Solder-spreading tests suggest that the spreading of high-tin-containing solders has far higher tolerance to the oxygen level in the process atmosphere (nitrogen) than other solders, as shown in Table 3.8 [Dong, Schwarz, and Roth 1977], for the same reason. The combined effect of the presence of surface oxide on both the substrate and solder can be seen in Fig. 3.16, which shows the contact angle for In48Sn solder on various substrates, after heating for 5 min at 250 C (480 C), as a function of oxygen partial pressure. Because the quality of the atmosphere used is nowhere near adequate to affect oxide reduction on the solder, the conclusion is that spreading is being achieved through a combination of oxide dissolution and other physical mechanisms that locally rupture the oxide lms and facilitate sites of direct metal-to-metal contact from which spreading can proceed. The ability of solders to dissolve their own oxides has only a limited benet for most uxless soldering applications. While wetting and spreading characteristics may supercially appear quite satisfactory, the degradation of the solder resulting from oxide dissolution is manifested in increased viscosity and impaired mechanical properties. As a process option it is therefore only really applicable to circumstances
where the joint contains a large volume of solder, a high ratio of the volume-to-surface area of the ller metal, and large superheats can be used.
3.3.3
For the reasons that are explained in section 3.1.3.2, whether or not an oxide can theoretically be reduced by hydrogen depends on the stability of the oxide at the process temperature and the hydrogen/water partial pressure ratio in the soldering atmosphere. Once these fundamental conditions are satised, one then needs to consider the oxide reduction rate. Even when an oxide is thermodynamically unstable, the rate of transformation of the surface oxides might be so slow as to render the process practically ineffective. In a normal soldering process, one might expect the initial layers of oxide on the surface of the ller metal to be of the order of, for example, 2 nm thick (0.8 in.) and the process cycle time available no more than 1 min. In this case, for reduction by hydrogen to be effective, the dynamics of oxide removal has to exceed 2 nm/min (0.8 in./min).This means that either the quality of the atmosphere must exceed the thermodynamic minimum value and/or the temperature must be increased above the base point. From measurement of the reduction rate of solder oxides as a function of temperature in hydrogen gas containing about 10 ppm combined oxygen and water vapor, it is possible to dene a threshold temperature above which the reduction rate exceeds 2 nm/min (0.8 in./min). As can be seen from Fig. 3.17, for solders covered with tin oxide the threshold temperature is approximately 430 C (800 F) and for solders covered with indium oxide it is slightly higher at 470 C (880 F). The initiation temperature for the reduction of zinc oxides exceeds 500 C (930 F), at which point the volatility of zinc will start to cause other problems. Hence, hydrogen gas is
Oxide growth on molten solders at an oxygen partial pressure of 1 Pa (1.5 104 psi). Adapted from Kuhmann et al. [1998]
Fig. 3.15
10 18 19 77 93 >300
Table 3.8 Tolerance of molten solders to the oxygen level in the atmosphere, as indicated by the excess temperature above the melting point necessary to achieve solder spreading
Superheat, C above melting point Solder O2 level 10 ppm O2 level 100 ppm O2 level 1,000 ppm O2 level 10,000 ppm
18 18 19 87 ... ...
No spread <300 C (<570 No spread <300 C (<570 No spread <300 C (<570 No spread <300 C (<570 ... ...
F) F) F) F)
not an effective substitute for ux unless the soldering process can be undertaken at very high excess temperatures. In an investigation of the oxidation and reduction kinetics of selected lead-tin, gold-tin, and indium-tin eutectic solders, it was possible to use hydrogen to achieve self-alignment of solder bumps during ip-chip interconnection [Kuhmann et al. 1998] (further information on ip-chip interconnection can be found in Chapter 5, section 5.2). However, in order to reach this satisfactory result it was necessary to achieve hydrogen/water ratios in excess of the limiting values derived from the Ellingham diagram. The investigators satised the requisite conditions by pumping down the soldering oven, which was an ultrahigh vacuum (UHV) system, to a pressure of 109 Pa (1.5 1013 psi). This pressure will have dened the residual partial pressure of water vapor present. Then, the introduction of sufficiently pure hydrogen at a pressure 1 Pa (1.5 104 psi) was able to satisfy the thermodynamic requirements for reduction of oxides, including that of indium, at a sufficiently fast rate that the oxides were actually removed within a few minutes at normal process temperatures. While the ability to achieve the stringent conditions required for the reduction of solder oxides by hydrogen has been demonstrated in a laboratory context, it is doubtful whether this would be realistic or economic in a manufacturing environment.
3.3.4
Although molecular hydrogen is not practicably capable of reducing solder oxides at normal
soldering temperatures, the same does not apply to atomic hydrogen. Atomic hydrogen is far more reactive than the molecular species. In essence, the energy provided externally to split the diatomic hydrogen molecule into single atoms then becomes available for accomplishing chemical work. Atomic hydrogen reacts with surface oxides on solders and joint surfaces to form hydroxides, hydrogenated complexes, or water, all of which are volatile at typical soldering temperatures and can therefore easily be removed from the area of the joint. Molecular hydrogen can be dissociated into atomic hydrogen using a variety of methods including a heated lament, photo dissociation, electrical discharge, microwave radiation, and, probably the most practical, low-frequency alternating current ionization. Commercial atomic hydrogen soldering systems are available from at least one vendor [ATV 2003]. The concentration of any monatomic hydrogen produced decays rapidly with distance from the source so that it must be generated near the workpiece. However, because atomic hydrogen is such a reactive species toward solder oxide, a commercial argon/hydrogen gas mixture can be used to feed the plasma generator and at a hydrogen concentration that is sufficiently low to obviate the need for hydrogen-monitoring equipment and the associated safety features. Because the species in a plasma are not in thermodynamic equilibrium, application of chemical thermodynamics to the analysis of such a system requires modications. A quantitative process description has been formulated that has allowed a line corresponding to the following chemical reaction to be added to the Ellingham diagram [Jacob, Chandran, and Mallya 2000]:
4H O2 2H2O
Under standard conditions, monatomic hydrogen can reduce almost all metal oxides at temperatures below 500 C (930 F). Currently, it is not possible to produce monatomic hydrogen at atmospheric pressure (Po 105 Pa, or 15 psi), and the dashed line included in Fig. 3.18 represents the reduction potential of atomic hydrogen at a base pressure of 1 kPa (0.15 psi). To exploit fully the reduction potential of monatomic hydrogen, it is necessary to remove the products of reaction and prevent back-reaction resulting from the formation of molecular hydrogen and oxide. This is achieved using a ow system where the hydrogen gas is passed through a dissociator and then over the workpiece before being transported away.
Using monatomic hydrogen, ller alloy and component surfaces can be cleaned in situ and excellent quality uxless joints obtained at normal soldering temperatures. The process cycle time can be less than 1 min, the time being limited more by the mechanics associated with passing the part through a heated zone containing the monatomic hydrogen rather than by the kinetics of the reduction process itself. Figure 3.19 is an example of application of hydrogen plasma to lead-tin solder using a commercial solder reow oven equipped with a low-frequency plasma generator. Similar ndings are reported in the literature [Hong and Kang 2001]. Reduction of metal oxide by hydrogen plasma has been followed by direct measurement of the oxygen/silver ratio as a function of process time (see Fig. 3.20). The substrate was 40 nm (1.6 in.) thick silver, the base pressure 1 Pa (1.5 104 psi), and the plasma intensity 1 eV. The kinetic energy of the hydrogen ions was below the threshold for sputtering processes, verifying the chemical nature of the reductive mechanism [Bielmann et al. 2002].
3.3.5
Contact angle for In-48Sn solder on four metal substrates, after heating for 5 min at 250 C (482 F), as a function of oxygen partial pressure. Adapted from Preuss, Adolphi, and Werner [1994]
Fig. 3.16
The native oxides on the surfaces of the solder and components can, of course, be removed physically. Although an obvious form of mechanical abrasion is to use abrasive paper or a scalpel, more rened and controlled methods exist. For example, fast atom bombardment using a neutral beam can mill surfaces at rates of tens
Fig. 3.17
Reduction rate of solder oxides in hydrogen, containing 10 ppm total oxygen and water vapor, for a range of solders as a function of process temperature. Adapted from Dong, Schwarz, and Roth [1977]
of nm per minute. The effectiveness of the process is demonstrated in Fig. 3.21, which refers to ip-chip joints made with high-lead Pb-3Sn solder. In this graph, joint strength is held to be an indicator of surface oxide thickness. The simi-
Simplied Ellingham diagram for oxides with a line (dashed) on it representing the reduction potential of monotomic hydrogen at a partial pressure of 1 kPa (PO is dened as this pressure as it is not possible to generate monatomic hydrogen at atmospheric pressure by any method currently available.)
Fig. 3.18
Balls of lead-tin solder, approximately 100 m diameter, reowed using an atomic hydrogen plasma. The solidied surface is exceptionally smooth and shiny and, at high magnication the duplex phase microstructure solder can be readily seen, attesting to the surface cleanliness. Source: Agilent Technologies
Fig. 3.19
larity of form between Fig. 3.21 and Fig. 1.27 (showing the effect of mechanical wiping) is noteworthy as they are essentially the same physical process. It is reported that an addition of 10% hydrogen into the sputtering plasma considerably boosts the oxide-removal rate. It is likely that the hydrogen ions contribute to the oxide removal through a chemical mechanism, as discussed in the preceding section [Hong, Kang, and Jung 2002]. The main drawback of all mechanical methods of cleaning surfaces, which is illustrated in Fig. 3.22, is that the oxide lms promptly regrow if the parts are exposed to air or, indeed, left in a nitrogen atmosphere of even the highest quality for more than a few minutes. Therefore, a uxless soldering process, not involving an atmosphere containing atomic hydrogen, must be able to cope with a thin layer of native oxide on, at least, the solder. Strategies for making a robust uxless soldering process are given in the following sections. An automated version of physical abrasion to remove surface contamination is ultrasonic agitation, which is more correctly called ultrasonic tinning. This technique is highly effective in removing even the most tenacious oxides and is therefore applicable to virtually all metals, but is most widely used with aluminum [Jones and Thomas 1956]. It is not a new technology, and there exists a patent reference to an ultrasonic soldering iron that predates World War II [Antonevich 1976]. Ultrasonic uxing normally involves applying an ultrasonically activated soldering iron to tin the workpiece [Schaffer et al. 1962]. In a variant arrangement, a solder bath is ultrasonically excited and the workpiece is dip coated with the solder. These ultrasonic systems typically operate at 20 to 80 kHz and 0 to 300 W of electrical power. Most systems operate at a single frequency, but some provide for application of high- and low-frequency acoustic energy simultaneously. A more recent development is to incorporate an ultrasonic transducer into wavesoldering equipment. The ultrasonic sonodes are an integral part of the solder wave and impart sonic action while the board travels through the wave [Swanson 1995]. In operation the sonodes remove the oxide from the component leads and PCB lands, thereby enhancing wetting of the solder. The oxide removal is accomplished by the cavitation action of the ultrasonic waves on the oxidized surfaces. The system operates with a frequency in the 20 kHz range and an amplitude in the micron (sub-mil) range. These pa-
rameters are selected in order not to adversely affect the integrity of the components and circuit board material. Ultrasonic uxing is a uxing and tinning operation for applying a metal coating to a clean surface of the workpiece. Once the metal coating has been applied, it protects the workpiece surface against reoxidation, and for that reason is akin to chemical uxing, as described previously. The coated part can subsequently be soldered (or brazed) using a conventional process. At present, the method is limited to tinning of passive electronic and optical devices. Active devices generally do not respond well to exposure to high-intensity acoustic waves. Vianco and Hoskin [1992] have studied the wetting of copper by molten tin when assisted by a point source of ultrasonic energy (20 kHz, 20 70 W). Some of their results are reproduced in Fig. 3.23. Visual inspection of ultrasonically tinned substrates revealed tin lms on both the front and back surfaces of the copper coupons. This implies wetting is not solely due to direct mechanical erosion by the incident pressure wave, but also secondary propagation within the substrate and molten metal. Ultrasonic uxing is attractive as a combined method of uxing and tinning since it is uxless and residue-free. With the progressive improvements being made to the understanding of the process and the push toward greener and cleaner manufacturing, ultrasonic soldering is reentering more widespread use, including for
applying metallizations to refractory materials such as alumina [Naka and Okamoto 1989].
3.3.6
Working within the denition of a uxless process being one that does not use a solid or liquid ux, it is possible to include the option of using reactive gases to chemically remove surface oxide lms. A number of gases are sufciently active to be capable of reducing even indium oxide. Reactive atmospheres that have been examined and reported in the literature include formic acid and acetic acid vapor, carbon monoxide, and halogen gases (CF2Cl2, CF4, SF6, etc.) [Moskowitz, Yeh, and Ray 1986]. Formic and acetic acid are probably the most commercialized reactive gas for uxless soldering. Carbon monoxide and halogen gases require relatively high temperatures for the reduction of tin oxides in any reasonable timescale, for example, 350 C (660 F) maintained for 3 to 5 min, which precludes their use in many soldering processes. In this regard, these gases are similar to hydrogen gas. Formic acid vapor is active at typical soldering temperatures, and the process implementation is relatively simple. Its effectiveness can be judged from Fig. 3.24, which illustrates the improved self-alignment of ip-chip assemblies as formic acid is introduced at progressively higher concentrations, to the point of reducing surface
Fig. 3.20
processes.
Reduction of silver oxide by atomic hydrogen generated in proximity to the compound using a plasma. The energy of the plasma was deliberately chosen to be below the threshold at which oxide removal could be attributed to sputtering
oxide on the solder spheres [Lin and Lee 1999; Deshmukh et al. 1993]. The reaction between gaseous formic acid and metal oxide can be described by:
MO 2HCOOH M(COOH)2 H2O (where M is the metal)
This reaction occurs at a signicant rate at temperatures above about 150 C (300 F) and is therefore effective at stripping surface oxide lms prior to the solder melting. When the temperature is higher than about 200 C (400 F), the metal formate decomposes to liberate carbon dioxide and hydrogen:
M(COOH)2 M CO2 H2
those of lead, are toxic. Appropriately specied materials are required for furnacing equipment capable of utilizing formic acid vapor, owing to its corrosiveness, but the price premium appears to be relatively small and off-the-shelf equipment designed to work with it can be readily purchased. Acetic acid vapor is effective at reducing tin oxide on solder. This is evident by the difference in the time to wetting for Pb-62Sn solder on copper and gold-coated nickel (see Fig. 3.25). Short wetting times are always obtained on the clean gold/nickel metallization, but longer process times or higher concentrations of acid are necessary to clean bare copper and obtain similar rapidity of wetting [Frear and Keicher 1992].
3.3.7
The concentration of formic acid vapor in the atmosphere needs to exceed 0.5% in order to have a useful practical effect. Appropriate precautions need to be exercised with this process, as many organic metal compounds, especially
As an alternative to cleaning the faying surfaces on the components and the ller metal in situ, processes have been developed that convert the surface layers of oxide to stable compounds that protect against reoxidation and that are readily removed during the soldering cycle. One such process exploits uorine chemistry. This is the PADS (plasma-assisted dry soldering) process developed at The University of North Carolina. It uses a plasma-assisted reaction to dissociate CF4 or SF6 within an appropriate carrier gas to produce monatomic uorine, which, it is claimed, converts tin oxide to tin oxyuoride:
SnOx yF SnOxFy
Effect of fast atom cleaning on the strength of high melting point Pb-3Sn solder joints made without breaking vacuum between cleaning and soldering. Adapted from: Kohono et al. [1996].
Fig. 3.21
The reaction time is a few minutes in a reduced-pressure atmosphere, and the component
Effect on joint strength of deliberate delay, in a high-vacuum environment, between mechanical cleaning and assembly for high-lead solder (Pb-3Sn) ip-chip bonds
Fig. 3.22
Wetted area of copper coupons by molten tin at 245 C (473 F) as a function of ultrasonic power at 20 kHz and a process time of 30 s. The degree of wetting was also found to be directly proportional to the process time for the process temperature of 245 C (473 F), but the relationship was less clear at other test temperatures. Adapted from Vianco and Hoskin [1992]
Fig. 3.23
temperature need not exceed 50 C (122 F). Treated parts can be stored in air and on subsequent heating in a nonoxidizing atmosphere the tin uoride effectively volatilizes, leaving clean metal surfaces that are readily soldered. The solderability shelf life for parts treated in this manner is up to 1 week when stored in normal laboratory air or 2 weeks in a nitrogen-purged desiccator [Marczi, Bandyopadhay, and Adams 1990]. Although pH-buffered chemical treatments are available commercially that are capable of modifying oxide lms on copper, few are available that work well with solders or other base metals. One innovation that appears able to overcome this limitation is the ROSA (reduced-oxide soldering activation) process (Rockwell Scientic). In this process, an acidic solution containing vanadium ions and associated hydrogen is used to reduce oxides. The vanadium ions, which also supply hydrogen ions by virtue of their multiple valency states, are regenerated electrochemically in a closed loop, so the only effluent is oxygen gas. This process only removes surface
oxides, and therefore the treated components must be used before the oxide substantially regrows [Tench et al. 1995].
3.3.8
While the methods described in the preceding sections may be used to minimize the inhibiting effect of surface oxides on soldering, even when they are applied in combination it is normally not possible to make a uxless soldered joint of the same quality as a uxed joint. The reason is simply that these agents are not as pervasive at reaching the critical interfaces in the assembly and are not as active as a liquid chemical in rendering all metal surfaces clean at the soldering temperature and simultaneously protecting the cleaned surfaces from reoxidation. Additional means are therefore necessary to encourage wetting and spreading of molten ller metals during uxless processing. Generally, such meth-
Fig. 3.24
Effect of formic acid vapor concentration (0.35 to 1.7%) on the pull-in alignment of ip-chip components deliberately misaligned by 30 m (1.2 mil). Adapted from Deshmukh et al. [1993]
Fig. 3.25
Fluxless wetting of Pb-62Sn solder on copper and gold-on-nickel metallizations, using acetic acid vapor of varying concentration as an active reductant in the atmosphere
ods involve utilizing nonoxidizable metallizations on the parent materials, minimizing the surface-area-to-volume ratio of the ller metal, applying mechanical means of enhancing solder ow and, sometimes, metallurgical modication of the contact angle.
previously cleaned component surfaces. This approach is considered in the next section.
3.3.8.2
Preform Geometry
3.3.8.1
Because oxide lms grow so rapidly on most base metals, as shown in Fig. 3.13, in practice clean and solderable component surfaces can only be achieved if they are of gold. Accordingly, for uxless soldering, a gold coating is generally applied that will be of adequate thickness and sufficiently pore-free to ensure good solderability over its specied storage life. It is important to bear in mind the fact that the solderable shelf life afforded by a gold coating will be a function of the roughness of the underlying surface, the method of application of the coating and its thickness (see Chapter 4, section 4.1.2). Thus, a 0.5 m (20 in.) thick gold layer deposited by sputtering can be relied on to maintain excellent solderability for several months, even where the coated surface is relatively rough. On the other hand, the solderability of a gold layer of the same thickness, but applied by electroplating to a rough surface (Ra > 3 m, or 120 in.) may not offer adequate protection to an underlying base metal from atmospheric oxidation for more than a few days. The use of gold-coated surfaces imposes constraints on the solders that can be used. In particular, most tin-base solders, including leadtin eutectic, are largely incompatible with thick gold metallizations, on account of the high solubility of tin in gold, which results in the formation of AuSn4 and consequential embrittlement of the joints if this phase becomes dominant. This topic is discussed in further detail in Chapter 2, section 2.3.2. This restriction can only be overcome by applying high-quality gold coatings of less than a certain thickness to the joint surfaces, so as to prevent the formation of AuSn4 as the primary phase. Where thicker gold coatings are used, uxless soldering tends to be conned to high-gold and indium-base solders, which do not form catastrophic embrittling phases with gold. The necessity for a noble metal surface on the parent materials can be eliminated if the solder itself is applied as the barrier coating to
The form in which solder is admitted into a joint gap can make a profound difference to the success of uxed and especially uxless soldering processes. Notwithstanding the condition of the faying surfaces, it is a general rule that the greater the solder volume is in relation to its exposed surface area, then the more readily the process will work. This is simply because there is proportionally less oxide to impede wetting and spreading. It is therefore perhaps not surprising that manufacturers of solder paste go to quite considerable lengths to ensure that the solder balls used in its preparation are perfectly spherical, have exceptionally high surface smoothness (low Ra), and that the most common form of solder is round wire. The most appropriate geometry of preform for admitting solder into a joint gap depends on the shape of the components being joined. Ideally, the solder preforms should be designed to have not only a high volume-to-surface-area ratio but also to be orientated so that the advancing front of molten solder will sweep trapped gas out of the joint gap (see Chapter 4, section 4.3.1.1). In situations where it is desired to make joints with very low aspect ratiothat is, thin in relation to the plan areait is often not possible to achieve sufficient solder spread to reliably obtain complete joint lling. This scenario is encountered commonly in the microelectronics and photonics industries where there is a need to join planar components but with extremely narrow joint gaps on account of the relatively poor thermal conductivity of most solder alloys compared with many other metals. The thinnest solder preforms that can be economically purchased are 15 m (0.6 mil) thick. Handling these foils is almost an art form, and mechanically cleaning them is virtually impossible. The high surface-area-to-volume ratio of such preforms also runs counter to the need to minimize native oxides. A growing number of companies are now offering a solution to this problem in the form of substrates that are sold with the solder composition of choice preapplied to surfaces. The solder is usually applied by either electroplating
or a vapor-phase technique, through a mask, so that only the required areas of the substrate are coated. The solder thicknesses available range from 2 to 50 m (0.08 to 2 mil), and almost every common composition is available. These solder coated substrates offer a number of distinct advantages compared with a solder foil or wire:
Jigging is likely also to be simpler. The thickness of the solder joint is decreased
because the solder layer can be substantially thinner than the minimum practicable thickness of approximately 25 m (1 mil) required for a solder preform. Soldering behavior is improved by eliminating two joint surfaces with all the attendant problems, including oxide layers, from the joint gap. Solder spread is automatically conned to a precise area, and the responsibility of substrate wettability is also passed on to the substrate producer. The quality of substrates prepared in this manner has improved substantially, and, although there is a price premium compared to traditional foil, they are now available to aerospace and telecommunications qualied standards. Because the success of a uxless soldering process is critically reliant on the absence of surface oxide on the molten solder, it is good practice to test whether this condition is being achieved. This may be readily accomplished if the furnace has a viewing port. On heating to the process temperature, a source of clean solder will melt and its surface acquires a shiny, mirrorlike nish. This is often referred to as the liquid lake condition. Any defects in the liquid solder lm, such as texture, a gray bloom, or brown spots indicate that some element of the process, usually either the solder itself or the reow atmosphere, is in some way decient and must be corrected. 3.3.8.3 Mechanically Enhanced Solder Flow
cased in a skin of oxide. If it is possible to extrude virgin metal through ssures or other defects in the solder, then there is an improved prospect of achieving a sound joint. One method of doing this is to apply a compressive force to the joint gap. The effectiveness of this approach is illustrated in Fig. 3.26. Clearly, the higher the compressive pressure, the more effective it is, with the optimal load in the region of 10 g/mm2 (14 psi), or more. A loading such as this is relatively easy to achieve with weights or spring-loaded jigs for all but the largest components. Precautions need to be taken to ensure that the load is applied uniformly and parallel to the joint gap, and that the method of application does not pose a thermal sink on the assembly that would give rise to adverse temperature gradients. Figure 3.27 shows application of this approach to uxless soldering of a GaAs semiconductor die to a gold thick-lm metallized alumina substrate. The back of the semiconductor is metallized with gold, and the ller metal is a preform of Ag-96Sn solder, 15 m (0.6 mil) thick. The process superheat was 30 C (86 F). Xradiography reveals the joint to be free of voids (the black circles are blind vias in the GaAs, which are included for functional reasons). 3.3.8.4 Metallugically Enhanced Solder Flow
Occasional reports appear in the published literature of the signicant difference that low concentrations of other metals can make in promoting wetting and spreading of molten solders.
No matter what precautions are taken, in an industrial uxless soldering process it will always be the case that by the time the components and preform have been set in jigs, loaded into the protective atmosphere, and heated to the process temperature, then the solder will be totally en-
Void level (as a percentage of the plan area of the joint) versus the load applied to preforms of three solders during uxless processing. The soldering conditions used for the three solders were a superheat of 25 C (or 77 F) and dry nitrogen as the joining atmosphere. The joints measured approximately 10 mm 10 mm 25 m (0.4 in. 0.4 in. 1 mil).
Fig. 3.26
Some cases are cited in Chapter 2, section 2.2 and Chapter 5, section 5.8. Virtually all uxless aluminum brazes make use of the fact that ppm levels of bismuth and beryllium have a marked effect on the wetting and spreading characteristics of the Al-12Si eutectic alloy. Further information on uxless aluminum brazing is to be found in the planned companion volume Principles of Brazing. The authors have investigated improving the uidity of indium by minor additions of other elements with a view to enhancing the use of indium ller metals in uxless processes. Figure 3.28 shows the wetting angle as a function of heating time for pellets of indium solders of controlled weight and geometry wetted on to silver substrates at 200 C (392 F) in a vacuum of 1 mPa (1.5 107 psi). The measurements were taken from video stills of the substrate viewed edge-on, with the timeline beginning at the onset of observed melting. Of the additions, zinc has an adverse effect. Antimony is initially neutral, but appears to impede any additional wetting. Bismuth and gold are benecial additions that improve wettability, which requires an extended dwell at the soldering temperature to take effect, while the rare earth metal cerium produces a marked and immediate reduction in contact angle. This could be a fruitful area for further research in developing new solder alloys (see Chapter 5, section 5.8).
3.3.9
out resorting to the use of chemical uxes. The joint measured approximately 10 by 5 mm (0.4 by 0.2 in.), and the components were both metallized with 3 to 5 m (120 to 200 in.) of nickel overlaid with 3 m (120 in.) of gold. In the industrial environment where this work was carried out, it was necessary to demonstrate the benecial effect of each change in the process before other alterations could be investigated. Shear strength was used to assess the quality of the soldered joints. The rst change to be considered was to apply a load to the joint gap during the heating cycle. Application of a compressive stress helps ensure that all free surfaces within the joint are in intimate contact. This has the effect of minimizing temperature gradients across the joints and, as explained previously, promotes rupturing of the oxide skin on the solder when it melts. The data shown in Fig. 3.29 (air) indicates the relationship between the applied load and resulting joint strength, with the joint shear strength increasing by roughly 1 kg (2.2 lb) for every 100 g (0.2 lb) of applied weight, for the range of applied pressure shown (0 to 6 g/mm2, or 0 to 75 psi). A more signicant improvement to the process was made by carrying out the joining operation in a furnace maintained under dry nitrogen, giving a combined oxygen and water vapor content of less than 10 ppm in the working atmosphere. This measure provided some protection against oxidation of the parts and solder and delivered a stepwise improvement in joint strength; see Fig. 3.29 (nitrogen). It was ob-
The requirement was to improve the quality a soldered joint that was being made using a 25 m (1 mil) thick preform of the In-48Sn solder, with-
Fluxless soldering of a GaAs monolithic microwave integrated circuit, approximately 3 5 mm (0.12 0.12 in.) achieved by application of a compressive load of 100 g/mm2 (140 psi) during the heating cycle. Source: BAE Systems
Fig. 3.27
The effect of different doping additions on the uxless wetting angle of indium on silver substrates, as a function of time following the commencement of melting of the solder
Fig. 3.28
served that as a result of changing to the nitrogen atmosphere, considerably less solder was exuded from the joint gap, implying that the increase in joint strength is largely due to improved wetting of the component surfaces. Hitherto, the foil preforms of solder had been used as received from the supplier. These are covered with a signicant layer of native oxide. Mechanical abrasion of the preform surfaces with a glass-ber brush immediately prior to use led to a further improvement in joint strength, as can be seen in Fig. 3.29 (cleaned). A thin foil solder preform is not a particularly attractive method of admitting solder into a joint because:
(wire). The cruciform conguration also provides for mechanical stability during jigging, and the direction of solder ow after melting helps sweep out any trapped gas from the joint gap (see Chapter 4, section 4.3.1.1). Examination of the joint surfaces of disassembled parts revealed good wetting of the components and a continuous bead of shiny solder formed at the edge of the joint.
3.3.10
The foil is difficult to handle and clean if thin. There is a high risk of trapping pockets of
vapor when the parts and foil are jigged together (see Chapter 4, section 4.3.1.1). Thin foils also command a cost premium, especially when they have to be purchased to custom dimensions. Solder in the form of round wire avoids these drawbacks. The preform was prepared from two equal lengths of wire, cold compression welded so as to form a symmetrical cross, as shown in Fig. 3.30. The surface area-to-volume ratio of this cruciform conguration is much lower than for a foil, and wiping the wire several times with a paper tissue soaked in solvent is an easy method of striping the solder oxide (see Fig. 1.27). The benecial effect of substantially removing the oxide lm from the solder by these means is attested by the results presented in Fig. 3.29
From the preceding discussion it might be assumed that to attempt uxless soldering of aluminum would be ineffectual. However, uxless soldering of aluminum is practiced as a repair technique. It was developed during World War II as a method of patching small bullet holes in airplane skins. The original process simply used zinc as the joining material, but the joints were weak and susceptible to corrosion. The modern variation employs a ller alloy containing about 90% Zn, 7% Al, with the balance being magnesium, manganese, and other elements. The alloy melts at approximately 380 C (716 F), and therefore it qualies as a solder. The joining method involves heating the aluminum part, often with a standard blowtorch and puddling (rubbing) the solder until a wetted surface lm is developed over the area to be joined. No ux is necessary because the alumina is physically removed and the exposed aluminum overcoated with a more oxidation-resistant alloy. Both joint surfaces are prepared in this manner, placed in
Shear strength of joints approximately 10 5 mm (0.4 0.2 in.) in area made uxless using In-48Sn solder at process temperature of 150 C (302 F) to goldmetallized components, as a function of the applied compressive load, showing also the effect of atmosphere quality and condition of the solder (see text for details)
Fig. 3.29
Cross-shaped preform of In-48Sn solder prepared by cold welding of two 300 m (12 mil) diam wires at the common intersection. Source: BAE Systems
Fig. 3.30
contact and the joint made by reheating without further addition of the ller metal [Phillips 1994]. Additional mechanical agitation helps meld the two molten liquid skins together. The aluminum in the solder helps to prevent corrosion of the joint, the zinc addition lowers the melting point, while the minor elements assist wetting and spreading. It is reported that joints made in this manner can have shear strengths around 90% of that of the parent metal. If the layer of applied zinc-base solder can be made extremely thin, it is possible to diffuse it completely into the base material if the process cycle is suitably extended. This technique is properly known as diffusion soldering, but it does enable uxless, effectively corrosion-resistant joints to be made to aluminum, which are not a source of mechanical weakness [Ricks et al 1989]. The applicability of the approach is limited by the long heating cycle times and the need to either use precisely machined parts or apply pressure to force the abutting surfaces into intimate contact. This latter requirement arises from the very limited volume of zinc in the joint and hence the small quantity of liquid ller metal available to ll the joint gap.Adiscussion of diffusion soldering, is to be found in Chapter 5, section 5.9 and diffusion brazing in the planned companion volume Principles of Brazing.
tion and physical meaning of internal energy and entropy are explained below. As is shown, an important property of the Gibbs free-energy function is that it is always a minimum at equilibrium, and the extent of its departure from the minimum value provides a measure of the tendency of a reaction to proceed spontaneouslythat is, of the driving force for the reaction.
Appendix A3.1: Thermodynamic Equilibrium and the Boundary Conditions for Spontaneous Chemical Reaction
The thermodynamic function that provides a measure of the driving force of a chemical (including metallurgical) reaction is the Gibbs free energy, which is dened as:
G E PV TS (Eq A3.1)
where E is the internal energy, S the entropy, T the absolute temperature, P the pressure, and V the volume of the materials system. The deni-
The minus sign denotes that work is done on the system to raise its internal energy. This expression provides a thermodynamic denition of internal energy. The internal energy of a system depends only on the state of the system (dened in terms of macroscopic or thermodynamic properties, such as the pressure and temperature of the system). For this reason, internal energy is termed a function of state.
Where work is done in changing the volume of a chemical system by an increment dV through the application of external pressure P, then:
dW PdV dE PdV
ings. It follows that a reciprocating engine that operates by extracting heat from one source must reject some of this heat to a sink at a lower temperature. If the operating cycle of the engine is reversible, such that work can be performed to pump heat from the sink back to the source, it is possible to show that in accordance with the Second Law, the integrated ratio dQ/T over one complete cycle is zero:
0 dQ T
In practice, most systems are not totally insulated from their surroundings so that thermal energy may be exchanged between them. If an increment of work dW is done on the system and an increment of heat dQ is exchanged with the surroundings, then the internal energy dE will change by the amount:
dE dQ dW
This equation is a mathematical expression of the First Law of Thermodynamics which, for a chemical system, may be written:
dE dQ PdV (Eq A3.2)
The circle through the integral sign denotes that the integration is to be carried out over the complete cycle, and the letter R is a reminder that the equation applies only if the cycle is reversible. This result is known as Clausius theorem. If the integration is carried out over only part of the cycle, say between two states 1 and 2, then the integrated ratio dQ/T is not zero, but equals the difference between the values of a thermodynamic function at the two states:
dQ T
S1 S2
This thermodynamic function of state is called entropy. If the two states are innitesimally close, then the relationship can be written:
( )
dQR T
dS
(Eq A3.3)
Subscript R indicates that this equation only holds if the heat increment dQ is transferred reversibly. This equation provides a mathematical expression of the Second Law of Thermodynamics. A consequence of the fact that entropy is a function only of state, a system that has changed from state 1 to state 2, always has entropy S2, which differs from that of the initial state S1 by S1,2 S2 S1, irrespective of the means used to drive the system. Thus, for example, the system may have been set in motion, and some of the kinetic energy converted into heat in overcoming frictional forces, thereby raising its temperature to a value that takes the system from state 1 to
state 2. In this irreversible process, the energy was not supplied to the system as heat so that:
dQ T
Gibbs free-energy function, G, to determine the temperatures and pressures under which chemical reactions are thermodynamically favorable, as well as the direction of the reactions. In incremental form, Eq A3.1 can be written:
dG dE PdV VdP TdS SdT
The letter I denotes that the process is irreversible. However, the entropy change S1,2 is still the same as that obtained by a reversible change between states 1 and 2, because it depends only on these states and not on the process connecting them; that is, here too S1,2 S2 S1. Thus, in all irreversible processes, the entropy change is greater than (dQ/T), where dQ is the heat absorbed at each incremental step in the irreversible change. This result can be generalized to the statement that in a spontaneous irreversible change, the entropy of an isolated system will increase, and when in equilibrium, it will remain constant. Considering the system and its surroundings together (i.e., the universe), any kind of process can be represented in entropy terms by:
dS (universe) 0
Substituting for dE and TdS from Eq A3.2 and A3.3 (all chemical/metallurgical processes being reversible) gives:
dG dQ PdV PdV VdP dQ SdT VdP SdT
For a reversible process at constant temperature (isothermal) and constant pressure (isobaric), that is, when the system is in equilibrium:
dG 0
Therefore, from a thermodynamic viewpoint, which is macroscopic, entropy can be understood as the propensity of a system to undergo a change, such as a chemical reaction. A clearer physical picture of entropy can be obtained at the microscopic level, where a system may be regarded as an ensemble of atoms or molecules. On this basis, it can be shown that entropy provides a measure of the degree of atomic or molecular disorder that exists in the system, and this will always tend to increase. This concept is consistent with the observations that all metals are intersoluble, albeit in some cases only to a small extent, and that all liquid metals will wet the clean surfaces of solid metals.
and G is constant and has a minimum value. This is an important result for metallurgical reactions, because these can be considered as taking place usually at constant temperature and pressure. More generally, at constant pressure, dP 0, and then:
dG S dT
applies, where n is the number of moles of gas and R is the gas constant. Then:
dG nRTdP/P at constant temperature
so that the Gibbs free-energy change resulting from a change from state 1 to state 2 at constant temperature is:
G2 G1 nRT ln P2/P1
The Gibbs free energy, like any other measure of energy, must have some reference point. By convention, a zero value of G is assigned to the stable form of elements at 25 C (77 K) and 1 atm of pressure. Then the Gibbs free-energy change of a gas at constant temperature from its value Go at atmospheric pressure, which is dened as its standard state value, is given by:
G G o nRT ln P (Eq A3.4)
where P is the pressure corresponding to the free-energy state G, expressed in atmospheres. Although Eq A3.4 is strictly valid for ideal gases, it is also approximately applicable to real gases and can be used for them at pressures close to normal atmospheric pressure (1 atm). In the case of solids, the molar volumes are small compared with those of gases, so that the change in the Gibbs free energy of solids resulting from small pressure excursions, P, such that P 1 atm (100 kPa) at constant temperature is small and, to a rst approximation, may be neglected in reactions involving solids and gases. It is also assumed that the solubility of the gaseous species in the solid phases is negligible at the temperatures of interest, as is largely the case in practice. It is now possible to determine the pressure dependence of the Gibbs free energy of the reagents that participate in a chemical reaction. Consider a reaction involving four gases, A, B, C, and D and two solids, X and Y, all at constant temperature T, as follows:
xX aA bB yY cC dD
where G(A), G(B), etc. are the Gibbs free energies of 1 mole of the reagents A, B, etc. at pressures P(A), P(B), etc., and Go(A), Go(B), etc. are the corresponding values at 1 atm. The free-energy change for the reaction is, from Eq A3.4:
G G(products) G(reactants) cGo(C) dGo(D) aGo(A) bGo(B)
P(C)c P(D)d P(A)a P(B)b P(C)c P(D)d P(A)a P(B)b
RT ln
where a, b, c, d, x, y are the number of moles of each of the reagents. The gaseous reagents are assumed to behave as though they are ideal gases. The Gibbs free energies G(X) and G(Y) of the solid constituents at moderate pressures are approximately equal to their values at atmospheric pressure, as explained previously. Therefore:
Free energy of x moles of solid X xG(X) xGo(X)
Go RT ln
The Gibbs free-energy changes of the solid reagents can be neglected, for the reasons given previously. Under equilibrium conditions, temperature and the respective pressures P(A), P(B), and so forth, are constant, and:
G 0
Hence, the Gibbs free-energy change when the gaseous reactants A and B in their standard states are transformed to the products C and D in their standard states may be expressed in terms of the partial pressures of the respective reactants in equilibrium, thus:
P(C)c P(D)d P(A)a P(B)b
per mole of oxygen participating in the reaction. That is, the driving force needed to oxidize a metal, as expressed by the Gibbs free-energy change, is directly related to the oxygen partial pressure of the atmosphere according to Eq A3.6.
G RT ln
o
Since the Gibbs free-energy change Go, for a particular reaction at a xed temperature and at atmospheric pressure has a xed value, so too does the argument of the logarithm. This constant is called the equilibrium constant KP, because it can be used to determine the equilibrium state that a reacting system will attain.
P(C)c P(D)d P(A) P(B)
a b
KP
The subscript P denotes that the equilibrium constant is specied in terms of pressure. Equation A3.5 becomes:
Go RT ln KP
there is one gaseous constituent and two solids, so that the equilibrium constant is simply:
KP
PO2
M
y/2
M where PO is the partial pressure of oxygen re2 quired to effect the oxidation reaction, or the dissociation pressure of the oxide, and
GoRT ln PO2
(Eq A3.6)
Hong, S. and Kang, C., 2001. Fluxless Bumping in Flip-Chip Package by Plasma Reow, Proc. Conf. International Symposium on Electronic Materials and Packaging, 1922 Nov (Jeju, Korea), p 139144 Hong, S., Kang, C., and Jung, J., 2002. FluxFree Direct Chip Attachment of SolderBump Flip-Chip by Ar H2 Plasma Treatment, J. Electron. Mater., Vol 31 (No. 10), p 11041111 Hunt, C., 2000. Development of Surface Insulation Resistance Measurements for Electronic Assemblies, Report MATC(A) 070, National Physical Laboratory Hyland, K. and Rao, S., 1996. An Environmentally Friendly Manufacturing Solution, Circuits Assem., Vol 7 (No. 7), p 3840 Jacob, K.T., Chandran, A., and Mallya, R.M., 2000. An Assessment of the Reduction Potential of Hydrogen Plasma, Z. Metallkde., Vol 91 (No. 5), p 401405 Jones, J.B. and Thomas, J.G., 1956. Ultrasonic Soldering of Aluminium, Proc. Symp. Symposium on Solder, 1920 June (Atlantic City, NJ), American Society for Testing and Materials, p 1529 Klein Wassink, R.J., 1989. Soldering in Electronics, 2nd ed., Electrochemical Publications Kohono, A., et al., 1996. Si Circuit Chip Joining Technology Using Ar Atom Bombardment, Proc. Conf. EuPac 96, 31 Jan2 Feb (Essen, Germany), p 4147 Kuhmann, J.F., et al., 1998. Oxidation and Reduction Kinetics of Eutectic SnPb, InSn and AuSn: A Knowledge Base for Fluxless Solder Bonding Applications, IEEE Trans. Comp., Pack. Manuf. Technol., Part C, Vol 21 (No. 2), p 134141 Lea, C., 1991. After CFCsMaking the Economic and Technical Choice, Circuit World, Vol 18 (No. 1), p 2833 Lin, W. and Lee, Y.C., 1999. Study of Fluxless Soldering Using Formic Acid Vapour, IEEE Trans. Adv. Pack., Vol 22 (No. 4), p 592601 Manko, H.H., 2002, Solders and Soldering, 4th ed., McGraw-Hill Marczi, M., Bandyopadhay, N., and Adams, S., 1990. No-Clean, No Residue Soldering Process, Circuit. Manuf., Feb, p 4246 Martin, M. and Fromm, E., 1977. LowTemperature Oxidation of Metal Surfaces, J. Alloy. Compd., Vol 258, p 716 McLaughlin, M.C., et al., 1998. The Aqueous Cleaning Handbook, Morris-Lee
Publishing Group
Vianco, P.T. and Hoskin, F.M., 1992. Analysis of Ultrasonic Tinning, Proc Conf., NEPCON West 97, 2327 Feb (Anaheim, CA), p 17181729 Wakelin, R.J., 1993. SMARTA Novel Lead-Free Soft Solder for Stainless Steel,
Proc. Conf. Innovation Stainless Steel, 1114 Oct (Florence, Italy), p 3.3273.332 Wicks, C.E. and Block, F.E., 1963. Thermodynamic Properties of 65 Elements Their Oxides, Halides, Carbides, and Nitrides, Bulletin 605, U.S. Bureau of Mines, U.S. Government Printing Office
CHAPTER 4
Thus, in certain situations, catastrophic mismatch stresses may be overcome by modifying the stress distribution in the vicinity of the joint. Problems associated with the formation of deleterious intermetallic phases in the joint, lack of wetting, and, at the other extreme, excessive erosion of one or more of the parent materials can usually be circumvented by interposing a layer of a different metal between the ller and
the parent material and thereby altering the metallurgical constitution of the joints. Fluxes and active ller alloys can be used to improve wetting and reduce void levels. Some of the possible remedies that may be brought to bear on these and other problems are detailed in the following sections. If, on the other hand, no joining solution proves technically tractable, or if the solutions
Table 4.1
Phase of development
Scale of production Cost constraints Size and weight limits Statutory requirements Thermal environment Stress environment Chemical environment Joining temperature Joining atmosphere Mismatch stress
Service conditions
Parent materials
Filler alloys
Joining temperature Joining atmosphere Joint geometry Filler geometry and form Metallurgical constraints
Process assessment: identication of critical materials problems Mismatch Wetting (Mechanical solutions) Identication and achievement of solutions Interlayers Multilayers Graded structures Compliant structures Diffusion soldering/brazing Determination of process viability
Alloying Erosion (Metallurgical solutions) Surface processing Wettable coatings Barrier coatings Active ller alloys Fluxes Diffusion soldering/brazing Phases
Process specied
are not economically justiable, then changes earlier in the decision-making chain are required. In extreme situations, this may require drastic revision, perhaps even to the extent that other means of assembly have to be considered or that the functional requirements of the product must be relaxed.
4.1
In principle, most metals can be joined using ller alloys. However, when there is a requirement to join two different parent materials together, the available choice of llers that are compatible with both is narrowed somewhat, especially when the constraints on the solder, processing conditions, and properties of the joints mentioned previously are imposed. The problem is often made more acute by the fact that the joining processes tend to be left to a later stage of product design, which further reduces the available options. The joints are always considered as an integral part of the overall design of an assembly if manufacturing is to be facilitated. Metallurgical incompatibility of materials and processing conditions will manifest itself through poor wetting, excessive erosion of the parent materials, and/or the formation of undesirable phases. Means for eliminating or suppressing these deleterious characteristics are described as follows.
4.1.1
Restrictions applied to the choice of joining temperature and atmosphere, including the use of uxes, can result in poor wetting of the component surfaces by the molten ller if the permissible process conditions are inadequate to ensure that the joint surfaces are sufficiently clean. Active solders (described in section 4.1.2.2 of this chapter) can often help to overcome this problem, but it is seldom possible to use solders containing active wetting agents. Alternative remedies are then necessary. Poor wetting is a particularly serious problem when the parent materials are refractory metals. Their reactivity with oxygen, and, in some cases, with other elements in the atmosphere, and the stability of the reaction products on the surfaces of these materials cause poor wetting. Nonme-
tallic phases present at the surface of materials, such as graphite inclusions in cast iron and nonmetallic components in metal-matrix composites, can also inhibit wetting. These problems are not restricted to the parent metals but can also encompass the ller metal. A prime example is provided by zinc-base solders when these need to be used without an appropriate uxing agent. Foils and other preforms of these alloys tend to produce poor wetting and spreading over the joint surfaces. However, by making small additions of elements that lower the surface tension of the molten ller and that also help to destabilize the native surface oxide layer, wetting can be considerably improved. This is illustrated for indium in Chapter 3, Fig. 3.28, and other examples are outlined in Chapter 2, section 2.2. The concentration of the individual additions must be restricted to a maximum of approximately 1% in order to avoid perceptibly altering the bulk metallurgical characteristics of the ller alloy. If the joining environment, which includes uxes when these are approved, cannot adequately clean and protect the components from oxidation during the process cycle, a favored solution is to apply a coating of a more noble metal to the precleaned joint surfaces. This coating may be sacricial in that it is subsequently dissolved by the ller, which then wets the clean surfaces of the parent material. At the same time, the composition and thickness of the coating should be such that when dissolved, it does not give rise to brittle phases by reaction with the ller and parent materials. Copper, silver, and gold are the principal elements used as wettable metallizations because of their nobility, metallurgical compatibility with most solder alloys, and ease of deposition. Tin is also widely used in conjunction with soldering processes, because tin oxide is readily displaced by an advancing wave of the molten solder, which itself often contains tin. The benets of tin are that it is relatively inexpensive and that it presents minimal risk of unfavorably altering the constitution of the solder. In the semiconductor and optoelectronic industries, platinum overlaid with a gold ash is a frequent choice of wettable metallization. Thinlm deposition equipment with a platinum sputtering source is often readily available, and the materials cost is insignicant compared to the process cost. Platinum has the benet of exceptionally low solubility in tin- and indium-base solders and will not oxidize when exposed to air
(Table 4.2). The gold ash ensures good spreading and aids visual inspection of the faying surfaces prior to use. The reaction product with tin-base solder has been identied as an exceptionally thin interfacial layer of PtSn4 [Kuhmann 1977]. This metallization scheme confers excellent wetting and spreading with precisely controlled reproducibility, which are useful attributes when joining high-value components. Solderable metallizations are usually applied by either wet plating or vapor deposition techniques. Tin coatings can be applied by dipping methods. The choice of deposition method is determined by various factors, such as the size of the component, its geometry, the required scale of production, the capital and running costs of the deposition equipment, and the thickness of the metallization.Additional details of the merits and limitations of each of these coating techniques are given in Appendix A4.1 in this chapter. Any metallization applied needs to be sufficiently thick for it to protect the underlying material from corrosion and to maintain wettability of the component over a reasonable storage period. The uniformity and density of sputterdeposited coatings of gold afford such protection at a thickness of typically 0.3 m (12 in.), whereas electroplated coatings need to be more than an order of magnitude thicker to provide the same shelf life [Humpston and Jacobson 1990]. This point is illustrated in Fig. 4.1, which shows the wettability, in terms of the wetting force measured on a wetting balance, of chromiummetallized coupons coated with different thicknesses of gold by sputtering and electroplating, as a function of the storage time in an open atmosphere. The porosity of thin electroplated or evaporated coatings is frequently overlooked, especially when a process is transferred from development to a less rigorous manufacturing environment. Further proof of the ineffectiveness of noble metal coatings in preventing oxidation of an underlying base metal comes from the benecial exploitation of their use to grow oxide lms of controlled thickness in the manufacture
Table 4.2 Dissolution rate of platinum in tin-base solder at selected temperatures
Temperature C F Dissolution rate
of solar cells. In particular, 5 m (200 in.) of silver, applied by thermal evaporation, to a clean copper surface permits the growth of Cu2O at the common interface at a rate of roughly 0.1 m/h (4 in./h) at 500 C (930 F) [Rosenstock and Riess 2000]. The alloying behavior of the solder with the coating metal may establish a maximum limit to the thickness of the metallization beyond which joint embrittlement ensues, as explained in sections 4.1.3 and 4.1.4 of this chapter. Where economic and size factors dictate that electroplating is the preferred method for applying gold coatings, the components should be provided with typically 3 to 4 m (120 to 160 in.) thick platings of gold to confer an adequate solderable shelf life. Then, immediately prior to the soldering operation, the level of gold can be considerably reduced to suit, in particular, high-tin solders by wicking this metal off the joint surfaces. The removed gold can be reclaimed from the discarded solder. Metallizations that promote wetting can also be used to advantage to conne the molten solder to specic areas on the surfaces of components. By selectively applying the coating to prescribed areas, the solder spread can be restricted accordingly. If the wettable coating is not adherent as applied, it often can be secured on the surface of a metal component by a subsequent heat treatment in a nonoxidizing atmosphere. This process can promote interdiffusion between the metal and the
Solderability shelf life of gold-coated components. Thicker and denser coatings are more impervious to oxygen and water vapor and therefore confer greater protection to the underlying metal.
Fig. 4.1
coating, resulting in a graded interface akin to that obtained using a carburizing (carbonenriched) or sheradizing (zinc-enriched) surface treatment. For example, titanium is rendered solderable by coating with gold (as applied, the gold layer will brush off if touched) and then heat treating at 750 C (1380 F) for 30 min in a nitrogen atmosphere. Copper lands on printed circuit boards and electronic component leads are often coated so as to maintain solderability in storage and to permit extremely rapid wetting and spreading by the molten solder at the appropriate time. Speed is important, because printed circuit boards pass through the solder wave of a wave-soldering machine in 2 s or less. The coatings must also entrain low materials and application costs. There are basically three main types: fusible coatings (tin, lead-tin, and other tin alloys); soluble coatings, that is, noble metals that are readily soluble in molten solder (silver, gold); and organic coatings (1,2,3-benzotriazole, 1,3-benzodiazole, imidazole, and similar compounds). Often, these coating strategies are combined. For example, one commercial product system designed to maximize the preservation of the solderability of copper involves application of a thin (0.07 to 0.1 m, or 3 to 4 in.) layer of silver, which is itself protected from damaging environmental effects by an organic layer, typically also 0.1 m (4 in.) thick. Silver is a good choice of overmetal because it is more noble than copper, and unlike lead/tin coatings, it cannot form an intermetallic structure with copper, which generally has poor solderability. The organic coatings are designed to form a chemical bond to a clean substrate surface, and hence protect it during storage, and then either dissolve in ux or decompose cleanly just below the soldering temperature (typically 120 to 150 C, or 250 to 300 F). Nickel/gold and nickel/palladium nishes are more correctly termed barrier coatings, because the nickel and palladium protect against reaction between the solder and the substrate. The gold overcoat preserves the solderability of the nickel or palladium. A comprehensive review of surface nishes relevant to electronic circuit assembly on printed circuit boards is provided by Vianco [1998].
when their surfaces are scrupulously clean. This is because they are chemically very stable, with their atoms strongly bound to one another. Therefore, these materials will not react with and be wetted by molten solder unless the latter contains an active element, which can attach itself to the anionic species of the nonmetallic material. In a compound or complex, this is normally oxygen, carbon, nitrogen, or a halide element. There are two solutions to soldering to ceramics. One is to use an active ller metal, as discussed in section 4.1.2.2 of this chapter. Active metal joining is only effective if sufficiently high temperatures, typically above 800 C (1470 F), can be used for the joining operation, so that the active ingredient is able to react with the nonmetal. Where activated llers produce successful joints, it is observed that the active element concentrates at the interface with the nonmetallic base material and imparts metallic characteristics to the surface of the ceramic. Where this approach is incompatible with the joining process, a very similar result can be achieved via a different route. This involves coating the joint surfaces with a metal that will bond strongly to the underlying nonmetal and that is at once wetted by the ller while not entirely dissolving in the process. Any erosion through to the original component surface would result in dewetting. This is exemplied by the dissolution of chromium metallizations by bismuth-containing solders observed at temperatures above 261 C (502 F). Bismuth and chromium react to form a low-melting-point eutectic alloy at this temperature [Humpston and Jacobson 1990]. Further details of solderable coatings for nonmetals are given in the following section. 4.1.2.1 Solderable Coatings on Nonmetals
4.1.2
Nonmetalsnamely, ceramics, glasses, and plasticsare not wetted by most solders, even
Wettable coatings can be applied to the nonmetallic components by methods similar to those used on metalsnamely, physical vapor deposition, chemical vapor deposition, and wet plating. Also widely used are red-on glass frits loaded with particles of metal powder or ake, often referred to in the literature as thick-lm metallization techniques. The vapor phase deposition route tends to be favored wherever thin coatings of high quality are required. It is versatile and permits a wide range of metal and alloy coatings to be applied. When metallizing a nonmetallic material, it is usual to deposit more than one layer onto the
joint surface. It is essential that the layer in direct contact with the nonmetal (henceforth referred to as the foundation layer) is an active metal, which is usually nickel, chromium, or titanium. The choice is largely dictated by the solubility of this metal in the ller alloy, which must be low but nite, and the ability of these active elements to bond strongly to nonmetallic materials, provided that the surfaces are clean. According to a simple thermodynamic model, it might be expected that the adhesion of a metal to, say, an oxide is given by the Gibbs free energy of formation of the oxide of the metal: The greater the extent to which the free energy is reduced when the metal reacts to form an oxide, the more active is the bonding. That this model is an oversimplication of reality is clear from the observation that chromium forms an adherent bond to silica (quartz), even though the Gibbs free energy of formation of silica is higher in magnitude than that of chrome oxide. Evidently, the Gibbs free energy of oxide formation is only one factor. This example merely highlights the complexity of the subject of metal-to-nonmetal bonding and accounts for the considerable literature that it has produced [Peteves 1988]. Nevertheless, the desired end result is that the component surface is rendered sufficiently metallic to enable wetting by the ller metal. A sputtering process usually incorporates a capability for cleaning surfaces prior to deposition, by being operated in a reverse-bias mode, so that the atoms of the inert carrier gas in the deposition chamber are made to bombard the surface and physically remove lms of contaminant. A separate ion source is often included in evaporation systems to perform the same function. An in situ preclean stage is essential to remove absorbed species and achieve good adhesion by the foundation metal. Chromium, or a mixture of nickel with chromium (Nichrome), is frequently recommended as the reactive metal for the foundation layer on nonmetals [Holloway 1980]. Nichrome has a lower intrinsic stress than pure chromium, which is benecial for some strain-sensitive electronic and optical components. Chromium reacts with many nonmetallic compounds to form complex chromates. These are not only strongly bonded to the nonmetal but also act as a barrier to further interaction taking place by diffusion between chromium and the nonmetal [Mattox 1973]. Hence, it is possible to use chromium as a highintegrity metallization on glass and intrinsically stable ceramics such as alumina and quartz. Ten-
sile strengths are reported to be in the region of 70 MPa (10,000 psi) [Vianco, Sifford, and Romero 1997]. The benet of this type of chemical bonding is evident when metallizing relatively unstable ceramics, such as zinc oxide. Zinc oxide is widely used in electrical voltage surge suppressors, known as varistors [Leite, Varela, and Longo 1992; Wersing 1992]. Heating a zinc oxide ceramic metallized with titanium or zirconium causes the metallization to blister and spall off due to the oxidizing reaction:
2ZnO Ti TiO2 2Zn G2ZNOTi (573 K) 272 kJ (65 kcal)
A by-product of this reaction is the volatilization of free zinc from beneath the coating as soon as the coated part is heated, which results in the detachment of the metallization. On the other hand, a chromium metallization is entirely benign toward zinc oxide, even upon heating to 400 C (750 F) for prolonged periods, due to the presence of the intervening chromate barrier layer. Chromium is also marginally easier to apply than titanium, because it is slightly less reactive toward oxygen in the air, and hence, the quality of the vacuum chamber in which the deposition is conducted can, accordingly, be slightly relaxed. The high reactivity of titanium and other active metals toward oxygen and nitrogen in air means that they rapidly lose their wettability even after a brief exposure to the atmosphere. To overcome this problem, a more noble metallization that offers good wettability to ller metals must be deposited over the active-metal foundation layer immediately, while the deposition chamber is maintained under a protective (low-oxygen) atmosphere. A metallization system that has been found to be effective for many soldering applications is one that comprises a 0.1 m (4 in.) thick layer of a foundation metal and a 0.3 m (12 in.) overlay of a wettable metallization, such as gold, both applied by sputtering. The chosen foundation-metal thickness is specied to provide good step coverage even after taking into account parameter variations in an industrial process that may result in local thickness variations up to 25%. The thicker gold layer provides a reasonable shelf life and usually permits joints to be made using tin-base solders without risk of embrittlement (see Chapter 2, section 2.3.2).
An alternative method of coating a nonmetallic material with an adherent metal coating is to use a two-stage joining process in which the rst step is to allow an activated solder or braze, capable of providing a solderable surface, to wet and spread over the component surfaces. The tinned surface is then usually mechanically dressed to present a at metal surface for the actual joining step, which is performed at a temperature below the solidus point of the activated ller metal. Because of the number of elements that are likely to be present in the joint, the resulting alloy constitution can be somewhat complex unless the process is designed such that the alloy used for the coating and the solder alloy have constituents in common. A somewhat different approach is to re on a relatively thick (typically, 1 to 10 m, or 40 to 400 in.) metal coating. This type of metallization process tends to be used only on glass and ceramic materials because of the high temperatures involved [Bever 1986]. In very simple terms, a thick-lm paste can be considered as a mixture of metal and ceramic or glass particles. On ring, the glass or ceramic phase wets the component surface, while the metal particles oat to the surface. Adhesion between the solidied glass or ceramic phase and the metal particles is by a combination of chemical and physical mechanisms. Further details of the range of commercial processes that are available are given in Appendix A4.1 in this chapter. Reference may sometimes be found in older technical literature to the active hydride process. It involves applying a metal hydride, usually of titanium or zirconium, in the form of a paste to the component surface. On heating, the hydride thermally decomposes, liberating hydrogen to leave a metal lm on the component surface [Pershall 1949]. This method has lost favor in recent years to alternatives, because the quality of the resulting metallization tends to be variable, as reected in the mechanical properties of the joints [Mizuhara and Mally 1985]. The unpredictable nature of the metallization quality is due to its high sensitivity to variations in the atmosphere during the ring stage. Oxygen and water vapor contents, in particular, affect the extent to which the highly reactive metallic constituent oxidizes after decomposition of the hydride and hence the ease of wetting by the molten ller. A widely used method for metallizing oxide ceramics, in particular, alumina, is the so-called moly-manganese process. It has the advantages
of being relatively straightforward to perform, highly reproducible, not requiring a vacuum environment, and thus amenable for processing large parts. In this process, a slurry of powders of molybdenum, manganese, and various glassforming compounds is applied as a paint to alumina components. Then, the coated ceramic is red in a wet hydrogen atmosphere (with a wellcontrolled dewpoint) at a temperature of 1450 to 1500 C (2640 to 2730 F). The ring operation results in chemical reaction between the glassy phase in the alumina and the manganese, while the molybdenum is established as a surface layer on the ceramic [Mattox and Smith 1985]. Because molybdenum is not readily solderable, especially when the process is limited to mild uxes, it is common to complete this metallization scheme with a coating of electroless nickel, followed by gold. An important point to be aware of is that metallizations are often put down in a stressed condition. This is an inherent feature of most deposition processes and can be a source of critical weakness. An example of the manifestation of excessive stress in a metallization layer is shown in Fig. 4.2. Here, a silver layer 20 m (800 in.) thick has cracked away from a ceramic material through such stress. The stress concentration at the interface between the component and the metallization can be controlled by limiting the thickness of the metallization layers, modifying the deposition parameters, and using conventional, postdeposition, stress-relief heat treatments. Particularly in microelectronics and photonics manufacturing, there is an ongoing trend toward product miniaturization. This is demanding the
A porous ceramic material metallized with a thick silver electroplate. The residual stress in the metallization has resulted in a peel failure through the near-surface layer of the ceramic.
Fig. 4.2
development of ever-smaller joint geometries and hence the volume of solder in each joint. At the same time, pricing pressures mean that any applied metallizations must be as thin as possible, because their cost is usually a direct function of machine time. In theory, a metallization need only be sufficiently thick that it ensures complete coverage of the faying surface and, if it is a barrier metal, will not be eroded back to the underlying material by reaction with the solder during the joining process or in service. While it is possible to give general rules of thumb regarding metallization thicknesses, attempts are now being made to derive more rigorous values by calculation. One such approach uses a combined thermodynamic and diffusion-kinetic model, and it is sufficiently broad in its scope to be able to deal with real systems [Ronka, Van Loo, and Kivilahti 1998]. No doubt, further advances will be made in this area over the next few years. 4.1.2.2 Active Solders
Reactive ller metals are mostly brazes, and further details of their design and function can be found in the planned companion volume Principles of Brazing. The basis of the approach is to incorporate into the ller metal small quantities of elements that are highly reactive. Provided that at least one of the products of reaction with the base material is metallic and remains as a layer on the surface of the nonmetal, then the ller alloy can wet and form sound joints. When selecting a reactive ller alloy, the choice of active ingredient, its optimal concentration, and the appropriate processing conditions need to be considered in relation to the nonmetal of the component. One of the most commonly used active constituents is titanium. Less reactive elements, such as chromium, and more reactive elements, such as hafnium, are also used. The addition of metals such as titanium to solders will enable them to directly wet nonmetals such as ceramics, provided the process temperature is sufficiently high to provide the necessary activation for the chemical bond to form with the nonmetal. Typically, this activation takes effect above 750 C (1380 F), which somewhat negates the benet of the low processing temperature characteristic of solders [Xian and Si 1992]. The process temperature is governed by the reactivity of the active ingredient, which increases with temperature. If the
temperature margin to the liquidus temperature of the solder is substantial, the heating cycle should be kept short in order to prevent extensive erosion of the joint surfaces or the growth of thick interfacial phases. It has recently been reported that melding small quantities (0.5 to 2%) of lutetium into solders has a similar effect but enables a reduction in the process temperature to more normal values. Rare earth elements have poor solubility in solder and therefore tend to agglomerate in small islands within the matrix, where they are protected from oxidation by the surrounding metal. However, in the molten state, sufficient rare earth metal is conveyed to the joint surface, where it forms a chemical bond with the nonmetal. The interface layer is 1 to 5 nm (0.04 to 0.2 in.) thick and appears to comprise lutetium oxide (Lu2O3) and lutetium-rich solder. By this means, tensile joint strengths in the region of 800 MPa (120,000 psi) and shear strengths of the order of 10 MPa (1500 psi) can be achieved to silica-based materials [Ramirez, Mavoori, and Jin 2002]. Further information on solders containing rare earth elements can be found in Chapter 5, section 5.8. Hitherto, active llers were found only to be effective above threshold temperatures of approximately 750 C (1380 F), and therefore, they have only been used for brazing. Recently, a route for substantially reducing the temperature threshold has been identied, making active solders a reality. This involves the addition of low concentrations of gallium and lanthanide elements to a silver-tin eutectic alloy containing 4% Ti [Smith 1998]. Although the exact role of the minor additions is unclear, they appear to fulll two functions. First, they modify the nature of the surface oxide that limits its growth. Second, the minor additions appear to increase the activity of the titanium, which enables the solder alloy to wet and bond directly to nonmetals such as alumina, aluminum nitride, beryllia, silicon carbide, silicon nitride, boron nitride, carbon/ carbon composites, aluminum/silicon carbide composites, copper, silicon, diamond, titanium and stainless steel, to name but a few, using a peak process temperature of only 250 C (480 F). The titanium and other additions increase the strength of the bulk solder and, interestingly, confer it with signicantly enhanced resistance to corrosion from aqueous media containing chlorides. The activated solder can be used in air without ux, but in order to facilitate wetting, it is necessary to use some form of mechanical activation, such as scrubbing. This locally rup-
tures the oxide skin and enables wetting and spreading to proceed from points of contact between the nonmetal and molten solder.
4.1.3
When a molten ller wets the surface of a parent material, alloying occurs, leading to a degree of dissolution of the parent material (or metallization), commencing at the joint interface. Dissolution of the parent materials occurs because the materials system encompassing the joint is not in thermodynamic equilibrium. This provides the driving force for wetting and spreading. The maximum solubility of the parent materials in the molten ller can be predicted by reference to the appropriate phase diagram, as stated in Chapter 2, section 2.3. In summary, extensive erosion is likely where the liquidus surface on the phase diagram between the llermetal composition and that of the parent material has a shallow slope and where the alloying depresses the melting point of the ller in the joint. If the phase diagram exhibits either of these features, then it is only possible to limit erosion by lowering the process temperature, shortening the heating cycle, and/or restricting the volume of molten ller metal. Such changes must obviously not compromise the integrity of the joints by, for example, reducing the effective uidity of the molten ller alloy, which in turn will impede wetting, spreading, and joint lling. Erosion can be reduced somewhat if intermetallic phases form along the joint interface so as to attenuate the rate at which solid material is transported from the components into the molten ller, or vice versa. This approach is only successful if the intermetallic phases formed are reasonably ductile, as exemplied by the case of tin-base solders used in conjunction with silver metallizations that is described in Chapter 2, section 2.1.2; otherwise, joint embrittlement results. In other cases, it is possible to protect the components against erosion by interposing a metalTable 4.3 solders
lization that will act as a barrier. An example is the application of a layer of nickel, typically 2 to 5 m (80 to 200 in.) thick, on aluminum components to prevent reaction with lead-tin solder. Although nickel-tin compounds form at the common interface, in normal soldering heating cycles, the intermetallic layer is quite thin and therefore does not adversely affect the mechanical properties of the joints. The extent to which dissolution occurs in a given heating cycle will also depend on the kinetics of reaction. The rate of dissolution of common engineering materials and metallizations in eutectic lead-tin solder is given in Chapter 2, Fig. 2.53. These data show that decreasing the process cycle time and temperature can be used to reduce erosion, although not always to acceptable levels in all cases. The propensity for dissolution (erosion) of the parent metal by a ller alloy can be reduced by preloading the ller with this metal so that it is already saturated and, on becoming molten, will not dissolve any further quantity. This approach is used for soldering to gallium arsenide devices using the Au-20Sn solder modied by the addition of 3.4% Ga. The addition reduces the solubility of gallium in the solder by half, with consequential benets to the process yield [Humpston and Jacobson 1989]. Metallurgical reaction between a solder and the materials of the components being joined (or their metallizations) can substantially alter the characteristics of the ller, which may in turn have unexpected consequences for the resulting joints. By anticipating the effects of these reactions, they either can be exploited to advantage or at least forestall embarrassing problems. Two examples are described as follows. Changes to the Melting Point of Soft Solders. Gold is widely used as the surface layer in metallization schemes because of its chemical inertness and ease of visual inspection. The change in the melting point (solidus temperature) of selected binary tin-base solders, when these have dissolved gold, is given in Table 4.3. The effect
The effect of gold on the solidus temperature of common, binary tin-base eutectic
wt% gold to form AuSn4 as the primary phase wt% gold that will dissolve in solder at 50 C superheat Change in melting point of solder by dissolving 1% Au, C (F)
Solder composition
<1 4 13 30 30
is particularly marked in the case of the Ag97.5Pb-1Sn solder, where the solidus temperature is reduced from 309 to 217 C (588 to 423 F) when 4% Au is added. For a layer of the solder 50 m (2 mil) thick, this radical depression of the melting point corresponds to a mere 1 m (40 in.) of gold coating being dissolved from each side of the joint. An example of a change in the contrary direction is the dissolution of gold in the In-18Pb-70Sn solder (melting range 136 to 182 C, or 277 to 360 F), where a 5% Au addition raises the solidus temperature by 14 C (57 F). Partial isopleths through these quaternary systems as a function of gold concentration are given in Fig. 4.3 and 4.4. Changes to Rate of Erosion of Parent Materials. Minor changes to the composition of the ller alloy are capable of inuencing the rate at which dissolution occurs. This is exemplied by the effect of small additions of silver, typically 2%, on the rate of dissolution of silver by lead-tin solders, as can be seen from Chapter 2, Fig. 2.3. The extent of the reduction could not be predicted from the Ag-Pb-Sn phase diagram and was determined empirically. Changes of this sort do not signicantly alter other properties of the ller, or those of the resulting joints, and can be readily implemented.
properties of joints. This is particularly true where the compound has low fracture toughness and forms as a continuous interfacial layer between the component surfaces and the ller alloy. It is sometimes possible to restrict the coarsening of these phases by carrying out the joining operation under conditions that are unfavorable for their growthnamely, restricting the heating cycle duration and peak temperature. In some
4.1.4
Phase Formation
Alloying between a molten solder and parent materials more often than not results in the formation of intermetallic compounds. This is particularly more true of solders, because the major constituents of most solders are elements with low crystallographic symmetry, and hence, they do not readily form solid solutions with engineering materials that tend to be based on simple bodycentered cubic (bcc), face-centered cubic (fcc), or hexagonal close-packed (hcp) crystal structures. The distribution, morphology, and proportion of these phases will depend on several factors, as explained in Chapter 2, section 2.3. Because intermetallic compounds generally possess a higher elastic modulus (i.e., they are stiffer) than many ller alloys themselves, welldispersed intermetallic phases are often benecial to the stress-bearing capability of the ller metal. This effect is shown for gold additions to the silver-tin eutectic solder in Chapter 2, Fig. 2.52. By and large, agglomerations of intermetallic compounds are deleterious to the mechanical
Section through the Ag-Au-Pb-Sn quaternary system showing the effect on the melting range of adding gold to Ag-97.5Pb-1Sn solder. Adapted from Evans and Prince [1982]
Fig. 4.3
Section through the Au-In-Pb-Sn quaternary system showing the effect on the melting range of adding gold to In-18Pb-70Sn solder. (This composition is not a ternary eutectic, as is sometimes stated in older literature.)
Fig. 4.4
cases, minor additions can be made to the ller alloy that will break up agglomerations of intermetallic phases present in the joint. This is one of the reasons for some minor additions made to zinc-base solders formulated for soldering to copper, which is described in Chapter 2, section 2.1.3. Barrier metallizations applied to joint surfaces can be used to prevent alloying with the parent materials and the consequential formation of undesirable phases. Some ller alloys are themselves hard because they contain intermetallic phases, yet form joints of high strength with certain parent materials, provided that the intermetallic phases are nely divided within the joint microstructure. Gold-tin solders fall into this category, with rapid solidication of the solder being a prerequisite for robust joints. Further details are given in Chapter 2, section 2.1.4.
4.1.5
Filler-Metal Partitioning
Partitioning of ller metals is a fabrication method that is becoming more common. It is an approach that can be benecial if an alloy composition is relatively brittle when prepared via conventional casting of ingots, but selected combinations of the constituents are ductile. Some examples taken from the published literature are listed in Table 4.4. The objective is either one of attempting to improve the mechanical properties of the ller metal prior to melting or one of cost reduction through decreasing the number of stock alloys required [Mackay and Levine 1986]. It is usual to have the lower-melting-point combination in the cladding, with a higher-melting-point mixture of ingredients in the core. The lowmelting-point fraction then melts at the soldering temperature and wets the joint interfaces. The Au-20Sn solder, as produced in the form of wire or foil, is very costly, partly because of its precious metal content but mostly because of the limited reduction in thickness that can be achieved on successive passes during hot rolling. This means that the fabrication costs are high.
Table 4.4
Filler metal
Silver-tin (mpt, 221 C) Lead-antimony-tin (mr, 245280 C) Silver-lead-tin (mr, 303310 C) Gold-tin (mpt, 280 C)
mpt, melting point; mr, melting range
Sn Sb-Sn Ag-Sn Au
Ag Pb-Sb-Sn Pb-Sn Sn
Because gold and tin are both very ductile metals, composite foil can be prepared by cold rolling. The noble nature of gold is exploited, and, by placing it on the outside of the sandwich, it is then able to prevent oxidation of the underlying tin during storage and heating to the process temperature. A postfabrication heat treatment step is claimed to be benecial in consolidating the foil and initiating interdiffusion between the constituents prior to use [Tokuriki Hoten 1981]. An alternative approach is to modify the surface of a ductile foil of tin through electrodeposition of a layer of gold. In this case, it is the higher-melting-temperature constituent that constitutes the surface layer. If the tin foil is 25 m (1 mil) thick, and 12 m (0.5 mil) of gold plating is applied on both sides the calculated composition of the solder will be hypereutectic at 70 wt% Au, and 30 wt% Sn. From the phase diagram for the gold-tin system, it may be deduced that if this composition were prepared as a homogeneous alloy, it would have a melting range of 280 to 370 C (536 to 698 F). The melting behavior of this alloy has been characterized using differential scanning calorimetry. The results are given in Fig. 4.5 and 4.6. On initial heating (Fig. 4.5), there is an endothermic reaction as the solid-state alloy formed by the interdiffusion of gold and tin melts at the binary eutectic temperature of 217 C (423 F). This is followed by a tin-rich transition reaction at 252 C (486 F) before onset of the gold-rich eutectic transformation at 280 C (536 F). Thereafter, melting continues until all the solid is consumed, which occurs at approximately 380 C (716 F). Cooling of this sample back to 150 C (302 F) and repeating the thermal analysis cycle (Fig. 4.6) shows that the sample is now homogeneous, with a solidus temperature of 280 C (536 F) and a liquidus temperature of approximately 380 C (698 F). This result accords well with the predicted melting behavior deduced from the alloy phase diagram. Partitioning of solders can sometimes also be benecial where the maximum permissible process temperature is only marginally above the melting point of the ller metal. Phased reow soldering, as this approach is described, has been exploited in some of the lead-free solder paste formulations based on Bi-Cu-Sn alloys. These replacement solders for lead-tin eutectic have to function at very low superheats, because most printed circuit boards and components have rated
20.00
mcal/s
10.00
0.00 160.00 190.00 220.00 250.00 280.00 Temperature, C 310.00 340.00 370.00 400.00
Fig. 4.5
Fig. 4.6
maximum process temperatures of 235 C (455 F). Copper-tin eutectic melts at 227 C (441 F), and bismuth-tin at 139 C (282 F). By partitioning the ternary composition so that it includes some low-melting-point tin-bismuth solder, liquid phase sintering will commence on heating above 139 C (282 F). The presence of the liquid phase improves heat transfer to the unmelted particles and helps minimize temperature gradients through the solder paste, both of which improve the reliability of ller-metal joining processes performed at low homologous temperatures. The progressive melting of the solder is also reported to be benecial in reducing the incidence of tombstoning [Warwick 2002]. This is where one land on a double-ended surface mount component wets more effectively than at the other, and the imbalance of surface tension forces causes the component to lift up vertically on one land. Several solder manufacturers have led patents on phased reow soldering. There have also been attempts to effect soldering by liquid phase sintering, that is, to join metal components without melting the highermelting-point constituent of a partitioned llermetal paste. By this means, it has been found possible to produce joints with reasonable mechanical integrity but nowhere near as sound as that achieved by melting the solder [Palmer, Alexander, and Nguyen 1999].
differing thermal expansivities that develops on cooling from the freezing temperature of the ller metal can be approximated by the following equation [Timoshenko 1925]:
Stress E 1 E2 E1 E2 (X1 X2)(Tf Ts )
where E is the modulus of elasticity of materials 1 and 2, X is the coefficient of thermal expansion of 1 and 2, Tf is the freezing point (solidus temperature) of the ller alloy, and Ts is the temperature of the assembly corresponding to the stress. In the derivation of this equation, it is assumed that the materials are only deformed within their elastic limits and that the joint is innitely thin. Despite these simplifying assumptions and the inaccuracies that they introduce, this expression is useful in providing an indication of whether the stress due to thermal expansion mismatch is close to or exceeds the failure stress of either of the abutting materials, that is, whether failure of the assembly is likely to occur. Some worked examples are described by Haug, Schaefer, and Schamm [1989]. In an assembly with a planar joint between two elastic but different materials 1 and 2, the magnitude of the bow distortion in one dimension can be estimated from the physical properties of the materials using a simplied model [Timoshenko 1925]. With reference to Fig. 4.7.
Bow distortion, B (L2/ 8) / R Radius of curvature (A1 A2)[3(1 M)2 (1 MN)(M 2 1/MN)] 6(X1 X2)(Tf Ts)(1 M)2 (Eq 4.1)
4.2
In an assembly composed of heterogeneous materials, there is usually a thermal expansion mismatch between the abutting components. This manifests itself as stress on cooling from the solidus temperature of the ller metal and is a maximum at the lowest temperature that the assembly experiences. Materials with a relatively low elastic modulus can accommodate strain and will tend to deform under the inuence of this stress, while brittle materials, notably glasses and ceramics, have a tendency to fracture, particularly if the stress distribution places the component in tension. Even if a heterogeneous assembly survives the joining operation, the stresses arising from the thermal expansion mismatch can cause it to fail by fatigue during subsequent thermal cycling in service. The stress in the region of the joint between two isotropic materials, designated 1 and 2, with
where M A1/A2 and N E1/E2, B is bow distortion L is the length of the joint R is the radius of curvature A1 and A2 are the thicknesses of materials 1 and 2 X1 and X2 are the coefficients of thermal expansion of materials 1 and 2 E1 and E2 are the moduli of elasticity of materials 1 and 2 Tf is the freezing point (solidus temperature) of the ller alloy and Ts is the temperature of the assembly corresponding to the bow distortion
Equation 4.1 assumes that the joint in the heterogeneous assembly is innitely thin and totally inelastic. From Eq 4.1, it can be seen that it is possible to effect some reduction in expansion mismatch stress, that is, decreasing R, by applying one or more of the following measures:
Fig. 4.7
in opposition, and the optimal condition for reducing the distortion of bonded components is a compromise between the two. In the absence of other indications, a good starting point is to use a temperature that is approximately 75% of the melting point of the ller metal, expressed in degrees Kelvin. Some further improvement in the residualstress level can be obtained by using a more compliant ller, especially when the joint is reasonably wide (>25 m, or 1 mil). Solders, almost without exception, have low elastic moduli, which is related to their low melting points, but there are only minor differences between individual alloys. Hence, the only possibility for obtaining a joint with improved compliance, short of redesigning the assembly, is to replace a hard solder with one that will creep. The benet of such a change might be offset by inferior dimensional tolerancing, and the joint design and operating environment should take this trade-off into consideration. The thermomechanical properties of most of the family of indium solder alloys (pure indium, silver-indium, indium-lead, indium-tin) are dominated by creep behavior at all normally experienced temperatures and strain rates. Thus, a soldered joint made with one of these alloys will always creep when subject to stress. The fact that they creep so readily, even at cryogenic temperatures, means that joints made with indium solders are substantially more resistant to failure by fatigue than might otherwise be expected (see Chapter 2, section 2.1.6). Wide joint gaps (>500 m, or 20 mil) can sometimes be used to minimize the effects of expansion mismatch between two components. The solder must have high viscosity in order to ll such wide joints. This is achieved by either using a ller metal with a wide melting range and performing the joining process below the liquidus temperature, so that the alloy is not fully molten, or by mixing in metal powder with a higher melting point. Spacers are required to control the joint gap. Wide joints can also be achieved by inserting porous shims, as described in section 4.3.4 of this chapter. One particular merit of wide joints to ceramic components is that they are tolerant to variations in the width of the joint gap, thereby obviating the need to closely machine the mating surfaces of the components, which tends to be costly and can weaken the material by creating subsurface cracks. Where the joint is wide, its mechanical properties are essentially those of the bulk solder (see section 4.3.3 of this chapter).
A variety of mechanical schemes are available to assist in overcoming the problem of thermal expansion mismatch. Several approaches that have proved successful are described follows.
4.2.1
Fabrication of most products requires the use of many different materials in their assembly, each selected for a particular property or combination of properties it offers. Ashby and coworkers have devised a scheme of materials selection charts that pictorially represent materials according to their properties, which is intended to facilitate selection [Ashby 1994]. Because silicon and most other semiconductor and optical materials, as well as engineering ceramics, have
low thermal expansion coefficients, compared to metals, there is always interest in controlled expansion alloys that can be used to bridge dimensional changes between these materials and metal components in the same assembly (Fig. 4.8). Single-phase materials, which include many engineering ceramics, such as alumina and aluminum nitride, change dimension in a fairly linear manner with temperature. This is certainly true over the temperature range of interest for consumer products (50 to 150 C). For common metals, their coefficients of thermal expansivity (CTE) at room temperature are directly proportional to their melting points. The expansion effect is attributable to the atomic vibrations of the crystal lattice. Raising the temperature increases the vibration, which means that each
Fig. 4.8
An Ashby materials selection chart. The linear expansion coefcient, , plotted against the thermal conductivity, . The contours show the thermal distortion parameter /.
atom occupies a greater space, and hence, the material grows in size. Because the maximum vibration is restricted by a material melting temperature, a low-melting-point material has a smaller expansion temperature range and will exhibit a higher CTE value than a metal with a relatively higher melting point, as demonstrated in Fig. 4.9, using the data in Table 4.5. The principal low-expansion metals are tungsten and molybdenum, which have CTE values at 20 C of 4.5 and 5.1 106/K, respectively. These metals, especially tungsten, are hard and stiff, offering little compliance, and they are also relatively difficult to machine. Both are dense materials, so components fabricated from them will be heavy. Their high melting point means that their atoms are strongly bonded, and, consequently, their machining requires considerable energy. Consequently, the net cost of parts fabricated from these metals can be signicant. Titanium has a low CTE (5.6 106/K) for a metal, low density, but a poor thermal conductivity of only 15 W/m K. This rules it out for any application where substantial heat transfer by conduction is an additional requirement. Alumina and aluminum nitride also suffer from energy-intensive production methods and difficulty of machining. These ceramics have to be fabricated in near-net shape forms, so the tooling cost of new components is quite high. Furthermore, their thermal conductivity is inferior to that of most metals. The lack of single-phase materials that offer the combination of low thermal expansivity, high thermal conductivity, and, preferably, low density has led to the development of several families of multiphase materials with tailored properties. These include iron-nickel alloys, copper-tungsten and copper-molybdenum alloys, metal-metal laminates, metal-ceramic laminates, metal-matrix composites, and metal-metalloid alloys (in particular, a family of aluminum-silicon alloys).
These materials and their key physical properties are listed in Table 4.6. Further information is given in the following section. 4.2.1.1 Iron-Nickel Alloys
General relationship between coefcient of thermal expansion, or CTE (between 273 and 373 K), and melting point for metals, Tm. Adapted from Li and Krsulich [1996]
Fig. 4.9
Most readers will probably know iron-nickel alloys by trade names that include Invar, Kovar, Nilo-K, Nilo Alloy 42, or by the UNS number K94610 or DIN WNr 1.3981. Kovar and Nilo-K, for example, are essentially the same Co-Fe-Ni alloy, of approximate composition 17Co-54Fe29Ni, that has a CTE at 20 C (68 F) of 5.8 106/K. These alloys are readily available from a number of manufacturers in many shapes and forms and are competitively priced. Iron-nickel alloys are widely used in electronic packaging. Being fairly soft and ductile in the annealed condition, they are also used in shims for joining low-expansion ceramics to higher-CTE metals, where they are capable of distributing and absorbing expansion mismatch stresses. Iron-nickel alloys offer abnormally low thermal expansion compared to their constituents. Indeed, over a limited range of temperature, it is possible to design material that has zero expansivity. The unusual expansion characteristics of these alloys can be ascribed to the fact that they are ferromagnetic. At temperatures above the Curie point (approximately 450 C, or 840 F), these alloys are ferromagnetic and exhibit normal thermal expansion characteristics. Below the Curie temperature, when they are in their ferromagnetic domain, the actual expansion is the sum of the normal positive expansion due to lattice vibration, counteracted by the negative expansion due to the ferromagnetism. The latter is termed magnetostriction. The result is reduced CTE below the Curie temperature that can be adjusted by controlling the state of cold work in the alloy. At very low temperatures, below 100 C (148 F), the expansion coefficient of iron-nickel alloys reverts to more normal values (15 106/K). Expansion curves for some iron-nickel alloys are given in Fig. 4.10 and 4.11. It is important to recognize that iron-nickel alloys only possess low and controlled expansion coefficients over a limited range of temperature. This is due to a combination of the limited temperature interval over which magnetostriction compensates for normal thermal expansion, plus the necessity for the alloy to be in a very particular state of cold work. Thus, further processing that involves either
mechanical working or a temperature excursion will change the internal stress in the material and thereby its expansion coefficient. Figure 4.12 illustrates the expansion coefficients of Fe-36Ni and Fe-42Ni alloys as a function of temperature and state of anneal. As can be seen, the lower-nickel-content alloys can be fabricated to have a smaller overall thermal expansion, but it is stable over a more limited range of temperature. These low-expansion alloys are notoriously difficult to machine, particularly in thin sections to a high surface nish. Specialist metal working companies have developed the necessary skills to deliver products of consistent and high quality. It is not possible to solder directly to iron-nickel alloys using electronicgrade uxes, and parts are normally plated with nickel and gold. Brittle intermetallic compounds form at the interface when cobaltTable 4.5
Metal
containing mixtures are soldered directly with tin-base solders. 4.2.1.2 Copper-Molybdenum and Copper-Tungsten Alloys
Low-expansivity molybdenum (CTE 5.1 106/K at 20 C, or 68 F) or tungsten (CTE 4.5 106/K at 20 C, or 68 F) is added to copper (CTE 17.6 106/K at 20 C, or 68 F) to produce controlled-expansion alloys. Molten copper is virtually insoluble in both molybdenum and tungsten, and various techniques have been devised to enable the manufacture of 100% dense alloys of these materials. These techniques include powder metallurgy as well as liquid inltration casting. Controlled-expansion alloys can be produced with a continuum of properties that range from essentially pure copper to pure refractory metal. By this means, families of ma-
Table 4.6 Indicative physical properties for selected semiconductor and low-expansion materials at 20 C (68 F). Exact values depend on the composition of the material, method of manufacture, test method, and test conditions. Reference should be made to suppliers data sheets for precise values.
Material Through-thickness thermal conductivity, W/m K In-plane thermal expansion coefficient, 106/K Density, g/cm3
Gallium arsenide Silicon Alumina Aluminum nitride Beryllia Molybdenum Titanium Tungsten Copper-alumina-copper Copper-molybdenum-copper Copper-85% tungsten alloy Copper-85% molybdenum Invar (Fe-36Ni alloy) Kovar (Fe-29Ni-17Co alloy) Aluminum-50% silicon alloy(a) Aluminum-70% silicon alloy(a) Aluminum-68% silicon carbide composite Beryllium-30% beryllia composite (E20)(b) Beryllium-51% beryllia composite (E40)(b)
42 84 20 165 260 140 15 174 26 166 180 160 14 17 150 120 150 210 220
6.5 2.5 6.7 4.5 7.2 5.1 5.6 4.5 7.3 5.5 7.2 6.7 2.2 5.8 11.0 7.4 7.2 8.7 7.5
5.3 2.3 3.9 3.3 2.9 10.2 4.5 19.3 4.1 10.0 16.1 10.0 8.1 8.4 2.5 2.4 3.0 2.1 2.3
(a) As supplied by Osprey Metals Ltd. (b) As supplied by Brush Wellman Inc.
terials have been developed having a range of controlled CTE values. A common ratio is 15%
18 16 Coefficient of expansion, 106/K 14 12 10 8 6 4 2 0
1.8 1.6 1.4 1.2 1.0 34 35 36 37
10
20
30
40
50
60
70
80
90
100
Cu, which gives a reasonable boost to the thermal conductivity without greatly impairing the thermal expansivity of the base metal. More importantly, perhaps, is that the addition of a soft copper phase in the otherwise refractory metal matrix greatly improves the machinability. These alloys still require considerable care to machine, because the copper is much softer than the refractory metal constituent, and it is easy to cause damage in the surface region. The principal drawback of these alloys is their high density, although this is partly offset by their high modulus, which means that thinner sections can sometimes be used, depending on the functional requirements and design of the product. When used in precision assemblies, these alloys need to be in the annealed condition to ensure dimensional stability on thermal cycling. 4.2.1.3 Copper-Surface Laminates
Nickel, %
Fig. 4.10
nealed state
Fig. 4.11
Copper can be attached directly to alumina via the copper/copper oxide eutectic reaction. These products are often marketed as direct-bonded copper. The copper can be patterned so that it can also fulll the function of, admittedly, a very low-density printed circuit board, and this has made it a very popular substrate for electronic power modules. A more recent development of this material has been its commercialization in the production of arrays of through-thickness vias, which are lled with copper. This arrangement does not change the in-plane thermal expansivity of the alumina but signicantly improves the average through-thickness thermal conductivity. Obviously, care must be taken when using this approach with high-intensity heat sources, typied by optoelectronic and microwave devices, to ensure that the pattern of heat studs matches the point sources in the semiconductors.
Fig. 4.12
Thermal expansion characteristics of Fe-36Ni and Fe-42Ni alloys as a function of temperature and state of anneal
Copper-molybdenum-copper and copperInvar-copper laminates are also available but provide a subtly different balance of properties. They do not offer the same stiffness and patterning abilities as the ceramic-cored alternatives and are less widely used. 4.2.1.4 Composite Materials
Some examples of composite material are included in Table 4.6. Each is representative of a different family: metal-metalloid, metal-ceramic, and metalloid-metalloid. These are all relatively new materials, and many variations exist on the market. Metalloid-metalloid composites are very immature products, and citing reliable property values from an assessment of published values has proved problematic. Controlled-expansion materials comprise two or more components: one, a metal or metalloid, and the other a lower-expansion metal, metalloid, or nonmetal. By varying the relative proportions of these constituents, the CTE values can be adjusted over quite a wide range. Alloys of copper-tungsten and copper-molybdenum, made by solid-state sintering together of powders of the separate constituents and referred to previously, represent controlled-expansion composites of metal components. Metal-ceramic composites are typied by aluminum/silicon carbide and beryllium-beryllia. These are of lower density, but their principal limitation stems from an inability for machining at high speed and obtaining an acceptable surface nish, because of the substantially different hardness of the constituents. Diamond tools need to be used on aluminum/silicon carbide because of the presence of extremely hard silicon carbide particles. After all, silicon carbide is itself a traditional cutting tool material, and silicon carbide in a matrix of aluminum constitutes a cemented carbide. In the case of the Be-BeO materials, the particles of BeO are poorly bonded to the beryllium metal matrix, and machining tends to result in surfaceopening cavities. There is also a potential health hazard should beryllia dust be generated. For these reasons, parts fabricated in these metal-ceramic composites tend to be produced using near-net shaping methods. Care is taken during the production to ensure that the part has an outer skin of metal and that the nishing machining does not expose ceramic particles. If it does, then there is a high chance that the particles will either crack or pull out, and if neither of these occur, then the presence of an exposed
ceramic phase tends to cause local adhesion problems if the part needs to be metallized. Metalloid-metalloid composites are represented by carbon-carbon ber composites, which were originally developed for exceptionally hightemperature applications such as aircraft brakes and rocket motor nozzles. They can be metallized and also impregnated with copper. This is desirable because it facilitates soldering and allows the in-plane thermal expansivity to be increased to more normal values. Care needs to be taken when selecting the carbon bers and the weave in the component. Carbon ber is a highly isotropic material and hence, the properties of parts fabricated by using it can vary accordingly. Likewise, there are many grades of carbon ber, and, as one might expect, it is the more expensive ones that possess the most desirable and stable properties. Carbon-ber-reinforced aluminum and copper composites are also available, but these are highly anisotropic and tend to be dimensionally unstable. Metal-metalloid composites are perhaps best represented by the series of silicon-aluminum alloys, marketed as controlled-expansion (CE) alloys and containing between 27 and 70 wt% Si, to achieve a range of CTEs, as can be seen in Fig. 4.13 [Jacobson 2000]. These are strictly alloys, not composites, produced from the melt by the Osprey spray-forming process. A relatively ne, two-phase, or duplex, microstructure of continuous, interwoven matrices of silicon and aluminum forms naturally by eutectic solidication during the spray-forming process (Fig. 4.14). The silicon phase constrains the expansion of the aluminum, while the latter fraction is largely responsible for the transport of heat. Both silicon and aluminum are elements of low density, with silicon being lighter than aluminum. The CE7 alloy (Al-70Si) is closely expansion matched to
Coefcient of thermal expansion (CTE) of Osprey controlled-expansion alloys (based on aluminumsilicon) as a function of the proportion of silicon, in weight percent
Fig. 4.13
alumina over a wide range of temperatures, making it suitable for packaging microwave/radio frequency circuitry.
4.2.2
Interlayers
One route toward reducing the mismatch stress concentration that develops in soldered and, in particular, in brazed assemblies involves a redesign of the joint to accommodate one or more interlayers. There are two basic congurations that are described in the literature. In the rst approach, a compliant interlayer is inserted that will yield when the joint is placed under stress and thereby reduce the forces acting on the components. This approach is not particularly effective with soldered joints for the following reasons. First, the moduli of compliant metals that are most effective in accommodating stress (in particular, silver and copper) are too close to those of many solders to provide much stress relief, so the solder will tend to yield in preference to the interlayer. Second, solders tend to form hard, interfacial phases with most engineering metals and alloys, which will confer a high modulus to the adjacent interlayer and actually exacerbate the situation. The approach is therefore usually only relevant where hard gold-tin solder is mandated. Filler metals with intrinsically low and controlled thermal expan-
sion coefficients are discussed in section 4.3.3.2 of this chapter. An alternative approach for reducing mismatch stress concentrations is to redistribute the stresses across a much wider zone, so that the stress is within tolerable levels everywhere in the assembly. A graduated redistribution of stress may be accomplished by inserting into the joint one or more thick shims or plates that have CTEs that are intermediate between those of the abutting components. The plates must be sufficiently thick so that they are not signicantly distorted by the imposed stresses and therefore are not usually less than a few hundred microns (several mils) thick. An assembly containing a single plate with an intermediate thermal expansivity is shown in Fig. 4.15. This approach is particularly suitable where there is a need to join metals to ceramics and other ceramic-like nonmetals. If the intermediate plate is selected to have a thermal expansion coefficient that is close to that of the nonmetal, then it is possible to transfer the major proportion of the stress to the more robust metallic part of the assembly. Where the two components have greatly different thermal expansivities, it may be necessary to use a graduated series of plates to reduce the mismatch stresses in each joint to an acceptable level. A monolithic plate of graded composition and thermal expansivity can be used in place of a series of discrete homogeneous plates. Typi-
Photographs of the microstructure (micrographs) of two controlled-expansion (CE) alloys produced by spray forming, showing their uniform phase distribution. The lighter and darker phases are primary aluminum and silicon, respectively. (a) CE17 (Al-27wt%Si. (b) CE7 (Al-70wt%Si). Courtesy of Osprey Metals
Fig. 4.14
cally, these may be prepared from powder compacts. Copper/tungsten components of this type are made by inltrating a loose compact of tungsten powder with molten copper. By adjusting the packing density of the powder, the relative proportions of copper and tungsten will vary, and the properties of the component can vary from those of essentially pure copper to approximately 95% W. This enables one side of the component to be made tungsten-rich, with a low expansion coefficient, and the other side copper-rich, with a much higher expansion coefficient. Because there are no abrupt interfaces in such a component, it can survive thermal cycling over wide ranges of temperature almost indenitely without suffering distortion through creep or fatigue fracture. A graded copper-tungsten plate is shown in Fig. 4.16.
If one or both of the components is highly brittle and vulnerable to fracture under a tensile or shear stress, it is often the practice to provide reinforcement by attaching it to a more mechanically robust metal plate of similar expansivity. This subassembly can then be joined via the reinforcing plate to further components of different expansivities, using approaches detailed previously. Such a conguration involving a reinforced subassembly is used for mounting of semiconductor edge-emitting lasers, because the wavelength of the light produced is a strong function of the strain in the semiconductor material. The following disadvantages are associated with the use of graduated joint structures based on the use of intermediate plates to accommodate mismatch stresses:
Use of a plate of intermediate thermal expansivity to reduce the stress due to thermal expansion mismatch in an assembly between an aluminum alloy mount and the body of a solid-state laser
Fig. 4.15
4.2.3
Compliant Structures
Monolithic plates of graded composition, varying from essentially pure copper to approximately 95% W. The thermal expansivities of the two surfaces differ by approximately 16 106/C (29 106/F).
Fig. 4.16
Equation 4.1, given previously for calculating the bow distortion of a bimetallic assembly, implies that the mismatch stress is a sensitive function of joint dimension or, more precisely, joint area. Although the overall size of the assembly is likely to be xed by the functional requirements of the product, it may be possible to replace one of the monolithic components with a lamentary, brushlike structure. Then, the dimensions of each individual bimetallic joint can be made as small as necessary, thereby effectively eliminating the mismatch stress from this
source, while the high aspect ratio of the laments confers a degree of lateral compliance that can accommodate the mismatch strain. Examples of these highly compliant structures are illustrated in Fig. 4.17 and 4.18, and others are described in the scientic and technical literature [Huchisuka 1986]. The so-called ip-chip bonding process used in semiconductor assembly also results in compliant contacts between the components, although in this case these are provided by the actual solder joints [Yung and Turlik 1991]. Flip-
chip bonding involves electroplating or vapor depositing solder bumps on the contact pads of an electronic component. These bumps are heated to reform the solder as hemispherical balls, which are typically 0.1 mm (0.004 in.) high. In a subsequent stage, the components are joined to circuit boards by reowing the solder bumps. The entire sequence is shown schematically in Fig. 4.19. The substrate is prepared in such a manner that solder wetting is laterally conned, in order to maximize the height of the solder pillars that constitute the joints. The Pb-5Sn solder is widely
Fig. 4.17
Examples of compliant structures for mitigating mismatch expansivity () of the abutting components
Fig. 4.18
(a) Longitudinal and (b) transverse sections through a compliant structure that is capable of accommodating a thermal expansivity difference between joined components. (a) 99. (b) 450
used for ip-chip bonding because it has a high compliance, although its fatigue resistance is relatively low for a metal. The ip-chip process is described in detail in Chapter 5, section 5.2. The use of compliant structures of the forms shown obviously incurs a cost penalty, due to the greater complexity of manufacture. The conductance between the components via the lamentary member will also be impaired. Even with laments having a hexagonal cross section to produce a close-packed structure, it is difficult to obtain a compliant structure that will work effectively with a packing density of greater than approximately 85% [Glascock and Webster 1983]. Furthermore, the ability to simultaneously make large numbers of small-area joints is by no means a trivial exercise but one that demands stringent control of tolerances and highly specied joining processes.
strengths by as much as an order of magnitude, the value depending on the geometry of the joint and the mode of stressing. This is illustrated for resistance to peel initiation in Fig. 4.20 to 4.22. The improvement in mechanical properties may be attributed to the gradual transition in geometry that the llet provides in minimizing stress concentration at the joint periphery and thereby joint failure through crack and peel initiation at the surface. The stress concentration can be calculated and is presented as a function of contact
4.2.4
Wherever practicable, it is good practice to design a joint so as to encourage the formation of a llet. Then, even if the joint contains voids, for whatever reason, the llet will serve to seal the joint because there is a higher probability that where llet formation is promoted, these tend to be continuous and void-free. Fillets also have a benecial effect on the mechanical properties of joints. Well-formed llets of ller metal can enhance the measured tensile, shear, and peel
Typical peel force (P) prole of original geometry joints of a at-pack module on a circuit board. The peaks in peel strength are associated with the llets at each end of the joint.
Fig. 4.20
Fig. 4.19
Experimentally derived relationship between the stress required to initiate peel fracture and the height of the solder llet
Fig. 4.21
Fig. 4.22
Typical peel force (P) proles of joints modied by increasing the solder volume and providing for a llet at both ends. The joint strength is unchanged, but the resistance to peel initiation is greatly improved.
angle in Fig. 4.23 [Eley 1961]. Provided the solder wets to form a llet with a contact angle below 30, there is no appreciable stress concentration at a change in geometric prole. The subject of stress concentration is discussed in section 4.3.3 of this chapter. Because it is difficult to form llets of exactly reproducible geometry, testpieces for mechanical testing are often designed to exclude llets, either by preventing their formation through the use of nonwettable surfaces outside the edge of the joint or by removing any that happen to form. Although this practice makes the measurements more readily reproducible, it modies joint strengths to an extent whereby they may not be representative of most practical situations. Examination of edge llets can provide an indication of the ller/substrate contact angle at the onset of solidication. This is usually taken
to be a good indication of the overall quality of the joint. While this is generally true, visual inspection of edge llets can be misleading as to the quality of the interior of the joint, as discussed in Appendix A4.2 in this chapter. Information on internal integrity can only be obtained by radiography, scanning acoustic microscopy, transient thermography, or destructive methods.
4.3
A large-area soldered joint may be dened as one where the total joint area exceeds approximately 20 mm2 (0.3 in.2) and the length in any direction is greater than approximately 5 mm (0.2 in.). This denition is based on the following practical criteria:
Fig. 4.23
Role of llets in reducing stress concentration at the changes in section between abutting components
Distortion of assemblies can arise from a number of causes, some of which were discussed in Chapter 1, section 1.3.2. Apart from CTE mismatch, the most common sources of warping or bowing are uneven heating, which leads to temperature gradients in the components, and residual stress from earlier stages of fabrication, which is relieved in the heating cycle. Distortion from these causes can be avoided by performing the joining operations under carefully controlled conditions. Notably, heating should only be carried out in furnaces that provide highly uniform temperature zones, with the rate of heating tailored to the thermal mass of the components, and after appropriate stress-relief routines have been performed.
4.3.1.1
Trapped Gas
4.3.1
Joint Area
The strength per unit area of a joint between components of the same material tends to reduce in proportion to the area above some lower threshold, often approximately 20 mm2 (0.3 in.2) for a soldered joint. This effect can be seen in Fig. 4.24, which shows the percentage of the joint by plan area that comprises voids as a function of the (square) component size. There is clearly an increasing tendency for voids to accumulate in the joint as its area is increased, with a threshold joint length of approximately 1 mm (40 mil). The voids have two causes:
Gas trapped or generated within the joint Solidication shrinkage of the molten ller
metal Each of these is considered in further detail in the following sections.
By far, the largest source of voids in soldered joints is trapped gas, even for joints made in high vacuum using both components and ller alloys that have been given a vacuum bakeout prior to the joining cycle. Void levels in soldered joints greater than approximately 100 mm2 (1.5 in.2) in area and that are no narrower than 5 mm (0.2 in.) can typically reach 50% of the joint volume, which is consistent with the entrapment of air or an evolved gas. Figure 4.25 shows a radiograph of a joint between a silicon chip and a metallized ceramic package, made in high vacuum using a solder preform, that graphically demonstrates the problem. Figure 4.26 is a scanning acoustic image of the same component and conveniently illustrates the correspondence that can be obtained between the two analytical techniques. Air becomes trapped when the components are assembled, with the mating surfaces forming an effective seal. As the temperature of the assembly is raised to the peak process temperature, the trapped air is augmented by gas (usually moisture) evolved from the joint surfaces. The total gas volume will increase as the temperature is raised and the ambient pressure is reduced, in accordance with the gas law (pressure volume constant absolute temperature). If the path length between a gas bubble and the joint periphery is small, the gas pressure can normally exceed the hydrostatic force exerted by the molten ller metal, allowing the gas to escape
Void content versus joint length for a range of representative solders. The substrates were square coupons of polished alumina metallized with thin-lm titanium/ gold, applied by sputtering. The joints were made at a superheat of 25 C, and the void level was assessed by quantitative xradiography.
Fig. 4.24
Radiograph of a silicon chip (10 10 mm) soldered into a metallized ceramic package. Voids in the joint gap are evident as the light areas.
Fig. 4.25
to the surrounding atmosphere. However, the limit to the path length for this to occur is of the order of 1 mm (40 mil) for soldered joints (Fig. 4.24), but it is signicantly longer for brazed joints, because the process temperature, and hence the pressure developed by trapped gas or vapor, is higher. Adsorbed water is particularly detrimental in this regard, because it expands rapidly as it vaporizes. One method of allowing the trapped gas to escape is to momentarily split the joint apart during the reow process. While this has been demonstrated as being highly effective in laboratory trials, it is not readily applicable to volume manufacture [Xie, Chan, and Lai 1996].
Scanning acoustic microscope image of the soldered joint shown in Fig. 4.25. Voids in the joint gap correspond to the light areas.
Fig. 4.26
When undertaking soldering using preforms, a common mistake made when admitting the solder into the joint gap is to use a foil preform of similar dimensions to the plan area of the joint. This approach typically results in a high level of voids because of the large surface area exposed to the atmosphere. An effective method of removing trapped air is to design the joint in such a manner that the molten solder is made to ow from the center of the joint out toward the periphery or through the joint from one edge, as tends to occur when feeding the ller into the joint from a rod or a wire preform. Suitable arrangements for achieving this type of ow are illustrated schematically in Fig. 4.27. The advancing front of molten solder is then able to displace the vapor and air ahead of it as it ows into the joint gap. However, neither approach is entirely satisfactory. A preform of increased thickness and reduced area should not exceed 2 mm (80 mil) in plan, which can make jigging of the components difficult. Moreover, the solution of introducing the solder from one side of the joint is only effective with ller alloys that do not react strongly with the substrate materials to stie ow of the molten alloy in its path through the joint (see section 4.1 of this chapter). There are a number of solutions to this problem. A rst modication is to introduce the solder from more than one side. The preferred conguration is to place two preform discs at either side of the joint, as illustrated in Fig. 4.28. Disc preforms are readily available in a variety of sizes, and, by using a pair, the necessity to have a highly uid solder is diminished. The preforms can be placed so that the voids that form will be
Fig. 4.27
Two congurations showing ow by a molten ller designed to sweep trapped gas out of a joint
in prescribed locations and their presence can be allowed for in the design of the assembly [Weston 1974]. An improvement on this technique uses a single foil preform in the shape of a cross, with the arms orientated along the longest diagonals of the component. By doing so, the jigging requirements are greatly simplied [Socolowski 1987]. However, this method requires a custom-made preform for each application, and foil preforms are difficult to clean mechanically (to remove the native oxide) by mechanical abrasion prior to use (see Chapter 3, section 3.3.5). Admitting the solder in the form of round wire preform overcomes three of the fundamental deciencies associated with using foil preforms:
The surface-area-to-volume ratio is considerably reduced. Correspondingly, the detrimental effects of surface oxide on the surface of the solder are greatly diminished. Furthermore, the solder oxide can be easily removed by wiping the wire several times with a paper tissue soaked in solvent (see Chapter 1, Fig. 1.27 and Chapter 3, section 3.3.5). If this operation is conducted immediately prior to the heating cycle, then voids stemming from solder oxide can be effectively eliminated. A round wire has only one small area of contact with each of the faying surfaces. This precludes gas pockets being trapped during jigging, and the advancing solder front is always in the optimal location to sweep out air from the joint gap. As a side benet, solder wire is readily available at a low price premium over the metal content, compared to foil preforms, and a few stock sizes can be used for a wide variety of joining applications.
Original preform
Solder flow
Fig. 4.28
Dual discs of preform used to reduce the incidence of voiding in large-area joints
Schematic illustration of the outward ow by a central cross of ller metal, a conguration that helps to prevent entrapment of vapor in pockets in a large-area joint
Fig. 4.29
Substrates with solder already applied in dened areas are now available commercially from a number of suppliers. These have the advantages of:
present), the solder, or the ux contain constituents that volatilize during the heating cycle. A vacuum bakeout immediately prior to the joining cycle can help prevent subsequent outgassing from porous materials. Gas evolution from polymeric materials is usually caused by thermal decomposition. The only practical remedy in this instance is to use either a lower-temperature process or to change the polymeric material to one that has superior thermal stability. Some materials used for printed circuit boards are less stable than others, often in relation to their cost. Similar considerations apply to metallic components and ller metals, where these contain volatile elements such as zinc, magnesium, cadmium, and, to a lesser extent, manganese, and also to uxes and pastes, which generally contain volatile constituents. One common method of introducing a solder into a joint gap is in the form of a uxed paste. If not correctly formulated and strict process controls observed, joints made with uxed paste tend to contain a higher proportion of voids than joints made with a uxed preform. This relationship has been studied. The voids are primarily caused by the volatilization of entrapped ux. The propensity of voids to form increases with decreasing wettability of the faying surfaces, decreasing
For some large components, it may be possible to incorporate vents through the components to provide a passage for trapped gas to escape from within the joint. By careful design, this vent can be further exploited to increase the effectiveness of the pressure variation process described subsequently in section 4.3.2 of this chapter [Humpston et al. 2001]. Evolved vapor or gas can originate from several sources, in particular:
Example of a large area, 10 10 mm (0.4. 0.4 in.), made uxless, at 10 C (18 F) superheat, using In-15Pb-5Ag solder introduced in the form of a wire cross, shown in Fig. 3.30. The joint ll is revealed by x-radiography. A line of residual voids marks the location of the original wire cross that the surface oxide regrew in the interval between cleaning and melting. Courtesy of BAE SYSTEMS
Fig. 4.30
ux activity, increasing ux volume, and increasing plan area of the joint. The extent of voiding is also inuenced by the paste design, in that the sooner coalescence of the metal occurs, relative to the ability of the ux to deal with the oxide present on the faying surfaces, then the higher will be the resulting void content. Likewise, the use of ux media with high boiling points relative to the coalescence temperature of the solder paste tends to exacerbate the formation of voids [Hance and Lee 1992]. 4.3.1.2 Solidication Shrinkage
Table 4.7 Solidication shrinkage of selected elements common to widely used solders
Volume contraction on solidication, % of solid Solid expansivity, linear(a) 106/K Liquid expansivity, cubic, 106/K
Element
31 14 19 17 29 23 25 10 13
In any soldered joint, a fraction of the residual voiding does not derive from trapped air, moisture, or gas. These residual voids are extremely difficult, if not impossible, to remove because they are intrinsic to the ller metal, being caused by the shrinkage when it solidies and further cools. Table 4.7 lists values for the shrinkage volume contraction of elements common to many solders. The reservoir of solder represented by the edge spillage fraction is seldom able to feed large-area joints and compensate in part for the contraction, because the outer extremities of joints usually solidify rst through radiative heat losses to the surroundings. The magnitude of solidication shrinkage, as given in Table 4.7, accounts for the fact that it is difficult to make joints of large area that contain less than approximately 3 to 5% voids. Shrinkage voids tend not to occur in small or narrow joints (<2 mm, or 0.08 in., in one of the joint area dimensions). This is because the thermal gradients that usually develop along a joint when the assembly is cooled from the joining temperature are large in relation to the joint dimensions, and this causes the ller to directionally solidify from one edge to the other, thereby preventing voids from forming within the joint. It is possible to achieve the same effect in large-area and wide joints by imposing a temperature gradient on the assembly, from either center-to-edge or edge-to-edge, such that some periphery of the joint is always the last portion to solidify. However, this is not always easy to achieve, especially when large numbers of components are involved. Furthermore, the imposition of a temperature gradient on a large assembly may produce stress gradients and thereby dimensional distortion of the components, which becomes xed when the solder solidies. Where the parent materials and solder are of similar composition, maintaining the assembly
(a) CTEs of the solids are the average values over the range 0100 C (0212 F), while the liquid CTEs are just above their melting points.
at elevated temperature but below the solidus temperature of the ller can result in a gradual reduction in void levels arising from solidication shrinkage by vacancy diffusion. This is the mechanism by which dry joint interfaces are removed in diffusion bonding (see Chapter 1, section 1.1.7.2). The process times then become relatively long and not less than 1 h. Bismuth and, to a lesser extent, antimony are exceptional among metals in having a volume expansion rather than a volume contraction on freezing, as shown in Table 4.7. Therefore, by combining bismuth and/or antimony with other elements, it is possible to produce solders with essentially zero volume change on solidication. The Bi43Sn solder (melting point of 139 C, or 282 F) is an example of one such alloy that has been recommended for applications in the electronics industry, where joints of guaranteed hermeticity are required [Dogra 1985]. However, the volume change that occurs is not instantaneous but often takes place over several hours after the solder has solidied [Manko 2002]. The forces accompanying the volume expansion can be signicant and must be allowed for in the joint design.
4.3.2
Void-Free Soldering
Regardless of the origin of gaseous voids, a very successful method has been devised to compress the trapped gas so that it occupies a smaller volume fraction of the joint. Because this procedure works while the solder is molten, it also helps reduce the volume of voids arising from the liquid-to-solid phase change through solidication shrinkage. The pressure variation method was developed specically to reduce void levels arising from trapped gas in adhesively bonded joints [Bascom and Bitner 1975]. It can be used to make large-area joints using solder preforms
that have void levels consistently below 5% [Mizuishi, Tokuda, and Fujita 1988]. Most of the residual porosity is due to solidication shrinkage, as discussed in the preceding section. The principle of the pressure variation method is to use the pressure of an external atmosphere to compress the gas trapped in the joint. The procedure, as applied to vacuum joining, is as follows (Fig. 4.31):
ality, if the initial void volume at pressure P1 is V1, then, at constant temperature, the volume at pressure P2 is:
V2 V1 P1 P2
Hence, the greater is P2 in relation to P1, the more effective is the method. This condition is also found to apply qualitatively to practical situations. The experimentally derived relationship between pressure variation (P2/P1) and the volume of voids, V2, obtained using this process is shown in Fig. 4.32. The nonlinearity of the relationship at large P2/P1 ratios is due to departure from ideality of the gas and the hydrostatic friction at the interface between the component surface and the liquid solder. Solder reow ovens are available commercially that automatically perform the necessary pump/pressure cycles to achieve joints with few and controlled voids.The pressure variation method for minimizing voids is obviously not suitable for situations where vapor is continually being evolved from volatile constituents. A general approach that has been found to be effective in producing well-lled and hermetic joints is one in which strong metallurgical reactions occur across the joint during the heating cycle while a compressive force is applied (see Chapter 3, section 3.3, for a discussion on uxless
Fig. 4.31
Pressure variation method for reducing void levels due to trapped gas. T, temperature; P, pressure
Experimentally derived relationship between pressure variation and void level obtained in large-area soldered joints using the pressure variation method. Adapted from Mizuishi, Tokuda, and Fujita [1988]
Fig. 4.32
soldering).The void-free joints obtained using the diffusion soldering and diffusion brazing processes are associated with such reactions. The diffusion soldering process is described in Chapter 5, section 5.9, and diffusion brazing in the planned companion volume Principles of Brazing.
The key to making high-strength joints is to prevent the development of stress concentrations and, at the same time, the strength of the ller alloy must be maximized. These aspects are considered in turn.
4.3.3
New materials that have enhanced strengths are continually coming onto the market. Recent examples are composite materials such as metalmatrix composites (MMCs) and precipitationstrengthened and dispersion-stabilized alloys. Both of the latter types of strengthening have been exploited in high-carat gold, suitable for use in jewelry and bond wire for electronics [Humpston and Jacobson 1992; Jacobson, Harrison, and Sangha 1996; du Toit et al. 2002]. While there is a desire to exploit these materials in a range of applications, widespread adoption is contingent on being able to use the favorable bulk strength levels in joined assemblies. In general, the strength of a joint, even when prepared by welding, is inferior to that of the materials in monolithic form. Moreover, the heating cycle used in the joining process can itself degrade the properties of these materials. For example, aluminum/SiC MMCs are susceptible to degradation when heated above approximately 500 C (930 F) due to reaction between the constituents, which results in the formation of a brittle interfacial layer of Al3C4 [Iseki, Kameda, and Maruyama 1984]. Although solders are mechanically inferior to welds and brazes, they are nevertheless sometimes used to make joints that are required to sustain moderate forces. The primary example is joints in copper water pipes, where the pressure can easily be 0.7 MPa (100 psi) with a very modest pump. In most circumstances, application of stress to a joint does not result in all regions of the joint sharing an equal proportion of the load. The unevenness of the stress distribution is referred to as the stress concentration, K. Mathematically, this is a dimensionless number that simply describes the magnication factor of the actual stress at one location compared to the uniform stress that would prevail in the absence of any stress concentrations. Expressed as an equation:
K (stress concentration) Local stress at a specified location Applied force/Joint area
4.3.3.1
Stress concentrations can be reduced by using joint congurations that distribute the load away from the joint. Further details of this subject are given in the planned companion volume Principles of Brazing, because brazed joints tend to be used more often than ones made with solder for load-bearing applications. In order to understand the origin and magnitude of stress concentrations that can arise, reference is made to a single-lap joint loaded in tension. Stress concentrations arise from two sources: namely, the differential straining of the components and ller, and the eccentricity of the loading path. In lap joints, the shear strength of the joint per unit area (length) actually decreases with increase in the joint length. This apparent anomaly can be explained by the fact that the shear stress is highest toward the ends of the joint, so that if the length of the joint is increased beyond a certain limiting value, the ller in the central portion of the joint will carry little or no stress, with the applied stress concentrating at both ends, as depicted in Fig. 4.33. This explains why simply increasing the length of the overlap does not improve the strength of this type of joint beyond a certain level. The stress concentration in the joint is proportional to the length of overlap, up to a limiting value, the thickness of the members, and to the thickness of the joint. Therefore, the stress concentration is least in thin joints of short overlap. Far more relevant to the strength of singlelap joints are the tensile or peeling stresses that act normal to the ends of the joint and originate from the eccentricity of the loading of the assembly. The elastic analysis is relatively complex, but the result obtained is that longitudinal loading of a single-lap joint effectively applies a perpendicular tensile stress of approximately four times that amount to the ends of the overlaps [Harris and Adams 1984]. These perpendicular tensile forces initiate failure of the joint by peel. With the continued application of stress, the sample rotates in an attempt to correct for the axial misalignment, and
Fig. 4.33
Schematic illustration of the stress distribution in the ller metal of lap joints of short and long overlap. When stressed in shear, the central portion of a long lap joint carries little or no load.
the fracture continues to propagate due to peeltype debonding. The stress concentrations in a simple lap joint and their inuence on its resulting failure mode are illustrated in Fig. 4.34. Fillets at the edges of a joint act to reduce the stress concentration in that region, as indicated in Fig. 4.23. They do this by coupling some of the applied stress into the ends of the laps, thereby reducing the differential straining between the
components and the ller in the joint and also shifting the position of the maximum perpendicular tensile stress (originating from the eccentricity of loading) to outside the joint. The magnitude of these effects depends on the radius of the llets, R, the step height, H, and the elastic properties of the ller. To be effective, the radius of the llets must exceed the step height, that is, R > H, and hence, it is desirable for soldered joints to have large and well-rounded llets at their periphery (Fig. 4.35). The ideal joint is one in which, under all practical loading conditions, the ller metal is stressed in the orientation in which it best resists failure. The complexity of the joint should also take into account the load intensity to be sustained and any aesthetic considerations. In general, simple, lowcost joint designs work well with unobtrusive joints and low-level loads, while conspicuous joints with higher and more complex loading situations demand more elaborate and expensive congurations. Strategies for some of the more common joint geometries are presented as follows. Lap joints are probably the most common conguration because of their use in electronics (surface mount) and plumbing (pipe joints). Increasing the length of overlap will improve the ability of the joint to resist load along its length, but, following the law of diminishing return, this is
Failure in a simple lap joint loaded in tension. (a) Stress concentrations. (b) Initiation of failure. Edgeopening crack (free arrow) formed and propagated by the high normal stress concentration. (c) Progression of joint rotation to fracture. Plastic bending of the joint region results in the majority of the failure being due to peel-type debonding. Adapted from Dunford and Partridge [1990]
Fig. 4.34
Fig. 4.35
A lap joint showing step height, H, llet radius, R, and contact angle,
because the center of the joint carries no effective load. Further improvement can be made by tapering the ends of the overlap, which is easily achieved if the components are thin and a llet is encouraged to form in this region. Lap-joint styles for different stress regimes are illustrated in Fig. 4.36(a). Butt Joints. For a butt joint between two circular rods subject to tension, there is no stress concentration. The strength of such joints is therefore proportional to area. However, the plain butt joint is only suitable for the least demanding of applications. The main reason for this is that the joint has very low resistance to bending forces. The scarf butt joint has the merit of only requiring simple machining to prepare the faying surfaces yet is highly efficient at resisting deformation under load. Scarng results in differential strains and hence the stress concentrations at the ends of the joint being considerably reduced, while the landed or step joint relies on the step sizes being small to achieve the same effect. Both congurations are symmetrical, and therefore, axial stresses will be balanced over the joint. They are illustrated in Fig. 4.36(b). By making the scarf angle sufficiently small, the joint strength can be made to approach that of the parent materials; that is, when the scarf angle is 90, the joint is a butt joint, whereas if the scarf angle is 0, there is no joint, just two
parallel pieces of parent material. From a theoretical perspective, a radially symmetric tongueand-groove joint should be the best able to resist loads, but unless the components being joined are particularly large in diameter, achieving adequate lling of the joint could be problematic, and the component preparation costs could be quite high. Strap joints are often used as cheap alternatives to butt joints, because the component pieces are generally simpler, and less precision machining is required, although the thickness and weight of the assembly are increased, and its aero/uid dynamic performance is often impaired. Some recommendations for different stress regimes are shown in Fig. 4.36(c). The main problem with this style of joint, as with lap joints, is that any asymmetry in thickness or material properties results in stress concentrations that cause the parent material to fail prematurely just outside of the joint region (this can erroneously be taken as an indication that the joint is stronger than the parent material). The stress concentration can be reduced by adding taper to the straps by machining and, ideally, allowing generous and continuous llets to form. In the preceding discussion, the assemblies were considered to be loaded solely in uniaxial tension. The location of any stress concentration and its magnitude will change as the stressing
Fig. 4.36
Recommended designs of (a) lap, (b) butt, and (c) strap joints for different stress environments
mode is altered, and hence, the optimal style of joint varies depending on the stress environment in which the component is required to operate. 4.3.3.2 Strengthened Solders to Enhance Joint Strength
One of the limiting parameters of joint strength, especially of thick joints, is that of the solder itself. Solders can be strengthened by metallurgical mechanisms involving elements placed in solid solution, microscopic second-phase particles (of either intermetallic precipitates or a dispersed refractory phase), and renement of the grains of the ller. These novel solders offer signicantly improved mechanical properties, particularly at room temperature. However, hardly any have yet been developed to the point of commercialization. An overview of the research in this area is presented in Chapter 5, sections 5.5 and 5.8. Another approach is to load the solder with a uniform distribution of coarse particles or bers (typically, 100 m to 1 mm, or 4 to 40 mil, in size) of a refractory or nonmetallic material. The dimensions of the reinforcement dictate that the joint gap must be relatively wide. In laboratory tests using chopped carbon bers as the reinforcement, substantial enhancement of the shear and tensile strength of the joints with respect to the unmodied solders has been demonstrated and, more particularly, a signicant reduction in the thermal expansivity of the solder [Ho and Chung 1990; Cao and Chung 1992]. Further details of these investigations can be found in Chapter 5, section 5.6.
It is perfectly practicable to make soldered joints that are as thin as 2 m (80 in.), even in volume manufacturing. Joints of this thinness require that the solder is preapplied to one of the joint surfaces as a high-quality lm. Ion-assisted vapor phase deposition and sputtering are the only reliable methods of achieving this. Electroplated and thermally evaporated lms are not sufficiently dense, and the residual porosity results in the proportion of oxide being large enough to interfere with wetting and spreading. It is generally not possible to make thin soldered joints by simply using a narrow joint gap and hoping that the solder from an adjacent reservoir area will run in and ll it. This is, again, a function of the very small volume of ller metal whose composition will change on alloying with the faying surfaces. Where the joint is thin, the composition of the advancing solder front rapidly becomes uniformly alloyed with material from the joint surfaces, and isothermal solidication ensues. In thin joints, liquid uxes interfere with wetting and spreading, because the volatile species have trouble escaping from a narrow gap, as can be seen from Fig. 4.37. Narrow joints are therefore best made using a gaseous ux or uxless (see Chapter 3, sections 3.3.6 and 3.3.8, respectively). Another method of making a thin joint is to use a more conventional quantity of solder, appropriate to a wider joint,
4.3.4
Under normal circumstances, a solder joint will naturally tend to be a few tens of microns (approximately 1 mil) thick. Sometimes, it is necessary to create joints that are either signicantly thinner (<10 m, or 0.4 mil) or thicker (>50 m, or 2 mil). Solders do not have particularly good thermal conductivity, so that if a joint is required to transport heat through its thickness, then thin joints are obviously desirable. Thick joints tend to be encountered where the mechanical tolerance of the components does not allow for joints to be consistently made narrower, or where creep is desirable in order to relieve mechanical stress.
Shear strength of soldered joints in brass testpieces as a function of joint thickness. Narrow joint gaps are progressively more difcult to ll, thus decreasing the measured shear strength of thin joints. A gaseous ux is better able to penetrate narrower joint gaps than a liquid ux; consequently, thinner joints can be made before the joint-lling problem appears. Adapted from Manko [1992]
Fig. 4.37
and, once the joint surfaces have been wetted by the molten ller, apply sufficient compressive stress to overcome the hydrostatic pressure of the solder to extrude surplus material from the joint gap. Physical stops can be used to control the nal joint gap. If the lower component is larger than the upper one, lands can be provided to catch the overspill in a controlled manner. The stress required to do this reliably is of the order of 100 g/mm2 (1 Pa, or 0.2 lb/ft2), which may damage some components, but it does usually enable the soldering process to be uxless (see Chapter 3, section 3.3.8.3). A frequently overlooked consideration when attempting to make thin joints is the cleanliness of the components and, particularly, the environment in which the assembly joining is conducted. When the desired joint gap is just a few microns wide (typically, 100 in.), there is no point jigging the components in a room where the airborne particles are larger! For this reason, the soldering process must be undertaken in a semiconductor-grade clean room, and close attention must be paid to the particulate content of all process gases, cleaning chemicals, tools, and so on. Table 4.8 shows the correlation between the various classes of clean room and their particle size distributions. Clearly, if the requirement for a joint gap is below 5 m (200 in.), then a class M4 (class 100) or better clean room is required. When endeavoring to make particularly thick solder joints, the problem encountered is how to retain the solder in the joint gap, particularly if there is compressive stress acting on the molten ller. The traditional method of solving this problem has been practiced for generations in the
plumbing industry. The approach is to select a solder that has a wide melting range and to conduct the joining operation below the liquidus temperature, when the ller alloy is in a pasty state (a mixture of solid and liquid). The presence of the solid phase drastically modies the viscosity of the alloy and prevents it from owing out of a wide joint gap. The same result can be achieved by loading a solder with solid particles. Inserting thin parallel shims, for example, of copper, into the joint effectively partitions the joint gap into a series of much thinner joints and enables conventional joining methods to be employed. Brazing of wide joint gaps is regular practice as a crack repair technique in the aerospace industry and involves inserting a honeycomb into the joint gap, again to partition the joint into a number of cells of more conventional dimensions. Further details can be found in the planned companion volume Principles of Brazing. The successful use of this approach with solders has not been documented. Flip-chip interconnects sometimes make use of joints that are thick in relation to their plan area. For example, the solder interconnects on ball-grid array integrated circuits (BGAICs) can easily be 1 mm high. Here, surface tension forces are exploited so that the solder interconnect adopts the shape of a truncated sphere and can therefore be tall in relation to its diameter. This approach to thick joint gaps only works because the packaged IC is light in relation to the total joint area (i.e., the sum of all the individual solder balls). Hence, the total surface tension force is sufficient to levitate the IC and hence achieve a thick joint gap. Flip-chip technology is discussed further in Chapter 5, section 5.2.
Table 4.8
Relationship between clean room class designation and airborne particle size distribution
Class limits
Class name Sl M1 M1.5 M2 M2.5 M3 M3.5 M4 M4.5 M5 M5.5 M6 M6.5 M7 English ... 1 ... 10 ... 100 ... 1000 ... 10,000 ... 100,000 ...
curring from a gaseous phase on or immediately adjacent to the surface of a substrate is known as chemical vapor deposition (CVD). The substrate is usually heated to generate the reaction. Chemical vapor deposition may be classied according to the type of chemical reaction involved. In a decomposition reaction, a gaseous compound AB may be decomposed into a solid condensate A and a gaseous product B when placed in contact with a colder substrate. If the compound AB instead dissociates into a solid phase A and a gas phase AB2, say, then the CVD process is referred to as one involving a disproportion reaction. Oxidation and reduction of halides constitute the two other types of reaction that are widely employed. Wet plating of metallic layers encompasses processes where coatings are deposited on a substrate through immersion of the substrate in a liquid, usually aqueous, containing the appropriate metallic ions. The deposition often functions by ionic discharge, with the metal deposited onto an electronegatively charged, conductive substrate (cathode). The plating process can introduce organic compounds into the metal coatings, although these can often be minimized by judicious choice of the bath formulation. Thin coatings can be grown autocatalytically (i.e., without an applied electric eld) through a reduction of metal ions in the plating bath by the immersed substrate. This process is known as chemical displacement and also as immersion plating. Another autocatalytic method, commonly referred to as electroless plating, involves the deposition of metal from a plating bath containing the metal ions together with a reductant. This process differs from chemical displacement in that no signicant reaction occurs within the volume of the liquid, and the depositing metal catalyzes further deposition, so that thicker lms can be grown. It is usually necessary to activate nonmetallic substrates by chemical treatment for them to generate the catalytic reaction. Nonmetallic elements, principally, phosphorus and boron, tend to be incorporated into the metallic coating from the reductant. Thick-lm formulations usually comprise a slurry, containing the metals or metal compounds and sometimes a glass in an organic carrier, which is intended to be applied by painting or screen printing onto the desired areas. Subsequent ring drives off the organic fraction and stabilizes the metallization by producing a diffused interface with the nonmetal substrate. It is
usual practice to apply and re each thick-lm metallization separately, although processes have been developed whereby at least two thick-lm layers are red together. Common thick-lm metallizations are discussed as follows. Systems based on reactive metals (zirconium, tungsten, titanium, manganese, molybdenum). These formulations are red at approximately 1600 C (2900 F) in a reducing atmosphere. Because of the relatively refractory nature of the resulting metal surface, either a strongly reducing environment is required to effect subsequent wetting by solder, or a wettable metallization should be applied on top. Alternatively, the wettable surface layer may be applied over the reactive metal layer and the two layers red together. An example is a tungsten-loaded frit overcoated with a nickel paste, which is cored at 1300 C (2370 F). Even at this temperature, the interdiffusion between the two metals is slight, so that a discrete layer of nickel forms on the surface after the heating cycle [Kon-ya et al. 1990]. Systems based on noble metals (copper, silver, gold, palladium, platinum). These materials are red between 850 and 950 C (1560 and 1740 F). The silver, gold, and platinum metallizations can be red in air, while a reducing atmosphere is generally required for the less noble metals. Metal-loaded glass frits are red on the surfaces of components above 400 C (750 F) to form a glaze that is strongly adherent to the non-
metal substrate. The concentration of the glass at the interface with the component means that the outer layer of the coating is sufficiently metallic in character for it to be electroplated or directly soldered. Thick-lm metallizations are supplied as complex proprietary formulations and are available in different physical forms, each tailored for a limited range of substrate materials. It is advisable to consult the supplier on their conditions of use and likely properties. As might be expected, there are advantages and disadvantages associated with the different metallization techniques. Vapor deposition is generally superior to wet plating in offering better control of impurities and reduced porosity in thin coatings. Wet plating, by comparison, tends to be faster and cheaper and can provide thicker coatings. Thick-lm metallizations may be more readily applied to selective areas and are more tolerant of substrate topology. The technique that will normally be chosen will be the one best suited to the particular application on the grounds of its tness for purpose and cost. A brief comparison of the characteristics of the principal methods used for applying metallizations, together with those of the coatings that they are capable of producing, is presented in Tables 4.9 to 4.12. It must be pointed out that the entries in the tables represent the general situation; particular cases might be out of the ranges indicated.
Table 4.9
Elemental metals
m-mm
Electroplating
Elemental metals and some binary and ternary alloys Elemental metals and a few binary alloys Wide range of elemental metals and alloys
Electrical conductors
Moderate
m-mm
Wide range of materials Materials that can withstand the ring temperatures
m m-mm
Good, but need to stringently control several process variables simultaneously Generally less precise than for vapor deposition Generally poor Moderate
Very high, can be continuous High, can be continuous High, batch or conveyor belt
Further information on these metallization techniques can be found in the literature, which is extensive. A comprehensive review that covers the chemical bonding, chemical reaction,
Table 4.10
Process
interface structure and properties of metal/ ceramic interfaces, property measurement, and their fracture behavior is given by Howe [1993]. Other useful sources of information are
Vacuum evaporation
Sputtering
Relatively simple equipment required for resistance heating evaporation, which is suitable for coatings of most elemental metals Possible to coat a wide range of compositions Dense coatings and good adhesion obtainable High-quality coatings are obtainable. Output is generally high. High throughput. Large areas can be coated with uniform thickness. Limited only by the size of the plating bath. Relatively easy to control Large areas can be coated with uniform thickness. Good throwing power. Only very basic equipment is required. Requires simple equipment. Lends itself to high volume production using screen printing and ring in belt furnaces
Not suitable for alloys that have constituents with greatly differing vapor pressures. Meticulous substrate cleaning prior to deposition is required. Requires sophisticated equipment. Low throughput. Heating of substrates and low deposition rates in conventional diode or triode sputtering Equipment is sophisticated and is usually specic to particular coatings. Chemical handling, vapor, and effluent problems. Film impurities and imperfections can also present problems. Can only apply coatings to electrically conductive materials. Thorough cleaning and chemical activation of substrates are required prior to plating. As above, except that nonconducting materials can be plated. Range of available coatings is restricted. Relies on manufacturers proprietary formulations. Relatively high process temperatures are used. Only thick lms can be applied by this technique.
Electroless plating
Thick lm
Table 4.11
Process
Vacuum evaporation Sputtering Chemical vapor deposition Electoplating Electroless plating Thick lm
0.0015 0.0051 5100 0.1100 0.11 1,00010,000 (does not include ring times)
Substrate is often heated to 200 C (390 F) to promote adhesion Mostly below 100 C (212 F) 2002000 C (3903630 F), but usually 400800 C (7501470 F) 10100 C (50212 F) 10100 C (50212 F) 4001800 C (7503270 F)
Table 4.12
Process
Vacuum evaporation Sputtering Chemical vapor deposition Electroplating Electroless plating Thick lm
Variable; determined by source-substrate geometry Higher uniformity possible than for vacuum evaporation Good uniformity possible; depends on design of the deposition chamber Good uniformity on ats, nonuniform at edges Fair uniformity Variable
Moderate to low porosity Low porosity Dense and essentially pore-free Susceptible to porosity and blistering Susceptible to porosity and blistering Dense coatings are achievable
Purity limited by source materials and deposition atmosphere Purity limited by source materials and deposition atmosphere Purity is that of the starting materials or even better May incorporate salts and gaseous inclusions May incorporate salts and gaseous inclusions Often contain glass and possibly organic residues
Fair Generally excellent Variable; dependent on materials and processing conditions Variable; often excellent Variable; often excellent Variable; dependent on materials and processing conditions
A single void equal to the length of the joint A single void that traverses the width of the
joint Again, these criteria permit a joint to contain large voids that potentially can be responsible for catastrophic failure yet is able to pass inspection. For application where void-free joints are essential to the functionality and reliability of the product, proposed inspection criteria are as follows: Visual Inspection. A sensible inspection criterion for void-free joints that relies only on visual examination is the presence of a complete and uninterrupted llet around the joint periphery. Although this condition does not guarantee zero voids within the joint, it does at least provide assurance that the external surfaces achieved and maintained wettability during the process cycle. It also indicates that the joint is leaktight. According to the same rationale, it is desirable that the llets are smooth, shiny, and have low contact angles. A smooth and shiny solder surface is an indicator of the absence of an oxide skin. Any blemishes, such as nonuniform hue, discoloration, bumps, cracks, craters, or foreign material, give cause for concern as to the integrity of the hidden volume of the joint. Low llet contact angles are desirable as an indicator of good wetting, but this inspection criterion is not usually available because solder spread often is conned by lands, causing the contact angle to be higher than it would be if the solder were able to spread freely. X-Ray or Scanning Acoustic Microscope (SAM) Inspection. The most demanding target that can sensibly be set for joints larger than a few millimeters per side is <5% voids by plan area. A truly void-free joint is not seen to be an achievable target in volume production, not least because even state-of-the-art assessment techniques can lead to ambiguity in identifying a void, especially when using thin, predeposited solder and thick components. For each application, it is should also be possible to dene a maximum acceptable void size (either by area or linear dimension) in various areas of the joint. This gure can be obtained from thermal and physical modeling. Void fraction, maximum void size, and their impingement into areas where solder is essential can be gen-
>75% of the perimeter to exhibit a llet >75% of each side to exhibit a llet Two or more sides to exhibit complete llets <50% voids in the joint No single void to traverse either the length or width of the joint, or to be >10% of the joint area
Although these standards provided welldened criteria for judging an assembly process, they are not well suited for the purpose of assessing void-free joints. In particular, there is no external indication of dewetting in the joint gap, and 49% voids is arguably a somewhat high level for a joint that is judged to be void-free. The aforementioned inspection criteria relate mostly to visual inspection of the joint and therefore its external appearance. To properly assess whether or not a joint is free of voids requires inspection of the joint interior. This is often possible using scanning acoustic or x-ray techniques (see Chapter 5, section 5.10). Standards exist that dene what constitutes a defect when voids are detected using these inspection methods. For example, ESA x-ray
erated automatically by modern x-ray and SAM equipment in real-time. The bottom line in devising an inspection strategy is to dene a set of criteria that can offer assurance of product integrity and thereby add to its value and reputation for reliability.
and contaminate the package atmosphere. Solder, particularly when used uxless, is ideal for this task. Gold-silicon solder, which is frequently used for semiconductor die attach applications, acts as a desiccant after the package is sealed, because the silicon in the alloy reacts with water vapor. Tests show that preforms of this material reduced the moisture content in the package from 15,000 ppm for control samples to 125 ppm [Carley, Nearhoff, and Dennin 1984]. The hydrogen content in the package was found to increase from 1800 to 7000 ppm, implying that the likely reaction taking place is:
Si 2H2O SiO2 2H2
Adhesives have been formulated that do not greatly outgas. However, care must be taken with their use in sealed enclosures, because they are only stable in this regard provided they are not heated beyond the curing temperature. For example, if a semiconductor die is attached using an epoxy adhesive, but the package is then sealed using a high-temperature solder, the adhesive will liberate moisture as the sealed package cools from the lid-seal process temperature. Next, the package needs to be dried before it is sealed. Water adsorbed onto metals needs to be heated to 115 C (239 F) in order to force it to desorb. One widespread misconception is the belief that a low moisture reading on a sealing box hygrometer will ensure a dry ambient inside the package. The fallacy arises because the major source of moisture is that adsorbed onto the package walls, and thus, the dry box reading has no correlation with the resulting package ambient atmosphere. Removing the water of hydration alone is not sufficient. Although this moisture will desorb at 115 C (239 F), it is not until a metal surface is heated above approximately 160 C (320 F) that the absorbed hydroxide species will convert to metal oxide, to the accompaniment of further evolution of water [Swartz et al. 1983], according to the reaction:
M(OH)2 MO H2O
Predicted time for moisture to permeate various packaging materials in one geometry. Adapted from Traeger [1976]
Fig. 4.38
where M represents the metal. Figure 4.39 shows the moisture content in a metal package as a function of the bakeout time and temperature. Low residual moisture content is obviously favored by high bakeout temperatures sustained for long periods.
A problem frequently encountered is that the package contents or assembly method do not permit the use of elevated temperature for the bakeout, while manufacturing economics are not compatible with extended bakeout times at a reduced temperature. One solution to this problem is to place the package in a sealed chamber and, at the highest permissible temperature, repeatedly saturate the atmosphere with dry nitrogen gas, and then pump it out. Nitrogen gas obtained from a cryogenic source and conveyed by stainless steel pipework will have a moisture content of 2 ppm or less. The moisture on the package walls will equilibrate with that in the atmosphere, so that when the nitrogen gas is removed, it will take moisture with it. Twenty such cycles, conducted over a period of a few hours, will reduce the moisture content of the package to a few parts per million. This approach has the additional advantage of reducing the time that the package is exposed to vacuum, which can result in hydrocarbon contamination of surfaces arising from backstream-
Moisture content of metal packages as a function of bakeout time and temperature. Adapted from Thomas [1976]
Fig. 4.39
ing of vacuum pump oil vapors. Note that the package must be sealed while still in the bakeout chamber because, if it is exposed to air, the surfaces will instantly become saturated with water again, even if they are kept hot. Having hermetically sealed the package, it might be expected that the contents would now be permanently protected from the atmosphere outside. This is not true because, in practice, hermeticity can only be specied in terms of a nite leak rate. Small gas molecules will enter the sealed cavity by diffusion or permeation until, ultimately, equilibrium with the atmosphere outside is reestablished. Package leak rates are measured using helium as a tracer gas, and the acceptable leak rate is dened for the specic application. For example, a package having an internal volume of 100 mm3 and a leak rate of 5 107 Pa m3/s (4 108 ft lb/s) will have a theoretical water buildup rate of roughly 104 ppm per day, indicating that the package atmosphere will be in equilibrium with the air in a matter of days. At a leak rate of 109 Pa m3/s (8 1011 ft lb/s), equilibrium will be achieved within a year. Figure 4.40 shows predicted moisture penetration rates for a helium leak rate of 108 Pa m3/s (8 1010 ft lb/s) for ambient air at 70% relative humidity [Stroehle 1997]. A helium leak rate of 109 Pa m3/s (8 1011 ft lb/s) is generally considered to be a minimum acceptable level of hermeticity, and 1011 Pa m3/s (8 1013 ft lb/s) to provide adequate protection against moisture ingress for all but the most sensitive components. The latter leak rate is also the lower detection limit of low-cost helium leak check equipment, while 1014 Pa m3/s (8 1016 ft lb/s) is about the limit of specialized laboratory systems.
REFERENCES
Calculated ingress of water vapor into a hermetically sealed package having a leak rate of 108 Pa m3/s (8 1010 ft lb/s). The package volume is 2 mm3 (1.2 104 in.3), and the ambient air is assumed to be at 70% relative humidity.
Fig. 4.40
Huchisuka, T., 1986. Bonding of Sintered Alloys, Met. Technol., Vol 56 (No. 5), p 2127
Annual Reliability Physics Symposium, April, p 101106 Swartz, W.E. et al., 1983. The Adsorption of Water on Metallic Packages, Proc. 21st Annual Reliability Physics Symposium, p 5259 Thomas, R.W., 1976. Moisture, Myths and Microcircuits, IEEE Trans. on Parts, Hybrids, and Packaging, Vol 12 (No. 3), p 167171 Timoshenko, S., 1925. Analysis of Bi-Metal Thermostats, J. Opt. Soc. Am. Rev. Sci. Instrum., Vol 11 (No. 9), p 233255 Tokuriki Hoten, K.K., 1981. Composite AuSn Eutectic Type Solder Alloy, Japanese Patent JP 81199063 Traeger, R.K., 1976. Hermeticity of Polymeric Lid Sealant, Proc. 25th Electronics Components Conference, April, p 361367 Vianco, P.T., 1998. An Overview of Surface Finishes and Their Role in Printed Circuit Board Solderability and Joint Performance, Circuit World, Vol 25 (No. 1), p 624 Vianco, P.T., Sifford, C.H., and Romero, J.A., 1997. Resistivity and Adhesive Strength of Thin Film Metallizations on Single Crystal Quartz, IEEE Trans. Ultrasonics Ferroelectrics Frequency Control, Vol 44 (No. 2), p 237249 Warwick, M., 2002. Tombstone Reduction via Phased Reow Soldering, Surf. Mount Technol., Vol 16 (No. 9), p 44-54 Wersing, W., 1992. Improved Ceramics Through Multilayer Technology, Metals and Mater., Vol 8 (No. 6), p 326331 Weston, A.D., 1974. Mounting Semiconductor Bodies, U.S. Patent 3,786,556 Xian, A. and Si, Z., 1992. Interlayer Design for Joining Pressureless Sintered Sialon Ceramic and 40Cr Steel Brazing with Ag57Cu38Ti5 Filler Metal, J. Mater. Sci., Vol 27 (No. 3), p 15601566 Xie, D.J., Chan, Y.C., and Lai, J.K.L., 1996. An Experimental Approach to Pore-Free Reow Soldering, IEEE Trans. Components Hybrids Manuf. Technol., Vol 19 (No. 1), p 148 153 Yung, E.K. and Turlik, I., 1991. Electroplated Solder Joints for Flip-Chip Applications, IEEE Trans. Components Hybrids Manuf. Technol., Vol 14 (No. 3), p 549559
CHAPTER 5
5.1
Lead-Free Solders
The literature on lead-free solders is almost overwhelming. A computer search on the term lead-free quickly reveals tens of thousands of technical papers and references to conference presentations on the subject. This depository of knowledge is augmented by entire issues of journals, chapters in several books, and entire books [e.g., Hwang 2001]. Thus, there is much valuable
technical information on the metallurgy, physical and chemical characteristics, and reliability of joints made using lead-free solders. Extensive as it is, these collective works often do not adequately describe the process window for a particular joining problem of interest nor predict the reliability of the resulting joints. This is, perhaps, not surprising, given such key information is still not immediately available for lead-tin solder, despite many years of endeavor. Many of the leadfree alternatives are technically more complex materials. The principles of lead-free soldering are not fundamentally different from those of lead-tin soldering. Lead-free soldering processes are now a commercial reality, and within very few years, it is likely that the vast majority of electronics and optoelectronics products will be manufactured using this new technology.
5.1.1
The drive for lead-free soldering for massmarket electronics assembly rst came to prominence in the late 1980s with Senator Reads bill in the United States Senate [U.S. Senate bill S391 1990; U.S. Senate bill S729 1993]. This bill sought to address the concern for human health posed by the rapidly increasing quantities of discarded electronics equipment accumulating in landll waste disposal sites. Although battery manufacture accounts for the vast majority of the 5.5 million tonnes (6.06 million tons) of lead consumed each year, battery recycling is almost 100% effective. However, virtually none of the 60,000 tonnes (66,000 tons) of lead used by the electronics and lighting industries is recycled. The waste in landll sites is subject to chemical attack by rainwater, from where leached-out constituents, including lead, eventually nd their way into drinking water supplies [Smith and Swanger 1999]. In 1998, the European Union introduced a draft directive, which is a precursor to law, entitled Waste Electrical and Electronic Equipment (WEEE) [The European Commission 2000; The European Commission 2001]. This followed Senator Reads initiative and called for a ban on lead in all electronics, except for automotive use, by 1 January 2004. The WEEE directive was intended to ban the selling, importing, and exporting of electrical/electronic equipment containing lead. The extent to which this directive was politically motivated may be the subject of
informed debate, but it is noteworthy that in the preceding January, the Japanese Electronic Industry Development Association (JEIDA) and the Japanese Institute of Electronic Packaging (JIEP) presented a roadmap to totally lead-free technology by April 2001. The combined effects of these proposals was to put pressure on the rest of the world to follow suit and generated a large investment in soldering process, equipment, and materials development. Many tens of thousands of hours were committed to research and development of leadfree solders during the 1990s. Although the efforts were somewhat uncoordinated, the conclusions of these studies conducted across the globe were remarkably similar: Lead-free soldering was seen to be a technical possibility, and indeed, further work has resulted in the commercial availability of lead-free solders and lead-free electronics products [IDEALS 1999]. The threat of legislation has now receded. Senator Reads bill was withdrawn, and the European Union has greatly expanded the category of exceptions and pushed back the implementation date to 2006 or 2008 (the date is not rmly xed). Nevertheless, most responsible companies now have accepted a commitment to clean and green manufacturing, whereby no new electronic products may contain lead. The authors are of the opinion that there will be a progressive transition to lead-free manufacturing for electronics products over the next decade or so, driven largely by companies wishing to promote an environmentally aware image. The majority of the work on lead-fee solders has concentrated on the need to nd a replacement for lead-tin near-eutectic alloys. While this endeavor has been successful, no alternatives to the high-melting-point, lead-rich family of alloys with melting points of approximately 300 C (570 F) have been identied. In recognition of this fact, the pending legislation exempts highmelting-point solders. Section 5.1.8 of this chapter outlines some of the reported attempts to nd replacements for these alloys. One positive benet of the switch to lead-free solders is elimination of the need to source low-alpha-emission solder for electronics applications. Lead, being a heavy metal, contains a proportion of radioactive isotopes, and the alpha particles emitted by spontaneous decomposition can cause electronic circuits to generate spurious digital signals. Although lead is nominally cheap, being only an eighth of the cost of tin, certied low-alpha lead is much more ex-
pensive. The price of low-alpha lead roughly doubles with every tenfold reduction in the counts per hour of emitted particles, so its removal from premium-quality solders represents a potential cost saving.
5.1.2
Lead-tin eutectic solder has many desirable characteristics, namely, excellent wetting and spreading behavior and a relatively advantageous range of physical properties. Moreover, because hitherto the entire electronics and lighting industries had developed around this alloy system, all associated materials, components, and subsequent processes were developed to be compatible with its properties and limitations. The initial goal of the work on lead-free solders was to nd a drop-in replacement for lead-tin that could be used at the same process temperature, on the same equipment, with the same substrate materials and uxes, and could deliver comparable performance and reliability. For the reasons that are elucidated, it is now recognized that this is not technically possible. Instead, there now exist several families of lead-free solders, each suited for a different group of applications and compatible with only a limited subset of metallizations and processes. A further consideration, of which any practitioner of soldering needs to be aware, is that many of the lead-free solders are incompatible with lead-tin solder. This will present difficulties for many years ahead, because an attempt to rework or repair a lead-free joint with lead-tin solder, or vice versa, can result in a joint with severely compromised functionality and reliability. This problem will be further exacerbated, because it is not easy to distinguish a lead-tin joint from one made with lead-free solder without resorting to analytical techniques.
5.1.3
In the mid-1990s, it was estimated that the electronics industry uses approximately 60,000 tonnes (66,000 tons) of lead-tin solder each year and the lighting industry somewhat more. This is used to make approximately 1013 soldered joints per annum [Vincent and Humpston 1994]. For any alloy to be a worthwhile solder for the electronics industry, it must meet specic qualities under the following criteria:
not subject to supply or price constraints that would prevent their widespread adoption. Based on the current world consumption of lead-tin solder, each 1% substitution for leadtin solder by another element represents a new annual consumption for that element of the order of 500 to 600 tonnes. This means that supply levels should be higher by severalfold. There are, for example, some excellent alloys that contain indium, but only 100 tonnes of indium are produced each year, and most of that quantity is recovered as a by-product of zinc extraction. World production of elements potentially suitable as constituents of lead-free solder is listed in Table 5.2. Taking into account availability, cost, melting point, and environmental issues leaves tin as the only element on which new lead-free solders for the mass market can be developed. Because tin melts at 232 C (450 F), alloying additions are necessary to depress the liquidus temperature and, preferably, to reduce the cost, because tin is relatively expensive compared to lead. Gallium
Mercury Cesium Gallium Rubidium Potassium Sodium Indium Lithium Tin Bismuth Thallium Cadmium Lead
38.9 28.5 29.8 38.9 63.7 97.8 156.9 179.0 231.9 271.3 303.5 320.8 327.5
Toxic Highly reactive Extremely rare Highly reactive Highly reactive Highly reactive Low availability and high cost (more expensive than silver) Highly reactive Nontoxic. Ample supply Obtained as a minor by-product of lead Toxic Toxic vapor Toxic
and indium are simply too scarce and too expensive to use for other than niche applications. The situation with regard to bismuth, silver, copper, antimony, and zinc is less restrictive, and virtually all of the lead-free solders available commercially are based on tin with selected additions of these elements. Almost without exception, lead-free solders are based on the silver-copper-tin eutectic with small optional additions of bismuth, antimony, indium, and zinc. Many of the most promising alloys have been assigned patent protection. Some of the key compositions are listed in Table 5.3. Bismuth is often recommended as an addition to lead-free solders because of its benecial effect on melting point and mechanical properties, as shown in Fig. 5.1 and 5.2. Users of bismuth-containing lead-free solders need to be aware of a potential incompatibility when undertaking repair work or using it to solder to lead-tin component or board nishes. The presence of lead in the alloy mixture results in nal solidication of the alloy at temperatures that can be below 100 C (212 F), depending on the composition and composition gradient in the soldered joint. Obviously, this could have a catastrophic effect on the reliability of an assembly, because it would normally be expected that the joint would remain solid up to at least 150 C (300 F). Lead-free solders are more expensive than lead-tin solder, because lead is substantially cheaper than all of the other constituents, and only zinc comes close to it in price. While this is important for wave-soldering processes where an industrial machine will require tens of kilograms of solder to ll the solder bath, the price premium of lead-free solders over lead-tin solder is minimal, when in the form of dispensable
Table 5.3 Composition ranges (wt%) claimed for some lead-free solder alloys
Tin
Silver
Copper
Bismuth or antimony
Indium
Zinc
(a) Source: U.S. Geological Survey data. To convert to U.S. (short) ton, multiply by 1.10
Note: The obvious overlap in composition ranges makes the legal fortitude of many of the patents questionable. However, cross-licensing of nearly all the silver-copper-tin family of lead-free solder alloys now appears complete, meaning that most formulations are readily available from the majority of suppliers.
pastes. The economics of paste production are such that they are dominated by the costs of preparing clean, spheroidized solder powder in precise size distributions and those of the specialist organic chemicals that comprise the binder and ux. Thus, while silver-copper-tin eutectic alloy costs approximately ve times that of leadtin eutectic, when based on metal prices, the premium paid for paste is typically only of the order of 1.1 to 1.
5.1.4
cooling-rate dependent. It is now generally considered that the composition for the ternary eutectic under equilibrium conditions is 3.5Ag0.9Cu-95.6Sn, with a eutectic point at 217.2 C (423.0 F). A complete liquidus projection of the ternary system is given in Fig. 2.14. Substitution of 0.9% Sn in the binary Ag-96.5Sn alloy by copper reduces the eutectic temperature by almost 4 C (7 F). Even this small temperature reduction can be signicant for electronic assemblies with components tailored for use with the lower-melting-point lead-tin solder, which can be detrimentally affected by higher soldering temperatures.
Because of the importance of silver-copper-tin alloys to lead-free soldering technology, it is imperative to have a reliable determination of the constitution of the silver-copper-tin phase diagram. With the aid of modern analytical techniques, the phase relationships in this system have recently been revisited and precisely determined. An excellent publication on this topic is one by Loomans and Fine [1999], which is well worth consulting as a model on how to determine a phase diagram using thermal analysis and metallography, particularly for a system where the composition of invariant reactions is
5.1.5
Reduction in the liquidus and solidus temperature of an off-eutectic silver-tin alloy as a function of the bismuth addition
Fig. 5.1
The metallurgy, physical, and chemical properties of lead-free solders are not markedly different from other solders. They wet, spread, and alloy with substrate metals in a similar fashion, and all other properties are broadly similar. However, there are small differences and various nuances that have been identied that may or may not be critical for certain processes. Some of the key parameters are presented in the following section, which draws extensively on the review by Glazer [1995]. As explained previously, the term lead-free solders is generally taken to mean alloys based on the silver-copper-tin ternary eutectic, although the data used for comparison also include other low-melting-temperature leadfree alloys. In attempting any comparison of solder properties, it must be remembered that for common solder alloys, room temperature represents close proximity to their melting points, in terms of chemical thermodynamics. Thus, in any test, it is virtually impossible to standardize factors that are governed by diffusion. This makes it extremely difficult to compare work by different investigators. While the data presented here may appear to be quantitative, it is best interpreted as semiquantitative and as an indicator of trends.
5.1.5.1
Surface Tension
Fig. 5.2
Effect of additions of bismuth on the tensile strength of bulk samples of silver-tin solder
The surface tension of tin is either raised or lowered by the addition of other elements in low proportions. Lead, bismuth, and antimony cause a reduction, while silver and copper increase sur-
face energy. Values for surface tension are given in Table 5.4. Surface tension values of solders exposed to air are typically lower than in an inert atmosphere, because oxidation lowers the surface energy of the molten solder. 5.1.5.2 Other Physical Properties
Data for the density, electrical resistivity, thermal conductivity, and coefficient of thermal expansion (CTE) of a few solders are given in Table 5.5. Density and thermal conductivity largely follow the rule of mixtures. On the other hand, resistivities and CTEs are all closely similar for the different solders, with the exception of Bi43Sn. Bismuth is a semimetal, rather than a normal metal, that is present in near-equal proportion by volume as the tin phases in the microstructure of the Bi-43Sn solder, which accounts for its higher resistivity and lower thermal expansivity. 5.1.5.3 Mechanical Properties
Because solders creep rapidly at room temperature, measurement of mechanical properties is extremely sensitive to strain rate. An additional complication arises because mechanical properties are also very sensitive to grain size,
Table 5.4 Measured values of surface tension for binary solder alloys in air and nitrogen
Alloy Air Surface tension, mN/m Nitrogen
Test conditions: test temperature, 250 C (482 F); polytetrauoroethylene substrate; SNMA, synthetic, mildly activated. (RMA-type, rosin, mildly activated) ux
joint thickness, and alloying with the parent materials. For solder alloys that are composed primarily of a phase mixture of two pure metals, the elastic modulus can be estimated from the rule of mixtures, based on the volume fraction of each phase present. Of the common lead-free solder alloys, only the elastic modulus of indium-tin alloys cannot be deduced from data on the respective metals but must be measured, because the indium-tin eutectic is formed between two intermetallic compounds of indium with tin. The yield strength of solders is a strong function of temperature, because thermal activation assists dislocation movement, while testing at a high homologous temperature renders the results particularly sensitive to strain rate. However, because solder generally has very low yield strength and is mostly used in a temperature regime where creep is important, this parameter is often of little practical value. More relevant is ultimate tensile strength, although measurement of this property is equally unreliable for the same reasons. Shear strength normally correlates closely with tensile strength, except where plastic ow plays a signicant role, as it does in most solders at room temperature. Therefore, for all three of these propertiesyield strength, tensile strength, and shear strengthit is probably fair to deduce that the differences between the solder alloys are small, with the exception of indium-tin alloys, which are noticeably weaker. Elongation is an important property, because high ductility enhances the low cycle fatigue resistance of a material. Also, reasonable elongation to failure at high strain rates is benecial in preventing catastrophic failures if a joint is inadvertently overstressed. The Bi-43Sn and In49Sn eutectic solders are greatly superior to Pb62Sn and Ag-96Sn in this respect. Comparative mechanical properties for selected solders are given in Table 5.6.
Table 5.6 Comparative mechanical properties for selected solders at room temperature
Elastic modulus, GPa Ultimate tensile strength(a), MPa Shear strength(a), Elongation(a), MPa %
Solder alloy
Solder alloy
Density, kg/m3
Resistivity, ohm cm
14 11 32 15
33 50 21 34
25 26 15 20
39 50 42 24
50 60 70 20
35 30 30 10
50 70 10 100
The resistance of metal to creep can be measured in many ways. Creep is a complicated process, and most solders exhibit at least three types of creep, denoted as primary, secondary, and tertiary. To provide a very rough guide to the relative ranking of different solders in their resistance to creep, stress-rupture life is a useful parameter, because there is a linear relationship between the applied stress and the life of the testpiece when plotted on logarithmic scales. Relevant data for some solders at room temperature are given in Fig. 5.3. Silver-tin solder is the most resistant to failure by creep, whereas indium-tin solder is the least resistant among the four solders represented. The fatigue resistance of different solders is one of the most relevant parameters, from the point of view of their application, but also the most difficult to determine reliably. The difficulty stems from the fact that the results are extremely sensitive to the specimen conguration and test conditions, and these factors can greatly affect their relative ranking. However, there is a general consensus that fatigue resistance of the four solders may be ranked as follows: In-Sn < Bi-Sn < Pb-Sn < Ag-Sn. Trials have shown that, for surface-mount assembly of components on FR4 printed circuit board, both silver-tin and coppertin lead-free solders are superior to lead-tin eutectic under most conditions. However, if the application involves operation above approximately 150 C (300 F), then, surprisingly, the lower-melting-point lead-tin solder actually exhibits the superior fatigue performance. This is because the almost equally divided and fairly
ne two-phase lamella structure of lead-tin solder provides a modicum of strength right up to the melting point, whereas above 150 C (300 F), the large crystals of the dominant tin fraction in silver-tin and copper-tin eutectics are less resilient to fatigue. 5.1.5.4 Corrosion Resistance
Lead-free solders generally do not suffer from corrosion in normal environments. The exception is alloys that contain zinc, because of the large difference in electrochemical potential between this metal and the other constituents. It is for this reason that zinc-containing alloys are not considered as suitable replacements for lead-tin solder for most applications. A further disincentive to the use of zinc was that, until recently, the wetting and spreading on copper surface was extremely poor, compared with lead-tin. However, recent work using uxes containing tin 2-ethylhexanoate shows considerable promise. This compound decomposes at the soldering temperature to leave metallic tin on the faying surface just ahead of the advancing solder front. By this means, contact angles as low as 20 can be achieved, which is low enough to be widely acceptable [Vaynman and Fine 2000]. 5.1.5.5 Susceptibility to Tin Pest and Tin Whiskers
Tin undergoes an allotropic transformation from its normal metallic white form, which has
Fig. 5.3
Stress-rupture life of joints made with low-melting-point solders, tested at room temperature. Silver-tin solder is more resilient than lead-tin eutectic, while indium-tin alloys are less able to resist creep.
a body-centered tetragonal crystal structure, to a powdery gray material, with a diamond cubic crystal structure, at, nominally, 13 C (55 F) (see Chapter 2, section 2.2). However, in order to initiate the transformation, it is generally necessary for white tin to be subject to strain, and the maximum rate of conversion occurs at approximately 40 C (40 F). Because the formation of gray tin from the white allotrope is accompanied by a 26% expansion in volume of the metal, the individual grains separate, and the solder disintegrates into ne powder. The transformation is termed tin pest. Clearly, with most lead-free solders having a higher proportion of tin than the lead-tin eutectic alloy they are replacing, there exists the possibility that tin-rich lead-free solders will be vulnerable to failure by this mechanism. Lead-tin solders are carefully formulated with controlled impurity additions to prevent the transformation from taking place. This is explained further in Chapter 2, section 2.2. The susceptibility of three pure tin-base solders to tin pest has been studied [Kariya, Gagg, and Plumbridge 2000]. It was found that on storage at 18 C (0.4 F), a proportion of the tin-rich phase in silver-tin, silver-copper, and tinzinc solders transformed to gray tin. The incubation period was shortest for the zinc-containing alloy and longest for the silver-containing solder. Further trials are necessary to establish whether this allotropic transformation can be suppressed by controlled doping, as it can with leadtin solder (see Chapter 2, section 2.2). Pure tin, under certain conditions, has a tendency to grow long crystals, or whiskers, that can cause electrical shorts between adjacent conductors on PCBs. The elimination of lead from solders applies equally to the solderable nishes applied to component leads and lands on PCBs. Some of the replacement nishes are tin-base. Therefore, both lead-free solders and some PCB nishes are potentially susceptible to tin whisker growth. Work is believed to be underway to assess this vulnerability, and the results are awaited with interest.
approximately 30 to 50 C (54 to 90 F) higher than for lead-tin solder. However, the upper temperature for soldering of electronics and optoelectronics assembly is xed at 260 C (500 F) by the design of existing parts. The peak temperature of 260 C (500 F), historically, gave an adequate margin for use with lead-tin solder. Lead-free solders are therefore required to wet and spread with considerably less superheat than was previously considered good practice for leadtin solder. This has two consequences:
5.1.6
Lead-free solders intended to be as close as possible form-t-function replacements for leadtin solder generally have a melting range of approximately 210 to 230 C (410 to 446 F) [Bradley, Handwerker, and Sohn 2003]. This is
processes with unacceptably narrow windows, to which the response has been the widespread adoption of nitrogen-inerted soldering. The nitrogen atmosphere retards oxidation of the solder and substrate before the ux becomes active and thereby reduces the work that the ux has to do in maintaining oxide-free interfaces ahead of the advancing solder front. If large superheats can be applied, the benets of a protective atmosphere are reduced, because solder is able to spread further and faster (Fig. 1.14 and 5.4) [Buckley 2000]. In recent years, much effort has gone into the design of nitrogen-inerted soldering equipment, and remarkably good-quality atmospheres can now routinely be achieved on open-access systems for surprisingly modest rates of gas consumption. Equipment with specications of 10 ppm oxygen, or below, in the working zone is available commercially.
5.1.7
The wetting and spreading characteristics of lead-free solders are not straightforward. There have been many studies on this topic, and much data exist in the published literature. Under certain regimes, it is possible for lead-free solders to wet and spread better than lead-tin solder, but for many it is not. The root of this complexity is that the wetting and spreading of lead-free solders is inuenced by the alloy composition, the process temperature, the process atmosphere, the ux, and the substrate metallurgy. This difficulty is exemplied by the data presented in Fig. 5.5 and 5.6, which show the time required for the wetting force to reach the acceptance value (23
of the theoretical maximum) in wetting balance tests for a range of process and materials combinations [Lau and Ricky Lee 2001]. It is impossible to discern any consistent trends in the data. Fortunately, the manufacturers of the various lead-free solder alloys have undertaken sufficient development to know what materials and process conditions are necessary to obtain wetting and spreading behavior that is at least as good as for the lead-tin process being substituted. Therefore, when making the change to lead-free soldering, it is advisable to adopt one alloy, ux, process window, and recommended component nish as a complete and integrated package. It is the authorsexperience that a surface-mount PCB assembly line can be switched from using leadtin to a lead-free solder in less than 20 min. This line was used for the mass production of circuit boards, and the manufacturing yield and inservice reliability obtained from lead-tin and lead-free solder were indistinguishable. The transition time was essentially that required to swap the solder paste print cartridges, allow the reow ovens to stabilize with a different temperature prole, and change some of the reels on the component placement machines, because some parts come supplied in a choice of lead-free or lead-tin plating.
5.1.8
Wetting speed of lead-tin solder on copper using a rosin-based ux in air and nitrogen atmospheres. Nitrogen reduces the propensity for the solder and substrate to oxidize and thereby decreases the cleaning action demanded of the ux to effect wetting.
Fig. 5.4
High-lead solders (>85% Pb) are currently granted an exception, subject to review, to the ban on lead in solders [European Union Directive 2001]. This dispensation has been granted on the grounds that there are no technically and economically viable alternatives. Lead-rich solders are used in large quantities by the lighting industry, and most discarded light bulbs are disposed of in landll sites, which provides an interesting perspective on the merits of the ban on lead-tin eutectic solder. Extensive phase diagram modeling suggests that the prospects for nding a new multicomponent solder with a melting point in the region of 300 C (570 F) are low [Lalena, Weiser, and Dean 2002]. Shimizu et al. [1999] have proposed the quaternary alloy Zn-4Al-3Mg-3.2Ga as one possibility. This alloy exhibits acceptable melting behavior (309 to 345 C, or 588 to 653 F), but there are difficulties associated with its use be-
cause of the unfavorable combination of high hardness, low ductility, and ease of oxidation. Nevertheless, this solder has a higher thermal conductivity (77 W/m K) than high-lead solders, lower thermal expansivity (20 106/K), and, moreover, foil can be produced by hot rolling at 250 C (482 F). Fluxless attachment of gold-metallized silicon die to silver-plated copper leadframes has been demonstrated with this solder using a die-bonding machine that provides for a mechanical scrubbing action during the heating cycle. The resulting joints, which measured 5 by 5 mm (197 by 197 mil), were well lled and passed standard thermal cycling and wet-high temperature storage tests without difculty. Another candidate for a lead-free highmelting-point solder is the ternary alloy Bi-11Ag0.05Ge, which has a liquidus of 360 C (680 F) and a solidus of 262 C (504 F). Although the wide melting range will be unattractive for some
applications, this alloy does have the merit that ingots can be mechanically processed by hot working into foil and wire. Molten bismuth oxidizes readily in air to an extent that renders the binary Bi-11Ag alloy unsuitable as a solder. Germanium has a higher oxygen affinity than bismuth and is largely insoluble in the binary base alloy. Because GeO sublimes only at 480 C (896 F), it is reported that at process temperatures in the region of 400 C (750 F), wetting by this solder can be obtained on nickel, silver, and gold but not copper, in air, using an inorganic aqueous ux. At room temperature, the alloy is softer than the Au-20Sn solder and possesses similar, limited ductility but has a thermal conductivity of only 9 W/m K, which may be restrictive for some applications [Lalena, Dean, and Weiser 2002]. Because of their limitation, both of these alloys are likely to be of interest only for niche applications.
Wetting rate, as measured by the time for the wetting force to reach an acceptable value, for a range of solders on different component and board nishes. There is no conclusive trend of superiority for any solder or solderable coating, emphasizing the need to tailor the lead-free process to each situation. OSP, organic surface preservation
Fig. 5.5
Fig. 5.6
Wetting rate, as measured by the time for the wetting force to reach an acceptable value, of solders on different substrates in different atmospheres. The relative ranking varies greatly, depending on the metals and process atmosphere involved.
5.2
Flip-Chip Interconnection
Flip-chip interconnection is a method of making electrical connection and providing physical attachment between two components, often a semiconductor die and substrate. Although the process was established at the birth of the semiconductor revolution, it was largely abandoned in favor of wire bonding. Flip-chip bonding was relegated to niche applications but nevertheless continued to evolve and develop to a high degree of technological sophistication. With the ongoing drive toward smaller, lighter, and cheaper electronics systems of greater functionality, ip-chip bonding experienced a renascence as an essential enabling technology. Extrapolation of current trends suggests that within ten years, the majority of semiconductor die may be ip-chip bonded. Flip-chip interconnection is being advocated or used in virtually every advanced packaging conceptball grid arrays (BGAs), chip-scale packages (CSPs), direct chip attach (DCA), and multichip modules (MCMs), to name but a few. The invention, or, more strictly, the commercialization, of ip-chip bonding is credited to IBM Corporation in the early 1960s [Miller 1969]. The process was developed at the same time as the rst integrated circuits began to appear, in response to the need to simultaneously make a large number of extremely miniaturized connections. The technique involved soldering a copper ball at every joint to provide the physical separation and electrical interconnection between the transistors on the chip and the substrate. Solder-based ip-chip technology came about from the realization that the copper ball was not required and that precise z-axis separation is maintained by the relatively high surface tension of solder alloys. This modication, known as the C4 process, standing for controlled collapse chip connection, forms the basis of the two ip-chip technologies practiced today. There now exist many variants of this process, and only the key features are elucidated in the following sections. For a more complete description of ipchip technology and methods of its implementation, the reader is referred to Lau [1996, 2000]. An overview of ip-chip technology, with particular emphasis on optoelectronic applications, is provided by Lee and Basavanhally [1994].
joined are rst patterned with mirrored, matching arrays of a solderable metal. One of these parts is then further processed to deposit interconnect metal (solder) on every pad. Either part is then ipped over (hence, the process name) and the metal patterns aligned. Finally, they are brought together to form a joined assembly. The ip-chip process ow is set out in greater detail in Table 5.7. It is described by reference to two basic inputs, an electronics wafer and a substrate, although either or both of these parts could be substituted by other parts of different functionality, such as a semiconductor component package and a PCB, as in a BGA mounting scheme. Processed semiconductor wafers are supplied coated with a passivation layer, typically either a polymeric or oxide/nitride lm, to provide some protection to the devices during nal assembly and packaging. Apertures are cut through this passivation layer at discrete points to allow contact to the top metal of the semiconductor fabrication process. This metal tends to be aluminum or copper on silicon devices and gold on III-V devices, such as gallium arsenide. The substrate will usually have a track pattern in a conductive metal such as copper or gold, with pads at the connection points. The rst step of the ip-chip process involves application of a wettable metal, or underbump metal, on the areas of the chip and substrate to be mated (Fig. 5.7). The deposit is often a multilayer metallization, the functions of which are to provide ohmic contact to and metallurgical compatibility between the contacts on the chip and substrate and the interconnect metal. Typically, it comprises a sequence of up to three metal layers. The rst metal is called the foundation metal, and its role is to provide the initial ohmic contact and a strong physical bond to the faying surfaces of the components. The choice of foundation metal varies with the nature of the comTable 5.7 Flip-chip process ow
Individual processes differ greatly in their details, but the general scheme is common to both solder bump bonding and compression bump bonding. The steps shown in parentheses are optional. Dene underbump metal pattern areas Deposit underbump metals Dene interconnect pattern areas Deposit interconnect metal Clean and inspect (Fuse interconnect deposit to form alloy) Align mating parts Bond (Underll) Inspect
5.2.1
A schematic illustration of the solder ip-chip process is given in Fig. 4.19. The two parts to be
ponents being joined and the method of deposition. Common examples are titanium and chromium, if applied by a vapor-phase technique, and zinc, if applied by wet plating. The second metal is the barrier metal and is included to prevent metallurgical reaction between the adhesion metal and the interconnect metal. The choice of barrier metal is extremely wide and includes both pure elements, such as platinum, nickel, and copper, as well as mixtures of metals, such as titanium-tungsten (oxy-nitride) and titanium nitride. The barrier metal needs to be of sufficient thickness to ensure that it remains intact during the making of the joint and the service life of the assembly, while remaining electrically conductive. The former requirement means that this barrier layer must be able to satisfy any dissolution into the molten solder and intermetallic growth in the solid state without being depleted. Finally, it is common practice to complete the underbump metal with a sacricial metal, which is often gold, so as to maintain the solderability shelf life and/or aid visual inspection. The underbump metal typically has a total thickness of less than 1 m (40 in.). One of the two parts is removed from the processing chain at this juncture, because it is usually neither cost-effective nor often necessary to have interconnect metal on both parts. The interconnect metal is mostly applied by screen printing or electroplating, although for very small bumps (<10 m, or 400 in., in diameter), vaporphase techniques are generally necessary. Historically, lead-tin alloys and pure indium were used for the interconnect metal, but the process works equally well with lead-free compositions and other solders [Lau 1996]. Finally, the components are cleaned, ready for bonding. A num-
Schematic illustration of a silicon semiconductor device prepared for ip-chip bonding. A hole has been cut through the passivation layer, and the three-layer underbump metal applied. The foundation metal, M1, is followed by a barrier metal. The interconnect metal has been deposited on top and fused to form a hemispherical bump.
Fig. 5.7
ber of direct deposition processes have been developed for applying the interconnect metal. One of the more interesting of these uses a jetting system to generate a stream of precisely sized molten droplets. These are steered by an electrostatic deection system, and an exact number of droplets can be deposited at any location over the entire area of a 6 in. diameter substrate (Fig. 5.8) [Priest, et al. 1994; Hayes, Wallace, and Boldman 1996]. The interconnect metal is frequently an alloy. Sometimes, for ease of process control, the solder is deposited as sequential layers of the constituent metals. It is then common practice to perform an additional fusion step immediately after deposition, whereby the part is heated until the layers constituting the solder melt and alloy, forming a hemispherical bump in the process. This opens the possibility of an interesting process variant, called the wet back process, whereby the diameter of the interconnect metal areas, as deposited, can be up to twice that of the underlying wettable metal pad. During fusion, surface tension forces pull the molten droplet back to the wettable pad, thereby increasing the height of the bump while limiting its width. This trick simplies the creation of particularly tall bumps. By making any tracks connected to the bump pad sufficiently narrow, no masking or other measures are required to prevent lateral solder ow along the tracks. The interconnects are not by any means restricted to a spherical geometry. Indeed, advances in photoresist technology make it possible to fabricate interconnects with aspect ratios of 5 to 1, even at tiny feature sizes. One published example shows solder columns 5 m (200 in.) in diameter and 20 m (800 in.) high at 10 m (400 in.) pitch [Yamada et al. 1998]. The column grid arrays so formed have substantially improved fatigue life compared with ball grid arrays, because the thermal expansion mismatch strain between the abutting components is distributed over a greater height of material. A threefold improvement in fatigue life is thereby realized [Sturcken 2000]. During the bonding step, the mating interconnect areas on the two components are aligned in plane, translation, and rotation and brought into contact. The connection is made by either cold compression welding, commonly referred to as indium bump bonding, or by heating to melt the interconnect metal, known as solder bump bonding (Fig. 5.9). Although ip-chip compression bonding and ip-chip solder bonding super-
Piezo crystal
Catcher
Temperature controller
CAD/CAM input
Fig. 5.8
Schematic layout of an apparatus designed to deposit individual solder balls at any desired location on a 6 in. diameter substrate. CAD/CAM, computer-aided design/computer-aided manufacturing
cially appear to be similar processes, they are implemented by exploiting different joining processes. Consequently, the resulting interconnects have signicantly different characteristics. The key features of each process and the characteristics of the resulting interconnects are summarized in Tables 5.8 and 5.9. It should be pointed out that indium and indium alloys are sometimes used as solder interconnects, and lead-tin alloys can be used as compression interconnects, so some care needs to be taken with terminology. Some companies are offering ip-chip processes based on conductive polymeric media. Generally, these have characteristics similar to compression interconnects, but they are not able to achieve interconnect diameters much below 100 m (4 mil), and of course they lack the self-alignment feature, which is a characteristic of molten metal. The self-aligning feature of solder bump bonding requires some explanation. The surface energy of a molten solder pillar between two contact pads will always attempt to minimize the surface area of the liquid. This results in a restoring force that regulates the relative locations of each mating pair of pads. The force balance derives from the contact angle of the solder onto the wettable pads, while the weight of the upper component controls the planarity and magnitude
Flip-chip interconnect schemes. The prepared components are aligned and joined by either applying pressure (solid-state compression bonding) or heat (soldering).
Fig. 5.9
of the interconnect height. A schematic illustration of the self-aligning process is given in Fig. 5.10. Typical alignment accuracy is better than 2 m (80 in.) in-plane and 0.5 m (20 in.) vertical, and these gures can be improved by nearly an order of magnitude by careful geometrical design of the interconnects and attention to process detail. Methods of achieving precise alignment between ip-chip bonded parts are discussed in section 5.2.6 of this chapter.
5.2.2
Flip-chip bonding can be dened as a method of providing simultaneous electrical connection and physical attachment between two components. It is attractive for low-cost, volume manufacturing because it is an inherently highyielding, wafer-scale process. The key features and associated benets of this technology are:
Attachment and interconnection are performed simultaneously. This provides a major cost and yield advantage over other forms of interconnection where the component must rst be rigidly xed to the substrate before proceeding with wire bonding or lead soldering. Fully automated ip-chip bonding machines are commercially available that can ip, align, place, and reow (if required) parts at throughput rates well in excess of 500/h. Self-alignment occurs between the components and substrates, through surface tension forces in the molten solder (this characteristic applies only to solder bonding, not compression bonding). Thus, optical components can be assembled in which a semiconductor laser or detector is reliably and accurately positioned relative to other parts, such as optical bers. Flip-chip interconnects provide a low and repeatable inductance, essentially a consequence of the short bond length. This charTable 5.8
Key parameter
Process characteristics
Compression bump bonding Solder bump bonding
Solid-state diffusion In, Au, Pb, Pb/Sn alloys 20200 C (68392 F) 10100 MPa (1.515 103 psi) 1:5
acteristic is particularly important for highfrequency circuits. A typical 25 m (1 mil) diameter bond wire interconnection 0.5 mm (0.02 in.) long will have an inductance of 0.25 nH, compared with approximately 0.05 pH for a typical ip-chip solder bond. Parasitic capacitance is small, predictable, and reproducible. Again, this becomes a most important issue as the operating frequency increases. While the interconnection capacitance of a typical bond wire at 150 m (6 mil) pitch is 25 fF, it is only 2 fF for ip-chip bumps at 30 m (1.2 mil) pitch. Interconnections can be accommodated over the entire area of the component. Wire bonding and tape interconnection can only be made around the periphery of chips. This restricts the total number of interconnections that can be made and requires all signal paths on the chip to be suboptimally routed out to the chip edge. Flip-chip bonding is, for similar reasons, the ideal interconnection technology for pixelated component architectures, one example of which is given in Fig. 5.11. The interconnection area is smaller than the component footprint. Close packing of digital information processing chips is necessary in order to minimize bus delay times. All ip-chip connections are located underneath the chip, whereas all other interconnection schemes tie up substantial space all around the chip periphery. This interconnection method results in low prole assembly, with ip-chip bumps just a few microns (<1 mil) high (Fig. 5.12). Therefore, the chip packages and the whole electronics system can be made correspondingly
Table 5.9
Technology characteristics
Solder bump bonding
Short (coin) interconnects Filled joint gap Fluxless Low residual stress Service temperature > bonding temperature Alignment as placed
Tall (pillar) interconnects Open joint gap Flux usually required Residual stress Service temperature < bonding temperature Self-aligning
Fig. 5.10
Fig. 5.11
(a) Semiconductor component of a pixelated imaging device. (b) Close-up view of a single solder ball. The imaging device utilizes a 16 by 128 array of 33 m (1.3 mil) diameter solder bumps for the interconnects. Each pixel is connected to its own amplier and processing circuitry electronics. Courtesy of CERN
thinner. This characteristic is particularly valuable in portable electronics products. Flip-chip processing is compatible with delicate materials. The upper frequency limit for silicon-base semiconductors is approximately 5 GHz. Above this frequency, compound semiconductors such as gallium arsenide (GaAs) or gallium nitride (GaN) must be used. Compared with silicon, these materials are extremely fragile, so that alternative interconnection techniques are preferred to mechanical wire and tape bonding. The latter can generate surface damage on the semiconductors because of the forces involved in these processes. The interconnect is packageless, and the joining medium is applied during component processing. Therefore, application of the materials required for ip-chip bonding only requires a few additional processing steps, and all components may be processed simultaneously. Removal or minimization of packaging costs is always attractive.
An array of exceptionally small ip-chip solder bumps ready for bonding. Each is mounted on a pedestal of polyamide for functional reasons. Solder bumps of this dimension can only be realized using semiconductor processing techniques. 5000. Courtesy of BAE Systems
Fig. 5.12
Enhance heat sinking of the die Extend the fatigue life of the interconnects
The process generally involves placing an accurate volume of material along one or more edges of the die and allowing it to be pulled into the gap by capillary action. The adhesive is cured by the action of time, temperature, exposure to ambient atmosphere, or a combination of these. Provided that the modulus and thermal expansion coefficient of the underll material are judiciously chosen, then the fatigue life of solder bumps between a silicon chip and an FR4 PCB can be increased by as much as two orders of
5.2.3
Underll
To improve the operational life of ip-chip assemblies, it is common to apply an underll adhesive into the gap between the die and the substrate on which it is mounted. The objectives of using an underll are fourfold:
magnitude on thermal cycling between 65 C (85 F) and 100 C (212 F) [Doi et al. 1996]. Flip-chip bonding is a rapidly evolving eld, and a number of underll materials are being offered that use different chemistries. Some of these are applied prior to assembly of the parts and actually ux the solder during the heating cycle before providing conventional postbonding functions.
5.2.4
Inspection
One of the challenging features of ip-chip bonding is that the interconnects are not easy to inspect. Some process engineers consider this to be a major benet, because it restricts the scope for quality inspectors to reject assemblies!Acomparative review of ip-chip inspection methods is given by Burdett, Lodge, and Pedder [1988]. The problem is that only the peripheral bonds can be observed visually, and then often only with some considerable difficulty. The other bonds are hidden from view beneath the components. Special microscopes have been developed that are able to look a considerable distance into the gap between bonded parts (see section 5.10.3 in this chapter). X-ray inspection is an obvious solution to this problem, stemming from the ability of x-rays to penetrate most materials. Fortuitously, the materials making up most components and substrates are lighter and less absorbent to x-rays than the solder, which therefore shows up clearly. The main defects that can be isolated using x-ray equipment include missing bumps, bridged bumps, excess bump material, and porosity in the interconnects. However, it cannot easily detect the most usual cause of yield loss in ip-chip technology, which is interconnects that have failed to produce joints, that is, leave gaps. The reason for this is that small air gaps negligibly affect x-ray absorption, particularly in comparison with solder balls. Further information on x-ray inspection is given in section 5.10.2 in this chapter. Infrared microscopy can be used to inspect certain types of ip-chip assemblies. Infrared microscopy is similar to traditional optical microscopy but uses much longer wavelengths. Silicon and many III-V semiconductor materials are transparent in this portion of the electromagnetic spectrum. Infrared microscopy permits the tracks and the backside of the underbump pads to be examined but not the detail of the bumps them-
selves, because solder bumps are opaque to infrared radiation. Arguably, the most appropriate technique for inspection of ip-chip bonded components is acoustic microscopy. An acoustic microscope generates and focuses an acoustic pressure wave into the device under examination and uses the reected echoes to construct an image of the object at a specic depth. The method is sensitive to boundaries between regions of different density and is therefore ideally suited to discriminating planar joints and transverse defects such as unwetted joints, missing bumps, cracks, and voids. Acoustic microscopy is a rapidly evolving eld, and instruments with truly remarkable resolution are available commercially. Defects in 20 m (800 in.) diameter solder bumps can be resolved in real-time. Further information on acoustic microscopy can be found in section 5.10.1 in this chapter.
5.2.5
Rework
Successful rework of a ip-chip mounted die is difficult but not impossible. Local hot gas nozzles can be used to heat an individual chip until the interconnect metal melts, whereupon the component can be lifted clear. The difficulty lies in dressing the used bond pads on the substrate so they are left in an adequate state of cleanliness and parallelism to accept another chip. For these reasons, rework tends not to be practiced for ne-pitch interconnects or with die that have been underlled, although it is perfectly practicable for larger bump bonds such as those used with BGAs, which have coarser geometries.
5.2.6
One of the remarkable features of solder ipchip interconnects is their ability to self-align. Accuracies at the 100 to 200 nm (4 to 8 in.) level, with respect to the underbump pattern accuracy, are possible. The factors that need to be considered to achieve these precisions are now reviewed. Self-alignment of ip-chip interconnects was noted almost from the inception of this soldering method, and subsequently, there was a lot of interest in its exploitation for the assembly of optical components. This interest has been reviewed by Tan and Lee [1996]. The best alignment from self-alignment is quoted as better than 0.8 m (30 in.) for four 130 m (5 mil) diameter
bumps, while Hayashi [1992] and Tsunetsagu et al. [1997] suggest that better than 0.25 m (10 in.) can be obtained with a larger number (20) of smaller bonds (25 m, or 1 mil, diameter). The best claimed alignment accuracy for systems with deliberately misaligned solder bumps pulling the device against a mechanical stop was <0.3 m (12 in.) in the stop axis but greater in other directions [Lin et al. 1993]. Attempts to use multiple stops in different orthogonal directions were unsuccessful. This is because the friction of moving the device past the rst stop is greater than the available restoring force. As stated previously, the driving force for selfalignment of solder bumps is surface tension, which seeks to minimize itself by adopting the shape of a perfect sphere, having the minimum surface area for a given volume. At that point, the surface tension is balanced by the internal pressure of the solder. Distorting the sphere, by squashing, stretching, or shearing sideways, will be resisted by a restoring force that is approximately proportional to the additional surface area created. When a solder bump is used to bond two components together, the situation is complicated slightly by the presence of at, wettable pads, plus the weight of the top chip. The solder will then try to adopt a shape as close to a truncated sphere as the volume of solder and the size and shape of the pads allow. The effect of the weight of the top chip is to slightly increase the internal pressure in the solder, and this will tend to reduce the restoring force. Software is available that allows the restoring force to be calculated [Brakke 1990], from which the following geometric design guidelines emerge for achieving good in-plane alignment:
area ratio becomes progressively more unfavorable with smaller bonds. Use smaller joints because they have slightly higher restoring force than a larger joint of the same aspect ratio. This is because a given misalignment distorts a small sphere much more than a larger one. On the other hand, because many more ball bonds can be accommodated in the same area than can larger bonds, there is less loading per joint and additional restoring force from each additional bond. Design the pads so that at the neutral position, all bonds are equally offset. Because the restoring force diminishes rapidly as the axial alignment improves, there can be insufficient net force available to move the upper component, particularly if the joint gap is lled with liquid ux. By offsetting the bonds, the available force for a given displacement is larger and is never zero. The other route of increasing the restoring force and thus the accuracy of self-alignment is to increase the surface tension of the molten metal. In fact, this can be changed reasonably easily. Lead-tin eutectic alloy has a surface tension of 0.483 N/m (2.8 103 lbf/in.) at 350 C (660 F) in inert atmosphere, compared to 0.55 N/m (3.1 103 lbf/in.) for tin at the same temperature and 0.44 N/m (2.51 103 lbf/in.) for lead. Replacing the lead with cadmium (0.59 N/m, or 3.37 103 lbf/in.), zinc (0.78 N/m, or 4.45 103 lbf/in.), or silver (0.92 N/m, or 5.25 103 lbf/in.) should increase the surface tension of the solder alloy and hence its restoring force against misalignment. The other major factor altering surface tension is the presence of ux. The surface tension of eutectic solder at 350 C (660 F) drops from 0.48 N/m (2.74 103 lbf/in.) in inert atmospheres to 0.41 N/m (2.34 103 lbf/in.) in the presence of nonactivated Rtype ux. Additions of chlorine to this ux drop the surface tension even further, so that using a 1% Cl activated ux lowers the surface tension to 0.31 N/m (1.77 103 lbf/in.). Liquid uxes should therefore be avoided and gaseous uxes used instead (see Chapter 3, section 3.3 on uxless soldering). Z-axis, or height, control is a function of the volume of the solder sphere, modied as necessary by the weight of the upper component. Achieving good z-axis control principally requires that accurate solder volumes are placed on pads whose surface areas are also well dened. These factors are primarily process con-
trol issues, although some design inputs will help. As before, many bonds mean that individual bond variations are less signicant [McGroarty et al. 1993]. For the vacuum deposition and wet back process, where solder is deposited over a greater area than the wettable metal pads, any variation in the thickness of solder deposited is exaggerated in the reowed solder bumps. Similarly, plating a thick solder deposit over a smaller part of the wettable metal will dilute these thickness errors. Solder deposition systems that work on mass change rather than thickness control tend to be superior in this context. Bond design also helps, with more spherically shaped bonds being less sensitive to deposited thickness variations than more columnar-shaped interconnects. Predicting the extent of self-alignment that can be achieved from a ip-chip structure is possible using the SURFACE EVOLVER computer code. Good agreement has certainly been demonstrated between the model and lead-tin solder bumps wetted onto copper pads [Josell et al. 2002]. This suggests that such models can be used as a tool for predicting the alignment and stand-off height of specied interconnect geometries.
5.2.8
Solder ip-chip interconnects can be made with essentially every known alloy. Step ipchip soldering is therefore both possible and practiced. Figure 5.14 is an example of a substrate that contains both wire-bonded and ip-chipmounted die. By using the high-melting-point
5.2.7
Surface Topography
One of the difficulties with ip-chip processes is that surface topography is difficult to accommodate in a reliable manner. Quite simply, the out-of-atness of the mating parts often represents a signicant proportion of the total height of the interconnects. One method of addressing this problem is to endow each of the mating bond pads with a solder land that is connected via a narrow path, as shown in Fig. 5.13. This provides a means whereby the excess solder on the rst pads to mate is progressively drawn off, reducing the total height of the interconnect. Thereby, shorter interconnects are able to form as the two parts are drawn together. The narrow land between the pad and overspill area greatly slows the rate at which the solder height can decrease and therefore ensures sufficient material is available to make all of the interconnects. Obviously, this system demands additional surface area to accommodate these solder drains, and the additional electrical capacitance that arises must be within design limits.
Flip-chip land designed to cope with surface topology. The solder is initially conned by surface tension to the central circle but will slowly ow through the constricting necks into the overspill areas. The ensuing reduction in solder volume, and hence bond height, enables adjacent interconnects of reduced height a chance to reach and wet their mating pads.
Fig. 5.13
Radio frequency die mounted using high-temperature ip-chip interconnects onto a substrate, which is itself populated with lower-melting solder balls ready for direct attach to a printed circuit board. Courtesy of Intarsia Corporation
Fig. 5.14
Au-20Sn solder (melting point 280 C, or 536 F) for this application and making the interconnects as short as possible (30 m, or 1.2 mil), it is possible to use a second ip-chip process to attach the substrate to a PCB, using, in this case, solder bumps of lead-tin eutectic solder (melting point 183 C, or 361 F).
5.3
The formation of strong joints is contingent on the ability of the molten ller to wet the joint surfaces over their entire surface area. Wetting of metal surfaces tends to be inhibited by surface oxides and contaminants. Visual inspection is not reliable for ascertaining the condition of surfaces with regard to solder wetting. The surface appearance of metallizations, in particular, can be deceptive, because color and reectivity are dependent on a variety of factors, such as grain size and type, whether equiaxed or columnar, surface texture, and any lms present on the surface. Therefore, wettability has to be determined by a direct measurement involving exposure of a part representative of a component to be joined to molten solder. These wettability measurement methods can also be used to assess the effectiveness of a ux, the sensitivity of surfaces to exposure to various atmospheres, and also shelf life, in general. The evaluation of solderability shelf life tends to involve articially accelerated testing to obtain data within a reasonable timescale. Because soldering processes are dynamic in naturethat is, they are sensitive to the time over which the ller is moltentwo aspects of solderability need to be considered, namely: The readiness with which substrates and components are wetted by a ller The extent of spreading that is obtained by the end of the process cycle, when the ller solidies A number of evaluation procedures and tests have been developed to measure these characteristics and are described in the subsequent sections. Reference is also made to the production of calibration standards of solderability.
5.3.1
Assessment of Wetting
The simplest and most popular test used for assessing wettability is the dip-and-look (DNL) test. It is a standard test method for solderability used by the electronics industry for testing the
leads of integrated circuit devices with appropriate nishes (e.g., ANSI/J-STD-002 and IPC/ EIA J-STD-003A). In this test, the component leads are immersed into a suitable rosin-type ux for 5 to 10 s and then immersed into molten solder (generally, lead-tin eutectic) for 5 to 10 s. The component leads are removed from the solder, cleaned, and then visually inspected. The percentage of the lead surface that has remained unwetted by the solder is then visually assessed. The standards specify the acceptance condition that all terminations shall exhibit a continuous solder coating free from defects for a minimum of 95% of the critical surface area of any individual termination. The chief disadvantage of the DNL test is that the conditions used differ considerably from those used in a manufacturing environment. In a typical production line, an infrared or convection furnace or vapor-phase reow chamber is used to heat the assembly, with solder paste preapplied to the joints.The dynamic thermal environment is therefore very different and the relative volumes of solder and ux very much smaller than that afforded by the test. A variant of the DNL test, called the surfacemount solderability test (SMT), was devised to more closely simulate the actual environment that surface-mount devices encounter during the solder reow process. In this test, solder paste is screened onto a thin, unmetallized ceramic plate (0.9 mm, or 35 mil, thick) using a solder stencil. The paste print used is the pattern of the footprints of the leads to be assessed. The device to be tested is then placed onto the solder paste print. Next, the ceramic substrate is passed through a reow cycle and allowed to cool. The devices are removed from the ceramic substrate, and the leads are examined for solder wetting. The advantage of this test is that leaded devices are subjected to a similar thermal environment to that experienced in the actual assembly. Furthermore, the leads receive the same volume of solder paste as in the actual assembly operation.The SMTalso neatly avoids the occurrence of solder bridging of ne pitch leads, to which the conventional DNL test is prone. This test method has been adopted as an Electronics IndustriesAssociation (EIA) interim standard (EIA/ IS-86). The new SMT test method has been shown to be especially effective for providing a reliable assessment of the solderability of palladiumplated integrated circuit (IC) leads. Palladiumplated devices have been marketed since 1989 and now account for a signicant portion of the IC
market. The palladium is applied as a 0.076 m (3 in.) plated nish over a nickel-plated lead. Its function is simply to act as a protective barrier to oxidation for the nickel. During the soldering cycle, the palladium is required to completely dissolve into the solder with the joint being formed with the underlying nickel layer [Abbott et al. 1991]. Trials using the SMT method have demonstrated that the palladium coating is completely dissolved, as designed. By comparison, the DNL test, with its short solder immersion time and absence of preheat, hinders the dissolution of the palladium from the lead surface. Neither of the two DNL-type tests described previously provide information about the dynamics of wetting, and neither is quantitative. A test that has been devised to quantitatively measure wetting as a function of time is the wetting balance solderability test.The method has been adopted as a standard test for measuring the solderability of electronic component leads and substrates under a closely specied set of conditions that include a prescribed atmosphere and ux. Solderability testers, as they are known colloquially, are available from several manufacturers, one example of which is shown in Fig. 5.15. Atypical wetting balance comprises:
and provides automatic tareing of specimen weight A temperature-controlled solder bath A bath lift or specimen fall mechanism with speed and positional control A computer to display the force/time curve and derive key metrics from the data The specimen under test is held in a holder that is itself suspended from the load cell. The bath containing the molten ller metal is raised at a preselected speed to immerse the testpiece to a given depth (on some equipment, the testpiece is lowered into the solder reservoir). The bath is held in this position for a set dwell time and is then returned to its rest position. The values of these parameters are preselected before commencing the test cycle and are dened by standards. The resolved vertical forces acting on the specimen are recorded as a function of time over the whole test cycle. Figure 5.16 shows the typical form of the trace that is recorded, together with the corresponding position of the specimen relative to the solder bath at each stage. The wetting balance provides a measurement of the vertical component of the force exerted on the testpiece as it is lowered into a reservoir of the molten solder or braze, as a function of time.
Fig. 5.15
This force, FR, is theoretically equal to the sum of the vertical component of the surface tension force, F, between the ller and the testpiece and the buoyancy of the testpiece, FB. Figure 5.17 shows an equilibrium situation appropriate to partial wetting. The resolved force in the vertical direction, FR, is the parameter measured in the test. The variation of this force as a function of time provides information on the dynamics of the wetting process. Typical wetting balance force/time traces are given in Fig. 5.18. The graphs show the effect of different cleaning methods on the wetting behavior of leadtin eutectic solder, heated to 235
C (455 F), on mild steel uxed with a mildly activated rosin ux. In Fig. 5.18(a), the steel coupon is in the as-received condition, and wetting by the solder is consequently very poor. Abrasive cleaning of the steel improves wetting, as shown in Fig. 5.18(b), but chemical cleaning is necessary in order to meet the qualityacceptance criteria, which are indicated by the box in the lower left-hand corner of the graph (Fig. 5.18c). The acceptance criterion for electronic components specied in national standards is a wetting force that exceeds two-thirds of the theoretical maximum force, achieved in a time representative of the soldering process to be used. For wave soldering, this is set at 2.1 s. A wetting balance test can be performed rapidly, and the results are quantitative, inasmuch as reproducible numerical data can be obtained for a well-dened set of sample and instrumental parameters and operating procedures, as explained in Barranger [1989] and Lea [1991]. Furthermore, the change in wetting as a function of time can be monitored. The surface tension of the molten ller can be calculated from data obtained on the wetting balance using nonwetted ( 180) substrates, such as polytetrauoroethylene (PTFE) or ceramic coupons, using the equation:
FR P cos gV (see Fig. 5.17)
Typical trace of the wetting force during a solderability test cycle, with the corresponding position of the specimen relative to the solder bath
Fig. 5.16
Fig. 5.17
Forces diagram for a solid plate partially immersed in a liquid. P, specimen periphery length; , liquid surface tension; , contact angle, , liquid density; gravitational constant, g, 9.81 m2/s; V, immersed solder volume
From this value and the measured wetting force, the angle of contact between the molten ller and the testpiece can be calculated. Attempts have been made to correlate wetting balance data with the results given by other methods used for assessing wetting [Thwaites 1981; Wooldridge 1988]. Moreover, adaptations have been made to the wetting balance for solderability testing on specic types of components and, in particular, for surface-mounted electronic devices and also in controlled atmospheres, including vacuum [Gunter and Jacobson 1990]. The only other test in common use that evaluates the dynamics of wetting is the measurement of contact angle in an enhanced version of the sessile drop test described by Matienzo and Schaffer [1991]. In that test, a xed volume of solder is melted on a at coupon of the substrate of interest, which is held at a xed temperature. A ux is generally added if the test is performed in air.The contact angle of the molten pool of solder is dynamically observed and measured. The principle of the method is illustrated in Fig. 5.19.
There is not necessarily a close correlation between solderability test measurements and the wetting characteristics of actual joints. One of the reasons is that the conguration of the test does not strictly mirror that of an actual joint. In particular, the volume of the molten ller relative to the volume of the components, including surface metallizations that react with it, may be signicantly different in the two cases. For example, the extremely fast dissolution rates of gold coatings in solder, reported for samples dipped into the reservoirs of solder in a wetting balance, typically 1 m/s (40 in./s) or more, do not apply to joints soldered using foil preforms of much smaller volume. Thus, gold coatings that completely dissolve in the solder bath can partly survive in actual joints after the same time and temperature of the bonding operation. Another difference is the nonequivalence of the thermal environment, as in the DNL test. In many materials systems, the ller metal will react differently with the testpiece as the temperature changes. This is a critical aspect because, as explained in Chapter 1, section 1.2.2, metallur-
gical reaction and not surface tension is usually the dominant driving force for wetting in liquidsolid metal systems. Because such a metallurgical reaction occurs, to an extent that depends on the materials involved and their temperature, the wetted interface changes its composition and geometry with time. This, in turn, means that the measured wetting force tends to vary in the course of a test, and only in exceptional cases is the wetting force equation strictly valid, depending as it does on the classical model of wetting. Considerations such as this limit the value of wetting balance tests for absolute measurements. However, as a means of obtaining comparative data, and, in particular, for quality-control assurance purposes (pass/fail determination), this method is most useful [Thwaites 1981].
5.3.2
Assessment of Spreading
One of the simplest and most direct methods that have been devised for assessing wetting by liquid metals on solid substrates involves measuring the area of spreading by the molten metal.
Wetting behavior of mild steel by lead-tin eutectic solder, measured on a wetting balance at 235 C (455 F). (a) As-received condition. Wetting occurs slowly and at an inconsistent rate. (b) Following mechanical abrasion of the coupon surfaces immediately prior to testing. Wetting occurs more rapidly because of denudation of the oxide scale, but the pass condition, which is a wetting force of 4.5 mN achieved within 2.1 s of immersion, is not achieved in this case. (c) Following chemical cleaning. The component solderability now satises the acceptance criterion.
Fig. 5.18
This type of basic spreading test measures wetting by a molten solder of a solid over the interface of contact and not the enhancement that occurs in narrow joints through the action of capillary forces (see Chapter 1, section 1.2.4). The procedure that is widely adopted in areaof-spread measurements is to melt a ller-metal pellet of known volume in a specied atmosphere, with or without uxes, and to allow it to spread over the surface of a testpiece for a xed period of time under controlled conditions. These conditions, where possible, should be representative of the intended application, because the spreading of the ller metal is usually sensitive to componentspecic variables (e.g., surface nish) and to process variables (e.g., time at temperature). The area of spreading of the ller metal is measured; this provides a relative index of wettability for comparative purposes. This index is the spread ratio, which is dened as:
Spread ratio (Sr) Total plan area wetted by the molten metal Original plan area of a metal pellet (of a specic geometry)
Because the geometry of the solidied ller metal is seldom perfectly circular, image analysis should be used to provide an accurate measure of the total wetted area. The technique of image analysis is capable of providing an assessment in real-time, so that with suitable instrumentation, the spread area can be monitored as a function of time on a single testpiece. A perpendicular view also enables the contact angle to be similarly monitored. A sequence of spread tests made to evaluate small changes to the composition of a ller alloy is shown in Fig. 2.25. An alternative means of determining relative spreading involves quantitatively dening a spread factor in terms of the volume, V, of ller metal used and the maximum height, h, of the solidied pool:
(6V / ) 1/3 h (6V / ) 1/3
100
where (6V/)1/3 D, the diameter of a sphere corresponding to volume V of the metal pellet before the spreading test. This term can be calculated from the density of the metal pellet and its mass. The normal method used for measuring the height (h) uses a micrometer. However, this approach introduces errors because of the necessity to subtract the thickness of the substrate, which is likely to be much greater than h, from the measured height. A more accurate method of measuring h is to determine the peak height from metallographic sections or prolometer traces. Both of these methods also enable simultaneous measurements to be made of the contact angle between the resolidied pellet and the substrate. Assuming that the initial pellet of ller can be approximated to a sphere and the resolidied ller to a spherical cap of radius R on the surface of the substrate, the spread ratio (Sr) and spread factor (Sf) can be expressed in terms of the angle of contact (), as follows:
4 cot2 /2
Sr
Principle of the sessile drop test used to assess wettability. (a) A controlled volume of ller metal (solder) is melted onto the substrate under controlled conditions. (b) The contact angle is measured with a calibrated viewnder. In an enhanced form of this test, the contact angle is recorded dynamically as a function of time.
Fig. 5.19
(Eq 5.1)
Sf 1
(1 3 cot /2)
2
(Eq 5.2)
1/3
The contact angle, , can be written in terms of the radius, A, and height, h, of the resolidied pool of ller:
sin 2
(A/h) (h/A)
(Eq 5.3)
The mathematical derivations of Eq 5.1 to 5.3 are given in Appendix A1.2. Some numerical values relating spread ratio and spread factor to the wetting angle (contact angle) are given in Table 5.10. Obviously, the situation represented by these expressions is an idealized one, and the assumptions made to derive them become less realistic with increasing solder spread. Nevertheless, these relationships do provide a reasonably close concordance with measured values. Some results of spreading tests are given in Fig. 1.14, which provides spread ratio data for a range of solder alloys melted on thin chromium metallizations covered with a ash of gold, as a function of the process temperature. The spread ratio becomes progressively more sensitive as the contact angle declines and wetting improves, whereas the spread factor varies almost linearly with contact angle, from a value of 1 at 0 to 0 at 180, as shown in Fig. 5.20. Spread ratio, therefore, provides a better differentiation between small differences in measured contact angle when the values of the latter are small. However, the converse is true when the contact angle is greater than approximately 60. The results of spreading tests over a single surface must be treated with some caution when attempting to relate them to the wetting and lling of joints. Because a joint comprises a pair of facing solid surfaces, with which the ller metal can react, the capillary forces can govern the spreading characteristics; there are hydrostatic forces to consider as well. The relevance of the spreading test is also questionable when the joints
Table 5.10 Calculated values of spread ratio and spread factor corresponding to selected contact angles. The values are derived from the expressions given in the text (Eq 5.1 and 5.2).
Contact angle (), degrees Spread ratio (Sr) Spread factor (Sf)
are made using foil preforms. For this type of conguration, it is not necessary for the ller metal to spread signicantly in order to ll the joint, so that a low spread in the test does not necessarily mean that a joint formed under similar conditions will be poor. Indeed, a high degree of spreading can be detrimental to joint lling, because the ller metal can ow out of the joint, resulting in voids and unwanted coverage of other parts of the component. Nevertheless, it is often, but erroneously, assumed that low spreading in the test necessarily implies weak interaction between the ller and the substrate and therefore poor bonding and weak joints. An additional problem in attempting to extrapolate spreading test results to predict joint quality is that the ller metal reacts and spreads over a single surface in a conventional spreading test to produce a somewhat different microstructure from that of an actual joint. This is demonstrated in Fig. 5.21. Therefore, the wetting and spreading characteristics might be different in the two cases. Large-area testpieces are needed for measuring spread parameters in situations where there is good wetting.
5.3.3
There are occasions where it is necessary to consider solderability tests from another standpoint, namely, where the solderability of the testpiece is dened and some other property is under investigation, such as occurs during the
180 90 40 10 0
Fig. 5.20
development of a new ux or when comparing solders from different manufacturers. In other words, there is a need for solderability calibration standards. Perfectly wettable and nonwettable testpieces are readily obtainable; gold and anodized aluminum are examples of the former and latter, respectively. The problem arises where there is a need for testpieces having intermediate and known degrees of solderability. The published literature cites at least two methods of preparing reference materials [Norme Francaise 1987; Lea 1990]. Neither method is entirely satisfactory. Both depend on chemical treatments to modify a clean copper surface and thereby degrade the solderability. The modied surfaces are not chemically stable, so that the testpiece must be freshly prepared before each use. Hence, there is considerable batch-to-batch variability that becomes amplied when different chemists at different sites are involved.
(a) Solder spread sample on a gold-plated substrate. The solder is Pb-60Sn, and the substrate is copper plated with 5 m (200 in.) of nickel and then 5 m (200 in.) of gold. At least three distinct microstructural bands are visible. (b) Micrograph of a joint made using the same substrate and solder described in (a). The joint has a regular microstructure, and all of the gold coating has dissolved in the solder.
Fig. 5.21
In recognition of this deciency, several teams of researchers, funded by the United Kingdom Department of Trade and Industry, were given the goal of developing a solderability standard reference material. The solution nally devised was a two-layer electrodeposit consisting of 10 m (390 in.) pure nickel, applied from a lowstress sulfamate solution, overlaid with 5 m (197 in.) of pure, soft gold [Hunt and Wallis 1995]. The substrate can be any inert material (from a soldering point of view), and nickel was selected because it does not contaminate the nickel sulfamate plating bath. Gold applied directly to nickel strip was not successful, because the composition of readily available nickel strip is not sufficiently tightly controlled. Problems were encountered with impurities in the base metal, especially zinc, diffusing through to the outer gold plating. Plating baths are formulated to very precise specications from high-purity chemicals, so the quality of the metals obtained from this source tends to be highly repeatable. The as-prepared testpieces have perfect solderability, because the wettable surface is pure gold, with the plated nickel interlayer providing an effective barrier to diffusion of impurities from the nickel substrate to the gold. Because the gold coating is thick and the diffusion of nickel through gold at room temperature is negligible, the reference standards have a long shelf life. Trials indicated no change in solderability when stored for a year in an open laboratory environment. In order to degrade solderability of the testpieces in a controlled manner, the plated material was heat treated in air for 2 h. The long heat treatment time was chosen to ensure that errors arising from the heating and cooling stages, which tend to be difficult to control, are minimal. It is desirable to load plated sheets directly into a preheated furnace and, at the end of the allotted time, remove them to ambient to achieve the fastest possible temperature change without introducing surface contamination. The heat treatment temperature needs to be quite tightly controlled, because the tightly grouped process temperatures of 295, 310, and 328 C (563, 590, and 622 F) yield three intermediate degrees of solderability. Because of the small temperature interval between these three temperatures, some care needs to be taken to ensure that the furnace temperature is calibrated and the reference material experiences the specied set-point temperature. During the heat treatment, nickel will diffuse through the gold and, on reaching the free sur-
face, will oxidize. Because nickel oxide is only removed by the most aggressive of uxes, the reference material has variable solderability, depending on the proportion of nickel oxide in the gold. Figure 5.22 shows the concentration of oxygen (atomic percent) in the gold at a depth of 3 nm (0.1 in.). Figure 5.23 represents the wetting force measured for each solderability reference standard using three commercially available uxes having different activities. The results were found to be consistent, with scatter bars of 2 mN (0.007 ozf) able to fully accommodate all of the experimental results obtained from an interlaboratory comparison involving ve sources of solderability reference material, six test laboratories, and three different makes of solderability test equipment. The uxes used in all the trials were of the same type, obtained from one manufacturer, but were from different batches and
dates of production. This reference material with the plated layers therefore fullls the requirement of being a readily reproducible calibration standard for assessment of solderability.
5.4
Amalgams as Solders
Oxygen concentration at 3 nm (0.1 in.) depth in the gold surface of solderability reference standard material as a function of the heat treatment condition. The substrate material is nickel sheet, electroplated with 10 m (390 in.) nickel, overlaid with 5m (200 in.) gold.
Fig. 5.22
An amalgam is a mechanically alloyed mixture of liquid metal and solid powder. They are usually designed such that metallurgical reaction at room temperature or when slightly warmed results in isothermal solidication, and the mixture sets solid. Amalgams therefore merit consideration as solders. The molten constituents are a low-melting-point metaleither mercury, gallium, or indiumor a mixture of these, while the solid powder contains metals having a considerably higher melting point. A list of liquid metals and solid powders that have been explored as amalgams is given in Table 5.11. Because amalgams initially exist as pasty uids, they offer a number of unique advantages over conventional solders. The bonding process occurs at low temperature, without ux, so that it greatly reduces the equipment complexity and confers much process exibility. In particular, amalgams can accommodate large out-of-plane engineering tolerances, because they are semisolids, but yet can be dispensed precisely at the bond location. Furthermore, because amalgams take a nite time to set, often many tens of minutes, and are electrically conductive while molten, a product can be tested for electrical functionality while the amalgam is still liquid, with time to replace any faulty components before the bond is cured. This property is also likely to be of benet in the assembly of optoelectronic devices, where it is frequently necessary to underTable 5.11 Solid and liquid metals that have been evaluated as amalgams
Solid powders Liquid metals Melting point C F
Wetting force at 2 s, determined using a wetting balance, for three commercial uxes, as a function of the heat treatment condition used to produce the solderability reference standard
Fig. 5.23
Antimony Chromium Cobalt Copper Germanium Gold Iron Nickel Manganese Palladium Platinum Silver Vanadium
39 29 159 5 15 16
38 84 318 41 59 61
take active alignment before the parts are xed in place, yet one does not want the further movement that can occur when a higher-melting-point solder solidies and cools. Amalgams are also suited for integration in step-assembly processes, because the melting point of a solidied amalgam is typically 200 to 600 C (390 to 1110 F). Therefore, multiple joints can be made sequentially using the same process, which can be carried out below the upper service temperature of the product. The liquid component of common amalgam systems also possesses the characteristic of being able to wet directly many nonmetals, including ceramics, glass, and, of course, tooth enamel. There have been some attempts over the years to devise amalgam systems for electronic assembly. These have met with only partial success. It is not clear from this work whether there is a fundamental limitation that will preclude the widespread availability of amalgams with favorable functional properties and process characteristics or whether it is simply a lack of research effort. Certainly, dental amalgams have beneted from very detailed study and, as a result, are highly effective for the function for which they are designed.
5.4.1
Dental amalgams have been in existence since before 659 A.D.; the alloy rst appears in a Chinese book with a title translated into Latin as Materia Medica, written by Su Kung in the fourth year of the Tang emperor Hsien Ching (659 A.D.) [Chu 1958], but the formulation currently used in restorative orthodontology was not devised until the early 20th century. The basic method for making amalgam involves a diffusion reaction between silver powder and mercury in a ratio such that all of the mercury is eventually consumed in the formation of an intermetallic compound. The reaction may be simply written as:
2Ag 3Hg Ag 2Hg 3 (melting point 127 C, or 261 F)
tion of the liquid mercury and also results in an alloy that has a small, controlled expansion on setting. This is necessary to effect a good seal to the walls of the tooth cavity but obviously must not be so great as to cause pain or crack the tooth. Copper is added to tie up the tin as Cu6Sn5 intermetallic compound, because this phase is proven to have superior mechanical properties and resistance to corrosion by oral uids than the HgSn7 phase that would otherwise prevail. Amalgam metallurgy is a relatively complex subject, and a whole family of alloys is available with varying proportions of minor elements to tailor the properties of the amalgam for its position in the mouth and the type of cavity it is being used to ll. The mechanical properties of mercury-base amalgams cannot be expressed in the same language as used in mainstream metallurgical technology without distinct risk of misinterpretation. For example, the tensile and compressive behavior are totally different, and the creep behavior does not correspond with conventional behavior. The most accurate description of mechanical properties is obtained when using rheological concepts, which are applicable to viscoelastic materials. The deformation process involves dislocation climb and grain-boundary sliding, and, when overstressed, a dental amalgam fails by intergranular brittle fracture [Waterstrat 1990]. A stress-strain curve for dental amalgams at low strains is given in Fig. 5.24. The mercury in dental amalgams is bound up in the Ag2Hg3 intermetallic compound, but there is concern that some mercury might be slowly released through extended contact with saliva. Mercury is classed as a hazardous substance. Efforts have been underway to develop resin materials for dentistry that are color-matched to teeth, but despite recent advances in the technology, these remain inferior to conventional
The intermetallic compound is often referred to as the gamma phase in the solidied alloy. Modern amalgams contain approximately 25% Sn. The presence of tin accelerates the consump-
Stress-strain curve for a dental amalgam (mercurysilver base) at low strains. Adapted from Dickson, Oglesby, and Davenport [1968]
Fig. 5.24
amalgams, particularly in their durability. They mostly contain beryllium, which is also hazardous in some forms, including the oxide. The requirements of an amalgam intended for use as a solder are very different to those used for dentistry. A dental amalgam works by lling a specially shaped cavity in the tooth that is created by the dentist. On curing, the amalgam swells slightly to lock the cohesive mass of metal in place. By contrast, a solder must wet and join two parallel surfaces. Also, a dental amalgam must set in minutes and attain useable strength within 2 h. The amalgam used as solder must afford sufficient persistence of uidity to give sensible bench life, yet only slightly elevated temperature is desired to accelerate the hardening process. Nevertheless, the detailed understanding of mercury amalgams that now exists may facilitate the development of amalgams for manufacturing based on gallium and indium.
5.4.2
Gallium melts at 29 C (84 F) and is therefore a potential base for formulating very lowprocess-temperature amalgams without the toxic hazard associated with mercury. Gallium is not a particularly expensive metal, having a comparable price by weight to solder pastes. Most gallium amalgams exhibit signicant volume change during curing, but certain compositions have volume change that is similar to mercury amalgams. Indeed, because of toxicity concerns with mercury, gallium-base amalgams have been developed for dentistry and are commercially available on the Japanese and Australian markets [Reusch, Geis-Gerstorfer, and Zeigler 1988]. Over the typical service temperature range of electronics equipment, the electrical and thermal conductivities of gallium amalgams are similar
to other solder alloys, as can be seen from the list of properties cited for a range of gallium-nickelcopper amalgams in Table 5.12 [Mackay 1993]. Gallium forms amalgams with several metal powders, of which copper and nickel are the most intensively studied. Unfortunately, the published literature does not elucidate on the important parameters of powder condition, particle size and shape, nor often the amalgamation method. However, from the available micrographs, it is possible to deduce that the nickel and copper powders used with gallium amalgam were probably spherical and roughly 25 m (1 mil) in diameter. Gallium is also known to form amalgams with silver, and it is regrettable that this potentially interesting combination has not been evaluated further. To form a gallium amalgam, a mechanical mixture of the constituent metals is warmed until the temperature reaches 35 C (95 F) and then thoroughly blended. A convenient method of doing this is in a commercial dental amalgamator, which contains a pestle to assist in obtaining an intimate mixture of powder and liquid in a short period of time. The amalgam will then directly wet most common metallizations used in the electronics industry and also many nonmetals, including alumina. Due to the high proportion of solids in the amalgam, some mechanical assistance in the form of scrubbing or wiping is necessary to spread the mixture. Although gallium alloys will cure at room temperature, to achieve acceptable process times, the use of elevated temperature is benecial. The graphs in Fig. 5.25 show the cure behavior of Ga-5Ni-30Cu amalgams at different temperatures (for comparison, solidied lead-tin eutectic has an equivalent hardness, on the same Shore Durometer scale, of approximately 80). When fully cured, high-strength and welllled joints with good fatigue resistance can be
Table 5.12
Property
Set-up time at 35 C (95 F), min Curing time at 100 C (212 F), min Expansion on curing, % Shear strength, MPa (psi) Elongation, % Electrical resistivity, ohm cm Thermal expansivity, 106/K
(a) Balance gallium
obtained, which gives an indication of the potential of this soldering method. Gallium amalgams have been demonstrated as ller metals in a number of electronics-oriented applications, including semiconductor die attach, ip-chip interconnection, and via-lling in PCBs. In the latter example, the amalgam was applied by screen printing [Bhattachatya and Baldwin 2000]. However, it is generally the case that the results to date show considerable scatter in terms of joint quality, and further research is required to develop processes satisfactory for industrial use [Baldwin, Deshmukh, and Hau 1996].
joints to silver-metallized surfaces, and therefore, further endeavor in this area may prove to be fruitful.
5.5
Strengthening of Solders
5.4.3
Indium is another liquid metal that can be considered as a base for amalgam systems. Particularly after alloying with tin and bismuth, the melting point of the liquid can be depressed to below 100 C (212 F). Because indium alloys melt at temperatures that are substantially above the melting points of mercury and gallium, it is potentially possible to premix indium amalgams using liquid indium. Then, the curing reaction can be suppressed temporarily by quench cooling the liquid-powder mixture until it is needed. The authors are only aware of attempts to form indium amalgams using silver powder. As with the gallium amalgams, the very restricted trial did not yield a useable mixture, which was attributed to the availability of only relatively coarse crystalline silver powder. Amalgams appear to be more successful if the powder used is ne and spherical. Nevertheless, the mechanical properties of the silver-indium intermetallic compounds are known to be favorable in indium
Fig. 5.25
Compared with most metals and alloys in everyday use, soft solders have intrinsically poor mechanical properties, that is, low strength and inferior resistance to creep and fatigue. For the commonly used lead-tin solders, application temperatures can easily be up to 90% of the melting point, expressed in Kelvin. This means that the microstructure of solidied lead-tin joints tends to be unstable under typical service conditions, and frequently, grain growth will occur, to the detriment of the creep and fatigue resistance of joints. Many solders are also susceptible to stressinduced microstructural changes. The superposition of these microstructural changes on a cyclic strain regime results in concentrations of shear bands developing in the alloy microstructure. Such features are prone to initiating fatigue cracks. Hence, solders differ from common metals in that fatigue failure is initiated internally, whereas in engineering structures, fatigue cracks almost always start at an exterior surface. The poor mechanical properties become most apparent when the joint gap is wide (>25 m), so that a proportion of the ller metal through the joint width is free from the constraint of the parent materials. It is precisely because the failure mode of soldered joints in electronics systems is so complex and depends on many interrelated variables that the test methods used to assess their integrity and reliability are specic to that industry, with the testing usually being carried out on fabricated assemblies. Further details are given by Coombs [1988]. So long as the primary function of solders has been to provide electrical contact, their mechanical weakness could largely be tolerated, although creep and fatigue failure arising from thermal expansion mismatch of abutting components have caused occasional reliability problems. This was the situation in electronics prior to the adoption of surface-mount technology. This development and, in particular, the reliance on support structures of solder, coupled with the continuing trend toward miniaturization, have focused attention on the mechanical weakness of soldered joints.
The key to boosting strength and, in particular, conferring fatigue resistance to joints made with solder is the development of a thermally stable and ne-grained microstructure. Three strategies have been pursued in an attempt to address this need:
C, or 370 F) and 3.5Ag-95.5Sn-1Zn (melting point, 217 C, or 423 F) containing additions of silver or copper, respectively, as the source of the heterogeneous phase at concentrations in the region of 0.1 to 0.5% [McCormack and Jin 1994]. This method of improving resistance to failure by creep or fatigue of soldered joints has the benet that it is simple to achieve, because it merely involves controlled doping of the solder during manufacture.
5.5.2
Oxide-Dispersion-Strengthened Solders
5.5.1
Grain Renement
Grain renement is conventionally achieved by small additions (0.001 to 0.5%) of selected elements, such as lithium, beryllium, indium, and gallium, to lead-tin solder [Klein Wassink 1989; Wade 1999; Tribula and Morris 1990]. This approach operates by creating a ne dispersion of either oxide or nitride particles or stable intermetallic phases when the solder is molten. These then act as sites from which the solid phases can grow on solidication (heterogeneous nucleation). The presence of a large number of such ne particles when the alloy is molten is reected in a ne alloy microstructure. The grain size that can be achieved is typically 150 m (6 mils), compared with 300 m (12 mils) for pure leadtin solder solidied at the same rate. Although grain renement is benecial to the mechanical properties of metals, improving both strength and ductility, the ne microstructure is not thermally stable and will gradually coarsen by solid-state diffusion when thermal energy is provided by exposure to elevated temperature (typically, above 75 C, or 165 F). As can be seen from the example given in Fig. 5.26, the creep-rupture time of a solder doped in this manner is an order of magnitude longer than that of regular lead-tin solder. Similar improvements to the creep properties have been achieved for other binary tinbase solders by grain renement [McCabe and Fine 2000]. Similarly, examples of ternary alloy solders, which have been shown to benet from grain renement accomplished by minor additions, include 5In-87Sn-8Zn (melting point, 188
Dispersion strengthening involves the incorporation of ne and intrinsically insoluble particles into the solder, usually by mechanical means. Examples are TiO2, SiO2, and Al2O3, all of which are cheap and inert compounds (notwithstanding the normal hazards associated with handling dust) and readily available as ne powder with precisely controlled size distributions. The particle size is typically 0.1 m (4 in.) diameter but frequently much less. The loading of refractory compound in the solder is typically 3% by volume. These dispersoids provide some grain renement on solidication of the solder by heterogeneous nucleation, but their principal role in improving the mechanical properties is by other means. First, they inhibit coarsening of the microstructure; because they are insoluble in the solder, it is thermodynamically favorable for the particles to reside at regions in the alloy matrix where there are natural departures from a regular atomic lattice, such as the boundaries between the tin- and lead-rich phases in lead-tin solder. Secondly, provided the particles are sufficiently ne, they will, for similar reasons, impede the movement of dislocations through the solder ma-
Creep curve for lead-tin eutectic solder and a dispersion-hardened equivalent alloy containing 0.5 wt% Ag, 0.5 wt% Sb, 0.1 wt% Cu, and 0.003 wt% Ga at a constant stress of 10 MPa (1450 psi) and a test temperature of 60 C (140 F)
Fig. 5.26
trix. Dislocations are defects in the atomic lattice that can be induced to move through it by the application of mechanical stress; the movement of many such dislocations results in plastic ow of the material. Hence, by restricting the movement of dislocations, the mechanical properties of the solder are enhanced. This type of dispersion is less sensitive to thermal degradation, and therefore, the strengthening effect is more resilient to elevated temperatures than is simple grain renement from metallic precipitates. Table 5.13 demonstrates the boost to the creep resistance of lead-tin eutectic solder that can be obtained even at elevated temperature through dispersion strengthening. To be effective obstacles to dislocation movement, the particles must be within a certain size range, be stable in size and interparticle spacing, and have a higher ow resistance than the matrix. Dispersion strengthening is quite well understood, from a theoretical perspective, and the properties of suitable dispersoids can be calculated from rst principles. It transpires that ne oxide particles are a good choice for inclusion in solders. Because the dispersoids are insoluble in the solder, dispersion strengthening remains effective even when the alloy is heated almost to its solidus temperature. No change in microstructure is reported even after heating for 48 h at 120 C (248 F). Furthermore, because the particles are unreactive toward the constituents of the solder, the dispersion strengthening occurs equally in both the tin-rich and lead-rich phases. This results in the modied solder exhibiting improved ductility, tensile strength, and resistance to creep [Mavoori and Jin 1998]. With a sufficient dispersion of nanosized TiO2 particles, it has been possible to boost the creep resistance of soft solder at room temperature to a level comparable to that of the Au-20Sn alloy and increase its tensile strength fourfold, albeit with a corresponding reduction in ductility [Mavoori and Jin 2000]. One useful side benet from stabilization
of the alloy microstructure is that it helps signicantly with simplifying reliability prediction modeling. The principal impediment to the widespread adoption of dispersion-strengthened solders is the difficulty of manufacture. In order to be effective, the particles need to be present as a ne and uniform dispersion in the solidied joint. The problem is that particles tend to agglomerate, trap porosity, and adversely affect the viscosityand hence, spreading abilityof the solder. While solder alloys can be produced satisfactorily on a small scale in the laboratory, the specialized equipment and processes required for their preparation and use currently preclude them from being drop-in replacements for conventional solder in most applications.
5.5.3
Composite Solders
Composite solders are not fundamentally different from the other two types of strengthened solders described previously, in that they are conventional solder alloys with improved mechanical properties that arise from the presence of small, hard particles. Thus, the strengthening mechanisms are as mentioned previously, namely, a combination of grain renement, grainboundary pinning, and impediment of the movement of dislocations. What differentiates these materials is the choice of particle to provide the reinforcement, being predominantly copper-tin (Cu3Sn, Cu6Sn), silver-tin (Ag3Sn), nickel-tin (Ni3Sn4), or copper-nickel-tin (Cu9NiSn3) intermetallic compounds [Guo and Subramanian 2002; NEPCON West 1992; Betrabet, McGee, and McKinlay 1991; Marshall et al. 1991]. These additions satisfy the following conditions [Guo and Subramanian 2002]:
the molten solder at normal temperatures used for reow, so that the reinforcements are essentially stable during reow and aging. The density of the reinforcements and solder matrix are sufficiently similar so that gravitational separation does not occur, and the mixture remains homogeneous when the solder is molten. These reinforcing phases do not appreciably coarsen during normal service.
They tend to retard the growth of the intermetallic layers at the solder/substrate interface. Although such layers are essential to successful bonding, their growth in thickness results in deterioration in the mechanical properties of joints. Composite solders come in two avors. The more common contains ne particles (<5 m, or 0.2 mil, in diameter) at low volume fractions (0.5 to 20 vol%) to produce grain renement. However, it has been found that the reinforcing particles are most effective in stabilizing the microstructure of the solder when they are of the order of 1 m (40 in.) in size. Figure 5.27 and Tables 5.14 to 5.17 illustrate some of the reported benets to mechanical properties. In general, metallic additions are less effective in boosting creep resistance of soft solders than lower percentages of dispersed oxides (compare Tables 5.14 and 5.13). Metal additions have a benecial effect on creep resistance to high cycle fatigue but are detrimental for low cycle fatigue (Tables 5.15, 5.16). Copper particulate reinforcement is superior to silver from the point of view of creep resistance (Table 5.14). However, in situ Cu6Sn5reinforced solders are superior to those strengthened with copper or silver. Solders containing distributed Ni3Sn4 particles are better still, in respect of high cycle fatigue and tensile properties (Tables 5.15, 5.17). A second and less common type of composite solder is lled with larger particles (typically 5 to 25 m, or 0.2 to 1 mil, in diameter) at volume fractions of 10 to 40%. These are akin to metalmatrix composite materials and benet from enhanced tensile strength, resistance to creep, and improvements to other mechanical properties at the expense of ductility and ease of application. The combination of higher volume fraction and larger particle size is effective in strengthening
Yield strength of composite solders at room temperature plotted as a function of volume fraction of the added intermetallic powder [Yost, Hosking, and Frear 1993]
Fig. 5.27
the solder matrix by restraining the yield behavior of the soft matrix. The material combinations that have been evaluated are largely the same as those described previously for the rst type of composite. To date, both types of composite solders have been difficult to use in joints without a substantial concentration of micropores developing, which offset much of the gains in properties outlined previously. Indeed, while there is a reasonable body of literature largely extolling the mechanical properties of composite solders, the data almost exclusively pertain to specially prepared bulk samples of the ller metal and not to the properties of joints of typical geometry and aspect ratio. There are three common methods by which composite solders are prepared: powder blending, mechanical mixing, and rapid solidication. In powder mixing, powder of the reinforcement of interest is simply stirred into a vat of molten solder prior to use. The difficulty with this method is that it is difficult to obtain a uniform dispersion, and small particles tend to dissolve and reprecipitate on the surface of larger particles, resulting in a relatively coarse dispersion. Mechanical mixing entails placing solder powder and powder of the reinforcing constituent in a ball mill and working the mixture until the desired distribution is obtained. Obviously, some of the benets of preparation in this manner will be lost when the solder is melted to effect joining. The third method involves gas atomization of essentially off-eutectic composition alloys. The reinforcement arises from the ne dispersion of primary and secondary phases of intermetallic that precipitate out on cooling. Again, some coarsening is inevitable when the solder is melted in the joining operation. Mechanical mixing and gas atomization yield powder that then needs to be added to a binder and ux to form paste. Composite solders generally have inferior wetting and spreading characteristics, compared to normal solders, and the visual appearance of the joints is impaired. A frequent and very apt description is that the solder is gritty. Wetting balance tests conrm that the wetting time of composite solders is affected detrimentally and the spreading is decreased by as much as 25% [Steen and Becker 1986]. However, the wetting angles of molten composite solders, in sessile drop tests using ux, are not appreciably higher than the unmodied alloys [Subramanian, Bieler, and Lucas 1999], as shown in Table 5.14. Proponents of composite solders argue that the poor
Table 5.14
Solder (fluxed)
Wetting and creep properties of composite solders reinforced with metallic additions
Average wetting angle on copper substrates, degrees Creep rate at a steady stress of 17 MPa, s1 25 C 65 C 105 C
Sn-3.5Ag 15 vol% Cu 15 vol% Ag 20 vol% Cu6Sn5, (produced in situ from added Cu and Sn in the solder)
Adapted from Guo and Subramanian [2002]
2.62 105 4.70 106 1.91 105 7.6 106 (at 25 MPa)
1.50 104 2.03 105 8.37 105 9.8 105 (at 13 MPa and 85 C)
1.9 103 3.95 104 1.80 103 5.8 104 (at 12 MPa and 125 C)
spreading is not a great impediment for electronics assembly applications where solder (paste) is usually preplaced and minimal further spreading is required to effect joining. The stiffer rheology of the semimolten composites results in wider joints than for conventional solders, which may be advantageous where there is a desire to bridge gaps, although wider solder joints are intrinsically weaker, counterbalancing potential benets to mechanical properties. Wide joints are a decided disadvantage for most electronic and optical assembly because of the relatively poor electrical and thermal conductivity of solTable 5.15 High cycle fatigue life (expressed as cycles to failure) of composite solders, determined at different stress levels, reversed 30 times/min (0.5 Hz)
Solder 28 MPa 34 MPa 42 MPa
(a) Cu6Sn5 dispersoids, 0.42 m (16.5 in.) in diameter. (b) Ni3Sn4 dispersoids, 0.25 m (10 in.) in diameter
Table 5.16 Low cycle fatigue life (expressed as cycles to failure) of composite solders, determined at strain rates of 0.05 and 0.005 Hz and a total strain of 1%
Solder 0.05 Hz, 1% strain 0.005 Hz, 1% strain
2100 460 75
ders, compared with metal conductors, notably, copper, silver, and gold. The improvements in materials properties resulting from the inclusion of dispersoids in the ller metal do not greatly expand the possibilities for using lead-tin solders in load-bearing applications. This is because the reinforcing processes are only effective at low strain rates, as can be seen in Fig. 5.28. At higher strain rates likely to be experienced in load-bearing structures (102 s1 and above), properties such as tensile strength are no different between composite and conventional solder [Mavoori and Jin 1998, 2000]. Furthermore, although the dispersoids improve the resistance to creep, high cycle fatigue life, and stress-rupture life of solder, it is mostly at the expense of ductility, low cycle fatigue life, and fracture toughness, which are properties likely to be of greater importance in structural applications. An interesting variation of composite solders has been proposed that utilizes insoluble particles in the form of iron powder. Iron is wetted, but not consumed, by alloying at a particularly fast rate by tin-base solders and therefore provides a similar degree of reinforcement as other dispersants. However, because iron is a soft magnetic material, the natural shape of the solder, when molten, can be altered by application of an external magnetic eld, and the modied prole will be frozen when the solder solidies. Fields in the region of 0.05 to 0.5 T (500 to 5000 G) will produce signicant height change of molten solder spheres. It is suggested that this effect might be exploited to help remove the effects of joint
Table 5.17
Solder
35 52 63
37 60 73
12 21 24
48 18 15
331 37 37
Fig. 5.28
Tensile strength of lead-tin eutectic solder with and without 3 vol% of Al2O3 dispersoids as a function of strain rate, measured at 80 C (176 F)
gap variation when attempting to make multiple joints in parallel, for example, ip-chip interconnection [McCormack, Jin, and Kammlott 1994]. Iron-containing composite solders could be of interest for microelectromechanical systems (MEMS) fabrication. At the time of writing, neither grain-rened, dispersion-strengthened, nor composite solders have yet been adopted in commercial use by industry. Although they are undoubtedly attractive for some applications, because of the problems outlined previously, there is clearly a requirement for further research before the transition can proceed with condence.
5.6
Reinforced solders are ller metals that incorporate a reinforcing medium that is physically large in relation to the joint width. On a simplistic level, they can be considered as visually equivalent to the use of steel bars within reinforced concrete constructions. Because the strengthening mechanism is bulk physical constraint of the solder by the reinforcement, it is fundamentally different from the mechanisms for strengthening solders described in section 5.5 of this chapter. Reports in the published literature reveal some of the results of attempts made to load solder alloys with a uniform distribution of strong (highmodulus) particles or bers (typically, 100 m to 1 mm, or 4000 in. to 0.04 in., in size) of non-
metallic materials. Refractory metals have been tried, but these tend to agglomerate when the ller is molten, because of their higher density [Ho and Chung 1990]. The most successful avenue to date has been to incorporate into the ller chopped carbon bers, electroplated with nickel or copper so they are wettable by solder. With lead-tin eutectic solder, the results show a threefold enhancement of the shear and tensile strength of the joints with respect to the unmodied llers and, more particularly, a signicant reduction in the thermal expansivity of the ller. The published data show that the tensile strength of joints formed with lead-tin with reinforcement can be as high as 65% of the ruleof-mixtures for the relative proportions by volume of solder and reinforcement present. Approximately 15% by volume of bers is the maximum that can be incorporated into the solder while retaining acceptable workability in the molten state. With this level of ber loading, joint strengths in excess of 250 MPa (5.2 lb/ft2) at room temperature can be achieved [Ho 1996]. An alternative approach to attempting to ll a joint with a molten alloy loaded with insoluble bers is to pack the joint with dry bers and use a conventional solder, with or without ux, as required, to inltrate and ll the interstices. Obviously, the bers need to be metallized so they can be wetted by the solder. Much higher loadings of reinforcement can then be achievedup to 55% of the volume of the solder. As alluded to previously, because carbon bers have a small but negative coefficient of thermal expansion, at 42% by volume, the thermal expansivity of the solder composite declines to zero over the tem-
perature range of 25 to 100 C (77 to 212 F). Solders reinforced in this manner might therefore offer benets where there is a requirement to join low-expansivity materials, such as ceramics to metals, where a wide joint gap is mandatory. The high thermal expansivity of a conventional solder alloy (typically, >20 106/K, or 20 ppm/K) introduces a shear stress at the component/solder interface, which is overcome by using a carbon-ber-loaded solder. The reduction in the thermal expansivity brought about by the ber addition reportedly accounts for a threefold enhancement in fatigue life on thermal cycling that is observed in bonded assemblies of this type. An additional benet of reinforced solders is that carbon bers can have very high longitudinal thermal conductivity, so the presence of a mat of bers in a joint can help to redistribute local hot spots within the plane of the joint. Clearly, there is scope for further research in this area. When contemplating using ber-reinforced solders, one of the key targets is to obtain a void-free joint; otherwise, poor joint lling mitigates the strengthening effect. This end is greatly assisted when the inltration of solder into the ber bundle is promoted not only by metallurgical wetting of the solder, but when surface tension forces are exploited to achieve spontaneous inltration into the interstices. This has been studied from a theoretical standpoint, albeit simplied, and some of the key results are presented in Table 5.18. In summary, provided that the wetting angle of the solder to the metallization on the reinforcement material is below 45, then spontaneous inltration should take place irrespective of the aspect ratio of the reinforcement. If the reinforcement medium is not closely packed, then the critical wetting angle decreases accordingly. The corollary is that unless the minimum conditions given in Table 5.18 are achieved,
the resulting joint will contain voids, unless external pressure is applied to force the molten metal into the interstices of the reinforcement material [Yang and Xi 1995].
5.7
This section considers methods for quantifying the mechanical integrity of joints and predicting their dimensional stability under specied environmental conditions. These are really two independent issues that are addressed separately. Although there have not been recent signicant advances in methods of measuring the mechanical properties of solders and joints, values for various material parameters, such as Youngs modulus, are required for numerical modeling techniques. It is therefore important to have an appreciation of the limitations inherent in the derivation of parameters and the scope of their applicability. Modeling of the lifetime of joints, when subject to cyclic conditions, has made considerable progress in recent years and is also considered in this section.
5.7.1
Measurement of the mechanical properties of bulk solder specimens and soldered joints should be an uncomplicated process. Suitable test methods, specimen designs, and methods of preparation for metallurgical specimens are all dened by standards. However, even obtaining what might appear to be relatively straightforward property data on bulk solders is fraught with problems. A few of the factors that inuence the mechanical strength of solders include: Test method used Procedure used to prepare the samples
Table 5.18 Calculated critical angle for a liquid to spontaneously inltrate the interstices in selected close-packed structures, and the minimum packing density necessary to achieve lling even with perfect wetting. Above the minimum packing density, spontaneous inltration is relatively easy to achieve, even when the wetting is relatively poor.
Critical wetting angle for spontaneous inltration, degrees Minimum volume fraction of reinforcement for spontaneous inltration at 0 contact angle, %
Reinforcement type
Unidirectional bers Body-centered cubic packed mono-sized spheres Face-centered cubic/hexagonal close-packed mono-sized spheres
45 65 50
40 20 40
Precise sample geometry Test temperature Strain rate employed in the test Microstructure of the ller metal
These sensitivities help to explain why published values of the properties of solders can vary by at least an order of magnitude for what, to the inexperienced observer, might otherwise appear to be identical samples [Plumbridge 1996]. The complexity of solder behavior, especially that of the softer indium-containing alloys, arises partly from the fact that at room temperature these alloys are working close to their melting point, expressed in Kelvin. This means that atomic diffusion can occur rapidly, and hence, the processes of annealing, alloying, and precipitation are observed during testing, in addition to conventional bulk metallurgical phenomena such as necking. In actuality, the response of a solder to mechanical stress is a complex combination of elastic, anelastic, and viscoplastic behavior. Obtaining consistent values for the mechanical properties of soldered joints is even more difficult. In addition to the variables cited previously in respect to bulk solders, the strength of a joint is additionally inuenced by a number of factors, including:
Dimensions (thickness and area) Method used for making the joint Heating excursion (integrated temperature
and time) Cooling rate Substrate materials Impurity content of the ller Age of the joined assembly
made to meet load-bearing requirements but to effect electrical connectivity, thermal conductivity, or a hermetic seal between components. Correlation between these variables and basic mechanical properties (e.g., tensile and shear strength) is often obtuse. Failure of the joined assembly to pass a die-shear test is only an indicator that it falls short of a minimum requirement. A joint can possess high levels of voids and therefore possess impaired local thermal conductivity and leakage paths, giving rise to hermeticity failure long before this is reected in its shear strength. In the electronics and photonics industries, the difficulty of obtaining consistent mechanical property data and correlating this information with the parameters of real interest have been recognized, and, as a result, simple mechanical testing has largely fallen out of favor. Instead, it is more common to build either complete products or representative subparts and subject them to some form of accelerated or extended test that is considered to encompass the rigors of life in service. The components are then assessed for signs of functional degradation. Any deciencies that are detected and can be attributed to unsatisfactory joints are addressed accordingly. More information about mechanical property tests can be found in the planned companion volume Principles of Brazing. Brazed joints are often expected to carry mechanical loads, and hence, issues such as the choice of test method used to measure the mechanical integrity of a joint become more directly relevant.
5.7.2
Variables such as strain rate and temperature alone can be manipulated to give joint strengths that vary by more than two orders of magnitude from identical samples prepared under rigorous laboratory conditions [Jones et al. 1997; ITRI 1987]. Notwithstanding the previously mentioned considerations, a further issue then arises as to whether the mechanical properties measured are satisfactory in the context of the service requirements of the assembly. For example, a solder joint between a silicon die and a ceramic package may achieve a shear strength of, say, 40 MPa (5800 psi). Because the die weighs only a few grams and there are no mechanical contacts made to it, then clearly, the joint is more than strong enough to hold the die in place. The majority of soldered joints in electronic products are not
Computational methods are nding growing use in modeling how components and products will respond to various stimuli, such as changes in temperature, mechanical loads, and exposure to chemicals. These naturally include consideration of soldered joints. There are basically two types of model: those that calculate the dimensions of the product, and others that predict the lifetime of joints. Models have also been devised that describe the prole of molten solders wetted onto solid substrates. An example of this type of model is referred to in section 5.2.6 in this chapter. 5.7.2.1 Dimensional Stability of Soldered Joints
One of the better known methods of calculating the dimensional stability of joints is nite-
element analysis (FEA). In FEA, a component or assembly is modeled in a geometrical manner in terms of a mesh of smaller units or elements, with other dimensions of each element scaling to a set of properties of interest. The visible mesh usually represents two or three orthogonal dimensions of the part, while the other dimensions are used to set how each element can respond to stimuli. Common parameters are Youngs modulus for mechanical behavior, and thermal expansivity and thermal conductivity for thermal behavior. In a realistic model, these parameters are not xed but vary with temperature. Constraints are then applied to the mesh, with matrix algebra used to obtain a comprehensive and numerically convergent solution. The large number of calculations that are needed requires a computer for this task. There are now a number of commercial FEA software packages available to suit a range of applications. A FEA prediction of the deformation that occurs in a ceramic-metal bond on cooling from the solidus temperature of the ller is given in Fig. 5.29. Despite the complexity of the modeling techniques, simplifying assumptions have to be made. Often, an axis of spatial symmetry is dened, and only half or one-quarter of the assembly is modeled. The oversimplication represented by this assumption can easily be demonstrated for many situations, especially when transient conditions and thermal gradients are taken into consideration. Also, when these models are ap-
plied to soldered assemblies, it is normally assumed that the joints are uniform in their composition and physical properties. This is patently not correct. A further restriction of applicability arises because most modeling programs assume that the interface between the components and the solder is relatively abrupt, whereas, in practice, it is a highly complex region that has signicant compositional and microstructural differences, generated by reaction across interfaces, and often these are not stable with time, temperature, or stress. Fortunately, the greater sophistication of software and increased computing power are enabling these features to be built into FEA models. To successfully apply such models requires knowledge of the properties of materials in an assembly and how those properties change in different environments. For mechanical models, the environmental variables are usually stress, strain, and temperature. Thus, there is often a call on the joining technologist to provide data on the mechanical and other physical properties of ller metals. For the reasons outlined in the preceding section, this is almost an impossible requirement to fulll reliably. However, the potential savings afforded by modern computer modeling techniques in their ability to assist in achieving rightrst-time designs and thereby speed product time to market means that there is often considerable pressure to provide appropriate material property data! Table 5.19 represents an attempt to provide indicative data of the bulk properties of
Fig. 5.29
Finite-element analysis prediction of the geometry of a ceramic-metal brazed joint, at its periphery, at the solidus temperature of the ller alloy and on cooling to room temperature
a few common solders and some of their common constituents at room temperature. When running a model, the values given in Table 5.19 can be taken as a starting point and modied to allow for the joint thickness, intermetallic formation, compositional change, and the properties of the components on either side of the joint. Temperature-dependent terms can then be added as appropriate. A literature search will usually throw out values for temperatureand strain-rate-dependent terms that can be adopted if they appear valid for the situation under consideration. Often, it is necessary to design and test joints of simplied geometry, in order to provide condence in the values and their sensitivity to second-order effects. While acknowledging the many limitations of the data, as mentioned previously, Table 5.19 does reveal some interesting trends in comparing the alloys with the constituent pure elements. Almost invariably, solders possess greatly inferior thermal and electrical conductivity, compared with pure metals. These characteristics follow from their heterogeneous microstructure and phase boundaries, which impede the ow of electrons and phonons. The effect is more pronounced on thermal conductivities because of the dominant contribution of phonons, or lattice waves, to thermal conductivity at room-ambient and elevated temperatures. Phonons are more strongly scattered by the inhomogeneous microstructure than are electrons, which dominate electrical conductivity and low-temperature thermal conductivity. At rst sight, it might be expected that the gold-rich solders would have properties not dissimilar to pure gold. However, the large difference in density between gold, on the one hand, and the alloying elements, on the other, means that in terms of volume fraction, the gold-
silicon eutectic alloy, for example, possesses close to 20 vol% Si. The gold-tin eutectic solder, Au-20Sn, is not a mixture of the constituents but an alloy of two intermetallic compounds of gold and tin (AuSn and Au5Sn), both of which have very different characteristics to the pure elements. In general, if compound formation occurs in the creation of solders, then the properties cannot be predicted reliably without direct knowledge of the mechanical and physical properties of the constituent phases. 5.7.2.2 Prediction of Joint Lifetime
The traditional methods for predicting the lifetime of a fabricated part and the joints contained therein are accelerated and extended testing. Because these are time-consuming and expensive to undertake, much effort has been devoted to attempting to predict joint life by numerical modeling. The ability to do this is particularly relevant to the needs of the PCB industry, where it is crucial to know the lifetime of the product and also for manufacturers to convincingly demonstrate to customers that the soldered joints will meet long-term expectations of reliability. The approach to this problem started with classical fatigue theory, which was rapidly found to be wholly inadequate for describing the behavior of solders. Gradually, more and more rate- and temperature-dependent parameters were added to the models. Although the quality of the analysis steadily improved, unexpected outcomes were as common as accurate predictions. Some testing, albeit less extensive, remained necessary to validate the model. Over the last few years, an alternative approach that uses a concept termed strain energy density has been gaining favor [Morrow 1964;
Table 5.19 Indicative property values of selected solders and pure metals in bulk form at room temperature
Element or solder Hardness, HV Tensile strength, MPa Youngs modulus, GPa Poissons ratio Elongation to failure, % Electrical resistivity, cm Thermal conductivity, W/m K Thermal expansivity, 106/K
20 5 10 15 5 15 25 10 5 15 100 100
80 10 15 50 10 40 ... 50 ... 30 80 60
0.40 0.45 0.42 0.35 0.45 0.35 ... 0.35 ... 0.25 ... 0.30
40 50 50 30 50 30 1 30 50 30 1 5
2 10 20 15 10 15 35 15 30 15 20 10
350 80 35 70 50 50 20 50 20 40 30 60
20 25 30 25 25 25 20 25 20 25 15 10
Vaynman and McKeown 1993]. In very simplistic terms, this model attempts to examine whether the deformation of a joint on single or multiple stress cycles exceeds the ability of the solder to absorb and/or dissipate the energy imparted to it, through creep or microstructural changes. Where microstructural changes do occur, the model is able to account for the material change on subsequent cycles. Although its complexities are beyond the scope of this book, the important point to note is that the quality of the joint lifetime predictions now being made generally accord with experimental results. For examples of application of this method, reference may be made to Lau [2000], Lau and R. Lee [2001], Lau et al. [2002], and similar publications.
with the ore from which it was obtained: monazite, xenotime, or bastnasite. Mischmetal is sometimes given the chemical symbol M, and it contains, typically, 50% Ce and 30% La. The general symbol used for rare earth is RE. Rare earth elements have the common attribute that they are extremely reactive toward other metals and most atmospheres. The addition of rare earth elements to solders has a number of effects on properties that vary with concentration.
5.8.1
5.8
A relatively new area of solder research concerns doping of traditional ller metals with rare earth elements. Some results are reported in the literature, and patents have been led in respect of certain composition ranges. Rare earth doping of solders is of interest because it affords a possible means of favorably modifying the characteristics of solders in the molten and solid states and enabling them to approximate more closely to the materials properties desired by the manufacturing industry. A large proportion of the work done in this area has been in conjunction with lead-free solders. These solders are discussed in section 5.1 of this chapter. The rare earth elements are so described because they were originally thought to have a low abundance in the Earths crust and to be difficult to win from it and to separate from each other. It is now known that lanthanum, cerium, and neodymium are actually more abundant than lead, and vast ore reserves have been found in China and the United States. There are thirty rare earth elements, which is really another name for the elements contained in the lanthanide and actinide series of group three of the periodic table. However, one element of the lanthanide series (promethium) and most of the actinides are transuranium elements, that is, manmade and atomically unstable. Sometimes, reference will be seen to a rare earth called mischmetal. This is actually an alloy containing a high concentration of the rare earth elements in proportion to their natural abundance, and therefore, its composition varies
Recent research has shown that doping of solders with rare earths appears to enhance their properties while introducing few drawbacks, and the melting range of solder alloys is hardly affected (Fig. 5.30) [Ramirez, Mavoori, and Jin 2002; Wang, Yu, and Huang 2002; Chen et al. 2003]. Small additions of rare earths:
Fig. 5.30
Solidus and liquidus temperatures of 3.8Ag-0.7Cu-95.5Sn solder as a function of the added rare earth (RE) concentration
the rare earth elements increases the preference for discrete compound formation, which would explain why the creep properties peak at 0.1% RE addition in 3.8Ag-0.7Cu-95.5Sn solder. Exposure to elevated temperature reduces the benet to be derived from grain-
boundary pinning by the rare earth intermetallic compounds, because other material transport mechanisms take effect. Thus, at 65 C (149 F), the creep-rupture life is only double that of the simple ternary solder, and any advantage disappears completely above
Tensile strength and elongation to failure of Ag96.5Sn solders doped with cerium and lanthanum. The samples were chill-cast ingots of solder, 20 mm (0.8 in.) long by 10 mm (0.4 in.) in diameter, tested at room temperature and a strain rate of 4 103/s. RE, rare earth
Fig. 5.31
Stress-rupture life of Sn-3.5Ag and Sn-3.5Ag0.25RE solders for an applied stress of 20 MPa (2900 psi) at 50 C (122 F)
Fig. 5.33
Tensile strength and elongation to failure of Sn-9Zn solders doped with lanthanum. The samples were chill-cast ingots of solder, 25 mm (1 in.) long by 5 mm (0.2 in.) in diameter, tested at room temperature and a strain rate of 5 103/s. RE, rare earth
Fig. 5.32
Stress-rupture life of 3.8Ag-0.7Cu-95.5Sn solder as a function of the rare earth (RE) content for an applied stress of 16.5 MPa (2400 psi) at room temperature
Fig. 5.34
Fig. 5.35
Contact angle and spread area of Ag-96.5Sn solders doped with cerium and lanthanum melted on copper at 300 C (572 F) for 30 s under cover of RMA ux. RE, rare earth
100 C (212 F). This result puts a question mark on the practical benet of the observed improvement in creep resistance. Also, the available data pertain to tens of hours of test duration, not thousands of hours of service life, and this issue needs to be investigated. Decrease the wetting angle of the solder on metal substrates and generally enhance spreading. The effect on contact angle, wetting force, and wetting time is shown in Fig. 5.35 for silver-tin eutectic solder on copper. Rare earth doping of the Sn-9Zn solder improves its wetting on metal substrates in a similar way (Fig. 5.36) [Wu et al. 2002]. Likewise, small additions of rare earths have been found to reduce the contact angle of indium solders on silver under inert atmosphere, as described in Chapter 3, section 3.3.8.4 (Fig. 3.28). The tests described were conducted in air using ux. It is presumed that the role of the rare earths is to reduce the interfacial tension between the solder and ux [Wang, Yu, and Huang, 2002]. It would seem that the reactive rare earth elements also help destabilize oxide lms on the surface of the solder and substrate. Certainly, once the rare earth concentration exceeds approximately 0.2%, the modied solders are able to directly wet and bond to nonmetals, such as silica. This has been demonstrated for both silver-tin eutectic solder and the high-melting-point Au20Sn alloy. The bonding mode is as described in Chapter 4, section 4.1.2.2, namely, migration of the rare earth element to the solder/substrate interface and chemical bonding by reduction of the nonmetal surface. Unfortunately, the concentration of rare earth elements necessary to achieve this behavior results in a loss of many of the boosted
Wetting force of Sn-9Zn solders doped with lutetium on copper using rosin-activated ux at 245 C (473 F) in air. RE, rare earth
Fig. 5.36
mechanical properties referred to earlier and impairs the ductility of all solders [Ramirez, Mavoori, and Jin 2002].
5.8.2
Most results reported to date relate to bulk solder specimens. The strength of lap joints in copper testpieces was found to be indifferent to rare earth doping. This somewhat disappointing result was attributed to porosity in the joints, originating from the use of a simple organic ux for the joining operation. Whether this porosity is an intrinsic characteristic of rare-earthdoped solders or indeed merely due to inappropriate choice of ux is uncertain, because currently, there are few reported studies of mechanical properties of joints made and assessed under controlled conditions. As far as the authors are aware, solders doped with rare earth elements cannot yet be found in manufacturers catalogs.
5.9
Diffusion Soldering
5.9.1
Process Principles
In soldering, wetting of the component surfaces is not always easy to achieve, and when it does occur, the resulting alloying between the ller and components can cause excessive erosion of the parent materials, embrittlement of joints due to the formation of phases with inferior mechanical properties, and other undesirable effects. These problems notwithstanding, solders have the singular merit of being able to ll joints of irregular dimensions and produce well-rounded llets at the edges of the joint. Diffusion bonding sidesteps the need for wetting and spreading by a ller metal (see Chapter 1, section 1.1.7.2). Once formed, diffusionbonded joints are stable to high temperatures, so that the service temperature of the assembly can actually exceed the peak temperature of the joining process without risk of the joint remelting. While the formation of undesirable intermetallic phases can also occur in diffusion bonding, because there are usually fewer constituents involved, it is easier to select a safe combination of materials. However, diffusion bonding tends to be limited as a production process, because it is not tolerant to joints of variable width, and moreover, its reliability is highly sensitive to surface cleanliness. High loads (typically 10 to 100 MPa, or 1.4 103 to 1.4 104 psi) have to be applied during the bonding cycle to ensure good metal-to-metal contact across the joint interface. Also, the duration of the heating cycle is typically hours, compared with seconds for soldering, because solid-state diffusion is much slower than wetting of a solid by a liquid. These factors and the absence of any signicant llets to minimize stress concentrations at the edges of joints (see Chapter 4, section 4.2.4) considerably limit the applications of diffusion bonding. There exists a hybrid joining process that combines the good joint lling, llet formation, and tolerance to surface preparation of conventional soldering, together with the greater exibility with regard to service temperature and metallurgical simplicity that is obtainable from diffusion bonding. This process is called diffusion soldering, sometimes also referred to as transient liquidphase (TLP) joining. Its higher-temperature analogue, diffusion brazing, is an established joining and repair process that has been used for decades in the aerospace industry and is described in the planned companion volume Principles of Brazing.
Diffusion soldering uses a thin layer, typically 5 m (200 in.) or less, of molten ller metal to initially ll the joint clearance, but during the heating stage, the ller diffuses into the material of the components to form solid phases, raising the remelt temperature of the joint. At this stage, isothermal solidication occurs, and further reaction proceeds by solid-state diffusion until the process cycle is complete. Due to the generation of liquid in the joint, the necessary applied pressures are much less than those required for normal diffusion bonding and are typically in the range of 0.5 to 1 MPa (70 to 140 psi). Diffusion soldering therefore provides the ready means to ll joints that are not perfectly smooth or at (a feature of liquid-phase joining) while offering great exibility with regard to process temperature in relation to the service temperature of the product. Because of these features of the process and also the precise conditions used in implementing it, especially the controlled thinness of the layer of low-melting-point ller and specied loading applied to the joint, the following additional advantages are obtained: Good reproducibility of joints Excellent joint lling that applies to both small- and large-area joints. Joints free of voids ensure leaktightness, which is important in situations where the joint is made to provide a seal. Tight control of edge spillage, which can be kept to a minimum Attainment of very narrow joints, typically less than 10 m (400 in.), which benets the mechanical properties, as compared with conventional solder Exploratory work on diffusion soldering was done in the mid-1960s [Bernstein and Bartholomew 1966; Bernstein 1966]. This activity, which was described as solid-liquid interdiffusion bonding, was primarily concerned with lowtemperature bonding of power semiconductor die involving indium and gold. In the processes that Bernstein investigated, the emphasis was on low temperatures, typically below 160 C (320 F), and the endpoint of the process left thick intermetallic phases in the joint, which tended to compromise the mechanical integrity, although it is adequate for the intended function. Alloy systems suitable for diffusion soldering will possess a phase constitution that includes a relatively low-melting-point constituent ideally, a eutectic reactionto initiate the melt-
ing process and a higher-melting-point phase or solid solution on which to terminate solidication. Wilde and Pchalek [1993] identied the binary combinations of the precious metals (silver, gold, palladium) and also copper and nickel, together with either tin or indium as the lowmelting-point constituent, as among the most suitable systems for diffusion soldering. Some of the process details for these combinations, as reported in the published literature, are listed in Table 5.20. In most of the published studies of diffusion soldering, the emphasis has been on limiting the process temperature to make the joining operation suitable for the attachment of dies and other electronic components, which are temperaturesensitive. In consequence, a compromise has had to be struck whereby intermetallic phases are tolerated at the expense of mechanical strength. This trade-off is usually acceptable, because the strength requirement in die attachment is generally low. For example, the gold-indium joints obtained by Wang et al. [2000], which achieved a shear strength of 12 MPa (1740 psi), were sufficiently strong to satisfy MIL STD 883E, method 2019.7 acceptance criteria. At rst sight, the cost of the precious metals, especially gold, might be considered an impediment. However, the depth of interaction of these metals with the ller metal is shallow in diffusion-soldering processes, so that they can be apTable 5.20 literature
plied as thin metallizations (<10 m, or 400 in.) to the component surfaces without the danger of the component materials entering the reaction. The contribution to component cost from such thin layers is relatively small. The use of copper, silver, or gold as surface coatings confers the particular advantage that, being relatively noble, they are readily wetted, even in slightly oxidizing atmospheres. This makes diffusion-soldering processes using these metals relatively tolerant to the condition of the atmosphere in which the joining operation is carried out. In the published reports referred to in Table 5.20, the authors emphasize their ability to achieve satisfactory joints without the use of uxes in moderately protecting atmospheres, such as a shroud of nitrogen. The sequence of steps involved in making a diffusion-soldered joint is shown schematically in Fig. 5.37.
5.9.2
The authors have found the silver-tin system to be one of the most versatile for diffusion soldering of components for use in the electronics industry, because well-lled and ductile joints can be produced using compressive loads of as little as 0.5 MPa (70 psi) [Jacobson and Humpston 1992]. Silver-tin solder of eutectic composition reacts with silver to form the Ag3Sn phase as a continuous interfacial layer, which is both
Substrate
Filler metal
Process temperature, C
Time, min
Remelt temperature, C
Ref
Copper Copper
Indium Tin
180 280 300 300 250 250350 175 450 310 260 200 160240 300 400
0.5 0.51.0 Not specied Not specied 1.0 0.17 0.5 1.0 0.28 0.3 1.25 0.05 0.8 0.8
>307 >415 >676 >676 >600 >600 >880 >900 >278 >278 >495 >495 >400 >977
Sommadossi et al. [2000] Bartels et al. [1994] Kato, Horikawa, and Kageyama [1999] Kang et al. 2002 Jacobson and Humpston [1992] Wilde and Pchalek [1993] Jacobson and Humpston [1992] Humpston, Jacobson, and Sangha [1993] Matijasevic, Lee, and Wang [1993] Lee and Wang [1992] Wang et al. [2000] Wilde and Pchalek [1993] Khanna, Dalke, and Gust [1999] Khanna, Dalke, and Gust [1999]
Gold Nickel
Indium Tin
tough and strongly adherent to the other phases in this alloy system. The rate of reaction between silver and molten tin has been characterized and is represented graphically in Fig. 2.35. The controllable nature of the alloying reaction in the conventional soldering system is indicated by the general prole of the erosion curves, which show that the reaction is self-limiting in character, within the context of realistic processing cycle times and temperatures. If a thin layer of tin, typically 5 m (200 in.) thick, is sandwiched between two components, each covered with a 10 m (400 in.) thick layer of silver, and heated to 250 C (480 F), the tin will melt and react with the silver to form Ag3Sn. On continued heating, the tin is progressively converted to this compound until no liquid tin remains. By keeping the tin layer thin, it completely reacts to form solid phases at the joining temperature in less than 1 min. The remelt temperature of a silver-tin diffusion-soldered joint is determined by the phases that are present. Immediately after the liquid tin has been consumed by reaction, the remelt temperature is that of the Ag3Sn compound, which is 480 C (895 F). Longer heating times promote
Fig. 5.37
continued diffusion of tin from theAg3Sn reaction zone into the silver. Consequently, the width of this zone decreases as it is replaced rst by Ag5Sn () and, ultimately, by a solid solution of tin in silver, as anticipated from the silver-tin phase diagram given in Fig. 2.9. This progression is illustrated by the series of micrographs shown in Fig. 5.38. As the reaction with the silver proceeds, the remelt temperature rises progressively toward the melting point of silver (962 C, or 1764 F). Mechanical property measurements have shown that the shear strength of diffusion-soldered joints containing theAg3Sn phase is close to the 25 MPa (3600 psi) value for conventional soldered joints made with theAg-96.5Sn eutectic solder to silvercoated components. As the joint microstructure converts to a silver solid solution, the mechanical properties shift in tandem toward those of pure silver, with the shear strength increasing toward 75 MPa (11,000 psi). This can be signicant from an applications point of view, because the strength of silver is approximately three times that of the Ag96.5Sn eutectic alloy, which itself is superior in this respect to the common Pb-62Sn solder at room temperature by a factor of two to three [Harada and Satoh 1990]. Another attractive feature of the silver-tin alloy system for diffusion soldering is that there is negligible volume contraction as the reaction proceeds, which is a fortuitous consequence of the closely similar specic volumes of the various phases. Therefore, the tendency to form voids or cracks as a result of volume change is minimal. Diffusion-soldering processes are not routinely encountered but are used commercially. One example is as a method of attachment of silicon power devices to molybdenum heat sinks [Jacobson and Humpston 1992; Humpston et al. 1991]. Replacing brazed joints made using an industry standard process, involving the Al-12Si alloy, with silver-tin diffusion soldering provided a means for reducing the process temperature from over 600 to 275 C (1112 to 527 F), which decreased the bimetallic, center-to-edge bow by 240% for a typical 50 mm (2 in.) component. Besides silver-tin, a silver-indium diffusion-soldering process is an alternative, offering a lower process temperature [Humpston and Jacobson 1990]. However, the associated processing involved is more complex the plating of indium is less standardthanthatoftin,andthemorerefractorynature of indium oxides makes it necessary to apply special surface treatments to exposed indium surfaces prior to the bonding operation [Sommadossi et al. 2002].
5.9.3
The gold-tin alloy system has provided the basis for the diffusion-soldering process for joining items of high-carat gold jewelry below 450 C (842 F). The traditional gold jewelry manufacturing route involves the use of the so-called carat gold solders, which are actually brazing alloys with working temperatures above 800 C (1470 F) [Rapson and Groenewald 1978; Normandeau 1996]. The high temperatures involved are detrimental to the mechanical strength of high-carat gold jewelry, because they anneal and soften rapidly when heated above approximately 450 C (842 F). Further details on carat gold solders and the metallurgy of gold jewelry can be
found in the planned companion volume Principles of Brazing. Diffusion soldering provides a satisfactory alternative joining process. In trials, it was found that a tin coating 3 to 4 m (120 to 160 in.) thick was generally sufficient to ensure complete joint lling and the formation of small edge llets. Provided that the peak process temperature exceeds 419 C (786 F), the melting point of the AuSn intermetallic compound, the tin will transform initially to the high-gold intermetallic compound Au5Sn and, on continued heating, to gold solid solution. The Au5Sn compound contains approximately 90 wt% Au and so meets the 18 carat requirement of the jewelry item. Prolonged heating is undesirable, because it results in softening of the gold assembly, as
Fig. 5.38
Series of micrographs showing the progressive change in joint microstructure that occurs on making a diffusion-soldered joint using tin in combination with silver metallizations applied to copper substrates. 400
reected by the grain growth, and also in Kirkendall voiding in the centerline of the joint. One hour at 450 C (842 F) under a compressive loading of 1 MPa (145 psi) was found to be an acceptable compromise for the processing conditions [Humpston, Jacobson, and Sangha, 1993]. Figure 5.39 shows a bracelet and matching earring set that was assembled by this method and exhibited at the World Jewelry Trade Fair held in Basel in 1992.
5.9.4
Copper-tin and copper-indium are also suitable systems for diffusion soldering. However, when the copper-tin joining process is carried out below 676 C (1249 F) and the copper-indium process is operated below 631 C (1168 F), they result in the formation of planar intermetallic phases that have limited fracture toughness and are responsible for relatively weak joints. In the case of copper-tin, these intermetallic phases are Cu6Sn5 () and Cu3Sn (), while Cu7In4 () and Cu7In3 () are found in the copper-indium system. The Cu3Sn () and Cu7In3 () phases form adjacent to the surface of the copper layer [Kato, Horikawa, and Kageyama 1999; Kang et al. 2002]. In the copper-tin process, it has been shown that the formation of the brittle intermetallics can be suppressed by raising the joining temperature above the melting point of the Cu3Sn intermetallic (676 C, or 1249 F). The successful dif-
fusion-brazing process that has been developed using this approach is described in the planned companion volume Principles of Brazing (see also Sangha, Jacobson, and Peacock 1998). An interesting variant of this diffusionsoldering process has been investigated whereby the tin solder in the copper-to-copper joint is replaced by the In-49Sn eutectic alloy (melting at 120 C, or 248 F). The joining operations were carried out at temperatures up to 400 C (752 F) [Sommadossi, Gust, and Mittemeijer 2002]. In this case, two intermetallic compounds form by reaction, but these are different from the and phases produced in the absence of indium. They are both ternary alloys. One of these is based on the Cu10Sn3 () phase, which, in the binary alloy system, is only stable at elevated temperature but is stabilized at room temperature by the addition of indium. The other phase is designated (confusingly, here) as but is based on the Cu2In rather than the Cu6Sn5 intermetallic. This phase can also dissolve the third element, in this case, tin. Below 200 C (392 C), only the phase grows, and above this temperature, both grow together, with the phase steadily outgrowing its sister phase. The joints made were relatively thick (typically 50 m, or 2 mils), which did enable signicant dilution of the indium and tin in copper to occur, so as to dissolve the intermetallic phases. However, joints made at 350 C (662 F) using 10 m (400 in.) thick sputtered layers of In-49Sn that were etch cleaned prior to bonding achieved a shear strength in excess of 155 MPa (22,500 psi) and a tensile strength of 160 MPa (23,200 psi) [Sommadossi et al. 2002]. These relatively strong joints contained only the homogeneous [Cu10(Sn, In)] phase, which grows entirely by solid-state diffusion and has a relatively small grain size.
5.9.5
Practical Aspects
Parts of an 18 carat gold bracelet and matching earring set assembled using the gold-tin diffusionsoldering process at 450 C (842 F). The unusually low process temperature enables the face plates to retain much of their workhardened strength and thereby accept a particularly high surface polish. Each box of the bracelet measures approximately 8 mm (0.3 in.) wide. Courtesy of the World Gold Council
Fig. 5.39
There are often practical difficulties with applying the layer of ller as an electroplated or vapor-deposited coating to the intended joints, including the need to mask off other areas of the surfaces of the components. It has often proved more convenient to use a foil preform of the relevant precious metal, typically 25 to 100 m (1 to 4 mils) thick, that is coated with the necessary thickness of the low-melting-point ller (tin or indium). An appropriate area of the foil is cut out and sandwiched in the joint gap. The use of a foil of soft precious metal offers the further merit of acting as a stress absorber, which is most
useful in situations where the parts to be joined have local topography or signicantly different coefficients of thermal expansion. This latter aspect is treated in greater detail in the planned accompanying volume devoted to brazing. Because of the higher temperatures involved in brazing operations, thermal expansion mismatch stresses can be a more serious problem in that context.
5.9.6
Some attention has been devoted to the theoretical modeling of transient liquid-phase joining processes, but the published studies to date have been conned to the higher-temperature diffusion-brazing process [Isaac, Dollar, and Massalski 1988]. The analytical models generally assume that the process kinetics are governed by diffusion, so that the phases that solidify from the melt at the joining temperature grow at a rate that is proportional to the square root of the bonding time. Clear evidence for the classical diffusioncontrolled relationship has been provided for the copper-tin system at 300 C (572 F) [Kato, Horikawa, and Kageyama 1999]. However, there has been little systematic work in modeling the kinetics of the various solid phases that grow and subsequently are replaced by other solid phases, or of the primary solid solution, as usually occurs in diffusion soldering. Such an endeavor would greatly facilitate the design and development of this interesting joining method.
5.10
Of the many techniques available to assess the integrity of soldered joints, particularly those used to attach electronic components to PCBs, three have beneted greatly from the advent of computer technology. These are ultrasonic inspection, x-ray inspection, and optical inspection. 5.10.1 Ultrasonic Inspection (Scanning Acoustic Microscopy)
Ultrasound is dened as pressure waves with frequencies higher than sound waves and that cannot be heardin practice, in the range of 0.5 MHz to 5 GHz. The particular characteristic of
ultrasound that is exploited in nondestructive technology is its ability to travel through solid materials while obeying the same laws of reection and refraction as light. Because ultrasound travels with xed velocity through a given material, echoes produced by reection at different surfaces and interfaces will be temporally resolved, and the corresponding distances can then be calculated. This is the basic principle of pulse-echo ultrasonic inspection. In commercial instruments, ultrasound is generated by a piezoelectric transducer mounted as a probe that is coupled to the surface of the testpiece via a liquid or pasty coupling agent. Two or more probes tend to be used: one to transmit the pulse and the others to detect echoes. The higher the ultrasonic frequency, the higher is the resolution, but the stronger is the signal attenuation due to absorption by the materials through which it travels. Ultrasonic signals, like other forms of wave energy, can resolve features down to approximately the size of a single wavelength. Accordingly, ultrasound of 10 MHz frequency should be capable of detecting cracks down to 0.5 mm (0.02 in.) in cross section in metal. This is clearly inadequate for the inspection of defects in joints that may themselves be of comparable size or even smaller, for example, those made to surfacemount electronic components. To improve on this level of discrimination, higher frequencies must be used. The scanning acoustic microscopy (SAM) technique has been developed to operate in the frequency range of 20 MHz to 2 GHz and offers the nest level of resolution of the ultrasonic test methods, although the depth of penetration is limited to below 10 mm (0.4 in.). The SAM technique is capable of nondestructively assessing the distribution of voids, cracks, inclusions, and other hard phases over the area of essentially parallel-sided joints [Matuasevic, Wang, and Lee 1990; Kauppinen and Kivilahti 1991]. It involves focusing an acoustic wave, generated from a piezoelectric transducer, via a sapphire lens onto the specimen and scanning it in a raster fashion. Changes in the reected acoustic signal from boundaries between features having different acoustic properties are recorded and mapped to produce the image. The correspondence that can be obtained between the images of a voided joint produced by x-radiography and by SAM is illustrated in Fig. 4.25 and 4.26. The interpretation of a scanning acoustic micrograph of a joint can present difficulties, for
example, determining whether a certain feature corresponds to a crack, void, or intermetallic particle. Similarly, the need to mechanically scan the probe in very close proximity to the surface of the components restricts the geometries that can be examined. Moreover, the limited depth of sample from which clear images can be obtained means that one of the components must be thin. Modern SAM equipment overcomes all these difficulties by using a variable or multiple frequency system, often with multiple transducers, and by automating the probe positioning and image processing. A modern SAM system is able to discriminate between internal and external boundaries of components and present the user with a three-dimensional image that can be electronically rotated and sliced similar to a computer-aided design drawing. By this advance, SAM has changed from being a laboratory diagnostic tool to part of the suite of in-line qualityassurance methods essential to ensure lowdefect-rate manufacturing. Some examples of the use of SAM to inspect the interior of electronic components in a nondestructive manner are given by Adams [2001].
movement enables defects to be viewed at an optimal angle. An important innovation in x-radiography is the development of systems incorporating microfocus sources. Small focal-spot sizes can be achieved at the x-ray target by electromagnetically focusing the incident electron beam, exploiting a technique that is widely used in electron microscope technology. The focal-spot size can be as small as 1 m (40 in.) in diameter. The benets offered by a microfocus x-ray system are:
5.10.2
X-Radiography
The x-ray spectrum comprises electromagnetic radiation of short wavelengths in the range 1016 to 1021 Hz. The high frequency and energy of x-rays enables them to penetrate materials and reveal internal features, including defects, provided they absorb the radiation to a different extent than does the surrounding material. It is this characteristic that provides the contrast in the x-ray image. Accordingly, voids, inclusions of heterogeneous material, and cracks parallel to the x-rays will be more conspicuous to this technique than cracks and interfaces that are perpendicularly oriented. Radiographs revealing voids in joints are shown in Fig. 1.17 and 4.25. In a modern industrial x-ray machine, a television camera system is employed in place of traditional lm. This enables the x-ray image of the object to be viewed in real-time. The component undergoing inspection is held by a freespace manipulator. The combination of movement of the testpiece coupled with real-time viewing permits a comprehensive examination to be made rapidly. The nature of visual perception is such that an area that differs marginally in contrast from its surroundings is more easily detected when in motion, so that defects are more readily noticed. Additionally, the controlled
Fig. 5.40
Steady improvements in microfocus x-ray tube design have resulted in geometrical magnication of over 2000 being realized in commercial systems, with the tubes operating in the low-kilovolt range. With further digital processing of the optical image, total magnications of 7000 are now possible. This enables features of the order of 250 nm (10 in.) to be resolved, and the system is then said to be operating in the nanofocus mode. A diminution of the focal-spot size means that the geometric unsharpness is correspondingly reduced; that is, the precision with which edges can be resolved is improved. A consequential benet of spatially separating the image sensor from the testpiece is a reduction in the scattered ray fraction generated by the testpiece itself, which contributes to the image. The x-ray beam emerging from a microfocus source can be proled to give a controlled cone of radiation, again decreasing scatter and improving the sensitivity to aws. Finally, by simply moving the testpiece along the axis of the x-ray beam, it is possible to continuously zoom in on detail.
Multifocus x-ray tubes have been introduced that enable the user to choose between microfocus, nanofocus, and high-power modes, using software control and a single mouse click. The microfocus mode is preferred for applications where the area being examined is millimeters on a side, such as inspection of circuit boards and discrete modules. By contrast, the nanofocus mode is suitable for examining ip-chip assemblies and wire bonds. High-power settings are appropriate for denser samples, such as die mounted on heat studs and generally, bulkier items where the boosted x-ray energy provides deeper penetration of the radiation.
is vastly improved compared to even only a few years ago. Unfortunately, these improvements come with a price, and a standard objective lens can easily carry a four-gure (dollars) price tag. Notwithstanding the fundamental improvement in lens design and manufacture, two relatively recent innovations in optical inspection are having an impact on soldering technology. These are automated optical inspection (AOI) and endoscopes. Automated optical inspection is the product of a highly successful marriage between optical and digital electronic technology. In essence, the part to be inspected is placed under an optical microscope, and the image is electronically processed to identify certain features, usually against specied pass/fail criteria. The viability of this inspection method hinges greatly on the camera system. The camera used in an AOI system is a digital camera with a sensor having a minimum resolution of six million pixels. To maximize the quality of the information acquired by each pixel, it is exported directly to the computer in digital format, without conversion to a video signal. Very sophisticated analysis is now possible on the acquired digital image. Some of the outputs that are available from a commercial AOI system include:
5.10.3
Optical Inspection
Optical inspection of joints normally describes examination by eye, often with the aid of a magnifying device. Usually, this will be an optical microscope but may be stretched to include scanning electron microscopy operated in its usual backscattered mode. Despite being a very old technology, optical inspection methods continue to improve at a remarkable rate. The quality of the image that can be obtained from a modern microscope in terms of its resolution, depth of focus, eld area, and working distance
The continued development of improved optics, cameras, and more powerful software, coupled with improved processing speeds, can only enhance the adoption of AOI as an integral part of electronics and photonics assembly lines. Endoscopes are essentially miniaturized microscopes that use optical bers or other means to transmit the image from the location of the
examination to the viewing point. They are widely used in medicine for conducting internal examinations in a noninvasive manner. Endoscopes have been developed for examining hidden soldered joints. Examples include interconnects of ball grid arrays (BGAs), chip-scale packages (CSPs), and ip-chip components. Custom optical heads enable side-on viewing of features, while their small size and long working distance permit inspection of otherwise inaccessible areas. Inspection of solder llets, solder paste print proles, via-hole plating integrity, and conformal coating uniformity are but a few examples of their application.
REFERENCES
brids Manuf. Technol., Vol 13(No. 4), p 736 742 Hayashi, T., 1992. An Innovative Bonding Technique for Optical Chips Using Solder Bumps that Eliminate Chip Positioning Adjustments, IEEE Trans. Compon. Hybrids Manuf. Technol., Vol 15(No. 2), p 225230 Hayes, D.J., Wallace, D.B., and Boldman, M.T., 1996. Solder Jet Printing for Low Cost Wafer Bumping, Proc. Conf. International Society for Hybrid Microelectronics, 811 Oct (Minneapolis, MN), p 296301 Ho, C.T., 1996. Nickel and Copper Coated Carbon Fibre Reinforced Tin-Lead Alloy Composite, J. Mater. Sci., Vol 31, p 5781 5786 Ho, C.T. and Chung, D.D.L., 1990. Carbon Fibre Reinforced Tin-Lead Alloy as a Low Thermal Expansion Solder Preform, J. Mater. Res., Vol 5(No. 6), p 12661270 Humpston, G. and Jacobson, D.M., 1990. U.K. Patent 2,235,642; U.S. Patent 5,106,009 Humpston, G. et al., 1991. Recent Developments in Silicon/Heat-Sinks for High-Power Device Applications, GEC Rev., Vol 7(No. 2), p 6778 Humpston, G., Jacobson, D.M., and Sangha, S.P.S., 1993. Diffusion Soldering: A New Low Temperature Process for Joining Carat Gold Jewelry, Gold Bull., Vol 26(No. 3), p 90104 Hunt, C. and Wallis, D., 1995. Solderability Standard: A Gold-Plated Nickel Reference Material, Solder. Surf. Mt. Technol., Vol 21(No. 1), p 69 Hwang, J.S., 1996. Environment Friendly Electronics: Lead-Free Technology, Electrochemical Publications IDEALS, 1999. Improved Design Life and Environmentally Aware Manufacturing of Electronics Assemblies by Lead-Free Soldering, BE95-1994, Commission of the European Community, 30 June Isaac, T., Dollar, M., and Massalski, T.B., 1988. A Study of the Transient Liquid-Phase Bonding Process Applied to a Ag-Cu-Ag Sandwich Joint, Metall. Trans. A, Vol 19(No. 3), p 675686 ITRI, 1987. Solder Alloy DataMechanical Properties of Solders and Soldered Joints, Publication 656, International Tin Research Institute Jacobson, D.M. and Humpston, G., 1992. Diffusion Soldering, Solder. Surf. Mt. Technol., Vol 10(No. 2), p 2732
per-Tin Diffusion Brazing Process, Weld. J. Res. Suppl., Vol 77(No. 10), p 432s438s Shimizu, T. et al., 1999. Zn-Al-Mg-Ga Alloys as Pb-Free Solder for Die-Attaching Use, J. Electron. Mater., Vol 28(No. 11), p 1172 1175 Smith, E.B. and Swanger, L.K., 1999. Are Lead-Free Solders Really Environmentally Friendly?, Surf. Mt. Technol., Vol 12(No. 3), p 6466 Sommadossi, S. et al., 2000. Development of Cu/Cu Interconnections Using an Indium Interlayer, Proc. Conf. EuroMat 2000, 2730 Sept (Munich), p 16 Sommadossi, S. et al., 2002. Mechanical Properties of Cu/In-48Sn/Cu DiffusionSoldered Joints, Z. Metallkd., Vol 93, p 496 501 Sommadossi, S., Gust, W., and Mittemeijer, E.J., 2002. Characterization of the Reaction Process in Diffusion-Soldered Cu/In-48 at.% Sn/Cu Joints, Mater. Chem. Phys., Vol 77, p 924929 Steen, H.A.H. and Becker, G., 1986. Effect of Impurity Elements on the Soldering Properties of Eutectic and Near-Eutectic Tin-Lead Solder, Brazing Soldering, Vol 11(No. 4), p 411 Sturcken, K., 2000. Column Grid Array, High Reliability Option for Packaging, Adv. Packag., Aug, p 4550 Subramanian, K.N., Bieler, T.R., and Lucas, J.P., 1999. Microstructural Engineering of Solders, J. Electron. Mater., Vol 28(No. 11), p 11761183 Tan, Q. and Lee, Y.C., 1996. Soldering Technology for Opto-Electronic Packaging, Proc. 46th Electronic Components and Technology Conference, 2831 May (Orlando, FL), p 26 36 The European Commission, 2000. Proposal for a Directive of the European Parliament and of the Council on Waste Electrical and Electronic Equipment and on the Restriction of the Use of Certain Hazardous Substances in Electrical and Electronic Equipment, COM (2000) 347 nal - 2000/0159/ COD The European Commission, 2001. Restrictions on the Use of Hazardous Substances, Document 501PC316, 6 June, Legislation in preparation, Commission proposals Thwaites, C.J., 1981. Solderability and Some Factors Affecting It, Brazing Soldering, Vol 1(No. 3), p 1518
Doping with Rare-Earth Elements, J. Mater. Res., Vol 17(No. 12), p 31463154; The Properties of Sn-9Zn Lead-Free Solder Alloys Doped with Trace Rare Earth Elements, J. Electron. Mater., Vol 31(No. 9), p 921927; Microstructure and Mechanical Properties of New Lead-Free Sn-Cu-RE Solder Alloys, p 928932; The Wettability and Microstructure of Sn-Zn-RE Alloys, J. Electron. Mater., Vol 2(No. 2), p 6369 Xia, Z. et al., 2002. Effect of Rare Earth Element Additions on the Microstructure and Mechanical Properties of Tin-Silver-Bismuth Solder, J. Electron. Mater., Vol 31(No. 6), p 564567
www.asminternational.org
Index
A
Abbreviations and symbols, 243 Abietic acid, 116 Acceptance criteria, 231 Acetic acid vapors, 132(F) for Pb-62Sn solders, 131 as reactive atmospheres for uxless soldering, 130 toxicity of, 131 Acetylene, 34 Acoustic microscopy, 204 Activated ller metals, 151 Activated solders, 8 Activation energy, 8, 24, 124 Activation temperature of uxes, 117 Active brazes, 103 Active ller metals for ceramics, 149 contact angle in, 17 high temperatures of, 149 Active hydride process, 151 Active solders, 147, 152153 Additives, 135 Adhesive bonding, 23 Adhesively bonded joints, 3, 173 Adhesives, 184 Aesthetic requirements, 50 Airborne particle size, 179(F) Alcohol solvent carrier, 118 Allotropic transformation, 77(F), 195196 Alloy constitution, dened, 78 Alloy Constitution (reference text), 78 Alloy J, 60, 61 Alternative atmospheres for oxide reduction, 111 Alumina, 121 alloy matching thermal expansion, 164 and gallium, 62 nonmetallic bonding to, 152 Aluminum and aluminum alloys Al-94Zn solders with, 54 brazes for, 6 corrosion of, 54 diffusion bonding of, 10 diffusion soldering for, 137 uxes for, 121122 heat capacity of, 62 as impurity, 76, 77 oxides on, 9 and temperature uniformity, 50 thermal conductivity of, 62 thermal expansivity of, 62 thermal heat capacity of, 50 zinc alloys with, 54 zinc-bearing solders for, 61 Aluminum uxes, 121 Aluminum-gallium-magnesium-zinc solders, specic types Al-3Ga-3Mg-90Zn, lead-free solders, 66 Al-3Ga-3Mg-90Zn, zinc-base solders as, 66 Aluminum-germanium eutectic alloy, 8 Aluminum nitride, 152 Aluminum quaternary alloy, 83 Aluminum-silicon alloys, 6 Aluminum-silicon alloys, specic type Al-12Si, additions for wetting and spreading, 134 Al-12Si, diffusion soldering processes with, 232 Al-70Si, alumina matched thermal expansion, 164 Aluminum-silicon carbide composites, 152, 175 Aluminum-zinc phase diagrams, 66(F) Aluminum-zinc solders, specic types Al-94Zn, with aluminum, 54 Al-94Zn, zinc alloys, 54 Al-94Zn, zinc-bearing solders, 62 Amalgams advantages of, 214 based on gallium, 216217 based on indium, 217 based on mercury, 215216 dened, 214 dental, stress-strain curve for, 215(F) as solders, 214217 solid and liquid metals evaluated for, 214(T) Ammonia uxes and intergranular corrosion with brasses, 114 Anisotropically conductive adhesives, 3 ANSI/J-STD-002 (test method), 207 Antimony added to lead-tin solders, 57 as additive to indium, 135 effects of, on surface tension of tin, 193 as impurity, 76 Pb-Sb-Sn ternary system, 57 as solder constituent, 5354 and solid-solution strengthening, 5354 Antimony-tin phase diagrams, 60(F) Application methods, 148 Argon, 36, 109 Arrhenius-type rate relationships, 24 Ashby materials selection chart, 159(F) Asthma, 43 Atmospheres. See also inert atmospheres categories of, for joining, 103 chemically active, described, 104 chemically inert, described, 104 controlled gas, 104 effect of, on spreading, 20 and uxes, 8 and heating method, 33
2004 ASM International. All Rights Reserved. Principles of Soldering (#06244G) 246 / Principles of Soldering
www.asminternational.org
Atmospheres (continued) joining, interrelationship of, 104(F) nonoxidizing, and heat treatment for wetting, 148 and oxide lm reduction, 105106 reactive gas, and oxide reduction, 130131 reduced pressure, and zinc bearing-solders, 62, 65 soldering, costs and benets of, 116(T) soldering, thermal conductivities of, 114(T) types of, 104 Atmospheric corrosion, 37 Atmospheric quality, 136(F) Atomic diffusion, 224 Atomic fraction of constituents, 96 Atomic hydrogen creation of, 127 reduction of silver oxide by, 130(F) solder oxides reduction by, 127128 Atomically clean, 50 ATV 2003, 127 Automated optical inspection (AOI), 237
B
Bakeout temperature, 36 Ball grid array integrated circuits (BGAICs), 179 Ball grid arrays (BGAs), 199, 204 Balling up, 6263 Balls (of lead-tin solders), 129(F) Barrier coatings, 133, 149 Barrier metal, 200 Barrier metallizations, 155 Basic spreading test, 211 Berthoud equation, 25 Beryllia, 152 Beryllia dust, 163 Beryllium, 104, 216 Bimetallic expansion, 50 Bimetallic strip, 158(F) Binary alloys and phase diagrams, 79 Binary compounds, 84 Bismuth added to lead-tin solders, 57 as additive to indium, 135 Bi-Pb-Sn ternary system, 57 effects of addition to, on liquidus and solidus temperatures of silver-tin off-eutectic alloys, 193(F) effects of addition to, on tensile strength of silver-tin solders, 193(F) effects of, on surface tension of tin, 193 expansion on freezing, 53 in hermetic soldered points, 53 as impurity, 76, 77 as solder constituent, 53 Bismuth-antimony-tin system, 96 Bismuth-bearing solders dissolution of chromium metallizations to, 149 inferior uidity of, 53 inorganic uxes for, 53 Bismuth-lead-tin ternary eutectic solders, 93
Bismuth-lead-tin ternary system, 57 Bismuth-silver-germanium alloys, specic type Bi-11Ag-0.05Ge ternary, 197198 Bismuth-tin lead-free solders, 122 Bismuth-tin phase diagrams, 56(F) Bismuth-tin solders, specic types Bi-43Sn, 31 Bi-43Sn, elongation, 194 Bi-43Sn, for hermetic joints, 173 Bi-43Sn, melting point of, 173 Bi-43Sn, physical properties of, 194 Boiling/sublimation temperatures, 107 Boiling, temperatures of, 109(T) Bolometers (thermal imaging), 35 Boltzmann constant, 24, 78, 98 Bond formation in pressure welding, 9 Bond quality, 212 Bond wire, 175 Bond wire interconnections, 202 Bonding process at low temperatures, 214 stages of, 9 Bonding temperature, peak, 38 Boron nitride, 152 Bow distortion of bimetallic assembly, 165 of bimetallic strip, 158(F) equation for, 157 Brasses and ammonia, 105 intergranular corrosion and ammonia ux, 114 vacuum atmospheres, and zinc metallization, 107 Braze alloy families and melting ranges, 7(F) temperature ranges of, 6 Braze alloy, specic type Al-4Cu-10Si, 6 Brazes and brazing ller metal temperatures in, 5 service temperature of, 8 British Standards (BS) 9430 (void free joints), 183 Brittle failure, 28 Brittle intermetallic layers, 28 Brittle joints, 52. See also embrittlement Brittle materials, 28 Brittle phases, 49 Bulk lter mechanical properties, 29 Bulk properties of solders, 225226 Bulletin of Alloy Phase Diagrams (phase diagrams), 79 Butt joints, 177, 177(F) Butt welding, 9
C
C-charts, 42 C4 process, 199 Cadmium and cadmium alloys health hazard of, 63 restrictions in use of, 50, 51
www.asminternational.org
Index / 247
toxicity of, 51, 191 Cadmium-base solders melting point depression, 93 melting points for, 93 Cadmium-indium-tin-zinc quaternary eutectic system, 93 Cadmium-indium-tin-zinc quaternary system, 94(T) Cadmium solders as substitutes for mercury, 93 Calibration standards, 212214 Capillary action, 23 Capillary dams, 23 Capillary ow in narrow gaps, 15 time for molten tin and copper in, 27(F) Capillary forces, 15, 15(F), 212 Carbon-carbon ber composites, 152, 163 Carbon dioxide, 109 Carbon-ber-loaded solders, 223 Carbon bers, 222 Carbon monoxide as reactive atmospheres for uxless soldering, 130 uses of as reducing gases, 109 Carboxylic acids, 116 Cast iron, 147 Cathodic sputtering, 180 CE7 alloy, 163 Ceramics active ller metals for, 149 soldering to, 149 wide-gap joints in, 158 zinc oxide, metallization of, 150 Cerium, 227 as additive to indium, 135 CFC-free cleaning operations, 40 CFCs. See chlorouorocarbons Chadwick peel tests, 73 Channeling, 22 Chemical cleaning, 37, 209 Chemical displacement, 180 Chemical uxes. See uxes Chemical properties of lead-free solders, 193196 Chemical reduction of metal oxides, 105 Chemical vapor deposition (CVD) metallization techniques, 180 on nonmetallic components, 149150 Chemically active atmospheres, 104 Chemically inert atmospheres, 104 Chip-scale packages (CSPs), 199 Chloride-based uxes, 121 Chlorides, 111 Chlorouorocarbons (CFCs), 39, 111, 115, 117, 118, 196 Chopped carbon bers, 222 Chrome oxide, 110 Chromium, 150 Chromium metallizations, 149 Clausius-Clapeyron equation, 97, 98 Clausius theorem, 138 Clean room class designation, 179(F)
Cleaning alternatives to, 41 benets and costs of, 41 CFC-free, 40 chemical, 37 costs of, 118 electronic assemblies, 40 measure of effectiveness of, 119120 mechanical, 37 methods of, 40 postjoining, 3941 by reverse-gas bias mode, 150 Cleaning agents chlorides as, 111 uorides as, 111 relative effectiveness of, 41(F) Cleaning treatments, 37 Cleanliness of IEC board test coupons, 119(F) Coatings gold, and embrittlement, 133 gold, shelf life of, 133, 148(T) non-metallic removal of, 114 onto component surfaces, 37 solderability shelf life of, 133 solderable, 149152 soluble, 149 storage shelf life of, 50 thickness of, 148 thin, by autocatalytic method, 180 types of, 149 vapor deposition, 234 Cobalt metallization, 71 Coefficient of thermal expansion (CTE) of carbon bers, 222 of components, 26 of copper-molybdenum alloys, 161 of copper-tungsten alloys, 161 of iron-nickel alloys, 160 and melting point for metals, 160(F) of metals, and their melting point, 159 of molybdenum, 160 of non-nickel alloys, 162(F) of Osprey controlled expansion alloys, 163(F) of titanium, 160 of tungsten, 160 Coelectroplating, 32 Cold compression welding, 200 Compatibility, characteristics of, for solder, 49 Compliant structures accommodating thermal expansivity difference, 165167, 166(F) economics of, 167 for mitigation of mismatch expansivity, 166(F) Component cleaning methods, 172 Component surfaces dissolution of, by brazing, 8 dissolution of, by soldering, 8 Component testing, 224 Composite materials controlled expansion materials, 163164
2004 ASM International. All Rights Reserved. Principles of Soldering (#06244G) 248 / Principles of Soldering
www.asminternational.org
Composite materials (continued) families of, 163 Composite solders high and low-cycle fatigue life of, 221(T) intermetallic compounds in, 219 iron containing, 222 with large particles, 220 preparation methods of, 220 of sessile drop tests, 220 tensile strength of, 221(T) types of, 220 wetting and spreading characteristics of, 220 wetting properties of, 221(T) wide joints and, 221 yield strength of, 220(F) Compositional equilibrium, 78 Compound formation predictions, 79 Compound semiconductors, 105 Compression bump bonding, 202(T) Compressive loading uxless soldering by, 135(F) and joint shear strength, 136(F) Concentration of solid metal in liquid metal, 25(F) Conductive adhesives, 3(T) Conductive polymeric materials, 201 Conductivity, 30 Constraints imposed on by components and solutions, 168179 joint area, 169173 Contact angle of Ag-96.5Sn solders, 228(F) of copper-silicon brazes, 16(F) effect of on llet formation joint lling, 17(F) effective, 22 and llet formation, 17 in lap joint, 176(F) on lead-tin solders, 18(F) of lead-tin solders, 22(F) measurement of, 211212 metallurgical modication of, 133 of Pb-60Sn solder, 20 and quality of wetting, 17 rare earth doping of indium solders, 228 in reactive wetting, 16 and spread factor, 4445 and spread factor and spread ratio, relationship between, 212(F) and spread ratio, 4445 temperature effects of, 25 time dependence of, 15 and wetting area, 1314 Contact angle (of droplets), 45 Control charts, 42 Controlled expansion materials alloys, 164(F) components of, 163 composite materials, 163164 copper-molybdenum alloys, 160161 copper-surface laminates, 162163 copper-tungsten alloys, 161162
interlayers, 164165 iron-nickel alloys, 160161 mechanical constraints and solutions, 159164 Cooling rate, 38 Cooling stages, 39, 39(F) Copper diffusion soldering of, 234 effects of, on surface tension of tin, 193 lead-free tin-base solders for, 116 lead-tin solders for, 116 nonmetallic bonding to, 152 Pb-63Sn solder wetting using rosin ux, 115(F) rate of dissolution of, in molten Pb-60Sn solder, 84 as wettable metallizations, 147 wetting of, by Pb-63Sn solder, 115(F) Copper abiet, 116 Copper-base alloys, 10 Copper-base brazes, 6 Copper coupons, 131(F) Copper-indium alloy systems, 234 Copper-indium intermetallic compounds, specic types Cu2In, for diffusion soldering, 234 Copper-Invar-copper laminates, 163 Copper-lead-tin phase diagrams, 8485 Copper-lead-tin system diagram sector of, 86(F) isothermal section of, 86(F) liquidus surface of, 85(F) Copper metallizations, 132(F) Copper-molybdenum alloys coefficient of thermal expansion (CTE) of, 161 as composite materials, 163 controlled expansion materials, 160161 Copper-molybdenum-copper laminates, 163 Copper-nickel-tin phase diagrams, 84 Copper particle reinforcement, 220 Copper powder, 216 Copper-surface laminates, 162163 Copper-tin alloy systems, 234 Copper-tin intermetallic compounds effect of thickness on, 88(F) growth of, 87(F), 88(F) interfacial, rate of formation of, 87 from lead-tin solders, 84 presence of thick layers of, 88 properties of, 87 Copper-tin intermetallic compounds, specic types Cu10Sn, for diffusion soldering, 234 Cu3Sn, binary compounds, 87 Cu3Sn, precipitates at copper/solder interface, 89 Cu6Sn5, as stoichiometric compounds, 91 Cu6Sn5, binary compounds, 84, 85 Copper-tin phase diagrams, 85(F) Copper-tin solders, specic types Cu6Sn reinforced, 220 Copper-to-aluminum direct bonding, 162 Copper-to-copper joint, 234 Copper-tungsten alloys coefficient of thermal expansion (CTE) of, 161 as composite materials, 163
www.asminternational.org
Index / 249
controlled expansion materials, 161162 Copper-zinc compounds, 64 Cored silver, 80 Corrosion atmospheric, 37 in joints, 49 mechanisms of, 29 of Sn-9Zn eutectic alloy, 96 Corrosion resistance of lead-free solders, 195 Costs. See also economics of 3.6Ag-1.6Au-92.8Sn eutectic solders, 96 of cleaning, 41, 118 of forming gas, 111 of furnace joining, 104 of gold-germanium vs. gold-silicon solders, 70 of lead-free solders, 192 of low-alpha lead, 190191 of machining molybdenum, 160 of machining tungsten, 160 of nitrogen, 109 of precious metals in metallizations, 231 of solder coated substrates, 134 of solder elements, 52 Costs and benets of soldering atmospheres, 118(T) Crack repair, 179 Cracked ammonia, 109 Cracks from low-cycle-fatigue, 74 orientation of, and x-rays, 236 from volume contraction, 67 Creep and heat treating, 158 of indium-base solders, 73 to relieve mechanical stress, 178 resistance of metals to, 195 and strain rate measurement, 194 thermal cycling without, 165 types of, 195 Creep behavior of indium solder alloys, 158 Creep curve for lead-tin eutectic solders, 218(F) Creep properties of composite solders, 221(T) Creep rates of various solders, 219(T) Creep resistance effect of metal additions on, 220 of lead-tin eutectic solders, 219 of solder, 217 Creep rupture of indium-base solders, 73 and intermetallic thickness, 88(F) Creep stress, 10 Critical angle, 223(T) Critical temperatures, 107 Curie point, 160 Curing curves, 217(F)
D
De Gennes model, 20, 22 Decomposition reaction, 180
Defect-recognition software, 189 Defective items, 42 Defects. See also tests/testing; voids distortion, 38, 50, 168170 dross formation, 115, 115(F) inspection of, 235 joint embrittlement, 52, 90, 153 levels of, and oxide thickness effect of, 124(F) in liquid solder lm, 134 rates of, in joints, 42 and x-ray inspection, 204 Delay effect on joint strength, from cleaning to assembly, 131(F) Dendrites arm spacing and tensile strength, 32(F) bridging, 120 on circuit board, 120(F) growth of, 119120 primary, of silver, 80 Dendritic growth, 120(F) Dental amalgams, 215, 215(F) Depression of melting point, 9396 Dermatitis, 43 Design criteria of soldering processes, 2830 Design guidelines for various metals for in-plane alignment, 205 Diamond as intermediate plate material, 165 nonmetallic bonding to, 152 Die attach of gold-metallized chips, 71 Differential scanning calorimeter, 155 Differential thermal expansion steps to reduce stress from, 158 stress from, 157 Diffuse heating, 33 Diffusion activation energy for, 78 rate of, 78 surface, 10 volume, 10 Diffusion bonding of aluminum-base alloys, 10 of copper-base alloys, 10 of gold, 45, 43 of gold, temperature/pressure curve for, 43(F) of indium, 45 of indium, temperature/pressure curve for, 44(F) interlayers for dissimilar metals, 10 limitations of, 230 process, 9 with standard solders, 43 of tantalum, 10 of titanium, 10 of titanium alloys, 10 of tungsten, 10 of various metals, 10 Diffusion brazing, 10, 175 Diffusion rates inequality of, 10 in solids, 83
2004 ASM International. All Rights Reserved. Principles of Soldering (#06244G) 250 / Principles of Soldering
www.asminternational.org
Diffusion reaction, 215 Diffusion-soldered joints micrographs of, 233(F) steps involving, 232(F) strength of, 232 Diffusion soldering, 7 advantages of, 230 for aluminum, 137 binary combinations for, 231 of copper, 234 copper-indium alloy systems for, 234 copper-tin alloy systems for, 234 Cu10Sn intermetallic phase for, 234 described, 230 of gold, 233234 interlayers in, 10 modeling of, 235 practical aspects of, 234235 process principles of, 230231 of silver, 231232 Diffusion-soldering processes with Al-12Si alloy, 232 basis of, 80 for gold jewelry, 233 jewelry made by, 234(F) with silver-indium alloys, 232 Diffusion soldering systems, 231(T) Dimensional stability of soldered joints, 224226 DIN WNr 1.3981 (iron-nickel alloys), 160 Dip-and-look (DNL) test, 207 Dipping methods, 148 Direct-bonded copper, 162 Direct chip attach (DCA), 199 Disc preforms, 170 Dislocation climb, 215 Dispersion-stabilized alloys, 175 Dispersion-strengthened solders, 219(T) Dispersion strengthening, 218, 219 Dispersoids, 218, 219, 221 Disproportion reaction, 180 Dissimilar materials and thermal cycling, joints with, 73 Dissimilar-metal joints, 9 Dissimilar metals, 10 Dissolution of chromium metallizations to bismuth-containing solders, 149 of component surfaces, 8 of parent materials, 153 rate of, 2425 rates of, 153 Dissolution of parent materials and intermetallic growth, 2425 Dissolution rate of copper in molten Pb-60Sn solders, 84, 89 of gold in molten lead-tin solder, 89 of metals and metallizations in lead-tin solders, 87(F) of platinum, in gold-tin eutectic solders, 71 of platinum, in tin-base solders, 148(T) and saturation limit, 25
of silver with lead-tin solders, 53(F) Dissolution rate constant, 24 Distortion of assemblies, causes of, 168170 from bimetallic expansion, 50 during heating, 38 from thermal expansion mismatch strain, 50 Doping, 227229 Doping additions, 135(F) Drop-in replacement, 191 Dross formation interference of, with wetting and spreading, 115 oxygen concentration and rate of, 115(F) Ductile foils, 155 Ductility improvement of, in ller metals, 155 of indium-base solders, 73 of indium-bearing solders, 52 of silver-tin intermetallic phases, 52 of tin-based solders, 52 of zinc-based alloys, 64 Duration of diffusion bonding process, 9 Dwell stages, 38
E
Economics. See also costs of compliant structures, 167 of uxless soldering, 123 of forming gas, 111 of joining process, 145 of lead-free solders, 192 of low oxygen atmospheres, 115 of thin foil preforms, 136 of vapor-phase techniques, 171 Economics and availability of lead-free solders, 191193 Edge llets examination, 168 Effective contact angle, 22 EIA/IS-86 (test method), 207 Elastic modulus, 194 Electric conductance of mechanical fasteners, 12 Electrical conductivity of adhesive joints, 3 of composite solders, 221 joint requirements, 30 Electrical properties of joints, 50 Electrically conductive adhesives, 3 Electrode potential of selected elements, 54(F) Electrodeposition, 155 Electronic assemblies cleaning, 40 joining methods for, 116 Electroplating difficulties with, 234 as manner of solder deposition, 31 in metallizations, 37 tin-lead solder, 32
www.asminternational.org
Index / 251
Ellingham diagram adjustments for plasma state, 127 application of, 107111 free energy change for oxidation of several metals, 106(F) for oxides with hydrogen at a partial pressure, 129(F) for selected oxides, 108(F) temperature and O/H ratio for metal oxide reduction, 110(F) Elongation, 194 Elongation to failure of Ag-95Sn solders, 228(F) of Sn-9Zn solders, 228(F) Embrittlement caused by impurities, 49 of copper-zinc compounds, 64 and gold-based coatings, 133 liquid metal, 30 of solder, 23 tin-based solders and AuSn4, 133 Embrittling phases, 4 Endoscopes, 237 Engineering ceramics thermal expansion, 159 Enthalpy of fusion, 97 Entropy changes in, 9899 in chemical bond, 105 dened, 137, 138, 139 of fusion, 97 Environmental concerns, 191 Environmental considerations. See also health and safety issues chlorouorocarbons (CFCs), 196 volatile organic compounds (VOCs), 115 Environmental durability, 29 Environmentally friendly chemicals as cleaning agents, 111 Equilibrium constant and oxidizing reaction, 141 Equilibrium contact angle, 1718 Equilibrium partial pressure of oxygen. See oxygen partial pressure Equilibrium phase diagrams, 25 Erosion changes to rate of, in parent materials, 154 conditions for, 153 of gold, by molten indium, 81 of gold by molten tin, 82(F) of parent materials, 49 of silver by molten tin, 82(F) of substrate surfaces, 33 Erosion of parent materials, 153154 European Space Agency (ESA) Specications (void free joints), 183 Eutectic alloys and alloying benets of, 81 grain renement of, 81 melting point depression by, 9396 melting point of, 81 vs. noneutectic alloys, 20 spreading characteristics of, 19
spreading of, 81 theoretical modeling of, 9799 used for solders and brazes, 6 Eutectic solders composition of, 8 melting temperatures of, 52 Eutectiferous character of alloys, 51 Evolved vapor, 172 Exchange reactions, 122 Expansion coefficients. See coefficient of thermal expansion (CTE) Expansion mismatch compliant structures for mitigation of, 166(F) and cooling stages, 39(F) steps to reduce stress, 158 Expansion on freezing, 53 Explosion risk, 110 Explosive limit, 111 Exposure limits for hazardous dusts and vapors, 43 External magnetic elds of iron powder solders, 221 Extrapolation of results in spreading test, 212 Eye and nose irritation, 43
F
Fast atom bombardment, 128, 180 Fast atom cleaning, 131(F) Fatigue cracks internal initiation of, 217 wide-gap joints and, 217 Fatigue failure, 157 Fatigue fracture, 165 Fatigue life, 200 of carbon-ber-loaded solders, 223 Fatigue resistance of Alloy J, 61 high-cycle, 220 indium solder alloys and, 158 low-cycle, 220 ranks of lead-free solders, 195 of solder, 217 for soldered joints, 4 Fatigue situations, joint reliability in, 205 Fatigue theory, 226 Ferromagnetic alloys, 160 Filler metal temperatures in brazing and soldering, 5 Filler metals and dissolution with component pieces, 26 form of, 3133 with limited component solubility, 2526 molten, ow inuences to, 12 molten, solidication shrinkage of, 169 partitioned, 155(T) partitioning of, 155157 spread characteristics of, 12 spreading of, 20 surface area to volume, ratio of, 133 for welding, 4 Filler spreading characteristics, 1922
2004 ASM International. All Rights Reserved. Principles of Soldering (#06244G) 252 / Principles of Soldering
www.asminternational.org
Fillet formation for aesthetic requirements, 50 Fillet radius, 176(F) Fillets benecial effects of, 167 enhancements of, 167 with frosty appearance, 81 inspection of, 238 integrity of, 168 quality of, 29 to reduce stress concentrations, 176 role of, 167168 and stress concentration reduction, 168(F) and stress concentrations, 167 as stress reducers, 4 Finishes, 149 Finite element analysis (FEA) of ceramic-metal brazed joint, 225(F), 226(T) modeling assumptions in, 225 Fire risks, 43 First Law of Thermodynamics, 137138 Flammability, 43 Flip-chip assembly operation, 123 Flip-chip bonding, 200(F) Flip-chip bonding process dened, 201 described, 166 Pb-5Sn solders for, 166 Flip-chip components pull-in alignment, 132(F) Flip-chip daisy chain, 74(F) Flip-chip inspection methods, 204 Flip-chip interconnection high-temperature, die mounted using, 206(F) self-alignment of, 202(F) self-alignment of solder bumps during, 127 thermal expansion mismatch in, 200 Flip-chip interconnects, 179 Flip-chip joining process, 167(F) Flip-chip joints with high-lead solders, 129 Flip-chip lands, 204(F) Flip-chip process ow for range of solders, in different atmospheres, 199(T) Flip-chip processes Au-20Sn eutectic solders for, 207 interconnection schemes for, 201(F) lead-tin eutectic solders for, 207 step soldered ip-chip interconnects, 206207 surface topology of, 206 Flip-chip solder bumps, 204(F) Flip-chip structures, 204206 Flip-chip technology characteristics of, 202203 rework, 204 underll, 203 Flow, molten ller metal inuences to, 12 Fluid ow, 1819 Fluidity inferior levels of, with bismuth-bearing solders, 53 of molten ller metals, 49 Fluorides, 111 Fluorine chemistry, 131
Flux action mechanisms, 114 Flux activators, 118 Flux boil, 123 Flux carriers, 117 Flux chemistry, 116 Flux-cored solder wire, 31, 114 Flux-formulations, 114 Flux removal, 39 Flux residues, 123 Flux vapors, 2526 Fluxes activation temperature of, 117 activity of, 116117 for aluminum, 121122 and atmospheres, 8 based on organic compounds, 66 chemical activity of, 118 chemical, described, 111112 chloride-based uxes for, 121122 classication of, 118(T) commercial designations of, 117 containing 2-ethylhexanoate, 195 with gold-tin eutectic solders, 122 with high-lead solders, 122 high-molecular-weight hydrocarbons as, 122 high-temperature, 122123 IA type, 117 ingredients in, 116 for lead-tin eutectic solders, 122 for magnesium, 122 no-clean, 39, 40, 118119 OA type, 117 organic, 121 with oxide scale, 34 R, RMA, RA type, 117 requirements for, 113114 role of, in wetting and spreading, 114 SA type, 117 for stainless steels, 122 that require cleaning, 116118 for tin-base solders, 116120 for unsolderable materials, 120122 water soluble, 39, 40 WS type, 118 for zinc-bearing solders, 6566 Fluxless joining gold as solderable metallization, 89 using Au-20Sn eutectic solders, 71 Fluxless joining process, 3031 Fluxless process, 111 Fluxless soldering of aluminum, 136137 by compressive loading, 135(F) economics of, 123 of gallium arsenide (GaAs), 134 gold coating for, 124 process considerations for, 132133 reactive atmospheres for uxless soldering, 130 using 82Au-18In solders, 72 using In-48Sn solder, 135136
www.asminternational.org
Index / 253
Fluxless wetting on copper metallizations, 132(F) on gold-on nickel metallizations, 132(F) of Pb-62Sn solder, 132(F) Fluxless wetting angle of indium-silver substrates, 135(F) Foil preforms deciencies of, 171 production of by solidication casting, 32(F) production of by strip casting, 32(F) and spreading tests, 212 Foils and sheets, 23, 31 Force diagram for immersed solid plate, 209(F) Formic acid vapor, 111 concentration, effect of, 132(F) as reactive atmospheres for uxless soldering, 130131 in self-alignment of ip chip assembly, 130 Forming gas, 111 Foundation layers, 150 Foundation metal, 199 Free energies comparative values of metal oxide formation, 105(T) of oxide formation, 125 Free-energy change for oxidation of several metals, 106(T) for oxidation reactions, 106 Free energy of formation, 110 Fuel gases, 34(T) Furnace joining, 104 Fusible coatings, 149 Fusion, 97
G
Gallium and alumina, 62 amalgams based on, 215216 liquid alloys based on, 93 Gallium alloy systems, 93 Gallium arsenide (GaAs), 134 Gallium-base amalgams, 216 Gallium-copper amalgams, 216(T) Gallium-indium-tin alloys Ga-In-Sn, melting points of, 93 Gallium-indium-tin solders Ga-In-Sn, effect of additions to, 94(T) Gallium-nickel-copper amalgams, specic types Ga-5Ni-30Cu, 216 Ga-5Ni-30Cu, curing curves for, 217(F) Gamma phase intermetallic compound, 215 Gap size, 178179 Gaps and x-ray inspection, 204 Gas atomization, 220 Gas evolution from polymeric materials, 172 Gas Law, 139 Gas torch, 34 Gaseous uxes as uxless process, 111 narrow joints with, 178
Gaseous reagents, 140 Geometry of adhesive joints, 3 Germanium, 77 Gibbs free energy changes in, 105, 141 dened, 137 depression of, as a function of temperature increases, 98 pressure dependence of, 139141 reference point of, 140 Gibbs free energy function (G), 16, 139 Gold as additive to indium, 135 characteristics of, 43 as constituent of high-melting-point solders, 51 critical levels of, 91 diffusion bonding of, 45, 43 diffusion soldering of, 233234 effect of, on tin-base eutectic solders, 153(T) effects of addition to Ag-97.5Pb-1Sn solder, 154(F) effects of addition to In-18Pb-70Sn solder, 154(F) erosion of, by molten indium, 81 erosion of, by molten tin, 82(F) high rate of dissolution of, in molten lead-tin solder, 89 melting point of, 54 in oxidizing atmospheres, 35 as solder constituent, 54 as solderable metallization, 89 temperature/pressure curve for diffusion bonding of, 43(F) as wettable metallizations, 147 Gold-antimony alloys, 66 Gold-antimony phase diagrams, 69(F) Gold-base brazes, 66 Gold-base solders, 111 Gold-bearing solders for gold-metallized components, 66 melting point of, 66 used as solders, 67(F) Gold-coated components electrodeposition of, 155 solderability shelf life of, 148(T) Gold coating, 133 Gold ash, 10, 147148 Gold-germanium, 6671 Gold-germanium eutectic alloy, 70 Gold-germanium phase diagrams, 68(F) Gold-germanium solders, specic types Au-12Ge, characteristics of, 70 Au-12Ge, contact angle of, 72(T) Au-12Ge, iron germides, 70 Gold-indium intermetallic phases, 83(F) Gold-indium joints, 231 Gold-indium noneutectic alloys 82Au-18In solders, 72 Gold-indium phase diagrams, 52(F) Gold-indium solders, specic types 82Au-18In, uxless soldering, 72 82Au-18In, gold-indium noneutectic alloys, 72 82Au-18In, highest melting-point solders, 72
2004 ASM International. All Rights Reserved. Principles of Soldering (#06244G) 254 / Principles of Soldering
www.asminternational.org
Gold-indium solders, specic types (continued) Au-18In, contact angle of, 72(T) Gold jewelry, 233 Gold layers in metallization, 150 Gold-lead-tin ternary system gold limit in, 89 liquidus projection of, 89(F) liquidus surface of, in phase diagrams, 89 vertical section through, 90(F) Gold limit, 89 Gold metallizations characteristics of, 89 erosion of, by molten indium, 53(F) and indium-bearing solders, 5253 Gold-metallized components gold-bearing solders for, 66 In-48Sn solder to, 136(F) Gold-on-nickel metallizations, 132(F) Gold-plated tin foil, 156(F) Gold-silicon, 6671 Gold-silicon intermetallics, 66 Gold-silicon phase diagrams, 67(F) Gold-silicon solders Au-2wt%Si, 66 cost advantage of gold-germanium eutectic alloy, 6970 as a desiccant for hermetic packaging, 184 for silicon semiconductor chips to gold-metallized pads, 66 vapor-phase technique, 66 Gold-silicon solders, specic types Au-2Si, 71 Au-2Si, spreading behavior of, 69(F) Au-2Si, titanium as additive to, 69 Au-2Si, zinc as impurity of, 77 Au-2wt%Si, alloy additions for spreading, 67 Au-2wt%Si, gold-silicon solders, 66 Au-2wt%Si, high molten viscosity of, 66 Au-2wt%Si, silaceous dross, 66 Au-2wt%Si, tin as alloy element for, 68 Gold-silicon-tin alloy system phase diagrams, 69 Gold-silicon-tin phase diagram, 70(F) Gold-silicon-tin-titanium solders, specic types Au-2Si-8Sn-1Ti, silicon wetting, 69 Gold-tin, 68(F), 7172 Gold-tin alloy system, 233 Gold-tin alloys, specic type Au-20Sn, wetting effect or rare earth doping, 228229 Gold-tin eutectic solders uxes with, 122 high-melting-point solders, 72 intermetallic compounds of, 226 Gold-tin intermetallic compounds, 32 AuSn4, and embrittlement, 133 AuSn4, and tin-base solders, 133 AuSn4, as stoichiometric compounds, 91 Gold-tin intermetallic compounds, specic types Au-Sn4 phase, and joint embrittlement, 90 Au-Sn4 phase, in Au-Pb-Sn ternary system, 89 Gold-tin partitioned ller metals, 155 Gold-tin phases and joint embrittlement, 90
Gold-tin solder, 32 Gold-tin solders, specic types Au-20Sn, application of, 71 Au-20Sn, bismuth as impurity of, 77 Au-20Sn, cobalt metallization for, 71 Au-20Sn, fabrication costs of, 155 Au-20Sn, uxless joining of, 71 Au-20Sn, foils and preforms of, 71 Au-20Sn, for ip-chip process, 207 Au-20Sn, for soldering to gallium arsenide, 153 Au-20Sn, forms of, 71 Au-20Sn, hermetic sealing of ceramic semiconductor packages, 71 Au-20Sn, high-lead solders as alternate to, 72 Au-20Sn, intermetallic compounds of, 226 Au-20Sn, melted in controlled atmosphere, 113(F) Au-20Sn, palladium metallization for, 71 Au-20Sn, rapid solidication casting, 71 Au-20Sn, titanium as additive to, 69 Au-30Sn, foil melting point, 156(F) Graduated joint structures, 165 Grain-boundary sliding, 215 Grain renement, 81, 218 Graphite brazing, 30 Green manufacturing, 190
H
Halide atmospheres, 111 Halide uxes, 117 Hallmarking regulations, 50 Halogen-base uxes, 122 Halogen gases, 130 Health and safety issues asthma, 43 from beryllia dust, 163 of beryllium, 216 of cadmium and cadmium alloys, 50, 51, 63, 191 of electronics equipment disposal, 190 explosion risk of hydrogen atmospheres, 110 exposure limits for hazardous dusts and vapors, 43 eye and nose irritation, 43 ammability, 43 of hydrogen, 37 of lead, 50 of mercury, 93, 191, 215 of nickel, 50 of organic metal compounds toxicity, 131 soldering fumes, 43 of thallium, 191 Heat-affected zone (HAZ), 4 Heat capacity, 62 Heat treatment and creep, 158 in non-oxidizing atmosphere, 148 vs. oxygen concentration, 214(F) prior to joining, 3738 stress relaxation, 158 temperature of, 39
www.asminternational.org
Index / 255
Heat treatment temperature, 213 Heating cycles of joining operations, 3839 parameters of, 38 proles of, 38(F) Heating methods, 3334 Heating rate, 38 Heavy metal deposition, 122 Helium, 184 Helium leak rate, 184 Hermetic sealing Bi-43Sn solders for, 173 of ceramic semiconductor packages, 71 with compressive force, 174 dryness of, 184185 In-48Sn solder, 44 Sn-40Pb solder, 44 with standard solders, 43 High creep levels of high-lead solders, 72 High-cycle fatigue life of composite solders, 221(T) High-cycle fatigue resistance, 220 High-lead solder alternative, 197 High-lead solders 98Pb-2Sn solders, 73 as alternate to Au-20Sn solders, 72 uxes with, 122 high creep levels of, 72 mechanical properties and corrosion resistance of, 73(T) mechanical properties of, 73 step-soldering sequence for, 72 High-melting-point solders, 54 constituents of, 51 eutectic alloys of, 7(F) fast atom cleaning effect of, 131(F) lead-free, 197198 lead rich, 72(T) for thick lm metallizations, 72 High-molecular-weight hydrocarbons, 122 High-purity solders, 77 High-volume contraction, 64 Highest melting-point solders, 72 Highly active elements, 103 Holding time, 38 Homologous temperatures, 6 Hookes law, 2728 Hot shortness, 26 Hydrochloric acid (HCL), 116 Hydrogen explosion risks of, 37 in inert atmospheres, 36 solder oxides reduction by, 126127 uses of as reducing gases, 109 Hydrogen atmospheres and explosion risk, 110 and gold-base solders, 111 oxides with hydrogen at a partial pressure, 129(F) tin oxide reduction in, 111 Hydrogen plasma, 128 Hydrogen poisoning, 105
Hydrogen safety, 127 Hydrostatic forces, 15, 27, 212 Hydrostatic pressure, 178179 Hypereutectic alloys, 19 Hypoeutectic alloys, 19
I
IA type uxes, 117 Ideal substrate, 19 Image analysis, 211 Immersion plating, 180 Impurity aluminum as, 76, 77 antimony as, 76 bismuth as, 76, 77 cobalt and iron as, 77 germanium as, 77 silver as, 77 tin as, 77 zinc as, 77 Indentation welding, 9 Indicative property values of selected solders and pure metals, 226(T) Indium amalgams based on, 217 diffusion bonding of, 45 uidity of, 135 melting point of, 75(F) molten, and gold erosion, 81 in pressure-welded joints, 43 production levels of, 192 pure, as a solder, 74 temperature/pressure curve for diffusion bonding of, 44(F) and tin in intermetallic compounds, 51 Indium amalgams, 217 Indium and indium alloys, 201 Indium and lead, 5253 Indium-base solders alloy additions to, 74 benets of platinum in, 147148 composites and melting points of, 75(T) creep rupture of, 73 ductility of, 73 intermediate melting temperature solders, 75 low melting point of, 73, 74 in optoelectronic applications, 74 phase segregation failure of, 74 in photonic applications, 74 rare earth doping and contact angles of, 228 superheats for, 75 Indium-bearing solders ductility of, 52 and gold-metallizations, 5253 Indium bump bonding, 200 Indium-gold reaction, 82 Indium-lead alloys non-eutectiferous character of, 51
2004 ASM International. All Rights Reserved. Principles of Soldering (#06244G) 256 / Principles of Soldering
www.asminternational.org
Indium-lead alloys (continued) peritectic reaction, 75 Indium-lead phase diagrams, 61(F) Indium-lead-silver solders, specic types In-15Pb-5Ag, 172(F) In-15Pb-5Ag, indium-base solders, 75 In-15Pb-5Ag, intermediate melting temperature solders, 75 In-15Pb-5Ag, thermal fatigue performance of, 75 Indium-lead-tin solders, specic types In-18Pb-70Sn, gold addition, effect of, 154(F) In-18Pb-70Sn, gold addition effects on, 154 Indium oxides alloying additions to facilitate removal of, 7475 stability of, 125 temperature for reduction of, 126 Indium-silver substrates, 135(F) Indium solder alloys creep behavior of, 158 fatigue resistance and, 158 Indium solders, 7375, 116 Indium-tin phase diagrams, 55(F) Indium-tin solders for magnesium, 122 Indium-tin solders, specic types In-48Sn binary, melting point of, 96 In-48Sn, contact angle for, 128(F) In-48Sn, continuum between stress strain and creep data for, 74(F) In-48Sn, elongation, 194 In-48Sn, uxless soldering, 135136 In-48Sn, hermetically sealed enclosures-hermetically sealed enclosures, 44 In-48Sn, joining operations with, 135 In-48Sn, oxide growth on, 125 In-48Sn, preform of, 136(F) In-48Sn, stress-strain curve for, 74(F) In-48Sn, to gold-metallized components, 136(F) Indium-tin-zinc solders, specic types 5In-87Sn-8Zn, grain renement in, 218 In-46Sn-2Zn ternary, melting point of, 96 In-86Sn-9Zn, melting point of, 96 Inert gas atmospheres described, 3536 industrial quality, 107 soldering in, 107109 types of gases in, 36 uses of, 109 Infrared microscopy, 204 Inorganic acid uxes, 117 Inorganic acids, 116 Inorganic uxes, 53 Inspections of joint interior, 183 Instantaneous melting properties, 93 Interatomic force per unit area, 27 variations of, 27(F) Interdiffusion, 148149 Interfacial compound formation, 84 Interfacial compounds, 26, 103
Interfacial reactions, 19, 24 Intergranular brittle fracture, 215 Intergranular corrosion, 114 Interlayers controlled expansion materials, 164165 for diffusion bond dissimilar metals, 10 gold ash, 10 nickel foil, 10 silver foil, 10 Intermetallic compounds barrier metallizations to avoid, 155 in composite solders, 219 duration of growth, 25 effects of, 153 elastic modulus of, 154 formation of, 10, 51, 154 gamma phase, 215 growth of, 16, 25 indium and tin in, 51 mechanical and physical properties of, 87(T) in solders, 5 in ternary systems, 84 thickness of, and creep rupture, 88, 88(F) Intermetallic phase layer, 16 Internal energy, 137138 International Electrotechnical Commission (IEC), 119(F) International Organization for Standards ux classications of, 117 soldering ux classications, 118(T) International Programme for Alloy Phase Diagram Data (IPADA), 79 Invar, 160 Invariant reactions, 193 Ion-aided deposition process, 180 Ion-assisted vapor spray deposition, 178 Ion plating, 180 Ionic contamination, 119 Ionized-cluster beam deposition, 180 IPC/EIA J-STD-003A (test method), 207 Iron and tin dendrites, 77 Iron germides, 70 Iron-nickel alloys controlled expansion materials, 160161 machinability of, 161 trade names of, 160 Iron-nickel alloys, specic type 17Co-54Fe-29Ni, coefficient of thermal expansion (CTE), 160 17Co-54Fe-29Ni, iron-nickel alloys, 160 17Co-54Fe-29Ni, Kovar, 160 Fe-36Ni, coefficient of thermal expansion (CTE), 161 Fe-36Ni, thermal expansion characteristics of, 162(F) Fe-36Ni, total expansion of, 162(F) Fe-42Ni, coefficient of thermal expansion (CTE), 161 Iron-tin intermetallic compound, specic types FeSn2, as stoichiometric compounds, 91 Irreversible processes, 139
www.asminternational.org
Index / 257
J
Jetting system schematic, 201(F) Jewelry, 175, 233 Jigging, 3031 Joining atmospheres, 35, 103111 Joining environments, 103141 Joining methods, 112, 2(F) Joining operations, heating cycle of, 3839 Joining process cost tolerance of, 145 uxless, 3031 Joining process development, 146(T) Joining temperature, minimum practical (liquidus), 145 Joint defect rates, 42 Joint design dimensions, 12 dimensions and mechanical properties, 25 issues regarding, 30 to minimizes concentration of stresses, 175178 remedy of problems in, 146 strengthened solders, 178 trials of, 145 Joint embrittlement effects of, 153 effects of intermetallic phases on, 153 and gold-tin phases, 90 of indium-bearing solders, 52 of tin-based solders, 52 Joint lling requirements, 26 Joint gaps compressive forces applied to, 134 control of, 26 limits to, 2526 optimal balance of, 26 self regulation of, 27 signicance of, 2527 size of, 26 upper practical limit to, 26 voids in, 169(F), 170(F) Joint geometries, 29 for brazed joints, 4 length vs. void content, 169(F) of mechanical fasteners, 1 for soldered joints, 4 for welding, 4 Joint integrity advances in techniques for assessing, 235238 optical inspection, 237238 scanning acoustic microscopy, 235236 ultrasonic inspection, 235236 x-radiography, 236237 Joint quality degradation and power cycling, 65(F) Joint strength of Ag-96.5Sn eutectic solders, 232 of Ag3Sn intermetallic phase, 232 effect on, from delay from cleaning to assembly, 131(F) and wetting of component surface, 136 Joint weakness, 28
Joints, 175178 butt joints, 177, 177(F) cleaning for solid-state joining, 5 cleanliness of, 4 corrosion in, 49 dimensional stability of, 224226 with dissimilar materials and thermal cycling, 73 tness for purpose tests, 224 landed butt, 177 lap joints, 175, 176(F), 176177, 177(F), 229 large area, 168, 171(F), 174(F) lifetime prediction of, 226227 measurement of mechanical properties of, 223224 mechanical properties of, 26, 224 modeling lifetime of, 223 narrow, 178 numerical modeling of, 224227 peel force proles of, 168(F) quality of, 12 recommended designs for, 177(F) scarf butt joints, 177 shear strength and compressive loading, 136(F) shear strength as a function of thickness, 178(F) shear strength of, 136(F), 178(F), 222 solidication shrinkage, 173 step butt joints, 177 strap joints, 177, 177(F) strength factors inuencing, 224 surface roughness of, 22 tensile strength of, 222 tongue and groove joints, 177 trapped gas, 169173 trapped gas sweeping, 170(F) voids in, 49 wide gaps in, 158 Journal of Phase Equilibria (phase diagrams), 79
K
Kelvin-Planck statement, 138 Kinetics of reaction, 153 Kirkendall porosity, 10 Kirkendall voids, 71, 234 Kovar, 160
L
Landed butt joints, 177 Lanthanum, 227 Lap joints, 176177 failure in, 176(F) geometry of, 176(F) recommended designs for, 177(F) shear stress on, 175 strength of, and rare earth doping, 229 stress distribution in, 176(F) Large-area joints denition of, 168
2004 ASM International. All Rights Reserved. Principles of Soldering (#06244G) 258 / Principles of Soldering
www.asminternational.org
Large-area joints (continued) due to trapped gas, 174(F) pressure variation and void levels, 174(F) vapor pockets in, 171(F) voiding in, 171(F) Latent heat of fusion, 97 Lattice waves, 226 Lead as constituent of high-melting-point solders, 51 effects of, on surface tension of tin, 193 health restrictions on, 50 in landlls, 190 pure, as solder, 7273 Lead-antimony-tin system, 62(F) Lead-antimony-tin ternary system, 57 Lead-free solder paste, 155 Lead-free solders Ag-20In-77Sn alloy for, 96 Al-3Ga-3Mg-90Zn solders, 66 Alloy J, 60 availability of potential alloying elements for, 192(T) composition ranges for, 192(T) corrosion resistance of, 195 costs of, 192 drive for, 190191 economics and availability of, 191193 economics of, 192 elongation of, 194 fatigue resistance ranking of, 195 and uxes, 196 health, safety and environmental aspects of, 191 high-melting point, 197198 history of, 190 In-48Sn binary eutectic solders, 96 In-86Sn-9Zn solders, 96 lead-tin solder, alternatives to, 191193 lead-tin solder, compatibility with, 191 literature on, 189190 mechanical properties of, 194195 melting point depression, implications for, 9596 melting ranges of, 191, 196 metallurgical, physical, and chemical properties of, 193196 other physical properties of, 194 physical and chemical characteristics of, 191 plastic ow of, 194 for plumbing applications, 51 for printed circuit board (PCB) components, 95 process window for, 196197 reduction in superheat for, 196 shear strength of, 194 silver-copper-tin ternary phase equilibria, 193 strain rate of, 194 surface tension of, 193194 tin-based solders as key to replacement for, 192 tin pest and tin whiskers, 195196 ultimate tensile strength of, 194 wetting and spreading characteristics of, 197 yield strength of, 194 Lead-free tin-base solders, 116
Lead-rich alloys, 6 Lead-silver solders, specic types 1.5Ag-92.5Pb-5Sn, micrograph of, 73(F) Lead-tin alloys as compression interconnects, 201 tensile strength of, 81(F) Lead-tin-copper solders, specic types 61.75Sn-38.05Pb-0.2Cu(wt%), lead-tin eutectic solders, 84 Lead-tin intermetallic compounds, specic types Pd3Sn2, with palladium metallization, 71 Lead-tin phase diagrams, 58(F) Lead-tin solders additions to, 57 advantages of, 56 antimony in, 57 application of hydrogen plasma to, 128 balls of, 129(F) bismuth in, 57 change of dissolution rate of silver, 154 characteristics of, 191 contact angle of, 22(F) for copper, 116 and copper substrate, reaction phases formed by, 87(F) copper-tin intermetallics from, 84 creep curve for, 218(F) creep-resistance of, 219 dissolution rate of metals and metallizations in, 87(F) dissolution rate of silver in, 53(F) drop-in replacement for, 95, 96, 191 for ip-chip process, 207 uxes for, 122 grain renement, 218 history of, 56 lead-free solders, alternatives to, 191193 lead-free solders, compatibility with, 191 molten, high rate of dissolution of gold in, 89 repairs to, 191 shear strength of joints with, 54(F) silver in, 57 vs. silver-tin solders, 60 solder alloy systems, 5660 tensile strength of, 222(F) wetting angle of, 14(F) wetting behavior of, on mild steel, 210(F) wetting speed of, 197(F) Lead-tin solders, specic types 98Pb-2Sn, Chadwick peel tests of, 73 Pb-3Sn, effect on joint strength on delay from cleaning to assembly delay, 131(F) Pb-3Sn, fast atom cleaning effect of, 131(F) Pb-3Sn, ip-chip joints, 129 Pb-3Sn, high melting point, 131(F) Pb-4Sn, 30 95Pb-5Sn, oxide growth on, 125 Pb-5Sn, for ip-chip bonding process, 166 Pb-26Sn, eutectic composition, 89 Pb-60Sn, 51 Pb-60Sn, contact angle of, 20 Pb-60Sn, effect of major ternary additions to, 76
www.asminternational.org
Index / 259
Pb-60Sn, impurity concentrations producing detrimental effects for, 76(T) Pb-60Sn, lead-tin eutectic solder, 51 Pb-60Sn, molten, rate of dissolution of copper in, 84 Pb-60Sn, spreading of, on gold-plated sample, 213(F) Pb-62Sn, elongation, 194 Pb-62Sn, uxless wetting of, 132(F) Pb-62Sn, for copper components, 131 Pb-62Sn, for gold-nickel components, 131 Pb-63Sn, copper wetting using rosin ux, 115(F) Pb-80Sn, effect of alloying additions on wetting of, 76(T) Pb-80Sn, effect of major ternary additions to, 76 Leadless ceramic chip carriers (LCCs), 40 Lever rule, 80, 80(F), 9293 Lifetime prediction of joints, 226227 Limitations of solderability test measurements, 210 Linear expansion coefficient vs. thermal conductivity, 159(F) Liquid ow, 19 Liquid inltration casting, 161 Liquid lake condition, 134 Liquid metal embrittlement, 30 Liquid nitrogen, 109 Liquid phase sintering, 157 Liquid solder lm, 134 Liquid-solid metallurgical reactions, 34 Liquid spreading, 19 Liquidus projection, 89(F) Liquidus surface of Ag-Cu-Sn system, 63(F) of Ag-Sb-Sn system, 64(F) of Cu-Pb-Sn system, 85(F) of Pb-Sb-Sn system, 62(F) of Si-Pb-Sn system, 65(F) Liquidus temperature, 49 of 3.8Ag-0.7Cu-95.5Cu solders, 228(F) calculating depression of, 98 of eutectic silver solders, 80 increases of, from silver additions, 53 Load applied to preforms vs. void level, 134(F) Local heating, 33 Local interfacial mismatch stresses, 28 Low-alpha-emission solders, 190 Low-alpha lead, 190191 Low-cycle fatigue life, 221(T) Low-cycle fatigue resistance, 220 Low-expansion materials, 161(T) Low-expansion metals, 160 Low-expansivity materials, 223 Low-melting-point eutectic alloys, 55(F), 149 Low-melting-point metals, 192(T) Low-melting-point solders Ag-96Sn properties of, 53 eutectic alloys of, 7(F) of indium-base solders, 73 stress-rupture life of, 195(F) Low-solid uxes, 118 Low spreading and bond quality, 212 Lower-melting-point solders, 5152
M
Machinability and machining costs, 160, 161 Magnesium, 122 Magnetostriction, 160 Maps of brazes and solders, 7 Materials storage of, 43 strength of, 2728 Materials systems approach, 146(T) Maximum exposure limits, 43 Mechanical cleaning, 37, 209 Mechanical constraints and solutions, 157168 Mechanical fasteners and fastening, 12 Mechanical integrity, 29 Mechanical properties. See also tests/testing of intermetallic compounds, 87(T) of joints, 26 of joints, measurements of, 223224 of lead-free solders, 194195 of selected solders, 194(T) of soldered joints, 224 tests/testing of, 224 Mechanical strengths in solders, factors in, 223224 Melting, instantaneous, 93 Melting point of Ag-20In-77Sn solders, 96 of Ag-Cu-Sn ternary system, 96 of aluminum-silicon alloys, 6 effect of, on multiple alloying additions, 9798 of eutectic alloys, 81 of eutectic solders, 52 gold addition effects on, 153154 of gold-bearing solders, 66 of In-46Sn-2Zn ternary eutectic, 96 of In-48Sn binary eutectic, 96 of In-86Sn-9Zn solders, 96 of indium, 75(F) of liquid metal, 15 of lower-melting point solders, 51 of metals and their coefficient of thermal expansion (CTE) of, 159, 160(F) and peak operating temperature, 29 of silver-tin binary system, 96 of solid metal, 15 Melting point depression behavior of, 9395 cadmium-base solders, 93 by eutectic alloying, 9396 general features of, 9395 lead free solders, implications for, 9596 liquid alloys based on gallium, 93 Melting ranges of Alloy J, 60 and braze alloy families, 7(F) of lead-free solders, 191
2004 ASM International. All Rights Reserved. Principles of Soldering (#06244G) 260 / Principles of Soldering
www.asminternational.org
Melting ranges (continued) and solder alloy families, 6(F) Mercury amalgams based on, 215216 health hazards, 215 health safety issues of, 93 melting point of, 93 and silver powder, diffusion reaction of, 215 substitutes for, 93 toxicity of, 191 Mercury-base amalgams, 215 Metal-ceramic composites, 163 Metal-loaded glass frits, 181 Metal-matrix composite (MMC), 175 Metal-metalloid composites, 163 Metal oxide reduction, 110(F) Metal oxides bond strength of, with parent metals, 106107 chemical reduction of, 105 and formic acid, reaction between, 131 Metal-to-oxygen chemical bond, 105 Metallic impurities, 7577 Metallization layer, 151 Metallization techniques advantages and disadvantages of, 181 as a barrier, 153 characteristics of, 181(T) chemical vapor deposition, 180 coating quality of, 182(T) of porous ceramic materials, 151(F) process parameters of, 182(T) relative merits of, 182(T) with silver electroplating, 151(F) thick-lm formulations, 180181 wet plating, 180 Metallizations of alumina, 151 and control of spreading, 148 costs of, 152, 231 ring on glass and ceramics, 151 gold layers in, 150 in layers, 150 and metals in diffusion bonding, 11(T) moly-manganese process of, 151 noble metals, 181 of oxide ceramics, 151 of parent materials, 49 for refractory materials, 130 solderable, gold as, 89 stresses in, 151 with titanium, 150 widely used, 37 with zirconium, 150 Metalloid-metalloid composites, 163 Metallurgical considerations for solder selection, 78 Metallurgical constraints and solutions, 147157 Metallurgical incompatibility of materials and processing conditions, 147 Metallurgical modication of contact angle, 133 Metallurgical properties of lead-free solders, 193196
Metallurgical reactions, 12 erosion rate changes to parent materials, 154 liquid-solid, 34 melting point changes, 153154 Metallurgical stability of soldering processes, 29 Metals cohesive strength of, 27 and their properties, 161(T) wetting of by solders, 147157 Metals and metallizations in diffusion bonding, 11(T) Microcracks, 28 Microelectromechanical systems (MEMS), 123, 222 Microame torch, 34, 34(F) Microfocus x-ray systems, 236(F), 236237 Micrographs of Ag3Sn intermetallic phase, 92(F) of controlled expansion (CE) alloys, 164(F) of diffusion-soldered joint, 233(F) of lead-tin solder on gold-plated copper substrate, 213(F) of peritectic transformation, 83(F) Microstructural coring, 75 Military Standards (MIL-STD) 883D, (void free joints), 183 883E, method 2019.7 (acceptance criteria), 231 Mischmetals, 227 Mismatch stress concentration, 164 Mismatch stresses, 39 elimination of, 165166 Modeling joint lifetime, 223, 224 of lifetime of joints, 224 Modeling assumptions in nite element analysis (FEA), 225 Modeling elements, 226 Modulus of elasticity, 157 Moisture content as a function of bakeout time and temperature, 185(F) as a function of bakeout times, 184 Moisture in semiconductor failure mechanisms, 123 Moisture permeation, time predicted for, 184(F) Moisture removal, 184 Molar volumes, 140 Molecular hydrogen, 127 Molten acetamid uxes, 122 Moly-manganese process, 151 Molybdenum of coefficient of thermal expansion (CTE), 160 joining atmospheres for, 104 as low-expansion metals, 160 Monatomic hydrogen, 127128 Monolithic plates, 165(F) Montreal Protocol on Substances That Deplete the Ozone Layer: 1991, 111 Multichip modules (MCMs), 199 Multilayer metallic coatings, 50 Multilayer metallization, 199 Multiphase materials, 160 Multiple frequency SAM, 236
www.asminternational.org
Index / 261
N
Narrow joints, 178 Native oxides, 126(T) Neodymium, 227 Neutral ame, 34 Nichrome, 150 Nickel health restrictions on, 50 toxicity of, 191 Nickel foil interlayers, 10 Nickel powder, 216 Nickel-tin intermetallic compounds, specic types Ni3Sn2 phase, binary compounds, 84 Ni3Sn4 phase, binary compounds, 84 Nilo-Alloy 42, 160 Nilo-K, 160 Niobium (columbium), 104 Nitrogen costs of, 109 dry, for moisture removal, 184 in inert atmospheres, 36 Nitrogen-based soldering, 197 No-clean uxes, 39, 40, 118119 Noble-metal metallizations, 181 Noble metals embrittlement from, 147 oxides of, 105 stability of, 105 Non-nickel alloys, 162(F) Nonmetallic bonding, 152 Nonmetallic components chemical vapor depositions on, 149150 physical vapor depositions on, 149150 wet plating on, 149150 Nonmetallic materials, wetting and spreading with, 103 Nonmetallic phases and wetting problems, 147 Nonmetallic surface coating, 114 Nonmetals foundation layers for, 150 solderable coatings on, 149152 wetting of by solders, 149153 Nonoxidizable metallizations, 133 Nonwettable patches, 17 Numerical modeling, 223224
O
OA type uxes, 117 Off-eutectic compositions, 54 Optoelectronic devices, 214215, 221 Organic acids, 116 Organic acids uxes, 117 Organic coatings, 149 Organic lms, 37 Organic uxes, 121 Organic metal compounds, 131 Osprey spray-forming process, 163
Outgassing, 30 Oxidation rate of, 107 water vapor as source of, 36 Oxidation reactions, 106 Oxide-dispersion-strengthened solders, 218219 Oxide lms destabilization of, and rare earth doping, 228 effects of, 111 reduction of, 105106 regrowth of, 129 removal of, 35 Oxide formation and removal, 124125 Oxide growth on 95Pb-5Sn solders, 125 on base metals, temperature dependence, 125(F) on base metals, time dependence, 124(F) equation for, 124 on molten solders, 126(F) Oxide layers, 122 Oxide reduction alternative atmospheres for, 111 dynamics of, 126 thermodynamic aspects of, 106107 thermodynamic principles for analysis of, 106 Oxide reduction rate by hydrogen, 126 Oxide scale, uxes with, 34 Oxide thickness and defect levels, 124 effects of, on defect levels, 124(F) vs. oxidation time, 125(F) temperature effects on growth of, 124 Oxides on aluminum, 9 growth of, 125, 148 with hydrogen at a partial pressure, 129(F) mechanical removal of, 128130 native, superheats to dissolve, 126(T) reduction of, by reactive gas atmosphere, 130131 surface, removal of, 37 Oxidizing atmospheres gold in, 35 graphite effect in, 30 platinum-group metals in, 35 Oxidizing ame, 34 Oxygen atmospheres and solder spreading, 127(T) Oxygen concentration effect of, on rate of dross formation, 115(F) vs. heat treatment, 214(F) Oxygen partial pressure, 3536 to effect oxidation reaction, 141 and metal oxide bond strength, 106 method for reduction of, 107 and water vapor desorption, 108 Oxygen, residual levels of, 35 Ozone depletion, 39
2004 ASM International. All Rights Reserved. Principles of Soldering (#06244G) 262 / Principles of Soldering
www.asminternational.org
P
P-charts, 42 Package leak rates, 184 Palladium metallization, 71 Palladium oxide, 107 Palladium plated devices, 207208 Parent materials erosion of, 49, 153154 metallization of, 49 wettability of, 49 Partial pressure. See also oxygen partial pressure, 36 Partitioned ller metals, 155(T), 155157 Pastes, 31 Path length, 169170 PCBs. See printed circuit boards (PCBs) Peak bonding temperature, 38 Peak operating temperature, 29 Peak process temperature, 33 Peel force, 167(F) Peel force proles, 168(F) Peel fracture and height of the solder lm, 167(F) Peeling stresses, 175 Peritectic reaction of aluminum quaternary alloy, 83 described, 83 indium-lead alloys, 75 Peritectic transformation, 83(F) Phase diagrams availability of, in literature, 79 binary alloy systems, 7983 and binary alloys, 79 binary eutectic composition solder with intermetallics, 8183 binary eutectic composition solder with no intermetallics, 7981 binary peritectic solder, 83 described, 78 distributed compound between eutectic solder and component metals, 8992 higher order systems, 9293 interfacial compound between eutectic solder and component metals, 8489 limitations of, 79 non-metallic systems, 9293 overview of, 49 soldering applications of, 7779 ternary alloy systems, 8392 of ternary systems, 84 uses of, 79 in weight percentages, 78 Phase diagrams, specic alloy systems of aluminum-zinc, 66(F) of antimony-tin, 60(F) of bismuth-tin, 56(F) of copper-tin, 85(F) of gold-antimony, 69(F) of gold-germanium, 68(F) of gold-indium, 52(F) of gold-lead-tin, 89
of gold-silicon, 67(F) of gold-silicon-tin, 69 of gold-tin, 68(F) of indium-lead, 61(F) of indium-tin, 55(F) of lead-tin, 58(F) of silver-indium, 57(F) of silver-lead, 61(F) of silver-tin, 59(F) Phase formation, 154155 Phase segregation failure, 74 Phased reow soldering, 155, 157 Phases instability of, 29 rate of growth of, 81 Phonons, 226 Phosphoric acid uxes, 122 Physical abrasion for surface cleaning, 129 Physical properties of Ag-96Sn solders, 53 of copper-tin intermetallic compounds, 87 of intermetallic compounds, 87(T) of lead-free solders, 193196 for low-expansion materials, 161(T) of selected solders, 194(T) for semiconductors, 161(T) Physical vapor depositions, 149150 Piezoelectric ceramic elements, 31(F) Planar joints, 50 Plasma, 127 Plasma-assisted dry soldering (PADS), 131 Plastic ow, 194, 219 Plastic leaded chip carriers (PLCCs), 40 Plastic yielding, 10 Platinum, 147148, 148(T) Platinum-group metals, 35 Platinum-tin intermetallic compounds, specic types PtSn4 interfacial layer, 148 Polymeric materials, 172 Polymers, 3 Porosity. See voids Porous ceramic materials, 151(F) Powder metallurgy materials (P/M), 161 Powdered solders, 31 Power cycling and joint quality degradation, 65(F) Precious metals binary combinations for diffusion soldering, 231 hallmarking regulations for, 50 Precipitation hardening, 49 Precipitation-strengthened alloys, 175 Preform geometry, 133134 Preforms Alloy J for, 61 disc shaped, 170 dual disc, 171(F) foil, deciencies of, 171 hermetically sealed enclosures-hermetically sealed enclosures, 44 round wire, cross shaped, 171 thickness of, 133
www.asminternational.org
Index / 263
Pressure dependence of Gibbs free energy, 140 Pressure in diffusion brazing, 10 Pressure variation method to minimize voids, 174 for reduction of level of voids, 173 for void level reduction, 174(F) Pressure welding, 45 of aluminum, 9 and dissimilar-metal joints, 9 indium in, 43 joint strength of, 9(F) with standard solders, 43 Pressure welding and diffusion bonding, 812 Primary dendrites of silver, 80 Principle of Conservation of Energy, 137 Printed circuit boards (PCBs) assembly line change to lead-free solders, 197 lead-free solders for, 95 reow soldering of, 36 solders for low process temperature, 155, 157 tin whiskers on, 196 Process control, 42 Process control chart of peak reow temperature, 42(F) Process cycle time, 33 Process variability (r-chart), 42 Process window for lead-free solders, 196197 Processing aspects, 3042 Product miniaturization, 151 Prolometer traces, 211 Progressive alloying, 98 Progressive eutectic alloying, 98 Progressive melting, 157 Propane, 34 Pulse-echo ultrasonic inspection, 235 Pure metals, 14 Pyroelectric elements, 34 Pyrometers, 35 Pythagorean theorem, 45
Q
Quality acceptance criteria, 209 Quality-control testing, 210 Quality of soldered joints, 12 Quality of wetting and contact angle, 17 Quasi-binary alloy systems, 92 Quasi-ternary alloy systems, 92 Quenching stages, 39
R
R-chart (range chart), 42 R, RMA, RA type uxes, 117 Radiographs of Ag-96Sn foil solders, 23(F) of joint integrity, 236237 of silicon chips, 169(F)
Range chart (r-chart), 42 Raoults Law, 97, 98 Rapid-solidication process, 66 Rapidly solidied alloys, 31 Rapidly solidied ller metals, 6061 Rapidly solidied processes, 31 Rare earth doping contact angle of indium solders, 228 effects of, on solders, 227 oxide lm destabilization, 228 of Sn-9Zn solders, 228 Rare earth elements effect of additions on solder properties, 227229 function of, in nonmetallic bonding, 152 implications of, for soldering technology, 229 solders doped with, 227229 Rate of dissolution. See dissolution rate Rate of reaction, 78 Reactive ller alloys, 152 Reactive ller metals, 152 Reactive gas atmospheres, 130131 Reactive ion etching, 172 Reactive-metal metallizations, 181 Reactive metals, 150 Reactive solders, 103 Reactive wetting, 16 Reduced-oxide soldering activation (ROSA) process, 132 Reducing atmospheres, 8, 3637, 109111 Reducing gases, 109 Reduction, 130(F) Reduction atmospheres, 34 Reduction ame, 34 Reference standards, 213214 Reow soldering, 3334, 36 Reow stage, 38 Refractory metals list of, 104 oxides of, 105 stability of, 105 wetting problems with, 147 Reinforced solders (solder composites), 222223 Reinforcing plates, 165 Relative spreading, 211 Resin materials, 215 Respiratory problems, 43 Restoring force, 205 Reverse-gas bias mode, 150 Reversible chemical reactions, 105 Reversible processes, 138, 139 Rheological concepts, 215 Richardson-Jeffes diagram. See Ellingham diagram Roll-bonding, 9 Rosin, 116 Rosin uxes, 115(F), 117 Rule of mixtures, 194
2004 ASM International. All Rights Reserved. Principles of Soldering (#06244G) 264 / Principles of Soldering
www.asminternational.org
S
S-charts (standard deviation chart), 42 SA type uxes, 117 Sacricial metal, 200 Safety. See health and safety issues Scanning acoustic microscopy, 170(F) Scanning acoustic microscopy (SAM), 189, 235236 Scanning acoustic techniques, 183 Scarf angle, 177 Scarf butt joints, 177 Screen printing, 31 Second Law of Thermodynamics, 138 Self-alignment of ip chip interconnection, 202(F) predictions of, 206 of solder bump bonding, 201 Semiconductor die attach, 69 Semiconductors, physical properties for, 161(T) Service requirements of joints, 50 Service temperature range, 3 Service temperatures of brazed and soldered joints, 4 of brazes and solders, 8 of solid-state joining, 5 Sessile drop tests, 211(F) of composite solders, 220 tests/testing, 209 Shear strength of uxless joints, 37(F) as a function of joint thickness, 178(F) of gold-indium joints, 231 of joints, 222 of joints with lead-tin eutectic solders, 54(F) of lead-free solders, 194 of soldered joints, 178(F) Shear stress, 175 Shelf life and storage requirements of uxed components, 114 Shelf life of multilayer metallic coatings, 50 Silaceous dross, 66 Silica, 69 Silicon, 152 Silicon-aluminum alloys coefficient of thermal expansion (CTE), 163 Osprey spray-forming process, 163 Silicon carbide, 152 Silicon chip radiograph, 169(F) Silicon-lead-tin system, 65(F) Silicon semiconductor chips to gold-metallized pads, 66 Silver coring of, 80 diffusion soldering of, 231232 dissolution rate of, in lead-tin eutectic solders, 53(F) effects of, on surface tension of tin, 193 erosion of, by molten tin, 82(F) as impurity, 77 oxides of, 38 primary dendrites of, 80
as solder constituent, 53 suldes of, 38 as wettable metallizations, 147 Silver additions, 53 Silver-antimony-tin solders, specic types Alloy J, 60 Silver-antimony-tin system, 64(F) Silver-base brazes, 6 Silver-base ller metals, 57 Silver-coated components, 81 Silver-copper-phosphorous brazes, 5 Silver-copper-tin alloys lead-free solders, 193 superheats of, 126 Silver-copper-tin lead-free solders, 115 Silver-copper-tin phase diagrams, 193 Silver-copper-tin solders, specic types 3.5Ag-0.9Cu-95.6Sn ternary, cooling rate dependency of, 193 3.5Ag-0.9Cu-95.6Sn ternary, melting point of, 96 3.8Ag-0.7-95.5Sn, stress-rupture life of, 228(F) 3.8Ag-0.7Cu-95.5Cu, liquidus temperatures of, 228(F) 3.8Ag-0.7Cu-95.5Cu, solidus temperatures of, 228(F) Ag-1.7Cu-93.6Sn, cobalt and iron as impurity of, 77 Silver-copper-tin system, 63(F) Silver-copper-tin ternary system, 96 Silver electroplating, 151(F) Silver foil interlayers, 10 Silver-gold-tin solders, specic types 3.6Ag-1.6Au-92.8Sn, costs of, 96 Silver-gold-tin ternary systems liquidus surface of, 91(F) vertical section through, 91(F) Silver-gold-tin ternary systems phase diagrams, 90 Silver-indium alloys, 232 Silver-indium intermetallic compounds, specic types AuIn2, 75 Silver-indium-lead solders, specic types Ag-80In-15Pb, melted in controlled atmosphere, 112(F) Silver-indium phase diagrams, 57(F) Silver-indium-tin solders, specic types Ag-20In-77Sn, melting point of, 96 Silver-lead phase diagrams, 61(F), 80 Silver-lead solders, 6, 7980 Silver-lead solders, specic types Ag-97.6Pb, silver-lead phase diagrams for, 80 Silver-lead system diagram, 80(F) Silver-lead-tin solders, specic types Ag-97.5Pb-1Sn, gold addition, effect of, 154(F) Ag-97.5Pb-1Sn, gold addition effects on, 154 Silver-mercury intermetallic compounds, specic types Ag2Hg3, 215 Silver oxide, 130(F) Silver particle reinforcement, 220 Silver powder in gallium-based amalgams, 216 indium amalgams with, 217 and mercury, diffusion reaction of, 215 Silver-tin alloys, 232
www.asminternational.org
Index / 265
Silver-tin alloys, specic type Ag-96Sn, joints made by, 24 Silver-tin binary system, 96 Silver-tin eutectic alloy, 152 Silver-tin eutectic solders 3.6Ag-1.6Au-92.8Sn eutectic solders, 96 effect of gold additions to, 154 lead-free solders, 60 reaction of, to silver, 231 remelt temperature of, 232 wetting effect or rare earth doping, 228229 Silver-tin intermetallic compounds, specic types Ag3Sn, 57 Ag3Sn, as continuous interfacial layer, 231232 Ag3Sn, as stoichiometric compounds, 91 Ag3Sn, characteristics of, 9192 Ag3Sn, joint strength of, 232 Ag3Sn, micrograph of, 92(F) Ag3Sn, rate of growth of, 92 Ag3Sn, thickness of, 92(F) Au5Sn, creation of, 233 Silver-tin intermetallic phases, 52 Silver-tin off-eutectic alloys effect of bismuth additions to, 193(F) liquidus and solidus temperatures of, 193(F) Silver-tin phase diagrams, 59(F), 81 Silver-tin solders eutectic alloys, 60 vs. lead-tin solders, 60 Silver-tin solders, specic types Ag-95Sn, elongation to failure of, 228(F) Ag-95Sn, tensile strength of, 228(F) Ag-96.5Sn, 90, 193 Ag-96.5Sn, contact angle of, 228(F) Ag-96.5Sn, for silver-coated components, 81 Ag-96.5Sn, joint strength of, 232 Ag-96.5Sn, spread area of, 228(F) Ag-96Sn, elongation, 194 Ag-96Sn, gallium arsenide (GaAs) with, 134 Ag-96Sn, in foil, radiograph of, 23(F) Ag-96Sn, mechanical properties of, 91(F) Ag-96Sn, properties of, 53, 90 Silver-tin systems, 231 Silver-tin-zinc solders, specic types 3.5Ag-95.5Sn-1Zn, grain renement in, 218 Single phase materials, 159 Solder compatibility of, 49 creep resistance of, 217 deposition methods, 3132 fatigue resistance of, 217 Solder alloy families and melting ranges, 6(F) temperature ranges of, 6 Solder alloy systems antimony in, 5354 bismuth in, 53 gold in, 54 indium and lead in, 5253 overview of, 49
silver in, 53 survey of, 5175 tin in, 52 zinc in, 54 Solder alloys constituents of, 51 surface tensions of, 15 volatile constituents of, 6 Solder bridging, 207 Solder bump bonding process characteristics of, 202(T) process ow for, 199(T) self-aligning feature of, 201 semiconductor components in, 203(F) technology characteristics of, 202(T) Solder bumps, 205 Solder coated substrates, 134 Solder composites (reinforced solders), 222223 Solder drains, 206 Solder elements, 52 Solder ow mechanically enhanced, 134 metallurgically enhanced, 134135 Solder foil vs. solder coated substrates, 134 Solder oxides reduction of, by atomic hydrogen, 127128 reduction of, by hydrogen, 126127 reduction rate of, in hydrogen, 128(F) self-dissolution of, 125126 Solder pastes, 31 Solder preforms. See preforms Solder reow ovens, 174 Solder reinforcement, 178 Solder selection, 78 Solderability evaluation of, 207 of selected metals and alloys, 121(T) Solderability calibration standards, 212214 Solderability shelf life, 132 of gold-coated components, 148(T) of gold coating, 133 Solderability test cycle, 209(F) Solderability test measurements, 210 Solderability test methods, 207214 Solderability testers, 208 Solderable component surfaces, 133 Soldered joints. See joints Soldering and brazing, distinction between, 6 chemical uxes for, 111123 design criteria, 2830 ller metal temperatures in, 5 ip-chip-interconnections, 199207 uxless, 123137 functional process of, 2830 health, safety and environmental aspects of, 4243 key parameters of, 1228 processes of, 2845 Soldering and brazing, 38 Soldering fumes, 43
2004 ASM International. All Rights Reserved. Principles of Soldering (#06244G) 266 / Principles of Soldering
www.asminternational.org
Soldering iron, 129 Soldering standards. See tests/testing Solders. See also specic solders active, 152153 bulk properties of, indicative data on, 225226 for domestic water pipes, 51 eutectic compositions of, 8 ow enhancement, 133 mechanical strengths, factors in, 223224 metallurgy of, 4999 powders, 31 service temperature of, 8 strengthening of, 217222 theoretical viscosity of, 19 wetting of metals by, 147157 wetting of nonmetals by, 149153 Solid-liquid interdiffusional bonding. See diffusion soldering Solid-liquid interfacial reactions, 24 Solid-liquid reactions, 25 Solid-solution strengthening, 5354 Solid solutions, 5 Solid-state diffusion tin-base intermetallic phases growth of by, 88(F) tin intermetallics by, 88 Solid-state diffusion processes, 25 Solid-state joining characteristics of, 5 with gold, 4344 with indium, 4344 joint cleaning for, 5 with solder constituents, 4344 temperature levels of, 5 Solid surfaces, 12 Solidication casting, 32(F) Solidication shrinkage control of voids from, 173 magnitude of, 173 of molten ller metals, 169 of selected elements used in solders, 173(T) stress concentrations from, 50 by vacancy diffusion, 173 Solids content, 118 Solidus temperatures, 29, 62, 80, 228(F) Soluble coatings, 149 Soluble halides, 118 Solution treatment stages, 39 Spherical cap, 44 Spherical cap geometry, 44(F) Spontaneous chemical reaction, 137 Spontaneous ltration, 223 Spread characteristics on binary solder alloys, 21(F) Spread factor and contact angle, 4445 dened, 45, 211 and spread ratio and contact angle, relationship between, 212(F) values for, 212(T) Spread ratio and contact angle, 4445
dened, 44, 211 for a range of alloys, 212 and spread factor and contact angle, relationship between, 212(F) values for, 212(T) Spread tests, 112113(F) Spreading area of, 20 assessment of, 210212 atmosphere effects on, 20 capillary action of, 23 classical model of, 12 driving force for, 84 of eutectic alloys, 81 irreversible nature of, 15 metallization control of, 148 of Pb60Sn solders, on gold-plated sample, 213(F) rate of, 22(F) vs. wetting, 49 Spreading behavior, 70(F) Spreading characteristics of lead-free solders, 197 Spreading test, 212 Sputter-deposited coatings, 148 Sputtering, 178, 180 Sputtering process, 150 Stainless steels bismuth-tin lead-free solders for, 122 uxes for, 122 halogen-based uxes for, 122 nonmetallic bonding to, 152 oxide layers of, 122 phosphoric acid uxes for, 122 and temperature uniformity, 50 thermal conductivity of, 50, 122 Standard deviation chart (s-charts), 42 Standards Scanning acoustic microscope, 183184 in soldering, 183184 visual inspection, 183 x-ray inspection, 183184 Statistical process control (SPC), 42 Step butt joints, 177 Step height, 176, 176(F) Step-joining process, 33 Step soldered ip-chip interconnects, 206207 Stirlings Formula, 98 Stoichiometric compounds, 91 Storage of multilayer metallic coatings, 50 Strain energy density, 226227 Strain rate, 194 Strap joints, 177, 177(F) Strength of lap joints, and rare earth doping, 229 mechanical, of brittle materials, 28 of pressure welded joints, 9(F) Strengthened solders, 178, 217222 Strengths of metals, practical, 28 Stress from differential thermal expansion, 157 reduction in, from differential thermal expansion, 158
www.asminternational.org
Index / 267
reduction of, by creep, 178 Stress concentrations dened, 175 and llets, 167 llets to reduce, 176 by high thermal gradients, 4 of metallization layer, 151 origins and magnitude of, 175 from solidication shrinkage, 50 Stress cycles, 227 Stress distribution, 176(F) Stress relaxation, 158 Stress-rupture life of 3.8Ag-0.7-95.5Sn solders, 228(F) of low-melting-point solders, 195(F) as ranking for creep resistance, 195 of Sn-3.5Ag-0.25RE, 228(F) of Sn-3.5Ag solders, 228(F) Stress-strain curves for dental amalgams, 215, 215(F) Strip casting, 32(F) Sub-zero service temperature, 76 Sublimation, 109(T) Super Solder system, 3233 Superheat dened, 8 to dissolve native oxides, 126(T) to dissolve surface oxides, 126 for indium-base solders, 75 and oxygen levels for melting point, 127(T) and solder spreading, 115 Surface-area-to-volume ratio, 133 Surface cleaning, 129 Surface condition of components, 12 requirements of, in diffusion brazing, 10 Surface conditioning, 131132 Surface diffusion, 10 Surface energy diagram of, 12(F) of a liquid, 13 of pure metals, 14 of a solid, 12 and surface tension, 13 and surface tension, diagram of, 13(F) Surface energy and surface tension, 1213 Surface erosion, 3 SURFACE EVOLVER (software), 12, 206 Surface nishes, 149 Surface insulation resistance test (SIR), 119 International Electrotechnical Commission (IEC) test coupons for, 119(F) proposed changes to, 120 test plot of, showing dendritic growth, 120(F) Surface metallization, 4 Surface mount solderability test (SMT), 207 Surface-mount technology, 217 Surface-opening cavities, 163 Surface oxides removal of, 37 superheats to dissolve, 126
Surface roughness of cold-rolled copper, 14(T) of components, 2225 effect of, on fracture toughness, 24(F) and fracture toughness, 2324 and spreading, 23 Surface temperature, 25 Surface tensions of binary solders, 194(T) changes to various metal alloys, 205 diagram of forces of, 13(F) of lead-free solders, 193194 between liquid and vapor, 13, 14 of molten ller metals, 209 of solder alloys, 15 solder bump self-alignment, 205 between solid and liquid, 13, 14 between solid and vapor, 13, 14 and surface energy, diagram of, 13(F) Surface topography, 206 Surfactants, 117 Surroundings, dened, 103 Sweeping of trapped gas, 133 Symbols and abbreviations, 243 Synthetically activated uxes, 117
T
Tantalum diffusion bonding of, 10 joining atmospheres for, 104 Tape bonding, 203 Tape interconnections, 202 Temperature and O/H ratio, 110(F) Temperature effects of contact angle, 25 of oxide thickness growth, 124 of surface temperature, 25 of viscosity, 25 Temperature gradients, 12 Temperature levels of diffusion bonding process, 9 of solid-state joining, 5 Temperature limits for brazing and soldering, 4 Temperature measurements, 3435, 38 Temperature/pressure curve for diffusion bonding of gold, 43(F) for diffusion bonding of indium, 44(F) Temperature uniformity, 50 Temperatures. See also melting point activation, of uxes, 117 active ller metals requirements, 149 of boiling, 109(T) of boiling/sublimation, for selected elements, 107 control of, 38 in diffusion brazing, 10 of eutectiferous phase transformations, 94(T) of heat treatment, 39 of heat treatment for reference standards, 213
2004 ASM International. All Rights Reserved. Principles of Soldering (#06244G) 268 / Principles of Soldering
www.asminternational.org
Temperatures (continued) peak operating, 29 peak process, 33 peak reow, process control chart of, 42(F) of reactive ller metal bonding, 152 for reduction of indium oxides, 126 for reduction of tin oxides, 126 for reduction of zinc oxides, 126 of sublimation, 109(T) Tensile strength of Ag-95Sn solders, 228(F) of composite solders, 221(T) and dendrites arm spacing, 32(F) of joints, 222 of lead-tin alloys, 81(F) of lead-tin eutectic solders, 222(F) of Sn-9Zn solders, 228(F) Tensile stresses, 175176 Ternary systems intermetallic phases in, 84 liquidus representation in phase diagrams of, 84 representation of, in phase diagrams, 84 Tests/testing acceptance criteria, 209, 231 ANSI/J-STD-002 (test method), 207 basic spreading test, 211 bolometers (thermal imaging), 35 Chadwick peel tests, 73 component testing, 224 dip-and-look (DNL) test, 207 edge llets examination, 168 EIA/IS-86 (test method), 207 tness for purpose tests, 224 IEC board test coupons, 119(F) inspection methods for voids detection, 183 inspections of joint interior, 183 International Electrotechnical Commission (IEC), 119(F) IPC/EIA J-STD-003A (test method), 207 of mechanical properties, 224 Military Standards (MIL-STD), 183, 231 of palladium plated devices, 207208 quality assurance testing, 210 quality-control testing, 210 scanning acoustic microscopy (SAM), 170(F), 189, 235236 scanning acoustic techniques, 183 sessile drop tests, 209, 211(F), 220 solderability calibration standards, 212214 solderability test measurements, 210 solderability test methods, 207214 solderability testers, 208 spread tests, 112113(F) spreading test, 212 standards, 183184 surface insulation resistance test (SIR), 119, 120 surface insulation test plots for dendritic growth, 120(F) surface mount solderability test (SMT), 207 test coupons, 119(F)
test plot showing dendritic growth, 120(F) wetting balance solderability test, 208209 wetting force during solderability test cycle, 209(F) Thallium, 191 Theoretical viscosity of solders, 19 Thermal activation energy, 8 Thermal characteristics of fuel gases, 34(T) Thermal conductance of mechanical fasteners, 12 Thermal conductivity of adhesive joints, 3 of aluminum and aluminum alloys, 62 of carbon bers, 223 of composite solders, 221 in nite element analysis, 225 of lead free solders, 198 limits to, 30 vs. linear expansion coefficient, 159(F) and rules of mixtures, 194 of soldering atmospheres, 114(T) of stainless steels, 50, 122 of titanium, 160 of zinc-bearing solders, 63 Thermal cycling without creep, 165 without fatigue fracture, 165 Thermal distortion, 33 Thermal distortion parameter, 159(F) Thermal expansion of engineering ceramics, 159 mismatch in, 145 of single phase materials, 159 Thermal expansion mismatch effects of, 157 in ip chip interconnection, 200 stress reduction technique, 165(F) Thermal expansion mismatch strain, 50 Thermal expansivity of aluminum and aluminum alloys, 62 of bismuth solders, 225 of engineering materials, 26(T) reduction of, 222223 Thermal fatigue, 29 Thermal heat capacity, 50 Thermal imaging bolometers, 35 Thermal properties of joints, 50 Thermally conductive adhesives, 3 Thermally induced distortion, 62 Thermocompression bonding, 5(F) Thermocouples, 3435 Thermodynamic and diffusion-kinetic model of metallizations, 152 Thermodynamic equilibrium, 137 Thermodynamic principles for analyzing oxide reduction, 106 Thermodynamics rst law of, 137138 second law of, 138139 Thick-lm formulations, 180181 Thick lm metallizations, 72 Thick-gap soldering, 178179
www.asminternational.org
Index / 269
Thick solder joints, 179 Thickness of coatings, 148 Thin coatings, 180 Thin foil preforms disadvantages of, 136 economics of, 136 shapes of, 136 Thin-gap soldering, 178179 Thin (narrow) joints with hydrostatic pressure, 178179 in semiconductor grade clean room, 179 Threshold formation, 9 Through-thickness vias, 162 Time requirements of diffusion brazing, 10 Tin activation energy of, 124 allotropic transformation of, 76, 77(F), 195196 allotropic transformation suppression, 196 as alloy element for Au-2wt%Si solders, 68 as impurity, 77 and indium intermetallic compounds, 51 as solder constituent, 52 Tin-base eutectic solders, 153(T) Tin-base intermetallic phases, 88(F) Tin-base solders 3.5Ag-0.9Cu-95.6Sn ternary eutectic solders, 96 Alloy J, 6061 dissolution rate of platinum in, 148(T) failure modes of, 74 uxes for, 116120 and gold coatings, 133 as key to lead-free solders replacement, 192 silver-tin, 60 Tin-bismuth solders, specic types Sn-65Bi, 15 Tin dendrites, 77 Tin-indium solders, specic types Sn-40In, 15 Tin intermetallics, 88 Tin-lead solder, 3132 Tin-lead solders, specic types 61.75Sn-38.05Pb-0.2Cu(wt%), 84 Sn-40Pb, hermetically sealed enclosures, 44 Tin oxides reduction in hydrogen atmospheres, 111 temperature for reduction of, 126 Tin pest described, 7677, 195196 lead-free solders, 195196 sub-zero service temperature, 7677 susceptibility of puree tin-base solders to, 195196 Tin-silver-rare earth solders Sn-3.5Ag-0.25RE, stress-rupture life of, 228(F) Tin-silver solders, specic types Sn-3.5Ag, stress-rupture life of, 228(F) Tin whiskers lead-free solders, 195196 on PCBs, 196
Tin-zinc alloys, specic type Sn-9Zn eutectic, corrosion of, 96 Tin-zinc-silver solders, specic types Sn-8.5Zn-1Ag, aluminum as impurity in, 77 Tin-zinc solders Sn-9Zn eutectic alloy, 96 Tin-zinc solders, specic types Sn-9Zn, elongation to failure of, 228(F) Sn-9Zn, rare earth doping of, 228 Sn-9Zn, tensile strength of, 228(F) Sn-9Zn, wetting force of, 228(F) Tinned surfaces, 151 Titanium as additive to Au-20Sn solders, 69 as additive to Au-2Si solders, 69 of coefficient of thermal expansion (CTE), 160 diffusion bonding of, 10 joining atmospheres for, 104 as low-expansion metals, 160 nonmetallic bonding to, 152 thermal conductivity of, 160 for zinc oxide ceramic metallization, 150 Titanium alloys, 10 TO-220 semiconductor die, 66(F) Tombstoning, 157 Tongue and groove joints, 177 Toxicity, 191 Transient liquid phase (TLP) joining. See diffusion soldering Transition reactions, 83 Transuranium elements, 227 Trapped gas in adhesively bonded joints, 173 described, 169173 pressure variation method for, 174(F) reduction, by vapor-phase techniques, 171 sweeping of, 133, 170(F) volume reduction of, 171 Triode sputtering, 180 Tungsten of coefficient of thermal expansion (CTE), 160 diffusion bonding of, 10 as low-expansion metals, 160
U
Ultimate tensile strength of lead-free solders, 194 Ultrahigh vacuum system (UHV), 127 Ultrasonic uxing, 129, 130 Ultrasonic power and wetted area, 131(F) Ultrasonic soldering, 128130 Ultrasonic systems, 129130 Ultrasound, dened, 235 Underbump metal, 200 Underll adhesive, 203 Universal gas constant, 97, 124 UNS K94610, 160 Unsolderable materials, 120122
2004 ASM International. All Rights Reserved. Principles of Soldering (#06244G) 270 / Principles of Soldering
www.asminternational.org
V
Vacancy diffusion, 173 Vacuum atmospheres brasses and zinc metallization, 107 industrial quality, 107 soldering in, 107109 Vacuum bakeout, 172 Vacuum deposition process, 206 Vacuum evaporation, 180 Vacuum furnace, 104 Vacuum joining, 174 Vacuum system, 36 Van der Waals forces, 15 Vanadium, 104 Vanadium ions, 132 Vapor deposition, 31 Vapor deposition coating, 234 Vapor deposition technique, 148 Vapor-phase techniques, 171 Velocity of liquid ow, 19 Ventilation, 43 Viscosity and uid ow of solders, 18 and molecular weight of metals, 19 of selected conductive adhesives, 3(T) temperature effects of, 25 theoretical, of solders, 19 Void content vs. joint length, 169(F) Void-free joints, 183 Void levels due to trapped gas, 174(F) vs. load applied to preforms, 134(F) and pressure variation, 174(F) Void size, maximum acceptable, 183184 Voids and carbon-ber-loaded solders, 223 causes of, 169 control of, 173 and diffusion bonding, 910 effect of, on joint mechanical integrity, 29 with uxed paste solder, 172173 formation of, through gas entrapment, 25 as function of component size, 169 and gap width, 15 incorporation of, 168 inspection methods for detection of, 183 in joint gap, 169(F), 170(F) joint interfaces as source of, 28 in joints, 49 Kirkendall voids, 71, 234 in phase segregation, 74 relationships of, to ux conditions, 172173 Volatile organic compounds (VOCs), 118 environmental considerations from, 196 environmental considerations of, 115 Volatilization, 107 Volume contraction cracks from, 67 of silver-tin alloys, 232
W
Water-soluble uxes, 39, 40, 118 Water vapor bakeout temperature of, 36 in hydrogen reducing atmosphere, 37 ingress into hermetically sealed package, 185(F) as source of oxidation, 36 Water vapor desorption and oxygen partial pressure, 108 Wave soldering, 33, 114 Wave soldering machines, 77 Weight fraction of constituents, 96 Welding, 4 Welding process, 4 Wet back process, 206 Wet plating for application of solderable metallizations, 148 metallization techniques, 180 on nonmetallic components, 149150 Wettability of parent materials, 49 Wetted area and ultrasonic power, 131(F) Wetting along surface valleys, 2223 assessment of, 207210 classical model of, 12, 14 difficulties of silica in, 69 driving force for, 16 effects of, by rare earth doping, 228229 and llet formation, 6 of metals by solders, 147157 of nonmetals by solders, 149153 problems with, 17, 147 rate of, 18 refractory metals problems with, 147 with refractory parent material, 147 Wetting and contact angle, 1318 Wetting and spreading, 20, 135 Wetting and spreading characteristics, 220 Wetting angle, 14(F), 212 Wetting area, 1314 Wetting balance, 208(F) Wetting balance solderability test, 208209 Wetting balance tests, 197 Wetting behavior, 210(F) Wetting characteristics, 197 Wetting equation, 13, 18 for binary metal systems, 15 Wetting force for commercial uxes, 214(F) of Sn-9Zn solders, 228(F) during solderability test cycle, 209(F) time of, to reach acceptance value, 197 Wetting front, 23 Wetting properties, 221(T)
www.asminternational.org
Index / 271
Wetting rate, 198(F) Wide-gap joints brazing of, 179 in ceramic components, 158 composite solders and, 221 Wire bonding, 203 Wire cross preform, 171(F), 172(F) Work-hardened alloys, 49 Work hardening, 49 WS type uxes, 118
X
X-bar chart, 42 X-radiography, 172(F) X-ray inspection and defects, 204 X-ray systems ne-focus, 189 microfocus, 236237 X-ray techniques, 183 X-rays, cracks and orientation of, 236
Y
Yield strength of composite solders, 220(F) of lead-free solders, 194 Youngs equation, 13, 15 Youngs modulus, 223, 225
Z
Z-axis control, 205 Zinc as additive to indium, 135 as impurity of Au-2Si solders, 77 as solder constituent, 54 Zinc alloys Al-94Zn solders, 54 with aluminum, 54
galvanic corrosion of, 64 Zinc-aluminum alloys, specic type 90% Zn, 7% Al ller metal, 136 Zinc-aluminum-magnesium-gallium alloys, specic type Zn-4Al-3Mg-3.2Ga quaternary, high-melting-point solders, 197198 Zn-4Al-3Mg-3.2Ga quaternary, high thermal conductivity of, 197198 Zinc-base alloys, 64 Zinc-base solders high-volume contraction of, 64 as lead-free solders, 66 limitations of, 50 stress concentrations of, 64 wetting additives for, 147 Zinc-bearing alloys, 67(F) Zinc-bearing solders, 6164 advantages of, 63 Al-94Zn eutectic solders, 62 for aluminum and aluminum alloys, 61 cadmium alloy of, 63 common alloys of, 6263 uxes for, 65 limitations of, 64 and reduced pressure atmospheres, 62, 65 solidus temperature range of, 62 thermal conductivity of, 63 Zinc chloride, 117 Zinc-containing alloys, 195 Zinc metallization, 107 Zinc oxide ceramic metallizations of, 150 metallized with titanium, 150 metallized with zirconium, 150 Zinc oxides stability of, 125 temperature for reduction of, 126 Zinc solders, 116 Zirconium joining atmospheres for, 104 for zinc oxide ceramic metallization, 150
engineers and scientists, a worldwide network dedicated to advancing industry, technology, and applications of metals and materials. ASM International, Materials Park, Ohio, USA www.asminternational.org This publication is copyright ASM International. All rights reserved. Publication title
Principles of Soldering To order products from ASM International: Online Visit www.asminternational.org/bookstore Telephone 1-800-336-5152 (US) or 1-440-338-5151 (Outside US) Fax 1-440-338-4634 Mail
Product code
#06244G
Customer Service, ASM International 9639 Kinsman Rd, Materials Park, Ohio 44073-0002, USA
Email CustomerService@asminternational.org
American Technical Publishers Ltd. 27-29 Knowl Piece, Wilbury Way, Hitchin Hertfordshire SG4 0SX, In Europe United Kingdom
Telephone: 01462 437933 (account holders), 01462 431525 (credit card)
www.ameritech.co.uk Neutrino Inc. In Japan Takahashi Bldg., 44-3 Fuda 1-chome, Chofu-Shi, Tokyo 182 Japan Telephone: 81 (0) 424 84 5550
Terms of Use. This publication is being made available in PDF format as a benefit to members and
customers of ASM International. You may download and print a copy of this publication for your personal use only. Other use and distribution is prohibited without the express written permission of ASM International. No warranties, express or implied, including, without limitation, warranties of merchantability or fitness for a particular purpose, are given in connection with this publication. Although this information is believed to be accurate by ASM, ASM cannot guarantee that favorable results will be obtained from the use of this publication alone. This publication is intended for use by persons having technical skill, at their sole discretion and risk. Since the conditions of product or material use are outside of ASM's control, ASM assumes no liability or obligation in connection with any use of this information. As with any material, evaluation of the material under end-use conditions prior to specification is essential. Therefore, specific testing under actual conditions is recommended. Nothing contained in this publication shall be construed as a grant of any right of manufacture, sale, use, or reproduction, in connection with any method, process, apparatus, product, composition, or system, whether or not covered by letters patent, copyright, or trademark, and nothing contained in this publication shall be construed as a defense against any alleged infringement of letters patent, copyright, or trademark, or as a defense against liability for such infringement.