SOLDER JOINT

RELIABILITY
SOLDER JOI NT
RELIABILITY
Theory and Applications

Edited by John H. Lau

ItimI VAN NOSTRAND REINHOLD

~ _ _ _ New York
Copyright © 1991 by Van Nostrand Reinhold

Library of Congress Catalog Card Number 90-12968

ISBN 0-442-00260-2

All rights reserved. No part of this work covered by

the copyright hereon may be reproduced or used in any
form by any means-graphic, electronic, or
mechanical, including photocopying, recording, taping,
or information storage and retrieval systems-without
written permission of the publisher.

I(T'iP Van Nostrand Reinhhold is a division of International Thomson

_c/ Publishing. ITP logo is a trademark under license.

Van Nostrand Reinhold

I 15 Fifth Avenue
New York, NY 10003
International Thomson Publishing
Berkshire House
168-173 High Holborn
London WCI V 7AA, England
Thomas Nelson Australia
102 Dodds Street
South Melbourne 3205
Victoria, Australia
Nelson Canada
1120 Birchmount Road
Scarborough, Ontario
M I K 5G4, Canada

16 15 14 13 12 II 10 9 8 7 6 5 4 3 2

Library of Congress Catalogjng-in-Publication Data

Solder joint reliability-theory and applications / [edited by] John

H. Lau.
p. cm.
Includes index.
ISBN 0-442-00260-2
I. Welded joints. 2. Solder and soldering. I. Lau, John H.
TA492.W4S63 1991
671.5'2042--{jc20 90-12968
CIP
 Contents

Foreword xv
Preface xvii'
Acknowledgments xix

1. Flux Reactions and Solderability 1

1.1 Flux History 1
1.2 Solderability Tests 2
1.2.1 Visual Assessment 3
1.2.2 Area of Spread Test 3
1.2.3 Edge Dip and Capillary Rise Tests 3
1.2.4 Globule Test 3
1.2.5 Rotary Dip Test 4
l.2.6 Surface Tension Balance Test 4
1.3 Flux Action from Solderability Measurements 4
1.4 Flux Types 5
1.4.1 Mechanistic Studies for Inorganic Fluxes 5
1.4.2 Mechanistic Studies for Rosin-based Fluxes 11
References 36

2. Solder Paste Technology and Applications 38

2.1 Chemical and Physical Characteristics 40

2.2 Fluxing and Fluxes 42

v
vi CONTENTS

2.3 Solder Alloys 44

2.4 Solder Powder 46
2.5 Paste Fonnulation 51
2.6 Paste Rheology 54
2.7 Rheology Behavior Characterization 56
2.8 Viscosity and Measurement 63
2.9 Printing Technique 64
2.10 Dispensing Technique 68
2.11 Soldering Principle 69
2.12 Solderability 70
2.13 Soldering Methods 72
2.14 Controlled Atmosphere Soldering 73
2.15 Solvent Cleaning 80
2.16 Aqueous Cleaning and Aqueous Cleaning Paste 84
2.17 No-clean Paste 85
2.18 Fine Pitch Paste 87
2.19 Quality 90
2.20 Conclusion 90
References 90

3. Technical Considerations in Vapor Phase and Infrared

Solder Reftow Processes 92
3.1 Introduction to Surface Mount Reftow Soldering 92
3.2 Type I 93
3.3 Soldering Requirements for Surface Mount
Technology 94
3.4 Reftow Process Phases 95
3.5 Reflow Equipment 98
3.5.1 Infrared 98
3.5.2 Vapor Phase 101
3.5.3 Convection 105
3.5.4 Conductive Belt 105
3.5.5 Laser Soldering 105
3.6 Prereflow Solder Paste Bake 106
3.7 Maximizing Solder Joint Yield 106
3.8 Reflow Processing 109
3.8.1 Vapor Phase 110
3.8.2 Infrared 112
3.8.3 Cost Comparison 113
3.9 SMT Reliability 113
References 116
 CONTENTS vii

4. Optimizing the Wave Soldering Process 117

4.1 Basic Wave Soldering Process Overview 117
4.2 Wave Soldering Process Hardware 119
4.2.1 Fluxing 119
4.2.2 Fluxers 119
4.2.3 Fluxer Measurement Parameters 124
4.2.4 Fluxer Optimization 124
4.2.5 Preheating 125
4.2.6 Preheaters 126
4.2.7 Preheat Measurement Parameters 128
4.2.8 Preheat Optimization 130
4.2.9 Wave Soldering 131
4.2. to Solder Waves 131
4.2.11 Solder Wave Measurement Parameters 133
4.2.12 Wave Soldering Optimization 134
4.2.13 Solidification 135
4.2.14 Conveyors 135
4.3 Wave Soldering Process Parameter Optimization 136
4.3.1 Optimization Procedure Test Study 137
4.4 Results 141
4.5 Conclusion 141
References 142

5. Post-Solder Cleaning Considerations 143

5.1 Purpose and Chapter Description 143
5.2 Environmental Concerns 143
5.3 Definition of Soldering Flux 144
5.4 Specifications 144
5.4.1 Test Methods 144
5.4.2 Institute for Interconnecting and Packaging Electronic
Circuits (IPC) 153
5.4.3 U.S. Military 153
5.4.4 Telecommunications 154
5.5 Flux Materials and Associated Cleaning 155
5.5.1 Rosin 157
5.5.2 Water Soluble 158
5.5.3 Synthetic Activated 159
5.5.4 Low Solids (No-Clean) 160
5.5.5 Controlled Atmosphere Soldering 162
5.6 Flux Application Methods 163
5.6.1 Wave 163
viii CONTENTS

5.6.2 Foam 164

5.6.3 Spray 164
5.6.4 Rotating Drum Spray 166
5.6.5 Application Issues for Low Solids Fluxes 166
5.7 Process Issues Associated with Reliability 166
5.7.1 Flux Residue 167
5.7.2 Solder Ball Formation 167
5.7.3 Top-Side Fillet Formation 167
5.7.4 Conformal Coating Compatibility 168
5.8 Non-Liquid Fluxes 168
5.8.1 Core Solder Material 168
5.8.2 Solder Paste Material 169
5.9 Trends 170
References 170
Additional Readings 172

6. Scanning Electron Microscopy/Energy Dispersive X-Ray

(SEMlEDX) Characterization of Solder-Solderability and
Reliability 173
6.1 Scanning Electron Microscopy/Energy Dispersive X-ray
Analysis 173
6.2 Other Methods-WDX 173
6.3 Detection Modes 174
6.4 Sample Preparation 174
6.5 Different Phases in Alloys 175
6.6 Intermetallics 175
6.7 Scope of the Chapter 176
6.8 SEMlEDX Characterization-General 176
6.8.1 Tin-Lead Solders 176
6.8.2 Two Percent Silver Solder 178
6.8.3 Gold- and Silver-Based Solders 178
6.8.4 Indium Solders 180
6.8.5 Bismuth Solders 183
6.8.6 Miscellaneous 186
6.9 Solderability Issues 186
6.9.1 Maintaining Solderability 186
6.9.2 Inadequate Tin Protective Coatings 189
6.9.3 The Dangers of "Forcing" Poor Solderability 189
6.10 Reliability Issues-Leaching of Substrate 194
6.11 Reliability Issues-Gold Embrittlement 205
6.12 Reliability Issues-Fatigue 213
References 222
 CONTENTS ix

7. The Role of Microstructure in Thermal Fati2ue of Pb-Sn

Solder Joints 225

7.1 Experimental Details 226

7.2 Eutectic Microstructures 229
7.2.1 Lamellar Eutectics 230
7.2.2 Degenerate Eutectics 231
7.2.3 Solder Joint Microstructures 233
7.2.4 Effects of Composition 238
7.2.5 Recrystallized Pb-Sn Microstructure 241
7.2.6 Coarsening Behavior 242
7.3 Mechanical Properties 243
7.3.1 Eutectic Structures 243
7.3.2 Deformation Mechanisms 246
7.4 Microstructural Evolution under Thermal Fatigue 248
7.4.1 Thermal Fatigue in Shear 249
7.4.2 Microstructural Mechanisms of Thermal
Fatigue 252
7.4.3 Other Microstructures 255
7.5 Conclusion 261
7.6 Acknowledgments 262
References 262

8. Microstructure and Mechanical Properties of Solder

Alloys 266
8.1 Thermal Cycling Fatigue 267
8.2 Precipitation and Dissolution in Pb-Sn Alloys 268
8.3 Discussion 276
References 277

9. The Interaction of Creep and Fatigue in Lead-Tin

Solders 279

9.1 Current Approaches to Accelerated Testing 279

9.2 Damage by Fatigue and Creep Mechanisms 280
9.3 Assessing Actual Joint Damage 283
9.3.1 In-service Testing 285
9.4 Understanding the Damage Mechanisms 286
9.4.1 Creep and Tensile Test Results 287
9.4.2 Cyclic Creep 287
9.4.3 Hold Time Effects 293
x CONTENTS

9.5 Interpretation for Packaging Applications 301

9.5.1 Deformation 301
9.5.2 Thermomechanical Test Guidance 302
9.6 Concluding Remarks 303
References 304

10. Creep and Stress Relaxation in Solder Joints 306

10.1 Ideal Expansivity of a Substrate 307

10.1.1 No Temperature Gradients, No Transients 307
10.1.2 Power Dissipation in the Component 309
10.1.3 Z-Gradients in the Substrate 309
10.1.4 In-Plane Gradients 311
10.1.5 Temperature Shock 311
10.1.6 Solder-Substrate Expansivity Mismatch 312
10.1.7 Overall Judgment 313
10.2 Creep and Stress Relaxation 313
10.3 Solder Properties 318
10.4 Constitutive Relations 319
10.5 Temperature Cycling 322
10.5.1 Small Temperature Range Cycling 323
10.6 Larger Temperature Cycles 326
10.7 Acknowledgments 330
References 330

11. Effects of Strain Range, Ramp Time, Hold Time, and

Temperature on Isothermal Fatigue Life of Tin-Lead Solder
Alloys 333
11.1 Definition of Failure, Specimen Design, and Mode of
Loading 333
11.2 Effect of Strain Range on Fatigue Life 334
11.3 Effect of Frequency on the Fatigue Life 338
11.4 Effect of Hold Time on Fatigue Life 343
11.5 Effect of Temperature on Isothermal Fatigue of
Solders 355
11.6 Conclusion 357
References 357

12. A Damage Integral Methodology for Thermal and

Mechanical Fatigue of Solder Joints 361

12.1 Inelastic Deformation and Stress Calculation 363

12.1.1 Governing Equation for Solder Stress 363
 CONTENTS xi

12.1.2 Inelastic Defonnation Behavior and Constitutive

Relations 365
12.1.3 Stress Calculation 370
12.2 Damage Rate Fonnulation 372
12.2.1 Damage Mechanisms 372
12.2.2 A Phenomenological Fonnulation for Crack Growth
Rates 374
12.3 Damage Integration and Failure Criterion Effects 377
12.3.1 Thennal Fatigue Life Estimation 377
12.3.2 Failure Criterion Effects 379
12.4 Discussion and Conclusions 380
12.5 Acknowledgments 380
References 380

13. Modern Approaches to Fatigue Life Prediction of SMT

Solder Joints 384

13.1 Mechanical Testing 385

13 .1.1 Detennination of Elastic Properties 385
13.1.2 Mechanical Properties 386
13.2 Life Prediction Techniques 388
13.2.1 Fatigue Models 388
13.3 Hybrid Life Prediction Techniques 395
13.3.1 Strain Range Partitioning Rule 397
13.4 Model Joints 398
13.4.1 Quality Control 398
13.4.2 Lap Joint Specimens 401
13.4.3 Straddle Board Specimens 401
13.5 Expert Systems 404
13.6 Conclusions 404
13.7 Acknowledgments 405
References 405

14. Predicting Thermal and Mechanical Fatigue Lives from

Isothermal Low Cycle Data 406
14.1 Low Cycle Fatigue (LCF) 411
14.2 Low Cycle Fatigue of Solders-Influence of the Definition
for Failure 413
14.3 Influence of the Temperature 423
14.4 Influence of Hold Times and Cycling Frequency 425
14.5 Influence of the Environment 436
14.6 Microstructural Changes 437
xii CONTENTS

14.7 Determination of the Displacement and Strain Distribution

in a Solder Joint 438
14.8 Prediction of the Fatigue Life of Solder Joints 443
14.9 Inherent Limitations to Fatigue Life Predictions 446
14.10 Necessary Further Work 447
14.11 Acknowledgments 449
References 449

15. Static and Dynamic Analyses of Surface Mount Component

Leads and Solder Joints 455
Stiffness of Gull-Wing and J Leads and Solder Joints for Surface
Mounted Chip Carriers 456
15.1 Boundary-Value Problem 458
15.2 Finite Element Methods 460
15.3 Stiffness of Gull-Wing Lead and Solder Joint 464
15.4 Stiffness of J Lead and Solder Joint 472
15.4.1 Unit Displacement (0.0001 in.) in the
I-Direction 472
15.4.2 Unit Displacement and Rotation in Other
Directions 477
15.4.3 Comparison of the Stiffness Matrices between the
PQFPs and PLCCs 477
Solder Joint Reliability Under Shock and Vibration
Conditions 478
15.5 Free Vibration of Soldered and Unsoldered Leads 478
15.5.1 Vibration Results for Wide SOICs 480
15.5.2 Vibration Results for Narrow SOICs 481
15.5.3 Vibration Results for PLCCs 483
15.5.4 Vibration Results for PQFPs 483
15.5.5 Experimental Verification 484
15.6 Free Vibration of a Constrained PCB with a SMC 486
15.7 Acknowledgments 493
References 493

16. Integrated Matrix Creep: Application to Accelerated Testing

and Lifetime Prediction 508
16.1 General Form of the Constitutive Relation 510
16.2 Development of the Constitutive Relation 511
16.2.1 Description of Data 511
16.2.2 Steady-State Creep Strain Component 511
 CONTENTS xiii

16.2.3 Elastic Strain Component 514

16.2.4 Time Independent Plastic Strain
Component 515
16.3 Summary of Constitutive Equation 517
16.4 Comparison of the Steady-State Creep Equation to
Published Data 517
16.5 Application of Constitutive Equation to Data of
Reference 2 520
16.5.1 Description of Numerical Procedures 520
16.5.2 Results 522
16.6 Multiaxial Stress States 525
16.6.1 Derivation of Constitutive Equation in Three
Dimensions 525
16.7 Fatigue Calculations and Mechanical Shear Tests 526
16.7.1 Correlation of the Data of Reference 2 528
16.7.2 Correlation of the Data of Wild and
Solomon 529
16.8 Analysis of Leaded Solder Joints 532
16.8.1 Extension of Matrix Creep Failure Indicator to
General Case 534
16.8.2 Description of Model 536
16.8.3 Results 540
16.9 Conclusions 542
16.10 Acknowledgments 543
References 543

17. Solder Joint Reliability, Accelerated Testing, and Result

Evaluation 545
17.1 The Reliability of Electronic Assemblies and Solder Joint
Reliability 545
17. 1. 1 "Bathtub" Reliability Curve-Hazard Rate
Model 546
17.2 Solder Joint Loading Conditions and Reliability 549
17.3 Reliability and Accelerated Tested-Overview 551
17.4 The Thermal Expansion Mismatch Problem 553
17.4.1 Solution 1: CTE-Tailoring to Reduce Expansion
Mismatch 554
17.4.2 Solution 2: Attachment Compliancy to
Accommodate Expansion Mismatch 555
17.5 Analytical Model of Solder Shear Fatigue 556
17.5.1 Solder Joint Fatigue 556
17.5.2 Leadless Solder Attachments 559
 xiv CONTENTS

17.5.3 Leaded Solder Attachments 564

17.5.4 Acceleration Transform 567
17.5.5 Failure Statistical Considerations 569
17.6 Accelerated Fatigue Reliability Testing 571
17 .6.1 Testing Considerations 571
17.6.2 Accelerated Test Conditions 578
17.6.3 Test Vehicle Design 581
17.6.4 Sample Statistics and Solder Joint Defects 581
17.6.5 Test Vehicle Assembly and Conditioning 582
17.6.6 Failure Definition and Detection 583
17.7 Prediction of SM Solder Joint Reliability 583
17.7.1 Simple Cyclic Load Histories 583
17.7.2 Multiple Cyclic Load Histories 584
17.8 Acknowledgments 585
References 585

18. Surface Mount Attachment Reliability and Figures of Merit

for Design for Reliability 588
18.1 Mechanics and Fatigue of SM Solder Joints 589
18.1.1 SM Leadless Attachments 589
18.1.2 SM Leaded Attachments 591
18.1. 3 Failure Distribution 594
18.2 Figures of Merit for Attachment Reliability 596
18.2.1 Derivation of FM Formulas 598
18.2.2 FMs for Multiple Thermal Fluctuations 599
18.2.3 Graphical Interpretation and Reliability
Charts 600
18.3 Examples 601
18.3.1 Example 18-1: Chip Components on FR-4 Printed
Wiring Boards 602
18.3.2 Example 18-2: 50-Mil Pitch Ceramic Leaded
Devices on FR-4 603
18.3.3 Example 18-3: 25- and 50-Mil Pitch Plastic Leaded
Components on FR-4 604
18.4 Concluding Remarks 607
18.5 Acknowledgments 607
Appendix 18-A FM Formulas and Scaling Constants 607
Appendix 18-B FMs for Multiple Thermal Fluctuations 610
References 611

Authors' Biographies 615

Index 627
 Foreword

Solders have given the designer of modern consumer, commercial, and military
electronic systems a remarkable flexibility to interconnect electronic components.
The properties of solder have facilitated broad assembly choices that have fueled
creative applications to advance technology. Solder is the electrical and me-
chanical "glue" of electronic assemblies.
This pervasive dependency on solder has stimulated new interest in applica-
tions as well as a more concerted effort to better understand materials properties.
We need not look far to see solder being used to interconnect ever finer geo-
metries. Assembly of micropassive discrete devices that are hardly visible to the
unaided eye, of silicon chips directly to ceramic and plastic substrates, and of
very fine peripheral leaded packages constitute a few of solder's uses.
There has been a marked increase in university research related to solder. New
electronic packaging centers stimulate applications, and materials engineering
and science departments have demonstrated a new vigor to improve both the
materials and our understanding of them. Industrial research and development
continues to stimulate new application, and refreshing new packaging ideas are
emerging. New handbooks have been published to help both the neophyte and
seasoned packaging engineer.
A critical element in the continued widespread use of solder is the reliability
of the solder joint. Will solder provide the characteristics necessary to allow the
world to depend on it in the future? This book, Solder Joint Reliability-Theory

xv
xvi FOREWORD

and Applications, edited by Dr. Lau and written by experts in the field, provides
a focal point of current understanding. This will help all participants in the
soldering world better plan for the future. It is an exciting time for solder, and
reliability will be a key element in its development.

Donald W. Rice
Hewlett-Packard Company
 Preface

Soldering has a history of several thousand years. It was considered an art until
the electronic age, when it was recognized as a technology. Until very recently,
however, research efforts into soldering technology have not been particularly
aggressive because solder joints in plated-through hole technology do not cause
serious reliability problems.
The 1980s have witnessed an explosive growth in the research efforts devoted
to soldering science as a direct result of the rapid development of surface mount
technology and the growing interest of miltichip module technology. (For a list
of papers published after 1985 see references I through 202 in chapter 15; for
a list of solder-joint papers published before 1985 see reference 185 in chapter
15.) Soldering is the joining method of choice for attaching component to printed
circuit board or chip to substrate. In both cases the solder-joint functions si-
multaneously as the electrical and mechanical attachment medium. Thus, solder-
joint reliability is one of the most critical issues in the development of these
technologies.
We have now begun to obtain useful insights into the mechanical properties
and microstructure of bulk solders and joints under fatigue, creep, and stress-
relaxation conditions. Some effective methods of accelerated testing, failure
analysis, and life prediction of solder joints have also been reported. We are
also achieving a new understanding of how solder-joint reliability is influenced
by flux reactions, solder paste, solderability, reflow methods, wave soldering,
and cleaning. These results have been disclosed in diverse journals or more

xvii
xviii PREFACE

incidentally in the proceedings of many conferences, symposia, and workshops

whose primary emphasis is material science or electronic packaging and inter-
connection. Consequently, there is no single source of information devoted to
the state of the art of soldering science and technology. This book aims to remedy
this deficiency and to present, in one volume, a timely summary of progress in
all aspects of this fascinating field.
This book begins with the concerns of solder joint formation and ends with
concerns over long term solder joint reliability. The book is divided into four
basic parts. Chapters 1 and 2 describe the application of flux, solder paste, and
solderability to solder joints. Chapters 3 through 5 review the various soldering
and cleaning methods and their effects on solder joint reliability. Chapters 6
through 12 explain the failure mechanisms of bulk solders and joints under
fatigue, creep, and stress-relaxation conditions. Chapters 13 through 18 provide
the data and methods for the test, design, analysis, and life prediction of solder
joints sUbjected to mechanicals thermal, shock, and vibration conditions.
Some duplication of material among chapters has been necessary if each
chapter is to offer the reader all the information essential for understanding the
subject matter. An attempt has been made to provide a degree of uniformity in
perspectives, but diverse views on certain aspects of solder joint reliability are
a reality. I hope that their inclusion here may be seen as an accurate reflection
of the state of the art and a useful feature of the book.
For whom is this book intended? Undoubtedly, it will be of interest to three
groups of specialists: (1) those who are active or intend to become active in
research on soldering science and technology; (2) those who have encountered
a practical soldering problem and wish to understand and learn more methods
of solving such problems; and (3) students and professors at universities, in view
of the fact that today's engineer receives on average less than one hour of
instruction on soldering science during his or her university studies. I hope that
this book will serve as a valuable reference to all those faced with the challenging
problems created by the ever expanding use of solders in engineering practice,
and that it will aid in stimulating further research on solder materials, testing
and analytical methods, and the sounder use of solders. Reliability of the solder
joint is limited only by the ingenuity and imagination of researchers, engineers,
and management.

John H. Lau
Hewlett-Packard Company
 Acknowledgments

Development and preparation of the manuscript was facilitated by the efforts of

a number of dedicated people at Van Nostrand Reinhold. I would like to thank
them all, with special mention to Stefania Taftinska for her coordination of the
publication process and to Stephen Chapman for his unswerving support and
advocacy. My special thanks to Marjorie Spencer who made my dream of this
book come true by effectively sponsoring the project and solving many problems
that arose during the book's preparation. It has been a great pleasure and fruitful
experience to work with them.
Most of the materials in this book have been presented at various ASMIASMEI
IEEElISHM conferences, symposia, and workshops in the past few years. I want
to thank these respected societies for allowing me to organize technical sessions
so I can have the privilege to invite the contributing authors to present their
research results.
Each chapter of the book was reviewed by at least three individuals who are
experts in soldering areas. According to their specialties, each individual re-
viewed at least three chapters of the book. These reviewers are: Dr. Birendra
N. Agarwala of IBM, Professor Donald Barker of the University of Maryland,
Dr. Kirk Bonner of Allied-Signal Inc., Professor Hans Conrad of North Carolina
State University, Dr. T. Dixon Dudderar of AT&T Bell Laboratories, Dr. Darrel
R. Frear of Sandia National Laboratories, Dr. Sung K. Kang oflBM, Dr. Joseph
Kevra of Alpha Metals, Dr. Ken Kinsman of Intel Inc., Dr. Larry Moresco
of Fujitsu Computer Packaging Technologies, Inc., Dr. Yi-Hsin Pao of Ford

xix
xx ACKNOWLEDGMENTS

Motor Company, Dr. Donald Rice of Hewlett-Packard Company Dr. Charles G.

Schmidt of the Stanford Research Institute, Professor Karl K. Stevens of Florida
Atlantic University, Dr. Boon Wong of Hughes Aircraft Company and Dr.Chee
C.Wong of AT&T Bell Laboratories. I want to thank them for their many help-
ful comments and constructive suggestions that added significantly to this book.
I express my deep appreciation to all 30 contributing authors, experts in their
respective fields, for their many helpful suggestions and cooperation in respond-
ing to requests for revisions. Their depth of knowledge, dedication, and patience
have been demonstrated throughout the process of preparing this book.
I thank Dr. Donald W. Rice for bringing me to this wonderful world of
electronic packaging and interconnection and for many fruitful discussions and
much strong support in the past six years at Hewlett-Packard (first HP Labora-
tories and now HP Corporate Manufacturing). I also want to thank my daughter
Judy and my wife Teresa for their consideration and patience by allowing me
to work on this project on many weekends.
John H. Lau
Hewlett-Packard Company
SOLDER JOINT
RELIABILITY
 1
Flux Reactions and Solderability
Colin MacKay

This chapter attempts, on the basis of experimental results available, to explain

the mechanism, for those fluxes that have been studied, by which they appear
to work. This chapter is not a discussion of what fluxes are available for elec-
tronics soldering, of how they are classified, or how they are used or applied.
It does not comment about the perceived relative merits of the various types
available.

1.1 FLUX HISTORY

Originally the idea of a flux as a soldering aid developed from the metal coatings
industry where the essential nature of having a clean surface before tin or zinc
coating was well known. Whether iron or copper, sheet metal was immersed in
an acid pickle solution prior to dipping in the molten metal bath. When soldering
of two components was required, it was obvious that similar cleanliness require-
ments applied. Consequently it became the practice to use the same chemicals
as for hot dipping, applied dropwise onto the joint area, to clean the joining
components of oxide or other contaminating reaction product.
The other consideration when hot dipping was to protect the molten metal
bath surface from oxidation. This was done with a "pot cover" of mixed fused
salts that floated on the molten metal. The second property required of a solder
flux is to provide a similar protective covering to the cleaned metal of the joint
components between the time that the surfaces are cleaned and the molten solder
2 FLUX REACTIONS AND SOLDERABILITY

wets the joint interface. This is exactly analogous to the pot cover operation and
it is not surprising that a mixture of the materials used for the two operations
developed for engineering soldering applications. The liquid acid component
cleaned the joint surface, and as the temperature continued to rise and the liquid
evaporated, the remnant salts fused in situ to provide the pot cover. Wetting of
the joint metals by a solder with subsequent capillary penetration completed the
soldering operation. Copious washing to remove the fused salt residues was, at
that time, no disadvantage.
Under these conditions it was easy to show that as the acid concentration
dropped, the time taken to clean the surface increased; consequently it took
longer for the soldering operation, capillary rise, and penetration to initiate and
complete. Soldering in the fledgling electrical industry used the same technology,
but it soon became apparent that washing to remove the residues had to be more
discrete, localized, and thorough, since water and salt residues represented a far
more serious hazard in electrical circuitry than did corrosion in a heavy, over-
engineered mechanical assembly. The natural precautions of only fluxing a small
area around the joint and carefully washing the residues from that area carried
over into the earliest electronics assemblies where heavy wiring harnesses of
individual wires and large components were involved. With the advent of the
printed circuit board and the associated wave soldering operation, residues were
distributed over the whole surface of the electronics assembly and their presence
became a serious consideration.
It was at this time that rosin-based fluxes became serious candidates for de-
velopment and rapidly took over the market as electronics fluxes. The assump-
tions about how fluxes worked and what controlled their efficacy were simply
carried over from the earlier ideas. As the complexity of the formulations of
fluxes developed, so did their proprietary nature. Black magic pays when there
aren't many witch doctors, so there was not a great effort to explain how these
improved fluxes worked. Flux classification became a matter of considering the
aftereffects of soldering; therefore tests and activity classifications became in-
creasingly concerned with the activity of the residues and not with the effec-
tiveness of a flux in promoting the soldering process. Little concern and less
explanation was available about how these complex formulations worked or what
controlled the magnitude of their effectiveness.

1.2 SOLDERABILITY TESTS

The most important requirement of any attempt to measure flux effectiveness is

a quantitative solderability test that would allow discrimination between different
formulations. Many tests have been reported in the literature, most of them based
upon some aspect of the soldering process. These fall under the headings:
 SOLDERABILITY TESTS 3

1. Visual assessment
2. Area of spread tests
3. Edge dip tests
4. Capillary tests
5. Globule tests
6. Rotary dip tests
7. Surface tension balance tests

1.2.1 Visual Assessment

Visual assessment relies upon inspector experience and memory. It offers no

quantatitive evaluation but with experienced personnel offers a rapid comparison
of a result with previously agreed visual standards.

1.2.2 Area of Spread Test

The area of spread test is a quantitative test that involves the spread, when
melted, of a weighed cylindrical tablet of solder, fluxed and heated on a square
coupon of substrate. The area is usually determined by planimetric methods.
This test does not have a time element; therefore it cannot indicate anything
about wetting rates. The test is allowed to continue until the spreading action
has stopped. It indicates the combination of the maximum substrate-solder in-
teraction with the minimum solder-air or solder-flux surface tension.

1.2.3 Edge Dip and Capillary Rise Tests

Both the edge dip and capillary rise tests involve the edgewise dipping of either
a single fiat piece of substrate material or a piece bent so that it creates a capillary
gap. Assessment is usually qualitative, the edge dip test in particular being used
to indicate solderability quality aspects (i.e., nonwetting, dewetting, pinholes,
etc.) The capillary rise variant could be used quantitatively if the meniscus rise
height were measured. These tests, like the previous one, measure an equilibrium
situation, stressing to a different extent the same interfacial forces.

1.2.4 Globule Test

The globule test is a very geometry-specific test designed for solderability as-
sessment of wires (preferably round wires). In this a weighed cylindrical tablet
of solder is heated on an anvil. When the solder is molten, a wire is lowered
into the solder bead. The time from initial contact until the bead coalesces around
4 FLUX REACTIONS AND SOLDERABILITY

the wire is a measure of wetting time. This test also includes an element of
applied pressure to the wetting interface resulting from the hydrostatic pressure
of immersion and is slightly biaxial in that the bead may wet along the wire as
well as around it. This is a rate measuring test.

1.2.5 Rotary Dip Test

Rotary dip testing is designed to measure the wetting time of flat sheet material
in a system that emulates the wave soldering process. It copies the relative
motion of solder and substrate by sweeping the test coupon tangentially across
the surface of a solder bath (as opposed to the solder wave sweeping the circuit
board surface). A contact timer indicates the wetting time for each test and a
series of samples is required for a series of different rotation speeds of the sample
arm. For a determination of optimum rotary dip solderability, the minimum
solder contact time that gives complete coverage of the test coupon is determined.
This test involves an element of squeegee action, a complex combination of
applied pressure, and enhanced thermal contact behavior that aids and speeds
the wetting process. Clearly this is a dynamic test.

1.2.6 Surface Tension Balance Test

The surface tension balance or wetting balance test measures the vertical force
acting on a test coupon from the time wetting is initiated until it is completed
or the force profile up until the time exceeds some preset maximum. Both
dynamic and equilibrium parameters are measured by this test in that the time
to reach any specified position in the wetting process can be determined. Also,
provided the wetting interface reaches equilibrium within the duration of the
test, the wetting force at the equilibrium wetting condition can likewise be
determined. Commonly the endpoints chosen are the time for the wetting force
to equal the upthrust force for a particular set of immersion depth and immersion
rate conditions, the time to initiate wetting, and the time for wetting to be
completed. Wetting rates at any position in the process can be calculated from
the profile and the final wetting force is measured directly.

1.3 FLUX ACTION FROM SOLDERABILITY MEASUREMENTS

A quantitative test is required in order to make an effective measurement of flux

action, and the type of test used determines which aspect of flux action is being
quantified. The area of spread test determines factors independent of time that
 FLUX TYPES 5

relate to the advance of the wetting interface across a surface. The surface tension
balance test determines rate factors that relate to the speed of wetting. Both the
rotary dip and the globule tests are more complex tests and are probably not
good tests for determining flux mechanisms because the results tend to be test
parameter dependent. The globule test is influenced by the size of the wire and
the size of the solder pellet, and the rotary dip test is influenced by the rotation
speed of the test arm.

1.4 FLUX TYPES

1.4.1 Mechanistic Studies for Inorganic Fluxes

On the basis of the historical development of fluxes, the purpose and action of
the flux was to clean the substrate of all contaminating films and then to protect
the clean surface from reoxidation until the coating metal or solder melts and,
by reacting with it, wets the surface. According to this description ofthe process,
any resulting response to a solderability test describes the effectiveness of the
cleaning (and/or the protection). Implicit in this would be that once the surface
was fully clean, the solderability would be optimum and could not be improved.
To test this, Turkodan and Zador l used a reducing atmosphere furnace to
reduce the oxide film on a degreased and cleaned steel strip. After flooding the
vessel with a protective, nonreducing atmosphere, a weighed quantity of tin was
placed on the strip and the area of spread measured. Optimization of the reducing
atmosphere and experiment parameters showed that the spread area reached a
maximum and thereafter stayed at this value, which represented the ultimate
spreading value possible with fully cleaned steel. When a weighed quantity of
inorganic flux salts was applied with the tin or solder under the same optimum
cleaning conditions, the spread area increased to being greater than the value
that had been obtained for clean steel. Clearly flux in this form provided some
additional aid to the wetting operation beyond what was possible from simply
applying tin to a clean metal surface.
The explanation originally offered was on the basis of a classical triangle of
forces equilibrium involving the three surface tension components,
'YH2Isolid, 'Y solidlliquid, and 'YH2/liquid (Fig. I-I):

'Y(H2/solidl = 'Y(solidlliquid) + 'Y(H2/IiquidICos 6

where 6 is the contact angle measured up from the solid interface to the surface
of the liquid. The area of spread for the fluxed system was greater than for the
hydrogen atmosphere system because 'Y(fluxlliquidl was smaller than 'YH2Iliquidl.
6 FLUX REACTIONS AND SOLDERABILITY

y H2 ILIQUID

HYDROGEN or FLUX

SUBSTRATE
SOLID
Figure 1-1 Wetting force equilibrium.

Effect of Impurities on Solderability

Later workers2 - 9 who used the area of spread test attributed the performance
improvement that they found to the reaction of the flux salts with the molten
solder, creating an alloy of the metal of the flux salt with the tin or solder, and
suggested that the alloy produced spread more readily than the tin or solder.
This, however, ignored results of other workers lO who had studied the effects

o 02 0·4 0·6
Sb IN 60% Sn - 40% Pb

Figure 1-2 Effect of antimony on area of spread of 60% Sn-40% Pb solder on brass base
metal (zinc ammonium chloride flux).
 FLUX TYPES 7

I I

i 10h-!~_!- -
a:
c..
(f)
LL.
OO'5~
«
-
UJ
a:
«

I I
o 0,2 0,4 0·6
Cu IN 60% Sn - 40% Pb

Figure 1-3 Effect of copper on area of spread of 60% Sn-40% Pb solder on brass base metal
(zinc ammonium chloride flux).

of deliberate additions of a range of impurities to 63% Sn-37% Pb and 60% Sn-

40% Pb alloys and had shown that in all cases this resulted in a decrease in
solderability. Some effects of impurities on area of spread results are shown in
Figures 1-2, 1-3, and 1-4 for the cases of low levels of antimony, copper, and
zinc impurity on the solderability of brass with 60% Sn-40% Pb when a zinc
ammonium chloride flux was used.

o 02 04 0·6 De 1-0
Zn IN 60% Sn - 40% Pb

Figure 1-4 Effect of zinc on area of spread of 60% Sn-40% Pb solder on brass base metal
(zinc ammonium chloride flux).
8 FLUX REACTIONS AND SOLDERABILITY

Only for the case of copper with brass as the substrate was any increase in
spread area observed. This was attributed to reaction of the copper with the zinc
dissolved from the brass surface to form small particles of copper-zinc compound,
which was observed in microstructure of the joint, and not the formation of a
modified solder alloy. The copper actually removed impurity arising from dis-
solution of the base material. In all other cases, for the brass, as shown, for
copper, and for steel substrates the area of spread was decreased. Similar effects
were also observed for wetting balance tests.
Table 1-1 shows a synopsis of these results for all impurities when fused salt
fluxes were used. The percentage column gives the impurity level percentage
that caused a 25 percent reduction in the area of spread value. Decreased areas
were observed both in systems that reacted with the solder to produce intermetallic
compounds with the impurity, as with copper, and in systems that dissolved the
impurity to form a solution in the melt. Only 0.4 percent copper was required
to initiate the formation of CU6SnS intermetallic in 60% Sn-40% Pb alloy, in
which it formed a fine suspension of needles of the compound. This in turn
caused the viscosity of the melt to increase with the corresponding inability to
flow and spread. Silver reacted in a similar manner, forming tiny platelets of
Ag3Sn. The effect of this compound was less marked than for copper. Antimony
compound does not form until the composition reaches about 8 percent, so at
the low levels used in these experiments it appeared to effect the spreading
properties only as a fully dissolved element.

Table 1-1 Effect of Impurity on Area of Spread Values on Various Substrates

Fluxed with Zinc Ammonium Chloride
Percentage of Impurity
to Produce 25%
Reduction in Area of
Impurity Test Piece Spread
Antimony Brass 0.9
Steel No effect
Arsenic Brass No effect
Steel 0.2
Bismuth Brass 5.1
Steel 2.8
Cadmium Steel 1.9
Copper Steel 0.5
Zinc Brass 0.04
 FLUX TYPES 9

An Electrochemical Mechanism

From the results mentioned earlier 2- 9 it was clear that a large range of inorganic
salts could be formulated into fused-salt-type fluxes and would demonstrate very
significant improvements in solder flow characterisitics. The most probable mech-
anism for the fluxing effect with inorganic fluxes, as put forward by Latin ll and
supported by both Bailey and Watkins l2 and Turkodan and Zador, 1 was elec-
trochemical. All of these studies detected an electrochemical potential between
either tin or solder and the substrate material in a cell with the fused salt as
electrolyte. Experiments l3 were made with a cell of the type shown in Figure
1-5, using fused salt mixes of ammonium chloride with stannous chloride and
ferric chloride, with tin or solder as one electrode and iron or copper as the
other. The gray film referred to by Latin ll and Baileyl2 was analyzed by scanning
electron microscopy and was found to be crystallites of CU6SnS on the substrates,
when either formulation was used at a temperature above the melting point of
the solder or tin . When lower temperatures were used, a thin film of solid tin
was observed. With steel as the substrate a thin film of FeSn2 was found. This
would indicate that the flux acts as an electrolyte of a fused salt electrochemical
cell.
The positive effects of a tin coating layer on solderability are well known, so
the deposition of tin had a clear beneficial effect since the substrate surface was
replaced by one that was more solderable. The reason for other materials having
beneficial action could be attributed to the extreme reactivity of freshly reacted
materials. Nascent hydrogen, freshly evolved, is a well-known example of this
phenomenon and freshly deposited metals also show this property. In the case
of a soldering situation any freshly deposited metal would be laid down im-
mediately before the advancing solder front at the wetting interface, ideally
situated for establishment of a metal-metal interaction layer and essential for
surface wetting.
When the work with a wide range of inorganic salts was examined, clearly
some large effects were observed. Although the explanation given was faulty,

FUSED SALT
FLUX

Figure 1-5 Test cell demonstrating electrochemical effect with fused salt fluxes .
10 FLUX REACTIONS AND SOLDERABILITY

the effects were real. Scrutiny of these results in the light of the material properties
of the compounds used showed a clear correlation between the metal ion poten-
tials and the free energies of the compounds, with the observed wetting responses.
For all the systems tested (chloride, sulphate, and nitrate compounds), those
with metal ion potentials below about -0.75 V had little, if any, effect on area
of spread. Metal ion potentials above this showed a progressively larger effect
(Fig. 1-6).
Similarly, compounds with high free energies were found to be less effective
than those with low values, with the lower energies showing a progressively
larger area of spread. The free energies of chlorides and nitrates, at which an
effect began to be seen, were similar, at about 100 kcals/mol (Fig. 1-7). Sulphates
in general had high free energy values and were effective at values below about
250 kcals/mol. Other properties included in the analysis were the valency of the
metal, the compound melting point, and the atomic weight. All these showed
no obvious correlation.
It seems, therefore, that for spreading capability of solder with inorganic salt
fluxes, the prime controlling factors are the combination of the electropotential

10
o

o o o
o

o
200
o
o
o
o
0~~--r-~-4--+--+--+--r--r-~-4--+--+--+-~~~4-~--+--+
-3_0 -2_8 -2.6 -2.4 -2_2 -2.0 -1.8 -1.6 -1.4 -1.2 -1.0 -0.8 -0_6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0

Metal Ion Potential

Figure 1-6 Effect of nitrate metal ion potential on area of spread of 60% Sn-40% Pb solder
on copper base metal.
 FLUX TYPES 11

5.5

5.0

0
4.5 0

4.0 0

3.5 0
~
~

1il 3.0
~
Q.
en
'0 2.5
'~"
c(

2.0
0

1.5

1.0 0
0
0
@
0.5
-200 -180 -160 -140 -120 -100 -80 -60 -40 -20

Metal Ion Potential

Figure 1-7 Effect of chloride metal ion potential on area of spread of 60% 5n-40% Pb
solder on copper base metal.

of the metal ion and the ability of the compound to dissociate in a fused state.
This agrees with an electrochemical mechanism being involved in the deposition
of a metal ion onto the substrate in front of the advancing solder.

1.4.2 Mechanistic Studies for Rosin-based Fluxes

As mentioned previously, the reasons for using fluxes are well known. First,
they clean the surface to be soldered by removing the oxidation and/or other
chemical contamination. Second, they form a protective blanket over the clean
surface to prevent reoxidation until the temperature becomes high enough for
soldering. However, fluxes for electronics applications rarely consist of simple
chemical solutions; usually they contain complex formulations designed to be
both as wide ranging and process specific as possible. For example, a flux might
be chosen so that it can cope with a particularly difficult surface or perform
more quickly.
For electronics applications, fluxes are based on an organic rosin vehicle and
contain the minimum possible activity so that any residues are noncorrosive.
12 FLUX REACTIONS AND SOLDERABILITY

Assessment of a flux's activity and assignment to a designated activity level

among the classes acceptable for electronics applications depend upon a com-
bination of criteria that often reduces soldering performance to a secondary
consideration. Some classifications are by percentage of activator, others distin-
guish fluxes by the activator used, while still others combine water extract
resistivity and flux corrosiveness on a thin copper film deposited onto glass as
the definitive classifier.
Nevertheless, it is not realized that equivalent formulations conforming to all
activation levels can be compounded from different amounts of activators that
are not necessarily related. These fluxes could behave differently on a series of
clean metals or on a range of increasingly contaminated or oxidized copper
surfaces. Indeed, fluxes are called upon to operate on a wide range of oxidation
levels, not simply on a set of carefully controlled surfaces. Both the speed and
quality with which a flux renders such a range of surface solderable reflect the
flux's usefulness.

Rosin

The principal constituent of an electronic flux is rosin. This is a natural product

derived from the resinous excretion of trees. It is a glassy mix of abietic acid with
some polymeric compounds of abietic acids and numerous hydrogenated modifi-
cations of that acid. The basic chemical structure is shown in Figure 1-8.
Since it is a natural product, the proportions of the various constituents will
vary from genus to genus and also for different variations within a botanical
group. In order to obtain as consistent a material as possible and thus produce
a consistent, reliable flux product, rosin from a single species is used. For

Figure 1-8 Basic structure of abietic acid.

 FLUX TYPES 13

example, in Europe, the material commonly used is derived from a specific type
of Portuguese pine; in America, a North Carolina variant is specified. In addition
to the gross differences in product constitution found between types of trees,
trees of the same type grown in different locations, even material fr<}m the same
woods in different years, can show significant variations. This obviously affects
product uniformity and repeatability.
Because of its solid state and glassy structure the acid action of rosin is not
apparent until it melts. Only then do the ions of the material have sufficient
mobility to enter into reactions with substrate surfaces. The material has a pK
rather than a pH number; that is, it exhibits ionic character only in the fused
state.

Rosin Flux Formulations

Although commercial fluxes are formulated as solutions in alcohols so that

the dispersed ions do have some mobility at room temperature, there is very
little evidence of room temperature reaction on an oxidized copper surface. The
main purpose of the solvent is to allow the application of a thin uniform film of
the flux or, with further formulation, to allow a specific process application
method (for example, foaming), as in the foam fluxing method. Preheating of
the fluxed substrate, while serving to reduce thermal shock as the substrate
contacts the molten solder surface, also serves to remove the solvent of the flux
system and create a thin uniform film of rosin together with whatever activators
have been added.
The rosin starts to soften at between 50° and 70°C and is fully fluid at about
120°C. In this condition it is weakly acidic and, without any augmentation, will
clean lightly oxidized copper. However, printed circuit boards and components
do not come in only one grade of oxidation. Storage conditions and controls
produce a range of differently oxidized surfaces and the soldering assembly
process is expected to easily cope with all of these and produce 99.99 + percent
yields.
Another aspect of the soldering process is the production rate. Line speeds
that allow economical production rates have to be achieved, so that although a
plain rosin flux might be able to fully clean a surface eventually, the time required
to do this could be unacceptably long. In order to improve both aspects of a
plain rosin flux, ionic additives (usually chlorides) have been formulated on the
basis of the same idea about fluxing mechanisms as applied for inorganic fluxes.
Therefore the flux series designated rosin mildly activated (RMAs) contains less
activator than does the rosin activated (RAs) series, and so on up to rosin super
activated (RSA).
14 FLUX REACTIONS AND SOLDERABILITY

The problem with unbridled application of the most activated materials was
that the residues remaining after soldering were corrosive to varying degrees.
This led in tum to specification of materials according to the amount and activity
of the residue and the inclusion of tests such as the copper mirror test of residue
extracts within the activity specification for electronics fluxes.
Technological developments within the formulation industry also served to
make any soldering performance-composition correlation more difficult. By pre-
paring formulations with two-part activators that reacted to form the active
constituent in situ, the temperature at which reaction with the substrate began
was raised from being at room temperature to being above that of the reaction
temperature of the two parts. Activators in which the active halogen is hydrogen
bonded to a second molecule (usually amine) are also used. The temperature
required to break the bond and release the halogen compound represented the
minimum activity temperature. Yet higher activity temperatures could be realized
with compounds that incorporated halogens covalently; the more stable bond
required higher dissociation temperatures. Implied in these developments was
the idea that because they required higher temperatures to initiate activity, the
residues would somehow be less active after soldering. This completely ignored
the fact that often the compound used as a carrier was volatile and was not
available for recombination after soldering; therefore, one-half of the initial
compound was no longer available for recombination after soldering. Only in
the case of covalently bonded compounds was there evidence that residues were
less active.

Wetting Balance Performance of Rosin Fluxes

So how well do rosin fluxes of different types perform? Wetting time tests
using the standard surface tension balance procedures, for a range of different
surface oxidation levels, achieved by immersion in an oxidizing solution, show
the following results. 13
The curves of Figure 1-9 essentially exhibited the same form, consisting of
an initial sharp increase in wetting times followed by a plateau that remained
nearly horizontal as oxidation levels, represented here by immersion times in an
oxidizing solution, increased. Further increases produced an abrupt rise in wetting
time immediately prior to the surfaces becoming nonwettable. In this region the
results diverge considerably.
Certain features and the shape of the wetting time-immersion time curves
become significant if it is assumed that all fluxes deposit an equal film over the
test coupon so that the same volume of active constituent is involved in all cases.
In this way a controlled quantity of rosin plus activator was available to react
 FLUX TYPES 15

3 I 2
12 5

11 2

10 6

9
9

/
U8
Q)

-;7
(/)

E
F 6
OJ
c:
.~ 5 11
~

20 40 60 80 100 140 180 220 260 300 340

Immersion Time (min)
Figure 1-9 Wetting time curves on oxidized copper prepared by immersion in an oxidizing
manganate solution, for the range of fluxes shown in Table 1-2.

with the oxide layer and made each sample representative of the bulk flux from
which it was taken. Results then became comparable.
The fluxes tested were all of the rosin type with grades RMA and RA. They
were also all from a single commercial supplier and the numbers quoted are for
that generic product series. Details of the application and performance of these
fluxes is given in Table 1-2. Plain rosin flux with no activator was also examined,
but since the maximum immersion time treatment that this flux was able to solder
was only 1 min, this would not fit on the scale of Figure 1-9.
All fluxes exhibited essentially the same features, namely an initial zone in
which the wetting time increased sharply, followed by a level plateau, and then
a second sharply increasing region that rose to an infinite wetting time (i.e.,
nonwetting). In some cases, the plateau was vestigial, as for RMA-l, while for
others, such as RA-12, it was extremely extended. In general, the longest plateaus
were for the most active (i.e., RA) fluxes. The plateau levels also tended to
mirror the activity levels, occurring at the shortest wetting times for the RA
16 FLUX REACTIONS AND SOLDERABILITY

Table 1-2 Properties of Fluxes with Test Results in Figure 1-9

Flux Ref. No. Classification Properties
R 40% pure water white rosin in alcohol. Not
activated, used for electronic soldering
where residues are left on boards.
2 RMA Rosin in alcohol, 37% solids activated with
a nonionic agent. Used for electronic
soldering by both foam dip and wave
fluxing. Residue removal necessary only
in most critical applications.
3 RMA Most active rosin in alcohol flux in its
class. 37% solids activated to higher
level than (2) with different non ionic
activating agent. For electronic soldering;
also for dip and wave applications.
Residue removal as above.
4 RMA Rosin in chlorinated solvent. Medium
activity flux as (2). Activated with low
level of ionic activator (37% solids).
5 RMA Special low solids content rosin in alcohol
activated with nonionic agent (25%
solids).
6 RA Rosin in alcohol, organic acid activated
flux. Completely halide free. Reduced
residue corrosivity (15% solids).
7 RA Rosin in alcohol, high solids content (57%)
with an ionic activating agent. Does not
degrade during extended soldering times.
8 RA Rosin in alcohol, fully activated with an
ionic activator. Residues generally must
be removed (25% solids).
9 RA Rosin in alcohol, low solids (13%) with an
ionic activator. Easy cleaning flux.
10 RA 35% rosin in glycol ether solvent with an
ionic activator. Low volatility, low
maintenance flux.
11 RA Rosin in alcohol (25% solids), moderate
activity. Residues efficiently removed by
chlorinated solvents.
12 RA Rosin in alcohol spirit, special low wicking
formulation with an ionic activating
agent (25% solids).
 FLUX TYPES 17

fluxes and highest for the RMA and R grades. Some RA fluxes, however,
exhibited the same plateau wetting times as the RMA materials; compare, for
example, RA-12 and RA-6 with RMA-3.

An Active Constituent, Concentration Dependent Mechanism

From classical chemical studies it is known that the typical oxidation process
follows an exponential growth curve and is particularly flat for copper (refer to,
for example, Metals Handbook, vol. 6) over an extended thickness region. The
immersion time scale on these graphs would therefore transpose to an essentially
linear increase in oxide thickness. With permanganate treatment, it seemed that
any initial nonlinear oxidation part was short, on the order of I-min immersion
time, as indicated by the similarity of the wetting time profile for plain rosin
flux with those for the activated series. The treatment still gave evidence of
increasing linear oxidation out to 6-h immersion time.
While these results were indicative and comparing them with normal soldering
should be done with some caution, they do serve to clarify how the wetting rates
are modified by the various compositions and they do contain solid evidence
about how these fluxes work. Electron spectroscopy for chemical analysis (ESCA)
clearly showed that the oxide film produced by permanganate solution differed
from the natural oxide. In the test specimens, proportions of CU20 and CuO
appeared to be nearly equal, whereas no CuD occurs in a natural film grown at
ambient temperatures. The test films contained more CuD than would be expected
from a film formed in air at high temperature. Precisely how this disparity affected
fluxing action was difficult to assess.

Effect of Impurities in Solder with Rosin Flux

As with fused salt fluxes the general effect of impurities in solders was to
decrease the area of spread results, as for example Figure 1-10. The effects on
surface tension balance results were mixed and depended upon the phase diagram
of the ternary systems that the impurity elements formed with the tin-lead alloy.
For cadmium, whose results for both wetting rate and equilibrium wetting force
are shown in Figure 1-11, wetting times increase and the equilibrium wetting
force decreases. When antimony was the impurity (Fig. 1-12), there was no
significant change in wetting time, while the equilibrium wetting force showed
a significant decrease. In this situation the area of spread would have shown a
decrease. By comparison, when bismuth was present (Fig. 1-13), wetting times
actually decreased while the wetting force increased. This represented a smaller
area of spread, which was reached more quickly.
 Figure 1-10 Effect of cadmium (
area of spread of 60'X. Sn-40% I
o 05 , 0 , 5 20 solder on copper base metal (nona
Ca IN 60% Sn - 40% Pb tivated rosin flux).

-
1:7
~ I
ftlhi~r___r----r--
T •

0·6

2{) 'T

I 1/
[I
~ t:1i
,.~
u
Q)
C/)

u.i
::;';
~ ',0
CJ
Z
~
~r;L
f-
~ O~
1 Figure 1-11 Effect of cadmium or
wetting properties of 60% Sn-40°j,
o 2 .: 5 Pb solder on copper base meta
Cd IN 60% Sn - 40% Pb (nonactivated rosin flux).

18
 T

09~

()
OJ
, 5 -
en
W

~lJ.-~t
~ '·0 -

j
I-
(')
z
f=:
t;J 05
:s: t •
Figure 1-12 Effect of antimony on
I I I I wetting properties of 60% Sn-40%
0 2 3 4 5 Pb solder on copper base metal
Sb IN 60% Sn - 40% Pb (nonactivated rosin flux).

!: ~%t-I-I-I-f-
a? 08

05
()
OJ
en
w0 5
::2:

..
f=:
~~- •
(')
Z
f=: t- l-
I- Figure 1-13 Effect of bismuth on
W
:s: wetting properties of 60% Sn-40%

° 2 3 5 Pb solder on copper base metal

4
Sf IN 60% - 40% Pb (nonactivated rosin flux).

19
 20 FLUX REACTIONS AND SOLDERABILITY

Thermodynamics of Flux Action

Thermodynamic studies of the processes taking place have been reported using
differential scanning calorimetry (DSC),14 thermogravimetric analysis (TGA),14
and Fourier transform infrared analysis. 16 When the DSC and TGA results were
examined in combination, so that a peak in the DSC scan, that occurred at the
same temperature as an arrest in the DT A scan, could be assigned to a weight
loss mechanism, and comparisons were made of results from a progressive series
of mixes of components used in a typical flux formulation until a complete model
formulation was assembled, certain features of the fluxing-soldering process were
seen.

Starting Materials
Abietic Acid The basis material used for the flux vehicle was abietic acid,
which was examined by both thermogravimetric analysis and differential scan-
ning calorimetry with traces as shown in Figure 1-14.
The decomposition of this material began at about 150a C, as shown, with
most of the reaction as shown by the sudden, sharp weight decrease, taking place

220

200
DSC oxo
180

160

1<10

.
E
s:
!:
120

100
D
Z

80

60
ondo
<10

20

0
0 50 100 150 200 250 300 350 500
Tellllerllture ,·cl
Figure 1-14 TGA and DSC scans of abietic acid.
 FLUX TYPES 21

at about 275°C. By about 475°C the material had fully decomposed, yielding a
residue weighing only about 3 percent of the original sample weight.
DSC scans of samples sealed maluminum pans (Fig. 1-14) showed a twin
endothermic peak, the first beginning at 49°C, the second at 52°C, and both
persisting to 148°C. These did not correspond to any weight loss temperatures
indicated by the TGA scan; therefore they must have represented the complex
melting behavior of abietic acid. The twin peaks at about 250°C corresponded
to the position of a maximum rate of weight loss in the TGA and represented
the decomposition of abietic acid into various molecular fragments. The DSC
scan also showed two well-resolved endotherms at 380°C and at 425°C, indicating
reaction between fragments of the molecule, possible evidence of polymerisation
to dimeric or trimeric products.
When a sample with a vent hole in the lid was scanned in the DSC module
(Fig. 1-15) the twin melting peaks were once again apparent. A similar rise was
observed at 149°C in the scan for the sealed sample, but instead of decaying
gradually, this peak persisted until suddenly decaying at 275°C. This sudden
decay coincided exactly with the complete decomposition of the abietic acid in
the TGA scans. This supports the contention that the presence of volatiles in the
pan (sealed samples) produced further interaction peaks in the DSC scan rep-

5.-----------------------------------------------~

810

...•
o

...•
II.

8ndo

-240----~~---1'OO----1'~----2'OO----2'50----3~OO--~~~O~~~IOO~~~5T.O~~500
Tellllereture ,·el
Figure 1-15 Comparison of DSC-scans of abietic acid with sealed and vented sample pans.
 22 FLUX REACTIONS AND SOLDERABILITY

resenting some other interactions occurring among the volatiles or with the
resulting residues.

Isopropyl Alcohol (IPA) A TGA scan of IPA, using a lOoC/min heating

rate, starting at room temperature, resulted in a smooth, continuous weight loss
curve, ending with a 0 percent weight residue at 90°C. The vented sample in
the DSC results showed a curve having one single endothermic peak beginning
at room temperature and ending at about 80°C. The two curves were consistent
with a simple evaporation process.

Triethanolamine-Hydrochloride (TEA-HCI) Using a lOoCimin heating rate

on the TGA, the TEA-HCI weight loss scan of triethanolamine-hydrochloride
showed a complex arrest starting at 51°C, with possible side reactions starting
at lOO°C and ending at 185°C. This arrest represented a weight loss of about
17.5 percent, which agreed relatively well with the theoretical percentage of
hydrochloric acid (19 percent) in TEA-HCl. The slightly low value could be
caused by adsorbed moisture in the deliquescent TEA-HCl, making the apparent
molecular weight higher than 185.7. Indeed, the first, small arrest between 51
and lOO°C could be the evolution of water vapor. The major arrest at 275°C

220

200
DSC
exe
180

160

140 ..
I

I
E 120 <--- TEA Inatantanaoua Malt

..•
~
s:
DO

z
100
<HCl Evolution>
TaA
TEA D8co~oaltlon

80

60
snde
40

20

0
0 50 100 150 200 250 300 350 400 450 500
T8~aratura I·e)

Figure 1-16 TGA and DSC scans of triethanolamine hydrochloride.

 FLUX TYPES 23

represented either the boiling or decomposition (or both) of the triethanolamine

that had formed immediately following hydrochloric acid evolution.
The DSC scan for this material (Fig. 1-16) showed an endotherm starting at
room temperature, probably associated with adsorbed water vapor, followed by
a second, which occurred between 75°C and 90°C. This latter peak was probably
caused by the evolution of any water of crystallization. The large peak at 177°C,
while indicative of the evolution of hydrogen bonded HCI molecules, was caused
by the spontaneous melting of triethanolamine, which was formed at above its
melting point when the HCl molecule broke free. This mechanism must take
place if the reported physical properties of triethanolamine and its HCl adduct
were correct. The dissociation temperature of the adduct was 177°C, while the
melting point of the parent compound was only 25°C. The boiling point of
triethanolamine, 177°C, was shown in the endotherm starting at about 250°C.
A superimposed fine structure on this peak indicated that this material also
decomposed at these high temperatures.

Copper Oxide (CuD) The TGA scan for this material, over the entire tem-
perature range, was a continuous horizontal line at 100 percent weight, showing
no weight loss. The DSC scan was also a featureless horizontal line, indicating
no reactions taking place.

Solder Powder The TGA scan for solder powder was essentially a continuous
horizontal line, as for copper oxide. The DSC scan at a lOoC/min heating rate
showed a single melting point endotherm at 179°C, which was in excellent
agreement with the 179°C melting point of the 62% Sn-36% Pb-2% Ag alloy.
As with all other scans, higher heating rates shifted the apparent melting point
to higher temperatures so, using a 50°C/min heating rate, the melting point
appeared to be 183°C.

Materials Combinations
Abietic Acid Plus IPA In combination, the abietic acid and isopropyl alcohol
system represented the solvation of the basic rosin vehicle in a typical flux
solvent. These components are considered inert and unreactive with one another.
For plain rosin (R grade) flux, which is this system, the active constituent is the
rosin, which when molten, behaves as an organic acid. The solution examined
in this and subsequent experiments is a 45 percent solution of abietic acid in
IPA.
The TGA scans were performed initially with open (no lids) pans. In all cases
the scans showed an intermediate arrest, indicating the system was not simply
one of classical evaporation. The presence of the arrest was evidence of some
type of bonding between the abietic acid and the solvent, resulting in additional
thermal energy being required to disrupt the association. This showed up in the
 24 FLUX REACTIONS AND SOLDERABILITY

220

200
,
---
DSC

180
\
\
\ ;----- 0>0

~v
160

1040 <-- IPA Evaporation and :r

x AbietiC Acid Malt '"
~
.§
120 "<;:I:
...•
..••
D
I&.
100
TGA 3'
~
:J:
80

60
ondo
040

20

0
0 50 100 150 200 250 300 350 0400 450 500
Te.pereture I·C)

Figure 1-17 DSC and TCA scans of abietic acid plus isopropyl alcohol.

traces as an arrest. When the bonds were broken, the solvent was released and
a sudden weight loss was registered. The precise temperature at which this
occurred was 150°C. The final residue in all cases was only 3 percent to 5 percent
of the original sample.
The associated DSC scans for this system, illustrated in Figure 1-17, indicated
a large composite endothermic reaction peak consisting of at least two compo-
nents. The first initiated at a temperature corresponding to the first weight loss
arrest in the TGA scan and represented the evaporation of solvent. The second,
which merged progressively with the first, occurred in the middle of the weight
loss arrest and represented the melting curve of abietic acid. This was followed
by an extended, more shallow peak corresponding with the decomposition of
abietic acid. The melting peak also showed a small satellite at the two lowest
heating rates, indicating reactions occurring at -120°C.

Abietic Acid Plus IPA Plus Activator The next stage in the sequence in-
volved adding a typical activator, in this case 0.3% TEA-HCl. The TGA scans
were almost identical in comparison to the trace with no activator. The slight
subsidiary arrest observed corresponded with the total evaporation of the IPA
solvent. The only effect of the activator addition appeared to be that the decom-
position temperature of the abietic acid was delayed by 5°C to lOoC. The only
 FLUX TYPES 25

220

200
0.0
PSC
180

160

I
1.010
'"
'"
...,
~

..•.
120 0

...z: 3''"
100 TGA ~
x
80

60
ondo
.010

20

0
0 50 100 150 200 250 300 350 .0100 .0150 500
Te!lpareture ,-C)

Figure 1-18 DSC and TGA scans of abietic acid plus isopropyl alcohol plus triethanolamine
hydrochloride.

other evidence of the small addition of activator was seen on the 10°C/min
heating rate curve where a noticeable change in the slope of the weight loss was
indicated at -150°C, which would correspond to the weight loss from the ev-
olution of HCI, as shown in the TEA-HCI scan.
On the DSC scans, illustrated in Figure 1-18, the IPA evaporation peak merged
into the abietic acid melt peak. The effect of the TEA-HCI addition could be
seen only on the 10 and 50°C/min scans, where the TEA-HCI appeared as a
satellite initiating at 130°C, in agreement with the temperature assigned to the
evolution of HCI on the pure TEA-HCI curve and the associated large melting
peak of tetraethanolamine, which formed spontaneously.

Abietic Acid Plus IPA Plus Activator Plus Copper Oxide Copper oxide was
then added to the mixture as a representation of the oxide normally present on
a copper surface of an assembly. It was added in excess in an effort to maximize
any reaction occurring. When a major problem associated with splattering of the
slurry that formed as the solvent evaporated had been overcome, continuous
curves were obtained, typically showing three arrests. The first ended with a 68
percent weight loss, corresponding to complete evaporation of the IPA solvent.
The second corresponded to the decomposition of the abietic acid and the third,
 26 FLUX REACTIONS AND SOLDERABILITY

occurring at -420°C, was associated with the presence of copper oxide in the
sample.
Calculations using values from the traces with and without CuO indicated that
probably the CuO had not been reduced to copper metal during this experiment.
DSC scans of this fonnulation with vented samples (Fig. 1-19) showed the
same trends as both abietic acid plus IPA and abietic acid plus IPA plus TEA-
HCI traces. The twin peaks of the IPA evaporation and the abietic acid melting
were the same as for the above two systems with the melting peak being larger
than the evaporation. The abietic acid decomposition endothenns were also
similar to those observed previously. One additional large feature was the ex-
othenn at 400 to 450°C associated with the presence of copper oxide. There was
also a small additional satellite peak at -140°C.
Sealed DSC samples, while essentially exhibiting the same features as the
vented, did reveal some differences. In these, the IPA evaporation endothenns
were larger than the abietic acid melting endothenns. The exothenn associated
with copper oxide was not so pronounced and occurred at slightly lower tem-
peratures. These two series of DSC scans also indicated that there were differ-
ences in the reactions that took place when the volatile components of the
reactions were allowed to escape and when they were contained.

~o,-------------------------------------------------~

••0
See led SallPle

..•
o
II. Vented SallPle
•
~

I
.ndo

-~~O--~~50r-~-tOrO~--t~~---~~O--~2'5-0~-3~OCO~-3=5CO~~~OLO~~~5T.O~~5~OO
TellPereture I·C)

Figure 1-19 Comparison of DSC scans of abietic acid plus isopropyl alcohol plus trietha-
nolamine hydrochloride plus cupric oxide in sealed and vented sample pans.
 FLUX TYPES 27

100,---------------------------------------------------.

elo

'vv------------ Run f1 / - '......_ _

__-------Ru-n-f-2-------------~
...•o
...•
I&.
Run f3
-----..
_---
....--_.... ..- .._----
.
J: "'\ /
~ .. ~ ...

\ I
\j
\ .
endo
I I
\.
I~

-1004-~_.r-~-.-,--._~_.--r-,_~_.--~_.~--~~~~~
o ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Te.perature ,-C)

Figure 1-20 DSC scans indicating the progressive reduction of copper oxide with repeated
application of flux.

The presence of the peaks associated specifically with the copper oxide, al-
though absent from the raw copper oxide scans, indicate a reaction taking place.
The exothermic nature of these peaks suggests a re-oxidation process, although
the weight loss seen in the TGA scans seemed to show continued reduction.
When the pans were opened after performing the scans, the black CuO had
changed to a coppery red color. Microscopic examination of this showed gray-
black films over parts of the residue, which might confirm the reoxidation theory.
In an additional set of experiments, the residues from a DSC scan were
repeatedly refluxed and rerun. The successive scans (Fig. 1-20) show the gradual
reduction of the peak at 400°C and its eventual disappearance. Visual examination
of the residue after the final scan showed a sintered coppery mass indicative of
complete reduction of the oxide. From these observations it appeared that the
reaction progressed in two steps. Copper oxide was first reduced to cuprous
oxide, which is reddish in color, and then was further reduced to copper metal.

Abietic Acid Plus IPA Plus Activator Plus Solder Powder As an interme-
diate for the full simulation of the complete soldering process, small additions
of 62% Sn-36% Pb-2% Ag alloy solder powder (-325 mesh) with a melting
point at 179°C were added to the flux formulation without any copper oxide
 28 FLUX REACTIONS AND SOLDERABILITY

being present. The TGA traces of this system showed a minor arrest occurring
at a slightly higher weight percent than the weight loss corresponding to loss of
all the solvent and a major arrest corresponding to the decomposition of the
abietic acid when TEA was present. The weight loss percentage at this point
corresponded exactly with that expected for decomposition of the abietic acid
when allowance was made for the solder powder.
The main features of the DSC scans, both vented and sealed, were the abietic
acid melting and decomposition peaks together with the solder melting peak. In
the traces for the two lowest heating rates, vestigial peaks at -130°C were
observed, indicating the evolution of HCI from the small amount of activator
present. The interaction of the flux with oxide on the solder powder surface did
not show up as a thennodynamic peak in these traces.

Abietic Acid Plus IPA Plus Activator Plus Copper Oxide Plus Solder
Powder When vented samples of these full soldering process simulation sam-
ples were scanned in the DSC, the major features were again the melting and
decomposition of abietic acid and the melting peak of solder. A peak at -400°C
associated with CuD was also present. At the lowest heating rate there was a
side peak at -125°C, as seen in Figure 1-21.

220

200 Pre8ence of Copper Oxide

DSC axo
180

160
~ <-- Solder Powder Melt
1..0
IPA Evaporation. Abietic Acid
is 120
...s::. Malt II TEA Malt

.....'" 100 TGA

z
80 ~ _ _ _ _ _ _ _-=Ab~ietlc Ac1d II TEA Deco&po8itn

60 IPA Evaporatn. II HCl Evolutn.

aodo
CuD Reductlon
..0

20

04----,--__,-__-,--__,-__-,--__ ~~_.~~~~_r~~

D ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Ta.peretura I·C)

Figure 1-21 DSC and TGA scans of a simulated soldering process abietic acid plus isopropyl
alcohol plus triethanolamine hydrochloride plus cupric oxide plus solder.
 FLUX TYPES 29

The scans from samples with sealed pans were more complex than those from
vented pan samples. A complex of satellite peaks were observed between -120
and 150°C; also the single exothennic peak from CuO was split into two, pre-
sumably due to an effect of reaction with the constrained vapors.
It was clear from comparison of the results of the vented and sealed DSC
experiments that the exact sample conditions have an effect on the reactions
taking place. In a real soldering situation obviously the assembly to be joined
would not be contained, and volatiles would be able to escape. However, the
volume of flux used would be considerably more than used in these experiments;
therefore it would be entirely possible that with some of the flux acting as a
blanket, conditions similar to those existing in our sealed tests could occur at
the flux-metal interface.
The testing using CuO as the simulated surface oxide was clearly too severe
for comparison with practical soldering reactions, but it showed some interesting
features of the fluxing mechanism, in particular a stepwise process for reducing
higher oxides. Exoelectron studies by Momose and TamilS had suggest that the
lower oxide, CU20, could have been reduced at -180°C so that CU20 could
have been a better choice of surface oxide simulation. These results indicated
that CuO had a somewhat more complex behavior, possibly associated with
reoxidation taking place at -400°C.

Abietic Acid Plus IPA Plus Activator Plus Copper Powder Plus Solder
Powder The results of experiments using copper powder are indicated in Figure
1-22. In these, the peak at -400°C associated with the CuO was absent. Two
shallow extended peaks corresponding to the fonnation of CU6SnS plus CU3Sn
fonned at temperatures consistent with the fonnation temperatures of these com-
pounds as shown in the phase diagram for this system.
The study was able to show many aspects of the sequence of events taking
place during a complete soldering operation. However, it had not been able to
show all the sequential details of the reactions involved. In part this was because
of the magnitude of the thennodynamic energies involved. The large energies
involved in change of state reactions such as melting and evaporation tended to
mask the smaller energies associated with chemical reactions such as the for-
mation of copper abiate. The glassy nature of the molten rosin that produced
the melting peak over such a broad temperature range was obviously important
since it tended to mask any sign of the other reactions taking place in that
temperature range in the mixtures studied.
A fonnulation with only 0.3% activator addition, while it corresponded to
the practical situation of an RMA grade flux, did not really allow observa-
tion of activator reactions in the flux systems. A higher level of activation,
although not necessarily a practical composition, could have been more in-
fonnative.
30 FLUX REACTIONS AND SOLDERABILITY

30~-------------------- __________________________--;

Copper Powder

\I
.,.,-----------~------
elo

r ---11
r---_ . . .
X
.!! V~ 1/V
\ Old Copper Powder
•o
........ ------------------~
••
41

:x:

I ,
/ ..........
Copper Ox ide __---/" .,____ _
,
....... .
.--.-. r---------....--------.. --...-..-
Ii
eodo
\. \I
•......1 ......, :
V
V

100 150 200 250 300 350 400 450 500

Temperature I-C)

Figure 1-22 Comparison of DSC scans with fresh and aged copper powder with copper
oxide.

Fourier Transform Infrared Analysis

Diffuse reflectance Fourier transfonn infrared analysis (DRIFT) has been used
to study the surface reactions of rosin with a variety of metal oxides common
to soldering systems. 16 By measuring the diffuse scattering from the surface of
a sample and comparing it to the scattering from a reference surface (usually
KBr), the absorbance characteristics of the sample can be measured. In the case
of flux-oxide powder interactions the diffuse reflectance spectra showed changes
arising from reactions within the flux film on the surface of the oxide powder.
Copper, tin, and lead oxides were studied, together with salts of copper and
lead. In the case of copper and stannous oxide (SnzO) the absorption band at
1600 cm- I , arising from the acid group [COOt I , was monitored. From these
spectra the presence of copper and tin carboxylate compounds were detected at
the lowest test temperatures used, 260°C. With stannic oxide (SnO), no reaction
was observed, even up to the highest temperatures used (350°C). This is not
really surprising considering the stability of this last compound. The same gen-
eralized conclusions did not hold for the lead compounds, in that Pb 3 0 4 , a highly
stable oxide, showed evidence of reaction at the lowest temperature (260°C) but
with the absorption band shifted to 1520 cm- I . Other lead compounds reacted
 FLUX TYPES 31

differently, in that lead chloride did not react below 300°C, while lead sulphate
was unreactive even at 350°C. Lead sulphate is highly stable while lead chloride
has a free energy that is similar to stannous oxide.
It seemed from these studies that the reactivity with the flux and the solubility
of the oxide or the reaction product with the flux were important factors and
that these reactions in the case of lead compounds were complicated.

Rosin Flux Action Inferred from Solderability Measurements

The wetting time experiments mentioned earlier l3 allow some direct inferences
as to the mechanism of flux action in a rosin-based system. If flux action were
dependent upon reactant concentration, either rosin itself or the activator or both,
as has been supposed, from classical considerations the reaction rate of the flux
with an oxide film would be simply dependent upon the concentration of the
active species in the solution. In physical chemical terms, for a system that
depends only upon a single reactant, which is a first order reaction, this can be
expressed as

dc
- = kc
dt

where c is the concentration of the reacting substance, which is a mathematical

representation stating that at any instant the rate of disappearance of the reactant
is proportional to the concentration at any instant.
If Cj is the starting concentration and this has reduced to c, after time t, the
remaining reactant concentration is Cj - Ct. In that case the above equation
becomes

d(Cj - c,)
--.:....-=---.:.:. = k( Cj - c,)
dt

or

d(c,) = k(cj - c,) dt

The reaction rate constant shown here is called the specific reaction rate. From
this it can be seen that the concentration of the reactant at any time is a function
of the form

which is a curve of the form shown in Figure 1-24.

32 FLUX REACTIONS AND SOLDERABILITY

soo

400

JOO

zoo

100

OXIDATION IMMERSION TIME (mins)

Figure 1-23 Oxide dissolution rates calculated from wetting time curves on oxidized copper
prepared by immersion in an oxidizing manganate solution, for a range of fluxes.

For a reaction involving two components A and B the basic equation becomes

(]x = k(A - x)(B - x) dt

where A the concentration of A at t = 0

B the concentration of B at t = 0
x = the reduction in concentration of A and B at t =
In the case of a solid film dissolving into a solution being one of the constit-
uents, the concentration of the dissolved solid remains constant until all of the
solid has dissolved. Only then does the concentration of that component begin
to decrease.
In this case the first part of the reaction behaves like a first-order reaction,
while the last part behaves as a second-order reaction. In either case the general
form of the relationship between concentration and time is a negative exponential,
as before.
The effect of having only a restricted flux volume is to accelerate the rate of
decrease of concentration so that the reaction rate drops more sharply. Higher
order reactions would have the same general negative exponential form
No matter which precise order of reaction is occurring, the expected wetting
time curve under this type ofregime would be a continuously increasing wetting
time as oxide thickness increased. But this is clearly not what is observed.
If we consider what is actually happening for the points across the wetting
time versus oxidation curves (Fig. 1-9), the oxide is steadily being moved until
a clean metal surface is exposed. When the metal is clean, the solder metal reacts
 FLUX TYPES 33

4.4

4.2

4.0

3.8

3.6

3.4

3.2

3.0

2.8
c:
gc:
0
2.6

Q)
2.4
u
c: 2.2
0
()
2.0
1.8

1.6
1.4

1.2

1.0

0.8
0.6
0 500 1000 1500 2000 2500 3000 3500 4000

Time

Figure 1·24 Concentration profiles for first and second order reaction mechanisms.
= D= first-order reaction; = 0 = second-order reaction.
34 FLUX REACTIONS AND SOLDERABILITY

with it and wets the clean copper. The basic wetting time for clean copper is
indicated by the intercept of the wetting curves with the wetting time axis when
oxidation immersion time is zero. The graph shows this to be -0.5 sec. However,
other experiments indicated that the time shown here was limited by the rise
time of the equipment used and that a time between 0.1 and 0.15 sec was probably
the true value for clean copper with 60% tin-40% lead solder. The increase in
wetting time above this minimum value would then be due to the time taken
with the test flux, for the oxidation to be removed. The ratio of the wetting time
to the oxidation time at any point on the curve would then be a measure of the
reaction rate for dissolution of the oxide in the flux. When these ratios were
measured for all points on the curves, it was found that they all fell on a single
straight line as shown in Figure 1-23. As can be seen, this shows a positive
slope with increasing reaction rate with increasing time, completely the opposite
of what was indicated by the first- or second-order concentration driven mech-
anisms indicated by Figure 1-24.

Autocatalysis and Chain Reaction Equations

As a general rule the rate of a catalytic reaction is proportional to the con-

centration of the catalyst present. An example of this would be the catalysis of
ether by iodine vapour where the equation for the rate of decrease in the amount
of ether in the presence of iodine could be written

-d[ether] [11} k 1 [ci2 ] k 2 [cether]

dt

With the large acceleration factors that catalysts impart, kl is very much bigger
than k2 and becomes the controlling factor in the reaction. When one of the
products of a reaction is itself a catalyst for part of the reaction, the process is
called autocatalysis. The effect of this process on the rate of a reaction, A=> P
say, that obeys the law

would be that if the concentration of A at t = twas Ao - x and of P was Po

- x, where x = 0 at t = 0, then

dx
dt = k[Ao - x] [Po - x]
 FLUX TYPES 35

and the fraction of P produced by time t would be

x (eat - 1)
Po (1 + beat)

where a = (AD + Po) / k and b = Po / AD. This is a steadily increasing function

with time and not one that decreases exponentially.
An alternative possibility would be a chain reaction in which an intermediate
product of the reaction reacted with one of the constituents to create yet more
of the intermediate. This could be represented as

which could proceed via the intermediate reactions

A2 ~2A
A + B2 ~AB + B
B + A2 ~AB + A

with the reaction quenched by the process

B + AB ~B2 + A
and 2A ~A2

This gives eventually a relationship of the form

d[AB] Ka [cA,2] [cB,2]

dt Kb + [CAB] [CB,2]

This also represents a continuously increasing function,

On the basis of the form of the reaction rate curve that is observed in the
surface tension balance experiments (Fig. 1-23), it seems clear that the mech-
anisms that would support a process as indicated here would be either autoca-
talysis or chain reactions. In the first case the reaction could be catalyzed by the
same ions, either copper or hydrogen, that were being generated during oxide
removal. A combination involving the halogen ions from the activator system
would also be possible. Such driven catalytic reactions are not unknown.
A chain reaction system could be such that a secondary reaction involving
copper ions, possibly even involving transitions from valence state Cu + + to
Cu +, assumes the rate determining function. Combinations of these could be
possible together with the possibility that an exothermic reaction could further
accelerate reaction rates.
36 FLUX REACTIONS AND SOLDERABILITY

The possibility that the increase in temperature arising from an exothermic

reaction increased reaction rates to the extent indicated would tend to be dis-
counted, since calorimetric results do not show large exothermic peaks.

Implications of Alternative Mechanisms

The implication of the differences in fluxing mechanism from a simple con-

centration dependence could be wide reaching. Perhaps injections of less cor-
rosive compounds than currently used activators, but which contain the necessary
ions, could replace the corrosive residue inorganic halides currently used. A
source of copper ions, for example, could show a significant activity increase.
In this way very highly activated systems that have no reactive residues could
be formulated. In a similar vein questions involving the use of other ionic species,
possibly other metal systems, certainly deserve consideration and study.
An example one can cite of where this type of study could lead, is the problem
currently looming in the field of solder creams for surface mount applications
where finer and finer detail requires finer solder powders for satisfactory screen
printing performance. This in tum means higher oxide percentages in the powder
requiring, in tum, more flux activity to allow solder reflow during the soldering
operation. Finer printed circuit board detail is usually associated with smaller
printed surface board-component joint gaps making it more and more difficult
to clean residues from increasingly more activated solder creams. This seems to
be a classical vicious circle that only a more enlightened approach to studies on
how flux systems work can break.

REFERENCES

1. Turkdogan, E. T., and S. Zador, J. Iron and Steellnst., March, 1961, pp. 233-
239.
2. Onishi, I., I. Okamoto, and A. Omori, Trans. Japan Welding Inst., 1 (1), 1972,
pp.23-27.
3. Ibid., 2 (1), 1973, pp. 113-119.
4. Ibid., 2 (2), 1973, pp. 97-102.
5. Ibid., 2 (2), 1973, pp. 103-110.
6. Ibid., 3 (1), 1974, pp. 99-103.
7. Ibid., 3 (1), 1974, pp. 105-109.
8. Ibid., 4 (1), 1975, pp. 79-84.
9. Ibid., 4 (1), 1975, pp. 85-90.
10. Ackroyd, M. L., C. A. MacKay, and C. J. Thwaites, Metals Technology, February
1975, pp. 73-85.
11. Latin, A., Trans. Faraday Soc., 34, 1938, pp. 1384-1395.
12. Bailey, G. L. J., and H. C. Watkins, J.lnst. Metals, 19, 1951,J:lp. 57-76.
 REFERENCES 37

13. MacKay, C. A., Proc. ASM Conf. "Microelectronics Packaging Technology," April
1989, Philadelphia.
14. Sherman, K., and C. A. MacKay, Proc. ASM Conf. "Microelectronics Packaging
Technology," April 1989, Philadelphia.
15. Momose, Y., and Y. Tamia, J.I.M., 93, 1970.
16. Snyder, R. W., Applied Spectroscopy, 14(3), 1987, pp. 460-463.
 2
Solder Paste Technology
and Applications
Jennie S. Hwang

The electronics industry, since the invention of the transistor in 1947, has con-
tinued to grow and change through evolutionary and revolutionary developments.
Most recently, the strongest trend in the design and manufacture of electronics
packages is the application of surface mount technology in lieu of conventional
through-hole technology in populating printed circuit boards. As the name im-
plies, surface mount technology involves placing components and devices on
the surface of the circuit board, instead of through the board, by using engineering
and scientific principles. This concept has been adopted in hybrid assemblies
since the 1960s, yet its actual application in printed circuit assemblies was not
fully developed until the early 1980s.
The aggregate merit of surface mount technology is superior performance-
cost ratio for printed circuit board manufacture, in relation to conventional through-
hole process. This is a result of increased circuit density, decreased component
and board size, reduced weight, shorter interconnection and therefore improved
electrical performance, and adaptability to automation.
For surface mounting, the most practical interconnecting material used today
is solder paste because of its virtues of "deformable" viscoelastic form, readiness
for automation, and tacky characteristic. The deformable form renders it appli-
cable in any selected size and shape, and the tacky nature provides the capability
of holding components and devices in position without the need of adhesive
before the permanent metallurgical bonds are formed. The bonds are metallic in
nature, providing electrical, thermal, and mechanical functions. Therefore, per-
formance and quality of solder paste are critical to the integrity of solder joints,
which, in tum, are critical to the overall reliability of the electronics assembly.

38
 SOLDER PASTE TECHNOLOGY AND APPLICATIONS 39

Solder paste, by one definition, is a homogeneous and kinetically stable mix-

ture of solder powder, flux, and vehicle, which is capable of forming metal-
lurgical bonds at a given set of soldering conditions and can be readily adapted
to automated production in making reliable and consistent solder joints.
From the technological point of view, solder paste involves the interplay of
several scientific disciplines. Sciences and technologies used include metallurgy
and particle technology, chemistry and physics, rheology, and formulation tech-
nology as shown in Figure 2-1.1 The three functional components of a solder
paste are solder powder, flux system, and vehicle system. The solder powder is
the only permanent component forming the metallurgical bond; the flux causes
the solder powder and the substrate surfaces to be joined in a clean and metallic
state so that good wetting and metallic continuity between solder and the surfaces
can be formed; the vehicle is a carrier for solder powder and flux to provide a
designated rheology. The final performance of a solder paste can be grouped as
applicability, solderability, residue characteristics, and solder joint integrity.
Figure 2-2 illustrates the correlation between the major components, the final
performance, and the fundamental sciences and technologies. It is indicative that
the solder paste is a genuine example of an interdisciplinary product and therefore
demands an interdisciplinary approach.

Metallurgy
Particle Rheology
Technology

Figure 2-1 Interdisciplinary technology.

40 SOlDER PASTE TECHNOLOGY AND APPLICATIONS

.01
.. r::
J .....
o r::QI
"0
S/lo
5
"
J..~~!.c:
... r::o
Ullr::

Jr
U
.. <II

/lo

Figure 2-2 Solder paste technology.

Under the main thrust of surface mount technology, specific trends and driving
forces prevalent in the electronics industry include increasingly complex and
dense integrated circuitry, interconnection reliability, the chlorofluorocarbon-
ozone issue, and cleanability and cleanliness measurement. These forces call for
increased knowledge and continued new product development in all aspects of
design, material, process, and equipment. For solder paste, finer and finer pitch
paste, aqueous-cleaning paste, no-clean paste, and controlled atmosphere sol-
dering are the new products and new processes to be developed and/or imple-
mented.
With this demand in mind, this chapter is intended to provide an integrated
knowledge of a full spectrum of technologies, techniques, and application of
solder paste with emphasis on the new products, new processes, and new issues,
as previously mentioned.

2.1 CHEMICAL AND PHYSICAL CHARACTERISTICS

In addition to the characteristics of solder powder, which are discussed later,

the performance of a solder paste equally depends on the chemical and physical
characteristics of the flux and vehicle systems. In this regard, the flux and vehicle
have to be treated as one entity because in solder paste the functions of flux and
vehicle are closely related and inseparable. A typical composition of the flux-
vehicle system consists of a range of 5 to 20 ingredients. Each ingredient or
 CHEMICAL AND PHYSICAL CHARACTERISTICS 41

group of ingredients provides a necessary function, such as binding, fluxing,

controlling rheology, suspending, or other targeted functions. The specific char-
acteristics to be considered are:

Boiling point of component

Boiling point of liquid phase
Melting point of component
Melting point of solid phase
Softening point of resin
Vapor pressure of component
Vapor pressure of system as a whole
Chemical functional groups and reactivities
Chemical reactivity versus temperature
Tackiness of resin versus surrounding medium
Environmental stability of component
Environmental stability of the system as a whole
Hygroscopicity of component
Hygroscopicity of system as a whole
Thermal stability of component
Thermal stability of system as a whole
Cohesive force of system
Viscosity of system
Metal load acceptability
Surface tension of liquid phase

Among the performance parameters of solder paste listed in Table 2-1, the
relationship between the performance parameters and the above-listed charac-
teristics is not always straightforward because of the potential interaction among
ingredients and the interrelationships among the characteristics. For example,
the boiling point of the component and the liquid phase, chemical functional
groups and reactivities, tackiness of resin, and thermal stability of component
and system all contribute to the residue properties in physical form (hard or soft),
level of corrosivity, and deanability. The tack and open times are highly de-
pendent on the boiling point and vapor pressure of component and system,
softening point and tackiness of resin, chemical functional groups and reactiv-
ities, environmental stability, and hygroscopicity of component and system. The
rheology, wetting ability, and solder balling rely on the delicate balance of all
the above characteristics. The cohesiveness of the flux-vehicle system is partic-
ularly important to the printing and dispensing efficiency. The metal load ac-
ceptability is most associated with paste appearance and the relationship between
viscosity and printability-dispensability, as well as with the tack and open times.
42 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

Table 2-1 Solder Paste Performance Parameters

Stage Parameter
Before soldering Physical appearance
Stability and shelf life
Cold slump
Dispensability through fine needle
Screen and stencil printability
Tack time
Open time
Adhesion
Quality and consistency
During Compatibility with surfaces to be joined
soldering Flow property before becoming molten
Flow property at and before molten
Wettability
Dewetting phenomenon
Solder balling phenomenon
Bridging phenomenon
Wicking phenomenon
Leaching phenomenon
Post soldering Residue quantity
Residue cleanability
Residue corrosivity
Electrom igration
Joint appearance
Joint voids
Joi nt strength
Joint microstructure
Joint integrity versus mechanical fatigue
Joint integrity versus thermal fatigue
Joint integrity versus thermal expansion coefficient
differential
Joint integrity versus intrinsic thermal expansion anisotropy
Joint integrity versus creep
Joint integrity versus corrosion-enhanced fatigue
Joint integrity versus interfacial intermetallics
Joint integrity versus bulk intermetallics

2.2 FLUXING AND FLUXES

The function of fluxes in solder paste is to chemically clean the substrate surfaces
to be joined, to clean the surface of solder powder, and to maintain the cleanliness
of both substrates and solder powder during soldering so that a metallic continuity
at the interface and a complete coalescence of solder powder can be achieved.
 FLUXING AND FLUXES 43

Broadly, chemicals that contain functional groups that are active or become
active at soldering temperature can be considered as fluxes. Organic chemicals
containing active carboxylic groups (-COOH), amine groups (- NH 2 , - NHR,
- NR2), and organic salts, such as hydrochloride or hydrobromide salts of amines,
are common fluxing agents.
In addition to functional groups, the fluxing strength depends on the molecular
structure, the surrounding medium, and the temperature in accordance with the
organic chemistry principles. 2 For practical applications, this leads to the "flux
activation temperature" and "flux activation time" for fluxes individually and
fluxes in solder paste. It should be noted that activation temperature and time
for the flux alone may be different when the same flux is incorporated in the
solder paste composition. Some fluxes are active at ambient room temperature
and others need elevated temperature to be activated. The normal activation
temperature can range from 20 to 200°C, and the activation time can range from
instantaneous to several minutes. Kinetically, flux activation temperature and
activation time are always interrelated.
Rosin-based flux has been established as a reliable chemical and is widely
used in the industry. The beauty of rosin is that it can provide multiple functions
as a binder, a tackifier, a protecting layer, and a flux in the same system. The
fact that its fluxing function is activated at an elevated temperature, but not at
ambient room temperature, makes it a most desirable flux ingredient.
Rosin, a natural resin obtained from pine trees, is classified as gum, wood,
and tall oil rosin, depending on its origin and processing. The unmodified rosin
acids consist of abietic acid, isopimaric acid, neoabietic acid, pimaric acid,
dihydroabietic acid, and dehydroabietic acid, with the abietic acid as a major
component. The abietic acid, which is structurally characterized by one carbox-
ylic group, a condensed three-ring structure, and a conjugated double bond, can
be readily modified through different processes. 1
With numerous grades of rosin available, rosin can be conveniently charac-
terized by the softening point, acid number, saponification number, and color.
Federal Specification LLL-R-626D calls for the following specifications for Class
A Type I Grade WW or WG:

Characteristic Specification
Insoluble matter in toluene (maximum percent 0.05
by weight)
Softening point (DOC, minimum) 70
Acid number (minimum) 160
Saponification number (minimum) 166

For soldering, rosin is normally accompanied by an active chemical system,

which is called activator, to enhance the flux strength. Customarily, several flux
44 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

types have been used in the industry. They are Type R, Type RMA, Type RA,
Type OA, and Type SA with the following definitions and properties:

Type R: Rosin without the presence of activator; mildest flux

Type RMA: Rosin mildly activated; containing both rosin and activator; no
halides
Type RA: Rosin activated; higher flux strength than RMA; containing halides
Type OA: Organic acid; high fluxing strength and highly corrosive; aqueous-
cleaning
Type SA: Synthetic activator; equivalent fluxing activity to organic acid; chlo-
rofluorocarbon and alcohol cleaning

The flux activity and its potential corrosivity can be measured by water extract
resistivity, copper mirror test, halide test as specified in Federal Specification
QQ-S-57IE, and surface insulation test as specified in Military Specification
MIL-P-5511OD and IPC specification IPC-TM-650 method 2.6.3.

2.3 SOLDER ALLOYS

An alloy is a material formed from a combination of two or more elements by

means of dissolving and mixing them in the liquid state and then freezing them
into the solid state through a proper solidification process. Solder generally refers
WEIGHT PERCENT LEAD
10 20 30 40 50 60 70 80859095
350
3270
-- ~

- V
300
L V
~,
V
/
250
232 0
V L+ (Pb)
UJ
~ 200
~ ./ (Pb)
+ Sf"" ~ la3° II
'4a: 1~.45 2.5) 26.1138.1) 71"""

'" \
150 (5n' (all
UJ
a. ~
~ 100
I-- (Sn + (Pb)
50
\
10 20 30 40 50 60 70 80 90 100
ATOMIC PERCENT LEAD Pb
Figure 2-3 Phase diagram of lead-tin.
 Table 2-2 Physical Properties of Solder Alloys
Ultimate 0.2% 0.01% Uniform
Liquid Solidus Tensile Strength Yield Strength Yield Strength Elongation
Alloy Composition eC) (OC) (ksi) (ksi) (ksi) (%)
42% Sn-58% Bi 138 138 9.71 6.03 3.73 1.3
43% Sn-43% Pb-14% Bi 163 144 5.60 3.60 2.77 2.5
30% In-70% Sn 175 117 4.67 2.54 1.50 2.6
60% In-40% Sn 122 113 1.10 0.67 0.53 5.5
5% In-95% Pb 314 292 3.66 2.01 1.79 33.0
30% In-70% Pb 253 240 4.83 3.58 3.08 15.1
60% In-40% Pb 185 174 4.29 2.89 2.06 10.7
80% Sn-20% Pb 199 183 6.27 4.30 2.85 0.82
.;.
63% Sn-37% Pb 183 183 5.13 2.34 1.91 1.38
en
60% Sn-40% Pb 190 183 4.06 2.06 2.19 5.3
25% Sn-75% Pb 266 183 3.35 2.06 1.94 8.4
10% Sn-90% Pb 302 268 3.53 2.02 1.98 18.3
5% Sn-95% Pb 312 308 3.37 1.93 1.83 26.0
15% Sn-82.5% Pb-2.5% Ag 280 275 3.85 2.40 1.94 12.8
10% Sn-88% Pb-2% Ag 290 268 3.94 2.25 2.02 15.9
5% Sn-93.5% Pb-l.5% Ag 301 296 6.75 3.85 2.40 1.09
1 % Sn-97.5% Pb-l.5 Ag 309 309 5.58 4.34 3.36 1.15
96.5% Sn-3.5% Ag 221 221 8.36 7.08 5.39 0.69
95% Sn-5% Ag 240 221 8.09 5.86 3.95 0.84
95% Sn-5% Sb 240 235 8.15 5.53 3.47 1.06
85% Sn-lO% Pb-5% Sb 230 188 6.45 3.63 2.62 1.40
5% Sn-85% Pb-l0% Sb 255 245 5.57 3.67 2.26 3.50
95% Pb 5% Sb 295 252 3.72 2.45 1.98 13.70
46 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

to fusible alloys with a liquidus temperature below 4()()OC (750°F). The elements
constituting solder alloys include tin (Sn), lead (Pb), silver (Ag), bismuth (Bi),
indium (In), antimony (Sb), and cadmium (Cd). The commonly used solder
alloys are binary systems: tin-lead, tin-silver, tin-antimony, tin-indium, tin-bis-
muth, lead-indium; and ternary systems: tin-lead-silver, tin-lead-bismuth, tin-
lead-indium, and tin-lead-antimony.
Regarding an alloy, a phase diagram, represented by composition versus
temperature, is a useful tool for anticipating the metallurgical behavior of the
alloy over a range of temperatures. Depending on the mutual solubility and
miscibility of the elements involved in both the solid and liquid states, and
on other metallurgical interactions, a phase diagram with different features will
result. For example, Figure 2-3 shows the phase diagram of a tin-lead binary
alloy showing the solid solution regions (a or 13), the combined liquid solu-
tion and solid solution region (L + a or L + 13), a solid solution mixture
region (a + 13), and a liquid region (L). The presence of solid solution re-
gions indicates that a tin-lead alloy will respond to designed heat treatment or
to unpurposed thermal excursion.
Some mechanical properties of selected alloy composition are summarized in
Table 2-2.

2.4 SOLDER POWDER

The technique of producing solder powders is largely based on an atomi-

zation process; that is, the disintegration of molten alloy into particulates by
means of an energy-rich stream of inert gas or a centrifugal rotating disc.
The process involves a number of operating parameters. To produce good
quality powders with reasonable yield requires the understanding of each pa-
rameter in relation to the powder properties and also requires a balance among
the operating parameters. The major properties of solder powder to be con-
sidered are:

Melting range Particle size distribution

Flow rate Oxide content
Apparent density Dross content
Particle morphology Elemental impurities
Particle size Solderability
The melting, or mushy. range is the temperature difference between solidus and
liquidus temperatures. For eutectic compositions where the solidus temperature
coincides with liquidus temperature, there is no melting range. The melting range
of tin-lead solder compositions can range from ooe for eutectic composition 63%
 SOlDER POWDER 47

Sn-37% Pb to 83°C for 25% Sn-75% Pb. The melting range can readily be
determined by using a differential scanning calorimeter (DSC). However, con-
sidering that a phase diagram is derived at equilibrium conditions and that most
processes and techniques used are under nonequilibrium conditions, the ther-
mogram obtained should be referenced to the corresponding phase diagram for
thermodynamic phase transitions and liquidus and solidus temperatures. The
thermograms as shown in Figures 2-4 and 2-5 can serve as thermal fingerprints
for a specific alloy.
The flow rate of solder powder is measured by a Hall flowmeter, in accor-
dance with ASTM Method B-2I3. The flow rate can have a range of 8 to 14
sec/50 g, depending on the particle size and distribution, as well as on the
surface condition of the powder. Generally, very fine powder size may tend
to agglomerate and, therefore, hinder the flow efficiency. A "free flowing"
condition is a minimum criteria for an acceptable powder. Apparent density
is measured according to ASTM Method B-212; its value is significantly lower
than that of bulk solder and falls in the range of 4.5 to 7.0 g/cc, varying with
alloy composition.
The desirable particle morphology of solder powder includes a spherical shape
and a smooth surface without satellites as shown in Figures 2-6 and 2-7. The
particle size and distribution should be tailored to fit the needs of applications

Sa.pI IItI 838

Size, 45.0000 _9 Kcell. 1.1700
Method. CoMPNT CHARACTER I ZA Tl 01>1
CD_ant, NZ 55CCIMIN 5. 00 "C/MtN 00"C-400"C
10

,.
~

.! -10

•
.....2.,
a -20
!

-3D

-40
IDo 260 360 4 o
Tamparatur.. COC) ~uPont 9000

Figure 2-4 DSC thermogram of 63% Sn-37% Pb solder powder.

48 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

S""'pla, 10/88/2
Size, ~5. 0000"'9 Keel II 1. 1700
Nathod, COMPNT CHARACTERIZATION
COII.ent, N2 55CC/N I N 5. CO·C/NIN OO·C-~OO·C
4

,."
~ C
;0
0

..
u.

0
II
-2
X

-4

-B+-----~--~r--------~~~------~~O--------~40
("el OuPont 9000

Figure 2-5 DSC thermogram of 105%-88% Pb-2% Ag solder powder.

with respect to rheology and solder paste disposition land pattern. The required
particle size for different groups of applications is discussed later. For particles
coarser than 44 f.Lm (i.e., + 325 mesh), the conventional sieve analysis according
to ASTM Method B-214 is adequate for separation and specification. Below 325
mesh, a special technique is needed to separate different ranges of powder particle
size.
The oxide content of solder powder can be determined by an oxygen de-
tection apparatus. Truly oxide-free powder is hardly attainable. Although ex-
cessive oxide content is detrimental to solderability performance, with tk' state-
of-the-art technology, powder containing oxygen in the range of 50 to 200
ppm is considered satisfactory. The dross content, a measurement of cleanli-
ness of the powder, is another indicator of the quality of powder; a content
below 0.5 wt %, as determined by wet chemistry technique, is a satisfactory
level.
The limits of elemental impurities Jor tin (for nonantimony solder), bismuth,
copper, iron, zinc, aluminum, arsenic, and cadmium are specified under Federal
Specification QQ-S-571E as shown in Table 2-3. Their measurements can be
carried out by means of wet chemistry techniques in conjunction with spectros-
copy.
49
,{"~/ ~_.' r ,,~,IJIIt., . , ';"', .,1
.'.
.!.
::J
o
",'. ,. !~ . . . . . . ~

~,.'-~ , ';'~1.''''
.: - . I(').
Q;
1g,

• "
~"",!·7\!> .
,.
r ,,(">.J
' :'
, :"
~ ...
::' ' :' .
'
,
" <~ I
Q;
""0'"
L

.~'",..
".
• . ~ . ..,<.",~.~.",..
~ ~~. o
o
~
,.,:-' .. ~. --+
Lt')

,,' " ."{. J Lt')

,i ...".- " ,::... -1 (S) 1 N
M

~~; ",.. ~ ".

, ' I
.... '
-5:'
" ,~ . " In,i- '

,. "" ~ .,
?!-
.. '
"C
M

v-.
.~. ?!-
M
..r>
'0
.

"
.. -§.
~., =:>~ ~
eu

~ ~,,~,
'E
~--:-
w u
) '~ Vl,E

"". ",
.... 1-'
N~

,~:. ~.~., ~
41<
.. ,+-

Si,o
.. _. ~' .~~- u::: ~

.~
" , . • (S)
.2:i
- ,,' ¥--' .:

50
 PASTE FORMULATION 51

2.5 PASTE FORMULATION

Based on the flux activity and residue cleaning requirements, today's solder paste
for electronics packaging may be classified as:

R-grade Water saponifiable

RMA-grade Water soluble
RA-grade No-clean

The first three grades are rosin based and solvent cleaning and have been used
in the industry for many years. The water-cleaning and no-clean pastes are
discussed near the end of the chapter.
Solder paste is a homogeneous and kinetically stable mixture of solder alloy
powder, flux, and vehicle that is capable of forming metallurgical bonds at a set
of soldering conditions and can be readily adapted to automated production in
making reliable and consistent solder joints. In terms of functionality, a solder
paste can be considered as being composed of three major components. They
are solder alloy powder, vehicle system, and flux system. The vehicle is primarily
a carrier for alloy powder, as a compatible matrix for flux system, and as a basis
for desirable rheology. The flux works to render the alloy powder and substrates
to be joined in a clean and metallic state so that a metallic continuity and good
wetting can be obtained. Both vehicle and flux are fugitive in nature at the
completion of soldering; nevertheless, they are crucial to the formation of reli-
able, permanent bonds. On a permanent basis, the alloy powder part is the only
functional component in forming a metallurgical bond.
As an example, a typical RMA solder paste formula may contain 12 ingre-
dients, providing various designated functions:

Ingredient Function
Rosin 1 Rosin system for designated softening point, acid
Rosin 2 number, thermal stability, fluxing, tackiness
Rosin 3
Nonhalogen activator 4 Activator system for accomplishing fluxing action
Nonhalogen activator 5 over a wide range of temperature, rheology, open
Nonhalogen activator 6 time
Solvent 7 Solvent system to accommodate solubility,
Solvent 8 rheology, temperature compatibility, chemical
Solvent 9 compatibility, tack time
Binder 10 Providing compatible viscosity, rheology, tackiness
Fluxing modifier 11 Stabilizing and modifying flux activity and strength
Rheology modifier 12 Contributing to targeted rheology
 Table 2-3 Solder Paste Trace Elements
Total of
all
Iron, Zinc, Others,
Tin Lead Antimony Bismuth, Silver Copper, max. max. Aluminum, Arsenic, Cadmium, max.
Composition % % % max. % % max. % % % max. % max. % max. % %
5n96 Remainder 0.10, 3.6 0.20 0.005 0.05 0.005
max. to
4.4
5n70 69.5 Remainder 0.50 0.25 0.1 0.08 0.02 0.005 0.005 0.03 0.08
to
71.5
5n63 62.5 Remainder 0.50 0.25 0.1 0.08 0.02 0.005 0.005 0.03 0.08
to
63.5
<n
5n62 61.5 Remainder 0.50 0.25 1.75 0.08 0.02 0.005 0.005 0.03 0.08
'"
to to
62.5 2.25
5n60 59.5 Remainder 0.50 0.25 0.1 0.08 0.02 0.005 0.005 0.03 0.08
to
61.5
5n50 49.5 Remainder 0.50 0.25 0.1 0.08 0.02 0.005 0.005 0.025 0.08
to
51.5
5n40 39.5 Remainder 0.50 0.25 0.1 0.08 0.02 0.005 0.005 0.02 0.08
to
41.5
5n35 34.5 Remainder 1.6 0.25 0.1 0.08 0.02 0.005 0.005 0.02 0.08
to to
36.5 2.0
5n30 29.5 Remainder 1.4 0.25 0.1 0.08 0.02 0.005 0.005 0.02 0.08
to to
31.5 1.8
 Sn20 19.5 Remainder 0.80 0.25 0.1 0.08 0.02 0.005 0.005 0.02 0.08
to to
21.5 1.2
SnlO 9.0 Remainder 0.20, 0.03 0.17 0.08 0.005 0.005 0.02 0.10
to max to
11.0 2.4
Sn5 4.5 Remainder 0.50, 0.25 0.1 0.08 0.02 0.005 0.005 0.02 0.08
to max
5.5
Sb5 94.000 0.20, 4.0 0.1 0.08 0.08 0.03 0.03 0.05 0.03 0.03
min. max to
6.0
Pb80 Remainder 78.5 0.20 0.25 0.1 0.08 0.02 0.005 0.005 0.02 0.08
to to
en 80.5 0.50
w
Pb70 Remainder 68.5 0.20 0.25 0.1 0.08 0.02 0.005 0.005 0.02 0.08
to to
70.5 0.50

Pb65 Remainder 63.5 0.20 0.25 0.1 0.08 0.02 0.005 0.005 0.02 0.08
to to
65.5 0.50
Agl.5 0.75 Remainder 0.40, 0.25 1.3 0.30 0.02 0.005 0.005 0.02 0.08
to max to
1.25 1.7
Ag2.5 0.25, Remainder 0.40, 0.25 2.3 0.30 0.02 0.005 0.005 0.02 0.03
max max to
2.7
Ag5.5 0.25, Remainder 0.40, 0.25 5.0 0.30 0.02 0.005 0.005 0.02 0.03
max max to
6.0
54 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

The formula may appear to be somewhat straightforward. However, designing

a viable product capable of delivering all desired performance characteristics as
shown in Table 2-3 is complex and requires understanding of the technologies
and skills in formulation. A product design can be accomplished by following
design and manufacturing steps:

Define performance objective

Use fundamental technologies
Select raw materials
Understand and anticipate potential synergistic or antagonistic interactions
Balance performance parameters
Fine tune the formula to meet designated specification
Develop production process
Produce consistent product

Among the performance parameters, some are trade-offs. For example, high
metal content is beneficial to a solder joint with respect to solder volume and
void and to residue yet may make the paste more prone to crusting and drying
and difficult to apply. High viscosity paste may improve flow control in resisting
to rising temperature, yet the paste is difficult to apply. Using highly active
fluxing chemicals may improve solderability in some cases, yet its use may leave
more corrosive residue and render the paste less stable to shelf storage and to
external forces. In such a case, improving the solderability by selecting proper
ingredients without the use of highly active fluxing chemicals requires skill. It
should be noted that increasing flux content does not always improve solderability
in terms of wetting and/or solder balling elimination.
While the product is being designed, developing a practical and reproducible
process for making the paste with consistent characteristics is equally important.
It is not an exaggeration, but an indication of the importance of the role of
process, to state that identical composition formulas can produce different prod-
ucts when the process is allowed to vary.

2.6 PASTE RHEOLOGY

Rheology of solder paste l •3 is an important characteristic with respect to appli-

cation, since the flow and deformation behavior of solder paste directly affect
the quality of paste deposition on the intended solder pads. Therefore, the cor-
relation of paste rheology with the application and its specific parameters is
crucial to the control of rheological quality when manufacturing paste, as well
as to the setting and control of application parameters when using paste.
In many rheological aspects, there are similarities between coating products
where fine oxide particles are dispersed in an organic carrier phase and solder
 PASTE RHEOLOGY 55

pastes. Yet, in solder pastes, different inherent factors are involved, making the
solder paste rheology more of a challenge. Factors such as the following are
influential:

Drastic difference in density between organic matrix and metallic particles

Suspended particles of a metallic nature
Relatively large particle size
Spherical particles
High load of particles

The drastic difference in density between the organic matrix (0.7 to 1.2 g/cm3)
and the metallic particles (7 to 12 g/cm3) facilitates the settling of particles, and
the large size (up to 75 J.Lm) of particles reduces the interactions between the
particle and matrix, which further facilitates settling. The metallic nature of
particles renders the mixing or dispersion technique more delicate. A high load
of particles, commonly more than 40 vol %, contributes to the complexity of
the rheological behavior.
The primary driving forces underlying the rheology of solder paste include
both kinetic and thermodynamic contribution. In principle, these cover thermal
motion with magnitude determined by the mean free path of molecules or particles
and by temperature, in conjunction with intermolecular or interparticle forces.
The forces can be viewed as electrostatic repulsion and attraction between par-
ticles and the affinity between the chemical matrix and the particle surface. The
particles will not aggregate when the affinity of particle surface for the chemical
matrix exceeds the attractive forces between particles or when steric effects
through absorption hinder the aggregation. Therefore, the rheology of solder
paste may be affected by the following factors:

The composition, shape, and size of suspended particles

The chemical composition of suspending matrix
The relative concentration of effective ingredients in the matrix
The structure of ingredients in the matrix
The interactions between matrix and suspended particles, either of a physical
or chemical nature, including wetting and solvation
The volume fraction occupied by the suspended particles; usually, the higher
amount of particles, the more deviation from viscous flow
The internal structure and its response to external forces
The interactions among particles and resulting aglomeration and flocculants
Temperature

It is difficult to predict the rheology of such a system because of the lack of

knowledge of the exact structure and the nature of forces exerted by molecules
or particles. However, rheological behavior can be characterized, and the char-
56 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

acterization can then be correlated to performance, as discussed in the next

section.
The ideal rheology of solder paste for both dispensing and printing applications
from an application point of view is expected to deliver the following properties:

Specified shelf stability

No separation or settling
No dripping
No stringiness but with enough tackiness to hold components
No slumping or sagging but with capability of leveling
Perfect transfer through printing pattern
Clean release from the screen
Clean break off through fine needles
Nearly instant recovery after transfer
Temperature insensitive

In practice, true instant recovery after paste transfer and temperature insensitivity
is difficult to attain. It is thus obvious that trade-off and balance are essential to
the design of pastes having a rheology able to accommodate the application
processing parameters.

2.7 RHEOLOGY BEHAVIOR CHARACTERIZATION

Solder paste is neither a purely viscous nor a purely elastic material; it exhibits
viscoelastic properties. Therefore, solder paste is characterized by its flow be-
havior as well as its elastic properties.
For viscous materials, Newton's law defines

T = 1]1
where 1 is the shear stress, 1] is the shear rate, and T is the apparent viscosity.
Viscous flow described by this relationship is called newtonian flow, as shown
in Figure 2-8.

T TJ t - - - - - - -

Figure 2·8 Newtonian flow.

 RHEOLOGY BEHAVIOR CHARACTERIZATION 57

T 77L-----
Figure 2-9 Plastic flow.

In addition to idealized newtonian flow, nonnewtonian types are exemplified

by plastic, pseudoplastic, dilatant, thixotropic, and rheopectic flow. As illustrated
in Figures 2-9 through 2-l3, the viscosity of plastic or pseudoplastic material
decreases with increasing shear rate, due primarily to the alignment of molecules
or particles along the direction of shearing. The dilatant material displays in-
creased viscosity as shear rate increases. It is believed that such flow resistance
is attributed to the formation of a close packed structure. It often occurs in
systems containing very small solid particles suspended in a highly defluocculated
manner. The plasticity of materials is characterized by the presence of a yield
value, which is the threshold shear stress required to overcome the forces of
internal structure to start flow. In addition to shear thinning effect, thixotropic
material exhibiting a time-dependent reestablishment in viscosity results in a
hysteresis loop in the ramping up and down cycle. Rheopectic flow is not well
understood, but it is postulated that a perfect structural orientation is associated
with this phenomenon, assuming in-site chemical reaction and evaporation are
absent.
Parallel to Newton's law for viscous fluids is Hooke's Law for solids. Hooke's
law relates shear stress T and shear strain -y in the following equation:

T = G-y

where G is the shear modulus and -y is the strain. Under stress, the deformation
of most materials differs substantially from the idealized case described by
Hook's law.

T TJ

Figure 2-10 Pseudoplastic flow.

58 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

T 1]

r
Figure 2·11 Dilatant flow.

Flow behavior can be characterized by several readily measured parameters:

shear stress versus shear rate, yield point, viscosity versus shear rate, viscosity
versus time, and shear stress versus time.
The relationship of shear stress to shear rate indicates the type of flow as
shown in Figures 2.8 through 2·13. There is more than one way to determine'
the yield point. For convenience, the yield point may be estimated by the ex-
trapolation of the flow curve to intercept the shear stress axis or may be taken
as the shoulder point (inflection) of the curve, if present. These techniques are
appropriate only for comparison purposes. When using these techniques to derive
yield point, it is more reliable to measure flow at low shear rate conditions «50
sec-I). The relationship of viscosity to shear rate is equivalent to that of shear
stress to shear rate. However, it provides direct readings on viscosity change in
response to shear rate change.
Figure 2-14 shows the flow profiles of a dispensing solder paste, indicating
very slight shear rate dependency. The shear rate ramping up and down curves
indicate a very small amount of thixotropy. It also shows a very low yield
point.
In comparison, Figure 2-15 is the flow profile of a printing paste, showing a
relatively large yield point, moderate thixotropy, and a high shear rate depen-

T 1]

r
Figure 2·12 Thixotropic flow.
 RHEOLOGY BEHAVIOR CHARACTERIZATION 59

T 7J

Figure 2-13 Rheopectic flow.

dency. The viscosity changes with time, after a constant shear rate is applied to
a dispensing paste, as is shown in Figure 2-16. The stress relaxation curve of a
printing paste also reflects the characteristics of a solder paste such as the mag-
nitude of yield value, leveling viscosity, and the internal structure change of the
paste.
At a constant temperature, the dynamic mechanical measurements that char-
acterize the elastic properties include storage modulus, loss modulus, and com-
plex viscosity in response to operating parameters of frequency, strain amplitude,
and time.
Under dynamic measurements, the stress is applied sinusoidally at a given
frequency, and the strain also alternates sinusoidally but is out of phase with the
stress at a given phase angle. The theoretical and mathematical treatment is
covered in the literature. 4

Tau = , (I)

Substance:

SUlple Ho.:
6

1-----+----11------. - - - ------

.~~==~==:J=~~
12 1.8
________
24 36 J) [lId

Figure 2-14 Flow curve of a dispensing solder paste.

60 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

Tau [Pal Tau = r (D)

4488r----..,----.------,-----.-----,

8illlple "0.:
3
neas. systetn:
n5,PKH.8

8'-----'---~
8 £, 12 18 38 D [tlsl

Figure 2-15 Flow curve of a printing solder paste.

Eta [Pasl Eta : , (t)

118

88

8i111ple "0.:

66 \ 2

""-- ~
neas. systell:
n5, PK1-1. 8

22

8
8 1 2 3 4

Figure 2-16 Viscosity versus time curve of a dispensing solder paste.

 RHEOLOGY BEHAVIOR CHARACTERIZATION 61

The elastic characteristics of paste may be defined by analyzing the following

features;

Magnitude of the storage modulus, loss modulus, and complex viscosity

Relative magnitude of loss and storage modulus
Relative changes of loss and storage modulus in relation to frequency change,
strain amplitude, or time
Relative rate of change in complex viscosity versus frequency change, strain
amplitude, or time

The storage modulus (G'), loss modulus (Gil), and complex viscosity (TJ*)
can be monitored in relation to frequency sweep under constant strain. These
three properties can also be characterized in relation to strain sweep under
constant frequency. Figure 2-17 is a measurement of storage modulus, loss
modulus, and complex viscosity of a printing paste in relation to frequency
under a constant strain of 5 percent. The relative weight of viscous and elastic
portions in solder paste can be directly measured in loss tangent value, as shown
in Figure 2-18.
Combining flow measurement and elastic measurement, a solder paste can be

FREQUENCY SWEEP
I. DEli

DYNAMIC
STRAIN - S %

~
. ~
0

.. ..
0
d

~
.. • •
•
<
i> ...
l-

I.DE2
I. DE-I FREQ radio I.DE3

Figure 2-17 Measurement of storage modulus (G'), loss modulus (Gil), and complex vis-
cosity (1]*) in relation to frequency for a printing solder paste.
62 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

tan delta - f (w [I/s1 I

10

Substance:
PASTE-3

Sample No.:
3
+
+ Meas. system:
M50sc,PK1-1.0
+ + +
Temperature:
+
25
+ +
+ +
+
+ + + +
+ +
+ +

w [1/s1
.1 10 100

Figure 2-18 Loss tangent of a printing solder paste in relation to frequency.

adequately characterized. The characterization of the paste has to be correlated

with the processing parameters as well as with the paste performance under a
specific set of conditions.
Under two main paste application methods, fine dot dispensing and pattern
printing, a paste possessing a low yield point and slight plastic behavior is found
most suitable for dispensing applications. This is consistent with the formulation
requirement that a dispensing paste normally has a lower upper limit in volume
fraction of metal load as well as a lower viscosity range, in comparison with
pastes designed for stencil or screen printing applications.
Low yield value physically signifies that paste will start to flow under
low shear stress. Too high a yield point may result in poor leveling, which
leaves an uneven surface, and too low a yield point may cause running and
slumping.
Printing paste normally demands a higher viscosity range. A moderate yield
point and moderate thixotropy are found to be suitable for printing mechan-
ics. Too high a degree of thixotropy may cause slumping and sagging.
However, shear thinning is desirable, especially at high shear rate, as en-
countered while a paste is being transferred through the opening during a
printing process.
 VISCOSITY AND MEASUREMENT 63

2.8 VISCOSITY AND MEASUREMENT

Table 2-4 lists the typical viscosity range and metal load for the dispensing and
printing application techniques, although some exceptions may exist.
A single-point viscosity measurement as a criterion of the rheology of solder
paste is prevalent in the industry. It is thus appropriate to discuss some practical
aspects of single-point viscosity measurement. One of the common techniques
adopted is to use the Brookfield viscometer with a helipath spindle TF at 5 rpm
in accordance with Federal Specification QQ-S-571E. Measurement obtained by
this technique is simple and convenient. However, single-point viscosity is valid
only when the following conditions are met:

The rheology of the paste is known.

The paste handling technique immediately prior to measurement is standard-
ized.
The measurement technique of the paste is standardized.
The paste container is specified in size and dimensions.

To illustrate the point, Table 2-5 is a summary of viscosity data measured

using a Brookfield model RTV viscometer with helipath spindle TF at 5 rpm for
two pastes, paste A and paste B. The variables in this table are paste handling
prior to descending spindle, as expressed by the mixing time, and the measuring
time (time to read) (i. e., the time interval from the start of the spindle descending
to obtaining a dial reading). This time interval also indicates the spindle position
in the paste container.
The data clearly indicate the effect of premixing. Longer mixing generates a
lower viscosity number across the full range of reading time, from l to 3 min.
Under the test conditions, after the paste is mixed thoroughly for 2 min, further
mixing does not significantly affect viscosity. The data also indicate that the
viscosity reading increases with increasing time to read, because of increased
friction as the spindle is descending further down.
An equivalent test is carried out for paste A. The comparison between pastes
A and B indicates that paste A has much less variation in response to mixing

Table 2-4 Viscosity Range of Dispensing and Printing Pastes

Viscosity* Metal Load
Solder Paste Type (CPS) (wt %)

Fine dot dispensing 200,000-450,000 Up to 88

Screen printing 450,000-750,000 Up to 92
Stencil printing 650,000-1,400,000 Up to 94
'Brookfield Model RVT, helipath TF spindle, 5 rpm, 3 min mixing and 2 min reading.
64 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

Table 2·5 Single-Point Viscosity of Pastes A and B in Relation to Premixing Time

and Time to Read 1
Viscosity (KCPS), time to read (sec)
Mixing Time 30 45 60 90 120 100 180
Paste B:
0 1440 1550 1700
1 840 950 1100
2 600 660 720 800 900
3 580 640 700 800 880 900
Paste A:
0 200 220 220 280 400 430
1 190 210 220 240 270 290
2 190 200 210 220 230 250
3 170 180 200 200 220 230

time and time to read. Again, as the paste is mixed for 2 min, further mixing
has no significant effect.
The comparative results demonstrate the relative sensitivity of pastes to han-
dling (mixing) and measuring techniques. With these variations in mind, it is
not surprising to see the inconsistency and lack of agreement from laboratory to
laboratory and operator to operator in viscosity data. Nonetheless, with the proper
technique and control, viscosity data can be consistently obtained. In practice,
realizing such variations, as well as being able to control them, is important to
the establishment of specifications and to communication between users and
vendors, which in turn is important to overall quality and production yield.
Single-point viscosity can also be used to develop an indicator for shear
thinning factor by measuring viscosities at two significantly different shear rates
(revolutions per minutes, or rpm), such as viscosities at 1 and 20 rpm. The ratio
of these two viscosities indicates the shear thinning factor. Again, the paste
handling has to be kept consistent in order to obtain meaningful results.

2.9 PRINTING TECHNIQUE

Today, the printing technique is most widely used for transferring solder paste
onto the intended solder pads on the surface mount production floor. Whether
using mesh screen or metal mask stencil, the forces contributing to the paste
transfer process are the relative adhesive between paste and screen interface and
the paste and substrate interface, the push force by the squeegee along the
direction parallel to screen, the pressure exerted by the squeegee along the
direction perpendicular to the screen, the hydraulic pressure created in the paste
when it is compressed between the squeegee and the screen, and the gravity
force.
 PRINTING TECHNIQUE 65

squeegee motion ~
paste
mesh screen
:J snap off distance
---------"""\ substrate

1 mesh screen

substrate

1 WIB"""""'" wet print

'--_ _ _ _ _ _ _....,1 substrate

Figure 2-19 Schematic of printing operation.

During printing, a sequence of motions as illustrated in Figure 2-19 is created;

these are:

The squeegee makes a line contact with the screen and a close contact with
a given amount of paste.
While the squeegee is in motion, the paste moves along in front of the squee-
gee.
The paste is released through the openings onto the substrate behind the
squeegee.
The screen behind the squeegee "instantaneously" snaps back after the squee-
gee passes over.

There are a number of printing variables that can be grouped into squeegee
system, substrate, screen system, paste, and snap-off parameter.

Squeegee system
Blade material type Squeegee size
Blade hardness Squeegee speed
Blade edge angle Downstop pressure
Blade edge flatness External pressure
Blade configuration
66 SOlDER PASTE TECHNOlOGY AND APPLICATIONS

Substrate
Material type Surface flatness
Surface condition Substrate size

Screen systems
Mesh screen Print pattern
Wire material type Screen size
Wire diameter Metal mask stencil
Mesh tension Material type
Mesh count Flatness
Mesh orientation Thickness
Emulsion type Print pattern
Emulsion thickness Stencil size

Paste
Paste consistency Chemical compatibility with the screen and squeegee
Paste rheology
Paste stability Compatible thickness with printing pattern
Paste tackiness
Snap off

After the equipment is selected and the design of assembly is finalized, there
are four major operational parameters to be monitored. They are the squeegee
speed, squeegee downstop, squeegee pressure, and snap-off. In time of paste
performance, these parameters are interrelated; changing one will change the
effects of the others in a nonproportional manner. Nonetheless, with an under-
standing of the general relationship among these parameters and their effects on
printing results, in conjunction with the systematic testing on the specific system
(printer, board, paste), a set of optimized parameters can be established.
For a printing operation, the balance between the two primary mechanisms,
paste transfer and screen release, dominates the printing results. Practically,
these two mechanisms require different sets of parameters. In other words, the
maximization of paste transfer efficiency may hinder the screen release. Figure
2-20 illustrates the relationship between the paste transfer (emptying of screen)
as described by Equation 2-1 and screen release (no screen sticking) as defined
by Equation 2-2.

D2 . (S - B) . Pa
TJ V :::;:; --'-----'--- (2.1)
128· dw . a
 PRINTING TECHNIQUE 67

'IT • a .m . (1' Lsr . (m + dR ) • Pa

1]V:S;;; s (2.2)
256· Lsr a

where 1] is viscosity, V is the squeegee speed, D is the diameter of mesh opening,

d is the wire diameter of mesh, S is the screen size, B is the length of printer
pattern, a is the snap-off distance, Pa is the atmosphere pressure, m is the wire
pitch, dR = (m - dw), and Lsr is the distance from squeegee edge to screen
frame. 5
Current surface mount printed circuit boards mostly consist of a mixture of
standard land pattern (50 mil pitch) and fine pitch land pattern. To accommodate
both patterns, using a stencil with a differential thickness is one approach to
deposit paste on the board with one-time printing. The metal mask stencil with
a given foil thickness is recessed in areas for fine pitch patterns. The recessed
areas with reduced foil thickness provide thinner print as required for fine pitch
patterns. When design requires thicker and thinner print patterns close together
(typically :s;;; 200 mil), a two- or multiple-foil process is needed. The difference
in thickness of paste print is obtained by joining foils together through a lami-
nation technique.
Another important factor to be considered is ambient environment. Solder
pastes, although to different extents, are responsive to temperature as represented
by the viscosity versus temperature data as shown in Figure 2-21. Maintaining
a reasonably constant temperature under which the printing operation is carried

20 ~-----r--~----------'----------'~--------'

-
E
:1;. 10./------1~-----+------=7~=-------1

o
o 0.01 0.02 0.03 0.04

Relative Snap-Off Distance

Figure 2-20 Balance of paste transfer and screen release in relation to snap-off distance.
68 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

2
1.90
1.80
1.70
1.60
1.50
1.40

U5 1.30
0..
1.20
~ +
>-g 1.10

~! 0.90
0
U 0.80
(f)
:> 0.70
0.60
0.50 +
0.40
0.30
+
0.20
0.10
+
0 I I 1
0 10 20 30 40 50

TEMPERATURE (DEGREES CELCIUS)

Figure 2-21 Variation of viscosity with temperature.

out is a recommended practice. The humidity is also a iactor; its effect on paste
performance depends on the properties of the paste. Controlled humidity, again,
is a necessary condition to achieve desired quality.

2.10 DISPENSING TECHNIQUE

For the assemblies requiring solder connection on an irregular surface or on

hard-to-reach areas, dispensing technique becomes a viable approach. The paste
is delivered on the designated solder pads by means of needles with different
length and inner diameter through pneumatic or volume displacement mecha-
nistics. The dispensing needle gauge and corresponding inner diameters 1 are
listed in Table 2-6.
 SOlDERING PRINCIPLE 69

Table 2-6 Dispensing Needle Gauges and Inner

Diameters
Inner Diameter
Needle Gauge in. mm
15 0.054 1.37
18 0.033 0.84
20 0.023 0.58
21 0.020 0.51
22 0.016 0.41
23 0.013 0.33
25 0.010 0.25
30 0.006 0.15

Each increment in gauge number can affect paste dispensability drastically. The
following is a list of some practical points:

In the dispensing system, set up the shortest distance possible that the paste
needs to pass through
Use a size of paste syringe only large enough to fit production run (avoid
excess size of syringe).
Select the right needle gauge that matches the targeted amount and size of
paste deposit (avoid too small a needle size).
When the system is new, flush it with a compatible lubricant to wet the inner
wall of passage that paste will be in contact with.
When the system will have extended downtime, flush the system thoroughly
with a compatible lubricant.
Select a proper dispensing paste.
For a syringe stored at low temperature, bring it to ambient temperature prior
to starting the dispensing operation.
Make sure no foreign particles fall in the system.
Keep the applied pressure at a minimum level as needed, usually 15 to 40
psi.
Maintain a correct and consistent gap between the needle and the substrate
because the gap affects the size and shape of the paste deposit.
Do not leave paste in the system for an extended downtime.
Check the chemical compatibility of the paste with the materials used in the
dispensing system.

2.11 SOLDERING PRINCIPLE

Soldering is the process of making solder joints by applying heat to the solder
material and to the substrates to be joined at a temperature normally below
70 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

400°C. When the solder material is a solder paste, a series of reactions and
interactions in sequence or in parallel are expected to occur in response to the
applied heat. These may be represented in different stages as follows, where
temperature increases as arrows indicate:

Preheat the ~ Volatile ~ Volatile ~ Chemical

substrates and evaporation evaporation pyrolysis
solder paste Flux Fluxing
concentration substrates
Flux Fluxing solder
activation powder
(I) (II) (III) (IV)

Solder melting ~ Flux or vehicle breaking ~ Solder

Solder wetting substrate out of molten solder solidification
Flux or vehicle protecting Chemical pyrolysis
substrates and breaking
out of molten solder
Chemical pyrolysis
(V) (VI) (VII)

The extent of each of the above phenomena in a given stage and between stages
varies with the specific makeup of the solder paste. For example, paste contain-
ing more low boiling volatiles may need an extended stage II at moderate temp-
erature (100 to 130°C) in order to mildly expel the volatiles without causing
violent paste spattering; the pyrolysis process in stages IV through VI can be
controlled through paste makeup and by the temperature profile (temperature
versus time) so that excessive pyrolysis will not occur and cleaning difficulty
will be avoided. For no-clean paste, stages II through VI may demand a
different temperature profile to achieve minimum residue. With these expected
phenomena in mind, the optimum temperature profile and compatible solder
paste can be established.

2.12 SOLDERABILITY

The solderability of solder paste is the ability to achieve good wetting of molten
solder on the surface of substrates without forming undesirable solder balls
outside the main solder pool so that sound solder joints are formed. The substrates
to be dealt with in soldering field include:
 SOLDERABILITY 71

Gold (Au) Platinum-gold (Pt-Au)

Silver (Ag) Tin-lead (Sn-Pb)
Palladium-silver (Pd-Ag) Tin (Sn)
Platinum-silver (Pt-Ag) Copper (Cu)
Palladium-platinum-silver (Pd-Pt-Ag) Nickel (Ni)
Palladium-gold (Pd-Au)

In order to achieve good solderability, the demand on the fluxing activity depends
on the intrinsic wettability by molten solder, the compatibility of the flux chem-
istry with the substrate, and the surface condition of the substrate, which may
change with the time and exposure condition. In terms of intrinsic wettability,
the substrates are ranked in descending order as tin, tin-lead> copper> pal-
ladium-silver, platinum-silver, palladium-platinum-silver> nickel. The surface
of component leads is normally coated with tin/lead composition (100% Sn -
5% Sn-95% Pb) by using either electroplating or molten solder dipping. The
coating generally is more uniform but more porous from the electroplating process
than from dipping, where a thicker and denser coating is obtained.
The tin-lead coating is relatively stable under ambient environment and its
oxides can be easily handled through fluxing. However, its long shelf time and
heat excursion may cause solderability deterioration through other mechanisms,
such as copper-lead intermetallics formation when leads are made of copper,
and lead material corrosion when leads are made of alloy 42 or Kovar. To assure
consistent and good solderability of components, the shelf time and heat exposure
must be minimized.
With respect to solderability measurement, the edge dip test in reference to
ANSIIIPC-S-804A is commonly used for solder pads and the MIL-STD-202
method for the component leads. A more quantitative test, using a wetting balance

unacceptablG melting

Fw
acceptable melting
Figure 2·22 Wetting phenomena of substrate dipping into molten solder.
72 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

(+) 0 Sp~cim~n I
M---tt------::::~--------

L Specimenll
K----tl~~--~~~---------

- - - - - - - --~ (I)
H-- - - ---FeW)

(-) A

Figure 2-23 Wetting force versus time.

method, is being developed by IPC. The wetting balance measures force versus
time. When a specimen is in contact with the molten solder, as shown in Figure
2-22, the forces involved are related as:

Where Fw is the resultant wetting force, Fg is the interfacial force between

specimen and molten solder, FB is the buoyancy force, 6 is the contact angle,
P is the density of molten solder, "g" is the gravitational acceleration constant,
and V is the immersed volume of specimen. Several characteristic parameters
of the wetting curve, as shown in Figure 2-23, are considered as the measurement
criteria. These are its minimum force, the time for zero net force corrected for
buoyancy, the time for two-thirds of the wetting force, the time for 200 ....NI
mm, the maximum force at 2 sec, and theoretical maximum force. Specific
testing conditions are also being developed.

2.13 SOLDERING METHODS

Soldering can be achieved through different heat transfer modes, including con-
duction, convection, infrared radiation, infrared-convection, vapor condensation,
hot gas, resistance, induction, and laser beam. Each solder method has its unique-
ness and beneficial features in cost, performance, and operational efficiency.
Table 2-7 summarizes the strengths and limitations of each method. 1
 CONTROLLED ATMOSPHERE SOLDERING 73

Table 2-7 Reflow Methods-Feature Comparison

Reflow Method Benefits Limitations
Conduction Low equipment capital Planar surface and single-
Rapid temperature side attachment
changeover requirement
Visibility during reflow Limited surface area
Infrared High throughput Mass, geometry dependence
Versatile temperature
profiling and
processing
parameter
Vapor phase Uniform temperature Difficult to change
condensation Geometry temperature
independence Temperature limitation
High throughput Relatively high operation
cost
Hot gas Low cost Temperature control
Fast heating rate Low throughput
Localized heating
Convection High throughput Slow heating
Induction Fast heating rate Applicability to
High temperature nonmagnetic metal parts
capability only
Laser Localized heating with High equipment capital
high intensity Specialized paste
Short reflow time requirement
Superior solder joint Limit in mass soldering
Package crack
prevention

2.14 CONTROLLED ATMOSPHERE SOLDERING 6

The control of heating rate, temperature uniformity across the mass under heating,
and temperature reproducibility, as well as the need for the compatibility of
heating profile with the characteristics of the solder paste, are important to the
ultimate yield of the soldering process and to the properties of the resulting solder
joint, which have been well covered in the literature.l.7.8.9.10 The atmosphere
under which the soldering is performed is, however, equally important and has
not been studied in the soldering field.
The atmosphere, based on its function, can be considered as being of two
types: protective and reactive. A protective atmosphere normally is inert toward
the specific soldering material under the specific conditions, and a reactive at-
 Table 2-8 Composition of Atmospheric Gases
Composition (vol %)
Carbon Carbon
Dioxide Oxygen Monoxide Hydrogen Methane Nitrogen
Atmosphere (CO 2) (0 2 ) (CO) (H 2) (CH 4 ) (N 2 ) Trace
Air 21.0 78.1 0.9
Nitrogen 99.8-100.0 0-0.2
.....
Hydrogen 99.8-100.0 0-0.2
"" Dissociated
methanol 33.3 66.7
Dissociated
ammonia 75.0 25.0
Exothermic gas
(air/gas = 6/1) 5.0 10.0 14.0 1.0 70.0
Endothermic gas
(air/gas = 2.411) 20.0 38.0 0.5 41.5
 CONTROLLED ATMOSPHERE SOLDERING 75

mosphere may involve oxidizing, reducing, or dissolving functions in the subject

system.
Broadly, the chemicals that can provide oxidizing or reducing potential in
relation to the specific metal and metal oxide system and that can generate
significant vapor pressure at an operating temperature are expected to contribute
to the functional nature of the total atmosphere. The following list shows some
commonly used atmospheres, and their corresponding nominal compositions are
summarized in Table 2-S .

Dry air
Nitrogen
Hydrogen
Nitrogen-hydrogen blends at different ratios
Dissociated methanol-nitrogen blends at different ratios
Dissociated ammonia
Exothermic gas
Nitrogen and dopants at different concentrations

Thermal cracking of methanol essentially yields hydrogen and carbon mon-

oxide at high temperature , as represented by the chemical equation:

At low temperature (below SOO°C), side reactions may occur, leading to the
formation of H20, Cf4, CO 2 , and Cs. Figure 2-24 shows the resulting com-
ponents of methanol decomposition over a range of temperature. 11 The com-
position of exothermic and endothermic gases also varies with the air-gas ratio.

80

;i
e..., 60
(/)
H2 O
C
Q)
c 40

E
0
a.. ~
0
0
20

a DC 200 400 600 800

DF 392 752 1112 1472
Figure 2-24 ThermodynamiC equilibrium composition of methanol decomposition.
76 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

Each component of atmosphere gases may function as an oxidant or reducing

agent depending on the temperature and its oxidation-reduction potential relative
to that of the materials involved. Among the components of common atmosphere
gases, oxygen, water vapor, and carbon dioxide normally serve as oxidant to
most metals and metal oxides, and hydrogen and carbon monoxide serve as
reducing agents. The ratio of oxidant content to reducing agent content in relation
to that ratio at equilibrium indicates whether the resulting atmosphere is oxidizing
or reducing.
Since water vapor is oxidizing in nature, its partial pressure in the furnace
contributes to the overall nature of the atmosphere. The partial pressure of water
vapor in an atmosphere gas is conveniently expressed as the dew point, that is,
the temperature at which condensation of water vapor in the air takes place. The
dew point can be measured by a hygrometer or dewpointer by means of a fog
chamber, chilled mirror, or the aluminum oxide technique. The relationship of
dew point to the vapor pressure is shown in Figures 2-25 and 2-26. The relative
humidity (RH) is related to the actual vapor pressure of water (or represented
by dew point) (P w) and the saturated vapor pressure at the prevailing ambient
temperature (P s ) as follows:

During soldering, the reactions and interactions of chemicals in solder paste

and those between chemicals and metal surface can be quite complex. In simple

100

80

0
0
60

~
~
40
"''-iii
(1)
0.
E 20
~

0
0 200 400 600 800
Vapor Pressure (mm Hg)

Figure 2-25 Relationship of vapor pressure and dew point from 0 to 100°C.
 CONTROLLED ATMOSPHERE SOLDERING 77

0
-1 0
- 20
[)
- 30

(
0

f-
- 40 - 50
-50 - 60
0.1 00 0.200 0.300
- 60
0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 5.00
Vapor Press ure (mm Hg)

Figure 2-26 Relationship of vapor pressure and dew point from -60 to O°C.

tenns, the mechanisms may include evaporation, pyrolysis, oxidation, and re-
duction. The generalized oxidation and reduction reactions can be expressed as:

XM + Y O2 ~ Mx 02y
MxOy + Y H2 ~ XM + Y H20
XM + Y CO2 ~ MxOy + Y CO
To obtain the thennodynamic equilibrium constant K for each of the preceding
reactions, proceed as follows:

(2.3)

XM(s) + Y CO2 (g) ~ Mx Oy(s) + Y CO (g) (2.5)

 78 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

Assuming the compositions of solids remain constant and the gases behave
ideally, we have:

Where "a" represents the activity of each reactant and product of reactions (2-
3), (2-4), and (2-5), and P H20 , P H2 , Pea, PC02 represent the partial pressure of
H20, H2, CO, and CO 2, respectively.
By introducing the relationship between free energy GO and the equilibrium
constant, namely,

GO = -RT In K

it is shown that the reactions (2-3), (2-4), and (2-5) can proceed in either the
forward or reverse direction, depending on the temperature and the ratio of
PedPco2, PH2dPH2' and P 02 .
Figure 2-27 shows the standard free energy of formation for some metal and
metal oxide systems and CO/C0 2, H20/H 2, CO 2/C atmospheres as a function
of temperature. Assuming it is under equilibrium condition and at soldering
temperature of 250°C, lead oxide and copper oxide can be reduced by hydrogen;
however, hydrogen is not effective on tin oxide until the temperature reaches
600°C. Equation (2-4) also indicates that the presence of too much water vapor
in the furnace atmosphere will cause oxidation to certain metals. The partial
pressure of water vapor should therefore be maintained at a constant and defined
value.
At a given atmosphere composition and dew point, the gas flow rate and the
flow pattern of the exhaust system in the furnace are also important factors in
soldering performance. A typical flow pattern in a furnace chamber is shown in
Figure 2-28. The gas flow rate should be high enough to avoid localized at-
mosphere buildup as a result of local reactions. In order to achieve the best
performance and cost results, the required flow rate is determined by the char-
acteristics of solder paste being used, furnace belt speed, loading pattern, belt
width, and other furnace parameters. The exhaust efficiency and its flow pattern
in combination with the flow rate dominate the removal of volatile components
 CONTRO LLED ATMOSPHERE SOLDERING 79

0
-20
-40
-60
,.......,
co -80
()
~
......... -100

0
N
-120
a... -140
c
-160
I-
a: -180
II
0
0 -200
<l -220
- 240
-260
-280
-300
0 400 800 1200 1600 2000 2400
Temperature ( °C )

Figure 2-27 Standard free energy of formation of selected metal and metal oxides.

Exhaust

Belt
Figure 2-28 Typical flow pattem of conveyorized furnace .
80 SOlDER PASTE TECHNOlOGY AND APPLICATIONS

generated from the pyrolysis and evaporation of chemicals in the solder material,
which in tum affect solderability.
It should be noted that a complete burnout process is normally infeasible at
the eutectic temperature of tin-lead or the eutectic tin-lead-silver solders, when
using a solder paste.
In order to obtain good solderability and quality solder joints, a metallic
continuity at the interface(s) between the solder and the substrate must be formed
during soldering. When using a solder paste, a complete coalescence has to occur
in synchronization with the development of metallic continuity at the interface(s).
At soldering temperature, the atmosphere surrounding the workpiece protects or
interacts with the surface of the substrates, with the solder alloys, and with the
chemical ingredients in the flux or vehicle system. These interactions affect the
resulting chemical and physical phenomena in terms of volatilization, thermal
decomposition, and surface-interfacial tension.
A controlled atmosphere is expected to deliver more consistent chemical and
physical interactions; therefore, a more consistent soldering process results.
In addition to consistency, the inert or reactive atmospheres possess further
merits, which include:

Solderability enhancement
Solderability uniformity
Irregular residue charring prevention
Polymer-based board discoloration prevention
Wider process window
Overall quality and yield improvement

The inert and reactive atmospheres 'are expected to facilitate conventional

fluxing efficiency during soldering. It should be noted, however, that the per-
formance results highly rely upon the specific atmospheric composition and its
compatibility with the solder material, substrates, and chemicals incorporated in
the system, which must be also compatible with the soldering temperature profile.

2.15 SOLVENT CLEANING

Until recently, the solvents used for cleaning the after-reflow residue were pri-
marily the azeotropes or blends of I ,I ,2-trichlorofluoroethane with alcohol, meth-
ylene chloride, or accetone or blends of I, I, I-trichloroethane with one or several
alcohols. Those solvents have delivered successful results from a cleaning point
of view. However, recent scientific data further substantiate the earlier findings
that the chemicals under the group of chlorofluorocarbon (CFC) are the main
contributors to the depletion of ozone (0 3) in the stratosphere, which results in
biological and agricultural hardship from the transmission of excess ultraviolet
 SOLVENT CLEANING 81

radiation onto the earth. It is theorized that the stratospheric chlorine free radial
(CI·) is an active species depleting ozone through a catalytic chlorine oxide
cycle, as shown in the following reactionsY

C)· + 0 3 ~ CIO + O2
CIO + O· ~ CI· + O2

In addition to natural ozone removal mechanisms occurring in the stratosphere,

the chlorine radical can be derived from sources such as CFC. This is because
of the extraordinary stability of CFC compounds that are released to the trop-
osphere and are not broken down by tropospheric processes. The accumulated
CFC in the troposphere can then diffuse into the stratosphere where CFCs are
photodissociated by ultraviolet radiation, generating chlorine.
These findings give rise to global concern about the use of CFCs and lead to
the restriction of CFC production with a designated schedule set by the Montreal
Protocol. Many countries and regions have actually started implementing more
stringent regulations limiting the use of CFC chemicals.
Under today's limited availability and a likely future unavailability of CFC
solvents, different alternatives to CFC solvent cleaning of electronic assemblies
will become available through the efforts of solvent manufacturers, related ma-
terial suppliers, and electronic packaging manufacturers. The alternatives and
the corresponding status as of this writing are:

Cleaning Medium Status

Hydrogenated CFC solvent Available, in field testing
Water plus saponifier In use
Water In use
Terpene plus water Available, in field testing
Other CFC substitutes In development

The hydrogenated CFC solvents essentially possess similar properties in terms

of cleaning as do the CFC solvents and perform effectively to clean conventional
rosin-based residue. Table 2-9 shows some preliminary testing results of a hy-
drogenerated CFC solvent, HCFC-141b (CH3 , CCl2 F), for cleaning RA-type
fluxes and RMA-type solder paste. 13 The terpene solvent is able to clean rosin-
based residue because of its natural compatibility with rosin. However, terpene
hydrocarbon solvent, having inherent low vapor pressure, has to be used in
conjunction with subsequent aqueous rinsing in order to remove the residual
solvent through emulsification, and it obviously lacks vapor-degreasing capa-
bility. Aqueous cleaning is discussed in the next section.
82 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

Table 2-9 Cleaning Ability of HCFC in Relation to CFe 12

Solvent
CFC-113/CH 3
OH/acetone HCFC-141 b/CH 3 0H
Remaining Contaminate isohexane/CH 3 N02 CH 3 N02
Flux 1 (ugNaCl/in.2)* 7.2 5.2
Flux 2 (ugNaCl/in. 2)* 21.2 17.4
Flux 3 (ugNaCl/in.2)* 7.0 6.7
Paste (ug/in2)** 1198.0 1105.0
"In reference to MIL-P-28809 A 4.8.2.3.
""By high performance liquid chromatography.

Despite today's dynamic nature of cleaning solvent and technique selection

and electronic assembly design, the principle of cleaning will not alter. The
knowledge and infonnation to answer the following questions are essential to
perfonn an effective cleaning process:

What are the characteristics of soldering residue to be cleaned?

What are the properties of cleaning agent?
What are the features of cleaning equipment?
What is the compatibility among residues, cleaning agent, and equipment?
How is the cleanliness measured?
What are the criteria of cleanliness?

During cleaning, the action steps involve wetting, spreading, softening, solu-
bilizing, and carrying away, which operate in parallel or in sequence. The
physical properties of the cleaning agent that affect its efficiency are boiling
point, density, viscosity, surface tension, solvency, latent heat of vaporization,
vapor pressure, and solubility parameters. The relationship between each property
and cleaning perfonnance follows the general trends listed below:

Increasing temperature increases cleaning efficiency.

Increasing density increases wetting ability.
Increasing viscosity decreases wetting ability.
Increasing surface tension decreases wetting ability but increases capillary
ability.
Increasing latent heat of vaporization decreases dryability.
Increasing vapor pressure increases dryability.
Increasing solvency increases solubilizing efficiency.
Closer solubility parameter between cleaning agent and gross residue facilitates
cleaning.
 SOLVENT CLEANING 83

Table 2-10 Solubility Parameters of Selected Materials*

Solubility Parameter (cal 1l2 cm- 3/2 )
Material 8d 8p 8h 8
Water 6.4 15.4 16.8 23.8
Acetone 7.6 5.0 3.4 9.8
Methylene chloride 8.9 3.1 3.0 9.9
Methanol 7.4 6.0 10.9 14.5
Ethanol 7.7 4.3 9.5 12.9
2-Propanol 7.8 3.0 8.0 11.5
l-Butanol 7.8 2.8 7.7 11.3
2-Butanol 7.7 2.8 7.1 10.8
n-Hexane 7.3 0 0 7.3
n-Heptane 7.5 0 0 7.5
n-Octane 7.6 0 0 7.6
n-Nonane 7.7 0 0 7.7
n-Decane 7.7 0 0 7.7
n-Hexadecane 8.3 0 0 8.0
Cyclohexane 8.2 0 0.1 8.2
Benzene 9.2 0 1.0 9.1
Toluene 8.8 0.7 1.0 8.9
Biphenyl 10.5 0.5 1.0 10.6
1,1,1-Trichloroethane 8.3 2.1 1.0 8.7
l,1,2-Trichlorotrifluoroethane 7.2 0.8 0 7.2
Prelete 8.6
Freon TMS 7.4
Freon SMT 7.9
Genesolv DMS 7.4
Freon TA 7.4
Genesolv DA 7.4
Freon TE 7.4
Genesolve DE 7.4
Di-n-butyl phthalate 8.7 4.2 2.0 9.9
N, N-Dimethyl formam ide 8.5 6.7 5.5 12.1
Ethylene glycol 8.3 5.4 12.7 16.1
Glycerol 8.5 6.0 14.3 17.6
Propylene glycol 8.2 4.6 11.4 14.8
·Courtesy of Allan F. M. Barton, Handbook of Solubility Parameters and Other Cohesion Parameters (CRC
Press, Inc., 1983).

The solubility parameter is a cohesive property of materials, also known as

the Hildebrand parameter, named after Joel H. Hildebrand for his pioneering
fundamental concept on the solubility of nonelectrolytes. Assuming the cohesive
energy (E) between molecules is a linear combination of three contributors-
84 SOlDER PASTE TECHNOlOGY AND APPLICATIONS

dispersion interaction (Ed)' polar interaction (Ep), and hydrogen bonding (Eh)---
then

Since cohesive pressure p is defined as cohesive energy per unit volume V,

-E
p=-
V

and solubility parameter (5) is related to the cohesive pressure by

5 = pll2

thus,

Table 2-10 lists the solubility parameters with three contributing portions for the
selected chemicals shown.
With respect to cleaning technique and equipment, the techniques used include
vapor degreasing, liquid spray, immersion, high pressure spray (more than 50
psi), and immersion with ultrasonic aid. The single step of vapor degreasing is
the mildest cleaning process and is considered inadequate to clean highly dense
surface mount assemblies. A combination of steps, such as

vapor --+ spray --+ immersion-ultrasonic --+ spray --+ vapor or vapor --+ spray
--+ vapor-immersion-high pressure spray --+ spray --+ vapor is generally required.

2.16 AQUEOUS CLEANING AND AQUEOUS CLEANING PASTE

Two types of aqueous cleaning prevail, namely, water saponification and water
solubilization. For water-saponification cleaning, the water medium contains a
saponifier that changes the water-insoluble rosin into a soluble salt as represented
by the following reactions:

©-COOH + N'OH~©-COO- N.+ + H2 0

©-COOR + N'OH~©-COO- N.+ + ROH

 NO-CLEAN PASTE 85

And water-solubilization cleaning refers to pure water as the cleaning agent,

without incorporation of any additives.
In addition to saponifer or saponifiction, the additives and reactions involved
in different chemical nature can be considered. The specific additive obviously
depends on the chemical makeup of the paste.
For cleaning with water and saponifer, the analog of conventional RMA-type
paste, which generally contains rosin or rosin derivatives and essentially main-
tains the characteristics of the RMA-type pastes, meets the requirement. The
cleaning effectiveness highly relies on control of the quality and quantity of the
saponifier used during the cleaning process. A water-soluble product, however,
requires a different chemical makeup. The conventional OA-type paste is not
considered a proper candidate. Technically, the system can be designed to deliver
adequate performance on the production floor; however, tighter process control
and smoother operation flow are generally needed.
It should be recognized that cleaning under the tight clearances between large
components and boards, as well as being able to nondestructively measure clean-
liness, has been and will continue to be a challenge, regardless of whether solvent
or aqueous cleaning is used. Water-soluble products are designed to be com-
patible with plain water for easy removal. In addition to the chemical makeup
of the water-soluble product, cleaning efficiency depends on the design and
layout as well as the execution and control of the cleaning process. Furthermore,
the temperature profile being used during soldering needs to be compatible with
the paste characteristics.
Aqueous cleaning eliminates the use of CFC solvents; but, waste water man-
agement and disposal or recycling come into play. The waste water should not
be discharged directly into public sewage systems without testing and treatment.
Ion exchange and chemical treatment have been used to precipitate out heavy
metals. Federal, state, and local environmental regulations should be observed
for implementing proper waste disposal.

2.17 NO-CLEAN PASTE14

To combat the potential occurrence of "imperfect" cleaning, using the mildest

amount of activity in the flux or vehicle system (nonsolder powder portion in
solder paste) possible is a logical strategy. This leads to a discussion ofno-clean
pastes, which are expected to solve the CFC issue as well as to alleviate clean-
ability and cleanliness-measurement problems.
No-clean paste, therefore, is poised as an elegant and well-rounded approach.
Nevertheless, the no-clean route is not a simple drop-in replacement-it not only
places new demands on the paste design but also on the reflow process. To
achieve the desired results, particularly the minimal residue, controlled atmo-
sphere soldering constitutes an integral part of the system. Controlled atmosphere
soldering is discussed in an earlier section. Figure 2-29 depicts an example of
 86 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

Figure 2-29 Solder joint and reflow residue of (a) No-clean paste after N2 reflow; (b) no-
clean paste after ambient air reflow; (e) RMA-type paste after N2 reflow.

the effect of controlled atmosphere soldering on solder joints prepared from a

standard RMA-paste and a no-clean paste after being reftowed under an N2
atmosphere. It also illustrates the solderability enhancement of an N2 reftow,
exhibiting good wetting characteristics, minimal and inactive residue, and no
solder balls, while the joint reftowed in ambient atmosphere shows solder balls
and excess residue.
 fiNE PINCH PASTE 87

2.18 FINE PITCH PASTE

As the pin count of surface mount components increases, the pitch between pins
becomes finer. As of this writing, a fine pitch pattern of 0.025 in. (0.63 mrn)
is considered the state of the art. When the pitch becomes smaller than 0.025
in. (0.63 mrn), the processes and materials must be modified. What is needed
in order to meet the demands of finer pitches covers several aspects of perfor-
mance. Paste flow control in response to temperature becomes more important.
The paste needs to retain its shape as temperature rises from ambient to near
reflow, in order to avoid pad bridging. In Figure 2-30, paste 1 shows the desirable
flow control and paste 2 imparts significant hot slump. The size of the solder
powder incorporated in the paste will shrink from the 44- to 74-~m range to the
5- to 45-~m range . This requires practical powder production technology coupled
with affective sizing technique. Solder ball-free solder paste also becomes more
important for fine pitch as demonstrated in the solder ball test of Figure 2-31.
In addition, the effects of the metal content, tackiness, and viscosity will become
more pronounced. Overall, precision and control are key factors in the success
of applying fine pitch solder paste. The following proposes some parameters in
the paste and process suitable for three ranges of component pitches:

25 mil ~ pitch < 50 mil

Standard solder paste
Powder particle size 44 to 74 ~m
Selected printing parameters
Slower squeegee speed, 0.5 to 2.0 in.lsec
Harder squeegee durometer, ~80
Thinner stencil , 6 ± 2 mil
Precision and control

Figure 2·30 Flow dynamics of solder paste: (1) high flow restrictivity; (2) low flow restric·
tivity.
88 SOLDER PASTE TECHNOLOGY AND APPLICATIONS

Figure 2-31 Solder ball test on alumina substrate (a) solder ball free; (b) excessive solder
ball.
 FINE PINCH PASTE 89

Table 2-11 Summary of Quality Assurance Tests for Solder Paste

Part Tests
Paste 1. Appearance
2. Metal content and flux and vehicle percentage
3. Density
4. Viscosity
5. Viscosity versus shear rate
6. Cold slump
7. Hot slump/open time
8. Molten flow
9. Tack time
10. Dryability
11. Dispensability
12. Printability
13. Shelf stability
14. Storage, handling, and safety
Flux and vehicle 15. Water extract resistivity
16. Copper mirror corrosion
17. Chloride and bromide
18. Acid number
19. Infrared spectrum fingerprint and other
spectroscopies
Solder powder 20. Alloy composition
21. Particle size-sieve
22. Particle size distribution-SediGraph
23. Particle shape
24. Particle surface condition
25. Dross
26. Melting range
Reflow 27. Solder ball
28. Solderability
29. Exposure time
30. Soldering dynamics
Post-reflow 31. Cleanliness-resistivity of solvent extract
32. Surface insulation resistance, before and after
cleaning
33. Solder joint appearance
34. Solder voids
35. Joint strength
36. Power cycling
37. Temperature cycling
38. Vibration test
39. Simulated aging
40. Thermal shock
90 SOlDER PASTE TECHNOlOGY AND APPLICATIONS

10 mil < pitch < 25 mil

Solder paste with powder particles 5 to 44 t-Lm
Stencil thickness 4 ± 2 mil
Designed paste deposition pattern
Pitch 10 mil
Fine mesh screen, 200 mesh or finer
Fine needle dispensing ~ 25 gauge
Designed paste deposition pattern
Prebumping
Solder paste with fine powder particles, 5 to 25 t-Lm

2.19 QUALITY

With quality defined as continuous improvement, a major effort needed today

in the solder paste area is to continue to advance and perfect the performance
and the consistency of solder paste. Table 2-11 summarizes the parameters that
affect the quality and performance of a solder paste. The detailed test procedures
of each parameter are illustrated in the reference. 1

2.20 CONCLUSION

In view of new trends and new demands, the question "can solder paste tech-
nology coupled with soldering technology continue to provide reliable intercon-
nections for board-level packaging in the next decade and beyond?" has to be
asked and answered. The answer relies on how promptly and how well the
concerned scientists and engineers continue to advance and perfect the design-
material-process and how committed the fund decision makers are to the advanced
technology development. This is not to imply that a technology has no limits,
but there are still advancements and innovations in solder paste yet to come.
This has to be implemented by effectively providing a system solution involving
design, material, equipment, and process.

REFERENCES

l. Hwang, J. S., Solder Paste in Electronics Packaging-Technology and Applications

in Surface Mount, Hybrid Circuits, and Component Assembly, Van Nostrand Rein-
hold, New York, 1989.
2. March, J., Advanced Organic Chemistry, Reactions, Mechanisms, and Structure,
McGraw-Hill, New York, 1968.
3. Hwang, J. S., "Solder Paste Rheology-Principle and Practice," Technical Pro-
ceedings, Expo SMT 88, pp. 171-177.
4. Ferry, J. D., Viscoelastic Properties of Polymers, Wiley, New York, 1970.
 REFERENCES 91

5. Riemer, D. E., "Analytical Engineering Model of the Screen Printing Process," Solid
State Technology, September 1988, p. 85.
6. Hwang, J. S., "Controlled Atmosphere Soldering-Principle and Practice," Technical
Proceedings, NEPCON-West, 1990.
7. Vandervoort, S., "Lamp IR Reflow Provides Repeatability Between Units," Elec-
tronic Packaging and Production, May 1989, p. 58.
8. Martel, M. L., "New Wrinkles in Reflow," Circuits Manufacturing, June 1989, p.
33.
9. Dow, S., "Reflow Soldering Survey," Assembly Engineering, December 1988, p.
18.
10. Glynn, M., "Full Spectrum IR Reflow through Energy Management," Printed Cir-
cuits Assembly, February 1989, p. 22.
11. McGinn, K. S., "Effects of Fumace Atmosphere Chemistry in Sintering," Industrial
Heating, May 1985.
12. Stolarski, R. S., and R. J. Cicerone, Can. J. Chem., 52, 1974, p. 161.
13. Bonner, J. K., "Searching for a Substitute," Circuits Manufacturing, July 1989, p.
42.
14. Hwang, J. S., "Soldering and Solder Paste Prospects," Surface Mount Technology,
October 1989, p. 56.
 3
Technical Considerations in
Vapor Phase and Infrared Solder
Reflow Processes
Charles L. Hutchins

3.1 INTRODUCTION TO SURFACE MOUNT REFLOW SOLDERING

Based on the method of soldering technology that is used, there are three general
types of assemblies. Some publications list other type designations, but the
convention used in this chapter is the one that is most common.) Since the
manufacturing or soldering technology is the basis for the designation, the types
given below are unambiguous and clearly understandable.
The two basic methods for soldering are reflow, using either presoldered
components, solder preforms, or solder paste, and wave or flow soldering,
commonly used for through-hole assembly. Once the type of process is chosen,
the types of components that can be used are determined. These two soldering
techniques are used for the general types that are discussed below:

Type I. Surface mounted components only using reflow soldering, either single
sided or double sided.
Type II. Both surface mounted components, using reflow soldering, and through-
hole components, using wave or flow soldering, that are mounted on top
of the printed circuit board (PCB). By using a modified or dual wave,
small surface mounted components may be mounted on the bottom with
adhesives and soldered at the same time as the through-hole components
(see type III below).
Type Ill. Surface mounted components are mounted on the bottom with ad-
hesives, through-hole components are inserted from the top, and both
are soldered with a single pass through a modified or dual wave.

The details of the type I (and the first part of the type II) process are discussed
in the following section because of their applicability to this chapter.

92
 TYPE I 93

3.2 TYPE I

This process uses only surface mount components, which are attached with a
reflow solder process. These include actives, such as small outline transistors
(SOTs), small outline integrated circuits (SOICs), ceramic chip carriers (CCCs),
and plastic leaded chip carriers (PLCCs), as well as a wide assortment of passive
components. Included in this group are chip capacitors and resistors, various
forms of larger filter capacitors, sockets, connectors, coils, switches, and other
special hardware to complete the electronic function of the surface mount as-
sembly. These can be mounted on either one side or on both sides .
A ·typical single sided type I assembly is shown in Figure 3-1. It uses several
different types of surface mount packages. Surface mount assemblies are found
in many different sizes and different substrate materials, from large epoxy-glass
computer "mother boards" to small and medium ceramic substrates for military
and avionics uses.
The basic process flow is shown in Figure 3-2. It shows the four major steps
for the reflow process. The first step is the placement of solder paste on the
component pads or lands on the PCB . This is typically done with either a syringe
dispenser or a screen printing procedure. The amount of paste can be easily
controlled and registration to ± 2 mils is usually achievable with standard equip-
ment.

Figure 3-1 Typical single-sided type I surface mount assembly. (Photo by author)
94 TECHNICAL CONSIDERATIONS IN VAPOR PHASE

YES (TYPE 1B DOUBLE SIDED)

TO TEST

Figure 3-2 Process flow for type I A and B.

The next step is the placement of components in the paste, with particular
care to align them to the PCB lands. The paste has a sticky consistency, holding
the components in place during careful handling. Although during the reflow
step the surface tension of the molten solder will provide some self-alignment
of smaller components, minor yield problems can result from component mis-
registration.
The third step is the reflow of the solder paste. This can be achieved in several
types of hot ambients, the most popular of which are the vapor phase (or con-
densation heating) and infrared (or infrared-convection) reflow systems. These
latter two steps will be discussed extensively in this chapter. The final step is a
clean-up operation to remove flux residues and other contamination.

3.3 SOLDERING REQUIREMENTS FOR SURFACE MOUNT

TECHNOLOGY

The two major requirements for soldering SMT assemblies are coplanarity and
solderability. The leads of the component and the land area of the substrate or
PCB must be coplanar and wettable with the solder used in their joining. As
these have a significant impact upon the solder joint strength, yield, and relia-
bility, it is important that we understand the effect that the reflow soldering step
has on both of them.
 REFLOW PROCESS PHASES 95

There are three major requirements for the solder alloy and the method used
in the soldering step. First, the reflow temperature of the solder alloy must be
compatible with the metallization used on the components, the substrate, and
the reflow system. Generally, this means that the solder alloy melts at approx-
imately the same temperature as the solder coatings on the component leads and
substrate lands. Also, the constituents of all three must not form undesirable
metallic compounds, such as intermetallics with brittle or weak mechanical
properties or high electrical resistance.
Next, the surface tension of the molten solder should be sufficiently strong to
provide a continuous surface between the component lead and the land on the
PCB during the reflow process, resulting in fillet formation upon cooling. This
attribute has not received much attention in previous solder research but will
find increased attention as SMT implementation increases. A solder menisco-
graph or wetting balance can be used as a quantitative test instrument.
Surface tension of the molten solder is what makes surface mounting a man-
ufacturable process. It is extremely difficult and costly to form component leads
into an absolutely planar arrangement, so a soldering system (alloy, flux, solvent,
and reflow technique) that can provide a measure of compensation for nonplan-
arity is very desirable and valuable. It will prevent yield losses from open solder
joints and, thus, the costly manufacturing rework.
Thirdly, the strength of the alloy formed after the soldering step must be strong
enough to provide reliable joints. In some cases a "flexible," or compliant, joint
is desired, while in others a strong rigid connection is required. Obviously, this
property is dependent upon the choice of the alloy constituents as discussed
above with respect to the components and substrate, but it also can be an in-
dependent choice. Further, it can be affected by the time and temperature history ,
reflow operation plus any rework or repair of the total assembly. The alloys and
intermetallics can continue to change if the conditions are favorable, such as
extremely hot leads of high power dissipating components. In general, this is
not a serious problem, but considerations for this potential must be constantly
maintained.

3.4 REFLOW PROCESS PHASES

The primary purpose of the reflow process is to melt the solder particles, wet
the surfaces to be joined, and then solidify the solder into a strong metallurgical
bond. However, prior to this melting phase, several other phases occur. The
sequence of these actions is diagrammed in Figure 3-3.
During the five phases shown, there are several local actions, followed by a
global action, and then a strong metallurgical bond and fillet formation. The
local actions start with the solvent evaporation. If this is too violent, there will
be "mini-explosions," giving rise to the formation of small solder balls being
scattered all over the board. This reduces the amount of solder volume available
 96 TECHNICAL CONSIDERATIONS IN VAPOR PHASE

200

~ SOLDER MELTING
w
g; 150
COMPLETES,
SURFACE TENSION
~ SOLDER BALLS
TAKES OVER
0:
w MELT, WETTING
~ 100 ANDWICKING
w BEGIN
I- FLUX REDUCES COOL
50 METAL OXIDES DOWN
PHASE

TIME

Figure 3-3 Solder reflow phases.

and can result in weak or open solder joints, or the solder balls can cause
extraneous shorts if they happen to end up between two metal conductors.
Next the flux melts and begins the chemical reaction with the metal oxides,
preparing a "clean" metal surface for the solder. As this metal oxide reduction
action of the flux is proceeding, the solder spheres in the paste begin to melt at
the hottest spots, replacing the flux and wetting the clean metal surfaces. This
wetting action results in the wicking of the solder onto the hottest and cleanest
metal surfaces. The speed of the action depends strongly upon the fluxing action,
the local temperature, and the metal alloys of the component lead, solder paste,
and substrate. Under most conditions, temperature nonuniformity is the major
factor. High temperature helps to reduce surface tension and can promote sig-
nificant wetting and wicking action.
The heat source should be 25 to 40°C greater than the melting, or liquidus,
temperature of the selected alloy to assure complete melting of the solder and
good metallurgical bond formation. The actual time for the solder to completely
melt and wet the metals to be joined depends upon the localized temperature of
the assembly, and at the typical reflow temperature of 215°C (Sn63 or Sn62) it
will be 1 to 2 sec,2 as shown in Figure 3-4. The time that the assembly is at the
maximum temperature of the reflow system should be limited to approximately
30 to 50 sec, depending upon the thermal mass of the assembly. This will allow
reflow to take place but will minimize the formation of undesirable intermetallics
or damage to the board. This will also allow time for the trapped flux and residual
solvents to escape, minimizing void formation as the solder cools.
After all of the solder spheres have melted, a liquid solder "volume" forms.
The global action of the surface tension of the molten solder then prevents further
wicking action. Thus, the surface tension action, which minimizes surface area,
 REFLOW PROCESS PHASES 97

190 195 200 205 210 215 220 225 230 235

TEMPERATURE (CO)

Figure 3-4 Solder wetting time versus temperature.

is beneficial in that it helps hold the solder in place and bridge a gap between
the component lead and the substrate, forming a fillet upon cool down. The
volume of solder, the metal alloys being joined, and the physical geometry of
the metal (lead, land, via, etc., shape, and spacings) determine the resultant
final shape, fillet formation, and strength of the solder joint.
If all the factors of the process are sufficiently matched, there is good solder
wetting and fillet formation between the component lead and the land on the
substrate. If this is not the case, several things can happen at the solder joint.
There can be an open joint due to an out of specification coplanarity gap or due
to a contamination-wettability problem. There could be a marginal (insufficient)
solder joint formed for several reasons-volume loss due to solder ball formation,
a small (not out of spec.) coplanarity gap, insufficient solder paste print, or a
wicking away of the solder from the desired joint (either vertically or horizon-
tally).
Thus, the matching of the solder paste properties and the reflow environment
must comprehend several actions. These are the solvent evaporation, flux action,
solder wetting and wicking, and surface tension of the molten solder. Further,
minimizing temperature nonuniformity and its resultant effects on the solder is
also very important. To fully optimize the process, one must control each of the
actions listed above with the raw material introduced into the operation and the
time-temperature profile.
The cool down after reflow should take place within 30 to 50 sec to enhance
small grain formation but should be no greater than 3°e/sec to prevent thermal
98 TECHNICAL CONSIDERATIONS IN VAPOR PHASE

shock damage to some components. Larger grain formation in solder usually

promotes voids and joints that are slightly weaker.

3.5 REFLOW EQUIPMENT

The major types of equipment used in reflow soldering include infrared furnace,
vapor phase system, convection oven, hot platen or conductive belt, and laser.
The order given is also the approximate order of popularity or frequency of use.
Each is an acceptable method in principle, with the differences being primarily
in the type of assembly being produced and in yield and cost trade-offs. Each
of these will be discussed in the following sections, with major emphasis on
infrared (lR) and vapor phase soldering (VPS).

3.5.1 Infrared

The IR process primarily uses the transfer of heat by absorption of radiant energy.
Infrared radiation is energy transported by electromagnetic waves. It is the portion
of the electromagnetic spectrum with wavelengths between 0.72 and 1000 ILm.
Radiation wavelengths are a function of the temperature of the emitter source
and possibly of the material used as a "reflector," which really absorbs and
reradiates the original energy.
All bodies radiate electromagnetic energy as a function of their absolute tem-
perature. This relationship is described by the Stefan Boltzmann law:

where W = total emissive power

K = Boltzmann's constant
T = absolute temperature

Thus as the temperature of the emitter changes, the radiated power changes by
the fourth power. This is very useful for quick response to changes to loading
of an IR furnace but can also make the control of the desired temperature very
difficult.
The useful range for heating occurs between the wavelengths of 1 to 8 ILm.
They are generally classified as short or near wave (0.7 to 3 ILm) and medium
to far wave (3 to 8 ILm). The temperature of the radiating element determines
the wavelength. Thus, the shorter wavelengths usually are obtained from lamp
emitters that have a tungsten filament, while longer wavelengths come from
panels that use resistive heating elements.
Heating takes place indirectly as the IR energy is absorbed within the molecular
structure and causes an increase in local temperatures. Thus it is a "volume"
heating phenomenon. IR energy is absorbed primarily in organic materials such
 REFLOW EQUIPMENT 99

as the PCB, solder paste flux and solvents, and plastic components. The depth
at which it is converted to heat is dependent upon the wavelength, with shorter
wavelengths penetrating deeper. Metal reflects most of the IR energy but does
act as a conductor to carry heat from hot to cool spots.
Since most of the energy is absorbed in the PCB, it gets hot first and then
transfers this heat to the lands. The flux and solvent also absorb a small amount
of the energy. Thus, the land gets hot first, the solder melts and wets the land,
then wets the lead. This reduces the amount of solder that wicks away, allowing
a given volume of solder to span a greater coplanarity gap if one exists, reducing
the probability of an open solder joint. However, since the board and conductors
are the hottest, solder can wick laterally away from a joint to a large conductor
or via if the solder is not constrained by proper PCB layout or construction.
The major difficulty with IR reflow is that it is not an "equilibrium" method
of heating, (i.e., the heating object will never reach the same temperature as the
heating source). With IR sources of quartz lamps or resistance panel heaters,
the source "temperatures" range from 400°C for panels to lOOO°C for quartz
lamps. To reach a maximum safe reflow temperature of 220°C, one must depend
on accurate pulse operation of the heaters and the belt speed to be adjusted such
that the heated object exits the heating chamber at the correct time-temperature.
For a given mass, too slow means that there will be higher energy absorption
and too high a temperature, and vice versa.
Also, since IR heat results from unequal energy absorption, variations in
absorptivity or mass, either across the board or from assembly to assembly,
could produce significant temperature variations. For boards of large size or large
mass variations, temperature differences in excess of 20°C can exist between
the comer (low mass area, typically the hottest) and the center (or high mass
area). This is generalized in the typical IR reflow profile shown in Figure 3-5.
This curve shows a five-zone system with appropriate temperature plateaus to
optimally achieve the actions shown in Figure 3-3.
Note that this curve is also dependent upon the spacing, or loading, of the
assemblies going through the heating zones. Different spacings between assem-
blies could result in significantly different temperature profiles. Excessive tem-
perature, with respect to the individual components or PCB material that are
used, can cause low yield and/or reliability problems. Inadequate temperature
could produce umeflowed solder joints.
A typical IR system will have three or more heating zones, and some will
have independent top and bottom heat source control. Other features will include
microprocessor control, with a memory to store profile information, and auto-
matic setup from memory. This latter feature is useful for production of SMAs
that are different in size and mass. A typical system that provides these features
is shown in Figure 3-6.
With temperatures up to lOOO°C being used to generate the radiant energy,
more care must be exercised. The geometry, mass, absorptivity, and thermal
conductivity of the material being soldered must be considered. With a short
100 TECHNICAL CONSIDERATIONS IN VAPOR PHASE

250
CORNER (LOW MASS) ,. f .· ·· · ·· ··.
TEMPERATURE

"'<.\. . . . . .
200 '\.. .,:'"
~ .... ... . . ,'
o
~
,, ,.'
~ 150
, " .'
::>
!;;:
a:
w
..,, ,
•
'
CENTER (HIGH MASS)
~ 100 TEMPERATURE
w
r-
ZONES

50
2 3 4 5

o 30 60 90 120 150 180 210 240

TIME (SEC)

Figure 3-5 Typical five-zone infrared profile.

Figure 3-6 Infrared reflow system. (Courtesy of Vitronics Corporation, Newmarket. NH)
 REF LOW EQUIPMENT 101

wave (lamp) furnace, ceramics, organics, and glasses are all transparent to some
degree; thus the energy is more penetrating. With a longer wave (panel) system,
there is more surface heating. In both types there is usually some portion of heat
energy that is transferred to the assembly by convection. With lamp systems it
is in the 20 to 40 percent range, while with a panel system it can be greater than
50 percent. With both systems, slower belt speed and lower power settings will
increase the ratio of convection heat.
Achieving a time-temperature profile for a safe and complete reflow is an
individual effort for each and every assembly, and to a lesser degree the profile
must be adjusted for the solder paste being used. Some surface mount assemblies
have a nonuniformity of mass due to component placements or cutouts in the
board for mechanical components. These make it difficult to find the correct
power settings and belt speed, and resultant zone temperatures, that will produce
acceptable temperature uniformity across the PCB to prevent problems with the
day-to-day range of manufacturing variables.
Since IR ovens for reflow soldering generally use three or more heating zones,
they provide a wide range of heating profiles for reflow of surface mount as-
semblies. As the assembly passes through the oven on the belt, the temperature
profile is established by the speed of the belt, the energy levels of the sources,
the distance of the assembly from the sources, and the absorptive characteristics
of the components and the substrate.
The IR system provides an opportunity to tailor a specific heat profile to
provide optimum yields but requires more attention to initial profile setup and
maintenance of the equipment. It must also have a "low mass" characteristic in
order to respond quickly to changes in mass loading, such as would happen
when the output of a pick-and-place changes due to component feeder problems.

3.5.2 Vapor Phase

Vapor phase soldering (VPS) was developed by the Western Electric Company
in the mid-1970s as a technique to solder large backplane assemblies that were
impractical to wave solder. In its simplest form, the process is carried out in a
deep container that contains boiling liquid in the bottom and a condensing coil
near the top. The part is introduced into the vapor just above the liquid, and the
vapor condenses on its cooler surfaces, giving up its latent heat of vaporization.
This is a very efficient process that transfers heat from the vapor concentration
gradient as well as a temperature gradient (convection heat) within the reflow
chamber.
The fluid used most commonly today is an inert perftuorinated liquid, a de-
rivative of compounds originally developed to provide vaporization cooling of
power transformers. These fluids have a high boiling point, are thermally and
chemically stable and compatible with most materials of the soldering process,
and have a vapor that is significantly heavier than air. The most commonly used
102 TECHNICAL CONSIDERATIONS IN VAPOR PHASE

fluids boil at approximately 215°C. There are other fluids available for higher
and lower temperatures for noneutectic tin-lead or low temperature solder alloys.
VPS is the easiest reflow technique to use. It is defined as an equilibrium
heating system. The object being heated eventually attains the same temperature
as the heat source. This means that the vapor phase is also a very safe process
in that the assembly will never see excess temperatures.
Since the temperature of a saturated vapor is the same as the boiling point of
its fluid, the soldering temperature is predetermined by the fluid chosen. For any
given fluid, this temperature is constant within a few degrees, resulting in a
process both precisely and repeatedly maintained. This provides adaptability of
the process to a variety of applications, independent of product geometry or
uniformity of thermal mass of the SMT assembly.
VPS is achieved by transferring the latent heat of the vapor to the surface
upon which it condenses. Thus, it is a "surface" heating phenomena. Objects in
the reflow environment are heated on their surfaces first. If an object has a higher
surface-to-volume ratio than its neighbor, it will heat up faster. Also, if it has
a higher thermal conductance value, it will reach higher internal temperatures
faster and conduct heat away from hot spots to cooler areas. This is important
where there is considerable metal, such as a "]" or "I" lead. The leads will get
hot much faster than the metallization on the substrate, which is thermally
constrained by the PCB. This is shown in the typical VPS reflow time-temperature
profile in Figure 3-7. Note that both eventually achieve the same maximum
temperature, a key safety feature of VPS.
Thus, with no solderability problems, the solder will wet to the lead first.
Since the lead is hotter away from the solder paste, there will be a tendency for

250~-----------------------------------------'

,
LEAD
TEMPERATURE ..... .

~,' ..
, LAND
TEMPERATURE

o 30 60 90 120 150 180 210 240

TIME
(SEC)

Figure 3-7 Typical in-line vapor phase profile.

 REFLOW EQUIPMENT 103

WATER

SCRUBBER

HEAT
EXCHANGER WATER

PUMP
IMMERSION HEATER ~
Figure 3-8 Cross section of a typical dual vapor batch reflow system.

the solder to wick up the lead. Eventually after the solder is completely molten,
the surface tension effects will take over, bridging between the lead and the land
and allowing fillet formation on cool down. Any disruption in the surface tension
due to solvent splatter or a coplanarity gap between component lead and land
greater than the solder can bridge with the surface tension will cause the solder
to separate. The fillet will not form and an open solder joint will be the result.
There is a relationship between the surface tension of the solder alloy, the volume
of solder paste, the temperature uniformity, and the maximum allowable gap
between the lead and land for successfully forming solder joints that can only
be determined empirically.
The most common version of the vapor phase system uses a secondary vapor
blanket to control loss of the primary fluid. A cross section of a typical batch
reflow system is shown in Figure 3-8. The secondary vapor, formed by the heat
of the primary condensing coil and controlled by the secondary condensing coil,
forms a cap just above the primary fluid. Another type of batch system does not
use the secondary vapor but provides vapor containment by the system design.
A photo of a this type of batch system is shown in Figure 3-9.
Traditionally, the secondary fluid was a mixture containing primarily chlo-
rofluorocarbon 113 (CFC113). Since the interface contains superheated vapors
of this fluid, it is an inherent acid generator (HF and HC!), requiring considerable
maintenance to minimize the effects on the system. Newer lower boiling tem-
perature perfiourinated compounds are now available that were developed to
reduce this acid generation problem.
Other systems have been designed to operate in an in-line fashion for automated
manufacturing lines. The reflow time for modules of different mass is adjustable
104 TECHNICAL CONSIDERATIONS IN VAPOR PHASE

Figure 3-9 Typical single vapor batch reflow system. (Courtesy of Centech Corporation,
Minneapolis, MN)

by changing the belt speed. Systems are available in both single and dual vapor
models. Dual vapor systems are generally more expensive to purchase but use
less primary fluid and are therefore somewhat cheaper to operate. They do require
regular maintenance to control the acid generation problem.
Newer in-line vapor phase systems have built-in preheaters in the entrance
area to dry the solvents in the solder paste and provide a more gradual heating
of the assembly. A picture of a typical system is shown in Figure 3-10. This

Figure 3-10 In-line vapor phase solder reflow system . (Courtesy of Centech Corporation,
Minneapolis, MN)
 REFlOW EQUIPMENT 105

system has the additional feature of a special "conveyor," rather than a belt, that
supports the assembly in a horizontal position during reflow and cool down.

3.5.3 Convection

This process exposes the surface mount assembly to a hot oven and primarily
uses the convection method to transfer heat to the solder paste for reflow. The
temperature of the atmosphere in the oven and the speed of the belt establishes
the temperature profile. Secondary heating occurs as a result of radiation and
conduction from the hot surfaces within the furnace. The temperature-time profile
will be similar to the IR oven profile shown in Figure 3-7.
Typical convection furnaces employ carefully designed entry and exit baffles
to provide close control of the air temperature and flow and thus the solder reflow
ambient. They can have several heat zones and possibly a cooling zone. Similar
to IR ovens, inert or reducing atmospheres can be used to prevent oxidation.
These systems generally operate at much slower conveyor speeds per unit length
of heat tunnel than IR systems.
One of the potential benefits of the convection-based reflow system is its
ability to accommodate several different board sizes, masses, and materials with
one time-temperature profile. One reported test used the same profile for two
assembly weights that varied by more than a 10 to 1 ratio. The temperature
variation between the two was less than 25°C, more than desirable but dem-
onstrating the versatility of the equipment.

3.5.4 Conductive Belt

This technique uses a series of temperature controlled platens, with a heat con-
ductive belt transporting the assembly. The heat is transmitted through the belt
to the assembly to accomplish solder reflow. Soldering temperature profiles are
controlled by varying the temperature of the platens and the speed of the belt.
This is the only reflow process that allows a continuous visual inspection of the
process. However, to prevent oxidation, an inert gas hood is required. This
process is widely used for soldering ceramic assemblies and generally has limited
use in epoxy or glass PCB reflow applications.

3.5.5 Laser Soldering

Reflow soldering with a laser is achieved by focusing the beam on a desired

solder joint, pulsing the beam, then moving the assembly, and refocusing and
pulsing the beam on the next solder joint. This movement and focusing requires
computer control and detail programming for each solder joint. This is costly in
106 TECHNICAL CONSIDERATIONS IN VAPOR PHASE

both time and money for use as a general production reflow but could be cost
effective in some situations.
Potential applications include specialty soldering of unusual geometries or heat
sensitive components that could not withstand the reflow process or for repair,
soldering individual leads using computer control. A manually controlled beam
locator could also be a practical solution for a repair operation.

3.6 PREREFLOW SOLDER PASTE BAKE

Baking the solder paste before reflow removes the solvents and absorbed water,
reducing the tendency for splatter and solder ball formation during reflow. A
typical drying operation would be approximately 30 min at 80DC. Other rec-
ommendations of different times (20 to 60 min), and different temperatures (75
to 100DC) have been found useful, with the variations generally not critical to
yield.
Solder ball formation can be a significant problem if the balls get trapped in
the smaller spaces between ceramic chip carriers and substrates or possibly
between the leads of small pitch leaded ICs. The solder balls also reduce the
solder volume available for the fillet formation, and the action of splatter inter-
rupts the holding power of the surface tension. Both of these have minor dele-
terious results on the reflow and connection process and hence on yield.
Baking before vapor phase reflow has been observed to increase yield by
reducing the incidence of solder joint opens on J lead components. 3 The first
heating zone in an IR reflow system can provide a short, but effective, solvent
drying step. This can be particularly important on large assemblies with over
1000 solder joints. Conversely, it can be a marginally beneficial and costly
process for in-line vapor phase reflow production of small modules of less than
a few hundred solder joints. The cost of equipment and handling for a separate
in-line bake could be greater than the yield improvement achieved. IR preheaters,
built into vapor phase systems, offer the most promise and are built into most
currently sold systems.

3.7 MAXIMIZING SOLDER JOINT YIELD

One subtle but important difference between reflow and wave soldering is the
management of the surface tension of the molten solder. In reflow soldering, it
is necessary to maximize surface tension, while most wave soldering systems
use temperatures well above the liquidus point of the solder and oil intermixes
to reduce surface tension. This helps to prevent solder bridging, the major yield
loss problem with wave soldering. Conversely, the major problem in SMT reflow
soldering is open solder joints. Thus, as the reflow solder process is developed,
it is necessary to look for ways to increase surface tension to maximize yields.
 MAXIMIZING SOLDER JOINT YIELD 107

Figure 3-11 Solder joint showing representative force vectors during reflow.

To further understand the necessity for maximum surface tension, let's use
the simplified vector drawing in Figure 3-11. The paste thickness will be ap-
proximately as shown. When the vapor phase reflow process starts to take place,
the solder will typically melt first at the interface with the lead. The vectors
shown represent the forces on the solder as the reflow starts to take place.
If there is a positive thermal gradient up the lead, vector A represents the
force of the solder wicking up the lead. Vector B represents the force of the
surface tension attempting to keep the solder intact between the component lead
and the PCB land. To maximize the probability of a fillet formation upon cool
down, reflow conditions should be such that thermal gradients are minimized
and surface tension is maximized. 4 Techniques to accomplish this will be covered
later in this chapter.
The wicking action of the solder can be better visualized by examining the
picture in Figures 3-12 through 3-14. The first picture shows typical solder joint,
two good ones in the background and two open joints in the foreground. Note
the good fillet formation on the good joints and the extra solder on the shoulders
of the bad joints.
The next two pictures show cross sections of similar joints. Again, note the
fillet formation and the contact between the lead and land on the good joint. On
the bad joint, note the excess solder on the shoulder and inside the lead. Also
note the space between the lead and the land, which is approximately 4 mils.
For the solder paste and reflow conditions used, the surface tension could not
offset the wicking action. The wicking action is enhanced by temperature non-
uniformities (i.e., the lead is hotter than the land during the initial reflow action).
Both joint exhibit good wetting, as evidenced by the small angle between the
wetted surfaces and by minimum formation of voids.
With infrared reflow processing, the wicking action is different. Since IR
energy is absorbed primarily in the PCB and is reflected by metal, the PCB gets
108 TECHNICAL CONSIDERATIONS IN VAPOR PHASE

Figure 3·12 Scanning electron microscope photo of good and bad solder joints. (Courtesy
of Texas Instruments, Austin, TX)

Figure 3·13 Cross section of a good solder joint. (Courtesy of Texas Instruments, Austin,
TX)
 REFlOW PROCESSING 109

Figure 3-14 Cross section of a bad solder joint. (Courtesy of Texas Instruments, Austin,
TX)

hot first. This in tum heats the lands first, melts the solder paste, which then
wets the leads. The temperature nonuniformity is favorable to reduce the wicking
of the solder up the lead, but there can be significant nonuniformity in the plane
of the PCB. This can cause wicking laterally along any solder coated metalli-
zation. This is controlled somewhat by the design technique. 1
In summary, the key process attributes to maximize yield are coplanarity
between components and the substrate, high surface tension solder, minimal
"boiling" action of the solvent and flux in the solder paste, temperature uniformity
between the component lead and the land, and wetting action of the solder that
favors the land first, then the lead.

3.8 REFLOW PROCESSING

The solder reflow process can be achieved by using almost any available practical
heat source. Hot air guns, hot plates, small soldering irons, and toaster ovens
(and even a kitchen stove oven but only once for safety sake) have been used
to build small experimental or prototype surface mount assemblies. The maxi-
mum temperature and time are controlled manually. While this is useful to get
an initial familiarization, it will not provide the quality and reliability needed
for a high volume commercial product. To do this will require a repeatable,
controllable, and cost effective heat source.
110 TECHNICAL CONSIDERATIONS IN VAPOR PHASE

It is important that the reflow time-temperature profile have a sufficiently slow

warm-up to evaporate the solvent safely without a massive solder ball problem.
This is usually handled with a bake or a preheat step prior to VPS reflow or
with the first plateau near 100°C with IR. Higher viscosity solder paste (higher
metal content) will have less solvent to evaporate, making this step less critical.
The flux activation level of the solder paste should be as mild as possible to
minimize clean-up problems. Thus, potential contamination problems should be
solved at the source, not at the soldering step. With this in mind, the flux
activation step of the reflow profile should be appropriate for the amount and
type used in the solder paste. Some large assemblies require a long convectionlIR
warm-up step to minimize the min-max temperature differences on the PCB
during reflow, such that the flux can be activated and used up before the solder
starts to melt. When this happens, there is no wetting action and the solder does
not flow together to form a joint.
Excessive temperature, with respect to the individual components used can
cause low yield and/or reliability problems. Inadequate temperature could pro-
duce unreflowed solder joints. Achieving a time-temperature profile for safe and
complete reflow is an individual effort for each assembly and solder paste com-
bination. Each component is sensitive to some degree to the temperature ramp-
up and to the maximum temperature. The nonuniformity of absorption and mass
make it difficult to find the correct power settings and belt speed in an IR system
that will produce acceptable maximum temperature and uniformity control across
the PCB to prevent problems with reliability and solder paste.

3.8.1 Vapor Phase

Vapor phase reflow, as discussed previously, is perhaps the safest and easiest
to use. The formulation of the fluid controls the reflow temperature and, if the
equipment is properly adjusted and maintained, the process is extremely re-
peatable. Two types of equipment exist, a top loaded system for batch processing
and an in-line system for automated manufacturing. The batch system is an
excellent choice for prototype development and low volume production of various
sizes and shapes of SMT assemblies. The in-line system will work well for all
high volume production requirements, regardless of size, shape, mass, or choice
of components.
The time-temperature profile of an in-line vapor phase system easily meets
most reflow requirements. There is generally a smooth transition from room
temperature to reflow temperature and back to room ambient. By adjusting the
preheater and belt speed, this transition can be easily controlled. The belt speed
should be adjusted to provide adequate time for reflow to occur for the largest
assembly to be processed. Because of the broad process range provided by vapor
phase, this is easy to achieve.
 REF LOW PROCESSING 111

VAPOR CONDENSING COILS

,/

FLUID

Figure 3-15 Cross section of an in-line vapor phase system.

A cross section of a typical continuous belt system is shown in Figure 3-15.

Note the small slope required on both the entrance and exit to control the vapor
region. This can cause large components to shift slightly while exiting the reflow
region, depending upon the amount of surface tension of the solder joints and
the weight of the component. This will generally only be a minor visual concern
for a 68 lead PLCC.
A common complaint regarding the operation of vapor phase systems is the
cost of the perfluorinated fluid. The initial cost, plus replacement cost of fluid
lost to evaporation and surface adsorption on the assemblies as they move through
the reflow area can become a significant, and easily identifiable, part of the
operational cost of the manufacturing area.
Vapor recovery systems have recently been developed to control the exhaust
vapors and to reduce this fluid loss. To obtain the advantages of the inherent
control and flexibility of the vapor phase process, one must exercise care to
select the proper size and type of equipment and to maintain it according to the
manufacturer's specifications.
The simplest process to control, and hence to optimize, is VPS. Even if not
fully optimized, the worst that generally happens is a few more open solder
joints that can be easily touched up. Table 3-1 gives the steps that can be taken
to minimize the solder wicking and open joint problem. If the maximum reflow

Table 3-1 Vapor Phase Soldering Optimization

Action Benefit
Prereflow bake Reduce boiling and splatter of solvent before reflow
Preheat Partial bake, reduces temperature nonequilibrium
Sn62 (2% Ag) Silver additive increases surface tension of solder
Two-part alloy paste Produces a delay in melting and wetting of solder
SMOBClHASL * PCBs Contamination free solder, wets at lower temperature
·Solder mask over bare copper/hot air solder level.
112 TECHNICAL CONSIDERATIONS IN VAPOR PHASE I

temperature is matched to the solder alloy, generally 215°C for Sn62 or Sn63,
and the time is sufficiently long, approximately 60 sec, very few reflow related
problems occur. Most defects will be the result of component or pick-and-place
problems.

3.8.2 Infrared

While the vapor phase provides inherent control based upon the formulation of
the fluid, the necessary set-up and control of the IR reflow system must be
provided manually or by machine. The flexibility for individual control of the
top and bottom heat source, and thus the zone temperature, also provides an
opportunity to provide too little heat and under-reflow the solder causing a weak
joint or to overheat the assembly causing heat related damage to the PCB or
components. If properly controlled, however, the yield can be slightly higher,
under worst-case conditions, due to a reduction in open solder joints.
The typical IR reflow profile can easily meet the reflow requirements given
earlier. First there is a rapid preheat zone where the board is heated to approx-
imately lOO°C. The second zone provides more convection heat than IR, causing
an effective "plateau" in the profile. This allows the temperature differential to
partially equalize (Fig. 3-5) and evaporate the solvent. The third zone heats the
board to just below reflow temperature, activating the flux. The fourth zone is
another equalization step to minimize temperature differentials. The fifth zone
heats the assembly to the reflow temperature, followed by a cool down step.
This profile has been observed to give the best yield and reliability by reducing
the temperature nonuniformity between the hottest and the coolest points.
By reducing the belt speed, and correspondingly the source power to achieve
the correct zone temperature, the heat retained in the surrounding walls and the
belt of the IR system will provide a small amount of convection heating. This
will reduce temperature variations between different components due to mass
variations and differential absorption of the range of wavelengths of the IR
radiation.
Profile setup and adjustment can be achieved by several techniques. Most
commonly, a thermocouple and a chart recorder are used along with an assembly
that represents the production unit. Thermocouples should be placed at the comers
of the board as well as in the middle. Maximum temperature should not rise
above 220°C, and the variation over the board should be less than lOoC. A silicon
diode in an IC on the test assembly can be used to sense the temperature inside
the package or as an alternative to a thermocouple. This temperature will typically
be a few degrees less than the area near the IC package lead.
Type I B assemblies can be reflowed if a proper mechanical support is provided
during the second reflow step. This support should prevent any physical contact
to the components on the bottom of the assembly. It is not necessary to use two
different solder alloys (different temperatures) since the belt in an IR oven runs
 SMT RELIABILITY 113

Table 3-2 Infrared Reflow Soldering Optimization

Action Benefit
Uniform mass on PCB Minimize temperature differential across PCB
Large component on corner Absorb more heat from hotter area
Metal border Reflect IR, spread heat from hotter edge area
Low temperature and slow Reduce rate of IR energy absorption
.belt speed
Increased convection heat Surface heat to minimize temperature differentials

tlat and relatively smoothly and does not cause any vibration problems. The
same time-temperature profile can be used to retlow both sides, although the
heat source may need to be increased (or the belt speed reduced) for the second
retlow to compensate for the additional component mass.
The techniques of optimization for an IR retlow process are shown in Table
3-2. These are primarily related to the design phase but will allow more latitude
once the manufacturing process has started.

3.8.3 Cost Comparison

Sometimes IR, when compared to vapor phase, is given too much credit for cost
reduction of the manufacturing operation. With vapor phase the cost is direct
and identifiable. With IR, the cost is indirect and possibly not identifiable with
IR operation. If the assembly production is typically a low volume and high mix
operation, there will be some lost production time for profile changeover and
checkout. This lost time and labor cost is usually not tracked as an IR cost. If
the setup is not properly performed, there will be poor solder joint formation or
damaged boards and components, both a latent reliability hazard. These could
result in very large expenses, much greater than a year's supply of vapor phase
tluid. In summary, if vapor phase is not fully optimized, there will be a few
open solder joints that can be easily reworked. If IR is not fully optimized and
controlled, whole production lots of boards must be scrapped.
If proper controls are maintained, the cost of the two operations (at the bottom
line) are approximately the same for a low volume-high mix operation. 5 For a
high volume-low mix, where the profile is setup within the IOoe min-max
requirement, is rarely changed, and is monitored frequently, IR can be more
cost effective.

3.9 SMT RELIABILITY

There are three areas of concern with respect to SMT retlow solder processing
and the interaction with reliability. First there is the ramp-up of temperature
114 TECHNICAL CONSIDERATIONS IN VAPOR PHASE

versus time. Next there is the "integral" of the temperature and time effects.
Thirdly, there is the impact of maximum temperature of the process on the
components. Each has its potential detrimental effects and will be discussed in
turn.
The surface mount solder reflow process takes the board and components from
room temperature (25°C) to soldering temperature (215°C) in approximately I
min (Figures 3-5 and 3-7). Then the board is held at reflow temperature for 20
to 50 sec, depending on the mass of the assembly. The thermal cycle stress and
the maximum temperature are certainly higher than normal operation, but they
are within the capabilities of the basic structure. For example, consider the
thermal stress of the temperature ramp upon entering the reflow system. The
temperature rise is given by:

60 sec 60 sec

The vapor phase profile curve has an "s" shape, so the worst-case slope will
be midway between the extremes. That value is typically 6 to 8°C/sec, with a
desired upper limit for some ceramic-based components of less than 4°C/sec.
This can be quite detrimental for multilayer ceramic chip capacitors.
For ICs, it is not as critical. This thermal cycle stress can be compared to that
given in the IC tests reported in typical reliability testing. There the components
are transferred from "hot ambient to cold ambient" in approximately 3 to 4 sec.
Thus the temperature cycle gradient is:

4 sec 4 sec

Thus during IC qualification testing, the value is an order of magnitude worse

and is repeated 1000 times with a component failure rate of less than 1 percent
defective. Thus the inherent strength of the plastic SMT IC package is sufficiently
reliable to withstand the thermal ramp.
The next concern is the amount of time that the IC components spend at the
soldering temperature. As stated above, this can be as much as 50 sec for large
assemblies. Further, there may be several cycles for the double-sided board and
a couple of repair cycles.
The IC experiences several time-temperature variations during its lifetime.
The plastic package is molded at approximately 300°C for a few seconds. Af-
terward, it is cured for several hours at approximately 175°C. If the electronic
assembly were subjected to a fire, the flame retardant constituents of the mold
compound would start to break down at 300°C in a few seconds. This would
extinguish the fire but would be very detrimental to the IC.
 SMT RELIABILITY 115

Solder dipping of component leads, which sometimes also includes immersing

the body in molten solder, has been done for many years with no observed
significant component degradation (assuming corrosive flux is not used). This
takes a few seconds at 260°C. Further, type III surface mount assemblies have
been successfully produced for several years, which also take several seconds
of immersion in the solder wave.
We know from long experiences with electronics that there is a temperature
that will allow ICs to last a long time. Further, if the temperature is too high,
there will be an excessively high failure rate. All of the information above, plus
the simplistic (and safe) assumption of a decaying parabolic (or exponential)
relationship between time and temperature, is used to construct the generalized
plastic component degradation curve shown in Figure 3-16.
Note that the total elapsed time for type I processing fits safely under the
curve, even for many reflow cycles. Thus, in general, the plastic IC package
suffers no significant damage during the surface mount reflow processing from
either thermal ramp or cumulative temperature exposure.
One of the more important problems in the SMT industry lately has to do
with the maximum temperature achieved. 6 The plastic body of the IC has a TCE
of approximately 16 ppm;oC below the glass transition point (approximately
160°C) and about 50 ppml°C. This generates very large stresses inside the package
;•• 'l-te maximum soldering temperature and depending on how these stresses are
managed, internal delamination or plastic package cracking can result.
The major factors in determining the internal stress and its resultant action
are plastic mold compound material, mold compound curing, mechanical design
of the package, silicon chip size, and absorbed moisture. A search for a low

300 ....-.-_- FRMC BREAKDOWN

POINT
UNSAFE REGION
S2. 250 ~""r---SOLDER DIP LEADS
w
a:
~

!;t
a: 200
w
0..
~
W
I- 150
SAFE REGION

100
o 5 10 15 20 25 30 120 240
TIME (MIN)

Figure 3-16 Generalized plastic component degradation curve.

116 TECHNICAL CONSIDERATIONS IN VAPOR PHASE

stress mold compound and its proper time-temperature cycle for curing to min-
imize internal stress buildup during SMT reflow temperatures has been the subject
of an intense study over the last few years. Improvements have been made, but
there appears to be a practical physical limit. Internal package design should be
such that stress build up is minimized or controlled. Perforations or indentations
of the silicon chip mount pad can help control the plastic expansion. The silicon
chip size is limited by the electronic function desired and generally cannot be
reduced just for the benefit of package reliability.
Moisture absorption in plastic packages has received a lot of attention over
the last couple of years. It is really a minor contributor to this problem, but,
unfortunately, it is the only variable that the user can control. By using a drying
step just before reflow, a marginally bad package can be improved into a mar-
ginally good package. Since this is the only alternative to the user, it can be
used for its pragmatic benefit until sufficient improvements can be made in the
mold compound and mechanical design. Baking or drying is generally beneficial
with the larger PLCC/PQFP (1 in. or larger) packages with silicon chips larger
than 0.300 in. on a side. Using single IC reflow soldering with a "hot bar" or
"hot collet," rather than mass reflow soldering, is another method to minimize
package damage.

REFERENCES

1. Hutchins, C. Surface Mount Technology-How to get started. Sugar Land, TX: C.

Hutchins & Associates, 1989.
2. McLellan, N., and W. Schroen. 1987. "TI Solves Wicking Problem." Circuits Man-
ufacturing, September 1987.
3. Hutchins, C., and J. Wallace. 1985. "Surface Mount Soldering Yield Improvement
Test." Fifth International Electronic Packaging Conference Proceedings, October
1985.
4. Klein Wassink, R. J., and J. A. H. Gerven. "Displacement of Components and Solder
during Reflow Soldering," Soldering and Surface Mount Technology, February 1989.
5. Hutchins, C., and P. Marcoux. 1988. "Optimizing the Vapor Phase and IR Reflow
Processes." Electronic Packaging and Production, February 1988.
6. McKenna, R. "Surface Mount Device Package Cracking: An Overview." Proceedings
of the Technical Program, Surface Mount '89, August 1989.
 4
Optimizing the Wave
Soldering Process
Janet R. Sterritt

Wave soldering is a very effective technique for joining components to printed

circuit boards (PCBs) in large quantities. A printed circuit assembly (PCA)
consists of the printed circuit board, connectors, and all components. As in all
metallic joining applications, there are three basic steps to the soldering process.
The first step prepares the surface for soldering. The next step applies molten
solder to the bottom side of the assembly. The final step consists of allowing
the joint to solidify by cooling. Defects that result in joint reliability issues can
result from all of these stages. The goal of wave soldering process optimization
is the reduction of joint defects. Several wave soldering process hardware en-
hancements are available to reduce certain types of defects. The process param-
eters should also be optimized to truly capitalize on the machine investment.
Some examples of process parameters are solder temperature, solder wave height,
conveyor speed, and preheat temperature setpoints. The optimal wave soldering
machine and wave soldering process depend closely on the peA design intended
for production. Both wave soldering process parameter identification and process
hardware selection must be examined to optimize the total process and reduce
solder joint defects.

4.1 BASIC WAVE SOLDERING PROCESS OVERVIEW

The first step, surface preparation, consists of two phases: cleaning all of the
metallic surfaces that have been designated for soldering and bringing these

117
118 OPTIMIZING THE WAVE SOlDERING PROCESS

surfaces to the proper temperature to accept solder. Cleaning is generally ac-

complished by chemically etching the surface with an acidic compound such as
flux. This step removes any oxidation or dirt that may have collected on the
board prior to the wave soldering process.
The next phase in surface preparation heats the board. Heating accom-
plishes two tasks: flux activation and a reduction in temperature between the
printed circuit assembly and the solder wave. Most fluxes require heat for
thermal activation to break down the oxides. Manufacturers' directions specify
the amount of heat and time needed for activation. The second role in pre-
heating the board and components is that of bringing the PCA temperature to
near solder temperature. Solder is typically used in wave soldering applica-
tions at a temperature of 5000 P (260°C). If a board and components at room
temperature enter the solder wave that is at a much higher temperature, the
instantaneous thermal increase could cause differential thermal expansion re-
sulting from the thermal expansion rates of the differing materials used inside
components. Surface mount technology (SMT) devices are more susceptible
to this expansion differential than leaded through-hole (LTH) devices, al-
though both types of components can exhibit the related defect. Differential
expansion related damage can cause the component to randomly fail in the in-
line test stage or later as the PCB is being used. Poor solder joints can also
be created if the thermal imbalance between the board temperature and solder
wave is too great. In this case the solder thermal energy is used to bring the
board to the wave temperature rather than to ensure that good wetting is oc-
curring. The poorly soldered defects in this case look like cold joints or joints
with rough or dull surfaces. These joints have a higher tendency to crack when
the board is being used in the field since hairline cracks in a rough surface
can later develop into cracked joints. Differential heat expansion between me-
tallic surfaces and the PCA create a randomized soldering process that is eas-
ily overcome by a well-controlled preheating section.
The next step in a basic wave soldering process is the application of the solder
to the PCA via the solder wave. The PCA is transported through the wave. This
fully immerses the component leads or surface mount component and pads on
the bottom side of the board in solder. The cleaned metal surfaces form a metallic
bond with the solder, creating the soldered joint. As it exits the wave, the solder
joint is in molten stage. Excess solder flows off the joint back into the wave.
The final step is joint solidification, which is achieved by allowing the joint
to cool. Any vibration to a nonsolidifiedjoint can cause surface roughness, which
can result in cracks that cause reliability problems. Therefore, a wave soldering
machine conveyor often extends a good distance beyond the solder wave to
ensure that the solder joint temperature is below the liquid to solid solder tran-
sition temperature of 361°P (183°C) , prior to removing the board from the
conveyor.
 WAVE SOLDERING PROCESS HARDWARE 119

4.2 WAVE SOLDERING PROCESS HARDWARE

The soldering process is actually performed by four subsystems. Board prepa-

ration is accomplished in two separate subsystems, flux application and preheat.
Soldering is performed in a solder delivery system. Joint solidification in a basic
wave soldering machine is performed by allowing the board to cool naturally
while still in the conveyor. Therefore, the conveyor is in essence the cooling
system. A complete wave soldering machine is shown in Figure 4-1.

4.2.1 Fluxing

Flux is formulated as active solid particles suspended in a liquid alcohol base.

Key issues germane to fluxing performance are ensuring bottom surface board
coverage and cleaning issues. Flux is used to remove the oxides from the metal
surfaces so that the solder can form a better bond with the metal surfaces under
the joint. There is no need or desire to apply flux to the top side of the board
since top side flux residue will only result in a cleaning issue. The primary reason
to clean a printed circuit board is to remove flux residue. Certain types of flux
residues can continue to etch the board surface after the board is in use. The
residues have also been found to act as a host for the growth of mold in humid
climates. Both can result in PCB field failures. Therefore, the objective of fluxing
is to apply an even coating to the bottom of the board with minimal excess flux.

4.2.2 Fluxers

There are three commonly used types of fluxers: foam, wave, and spray. A foam
fluxer operates by pushing air through a diffuser to create bubbles that then rise
through a chimney to form a flux head. The diffuser is typically a porous plastic
or pumice stone. Foam fluxing is dependent on surfactants that are added to the
flux enabling it to foam. The flux head is capable of being adjusted over a range
of depths. The process parameter that can be used to control a foam fluxer is
air flow. Air flow or pressure adjusts the amount of flux aeration as it passes
over the flux stone, causing the depth of the flux head to vary. Air pressure can
also cause the size of the flux bubbles to change. Higher pressures force the air
through the flux at a more rapid rate, resulting in smaller bubbles. Conversely,
lower air pressure can yield larger bubbles. Large bubbles imply that there is a
noticeable air space in the center of each bubble. The air space results in an
area of the printed circuit assembly being exposed to air and not flux; which in
tum results in an uneven flux deposition. Small bubbles are the preferred for-
mation for a good foam fluxer. Figure 4-2 shows a typical foam fluxer in position
120 OPTIMIZING THE WAVE SOLDERING PROCESS

Figure 4·1 Wave soldering machine. (Courtesy of Ho"is Automation)

 WAVE SOLDERING PROCESS HARDWARE 121

Figure 4-2 Foam f1uxer. (Courtesy of Hollis Automation)

in a soldering machine. This particular fluxer has black brushes in the photograph.
The brushes enable better foam head support.
A wave fluxer propels a large amount of flux up a chimney to create a wave
of flux . A flux wave can be adjusted over a wide range of depths. Since the
bottom side of the printed circuit assembly is submerged in flux, wave fluxers
tend to be capable of applying more flux to the board than foam fluxers. Aux
used in a wave fluxer does not require the addition of a foaming agent, which
can be an advantage when using certain types of flux. Some fluxes containing
a heavy foaming agent will leave additional residue on the board after soldering.
This residue can cause additional cleaning issues. The process parameter that
controls wave flux application is pump speed, which is used to adjust wave
height. A wave fluxer is shown in Figure 4-3 .
A spray fluxer, as shown in Figure 4-4, is used to thinly coat the printed
circuit assembly with flux. An oscillator is used to inject flux into an air stream,
which is then diverted to form a spray pattern onto the peA. This method applies
the least amount of flux onto a board surface. The process parameter in spray
fluxing adjusts the amount of flow from the sprayer. Specific gravity is checked
when first loading the flux reservoir but not required for routine checking. A
122 OPTIMIZING THE WAVE SOLDERING PROCESS

Figure 4-3 Wave fluxer. (Courtesy of Hollis Automation)

spray fluxer draws from an enclosed tank; flux that is not used does not recirculate
back to the flux reservoir. Unlike the open tank used in wave and foam fluxers,
the closed reservoir greatly reduces the evaporation rate.
Additional process enhancements are available for the form and wave fluxing
subsystems. A fluxer air knife uses compressed air at ambient temperature to
wipe excess flux off the printed circuit assembly after it has been coated by the
fluxer. Some of the excess flux can drip from the board onto the preheat module
as it travels through preheat on the conveyor. This increases preheat maintenance
time since it requires cleaning. Most of the excess flux is baked on the board in
preheat and not removed in the solder wave. The baked-on flux remains on the
board as a gummy residue that is difficult to clean. Generally, flux containing
higher solids will yield a flux residue by-product that should be cleaned. Low
solids or no-clean fluxes yield significantly less residue issues. Fluxer air knife
angle and air pressure become two more process parameters that need to be
included in the optimization process.
Another foam or wave fluxer feature is closed-loop specific gravity control.
Since both foam and wave fluxers reside in an open flux reservoir, the flux in
the reservoir is subject to evaporation. As the alcohol in the flux evaporates, the
solids content in the flux changes. The change in solids content can be monitored
 WAVE SOLDERING PROCESS HARDWARE 123

Figure 4-4 Spray fluxer. (Courtesy of Ultrasonic Systems Engineering)

through measuring specific gravity. As the specific gravity increases, thinner

needs to be added to lower the solids content. Closed-loop control consists of
using pumps to add flux or flux thinner to the flux reservoir as the specific gravity
changes over time. Closed-loop control enables specific gravity to be used ef-
fectively in a parameter optimization.
124 OPTIMIZING THE WAVE SOLDERING PROCESS

4.2.3 Fluxer Measurement Parameters

The parameters crucial to establishing good fluxing characteristics are specific

gravity, air pressure for foam fluxers, and pump speed for wave fluxers. Specific
gravity is most commonly measured by taking a sample of flux, placing it in a
graduated cylinder, and placing a hygrometer in the cylinder. The hygrometer
is a buoy that floats to the level determined by the specific gravity. An automated
system consists of bobs that float in a flux tanle The bobs display a relative
displacement, which is translated into specific gravity. The displacement is read
by a sensor and displayed. Closed-loop control systems can add either flux or
thinner to adjust concentration and specific gravity to the desired number. The
desired level is dependent on the type of flux used, varying according to the
manufacturer's specification. The optimal flux specific gravity is controlled to
maintain the value that the manufacturer has specified.
Foam fluxer air pressure regulates the size of the bubbles in the flux head and
the height of the flux head. A sensor and pressure regulator mounted in the
incoming air line enable measurement and adjustment, respectively. The air
pressure is adjusted by placing a setup board in the conveyor with the lead edge
immediately over the flux head. The conveyor is then stopped. The air pressure
is increased until the foam head extends up to a level that is two-thirds of the
way up the thickness of the board. The air pressure level that corresponds to
this point is the setpoint. As a further aid in setting up the process, a glass board
can be substituted for a regular board. The glass lends itself to estimating flux
head bubble size for fine air pressure adjustment as a visual verification of flux
bubble size when in contact with the board surface.
Wave fluxer height is measured as a function of pump speed. In a manual
mode, this is achieved by using a strobe to determine the revolutions per minute
(rpm) of the pump. In an automated system, a tachometer is placed on the pump
and the value is displayed. The wave fluxer height is found in a method similar
to the one used for foam fluxers. A setup board is placed in the conveyor with
the lead edge immediately over the 'peak wave height position. The conveyor is
stopped. Pump motor speed is increased until the wave extends up to a level
that is two-thirds of the way up the thickness of the board. The pump speed that
corresponds to this point is the setpoint.

4.2.4 Fluxer Optimization

The greatest challenge in fluxing is ensuring that all metallic surfaces are covered
while using a minimum amount of flux. Excess flux results in cleaning issues
after the wave soldering. The type of PCA product being manufactured deter-
mines the fluxer requirements. Key board descriptors that influence this decision
are board depth, complexity, and line throughput rate. Board depth is the distance
 WAVE SOLDERING PROCESS HARDWARE 125

Table 4-1 Fluxer Hardware Optimization

Printed Circuit Board Optimal Wave Soldering
Factors Process
Maximum depth: short Foam fluxer
Board complexity: high or low
Line speed: low to medium
Maximum depth: short or long Spray f1uxer
Board complexity: low
Line speed: low to medium
Maximum depth: long or short Wave fluxer
Board complexity: high or low
Line speed: low to high

from the bottom board surface to the longest or highest lead or component,
respectively. Board complexity is a metric used to describe the combination of
shadowed components and solder joint density. Highly complex boards have
large areas that are shadowed by other components and/or very closely spaced
leads and pads.
Wave fluxers are used most often with longer leaded devices since the wave
depth can be extended to a deeper range than a foam fluxer. Wave fluxers are also
capable of depositing a good layer of flux at very high throughput speeds (Le., in
the range around 10 ftlmin.). A foam flux head can maintain an adequate depth
for most applications. Wave and spray fluxers are used when flux additives are a
concern. The spray fluxer provides a very thin coating of flux, but shadowing can
occur with some board layouts. A spray fluxer does not gain any benefit from
closed-loop control since the reservoir is sealed and little or no evaporation takes
place. Since the spray fluxer deposits very little flux on the board, a fluxer air knife
is not required either. Some excess spray residue can collect in the machine ex-
haust and other wave soldering machine parts over time, increasing maintenance.
All of these factors need to be considered when selecting fluxer hardware. A quick
reference for making these trade-offs is provided in Table 4-1.

4.2.5 Preheating

Preheat brings the printed circuit assembly temperature to a level near wave
solder temperature. The key to obtaining good preheat characteristics is raising
board temperature with minimal variation across the board or over the length of
time for a production run. Heating issues include thermal shock, color sensitivity,
and thermal variation. Thermal shock is the rapid absorption of a large quantity
of heat in a very short space of time. This can cause differential expansion within
126 OPTIMIZING THE WAVE SOLDERING PROCESS

components, which can result in stretching internal component connections and

thus lead to component failures in the field. Color sensitivity is exhibited when
a component encased in a black covering heats more rapidly than one encased
in white. Thermal variation is the heating envelope provided by the wave sol-
dering preheating section, shown as hot and cool areas on the same board. This
results in a board that has various ~mperatures over the surface of the board as
it is immersed in the solder wave and causes variations in the surface appearance
of solder joints since the wave is used in a differing manner across the surface
of the board. In cooler areas, the wave is actually being used as a final preheat
station, while in areas that are too hot, the flux may have been evaporated off
so that some oxidation can occur prior to entry into the solder wave.

4.2.6 Preheaters

Preheating can be accomplished by several different techniques. Infrared heat

and convection are the most common techniques. Infrared devices are divided
into three types: short, medium, and long wave length. Short wave length is
defined as 1 to 3 IJ.m. Medium wave length extends from 3 to 5 IJ.m, and long
wave length is defined as above 5 IJ.m. Printed circuit assemblies heated with
short wave length radiation tend to heat components unevenly in a color-sensitive
pattern. Color sensitivity is not characteristic in medium- and long-range wave
length units. Long-range wave length sources tend to require more time to
stabilize at temperature, yielding longer process setup time. Medium-range wave
length infrared sources are the most commonly used preheat method. A typical
infrared preheater module with medium wavelength quartz lamps is shown in
Figure 4-5.
Strip heaters are black body radiators and fall into the long wave length infrared
heater category, most commonly used in strips, rods, or serpentine shapes. When
used as an exposed rod or serpentine shape, this method can give a preheat
pattern that has noticeably hot and cold spots. The spotting is caused by the heat
distribution from the source. The amount of heat delivered is reduced by the
distance away from the source. In a serpentine-shaped heater, for example, the
region near the curved ends has a higher concentration of heat and is therefore
hotter than the straight rods that connect them. A black body radiation source
that is implemented in contiguous pieces would not exhibit this problem. Black
body radiator preheaters are also assembled into self-contained units with a quartz
or glass top. In this configuration, the preheating module is referred to as a panel
heater. The cover provides a thermal barrier that makes the resulting heat pattern
more uniform across the module since the barrier acts as a method of capturing
and controlling convective heat from the sources along with infrared heat. Ce-
ramic plates that have heating coils laced through the plate also fall into the
 WAVE SOLDERING PROCESS HARDWARE 127

Figure 4·5 Infrared preheat module. (Courtesy of Hollis Automation)

panel heater category. Strip heaters and panel heaters require time to warm up
and stabilize prior to usage.
Convective preheat is achieved by heating the air surrounding the board, which
in tum heats the PCA. This method gives very even and thorough preheating;
however, convective heating alone is a very slow process. Convection occurs
in combination with all of the previously described process techniques. Preheaters
not only heat the printed circuit assembly but also heat the air in the region
surrounding the module. Some wave soldering equipment capitalizes on this
effect by surrounding the preheat subsystem with a tunnel as a means of com-
bining convective heat with infrared heat. Convective heating used in combi-
nation with other methods helps to make the preheat process more uniform.
In the actual use of all preheat technologies, individual preheaters are assem-
bled into larger modules that are in tum controlled as one set. The modules can
then be configured in different patterns, where one module represents one zone
of preheat. Generally, each separate pattern defines a distinct wave soldering
machine. The most common configurations used consist of mounting varying
lengths of modules on the bottom side of the conveyor only or mounting preheater
modules to both the top and the bottom sides.
The most significant preheat process additions are closed-loop control and
zone adjustable preheat. Closed-loop control is achieved by placing a thermo-
128 OPTIMIZING THE WAVE SOLDERING PROCESS

couple in each preheat zone. Thennocouple placement depends on the preheating

method: above the zone for convective preheaters or combination convective
and infrared preheaters, in an infrared bulb, or in a panel. The preheat module
is designed to provide heat to the board over a length of time. In practice, the
production line throughput rate specifies the amount of time. Therefore for
different mass boards varying lengths of preheat are needed to be able to fully
heat the board within the time constraint. The length is further divided into
separate preheat modules that are individually controlled preheat zones. Preheat
modules may be configured both on the top and the bottom sides of the conveyor
to give more even heating to both sides of the board.

4.2.7 Preheat Measurement Parameters

Preheat process parameters are zonal temperature setpoint, temperature unifonn-

ity, heating rate, preheat exit temperature, and delta T. The difference in tem-
perature between preheat exit and the solder wave temperature that the printed
circuit assembly experiences as a rapid change is the change in temperature or
delta T. The preheat exit temperature is the average board temperature at preheat
exit, which is the maximum board temperature prior to entering the solder wave.
The heating rate is the slope of the thennal heating rate curve that the board
experiences. Temperature unifonnity is the maximum allowable variation in
temperatures that occur over the surface of a peA. The zonal setpoint temper-
atures are the setpoints for each individual preheat zone.
A thennal profile is shown in Figure 4-6. The profile is achieved by attaching
a thennocouple to the bottom surface of a board. The profile is a graph of
temperature over the length of time in the conveyor. The profile shows the
gradual rise as the board passes over preheat. Immediately upon preheat exit,
the board enters the first wave. Then it cools as it passes in the space between
waves, rapidly rising again as it enters the laminar wave. After exiting the wave,
the board displays a nonnal natural cooling effect. The profile is superimposed
over a block diagram of the machine subsystems. This machine has 10 zones
of preheat between the fluxer and solder waves. The profile also shows a further
warming level at the air knife position. In the example, the temperature ramp
rate is the slope of the line plotted on the graph. Delta T is the difference between
the solder temperature of 500°F in this case and the maximum temperature at
the exit of preheat.
Temperature uniformity is found by using a multiplicity of thennocouples and
observing the collective results. Some key locations are top side surface tem-
perature, inside bottom side fillets, under components, and in or near heat sinks.
Thennocouple placement is determined by board layout and component selection
issues for each individual board. The result of using several thennocouples
simultaneously is shown in Figure 4-7. This unifonnity profile is measured by
 WAVE SOLDERING PROCESS HARDWARE 129

217° ______________- ,
74°C (delta T)
•
-=::::::: 227"
153°-------.

Zone 6
41 e'f'lIi

Machine
Entrance

.. ..
Preheat Profile

~ ~~~1 '~~\Ei'@J
Jj I
I J
--==~ Zone 5 Zone 7 ZoneS

.I ~-. : ...'"
I Zone 1
'~Zone3
Zone2
HSS Zone 4
S2 Zone9 ;83 -

I L - 61 Solder Pot ~
L - ~er HPAK
Figure 4-6 Preheat profile in relationship to subsystems. (Courtesy of Hollis Automation)

375 ~--------~----------~--------~----------~---------.

275 I I I I I I I I I I I I I I I I , I I I

w
a:
~
!;;:
a:
w
Il.
::;:
W
f- 175 I I I I I I I I I I I I I I I I I I I I I I I I I I I • I I I I I I I I I , , I I I I I I I I I I

75~=~J.. __-.JL__J..__..J___J
TIME IN CONVEYOR (MIN)

Figure 4-7 Uniformity profile. (Courtesy of Hollis Automation)

129
130 OPTIMIZING THE WAVE SOLDERING PROCESS

placing five thennocouples in a row perpendicular to the direction of motion.

All are reading bottom side surface temperature. The unifonnity range is the
difference between maximum and minimum temperature.
The preheat setpoints are established by using a setup board that has ther-
mocouples attached in the desired locations. The board is run at the desired
conv"yor speed and the thennal profile recorded. When the optimal profile is
obtained, the temperature setpoints for each zone are noted. These are the set-
points for preheat.

4.2.8 Preheat Optimization

When considering the many preheat options, the key printed circuit product
descriptions that detennine the preheat choice are product size, mass, and number
of board layers. The maximum amount of preheat required is a function of the
size and mass of the board. Thin, small boards require less heat to raise the
temperature to the desired level than heavy, thick boards. An overall metric used
to detennine this level in an optimization is the number of layers in the board.
Multilayer PCBs require more heat to achieve the same thennal characteristics
as single-layer boards. A good overall cutoff is four or more layers being con-
sidered to require more preheat. Production throughput goals result in line speed
requirements that also affect overall preheat length. Printed circuit assemblies
that require short preheat lengths at relatively low line speed rates of 3 to 4
ftlmin can require twice the length to satisfy the same demands at 6- to 8-ftlmin
rates.
All preheat varieties meet the overall preheat goals as long as they are capable
of providing even, non-color-sensitive heat. Single-layer or multilayer boards
of up to four layers with leaded through-hole components at relatively low
conveyor speeds are prime candidates for short length, bottom side preheaters.
A short length preheater is considered to be from 1 to 3 ft since the boards in
this category can be properly warmed in this distance at standard conveyor speeds
(i.e., 3 to 4 ft/min). Higher mass, multilayer boards that are small in overall
surface area and contain leaded through-hole components are matched to short
lower preheat with a top side preheater module to assist in overall heating. Large
multilayer, high mass boards, with leaded through-hole components are matched
to longer preheaters with both top and bottom side preheating units. Long pre-
heating length is considered to be over 3 ft. Any board containing SMT devices
is matched to long preheat length on both the top and the bottom side. Since
LTH devices are less susceptible to thennal shock, there is less emphasis on
reducing slope. The longer-length preheat can be used to gradually increase
board and component temperature. If higher conveyor speeds are desired, even
longer preheater subsystems are required. If a large, massive SMT assembly is
the identified product and a throughput requirement leads to line speeds in the
 WAVE SOLDERING PROCESS HARDWARE 131

Table 4-2 Preheater Hardware Optimization

Printed Circuit Board Optimal Wave Soldering
Factors Process
Thin LTH Short bottom side preheat
Multilayer LTH Long bottom side preheat or short bottom side with
top preheat
SMT Long bottom side preheat and top side preheat

7- to 1O-ft/min region, as much as 10 ft of top and bottom side preheat is needed

to meet these requirements.
The greatest key to optimizing performance is understanding thermal profiling.
The thermal profiling concept encompasses the preheating and soldering sub-
systems in the sense that the solder temperature defines the peak process tem-
perature. Reducing the peak temperature changes solder viscosity. Even though
lowering solder temperature reduces operating latitude from the eutectic ratio,
tight control of the tin-to-Iead solder content ratio is not required since there still
remains ample operating room between the liquidus limits.l
The longer-length machine accommodates all types of printed circuits being
manufactured; however, cost limitations in both soldering machine price and
floor space may force the use of short-length machines. Optimal profiling can
still be achieved by running at slower throughput speeds. The preheat trade-off
optimization is shown in Table 4-2.

4.2.9 Wave Soldering

The wave soldering subsystem applies solder to the bottom surface of a PCA.
It is a mass of solder that is pumped to a height above the level of a reservoir
and then cascades back into the tank, creating a moving wave of molten solder.
The expanded definition of a wave soldering subsystem refers to all hardware
that is intended to shape solder joints while still in the molten state.

4.2.10 Solder Waves

There are many forms of solder waves. The basic solder wave provides a wide
laminar flow of solder that moves in a direction opposite to the conveyor motion
at the entrance to the wave and with the conveyor direction at the exit of the
board from the solder wave. The type of solder flow lends its name to describe
each wave. A bidirectional wave parts in the center with half of the solder
flowing to the entrance and the other half of the solder flowing toward the exit.
A high volume laminar wave, which is optimized to run product at faster through-
132 OPTIMIZING THE WAVE SOLDERING PROCESS

Figure 4·8 Dual solder wave. (Courtesy of Hollis Automation)

put rates, is a bidirectional-type wave but has a disproportionately larger mass

of solder flowing toward the entrance than toward the exit. The laminar wave
gives excellent results on most leaded through-hole technology and simplistic
bottom side surface mount technology. However, component shadowing as seen
in some complex designs and deep leaded through-hole barrels prevalent on
multilayer boards can result in lack of penetration, skips, and misses.
In some product applications, a narrow highly turbulent wave precedes the
laminar wave. The turbulent wave has a high vertical velocity that can drive up
long barrels or fill in behind a large surface mount component. Used in com-
bination with the laminar wave, the dual wave is more robust in that it can easily
solder more difficult leaded through-hole technology as well as bottom side
surface mount devices. Figure 4-8 depicts a typical dual wave system with a
turbulent wave on the left and a laminar wave on the right.
A unique tool that helps to sculpt solder joints while they are still molten is
the wave soldering air knife . The air knife produces a low pressure warm air
mass , which is directed toward the bottom surface of the board upon immediate
exit from the wave. This removes excess solder from molten joints, reducing
bridges, excess solder, and icycling. Figure 4-9 shows an air knife in a soldering
machine. It is the square tube positioned immediately after the wave in the
picture.
 WAVE SOLDERING PROCESS HARDWARE 133

Figure 4-9 Air knife. (Courtesy of Hollis Automation)

Another wave soldering process addition is mixing oil in with solder in the
wave. Oil intermix reduces the solder alloy surface tension. This changes the
soldering entrance and exit velocity latitude, making these items less critical.
Opening the solder wave setup latitude enables greater ease of setup in the
production environment. The oil has proven not to cause difficulty in solder joint
formation since the oil does not impede solder application to wetted surfaces
but rather reduces the surface tension on the solder wave. Oil intermix reduces
bridging, excess solder, and icycling. Waste oil concerns offer the printed circuit
manufacturer a trade-off between improved process latitude and waste removal.

4.2.11 SQlder Wave Measurement Parameters

Wave soldering process parameters are solder temperature, solder wave height,
and wave height uniformity across the process width. Solder temperature refers
to the temperature of the solder in the wave. Since solder is a metal and the
mass of solder is continually overturned as it is pumped through the wave, wave
temperature rarely differs from solder reservoir temperature by more than a
degree. The time in solder is actually monitored by finding the contact length
of the printed circuit assembly and matching it with conveyor speed. The relative
velocity between the solder wave and the peA as it enters and exits the wave
shapes the joint. Surface roughness on the smooth wave is limited so that areas
of the board are not missed.
Solder wave height can either be measured directly via a wave height sensor
or empirically by using solder pump speed. A wave height sensor is mounted
on the wave and uses closed-loop control to monitor and control solder wave
134 OPTIMIZING THE WAVE SOLDERING PROCESS

height. Pump speed measurements and setup are performed in a manner similar
to the wave fluxer setup and measurement.
A very thin layer of dross can form on top of the wave. It appears to break,
falling over both edges. The fine top dross layer actually floats on top of the
wave. Since it is floating, it does not have the same properties as the solder in
the wave. Therefore, it is not a precise indicator.
If an air knife is used, the key parameters are distance from the solder wave
exit, distance to the board, angle to the board, air temperature, and air pressure.
The air knife should be positioned as close to the wave as product depth limi-
tations enable. The knife should also be set 1116 in. lower than the component
with the maximum depth on the assembly. The angle varies as a function of
product type and board layout. The lower angle ranges from 38° to 50° are used
for leaded through-hole boards in general. Higher angles, ranging from 50° to
90°, are generally used in SMT soldering. Extremely tight LTH components
require higher air knife angles and can extend to 90°. Air knife temperature is
optimized when set to any value in the range from 500"F (260°C) to 8000 P
(426°C) as the air exits the orifice. The high temperatures are used when excessive
board mass can deplete air temperature. The effects of air pressure are determined
by soldering some sample boards and examining the results. An air gauge is
placed in line· to measure air pressure at the knife body. The air knife is setup
by repetitively setting pressure and finding performance by running a board. A
trade-off between bridging and excess solder defects and blow holes must be
made. In general, higher pressure is used with SMT devices.
The oil intermix process control parameter is the amount of oil being fed into
the wave. This is gauged through a metering pump. When set up properly, the
oil should break the wave in an even array of very small oil droplets.

4.2.12 Wave Soldering Optimization

Soldering requirements leads to determining the need for a single laminar wave,
dual wave, or air knife addition. The factors used to determine the hardware
trade-off for the solder wave are joint type, density, and number of board layers.
The joint type is either LTH or SMT. Density refers to the spacing between
joints. As joints are more closely located, the tendency to create bridges is
increased. Single-layer boards or relatively few layer boards with leaded through-
hole technology at a low level of complexity (30 joints per square inch or less)
can be soldered effectively on a single laminar wave in a non-SMT application.
SMT devices by their design present a large surface to the wave. The leading
edge of the device shadows the trailing edge; the vertical velocity component in
the turbulent wave fills in the regions that would be shadowed in a laminar wave.
Leaded through-hole technology assemblies with long barrels for multilayer
boards require a mechanism to force solder up into the barrels. This results in
 WAVE SOLDERING PROCESS HARDWARE 135

Table 4-3 Solder Wave Optimization

Printed Circuit Board Optimal Wave Soldering
Factors Process
Low density, few layer LTH Single laminar wave
High density, few layer LTH Single laminar wave and air knife
Low density SMT Dual wave
Low density, multilayer LTH
High density SMT Dual wave and air knife
High density, multilayer LTH

the addition of a turbulent wave prior to the laminar wave. If the complexity is
further increased on a leaded through-hole board with a level of joints per square
inch higher than 30, the addition of a hot air knife is matched to the process.
All boards containing SMT devices benefit from a dual wave system with the
addition of a hot air knife; however, cost limitations force a trade-off to be made.
The solder wave trade-off optimization is described in Table 4-3.

4.2.13 Solidification

In a basic wave soldering process, joint solidification is accomplished by allowing

the PCA to cool naturally. The solder joint is molten as it exits the soldering
subsystem. The joint must not be disturbed as it cools, or surface roughness can
result. The hardware that accomplishes this task is the conveyor, by virtue of
the length after exiting the solder wave.

4.2.14 Conveyors

A conveyor can be set up for use directly with the PCA or for use with a pallet
system. A finger conveyor grips the exterior of the board to carry it through
fluxing, preheat, and soldering. The fingers contact the materials in each of these
subsystems, which results in a gradual build-up over a period of time. If the
finger residue is not removed, board jams or conveyor vibrations can result.
Conveyor vibrations can cause the conveyor to shake, which results in movement
in the solder joint as the solder cools. This can result in defective joints. The
fingers can be cleaned through routine preventive maintenance or the imple-
mentation of a finger cleaner, which scrubs the fingers after the solder wave.
Another process enhancement that can be added to this subsystem is an active
cooling module such as fans or vortex tubes. A massive PCA can take a great
deal of time to cool. If this is further complicated by a desire to run at high
throughput speeds, a very long conveyor would be required. A slight amount
136 OPTIMIZING THE WAVE SOLDERING PROCESS

of cooling using ambient air cooling fans can reduce the overall length of con-
veyor needed to allow the solder to solidify without damaging solder joints. This
enables a more favorable trade-off between overall machine length, board mass,
and conveyor speed.

4.3 WAVE SOLDERING PROCESS PARAMETER OPTIMIZATION

Once the optimal wave soldering hardware set has been identified, the process
parameters must be optimized. The subsystem parameters described in the pre-
vious sections must be set up prior to a system optimization. The system opti-
mization is primarily focused on the thermal management throughout the wave
soldering machine in preheat, wave soldering, and through the air knife.
The surfaces intended for soldering wet more easily if they are all near the
solder temperature. In a nonoptimized wave soldering process, it is possible that
the components and PCB are hundreds of degrees lower in temperature than the
solder wave. When a cooler board enters the wave, it is actually brought up to
temperature by the wave itself. This very fast thermal transition is referred to
as thermal shock. An optimized wave soldering process increases preheat exit
temperature to minimize thermal shock.
Another system interactive factor contributing to good wetting is solder con-
dition. Alloy purity strongly affects joint quality and determines the acceptable
range of fluxing required. An interesting aspect of solder condition is its tem-
perature. Typical recommended wave solder temperatures range from 500°F
(260°C) to 525°F (274°C). These temperatures are well above the melting point
of solder and are the accepted industry standards because they offer a wide
process operating range over a wide latitude of solder purity levels. The minimum
is limited to 500°F (260°C) to ensure that the solder is well above the liquidus
and solidus temperature range. However, for alloy composition near 37 percent
lead and 63 percent in (eutectic) and near eutectic, 500°F (260°C) is 139°F (77°C)
over the minimum melting point. Therefore, if solder alloy quality is maintain~d
to a level near eutectic, the minimum solder temperature can easily be lowered
by as much as 75°F (41°C) without jeopardizing good wetting conditions. Low-
ering the soldering temperature reduces the peak temperature to which the board
is subjected. Overall, this results in less thermal stress on the assembly com-
ponents.
Optimal system profiles are produced by limiting three factors: peak temper-
ature, delta T, and temperature ramp rate or slope. The main impetus for these
limits stems from SMT component manufacturers. For example, one recom-
mendation limits the slope to not more than 3.6°F (2°C) per second. Another
manufacturer recommends limiting thermal shock to less than 212°F (100°C). 2.3
Both manufacturers' limitations are based on packaging material expansion rates.
As the technology matures, recommendations may be relaxed. Even though
 WAVE SOLDERING PROCESS PARAMETER OPTIMIZATION 137

recommendations may be relaxed, process optimization from the standpoint of

improved solderability will continue to enforce these concepts.
It is important to note that all of the temperatures controlled by this optimization
are measured on the bottom surface of the board. One popular method of quan-
tifying thermal performance in wide use today is the overall temperature equiv-
alent for the top side. Typically this rule limits top side surface temperature to
a range from 180°F (82°C) to 250°F (121°C). While this is an effective overall
metric, it is not capable of delineating bottom side temperature performance.
On an average board the top side is 30°F (16°C) to 40°F (22°C) cooler than the
bottom side; however, this is greatly affected by the mass of the board, the
number and type of ground planes, the presence of heat sinks, and the soak: time
in preheat. The longer the preheat section the less temperature differential be-
tween top and bottom side surfaces. SMT components can also shadow pad
areas so that the surface to be soldered is not at the desired temperature, even
though the average bottom side surface temperature may be within the desired
limits. An actual bottom side surface measurement is necessary to properly
optimize the profile. The ideal system thermal profile minimizes all temperature
gradients throughout the process.

4.3.1 Optimization Procedure Test Study

To evaluate these theories, a test was run to compare the results of running
mixed technology product on two machines, one with hardware optimized for
LTH and the other with hardware optimized for SMT. One other more subtle
difference existed between the machines. The LTH machine was not run with
optimal thermal management while the SMT machine was. The boards chosen
for this test were both mixed technology, enabling the test to produce leaded
through-hole and surface mount results with fewer boards. However, one board
is primarily SMT while the other is primarily LTH.
Using the hardware optimization guide, the standard LTH machine consisted
of foam fiuxer, short preheat, and dual wave soldering system. The hardware
optimization process recommends lengthening the preheat and adding an air
knife to better optimize the machine for mixed technology boards. A wave
soldering machine matching the optimal description was produced and used in
a comparison test to verify the optimization process.
The test was run using commercial boards in a real production environment.
This type of real manufacturing production facility test is perfect for verifying
hypotheses and enabling comparisons. In the full production process the boards
were top side populated, refiowed, and cleaned. Then the boards were populated
by pick-and-place for bottom side SMT into epoxy. The epoxy was cured and
then populated with leaded components by pick-and-place equipment. Finally,
off-sized parts were hand inserted. Printed circuit assemblies were built and
138 OPTIMIZING THE WAVE SOLDERING PROCESS

Table 4-4 Test Machine Configurations

Parameter Standard Machine Optimized Machine
Conveyor Finger Finger
4 ftlmin 4 ftlmin
Fluxer Foam Foam
SMC grade OA SMC grade OA
Preheater I/R I/R
4 ft bottom 5 ft bottom
Oft top 2 ft top
Solder Dual wave Dual wave
500°F (260°C) 470°F (243°C)
SMC grade solder SMC grade solder
Air knife None 1.5 in. to wave
450°F (232°C) at PCA
Finger cleaner Yes Yes

processed in batch mode. Both machines used the same batch of flux and solder,
enabling a wave soldering process machine test as opposed to a soldering ma-
terials test. Boards were run simultaneously on both machines from the same
batches. The only differences between machines were the inherent system tem-
perature profile and hot air knife.
Machine configuration is described in Table 4-4. Both machines were confi-
gured similarly. Both ran at 4 ftlmin, with foam fluxers and dual solder waves.
The main differences were preheater length, solder temperature, and the addition
ofthe hot air knife in the SMT optimized unit. In order to meet the 3.6°F (2°C)
per second slope and 212°F (lOO°C) delta T rule at 4 ftlmin conveyor speed, the
bottom side preheater was lengthened to 5 ft under the conveyor and 2 ft were
added to the top of the conveyor. The top side preheater was directly over the
bottom side preheater section at the edge nearest to the solder pot. The optimized
machine soldered in a "low" temperature mode at 470°F (243°C). The air knife
was set at the same temperature for both boards, but angle and pressure were
further optimized for each. The LTH board used a 45° angle at high pressure,
12 psi. The SMT board used a high angle of 85° and low pressure of 6 psi. The
pressure differential between boards types was mainly driven by the maximum
board depth. The LTH board was 1/8 in. deeper than the SMT board.
The standard LTH machine profile is shown in Figure 4-10 and Figure 4-11
is the optimized profile. Figure 4-11 illustrates the elongated preheat thermal
profile capable of being achieved on an SMT-type wave soldering machine. Both
profiles are generated by using the same thermocouple placed on the bottom
surface of the board. Both figures show bottom side surface temperature with
solder waves on. The thermocouple was attached with high temperature solder,
effectively resoldering it in a hollowed out joint. The optimized machine meets
 WAVE SOLDERING PROCESS PARAMETER OPTIMIZATION 139

600 -------------.-------

500

400

u:-
"-
w
a:
::J 300
!;(
a:
w
0..
:::;
W
I- 200

Preheat
Entrance ' \
100

0 ..... , ....... 111. 1111,111"1,1,111."'"''"1'''''''''1',,11''''1','',11''1'''''''''''''''""I'lT\"T1"TTTfl'TfTITlTrT"TTTTI'TTT'f""lTTTTlTrpn"fT

0 10 20 30 40 50 60 70 80 90 100110120130140150

ELAPSED TIME (SEC)

Figure 4-10 Standard machine profile. (Courtesy of Hollis Automation)

600

500

400

tw
a:
::J
!;( 300
a:
w
0..
:::;
W
I-
200

Preheat
100 Entrance ""\

o 10 20 30 40 SO 60 70 80 90 100 110 120 130 140 15u

ELAPSED TIME (SEC)

Figure 4-11 Optimized machine profile. (Courtesy of Hollis Automation)

140 OPTIMIZING THE WAVE SOLDERING PROCESS

the 3.6°F (2°C) per second limit as well as the thermal shock limit of 212°F
(1 OO°C). The actual optimized rise rate is a maximum of 3.4of (1. 9°C) per
second and a maximum of 153°F (85°C) shock temperature, delta T. This com-
pares to the standard machine maximum temperature rise rate of 5.4°F (3°C) per
second and maximum thermal shock temperature of 198°F (1lO°C). The standard
machine was intentionally set to the best profile it could achieve on these boards.
The soldered boards were inspected by the same team. The boards were batch
coded and mixed so that the inspectors were not aware of the machine desig-
nation. Visual inspection focused on identifying bridges, excess solder webbing,
poor solder, no solder, blow holes, and solder balls. A bridge was defined as a
short or solder-filled space between any two leads or pads to leads in the case
of SMT joints. If, for example, two spaces between three leads were filled, two
bridges were recorded. Excess solder pertained to leaded and leadless SMT
devices as well as leaded through-hole components. For leaded components of
both varieties, excess solder was defined as a joint in which the lead could not
be clearly discerned. For leadless SMT devices, excess solder was determined
to exist when the knee of the solder fillet protruded more than half way up the
edge of the device. Webbing was defined as a solder streamer condition that
formed very fine joints between leads and/or pads. Poor solder was defined on
SMT components only as the condition when the knee of the solder fillet occurred
below the halfway mark on the edge of the device. While it was clear that some
solder may have wetted to the solder joint in a poor solder condition, the main
portion of the lead was exposed on leaded devices. No solder referred to a
complete lack of solder. A blow hole was defined as a void in the joint. Solder
balls were defined as coherently formed globules that were not attached to either
lead or pad but did attach to the surface of the board.
The test used two boards in an effort to establish the effectiveness of wave
soldering optimization on a wide range of product shown in Table 4-5. The first
board was a mixed technology with top side retlow surface mount, bottom side
wave soldered surface mount, and leaded through-hole components. It measured
2 by 8 by 0.128 in. and contained 347 bottom side joints. The bottom side
surface mount were leadless capacitors including some ceramic C-size ones,
small outline transistors (SOTs), and small outline integrated circuits (SOICs).
A total of 201 joints were surface mount and 146 were leaded through-hole.
Board 1 represented a very complicated SMT application.

Table 4-5 PCB Descriptions

LTH SMC Total Sample Size
Product joints joints joints No. of Boards
1 146 201 347 200
2 1980 22 2002 500
 CONClUSION 141

Table 4-6 Results of Comparative Processes

Standard Optimized Defect
Product Machine Machine Ratio
1 0.35 3X
2 0.07 14X

The second board was primarily leaded through-hole representing the other
side of the spectrum. It mainly consisted of leaded through-hole DIPs with a
few bottom side small chip capacitors. It measured 8 by 10 by 0.128 in. with
a total of 2002 joints, all on the bottom side. These were divided into 1680
leaded and 22 SMT joints.

4.4 RESULTS

The results for both boards are shown in Table 4-6. The defect per million (ppm)
rate has been calculated by averaging total defects and converting to a per million
joint unit basis. Since both products were assembled and run at the same time
under identical conditions, except for wave soldering, all defects not relating to
wave soldering can be assumed to be the same for all boards, making direct
comparisons possible. The ppm rates are shown normalized to the standard
machine. The normalization process enables us to see the effect of the optimi-
zation without showing the absolute value and exposing a private manufacturer's
industry secret. The ratio between the machines remains the same regardless of
the reference point.
Board type 1 showed an improvement of 3 times fewer defects on the optimal
machine versus the standard machine. Board type 2, the priniiuily leaded through-
hole board, showed 14 times fewer defects than the nonoptimal machine. This
study shows the enormous effect that optimizing process profile can have on
reducing joint defects.

4.5 CONCLUSION

Joint specifications place stringent solderability requirements on the process

physics. Surface mount components continue the trend toward finer pitches. This
increases the joint density and the likelihood of developing solder joint defects,
such as bridges. The bottom side surface mount components themselves are also
subjected to harsh environmental changes. Many SMT components cannot tol-
erate rapid dramatic temperature changes. Leaded through-hole components also
benefit from these thermal profiling trends. Basically, any surface that is brought
closest to the soldering temperature benefits by not being subjected to a large
thermal shock.
142 OPTIMIZING THE WAVE SOLDERING PROCESS

The important goal in optimizing the process is to make the best hardware
selection with respect to the expected product type first and then further optimize
the process by choosing the correct process setpoints. Wave soldering hardware
is dictated by board design and production throughput rate. These factors are
derived from two items; the first is describing the most difficult board that is
expected to be run and the second is identifying the maximum expected through-
put rate or line speed. The wave soldering subsystem hardware is selected. Then
the process parameters are optimized to achieve the lowest peak thermal profile
with the least thermal rise rate. This augments the wicking action as the board
is introduced to the solder in the wave. The study shows that wave soldering
process optimization increases yield for SMT devices and excels at increasing
performance for LTH technology. The study further shows that optimizing both
hardware and process parameters for a technology level higher than the lower
cases also benefits the lower technology. If there were no cost limitations on
purchasing hardware, it would always be more beneficial to optimize to a more
difficult product type. The effect of longer preheat, dual waves, and air knife
benefits all product types.

REFERENCES

1. Manko, H. H., Soldering Handbookfor Printed Circuits and Surface Mounting, Van
Nostrand Reinhold Company, Inc., New York, 1986.
2. AVX product literature.
3. Murata Erie product literature.
4. Woodgate, R., Handbook ofMachine Soldering, John Wiley & Sons, Inc., New York,
1983.
5. Leonida, G., Handbook of Printed Circuit Design, Manufacture, Components, &
Assembly, Electromechanical Publications, Ltd., 1981, pp. 331-463.
6. Bahr, K. E., "Solderability: A Critical Issue for Fabricators and Assemblers," fPC
Proceedings, January 1990.
 5
Post-Solder Cleaning Considerations
Leslie A. Guth

5.1 PURPOSE AND CHAPTER DESCRIPTION

The improper selection of either the soldering flux or the subsequent cleaning
materials, processes, or equipment can greatly affect circuit pack reliability. In
fact, it is necessary to choose the appropriate cleaning strategy by taking into
account the activity and the composition of the soldering flux used. Residue
remaining from either the flux or the cleaning agent or a combination of the two
can affect the reliability of the circuit board.
This chapter begins with a section on the existing specifications for soldering
fluxes and the associated test procedures for qualifying them. Following are
sections on flux composition, flux application methods, and soldering process
issues, all of which are directed toward wave soldering. The last two sections
discuss nonliquid fluxes and future trends. Throughout this chapter, the char-
acteristics of soldering fluxes and their use are described, with the underlying
theme relating to the importance or the lack thereof for postsolder cleaning.
Due to the ever-increasing environmental concerns, an emphasis is placed in
this chapter on those fluxes and associated soldering processes that require no
cleaning, especially low solids fluxes (LSFs). Again, flux selection should be
made realizing that this decision will directly affect the possible subsequent
cleaning materials and processes and their associated environmental concerns.

5.2 ENVIRONMENTAL CONCERNS

Chlorofluorocarbons (CFCs), widely used in electronics industry for cleaning,

have been implicated in the depletion of stratospheric ozone. In 1987, an inter-
national agreement, The Montreal Protocol, was signed; it went into effect in
1989 and called for a 50 percent reduction in CFC usage by 1998. In June 1990,

143
144 POST-SOlDER CLEANING CONSIDERATIONS

this document was amended based on a technology assessment and now includes
provisions for total phaseouts of CFCs and other chlorinated solvents by 2000
(methyl chloroform's phaseout date is 2005). Because of this agreement, cleaning
trends are toward nonhalogenated solvents (e.g., terpenes, aqueous saponifiers,
and water only) or no cleaning at all. The no-clean option is preferred since it
eliminates other environmental concerns such as waste disposal and energy con-
sumption, a contributor to the greenhouse effect.

5.3 DEFINITION OF SOLDERING FLUX

Soldering flux is defined in the flux specification of the Institute for Intercon-
necting and Packaging Electronic Circuits (IPC), ANSIJIPC-SF-818, "General
Requirements for Electronic Soldering Fluxes," as:

A chemically and physically active fonnula which promotes wetting of a metal surface
by molten solder, by removing the oxide or other surface films from the base metals
and the solder. The flux also protects the surfaces from reoxidation during soldering
and alters the surface tension of the molten solder and the base metal.

That is, flux prepares the metal surfaces to be soldered by:

• Removing oxides prior to soldering

• Assisting in uniform heating of the metal surfaces during preheat
• Maintaining an oxide-free surface during the soldering process
• Lowering the surface tension at the metal-solder interface

The vast majority of soldering is done with the aid of fluxes; choosing the
appropriate flux for the soldering process dictates the necessary postsoldering
cleaning process.

5.4 SPECIFICATIONS

Before examining the different specifications in detail, several test methods that
are cited in two or more specifications will be discussed in a general sense.
These methods are the copper mirror, silver chromate paper (halide), the surface
insulation resistance, British corrosion, and ionic conductivity tests.

5.4.1 Test Methods

The methods discussed below have specific purposes, advantages, and disad-
vantages. These, along, with their procedures, will be described.
 SPECIFICATIONS 145

Copper Mirror Test

The copper mirror test is referenced in a number of specifications 1,2 and is a

measure of a flux 's corrosivity to copper. This corrosivity is an indication of the
activity level of a flux. Both the speed and relative ease with which this test can
be performed make its use quite common. However, it also has several dr~w­
backs.
The copper mirror test involves placing two drops of the flux in question on
a copper mirror, a glass microscope slide on which 300 to 500 Aof copper has
been vapor deposited. Typically, a flux made of pure water white gum rosin and
isopropanol acts as a control and a drop of it is also placed on the mirror. The
mirror is then placed in a controlled environment of 23 ± 2°C and 50 ± 5 percent
relative humidity for 24 hours. Then the mirror is rinsed in isopropanol and
examined for any spots or areas where the copper has been removed by the flux.
Figure 5-1 illustrates two passing copper mirrors on the left and three failing
copper mirrors on the right.
Results of the copper mirror test indicate the potential corrosivity of the raw
or unheated flux . Although the corrosivity of unheated flux can be a concern, it
is, unfortunately, not always indicative of the corrosivity of a partially heated
flux (e.g., flux present on the top, or component side, of a circuit pack during

Figure 5-1 Results of copper mirror test.

146 POST-SOLDER CLEANING CONSIDERATIONS

wave soldering) or a fully heated flux (e.g., flux present on the bottom, or wiring
side, of a circuit pack during wave soldering). Also, the results can be difficult
to interpret sometimes since they are subjective and there is no middle ground
between pass or fail.
Because of these drawbacks, critics have questioned the relevance of this test
altogether. However, due to its speed, the relative ease of performing this test,
and the evidence that these results do correlate well with those of the more
lengthy and involved surface insulation resistance test, 3 this test is certainly
worthwhile and justified. It is especially useful as a prescreening test when testing
a new formulation or as a routine incoming materials check.

Silver Chromate Paper (Halide) Test

The silver chromate paper test is another fairly simple test to perform. Its
purpose is to detect significant amounts of ionizable chlorides and bromides in
a flux. Any flux that leaves the silver chromate paper unchanged has either no
Cl- or Br- , or at least, no detectable quantities.
The test procedure involves placing a drop of flux on silver chromatic paper.
After 15 s, the paper is rinsed or immersed in isopropanol and then dried for 10
min and examined for any color change. A color change means that halides are

Figure 5·2 Results of silver chromate paper test.

 SPECIFICATIONS 147

present in detectable quantities. Examples of silver chromate paper results are

shown in Figure 5-2.
Drawbacks of this test include possible subjectivity in interpretation of results,
lack of quantitative data, and interference from other nonhalide materials such
as amines, cyanides, or isocyanates.' Some chemicals can cause color change,
as can fluxes that are unusually acidic (pH < 3). Thus, if halides are not suspected
and a color change is detected, other tests should be performed to verify the
results.
Another problem with this test is that the results are pass/fail only and are not
quantitative. A proposal in a telecommunications specification had called for a
more quantitative assessment of the amount of chlorides and bromides in a flux
based on a color changes seen in the silver chromate paper.4 Still yet another
problem is that the silver chromate paper only detects chlorides and bromides,
not fluorides.
Even with these shortcomings, though, this test is beneficial for prescreening
or incoming material inspection for fluxes that should be noncorrosive.

Surface Insulation Resistance Test

Surface insulation resistance (SIR), also known as insulation resistance (IR)

and moisture and insulation resistance (M&IR), is a lengthy, but worthwhile,
flux evaluation test. SIR testing is also used to evaluate printed circuit fabrication
processes. Of all test methods, this one best simulates the relevant cause and
effect relationship between soldering fluxes and circuit boards. Even though SIR
coupons are not identical to actual circuit boards and SIR measurements are not
exactly like actual circuitry, SIR testing is better than other tests since (1) the
soldering process is part of the sample preparation, (2) the sample substrate is
itself a circuit board, and (3) the response variable, leakage current, is a measured
attribute of circuitry. Even so, this test method does have some shortcomings
that will be discussed along with the details about the types of test patterns,
sample preparation, temperature and humidity selection, necessary equipment,
and types of requirements.
The result of SIR testing is an electrical measurement of the effect of flux
and/or cleaning residues on a printed circuit board. The change in the SIR due
to the presence of some residue is measured assuming that it is much greater
than the bulk resistance of the printed circuit board. To maximize the measuring
sensitivity of this test, the leakage current passing between long parallel lines
placed at different electrical potentials is measured. Instead of long narrow
circuits, interdigitated combs that are electrically identical to two long lines are
typically used. Examples of two comb patterns are shown in Figure 5-3, the
telecommunications comb pattern with O.025-in. lines and O.050-in. spacings
and the IPC 'B' comb pattern with O.0125-in. lines and spacings.
148 POST-SOLDER CLEANING CONSIDERATIONS

Figure 5-3 Two SIR comb patterns: telecommunications and IPC; the smaller pattern is
from IPC.

Traditionally, data are reported as insulation resistance (in ohms and megohms)
rather than as leakage current. The conversion is simple via Ohm's law, V =
fR, where V is the test voltage and f is the leakage current. Diagrams in Figure
5-4 show the typical circuitry used to apply a 45 to 50-V dc bias voltage on the
left and to measure the resistance with a reverse polarity 100-V test voltage on
the right. SIR data are a function of a number of parameters:

1. The width of the spacing between the traces

2. The length of the traces
3. The testing voltage
4. The test voltage electrification time
5. The bias voltage, if any
6. The temperature
7. The relative humidity
8. The dielectric material of the test substrate
9. The specific solder resist coating, if any
10. The quality of the test leads, the electrical instruments, and the noise
shielding
 SPECIFICATIONS 149

ENVIRONMENTAl CHAMBER

100 V
(a) (b)

Figure 5-4 Circuit diagrams for biasing and measuring SIR comb patterns (a) Test, (b)
Measurement.

Although SIR is affected by both the total length and width between the traces,
it is a moot point when testing fluxes if the same pattern is always used. The
absolute values are only of interest if the test coupon's circuitry is identical in
both length and spacing as is the circuitry of the actual electronic product, which
is rarely, if ever, the case. However, as mentioned previously, there are many
comb pattern shapes and sizes. Theoretically, the data from all comb patterns
can be correlated by taking into account the number of squares that make up a
comb. A square is a unitless value equivalent to the total length of the comb
divided by the width of the spacing between two lines. So, a comb pattern that
has more lines will be longer and have more squares; additionally, a comb pattern
that has narrower spacing will have more squares. Several papers that have
discussed theory and concepts behind the use of squares are recommended for
further reading. 5 ,6 More recently, a study has been documented showing that
correlation is good within a range of squares of not more than one order of
magnitude. 7
The test voltage as well as the electrification time can also influence the ultimate
SIR measured. Moreover, the bias voltage and its value relative to the test voltage
can affect the results. These results have been documented. 8 Whether to use a
bias voltage and whether this bias voltage should be the reverse polarity of the
test voltage is an ongoing debate. Both the telecommunications and the IPC
150 POST-SOLDER CLEANING CONSIDERA liONS

specifications require a 100-V dc test voltage and a 45 to 50-V dc bias of reverse

polarity. Although these electrical parameters may influence the SIR results, it
is important to look at them in the context of the whole picture. That is, if one
simply maintains particular settings for these parameters, the changes in the SIR
resulting from changes in the electrical parameters can be disregarded . Estab-
lished SIR test methods specify these settings so the relative differences between
different fluxes can be observed without worrying about electrical parameter
effects.
Both temperature and humidity can be used to accelerate the ageing of samples.
Depending on the setting, however, these factors can also be specified to simulate
worst-case conditions. Figure 5-5 shows several typical temperature and humidity
settings. Also shown is the envelope in which circuitry typically must operate
relative to an example of a performance test condition (e.g., 35°C at 9O%RH)

100
PERFORMANCE TEST LFETIME TEST
CONOrTlON CONOITDNS
90
.'
,..A.

•
80
;?
~ 70
>-
I-
0 60
:2
=>
:r: 50
w
>
I-
ex:: 40
--'
w
a:
30

20

10

10 20 30 40 50 60 70 80 90 100

TEMPERATURE (CO)
Figure 5-5 Typical temperature and humidity conditions for SIR tests.
 SPECIFICATIONS 151

and several lifetime or accelerated test conditions (e.g., 55, 65, and 85°C, all
at 80%RH). Other conditions are also used such as 85°C/85%RH for Class ill
SIR testing per IPC-SF-818. As desirable as accelerated conditions may seem,
it is important to realize that an extreme environment may cause some flux
materials to fail by mechanisms that are nonexistent at normal operating con-
ditions.
Sample preparation simulates the assembly process, including fluxing the comb
pattern and then soldering it. Recently, the Bellcore specification2 has added an
additional test in which component side flux residues are tested by floating the
comb pattern face up on a solder pot or wave, rather than face down. This
additional test is important since some flux residues can potentially be more
harmful when exposed to less heat.
Some specifications require minimum absolute SIR values as a pass/fail cri-
terion whereas others require a minimum SIR relative to the SIR value of a
control. A relative value is preferred from a designer's point of view since it is
simply allowing a minimum degradation due to the presence of flux residues.
Obviously, some guideline for an absolute value for the control is a good idea.
In contrast, an absolute value for an SIR requirement is implying a knowledge
of the electrical needs for a particular product, a difficult if not impossible task
for widely used industry specifications. This is true since one requirement on
one particular comb pattern cannot simulate any and all use environments and
circuitry designs.

British Corrosion Test

Another test that is specified by a number of groups is a corrosion test, also

referred to as the British corrosion test. 1 It involves reflowing solder in the
dimple of a copper panel in the presence of flux solids and then ageing the panel
at 40°C and 93%RH for a number of days, depending on the specification and
the type of flux being tested. A flux is said to pass if no blue-green corrosion
can be observed after ageing. Figure 5-6 shows both passing and failing results
for the corrosion test.
This test has both benefits and drawbacks. On the plus side, the results are
independent of a flux's solids content since this quantity is held constant. Because
a concentrated flux material is used, the test is certainly a worst-case condition.
However, the test lacks an electrical bias, which is necessary to initiate some
corrosion mechanisms (e.g., electrolytic). Also, it creates an accentuated copper-
solder-flux interface. Lastly, the interpretation of results is difficult, subjective,
and qualitative only. An evaluation of "how green is green" and the extent of
green or blue-green coloration is difficult to quantify. As with the copper mirror
test, there is no middle ground between pass and fail.
152 POST-SOlDER CLEANING CONSIDERATIONS

.-

Figure 5-6 Results of corrosion test.

The difference between this test and the SIR test lies in their two distinct
purposes. The corrosion test evaluates the potential for an "open" failure, whereas
the SIR test determines the potential for an electrical "short."

Ionic Conductivity Test

The ionic conductivity tesf involves extracting those species on the circuit
board that are soluble in a water-alcohol solution. Those species that are ionizable
are measured either by their effect on the resistivity or conductivity of the
extracting solution. A number of different systems are commercially available
including one that constantly "cleans" the extracting solution, a "dynamic" mea-
surement, and several that measure the decreasing resistivity in a static situation,
"static" measurements. Some contain room temperature extracting solutions whereas
others both heat and spray the extracting solution to improve extraction efficiency.
Data are typically provided in J.Lg/cm2 or J.Lglin2 NaCI equivalents. This test is
useful for monitoring of the cleaning process and yields an average contamination
level of ionizable species on the board. Obviously, those species that are not
soluble or ionizable are not going to be counted, and information on localized
concentrations of residue cannot be collected. The ionic conductivity test is often
used to monitor water cleaning.
 SPECIFICATIONS 153

5.4.2 Institute for Interconnecting and Packaging Electronic Circuits

(lPC)

The IPC flux specification, IPC-SF-818, classifies fluxes in terms of activity

levels rather than in terms of material types. All electronic grade fluxes should
fit in one of three major categories of L, M, and H, denoting low, medium, and
high activity, respectively. Because of the further groupings for the SIR testing,
ultimately a flux would be categorized with L, M, or H, followed by the numeral
1, 2, or 3 (designating class of assembly) and by C, N, or CN depending on
whether a flux meets the SIR after being cleaned (C), not cleaned (N), or both
(CN).
Because of the classification system itself, virtually all fluxes can fit into at
least one category. It is up to the user to determine what product class is being
soldered and whether postsolder cleaning is part of the assembly process. After
these decisions, the activity level necessary or sufficient for the assembly process
must be chosen, remembering that with higher activity level, there is usually
increased soldering efficiency along with increased potential for corrosion if
proper cleaning procedures are not maintained.
The IPC specification requires (1) copper mirror test, (2) silver chromate
(halide) test, (3) British corrosion test, and (4) surface insulation resistance test.
The SIR comb pattern is a part of the IPC-B-25 test board and has O.0125-in.
lines and O.0125-in. spacings (see Fig. 5-3). The SIR requirement is an absolute
value of ~ 1 x 108 o.

5.4.3 U.S. Military

The U.S. military specification, MIL-F-14256E, is limited to only rosin-based

fluxes. There are three categories allowed, R, RMA, and RA designating rosin,
mildly activated rosin, and activated rosin, respectively. These are equivalent
to the R, RMA, and RA categories in QQ-S-571, the U.S. federal government
solder specification.
Like many other specifications, MIL-F-14256 requires a copper mirror test
(for R and RMA) and a silver chromate test (for R, RMA, and RA). A solder
spread factor test is also required for RMA and RA fluxes.
However, what sets the military specification apart from other specifications
is that it not only allows just rosin fluxes but also demands that a minimum of
51 percent of the solids content is rosin and that the rosin itself must meet a
certain minimum acid number. Another unusual requirement is a minimum re-
sistivity of water extract. For the first time, MIL-F-14256 has incorporated a
surface insulation resistance test in Issue E, June 1989. It requires a ~ lOO-MO
reading under 85°C and 85%RH conditions using the IPC B-25 'B' comb pattern
(same as IPC-SF-818 Class 3 test coupon, conditions, and requirement).
154 POST-SOlDER CLEANING CONSIDERATIONS

5.4.4 Telecommunications

The telecommunications industry has its own specifications and requirements.

After the divestiture of AT&T in 1984, the court set up a research group, Bellcore,
to work for the interests of the seven Bell operating companies. The Bellcore
specifications grew out of the original AT&T specifications and have in recent
years dictated U.S. telecommunication industry requirements.
The most recent Bellcore assembly document that specifies flux materials is
TR-TSY-OOOO78, Issue 2.2 It divides allowed fluxes into only two categories,
compliant and noncompliant.
Compliant fluxes are "noncorrosive fluxes" (formerly called "rosin noncor-
rosive fluxes"). Requirements that must be met include copper mirror, silver
chromate, fluoride, pH, and sm tests. Thesm test specifies a comb pattern
with 0.025-in. lines and 0.050-in. spacings (see Fig. 5-3) and an environment
of 35°C and 90%RH. Originally, Bellcore required that a comb pattern soldered
with a compliant flux be at least 75 percent of the sm of the control (like the
AT&T specification MS-58556). Even though Bellcore still requires an sm value
relative to the control, the requirement is really the equivalent resistance of two
parallel resistors, one being the sm for the control and the other, an ideal resistor
with a value of HPMO. Thus, the requirement is a sliding scale and the gap
between it and the control narrows as the control value decreases. This require-
ment must be met for three different types of comb pattern groups: (1) fluxed,
soldered, and cleaned, (2) fluxed, soldered, and not cleaned, and (3) fluxed,
soldered with comb face up, and not cleaned. The reason for the third group is
to simulate the exposure of the top or component side of a circuit board to flux.
This is necessary since some fluxes have been found to be more detrimental
when they are not heated to as high a temperature, a situation experienced by
flux present on the top side of a circuit board. There is a misconception that a
telecommunications noncorrosive rosin or compliant rosin flux is the same as
an RMA flux. This is not necessarily true since the requirements for each are
different.
The other category for telecommunication fluxes is noncompliant, which is
simply a classification for aggressive fluxes. That is, any flux that cannot meet
the requirements for compliant fluxes is then tested for the noncompliant re-
quirements. Aggressive fluxes are tested per these two requirements: (1) surface
insulation resistance of;;;a: 3 x 109 0 on a telecommunications comb pattern with
stripes of solder mask perpendicular to the comb's metal traces exposed to the
flux, solder, and an appropriate cleaning process and (2) ionic extraction value
of..:;; 1.0 f.Lglcm2 (6.45 f.Lglin. 2) NaCI equivalent for a bare copper comb pattern
exposed to flux, solder, and the cleaning process. Clearly, not all aggressive
fluxes can necessarily meet these requirements. Since the residues of these more
active fluxes are designed to be removed by cleaning, testing uncleaned boards
is not required.
 FLUX MATERIALS AND ASSOCIATED CLEANING 155

The inclusion of solder mask stripes on the SIR coupon has been debated
as have possible alternative patterns. Some people fear that the solder mask
stripes make the test vehicle unduly stringent and even unrealistic compared
to actual circuit designs. Effort to modify this geometry is ongoing. The ionic
extraction test is generally more well accepted by the industry as a cleaning
efficiency check for aggressive fluxes since it is easier; however, nonionizable
but hygroscopic and harmful species are undetected by ionic extraction even
though their effect is measurable by SIR. The extraction medium can be either
50:50 or 75:25.

5.5 FLUX MATERIALS AND ASSOCIATED CLEANING

Fluxes are composed of a number of different materials including:

• Solvent: the liquid carrier for the flux ingredients

• Vehicle: a thermally stable material that acts as the high temperature solvent
• Activator: ingredients that create a wettable surface for solder by removing
oxides and possibly other contaminants
• Surfactant: an ingredient that reduces the surface tension at the metal-solder
interface to further promote solder wetting
• Antioxidant: a material that prevents reoxidation of the metal surfaces

In this section, flux characteristics and the corresponding cleaning process

will be discussed in more detail and will be examined in terms of the following
flux categories: rosin, water soluble (WS), synthetic activated (SA), and low
solids (LS).
Before looking at each flux group individually, it is worthwhile to com-
pare the makeup of these fluxes in terms of generic constituents and propor-
tions, only looking at the three major components, solvent, vehicle, and ac-
tivator. As shown graphically in Figure 5-7, these four groups of fluxes consist
of different proportions of vehicle, activator, and solvent. Those rosin fluxes
that pass military RMA or telecommunication specifications typically have a
much lower activator to vehicle ratio than the other flux types shown. Water
soluble and synthetic activated fluxes may have about the same proportion of
activator relative to the amount of vehicle. All of these groups typically have
15 wt% to 35 wt% solids or nonvolatiles. In comparison, LSFs have much
less solids, and the amount of activator is typically the same or more than
the amount of vehicle.
Figure 5-8 shows the effect of flux selection on cleaning material choice. Not
all cleaning materials can remove all flux residue types, nor do all flux residues
need to be removed.
156 POST-SOLDER CLEANING CONSIDERATIONS

WEIGHT
100r-~--~~~--~------~--~~------~-r------~r-~

75

50

25

o
Rosin WSF SA LSF

FLUX TYPE
Acitvator _ Vehicle c::J Solvent

Figure 5-7 Comparison of major component proportions for four groups of fluxes.

fLUX TYPE
CLEANING Rosin Rosin WSF SA LSF
MEDIUM (benim) (alll!J'essive)

None X - . - X
(no clean)

ClIlorinated
solvents or X X . X -
(PC-1l3 based

Aqueous X X X - -
(saponifier)

Semi-aqueous
(terpeues)
X X X X -

Figure 5-8 Cleaning material choices as a function of flux type.

 FLUX MATERIALS AND ASSOCIATED CLEANING 157

5.5.1 Rosin

The vehicle in rosin fluxes is rosin or colophony, a naturally derived material

from pine trees. Rosin, consisting of a number of isomers, mostly abietic acid,
which are fairly large molecules and include one carboxyl group (molecular
formula: CIJl29COOH), acts as a flux vehicle and is also a mild activator at
soldering temperatures. Another unique property of rosin is that it is an extremely
good insulator at room temperature; in fact, its bulk resistivity is at least an
order of magnitude higher than the epoxy-glass printed wiring board material's
bulk resistivity. 9 Because rqsin hardens on cooling to room temperature, it serves
as an excellent encapsulant, prohibiting the movement of other flux residue
ingredients such as those from activators.
Depending on the final activity of a rosin flux, there are a number of different
types of activators typically used ranging from organic acids to organo-halides.
Again, depending on the quantities added, rosin fluxes mayor may not pass
copper mirror, silver chromate, surface insulation resistance, and corrosion tests.
"L" or "RMA" fluxes should pass these tests, whereas "RA," "M," or "H" will
not. Work has been carried out to maintain the noncorrosive nature of rosin
fluxes, yet maximize the soldering efficiency achievable with them. 9
Because of the insulating properties of rosin flux residue, it is not necessarily
cleaned off to avoid corrosion. In fact, many companies have left rosin flux
residue unremoved for years without problems. Nevertheless, there are a number
of reasons for removing rosin flux residue. Often rosin residue is sticky and
tacky. So, for aesthetic reasons or if the printed circuit boards may be exposed
to excessive dust and dirt,IO it may be necessary to clean. Another reason for
cleaning is to avoid problems during bed-of-nails testing or in-circuit testing;
rosin flux residue can cause false opens when the test pins are unable to push
through insulating residue covering a test pad. Sometimes it is necessary to
remove rosin residue because of its effect on the subsequent addition of conformal
coating to a printed circuit board. It is unfortunate when it is necessary to remove
rosin residue since its complete removal, although possible, is usually quite
difficult.
Cleaning processes for rosin residue removal include semiaqueous (e.g., ter-
penes, other hydrocarbons) ones, degreaser types using CFCs and HCFCs (par-
tially halogenated chlorofluorocarbons), and those using aqueous detergents or
saponifiers. For environmental reasons, CFCs and other chlorinated solvents
should be avoided. Aqueous detergents and saponifiers have been used for years
in both underbrush and total immersion spray cleaning operations. Since these
saponifiers are highly caustic and therefore corrosive, the wash stage must be
followed by a rinse stage in which the saponifier solution is rinsed off the boards
sufficiently. Semiaqueous processes, such as those with terpenes, have been
shown to be excellent cleanersY Component compatibility and environmental
impact should always be addressed when selecting a cleaning process. Rosin
158 POST-SOLDER ClEANING CONSIDERATIONS

residue removal can be accomplished with batch cleaners, underbrush in-line

cleaners, or total immersion spray impingement cleaners.

5.5.2 Water Soluble

Water soluble fluxes (WSFs) are typically very active and aggressive, and the
residue can be corrosive if not removed properly. However, since they are active
and designed to be removed in aqueous solutions, they offer a number of ad-
vantages over less active and more difficult to remove fluxes. 12 In fact, the
process window for soldering is relatively large, and extremely low solder defect
levels (less than 50 ppm) are attainable. Figure 5-9 illustrates the wider process
window and the lower solder defect level attainable with a WSF compared to
some rosin fluxes. With the improved soldering efficiency, some companies have
been able to eliminate the need for postsolder touch-up operations.
Often WSFs are referred to as organic acid, or OA, fluxes. Although it is true
that there are organic acids in many WSFs, it is also true that many other fluxes
including rosin, SA, and low solids include organic acids. Thus, the name OA
for water soluble fluxes is a misnomer and should be avoided.
The vehicle for WSFs is typically a polyethyleneoxide or polypropyleneoxide,
commonly referred to as polyglycols. These materials are nonionic but are hy-

30

25---r~r---~-----+-----+-----+-----r-----

1------~~--+_----~---4-----+----~_r~~
:
1'51--'--+-~--+--+-----+-~-
J
it'0-----+~~~--~~--~----~~~+_--~ WSI
c

5----~~--~--~~-----+-----+----~~~-

o 150 170 190 210 230 250 270'F

Figure 5-9 Solder defect levels as a function of preheat temperature for rosin fluxes RA2,
NR2, and RMA4 compared to water soluble flux WS1. 12
 FLUX MATERIALS AND ASSOCIATED CLEANING 159

groscopic, SO a deleterious, but synergistic effect is observed, when these po-

lyglycol residues are not removed along with the ionic residues from WSFS.13
That is, polyglycols are not going to encapsulate activator residues like rosin.
Usually the activity level is quite high in WSFs, meaning that fairly strong
organic acids and halide-bearing compounds are present or that relatively large
quantities are present. Some WSFs are halide-free and promoted as less corrosive,
yet their postsolder residue still needs to be removed. Also, there are some WSFs
that have a neutral pH since it is believed that these would be less corrosive
than those that are more acidic.
Because of their high activity level, most WSFs will fail the copper mirror
and silver chromate paper tests. If proper cleaning is employed, a number of
WSFs can pass SIR and corrosion tests. Both WSFs and SA (see next section)
fluxes must meet the requirements for noncompliant fluxes for the telecommun-
ications industry, and neither are presently allowed by the U.S. military since
they do not contain rosin.
WSFs, like other aggressive fluxes, must be removed in a spray impinge-
ment cleaning system, sometimes called "total immersion cleaning." For WSFs,
the cleaning medium is either water only or aqueous detergents. In-line spray
systems look much like car washes, with spray nozzles directed at the printed
circuit boards from both the top and the bottom. Because of the need to thor-
oughly remove all residues and because of the type of cleaning used, the
compatibility of components to immersion in water must be addressed. Also,
it is important to ensure that components are irrigable. The use of WSF and
aqueous cleaning for surface mount or mixed assemblies has been debated for
a number of years. However, a number of companies have demonstrated their
feasibility. 14-16,11

5.5.3 Synthetic Activated

In the 1980s, Du Pont developed a family of high activity synthetic activated

(SA) fluxes, analogous to WSFs and designed so that its residues were easily
soluble in the CFC-113/methanol azeotrope cleaning medium. SA fluxes are
based on a mono and di-isooctyl phosphate (lOP) chemistry and include com-
patible activator and solvent materials. 17 This technology was offered to flux
manufacturers who subsequently marketed these SA fluxes.
Due to their high activity, these fluxes perform much like WSFs in that they
will not pass the copper mirror or silver chromate paper tests. Also, SA flux
residues must be removed in a spray impingement or total immersion cleaning
system. Usually the cleaning medium is CFC based; however, it has been shown
that terpenes can adequately remove SA flux residues also. IS Thus, component
compatibility with the cleaning medium is an issue that must be addressed.
Unfortunately, a component that is compatible with water may not be compatible
160 POST-SOlDER CLEANING CONSIDERATIONS

with CFCs, and vice versa. If the cleaning process is sufficient, the requirements
for SIR and corrosion tests can be met with SA fluxes.
Like WSFs, SA fluxes typically yield lower solder defects than rosin fluxes
and their residues are more easily removable. Nevertheless, even with these
benefits, due to environmental concerns with CFCs, a decline in SA flux usage
is already being seen and will probably continue.

5.5.4 Low Solids (No-Clean)

Unlike the fluxes discussed in the previous sections, LSFs are formulated to
leave minimal or no postsolder residue. As the name implies, this category of
fluxes includes those that have a much lower solids content, typically from 2 to
5 wt%, compared to traditional fluxes.
Because the intent is to avoid cleaning, these LSFs should meet the LN
(preferably) or at least MN activity designations specified in IPC-SF-818. They
also should be able to meet the noncorrosive or compliant flux requirements in
the telecommunication specification. LSFs are not presently allowed by the U. S.
military.
The elimination of postsolder cleaning eliminates machine cost and mainte-
nance and reduces materials and operating costs. Along with cost reductions,
space savings and process simplification are also realized. Additionally, the need
for any postsolder assembly steps is eliminated in cases where components,
incompatible with the cleaning process, must be manually inserted and soldered
(this postsolder assembly of components is often necessary when aggressive
fluxes like WSF and SA, which require immersion cleaning, are used). Moreover,
the associated temporary solder mask used as a solder resist for plated through-
holes for postsolder assembly is not needed.
Concerns over EPA regulations, international restrictions, component quali-
fication, and waste disposal are precluded by the elimination of cleaning. The
environmental concerns are becoming increasingly important in making process
decisions as more is discovered about the effect of these processes on the eco-
systems and atmosphere.
An unexpected benefit has sometimes been realized in terms of improved
soldering yields with LSFs relative to those with rosin fluxes. 19 In fact, as much
as an order of magnitude drop in defect levels has been observed. 20 These defect
level changes are also dependent on the product code and process parameters.
Because of the lower viscosity and freer flow compared to rosin fluxes, LSFs
can more easily wet all surfaces of the wiring side of a board, which is especially
important when surface mount devices are wave soldered. 21
Ideally, a flux that does not need postsolder cleaning should be noncorrosive,
leave a nontacky, noncorrosive, colorless and/or minimal amount of residue,
have sufficient activity to yield acceptable soldering, and be compatible with
 FLUX MATERIALS AND ASSOCIATED CLEANING 161

existing fluxing equipment. Unfortunately, it is difficult, if not impossible, to

find a LSF that meets all of these characteristics. Instead, it might be necessary
to modify the flux application process to realize all of the above.
By maintaining a low solids content and ensuring sufficient flux activity, the
activator often outweighs the vehicle. Because of this unusual ratio, one cannot
rely on the vehicle being able to encapsulate activator residues.
The 'solvent used in LSFs is typically isopropanol, the same one used in the
other fluxes, the only difference being that there is a lot more solvent in a LSF
than in a traditional solids flux. The vehicle in LSFs is a synthetic resin material,
rosin, or modified rosin. Those that contain rosin are amber in color, rather than
clear, and can, like higher solids content rosin fluxes, leave residues that impair
bed-of-nails testing. One flux manufacturer developed a number of LSFs based
on a vehicle that is more thermally stable than rosin, pentaerythritol tetrabenzoate
(PETB).22 Besides being the flux vehicle, PETB supposedly acts as an antioxidant
also. Most LSFs are halide-free and contain only organic acid type activators.
Usually, the ones that have halides also include rosin. Preferably, a LSF will
be halide- and rosin-free. Many LSFs do not naturally foam due to their low
solids content and the characteristics of the particular constituents themselves.
LSFs that are foamable usually contain foaming agents that are either nonvolatile
or volatile. Some nonvolatile foaming agents are glycols that remain along with
the other flux residue after soldering. Like the glycols in WSFs, these can be
hygroscopic, creating an environment for potential corrosion. Some LSFs in-
clude, along with isopropanol, small quantities of other solvents that promote
foaming but volatilize during heating. In general, it has been seen that foaming
agents may have a deleterious effect on copper mirror, SIR, and corrosion test
results.
It has been noted that LSFs may be relatively difficult to remove with traditional
CFC azeOtropes; in fact, white residues can form so caution is necessary. 21
Consequently, it should be emphasized that there are certainly some cases when
cleaning does not necessarily improve a situation and can, in fact, do harm.
It has been shown that the SIR decreases as a function of the original quantity
of flux applied, inferring that excessive postsolder residue can cause electrical
problems. 2O Results indicated that excessive LSF residues are detrimental to
circuit integrity. In hindsight, it seems obvious that flux residues that are not
removed after soldering and do not contain a large proportion of highly insulating
water white rosin could cause problems. To better understand this phenomenon
and confirm this hypothesis, SIR studies were modified such that the quantity
of flux applied to comb patterns was carefully controlled. SIR results supported
the accelerated ageing results and demonstrated that heavy deposits of flux residue
cause high surface leakage currents. Figure 5-10 shows the dependence of SIR
on flux quantity for eight different LSFs. It is clear that, although all show the
same trend, the SIR for some fluxes are more strongly dependent on flux quantity
than others.
162 POST-SOLDER CLEANING CONSIDERATIONS

~
Flux A
Flux E
-+-
~
Flux B
Flux F
--
--8-
Flux C
Flux G
--a- Flux 0
---B- Flux H

SIR (Ohms)
1,000E+ 12
I Control
1,OOOE+ 11

1,000E+l0

1,000E+09

1,000E+08

10000000

1000000
2X 4X 40X 42x
Flux Quantity

Figure 5-10 Average SIR data for comb patterns soldered with eight LSFs as a function of
quantity of flux appliedY

The drawbacks of using LSFs compared to higher solids content fluxes are
summarized below:

• Quantity is critical to ensure minimal residue on the one hand and sufficient
fluxing activity on the other-
• Specific gravity is difficult to controL
• The effect of water absorption can be more pronounced.
• Foaming is difficult.
• The process window may be narrower.

Because of these drawbacks with LSF, the selection of application equipment

and solder process parameters is important. However, if these are chosen wisely,
the use of LSFs can yield enormous benefits.

5.5.5 Controlled Atmosphere Soldering

Controlled atmosphere soldering has also been called inert gas or fluxless sol-
dering. Inert gas soldering or fluxless soldering are misnomers since sometimes
more than inert gases are used, either reducing gases or vacuum atmospheres.
Furthermore, liquids consisting of a flux solvent and activator are commonly
 FLUX APPLICATION METHODS 163

used. Sometimes, for solder paste reflow, totally inert nitrogen atmospheres are
used (see Chap. 2)
Recently, several companies have developed wave soldering systems that
purportedly do not require the use of fluxes and therefore do not require postsolder
cleaning. Another advantage that is cited is the decreased amount of solder dross.
Dross is sometimes treated as a hazardous material. However, dross also can be
sold to recycling companies that refine it for customers desiring lower grade
solder; in this case, the amount of dross formed is not as problematic except
that it must be removed more often.
One system, in particular, uses a combination of nitrogen and formic acid
vapor to create an oxygen-free environment. However, to ensure proper through-
hole soldering, a liquid consisting of an organic acid and an alcohol is applied
by an ultrasonic atomizer. Although this material is not called a flux, it is, by
all definitions, a flux and should be treated as such in terms of its impact on
reliability. Another controlled atmosphere soldering system relies on vacuum
pumps and chambers to eliminate oxygen in the wave soldering module. Although
these new processes are promising, further evaluations of the reliability and
compatibility impact of the gases and liquids that the PC boards encounter need
to be performed.

5.6 FLUX APPLICATION METHODS

There are a number of ways to apply fluxes to printed circuit boards for wave
soldering, the most common being wave, foaming, and spraying.

5.6.1 Wave

A wave fluxer is similar to the solder wave itself; the flux is pumped continuously
through a wide nozzle located in an open reservoir or pool of flux. It is quite
simple to operate and maintain. Another advantage is that virtually any liquid
flux can be applied this way. Because a wave fluxer can apply excessive quantities
of flux, it is sometimes followed by brushes or an air knife to remove excess.
With a wave fluxer, maintaining the original specific gravity requires adding
solvent since the flux is in an open reservoir and exposed to the ambient. With
most fluxes, the specific gravity can be controlled adequately with an automatic
density controller that monitors the specific gravity and adds additional solvent
or flux when the specific gravity gets out of range. However, with LSFs fluxes,
this issue of changing specific gravity can be detrimental since small changes in
the flux solvent content can greatly change the composition due to the small
amount of solids (see the section "Application Issues for LSFs"). Water ab-
sorption is also an issue with wave fluxers due to open reservoir:
164 POST-SOLDER CLEANING CONSIDERATIONS

5.6.2 Foam

A foam fluxer consists of an open reservoir of flux containing a submersed

porous stone through which air is pumped. With the right combination of air
pressure, nozzle size and shape, and surface tension of the flux, air bubbles are
created in the pool of flux and pushed through a nozzle, creating a foam of flux
at the surface. Maintaining a level and uniform foam head with fine bubbles can
sometimes be difficult since it depends on a number of factors. In recent years,
modifications have been made to foam fluxers to make it easier to foam low
solid fluxes; one manufacturer has designed a foamer with a dual aerator stone
and an enlarged nozzle opening.
Like a wave fluxer, a foam fluxer causes changes in the specific gravity of
the flux since the flux is not only exposed to the ambient but is aerated. So,
either a density controller is necessary or, in the case of LSFs, caution is sug-
gested. Another issue, especially important with foam fluxes, is the possibility
of water absorption by the isopropanol or other hygrOSCOPIc flux constituents.
The addition of water can increase solder defects, cause spattering and solder
balls, and give a false confidence in specific gravity maintenance. That is, the
addition of water can lead one to believe that the specific gravity is correct when,
in fact, it is only in the right range due to the presence of an impurity (water).
Because of the nature of both wave and foam fluxers, fluxes applied with
these methods can easily become contaminated with water, dirt from the printed
circuit boards, conveyor fingers, or fixtures.

5.6.3 Spray

There are several different spray applications methods, high velocity spray,
ultrasonic spray, and rotating drum spray, all with advantages and disadvantages,
which are discussed in detail below.
A high velocity spray fluxer, which can have a nozzle similar to one on a
spray cleaner, can propel excessive flux onto the top side of the circuit board.
Moreover, units with large spray patterns can coat the inside of solder machines
and can create maintenance problems. Typical nozzles create spray patterns with
relatively higher concentrations in the center, making uniform application dif-
ficult. Also, because of the nozzle design and the evaporating solvent, typical
spray nozzles are easily clogged. Sometimes, a solvent after-spray is discharged
after the flux spray to prevent or, at least, minimize clogging. 24
Because of the importance of controlling the deposition of LSFs, a novel
means of applying them was developed and patented. 25 The theory behind ul-
trasonic atomization has been documented. 26 While the flow rate controls the
amount of flux that will be dispensed, the power into the nozzle controls the
 FLUX APPLICAITON METHODS 165

amplitude of the sound waves into the nozzle, affecting the droplet size. A low
velocity air stream directs the spray mist upward. A drawing of the flux mist
being generated and subsequently shaped in the mixing chamber and directed
upward is shown in Figure 5-11.
The ultrasonically generated spray fluxer offers several advantages over other
fluxers, especially pertinent for LSFs. 20 First, flux quantity can be controlled by

Figure 5·11 Ultrasonically atomized flux being shaped and directed in mixing chamber.25
166 POST-SOLDER CLEANING CONSIDERATIONS

adjusting the flux flow rate, ensuring product quality and reliability. Second,
this is a closed system, preventing the evaporation of the flux solvent and thereby
assuring constant flux composition, especially important for LSFs. Moreover,
the closed system prevents the possibility of water absorption. Because of the
nature of the low velocity flux mist, a minimal amount of flux is deposited on
the top side of the board, important for LSFs that can degrade the SIR more
severely when not heated fully to soldering temperature.

5.6.4 Rotating Drum Spray

Another spray method that has been in existence for a number of years has
recently been touted as an application method for LSFs. It consists of a partially
immersed mesh drum rotating in a flux tank. Located on its center axis is an air
knife, which, in combination with the rotating drum, creates a fine flux spray.
The quantity of flux that is sprayed depends on the mesh size, the rotation speed,
conveyor speed, and the air velocity of the air knife. Therefore, the quantity of
flux can be controlled with this application method. However, the issues re-
garding control of specific gravity and the possibility of water absorption exist
with this method since there is an open flux tank.

5.6.5 Application Issues for Low Solids Fluxes

A number of different application techniques are commercially available, each

with its own list of advantages and disadvantages. From a flux quantity stand-
point, a wave fluxer applies more than a foam fluxer, which applies more than
a spray fluxer.
Already discussed is the fact that the amount of LSF applied may affect the
SIR and ultimately the reliability of a printed circuit board. Consequently, it is
important to control the quantity of flux applied. Normally, a board's cleanliness
is controlled by the efficiency of the cleaning material and cleaning process;
here, the control must be directed at the flux application process step since there
is no postsolder cleaning. As important is the amount of flux that is deposited
onto the top side of the printed circuit board since partially heated LSF residues
may be corrosive.

5.7 PROCESS ISSUES ASSOCIATED WITH RELIABILITY

There are a number of process issues associated with the type of flux used. These
issues include flux residue, solder ball formation, top-side fillet formation, and
conformal coating compatibility, all of which are discussed below.
 PROCESS ISSUES ASSOCIATED WITH RELIABILITY 157

5.7.1 Flux Residue

Depending on the flux type, flux residue mayor may not be a reliability concern.
The more aggressive fluxes leave behind more corrosive residue and should
therefore be removed properly. Less active fluxes or those that meet certain tests
requirements without removing their postsolder residue may be removed for
aesthetic reasons or to avoid problems with testing, solder balls, or conformal
coating.

5.7.2' Solder Ball Formation

Material incompatibilities can cause excessive solder balls and/or other solder-
type residue. The number of solder balls is a function of the type of flux used,
the type of solder mask on the printed circuit board, and the soldering conditions.
The presence of solder balls becomes a real issue when there is no postsolder
cleaning (e.g., some rosin fluxes, LSFs, controlled atmosphere soldering) that
could remove the balls along with the other residue. In comparisons of a number
of solder mask types and flux types (rosin, SA, WSF, and LSF), the industry
experience is that the number of solder balls is greater with dry film solder masks
and with low LSFs. Work has been done showing that the following factors can
affect solder ball formation: dry film solder mask type, printed wiring board
cleanliness, surface roughness, and flux type.27 Khutoretsky's study included
dry film solder masks and both water soluble and rosin fluxes; the results indicated
that both the flux residue and the solder mask polymer can retain solder balls
on the board surface. Photomicrographs in Khutoretsky's paper illustrate this
"connecting chain" phenomenon.

5.7.3 Top-Side Fillet formation

Top-side fillets are required in some specifications, the definition of which varies
amongst specifications. Some specifications simply require positive wetting on
both the component leads and the barrel and a certain amount of barrel fill,
whereas other specifications might require a fillet on the top side that looks
almost identical to that on the bottom side. The exact amount of solder necessary
to create a reliable joint is debatable and obviously is affected by the conditions
under which a printed circuit board will ultimately be exposed. That is, an
electronic product that is made for military aircraft will be exposed to much
harsher temperature and vibration conditions than a household consumer product.
The formation of sufficient fillets on the component side of the printed circuit
board is affected by the board and component temperatures through the solder
wave, their solderability, board thickness, the component-to-hole aspect ratio,
168 POST-SOLDER CLEANING CONSIDERATIONS

the solder wave height, the conveyor speed, the flux application method, the
activity of the flux, and the solids content of the flux. If the circuit board and
components are not adequately heated, the solder will not rise properly in the
holes. Moreover, if solderability is poor or the activity of the flux is not adequate,
the solder will not readily wet. Differences in top-side fillets have been observed
between a 37 percent solids rosin flux and a 15 percent solids rosin flux; addi-
tionally, other parameters also affected top-side fillet formation in this study
including conveyor speed, height of flux foam, and the preheat profile. 28 For
LSFs, problems with top-side fillet formation can be more pronounced.
In summary, the solder parameters, the board design, the solderability, and
the type of flux can all affect top-side fillet formation.

5.7.4 Conformal Coating Compatibility

Conformal coating is commonly applied after soldering to printed circuit boards

that will be exposed to harsh or extreme conditions. These include military
applications and automobile electronics. Some flux residues underneath these
coatings can cause blistering, peeling, or measling. Any moisture or hygroscopic
residues underneath the coating can cause these problems. Obviously then, coat-
ing compatibility with a flux and its residue should be taken into account.
Although rosin fluxes are usually associated with product requiring conformal
coating, reports by members of the electronics industry indicate that coatings
exist that are compatible with other flux types including WSF and LSF; cleaning
before the coating process is usually necessary and depends on the type of flux
used.

5.8 NON-LIQUID FLUXES

5.8.1 Core Solder Material

Instead of liquid fluxes that are used for wave soldering, hand soldering is usually
performed with solder wire that contains a core or several cores of flux material.
This material is solid or semisolid at room temperature, but it flows and is active
at soldering temperatures to promote adequate wetting of the solder. Rosin-based
core solder has been the standard in the electronics industry but in the last few
years there has been an increase in water soluble core flux and, more recently,
LSF core solders. Activators may be organic acids and/or organo-halides.
Core materials can be tested just like liquid fluxes by obtaining the core material
from the manufacturer or extracting it from the solder wire. (Since using the
core material is easier than extracting it, this is preferred.) However, although
some specifications state that the core material may be removed by melting the
solder wire and then testing, this practice is emphatically not recommended. By
 NON-LIQUID FLUXES 169

melting the solder, the core material is also heated and can thus react with the
solder or simply undergo changes by itself. As a consequence, the material tested
will have undergone one more heating step than allowed or required in tests such
as copper mirror or SIR. It is best to use the unheated core material, apply it to
the appropriate substrates, and perform the tests as described earlier.

5.8.2 Solder Paste Material

Solder pastes are often used for attaching surface mount devices. The paste is
stenciled or screen-printed on pads on the printed circuit board, the components
are placed onto the pads, and the whole assembly is heated so that the solder
paste is reflowed. The solder paste is a combination of solder powder, rheology
modifiers, and normal flux constituents such as high temperature solvent, a
vehicle, and activators. It is difficult to find a good solder paste because of the
number of requirements imposed upon these materials. That is, along with ac-
ceptable flux and solder materials, a paste must also be formulated to include
other properties such as tackiness, printability, and minimal slumping. Therefore,
solder pastes are usually tested like other fluxes for chemical reliability purposes
and then undergo further tests to ensure the presence of certain physical prop-
erties. The commercial electronics industry specification for solder pastes is IPC-
SP-819.
Solder pastes based on any of the four major flux types exist. However, most
solder paste used today is rosin-based with minimal or no halide activators due
to cleaning concerns and question regarding need for higher activity. Since paste
is placed exactly where it is needed and the type of joint that is made is so
different for surface mount than for through-hole, the activity of a solder paste
is not as important as it might be for a liquid wave soldering flux. Because of
the additional materials in solder pastes and the localized concentrations of paste
residue compared to wave soldering flux residue, cleaning is more difficult and
more of a concern than cleaning after wave soldering.
Removal of solder paste residue is sometimes, but not always, necessary for
the same reasons as for liquid fluxes. However, with the increasing problems
with cleaning, lower solids content solder pastes are being developed. Water
soluble solder pastes are available for applications such as mixed technology
where water soluble liquid flux and solder paste residue will be removed in a
water cleaning system. The concern is that the residue be removable after the
heat of soldering and that the electrical properties (SIR) of the assembly have
not been excessively degraded. Even though the flux-type ingredients may be
water soluble, the rheological agents are not necessarily so. Hwang29 offers an
extensive discussion about solder pastes in general (also, see Chapter 2).
For liquid fluxes, core solders, or solder pastes, less flux solids means less
residue. An effective way to reduce flux residue when soldering with cored wire
170 POST-SOLDER CLEANING CONSIDERATIONS

solder is to use smaller core sizes. Typically the standard 3 percent flux (by
weight) can be reduced to 2, 1, or 0.5 percent. With less flux being used,
solderability and heat control become more important to ensure that reliable
solder connections are made. For solder pastes, it is difficult to truly make a
lower solids content paste and also maintain the proper physical properties. That
is, the less nonmetal material that exists, the more solder powder that is present;
obviously, the printability and tackiness properties will then be affected. Thus,
it seems a paste that leaves less or no residue should contain the usual proportion
of solder powder (approximately 90 wt %) but include other materials that
decompose during reflow. Low solids pastes reflowed in controlled atmospheres
are being investigated. 30

5.9 TRENDS

In 1990, the major push is away from CFC and, if possible , toward the elimination
of cleaning altogether. There are several possible ways to avoid cleaning in-
cluding the use of noncorrosive rosin fluxes, the use of LSFs, and the use of
controlled atmosphere soldering. According to Szymanowski,31 controlled at-
mosphere (inert gas) soldering and low LSFs offer the most attractive alternatives
to CFCs. In those cases where cleaning is still necessary, HCFCs, terpenes, or
aqueous cleaning are the choices. However, HCFCs are an interim solution only,
since their use may be restricted in the future. Presently, there are still no
soldering methods that are truly "fluxless" and there may never be any that do
not use some kind of fluxing material_ But with LSFs and controlled atmosphere
soldering, at least, we are getting closer to that ideal.
In the future, improvements in wave soldering LSFs will be seen as well as
improvements in low solids solder core wire and low solids solder pastes.
Other types of fluxes can and will still be used with appropriate non-CFC
cleaning media. For rosin flux residue removal, semiaqueous cleaners are rec-
ommended over CFCs and saponifiers for environmental reasons. If a completely
aqueous system is preferred, water soluble flux can be used if the assembly
allows complete removal of the flux residue with water. However, the addition
of a cleaning process brings additional expense, maintenance, capital equipment,
and floor space. With the ever-increasing restrictions on the wastes that can be
expelled from factories, minimizing or eliminating cleaning altogether is certainly
the preferred direction and the trend.

REFERENCES

1. IPC, ANSIJIPC-SF-818, "General Requirements for Electronic Soldering Fluxes,"

Lincolnwood, IL: Institute for Interconnecting and Packaging Electronic Circuits
(IPC), 1988.
 REFERENCES 171

2. Bellcore, TR-TSY-000078, Issue 2. "Generic Physical Design Requirements for

Telecommunications Products and Equipment," Red Bank, NJ: Bell Communications
Research, Inc., 1988.
3. Guth, L. A., "Low Solids Flux Material Characterization Studies," IPC Technical
Paper IPC-TP-735, 1988.
4. Bellcore, TA-TSY-000078, Issue 2. "Proposed Revision ofTR-TSY-000078: Generic
Physical Design Requirements for Telecommunications Products and Equipment,"
Red Bank, NJ: Bell Communications Research, Inc., 1987.
5. Zado, F. M., "Electrical/Electronic Reliability Considerations in Modem PWB Man-
ufacturing and Assembly Operations," Printed Circuit World Convention Ill, WCill-
70, 1984.
6. Gorondy, E. J., "Surface Insulation Resistance-Part I: The Development of an
Automated SIR Measurement Technique," IPC Technical Paper IPC-TP-518, 1984.
7. Gorondy, E. J., "SurfacelMoisture Insulation Resistance (SIRIMIR}-Part ill: Anal-
ysis of the Effects of the Test Parameters and Environmental Conditions on Test
Results," IPC Technical Paper, 1988.
8. Chan, A. S. L. , and T. A. Shankoff, "Interrelating Surface Insulation Resistance
Test Patterns," Circuit World, 15(4), 1989, pp. 34-38.
9. Zado, F. M., "Increasing the Soldering Efficiency of Noncorrosive Rosin Fluxes,"
The Western Electric Engineer (1), 1983, pp. 22-29.
10. Reagor, B. T., and C. A. Russell, "A Survey of Problems in Telecommunication
Equipment Resulting from Chemical Contamination," IEEE Transactions on Com-
ponents, Hybrids, and Manufacturing Technology, CHMT-9 (2) , 1986, pp. 209-
214.
11. Dickinson, D. A., L. A. Guth, and G. M. Wenger, "Advances in Cleaning of
Electronic Assemblies," Nepcon West 1989, 1989.
12. Chung, B. C., M. P. Cassidy, and G. W. Graham, "Evaluation of Flux Perfonnance,
Cleaning and Reliability," The Western Electric Engineer, (1), 1989, pp. 30-39.
13. Zado, F. M., "Effects of Non-Ionic Water Soluble Flux Residues," The Western
Electric Engineer, (1), 1983, pp. 40-48.
14. Wargotz, W. B., L. A. Guth, and C. V. Stroud, "Quantification of Cleanliness
Beneath Surface Mounted Discretes Assembled by Wave Soldering," Printed Circuit
World Convention IV, WCIV-n, 1987.
15. Danford, A., andP . Gallagher, "SMD Cleanliness in an Aqueous Cleaning Process,"
Nepcon East 1987, 1987, pp. 246-255.
16. Aspandiar, R., A. Piyarali, and P. Prasad, "Is OA OK?" Circuits Manufacturing,
April 1986, pp. 29-36.
17. Kenyon, W. G., "Synthetic Activated (SA) Flux Technology: Development, Com-
mercialization, Benefits and Future Applications," Proceedings ofInternepcon Japan
'86, 1986.
18. Hayes, M., and C. Hood, "Alternative Cleaning Methods: One Possibility," interview
in Printed Circuit Assembly, September 1989, pp. 33-35.
19. Sholley, C., "New Fluxes Eliminate Cleaning of PCB's," Electronic Manufacturing,
March 1989, pp. 32-33.
20. Guth, L. A., "Low Solids Flux Technology for Solder Assembly of Circuit Packs,"
Proceedings of the Electronic Components Conference, 1989, pp. 748-753.
21. Toubin, A., "Low Solids Content Fluxes," Circuit World, 15(2), 1989, pp. 17-18.
172 POST-SOLDER CLEANING CONSIDERATIONS

22. Rubin, W., "Some Recent Advances in flux Technology," Brazing and Soldering,
(2), Spring 1982, pp. 24-28.
23. Guth, L. A., "flux Selection and Application to Negate Cleaning Needs," presen-
tation at Nepcon East, Boston, MA, June 14, 1989.
24. Chung, B. C., W. A. Jancuk, and V. A. Corsaro, "Aqueous DetergentforRemoving
Rosin fluxes," The Western Engineer, (1), 1983, pp. 62-68.
25. Fisher, J. R., L. A. Guth, and J. A. Mahler, "Method and Apparatus for Applying
flux to a Substrate," U.S. Patent 4,821,948, April 18, 1989. Assigned to AT&T.
26. Berger, H., "Ultrasonic Nozzles Atomize Without Air," Machine Design, July 21,
1988, pp. 58-62.
27. Khutoretsky, M., "An Evaluation of Solder Balls on Printed Wiring Boards with
Dry Film Solder Resist," Proceedings of AMS's Third Conference on Electronic
Packaging: Materials and Processes & Co"osion in Microelectronics, 1987, pp.
225-233.
28. Wittmer, P., Conversation with the author. Unpublished Magnavox data, 1989.
29. Hwang, J. S. Solder Paste in Electronics Packaging, New York: Van Nostrand
Reinhold, 1989.
30. Morris, J. R., "No Clean Alternatives for Solder Paste Reftow," Proceedings of
SMART VI Conference, (2), 1990, pp. 97-106.
31. Szymanowski, R. A., "fluxing Options for CFC Elimination," IPC Technical Re-
view, April-May 1989, pp. 19-23.

ADDITIONAL READINGS

Deram, B., "Considerations for Use of No-Clean fluxes in Soldering PCB's," Electronic
Manufacturing, February 1989, pp. 32-34.
Elliott, D., "How to Avoid Problems with Wave Soldering and IR Reftow," Surface
Mount Technology, October 1989, pp. 47-54.
Johnson, R. 0., "Solderability Issues for Hand Soldering," Printed Circuit Assembly,
June 1989, pp. 30-32.
Klein Wassink:, R. J., Soldering in Electronics, Ayr, Scotland: Electrochemical Publi-
cations, Ltd., 1984.
Wade, R. L., "A View of Low-Solids fluxes," Printed Circuit Assembly, March 1989,
pp.31-34.
 6
Scanning Electron Microscopy and
Energy Dispersive X-ray (SEM/EDX)
Characterization of Solder-
Solderability and Reliability
James L. Marshall

6.1 SCANNING ELECTRON MICROSCOPY/ENERGY DISPERSIVE

X-RAY ANALYSIS

SEMlEDX is a powerful way to characterize solder in resolving solderability

and reliability problems. SEMlEDX is a combination tool, with two instruments
working in partnership. These two instruments operate simultaneously to com-
plement each other's data acquisition, guided by the instrumental operator. Scan-
ning electron microscopy (SEM) provides an image (i.e., morphological infor-
mation or surface features) on a magnified scale-XIO to Xl00,OOO, although
the usual range is perhaps X50 to X5000. 1 Energy dispersive x-ray (EDX)
determines the elemental composition of an area, with a sensitivity of perhaps
0.1 to 1 percent composition and with a spatial resolution of 1 J.Lm. 2 EDX is
commonly used for elements with atomic number ;;;:.11 (sodium), but thin window
EDX systems can also detect elements with atomic number ;;;:.5 (boron). With
these two instruments operating together, the instrumentalist can scan areas of
potential interest, zoom in with higher magnification, and determine elemental
compositions in selected areas of interest.

6.2 OTHER METHODS-WDX

A less commonly used method of x-ray analysis is wavelength dispersive x-ray

(WDX). One advantage of WDX is a much higher resolution, permitting the
clean separation of poorly resolved signals in EDX, such as indium and tin. This
higher resolution for WDX permits a higher sensitivity and the detection of
impurities that otherwise might be masked by major constituent peaks. Another
advantage of WDX is the facile detection of elements with low atomic number

173
174 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X·RAY (SEM/EDX)

(:::'5). Instrumentally, however, WDX involves a more complex mechanical

operation to select appropriate wavelengths (EDX uses a solid-state detector). 3
WDX also involves longer analysis times since the spectrometer is mechanically
scanned throughout its wavelength range--consuming minutes or even hours
versus seconds or minutes for EDX.2 EDX therefore is more useful for rapid
qualitative analysis, particularly for unknown specimens. Since EDX is more
commonly used than WDX in SEM, discussion in this chapter will be limited
to SEMJEDX.

6.3 DETECTION MODES

Two detection "modes" are available for SEM-secondary electron emission

and backscattered electron emission.! The secondary image generates high res-
olution images and is used for morphological studies. Hence, this mode of
detection is appropriate for minimally prepared samples for rapid analysis. For
example, a solder joint clipped from a circuit board can be quickly mounted and
inserted into a spectrometer and within minutes can supply important information
such as:

Where is the fracture?

Through the bulk solder?
Or is the solder peeling from the component lead?
What is the composition of the solder?
Of the component lead?

In contrast, the backscattered mode of detection emphasizes elemental identifi-

cation and presents higher visual contrast for compositional differences-dark
images for low atomic number, light images for high atomic number. The res-
olution for backscattered images is lower because the sampling volume of de-
tected electrons is much greater, perhaps 0.5 j.Lm.! Although the backscattered
mode may be used for rapid-mount samples, it is more commonly reserved for
carefully prepared specimens that present a fiat, polished surface. Backscattered
images are particularly useful in distinguishing different phases in metallurgical
samples, for example, the differentiation among various phases of copper-tin
elements and intermetallic compounds.! In this chapter, both secondary and
backscattered images will be presented to illustrate the suitability of each method
in different situations.

6.4 SAMPLE PREPARATION

For morphological images of solder specimens, a quick mount can customarily

be used for sample preparation because solder is conductive. For a cross section,
 INTERMETALLICS 175

a flat polished specimen must be prepared, using the steps of potting, cutting,
grinding, and polishing? The specimen is then gold- or carbon-coated to render
the sample conductive. Care must be taken in all these steps to prevent chemical
artifacts. For example, chemical etching can change the composition of the
surface and must be avoided. Smearing must be carefully avoided to prevent
misinterpretation of results; smearing can occur during the grinding or polishing
steps. If materials of different hardness exist in the sample (such as solder on
Alz°3 ) , the solder can be preferentially removed in a lapping operation, pro-
ducing a specimen with roughened solder recessed beside the durable ceramic.
To prevent such useless specimens, low pressure grinding and polishing of short
duration should be used. It may be useful to the instrumentalist to compare
secondary images with backscattered images or EDX to verify agreement of
observed structures, remembering that secondary images are more superficial
(which would include smearing), whereas backscattered images or EDX disclose
data from deeper portions. The instrumentalist may have to perfect diverse
techniques for different situations. For techniques of sample preparation, the
reader is referred to excellent literature distributed by Buehler, Ltd.4-8

6.5 DIFFERENT PHASES IN ALLOYS

Different compositional phases, identified in phase diagrams,9 can be differen-

tiated in SEMlEDX. Backscattered images, which emphasize composition, are
particularly useful to distinguish phases. EDX data are sufficiently quantitative
to determine compositions and thereby identify phases that differ only in stoi-
chiometry (e.g., the copper-tin phases CU3Sn and Cll(;Sns or the gold-tin phases
AuSI4, AuSn2, AuSn). Phases in alloys generally have specific stoichiometries,
although the composition of a given phase may have a wide range. These phases
are sometimes designated by Greek letters. However, binary compositions of
different combinations of elements may use the same Greek designation, and
ambiguity may result (e. g., two different a phases may exist in the same sample).
To avoid confusion, the description of elemental and compound phases will be
done by formula (e.g., Ag3Sn, C6 Sns, Sn) with the understanding that stoichi-
ometry is rarely exact. 10

6.6 INTERMETALLICS

The term "intermetallics" in this chapter has the special meaning generally at-
tributed in solder technology (Le., an alloy formed between a solder and the
base metal during the soldering operationll). This intermetallic forms the bond
between the solder and the substrate. Intermetallics are generally more brittle
than the base metal or the solder. Sometimes the intermetallic can grow needles
or prisms into the solder and can even fracture and drift into the solder.
176 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X·RAY (SEM/EDX)

6.7 SCOPE OF THE CHAPTER

This chapter will present typical SEM/EDX examples as an interpretative guide

for solving problems for the instrumentalist- We will assume an instrumental
system commonly used in solder work, a SEMlEDX with thick window (atomic
number ~ 11). For more detailed discussions of SEMlEDX instrumentation and
techniques, the reader is referred to specific treatises. 12,13,1-3

6.8 SEMlEDX CHARACTERIZATION-GENERAL

6.B.l Tin-Lead Solders

In a cross section of a typical 60% Sn-40% Pb solder joint to copper (Fig. 6-1),
the darker tin and the lighter lead domains are easily visible. 11,14 The intermetallic 19
is not so easily seen and it takes more effort to clearly distinguish this CU6SnS
compound (Fig. 6-2). For a quick-mount sample (external appearance), these tin
and lead domains can be easily seen in the secondary mode of detection if the
surface of the solder is clean and not heavily oxidized (Fig. 6-3).

Figure 6-1 Cross section of 60% 5n-40% Pb solder joint. Lower left, dark region: copper
substrate. Upper right: darker domains, tin; lighter domains, lead. X2000, backscattered .
 SEM/EDX CHARACTERIZATION-GENERAL 177

Figure 6-2 Cross section of 60% Sn-40% Pb solder joint, with careful polishing and careful
adjusting of the instrumentation, to manifest the copper-tin Cu6Sns intermetallic, about 1 jJ.m
thick. Top half, dark region : copper substrate. Bottom half: darker domains, tin; lighter domains,
lead . Intermediate region, CU6SnS intermetallic. X2000, backscattered.

Figure 6-3 Cleaned solder paste, 60% SN-40% Pb. X200, secondary.
178 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-4 Cross section of 62% Sn-36% Pb-2% Ag solder joint. Left, dark region : copper
substrate. Right: darker domains, tin; lighter domains, lead. The crystal dominating the view
is Ag 3 Sn, growing from the intermetallic . Because of extended heat treatment, some CU6SnS
is growing from the interface into the solder (darker small areas, e.g., top of connection of
Ag 3 Sn crystal to substrate interface). X2000, backscattered.

6.8.2 Two Percent Silver Solder

Ordinarily, the appearance of 62% Sn-36% Pb-2% Ag solder is virtually identical

to 60% Sn-40% Pb or 63% Sn-37% Pb solder, but sometimes one may observe
a silver-rich crystal9 (Ag 3Sn) growing from the solder joint interface (Fig. 6-4).
Chunks or crystals of binary alloys growing into the solder are indications of
extended heat treatment (also, see Figs. 6-28 through 6-30).

6.8.3 Gold- and Silver-Based Solders

Like tin-lead, silver solder (95% Sn-5% Ag) possesses two phases. 9 With a
different ratio of phases, however, the appearance is quite unlike that of tin-lead
(Fig. 6-5). Gold can form many phases with tin, but gold solder possesses only
one phase with the composition 80% Au-20% Sn (Fig. 6-6).9
 SEM/EDX CHARACTERIZATION-GENERAL 179

Figure 6-5 Cross section of 95% Sn-5% Ag solder. The light domains are silver-rich phases.
X7800, backscattered.

Figure 6-6 Cross section of gold solder joint. Upper left, dark region: silver wire. Bottom,
dark region : copper substrate. Solder: lighter regions, gold; darker regions, 80% Au-20% Sn
solder. X100, backscattered.
180 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

6.8.4 Indium Solders

The common solders of indium include lead and tin alloys. Indium-lead solder
of a given composition (e.g., 50% In-50% Pb) possesses only one phase9 and
a cross section of this solder is featureless. For indium-tin (Fig. 6-7), it is very
difficult to distinguish between the indium-rich and the lead-rich phases9 because
these two elements have almost identical atomic numbers. Hence, the intensities
of the two phases are virtually identical in SEM micrographs, even in the back-
scattered mode of detection. Obtaining a dot map, therefore, is laborious (Fig.
6-8". Figures 6-7 and 6-8 point out the caution that must be exercised during
interpretation of SEM micrographs: artifacts in Figure 6-7 could have been
misidentified as a different phase_
Although the SEM of indium solders is not very informative compositionally,
the indium-copper or indium-tin-copper intermetallic9 can be easily seen (Figs.
6-9 and 6-10)_
The resoldering of an indium solder joint with tin-lead solder can result in a
weakened joint (Fig. 6-11)- It may appear superficially (i.e., optical examination)

Figure 6-7 Cross section of 50% In-50% Sn solder joint. lower right, dark region: copper
substrate. Upper left: solder. The darker and lighter regions of the solder do not identify the
two different phases of the indium-tin solder but are artifacts. X2000, backscattered.
 SEMIEDX CHARACTERIZATION-GENERAL 181

Figure 6-8 Tin dot map of Figure 6-7. Darker regions of solder: tin-rich phases. Lighter
regions of solder: indium-rich phases.

Figure 6-9 Expansion of Figure 6-7. Lower right, dark region: copper substrate. Between
the solder and substrate lies the indium-tin-copper intermetaliic, about 2 j.Lm thick. X4000,
backscattered .
182 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-10 Cross section of 50% In-50% Pb solder joint. Upper right, dark region: copper
substrate. Center region: 50% In-50% Pb solder joint. Inbetween is the indium-copper inter-
metallic, about 2 J.l.m thick. X4000, backscattered.

Figure 6-11 Cross section of 50% In-50% Sn solder joint resoldered with 60% Sn-40% Pb.
The original indium-tin solder joint was positioned at the top, and the 60% Sn-40% Pb solder
intrusion was from the bottom . lightest phase, at bottom: lead-indium, high lead concentration.
Darkest phase, scattered throughout view: indium-tin. Gray phase, mostly at top: lead-indium,
low indium concentration. Commonly, only the darkest and lightest phases are observed in a
resoldered joint. XlOOO, backscattered .
 SEM/EDX CHARACTERIZATION-GENERAL 183

Figure 6-12 Cross section of failing 50% In-50% Sn solder joint resoldered with 60% Sn-
40% Pb. Lightest phase: lead-indium . Dark phase: indium-tin . X1000, backscattered .

that the solder joint is normal, but SEM shows an unusual combination of phases
that later may promote failure. In Figure 6-11, three different phases are shown,
but more commonly only two phases of indium-tin and indium-lead are observed.
A fracture develops between these two domains and propagates through the
sample (Fig. 6-12). The mechanism of this failure is similar to that for tin-lead
(see below)-a fracture develops mainly at the interface between different do-
mains-but the process is accelerated for indium-tin-Iead, which are composi-
tions of inferior fatigue resistance.

6.8.5 Bismuth Solders

The appearance of 43% Sn-43% Pb-14% Bi (Fig. 6-13) is very similar to that
of tin-lead solders, with two phases-tin and lead-bismuth.9 This solder has two
very attractive properties that ordinarily are mutually exclusive-a low melting
point (ca. 144 to 163°) and good fatigue resistance properties. IS Other bismuth
solders include tin-lead-bismuth (16%-32%-52%), melting at 96° and 42% Sn-
58% Bi melting at 138°. Although these latter two solders previously had a good
reputation for reasonable shear strengths and fatigue properties,16,17 they suffer
from the disadvantage that the addition of tin-lead (a common occurrence during
soldering to tin or tin-lead coated substrates) can actually lower the melting point
(by approaching the tin-lead-bismuth ternary eutectic) and cause rapid failure. IS
 184 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-13 Cross section of43% Sn-43% Pb-14%Bi. Darker domains: tin. Lighter domains:
lead-bismuth . X2000, backscattered .

Figure 6-14 Cross section of solder joint to cadmium-plated brass. Lower right, dark region :
brass. Solder region: tin and lead domains, heavily admixtured with cadmium . Interface be-
tween substrate and solder: copper-cadmium . Cadmium needles are reaching from the interface
into the bulk of the solder. X600, backscattered.
 SEM/EDX CHARACTERIZATION-GENERAL 185

Figure 6-15 Cross section of "weld" of two tinned copper wires. X360, backscattered.

Figure 6-16 Expansion of Figure 6-15. The boundary between the two copper wires is a
90%-10% composition of copper-tin and hence is not an ordinary solder. X6000, backscattered.
186 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-17 Cross section of alloy J. The lighter regions are silver-rich. Xl0,OOO, back-
scattered.

Miscellaneous

Soldering to cadmium plate is not as common as in the past since it is now real-
ized that this creates new compounds and intermetallics (Fig. 6-14). Tin-lead sol-
ders under special situations can actually create a new composition of "solder"
(Figs. 6-15 and 6-16)-with tin-plated copper wires, high temperatures (lOOO°C)
produced by resistance welding technique can produce an intermediate alloy with
a composition of 90% Cu-lO% Sn. 9 Alloy J, a high-melting rapid solidification
(RS) alloy,IS has a unique appearance with small domains (Fig. 6-17).

SOLDERABILITY ISSUES
Maintaining Solderability

Wetting of a copper substrate can clearly be prevented if the surface is oxidizedll ,14
(Fig. 6-18). However, a tinned surface can also experience loss of solderability
if copper from the substrate migrates to the surface with subsequent oxidation l9
(Fig. 6-19). For brass (Figs, 6-20 through 6-22), both copper and zinc mi-
grate. 20 ,21 Copper migrates via the growth of intermetallic CU6Sn5,19 and zinc
moves more rapidly via diffusion. 2o ,21 The migration of copper can be prevented
by an adequate thickness of tin or solder (perhaps 400 jJ, in.), and the diffusion
of zinc can be prevented by a copper barrier. Alternatively, migration of copper
and zinc can be prevented by a nickel barrier.
 SOLDERABILITY ISSUES 187

Figure 6-18 An attempt to solder a heavily oxidized copper surface. X300, secondary.

Figure 6-19 This tinned copper substrate, over a period of a year, experiences migration
of copper from the substrate in the form of Cu6Sns. These crystals of CU6SnS are sometimes
masked by extensive oxidation (darker regions) of the copper. X2000, secondary.
188 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-20 Cross section of tinned substrate with extensive base metal migration. Below:
brass substrate. Middle portion : tinned layer. X4000, backscattered.

Figure 6-21 Copper dot map of Figure 6-20. CU6SnS has migrated extensively through the
tinned layer.
 SOLDERABILITY ISSUES 189

Figure 6-22 Zinc dot map of Figure 6-20 . Zinc has diffused rapidly through the tinned
layer.

6.9.2 Inadequate Tin Protective Coatings

Figures 6-19 through 6-22 above exemplify the dangers of an inadequate tin
protective barrier for copper or brass since the base metal can migrate to the
surface and oxidize. Nevertheless, a freshly tinned surface will solder satisfac-
torily, even with minimum thickness. However, its appearance is unusual. Fig-
ures 6-23 through 6-27 show such a surface with a tin thickness gradient.

6.9.3 The Dangers of "Forcing" Poor Solderability

Occasionally a situation of poor solderability is "solved" by excessive heating,

to "force" the soldering process. This excess operation may be performed, for
example, by extended periods in direct contact with a soldering iron whose tip
temperature may far exceed usual wavesoldering, vapor phase, or infrared re-
flow operations. Evidence for this extended heat treatment is conspicuous: C~Sn5
chunks (Fig. 6-28) suspended in the solder. Nickel, also, can form intermetal-
lies; upon extended heat treatment, nickel-tin intermetallic can grow in the form
of prisms from a nickel barrier and can fracture and drift in the solder. If a
silver-containing solder is used, Ag 3Sn crystals sometimes develop (Figs. 6-29
and 6-30).
190 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-23 Inadequately tinned copper surface (fresh). EDX indicates only the presence
of tin-copper bimetallic. XlOOO, secondary.

Figure 6-24 Similar situation to Figure 6-23, with slightly more tin . The upper left-hand
corner indicates incomplete fusion . X1000, secondary.
 SOLDERABIlITY ISSUES 191

Figure 6-25 Backscattered view of Figure 6-24 . Upper left corner, dark: tin-copper. Re-
mainder: dark regions, tin; light regions, lead. X1000, backscattered.

Figure 6-26 Similar situation to Figure 6-24 but with even more tin. X1000, backscattered.
192 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-27 Backscattered view of Figure 6-26. The tin-copper bimetallic regions are now
confined to isolated areas. Xl 000, backscattered .

Figure 6-28 Cross section of solder joint, excessively heated to "solve" problem of non-
solderability. The large chunks are CU6SnS. X500, backscattered.
 SOLDERABILITY ISSUES 193

Figure 6-29 Expansion of Figure 6-28. The large vertical needle is Ag 3Sn . X1000, back-
scattered .

Figure 6-30 Expansion of Figure 6-29. Upper right corner, dark region: copper substrate.
Careful inspection can differentiate the lead (light regions) from the darker regions-the tin
(mosaic) and the Ag3Sn (needles) and the CU6Sn5 (smaller, darker needles, generally emanating
from the substrate interface. X2000, backscattered.
194 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX )

6.10 RELIABILITY ISSUES-LEACHING OF SUBSTRATE

Silver is commonly plated onto ceramic to permit soldering to this substrate.

Scavenging of this silver is possible when excess heating is experienced during
the soldering process, and failure of the silver-ceramic bond can occur. If the
component is originally defective, tin intrusion has already occurred into the
silver metallization and failure is difficult to avoid. Effective inspection may be
difficult because the silver metallization may superficially appear to be sound
(Figs. 6-31 and 6-32), and a silver dot map may show minimal scavenging of
the silver into the solder (Fig. 6-33). However, tin intrusion may be sufficiently
severe to scavenge the silver from the substrate (Fig. 6-34). Push tests of this
component easily shear the metallization away from the ceramic substrate-the
component failure rate is high. An acceptable metallization layer (Fig. 35) dis-
plays an intact silver layer with no tin intrusion. Push tests of this component
demonstrate a strong bond between the metallization and the ceramic, and the
component passes quality control inspection. After reftow, an intact silver me-
tallization layer remains (Fig. 6-36). An example of extended heat treatment
during reftow shows complete scavenging of the silver (Fig. 6-37).

Figure 6-31 Cross section of a failing ferrite inductor. Below, dark region: ferrite. Above,
62 % 5n-36% Pb-2% Ag solder. Between the ferrite and the solder lies the silver metallization,
apparently intact but nevertheless defective. X480, backscattered .
 RELIABILITY ISSUES-LEACHING OF SUBSTRATE 195

Figure 6-32 Expansion of Figure 6-31. The silver metallization appears to be sound. How-
ever, see Figure 6-34. X2000, backscattered.

Figure 6-33 Silver dot map of Figure 6-32 . The silver layer is not extensively leaching away
into the solder. However, see Figure 6-34. X2000.
196 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-34 Tin dot map of Figure 6-32. The tin of the solder has extensively intruded into
the silver metallization.

Figure 6-35 Cross section of acceptable ferrite inductor. Superficially, this view may even
appear to be inferior to the faulty inductor because the metallization layer is thinner. However,
tin dot maps demonstrate tin intrusion is minimal. X2000, backscattered.
 RELIABILITY ISSUES-LEACHING OF SUBSTRATE 197

Figure 6·36 Cross section of specimen of Figure 6-35 after reflowing. Minimal temperature
excursion was experienced, and the silver metallization is still intact.

Figure 6·37 Cross section of ferrite inductor with extended heat treatment. The silver has
been largely leached away into the solder, developing Ag3 Sn needles. This component quickly
fails. X480, backscattered.
198 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-38 Cross section of a faulty solder joint. Below, dark region: alumina ceramic.
Above, 62% Sn-36% Pb-2% Ag solder. Between the ceramic and solder is the silver-palladium
metallization . Xl 000, backscattered.

In other cases, the silver may be partially scavenged, and yet the component
is not defective. The intrusion of tin into the metallization is again the key. In
a defective reflowed solder joint (Fig. 6-38), a silver metallization appears to
remain between the solder and the alumina ceramic. However, a silver dot map
(Fig. 6-39) demonstrates that the silver is dispersing into the solder, and a tin
dot map shows complete tin intrusion to the ceramic substrate (Fig. 6-40). A
lead dot map (Fig. 6-41) shows no unusual distribution of this element. Push
tests of this component fail.
An acceptable solder joint (Fig. 6-42) superficially appears similar to the faulty
component, with a silver dot map (Fig. 6-43) demonstrating some silver mixing
with the tin. Nevertheless, a distinct layer of original conductor remains next to
the ceramic. The tin dot map shows that tin does not intrude into this original
conductor layer (Fig. 6-44). In this satisfactory solder joint, the lead distributions
(Fig. 6-45) are again not unusual. Push tests of this component pass because a
silver layer to the ceramic is intact.
A common solution to the silver scavenging problem is a nickel barrier. A
series of micrographs show how the integrity of this system is preserved in a
chip capacitor soldered to a metallized ceramic substrate (see Figs. 6-46 through
6-50).
 RELIABILITY ISSUES-LEACHING OF SUBSTRATE 199

Figure 6-39 Silver dot map of Figure 6-38. Extensive scavenging of the silver into the solder
is occurring. X1000.

Figure 6-40 Tin dot map of Figure 6-38. Tin intrusion is complete to the alumina ceramic.
X 1000.
 200 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-41 Lead dot map of Figure 6-38. Xl000.

Figure 6-42 Cross section of satisfactory solder joint. Below, dark region: alumina ceramic.
Above, 62% 5n-36% Pb-2% Ag solder. Between the ceramic and solder is the silver-palladium
metallization . Xl000, backscattered.
 RELIABILITY ISSUES-LEACHING OF SUBSTRATE 201

Figure 6·43 Silver dot map of Figure 6-42. Although the silver mixes with the tin, a distinct
layer of original conductor remains next to the ceramic. X1000.

Figure 6·44 Tin dot map of Figure 6-42. The tin does not penetrate to the ceramic. X1000.
202 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-45 Lead dot map of Figure 6-42.

Figure 6-46 Cross section of chip capacitor to metallized ceramic. Far left, dark region:
alumina ceramic substrate. Far right, gray region: silver metallization to chip capacitor (not
shown). Middle of micr.ograph: 62% Sn-36% Pb-2% Ag solder. Between the solder and the
alumina lies the silver-palladium metallization. Between the chip metallization and the solder
lies the nickel barrier. X1300, backscattered.
 RELIABILITY ISSUES-LEACHING OF SUBSTRATE 203

Figure 6-47 Silver dot map of Figure 6-46. The silver metallization to the alumina ceramic
is intact. X1300.

Figure 6-48 Tin dot map of Figure 6-46. The tin does not intrude to the silver metallization
to the left. X1300.
 204 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-49 Lead dot map of Figure 6-46. X1300.

Figure 6-50 Nickel dot map of Figure 6-46. This is the nickel barrier between the solder
and silver end cap of the chip capacitor. X1300.
 RELIABILITY ISSUES- GOLD EMBRITTLEMENT 205

6.11 RELIABILITY ISSUES-GOLD EMBRITTLEMENT

It is well understood that contamination can frequently cause solderability or

reliability problems, but it is not universally understood that gold-a common
protective coating to preserve solderability 22-can cause reliability problems. 23
Figures 6-51 through 6-57 display the morphological characterization of soldering
to gold-plated substrates. Superficially the solder joint has a monotonic com-
position (Fig. 6-51), but a backscattered micrograph (Fig. 6-52) manifests a great
deal of elemental redistribution, where tin from the solder is scavenged by the
gold underneath to produce a lead-rich area. Further magnification shows that
the interface between the solder and gold possesses gold-tin intermetallics (Figs.
6-53 and 6-54). In some regions the scavenging of the tin is sufficient to leave
only lead-rich metal, which does not wet well with the gold-plated substrate
(Fig. 6-55). Further up the solder slope, gold occasionally migrates to the surface
to produce AuSI4 needles (Fig. 6-56). A backscattered view of this region
displays the redistribution of AuSI4 not only in the form of needles but also in
the form of large aggregates among the lead spheroids (Fig. 6-57).
A cross-sectional characterization of soldering to gold-plated substrates is
shown in Figures 6-58 through 6-60. If the gold layer is thick (perhaps above

Figure 6-51 Solder joint to gold-plated substrate. Region A: gold-plated nickel substrate.
Region B: 62 % Sn-36% Pb-2% Ag solder. X60, secondary.
206 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-52 Backscattered view of Figure 6-51 . The solder region is divided into the " nor-
mal" region (B) and the tin-deficient region (C). Region C consists mainly of lead domains.
Xl 00, backscattered.

Figure 6-53 Expansion of Figure 6-51. Top region (to the right): lead spheroids. Bottom
region (to the left): gold-plated substrate. Intermediate region : gold-tin intermetallic region.
Xl000, secondary.
 RElIABILITY ISSUES-GOLD EMBRITTLEMENT 207

Figure 6-54 Expansion of Figure 6-53. Top region (to the right): lead spheroids. Bottom
region (to the left) : gold-plated substrate. In the intermediate region, the gold-tin intermetallics
are now clearly visible in the form of AuSn4 needles. X3000, secondary.

Figure 6-55 Interface area where tin has been thoroughly scavenged, leaving only lead-
rich solder, which does not wet well with the gold-plated surface. X1200, secondary.
 208 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-56 Area in solder where gold has migrated to the surface, producing AuSn4
needles. Left: lead-rich spheroids. X540, secondary.

Figure 6-57 Backscattered view of Figure 6-56. Left: light regions, lead; gray regions, AuSn4'
Right: dark regions, tin; light areas, lead; gray needles, AuSn4 ' X1000, backscattered.
 RELIABILITY ISSUES-GOlD EMBRITTlEMENT 209

Figure 6-58 Cross section of solder joint with excess original thickness of gold. The original
gold layer was 5 j.l.m thick, and this solder joint has been reduced to 3 j.l.m . Bottom, dark
region : nickel substrate. White horizontal strip: gold layer. Solder: dark regions, tin; light
regions, lead; gray regions, AuSn4. X1000, Backscattered.

Figure 6-59 Cross section of solder joint with excess original thickness of gold, and reflowed
with excess solder in an attempt to remove gold-tin intermetallics from the substrate. However,
large amounts of AUSn4 still adhere to the substrate. Bottom, dark region : nickel substrate.
White horizontal tongue, leading from right: gold layer. Solder: dark regions, tin; light regions,
lead; gray regions, AUSn4. X1000, backscattered .
210 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-60 Expansion of Figure 6-59 at the interface between the solder and gold. Bottom
white region: gold. Solder: light regions, lead; dark regions, tin; gray areas, AUSn4. In addition,
AuSn2 (arrows pointing right) and AuSn (arrows pointing left) appear in very minor amounts.
X2700, backscattered.

3 J.Lm), a residual layer can remain on the substrate (Fig. 6-58), with extensive
gold-tin intermetallics_ Even if excess solder is used so that all the gold is
redistributed (Fig. 6-59), considerable amounts of gold-tin intermetallics still
may remain_ Most of these intermetallics are AuSI4, although at the interface
of the gold and solder itself sometimes AuSn2 and AuSn are found (Fig. 6-60).
Reliability issues arise with soldered gold-plated substrates when gold-tin
intermetallics still adhere to the soldered substrate (Fig. 6-61). Since the gold-
tin intermetallics are brittle, they can crack through normal temperature excur-
sions and cause failure (Fig. 6-62). The examples given here are for gross failures;
however, deterioration in performance can be experienced even before this crack-
ing can be observed by SEM. The solution to the problem is to ensure that no
gold-tin intermetallics remain at the substrate_ If the soldering process guarantees
that the gold-tin intermetallics drift away from the substrate (Fig. 6-63) so that
the solder interface is substrate-tin (Fig. 6-64), gold embrittlement should not
be a concern_
 RElIABILITY ISSUES-GOLD EMBRITTLEMENT 211

Figure 6-61 Cross section of gold-embrittled, failing solder joint. The cracks are developing
at the gold-tin intermetallic. left, dark region: nickel substrate. Next to the substrate, white
vertical strip: gold. Solder: dark regions, tin; light regions, lead; gray regions, gold-tin inter-
metallics. X1000, backscattered.

Figure 6·62 Cross section of gold-embrittled, failed solder joint. The horizontal crack
between the gold and solder is complete. Bottom: copper substrate. Next to the substrate,
white horizontal strip: gold. Top: solder. X3000, secondary.
212 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-63 Cross section of well-soldered joint to gold-plated substrate. The AuSn4 la-
mellae have drifted well into the bulk of the solder. Dark regions: tin. Light regions : lead.
Gray regions: AuSn4' Xl 000, backscattered.

Figure 6-64 Cross section of well-soldered joint to gold-plated substrate at the solder-
substrate interface. The gold-tin intermetallics have drifted far from the interface, leaving the
tin to form the normal solder interface (arrows). Bottom, horizontal dark area: nickel substrate
with nickel-tin intermetallic. Solder: dark areas, tin; light areas, lead; gray areas, AuSn4 ' X3900,
backscattered .
 RELIABILITY ISSUES-FATIGUE 213

6.12 RELIABILITY ISSUES-FATIGUE

Solder fatigue is a process whereby mismatch of the thermal coefficients (TeE)

of different materials,24 coupled with an inherent design deficiency, causes stress
and/or strain cycling in a solder joint, ultimately resulting in failure. 25 ,26 This
fatigue can occur either isothermally16,17,27,28 or during thermal cycling.29 The
isothermal process can be more easily characterized,16,17 but the latter process
more closely describes the true field situation. 30 The duration of the stress is
important in determining the rate of fatigue. 31,32 The sequence of fatigue (Figs.
6-65 through 6-69) progresses from a pristine solder joint, through crazing (de-
velopment of a matted surface) and pitting (development of small cavities), to
cracking and failure. Hard solders can experience more catastrophic failure (Fig.
6-70) (catastrophic failure is the rapid development of failure after a long, quies-
cent induction period).
Surface SEMs (i.e., external appearance) during this fatiguing process have
different appearances for various solders, principally because of different duc-
tilities of these solders. 11 Tin-lead solders show the extensive working of the
material, leading to the crazing phenomenon (Figs. 6-71 and 6-72). Other soft
solders (Figs. 6-73 and 6-74) exhibit similar appearances, with a higher degree

Figure 6-65 Pristine solder joint (60% 5n-40% Pb). X40, secondary.
214 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-66 Fatiguing solder joint (60% Sn-40% Pb) progressing through the initial stage
of crazing. X40, secondary.

Figure 6-67 Fatiguing solder joint (60% Sn-40% Pb) progressing through the stage of pitting.
X40, secondary.
 RELIABILITY ISSUES-FATIGUE 21 5

Figure 6·68 Fatiguing solder joint (60% Sn·40% Pb) beginning to crack. X40, secondary.

Figure 6·69 Complete failure of fatiguing solder joint (60% Sn-40% Pb). X40, secondary.
216 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-70 Complete failure of 95% Sn-5% Ag solder joint. The superficial evidence of
fatiguing (crazing, pitting) observed with tin-lead solders does not exist here; instead, cata-
strophic failure occurs. X40, secondary.

Figure 6-71 Surface SEM of crazing 60% Sn-40% Pb solder. X500, secondary.
 RELIABILITY ISSUES-FATIGUE 217

Figure 6-72 Expansion of Figure 6-71. X4000, secondary.

Figure 6-73 Surface SEM of 50% In-50% Pb solder joint before fatiguing. X500, secondary.
218 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X·RAY (SEMIEDX)

Figure 6-74 Surface SEM of 50% In-50% Pb solder joint after failure. X2000, secondary.

of creep. The behavior of hard solders (Figs. 6-75 and 6-76) reflect the cata-
strophic behavior of this class of alloys: failure along a definite slip, with the
remainder of the surface relatively unchanged.
Cross-sectional analysis of tin-lead solders shows a well-defined and repro-
ducible fatiguing mechanism. 33 ,34,35 Failures develop along the area of mechan-
ical activity (movement) that enlarges the tin and lead domains, weakening the
structural material (Figs. 6-77 and 6-78). Long before failure occurs, this de-
veloping site of domain enlargement can be observed (Fig. 6-79).
The mechanism for tin-lead-silver solders (62% Sn-36% Pb-2% Ag) appears
the same as for 60% Sn-40% Pb or 63% Sn-37% Pb, with one additional feature-
silver concentrates into larger domains during the developing fatigue (Fig. 6-
80). Tin-lead-silver solders are more durable than tin-lead solders,30 probably
because of their greater shear and creep strengths. ll
For soft solders, cross-sectional analysis does not show this mechanism of
domain enlargement. The higher ductility of these solders can significantly affect
their proper application. 36 For hard solders, failure can frequently occur at the
intermetallic30 (Fig . 6-81).
 RELIABILITY ISSUES-FATIGUE 219

Figure 6-75 Surface SEM of 95% Sn-5% Ag solder joint during fatiguing. Xl 000, secondary.

Figure 6-76 Surface SEM of 95% Sn-5% Ag solder joint after failure. X1000, secondary.
220 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6·77 Cross section of fatiguing 60% Sn-40% Pb solder joint. Left, dark region: copper
pin. Solder: dark domains, tin; light domains, lead. The area of mechanical activity in this
example is occurring parallel to the solder joint, resulting in domain enlargement and eventual
failure of the solder material. X470, backscattered.

Figure 6·78 Enlargement of Figure 6-77. X780, backscattered .

 RELIABILITY ISSUES-FATIGUE 221

Figure 6-79 Cross section of solder joint at initial stages of fatiguing . Careful inspection
discloses an area of mechanical activity (middle of view, vertical zone). X300, backscattered .

Figure 6-80 Silver dot map of cross section of fatiguing 62% Sn-36% Pb-2 % Ag solder.
The silver is aggregating into domains. This domain enlargement occurs not only at the fatiguing
site but throughout the solder. Xl 000.
222 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

Figure 6-81 Cross section of failed 95% 5n-5% Ag solder joint. A major portion of failure
occurs at the intermetallic itself. Left: copper pin . Right: solder. X2000, backscattered.

REFERENCES

1. Verhoeven, I. D. , "Scanning Electron Microscopy ," Metals Handbook Ninth Edition

Volume 10 Materials Characterization, American Society for Metals, Metals Park,
OH, 1986, pp. 491-515 .
2. Heinrich, F. I., and D. E. Newbury, "Electron Probe X-Ray Microanalysis," Metals
Handbook Ninth Edition Volume 10 Materials Characterization, American Society
for Metals, Metals Park, OH, 1986, pp. 516-535.
3. Leyden, D. E., "X-Ray Spectrometry," Metals Handbook Ninth Edition Volume 10
Materials Characterization, American Society for Metals, Metals Park, OH, 1986,
pp.82-101.
4. Buehler Digest, vol. 25, no. 1. "Cross-Sectioning Techniques for Analysis of Mi-
croelectronic Components ." Lake Bluff, IL: Buehler, Ltd. , 1988.
5. - - - , vol. 25, no. 2. "Cross-sectional Analysis of Printed Circuit Boards." Lake
Bluff, IL: Buehler, Ltd., 1988.
6. - - - , vol. 25, no. 3. "Cross-sectional Analysis of Soldered Connections , Con-
nectors, and Electrical Components." Lake Bluff, IL: Buehler, Ltd., 1988.
7. Nelson, I . A., "Basic Steps for Cross Sectioning PCB's," Insulation/Circuits , Lake
Publishing Corporation, Libertyville, IL, May 1977.
8. Nelson, J. A., "How Microstructural Analysis of Printed Circuit Boards Provides a
Versatile Quality Control Method, " Electrionics, Lake Publishing Corporation, Lib-
ertyville, IL, July 1984.
 REfERENCES 223

9. Hansen, M., and K. Anderko, Constitution of Binary Alloys, McGraw-Hill, New

York, 1958.
10. Lyman, Ed., "Metallography, Structures, and Phase Diagrams," Metals Handbook,
vol. 8, American Society for Metals, Metals Park, OH, 1973.
11. Wassink, R. J. K., Soldering in Electronics. Electrochemical Publications Ltd., Ayr,
Scotland, 1984.
12. Goldstein, J. Practical Scanning Electron Microscopy, Plenum Press, New York,
1975.
13. Goldstein, J. Scanning Electron Microscopy and X-ray Microanalysis: A Text for
Biologists, Materials Scientists, and Geologists, Plenum Press, New York, 1981.
14. Manko, H. H., Solders and Soldering, 2nd ed., McGraw-Hill, New York, 1979.
15. Marshall, J. L., "Scanning Electron Microscopy Characterization of Solder Failure-
Standard and Nonstandard Joints," Proceedings NEPCON '90, Anaheim, CA, Feb-
ruary 1990, pp. 1361-1370.
16. Wild, R. N., "Some Fatigue Properties of Solders and Solder Joints," IBM No.
73ZOOO421, IBM Publications, Owego, NY, January 1973.
17. - - , October 1975. "Some Fatigue Properties of Solders and Solder Joints," IBM
No. 74Z000481, IBM Publications, Owego, NY, October 1975.
18. Olsen, D. R., and K. G. Spanjer, Solid State Tech., September 1961, pp. 121-125.
19. Kay, P. K., and C. A. MacKay, "The Growth of Intermetallic Compounds on
Common Basis Materials Coated with Tin and Tin-Lead Alloys," Publication 517,
International Tin Research Institute, Columbus, OH.
20. Carter, G. F., Principles of Physical and Chemical Metallurgy, American Society
for Metals, Metals Park, OH, 1979.
21. Askill, J., Tracer Diffusion Data for Metals, Alloys, and Simple Oxides, pp. 45-
54. New York: IFVPlenum Press, 1970.
22. Thwaites, C. J., "Soldering Gold Surfaces," Electroplating & Metal Finishing, no.
26, 1973, pp. 21-26.
23. Duckett, R., and M. L. Ackroy, "The Influence of Solder Composition on the
Embrittlement of Soft-soldered Joints on Gold Coatings," Publication 520. Inter-
national Tin Research Institute, Columbus, OH.
24. Wolverton, W. M., Brazing and Soldering, Vol. 13, Autumn 1987, pp. 33-38.
25. Manson, S. S., Thermal Stress and Low-Cycle Fatigue, McGraw-Hill, New York,
1966.
26. Coffin, L. F., Jr., "Introduction to High-temperature Low-cycle Fatigue," Experi-
mental Mechanics, May 1968, pp. 218-224.
27. Solomon, H. D., "Fatigue of 60/40 Solder," Proceedings 1986 36th Electronics
Components Conference, 1986, p. 622.
28. Stone, D., H. Wilson, R. Subrahamanyan and c.-y. Li, "Mechanisms of Damage
Accumulation in Solder during Thermal Fatigue," Proceedings 1986 36th Electronics
Components Conference, 1986, p. 630.
29. Shine, M. C., L. R. Fox, and J. W. Sofia, "A Strain Range Partitioning Procedure
for Solder Fatigue," Brazing and Soldering, Vol. 11, no. 9, 1985, p. 11.
30. Marshall, J. L., and S. R. Walter, Int. J. Hybrid Microel., Vol. 10, 1987, pp. 11-
17.
31. Engelmaier, W. "Functional Cycles and Surface Mounting Attachment Reliability,"
Surface Mount Technology, 1984, p. 87.
224 SCANNING ELECTRON MICROSCOPY AND ENERGY DISPERSIVE X-RAY (SEM/EDX)

32. Coffin, L. F., Jr., "Low-Cycle Fatigue: A Review," Applied Materials Research,
vol. 1, October 1962, pp. 129-141.
33. Lichtenberg, L. R., "Comparison of Environmental Thermal Cycle Tests on Reftow
Soldered Assemblies," ISHM 1984 Proceedings, The International Society for Hybrid
Microelectronics, Reston, VA, 1984, p. 65.
34. Wright, E. A., and W. M. Wolverton, "The Effect of the Solder Reftow Method
and Joint Design on the Thermal Fatigue Life of ~adless Chip Carrier Joints,"
Proceedings 1984 34th Electronics Components Conference, 1984, p. 149.
35. Smeby, J. M., "Solder Joint Behavior in HCCIPWB Interconnects," Proceedings
1984 34th Electronics Components Conference, 1984, p. 117.
36. Coombs, V. D., "Fatigue Properties of Pure-Metal Solders," ISHM 1976 Proceed-
ings, The International Society for Hybrid Microelectronics, Reston, VA, 1976, p.
67.
 7
The Role of Microstructure
in Thermal Fatigue of Pb-Sn
Solder Joints
J. w. Morris, Jr., D. Tribu/a, T. S. E. Summers,
and D. Grivas

The Pb-Sn eutectic alloy is widely used as a joining material in the electronics
industry. In this application the solder acts as both an electrical and mechanical
connection within and among the different packaging levels in an electronic
device. Advances in packaging technologies driven by the desire for miniatur-
ization and increased circuit speed result in severe operating conditions for the
solder joint and thus solder joint reliability problems. Specifically, the mis-
matched thermal expansion characteristics of the materials joined by the solder
and the cyclic temperature fluctuations normally encountered during service
constitute a condition of thermal fatigue for the constrained solder. This is
especially a problem in the surface mounting of leadless components where a
shearing of the solder joint occurs with each temperature excursion. Repeated
temperature cycling, such as that associated with Joule heating as the device is
turned on and off, and environmental temperature changes fatigue the solder
joint ultimately causing its failure. Increased solder joint reliability and the
development of new more fatigue-resistant solder alloys require a fundamental
understanding of the metallurgical mechanisms operative during the fatigue cycle
and, hence, an understanding of the solder microstructure.
From the metallurgical perspective the thermal fatigue problem in solder con-
tacts is at least as complex as any fatigue problem that has been researched.
Since the solder is mechanically soft and is used at a high homologous temper-
ature, deformation is introduced by plasticity and creep throughout the strain
cycle. In addition, the cyclic deformation is affected by the fact that the solder
contact is a thin layer bonded to relatively rigid materials with a complex in-
termetallic layer at the bonding plane. An understanding of the deformation is
especially complicated by the inherent complexity of the as-solidified eutectic

225
226 THE ROlE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOlDER JOINTS

solder microstructure and its instability as it is cycled and aged. The problem
of predicting the rate of thermal fatigue in a solder joint is formidable; there is
no reliable predictive theory for thermal fatigue under creep conditions for even
simple, single phase, monolithic materials. Nonetheless, mechanistic research
on the fatigue and related deformation modes of solder joints can provide valuable
information to help guide the design of accelerated tests and the development
of improved solders.
The work that is reviewed here concentrates on the role of the solder micro-
structure and its association with the mechanisms of thermal fatigue. Particular
emphasis is placed on a fundamental understanding of the microstructure-property
relationship in Pb-Sn alloys and its dependence on processing history. It has
been observed that in all cases where shear deformation at elevated temperatures
is a factor, the solder fails by the same basic mechanism-inhomogeneous de-
formation of the solder microstructure resulting in inhomogeneous recrystalli-
zation and softening and thus joint failure. The metallurgical causes of this
mechanism are outlined here. Possible paths toward the defeat of this failure
mechanism in the design of superior, more fatigue-resistant solder alloys are
discussed and evaluated.

7.1 EXPERIMENTAL DETAILS

This research on the microstructural mechanisms of fatigue in solder has included

the design of samples that permit controlled deformation of well-characterized
solder joints, the development of appropriate metallographic procedures, and the
design and construction of equipment to test joints under various loading con-
ditions including high temperature creep in shear, isothermal shear fatigue, and
thermal fatigue. The specimen design and mechanical testing procedures are such
that they are capable of producing the loading conditions and geometries of real
solder joints; since thermal fatigue failures are usually associated with shear
loading, these have been emphasized. The mechanical tests conducted as part
of this work include tests to failure in shear, isothermal and thermal fatigue
testing, and isothermal creep testing in shear. The specimens designed for these
experiments are such that they will produce loading in a well-defined stress state
under simple shear, for shear to failure, creep, and isothermal and thermal fatigue
conditions. Figures 7-1 through 7-3 show the specimens used in this work.
Metallographic preparation techniques were developed l - 3 that ensure flat solder
joint surfaces for detailed examination in both optical and SEM microscopes.
The specimen in Figure 7-1 is that used in shear to failure, isothermal shear
fatigue, and creep in shear experiments. The design is such that, under loading
in tension, the solder joints that correspond to the shaded regions in Figure 7-1
are loaded in nearly simple shear. Manufacture of the specimen consists of
assembling polished and fluxed copper plates with appropriate spacers to obtain
 EXPERIMENTAL DETAILS 227

..........
...... ......
.,
'",'" ,,,
,,, ,
~

,'"
,,, ............. ,
,", , ,"
,'"" ~::~:::::::::
,,,, , 6.4 mm
,,,
,'", ,
,'"
'"
3.2 mm ~::~:;~:~~,4,~,~,~,~,~,=r:::
~:~:~, ,............ ..., ..., .. ,
~;:;: ; ; : : : :: """""'/1

:;:;:,
"',',"
...... "...... ',"',",
.... " .... " ........
:':-: ;:;:;:::;~ I-i:~o 5 mm
,',', ;<":«~~~>I·
:;:;: ;:;:;:;:;:;:;~ :;:;:;
...' ...' ...

~:~:~: 89 mm

"", 't,·
~ .. , ...;.. '
.. .. ... 'to 'to .. ..

3.2 mm ~ .....

~ ~ ~ 11r~
',','" ;:~:;:;::::~;
()
:::::;
9.Smm
_
Sn-Pb
Sn-Pb in shear

3.2 mm
~: 1:1~ ~: 1: : ;
~1~1~: g1~g:~: ~:l:l:
:~{
..
;:;:;:;:;:;:; :;:;:;
, .. .. .... .. ,',;; ..
, ,' , ..., , ... ,
~;~;~ ;~:~;::::~::: :;~:::
:~:~: ,:,»:,:,:,~ :':':'
. . ;«««E; «<
~: ~; :' ,';~,: ' ::;:;:;:;:;:;~ :;:;:;
~:). {~~~~~~~~~~ ~~~~~~
;~;:::::::::: :;:;:;
:':':' ;:::::::::::: ::::::
~< . .'.. ,<:;~:" "«<

Figure 7-1 The specimen configuration used for shear strength, isothermal shear fatigue,
and creep testing. The specimen is a multilayer copper-solder sandwich. On axial loading,
the two solder joints, shaded regions in figure, are loaded in nearly simple shear. Dimensions
are in millimeters.
228 THE ROLE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOLDER JOINTS

desired solder joint width. The assembly is then immersed in a molten solder
bath. After cooling, the excess solder is removed and the block is sliced to yield
several specimens such as that shown in Figure 7-1. Figure 7-2 shows the second
specimen configuration used in the creep tests. Unlike the specimen in Figure
7-1, this geometry consists of 18 individual and separate solder joints within
each specimen (9 on either side of the central plate). Again, under loading in
tension, the solder contacts are deformed in simple shear. To make these spec-
imens, polymeric circuit boards with plated Cu pads are assembled together and
immersed in molten solder. These two specimen configurations differ most im-
portantly in the solder volume per joint and thus the cooling rate of the solder
material.
The thermal fatigue specimen appears in Figure 7-3. Specimen manufacture
is similar to the specimen in Figure 7-1, In this case, the central Cu plate is

0.5" (12.7 mm)

1/16" 0( ..

(1.59 mm)
o

2.5"
(63.5 mm)
o
•••
•••
•••
o

Solder pad: 50mil x 80 mil

(1.27 mm x 2.03 mm)
thickness: 7 mil (0.178 mm)

1/8" (3.18 mm)

~

Figure 7-2 Specimen configuration used for creep testing of rapidly solidified solder joints.
The specimen consists of two sets of nine individual solder contacts (on Cu) on a polymeric
circuit board. On loading, an appropriate spacer is inserted between the two boards on top
to ensure symmetric loading in shear of the solder contacts under axial loading of the specimen.
The small solder volume per joint results in faster cooling rates of the solder material. (All
dimensions are in inches. Pad size 0.03 by 0.07.)
 EUTECTIC MICROSTRUCTURES 229

80 mm

,.
0.25 mm t
13.2 mm

Figure 7·3 Specimen used for thermal fatigue experiments. The dark central regions are
the solder joints. The mismatched thermal expansion characteristics of the constraining copper
and aluminum strain the solder joints in shear with each temperature excursion. Since the
strain magnitude varies linearly from the center of the specimen to a maximum strain at the
edges, the holes along the length of the solder joints serve to isolate individual solder regions
at a particular strain level.

replaced with a plate of Al (Ni and Cu plated to allow wetting by the solder).
After polishing, fluxing, and assembly with an appropriate spacer, the assembly
is immersed in molten solder. Slicing the block yields the individual specimens.
A final drilling procedure results in isolated solder joints along the length of the
specimens. The thermal expansion difference between the Cu and Al layers
imposes a shear strain on the solder; the magnitude of this shear strain increases
linearly from the center to the ends of the specimen. The holes along the length
of the specimen serve to isolate individual solder regions at a particular shear
strain.

7.2 EUTECTIC MICROSTRUCTURES

A central principle of materials science states that a material's properties originate

not only from its atomic composition but also from the atomic arrangements
within that material-the material's microstructure. However, for a material of
a given chemical composition, the material's microstructure is not constant but
can vary greatly with its processing history. It follows that if one is to understand
and control the mechanical behavior of solder joints, one must begin not only
with an understanding of the solder microstructure but also with an understanding
of its origin in terms of various processing parameters. In the case of solder
alloys, the application dictates use in the as·cast state, and hence, the cooling
rate of the alloy is the only variable processing parameter. The local cooling
rates of a solder contact depend on many factors, including the soldering process,
230 THE ROLE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOLDER JOINTS

the specific solder joint thickness and geometry, the composition and configu-
ration of the carrier board, and the neighboring components. To understand the
effect of cooling rate on solder joint microstructure and thus mechanical prop-
erties, it is instructive to first consider more simple solidification scenarios and
the resultant solder microstructures .

7.2.1 Lamellar Eutectics

In the case of preferentially oriented plane-front steady-state growth, the eutectic

Pb-Sn material grows as alternating lamellae of the two constituent phases parallel
to the direction of growth. Eutectic solidification is a cooperative growth process
since the solute rejected ahead of one phase region becomes immediately in-
corporated as the solvent phase in the adjacent region, and the plates thus grow
at the same rate. Many lamellae constitute a lamellar grain and will continue to
grow in such a coupled manner until contact with a mold wall or a similarly
growing grain. A distinct boundary delineates two adjacent eutectic grains; an
example appears in Figure 7-4. The lamellar spacing within the eutectic grain
is determined by the cooling rate. A faster cooling rate results in fine lamellae,

Figure 7-4 SEM micrograph of a eutectic Pb-Sn solder slowly cooled from 250°C. Slow
cooling results in a highly lamellar structure as shown. In the figure, two eutectic grains are
seen. Each grain is distinguished by a specific direction of lamellar arrangement, and a distinct
boundary delineates each grain . The light regions are the Pb-rich phase; the dark regions are
the Sn-rich phase.
 EUTECTIC MICROSTRUCTURES 231

and, in the case of a multigrained structure, the eutectic grains will be smaller
and the lamellar spacing within each grain will be finer.4-6 The microstructural
features that make up such a structure are the eutectic grain size and the finer
phase structure contained within.
It is important to note that within such a morphology, many individual phase
regions (e.g., the lamellae) constitute a single eutectic grain. Within these eutectic
grains all phase regions of the same kind exhibit a single and constant crystal-
lography, and there is a unique and preferred crystallographic interface plane
between the two constituent phases.6-1l The lamellar grain just described is, in
fact, two single crystallites of the two constituent phases. 10-- 14 The crystallo-
graphic relationships within and between the two phases are those that minimize
growth and interfacial energies for a given set of solidification conditions. The
stability of the liquid solid interface determines the inter- and intraphase crys-
tallographic and morphological perfection. 7- 10

7.2.2 Degenerate Eutectics

A breakdown of the lamellar structure results from instabilities of the advancing

liquid-solid interface. The resultant morphology may be wavy, rod-like, or glob-
ular (degenerate lamellar) and lacks the long-range perfection of the regular
lamellar structures. 9 Interface instabilities may arise from growth along unfa-
vorably oriented growth directions, local supercooling effects, and rapid cooling
rates.8-10 Solidification under such conditions results in discontinuities and faults
of the individual phase regions and phase interfaces that disrupt the continued
growth of the aligned lamellar structure. Instead, only a short-range phase align-
ment is maintained and a colony substructure develops within the individual
eutectic grain. The colony is a subset of the eutectic grain and is characterized
as a short-range phase alignment over a region of the eutectic grain separated

Figure 7-5 Schematic representations of several features of a eutectic microstructure. (a)

Several lamellar grains showing distinct grain boundaries, (b) several lamellar grains showing
irregular degenerate material at the grain boundaries, (c) a single eutectic grain showing a
colony substructure.
232 THE ROLE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOLDER JOINTS

from other eutectic colonies by a small crystallographic mismatch. Figure 7-5

schematically illustrates several features of a eutectic microstructure including
the colony structure within the eutectic grain. 15 These degenerate structures lack
the structural perfection of the lamellar eutectics, especially at the colony bound-
aries. However, they do maintain a special and unique crystallographic alignment
with the growth direction and, as much as possible, between the two phases.
A dramatically different microstructure is obtained for quenching of the Pb-
Sn eutectic material from the melt. 13 In this case, the solidification occurs so
quickly that the eutectic colony fonnation is suppressed, and instead the micro-
structure appears as a dispersion of the Pb-rich phase in a Sn-rich matrix. The
quenched microstructure, unlike the slower cooled eutectic colony structure, is
unifonn in phase size and shape distribution throughout the entirety of the spec-
imen. And, unlike the eutectic grain structure described above, there is no
apparent crystallographic relationship between the two phases; each individual
phase region appears to be an individual phase grain. Figure 7-6 shows a bulk
Pb-Sn eutectic specimen water quenched from 230°C. In contrast to a more

Figure 7-6 SEM micrograph of a eutectic Pb-Sn alloy water quenched from 250°C. The
microstructure lacks the eutectic colony features characteristic of slower cooled specimens
and appears instead as a homogeneous dispersion of the Pb-rich phase in a Sn-rich matrix.
The light regions are the Pb-rich phase; the dark regions are the Sn-rich phase.
 EUTECTIC MICROSTRUCTURES 233

moderately cooled specimen the quenched microstructure is much finer and lacks
the morphological heterogeneity that characterizes the colony structure.
The description of a eutectic microstructure is not straightforward. Unlike a
single-phase material, a grain size is not readily apparent. Much of the published
work on solder microstructures uses the phase diameter (i.e., the average diameter
of a phase region in a two dimensional section) as the grain size. The phase
diameter may not be a true grain at all. By definition, a grain refers to an element
of a material within which a single crystallography exists. In a eutectic structure,
many individual phase regions may, in fact, constitute a single eutectic grain
(e.g., the two lamellar grains in Figure 7-4). The terminologies that apply to
eutectic microstructures, as established by Weart and Mack, 14 include the eutectic
grain, the eutectic colony, and the individual phase regions. The eutectic grain
refers to that portion of the structure that nucleated at a single site and that shares
a specific and unique crystallography. The eutectic colony may be a subset of
a eutectic grain and refers to a region within the microstructure where the phase
particles have a characteristic arrangement.

7.2.3 Solder Joint Microstructures

In a casting, such as a solder joint, solidification initiates from many different

nucleation events. A favorably oriented seed may not be present and the cooling
rate is such that there is insufficient time to select preferred growth directions.
The final microstructure will thus contain both lamellar and degenerate eutectic
features within a eutectic colony/grain arrangement. The distribution and size
of these will vary with global and local cooling rates, larger lamellar eutectic
colonies/grains dominating at slower cooling rates. Such a microstructure is
characterized by a local homogeneity-within a eutectic colony/grain and a global
inhomogeneity as the phase features differ from eutectic grain to eutectic grain
and, especially, at the eutectic colony/grain boundary. Faster cooling rates yield
progressively less distinct eutectic colony/grain structures and an increasingly
more degenerate and globally uniform structure.
In the solder microstructures that are described in this work no distinction is
made between a eutectic grain and eutectic colony. The term "eutectic colony"
here refers to a region of the microstructure that exhibits a specific phase particle
size and shape arrangement. For convenience, the term "colony" is being used
to distinguish it from the grain size referred to in the superplastic and highly
degenerate structures where an individual phase region constitutes in fact an
individual crystallographic grain. It should be noted that polarized light micro-
scopy of the colony or grain structures studied here reveals constant contrast
within a colony feature and contrast changes between these eutectic structural
features, thus indicating that the colonies are indeed individual and distinct
crystallographic grains.
234 THE ROLE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOLDER JOINTS

Figure 7·7 Optical micrograph of a typical 60% Sn-40% Pb solder joint. The dark regions
are the Pb-rich phase. Note the phase distribution into grainlike features (i.e. , colonies). This
solder microstructure is representative of the solder microstructures in the specimen configu-
ration shown in Figure 7-1. The light regions are the Sn-rich phase; the dark regions are the
Pb-rich phase.

The eutectic Pb-Sn solder joint microstructure shown in Figure 7·7 is repre-
sentative of the experimental joints used in this work. The bulk joint material
consists of lamellar and degenerate lamellar features distinguishable as eutectic
colonies with distinct boundaries separating the individual colonies. Note that
the eutectic colony size is far larger than the individual phase regions and is
almost on the order of the actual joint thickness. The complex and irregular
nature of the solidification is consistent with the many different microstructural
features observed. The microstructure in Figure 7-8 corresponds to a joint from
the specimen configuration in Figure 7-2. The solder volume in this joint is quite
small, and thus the solder experiences a much faster cooling rate. This is evi-
denced by the smaller occurrence of distinct colony features and the greater
overall degeneracy of the eutectic structure. Figure 7-9 shows a cross section
through a "real" solder joint. This particular joint attaches a resistor to a polymeric
circuit board and was made by the wave soldering method. 16 The microstructure
consists of individual and distinct eutectic grains that contain phase morphologies
of both lamellar and degenerate character. Solder joint microstructures vary
considerably with soldering processes and joint geometries. The regular lamellar
 EUTECTIC MICROSTRUCTURES 235

25tJm
Figure 7-8 Optical micrograph of a 60% Sn-40% Pb solder joint. This solder microstructure
is representative of the solder microstructures in the specimen configuration shown in Figure
7-2. The smaller joint solder volume results in a faster solder cooling rate. This is evidenced
by the slightly smaller scale of the colony features and a greater degeneracy of the eutectic
structure. The light regions are the Sn-rich phase; the dark regions are the Pb-rich phase.

colony features dominate with moderately cooled solder joints of reasonable

thickness, and the more degenerate dispersed structures are characteristic of faster
cooling rates and very thin joints. An understanding of the microstructural mech-
anisms of thermal fatigue should make clear the role of the initial microstructure
and its importance to fatigue life.
236 THE ROlE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOlDER JOINTS

Figure 7-9 SEM cross section of a "real" solder joint. This particular joint was made by the
wave soldering method. The microstructure contains both distinct lamellar colony features and
regions of more degenerate character. The light regions are the Pb-rich phase; the dark regions
are the Sn-rich phase.

Intermetallics

Since a Pb-Sn solder joint most frequently occurs on a Cu substrate, another

microstructural component of the Pb-Sn solder joint is the intermetallic layer at
the Cu-solder interface. The intermetallics form when the molten solder comes
into contact with the clean Cu surface. In low Sn solders, the intermetallic is CU3Sn
which grows as densely packed globular particles on the Cu surface. When the
Sn content is higher, such as in the near-eutectic compositions, the intermetallic
phase is a double layer with CU3Sn on the Cu substrate and CU6SnS in contact with
the molten solder. The C~Sns grows as hexagonal rods that protrude into the
molten solder. Turbulence in the melt causes these spikes to break off and they
appear as intermetallic particles in the bulk of the joint material. 17 Figure 7-10 shows
detail of the interfacial and bulk intermetallic phase in a 60% Sn-40% Pb solder
joint. It has been shown that the intermetallics have little influence on the solder
mechanical properties reviewed in this work.9 ,18-20
 EUTECTIC MICROSTRUCTURES 237

10 ~m

2 pm

Figure 7-10 SEM micrograph of a polished 60% Sn-40% Pb surface etched to reveal details
of the microstructure. (a) Interfacial intermetallics are clearly seen at the Cu-solder interface.
The interfacial intermetallic phase is in fact a bilayer consisting of a CU3Sn layer and a layer
of Cu&SnS adjacent to the solder. (b) Detail of a Cu&SnS intermetallic particle found in the bulk
solder joint material. These intermetallics originate at the Cu-solder interface. Turbulence during
solidification causes them to break off and locate within the bulk of the solder material. The
light regions are the Pb-rich phase; the dark regions are the Sn-rich phase.
238 THE ROlE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOlDER JOINTS

7.2.4 Effects of Composition

Other frequently used solder compositions are the off-eutectic, Pb-rich alloys.
The Pb-rich solders exhibit microstructures that are very similar to those of the
near-eutectic alloys and are thus also of interest to this work since a characteristic
failure mechanism operates as long as the microstructure contains a sufficient
volume fraction of the eutectic structure. The Pb-rich microstructures differ only
in the appearance of the primary Pb-rich phase, as seen in Figure 7-11 for several
Pb-rich alloys. The presence of this feature can be understood in terms of the

Figure 7-11 Optical micrographs of several Pb-rich solder alloys. (a) 50% Sn-50% Pb, (b)
40% Sn-60% Pb, (c) 20% Sn-80% Pb. Increasing the amount of Pb in the alloy from the near-
eutectic compositions results in increasing amounts of the primary Pb-rich phase. The remaining
microstructure consists of colony features as seen in the near-eutectic alloys. The light regions
are the Sn-rich phase; the dark regions are the Pb-rich phase.
 EUTECTIC MICROSTRUCTURES 239

Atomic percent tin, a/o

0 20 40 60 80 100

600
300
0 LL
0 232°C 0

- ...::J
- ...::JcP
Q)~

200 400
...
C\'S
Q)
...
C\'S
Q)
c- o.
E
Q)
E
I- 100 200 ~

o
o 20 40 60 80 100

Composition, w/o Sn
Figure 7-12 The Pb-Sn phase diagram. The eutectic composition occurs at 61. 9% Sn-
38.1% Pb.

binary Pb-Sn phase diagram of Figure 7-12. On cooling from the melt, the off-
eutectic compositions pass through the Pb-rich a plus liquid two-phase region.
At the temperatures that fall inside this phase field, the Pb-rich phase begins to
solidify in a dendritic morphology. Upon reaching the eutectic temperature of
183°C, the remaining liquid solidifies into the classic eutectic microstructure.
The final microstructure thus consists of primary Pb-rich dendrites in the eutectic
structure. The volume fraction of the primary Pb-rich dendrites is determined
by the initial composition and cooling rate.
A relaxation of the regularity and perfection of the eutectic lamellar structure
is also possible with a small addition of a third element. 21 During solidification,
the third element in the eutectic Pb-Sn alloy, acting analogously to an impurity
in a single component system, will be rejected from the solidifying phases into
the remaining liquid. (This analogy is valid when the partition coefficient is < 1,
as is expected here based on an analysis of the binary diagrams of the impurity
and constituent phases.)4-6 The resultant compositional changes result in an area
of constitutional supercooling directly ahead of the advancing solid-liquid in-
terface and allow for the perturbation and breakdown of the solidification front
into a colony-(cellular) structure. The coarser and irregular features at the colony
boundaries result from the slower growth rates encountered at the curved colony
interfaces. 6.14 ,15,22,23
240 THE ROLE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOLDER JOINTS

Figure 7-13 Optical micrographs of several ternary alloys. (a) 58% Sn-40% Pb-2% Sb, (b)
58% Sn-40% Pb-2% Bi, (c) 58% Sn-40% Pb-2% In, 58% Sn-40% Pb-2% Cd. All microstructures
are the result of identical processing histories; hence, the microstructural changes are due to
compositional changes only. Note the appearance of diffuse colony boundaries and the as-
sociated increase in the amount of degenerate material with the In and Cd additions. The light
regions are the Sn-rich phase; the dark regions are the Pb-rich phase.

Figure 7-13 shows microstructures of several ternary alloys, 58% Sn-40% Pb-
2%X where X is Sb, Bi, Cd, and In. All samples were prepared using equal
additions of the third element and identical processing procedures; hence the
microstructural changes are due to compositional changes only. The addition of
Sb does not dramatically alter the solder microstructure. However, the addition
of Bi, In, and Cd changes both the colony macrostructure and the eutectic
microstructure within. The boundary area between individual colonies is much
wider and consists of coarse irregular phase particles. Additionally, the phase-
 EUTECTIC MICROSTRUCTURES 241

Figure 7-13 (continued)

to-phase arrangement within the colonies is more faulted and the areal extent of
the lamellar eutectic arrangement is reduced. Of the alloys studied here, the Cd
alloyed solder exhibits the least distinct colony structure (i.e., this solder pos-
sesses the most degenerate, most globally uniform microstructure).

7.2.5 Recrystallized Pb-Sn Microstructure

A very different Pb-Sn microstructure is achieved after a mechanical working

of the as-solidified Pb-Sn material. The plastic deformation introduced with, for
example, rolling, swaging, or extrusion of the as-cast bulk solder introduces a
very high dislocation density into the Pb-Sn microstructure that in tum initiates
a recrystallization of the deformed material. The recrystallization process in-
242 THE ROLE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOLDER JOINTS

Figure 7-14 Optical micrograph showing a worked and annealed 60% 5n-40% Pb micro-
structure. The thermomechanical treatment results in a recrystallized microstructure in which
the colony features characteristic of the as-solidified structure have been completely replaced
by a uniform distribution of equiaxed Pb-rich and Sn-rich grains. The light regions are the Sn-
rich phase; the dark regions are the Pb-rich phase.

volves the nucleation of new strain-free grains in the deformed material that
grow until the latter is consumed. Thermal energy above some threshold is
necessary to initiate this process; however, since solder is already at a high
homologous temperature at 25°C, recrystallization may occur dynamically as the
material is being deformed. The recrystallized microstructure consists of equiaxed,
randomly oriented grains of the two constituent phases uniformly distributed
throughout the material. As shown in Figure 7-14, this type of Pb-Sn micro-
structure bears little resemblance to the eutectic colony structure that is char-
acteristic of the normally cooled, as-solidified material. The ease of recrystal-
lization and the dramatic difference in appearance of the recrystallized material
are well documented for this alloy. 24-27

7.2.6 Coarsening Behavior

The coarsening behavior of as-cast eutectics varies significantly with the various
possible morphologies. The perfect lamellar structures exhibit exceptionally high
 MECHANICAL PROPERTIES 243

morphological stability at elevated temperatures. 7,27-29 This contrasts greatly with

the rapid spheroidization and coarsening of the more faulted degenerate and
dispersed structures and the irregularly shaped material comprising the colony
boundaries. Coarsening always initiates in these less regular regions and very
slowly consumes the more perfect material. The difference is attributed to the
higher interfacial energies of the less favorably aligned structures and the en-
hanced diffusion afforded by the greater fault and grain boundary areas. 30-32

7.3 MECHANICAL PROPERTIES

Just as the processing history determines a material's microstructure, the micro-

structure in tum determines the mechanical properties. Since the Pb-Sn eutectic
structure exhibits dramatically different morphologies with different processing
histories, it is not surprising that these different microstructures exhibit signifi-
cantly different mechanical properties.

7.3.1 Eutectic Structures

The regular lamellar microstructures characteristic of the as-cast Pb-Sn eutectic

material possess the highest yield strengths and creep resistances and the lowest
elongations and strain rate sensitivities within this alloy system. 24 ,25,33-37 The
exceptional strengths are attributed to the large-scale, regular arrangement of
the lamellar structure and the associated difficulty of dislocation motion through
the aligned two-phase structure. 12 Deformation of the multigrained material is
most easily accommodated at the eutectic colony or grain boundaries where
these strengthening mechanisms are less effective. For a multicolony structure,
such as that observed in the joints in Figures 7-7, 7-8, and 7-9, the deformation
character is thus highly inhomogeneous. Deformation is more easily accom-
modated in the softer intercolony areas than in the stronger regularly aligned
phase arrangement inside the colony. This is readily apparent on observing the
patterns of surface relief that develop on a flat polished eutectic colony struc-
ture as it is deformed. Figure 7-15 shows such a pattern. Relatively little de-
formation occurs in the colony interior; rather, the surface relief indicates that
the strain is accommodated by the degenerate material at the colony bounda-
ries. Furthermore, rnicrohardness traces across such a eutectic structure, as shown
in Figure 7-16, illustrate that the irregular phase arrangement at the colony
boundary is softer than the regular phase arrangement within the eutectic col-
onies. Processing changes that reduce the colony size and thus increase the amount
of soft boundary material predictably affect the observed mechanical properties.
Castings of bulk specimens show a dramatic decline in the yield strength34 ,38
and creep resistance and an increase in strain rate sensitivity and elongation35
with faster cooling rates. Faster cooling rates increase the amount of degenerate
244 THE ROLE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOLDER JOINTS

Figure 7-15 Optical micrograph of a 60% Sn-40% Pb solder joint showing the surface relief
that develops as the joint is deformed. The deformation pattern shows that the response of the
solder microstructure is highly inhomogeneous; deformation occurs predominantly at the col-
ony boundaries where the strengthening mechanisms that characterize the interior of a eutectic
colony are lacking. The light regions are the Sn-rich phase; the dark regions are the Pb-rich
phase.

irregular material through reduced colony size and structural perfection and thus
lead to an overall softening of the material.
The shear strengths of the ternary alloys appear to follow a similar trend. The
alloys that exhibit the greatest volume of lamellar aligned eutectic material also
exhibit higher shear strengths. The Cd and In alloyed solders, those microstruc-
tures possessing larger fractions of irregular degenerate phase particles, exhibit
slightly lower yield strengths. Table 7-1 lists the shear strength data for the
various alloys. All data refer to the specimen configuration in Figure 7-1.
A softening also occurs within the simple binary system when the amount of
eutectic structure is decreased by increasing the amount of Pb in the alloy. In
the Pb-rich alloys, the primary Pb-rich phase replaces the classic lamellar eutectic
structure. Since the Pb-rich phase is relatively soft, its presence in significant
amounts leads to a decreased shear strength in the Pb-rich alloys. 39
An even greater difference in mechanical behavior is exhibited by the recrys-
tallized Pb-Sn material. Within this homogeneous, fine-grained microstructure
the strengthening mechanisms associated with the composite nature of the la-
mellar eutectic are lost and the deformation character is greatly altered. The fine
equiaxed grain size and associated large grain boundary area contribute to the
exceptional softness and ductility of the recrystallized microstructure. 24,40--42
 MECHANICAL PROPERTIES 245

Figure 7-16 Optical micrograph of microhardness traces across a multicolony eutectic Pb-
Sn microstructure. The hardness values illustrate the difference in mechanical response between
the regular phase arrangement inside a colony, relatively hard material (17 Vicker's micro-
hardness number), and the degenerate features at the colony boundary, relatively soft material
(11 Vicker'S microhardness number). The light regions are the Sn-rich phase; the dark regions
are the Pb-rich phase.

Table 7-1
Stress Exponent* Shear Strength (MPa)
60% Sn-40% Pb 6.0 22
61 .9% Sn-40% Pb 6.3 23
58% Sn-40% Pb-2% Bi 5.5 25
58% Sn-40% Pb-2% Sb 5.6 26
58% Sn-40% Pb-2% In 3.7 19
58% Sn-40% Pb-2% Cd 3.3 18
·Stress exponent refers to Equation (7-1).
Note: Right column : Shear strengths of several Pb-Sn alloys. These data were obtained on the specimen
configuation shown in Figure 7-1. left column: Stress exponents, as determined from creep experiments
according to Equation (7-1) of several ternary alloys. The simple binary Sb-alloyed and Bi-alloyed solders
exhibit stress exponents characteristic of conventional plasticity. The In- and Cd-alloyed solders exhibit
slightly lower stress exponents. This division is also consistent with the extent of lamellar eutectic nature of
the solder microstructure. The higher stress exponents are associated with the greatest amount of lamellar
eutectic material.
246 THE ROLE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOLDER JOINTS

In fact, under the appropriate deformation conditions, this type of microstructure

is superplastic, exhibiting elongations on the order of hundreds of percent. 24,43,44
The strength of the eutectic Pb-Sn material arises from the regular lamellar
eutectic arrangement of the individual phase regions. A reduction in the areal
extent and perfection of this classic eutectic structure also reduces the material's
strength. Faster cooling rates and appropriate compositional changes increase
the amount of degenerate material and thus soften the material. The microstruc-
ture that exhibits no trace of the colony structure (i.e., the recrystallized micro-
structure) is by far the softest. In the microstructures that exhibit a colony
structure, the colony size is an important microstructural parameter since this
determines the ratio of hard to soft material. This microstructural heterogeneity
also accounts for the highly inhomogeneous deformation that is observed in the
colony microstructures. In the degenerate and recrystallized material, defor-
mation is more uniformly distributed since the area over which the deformation
can be easily accommodated, the grain boundary regions, is larger.

7.3.2 Deformation Mechanisms

Solder service conditions entail operation at temperatures that represent high

solder homologous temperatures; creep deformation is thus an important defor-
mation mode, and the study of these low strain-rate high temperature deformation
mechanisms is thus warranted.
To investigate the predominant deformation mechanisms the steady-state creep
data of an alloy system are represented by the equation

(7-1)

where dy/dt is the steady-state shear strain rate and 'T is the shear stress. 27.43-51
In this representation, it is believed that, within a wide strain rate-stress regime,
there exist regions of constant nand Ea values within each of which a different
deformation mechanism dominates. For example, superplastic deformation of
fine equiaxed microstructures is characterized by a stress exponent in the range
of 2 and an apparent activation energy equal to the activation energy of grain
boundary diffusion.
Superplastic deformation of the eutectic Pb-Sn system has been extensively
studied and the values of these constants are well known. The microstructures that
exhibit superplasticity are not those of the as-cast material but rather those micro-
structures that result after a mechanical working of the as-cast alloy (i.e., a recrys-
tallized microstructure that is uniformly fine grained and equiaxed).26,27.44,51-60
Superplasticity of the recrystallized Pb-Sn microstructure has been observed for strain
rates between 10-7 to 10-2 sec-I with a stress exponent _2 26 ,27,44,51-59 and activa-
 MECHANICAL PROPERTIES 247

tion energy on the order of 12 kcal/mol (a favorable agreement with the activation
energy for grain boundary diffusion). 27.52,53,55,57
Other deformation mechanisms, such as those operative in nonsuperplastic
microstructures or in different strain rate regions, are characterized by different
values ofthese constants. For example, conventional dislocation climb controlled
plasticity is associated with much higher stress exponents, 4 to 10, and activation
energies on the order of those for self-diffusion. 38 Within the Pb-Sn system,
conventional plasticity (e.g., deformation of nonsuperplastic microstructures),
is associated with a stress exponent of ~6 to 7 and an activation energy of ~20
kcallmol. 27,44,49,61,62
The fine equiaxed microstructures necessary for superplastic deformation are
not characteristic of as-solidified joint solder microstructures, such as those
shown in Figure 7-7. Hence, these joints are not expected to deform superplast-
ically. Results of creep experiments on these solder joints are consistent with
this expectation and reveal deformation constants representative of conventional
plasticity. Figure 7-17 shows a plot of strain rate versus stress for these solder
joints. The slope of this curve, according to the representation of Equation 7-1,

10-3~----------------------~--------------~

10. 4

..-
.,u
!!.
10. 5
.,
~

..
c.
e 10. 6

10·8~----~~~~~~~~------~~--~~~~~
1 10 100
stress (MPa)

Figure 7-17 A plot of steady-state strain rate versus stress, at constant temperature, for the
eutectic Pb-Sn alloy. The slope of the line is equal to 6.0, indicating conventional plasticity.
248 THE ROlE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOlDER JOINTS

is the stress exponent; in this case, the stress exponent has a value of 6, indicating
conventional plasticity.
Creep experiments on the more rapidly cooled solder joints, where the smaller
colony size and decreased amount of lamellar eutectic material exhibit slightly
different, lower, stress exponents, suggest different deformation mechanisms.
The smaller joint volumes of the joints in the specimen configuration of Figure
7-8 result in faster cooling rates and thus a more degenerate, less "eutectic"
structure. Preliminary measurement of creep deformation of this solder micro-
structure reveals a lower stress exponent, approaching 2-the value normally
associated with superplastic deformation. 63 Superplasticity in an unworked un-
recrystallized rapidly solidified eutectic solder has not been unambiguously dem-
onstrated; the possibility has recently been suggested by Shine and Fox. 64 How-
ever, a different microstructure may indeed imply a different deformation
mechanism. Furthermore, it seems reasonable that as a microstructure acquires
more and more degenerate eutectic material (i.e., the more closely it resembles
a superplastic microstructure), the deformation character will too acquire a su-
perplastic component and thus exhibit a lower stress exponent. This area is
currently under investigation.
A survey of the work on the creep deformation of near eutectic Pn-Sn solder
in a joint configuration reveals stress exponents in the range - 2 to 6.64-67 These
data are not always accompanied by microstructural information; however, the
different joint geometries and joint manufacturing techniques used in these ex-
periments most probably result in a variety of joint solder microstructures. As
with the data presented earlier, the scatter may be due to the varying degree of
classic lamellar eutectic structure that occurs with the different joint manufac-
turing techniques.
A similar observation is made for the ternary alloys. The stress exponents
measured for solder joints of the configuration shown in Figure 7-1 appear in
Table 7-1. It is possible to separate these solders into two groups. The Cd and
In alloys exhibit a stress exponent between 3 and 4, and the Sb, Bi, and simple
binary alloys exhibit stress exponents of - 6. This classification is also consistent
with the observed microstructures. The Cd and In alloys, though solidified under
similar conditions, possess the more degenerate microstructures. The remaining
three possess the most lamellar structures. As with the binary alloys, it is possible
that the reduced stress exponent of the Cd and In is the result of the more
degenerate nature of the joint solder microstructure and an associated superplastic
component in the deformation.

7.4 MICROSTRUCTURAL EVOLUTION UNDER THERMAL FATIGUE

Thermal fatigue of solder joints is driven by the difference in the thermal ex-
pansion coefficients of the surfaces that bind the solder. Figure 7-18 schematically
 MICROSTRUCTURAL EVOLUTION UNDER THERMAL FATIGUE 249

Chip Carrier
Solder Joint

Upolyimide= 15 x 10- 6 fOC

Electronic Package at Zero Strain

,........................................................................
...... " ... , " " " " " " " " " " ' "
"~,""~"~"~"~"~"~"~"~"~"~"~"~,

'""" """
"" "
"," "'"
" """ "" ..." """ " "" "" ",
.." "
, ..... " "
""""""""";;""",,, ""
... , "" " ","" " "" ' "" " "",
... ""'"

Electronic Package with Negative Electronic Package with Positive

Temperature Excursion Temperature Excursion

Figure 7-18 Schematic illustration of the shearthat develops in a solder joint between chip
and substrate with high and low temperature excursions.

illustrates a typical joint geometry and the origin of the thermal fatigue problem.
As shown in the illustration, for a surface mounted leadless component, the
dominant load on the joint is in shear; even for more complex joint geometries,
initial failure of solder joints in service often occurs in that portion of the joint
that experiences shear loading. For these reasons thermal fatigue under shear
loading and associated microstructural evolution is of particular interest.

7.4.1 Thermal Fatigue in Shear

Figure 7-19 illustrates the microstructural changes that occur in a 60% Sn-4O%
Pb solder joint as it is cycled from - 55 to 125°C. The specimen configuration
is shown in Figure 7-3. The uppermost image is the initial microstructure; sub-
sequent images represent the same spot in the specimen with an increasing number
of thermal cycles. The initial microstructure is the classic eutectic structure
described earlier consisting of eutectic colonies made up of alternating phase
regions of the Sn- and Pb-rich phases. As the cycling proceeds, the shear de-
formation concentrates into bands that are clearly marked by an associated rapid
 Figure 7-19 A series of optical micrographs showing the microstructural evolution as a
joint is cycled between - 55 and 125°C. Relatively early on in the cycling, a band of coarsened
microstructure appears near the substrate-solder interface. This band becomes more pro-
nounced with increased thermal cycles, and eventually cracks initiate within this coarse mi-
crostructure causing the joint's failure. The light regions are the Sn-rich phase; the dark regions
are the Pb-rich phase. The top micrograph was taken prior to thermal cycling while the bottom
one after 1700 cycles.

250
 MICROSTRUCTURAL EVOLUTION UNDER THERMAL FATIGUE 251

microstructural coarsening. The band of coarsened material traverses the initial

microstructure parallel to the eu-solder interface, along the direction of maximum
shear. After 1700 cycles, cracking of the solder joint is clearly visible within
this coarsened band. Higher magnification SEM images of the coarsened band,
Figure 7-20, clearly show its definition along with the equiaxed and coarsened
individual Sn- and Pb-rich phase grains. As shown in Figure 7-19, failure occurs
exclusively through this coarsened material. An examination of published mi-
crostructural studies of thermally fatigued and failed solder contacts shows that
this fatigue behavior has been widely observed. \9.20.68-72

Figure 7-20 High magnification SEM micrograph showing the definition and detail of the
coarsened band during thermal fatigue. Within the coarsened band, individual Pb-rich and
Sn-rich phase grains are readily discernible. The light regions are the Pb-rich phase; the dark
regions are the Sn-rich phase. The uppermost micrograph corresponds to the initial microstruc-
ture after 1700 cycles.
252 THE ROLE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOLDER JOINTS

7.4.2 Microstructural Mechanisms of Thermal Fatigue

The fatigue failure of the solder joint is directly related to the formation of a
planar band of microstructurally coarsened material. The formation and growth
of the coarsened band are intimately related to the inhomogeneous deformation
character of the as-solidified solder. This fatigue failure behavior is also char-
acteristic of solder joint failures under thermal fatigue,19,20,66-72 isothermal fa-
tigue,6S,69,73 or high temperature creep,20,74 where the solder joint has reasonable

Figure 7-21 Optical micrographs of a polished solder joint deformed in shear. (a) The
surface relief that develops with deformation . The deformation is localized to a thin band near
the Cu-solder interface and along colony boundaries elsewhere in the joint. (b) The same
specimen as in (a) but polished to remove the surface relief. There is an exact correspondence
between the location of the major deformation band and a band of recrystallized and coarsened
microstructure. The deformation elsewhere in the joint corresponds to the irregular material
at the colony boundaries. The light regions are the Sn-rich phase; the dark regions are the Pb-
rich phase.
 MICROSTRUCTURAL EVOLUTION UNDER THERMAL FATIGUE 253

thickness and the lamellar eutectic structure predominates. A closer study of the
sequence of events that initiate and govern this failure mechanism is readily
possible with appropriate creep experiments. The sequence of metallurgical changes
that lead to fatigue failures can be described as follows.
The microstructural event that initiates the pattern of failure is the inhomo-
geneity of the plastic deformation of the as-solidified eutectic microstructure as
it is deformed in shear. Figure 7-21 shows the deformation pattern that forms
when the solder joint diagrammed in Figure 7-1 is deformed slowly in shear.
The deformation pattern is easily observed by polishing the surface of the solder
joint prior to testing and observing the surface relief that appears as the joint
deforms. The most common pattern is that shown in Figure 7-21 and consists
of a nearly straight band of deformation that runs slightly displaced and parallel
to the Cu-solder interface along the direction of maximum shear and more
irregular deformation features that lie approximately perpendicular to this direc-
tion. A comparison of the deformation pattern and the corresponding polished
solder joint surface reveals the direct correspondence between the deformation
and the location of microstructurally coarsened material. The definition of the
coarsened band and the microstructure within are seen in greater detail in Figure
7-22. The material within the coarsened band consists of equiaxed, coarsened

Figure 7-22 Detail of the coarsened band seen in Figure 6-21 (optical) showing its definition
and the presence of equiaxed Pb- and Sn-rich phase grains apparently formed by recrystalli-
zation . The light regions are the Sn-rich phase; the dark regions are the Pb-rich phase.
254 THE ROLE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOLDER)OINTS

grains of the Sn- and Pb-rich phases with no trace of the initial as-solidified
colony structure visible. Clo§er inspection shows that the parallel band in Figure
7-21 traverses eutectic colonies, while the perpendicular coarsening occurs at
colony boundaries. In all cases, the parallel defonnation originates at a lateral
free surface of the joint, a point that corresponds to the principal stress concen-
tration in this sample geometry (see Fig. 7-23).75
The initiation of the coarsened band seems straightforward. The joint solder
microstructure defonns inhomogeneously; defonnation occurs primarily along
colony boundaries where the softest material is found. The inhomogeneity of
the shear strain is further enhanced in the vicinity of a stress concentration. It
is well known that the Pb-Sn is unstable with respect to recrystallization if it is
sufficiently worked at a sufficiently high homologous temperature or exposed to
high temperature after defonnation. The inhomogeneous shear strain that de-
velops during defonnation creates regions of concentrated plastic defonnation.
This concentration is greatest near the principal stress concentration within the
joint configuration, and the eutectic recrystallizes there to relieve the accumulated
deformation. (Figure 7-23 shows the location of the stress concentration in this
joint geometry, the four lateral free surfaces.)
The mechanism of growth of this material into the recrystallized bands ob-
served also seems straightforward. Once a recrystallized region has fonned, it
grows through the progressive recrystallization of the defonned material. Since

Figure 7-23 Schematic of the double shear specimen showing the principal stress concen-
tration of this joint geometry.
 MICROSTRUCTURAL EVOLUTION UNDER THERMAL FATIGUE 255

Figure 7-24 The plastic deformation field ahead of a mode II crack in a Von Mises material
(after McClintock et aI. 76 ).

the recrystallized material is soft compared to the material surrounding it, it

behaves mechanically like a narrow crack. The applied stress is a simple shear,
and the associated strain field is characteristic of a mode II crack. The plastic
strain field for a mode II crack has been solved for a Von Mises material by
McClintock, et al.,16 and is diagrammed in Figure 7-24. Note that the deformation
field is narrow and tends to confine the deformation to the plane of the crack.
It follows that recrystallization and coarsening will always occur at the periphery
of the recrystallized region in the direction of the applied shear since this is the
most highly deformed material. Thus the recrystallized band tends to remain
planar and narrow while it extends itself along a shear plane (parallel to the Cu-
solder interface), as is observed.
Inhomogeneous shear deformation, resulting from the inhomogeneity of the
as-solidified microstructure, triggers inhomogeneous coarsening of the eutectic
microstructure that is concentrated in bands of maximum shear. Since the mi-
crostructure softens as it coarsens, shear deformation concentrates in the coar-
sened regions, which therefore coarsen more rapidly. Eventually, the accumu-
lated strain leads to cracks within the coarsened material, and the joint ultimately
fails.

7.4.3 Other Microstructures

The proposed mechanism and its dependence on the mechanical inhomogeneity

and instability of the solder microstructure suggest that it may be possible to
develop solders with improved fatigue properties. The guidelines for solder alloy
design are relatively clear. Since the coarsened band that precipitates failure is
the result of an initiation and growth process, steps should be taken to inhibit
its initiation and/or inhibit its propagation. Both approaches have been explored
and the results are consistent with our current understanding of the fatigue failure
mechanism.
256 THE ROLE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOLDER JOINTS

Ternary Alloys

High temperature creep deformation for certain ternary alloys is associated

with the development of relatively uniform deformation patterns. Surface relief
observations, such as those in Figure 7-25, show that the deformation in the
degenerate microstructures of the In- and Cd-containing alloys is accommodated
over a much greater area than in the case of the simple normally cooled eutectic
structures. The diffuse deformation behavior of the degenerate microstructures

Figure 7-25 The surface relief that is observed after creep deformation of the (a) In- and
(b) Cd-alloyed solders. The deformation patterns reveal significantly more uniform deformation
throughout the entire joint than was observed for the simple binary alloy. (Optical micrographs:
the light regions are the Sn-rich phase; the dark regions are the Pb-rich phase.)
 MICROSTRUCTURAL EVOLUTION UNDER THERMAL FATIGUE 257

can be rationalized through the greater volume of soft material. As with the
classic eutectic structures, deformation will always concentrate in the softest
material (i.e., the degenerate material at the colony boundaries). However, since
the areal extent of the degenerate material is greater in the In and Cd containing
alloys, the distribution of the deformation is also extended. Furthermore, the
coarsened band formation associated with creep in the lamellar eutectic structures
is not present in these degenerate, uniformly deforming morphologies. Figure
7-26 shows a Cd-containing specimen crept under the same conditions as the
simple binary alloy in Figures 7-21 and 7-22. As seen in Figure 7-26, the creep
deformation in the Cd alloy is not associated with coarsened band formation .
Since the coarsened band formation is intimately related to the thermal fatigue
failure of solder joints, these results indicate improved fatigue resistance.
In fact, in terms of inhibiting the coarsened band formation, the addition of
the third element to the binary alloy is potentially beneficial in two ways. First,
due to the creation of a more mechanically uniform microstructure, local plastic
strain levels are lower. The microstructural instability that leads to the coarsened
band formation may thus be suppressed since local strain concentrations may

..... 150pm~
Figure 7-26 The polished microstructure of the deformed Cd-alloyed specimen seen in
Figure 7-25. The deformation is not associated with microstructural changes, as was the case
with the binary alloys. (SEM micrographs: the light regions are the Pb-rich phase; the dark
regions are the Sn-rich phase.)
258 THE ROLE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOLDER JOINTS

"'200 ... m~
Figure 7-27 Optical micrographs of several thermally fatigued solder joints, 3000 cycles
between - 55 and 125°C. (a) A 60% Sn-40% Pb joint showing the definition and equiaxed
grain structure of the coarsened band material. Joint cracking has occurred exclusively through
this coarser microstructure. (b) 58% Sn-40% Pb-2% In and (c) 58% Sn-40% Pb-2% Cd solder
joints after the same thermal treatment. Both these solder joints are still intact. These alloyed
solder joint microstructures have coarsened with the thermomechanical exposure; however,
the coarsening is global and not confined to a band near the solder substrate interface. (Optical
micrographs : the light regions are the Sn-rich phase; the dark regions are the Pb-rich phase.)
 MICROSTRUCTURAL EVOLUTION UNDER THERMAL FATIGUE 259

not exceed the critical level needed to initiate the recrystallization. Furthennore,
the addition of certain elements, including In and Cd, are known to inhibit
recrystallization and creep defonnation in the Pb and Sn systems. 77- 80 Thus
independent contribution of these effects to the inhibition of coarsened band
fonnation in the crept ternary alloys is not yet clear.
Thennal fatigue of these ternary alloys reveals similar defonnation patterns.
Surface relief examinations of thennally fatigued ternary solder joints reproduce
the relatively unifonn defonnation patterns seen under creep loading conditions.
A thennal fatigue life comparison between the simple binary eutectic structures
and the degenerate ternary alloys reveals the enhanced fatigue resistance of the
latter alloys. Figure 7-27 shows a comparison of the three different compositions
after 3000 thennal cycles. The 60% Sn-40% Pb joint is clearly cracked and both
the In and Cd alloys appear intact. The cracking in the 60% Sn-40% Pb solder
is, as expected, confined to the coarsened band, and the coarsened band is clearly
visible and clearly microstructurally distinct from the remaining joint material.
However, unlike the creep defonnation, thennal fatigue defonnation does result
in a coarsening of the solder microstructures. But, in contrast to the 60% Sn-
40% Pb alloy, the coarsening is global and not confined to a band near a joint
interface. The addition of the third element has diffused the initial microstructures
and thus the defonnation patterns. The increased fatigue resistance of the ternary
alloys appears to be the result of a postponement of the fatigue recrystallization
and softening failure mechanism rather than a complete inhibition. The more
unifonn microstructures result in less localization of strain and thus defer the
initiation and propagation of the recrystallized band.

Pb-rich Solders

The isothennal fatigue behavior of several compositions within the Pb-Sn

system has recently been studied. 73 In these high Pb alloys, islands of the
relatively soft Pb-rich phase displace the harder eutectic material and cause the
surrounding eutectic to become more degenerate. It was found, however, that
as long as the joint microstructure contained a significant fraction of the classic
lamellar eutectic structure, failure occurred by the same mechanism as seen
previously in fatigue and creep of near-eutectic alloys-inhomogeneous defor-
mation resulting in the fonnation of a recrystallized and coarsened band of
material. Figure 7-28 shows joints fatigued with a 10 percent strain amplitude
at 75°C for several Pb-rich alloys. For Sn concentrations between about 40 and
63 wt %, fonnation of the coarsened band was not prevented or postponed with
respect to its fonnation in the near-eutectic alloys. Its propagation, however,
was inhibited by the presence of the Pb-rich proeutectic phase, with the 50%
Sn-50% Pb alloy showing the most resistance to fatigue. Since the 40% Sn-60%
Pb alloy has the most proeutectic phase and the least Sn of the alloys exhibiting
260 THE ROLE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOLDER JOINTS

200 um

Figure 7-28 Optical micrographs of several Pb-rich alloys isothermally fatigued with a 10
percent strain amplitude at 75°C. The (a) 50% Sn-50% Pb and (b) 40% Sn-60% Pb joints both
exhibit the formation of a coarsened band that eventually becomes the site of joint failure . The
increased fatigue life of the Pb-rich alloys is due to the more difficult propagation of the
coarsened band through a microstructure that contains many primary Pb-rich islands. (Optical
micrographs: the light regions are the Sn-rich phase; the dark regions are the Pb-rich phase.)
 CONCLUSION 261

coarsening, it is believed that the Pb-rich phase acts as a barrier to the propagation
of the coarsened band, but its effectiveness is determined by the amount of Sn
present. When a significant amount of the joint material consists of the primary
Pb-rich phase, its fatigue resistance relative to the eutectic-containing alloys
depends greatly on test conditions and it fails by a different mechanism.

Rapidly Solidified Joints

More rapidly solidified solder joints, such as those shown in Figure 7-8, possess
a smaller colony size and a larger fraction of degenerate eutectic material. These
microstructures also fail through the formation of coarsened bands that concen-
trate the strain and thus lead to failure. However, recent creep experiments
indicate that the failure in these more degenerate structures is delayed. The
presence of a larger amount of degenerate microstructure, as in the ternary
solders, is expected to diffuse the deformation and thus postpone the micro-
structural instability that leads to failure. The possibility that this postponement
is actually due to a superplastic deformation component is currently under in-
vestigation. The idea was suggested by Shine and Fox,64 who used it to formulate
a quantitative model of creep fatigue deformation. Since superplastic deformation
occurs without accumulating damage, it should postpone the plastic strain in-
tensities that initiate failure.

7.5 CONCLUSION

The problem of solder fatigue is a complex one; however, an understanding of

the metallurgical mechanisms that lead to joint failure clarifies the role of the
solder microstructure and thus indicates paths toward improved solder alloy
development. The complexity of the problem is, at least partly, attributable to
the complex nature of the two phase eutectic microstructure. An understanding
of the microstructure and its relationship to process parameters and influence on
properties is an important first step. The work outlined here reviews the eutectic
Pb-Sn microstructure, its microstructure-property relationship, and, finally, its
impact on the fatigue behavior.
As in any other material, the mechanical properties depend on the alloy's
microstructure. A Pb-Sn alloy of a given composition may exist with many
different microstructures depending on the processing history. Each microstruc-
ture has its own characteristic mechanical properties and the differences in me-
chanical behaviors may be significant. The exceptional strength and stability of
the lamellar eutectic microstructures arises from the long-range regularity and
alignment of the two-phase crystal structure. These strengthening mechanisms
are lost in more degenerate, less regular structures, and such microstructures
deform more easily.
262 THE ROLE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOLDER JOINTS

The solder joint microstructure contains both lamellar eutectic and degenerate
features and thus a tendency toward inhomogeneous deformation. In the eutectic
colony structures routinely observed in solder joints, deformation is most readily
accommodated at the colony boundaries where the long-range lamellar phase
arrangement terminates and a more faulted phase structure exists. The inhom-
ogeneous deformation character results in strain concentrations at the colony
boundaries as the joint is deformed. Since the Pb-Sn eutectic is unstable with
respect to recrystallization and coarsening under mechanical working, it will
recrystallize in these highly strained regions. The recrystallized material is very
soft relative to the remaining joint material and will further concentrate defor-
mation and eventually will lead to joint failure.
The metallurgical mechanism of fatigue involves two processes: shear defor-
mation develops in an inhomogeneous pattern that triggers recrystallization,
which forms soft, coarsened bands. It follows that a microstructure that deforms
more uniformly or one that is inherently more resistant to the initiation and
propagation of the recrystallized bands should show enhanced fatigue resistance.
It is possible to alter the amount of lamellar eutectic material within a solder
joint with appropriate compositional and processing changes and thus make the
joint more uniform in its mechanical response. Chemical additives that are known
to enhance the microstructural stability may also be used. In this work, several
such metallurgical techniques have been proposed and evaluated. The replace-
ment of the lamellar eutectic microstructure with a more equiaxed degenerate
one results in a microstructure that naturally deforms more uniformly and possibly
superplastically. Such microstructures have shown enhanced creep and thermal
fatigue resistances. In Pb-rich solders, the presence of the islands of proeutectic
Pb-rich phase results in a difficulty in propagating these coarsened bands and
thus an improved fatigue life.

7.6 ACKNOWLEDGMENTS

This work is jointly supported by the director, office of Basic Energy Science,
Materials Science Division of the U.S. Department of Energy under contract
No. DE-AC03-76SFOOO98.

REFERENCES

1. Grivas, D., D. Frear, L. K. Quan, and J. W. Morris, Jr., J. Electr. Mat., 15, 1986,
p.355.
2. Frear, D., D. Grivas, L.K. Quan, and J. W. Morris, Jr., MRS Symposium, 72,1986,
p. 181.
3. Frear, D., D. Grivas, and J.W. Morris, Jr., J. Electr. Mat., 17, 1988, p. 171.
4. McLean, M., Directionally Solidified Eutectics. The Metals Society, London, 1983.
 REFERENCES 263

5. Shewmon, P. G., Transformations in Metals, J. Williams Book Co., OK, 1983.

6. Elliott, R., Eutectic Solidification Processing, Butterworths, London, 1983.
7. Verhoeven, J. D., D. P. Mourer, and E. D. Gibson, Met Trans. A, 8A, 1977, pp.
1239-1247.
8. Hunt, J. D., J. Inst. Met. 94, 1964, pp. 125-129.
9. Hopkins, R.H., and R. W. Kraft, Trans. AIME, 242, 1968, pp. 1627-1633.
10. Labulle, B., and C. Petipas, J. Cryst. Growth, 28, 1975, pp. 279-287.
11. Blanchin, M. G., A. Guinier, C. Petipas, and G. Sauvage, Acta Met., 20, 1972,
pp. 1251-1258.
12. Cagnon, M., M. Suery, A. Eberhardt, and B. Baudelet, Acta Met. 25, 1977, pp.
71-75.
13. Chilton, J. P., and W. C. Winegard, J.lnst. Met., 89, 1960, pp. 162-164.
14. Weart, H. W., and D. J. Mack, Trans. Met. Soc. AIME, 10, 1958, pp. 664-670.
15. Quan, L. K., "Tensile and Shear Behavior of Alloyed Sn-Pb Solder Joints," M.S.
Thesis, University of California, Berkeley, 1988.
16. Specimen courtesy of J. Maki, EMPF, China Lake, CA.
17. Frear, D., D. Grivas, 1. W. Morris, Jr., J. Elect. Mat., 16, 1987, p. 181.
18. - - J . Elect. Mat., 17, 1988, p. 171.
19. - - J . Met., 40, 1988, p. 18.
20. Tribula, D., D. Grivas, D. Frear, and J. W. Morris, Jr., Weld. J. Res. Supp., 68,
1989, p. 404s.
21. Eady, J. A., and W. C. Winegard, Can. Met. Quart., 10(3), 1971, pp. 213-214.
22. Kraft, R. W., D. L. Albright, J. A. Ford, Trans. Met. Soc. AIME, 227, 1962, pp.
540-552.
23. Kraft, R. W., J. Met., 2, 1966, pp. 192-200.
24. Cline, H. E., and T. H. Alden, Trans. AIME, 239, 1967, pp. 710-714.
25. Avery, D. H., and W. A. Backhofen, Trans. ASM, 58, 1965, pp. 551-562.
26. Morrison, W. B., Trans. AIME, 242, 1968,2221-2227.
27. Kashyap, B. P., and G. S. Murty, Mat. Sci. and Eng. 50, 1981, pp. 205-213.
28. Frebel, M., and B. Otte, Scripta Met., 9, 1975, p. 1317.
29. Graham, L. D., and R. W. Kraft, Trans. AIME, 236, 1966, p. 94.
30. Frebel, M., G. Duddek, W. Graf, M. N. Faridani, and B. Otte, B. Bunsenges Phys.
Chern., 82, 1978, pp. 259-265.
31. Livingston, J. D., and J. W. Cahn, Acta Met., 22, 1974, p. 495.
32. Frebel, M., and G. Duddek, Mat. Sci. Eng., 32, 1974, p. 125.
33. Kashyap, B. P., and G. S. Murty, Met. Trans. A., 13A, 1982, pp. 53-58.
34. Fine, M., and R. L. Dowdell, Trans. ASM, 37, 1946, pp. 245-277.
35. Baker, W. A., J.lnst. Met .. 65,91939, pp. 277-300.
36. Pearson, C. E., J. Inst. Met., 54, 1934, pp. 110-124.
37. Jenkins, C. H. M., J. Inst. Met., 40, 1928, pp. 21-39.
38. Rack, H. J., and 1. K. Maurin, J. Test. and Eval., 2(5), 1974, pp. 351-353.
39. Inoue, H., Y. Kurihara, and H. Hachino, IEEE Trans. CHMT-J2, 1986, p. 190.
40. Hargreaves, F., J. Inst. Met., 40, 1928, p. 41.
41. - - , J. Inst. Met., 39, 1928, p. 301.
42. - - and J. Hills, J. Inst. Met., 41, 1929, p. 257.
43. Bird, J. E., A. K. Mukherjee, and 1. F. Dorn, Quantitative Relations Between
Properties and Microstructure, Israel University Press, 1969.
264 THE ROLE OF MICROSTRUCTURE IN THERMAL FATIGUE OF Pb-Sn SOLDER JOINTS

44. Zehr, S. W., and W. A. Backhofen, Trans. ASM, 61, 1968, pp. 300--313.
45. Mukherjee, A.K., J. E. Bird, and J. F. Dorn, Trans. ASM 62, 1969, p. 155.
46. Weertman, J., and J. R. Weertman Physical Metallurgy, R. W. Cabn, P. Haasen,
Eds., Elsevier Science Publishers, 1983, p. 1310.
47. Vaidya, M. L., K. L. Murty, and J. E. Dorn, Acta. Met., 21, 1973, p. 1616.
48. Mohamed, F. A., K. L. Murty, and J. W. Morris, Jr., Met. Trans., 4, 1973, p.
935.
49. Grivas, D., "Steady State Creep in Pb-Sn Eutectic Alloys," M.S. thesis, University
of California, Berkeley, CA, 1974.
50. Murty, G. S., J. Mat. Sci., 8, 1973, p. 611.
51. Grivas, D., "Deformation of Superplastic Alloys of Relatively Low Strain Rates,"
Ph.D. Thesis, University of California, Berkeley, CA, 1979.
52. Schneibel, J. H., and P. M. Hazzledine, J. Mat. Sci., 18, 1983, pp. 562-570.
53. Geckinli, A. E., and C. R. Barrett, J. Mat. Sci., 11, 1976, pp. 510-521.
54. Murty, G. S., and B. P. Kashyap, Trans. Ind. Inst. Met., 37(5), 1984, pp. 509-
513.
55. Mohamed, F. A., and T. G. Langdon, Phil. Mag., 1975, pp. 697-709.
56. Yadava, R. K., and K. A. Padmanabhan, J. Mat. Sci., 17, 1982, pp. 2435-2440.
57. Baudelet, B., and M. Suery, J. Mat. Sci., 7, 1972, pp. 512-516.
58. Kashyap, B. P., and G. S. Murty, Met. Trans. A., 12A, 1981, pp. 1923-1925.
59. Yadava, R. K., and K. A. Padmanabhan, Mat. Sci. and Eng., 37, 1979, 127-136.
60. B. P. Kashyap, A. Arieli, and A. K. Mukherjee, J. Mat. Sci., 20, 1985, pp. 2662-
2686.
61. Rohde, R. W., and J. C. Swearengen, Trans ASME, J. Eng. Mat. and Tech., 102,
1980, p. 207.
62. Weinbell, R. C., J. K. Tien, R. A. Pollack, and S. K. Kang, J. Mat. Sci., 1987,
p.3901.
63. Mei, Z., M. C. Shine, and J. W. Morris, Jr., submitted to J. Electr. Mat.
64. Shine, M. C., and L. R. Fox, Low Cycle Fatigue, ASTM STP 942, H.D. Solomon,
Ed., American Society for Testing and Materials, Philadelphia, 1988, p. 588.
65. Solomon, H. D., Braz. and Sold., 11, 1986, p. 68.
66. Hall, P. M., IEEE Trans., CHMT-12, 1987, p. 556.
67. Enke, N. F., and B. 1. Sandor, AS ME publication 88-WNEEP-12, 1988.
68. Wild, R. N. "Some Properties of Solder and Solder Joints," IBM Report no.
74z000481, Oct. 1975.
69. Bangs, E. R., and R. E. Beal, Weld. Res. Supp. Weld. J., 54, 1978, p. 377.
70. Wolverton, W. M., Braz. and Sold., 13, 1987, p. 33.
71. - - , "Know Your Enemy: How to Stop Solder Joints from Cracking," Circ.
ManuJ., Nov. 1972, p. 10.
72. Frear, D., D. Grivas, M. McCormack, D. Tribula, and J. W. Morris, Jr., Effect of
Load and Thermal Histories of Mechanical Behavior TMS, 1987, p. 113.
73. Summers, T. S. E., and J. W. Morris, Jr., Trans. of the ASME, Journal of Electronic
Packaging, 112 (June 1990) pp. 94-99.
74. Tribula, D., D. Grivas, D. Frear, and J. W. Morris, Jr., J. Elect. Pack., 1989, p.
83.
 REFERENCES 265

75. Chen, F. K. "Applications of the Finite Element Method to the Plastic Deformation
Analysis of Perforated Sheet Metals and the Process Simulation of Shape Rolling,"
Ph.D. Thesis, University of California, Berkeley, 1989.
76. McKlintock, F. A., and G. R. Irwin, "Plasticity Aspects of Fracture," ASTM STP
381 (1965) p. 84 as referenced in Elementary Engineering Mechanics, D. Broek,
Ed., Martinus Nijhoff Pub., 1983, p. 99.
77. Hanson, D., and E. J. Sandford, Trans AIME, 54, 1935, p. 159.
78. Sully, A. H., Metallic Creep and Creep Resistant Alloys, Butterworths, London,
1949.
79. Hopkins, L. M. T., and C. W. Thwaites, J.lnst. Met., 54, 1953, p. 181.
80. Jenkins, C. H., J.lnst. Met., 40, 1921, p. 21.
 8
Microstructure and Mechanical
Properties of Solder Alloys

Harold}. Frost

The reliability of solder joints often depends directly on the mechanical properties
of the solder alloy. The tensile strength of the alloy determines the resistance
of the joint to mechanical rupture under short-term direct loading, as might occur
during impact testing. If the solder joint is subjected to cyclic stresses and strains
generated by thermal expansions during thermal cycling, the reliability depends
on resistance to fatigue. Because the temperatures of application of solder joints
are typically a substantial fraction of the melting temperature of the alloy, the
low frequency fatigue behavior is related to stress relaxation and creep behavior.
It is commonly recognized that the mechanical properties of metals and alloys
depend on their microstructures. Solder alloys are no exception to this general
pattern. Solder alloys are often a special case, however, because unlike most
structural alloys, they are used at such high temperature (relative to the melting
temperature) that the microstructures can change and evolve during service. What
is more, application conditions that involve thermal cycling may (or may not)
produce cyclic microstructural evolution, depending on the temperatures and
frequencies involved. Finally, cyclic mechanical stresses or deformations may
influence or promote microstructural changes.
The elements of microstructure that are typically important are the grain size
and shape, the size, shape, and spacing of second phase particles, and the
dislocation density. In eutectic solders the grain size and the second-phase size
and spacing are directly related because (at least) two phases are always present.
There are some solders that are essentially single-phase solid solutions in which
the precipitation of a second phase is not important (for example, Pb-5 wt %
In). Finally there are some solder alloys that are predominately one phase, but
in which an important second phase may precipitate (for example, Pb-5 wt %
Sn).

266
 THERMAL CYCLING FATIGUE 267

8.1 THERMAL CYCLING FATIGUE

The standard mechanical reliability concern for solder joints is the measurement
of resistance to failure by low cycle fatigue due to thermal cycling. If the service
conditions involve any thermal cycles, strains due to thermal expansion mismatch
are very difficult to avoid. Even with matched thermal expansion coefficients
there is likely to be some strain imposed on the solder because of differential
rates of heating and cooling in different regions. In many designs the imposed
strains exceed the elastic limit of the solder and produce plastic deformation. In
these conditions we may expect eventual failure by low cycle fatigue. The
duration of the fatigue lifetime that is required in service typically far exceeds
the time available for reliability testing. This means that service reliability must
be estimated using some sort of extrapolation. The situation becomes complex
because of the relatively high temperatures involved. Low cycle fatigue at more
than half the melting point of the solder will involve creep fatigue interactions.
Under these conditions, the number of cycles to failure depends on the cycle
frequency and on the stress cycle shape (that is, on hold times at load).
An understanding of the thermal fatigue of a solder joint requires that we
know the stresses and strains experienced by the solder. In the typical case, the
solder joint is much less stiff and more compliant than the other components
that the solder joint joins together. In this case, a specified thermal cycle results
in a specified strain cycle on the solder. The specified strain is made up of both
elastic and plastic strains. The stresses reached in such a cycle depend on the
elastic and plastic compliances of the solder. Plastic compliance results from
both very rapid plastic deformation at high stresses and much slower plastic
creep at lower stresses. During long hold times, plastic creep deformation re-
places elastic deformation during stress relaxation. If the strain cycle history is
known, and if the relationship between creep rate and stress is known, it is
possible to predict the stress cycle history. This sort of modeling has recently
been reported by Wilcox, Subrahmanyan, and Li and Subrahmanyan, Wilcox,
and Li. 2 In their work it is necessary to include the temperature dependence of
the plastic creep behavior. Creep relaxation is much faster at high temperatures.
The magnitude of the stresses reached upon heating is therefore lower than the
magnitude reached upon cooling.
The eventual failure of the solder results from the initiation and propagation
of fatigue cracks. Crack initiation probably results from the accumulation of
creep cavitation damage, the rate of which is determined by the time spent at
stress. (The rate of damage accumulation may also depend on temperature.)
Propagation of the cracks after initiation may also depend on the time spent at
stress. The fact that the stress relaxation is different in the heating and cooling
parts of the cycle means that the accumulation of damage will also be different.
Although the situation is complicated, it can be modeled if we have two rela-
tionships: one for the rate of plastic deformation as a function of stress and
268 MICROSTRUCTURE AND MECHANICAL PROPERTIES OF SOlDER AllOYS

temperature and another for the rate of damage accumulation as a function of

stress and temperature. In the general case these two relationships would also
depend on the strain and the temperature histories because of the effects of work
hardening and annealing. In some cases it may be reasonably accurate to neglect
such history effects. In the case where these two relationships are valid, it is
possible to detennine the relationships from experiments at one set of conditions
(frequency, temperatures, and strain levels) and use the relationships to predict
behavior (fatigue lifetime) at another set of conditions. That is, the long-term
service reliability could be extrapolated from short-term accelerated tests.

8.2 PRECIPITATION AND DISSOLUTION IN PB-SN ALLOYS

The equilibrium phase diagram for binary Pb-Sn alloys is shown in Figure 8-1.
For alloys with less than 19 wt% Sn it is possible to achieve a single-phase solid
solution of Sn in Pb by annealing at a sufficiently high temperature. At lower
temperatures (below the solvus temperature) the equilibrium structure has two
phases: precipitates of the tin phase within a matrix of the lead phase. For a
range of low tin compositions the solvus temperature is within the temperature
range that may be expected during service or during reliability testing (Fig. 8-
2). For these alloys, temperature cycles may produce cycles of precipitation and

Atomic Percent Tin

0 10 20 30 40 50 60 70 80 90 100
350
32?.I502"C

300

250

....
~
200

.....
....::f

..E
Co 150

f-

100
(jlSn)

50

0
0 10 20 30 40 50 60 70 80 90 100
Pb Weigh l Percen l Tin Sn

Figure 8-1 Equilibrium phase diagram for binary lead-tin alloys.3

 PRECIPITATION AND DISSOLUTION IN P8-SN ALLOYS 269

WEIGHT r. TIN
0 5 10 15
200
175 SINGLE PHASE
u
:... 150 ex
w 125
a:
:::> 100
f-
« 75
a:
w
0- 50 TWO PHASE
:L
w 25
f-
ex+~
0
-25
0 5 10 15 20 25
ATOMIC r. TIN

Figure 8-2 Limit of solid solubility (solvus) of tin in lead as a function of temperature, from
Tu and Turnbul1. 4

dissolution, depending on the frequency. If the thermal cycle frequency is very

low, the precipitation process will go to completion at the low temperature and
the dissolution process will go to completion at the high temperature. The mi-
crostructure will therefore be different in different portions of the thermal cycle.
If, on the other hand, the thermal cycle frequency is very high, neither the
precipitation nor the dissolution will be able to occur during a cycle. In this case
the microstructure will achieve some approximate steady state that may differ
from the equilibrium structure at either temperature.
The precipitation reaction usually proceeds as cellular precipitation, in which
cells of lamellae of the tin phase nucleate at the grain boundaries and grow into
the matrix. This process, also known as discontinuous precipitation, has been
well studied in Pb-Sn alloys by Turnbull and Treaftis5.6 and Tu and Tum-
bull. 4.7 - 9 More recent studies include Afanas'yev and Yelsukova,IO·11 Varkey
and Padmini,12 Osorio,13 and Hackney and Biancaniello. 14 An example of the
progressive growth of cells is shown in Figure 8-3. 15 The fraction transformed
f as a function of time t may be expressed by the Avrami equation for a given
temperature:

f = 1 - exp ( - Kf')

where the constant K depends on temperature, and n is independent of temper-

ature. The exponent n is between I and 4 for most transformations; Lutender
found n = 2. Figure 8-4 shows the results as time-temperature-transformation
curves for 50 percent completion, as calculated from fits to the A vrami equation.
Also shown in Figure 8-4 are curves for 6 and 9 wt% Sn taken from Borelius,
Larris, and Ohlsson. 16 The temperature for the maximum transformation rate
increases with increasing tin content. For 3 to 5 wt% Sn, the maximum rate is
270 MICROSTRUCTURE AND MECHANICAL PROPERTIES OF SOLDER ALLOYS

Figure 8-3 Micrographs of 97% Pb-3% Sn aged at 23°e for various times after solution
annealing. (a) 1 hr, 0 percent transformed, (b) 71 hr, 23 percent transformed, (e) 99 hr, 45
percent transformed, (d) 174 hr, 78 percent transformed.

near room temperature; for 2 wt% Sn, it is at a lower temperature. For practical
purposes, transformation in the 2 wt% Sn alloy does not become significant for
many days. Conversely, transformation in the 5 wt% Sn alloy near room tem-
perature is completed within 1 day .
An alternative mode of the precipitation reaction has recently been described
by Frear, Posthill, and Morris. 17 They found that the alloy Pb-5 wt% Sn showed
cellular precipitation if it was slowly cooled from a high temperature but under-
went a homogeneous precipitation reaction when it was quenched to room tem-
perature. Cellular precipitation is usually the preferred reaction because heter-
ogeneous nucleation at grain boundaries is easier than homogeneous nucleation
within the lead matrix and because the cell boundaries provide fast diffusion
paths for the tin to agglomerate. The homogeneous nucleation observed by Frear
et al. probably occurred at vacancies clusters formed during the quench. Un-
fortunately the quenching rate that is required to produce the homogeneous
nucleation is not known.
 PRECIPITATION AND DISSOLUTION IN PB-SN ALLOYS 271

50% TRANSFORMATION
80
70 6 wt.% 2.0 wt.%
....... -- 9 wt.% ... 3.0 wt.%
u 60 -a- 3.5 wt.%
•
........ 50 ... 5.0 wt.%
UJ
a:: 40
::::>
J- 30
<:
a:: 20
UJ
Q. 10
I:
UJ 0
J-
-10
-20
-30
10- 2 10 1

TIME (HOURS)
Figure 8-4 Time required for a 50 percent transformation for the precipitation reaction in
Pb-Sn alloys of various compositions. The curves for 6 and 9 wt % Sn are taken from the
resistivity measurements of Borelius, Laris and Ohlsson. 16

Early studies of the cellular precipitate structure (e.g., Tu and Tumbu1l4 )

discovered that the average spacing between tin lamellae decreased with increas-
ing tin concentration and decreasing precipitation temperature. Frost et al. 18.20
confirmed these findings, as shown in Figure 8-5. This behavior is certainly the
expected behavior for diffusion controlled phase transformation. It becomes
important for our consideration here because the flow stress of the alloy depends
on interlamellar spacing. Figure 8-6 shows examples of stress-strain behavior
for different microstructures. When the interlamellar spacing is smaller than
about 1 f-Lm, the precipitate structure provides significant strengthening.
When the temperature is raised above the solvus temperature (Figure 8-2),
the precipitate structure begins to dissolve. This may be a complicated process.
One possible mechanism is the direct diffusion of the tin into the lead phase.
Another mechanism is the retraction of the cell boundaries as a reversal of the
cellular precipitation process. The grain boundaries here provide a fast diffusion
path for the dispersal of the tin into the lead phase. This process is dominant at
the relatively lower temperatures that are only slightly above the solvus. Higher
temperatures result in faster dissolution as shown in Figure 8-7. This temperature
272 MICROSTRUCTURE AND MECHANICAL PROPERTIES OF SOLDER ALLOYS

l!)
z
- 2 Sn
U
..... 3
1:]- wt.~
«
0-
(/)-
(/)
10 ....
-0-
Sn
wt.~
3.5 wt.~ Sn
5 wt.~ Sn
«0
~z ~
3.23 wt.~ Sn
-.J~ -0- 6.7 wt.~ Sn
-.JU
1J.J -
1:1:
«-
-.J
5
~
1J.J
~
Z
o t.5S~~~.....L.-...L.L.......l.....~
-20 -1 0 o 1 0 20 30 40 50 60
TEMPERATURE (·e)
Figure 8-5 Average interlamellar spacings as a function of transformation temperature, for
various tin concentrations.

61~--~----~----~--~----~
Pb-5 wt.% Sn
51 1. ODC, 95% trans.
l 2. 25 DC, 95% trans.
3. 25 °C, overaged
~ 41 4. solution anne~a~le~d!--_ _- - ,

!!!,. . . /.----_.:.:-:-::.:i
~
~ ::'- - :' -~.'-'-"""""-"''';
........._-
.. .. ..

I~--~~--~--~~--~--~
.• .a .U .a .~ .n
TRUE STRAIN
Figure 8-6 True stress and true strain curves for tensile tests on the 95% Pb-5% Sn alloy,
tested at 25°C, with an initial tensile strain rate of D.l/sec.
 PRECIPITATION AND DISSOLUTION IN PB-SN ALLOYS 273

--
-*'
en
Q)
(,)
c:
o

-
'';::
as
0-
o
Q)

-
a.
'-

o
as
....
Q)
o I
<C
o 100 200
TIME (hours)
Figure 8-7 Dissolution behavior of the cellular precipitate structure at various annealing
temperatures, as measured by the area fraction containing the cellular precipitates as a function
of time.

effect results from both temperature dependence of diffusion and the fact that
the thermodynamic driving force for dissolution also varies with temperature.
The apparent activation energy for the process is theref.Jre not directly comparable
to the activation energy of the rate-limiting diffusion process.
Tu and Tumbull4 observed that dissolution by cell boundary retraction would
occur even at temperatures slightly below the solvus. (The effect extends to
about lOoC below the solvus.) One driving force for this retraction is the surface
energy of the interfaces between the lead and tin phases. Frost, Howard, and
Stone 19 have reported that the rate of dissolution by cell boundary retraction
increases with smaller interlamellar spacings. This effect may be the result both
of shorter diffusion distances required to disperse the tin and of the additional
driving force for retraction provided by the interphase interfaces. This means
that the dissolution rate depends on the temperature at which the precipitate
structure initially formed, as shown in Figure 8-8.
During the process of dissolution by cell boundary retraction, there is also an
evolution in the structure of tin lamellae within the cells. The tin phase actually
 274 MICROSTRUCTURE AND MECHANICAL PROPERTIES OF SOLDER ALLOYS

-- Pb - 5 wt.% Sn
-*'
( /)

Q)
Precipitation Temperature
• -20 °e
()
A ooe
c:: c 23°e
o
)( 40 °C

-
"-
c " ......
o 20 _ x
ctS
Q)
t-
oe:(
a
a 100 200
TIME (hours)
Figure 8-8 Dissolution behavior of the cellular precipitate structure in the 95% Pb-5% Sn
alloy at the annealing temperature of 1oODe for four different ageing temperatures. The different
aging temperatures that correspond to different interlamellar spacings, as shown in Figure 8-
5, result in remarkably different dissolution rates.

occurs as blades with numerous rifts and holes instead of complete plates (Tu
and Turnbu1l4 ,9). They are therefore subject to instabilities due to capillarity
effects that lead to break-up of the lamellae and spheroidization. This pheroid-
ization was observed by optical metallography, 1,4 and is also evident in Figure
8-5. The extent of spheroidization is also shown by a comparison of Figures 8-
9 and 8-10. Figure 8-9 shows a high magnification view of the early stage of
dissolution by cell boundary retraction. The tin phase remains predominantly in
the form of well-defined lamellae. Figure 8-10 shows a much later stage of
dissolution in which a region of tin precipitates has spheroidized.
During the spheroidization process there is also some dissolution of the tin
phase by lattice diffusion directly into the lead matrix. For temperatures near
the solvus, however, the lattice diffusion is so slow that direct dissolution by
lattice diffusion is not completely accomplished before the cellular dissolution
is nearly complete. Cellular dissolution by cell boundary migration continues
 PRECIPITATION AND DISSOlUTION IN PB-SN AllOYS 275

after the spheroidization has become prominent, but the boundaries appear to
slow down. Spheroidization should reduce two components of the driving force
for the cell boundary migration. It reduces the area of lamella-matrix interface,
and it also disperses some tin into the lead matrix, thereby reducing the chemical
driving force for cell boundary migration.
Just as the precipitate microstructure is unstable to coarsening and spheroid-
ization during dissolution above the solvus, it is also unstable below the solvus.
The initial cellular precipitation process cannot produce a microstructure that is
completely at eqUilibrium. There must be some free energy of the interfaces
between the two phases, and the compositions of the two phases will deviate
slightly from their equilibrium values. These factors provide a driving force for
evolution toward a coarser microstructure. If the tin were distributed in small
spherical particles, then those particles should coarsen in the usual manner. Frear
et al. 17 report a decrease in hardness during ageing that is directly attributable
to such coarsening. If the tin were distributed in planar lamellae, without im-
perfections, then coarsening might occur in a discontinuous fashion by the pas-
sage of interfaces that consumed the fine lamellae and left behind coarse lamellae.

Figure 8-9 Cell boundary retraction in the 95% Pb-3.5% Sn alloys. The precipitate structure
formed at 23°C. The dissolution anneal was for 39.7 h at 90°C. The migrating grain boundary
is visible on one side of the denuded zone.
276 MICROSTRUCTURE AND MECHANICAL PROPERTIES OF SOLDER ALLOYS

Figure 8-10 Microstructure of 95% Pb-5% Sn after precipitation at room temperature,

followed by annealing for 17 h at 110°C. Most of the precipitate structure has undergone
cellular dissolution. In the region shown the lamellar structure of the tin precipitates has
undergone spheroidization.

Although the process of discontinuous or cellular coarsening has long been

recognized (e.g., Livingston and Cahn21), it has not yet been quantitatively
studied in Pb-Sn alloys.

8.3 DISCUSSION

The complexity of the dissolution and coarsening processes means that it is

somewhat difficult to predict exactly what microstructure would result from an
arbitrary cyclic thermal history. We may expect precipitation to proceed at the
lower temperatures and dissolution to proceed at the higher temperatures. If the
cycle time is long compared to the phase transition time, the microstructure
should cycle in a manner consistent with the simple experiments. The rate of
precipitation from the solid solution is fairly well known, and although the
dissolution rate depends on the precipitate structure, it may also be estimated.
If the cycle time is short, neither precipitation nor dissolution will be complete
 REFERENCES 277

in one cycle, and the microstructure will drift toward a structure determined by
whichever process is dominant.
The importance of microstructural evolution to the thermal fatigue problem
involves the boundary between very low frequency behavior (cyclic microstruc-
ture) and high frequency behavior (steady-state microstructure). If the cyclic
strain is specified, the cyclic stresses will depend on the microstructure. If the
initiation and propagation of creep-fatigue damage depends on stress, the fatigue
lifetime will also depend on the microstructure. It may therefore be inaccurate
to extrapolate across the frequency boundary between high and low frequency
behavior. Furthermore, for either high or low frequency cycling, the strength of
the alloy should depend on the temperature limits. This means that an extrap-
olation from thermal cycling from one temperature range will involve more than
just the extrapolation from one level of thermal strains to another level of thermal
strains.
From the discussion above it is clear that the effect of changing cycle frequency
is intertwined with the effect of change in the cyclic temperature limits. Because
both the precipitation and dissolution rates depend on temperature, we expect
that the boundary between low and high frequency depends on the temperature
limits involved. The situation is complicated by the fact that the dissolution rate
depends not just on alloy composition and dissolution temperature but also on
the temperature at which the precipitate structure was formed. A proper estimate
of the frequency boundary must take this into account. In conclusion, a proper
extrapolation of the mechanical behavior from one set of conditions to another
must take the microstructural evolution into account.

REFERENCES

1. Wilcox, J. R., R. Subrahmanyan, and Che-Yu Li, "Thermal Stress Cycles and
Inelastic Deformation in Solder Alloys," Proceedings of 1989 ASM Electronic Pack-
aging Conference-Materials and Processes. ASM International, 24-28 April 1989,
Philadelphia, PA., pp. 203-211.
2. Subrahmanyan, R., J. R. Wilcox, and Che-Yu Li. "A Damage Integral Approach
to Solder Joint Fatigue," Proceedings of 1989 ASM Electronic Packaging Confer-
ence-Materials and Processes. ASM International, 24-28 April 1989, Philadelphia,
PA, pp. 213-221; also "A Damage Integral Approach to Thermal Fatigue of Solder
Joints." Proceedings of 39th Electronics Components Conference, IEEE, 22-24 May
1989, Houston, TX.
3. Karakaya, I., and W. T. Thompson, "The Pb-Sn (Lead-Tin) System," Bulletin of
Alloy Phase Diagrams, 9(2), 1988, pp. 144-152.
4. _ _ , "Morphology and Kinetics of Cellular Dissolution of the Pb-Sn Alloy,"
Metallurgical Transactions, 2, 1971, pp. 2509-2515.
5. Turnbull, D., and H. N. Treaftis, "Kinetics of Precipitation of Tin from Lead-Tin
Solid Solutions," Acta Metallurgica, 3, 1955, pp. 43-54.
278 MICROSTRUCTURE AND MECHANICAL PROPERTIES OF SOLDER ALLOYS

6. _ _ , "Micrographic Investigation of Precipitation in Pb-Sn Alloys," Trans AlME

212, 1958, pp. 33-42.
7. Tu, K. N., and D. Turnbull, "Morphology of Cellular Precipitation of Tin from
Lead-Tin Bicrystals," Acta Metallurgica, 15, 1967, pp. 369-376.
8. _ _ , "Morphology of Cellular Precipitation of Tin from Lead-Tin Bicrystals-
n," Acta Metallurgica, 15, 1967, pp. 1317-1323.
9. _ _ , "Morphology and Structure of Tin Lamellae Fonned by Cellular Precipi-
tation," Acta Metallurgica, 17, 1969, pp. 1263-1279.
10. Afanas'yev, N. I., and T. F. Yelsukova, "Ageing Peculiarities of Lead-Tin Alloys
and Experimental Testing of Continuous Decomposition Theories," Phys. Met. Me-
tali., 53, 1982, pp. 129-134.
11. _ _ , "The Influence of Impurities on the Rate of Discontinuous Decomposition
of Lead-Tin Alloys," Phys. Met. Metall., 57, 1984, pp. 86-91.
12. Varkey, P. A., and A. R. K. L. Padmini, "Precipitation Studies in Lead-Tin Alloys,"
Acustica 52, 1983, pp. 290-295.
13. Osorio, F. S., "Morphology of the Cellular Precipitation Reaction in a Pb-3.0 wt%
Sn Alloy," Thesis, University of Missouri-Rolla, 1973.
14. Hackney, S. A., and F. S. Biancaniello, "Experimental Indications of a Low Un-
dercooling Instability in the Cellular Mechanism of Discontinuous Precipitation,"
Scripta Metallurgica, 21, 1987, pp. 371-376.
15. Lutender, S. D., "Micrographic Study of the Precipitation and Dissolution Kinetics
of Selected Pb-Sn Alloys," M.E. Thesis, Dartmouth College, 1986.
16. Borelius, G., F. Larris, and E. Ohlsson, "Kinetics of Precipitation in Pb-Sn Alloys,"
Arkiv For Maternatik, Astronomi och Fysik, 31A(10), 1944, pp. 1-19.
17. Frear, D. R., J. B. Posthill, and J. W. Morris, Jr., "The Microstructural Details of
~-Sn Precipitation in a 5Sn-95Pb Solder Alloy," Metallurical Transactions, 20A
1989, pp. 1325-1333.
18. Frost, H. J., R. T. Howard, P. R. Lavery, and S. D. Lutender, "Creep and Tensile
Behavior of Lead-Rich Lead-Tin Alloys," IEEE Trans. on Components, Hybrids and
Manu/act. Tech., 11, 1988, pp. 371-379.
19. Frost, H. J., R. T. Howard, and G. J. Stone, "The Effects of Thennal History on
the Microstructure and Mechanical Properties of Solder Alloys," Microelectronic
Packaging Technology, Materials and Processes. W. T. Shieh, Ed. ASM Internation,
1989, pp. 121-127.
20. Frost, H. J., P. R. Lavery, and S. D. Lutender, "Microstructure and Mechanical
Properties of Lead-Tin Alloys," in Electronic Packaging and Corrosion in Microe-
lectronics, M. E. Nicholson, Ed. ASM International, 1987, pp. 259-267.
21. Livingston, J. D., and J. W. Cahn, "Discontinuous Coarsening of Aligned Eutec-
toids," Acta Metallurgica 22, 1974, pp. 495-503.
 9
The Interaction of Creep and
Fatigue in Lead-Tin Solders
John K. Tien, Bryan C. Hendrix, and
Abbas I. Attarwala

For 2000 to 3000 years, Pb-Sn solders have found extensive use in plumbing,
and more recently, in electrical connections because of their good wetting prop-
erties and low melting temperature. But solder has been avoided for mechanical
connections because of its low strength at ambient operating temperatures, which
is above one-half of the absolute melting temperature (O.5Tm) of Pb-Sn solder.
However, during the past 15 years, the packaging of electronic components has
moved from technology that used solder predominantly as an electrical connec-
tion (plated through-hole) to technology that uses solder as both a mechanical
and electrical connection (surface mount technology, SMT). In addition, elec-
tronic, components are expected to withstand increasingly harsh temperature
cycles. Accordingly, thermal fatigue and creep of solder joints has emerged as
a source of failure of SMT components, l where temperatures playa role not
only in inducing the stresses but also in thermally activated deformation and
rupture.

9.1 CURRENT APPROACHES TO ACCELERATED TESTING

Thermal testing is used to measure electronic component reliability and to es-

timate component life. Testing for the thermal fatigue or cyclic creep life of
components with realistic temperature cycles would take as long as the design
life of the component. For this reason, components are often tested by accel-
erating the thermal cycles (i.e., increasing the heating rate and/or decreasing the
cycle hold time).

279
280 THE INTERACTION OF CREEP AND FATIGUE IN LEAD-TIN SOLDERS

In order to predict lives from accelerated tests, the damage stored during a
realistic thermal cycle must be related to the damage stored in the accelerated
thermal cycle. Damage in solder joints will accumulate from thermally activated,
time dependent mechanisms (creep), from athermal cyclic mechanisms (fatigue),
and from microstructural changes. These damage mechanisms are expected to
interact with one another and to have different relative magnitudes as a test is
accelerated. Ab initio, one would expect that for any thermal cycle where the
solder does not completely relax during the time at maximum temperature, the
creep damage inflicted will be hold time dependent. For subzero minimum
temperatures, the time dependent damage should be small compared to the
athermal damage and so hold time would not be expected to be important. By
the same reasoning, the heating or cooling rate could be accelerated in the cold
regions of the cycle but not in the hot parts without resulting in time or cycle
dependent complications. As will be discussed below, these complications in-
clude not only creep but also anelastic relaxation.
Coffin-Manson-type relationships have been modified to account for frequency
and temperature effects.2-6 These approaches are empirical methods for keeping
track of how much damage is stored in each cycle and relating this damage to
the life of the material. This empirical approach is effective as long as the model
is developed carefully within a particular temperature range and electronic pack-
age type. It can be argued that the most trustworthy models fit data from ac-
celerated thermal tests of components, 2,5 while others use bulk specimens3,4 and
isothermal tests.3,4 One of the problems with this empirical approach is that it
takes too much time to develop good tests for a particular application and then
tests need to be redeveloped for other applications.
The cyclic nature of the stress history of electronic packages implies at least
some sort of fatigue behavior, if not creep/fatigue behavior, and has led to the
current approach of a modified Coffin-Manson equation. This may still be in-
adequate because of the high homologous temperatures involved with solder
applications since, even at room temperature, creep mechanisms will also con-
tribute significantly to the damage. In order to improve the way in which models
are developed and applied, the fundamental behavior of the solder needs to be
examined at realistic temperatures and with the application of cyclic stresses.

9.2 DAMAGE BY FATIGUE AND CREEP MECHANISMS

The cyclic nature of the temperature variation causes the strain experienced by
the solder joints to be cyclic and, therefore, the damage in the solder to be a
function of the number of thermal cycles. By this definition, SMT solder joints
experience not only fatigue but thermal fatigue. At low temperatures (below
T/Tm = 0.5), fatigue can cause damage by the initiation and propagation of
microcracks. In the above equation, T is temperature and Tm is melting tem-
 DAMAGE BY FATIGUE AND CREEP MECHANISMS 281

perature. The usual equations used to describe fatigue life Nf are the Coffin-
Manson relation: 7

(9-1)

This is an empirical equation that predicts fatigue life. Low cycle fatigue is
controlled by the rate of crack growth. In general, low cycle fatigue controls
life up to around lQ4 cycles and high cycle fatigue controls life above 106 cycles.
The first term describes low cycle fatigue where the damage stored in each
cycle is related to the plastic strain range aEp raised to the negative A power.
The second term describes high cycle fatigue where most of the life of the
component is spent in the initiation of cracks. Here the damage stored in each
cycle is best described by the stress range aO" raised to the negative B power.
In equation (9-1), A is the low cycle fatigue Coffin-Manson exponent, B is the
high cycle fatigue Coffin-Manson exponent, Ef is the fracture strain, and 0"0 is
the characteristic stress for high cycle fatigue.
In general, surface mounted components see thermal cycling that can range
from O.4Tm to O. 75Tm' Accordingly, it can be argued that in most microelectronic
applications, the life of a component is spent at temperatures above 0.5 = TITm,
which is about room temperature for the highest melting point Pb-Sn solders.
As a rule of thumb, creep is an operative deformation mechanism whenever the
temperature is above 0.5Tm' Creep strain can result from any of a number of
thermally activated, time dependent mechanisms. These mechanisms can obey
a constitutive relation of the form:

1)0 1 (0" - O"b)n

(d Eo (9-2)
C kT E(T) exp kT

C is a constant, d refers to grain size and a to grain size sensitivity, 0" is the
applied stress with O"b being the back stress and E(T) the modulus as a function
of temperature, making n the stress exponent. The thermal activation of creep
is characterized by an activation energy Eo and k is Boltzman's constant.
At high stresses, creep is controlled by dislocation motion. When dislocation
entanglement and recovery reach an impasse where the rate of hardening is equal
to the rate of recovery, a quasi-steady state is defined that obeys Equation (9-
2). This creep rate is controlled by the rate at which edge dislocations can climb
out of their slip planes. In this case the grain size exponent a will be negligible,
the stress exponent n will be between 3 and 6, and the activation energy Eo will
be the energy for vacancies to travel to climb sites at dislocations S- 1O whether
by bulk diffusion or some short circuit path.
282 THE INTERACTION OF CREEP AND FATIGUE IN LEAD-TIN SOLDERS

At lower stresses, creep strain results directly from the motion of vacancies
from grain boundary to grain boundary, which results in the elongation of grains.
Here, the grain size sensitivity a becomes important. At the highest temperatures,
the vacancies move by lattice diffusion and the creep mechanism is referred to
as Nabarro-Herring creep. Nabarro-Herring creep is characterized by grain size
exponent a = 2, a stress exponent n = 1, and an activation energy Ea of the
lattice diffusion energy,u At lower temperatures, the vacancies move by grain
boundary diffusion and the creep mechanism is referred to as Coble creep. Coble
creep is characterized by a grain size exponent a = 3, a stress exponent n =
1, and an activation energy Ea of grain boundary diffusion. 12
In some cases at intermediate stresses, materials can undergo superplastic
deformation. Although as-cast solders do not typically exhibit superplastic be-
havior, solder alloys in the rolled and annealed condition are often superplastic,
and their superplastic deformation can provide some insight into the deformation
of as-cast solder. Superplastic deformation is characterized by extremely large
strains caused by steady stresses. In practice, strains of 500 to 1000 percent are
often necessary before failure can occur. Stress exponents are usually less than
3 regardless of the particular mechanism since higher stress exponents make
specimens sensitive to necking.
The most common type of superplastic flow involves grain boundary sliding
and migration in fine grained materials. This mechanism of superplastic flOW 13 .14
has the same grain size, stress, and temperature dependence of Nabarro-Herring
or Coble creep, depending on the temperature. However, the strain rate is a
couple orders of magnitude higher than Nabarro-Herring or Coble creep because
the shape change of the grains merely allows grains to slide past one another
and reorganize. No permanent elongation of the grains is required.
The damage that is stored during creep deformation can be of three basic
types: creep cracks, void nucleation and growth, and microstructural degradation.
Creep cracks are formed and grow at lower temperatures where flow mismatch
at grain boundary triple points or persistent slip bands cannot be accommodated
by the flow of the adjacent grains. Void nucleation and growth at grain boundaries
occur throughout creep under most conditions even when voids have little to do
with the final failure. There are many models for void nucleation,15.16 most of
which link nucleation to grain boundary sliding and impurities or particles at the
grain boundaries. As voids grow and link, the supporting cross section is de-
creased and the stress increases. Microstructural degradation is often the most
serious damage in alloys that depend on the phase morphology for creep resis-
tance. Stresses at high temperatures allow strengthening precipitates to coarsen
and change shape, weakening the alloy. Recrystallization can also occur as
dislocation tangles recover and grow. It should be noted that the localized weak-
ening by recrystallization can localize the creep strain, causing necking and
failure.
 ASSESSING ACTUAL JOINT DAMAGE 283

9.3 ASSESSING ACTUAL JOINT DAMAGE

The above reviews damage storage leading to failure in bulk specimens. The
next issue is the more realistic or at least the perceived realities of the damage
in actual joints. During a thermomechanical cycle, the SMT component, the
solder, and the wiring board expand or contract at different rates. Under this
action, damage is stored in the solder. At the lower temperature reaches of the

Leadless Carrier

D completely relaxed cycle

13 accelerated cycle

solder strain

.....1 - - - - strain mismatch between _ _ _ _ _••

substrate and component

Leaded Carrier
D completely relaxed cycle

13 accelerated cycle

solder strain

Figure 9-1 Stress-strain hysteresis in the solder are a measure of the amount of damage
stored. The figure compares accelerated hold times to long hold times. 2
284 THE INTERACTION OF CREEP AND FATIGUE IN LEAD-TIN SOlDERS

thennal cycle, at least some of the damage will be athermal dislocation entan-
glement. During the higher temperature parts of the cycle, nearly all of the
damage is stored in a time and temperature dependent manner by one of the
creep mechanisms.
The integration of the solder stress versus solder strain to give the energy
stored in the solder over a thermal cycle is shown in Figure 9-1 for generic
leaded and leadless packages. 2 The slope and plastic yield point of the on-load
ramp is a function of the temperature rate and the solder response. The consti-
tutive relation of the solder at that strain rate and each temperature determines
the shape of the stress-strain relation. In this way, any change in temperature
rate can change the entire shape of the stress-strain relationship. The shape of
the stress-strain relation during the relaxation of the solder during the upper
temperature hold time (see Fig. 9-1) is controlled by the bending compliance of
the SMT component-wiring board or SMT component-lead-wiring board system,
depending on whether the package is leadless or leaded, respectively. If the

20 -r---------------------------------------,

15

N
:I:
C')

e. ----- ....... ......

C')

-. 60% Sn - 40% Pb
CIS
a..
10
....... solder
(!)
II)
:::l FR-4 "",
"5 glass/epoxy \
"C
0
:::E \
\
5 \
\
\
\

"...... ........ _-
o-+------~------~------~------~------~

-50 o 50 100 150 200

Temperature (0C)

Figure 9-2 Schematic modulus as a function of temperature for solder and FR-4 glass or
epoxy board.
 ASSESSING ACTUAL JOINT DAMAGE 285

package system is very compliant, large amounts of mismatch can be accom-

modated at low stresses that store less damage in the solder joint. High com-
pliance can be achieved by using long, thin leads or thin, compliant boards. It
is important to note that the compliances of both solder and polymeric board
materials are quite temperature dependent as shown in Figure 9-2.
The stress relaxes toward zero during the upper temperature hold at a rate
depending on the temperature through a yet to be determined constitutive equation
for the solder joint. One problem with the development of a solder constitutive
equation is that it must incorporate both deviatoric and hydrostatic stresses. An
even more difficult factor is that the microstructure of a solder joint can change
during the thermal cycle. These changes can include coarsening of the Pb and
Sn phases, recrystallization, and grain growth.
The damage stored in an accelerated test can be compared with the damage
stored in a realistic thermal cycle by comparing the stress-strain hysteresis ex-
perienced by the solder. The accelerated test can be valid if the dominant damage
storage mechanism is the same as in the real thermal cycle (i.e., the mechanism
that controls failure during the application must be the active mechanism in the
test.) In addition, if the accelerated and realistic cycles are well defined, some
estimation of life might be made.

9.3.1 In-service Testing

There is always a need to measure the damage in a component after it has

experienced some amount of in-service use in order to determine the residual
life left in the component. There have been a number of proposed in-service
tests, some of which destroy the component in question and some which attempt
to probe the damage without destroying the joints.
The nondestructive methods focus upon measuring cracks that have developed
in the joints. Ultrasonic microscopy detects the discontinuity of the modulus in
the solder at the crack surface and can map out the cracks in all the joints. 17
Electrical conductivity techniques measure as a function of time the loss of joint
cross section by crack or void growth. Conductivity is also a function of the
growth of brittle intermetallic layers and microstructural changes in the solder.
Likewise, the thermal conductivity of joints can be examined by infrared mapping
of the temperature distribution during operation. These nondestructive techniques
are in the stage of initial development.
Another sort of nondestructive test that is being examined is the measurement
of the hysteresis in the temperature versus bending strain of the wiring board.
The bending strain is usually measured with a strain gauge mounted on the wiring
board surface opposite the SMT component. It has been found that the area of
the hysteresis decreases according to a log-linear relationship with the number
of cycles that a component has seen up to failure. 18
286 THE INTERACTION OF CREEP AND FATIGUE IN LEAD-TIN SOLDERS

The destructive examination of joints is being proposed for measuring mi-

crostructural damage in the joints. 19-22 A more recent development is the mon-
itoring of the voids on grain boundaries. 23 Grain boundaries can be exposed for
examination by embrittling the boundaries with liquid gallium. There is a cor-
relation between the void distribution on the grain boundary surfaces and the
amount of creep-fatigue damage that has been stored. The details are still under
investigation. 24,25

9.4 UNDERSTANDING THE DAMAGE MECHANISMS

It should be evident that in order to understand and implement accurate and

dependable accelerated testing and in-service testing, we need to more fully
understand solder damage storage behavior as a function of time, temperature,
stress, and microstructure. Conversely, as mentioned, a priority for mechanical
property understanding is the determination of what the dominant modes of
damage storage are in realistic temperature and strain rate regimes.
In what follows, a self-consistent set of experiments and their results are
presented. These experiments are designed to separate the different types of
strain in isothermal cyclic creep tests in order to understand the damaging strain
from benign strain. Bulk cast material was used in this development of the form
of constitutive equations for joints. Cyclic creep tests were performed on two
alloys of nominal composition 63% Sn-37% Pb (eutectic) and 81% Pb-19% Sn,

•
10 2
o
U
-
CD
en 10 1 ~

...as
CD
10 0
a::
c o
'i!...
tn
I
10-1

I~
eutectic
0 81-19
10-2
5.0 10.0 20.0 40.0

Stress (MPa)

Figure 9-3 Minimum creep rate as a function of stress for bulk eutectic and 19% 5n-81 %
Pb solders.
 UNDERSTANDING THE DAMAGE MECHANISMS 287

both of which have their solidus at the eutectic temperature (183°C). Experimental
details can be found elsewhere. 24-29

9.4.1 Creep and Tensile Test Results

The creep results shown in Figure 9-3 indicate that the minimum creep rate can
be described by a power law Equation (9-2) with a stress exponent n equal to
5 to 6 in the range of stresses tested at ambient temperature. Cyclic tests were

BALANCED WAVE FORM

.)

HOLD TIME ON·LOAD

b)

HOLD TIME OFF· LOAD

c)

FAST RAMP TRAPEZOID

d)
r 1

I ...
TIME

Figure 9-4 Waveforms used during cyclic testing. (a) Balanced wave form. (b) Hold time
on-load. (c) Hold time off-load. (d) Fast ramp trapezoid.
288 THE INTERACTION OF CREEP AND FATIGUE IN LEAD-TIN SOLDERS

100
0 81-190.2% V.S .

•
-
81-19 U.T.S.
0 eutectic 0.2% V.S.
as
•
-
50 eutectic U.T.S. I
D..
:E ~~ I! ~

•
-
til
til
...
CD 0
en 20

10~~~~m·r-I~~~~·I~~~~·r-I~~~W·I--~~~
10 -6 10 -5 10 -4 10 -3 10 -2 10 -1
Strain Rate (1/sec.)
Figure 9-5 Ultimate tensile and 0.2 percent yield stresses for bulk eutectic and 19% Sn-
81 % Pb solders.

designed with the waveforms shown in Figure 9-4 and stresses such that the
corresponding static tests had lives greater than 3600 sec (1 h). Initial creep tests
were performed on joint specimens at 27.6 and 55.2 MPa. The 27.6-MPa test
elongated immeasurably after 60 h and so 55.2 was chosen as the stress at which
to test the frequency dependence of the joint configuration. The 19% Sn-82%
Pb alloy was tested at 0 to 13.8 MPa since this was below the 0.2 percent yield
stress down to 8 x 10-5 sec-I. The tensile results in Figure 9-5 indicate a power
law material at lower strain rates and a less strain rate dependent behavior at
higher strain rates. The fast ramping trapezoidal waveform (Figure 9-4d) was
designed to have a strain rate around 10-2 sec-I. This served to minimize the
amount of time dependent strain being stored during the on-load ramp and
provides for the separation of strain storage mechanisms. Strain controlled cyclic
creep tests were designed at 200 ,....E for comparison to the 27-MPa stress con-
trolled tests. Higher alternating strains were used to find the effects of plastic
increments on life and stress relaxation.

9.4.2 Cyclic Creep

The rate of damage storage in a stress controlled cyclic creep test is the envelope
strain rate shown schematically in Figure 9-6. Envelope strain is the measure of
the nonrecoverable strain stored in each cycle. For each test, an envelope strain
curve was obtained by plotting the point of maximum strain in each cycle as a
function of time. The minimum slope of the resulting curve gives the minimum
 UNDERSTANDING THE DAMAGE MECHANISMS 289

o "envelope" strain
strain as a function of time

c:
.~

ii5

Time

Figure 9-6 Schematic envelope strain curve. 0 = envelope strain; - - = strain as a

function of time.

envelope strain rate that can be fit to a power law such as equation (9-2). The
envelope strain rate in a cyclic test can be compared with the strain rate of a
static creep test. In the case of zero minimum stress, the creep rate of the cyclic
test should be measured as a function of the time spent on-load rather than total
time, which includes time spent off-load. In the case of nonzero minimum stress,
a method is developed below for comparison of test results.
The cyclic creep results for the eutectic alloy specimens are shown in Figure
9-7 as a function of stress range and in Figure 9-8 as a function of mean stress.
A comparison of the two methods of plotting indicates that the rate of storing
permanent damage is a weak function of stress range but a strong function of
mean stress. This underscores the importance of the creep component as opposed
to a strictly fatigue component. A factor of 2 in stress range can make a difference
of perhaps a factor of 5 in damage storage rate or life; a factor of 2 in mean
stress produces a change of around 2 orders of magnitude in damage storage
rate.
If the stress exponent n in Equation (9-2) is other than 1, the effective mean
stress during a cyclic creep test should be defined as the average of each stress
raised to the n power.

(9-3)

where

(9-4)
290 THE INTERACTION OF CREEP AND FATIGUE IN LEAD-TIN SOLDERS

10 2
0
0 Mean S15=6.9, 3 sec

U
III 10 1
-• Mean Sls=1 0, 100 secs

Mean Sls=10.5, 3 sec

0
0

.
0
.....
III 0 Mean Sls= 13.8, 3 sec

-
~

- - .- •
0

III
( II
10 0 •
-
a:

-
(/J
c
...as 10- 1
- 0
0

10-2 -t--------r--------r----'
4 10 25

Stress Range (MPa)

Figure 9·7 Minimum envelope strain rate as a function of stress range.

10 3 y.r===============r-------,
0 Strss Rng=5, 100/sec

• Strss Rng=10, 100/sec

•
10 2 0 Strss Rng=13.8, 3/sec .0

• Strss Rng=27.6, 3/sec

• i•
10 1 0 •

III 10 0 -0
• 0

1ii 0 0
a:
c •
-
'l!
( /)
10 -1
0
0
0
o

10-2
0.6 1.3 2.5 5.0 10.0 20.0 40.0

Mean Stress (MPa)

Figure 9·8 Minimum envelope strain rate as a function of mean stress.

 UNDERSTANDING THE DAMAGE MECHANISMS 291

In the above equations, CJ'eff is the effective average creep stress, CJ'max is the
maximum stress, and CJ'min is the minimum stress. Hence, the cyclic creep data
from Figures 9-7 and 9-8 are replotted as a function of effective mean stress in
Figure 9-9 with n taken equal to 6. In these plots, the use of the effective mean
stress allows us to correlate the information of the specimen tested at a constant
mean stress (EllO) to that of the specimens tested at constant stress range (E103
and EI05). Effective mean stress also takes into account the "time on load" for
specimens with zero minimum stress. The plot of log strain rate versus log
effective stress shows that the data points from the same specimens are approx-
imately parallel to each other. This behavior with respect to effective mean stress
indicates that even the small stress range effect shown in Figure 9-7 is, in fact,
a result of creep damage and not an indication of some small amount of athermal
fatigue damage.
The IOO-sec hold time data is separated according to specimen so that the
scatter caused by different specimens is obvious. The IOO-sec hold time tests
were performed on only three specimens. Multiple tests were performed on single
specimens in order to minimize the effects of microstructural variations in the
initial castings. Tests were interrupted after they had reached a minimum strain
rate and then the specimen was used again for tests at other stresses. The influence

10 3 rr=========r------,
D 100 sec, specimen El03
• 100 sec, specimen El05
• 100 sec, specimen Ell0
D
•

-()
Q)
1/1 10 1 D
•

-
~
•
Q)
as 10 0 •
II: D •
.ca; • •
... lD- 1
tii a
hD D
De

10- 2
2.5 5 10 20 40

Effective Stress (MPa)

Figure 9-9 Minimum envelope strain rate as a function of effective mean stress defined by
Equation (9-4). El03 and El05 had constant stress ranges of 5 MPa and 10 MPa, respectively;
El10 had a constant mean stress of 10 MPa.
292 THE INTERACTION OF CREEP AND FATIGUE IN LEAD-TIN SOLDERS

of the prior stress history on the strain rate is demonstrated by comparing the
points a, e, and h in the lower left-hand corner of Figure 9-9 from specimen
E103. These three tests were performed at the beginning, middle, and end of
specimen EI03's test history. Comparing the scatter between points a, e, and h
to the scatter between specimens indicates that the scatter introduced by the
evolving microstructure is less than the scatter introduced as a result of the initial
microstructural variations between the different specimens. It should be noted
that the casting conditions for bulk specimens in the lab are quite well controlled
as compared to typical solder joint reftow conditions.
The temperature dependence of the envelope strain rate is shown in Figure 9-
10. The 27.6-MPa 3-sec hold time data show an apparent activation energy of
0.95 eV. The IO-MPa 100-sec hold time data show an apparent activation energy
of O. 86 e V. These activation energies are in line with the bulk diffusion energies
for Pb or Sn30 ,31 and with other creep studies on lamellar eutectic Pb_Sn.32 In
practical terms, this implies that for IOC variation in test temperature around
room temperature, the rate of nonrecoverable strain storage will change by about
12 percent.
In a strain controlled test, the strain range dependence of solder life obeys a
Coffin-Manson-type relationship when all other variables remain constant, as

-
10 3

• •c 27.6 MPa

-
()
CD 10.0 MPa
t/) 10 2
•
-~

-
10 1
CD
a:I
a: 10 0 c
c •
-
C c c
...
a:I
10 -1
c B
8 tI
en -:

10 -2 I I I

3.1 e-3 3.2e-3 3.3e-3 3.4e-3 3.Se-3 3.6e-3 3.7e-3

1 IT
Figure 9-10 Minimum envelope strain rate as a function of temperature for lOO-sec hold
time at 10 MPa and 3-sec hold time at 27.6 MPa.
 UNDERSTANDING THE DAMAGE MECHANISMS 293

10 5

~
Q)
CI
•
s:::: 10 4

...
CIS
a::
•
-
s::::
...
'iii
en
10 3
10 2 10 3 10 4 10 5
Cycles to Failure

Figure 9-11 Cycles to failure Nf as a function of strain range for 3-sec hold times.

shown in Figure 9-11. The failure of a specimen was declared when a measurable
drop in the apparent modulus resulted from crack growth reducing the cross
section of the specimen. All tests were cycled from zero strain to the maximum
strain of the test. The minimum stress started at zero but relaxed to a negative
value during a shakedown period until the mean stress approached zero. The
shakedown of the stress is a measure of permanent damage in the specimen, but
some of the creep strain stored on the positive side of the cycle is macroscopically
recovered on the negative side of the cycle even though it could still be quite
damaging. Likewise, damage is stored on the negative side of the cycle, though
the resulting strain is erased on the positive side of the cycle. The larger strain
range tests stored a significant amount of plastic strain on the ramp-up; therefore
the mean stress is essentially zero after the first cycle. In these large strain range
tests both athermal and thermal components of the nonrecoverable strain con-
tribute to the stress relaxation in the shakedown process.

9.4.3 Hold Time Effects

The acceleration of a fatigue or thermal fatigue test pivots about the cyclic hold
time and the ramp rate. The ramp rates determine the relative amounts of time
dependent and time independent strains that are stored during the ramp. The
hold time controls the amount of time dependent strain stored during the hold.
These tests were designed to help understand the effect of accelerating a fatigue
test. The earliest of these tests kept a constant wave shape (Figure 9-4a) and
varied the frequency. Once it became apparent that the damage controlling
mechanisms were time dependent, the ramp rate was kept constant and only the
hold time was varied (Fig. 9-4d).
The effect of hold time on the envelope creep is shown in Figure 9-12. As
the hold time decreases (i.e., as the test is accelerated), the creep rate decreases
294 THE INTERACTION OF CREEP AND FATIGUE IN LEAD-TIN SOLDERS

32

28

24
N'
.....,
.5 20
~CD
16
Q.
0
Q) 12
>
as 8

0.5 1 1.5 2
TIme on Load (hours)
Figure 9-12 Envelope strain as a function of time for a static test, a 300-sec hold time test,
and a O.3-sec hold time test.

and the life becomes much longer. This increase in life is shown as a function
of hold time for different stresses, temperatures, and alloys in Figure 9-13. All
cases show a decrease in damage storage rate by a factor of 5 to 10 as the test
is accelerated from static creep into the short hold time regime.
A similar increase in life with decreased hold time is also observed in the
joint specimens, as shown in Figure 9-14. In this case, life, rather than strain
rate, is plotted because the gauge length of a joint is too short to allow accurate
measurement of strain. Since the life of a joint increases significantly over the
same range of frequencies as the bulk solder, the effect should be examined
further in the bulk solders and applied to joint testing.
The detailed strain per cycle data is shown in Figure 9-15 for a 3-sec hold
time test. Quite a significant strain recovery is observed upon each unloading.
This strain recovery was found to range from 10 to 40 percent of the elastic
strain depending upon the temperature and stress range. This strain recovery
resembles an anelastic effect in that it is a time dependent recovery of strain.
Strictly, an anelastic effect should be linear with stress (Le., the elastic strain). 33
Confirmation or denial of the linearity of this strain storage with the stress requires
more exact control of the stress than was available at the time of testing. For
the lack of a better term, the effect will be referred to as anelastic in this work.
In order to determine whether this anelastic strain recovery or some cyclic
hardening effect is responsible for the decrease in strain rate with decreasing
hold time, tests were run that eliminated either the off-load hold time or the on-
load hold time, as shown in Figures 9-4b and 4c. All three tests had 3-sec hold
 UNDERSTANDING THE DAMAGE MECHANISMS 295

1.0
• 37-63Sn, 27.6 MPa, T=27C •
[] 37-63Sn, 13.8 MPa, T=27C
[]
0.8
CII 0 81-19Sn, 13.8 MPa, T=27C 0
'!ii
a:
• 37-63Sn, 9.0MPa, T=100C
0.6

0.4 •
•
0.2
•
[]
8
0.0
• 0

10-2 10-1 10 0 10 1

Hold Time (sec.)

Figure 9-13 Normalized minimum strain rate (Le., minimum strain rate divided by the
minimum strain rate for the static creep test) as a function of hold time for eutectic and 19%
5n-81 % Pb solders.

times and identical ramp rates. The time on load, which is defined here as the
time spent above the mean stress, is used rather than total test time or number
of cycles to compare the severity of the tests. The envelope strain curves for
these tests are shown in Figure 9-16. The rate of damage storage is much faster

...
CII
6
•
.a 5
•
--
~
0 4
"l!!

-
111::1
00 3
....I&:
C

•
0
CII 2
E
j::
1
.1 10 100 1000

Hold Time (sec)

Figure 9-14 Joint life as a function of cyclic frequency.

296 THE INTERACTION OF CREEP AND FATIGUE IN LEAD-TIN SOLDERS

30~-------------------------,

29
*
§
28

27+-~~-r--~~~~--,---~-i
592 594 596 598 600
Time (sec)

Figure 9-15 Strain as a function of time for a single cycle showing the large strain recovery
during the off-load hold time.

when there is relatively little time spent relaxing off-load. In fact, the on-load
hold time test behaved almost identically to the static creep test. This indicates
that the increased life with decreased hold time is the result of recovered anelastic
strain.
If the decrease in creep rate had been the result of cyclic hardening, the on-
load hold and the balanced hold time tests would have had the same strain rates.

COMPARISON OF WAVEFORMS
R=O 0.167 Hz 'D.6 t.tPa

'35"'
ON LDAD HOLD

30

Z 25
<C
a: 20
I- BALANCED
en
::-!!
0 15

10

" 7
2 3 5 6 8
TIME ON-LOAD (h)

Figure 9-16 Envelope strain as a function of time for different waveforms shown in Figure
9-4. Comparison of waveforms: R = OJ 0.167 Hz; 27.6 MPa.
 UNDERSTANDING THE DAMAGE MECHANISMS 297

Both tests see the same number of cycles for the amount of time they spend on-
load. In the case where no time is spent recovering the anelastic strain during
an off-load hold, the anelastic storage is unable to prevent the storage of non-
recoverable strain during the on-load hold. And so, nonrecoverable strain is
being stored right from the beginning of each on-load hold. The test with only
an off-load hold spends very little time storing time dependent strain and spends
much more time recovering anelastic strain. In this case all of the anelastic strain
that was stored can be recovered during the off-load hold time and therefore all
of the time during the on-load is spent storing recoverable strain.
As permanent creep damage is retarded by increasing the off-load hold time,
the reduction of creep rate with decreasing hold time in the balanced cyclic tests
must be caused by the recovery of the initial portion of the time dependent strain
that is stored during each on-load hold time. Therefore, the creep strain and the
anelastic strain are serial processes during the on-load hold. In other words, the
storage of anelastic strain at the beginning of the on-load hold retards the storage
of creep strain. However, as the anelastic process becomes saturated, creep strain
will be stored. Accordingly, for short hold time tests, almost all of the time
dependent strain that has been stored during the on-load hold is anelastic and
can be recovered during the off-load hold. Conversely, for longer hold time
tests, the reverse will be the case.
For the purpose of developing constitutive equations, the time dependent strain
stored in each cycle may be separated into the recoverable anelastic strain and
the nonrecoverable creep strain. The strain storage and recovery during the
steady-state stage of the creep tests can be analyzed using data from individual
cycles as shown schematically in Figure 9-17. The storage of strain can be
described by

The recovery of strain can be described by

(9-6)

In the above equations eEL is the elastic strain, IOCR is the creep strain, and lOAN
is the anelastic strain. Each cycle then has two time independent strain com-
ponents, one of which is recoverable and the other of which is nonrecoverable;
therefore they can be separated by comparing the strain stored during the on-
load ramp to the strain recovered during the off-load ramp. Likewise, the two
time dependent strain components, anelastic strain and creep strain, can be
separated by comparing the strain stored during the on-load hold time to the
strain recovered during the off-load hold time.
The temperature dependence of the anelastic relaxation is small, and small
drifts in the minimum stress during the off-load hold prevented accurate mea-
298 THE INTERACTION OF CREEP AND FATIGUE IN lEAD-TIN SOlDERS

c: EL
-~
Ci5
AN

Time
Figure 9-17 Schematic of the separation of the different type of strain in a single cycle.
The time independent strains, elastic EL, and plastic PL, are separated by comparing the strain
stored during loading to the strain recovered during unloading. Likewise, the time dependent
strains, creep CR and anelastic AN, are separated by comparing the strain stored during the
on-load hold time to the strain recovered during the off-load hold time.

surement of the relaxation rate. Relaxation times of 1 to 10 sec were measured

using a simple exponential decay model. These were measured over the tem-
perature range of 17 to 42°C. There was no systematic variation in the relaxation
time over this temperature range. Assuming that the change in relaxation time
is hidden in the scatter, the activation energy for the relaxation must be less than
0.73 eV.
Many joint loadings can be described as strain or displacement controlled.
Stress-strain hystersis loops are shown in Figure 9-18 for hold times of 0.3 and
300 sec with a 2000 IJ.E strain range. In all cases, although the minimum stress
started at 0, it became negative as nonrecoverable strain was stored. This shake-
down continued until the mean stress approached zero and, after cracks became
significant, even dropped below zero. At short hold times this relaxation takes
much longer than it does with long hold times. In the 3OO-sec hold time test,
the average stress is relaxed to close to 0 in the first cycle and relaxes little more
over the next 9000 sec. In the 0.3-sec hold time test, even after 9000 sec very
little relaxation has occurred by comparison. The rate of relaxation is plotted as
a function of time for three hold times in Figure 9-19.
A problem with strain controlled testing is in defining an easily measurable
rate of damage accumulation before cracks become measurable. The rate of
stress relaxation measures the storage of unrecovered plastic strain, but it is
unclear as to whether strain storage of this sort is damaging in terms of the life
of the specimen. The strain stored in the specimen during the on- or off-load
holds of the strain controlled test can be compared to the stress controlled test,
equations (9-5) and (9-6).
 UNDERSTANDING THE DAMAGE MECHANISMS 299

~
300sec

20

'C'
a..
6
III a
~
~
-20 o SEC
- --
9285 SEC_

~
a 500 1000 1500 2000 2500
MICROSTRAlN

20

'C'
a..
6
III a
~
~
-20

~~----~----r---~----~-----1
a 500 1000 1500 2000 2500
MICROSTRAIN
Figure 9-18 Stress-strain hysteresis for a strain range of 2000 jLE comparing the first cycle
of a test to a cycle after approximately 9000 sec of testing for 300-sec and 0.3-sec hold times.

The anelastic and creep strains cannot be separated in this case, and so there is
no way to determine at this time how much time dependent damage has been
stored in the material and how much of the strain was benign. But if hold times
are long enough, the anelastic portion of the strain will be insignificant, and a
simple comparison of the hysteresis areas between an accelerated cycle and a
real cycle could give a reasonable acceleration factor. In equation (9-7), Eo is
the total strain range.
 40,---------------------------------~

30
'0'
a..
-5

-
III
III
~
V ')
20 Frequency
0.00167 Hz
0.167 Hz

10~~~~~M_~~~~~~~~~~~~
0.001 0.01 0.1 1 10 100 1000 10000
Time On-Load (sec)

Figure 9-19 Maximum stress in a cycle as a function of time comparing different hold time
tests all with strain ranges of 2000 j..LE.

50~--------------------------------~

40

'0'
a.. 30
6
V')

~
~ 20
l-
V')

10

O~------~--------~------~------~
o 5 10 15 20
MICROSTRAIN (Thousands)
Figure 9-20 Stress as a function of strain for a tensile test at 10-1 sec 1 • Note the large
amount of plastic strain stored at low stresses.

300
 INTERPRETATION FOR PACKAGING APPLICATIONS 301

The effect of decreasing the hold time is the increase in the average stress in
a half cycle. This is a natural consequence of allowing less time for relaxation
in each cycle, both recoverable and nonrecoverable. At the shortest hold times,
the time dependent strain that is being stored during the hold should not be
storing any damage since it is all recovered during the hold of opposite sign by
the anelastic mechanism. But the shorter hold times also have higher ramp rates
that disallow the storage of time dependent strain during the ramp (these tests
were performed with a constant wave shape) and thus cause some greater amount
of athermal plastic flow to occur during the ramp. The storage of athermal plastic
strain occurs at stresses much lower than the 0.2 percent offset yield point even
at high strain rates, as can be seen in Figure 9-20.

9.5 INTERPRETATION FOR PACKAGING APPLICATIONS

9.5.1 Deformation

Creep deformation and anelastic recovery have been shown to be the dominant
deformation and recovery mechanisms through whose interaction damage is
stored near and above ambient temperature. As mentioned, the anelastic defor-
mation acts as the initial time dependent deformation during an on-load and acts
to recover the last bit of time dependent deformation that was stored prior to an
off-load. The relaxation time for the anelastic mechanism is less than 10 sec
over the range of tempertures tested, - 1 to 100°C. Whenever the hold time of
a cycle is on the order of or less than the relaxation time, the anelastic strain
acts to prevent the creep deformation from creating damage in the solder. Also,
the parallel nature of the creep and anelastic strains implies that they are me-
chanistically one type of event.
The eutectic and the 19% Sn-81% Pb alloys show nearly identical behavior
with respect to accelerating the cyclic test, as was shown in Figure 9-13. This
is despite the fact that the eutectic alloy has a matrix of Sn with lamellar Pb
phase and the 19% Sn-81 % Pb has a Pb matrix with Sn precipitates. The similarity
of the behavior implies that the anelastic effect is independent of the matrix
phase and matrix grain boundaries. Rather, it is probably the result of the in-
teraction between the Pb and the Sn phases, which must persist through the
change in microstructural morphology from one alloy to the other. This is con-
sistent with the large anelastic effects that have been identified in the superplastic
microstructure of eutectic Pb_Sn34 ,35 that is a random mixture of equiaxed Sn
and Pb phase grains. The anelastic and creep deformation behavior can be
mechanically modeled as two nonlinear viscoelastic phases, Pb and Sn, connected
by a viscoelastic phase boundary, as shown in Figure 9-21. This model of the
strain in Equation (9-5) is the beginning of constitutive equations that will be
life predictive.
302 THE INTERACTION OF CREEP AND FATIGUE IN LEAD-TIN SOLDERS

modulus
elastic
of Pb
modulus
of Sn

creep rate
creep rate of Pb
of Sn

boundary
boundary viscosity modulus

Figure 9-21 A mechanical model that can account for solder's deformation behavior.

9.5.2 Thermomechanical Test Guidance

In order to rank designs, an accelerated test must store damage by the same
mechanisms that a realistic thermal cycle stores damage. Also, in order to predict
component life, the damage stored during an accelerated test cycle must be
related to the amount of damage stored in a realistic cycle.
The effects of hold time and ramp rate on the amount and types of damage
stored during the cycle are illustrated in Figure 9-1_ In either the stress or strain
 CONCLUDING REMARKS 303

controlled case, the ramp rate controls the relative amounts of plastic and time
dependent strain. Faster ramp rates cause the accumulation of more plastic dam-
age and less anelastic strain and creep damage. Slower ramp rates minimize the
storage of athermal plastic strain. If parts of the thermal cycle are in a temperature
region where little or no time dependent damage is stored, acceleration of this
part of the cycle requires no modification of damage stored per cycle. Short hold
times of accelerated tests decrease the amount of relaxation and thereby decrease
the amount of damage stored, as shown in Figure 9-1. This change in damage
storage per cycle can be accounted for by considering the change in the energy
stored per cycle. But, if hold times are decreased below, say, around 3 to 10
sec, the storage and recovery of anelastic strain will retard the storage of creep
damage. The energy of the hysteresis caused by anelastic strain storage and
recovery is not damaging to the solder and so the calculation of a damage storage
rate from the hysteresis becomes meaningless. Hold times should be designed
long enough to cause the storage of creep damage to be dominant (Le., to make
the anelastic strain that is stored insignificant compared to the creep strain). By
this analysis, the practical effect of anelastic strain recovery is to define a limit
to the acceleration of a thermomechanical test such that hold times should be
significantly larger than the anelastic relaxation time of 3 to 10 sec.

9.6 CONCLUDING REMARKS

Accelerated thermomechanical test cycles and life or residual life predictive

equations should be carefully designed with the consideration of the type of
damage being stored at each temperature. At ambient temperature and above,
the damage will be controlled by cyclic creep. Acceleration of the ramping rate
can certainly be made in parts of the cycle where athermal damage is dominant.
Acceleration at high homologous temperatures should take into account the time
dependent nature of the creep damage storage and the anelastic recovery phe-
nomena. But a thermomechanical test should be accelerated only with caution
to the point where the anelastic storage and recovery of strain prevents the storage
of creep damage. Around ambient temperature, hold times should be kept above
3 to 10 sec in order to keep the anelastic effect from becoming a significant part
of the strain storage.
Further study should be aimed at the modes of deformation and fracture at
higher temperatures as a function of acceleration of the cycle. As temperature
increases, the frequency at which anelastic deformation suppresses creep damage
may change enough to allow more acceleration of the test than the limit advised
here of 3- to 10-sec hold times. The effects of acceleration on higher Pb content
solders (e.g., 90% Pb-lO% Sn, 95% Pb-5% Sn, or 98% Pb-2% Sn) should also
be tested. As the amount of Sn phase decreases, the anelastic relaxation should
become less pronounced, and it may be practical to accelerate tests much more
304 THE INTERACTION OF CREEP AND FATIGUE IN LEAD-TIN SOlDERS

in these cases. Acceleration of the strain controlled tests showed a clear depen-
dence of the relaxation rate on hold time, but further work needs to be carried
out on the effects of acceleration on life in strain controlled tests.
Small volume specimens should be developed and examined in order to un-
derstand how the details of solder behavior may change in going from bulk solder
to micrometer dimensioned solder joints. These details could affect the limit of
acceleration of a cycle.

REFERENCES

1. Engelmaier, W., "Surface Mount Solder Joint Long-Term Reliability: Design, Test-
ing, Prediction," Soldering and Surface Mount Technology, 1(1) Feb. 1989, p. 14.
2. "Functional Cycles and Surface Mounting Attachment Reliability," Circuit World,
J. Inst. Cir. Techn., 11(3), 1985.
3. Solomon, H. D., "Low-Frequency, High-Temperature, Low-Cycle Fatigue of 60/40
Solder," Brazing and Soldering, No. 11 Aug. 1986, pp. 68-75.
4. Vaynman, S., M.E. Fine, and D. A. Jeanotte, "Isothermal Fatigue of Low Sn Pb
Based Solders," in Load and Thermal Histories on Mechanical Behavior ofMaterials,
P. K. Liaw, Ed. Warrendale, PA, TMS, 1987.
5. Fox, L. R., J. W. Sofia, and M. C. Shine, "Investigation of Solder Fatigue Accel-
eration Factors," IEEE Transactions on Components, Hybrids, and Manufac. Tech.,
CHMT-8(2), June 1988.
6. Sinclair, N. A., "Thermal Cycle Fatigue Life of LSI Solder Interconnections," Proc.
Inti. Electrical and Electronics Conf. and Expo, Toronto, 1982, pp. 56-57.
7. Manson, S. S., Thermal Stress and Low Cycle Fatigue, New York, McGraw-Hill,
1966.
8. Weertman, J., J. Appl. Phys., 26, 1955, pp. 1213-1217.
9. Friedel, J., Dislocations, Ch. 11, London and Oxford, Pergamon Press, 1967.
10. Frost, H. J., and M. F. Ashby, in Fundamental Aspects of Structural Alloy Design,
R. I. Jaffe and B. E. Wilcox, Eds., New York and London, Plenum Press, 1977,
pp.27-58.
11. Herring, c., "Diffusional Viscosity of a Polycrystalline Solid," J. Appl. Phys., 21,
1950, p. 437.
12. Coble, R. L., "A Model for Boundary Diffusion Controlled Creep in Polycrystalline
Materials," J. Appl. Phys., 34, .(6), 1963, p. 1679.
13. Ashby, M. F., and R. A. Verrall, "Diffusion Accommodated How and Super plas-
ticity," Acta Met., 21, 1973, p. 149.
14. Sherby, O. D., R. D. Caligiuri, E. S. Kayali, and R. A. White, "Fundamentals of
Superplasticity and Its Application," in Advances in Metal Processing, J. J. Burke,
R. McHrabian, and V. Weiss, Eds., New York, Plenum Press, 1978.
15. Goods, S. H., and L. M. Brown, "The Nucleation of Cavities by Plastic Defor-
mation," Acta Met., 27, 1979, pp. 1-15.
16. Riedel, H., Fracture at High Temperatures, Chs. 6-9, New York, Springer-Verlag,
1987.
 REFERENCES 305

17. Burgess, J. F., R. O. Carlson, H. H. Glascock, C. A. Neugebauer, and H. F.

Webster, "Solder Fatigue Problems in Power Packages," IEEE Transactions on
Components, Hybrids, and Manufac. Tech. CHMT-7(4), 1984, pp. 405-410.
18. Honke, D., private communication.
19. Zubelewicz, A., R. Berriche, L. M. Keer, and M. E. Fine, "Lifetime Prediction of
Solder Materials," presented at the ASME Annual Meeting, Chicago, IL. Dec. 1988.
20. Frear, D., D. Grivas, M. McCormack, D. Tribula, and J. W. Morris, "Fatigue and
Thermal Fatigue of Pb-Sn Solder Joints," ASM Conf. Electronic Packaging, 1987.
21. Frear, D. R., D. Grivas, and J. W. Morris, "Thermal Fatigue in Solder Joints," J.
of Metals, June 1988, pp. 18-22.
22. O'Clock, G. D., M. S. Peters, J. R. Pater, G. A. Kleese, and R. V. Martini, "Pb-
Sn Alloy Microstructure: Potential Reliability Indicator for Interconnects," IEEE
Transactions on Components, Hybrids, and Manufac. Tech., CHMT-IO(1), 1987,
pp.82-88.
23. Greenwood, P. J., T. C. Reiley, V. Raman, and J. K. Tien, Scripta Met, 22, 1988,
pp. 1465-1468.
24. Hendrix, B. C., Ph.D. Thesis, Columbia University, 1989.
25. Hendrix, B. C., A. I. Attarwala, and J. K. Tien, unpublished work.
26. Weinbel, R. C., J. K. Tien, R. A. Pollak, and S. K. Kang, J. Mat Sci, 22, 1987,
pp. 3901-3906.
27. Tien, J. K., B. C. Hendrix, and A.1. Attarwala,lEEE Transactions on Components,
Hybrids, and Manufac. Tech. CHMT-12, 1989, pp. 502-505.
28. Hendrix, B. C., J. K. Tien, S. K. Kang, and T. Reiley, in Effects of Load and
Thermal Histories, P. K. Liaw, Ed. Warrendale, PA. TMS, 1987, p. 139.
29. Weinbel, R. C., E. A. Schwarzkopf, and J. K. Tien, Scripta Met., 21, 1987, pp.
1165-1168.
30. Oberschmidt, J., K. K. Kim, and D. Gupta, J. Appl. Phys., 53, 1982, p. 5672.
31. Askill, J., Tracer Diffusion Datafor Metals, Alloys, and Simple Oxides, New York,
Plenum Publishing Corporation, 1970, pp. 35-39.
32. Cagnon, M., M. Suery, A. Eberhardt, and B. Baudelet, "High Temperature Defor-
mation of the Pb-Sn Eutectic," Acta Met., 25, 177, pp. 71-75.
33. Nowick, A. S., and B. S. Berry, Anelastic Relaxation in Crystalline Solids,
New York, Academic Press, 1972.
34. Homer, C., and B. Baude1et, "Anelastic Behavior of the Pb Sn Eutectic Alloy,"
Scripta Met., 11, 1977, pp. 185-189.
35. Schneibel, J. H., and P. M . Hazzledine, ''The Influence of Anelasticity on the Stress
Relaxation Behavior of Superplastic Pb-Sn Eutectic," Scripta Met., 11, 1977, pp.
953-956.
 10
Creep and Stress Relaxation in
Solder Joints
Peter M. Hall

Stress-strain analysis of solder joints is a wonderful field. Stress and strain both
have six components, and there are thirty-six elastic "constants" (which of course
are not constant at all). All these depend on the position in the joint, the com-
position, purity, history, temperature, and geometrical configuration of the sol-
der, as well as the surrounding structure and its properties and temperature
distribution. In addition, they are functions of time, with time "constants" ranging
from milliseconds to centuries. All together, there are enough variables to keep
the journals supplied with papers for years to come. (A bibliography compiled
a few years ago l already contained some 372 references.) This provides oppor-
tunities for us to exercise our skills in materials science, mechanical engineering,
physics, chemistry, and metallurgy. The game of course, is to focus on one or
two independent variables and one or two dependent variables and study their
interaction. The hope is that the others can be kept more or less under control
or that they don't matter too much, and if we don't oversimplify something
important, perhaps some insights can be gained. The problem is that the reliability
of these joints is a deadly serious matter. They are created by the billions, and
all kinds of important (and unimportant) devices rely on their performance.
This chapter is a summary of some thinking and some experiments that we
have done over the last few years from a fundamental point of view. It is hoped
that the insights gained nay be helpful to those who design, test, and specify
these joints in production and those who use them in the field. The focus here
is on surface mount leadless solder joints and on the stresses caused by thermal
expansion mismatches.
The objective of this study is to provide some information for understanding
and interpreting thermocyclic testing of components involving solder joints. It

306
 IDEAL EXPANSIVITY OF A SUBSTRATE 307

is well known that solder creep and stress relaxation are important in such cycling,
but infonnation on these processes is difficult to come by, so it is usually ignored,
by assuming either that there is no relaxation or that there is complete relaxation.
Depending on the temperature range, either assumption can be quite misleading.
It is important to understand the stress relaxation kinetics during any temperature
cycle and especially during the "hold times."
For years we have complained to makers of printed wiring boards (PWBs)
and other substrates that most of our problems arise because they provide us
with substrates that don't have the optimum coefficient of linear thennal expan-
sion. (This is often acronymized as CTE, or as some prefer, TCE. My own
preference is for the tenn "expansivity," or the symbol IX.) Indeed, the number
of thennal cycles to failure is a strong function of the expansivity of the substrate
(e.g., Sherry and HalJ,2 Fig. 17). Recently, however, the substrate makers have
called our bluff and offered to provide us with whatever expansivity we feel is
optimum. This has forced us to reconsider the question, "Given a surface mount
component with a given expansivity, what is the ideal expansivity for the sub-
strate?" In the next section, we deal with this question, as always, by simplifying
it down to its fundamentals.

10.1 IDEAL EXPANSIVITY OF A SUBSTRATE

We treat this subject in various degrees of sophistication, looking in turn at six

of the effects that may be involved. In each case, we consider how much strain
would be imposed on the solder if it were completely compliant or fully relaxed
and how we could adjust the expansivity of the substrate to minimize that strain.

10.1.1 No Temperature Gradients, No Transients

Given a component surface mount soldered to a substrate, let its temperature be

changed from T to To as in Figure IO-la. Let there be neither temperature
gradients nor transients, and assume that the solder is fully compliant, or relaxes
fully. The temperature change will then result in an overall shear strain in the
joint of

. (10-1)

where 'Y is the shear strain, 2L is the in-plane dimension of the component to
be surface soldered. (In a chip carrier, we take 2L as the diagonal, to provide
a worst case.) H is the height of the solder, T is the temperature, To is the
temperature at which the shear strain is zero, IXc is the (given) expansivity of
 SOLDER~
r-ac
2L ---1
T I IH
77 SS
(a) ~ as T \f

I
II
ac 1". I
Ii
(b) ~ as T. ~

(c)

(d) STo To(

(f)
Figure 10-1 Idealized thermal strains. (a) Negligible temperature gradients. (b) Component
hotter than substrate. (c) Through-board temperature gradient. (d) In-plane temperature gra-
dient. (e) Thermal shock. (f) Solder-substrate mismatch.

308
 IDEAL EXPANSIVITY OF A SUBSTRATE 309

the component, and as is the expansivity of the substrate. Many authors have
quoted this equation. Here we have ignored nonuniformity of strain within the
joint as well as from one joint to the other (for a multijoint component). Never-
theless, for this case it makes good sense to have the expansivity of the substrate
equal to that of the component so as to minimize the shear (strain). So the answer
to our question for this case is simply as(ideal) = ac.

10.1.2 Power Dissipation in the Component

If the component is dissipating heat, there will be temperature gradients in the

system. A first order analysis assumes that the component is at a uniform tem-
perature Tc , whereas the substrate is at some other uniform temperature, Ts , and
generally, Tc > Ts (Fig. 1O-1b). Again, we let the solder relax completely, and
here we find that the shear strain is

(10-2)

In this case, to minimize the shear strain, we would set the substrate expansivity
equal to

(10-3)

Normally, Tc > Ts so we now conclude that the ideal coefficient of the substrate
is greater than that of the component. In one case in point, for a ceramic chip
carrier on a PWB, the ideal as was 1.6 times that of the ceramic chip carrier.
This concept is often referred to as "tailoring" the expansivity, and it is well
documented. 3.4 It was received with enthusiasm among many substrate engineers,
partly because a factor of 1.6 applied to a typical ceramic expansivity (6 ppm!
K) results in 9.6 ppm!K, which is readily achieved by adding certain metal
inserts into the PWB .5 It was soon realized, however, that not all the components
on a substrate would have the same power or temperature, so the best one can
do is optimize the expansivity for one of the components on the substrate. 4
Already, our answer to the question posed is becoming complex. But there are
other considerations.

10.1.3 Z-Gradients in the Substrate

If the ceramic chip carrier is dissipating power, as above, there will be thermal
gradients in the substrate normal to the plane of the substrate (Fig. lO-lc). Figure
310 CREEP AND STRESS RELAXATION IN SOLDER JOINTS

10-2 shows results of some thermocouple measurements of the temperature rise

above ambient due to 0.75 W dissipated in a 0.65- by 0.65-in. (1.65- by 1.65-
cm) ceramic chip carrier. The average temperature rise of the chip carrier was
24 K, and the average rise of the PWB in the vicinity was 15 K. There was a
temperature difference of 10 K through the thickness of the board (under the
center of the carrier). Waller, Fox, and Hannemann6 used a finite element analysis
on a similar geometry and also showed that there is a considerable temperature
drop across the board. This gradient causes the board to bow, as we have seen, 7
and remain bowed even after the solder has completely relaxed. The result is
that the solder joints along anyone side of the carrier will have a distribution
of tension and compression strains. Again we simplify the problem, assuming
the bowing will result in a spherical deformation of the substrate. Let the de-
flection of the substrate at the comers relative to that in the center of one side
be 8. If we assume that the algebraic sum of the deformations is zero, the comer
joints will be stretched by %8, whereas the joint in the center of a side will be
compressed by 1/38. The deflection difference is

8 = L2(TsT - TSB)«s
(10-4)
4hs

where hs is the thickness of the substrate and TST is the temperature of the top
of the substrate and TSB is the temperature at the bottom of the substrate. So the
solder tensile strain in the comer (worst case) joint is

(10-5)

Thus to minimize the tensile strain, we should have the substrate expansivity
equal to zero. We can't do this without increasing the shear strain gi~en by
Equation 10-2, so some engineering compromise must be made. One way would
be to minimize the sum of the squares of e and 'Y. We assume that Ts from
Equation 10-2 can be approximated by 1/2(TsT + TSB). This results in the
following "ideal" expansivity:

(10-6)

This will produce a result between those of minimizing the shear and minimizing
the tension. In our case, the numerical result was that

«s(ideal) = 0.88ac (10-7)

 IDEAL EXPANSIVITY OF A SUBSTRATE 311

Now we find the ideal substrate expansivity is less than that of the component,
an unexpected result.

10.1.4In-P/ane Gradients

In addition to through-board gradients, for the case of powered components there

will be in-plane gradients (Fig. 10-2). If there is a hot spot under each carrier,
there will be buckling of the substrate in those regions (Fig. 1O-1d), as shown
by Howard, Sobeck, and Sanetra. 8 The substrate might conceivably buckle in
either direction if this were the only effect, but as we already have seen, the
through-board gradient will cause convex bowing, so the buckling caused by
the in-plane gradient will add to it. We don't have a simple formula for this
buckling, but clearly it will t;>e proportional to the expansivity of the substrate,
so it causes us to lower the value of our ideal expansivity even more. We have,
then, for our particular case,

as(ideal) < O.SSac (10-S)

Depending on the circumstances, this mayor may not be a large effect, but the
direction of the effect is clear.

10.1.5 Temperature Shock

Another environment of interest is that of temperature shock. If the temperature

is changed so rapidly that the inner parts of the system lag well behind the outer

ABOVE
29K
AMBIENT

26K

22K

12K

DATA FROM R.A. DEIGHAN III

Figure 10·2 Temperature distribution with 0.75 W-dissipated.

312 CREEP AND STRESS RELAXATION IN SOLDER JOINTS

parts, we will find opposed bowing, as suggested by Engelmaier9 and by Figure

lO-le. Here both the chip carrier and the substrate have the temperatures of their
"exposed" surfaces changed, but their "interior" surfaces remain at the original
temperature. They will both bow but in opposite directions. The figure is for a
positive temperature change and positive expansivities. As usual, it is highly
simplified. We can obtain an idea of the strains involved by assuming a linear
gradient through each of the materials and setting the sum of the solder joint
deflections equal to zero. Here we are assuming that the solder relaxes faster
than the thermal time constants of the system, which is normally not the case.
Instead of the strains calculated here, there would be stresses built up. The
general conclusion, however, remains the same:

_ U (Te - To + Ts - To )~ (10-9)
E - 2H he ac hs as 3

where he is the thickness of the component (e.g., a chip carrier). To minimize

the tension due to this effect we would have to set the substrate expansivity to
less than zero. Specifically,

as(ideal) (10-10)

where he and hs are the thicknesses of the component and the substrate, re-
spectively. For the particular case we are interested in, the factor in brackets
was 1.5. Thus, if we are concerned about stresses induced by thermal shock,
we should lower our overall value of ideal substrate expansivity even farther
than indicated by Equation 10-8. The amount by which it should be lowered
depends on the perceived hazard of temperature shock relative to that of power
cycling for field conditions. As Engelmaier9 has pointed out, temperature shock
is more commonly encountered in testing than in the field.

10.1.6 Solder-Substrate fxpansivity Mismatch

Another effect that should be considered here is the expansivity mismatch be-
tween the solder and the substrate (Fig. 1O-1f). This mismatch has received little
attention until recently, 10 no doubt because the relevant dimension of a solder
joint is normally much less than that of the component (L). In terms of free
bodies, if the solder were not connected to the pads, the in-plane displacement
due to the expansivity mismatch between the component and the substrate would
be given by L(as - ad. On the other hand, if the substrate and component
expansivities are equal, and again the solder is not connected, the in-plane
 CREEP AND STRESS RELAXATION 313

displacement due to the solder-substrate mismatch is given by D(as - aSOLDER),

where D is the half-diagonal of the solder pad. The expansivity of solders is in
the range of 25 ppmlK,11 so there can be some large mismatches at the joints
even if the component and the substrate are perfectly matched (or even if there
is no component, only a "blob" of solder). To minimize this effect, of course,
the substrate expansivity should match that of the solder. This would normally
be impractical, however. Even if it were possible, there would still be the
mismatch of the solder at the component side of the joint. So optimizing the
substrate expansivity would only shift the failure to the other side of the joint.
Still, when solder joints are made to a low expansivity material like ceramic or
silicon, large local mismatches can and do occur, and they can and do cause
cracking, especially for large area joints, both during thermal stresses of soldering
and temperature cycling after the joints are made.

10.1.7 Overall Judgment

Thus we find there is no one answer to the question posed by the substrate
manufacturers. We are reduced to making an engineering compromise, based
on our best judgment. My own judgment, based on the above concepts, is that
the ideal substrate expansivity should be less than that of the component. If, for
example, the component has an expansivity of 6 ppmlK, an ideal substrate
expansivity might be in the range 2 to 4 ppmlK. It seems well to remember that
for the world to be ideal, it is insufficient for all expansivities to be equal. The
ideal world would have them all zero. And most of the time, we improve matters
by moving toward this ideal, that is by lowering expansivities.

10.2 CREEP AND STRESS RELAXATION

Solder at room temperature and above is a ductile material. A wire made of

60% Sn-40% Pb can be stretched at room temperature by more than 60 percent
before breaking (reference 12, pp. 9-71). Thus when stresses are applied to
solder joints, they yield, and they do so in such a way as to reduce the stresses.
When a sample's strain is held constant and the stress varies (decreases), we
call the process "stress relaxation." When the stress is held constant and the
strain changes in time, we call it "creep." The distinction between the two has
been useful in interpreting stress-strain data where one of the variables is con-
trollable. In both cases there is plastic flow, and microstructurally, the same
things happen. Curves relating plastic strain rate to stress normally give identical
results for creep and stress relaxation experiments. 13 In the field, solder joints
experience time-dependent changes in the strains and the stresses simultaneously.
314 CREEP AND STRESS RELAXATION IN SOLDER JOINTS

We have called the combined process "stress reduction,"14 but a better term
might be "creep-relaxation. "*
When thermal stresses are set up in a soldered assembly, the members applying
the stresses tei the solder (i.e., the components and the substrate) are normally
stressed within their elastic range. Thus, we can speak: of a "stiffness" (k) of
the assembly, which is the derivative of force with respect to displacement, as
in a spring constant.

dF
k= (10-11)
dx

So if the solder yields by a distance ax, the force on the joint is reduced by an
amount kax, where k is the stiffness of the assembly. There are six components
of stress and six of strain, so in general there are thirty-six stiffness constants
involved. In the following sections, we make the usual three assumptions (ref-
erences 15 and 16, p. 282): (1) The main effects are shear and bending in the
solder joint, (2) the configuration has axial symmetry, and (3) only radial forces
and tangential moments are applied to the solder (Fig. 10-3).
Let k be the stiffness of the assembly for these shear stresses. For constant
temperature we have ~..y = ax/H, or

-M
~:y =--
kH
for aT = 0 (10-12)

where F is the force per joint. We also know from Equation 10-1 that if F =
o (complete relaxation),

for F = 0, M = 0 (10-13)

(provided we have no temperature gradients). We can combine these two equa-

tions to obtain

(10-14)

This equation is important despite its simplicity since it relates the shear strain,
the shear stress, and the temperature in a linear equation. It simplifies the analysis
considerably, provided the stiffness k is not a function of temperature, which

*This author at one time proposed coining the term "creelaxation." Fortunately AT&T
management talked me out of it.
 CREEP AND STRESS RELAXATION 315

~ COMPONENT

...
Me
~

f
F

H SOLDER

L F _
I <..:.J I
I Ms I

~ SUBSTRATE (
Figure 10-3 Idealized surface mount assembly.

has been shown to be a reasonable approximation for our case, at least. 14 For
instance, the equation tells us that if the temperature is held constant, the force
will be linearly related to the shear strain, with slope - kH. This is the condition
of creep-relaxation mentioned above. If k is large (stiff), we have conditions
approximating stress relaxation. If k is small, we are closer to creep. In any real
case, we have simultaneous time dependent changes in stress and strain.
Another way to look at this equation is to realize that when the temperature
is changed, the expansion mismatch (the first term on the right) causes a change
in the force (the second term on the right) or a change in the shear strain (the
term on the left) or some combination, depending on the properties of the solder.
If the solder is highly compliant, the mismatch causes strain, as at high tem-
peratures. If the solder is stiff, the mismatch causes mostly stresses, as at low
temperatures. Equation 10-14 shows how the stresses and the strains together
make up the thermal mismatch.
316 CREEP AND STRESS RELAXATION IN SOlDER JOINTS

This equation is most useful in combination with strain measurements. Using

strain gauges, we can measure the shear strain fl:y. 15

(10-15)

where EST and ESB are the strains on the top and bottom of the substrate, and
Ecr and ECB are the strains on the top and bottom of the component. And if we
know the elastic constants of the substrate and component, we can obtain a
measure of the shear force by measuring changes in their bending. Their radii
of bending are given by

!l(~)e =
!l( ECB - Ecr)
he (10-16)

and

!l(~) s
!l(ESB - EST)
hs (10-17)

The change in the shear force per joint llF is then given by Hall,15

(10-18)

where Ee
and Es are the elastic moduli of the component and substrate, and Ve
and Vs P
are their Poisson's ratios. The quantity is the "pitch," or the center-to-
center spacing of the joints around the perimeter of the component. The reason
that the component and the substrate contributions appear differently in the above
equation is that the substrate extends over a larger region than just under the
component, so additional force is required to deform the portion of the substrate
that is not under the component.
This equation explicitly provides for the component and substrate to have
different radii of curvature, so it does not depend on the "bimetallic strip" model.
From strain gauge measurements and using Equations 10-15 and 10-18, we can
follow the shear force F and shear strain 'Y for any given temperature cycle as
a function of time. Of course, we have to know the temperature of each gauge
to correct for the apparent strain, so these equations are most useful for the
special case of negligible temperature gradients. One simple thing that can be
done is to measure the effective stiffness k by observing the change of shear
force as the temperature is changed under constant shear strain. This can be
 CREEP AND STRESS RELAXATION 317

readily done at lower temperatures, where the shear strain rate is negligible, and
Equation 10-14 reduces to

dF
ar = kL(as - ad at constant 'Y (10-19)

From this equation it is easy to determine the stiffness k. For our case,15 over
the range - 25 to + 4Q°C, the parameter k was remarkably constant at 1.7 lbl
mil, or 300 N/mm.
Figure 10-4 is a schematic plot of shear force F as a function of shear strain
'Y for a given constant temperature. If we start at a given value of force Fo, as
the solder yields the shear strain will increase, and the force will decrease along
the straight line shown. In principle, this will continue spontaneously until the
force reaches the elastic limit at the temperature in question. At high enough
temperatures, the elastic limit is small, and the force can be considered to go to
zero at some final shear value 'Yf. Similarly, if the force starts at a negative
value, Fl at the same temperature, the force will move toward zero, and the
strain will decrease, to arrive ultimately at 'Yf. So far, this analysis is independent
of the properties of the solder, except for the assumption that it creeps-relaxes

FO

SHEAR
FORCE

--~--------~-------SH~E~A~R~--~1
STRAIN

Figure 10-4 Creep-relaxation at constant temperature.

318 CREEP AND STRESS RELAXATION IN SOLDER JOINTS

to its end point. The slope and location of the line depend on the component,
the substrate, and the configuration. The solder properties only control how fast
the system moves along the straight line shown. We deal with this aspect next.

10.3 SOLDER PROPERTIES

The mechanical properties of ductile materials are often described in terms of

elastic and plastic (i.e., nonelastic) strain. Any change in strain is assumed to
be made up of these two components. Thus,

(10-20)

Note that when we hold 'Y constant (stress relaxation), the two parts are equal
in magnitude but opposite in sign. The elastic part is assumed to be reversible
and linear in the force. For shear, this is expressed as

(10-21)

where G is the shear modulus of the material and A is the area of the joint,
usually taken as the area of the pad. Here we have assumed that the shear stress
is uniform within the joint, which is certainly not the case. 17,18 We can, however,
speak of some effective or average shear modulus, which makes this equation
useful. Here we note that if we hold the force (or the stress) constant (creep
conditions), the elastic strain is constant, and the change in plastic strain equals
the change in total strain. In creep-relaxation conditions, we see from Equations
10-12 and 10-21 that both the total strain and the elastic strain are linearly related
to the force, so the plastic strain must be also. Combining Equations 10-20 and
10-21 we can eliminate il.'Ye and find that

(10-22)

We then substitute this into Equation 10-14 and solve for il.'Yp

(10-23)

With the strain gauges, we can measure il.'Y and 1l.F, so if we also measure il.T
(with thermocouples, for example), we are in a position to measure the plastic
shear strain il.'Yp and the shear force Il.F as a function of time and temperature.
The first term in parentheses in Equation 10-23 gives us an idea of how large
the stiffness must be to approach stress relaxation and how small to approach
 CONSTITUTIVE RELATIONS 319

creep conditions. When kH ~ GA, for constant temperature, the plastic shear
strain change is approximately equal to the total shear strain change, which is
the condition we know as creep. When kH ~ GA, (again for constant temperature)
the plastic shear strain change is much greater than the total shear strain change.
This seems anomalous until it is remembered that in creep-relaxation, the change
in elastic strain always has the opposite sign of the change in plastic strain. Thus,
in this condition, the plastic and elastic shear strain changes are approximately
equal and in opposite directions, which is the condition of stress relaxation. The
value of G for eutectic Sn-Pb solder is a function of temperature, 12 ranging from
0.78 Mlb/in. 2 at - 28°C to 0.43 Mlb/in. 2 at 97°C. (In metric units, this is 5.4
and 3.0 GPa). In using these values, we are near the middle of a wide range of
reported results for the elastic modulus of eutectic Sn-Pb solder, ranging from
0.8 Mlb/in. 2 (5.7 GPa)19 to 4.5 Mlb/in. 2 (31 GPa) (reference 11, p. 345) both
at room temperature. For room temperature, we use 2 Mlb/in. 2 (14 GPa) for the
elastic modulus. The shear modulus is then calculated from the standard relation
0.5E/(1 + v) where E is the elastic modulus and v is Poisson's ratio. Thus, for
our conditions, kH/GA varied from 0.11 to 0.18 over the temperature range of
the experiments. This is not small enough to be negligible, so we applied the
term as given in Equation 10-23 when processing data. If the "true" value of
the modulus is higher than we have assumed, the correction is even smaller than
0.18. If it is actually a factor of 2 lower than we have assumed, our values of
d'Ypfdt should be increased by 10 to 15 percent.

10.4 CONSTITUTIVE RELATIONS

The rate of plastic shear strain change is a function of the shear force and the
temperature. This relation is often expressed for ductile materials in terms of a
power-law constitutive relation. 20

d'Yp = b(~)N (10-24)

dt A

where t is the time, and band N are functions of temperature but (so it is
assumed) independent of time. Now as we have seen, we can measure fl'Yp and
!1F, so within a constant, we can measure F and thus evaluate the above equation.
This is especially easy when we hold temperature constant. We have done this
for several different temperatures, by changing the temperature of the sample to
a given temperature. We then hold it there while we observe the force as it
decays and the strain as it changes. 14 Figure 10-5 is a log-log plot of the force
as a function of time for several temperatures. These curves are similar to the
stress relaxation curves of Baker. 21 We cannot readily control the force at which
the creep-relaxation starts. We simply change the temperature, accept whatever
320 CREEP AND STRESS RELAXATION IN SOLDER JOINTS

ONE ONE ONE ONE

HOUR DAY WEEK MONTH

rn 0 j 1IIIIIill!ulilliiligllilillliiiMi
IlL
0.5 2.0

1.0
~
0.2

--
In
a.'I
....I
0.5 In
w 0.1 Z
U
a: o
I-
0 XX~
LL.
0.05 ?ex ==
W
Z
0.2
UJ
u
a:
0.1 oLL.
0.02

0.05
0.01
10 2 10 4 10 5
TIME (SECONDS)

Figure 10-5 Creep-relaxation in solder joints on a chip carrier.

is there, and watch it creep-relax at constant temperature. It is clear, however,

that higher temperatures generally result in smaller forces. That is expected
because creep-relaxation takes place during the time it takes to increase the
temperature. Previously, we derived an expression for the force as a function
of time, 14 assuming a power-law constitutive relation. It predicted that on a log-
log plot, the force per joint would asymptotically approach a straight line at long
times and that its slope would be - lI(N - 1). The curves in Figure 10-5 are
consistent with that relation. We did not see any leveling off. As near as we
can tell, for as far as the experiments were carried out, the creep-relaxation
process was still continuing and with the same relation. Of course, in practice,
the decay rate of the force eventually becomes too small to be of practical import,
but that does not mean that the force has gone to zero, even after half an hour
at 97°C or 24 h at 73°e. The curve at 33°C was continued to a full month, and
there was still a large force left (i.e., the solder was constraining the carrier and
the board in a bowed condition). The waviness at the far end of that curve is
unfortunate, because it might have obscured some deviation from the straight
line. The waviness had a period of 7 days, and we suspect it is associated with
 CONSTITUTIVE RELATIONS 321

1.0
0 0 20
.-.
III
0.5 D..
::!iii
.-. 10
co en
....I
33 DC
en
W
LL 0.2 a:
I-
~ 5 en
z a:
...,
(5
0.1
c:(
w
:I:
a: en
w
D.. 2 I-
W z
CJ 0.05 w
....I
a:
0
LL
1 ~
::::)
0
w
0.02
0.5

0.01
10-10 10-9 10-8 10-7 10-6 10-5
RATE OF CHANGE OF PLASTIC SHEAR STRAIN (1/sec)
Figure 10-6 Constitutive relations for solder joints.

the weekend room temperature change (the experiment was performed in De-
cember).
Figure 10-6 is a plot from the same experiment of the force as a function of
the rate of change of plastic shear strain d"'fJdt. (These curves start at the right
and proceed to the left.) In the author's opinion it would be better to plot d"'fJ
dt on the y axis, as it is normally the dependent variable. We bow to convention,
however, in Figure 10-6. According to Equation 10-24, for any given temper-
ature, the curve should be a straight line on this plot, and it should have slope
liN and x-intercept b!(AN ). Table 10-1 gives the results for band N. As nearly
as we can tell, the power-law constitutive relation holds for all these temperatures
for the time periods of the experiments.
The form of the constitutive relation is as expected, but the constants are not.
In this range of temperature and strain rate, solder is expected22 ,23 to show
"superplasticity," with values of N from 1.5 to 2.5 instead of the 5 to 22 shown
in Table 10-1. Our values are indicative of normal plasticity, which is expected
at larger values of shear strain rate. Solomon24 found values of N from 2.6 to
over 40 for strain rates from 10-5 to 0.1 sec- 1 at temperatures from - 50 to
150°C. Shine and Fox25 found a transition from normal to superplastic behavior
for strain rates below 10-5 sec-I. Their results are especially interesting since
322 CREEP AND STRESS RELAXATION IN SOLDER JOINTS

Table 10-1 Constants for Constitutive Relation 14

Duration of
Temperature b Experiment
(0C) N seLl (MPaj-N (h)
-27.7 21.9 1.2 x 10-35 BO
33.2 16.5 1.6 x 10-21 BOO
4B.3 11.5 3.7 x 10-14 24
72.7 5.4 7.7 x 1O~ 16
96.B 4.9 1.5 x 10-5 1

they were perfonned on real solder joints on chip carriers, not on bulk samples
or specially shaped test specimens. It is interesting to note that their values for
stress and strain rate are not far different from ours. In fact, they overlap.
Nevertheless, their experiment was quite different. They applied a mechanical
stress rather than a thennal stress, and they applied it cyclically, so they used
different data processing to extract the rate of change of plastic shear strain. We
still do not understand why we did not find a transition to superplasticity, whereas
they did. This would be a good area for future work.

10.5 TEMPERATURE CYCLING

We have been treating the special case of isothermal creep-relaxation. Similar

considerations apply to temperature cycling. Their application to temperature
cycling is of great practical importance because the number of cycles to failure
is often related to the strain range through the "Coffin-Manson" relation,26 strain
range partitioning,27 damage integral calculations,28.29 or other theories. In the
Coffin-Manson development, the number of mechanical strain cycles to me-
chanical failure is postulated to be inversely proportional to a power of the plastic
strain range during the cycle. Wild30 plotted the number of cycles to failure as
a function of total strain and found the power to be of the order of magnitude
of 2 but to depend on temperature and frequency. He reported nine data points
for eutectic Sn-Pb solder, involving two temperatures (25 and 100°C), two
frequencies (96 and 7200 cycles/day), and three shear strains (3, 8, and 20
percent). His results have been conveniently parameterized by Engelmaier· 31
for extrapolation to other temperatures, frequencies, and strain ranges. Wild
applied his strains mechanically, rather than by temperature cycling, so he knew
the total strain range. In temperature cycling, we generally know the temperature
range AT but not the strain range A'Y, let alone A'Ypo At high temperatures, where
the creep-relaxation process goes rapidly to completion, we can confidently use
Equation 10-1 to obtain the plastic strain range for the Coffin-Manson relation.
At lower temperatures, this is not valid. To apply the Wild or similar data to
 TEMPERATURE CYCLING 323

temperature cycling, we need to know (or assume) how Il:y varies with tem-
perature and frequency. Wild30 attributes his temperature dependence to the
changing ratio of ll-yp to ll-y. In temperature cycling, we must also consider the
temperature dependence of the total strain range, a factor that was not included
in Wild's experiment. In spite of this, the Engelmaier equation has been suc-
cessfully used in correlating temperature cycling results. 3.4,9

10.5.1 Small Temperature Range Cye/ing

When we consider temperature cycling over small temperature ranges, the elastic
constants can be considered constant. Figure 10-7 shows hysteresis loops of force
versus strain, measured with strain gauges, as before. The oven temperature was
cycled from 42 to 52°C, with ramp time equal to hold time. A schematic loop
is shown in the upper right comer. Based on our earlier concepts, if we cycle
between two temperatures, we should trace out a hysteresis loop between the
two creep-relaxation lines for the two temperatures. As we start the ramp from
one temperature to the other, we initially change the force and shear strain with
a slope GA. We see this from Equations 10-20 and 10-21. At the start of the
temperature ramp the change in the plastic shear is initially negligible, so the

0.25

0.20

g
t-
z 0.15
,0
a:
w
Q.
w
U 0.10
a:
0
LL

" ....... ,
0.05 ..............
.......
.......

,
.......
.....................
°0~---1~--~2---~3---~4~~--­ .l'YMAX - - - - - - - -...
At= ex>
--11
SHEAR STRAIN, 'Y (mrad)

Figure 10-7 Hysteresis loops between 42 and S2°e. The loops have been arbitrarily dis-
placed along the creep-relaxation slope to eliminate overlap.
324 CREEP AND STRESS RELAXATION IN SOLDER JOINTS

total shear strain change is simply flF/(GA). Handbook data 17 for the shear
modulus G at 47°C was used to draw the expected initial slope (Figure 10-7).
As mentioned, above, this number is subject to considerable uncertainty. The
literature contains values twice as large, as well as values half as large. If the
cycle were traversed infinitely fast, so that no creep-relaxation could take place,
the system would simply go up and down that line, the hysteresis loop would
have no width, and there would be no plastic strain. The total strain would then
be elastic and would be given by solving Equation 10-23 for iJ:.:y with iJ:.'Yp =
O. The result is 0.71 mrad. This is the minimum strain range expected at this
temperature for a iJ:.T of 10 K.
If we proceed up the ramp slowly, however, the creep-relaxation becomes
appreciable, and the curve deviates from the initial slope. After arriving at the
new temperature, the sample creeps or relaxes down the line until the temperature
is changed again. If we make the assumption that the initial slope extends at
least halfway up the ramp, we can measure the loop at its widest part and call
that the plastic strain range for the loop. The figure illustrates how the total range
A'Y is then divided up between plastic and elastic portions. Since we are dealing
here with ranges, the plastic strain range is always less than the total strain range.
This is a classical method for detennining plastic strain range for any hysteresis
loop (e.g., Solomon32).
Before discussing the data, we should review some experimental considera-
tions. These hysteresis loops are for five different frequencies: 0.6, 1.2, 6, 20,
and 60 cycles/day. For clarity, we have displaced them along the creep-relaxation
lines to remove overlaps. The periods are shown on the figure. The experiments
involved typical frequencies for electronic equipment in the field, substantially
lower than those used by Wild30 (viz. 96 and 7200 cycles/day). Note that the
faster cycles did not reach the creep-relaxation lines for 42 and 52°C. That is
because of the temperature lag in the sample with respect to the oven air. The
sample temperatures were measured with thennocouples, which indicated that
in the worst case (6-min ramp and dwell times), the temperature range was only
6.9K instead of 10K. In each case, the cycle was repeated six times, and only
the last cycle is represented here, as there was some "global" creep-relaxation
for the first few cycles, which was convoluted with the effects of interest. One
other confusing factor here is the unknown effect of temperature gradients, which
playa more significant role for the faster cycles. Despite this, we can learn a
lot from these results.
For example, the fastest cycle came close to approximating the elastic slope,
and it had a small plastic portion (Le., a small width). The temperature gradients
mentioned above may have caused the initial slope of this loop to be (slightly)
greater than the expected elastic slope. For slower cycles, the plastic part was
much greater. But even at a 40-h period (at this temperature) the strain range
was well below (less than half) that of a material with complete creep-relaxation
(i.e., A'YmaJ, which we postulate would be the case for an infinite period.
 TEMPERATURE CYCLING 325

PERIOD
ONE OAY ONE HOUR

:c
~ MAXIMUM STRAIN, from EQUATION 1

!. 4.0
w
CJ
Z
<C
a:
z 3.0
Ci
a:
t;
a:
<C 2.0
w
J:
CI)
C
W
~ 1.0
CI)
<C
w
::::iii
o ,--~~~~~~--~~~~~--~~~~~

0.1 1.0 10 100

FREQUENCY (CYCLES PER DAY)

Figure 10-8 Strain ranges during cycling from 42 to S2°e.

For each of these loops, we can measure the total strain range tl:y and (from
the widest part of the loop) the plastic strain range ll.'Yr We normalized the
results from the fast cycles to a ll.T of 10 K by multiplying by the correction
factor of 10/ll.T. The results are plotted on Figure 10-8. Both the total and plastic
strain range decrease as the frequency is increased. But all are well below the
value given by Equation 10-1 for a material in which the creep-relaxation process
has gone to completion. Note that at the lowest frequency the strain is mostly
plastic, while at the highest frequency it is about half elastic. As discusseq above,
the total strain range at this temperature is not expected to go below 0.71 mrad.
At 60 cycles/day it is already 0.8 mrad, so it must be leveling off. We cannot
extend it to higher frequencies because of the temperature gradients.
Figure 10-9 shows results 14 from similar small temperature cycles where we
have maintained the frequency at 6 cycles/day and varied the temperature mid-
point from 0 to 100°C. Here we see how sharply the temperature affects the
strain range. We found that at 80°C, the (total) strain range was 70 percent of
that obtained from Equation 10-1. At 20°C, it was only 12 percent. The plastic
strain range was 60 and 4 percent of that obtained from Equation 10-1 for the
326 CREEP AND STRESS RELAXATION IN SOLDER JOINTS

.-
"C
.~ MAXIMUM STRAIN, from EQUATION 1

·E 4.0
w
CJ
z
<C
a:
Z 3.0
C(
a:
Ii;
~ 2.0
w
J:
tn
C
W
a: 1.0
::I
~
w
:E
OL-~~ __L--L~__~-L~L-~~__L-~~~
-20 o 20 40 60 80 100 120
TEMPERATURE (DC)

Figure 10-9 Strain ranges during cycling with J1 T = 10K and frequency = 6 cycles/day.

same two temperatures. Judging from this figure, for temperature ranges above
100°C it is reasonable to assume that the strain range is proportional to I1T, as
given by Equation 10-1. If the temperature range under consideration includes
temperatures below 80°C, the plastic strain range is less than half that calculated
from Equation 10-1, so it would be well to realize that l1"{p is not proportional
to I1T. Since the lifetime is roughly inverse to the square of the strain range ,30
a factor of 2 in the strain range means a factor of 4 in the calculation of the
number of cycles to failure.

10.6 LARGER TEMPERATURE CYCLES

The situation is considerably more complicated for larger temperature cycles,

especially when they involve both temperature ranges: those where the mismatch
produces mostly forces and those where it produces mostly strains. We next
consider briefly such a case from our previous work. 14,15
A leadless ceramic chip carrier without castellations was surface mount sol-
dered to an FR-4 PWB, with strain gauges attached to the top and bottom of
the middle of the carrier and also to the top and bottom of the middle of the
 LARGER TEMPERATURE CYCLES 327

PWB. Experimental details are given in Hall. 15 The assembly was slowly cycled
in a chamber according to the schedule shown in Figure 10-1Ob. Using Equations
10-15 and 10-18, we obtained the shear strain and shear force during the cycle.
The measurements showed good repeatability after the first half cycle. 15 There-
fore, we only show the second cycle here. This cycle is considered moderately
severe. The assembly would begin to show electrical failure after a few hundred
of these cycles. 2
Figure 10-1 Oa shows the shear strain as a function of temperature. The almost-
straight line with the triangle data points is the strain expected from Equation
10-1, given the expansivities as a function of temperature as measured indepen-
dently from similar strain gauges. This would be the case of complete creep-

40
A B
0;
c 30
01
'6 12S0C-
~
! 20

z
cc
a: 10
Iii
a: -2S0C-
cW
0 "I- ·1·
::c
II) _ COMPLETE CREEP I 5 2 5
RELAXATION TIME (HOURS)
-10
-40 0 40 80 120
TEMPERATURE (OCI

I --1---.------,--'-
0.10
C
iii" 0 - --- -- -----
c
z
5 -0.10 . \

!!:. \

'$.
~ -0.20 'P
a:
",,
\
o \
\
LL -0.30 \
a:
c
\
\
\
"
\
~ -0.40 \
II)
\
,,
,,
-0.50 \

-'--'~-:'::-~'--'--::'::-'~L_~..I. •
_-,,--,-I _~ __ .L_.\..
-40 40 80 120 -10 o 10 20 30 40
TEMPERATURE (OCI SHEAR STRAIN (milliradians)

Figure 10-10 Temperature cycling with large ilT.

328 CREEP AND STRESS RELAXATION IN SOlDER JOINTS

relaxation as the temperature is changed. In this plot, the zero for the strain is
arbitrary. For the assembly curve, we used the lowest temperature value as the
zero. For the complete creep-relaxation curve, we adjusted its zero such that it
would match the assembly curve at high temperatures. At high temperatures,
the shear strain for the assembly follows the complete creep-relaxation curve
well. That is because the creep-relaxation is fast enough to be almost complete
as the temperature is changed. The deviation from the complete creep-relaxation
line as the temperature is reduced is opposite to that obtained as the temperature
is raised. That is because the shear force on the joints is in the opposite direction
in the two cases. As the temperature comes down, the deviation from the complete
creep-relaxation curve increases as the solder becomes stronger, and at - 25°C
the slope is close to zero, indicating high rigidity of the solder and little creep-
relaxation. During the 2-h dwell at - 25°C, there is little change in the shear
since the creep-relaxation is extremely slow at these temperatures (despite the
large forces). As the temperature is increased again, the shear strain stays almost
constant until it reaches the complete creep-relaxation line, at which point the
force reverses, creep-relaxation takes over, and the curve asymptotically ap-
proaches the complete creep-relaxation curve (as given by Equation 10-1).
Figure 10-1 Oc shows the force measured during the same cycle, using Equation
10-18. We also have to choose a zero value for the force since a strain gauge
only indicates changes from a reference state. In this case, we have a better
indication of where the zero is since we know that it always relaxes toward zero
and that it is small at the highest temperatures. The zero chosen for this plot is
the force at the end of the 2-h dwell at 125°C.
As the temperature is decreased from 125°C, the force gradually increases in
the negative direction, and its slope gradually increases also, as the solder gets
"stronger." The curve traced out during decreasing temperature is similar in
shape to that of the ultimate shear strength. 33 As the temperature goes below
room temperature, the forces get very large indeed. This puts the largest stress
on the metallization and components that are soldered together. At -25°C, the
2-h dwell causes a barely detectable decrease in the force. As the temperature
is increased again, the force increases quite linearly, all the way up to 40°C.
During this time, the large bend in the chip carrier and the PWB are steadily
decreasing, until they cross the point of zero force (i.e., no bend). This particular
part of the cycle is much like a classical bimetallic strip, where there is no shear
between the two members, but the force and the bend are linearly related to the
temperature. The slope of the line in this region is 0.0087 lb/K, and this is
consistent with an estimate of the shear forces expected for a bimetallic strip.14
Thus, we can roughly divide the temperature range into two regions: (1) Above
50°C, the solder creeps-relaxes enough that Equation 10-1 is adequate and (2)
below 50°C, the solder is stiff enough that the bimetallic strip analogy holds.
That is, the force changes but the shear strain does not. The lesson is that
temperature cycling in the high range is completely different from cycling in the
 LARGER TEMPERATURE CYCLES 329

low range. Thus we concur with the conclusion ofEngelmaief that if an assembly
is not to be used at low temperatures, it may be misleading to evaluate its
reliability from temperature cycles that include the low temperature region.
Figure lO-lOd displays the same data, plotting force as a function of strain.
Again the two regions are clearly shown. At high temperatures, the force is
small and the strain changes are large. At low temperatures the force is large
and the strain changes are small. In this plot we have added the creep-relaxation
isothermal lines (large dashed lines outside the loop) as predicted by Equation
10-14. They all have the same slope, and their separation is uniform. The slope
is - kH, where H is 0.011 in., and k was taken as 2.0 lb/mil, as calculated from
a bimetallic strip model. 14 We could also have obtained the slope from Figure
lO-lOc since we know that if !>.:y is negligible,

1 dF
k=-- (10-24)
Lao. ar
Using this relation, the value of k is 1. 7Ib/rnil, not far different from the bimetallic
strip value. We have used the value 2.0 in making the creep-relaxation lines of
Figure lO-lOd since that was used before, and it fits the data a little better. The
separation between the creep-relaxation lines is given directly from Equation 10-
I since it must be valid for the x-intercept (zero force) as assumed by Equation
10-1.
The dotted lines inside the loop were obtained as follows. As we cycle through
the loop, we cross any given temperature value twice; once while increasing and
once while decreasing the temperature. We have interpolated the raw data to
obtain the force artd strain at which we crossed even values of the temperature.
Then we drew a dotted line connecting each pair of points thus determined. This
created an experimental grid of creep-relaxation lines that can be compared with
the simple linear theory indicated by the dashed lines outside the loops. The
agreement is satisfactory. The slope at 40 oe, for example, is just as predicted,
although the location is slightly off. At - 20 oe, the location is slightly off, but
the slope is still in good agreement. At 100oe, the slope is slightly off, but the
location is as predicted. This is considered good evidence that the simple linear
relation among force, strain, and temperature is valid for this type of cycling.
One can, in principle, use the constitutive relations along with the temperature
profile (Fig. lO-lOb) to numerically generate this entire loop. This has been
done. 28 •29 ,34,35 Such approaches can be used to extrapolate these results to tem-
perature cycles of different frequencies, different aT's, and different geometries.
That is beyond the scope of this chapter. Suffice it to say that we have obtained
some insights into the behavior of these solder joints as we change the temper-
ature. The most important result is that low temperatures are fundamentally
different from high temperatures, and we should bear that in mind as we consider
the reliability of these joints.
330 CREEP AND STRESS RELAXATION IN SOLDER JOINTS

10.7 ACKNOWLEDGMENTS

I would like to acknowledge the contribution of R. A. Deighan, who wrote the

programs for processing the strain gage data, and F. L. Howland, who provided
stimulation, experience, and advice.

REFERENCES

1. Moore, R., Guidelines for Surface Mounting and Interconnecting Chip Carriers,
IPC-CM-78, Evanston, IL, Inst. for Interconnecting and Packaging Electronic Cir-
cuits, 1983.
2. Sherry, W. M., and P. M. Hall, "Materials, Structures, and Mechanics of Solder-
Joints for Surface-Mount Microelectronics Technology," in Proc. Conf. Intercon-
nection Technology in Electronics, Fellbach, W. Gennany, Deutscher Verb and fur
Schweisstechnik, 1986, pp. 47-61.
3. Engelmaier, W., "Functional Cycling and Surface Mounting Attachment Reliability,"
ISHM Technical Monograph Series, Int. Soc. for Hybrid Microelectronics, Silver
Spring, MD, Series 6984--002, 1984, pp. 87-114.
4. ___ , "Surface Mount Attachment Reliability of Clip-Leaded Ceramic Chip Car-
riers on FR-4 Circuit Boards," Proc. International Electronics Packaging Confer-
ence, 1987, pp. 104-113.
5. Dance, F. J., and J. L. Wallace, "Clad Metal Circuit Board Substrates for Direct
Mounting of Ceramic Chip Carriers," Electronic Packaging & Production, Jan. 1982,
pp. 226-237.
6. Waller, D. L., L. R. Fox, and R. J. Hannemann, "Analysis of Surface Mount Thennal
and Thennal Stress Perfonnance," IEEE Trans. on Components, Hybrids, and Man-
ufacturing Technology, CHMT-6(3), 1983, pp. 257-266.
7. Hall, P. M., T. D. Dudderar, and J. F. Argyle, "Thennal Defonnations Observed
in Leadless Ceramic Chip Carriers Surface Mounted to Printed Wiring Boards," IEEE
Trans. on Components, Hybrids, and Manufacturing Technology, CHMT-6(4) , 1983,
pp. 544-557.
8. Howard, R. T., S. W. Sobeck, and C. Sanetra, "A New Package-Related Failure
Mechanism for Leadless Ceramic Chip Carriers (LC-3s) Solder Attached to Alumina
Substrates," Solid State Technology, 26(2), 1982, pp. 115-121.
9. Engelmaier, W., "Is Present-Day Accelerated Cycling Adequate for Surface Mount
Attachment Reliability Evaluation?" IPC-TP-653, Proc. !PC. Lincolnwood, IL: In-
stitute for Interconnecting and Packaging Electronic Circuits, 1987.
10. Clech, J.-P., "Analysis of Thennomechanical Strains in Solder Joints of Surface-
Mount Assemblies," Abstract in Proc. IEEE International Electronic Manufacturing
Symposium, 1987, p. 89.
11. King, J. A., Materials Handbookfor Hybrid Microelectronics, Boston, Artech House,
1988.
12. Harper, C. A., Handbook of Materials and Processes for Electronics, New York,
McGraw-HilI, 1968.
 REFERENCES 331

13. Murty, G. S., "Stress Relaxation in Superplastic Materials," J. Materials Sci., 8,

1973, pp. 611-614.
14. Hall, P. M., "Creep and Stress Relaxation in Solder Joints of Surface-Mounted Chip
Carriers," IEEE Trans. on Components, Hybrids, and Manufacturing Technology,
CHMT-12 (4) , 1987, pp. 556-565.
15. _ _ , "Forces, Moments, and Displacements during Thermal Chamber Cycling
of Leadless Ceramic Chip Carriers Soldered to Printed Boards," IEEE Trans. on
Components, Hybrids, andManufacturing Technology, CHMT-7(4), 1984, pp. 314-
327.
16. Twnmala, R. R., and E. J. Rymaszewski, Eds., Microelectronics Packaging Hand-
book, New York, Van Nostrand Reinhold, 1989.
17. Sherry, W. M., J. S. Erich, M. K. Bartschat, and F. B. Prinz, ''The Effect of Joint
Design on the Thermal Fatigue Life of Leadless Chip Carrier Solder Joints," IEEE
Trans. on Components, Hybrids, and Manufacturing Technology, CHMT-8(6), 1985,
pp. 417-428.
18. Lau, J. H., D. W. Rice, and P. A. Avery, "Nonlinear Analysis of Surface Mount
Solder Joint Fatigue," Proc.lnt. Electronic Manufacturing Techn. Symp., New York,
IEEE-CHMT, 1986, pp. 173-184.
19. Payne, R. K., and B. Livesay, Task I Report, General Approach to SMD Fault
Identification, submitted to Summit Technologies, Inc., and U.S. Army Laboratory
Command, Georgia Institute of Technology, 1989.
20. Jones, J. W., "Theory of Creep Deformation," in Metals Handbook, Volume 8,
Mechanical Testing, Metals Park, OH, American Society for Metals, 1985, p. 309.
21. Baker, E., "Stress Relaxation in Tin-Lead Solders," Materials Sci. and Engineering,
38, 1979, pp. 241-244.
22. Kashyap, B. P., and G. S. Murty, "Experimental Constitutive Relations for the High-
Temperature Deformation of a Pb-Sn Eutectic Alloy," Mater. Sci. Eng., SO, 1981,
pp. 205-213.
23. Lam, S. T., A. Arieli, and K. Mukherjee, "Superplastic Behavior of Pb-Sn Eutectic
Alloy," Materials Sci. And Engineering, 1979, pp. 40, 73-79.
24. Solomon, H. D., "Fatigue of 60/40 Solder," Proc. Electronic Components Confer-
ence, 1986, pp. 622-629.
25. Shine, M. C., and L. R. Fox, "Fatigue of Solder Joints in Surface Mount Devices,"
in Low Cycle Fatigue, ASTM Special Tech. Publication 942, H. D. Solomon, G.
R. Halford, L. R. Kasand, and B. N. Leis, Eds., Philadelphia, American Soc. for
Testing and Materials, 1987, pp. 588-610.
26. Manson, S. S., Thermal Stress and Low Cycle Fatigue, New York, McGraw-Hill,
1966.
27. Halford, G. R., and S. S. Manson, "Life Prediction ofThermal-Mechanical Fatigue
Using Strain Range Partitioning," in Thermal Fatigue of Materials and Components
Spera and Mowbray, Eds., ASTM Special Technical Publication 612, 1975, pp.
239-254.
28. Subrahmanyan, R., J. R. Wilcox, and C.-Y. Li, "A Damage Integral Approach to
Thermal Fatigue of Solder Joints," Proc. Electronic Components Conference, 1989,
pp. 240-252.
332 CREEP AND STRESS RELAXATION IN SOlDER JOINTS

29. _ _ , "A Damage Integral Approach to Solder Joint Fatigue," Proc. ASM Elec-
tronic Packaging Conference-Materials and Processes, ASM International, 1989,
pp. 213-221.
30. Wild, R. N., "Some Fatigue Properties of Solders and Solder Joints," in Nat. Elec-
tronic Packaging Conf., 1974, pp. 409-414.
31. Engelmaier, W., "Accelerated Solder Joint Reliability Testing and Result Evalua-
tion," Chapter 16 of this book, 1991.
32. Solomon, H. D., "Low Cycle Fatigue of 60/40 Solder- Plastic Strain Limited vs.
Displacement Limited Testing," Proc. ASM Electronic Packaging: Materials and
Processes Conference, 1985, pp. 29-47.
33. Shamash, M. B., Development ofHighly Reliable Soldered Joints for Printed Circuit
Boards, NASA Contract NAS 8-21233, Final Report, 1968, pp. 4-60.
34. Wilcox, J. R., R. Subrahmanyan, and C.-Y. Li, "Thermal Stresses and Inelastic
Deformation in Solder Joints," Proc. ASM Electronic Packaging Conference-Ma-
terials and Processes, ASM International, 1989, pp. 203-211.
35. Clech, J. -P., and J. A. Augis, "Engineering Analysis of Thermal Cycling Accelerated
Tests for Surface-Mount Attachment Reliability Evaluation," Proc. 7th International
Electronics Packaging Conference, 1987, pp. 385-410.
 11
Effects of Strain Range, Ramp Time,
Hold Time, and Temperature on
Isothermal Fatigue Life of Tin-Lead
Solder Alloys
Semyon Vaynman and Moris E. Fine

An understanding of the mechanical and fatigue properties of solder alloys used

in electronics is required for better design of solder joints, accelerated test de-
velopment, and development of improved solders.
Solders of different compositions are used in electronic devices, and while
there are some similarities in fatigue behavior of different solders, they are
actually different materials with their specific microstructural, mechanical, and
fatigue properties. Therefore, application of data on solder alloys found in the
literature to a specific device must consider solder composition, microstructure,
specimen design, testing conditions, fatigue life definition, and other factors to
avoid erroneous assessment of fatigue reliability of the device.
Strain range, ramp time, hold time, and temperature are the most important
factors affecting isothermal fatigue of materials in the given atmosphere. The
effect of these parameters on the isothermal fatigue life of solder materials in
the air will be discussed in this chapter.

11.1 DEFINITION OF FAILURE, SPECIMEN DESIGN, AND

MODE OF LOADING

Fatigue data discussed here have been obtained for specimens of different de-
sign, ranging from a solder joint in a real device l - 3 and two plates joined by
solde~ to a bulk specimen9--J3 tested under different loading conditions (ten-
sion,10--12 shear,2-8 bending,14 and torsion l ,9). Different definitions of end of

333
334 EFFECTS OF STRAIN RANGE, RAMP TIME, HOLD TIME, AND TEMPERATURE

fatigue life have been used at different laboratories: visible cracking proportioned
from total fracture,14 a predetermined drop in load,l-s.9.ls start of the drop of
the maximum tensile stress, 10-13 or of the tensile stress-compressive stress ratio 13
and predetermined increase in electrical resistance,z·16 etc. Of coarse, a different
definition of failure leads to a different fatigue life. For example, when end of
fatigue life in near-eutectic solder was defined as the number of cycles to reduce
the tensile stress to 75, 50, or 10 percent of its initial value, the different
definitions of fatigue life led to different slopes of Coffin-Manson plots. IS Thus,
it is very often difficult, if not impossible, to compare data developed at different
laboratories and correlate these data to the fatigue life of a solder joint in the
device.

11.2 EFFECT OF STRAIN RANGE ON FATIGUE LIFE

Increasing the strain amplitude leads to a decrease in the number of cycles to

failure. Low temperature « 0.3 absolute melting point, Tm), low cycle fatigue
data for many metals may be fitted to the empirical Coffin-Manson relation: 17.18

N! Ep = C (11-1)

where Nt = number of cycles to failure

Ep = plastic strain range

13, C = constants

For most metals the constant 13 is equal to approximately 0.5.

The Coffin-Manson relation was found to be valid for 60% Sn-40% Pb solder4's
tested in shear in the 1 to 10 percent plastic strain range at 35 and 150°C (Fig.
11-1). The exponent 13 was approximately 0.5 at temperatures below 150°C and
decreased to 0.37 at 150°C. An intergranular mode of failure was observed at
all testing conditions.
Limited isothermal experimental bending fatigue data for 95% Pb-5% Sn
solder l4 was found to obey the Coffin-Manson relation. The exponent 13 was
found tp be equal to 0.43 to 0.46 in the strain range of 0.1 to 5.0 percent and
the temperature range of 25 to 85°C. For temperatures equal to or higher than
120°C at strain ranges higher than 1 percent this exponent became about 1. 95%
Pb-5% Sn solder also was found to approximately obey the Coffin-Manson
relation when tested in torsion from 0.05 to 2.0 percent (Fig. 11_2).9 In these
tests at 1.0 percent total strain at room temperature transgranular cracking was
observed. At 150°C the solder failed in the grain boundaries.
The Coffin-Manson relation was shown to describe approximately the fatigue
life of the Indalloy 151 solder (92.5% Pb-5% Sn-2.5% Ag) tested in shear (0.5
 EFFECT OF STRAIN RANGE ON FATIGUE LIFE 335

0=0.52

o
o

163~~~~~~~~~WL~~~~~~~-U~
10- 1 % 101 102 10 3 10 4 105
Nf (cycles to 1/2 load)
Figure 11·1 Number of cycles to failure versus plastic shear strain for 60% Sn-40% Pb
solder. Tests without hold time. (After H. D. Solomon 4 )

10.0

-
I'"'Q'\
~
5.0

~ 2.0
<1 ~
1.0
~ 0.5
<1
?' 0.2
<J 0.1
0.05
102 103 10 4 10 5
eyc I es to fai lure (N9
Figure 11·2 Number of cycles to failure versus plastic strain for 95% Pb-5% Sn solder.
Tests without hold time. (After M. Kitano, T. Shimizu, and T. Kumazava 9 )
336 EFFECTS OF STRAIN RANGE, RAMP TIME, HOLD TIME, AND TEMPERATURE

t r =0.1-2.5 sec
0.6

A 5°C
o 25°C
x 50°C
[] 80°C
0.2 + 100°C
103
Nf ~ycles)
Figure 11-3 Number of cycles to failure versus plastic strain for 96.5% Pb-3.5% Sn solder.
Tests without hold time. (After S. Vaynman, M. E. Fine, and D. A. leannotte lO )

to 100 percent plastic strain) with a cycling frequency of 0.3 HZ.15 Ductile
fracture was detected at all testing conditions.
However, for 96.5% Pb-3.5% Sn solderlOtested at low strain ranges in tension-
tension it is evident that the data for each temperature cannot be well fit to a
single log-log straight line (Figure 11-3). The data are much better represented
by two straight lines with a break point at approximately 0.3% plastic strain.
The best fit values of the constants ~ and C from Coffin-Manson equations for
this solder at different temperatures are summarized in Table 11-1.
In general, cracks in this low-tin lead-based solder could be observed under
all testing conditions very early in the fatigue life. Figure 11-4 shows that at
0.75 total strain range intergranular fatigue cracks as well as transgranular cracks
along slip bands are formed. At low strain ranges (e.g., 0.3 percent total strain
range) coarse slip lines and the transgranular cracks associated with them are
absent (Fig. 11-5). Thus, the break in the lines in the Coffin-Manson plot (Fig.

Table 11-1 Values for Constants in Coffin-Manson Relation for 96.5% Pb-3.5%
Sn Solder"
Low Strains High Strains
Temperature
(OC) ~ C ~ C
5 0.45 0.15 0.45 0.15
25 0.62 1.07 0.39 0.14
50 0.56 0.43 0.37 0.08
80 0.47 0.13 0.24 0.02
 EFFECT OF STRAIN RANGE ON FATIGUE LIFE 337

Figure 11-4 Scanning electron micrograph of surface offailed specimen. Total strain range
0.75 percent, 25°C. Ramp time 2.5 sec. No hold. 2550 cycles. Nt = 2500 cycles. (After S.
Vaynman, M . E. Fine, and D. A. Jeannotte 10)

Figure 11-5 Scanning electron micrograph of surface replica and failed specimen . Total
strain range 0.30 percent, 25°C. Ramp time 2.5 sec. No hold . 30,000 cycles. N t = 29,000
cycles. (After S. Vaynman, M. E. Fine, and D. A. Jeannotte 10)
338 EFFECTS OF STRAIN RANGE, RAMP TIME, HOLD TIME, AND TEMPERATURE

11-3) has been attributed to the change in the fracture mode from intergranular
at low strains to mixed transgranular-intergranular at high strains. 10

11.3 EFFECT OF FREQUENCY ON THE FATIGUE LIFE

At temperatures well below one-half of the absolute melting point, frequency

has little effect on the fatigue life of most metals. When the temperature is
increased to over half of the absolute melting point, a reduction in frequency
decreases the number of cycles to failure for many metals including sol-
ders. 4 ,5. 10-12. 19-30 Failure tends toward a dependence on the total time of stress
application rather than on the number of strain cycles. 31 This behavior arises
because at high temperatures creep failure such as void formation and environ-
mental attack, which are time dependent processes, play important roles in
damage accumulation.
A general trend in the effect of frequency on the number of cycles to failure
for a number of metals 23 ,24 is observed (Fig. 11-6); there is almost no frequency
effect at very high and very low frequencies but a large effect at intermediate
frequencies. Of course, the frequency effect depends on the material and on
testing conditions.
J. F. Eckel 19 studied the fatigue behavior of lead in a rotating bending test at
the strain ranges of 0.27 to 0.65 percent and found a marked influence of
frequency on fatigue life. The time to failure tf and the number of cycles to

D o
•

B~
. : . . - - -.....s.....!o~

10-3 16 2 161 1 10
Frequency Hz
Figure 11-6 Schematic representation of number of cycles to failure versus cycling fre-
quency. (After S. S. Manson 23 )
 EFFECT OF FREQUENCY ON THE FATIGUE LIFE 339

failure were related to frequency f by the following empirical relations:

ft = bf-m (11-2)
Nf = bf'-m (11-3)

where b and m are constants

The frequency exponent m for lead was found to be equal to 0.7 at 43°C and at
all strains tested. 20
Equations (11-2) and (11-3) were also found to be valid for chemical grade
lead and lead with 1 percent Sb29 tested at room temperature in reverse bending
(zero mean strain) in the strain range from 0.1 to 1.0 percent. However, the
value of m was found to depend on the strain range and to vary from 0.4 to 0.8.
The effect of frequency on fatigue life was found to be stronger at high strains.
It was also found that the stiffer the lead alloy the less its fatigue life was affected
by the variation in frequency.
The number of cycles to failure for 60% Sn-40% Pb solder tested in shear at
35 and 150°C was reduced when frequency of cycling was decreased below
approximately 3 X lQ4 Hz at 35°C and 3 x 10-3 at 150°C. 4 ,s The frequency
constant was 0.42 for 'Yp = 10 percent at 35°C (Fig. 11-7) and was equal to 0
at 5.0 and 10.0 percent of 'Yp' At higher frequency the decrease in the number
of cycles to failure with increasing time per cycle was less significant.

x Points at -50°C

Figure 11-7 Number of cycles to failure versus cycling frequency for 60% Sn-40% Pb
solder. Test run at 35°C (with some - 50°C data). (After H. D. Solomon 4 )
340 EFFECTS OF STRAIN RANGE, RAMP TIME, HOLD TIME, AND TEMPERATURE

The effect of frequency on fatigue life of 95% Pb-5% Sn solder was evaluated
in bending tests from 0.033 to 192 sec/cycle and in strain ranges from 0.5 to
4.0 percent. 14 Eckel's relations in Equations (11-2) and (11-3) were found to be
valid with a frequency exponent m equal to 0.25 at all strains tested.
For 96.5% Pb-3.5% Sn solder practically no change in the number of cycles
to failure was observed on varying frequency in the comparatively high frequency
region. IO ,1I A reduction in the number of cycles to failure with decreasing fre-
quency below 10-2 Hz was found at 25, 50, and 80°C (Fig. 11-8 through 11-
10) for all strain ranges tested. The behavior of this solder obeys the general
trends found by Eckel. The values of the constants m and b for 96.5% Pb-3.5%
Sn solder are summarized in Table 11-2. While the frequency exponent m at 50
and 80°C in tests with total strain of 0.60 percent are practically the same, it
differs from exponents found in tests with a total strain of 0.75 percent at room
temperature and in tests with a total strain of 0.50 percent at 80°C. This indicates
that the frequency effect on fatigue life of this solder may be practically inde-
pendent of temperature for a given strain range but is a function of the strain
range. The effect of frequency on the fatigue life of this solder is related to
intensification of damage in grain boundaries with increasing time per cycle
during fatigue (Fig. 11-11). At these strain ranges and temperatures some trans-
granular cracking can be observed; however, the grain boundary cracking dom-
inates. Thus, creep processes, such as grain boundary cavitation and grain bound-
ary sliding, may be the main modes of failure for this solder under conditions
of slow cycling. Since the failure of this solder at low strains and at high frequency
is already mainly intergranular, the effect of frequency on fatigue life is less
pronounced at low strains than at high ones where the mode of failure changes

~Et = 0.75%
106
'U'
Q)

-
..!:!J/

Z- 104
+-

102
~~------------~------------~--~

Figure 11-8 Effect of frequency on fatigue life of 96.5% Pb-3.5% Sn solder. (After S.
Vaynman, M. E. Fine, and D. A. jeannotte lO )
 EFFECT OF FREQUENCY ON THE FATIGUE LIFE 341

t:..Et = 0.60%
10 6
50°C
--0---
'In'

.,
~ 10 3 80°C
U'
cv
u ,--0---
,
- --
>. ~
~
,, ,-
,
Z \ 104
,~

102 J,, '

10-2 10°
f (Hz)
Figure 11-9 Effect of frequency on fatigue life of 96.5% Pb-3.5% Sn solder. (After S.
Vaynman, M. E. Fine, and D. A. leannotte lO )

/). E t =0.50%
80°C - 106
(j)'

.>-
Q) 'U'
u Q)

--
>. ~
~ 10 5
I- , -0---
-10 3
Z
,\
X
- 104
1 \ I

10- 2 10°
f (HZ)
Figure 11-10 Effect of frequency on fatigue life of 96.5% Pb-3.5% Sn solder. (After S.
Vaynman, M. E. Fine, and D. A. leannotte 10)

Table 11-2 Values of Constants m and b from Eckel's Equation for 96.5% Pb-
3.5% Sn solder. Frequency below 10-2 HZ10
Total Strain Temperature
Range (%) (0C) m b
0.75 25 0.70 36100
0.60 50 0.56 10070
0.60 80 0.59 8090
0.50 80 0.25 2820
342 EFFECTS OF STRAIN RANGE, RAMP TIME, HOLD TIME, AND TEMPERATURE

50 urn

Figure 11-11 Scanning electron micrograph of surface offailed specimen . Total strain range
0.75 percent, 25°C. Ramp time 60 sec. No hold. 1300 cycles. N f = 1230 cycles. (After S.
Vaynman, M. E. Fine, and D. A. leannotte lO )

from mixed transgranular-intergranular at high frequency to predominantly in-

tergranular at low frequency.
A combination of the Coffin-Manson Equation (11-1) and Eckel's Equation
(11-3) leads to the frequency-modified Coffin-Manson relation. The frequency-
modified Coffin-Manson relation may be used only when the parameter m is
independent of the strain range. However, in case of low-tin lead-based solder
tested at total strain ranges under 1 percent this relation cannot be used since
the m is a function of strain range. 10,11

11.4 EFFECT OF HOLD TIME ON FATIGUE LIFE

Hold (dwell) times introduced in the cycle at maximum and/or minimum strains,
stresses, or temperatures affect the fatigue life of materials depending on the
kind of material and testing conditions.
 EFFECT OF HOLD TIME ON FATIGUE LIFE 343

Tensile hold time is very damaging during high temperature fatigue of many
materials such as steels,n, 31-35 alloys,27,33,36,37 and solders, 10-13,38-40 The num-
ber of cycles to failure decreases when tensile hold time is increased; however,
a saturation in the number of cycles to failure with increasing hold time is detected
for a number of metals.
In a few cases, however, as in high temperature isothermal fatigue of such
metals as Rene 80,41 2.25% Cr-l% Mo alloy,42 Udimet 700,43,44 and nickel-
base superalloy In 738,45 tensile hold time gives greater fatigue lives due to
formation of a mean compressive stress during cycling.
Often the reduction in cycles to failure of metals due to tensile hold is
accompanied by a transition from the transgranular to the intergranular mode
of fracture. Cavitation on grain boundaries is observed; therefore, the effect
of tensile hold time on fatigue life was attributed mainly to creep (cavita-
tion).25-28,33,46,47
In some cases the reduction in fatigue life due to tensile hold was attributed
to the effect of environment. Oxidation was found to have a much higher effect
on fatigue life of some steels at elevated temperature than creep. 24
Hold time at maximum strain has a dramatic effect on the number of cycles
to failure of low-tin lead-based and lead-tin eutectic solders 10-- 13 (Fig. 11-12 and
11-13); a few minutes of hold time reduce the number of cycles to failure almost

=0.75"10
t:.€

2000 tr = 2.5sec
25°C 150 "U'
Q)
(i)'
Q) ~

-
(J .."
>-

-
100 0

-
~
1000
z
50 -

N'f
o L.----20..r...0---4-J0'-0---60..r...0~
tht (sec)
Figure 11-12 Effect of tensile hold time on fatigue life of 96.5% Pb-3.5% Sn solder. (After
S. Vaynman, M. E. Fine, and D. A. leannotte1o)
344 EFFECTS OF STRAIN RANGE, RAMP TIME, HOlD TIME, AND TEMPERATURE

63Sn -37Pb
1500 1% total strain
Ramp 1 sec
25°C

-
z

600
Tensi Ie hold (sec)
Figure 11-13 Effect of tensile hold time on fatigue life of 63 % Sn-37% Pb solder. (After S.
Vaynman and M. E. Fine 13 )

by an order of magnitude and a further increase in hold time eventually leads

to fatigue life saturation. There exists a linear relation between time to failure
and tensile hold time per cycle (when hold time ~ ramp time) (Figs. 11-12 and
11-13):

It = A + Btm (11-4)

where tf = time to failure

tm = hold time at maximum strain (tensile hold)

Since time per cycle te is equal to

te = 2tT + ttb (11-5)

where tT = ramp time

the number of cycles to failure can be found as:

_
Nf - tf A + Bttb
(11-6)
2tT + ttb 2tT + Bttb
 EFFECT OF HOLD TIME ON FATIGUE LIFE 345

When the ramp time is much less than hold time, as in the present study, it
can be neglected and finally

(11-7)

For long hold times, the first term in Equation (11-7) becomes negligible and

Nf = B = N'r (11-8)

From the Equation (11-8) the limit in number of cycles to failure under a
given set of conditions is equal to the slope of the time to failure versus tensile
hold time per cycle line (Figs. 11-12 and 11-13). Thus, when tth }} t" solders
will survive not less than B number of cycle with any tensile hold time in the
cycle under a given strain range and temperature. Equations (11-7) and (11-8)
were found to hold for other materials alsO. 48
Using Equation (11-7) the data for a few strain ranges were extrapolated to
higher hold times and approximately the same number of cycles to failure was
found for tensile hold times of 1 and 10 h and this essentially gives Nj values
(Fig. 11-14).10 It is obvious that the Coffin-Manson relation does not hold for
this solder. There are three regions in this figure; at high and low strains there
is dependence of the number of cycles to failure on the strain range, while at
intermediate strains the dependence is very weak. Such behavior of this solder,
similar to the behavior of a number of steels at high temperature,49 indicates a
strong creep-fatigue interaction in low-tin lead-based solder. Thus, the simple

0.6
o~
-
~
c. O•4
<l

0.2 25-C
J,. 1.0 - 2.5 sec

Figure 11-14 Number of cycles to failure versus plastic strain for 96.5% Pb-3.5% Sn solder.
Tests with tensile hold time. (After S. Vaynman, M. E. Fine, and D. A. Jeannotte lO)
346 EFFECTS OF STRAIN RANGE, RAMP TIME, HOlD TIME, AND TEMPERATURE

Coffin-Manson relation cannot be used for isothermal fatigue life prediction of

solders at strain ranges below 1 percent.
Tests performed for low-tin lead-based solder in air and vacuum50 ,51 at 0.80
percent total strain range indicate that only some of fatigue life reduction with
increasing tensile hold time may be attributed to environmental attack; the number
of cycles to failure saturated in vacuum as it did in air; however, fatigue life in
vacuum was found to be only slightly higher than in air (Fig. 11-15). Therefore,
most damage during fatigue with hold time at this strain range has to be attributed
to creep and in particular to the formation and growth of voids on grain bound-
aries. Voids were found on grain boundaries during fatigue of 98% Pb-2% Sn52
and 97.5% Pb-3.5% Sn50 •51 solders.
While mixed transgranular-intergranular fracture was observed in low-tin lead-
based solder in no-hold time tests at higher strains (Fig. 11-4), fracture of solder
along grain boundaries was characteristic for this solder in tests with tensile hold
time in air at all strain ranges tested lO (Fig. 11-16).
While tensile hold time is found to be very damaging for a number of metals,
including solders, compressive hold times are usually not nearly as damaging.
The processes of formation and growth of grain boundary cavities were found
to be detrimental in fatigue life under tensile hold time conditions. The less
damaging effect of compressive hold time is attributed to the processes of shrink-

3000~----------------------------~
o Vacuum Cp=2X16 5 Torr)
l::. Air CRH=50 %
)

~ 1000
z

-- ---e-----e
"-
--_ 0

100 200 300 400

Hold time (sec)
Figure 11-15 Effect of tensile hold time on fatigue life of 96.5%-3.5% 5n Pb solder in
vacuum. (After R. Berriche, S. Vaynman, M. E. Fine, and D. A. jeannotte50 )
 EFFECT OF HOLD TIME ON FATIGUE LIFE 347

Figure 11 -16 Scanning electron micrograph of surface replica . Total strain range 0.75
percent, 25°C. Ramp time 2.5 sec. Tensile hold time 90 sec. 200 cycles. Nr = 330 cycles.
(After S. Vaynman, M. E. Fine, and D. A. leannotte lO )

age and annealing of such cavities. For example, the number of cycles to failure
in tests with equal tensile and compressive hold times for Rene 8041 and for
60% Sn-40% Pb solder3 was found to be close to the number of cycles to failure
in continuous fatigue cycling with the same period per cycle and the same plastic
strain (Fig. 11-17).
Compressive hold time alone in a cycle was found to be very damaging for
Rene 80 and nickel-base superalloy IN 738, both tested at 1600°F.41 .45 Since
elastic strain in these alloys is much higher than the plastic strain, a significant
mean stress (tensile in the case of compressive hold time in the cycle) is built
in these alloys leading to enhanced crack growth. The build up of a significant
mean stress is not a case for most metals, including solders.
A limited number of tension-tension fatigue tests perfonned with 96.5% Pb-
3.5% Sn solder at 0.75 percent total strain range in air under conditions of fast
ramp time (2.5 sec) and hold time at zero strain (compressive stress hold time)
indicated practically no effect of such hold on fatigue life.1O However, the
compressive hold dramatically reduced the number of cycles to failure for this
solder when combined with a tensile hold. Figure 11-18 shows how variable
compressive hold time affects the number of cycles to failure in tests with fixed
tensile hold time of 120 and 360 sec. Line 1 in this figure corresponds to tests
348 EFFECTS OF STRAIN RANGE, RAMP TIME, HOLD TIME, AND TEMPERATURE

o Continuous eyel ing /V"..

x Hold time n..n. o

Figure 11-17 Influence of cycling frequency and hold time on fatigue life of 60% Sn-40%
Pb solder. (After H. D. Solomon s3 )

with only tensile hold time. By using Equation (11-8) the limit for tensile hold
tests under these conditions was found to be approximately 200 cycles. However,
adding compressive hold reduced the fatigue life below this value. As in tests
with only tensile hold, reduction in Nf is greatest during the first few minutes
of compressive hold, and the addition of more compressive hold time leads to

750 AEt = 0.75%

r.::-.. o- tensile hold time only

~ 500 x - 120 sec tht plus variable the
(,)
~
l:l. - 360 sec I I . . II ..

&
_ 250
Z NIf
2
Nil
f
o~------~--------~--------~----~
400 800 1200
Total hold time per cycle (seq)
Figure 11-18 Effect of combined tensile and compressive hold time on number of cycles
to failure for 96.5% Pb-3.5% Sn solder. (After S. Vaynman, M. E. Fine, and D. A. leannotte lO)
 EFFECT OF HOLD TIME ON FATIGUE LIFE 349

another limit Nj, where Nj is the minimum number of cycles to failure all
specimens made of this solder will survive at any combination of tensile and
compressive hold times at given conditions of strain range, temperature, and
ramp time.
The times to failure for tests with combined tensile and compressive hold
times were plotted versus total time per cycle spent in tensile and compressive
hold (Fig. 11-19). While experimental results for tests with only tensile hold
time in the cycle fallon line 1, experimental results for tests with combined
tensile and compressive hold time (the) fallon line 2. It is evident from this
figure that a linear relation between the time to failure and combined hold time
per cycle exists provided that compressive hold time is not too small:

(11-9)

Procedures identical to those for obtaining Equations (11-7) and (11-8) lead
to Equations (11-10) and (11-11) provided total hold time per cycle is long
compared to ramp time.

(11-10)

Nt = D = Nj (11-11)

The Nj is determined by D, the slope of line 2 in Figure 11-19.

AEt =0.75% t, = 2.5sec

<3'
~150

-
'0

- o}
)( 75
.... X the sama
6. as In Fig. 18

o 500 1000 1500

Total hold time per cycle (sec)
Figure 11-19 Effect of combined tensile and compressive hold time on time to failure for
96.5% Pb-3.5% Sn solder. (After S. Vaynman, M. E. Fine, and D. A. Jeannotte lO )
350 EFFECTS OF STRAIN RANGE, RAMP TIME, HOLD TIME, AND TEMPERATURE

"0' 0.041-
~
0-
L.

W0.02r-
<l

O'r I

400
I
1200
I

f.. )800
tr ~ec

Figure 11·20 Increase in plastic strain due to increase in ramp time. 96.5% Pb-3.5% Sn
solder. Total strain range 0.75 percent, 25°C. (After S. Vaynman 54 )

It is generally considered that damage per cycle during cycling is proportional

to the amount of plastic deformation per cycle. The small changes in plastic
strain range per cycle (for example, for tests with 96.5% Pb-3.5% Sn solder at
0.75 percent total strain range discussed previously) with increasing ramp time
(Figure 11-20) or hold time (Figure 11-21) are not significant enough to account
for observed large, order of magnitude reductions in the number of cycles to
failure. It is also well know that creep-fatigue damage is very sensitive to strain
rate. As will be shown later, the strain rates in tests were varied over a few
orders of magnitude; therefore, changes in fatigue life of solder with increasing
ramp or hold times are attributed to the changes in strain rates. 54 The fact that
hold is much more destructive than ramp was assumed to be related to the much
lower strain rates operating during hold time.

~ 0.04
'-'"
...a.
~ 0.02 6Et =0.75%
25°C
t, = 2.5sec
o 600 900
(sec)
Figure 11-21 Relaxation strain per cycle versus tensile hold time. 96.5% Pb-3.5% Sn solder.
Total strain range 0.75 percent, 25°C. (After S. Vaynman 54)
 EFFECT OF HOLD TIME ON FATIGUE LIFE 351

The strain rates during ramps in no-hold tests were calculated as:

(11-12)

where E = strain rate

Et = total strain range

The stress relaxation data for 96.5% Sn-3.5% Pb solder4 (Fig. 11-22) were
used to calculate the strain rates operating during hold times. It is well known
that

(11-13)

where Ec = creep strain range

Ep = plastic strain range

Eel = elastic strain range

III
c
~7

6~ __~________~________~~
4 12 20
Time (min)

Figure 11-22 Stress relaxation in 96.5% Pb-3.5% Sn solder. Total strain range 0.75 percent,
25°C. (After S. Vaynman 54)
352 EFFECTS OF STRAIN RANGE, RAMP TIME, HOLD TIME, AND TEMPERATURE

During stress relaxation the total strain range is constant and plastic strain is
zero; therefore

Ec = -Eel (11-14)

0-
Since Eel -
- -
E (II-IS)

(J'
Ec = - E- (11-16)

For each test performed, the strain rates, which are calculated by Equations
(1l-12) and (1l-16), are plotted in Figure 11-23 together with the values of the
accumulated strains to failure.
It is obvious that the damage in continuous wave tests occurs because of
inelastic (plastic and creep) strains, which are a sum of plastic and creep strains,
both operating at the same strain rate. Therefore, accumulated strains to failure
for these tests were defined as

(11-17)

• - no hoi d tests
x- tensile hold tests
1600

t1.I1200
t..
:J

-o
..... 800
0
c
o
'-
iil 400

O~~==~~~~ ____~__~____~__~
-7 -6 -1
Log strain rate

Figure 11-23 Effect of strain rate on strain to failure for 96.5% Pb-3.5% Sn solder. Total
strain range 0.75 percent, 25°C. (After S. Vaynman 54 )
 EFFECT OF HOLD TIME ON FATIGUE LIFE 353

In hold tests with very short (2.5 sec) ramp time, it was assumed that the
material experiences almost all the damage during hold time because strain rates
during hold are a few orders of magnitude lower than during ramp. As a result,
accumulated strains to failure for hold time tests are defined as

(11-18)

where Epr = relaxation (creep) strain per cycle (shown in Fig. 11-24).
It is evident that strain to failure is a complex function of strain rate (Fig. 11-
23). At high strain rates, the accumulated strain to failure does not vary with
strain rate. When strain rate is reduced below 10-3 sec-I, strain to failure in tests
without hold time starts to drop steadily. No tendency for saturation in the values
of strain to failure is observed for these tests when time per cycle is increased
up to 1 h. However, at longer than 1 h time/cycle, saturation in strain to failure
may be expected. In hold time tests, strain to failure is very low. It drops very
fast and reaches the plateau at hold times of 30 sec. As shown in Fig. 11-23,
the strain rates in tests with almost any hold times are lower than those in
continuous wave tests with long ramp time over the ranges usually investigated.
This drastic difference in the strain rate magnitude explains why hold time has
a much stronger effect on the fatigue life of solder than ramp time.
The variation of strain to failure with strain rate for low-tin lead-base solder
is very similar to that found for a number of engineering alloys such as 304,
316 and 20%-25% Ni-Nb doped stainless steels at elevated temperatures46 and
depicted in general form in Figure 11-25; decreasing strain rate leads to reduction
in strain to failure from a high plateau level via a transition region to a low strain
to failure level at lower strain rates. This behavior is explained by considering
different cavity growth (creep) mechanisms that may operate at various strain

- - - °min

a£pr

Figure 11-24 Typical hysteresis curve for tensile hold time tests.
354 EFFECTS OF STRAIN RANGE, RAMP TIME, HOLD TIME, AND TEMPERATURE

Plastic growth

c
o
+'
c:
C
L
+' Const ra i ned
til
growth

Log strain rate (sec-1)

Figure 11-25 Typical strain rate-strain to failure relation for a number of commercial alloys.
(After D. A. Miller, R. H. Priest, and E. G. Ellison 46)

rates.46 At high strain rates, cavity growth occurs by defonnation and at these
strain rates an alternative mechanism of fracture, such as transgranular fracture,
may operate. Both transgranular and intergranular modes of fracture are oper-
ational in tests with 96.5% Pb-3.5% Sn solder at high strain rates. 1O •12 When
the strain rate is reduced, the cavity growth mechanism changes from defonnation
to unconstrained (transition region in Figs. 11-23 and 11-25) and constrained
cavity growth mechanisms (lower plateau in Figs. 11-23 and 11-25). Constrained
cavity growth mechanism is considered to be very efficient in converting plastic
strain into cavities (Le., to be the most damaging mechanism during fatigue).
Fonnation of voids in low-tin lead-based solders was detected during fatigue
tests of low-tin lead-based solder. 51.52
In an electronic device, the strain rates below 10-7 sec- l operate during hold
times; therefore, the most damaging constrained cavity growth mechanism is
expected to be operational under these conditions. Solder joints experience strain
rates of an order of 104 to 10-5 sec- l during ramping and then less damaging
unconstrained cavity growth mechanism leads to solder fatigue failure. Thus,
the most attention in solder fatigue research should be paid to the hold time,
during which solder experiences the most damage, and therefore fatigue life
prediction methods should be based on hold time tests.
 EFFECT OF TEMPERATURE ON ISOTHERMAL FATIGUE OF SOLDERS 355

11.5 EFFECT OF TEMPERATURE ON ISOTHERMAL FATIGUE

OF SOLDERS

As a rule the isothermal fatigue life of metals decreases as temperature increases.

However, the degree of fatigue life change depends on material and testing
conditions. For near eutectic tin-lead solders (63% Sn-37% Pb!3 in the 25 to
80°C, 60% Sn-40% Pb in 35 to 150°C,4.5 and - 60 to 150°C9 temperature ranges)
the temperature effect is insignificant for all strain ranges and modes of loading
investigated.
The effect of temperature on low-tin lead-based solder is much more pro-
nounced than for near-eutectic tin-lead solders. The fatigue life of 95% Pb-5%
Sn solder at strains lower than 1.0 percent was found to be much smaller at
150°C than at room temperature or - 6O°C. 9 The fatigue life of 95% Pb-5% Sn
solder was shown to decrease when temperature was increased from room tem-
perature to 1200C. 14 The fatigue life for this solder at 120 at 150°C was shown
to be practically the same.
Fatigue life of 96.5% Pb-3.5% Sn solder at 80°C is practically the same as
at 100°C (Fig. 11-26).10 The fatigue life of this solder was found to be the lowest
at these temperatures. A gradual decrease in fatigue life with increasing tem-

T Coe)
100 80 50 25 5

t r = 0.1 - 2.5 sec

0.0030 0.0034
1fT
Figure 11-26 Effect of temperature on fatigue life of 96.5% Pb-3. 5 % Sn solder in tests
without hold time. Ramp time 0.1 to 2.5 sec. (After S. Vaynman, M. E. Fine, and D. A.
Jeannotte ll )
356 EFFECTS OF STRAIN RANGE, RAMP TIME, HOLD TIME, AND TEMPERATURE

Table 11-3 Apparent Activation Energy for 96.5% Pb-;3.5% Sn Solder. 25 to

80°c' 0.1-2.5-sec Ramp Time. No Hold. l l
Total Strain Apparent Activation
Range (%) Energy, (KJ/mol)
0.30 22.7
0.40 22.7
0.50 30.1
0.60 40.8
0.75 46.0

perature was observed in the 25 to BOaC region; however, fatigue life decreased
when temperature was further reduced to SoC. The solder fatigue life dependence
on temperature appears to follow approximately an Arrhenius equation only
between 25 and BOaC. The apparent activation energy for this temperature range
is summarized in the Table 11-3. As is evident from this table, the lowest apparent
activation energy is associated with grain boundary failure at low strains (Fig.
11-5).
At higher strain (0.6 percent) the temperature effect on fatigue life of this
low-tin lead-based solder is a function of frequency (Fig. 11-9) and hold time
(Fig. 11-27). The change in the number of cycles to failure with temperature at
the frequency less than 10-2 Hz is less than that at higher frequency (Fig. 11-
8). This converts to an apparent activation energy of 24 to 26 Kllmol at low
frequency and 40.6KJ/mol at high frequency testing conditions.

80 25

'0'

-
Q)
..$E./ 4
10
+- o - no hold
+- 15 sec tensile hold
0- 30sec " "
x-120sec
A-360sec II "

0.0029 0.0031 0.0033

1fT
Figure 11-27 Effect of temperature on fatigue life of 96.5% Pb-3.5% Sn solder in tests with
and without tensile hold time. Total strain range 0.60 percent. 0 = no hold; + = 15-sec;
o = 30-sec; x = 120-sec; 6. = 360 sec-hold time. (After S. Vaynman, M. E. Fine, and
D. A. leannottel l )
 REFERENCES 357

The temperature effect on fatigue life of low-tin lead-based solder is different

in tests with hold time and without hold time(0.6 percent total strain range, 2.5-
sec ramp time; Fig. 11-27); the effective activation energies are 24 KJ/mol and
40 KJ/mol, respectively.
Thus, grain boundary fracture in low-tin lead-based solder under low strain
rate conditions is characterized by low activation energy.

11.6 CONCLUSION

There is no unique relationship describing the dependence of isothermal solder

fatigue life on variables for all solder alloys. This is because of differences in
composition and microstructure. The relationship should be determined exper-
imentally for each solder material.
The following is the summary of the isothermal fatigue behavior of solder
materials:

1. While at strain ranges over 1 percent the Coffin-Manson relation may be

used for extrapolation of isothermal solder fatigue data; this relation is
found to fail at lower strains due to the strong creep-fatigue interaction in
solders during cycling.
2. Increasing time per cycle over approximately 60 sec in no-hold tests leads
to reduction in the number of cycles to failure.
3. Tensile hold time reduces the number of cycles to failure and at longer
hold times leads to the saturation in the number of cycles to failure.
4. The effect of temperature on isothermal fatigue of tin-lead near-eutectic
solders is negligible. Temperature affects the fatigue life of high-lead sol-
ders. The effect of temperature on fatigue life of these solders is a function
of testing conditions.

REFERENCES

1. Shine, M. C., and L. R. Fox, "Fatigue of Solder Joints in Surface Mount Devices,"
Low Cycle Fatigue, ASTM STP 942, H. D. Solomon et aI., Eds. Philadelphia,
ASTM, 1988, pp. 588-610.
2. Solomon, H. D., "Low Cycle Fatigue of Surface Mounted Chip CarrierlPrinted
Wiring Board Joints," Proceedings of39th Electronic Components Conference, 1989,
pp. 277-292.
3. _ _ , "Influence of Temperature on the Low Cycle Fatigue of Surface Mounted
Chip Carrier/Printed Wiring Board Joints," GE Information Series, September 1987.
4. _ _ , "Low-Cycle Fatigue of 60/40 Solder-Plastic Strain Limited vs. Displacement
Limited Testing," in Electronic Packaging: Materials and Processes, J. A. Sortell,
Ed., ASM, 1986, pp. 29-49.
358 EFFECTS OF STRAIN RANGE, RAMP TIME, HOLD TIME, AND TEMPERATURE

5. _ _ , "Low-Frequency, High-Temperature Low-Cycle Fatigue of 6OSnl40Pb Sol-

der," Low Cycle Fatigue, ASTM STP-942, H. D. Solomon et aI., Eds. Philadelphia,
ASTM, 1988, pp. 342-369.
6. Frear, D., D. Grivas, and J. W. Morris, Jr., "Thermal Fatigue Failures in Solder
Joints," Journal of Metals, 40, 1988, pp. 18-22.
7. Frear, D., D. Grivas, M. McCormack, D. Tribula, and J. W. Morris, Jr., "Fatigue
and Thermal Fatigue Testing of Pb-Sn Solder Joints," Proc. 3d Ann. Electronic
Packaging and Corrosion in Microelectronics Con/., M. E. Nicholson, Ed., ASM
International, 1987, pp. 269-274.
8. _ _ , "Fatigue and Thermal Fatigue ofPb-Sn Solder Joints," Proc. Effect ofLoad
and Thermal Histories on Mechanical Behavior Symp., P. K. Liaw and T. Nicholas,
Eds., 1987, pp. 113-126.
9. Kitano, M., T. Shimizu, and T. Kumazava, "Statistical Fatigue Life Estimation:
The Influence of Temperature and Composition on Low-Cycle Fatigue of Tin-Lead
Solders," Current Japanese Materials Research, 2, August 1987, pp. 235-250.
10. Vaynman, S., M. E. Fine, and D. A. Jeannotte, "Isothermal Fatigue of Low Tin
Lead Based Solder," Metall. Trans., 19A, 1988, pp. 1051-1056.
11. _ _ , "Prediction of Fatigue Life of Lead-Base Low Tin Solder," Proceedings of
37th Electronic Components Conference, IEEE 1987, pp. 598-603.
12. _ _ , "Isothermal Fatigue Failure Mechanisms in Low Tin Lead Based Solder,"
Proc. Effect of Load and Thermal Histories on Mechanical Behavior Symp., P. K.
Liaw and T. Nicholas, Eds., 1987, pp. 127-137.
13. Vaynman, S., and M. E. Fine, "Fatigue of Low-Tin Lead-Based and Tin-Lead
Eutectic Solders," Microelectronic Packaging Technology: Materials and Processes.
Proc. of the 2d ASM International Electronic Materials and Processing Congress,
W. T. Shieh, Ed., ASM, 1989, pp. 255-259.
14. Rathore, H. S., R. C. Yih and A. R. Edenfeld, "Fatigue Behavior of Solders Used
in Flip-Chip Technology," J. Test. Eval., 1, 1978, pp. 170-178.
15. Solomon, H. D., "Room Temperature Low Cycle Fatigue of a High Pb Solder
(INDALLOY 151)," Microelectronic Packaging Technology: Materials and Pro-
cesses, Proc. of the 2d ASM International Electronic Materials and Processing
Congress, W. T. Shieh, Ed., ASM, 1989, pp. 135-146.
16. Wild, R. N., "Fatigue Properties of Solder Joints," J. of Welding Research, 51,
1972, pp. 521s-526s.
17. Coffin, L. F., "A Study of the Effect of Cyclic Thermal Stress on a Ductile Metal,"
Trans, ASME, 76, 1954, pp. 931-950.
18. Manson, S. S., "Fatigue: A Complex Subject-Some Simple Approximations,"
Experimental Mechanics, 5,(7), 1965, pp. 193-226.
19. Eckel, J. F., "Influence of Frequency on the Repeated Bending Life of Acid Lead,"
Proc. ASTM, 51, 1957, pp. 745-760.
20. Solomon, H. D., "Frequency Dependent Low Cycle Fatigue Crack Propagation,"
Metall. Trans., 3, 1972, pp. 341-347.
21. James, L. A., "Effect of Frequency upon the Fatigue Crack Growth of Type 304
Stainless Steel at 1000°F," Stress Analysis and Growth of Cracks, ASTM STP 513,
ASTM, 1972, pp. 218-229.
 REFERENCES 359

22. Conway, J. R., J. T. Berling, andR. H. Stentz, "Strain Rate and Hold Time Saturation
in Low Cycle Fatigue: Design Parameter Plots," Fatigue at Elevated Temperatures,
ASTM STP 520, ASTM, 1973, pp. 637-647.
23. Manson, S. S., "The Challenge to Unify Treatment of High Temperature Fatigue-
A Partisan Proposal Based on Strainrange Partitioning," Fatigue at Elevated Tem-
pertures, ASTM STP 520, ASTM, 1973, pp. 744-782.
24. Sidney, D., and L. F. Coffin, "Low-Cycle Fatigue Damage Mechanisms at High
Temperature," Fatigue Mechanisms, ASTM STP 675, ASTM, 1979, pp. 528-568.
25. Hales, R., "A Quantitative Metallographic Assessment of Structural Degradation of
Type 304 Stainless Steel During Creep Fatigue," Fat. Eng. Mater. Struct., 3, 1980,
pp. 339-356.
26. Baik, S., and R. Raj, "Mechanisms of Creep-Fatigue Interaction," Metall. Trans.,
13A, 1982, pp. 1215-1221.
27. Tien, J. K., S. V. Nair, and V. C. Nardone, "Creep-Fatigue Interaction in Structural
Materials," Flow and Fracture at Elevated Temperatures, Ed. R. Raj ASTM, 1985,
pp. 179-214.
28. Raj, R., "Mechanisms of Creep-Fatigue Interaction," Flow and Fracture at Elevated
Temperatures, ASTM, 1985, pp. 215-249.
29. Gohn, G. R., and W. C. Ellis, "The Fatigue Tests as Applied to Lead Cable Sheath,"
Proc. ASTM, 51, 1951, pp. 721-740.
30. Solomon, H. D., and L. F. Coffin, "Effect of Frequency and Environment on Fatigue
Crack Growth in A286 at 1100°F," Fatigue at Elevated Temperatures, ASTM STP
520 ASTM, 1973, pp. 112-122.
31. Forrest, P. G., Fatigue of Metals, New York, 1962, Pergamon Press.
32. Wareing, J., "Fatigue Crack Growth in a Type 316 Stainless Steel and a 20% Cr/25%
NilNb Stainless Steel at Elevated Temperature," Metall. Trans., 6A, 1975, pp. 1367-
1377.
33. _ _ , "Creep-Fatigue Interaction in Austenitic Stainless Steel," Metall. Trans,
8A, 1977, pp. 711-721.
34. Maiya, P. S., and S. Majumdar , "Elevated-Temperature Low-Cycle Fatigue Behavior
of Different Heats of Type 304 Stainless Steel," Metall. Trans., 8a, 1977, 1651-
1660.
35. Majumdar, S., and P. S. Maiya, "A Mechanistic Model for Time-Dependent Fa-
tigue," Eng. Mater. Technology, 102, 1980, pp. 159-167.
36. Day, M. F., and G. B. Thomas, "Microstructural Assessment of Fractional LIfe
Approach to Low Cycle Fatigue at High Temperatures," Metall. Sci., 13, 1979, pp.
25-33.
37. Meleka, A. R., "Combined Creep and Fatigue Properties," Metallurgical Reviews,
7, 1962, pp. 43-93.
38. Ellozy, A. R., P. M. Dixon, and R. N. Wild, "Combined Low-Cycle Fatigue and
Stress Relaxation of Some Pb-In and 96/4 PbSn Alloy at Room Temperature," Proc.
of the 2d International Coriference on Mechanical Behavior of Materials, 1976, pp.
903-907.
39. Shah, H. J., and J. H. Kelly, "Effect of Dwell Time on Thermal Cycling of the
Flip-Chip Joint," ISHM, 1970, pp. 3.4.1-3.4.6.
360 EFFECTS OF STRAIN RANGE, RAMP TIME, HOLD TIME, AND TEMPERATURE

40. Agarwala, B. N., "Thermal Fatigue Damage in Pb-In Solder Interconnections," 23rd
Annual Proceedings, Reliability Physics IEEElIRPS, 1985, pp. 198-205.
41. Lord, D. C., and L. F. Coffin, "Low Cycle Fatigue Hold Time Behavior of Cast
Rene'80," Metall. Trans., 4A 1973, pp. 1647-1654.
42. Teranishi, H., and A. J. McEvily, "The Effect of Oxidation on Hold Time Fatigue
Behavior of 2.25Cr-1Mo Steel," Metall. Trans., lOA, 1979, pp. 1806-1808.
43. Wells, C. H., and C. P. Sullivan, "Interaction Between Creep and Low Cycle Fatigue
in Udimet 700 at 1400"F," Fatigue at Elevated Temperatures, ASTM STP 520,
ASTM, 1973, pp. 59-74.
44. Sadananda, K., and P. Shaninian, "Hold Time Effects in High Temperature Fatigue
Crack Growth in Udimet 700," J. Mater. Sci., 13, 1978, pp. 2347-2357.
45. Ostergren, W. J., "A Damage Function and Associated Failure Equations for Pre-
dicting Hold Time and Frequency Effects at Elevated Temperature, Low Cycle
Fatigue," J. of Testing and Evaluation, JTEY, 4, 1976, pp. 327-339.
46. Miller, D. A., R. H. Priest, and E. G. Ellison, "A Review of Material Response
and Life Prediction Under Fatigue-Creep Loading Conditions," High Temp. Mater.
and Processes, 6, 1984, 155-194.
47. Hong, J. W., S. W. Nam, and K-T. Rie, "A Model for Life Prediction in Low-
Cycle Fatigue with Hold Time," J. Mater. Sci., 20, 1985, pp. 3763-3779.
48. Yaynman, S., "On Fatigue Life Prediction Under Conditions of Tensile Hold,"
Scripta Metall., 21, 1987, pp. 1385-1387.
49. Skelton, R. P., Fatigue at High Temperature, London and New York, Applied
Science Publishers, 1983.
50. Berriche, R., S. Yaynman, M. E. Fine, and D. A. Jeannotte, "The Effect of En-
vironment on Fatigue of Low Tin Lead Base Alloys," Proc. 3d Ann. Electronic
Packaging and Corrosion in Microelectronics Conf., M. E. Nicholson, Ed., ASM
International, 1987, pp. 169-174.
51. Berriche, R., "Environmental Effects on Time Dependent Fatigue of Low Tin-High
Lead Solder," Ph.D. Thesis, Northwestern University, Evanston, n. 1989.
52. Greenwood, P. J., T. C. Reiley, Y. Raman, andJ. K. Tien, "Cavitation in a Pb/Low-
Sn Solder During Low Cycle Fatigue," Scripta Metall., 22, 1988, pp. 1465-1468.
53. Solomon, H. D., "The Influence of Hold Time and Fatigue Cycle Wave Shape on
the Low Cycle Fatigue of 60/40 Solder," 38th Electronic Components Conference,
IEEE, 1988, pp. 7-12.
54. Yaynman, S., "Effect of Strain Rate on Fatigue of Low-Tin Lead-Base Solder," 39th
Electronic Components Conference, IEEE, 1989, pp. 273-276.
 12
A Damage Integral Methodology for
Thermal and Mechanical Fatigue of
Solder Joints
Che-Yu Li, R. Subrahmanyan, l.R. Wilcox,
and D. Stone

The differential thermal expansion of the connected members of a soldered

assembly during temperature cycles produces mechanical displacements in solder
joints. The resulting cyclic stresses are the driving force for damage processes
that eventually cause fatigue failures. Conventional approaches to solder joint
fatigue rely on correlations between fatigue life of the solder joint and some
measurable or calculable parameter characteristic of the fatigue loading. These
approaches often require extensive fatigue data to establish such correlations.
Furthermore, their validity is often limited to the range over which the parameters
in the correlation are uniquely related to the stress in the solder joint. The plastic
strain range is an example of such a parameter. For a given plastic strain range,
the stress in the solder joint may vary, however, depending on the displacement
rate, such that the correlation between the plastic strain range and the number
of cycles to failure is not unique.
A variety of damage processes such as wedge cavitation, I grain boundary
voiding, 2 and crack growth 3 ,4 have been found to be operative. The accumulation
of fatigue damage is manifested as a reduction in the load bearing area of the
solder joint, which eventually leads to failure. The task of describing the ac-
cumulation of fatigue damage is essentially two-fold. First, the time profile of
the stresses due to the applied displacement profile must be known. Second, the
rate of the damage process as a function of stress must be estimated. The fatigue
life of the solder joint may then be determined by the integration of the rate of
damage process to some specified failure criterion. This is the basic concept of
a damage-integral approach. 5,6

361
362 A DAMAGE INTEGRAL METHODOLOGY

The total applied displacement is accommodated by both the solder joint and
other connecting members of the soldered assembly. The connecting members
usually deform elastically; hence the load supported by the solder joint is influ-
enced by their compliance. The solder joint, however, can also deform inelast-
ically. Therefore, the stress in the solder joint can depend on the applied dis-
placement rates and temperature. The damage processes mentioned above determine
the rate of fatigue damage and its dependence on the stress state, temperature,
and environment. While the rate of internal damage processes such as wedge
cracking and cavitation is determined by the stress state and temperature alone,
the growth of surface cracks is additionally influenced by the external environ-
ment.
As discussed previously, when the stress is a unique function of a particular
parameter, such as plastic strain range, it may be possible to correlate the fatigue
life of the solder joint in terms of that parameter. A good example is the Coffin-
Manson approach,7,8 which correlates isothermal mechanical fatigue life to the
plastic strain range, This approach is often used to describe low cycle fatigue
behavior of copper and steel at room temperature. The success of this approach
to such applications can be attributed to the fact that at low homologous tem-
peratures the strain rate dependence of stress is small; therefore the stress profile
is essentially determined by the applied plastic strain range. For these conditions
it is possible to show that the accumulated fatigue damage per cycle is a nearly
single-valued function of the plastic strain range. 9,10
At temperatures above roughly a third of the melting point, typical in solder
joint fatigue, the stress will not only depend on the plastic strain but also on the
plastic strain rate. Hence, the fatigue life becomes sensitive to the time profile
of the applied displacement. In fact, it has been experimentally shown ll - 13 that
for such conditions the number of cycles to failure depends on the frequency
and hold time at the displacement limits. For a given plastic strain range, de-
creasing frequency or increasing the hold time can result in a smaller number
of cycles to failure.
Thermal fatigue adds another level of complexity. In addition to the expansion
mismatch-induced displacements, the temperature can change dynamically as a
function of time. Since both the rate of inelastic deformation and the various
damage processes are temperature dependent, the calculation of the stress and
the damage rates must further include the effects of temperature cycles. During
such conditions, a Coffin-Manson-type relation in its original form may not be
able to adequately characterize the low cycle fatigue behavior (e.g. reference I4).
A damage integral approach, on the other hand, provides a convenient frame-
work to account for the aforementioned complexities. The time profile of the
stress and the damage rates are explicitly considered in such an approach. Since
the time variation of both the stress and the damage rate for arbitrary applied
displacement profiles can be complex, the use of damage-integral methods often
requires numerical computations.
 INELASTIC DEFORMATION AND STRESS CALCULATION 363

Damage integral approaches were first applied to isothermal "high cycle fa-
tigue" (see e.g., references 15, 16). During high cycle fatigue, since the bulk of
the specimen deforms elastically, the stress in the material is well defined.
Furthermore, the failure occurs only by crack growth. The phenomenon of crack
growth under mostly elastic loading has been extensively studied and is well
described by a fracture mechanics approach.
The application of the same approach to low cycle fatigue, however, is not
straightforward. This is to a large extent because the description of the stress at
temperatures high relative to the melting point is not simple. They are a sensitive
function of the specimen microstructure and composition, stress, and tempera-
ture. High temperature phenomena such as grain boundary sliding,2,17,18 grain
growth and the coarsening of dispersed phases,19 and solute strengthening20 can
significantly influence the inelastic deformation behavior. In addition, the relative
importance of individual damage processes and their interactions, particularly
for nonmonotonic loading, is not well established.
In this chapter, some basic concepts involved in the application of a damage
integral approach to solder joint fatigue are discussed. We shall show that a
phenomenological approach to the description of the inelastic deformation prop-
erties, such as a state variable approach, can be successfully applied to the
calculation of the solder joint stress profile. An equivalent approach is needed
to calculate the fatigue damage. Some preliminary efforts toward this latter
direction are presented here.

12.1 INELASTIC DEFORMATION AND STRESS CALCULATION

12.1.1 Governing Equation for Solder Stress

The solder joint can deform both elastically and inelastically. For any given
displacement profile, the stress in the solder joint is dependent on the compliance
of the assembly (including the solder joint) and the inelastic deformation prop-
erties of the solder joint. We can write the total displacement rate as the sum
of the total elastic displacement rate (in the solder joint and assembly) and the
displacement rate corresponding to the inelastic deformation of the solder joint.
A governing equation for the load P supported by the solder joint can be written
as 21

(12-1)

where k is the total displacement rate and k; is the inelastic displacement rate.
S, the total stiffness, relates the total elastic displacements to the load. It incor-
porates the stiffness of both the solder joint and the connecting members and is
364 A DAMAGE INTEGRAL METHODOlOGY

dependent on their geometry, moduli, and the loading mode. It may be exper-
imentally measured or calculated.
For the case of shear loading, the solder joint height h and area A remain
constant. Equation (12-1) can be rewritten in terms of the shear stress rate +and
the inelastic shear strain rate -y as

+A = S(X - -yh) (12-2)

Consider the thermal cycling of a solder joint as shown schematically in Figure

12-1. The applied displacement rate for this case arises from the differential
thermal expansion between the members of the soldered assembly; therefore X
is given, for example, by X = anT~ where an is the difference in thermal
expansion coefficients, T is the heating or cooling rate, and ~ is the distance
from the neutral point.
The calculation of the stresses in the solder joint for an arbitrary loading profile
requires the integration of Equation (12-2). The dependence of stress on the
inelastic strain rate and temperature must be known in the form of appropriate
constitutive relations to facilitate the integration. Experiments such as displace-
ment controlled cyclic loading, creep, and load relaxation are often used to
characterize the inelastic deformation properties of solder joints to establish
required constitutive relations. Inadequate constitutive relations can greatly in-
crease the margin of error in the stresses and, therefore, the damage rates in the
solder joints.
The inelastic deformation behavior of the solder joint can be significantly
affected by the composition and microstructure of the solder joint. These are in
tum changed by the cooling rates during solder joint reflow and the dissolution
of the adjoining metallization. The thermal history of the solder joint can sig-
nificantly influence the stability of the solder joint microstructure. 4 ,22.23 In ad-
dition, the deformation properties can depend on the contribution of multiple
mechanisms. The relative contribution of each varies depending on the stress

::'L:r=====i====:::I~=;SOLDER
JOINT

Figure 12-1 Schematic of lead less surface mount solder joint.

 INELASTIC DEFORMATION AND STRESS CALCULATION 365

and temperature as well as microstructure and composition. A comprehensive

theoretical basis is yet to be developed for the establishment of the required
constitutive relations.

12.1.2 Inelastic Deformation Behavior and Constitutive Relations

In addition to solder joint-specific concerns discussed above, some basic issues

remain to be addressed for the development of appropriate constitutive equations
for inelastic deformation at high homologous temperatures. At these temperatures
thermally activated structural and compositional changes usually accompany
inelastic deformation. It is, therefore, necessary to account for the effect of such
thermo-mechanical history on the deformation properties. On the other hand,
the need for reducing experimental effort has always been of paramount impor-
tance. A phenomenological framework is therefore needed to deduce constitutive
relations based on a minimum amount of experimental effort.
A state variable approach originally proposed by Hart24 for macroplastic de-
formation is preferred in the present work. Hart's model was later extended by
Jackson et al. 25 to describe transient deformation phenomena such as micro-
plasticity and the Bauschinger effect. Although the initial models treated inelastic
deformation due to dislocation-based processes, other deformation processes such
as grain boundary sliding 26 can also be included in a state variable treatment. It
has been shown l7 that a variety of test results involving inelastic deformation in
solders and solder joints, obtained by load relaxation, creep, and displacement-
controlled cyclic loading experiments can be represented in a unified manner
using a state variable concept. This demonstrates the applicability and versatility
of constitutive relations based on Hart's concept even in the presence of the
various complexities discussed previously.
A key feature of Hart's formulations for inelastic deformation is the presence
of a microstructural and mechanical state of the material. This state is uniquely
specified by the current deformation properties, for example, by the isothermal
stress-inelastic strain behavior. The deformation properties can be scaled from
one mechanical state to another at a given temperature or from one temperature
to another at a given mechanical state. It follows automatically that experimental
effort required to determine the deformation properties over a wide range of
states and temperatures can be reduced.
At high homologous temperatures deformation processes involving grain
boundary phenomena wiII be active as will dislocation-based processes in the
grain matrix. These phenomena include grain boundary sliding and migration.
We have assumed that the inelastic deformation of the solder joint, represented
in terms of tensor invariant, can be described by the rheological model shown
in Figure 12-2. This model assumes that steady-state macroplastic deforIilation
and transient deformation are separable. By transient deformation we refer to
 w
0'
0'

)-
o)-
3:
)-
C)
m
z-+
m
Inelastic C)

a2 CX 2 deformation s:
r
3:
m
Elastic -+
I
deformation oo
or
CX 1 (J oC)
/ - - - - ...... - - - - - -..---~ -<

a1

.....1- E: =-1

Figure 12-2 A rheological model representing the deformation behavior of solders and
solder joints.
 INELASTIC DEFORMATION AND STRESS CALCULATION 367

both anelastic (recoverable but time dependent) deformation and microplastic-

ity.25 Furthermore, for the present purposes, it is assumed that these elements
represent the combined contribution of the grain boundary and grain matrix
processes. The inelastic strain rate due to an applied stress rr is represented by
E. The macroplastic deformation behavior is represented by the element al' The
springs al and a2 represent the anelastic deformation; MI and M2 are the anelastic
moduli. The element a2 represents the microplastic element.
The operation of the rheological model may be illustrated based on a stress-
strain loop obtained during isothermal displacement controlled cycles (Figure
12-3). Upon loading at point A, only the elastic spring and the anelastic springs
are active. Therefore, the initial linear slope represents the combined elastic and
anelastic moduli. Initial deviation from linearity occurs when the microplastic
element becomes active (at point B). Further increasing the stresses will even-
tually cause the solder joint to flow macroplastically (point C). A similar behavior
is observed for loading in the reverse direction (along points D, E, F, and A).
Accurate and comprehensive experimental data are required to establish the
constitutive relations to describe deformation behavior such as that shown in

30 I I I I I

-
C. ..... ••••
20 0 -
"....... •••:
0
0... •• •• ••
10 - •• ••• -
•••
~
-....- •
l ••
en •• •••
en •
Q) 0 fo-
81• :'E -
+'
L
:• •
••
en •• ••
•• I•
•••
· •• -
L
-10
A"·...... •••••• .. -F
0 fo-
• ••
Q)
..c • •• ~
en
-20 '- -

I I I I I
-30
-3 -2 -1 0 1 2 3
shear strain (percent)
Figure 12-3 Measured room temperature stress-strain loop for surface mount type solder
joints 21 during isothermal cycling. This loop generated using a triangular wave form with a
cyclic period of 100 sec.
368 A DAMAGE INTEGRAL METHODOlOGY

Figure 12-3. In the following, an experimental method is considered that has

been useful in characterizing macroplastic deformation properties of solders and
solder joints over a wide range of loading conditions. We shall show that even
with the presence of several deformation mechanisms, simple and yet accurate
phenomenological constitutive relations can be established. The considerations
involved in the deconvolution ofthe individual deformation mechanisms involved
are discussed in references 27 .!7 and are outside the scope of this chapter.
Wilcox!7 has used load relaxation experiments to determine isostructural (con-
stant mechanical state) deformation properties of the macroplastic element. The
bulk eutectic alloy was prepared using high purity Sn and Pb. A cylindrical
tensile sample, with a gauge length and area of 20.5 mm and 32.2 mmZ, re-
spectively, was used to perform load relaxation and constant load creep exper-
iments at several temperatures. Initial tests were performed after ageing at the
maximum temperature of interest (125°C) for 24 h. Subsequent tests were per-
formed at progressively decreasing temperatures.
In each load relaxation experiment, the sample was typically loaded to an
additional 2 percent strain. The cross head was then fixed. The load and the
sample displacements, measured as a function of time, were used to determine

2.0~-----r------~----~------~----~

1.5

~ 1.0
~
(/)
~ 0.5
....
+-'
.!!!-
g> 0.0

-0.5
•
-1.0 1...-_ _--1._ _ _.....1.._ _ _........._ _ _...&...-_ _- - - '

-10 -8 -6 -4 -2 o
log (inelastic strain rate, 1/sec)
Figure 12-4 Inelastic deformation behavior of bulk eutectic tin-lead solder measured by
tensile load relaxation experiments.
 INELASTIC DEFORMATION AND STRESS CALCULATION 369

the stress and the inelastic strain rate of the solder specimen. The significance
of stress-strain rate relations obtained from the load relaxation experiments is
contingent on the stability of the microstructure throughout the relaxation run.
The validity of a constant structure assumption was confirmed by reloading
relaxation experiments. 28.17 Except at - 25°C, the extent of work hardening
during loading was found to be negligible. The measured constant structure
stress-inelastic strain rate data at various temperatures are presented in Figure
12-4.
The data in Figure 12-4 represent the dependence of the macroplastic flow
stress on the inelastic strain rate. It is seen that at each temperature, the slope
of the stress-inelastic strain rate curves is not constant. It is small at high strain
rates, then increases at intermediate strain rates, and decreases at small strain
rates. The specimen surface after completion of all the relaxation experiments
showed steps at the boundaries of the dispersed phase colonies and at the Pb-
Sn phase boundaries, suggesting boundary sliding at these sites. It was concluded 17
that a steady-state combination of boundary migration and sliding contributes to
the observed sigmoidal shape of the stress-inelastic strain rate curves. It is
important to recognize that the shape of the stress-inelastic strain rate curves in
Figure 12-4 is essentially the same for almost the entire temperature range. This
observation is an important basis for the temperature scaling approximations
made in this work and in Hart's phenomenological approach.
For the present purposes, we shall describe the macroplastic deformation
behavior represented in Figure 12-4 by the macroplastic element in Figure 12-
a
2. The macroplastic strain rate is expressed as a function of the stress (J'through
a constitutive relation of the form

(12-3)

where (J'* is the hardness parameter of the grain matrix, E* is the rate parameter,
and A is a shape parameter. As seen by Kashyap and Murty29 in their creep data,
different activation energies were found for different regimes of strain rate and
temperature. Using an activation energy Qd of 46 KJ/mol, the stress-strain rate
curves in Figure 12-4 have been translated to 25°C to illustrate the temperature
scaling described previously [~ln(E) = - (QjR) ~ (liT) for a given mechanical
state]. The result, shown in Figure 12-5, indicates reasonable temperature scal-
ing. This temperature scaling is facilitated by the fact that work hardening has
been found to be unimportant 17 in the temperature range involved for this solder
composition and microstructure. Simplification of the description of the inelastic
deformation behavior over the range of measured data is thus possible although
a variety of deformation mechanisms and related phenomena are involved.
It is similarly possible, using isothermal load relaxation and displacement
controlled cycling experiments, to determine the constitutive relations for the
370 A DAMAGE INTEGRAL METHODOlOGY

2.0~----r-----~----~----~----~----~

1.5 62% Sn - 38% Pb

........
ro 1.0
a.
~
(/)
0.5
-
(/)

....Q)
• -25°C
(/)
'-"
Cl
0 0.0 25°C
A~ 0

•.'
A 75°C
-0.5
• • 125°C
-1.0 I..-_ _ ~ _ _...I....._ _- - i -_ _- - ' -_ _- - " '_ _- - - '

-12 -10 -8 -6 -4 -2 o
log (inelastic strain rate, 1/sec)
Figure 12-5 Stress inelastic strain rate data for eutectic tin-lead solder (Figure 12-4) shifted
to 25°C from the actual test temperatures using a thermal activation energy of 46 KJ/mo!.

anelastic spring elements and the microplastic element in Figure 12-2. A detailed
description of experimental methods is given in the reference. 28

12.1.3 Stress Calculation

The applicability of the phenomenologically based constitutive relations devel-

oped above can be demonstrated by solder joint stress calculations during iso-
thermal displacement controlled cycling. Solomon30 •31 has reported such data
for 60% Sn-40% Pb solder joints obtained under total displacement control and
plastic strain limits. The solder joints, which connect brass or copper loading
blocks, have a nominal area and height of 32.25 mm2 and 0.19 mm, respectively.
He has reported stress-strain loops at several temperatures and plastic strain
ranges. The measured stress loops were simulated by numerical integration of
Equation (12-2) and the constitutive relations described above. The shear stresses
and inelastic shear strain rates are converted to their tensor invariants as (J' =
v3T and E = .y/v3. Details of the stress calculation are provided in the ref-
erence. 21 Figure 12-6 compares the results of the numerical integration to the
 INELASTIC DEFORMATION AND STRESS CALCULATION 371

measured data at two temperatures 150 and 35°C. An adequate description of

isothermal stress cycles appears to be possible. In a separate work21 ,l0 it was
shown that it may be possible to apply the constitutive relations similar to those
measured in bulk eutectic solder to small solder joints. This result is consistent
with the stress calculations described here.
It is also possible to use the same constitutive relations to calculate the stresses
during a thermal cycle. To simulate the thermal stress cycles in surface mount
solder joint used by Hall in Figure 12-7, his bimetallic strip approximation is
used (see Chapter 10 and the reference 32 ) to determine the assembly stiffness S
= 0.28 MN/m. The loading parameters are chosen to match his experimental
conditions. 34 Equation (12-2) is numerically integrated along the thermal profile
including the peak temperature holds. The resulting stress simulation is that
shown by the solid line in Figure 12-7. The calculated stress shows good agree-
ment with the experimental data at all temperatures.

40 ,..---r--,..--,...--,

35C 150C

20 20

-
a.
c
:::::i!
'-'
-
a.
c
:::::i!
'-'
1/1 1/1
1/1 1/1

....
~ 0 0 ....f!
1/1 1/1
L.. L..
C C
Cl) Cl)
.r:. .r:.
1/1 1/1

-20 -20

• data • data
simulation simulation

-40 -40
-4 0 4 -4 0 4
total strain (%) total strain (%)
Figure 12-6 Isothermal shear stress-shear strain loops from reference 31. The solid line is
the simulation of the stresses based on the assumed flow Equation (12-3).
372 A DAMAGE INTEGRAL METHODOlOGY

10r-------~--------_.--------_r------~

o
(f)
(f)
Q)
\....
-+-'
(f)
AT = -25 to 125°C
T = 30 cycles/h
\....
o -10
Q)
...c 0= data
(f)
- = simulation

-20~------~--------~--------~------~
-50 o 50 100 150
Temperature (CO)
Figure 12-7 Thermal stress cycles measured by HallY The solid line is the simulation.

It must be emphasized that damage rates, such as fatigue crack growth, are
approximate power functions of the applied stress, with exponents that are large.
Errors in stress calculations can therefore amplify the errors in the estimated
damage rates and the fatigue life. Accurate calculation of the stress during a
thermal or mechanical cycle is, therefore, the first task in the development of
an integral approach to low cycle fatigue of solder joints. It was shown that a
phenomenological framework based on a state variable concept can be used to
accurately estimate the stresses during both isothermal and thermal cycling of
solder joints. The development of a similar phenomenological approach that can
be applied to characterize the damage rate in the solder joint will be examined
next.

12.2 DAMAGE RATE FORMULATION

12.2.1 Damage Mechanisms

Common damage mechanisms that are observed during low cycle fatigue and
at high homologous temperatures are crack growth, grain boundary cavitation,
 DAMAGE RATE FORMULATION 373

and grain boundary wedge cracking. Sliding at phase and grain boundaries is
operative at these temperatures. These phenomena produce stress concentrations
at sites along the boundaries where flow incompatibility occurs. Cavity nucleation
and wedge cracking are processes that tend to relieve such stress concentrations.
The growth of wedge cracks and cavities act together with the growth of a surface
crack, producing the fatigue damage that eventually leads to failure of the solder
joint. Unlike inelastic deformation studies, the description of damage rates for
the above processes and their interactions is not well developed, particularly for
high homologous temperature fatigue.
Boundary cavitation and wedge cracking are related phenomena. Grain bound-
ary cavitation is a diffusive process that occurs at the stress concentration sites
along the grain or phase boundary.33 Wedge cavitation also occurs because of
the stress concentrations but at the grain boundary triple points. The latter process
is a more important phenomenon at strain rates higher than those corresponding
to boundary cavitation. Note that these damage mechanisms are driven by grain
and phase boundary sliding. The deconvolution of the observed deformation
behavior to deduce both the relative contributions of the various boundary phe-
nomena and the operative stress-strain rate windows is possible. Such an exercise
will, however, not be simple for solder joints because of the presence of com-
plicated microstructures.
Since the failure of a solder joint usually involves the growth of a major crack,
it may be possible to represent the combined effects of the various damage
processes by the rate of growth of such a major crack. The parameters of the
crack growth rate formulation will then contain the synergistic effects of all the
damage processes. The description of the rate of crack growth is phenomeno-
logically based on the applied stress intensity factors. While such an approach
has been successfully used to describe the crack growth behavior under mostly
elastic loading, the application of the same approach for the present interest
requires extensive examination.
Even without the presence of other damage processes, the use of the fracture
mechanics approach to describe crack growth under plastic loading is a contro-
versy in itself. In the following, an attempt at a phenomenological description
of fatigue damage under plastic loading will be discussed for the purpose of
stimulating further work in such a direction. The nominal stress intensity factor
K in the crack growth law should be considered an effective value. Although
nominally in shear, there will always be a mode I contribution to this effective
K, the extent of which depends on the exact loading conditions. It is further
assumed that the stress contained in such stress intensity formulations is uniquely
related to the operative bulk stress that is the driving force for the nucleation
and growth of wedge cracks and boundary cavities. Thus the parameters in the
crack growth rate formulation will reflect those for the growth of the major crack
as well as the synergistic effects of all the damage processes and the environment.
374 A DAMAGE INTEGRAL METHODOLOGY

12.2.2 A Phenomenological Formulation for Crack Growth Rates

The damage rate that results from the stress is calculated according to a phe-
nomenological crack growth rate formulation. 15,16 In general, crack growth rate
is a function of the stress intensity factor, the chemical environment, and the
temperature. The fatigue crack propagation rate is expressed in terms of a nominal
stress intensity factor K such that

de
v*K' (12-4)
dt

where dcldt is the time rate of crack growth. The nominal stress intensity factor
is given by K = TNVcY(e) where e is the instantaneous crack length and Y(e)
is a geometric factor, TN is the nominal stress based on the initial loading area,
and r is a crack growth exponent that can be considered constant over small
ranges of K. The rate constant v* is presumed to have the physical significance
discus/ied previously. It has an Arrhenius temperature dependence with an ac-
tivation energy Qc and a preexponential v~.
Extensive and good quality experimental data are needed to establish and
verify the phenomenological framework, The parameters in Equation (12-4) can
be deduced from isothermal fatigue data of Solomon30 ,31 for 60% Sn-40% Pb
solder joints. He has reported the peak load versus number of cycles N for
cracking solder joints at several temperatures and plastic strain ranges. If the
solder flow stress and the uncracked area remains constant and the inelastic strain
is sufficiently large to achieve macroplastic flow during every cycle, the peak
load directly reflects the uncracked area of the solder. Two linear crack fronts
originating from opposite ends are assumed (see Solomon30). Then, we can
calculate the crack length and crack growth per cycle from the published peak
load drop versus cycles data. If the crack length change per cycle is small
compared to crack length, the nominal stress and the stress intensity factor K
are nearly constant in any cycle. Then, a conversion to time rates of crack growth
dcldt = v(dcldN) is possible (lIv is the time period).
Since the load and crack length are now known as a function of cycles N, the
nominal stress intensity factor can be estimated for any cycle, The geometry
factor Y(e) is assumed to be unity. When the plastic strains are large, most of
the cycle is spent close to the peak stress intensity factor; hence the crack growth
rate estimated approximately corresponds to the peak stress intensity factors.
Thus the crack growth rate can be estimated as a function of the stress intensity
factor. Repeating this procedure for isothermal load drop data for large plastic
strain ranges at several different temperatures, we can deduce the relationship
between dcldt and K. These data are plotted in as the symbols in Figure 12_8. 5 •6
Figure 12-9 plots the crack growth rates versus reciprocal temperature at a
constant K (= 1 MPaVrn). From this plot the activation energy for crack growth
 ,, ,
,
I
I I

,
"

,
I
/ I I

I
I
I . ,
, .

I ,
I /
I
/

I , I /
/ / /
10- 4 - I I .-
I . / J
I I /

-- ,
/ I
/ /

---
/ I
(,)
I /
Q) /
en
E AI
/
/

/
I
I
/9
, q>
.....
Q)
co J I
¢:9'cd
I.... ~
• I .I /
10- 5 .-
.....
..c / I
~.

.~ I
~ I /

k
0 I / " 6/
I....
0> / I .~
~ I I. I' ~
c..:>
•
,
I ~
co I I •
~ I
I....
(,) I .I 6
I I
II
/ I I TOC tJ."Y p
I I
10- 6 l- I ·1· '0 -50 0.307. .-
I -50 0.103
I I. t:.
I
I I 0 35 0.260
/
/ / I • 35 0.082
l.t I I 125 0.102
•
/ I 150 0.097
/ I
•
I /
/ /
/ / I

100
stress intensity factor (MPa-ym)
Figure 12-8 Deduced crack growth behavior as a function of temperature and nominal
stress intensity factor. The dashed lines represent typical crack growth behavior. These lines
are shifted vertically through an activation energy Qc.

375
 376 A DAMAGE INTEGRAL METHODOlOGY

2200150 100 50 -50

10- I I I I I

\
10-3 r- \ -
\
, Qc = 79 KJ/mol

-
U
Q)
en
E 10- 4 r- Q V*o = 5.5 X 106 (MPawnr rm/sec -
~
........
I ..\

~CO ,
... \
..

a.. 10-5 r-
.....
...., '....':- -
~
..........
i<
~
·""'·6
.- .........
......... .........
........
10- 6 - Qc = 11 KJ/mol
..6) ... _ ... _ _
..........

10-7~ __~I____~I_____~I____I~__~I____~
2.0 2.5 3.0 3.5 4.0 4.5 5.0
1OOOIT (K)
Figure 12-9 Temperature dependence of crack growth deduced at K = 1 MPaYm indi-
cating that at least two regimes of crack growth exist, one above 1DO·C and one below 1DD·C.

Qc and the athermal rate parameter v~ can be determined. Different crack growth
behavior is observed above and below approximately l000 implying that at e
least two regions of damage behavior may be operative. For T greater than about
l00oe, Qc = 80 KJ/mol and v~ = 5.5 X 107 (MPa-2 . 3 m-O· 15 S-I) while below
l000 e Qc = 11 KJ/mol and v~ = 3.9 x 10-5 (MPa-2.3 m-O· 15 S-I). Solomon
has observed intergranular failures at high temperatures while at lower temper-
atures the crack path appears to be both transgranular and intergranular. However,
it remains to be resolved6 whether or not the observed difference is caused by
a change in crack path due to damage mechanisms such as boundary cavitation
or caused by interaction of crack growth process with the chemical environment
or both.
 DAMAGE INTEGRATION AND FAILURE CRITERION EFFECTS 377

12.3 DAMAGE INTEGRATION AND FAILURE CRITERION EFFECTS

It has been demonstrated that with good constitutive relations and accurate ex-
perimental data an adequate calculation ofthe stress in the solder joint is possible.
A phenomenological crack growth rate relation that describes the damage rates
for a range of temperatures and stresses has also been established. The application
of these considerations to thermal fatigue of solder joints will now be examined.
Separating variables and integrating Equation (12-4) over the fatigue life,

Iei
C;
dc
-----=:- -
(Y(c)Vcy -
J'f
0
0
(Qc)
dt v* exp - -
RT
Tr
N
(12-5)

where Ci and CI are the initial and final crack length, respectively and ti is the
time to failure. Each side of Equation (12-5) is a measure of the total damage
at failure, one in terms of the crack length and the other in terms of time. For
a given combination of Ci and cI' the crack length integral on the left will be a
constant. Furthermore, if the exponent r is greater than 2, this integral is essen-
tially independent of the final crack length CI and for a given solder joint design,
this integral will, in the limiting case, depend principally on the initial flaw size
Ci·
For some applications 5 where the damage per cycle is nearly constant, a
reasonable simplification is to consider the fatigue damage over a single cycle.
The cycles to failure NI is then given by

where'!} is the crack length integral, and lIv is the time period. This single
cycle approximation is valid, for example, where the changes in nominal stress
profile are small over most of the fatigue life.

12.3.1 Thermal fatigue Life Estimation

As a numerical illustration of the previous considerations we shall perform sample

calculations for a solder joint assembly at thermal cycle conditions for which
the thermomechanical fatigue life has been measured. Such data is available on
60% Sn-40% Pb solder joints. Sherry and Hall 14 have measured the thermo-
mechanical fatigue life of SMT solder joints that connect a ceramic chip carrier
to an epoxy-glass printed wiring board. The assembly geometry is similar to the
schematic in Figure 12-1. They have reported the cycles to failure for temperature
 378 A DAMAGE INTEGRAL METHODOLOGY

cycles with minimum temperature Tmin = - 20DC or Tmin = 30DC. These tests
were performed for various temperature ranges a T at a constant cyclic frequency
of 58 cycles/day. The measured fatigue data are reproduced in Figure 12-10 as
open symbols. These data vividly illustrates that the temperature range, which
is a measure of total thermal displacement range, and fatigue life are not uniquely
related.
The assembly details of these solder joints have been reported by Clech and
Augis. 34 The thermal expansion difference a is 3.5 ppm/DC. The temperature a
gradients in the solder joint are assumed to be negligible. The solder joint height
h = 200 /-Lm and the area A = 0.155 mm2 • The elastic moduli of the ceramic
and the circuit board are 255 and 12.3 GPa, respectively. Other assembly details
are listed in Appendices Band C of reference. 34 The approximate thermal profiles
for these tests are also given in reference. 34 The Tmin = - 20DC profiles have
nearly equal dwell times at the minimum and maximum temperatures while the
Tmin = 30DC profiles have longer dwells at Tmin. The temperature profiles for
each simulation are chosen to match these measured profiles.
The details of the damage integrations are provided in reference. 5 The stress
profiles are calculated in the same way as that for the data in Figure 12-7. The

200 \
I

-
~
~
00 \
...... ...... \
Q) 100 f- ~
......
-
C>
c: ...... \
ctS
'- ~~
Q)

-
'- ......
:::J ......
I---()-->-d
ctS ......
'- ......
a. 50
Q)
t- ...... -

-
......
E
Q)
......

Tmin -20°C +30°C~ . . .

C> ......
..Q data t:. 0
......
~
......
simulation .to
• ....

20 I

10 2 103 10 4
log (cycles to failure)
Figure 12-10 Comparison of measured thermal fatigue data for lead less surface mount
solder joint with simulations. Frequency is 58 cycles/day.
 DAMAGE INTEGRATION AND FAILURE CRITERION EFFECTS 379

results of the numerical integration perfonned based on Equations (12-2) and

(12-6) for each of these data points are shown as the filled symbols in Figure
12-10. The agreement between simulation and experimental data is apparent.
These results suggest that a phenomenological approach may be successfully
applied to describe the thennal fatigue behavior under conditions where more
than one damage mechanism may be operative.

12.3.2 Failure Criterion Effects

Failure of soldered interconnects is often specified in tenns of damage parameters

that represent the decrease in solder joint area during fatigue. For mechanical
fatigue tests, the failure criterion is usually specified in tenns of a load drop
(e.g., see reference 30). For thennal cycling or during service, the electrical or
thennal resistance increase is often monitored to determine failure. 35 ,4,36
The dependence of fatigue life on failure criterion specified in tenns of various
damage parameters is reflected by the profile of the parameter as a function of
cycles or time. This profile can be established based on two considerations: (1)
the relationship between the measured damage parameter and the solder joint
area and (2) the time profile of the solder joint area during fatigue. The fonner
is used to calculate the change in the measured damage parameter corresponding
to a change in solder joint area. The time or the number of cycles required to
produce a given solder joint area change can be estimated based on damage
integration [Equation (12-5)].
An interest in the effect of the failure criterion on fatigue life arises from the
need to correlate fatigue life data obtained from laboratory isothennal mechanical
tests to thennal fatigue life. As mentioned previously, the failure criterion in
each case is often specified in tenns of different damage parameters. Furthennore,
these criteria can correspond to different crack lengths. For such situations, the
relationship between the solder joint area and the corresponding damage param-
eter is required to compare fatigue lives.
The dependence of load drop on the solder joint area is a function of the
stiffness of the test machine or assembly relative to that of the solder joint and
the magnitude of the applied displacement. When the applied displacements are
small, the defonnation of the solder joint is essentially linearly proportional to
the load. During such conditions, the load will reflect the solder joint area only
if the solder joint stiffness is comparable to or smaller than the test machine or
the assembly stiffness, For an infinitely stiff machine, the stress in the solder
joint is proportional to the solder joint strain through a combination of the elastic
and anelastic moduli, Therefore, the load will be proportional to the solder joint
area. When the applied displacements are large, the extent of plastic defonnation
can be large. As discussed previously, the load drop will then be proportional
to the solder joint area in the absence of work hardening.
380 A DAMAGE INTEGRAL METHODOLOGY

In general, when the electrical resistance is used as a damage parameter, a

significant decrease in solder area is necessary for appreciable change in total
resistance. 37 ,38 At small crack lengths the dependence of the resistance on the
solder joint area is usually small. As the solder joint approaches final separation,
the total resistance increase can accelerate dramatically. Correspondingly, if the
failure is specified in terms of resistance increases corresponding to small crack
lengths, the effect of failure criterion on the cycles to failure can be significant.
On the other hand, the effect of failure criterion on fatigue life is expected to
be small if a resistance increase failure criterion corresponding to the final stages
of crack growth is specified.

12.4 DISCUSSION AND CONCLUSIONS

Deformation and damage during solder joint fatigue are distinct processes, and
hence they can be examined separately. However, both processes can involve
multiple mechanisms. A comprehensive treatment of the interactions of the
various mechanisms requires a framework that can be used to design relevant
experimental methods and to correlate experimental data. Phenomenological
approaches for the description of deformation and fatigue damage in solder joints
has been demonstrated. While for scientific knowledge such approaches provide
a basis for deconvoluting individual operative mechanisms, its merits in engi-
neering practice deserve further exploration. The development and application
of phenomenological approaches to inelastic deformation of materials has re-
ceived considerable attention. 39 However, significant efforts have not been made
toward the use of similar approaches to damage processes. It is shown here that
it may be possible, just as in the treatment of deformation phenomena, to view
damage processes phenomenologically. The success of the latter will further
enhance the power of integral approaches.

12.5 ACKNOWLEDGMENTS

This work has supported by IBM Corporation and by Semiconductor Research

Corporation.

REFERENCES

I. Homa, T. R., "Cracking of Electronic Packaging Joints due to Creep-Fatigue," in

Proc. of ASM 3d Conf. on Electronic Packaging: Materials and Processes and
Corrosion in Microelectronics, Minneapolis, MN, ASM, 1987, pp. 209-216.
2. Raman, Y., and T. C. Reiley, "Cavitation and Cracking in As-Cast and Superplastic
Pb-Sn Eutectic during High-Temperature Fatigue," Journal of Materials Science
Letters, 6(5), 1987, pp. 549-551.
 REFERENCES 381

3. Agarwala, B. N., "Thennal Fatigue Damage in Pb-In Solder Interconnections," in

23rd Annual Proc. Int. Reliability Physics Society, Orlando, Florida: IEEE, 1985,
pp. 198-205.
4. Inoue, H., Y. Kurihara, and H. Hachino, "Pb-Sn Solder for Die Bonding of Silicon
Chips," IEEE Transactions on Components, Hybrids and Manufacturing Technology,
CHMT-9(2), 1986, pp. 190-194.
5. Subrahmanyan, R., 1. R. WIlcox, and C.-Y. Li, "A Damage Integral Approach to
Thennal Fatigue of Solder Joints," IEEE Transactions on Components, Hybrids and
Manufacturing Technology, CHMT-12(4), 1989, pp. 480-491.
6. _ _ , "A Damage Integral Approach to Solder Joint Fatigue," in Microelectronic
Packaging Technology: Materials and Processes, Proc. 2d ASM Int. Electronic
Materials and Processing Congress, ASM, 1989, pp. 213-221.
7. Coffin, L. F., "Predictive Parameters and Their Application to High Temperature
Low-Cycle Fatigue," in Fracture 1969 (Proc. 2d Int. Conference on Fracture),
London, Chapman and Hall, 1969, p. 643.
8. Manson, S. S., "Interfaces between Fatigue Creep and Fracture," Int. Journal of
Fracture Mechanics, 2(1), 1966, pp. 327-363.
9. Majumdar, S. and P. S. Maiya, "A Unified and Mechanistic Approach to Creep
Fatigue Damage," in Proc. 2d Int. Conf. on Mechanical Behavior of Materials,
ASM, 1976, pp. 314-327.
10. Stone, D., H. Wilson, R. Subrahmanyan, andC.-Y. Li, "The Mechanisms of Damage
Accumulation in Solders during Thennal Fatigue," in Proc. 36th Electronic Com-
ponents Conf., IEEE, 1986, pp. 630-635.
II. Solomon, H. D., "Effect of Hold Time and Fatigue Cycle Wave Shape on the Low
Cycle Fatigue of 60/40 Solder," in Proc. 38th Electronics Components Conf., Los
Angeles, IEEE, 1988, pp. 7-12.
12. Shah, H. J., and J. H. Kelly, "Effect of Dwell Time on Thennal Cycling of F1ip-
Chip Joint," in Proc. 5th Int'l. Hybrid Microelectronics Symposium, 1970, pp. 3.4.1-
3.4.6.
13. Eckel, J. F., "The Influence of Frequency on the Repeated Bending Life of Acid
Lead," in Proc. American Society of Testing Materials, ASTM, 1951, pp. 745-760.
14. Sherry, W. M., and P. M. Hall, "Materials, Structures, and Mechanics of Solder
Joints for Surface Mount Microelectronics Technology," in Proc. 3d Int. Conf. on
Interconnection Technology in Electronics, Fellbach, West Gennany, 1986, pp. 47-
81.
15. Rolfe, S. T., and J. M. Barsom, Fracture and Fatigue Control in Structures, En-
glewood Cliffs, NJ, Prentice-Hall, 1977, pp. 232-267.
16. Hertzberg, R. W., Deformation and Fracture Mechanics of Engineering Materials,
New York: John Wiley & Sons, 1983, pp. 519-598.
17. Wilcox, J. R., "Inelastic Defonnation and Fatigue Damage in Metals at High Ho-
mologous Temperatures," Ph.D. Thesis, Cornell University, Ithaca, NY, 1990.
18. Lee, S.-M., and D. S. Stone, "Grain Boundary Sliding in Surface Mount Solders
during Thennal Cycling," to be published in Proc. 10th Electronic Components and
Technology Conf., Las Vegas, Nevada, May 1990.
19. Tribula, D., D. Grivas, D. R. Frear, and J. W. Morris, "Observations on the
Mechanisms of Fatigue in Eutectic Pb-Sn Solder Joints," Journal of Electronic
Packaging, ASME, 111(2), 1989, pp. 83-89.
382 A DAMAGE INTEGRAL METHODOLOGY

20. Stone, D., H. Wilson, R. Subrahmanyan, and c.-y. Li, "An Investigation of the
Creep Fatigue Interaction in Solder Joints," in Proc. National Electronic Packaging
Conf. East, 1986, pp. 175-180.
21. Wilcox, J. R., R. Subrahmanyan, and c.-y. Li, "Thermal Stresses and Inelastic
Deformation of Solder Joints," in Microelectronic Packaging Technology: Materials
and Processes, Proc. 2d ASM Int. Electronic Materials and Processing Congress,
ASM, 1989, pp. 203-211.
22. Frost, H. J., P. R. Lavery, and S. D. Lutender, "Microstructure and Mechanical
Properties of Lead Tin Solder Alloys," in Proc. ASM 3d Conf. on Electronic Pack-
aging: Materials and Processes and Corrosion in Microelectronics, 1987, pp. 259-
267.
23. Frost, H. J., R. T. Howard, and G. J. Stone, "Effects of Thermal History on
Microstructure and Mechanical Properties of Solder Alloys," in Proc. ASM 2d Int.
Electronic Materials and Processing Congress, 1989, pp. 121-127.
24. Hart, E. W., "Constitutive Relations for the Nonelastic Deformation of Metals,"
Journal of Engineering Materials and Technology, 98(3), 1976, pp. 193-202.
25. Jackson, M. S., C. W. Cho, P. Alexopoulos, and c.-y. Li, "A Phenomenological
Model for Transient Deformation Based on State Variables," Journal of Engineering
Materials and Technology, 103(4), 1981, pp. 314-325.
26. Li, C.-Y., "State Variable Theories for Nonelastic Deformation," in Metallurgical
Treatises, J. K. Tien and J. F. Elliott, Eds., TMS-AIME, 1981, pp. 469-486.
27. Stone, D., S.-P. Hannula, and C.-Y. Li, "The Effects of Service and Material
Variables on the Fatigue Behavior of Solder Joints during the Thermal Cycle," in
Proc. 35th Electronic Components Corif., IEEE, 1985, pp. 46-51.
28. Alexopoulos, P., R. L. Keusseyan, G. L. Wire, and c.-y. Li, "Experimental In-
vestigation of Nonelastic Deformation Emphasizing Transient Phenomena by Using
a State Variable Approach," in Mechanical Testing for Deformation Model Devel-
opment, R. W. Rohde and J. C. Swearengen, Eds., 1981, pp. 148-184.
29. Kashyap, B. P., and G. S. Murty, "Experimental Constitutive Relations for the High
Temperature Deformation of a Pb-Sn Eutectic Alloy," Materials Science and En-
gineering, 50(2), 1981, pp. 205-213.
30. Solomon, H. D., "Low Cycle Fatigue of 60/40 Solder-Plastic Strain Limited vs.
Displacement Limited Testing," in Proc. ASM's 3d Electronic Packaging: Materials
and Processes Conf., Minneapolis, MN, ASM, 1985, pp. 29-47.
31. Solomon, H. D., "Low-Frequency, High Temperature, Low Cycle Fatigue of 60
SnJ40 Pb Solder," Tech. Rep. 85CRD238, GE Corporate Research and Development,
1985; also published, in part, in Low Cycle Fatigue, ASTM-STP 942, H. D. Solomon
et aI., Eds., ASTM, 1985, pp. 342-371.
32. Hall, P. M., "Forces, Moments, and Displacements during Thermal Chamber Cycling
of Leadless Ceramic Chip Carriers Soldered to Printed Boards," IEEE Transactions
on Components, Hybrids and Manufacturing Technology, CHMT-7(4), 1984, pp.
314-327.
33. Garafalo, F., Fundamentals of Creep and Creep Rupture in Metals, New York,
Macmillan Company, 1965.
34. Clech, J. P., and J. A. Augis, "Engineering Analysis of Thermal Cycling Accelerated
Tests for Surface-Mount Attachment Reliability Evaluation," in Proc. 7th Int. Elec-
tronics Packaging Society Coriference, IEPS, 1987.
 REfERENCES 383

35. Wild, R. N., "Some Factors Affecting Leadless Chip Carrier Solder Joint Fatigue
Life II," Circuit World. 14(4), 1988, pp. 29-36,41.
36. Solomon, H. D., "Low Cycle Fatigue of Surface-Mounted Chip-Carrier/Printed Wir-
ing Board Joints," IEEE Transactions on Components. Hybrids and Manufacturing
Technology. CHMT-12(4), 1989, pp. 473-479.
37. Solomon, H. D., "Strain Life Behavior in 60/40 Solder," Journal of Electronic
Packaging. ASME, 111(2), 1989, pp. 75-82.
38. Wilcox, J. R., R. Subrahrnanyan, and C.-Y. Li, "Assembly Stiffness and Failure
Criterion Considerations in Solder Joint Fatigue," Journal of Electronic Packaging.
Transaction of the ASME, 112, (2), June 1990, pp. 115-122.
39. Korhonen, M. A., S.-P. Hannula, and C.-Y. Li, "State Variable Theories based on
Hart's Formulation," in Unified Constitutive Equationsfor Creep and Plasticity. A.
K. Miller, Ed., London/New York, Elsevier Applied Science, 1987, pp. 89-138.
 13
Modern Approaches to Fatigue Life
Prediction of SMT Solder Joints
Thomas J. Kilinski, Jon R. Lesniak,
and Bela I. Sandor

Surface mount technology (SMT) has been widely adopted in recent years,
brought on by the need to produce small, dense electronic packaging. This
technology has many advantageous characteristics:

• SMT allows for components to be mounted on both sides of the board.

• Holes through the board are only needed for communication between the
surface and the circuit planes; thus the size of the holes can be decreased.
• Since less room is needed for holes, a 50 percent increase in line density
from a similar through-hole board can be obtained. 1

Unfortunately, SMT also brings with it a new era of failures. The joint is
responsible for both electrical and mechanical connections. Solder does not have
adequate ductility to ensure the repeated relative displacements due to the mis-
match between expansion coefficients of the chip carrier and the circuit board,
as it has been believed.
A major finding from mechanical tests is that solder behavior involves a creep-
fatigue interaction, making it a poor material for use in mechanical connections.
These considerations lead to the conclusion that it will take a combined effort
in mechanics and metallurgy to control the failures of these joints. Even then,
finding a technique for increasing the service life of future connections is not
the total solution. A method must be developed for predicting the remaining
service life of the many SMT joints already in use.

384
 MECHANICAL TESTING 385

This chapter expresses our thoughts and research direction regarding SMT
solder joint reliability analysis, keeping in mind fundamental and practical issues.
There is no simple, complete solution to the problem at hand, but significant
progress can be made toward understanding and controlling it.

13.1 MECHANICAL TESTING

13.1.1 Determination of Elastic Properties

The elastic modulus (E), Poisson's ratio (v), and the shear modulus (G) are
determined with strain gages, or mechanical extensometers on a specimen of
uniform cross section in directions parallel and perpendicular to the applied
loading; E and v can be measured directly, while G is commonly computed from

E
G=--- valid for isotropic materials (13-1)
2 (1 + v)

Extreme caution must be taken because solders have both elastic and inelastic
response at the instant of load application. If testing is performed at low strain
rates, the resulting stress versus axial strain plot may appear fully linear. How-
ever, the slope of this curve does not provide an accurate value for the elastic
modulus because inelastic as well as elastic strains are present. To obtain an
accurate value, testing must be performed at low stresses so that minimal yielding
occurs and at a rate high enough to eliminate any viscoplastic effects. Screw-
driven testing machines cannot achieve high enough strain rates under proper
control to accurately perform these tests. Closed-loop, servo-hydraulic machines
may be needed to obtain accurate data. In order to know if the testing rate is
sufficiently high, the stress-strain slope must be obtained for increasing loading
rates. When there is no change in the measured slope with further increases in
rate, it is safe to assume that the solder is undergoing purely elastic deformations.
Using this technique, it was found that a ramp speed equivalent to 80 MPa/sec
was sufficient to minimize viscoplastic effects in 60% tin-49% lead solder at
25°C. It is important to realize that these elastic constants may vary with tem-
perature and are slightly dependent on microstructural properties such as eutectic
grain size, provided that the microstructure is isotropic (a reasonable assumption
for 60% tin-40% lead solder). Therefore, it is necessary to account for ambient
temperature effects while performing any solder behavior analysis.
The experimentally determined values for E and v are summarized in Table
13-1, where they are compared with other values found in the literature. Note
that there are large differences in E obtained by various researchers. Based on
our data, the average shear modulus at 25°C is G = 14.4 GPa.
As a further check of our results, handbook values for the elastic moduli of
tin and solder were obtained. Since the solder microstructure is near-eutectic
38& MODERN APPROACHES TO FATIGUE LIFE PREDICTION OF SMT SOLDER JOINTS

Table 13-1 Elastic Constants for 60% Tin-40% Lead Solder at 25°C
Reference Elastic Modulus (GPa) Poisson's Ratio
Authors' data 38.6 ± 4 0.36 ± 0.04
3 16
4 15 0.40
5 30
6 35

and consists of a random mixture of lead-rich and tin-rich grains combined with
a random distribution of lead-rich dendrites, a simple rule-of-mixtures calcula-
tion, used to estimate the elastic modulus of macroscopic composite materials, I
should provide a reasonable estimate for the elastic modulus of 60% tin-40%
lead solder. As seen below, this estimate is in agreement with our experimental
data: 2

Etin = 50 GPa (Ref.?)

Elead = 14 GPa (Ref. 8)
E6O% tin-40% lead solder = 0.6(50) + 0.4(14) = 36 GPa

13 .1.2 Mechanical Properties

Other kinds of laboratory testing are needed in order to fully understand the
behavior of solders during thermal-mechanical fatigue loading. Ambient tem-
perature fluctuations, stress and strain amplitudes, superimposed mean stresses
and strains, loading rates, and even the shelf life of the specimens contribute to
the solder behavior causing scatter in the fatigue life prediction.

Creep and Stress Relaxation

Stress relaxation and creep properties determined in the laboratory help model
the joint compliancy during thermal expansion cycles. Tests must be performed
on bulk solder as well as on model SMT joints. Creep tests are somewhat
unrealistic because real joints are constrained. Also creep tests may require much
time. The mechanistically equivalent stress relaxation tests, on the other hand,
can be performed swiftly because the primary information, stress relaxation rate
or rate of load drop, occurs "immediately" upon loading. In fact, with solders,
the initial strain amplitude must be applied rapidly; otherwise, the load will have
significantly relaxed by the time the desired strain amplitude has been reached,
resulting in an unrealistically small recorded peak load.
 MECHANICAL TESTING 387

If the stress-time data are fitted to a power equation of the form

(13-2)

where T is the nominal shear stress, t is time, and A and n are constants, the
stress relaxation rates can be determined by differentiating this equation.
The stress relaxation data can also be used to obtain stress versus steady-state
creep rate data by remembering that throughout a stress relaxation test the total
deformation (elastic plus inelastic) remains constant. The solder strain (in this
case shear strains) can be divided into its elastic and creep components,

P
'Y = 'Ye + 'Yc = AG + 'Yc (13-3)

where P is the load acting on the solder joint and A is the cross-sectional area
in shear. The steady-state creep rate is determined by substituting Equation (13-
2) into Equation (13-3):

'Y = tan- 1 (~) = (~) for 'Y ~ 1 (13-4)

where 8. is the displacement across the solder joint and H is the height of the
joint. Caution must be taken when using small angle approximations for deflec-

100
Stress relaxation data

r0-
o..
~

m,
If)
If)

....Q) 10 m
+-'
Cf) mm ~ Elm
....ctl EI EI

EI EI EI
Q)
.r:::
Cf)
Creep data

1 ~~-r~~~~~~-M~~~~~~
10- 9 10- 8 10- 7 10- 6 10- 5 10- 4 10- 3 10- 2

Strain rate (mm/mm/s)

Figure 13-1 Steady-state creep rates for 60% tin-40% lead solder at 25°C.
388 MODERN APPROACHES TO FATIGUE LIFE PREDICTION OF SMT SOLDER JOINTS

tions. Data have been reported stating shear strains at failure exceeding -rr12 rad,
which is unrealistically large. The resulting equation is then differentiated with
respect to time to obtain the steady-state creep rate:

(13-5)

Figure 13-1 shows the correlation between the steady-state creep rate calculated
from stress relaxation tests and the data obtained from numerous independent
creep tests.

13.2 LIFE PREDICTION TECHNIQUES

This section is devoted to describing the most popular empirical models available,
along with their applicability to SMT fatigue life prediction.

13.2.1 Fatigue Models

SMT solder joint fatigue failure is caused most often by a thermal expansion
mismatch between the surface mounted device and the substrate material. These
conditions result in the total inelastic strain range in the solder being a combi-
nation of instantaneous inelastic strains (plastic strains) and time dependent
inelastic strains (creep strains). Fatigue models can be divided into several major
categories. The two most prominent categories are:

Mechanistic models. The damage mechanisms involved are mathematically

described. These models tend to be quite sophisticated; in addition, a
given model is usually only applicable to a specific material. For example,
a model developed for 60% tin-40% lead solder may be inappropriate
for a tin-lead-indium solder.
Empirical models. A measurable physical quantity, usually the inelastic strain
range, is used as the basis for life prediction. The particular damage
mechanisms causing this physical quantity are usually not known. Despite
this, empirical models have been successful for several common mate-
rials.

Coffin-Manson Equation

The Coffin-Manson equation can be written as

(13-6)
 LIFE PREDICTION TECHNIQUES 389

where tlepl2 = plastic strain amplitude (Figs. 13-2 and 13-3)

ef = fatigue ductility coefficient

c = fatigue ductility exponent

'2Nt = reversals to failure (1 cycle = 2 reversals)

The equation was developed in the 1950s. 9,10 It represented a major breakthrough
in the field of low cycle fatigue. It is therefore not surprising that this is the
model most often seen in the solder fatigue literature.
It has long been recognized in the aerospace establishment that the Coffin-
Manson equation is inadequate for modeling in-service thermal fatigue conditions
in high temperature alloys. The primary difficulty is the inability to account for
variations in waveform. Thermal fatigue involves a combination of creep and
plastic strains, and each can have a different effect on the component life. The
Coffin-Manson equation does not allow differentiation between these two strains,
resulting in inaccurate life predictions.

0'

0' A_O'_
__
a 2

, ,
,,
J
J
,
,,,
J

,
,
~~ A£e
A£

Figure 13-2 Representative stress-strain hysteresis loop for a material undergoing low cycle
fatigue.
390 MODERN APPROACHES TO FATIGUE LIFE PREDICTION OF SMT SOLDER JOINTS

E0> C"
c f

Ql
"0
.~
a.
E
cO
c
~
(/)

U
.+'
(/)
C1l
a::
10 1 10 2 10 3 10 4
Reversals to failure, 2Nf

Figure 13-3 Schematic of Coffin-Manson equation.

Strain-Life Relationship (Total Strain Equation)

In this approach, the Coffin-Manson equation is combined with Basquin's

equation for high cycle fatigue:

(13-7)

where (Ta = a (T I 2 = true stress amplitude

(Tj = fatigue strength coefficient
b = fatigue strength exponent (Basquin's exponent)

The total strain amplitude, aE/2, can be written as the sum of elastic and plastic
components

(13-8)

Also, the elastic strain amplitude and the true stress amplitude are related by

(13-9)
2 E
 LIFE PREDICTION TECHNIQUES 391

Combining Equations (13-6), (13-7), (13-8), and (13-9) results in the following:

(13-10)

This equation, displayed graphically in Figure 13-4, is often referred to as the

strain-life relationship, or the total strain equation. The strain-life relationship
has the advantage of combining both high and low cycle fatigue life prediction
into a single equation. Note that it does not account for elastic and plastic
components of damage. In high cycle fatigue, elastic strains predominate and
plastic strains are too small to measure accurately; consequently, the elastic strain
amplitude is used as a means to estimate long fatigue life. Although the plastic
strains are much smaller than the elastic strains in high cycle fatigue, it is still
the plastic strains that are causing the actual fatigue damage. It should be realized
that, fundamentally, elastic strains do not cause damage.
To establish the strain-life relationship, tests performed under low cycle fatigue
conditions can be used to determine the empirical constants b, c, aJ, eJ, and E.
These constants are all that are needed to predict the fatigue response of the
material over the entire range of practical lives. It would be useful to have a
method that uses accelerated tests to predict in-service life. However, it is nec-
essary to account for creep strains, which is difficult to do reliably.

c'f
Cl
.2
(j' equation
a> f
"0
.~ E
a.
E
ct!
c
.~

(j)
,
"
" v Coffin-Manson equation

Reversals to failure, 2N f

Figure 13-4 Schematic of total strain versus life equation.

392 MODERN APPROACHES TO FATIGUE LIFE PREDICTION OF SMT SOLDER JOINTS

Frequency-Modified Coffin-Manson Equation

The Coffin-Manson equation can be modified to include a term that accounts

for the cycling frequency:

(13-11)

where v is the cycling frequency and k is a material parameter. For very low
life (2Nt < 100), frequency is often found to have an insignificant effect on the
life, and hence k = 1. In comparison, lowering the frequency at longer lives
(2Nt > 100) usually results in a decreased life, and k takes on a value less than
1. A plot of total inelastic strain amplitude versus frequency-modified life (2Nt
x V k- 1) often reveals a bilinear relationship. More details can be found in the
reference. 11
The frequency-modification approach is good for situations in which the in-
service loading is uniform for all cycles and is a sinusodial wave of constant
frequency. The in-service loading waveform must have the same shape as that
used in the laboratory testing. This represents a severe limitation of this approach:
There is no easy way to account for variations in waveform. Consider, for
example, the two hypothetical load waveforms that are assumed to result in the
same total inelastic strain range per cycle, shown in Figure 13-5. Although both
waveforms have the same frequency, the waveform shown in Figure 13-5b
involves a longer hold time at the maximum stress, resulting in more creep per

stress
stress

time
time

Figure 13-5 Hypothetical load waveforms and corresponding hysteresis loops. (a) Large
stress amplitude; no hold time. (b) Small stress amplitude; long hold time.
 LIFE PREDICTION TECHNIQUES 393

cycle than in the waveform of Figure 13-5a. Since creep strains are usually more
damaging than plastic strains, the waveform of Figure 13-5b will tend to result
in a lower life than the waveform of Figure 13-5a. However, because the total
inelastic strain range is the same in both cases, the method predicts the same
lives. This equation is not recommended for solder joint fatigue analysis, except
in well-defined special cases, although it has already found wide acceptance in
the SMT field.

Strain Range Partitioning (SRP) Procedure

The SRP approach accounts for different types of strains causing different
amounts of damage. The hysteresis loop for any waveform can be separated into
four possible strain range components l2 (Fig. 13-6):

aEpp = plastic strain in tension and compression (PP)

aEcc = creep strain in tension and compression (CC)
aE cp = creep strain in tension; plastic strain in compression(CP)
aEpc = plastic strain in tension; creep strain in compression (PC)

At most, only three of the four SRP components can exist in any hysteresis
loop, since there can never be both CP and PC loading in the same loop. For
example, if there is an excess of plastic strain in tension over that in compression,

Nep Nee Npe

Cycles to failure

Figure 13-6 Idealized hysteresis loops for strain range components.

394 MODERN APPROACHES TO FATIGUE LIFE PREDICTION OF SMT SOLDER JOINTS

there will be some PC loading but no CP loading. Each of the four components
is usually associated with a different microstructure, caused by creep and plastic
strains.
To apply the SRP approach, the hysteresis loop corresponding to the in-service
cycling conditions must be known. This complex hysteresis loop is resolved into
its strain range components. Next, an interactive damage rule is applied:

(13-12)

where Nf = predicted cycles to failure of the material for the

hysteresis loop being considered

Nij = cycles to failure for the material if the total inelastic strain range,
a fin> of the hysteresis loop were of type ij, where ij
is either PP, CC, PC, or CP (see Fig. 13-7)
Fij = fraction of the total inelastic strain range of the hysteresis
loop that is actually of type ij.

Note that to apply this damage rule, it is necessary to have strain range versus
life relationships for the four types of strain range components. A Coffin-Manson-
type relationship will usually fit the data for all four component types. In fact,
the strain-life equation describing PP-type loading is precisely the Coffin-Manson
equation in its commonly used form.

(J

L-,f-- E L-,f--E

Figure 13-7 Strain-life plots for strain range components.

 HYBRID LIFE PREDICTION TECHNIQUES 395

Total Strain Version of the SRP Procedure

A disadvantage of SRP is that it is difficult to use low cycle fatigue life results
to estimate high cycle creep-fatigue lives. It often predicts in-service lives much
longer than actually observed. It was this difficulty that lead to the development
of the total strain version of the SRP procedure. This approach combines the
advantages of low cycle SRP and high cycle strain life approaches. The technique
involves estimating the elastic strain range-life relationships for the PP, CC, PC,
and CP types of loading. For PP-type loading, the elastic strain range-life re-
lationship is simply Basquin's equation. For the other three types of loading,
the elastic strain range-life relationship is not as easily defined. Instead, it has
been found to depend on the waveform (PC, CP, or CC), frequency, and the
method by which creep strains are caused (stress hold, strain hold, or a com-
bination thereof). Saltsman and Halford now recommend accounting for these
effects by conducting cyclic stress-strain tests in order to obtain the flow char-
acteristics of the material under the conditions of interest. 13
Although the total strain version of SRP may help to more accurately predict
high cycle creep-fatigue life, it is considerably more complicated to use than the
basic version of SRP. In addition, some of the material constants required for
this approach actually depend on waveform, frequency, and other effects. Thus,
it is probably not practical to attempt to use the total strain version of SRP for
solder joint fatigue modeling at this time. A thorough analysis of the standard
SRP approach must first be tried in any case. In addition, the work by Halford
and others has shown that both versions of SRP, total strain and conventional,
give very similar life predictions (that is, both are usually accurate to within a
factor of 2 of the actual life) when the cycles to failure is approximately 10,000
or less. 14 Since many SMT uses fall within this realm, the standard SRP procedure
should prove adequate for most applications.

13.3 HYBRID LIFE PREDICTION TECHNIQUES

One goal of our program is to marry the fatigue prediction techniques with
mechanistic models. It is difficult to calculate the number of cycles to failure
from a purely theoretical model. However, it is meaningless to experimentally
determine a life plot if the test strategy is not driven by some knowledge of the
failure modes. It is imperative for those collecting life data to work closely with
those analyzing the failures.
Before determining which fatigue technique to use, mechanistic damage models
must be understood. It is useful to consider two types of cracks, those traversing
the bulk of the solder and those traveling through the intermetallic layer between
the solder and the substrate. Also, cracks can start in the interface but may
396 MODERN APPROACHES TO FATIGUE LIFE PREDICTION OF SMT SOLDER JOINTS

continue through the bulk solder. Each type of crack is influenced by different
mechanisms, and each mechanism is influenced by many parameters. 15
The total creep rate is a combination of matrix deformation and grain boundary
sliding and, in the case of near-eutectic solders, phase boundary sliding. Using
polycrystalline models developed by Hart and Ghahremani, Stone l5 predicts
solder to have three distinct regions of stress versus strain rate (Fig. 13-8). At
high strain rates the creep stresses are dependent on matrix properties. As the
strain rate decreases, grain boundary sliding becomes involved. At increasingly
lower frequencies of loading, matrix properties are again the limiting factor since
the matrix must deform to allow sliding at triple points (the junction of three
grain boundaries).
An SRP approach lends itself to the combination of mechanistic models with
life prediction functions. As discussed above, the intent of a partitioning approach
is to separate the various types of strains that drive various kinds of damages.
For low cycle fatigue of structural alloys at elevated tempertures this entails
separating plastic strains from creep strains. In the case of solder, it is particularly
important to separate creep by grain boundary sliding and matrix deformation.

grain
boundary

b
Cl
o

log £.
Figure 13-8 Stress versus strain rate curves that incorporate grain boundary sliding for
polycrystaliine material. (From D. Stone, "Creep-fatigue interactions in solders and solder
joints," unpublished paper, University of Wisconsin-Madison, 1988)
 HYBRID LIFE PREDICTION TECHNIQUES 397

13.3.1 Strain Range Partitioning Rule

Standard SRP handles the separation of different strains for a given situation
well. The cumulative damage rule suggested by Hirschberg and Halford 12 works
well for structural alloys at elevated temperatures. It is important to understand
why this relatively simple equation works if this technique is to be extended to
creep-fatigue of solder. For applying the simple damage rule, several things must
hold true:

• All kinds of strains must contribute to a common damage.

• The common damage must be the unique cause of failure.
• The damage versus life rates must be similar for different strain levels and
types of strain.

The first two criteria mean that if the damages caused by various types of
strains are to be analyzed in this simplistic fashion, they must be the same
damages. If more than one kind of damage occurs, one must predominate so
that the lives for all types of strains are dependent only on that mechanism.
Also, the presence of any other damage or condition should not interact with
the dominant damage mechanism. The third criterion states that the damage rates
need to be similar. This implies that if the pertinent damage parameter could be
plotted versus fraction of life exhausted instead of reversals, the amount of
damage would be identical for the same percentage of lives exhausted. It also
implies that failure will occur at the same level of damage for any strain range
or any type of strain. For example, consider a test where 80 percent of the life
is exhausted at a type 1 strain of range a Eb, and the condition is switched to a
type 2 strain of range a Ea (Fig. 13-9). Because the same amount of damage
would have accumulated by exhausting 80 percent of the life for a type 2 test,
damage accumulation continues ignorant of the type of strain that caused the
damage in the first part of the test. It is interesting to note that the damage versus
reversal curves must be similar, which does not demand the damage be accu-
mulated linearly. Of course, the damage function probably will not and need
not be exactly known for most practical applications of SRP. If the damage
versus cycles curves are not similar for all relevant types of strain or strain
ranges, these nonsimilarities must be accounted for.
In light of all that is known about solder and its damage mechanisms, a simple
damage accumulation rule cannot be expected to work. The nature of these
mechanisms violates many of the basic criteria. There is not a unique damage
occurring, several types of cracks are possible, the mechanisms interact, and
they are dependent on strain rate, temperature, and time.
Shine and Fox have contributed much to these ways of thinking. 16 They carried
SRP one step further by partitioning the creep strain into matrix and grain
398 MODERN APPROACHES TO FATIGUE LIFE PREDICTION OF SMT SOLDER JOINTS

Reversals. 2N f

Damage at
failure \ 1& 2
CD
g> ---------------------
E
<II 2: Damage at
Cl
i 80% life \
---~---------------------
I
I
I
I
I

Reversals. 2N Percent Life Exhausted

Figure 13-9 Strain range partitioning damage relationships.

boundary slip. Fatigue lives are correlated with percentages of these strains and
with strain rates. The creep strain rates are separated assuming the total creep
rate is the sum of matrix deformation creep rate and grain boundary sliding.
It is our intention to complement approaches such as that of Shine and Fox
by developing an interactive damage rule. This rule needs to incorporate the
interaction of time, temperature, strain rates, other components of strains, and
any parameter that influences the damage process. This relationship will take an
integral form, thus accounting for damage in each individual cycle, based on
the conditions at that time.

13.4 MODEL JOINTS

13.4.1 Quality Control

It is important to control the process of manufacturing model joints to maintain

a consistent representation of actual SMT connections. The process of main-
taining quality control can be assured through the following techniques ..
 MODEl JOINTS 399

Reflow Process

There are two important aspects to control while making model SMT joints:
metallization diffusion and flux burning. Metallization diffusion occurs when
the specimen is held at or above the reflow temperature for too long. This allows
the metallization to completely diffuse into the molten solder, weakening the
bond between the solder and the substrate. flux burning occurs when the solder
does not reach retlow temperature quickly enough. The tlux bums off before the
solder melts , leaving flux residue within the solder.

Cooling Rates

The cooling rate affects the grain size of the solder joint. Maintaining uniform
grain size is crucial for relating the fatigue life results obtained from model joints
to actual SMT connections. The cooling rates have to be adjusted to account for
the larger solder volume differential in the model joint. However, caution must
be taken so micro- or macro-cracks are not introduced by rapid cooling.

Inspection

Optical and SEM inspection techniques should be used as a method of quality

control for model joints both before (Fig. 13-10) and after testing (Fig. 13-11).

Figure 13-10 SEM backscattered electron image showing initial solder microstructure of a
typical lap joint. Light regions are lead-rich grains; dark regions are tin-rich grains.
 400 MODERN APPROACHES TO FATIGUE LIFE PREDICTION OF SMT SOLDER JOINTS

(a)

(b)

Figure 13-11 SEM backscattered electron images of fracture surfaces of several lap joint
specimens tested in creep showing regions of enhanced grain growth. Light regions are lead-
rich grains; dark regions are tin-rich grains. (a) Initial shear stress = 10.3 MPa. Time to failure
= 6 h. (b) Initial shear stress = 7.7 MPa . Time to failure = 48 h. (c) Initial shear stress =
3.4 MPa. Time to failure = 45 days.
 13.4 MODEL JOINTS 401

(c)

Figure 13-11 Continued

These periodic inspections of specimens ensure consistency of grain size and

void concentrations to help obtain consistent fatigue life data. Simple op-
tical inspection using high powered microscopes is sufficient for grain size
analysis.

13.4.2 Lap Joint Specimens

Single-ball solder specimens are used to isolate fatigue properties free of the
geometric influences found in multilead specimens. They offer the flexibility of
quickly exploring and isolating individual influences on fatigue lives. They lend
themselves to the collection of basic material properties. The specimens used in
our investigations have small metallic pads on an alumina substrate (Fig. 13-
12). Solder balls from 0.38 to 0.89 j.Lm in diameter are used.

13.4.3 Straddle Board Specimens

The straddle board approach is a new direction for bridging the gap between the
simplest model joints and complete manufactured boards. The idea is to perform
tests on real SMT connections to look for trends in life versus joint geometry.
 -----.,.,
402 MODERN APPROACHES TO FATIGUE LIFE PREDICTION OF SMT SOLDER JOINTS

...t.
.---, ~

iooII41--------51.0--------.t~ T

.---_----.-t
ceramic thickness = 0.9
(all dimensions in mm.) I • .. I ~
~12.7--+1·

~=============~~~===-==========~

,..141--------104.0--------t...,1
Figure 13-12 Ceramic surface mount lap joint design. Upper: top view. Lower: side view.

The simplest application for the straddle board is to enable an A to B comparison

of different lead, joint, and connection systems. The straddle board is actually
a fragment of a circuit board. A single chip carrier is mounted to each side of
a printed wiring board (Fig. 13-13). To ensure that the joints are of the same
specifications as a typical SMT joint, the boards are manufactured in a conven-
tional manner, using the same vapor phase reflow assembly process as for regular
components.
Aluminum stiffeners are attached to the straddle board before its placement
into the grips. After the board has been aligned in the grip fixtures, the remaining
mid-section of the board is cut free. The load and strain are zeroed, and the
stiffeners removed. The straddle board is now ready for testing, with load carrying
capabilities coming only from the leads.
Besides performing A to B comparisons, the straddle boards are used to obtain
fatigue-life prediction data for combinations of load, deflection, and temperature.
The majority of testing employs shear defonnations. This type of loading most
accurately resembles the thennal strains induced by the difference in thennal
expansions between the board and the chip carrier. A major obstacle to overcome
when using straddle board-type specimens is accounting for the extent of damage
occurring in individual joints. A typical straddle board configuration has eight
leads per chip, the comer leads of each side remaining connected. It is impossible
to measure the loads on each individual lead given this setup. However, for
correct fatigue analysis it is crucial to distinguish between a drop in the measured
load caused by the failure of a single lead and the load drop caused by the
general fatigue damage occurring in all the leads. An attempt has been made by
 MODEL JOINTS 403

Chip carrier
-,
.
"i
•
•
E
E
C\I
\()

.
·
•"

-.II+-
1.6 mm

Figure 13-13 Straddle board configuration.

other researchers to correlate the amount of damage occurring in individual leads

by measuring the resistance increase within that lead. The change in resistance
occurs because of the change in cross-sectional area from cracks propagating
within the lead. Although the concept is attractive, caution must be taken when
dealing with shearing forces. Even though the connection has failed completely
with respect to load carrying capabilities, electrical connection can still be made
through the contact of fracture surfaces sliding past one another. Therefore, the
resistance measurement will be determined by the contact area of the fracture
surfaces. Arguments can be made that the electrical attachment is the only critical
connection (i.e. , if the joint can no longer hold a load but is still passing a signal,
it is still connected). However, under real operating conditions the joint is subject
to a multi axial state of loading, not just shear loading. In this case a tensile force
acting on the connection, due to shock, vibrations, or thermomechanicalloading,
will tend to separate the joint causing an electrical failure. Separation for even
an instant is unacceptable.
404 MODERN APPROACHES TO FATIGUE LIFE PREDICTION OF SMT SOlDER JOINTS

13.5 EXPERT SYSTEMS

An expert system is a computer program designed to emulate an expert within

a given field. The software makes use of factual knowledge and rules of thumb
to perform given tasks. It consists of a knowledge base and an inference engine.
In the case of SMT life prediction, the knowledge comes in the form of a large
database of fatigue tests performed on actual joints and model laboratory spec-
imens. The inference engine defines the methods by which the stored information
is retrieved and used for problem solving.
The strength of the software is its capacity to store and analyze the huge
quantities of life prediction data obtained through mechanical testing. This means
that data from many investigators can be used in the data bank. A key aspect
of a complete SMT expert system is that it could perform nonlinear finite element
analysis to predict the stresses occurring in the joint connection for any given
loading situation.
A major obstacle in SMT life prediction is that many fatigue life curves are
needed to account for all possible combinations of joint geometries, loading
conditions, and temperature cycles. The expert system has the ability to interpret
much data to predict an outcome for any situation. This way, a finite number
of loading combinations and lead geometries can be analyzed in the laboratory
to determine their individual fatigue behaviors. The finite element method will
be used to predict the stress distribution for these same constraints. The advantage
of the expert system is that the geometry and loading schematic of a lead or
connection system (any geometry or loading) is entered by a user. The finite
element analysis predicts the stress distribution in the connection. The expert
system program will then make a prediction of the service life based on an
interpolation of known results.
Another advantage of the expert system is in optimizing a life prediction
procedure. A limited number of tests, such as three or four, could be performed
on the specimen in question and the results compared to the prediction. The
program may ask for additional tests or use the available data to make a self-
adjustment on its prediction. The expert system will learn from its mistakes
through this method and apply this knowledge to the next interpolation.

13.6 CONCLUSIONS

Our ultimate goal is not just to establish one life rule for one particular joint but
to develop and establish the most efficient test methodology to address both
present and future problems. By combining the knowledge of the pertinent
engineering disciplines, we are setting down guidelines by which material prop-
erties and fatigue data should be collected. Furthermore, this will be done in an
expert system format that provides researchers a means of combining or com-
 REFERENCES 405

paring their findings with those of others and will also serve as a medium
accessible to all electronic packaging engineers.

13.7 ACKNOWLEDGMENTS

This work has been funded by the Electronic Packaging and Interconnection
Program, an industrial consortium at the University of Wisconsin-Madison.

REFERENCES

1. Seraphim, D., R. Lasky, and C. Li, Principles of Electronic Packaging, New York,
McGraw-Hill, 1989.
2. Enke, N. F., et al., "Mechanical Behaviors of 60/40 Tin-Lead Solder," Proceedings
of the 39th Electronic Components Conference, Houston, 1989, pp. 264-272.
3. Engelmaier, W., "Test Method Considerations for SMT Solder Joint Reliability,"
IEPS, Oct. 1984, pp. 360-369.
4. Harper, C. A., Handbook of Materials and Processes for Electronics, New York,
McGraw-Hill, 1970.
5. Greenfield, L. T., and P. G. Forrester, "The Properties of Tin Alloys," Int. Tin
Research Institute Publication No. ISS, 1947.
6. Clatterbaugh, G. V., and H. K. Charles, Jr., "Thermo-mechanical Behavior for
Soldered Interconnects for Surface Mounting: A Comparison of Theory and Exper-
iment," Proceedings of the 35th IEEE Electronic Components Conference, Wash-
ington, DC, 1985, pp. 60-72.
7. Bever, M. B., (ed.), Encyclopedia of Materials Science and Engineering, vol. 7,
Cambridge, MA, MIT Press, 1986, p. 5070.
8. Blake, A. (ed.), Handbook of Mechanics, Materials, and Structures, New York,
John Wiley & Sons, 1985, pp. 181.
9. Coffin, L. F., Jr., Transactions, American Society of Mechanical Engineers, TASMA,
76, 1954, p. 931.
10. Manson, S. S., "Behavior of Materials Under Conditions of Thermal Stress," NACA
Techn. Note. 2933, 1954.
11. Coffin, L. F., Jr., "A Note on Low Cycle Fatigue Laws," Journal of Materials
JMLSA 6,(2), June 1971, pp. 388-402.
12. Hirschberg, M. H., and G. R. Halford, "Use of Strainrange Partitioning to Predict
High-Temperature Low-Cycle Fatigue Life," NASA Technical Note 0-8072, January
1976.
13. Saltsman, J. F., and G. R. Halford, "An Update ofthe Total-Strain Version of SRP,"
NASA Technical Paper 2499, October 1985.
14. Halford, G. R., and J. F. Saltsman, "Strainrange Partitioning-A Total Strain Range
Version," NASA Technical Memorandum 83023, 1983.
15. Stone, D., "Creep-Fatigue Interactions in Solders and Solder Joints," Unpublished
paper, University of Wisconsin-Madison, 1988.
16. Shine, M. C., and L. R. Fox, "Fatigue of Solder Joints in Surface Mount Devices,"
ASTM STP 942, Philadelphia, 1987, pp. 588-608.
 14
Predicting Thermal and Mechanical
Fatigue Lives from Isothermal
Low Cycle Data
Harvey D. Solomon

The surface mounting of electronic components on printed wiring boards has

exacerbated the problem of the fatigue of the solder joints that attach the com-
ponent to the board. With surface mounting, the joints serve as the mechanical
as well as the electrical connection for the component. As such, they must
accommodate any relative displacement between the component and the board
and support the stresses that are developed by this displacement.
Thermal displacements, resulting from the action of temperature changes on
the different coefficients of expansion of the board and component, are a major
problem. Cyclic temperature changes cause cyclic displacements, which can lead
to thermal fatigue failures. 1 Thermal displacements are not limited to cases where
the board and component have different coefficients of expansion. They can also
develop due to temperature gradients, even when the board and component have
the same coefficient of thermal expansion. The solder will also experience strains
due to the difference in its coefficient of expansion compared to that of the
component or board.
Fatigue failures are not limited to those produced by thermal displacements.
Purely mechanical displacements of the component relative to the board can
produce fatigue failures, even when the temperature is kept constant and there
are no temperature gradients. Such displacements are developed when vibrations
cause resonance or when impacts produce repetitive inertial forces. Fatigue, such
as that due to mechanical displacements, is termed "isothermal fatigue" when it
occurs at a constant temperature.

406
 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES 407

The long-term reliability of solder interconnections is determined by their

resistance to fatigue, corrosion, creep rupture, or other such processes. This
chapter will concentrate on solder fatigue but will also consider creep and en-
vironmental influences. Solder fatigue can be estimated by accelerated life testing
of actual components and joints or by mechanical testing of modified laboratory
specimens. The former approach has the advantage of testing the actual hardware
but the disadvantage of being time consuming, with the results not easily applied
to other designs or when the acceleration approach may not be appropriate. The
mechanical approach allows for the generation of more generic data and the
separation of the various variables that control the fatigue life. This chapter is
concerned with this mechanical approach, in particular to results obtained in
isothermal tests.
Isothermal, mechanical fatigue is important not only as a mode of solder joint
failure, such as that due to vibrations, but also insofar as it can be used to
describe and predict the more complex problem of thermal fatigue. Figure 14-
1 shows the development of a displacement .:1 due to the mechanical displacement
of a component attached to a board by simple solder posts. Figure 14-1a shows
the component and board at rest. Figure 14-1 b shows the displacement of the
component to the right by an amount .:1R and FIgure 14-1 c with a displacement
to the left by .:1L• For cycling from .:1R to .:1L the cyclic displacement is .:1, which
is just .:1R + .:1L. The shear strain -y developed by this displacement is, for this
very simple case, just

.:1
-y=- (14-1)
h

where h is the height of the component above the board. Equation (14-1) is very
simplified, using the simplest solder geometry and neglecting any constraint
effects at the solder component and solder board interfaces. More detailed cal-
culations will be discussed later. The important point to remember is that a cyclic
repetition of this strain can cause a fatigue failure.
Figure 14-2 shows the same joint subjected to thermal strains. Again Figure
14-2a shows the at-rest case (i.e., at a state where there are no thermal strains).
Figure 14-2b shows the case where the temperature is increased from To to T max
and the coefficient of thermal expansion aB is greater for the board than for that
of the component ac. This larger expansion is accommodated by shearing the
solder joint. The assumption is being made that the component and board are
not strained or bent. Later on we shall consider the more general situation when
this is not the case. The expanded length of the chip carrier is given by L[l +
ac(TIIUlX - To)] and that of the board by L[l + aB(Tmax - To)). The difference
between these two lengths defines the net displacement of both joints. If it is
assumed that both joints are equally displaced, the thermal displacement of each
joint is given by Ll2(aB - ad (TIIUlX - To). Figure 14-2c shows the displacement
408 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

~ L ~I

~b_~1-r;-h--C-C------..I",.,dJ +- Solder

PWB a

cc

PWB b

cc

PWB c

Figure 14-1 Development of mechanical shear strains.

developed by cooling from To to Tmin> with the thermal displacement of each

joint being given by LI2(aB - ad (Tmin - To)·
The thermal displacement A developed by heating to Tmax and then cooling
to Tmin is given by the difference in the displacements at Tmax and Tmin or by

(14-2)
 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES 409

b!~_!1_;-h--CC-----.t...",.lJ +- Solder
PWB a

cc

~I
b

Le= L [1 HIe (TMIN - To)]

J
14 ~I

~ cc
c

Figure 14-2 Development of thermal shear strains.

We are making the assumption that the coefficient of thermal expansion is not
a function of temperature. This is not true in the general case. With a = fiT)
Equation 14-2 becomes

f
T~

d = ~ (a8(T) - ac(T))aI (14-3)

Tmin
410 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

Equations 14-2 and 14-3 assume that the displacement is taken up equally by
both the solder joints (on opposite sides of the component). This need not be
the case. In general, the factor of 2 in Equation (14-2) or (14-3) can be replaced
by a partition coefficient P that varies from 1 to 2. For P = 2 the displacement
is shared equally by the opposite rows of joints, as is the case for Equations 14-
2 and 14-3. For P = 1, all of the displacement is taken by a single side of the
component. This can develop when the soldering is done asymmetrically,
strengthening one side and causing all of the displacement to be taken up by the
weaker side.
The shear strain associated with the displacement given by Equation (14-1)
or (14-2) is approximated as just the displacement divided by the distance h
between the component and the board, that is, for temperature independent
coefficients of thennal expansion and using the partition coefficient P

(14-4)

As with Equation (14-1) the simplest geometry is assumed and more complicated
questions of geometry are left for a subsequent discussion.
Equations 14-1 and 14-4 are the same for .:1 of Equation (14-1) equal to

(14-5)

The development of a thennal strain is thus assumed to be equivalent to a

mechanical displacement.:1, where.:1 is given by Equation (14-5). Actually there
are a number of additional factors that might alter this equivalence and these
will be discussed later. They shall be neglected, however, for most of this
discussion and the assumption shall be made that the fatigue produced by a
thennal strain can be approximated by that due to a mechanical strain. This
assumption, done to simplify the problem, is a customary procedure and serves
as the basis for many design approaches used for thennal fatigue. For example,
see the reference. 2 We shall concentrate on the factors that must be taken into
account in applying isothennal test data to describe the influence of thennal
strains and then consider how this approach might be altered to correct for the
differences between isothennal, mechanical fatigue and thennal fatigue.
This chapter is based largely on the authors' data and represents his view of
the problem. No attempt was made to synthesize all the other data available,
although an attempt was made to cite it and discuss it where necessary. The
reader is directed to the other chapters of this book to get an in-depth presentation
of other views on this subject.
 LOW CYCLE FATIGUE (LCF) 411

14.1 LOW CYCLE FATIGUE (LCF)

Fatigue life is related to the imposed cyclic stress or strain. Either stress or strain
can be used. We shall consider only the use of strains because they are more
relevant to thermal cycling. Equations (14-2) through (14-4) show that thermal
cycling imposes a displacement or strain. This' strain will generate a stress, so
stress could be used to define the problem, but strain is used because it is the
imposed variable; the stresses are only derivative. The fatigue life can be ex-
pressed in terms of the applied elastic strain, the plastic strain (including all
nonelastic strains such as the anelastic strain), or by the total elastic plus plastic
strain. The elastic dependence is given by the Basquin equation,3

(14-6)

where 13,08 = constants

Nf = cycles to failure
'Ye = elastic shear strain

The plastic strain dependence of the fatigue life is given by the Coffin-Manson
law,4-7

(14-7)

where a,OCM = constants

'Yp = plastic shear strain

Morrow 8 added the elastic and plastic strains to form the total strain and thus
was able to define the fatigue life in terms of the total strain,

'YT ='Ye +'Yp =...J

0
(N) -1113 +-.L
8
(N)
0
CM
-I/a
(14-8)

Rather than use a shear strain, these expressions are often written in terms of
the shear strain range, 1l'Y. This is the strain applied between the strain limits
rather than just the maximum strain. With a zero minimum strain, the strain
range and maximum strain are the same, but when equal positive and negative
strains are employed (as is generally the case when the plastic strains are large)
the strain range is twice the maximum strain.
It is important to remember that the fatigue life Nf in Equations 14-6 through
14-8 is the same for each expression. Remember that the fatigue life for a given
412 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

total strain is Nf and that this strain is composed of an elastic and a plastic
component, either one of which could be used to define the fatigue life. This is
illustrated in Figure 14-3, which shows the fatigue life Nf correlated by the
elastic, plastic, and total strains. It is generally the correlation with the plastic
strain [Le., the Coffin-Manson law, Equation (14-7)], or the total strain corre-
lation, Equation (14-8), that is used to predict the fatigue life in thermal fatigue.
This is because the elastic strain is generally small and is therefore a poor
parameter to use. Also, stress relaxation with a fixed total strain (such as that
occurring during a hold time at the maximum temperature or during slow tem-
perature changes) causes the load to drop and the plastic strain to replace the
elastic strain. In the limit, all the elastic strain will disappear and the total strain
will be all plastic. In this case, to be discussed later, the Coffin-Manson law
should be used. Experimentally, it is also better to use the plastic strain limits
and the Coffin-Manson law because the flow stress, and hence the elastic strain,
is dependent on the strain rate employed and this makes it difficult to correlate
laboratory tests run at one strain rate and actual design conditions using another.
The strain developed by thermal cycling [Equation (14-4)] is used with Equa-
tion (14-8) to predict the fatigue life. If all of the strain is plastic (as is the case
when stress relaxation causes all of the elastic strain to be replaced by plastic
strain), Equation (14-7) is used to predict the fatigue life. The use of Equation
(14-4), or a more elaborate version of this equation, coupled with Equation (14-
7) or (14-8), is the customary method of handling thermal fatigue problems. 9 . 10
The differences in approach center on the handling of the influence of the def-
inition of the fatigue life, the cycling frequency, hold times, peak temperature,
temperature and strain rate dependence of the mechanical properties, and the

N, (Log Scale)

Figure 14-3 Fatigue life as correlated by the elastic, plastic, and total strains.
 LOW CYCLE FATIGUE OF SOLDERS-INFLUENCE OF THE DEFINITION FOR FAILURE 413

environmental and metallurgical changes that occur during thermal cycling but
not necessarily in isothermal exposures. The following considers these variables
in so far as the fatigue of solder is considered. Both eutectic and high Pb solders
are considered.

14.2 LOW CYCLE FATIGUE OF SOLDERS-INFLUENCE OF THE

DEFINITION FOR FAILURE

The problem of the fatigue failure of solder joints has prompted numerous
studies,l1-l07 particularly in the past few years. This increased interest is due to
the increased problems encountered with surface mounted components. Strain-
Fatigue Life data of the sort described by Equations (14-7) and (14-8) have been
generated. An example of trend lines from two studies is shown in Figure 14-
4. Both data sets were generated with shear loading of solder joints rather than
bulk specimens. The Solomon data (the solid lines from the reference 29 ) were
generated with plastic strain limits, whereas total strain limits were used for the
Wild data (the dashed lines from the references 13 ,14). For plastic strains that are
much larger than the elastic strains, the plastic and total strains will be essentially
the same (see Fig. 14-3), enabling both to be compared in Figure 14-4. Solomon
used the drop in the hysteresis load to define failure, and Wild used the first
appearance of a crack or a joint resistance increase of 0.03 mO to define failure.
Figure 14-4 highlights two important points: The slopes of the Solomon and
Wild curves are different, and they are displaced according to the definition used

10~------------ ______________________________ ~

60/40

....... 0.03 milliohm

······.....resistance increase
··..•··•·..~WILD)
.......................
0.01

907. LOAD DROP

0.001 +-+-++++....-+-++I......-f--+f.....II-+-+++tIIfI!--+-t-++......-+-+++tfItt-+-+-t+HIII
0.1 10 100 1000 10000 100000 1000000
FATIGUE IlFE (CYCLES)

Figure 14-4 Strain life fatigue data for two studies. (From Solomon 29 and Wild13,14)
414 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

for failure. These differences make the use of the data of Figure 14-4 to predict
device lives, difficult at best. In this section we shall first consider the influence
of the definition of failure; then the differences in the Coffin-Manson slope will
be considered.
The difference in the position of the Wild curves is rather straightforward
when one considers that, contrary to what one might expect, dc resistance mea-
surements are not very sensitive to the presence of fatigue cracks. 104 Fatigue
cycling results in the nucleation and growth of fatigue cracks, which increase
the joint resistance. The resistance increase is not, however, a linear function
of the crack length. \02 The resistance increase depends on the amount of the
measured resistance that is due to the solder layer compared to that which is due
to the rest of path being measured and to the fraction of the solder Joint that is
cracked. The ratio of the thickness of the solder crack to the total resistance path
(corrected for conductivity changes between the various materials encompassed
by the resistance path) is a scaling factor that determines the sensitivity. This is
simply expressed by Equation (14-9), i.e., by

(14-9)

where MIRo is the change in resistance for a crack, or cracks, reducing the area
conducting the current from Ao to Ae, Le is the crack thickness (separation of
the crack faces), and Lo is the length of the resistive path. As Ae increases so
does MIR o, but the change is scaled by LjLo. When Ae goes to zero, the
resistance change becomes infinite, providing the faces do not touch (Le., Le is
not zero). Even a cracked joint can pass current if nonoxidized crack faces touch.
Thus, it is possible to visually observe a fatigue crack while the resistance change
might be barely resolvable or even unresolved above the background noise. A
larger resistance change might not be produced until the crack is quite large
(Le., Ae is large relative to Ao).I04 Hence, a crack can be observed before a dc
resistance increase of 0.03 mn is observed.
The growth of fatigue cracks also explains the difference between the Solomon
curves. Here, the load required to produce the plastic strain that is being used
in the test is the measure of the fatigue crack nucleation and growth. Figure 14-
5 (from the reference25 ) shows the hysteresis loops generated in tests run with
a constant plastic strain (fixed with an analog plastic strain computer). (For
simplicity we shall talk of the strain, but it is actually a displacement that is
being measured, with the strain being calculated by dividing by the thickness
of the solder layer.) The plastic strain includes the anelastic as well as the true
plastic strain. It is determined by subtracting out the linear stress-strain portion
of the hysteresis loop.
The tilted loops of Figure 14-5 used the signal directly from the extensometer;
the vertical loops were drawn with the output from the plastic strain computer,
 LOW CYCLE fATIGUE Of SOLDERS-INfLUENCE OF THE DEfiNITION FOR fAILURE 415

_ 6e1~~ 7 ", ,IN/..

,,,, ~/1!' CroItk< v.ttl .s
;It ....... 0 006-f1' "'"
T)..~ ~(~. "".U j

r... r Cl ,....1....,' 0
"' ... ....c{J~.,.l 'J t. <;0 (NU)
~ 1Oi' .. ,/...... . II .JOi'~ .jtl""".
• '.J.S<:' [... ftJ.tl.

II 11-

N.1 ;4 'foP; 0.082

N.252 - 25 "'liT = 0.105
N.50 . 92 "'li T = 0.038 sec'
N.75 ; 260
. N.9 0 a 320

Figure 14-5 Typical hysteresis loops. (From Solomon 25 )

where the elastic strain was subtracted out from the total strain signal. The plastic
strain computer keeps the loop width (plastic strain) constant throughout the test.
As cycling continues, the load required to produce the plastic strain decreases
(the cycle number is written above the hysteresis loop). Figure 14-6 plots the
load drop parameter cI> == I - (a P//)J>M) versus the cycle number; a P is the
a
hysteresis load range at any cycle and PM is the maximum hysteresis load
range (developed on the first complete cycle or after the first few cycles). In this
experiment, the test was stopped periodically and the specimen was removed
for ultrasonic microscopy . 2~.IU~ This technique allows the cracked areas (the
shaded areas in the inserts) to be determined. As can be seen, the load drop is
associated with the increase in the cracked areas. Using a larger load drop
definition for failure requires more crack growth and more cycles; hence the
strain-life curves are displaced to longer lives.
The degree to which the fatigue life is dependent on the load drop definition
used for failure is a function of the compliance of the testing machine, the plastic
strain t:mployed in the test, and the control limit used in the test. 107 With total
displacement testing limits, a relatively soft testing machine, and small plastic
strains, the load decrease is abrupt, with little difference between the number
of cycles for the initial load drop and larger load drops . This is caused by increases
in the plastic strain range with cycling. The data of Figure 14-6 through 14-10
were, however, generated with plastic strain limits. In effect, this is equivalent
to using an infinitely stiff testing machine (because the plastic strain computer
416 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

~\
( -
I 60/40 35°C
l1yP- 2.3%
0.3 Hz

Figure 14-6 Load drop parameter <I> versus the cycle number of an interrupted test. The
inserts show the cracked area as measured by ultrasonic microscopy. (From Solomon 25 )

10~-------------------------------------------------,

3S'C
(a defined with respect to FL)

0.1

0.01

507. L.D. 907. L.ll.

a=0.59 a=0.52
0.001 +--+--<I-+++ttMf--I-++++HII--+-+-t+f+ltt---t-I-+++ttMf--I-++++HII--+-+-t+f+ltt---+-;I-++HIII
0.1 10 100 1000 10000 100000 1000000
FATIGUE LIFE (CYCLES)

Figure 14-7 Low cycle fatigue life curves for 60% Sn-40% Pb with the fatigue life defined
by different load drops. (From Solomon 29 )
 LOW CYCLE FATIGUE OF SOLDERS-INFLUENCE OF THE DEFINITION FOR FAILURE 417

10

60/40 3S'C

~ tlyP=0.56
Il.
0
~
t=l
t=l
<
0
...1
0.1

10 100 1000 10000 100000

CYCLES

Figure 14-8 Load drop parameter !lJ versus the cycle number for different applied plastic
strains for 60% Sn-40% Pb solder.

35'C 8.3 Ifz

~
Il.
0 8.1
~ 6yP = 0.0033
~

~
0
...1

92.5% Pb - 5.0% Sn - 2.5% Ag

8.81
1 18 188 188B 10,000 100,000
CYCLES

Figure 14-9 Load drop parameter !lJ versus the cycle number for different applied plastic
strains for 92.S% Pb-S.O% Sn-2.S Ag solder. (From Solomon lO&)
418 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

1 8 : - - - - - - - - - - - - - - - - - - - - - - - - - - -________________-,

x o.~~92.5% Pb - 5.0%Sn -2.5%Ag

~
CIl
x .~
N . +

.... '~
U
xD~'
o. +', ++
~ 0.1 x D • +

~~~
8.81

LOAD DROP = 18~ Z5~ 58~ 98y.

= 8.95 8.B5 8.78 8.6B
8 .881 +-+-t-++liffl--+-++It-+i*I-iH-H+tHf--Hi--H'HIfIo-+++-++Itit-+++Hiill--+-+-+++++!I
8.1 18 188 1888 1B888 188888 1888888
FATIGUE IJFE (CYCLES)

Figure 14-10 Low cycle fatigue life curves for 92.5% Pb-5.0% Sn-2.5% Ag solder with the
fatigue life defined by different load drops. (From Solomon 106)

subtracts out the elastic strains and controls only on the plastic strains). The
plastic strains are kept constant and the load drop is gradual rather than abrupt.
The fatigue life is therefore a strong function of the definition of failure and this
must be accounted for before the data can be used in any meaningful manner.
The problem of the dependence of the magnitude of the fatigue life on the
definition of failure is inherent in all fatigue tests, but it becomes most critical
when one tries to predict the life of solder joints. This is because the actual
joints are generally smaller, often by an order of magnitude, than the laboratory
joints with which the strain life data was generated. With larger equipment, such
as turbine discs or aircraft frames, the laboratory specimens are generally smaller
than the actual device that they are meant to simulate. In these cases the lives
determined from laboratory specimens are generally conservative since a crack
growing through the laboratory specimen would be contained within the larger
device (provided brittle fracture does not intervene, in which case there may be
little or no difference between the laboratory test and the part life). This is not
the case when the device (in this case a solder joint) is smaller than the laboratory
specimen and brittle fracture does not occur.
Another factor that makes the dependence of the fatigue life on the definition
of failure critical for solder joint life prediction is the fact that the loading is in
shear on a solder layer. In this case, the crack growth rate is, to a first approx-
imation, only a function of the applied plastic shear strain and not also a function
of the crack length. 25 This conclusion is basedlO9 on Rice's analysis I 10 of the J
integral of a crack in simple shear, with semi-infinite boundary conditions. It is
supported25.29.89.106 by the observation that, for tests run with a constant plastic
 LOW CYCLE FATIGUE OF SOLDERS-INFLUENCE OF THE DEFINITION FOR FAILURE 419

strain, the load drop (which is proportional to the crack growth rate) does not
increase as the cracking gets more severe (for instance, see Figure 14-6). If the
applied plastic strain is kept constant, to a first approximation, the crack growth
rate will be constant. This means that increasing the specimen size increases the
fatigue life because the crack must grow longer to produce failure. It also means
that the life of a test specimen will be longer than that of a smaller solder joint
subjected to the same plastic strain.
With tensile loading (which is more typically used in other design situations),
the crack growth rate increases as the crack length gets larger (because the stress
intensity and J integral are functions of the crack length as well as the applied
plastic strain or stress level). As the crack grows longer, its growth rate increases.
Increasing the specimen size produces a less than proportional increase in the
fatigue life. For instance, doubling the length to which a crack must grow to
produce failure does not double the number of cycles required to achieve this
crack growth (because the crack growth rate is increasing as the crack length is
getting larger). This tends to minimize the influence of the laboratory specimen
size relative to that of the device.
With a crack growth rate independent of the crack length, the crack growth
rate can be easily determined from the LCF data. For a constant applied plastic
strain, which to a first approximation makes the crack growth rate constant, the
cyclic crack growth rate da/ MY is just the distance D over which the crack must
grow to produce failure, divided by Nf , the number of cycles to produce failure,
that is,

da D
-=- (14-10)
MY Nf

As will be seen, D can be thought of as a normalization parameter for the

different load drop data. The larger the load drop, the larger Nf and hence the
larger D to give a constant crack growth rate for a given applied plastic strain
range.
For a real joint subjected to the same plastic shear strain, the crack growth
rate should be the same. Failure would then be determined by the length Lover
which the crack must grow. Equating the crack growth rate in the laboratory
test (for the crack growing a distance D in Nf(lab) cycles) with that which should
occur in a real joint one gets

(14-11)

The fatigue life should scale according to the ratio of LID. The fatigue life,
developed in a joint that is smaller than that used in laboratory tests, will be
420 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

smaller by LID. The reverse is true if the joint of interest is larger than that of
the laboratory joint.
Distance D is determined not only by the geometry of the laboratory specimen
but also by the definition used for failure. For instance, for a 50 percent load
drop definition for failure, the cracked area will be one-half the joint area (as-
suming that all of the load drop is due to cracking of the joint and that fatigue
softening or hardening can be neglected). For a 25 percent load drop definition
for failure, the cracked area must be one-quarter of the joint area. The exact
determination of D is difficult because the cracks do not have simple linear fronts
that extend across the width of the specimen and grow in the direction of shearing.
The crack front is irregular and while it grows in the general direction of shearing,
it does not do so in a completely linear manner. However, some simple ap-
proximation can be made by assuming that the crack front is linear and goes
across the width of the joint.
Figure 14-7 shows the fatigue life curves for 60% Sn-40% Pb solder with
failure defined by 10, 25, 50, and 90 percent decreases in the hysteresis load
range required to produce a given plastic strain range. Two difficulties are
apparent in applying the idea of using a D value to normalize this data. First,
the curves are not displaced according to the load drop, that is to say, the 50
percent load drop curve is not at twice the fatigue life of the 25 percent load
drop curve or at 5 times the life of the 10 percent load drop data. Secondly, the
curves are not parallel; that is to say the Coffin-Manson exponent a is dependent
on the load drop definition for failure. These problems can be handled by as-
suming that a different number of cracks are growing at each load drop level
(an assumption that is supported by the ultrasonic microscopy data).
The joint used for the data of Figure 14-6 is 0.1 x 0.5 in (2.54 X 12.7 mm).
Assuming that the crack growth is in the 0.5-in (12. 7-mm) direction (the direction
of shearing), a single crack propagating to reduce the load by one-half would
require crack growth of 0.25 in (6.35 mm). Figure 14-6 shows that the actual
situation is more complex because there is more than one crack growing. If two
cracks were present, the average crack growth of each, D, is 0.25 / 2 = 0.125
in. (3.18 mm). Figure 14-6 also shows that the number of growing cracks
decreases, due to crack coalitions, as the test progresses. At the start of the test
there may be more than four distinct cracks present, but by the time the load
has dropped by 90 percent there are only one or two cracks present. It was
found 105 that the fatigue curves for different load drops could be normalized by
assuming that for a 25 percent load drop, there were six growing cracks, for a
50 percent load drop there were three growing cracks, and for a 90 percent load
drop, there were one and one-half growing cracks.
If the strain range in a joint is t:.-Yp over a length L, the life is given by Equation
(14-10) for Nf and D defined by the same load drop and Nf determined by t:.-Yp.
For instance, in Figure 14-7, for a plastic strain of 10 percent, the fatigue life
 LOW CYCLE FATIGUE OF SOlDERS--INFLUENCE OF THE DEFINITION FOR FAILURE 421

in this laboratory test will be 9, 50, 170, and 400 cycles for the 10,25,50, and
90 percent load drop curves, respectively. The value for D will be different for
each load drop, reflecting the different crack lengths required to produce this
load drop and differences in the number of cracks that may be present. Assuming
that 10, 6, 3, and 1.5 cracks are present at 10,25,50, and 90 percent load drop,
respectively, D will be 0.05 I 10 = 0.005, 0.125 I 6 = 0.021, 0.25 I 3 =
0.083, and 0.45 I 1.5 = 0.3, respectively. These values can be thought of as
correction factors to normalize the different fatigue curves shown in Figure 14-
7, that is, the values of Nfl D are roughly constant (at worst, within a factor of
2 or 3 for different load drops and plastic strain ranges). Furthermore, the exact
choices, for the number of cracks growing at any load drop, were made to
correlate laboratory data and tests run\05 on actual chip carrier-printed wiring
board (CC-PWB) joints. The ratio of the D values normalizes the different load
drop curves, but the choice of the exact D value was based on the CC-PWB
joint data of the reference. \05 D is not, however, a completely disposable curve
fitting parameter since the values used are based on actual joint geometries and
the number of cracks assumed to be growing is consistent with the ultrasonic
microscopy measurements.
If all the curves were parallel and displaced according to the load drop (Le.,
the 50 percent load drop curve was at just twice the fatigue life of the 25 percent
load drop curve), the curves would be normalized by assuming that the same
number of cracks was growing at different stages of the test (instead of using a
different number of cracks) and that a 50 percent load drop required just twice
as much crack growth as a 25 percent load drop. This was the sort of behavior
observed at 150aC. 29
The fatigue curves can also be normalized by calculating a crack growth rate
directly from the load drop curves. Figure 14-8 shows some of the load drop
curves used to develop Figure 14-7. Here the test was not interrupted and a
single straight line is developed for each experiment run with a different applied
plastic strain range. These curves fit to a simple power law expression, that is,

(14-12)

where the constants A and B are functions of the applied plastic strain range
according to

(14-13)

and

(14-14)
422 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

Assuming that the crack growth rate da / dN is proportional to the load drop
rate dcp/dN, and using Equations 14-12 through 14-14, it can be shown 29 that

(14-15)

If one further assumes that the fatigue crack extends across the width of the
joint, propagating in the length direction, the crack growth rate will be equal to
the right-hand side of Equation (14-15) instead of being just proportional to it.
Equation (14-15) assumes that a single crack is present and corrects for the
presence of multiple cracks or other factors that might influence the load drop
curves by altering the constants Ao, AJ, Bo, and B,. However, Equation (14-15)
can be difficult to use because unless B = 1 da/dN will be a function of N.
Since B is generally less than 1, this reflects a slowdown in the load drop, which
was handled in the previous approach by decreasing the number of cracks that
were assumed to be growing.
When B = 1, Equation (14-15) can be integrated to yield Equation (14-7)
with 0: = lI(A, + B,) and for B, = 0 and Bo = 1 (which gives B = 1) 0: =
lIA,. With B not equal to 1, the correlation of 0: with A and B is more complex
but describes the variation in 0: with the definition of failure. 29 This is the situation
that prevails for 60% Sn-40% Pb solder at 35°C. At 125 and 150°C B, is small
enough to be considered zero and Bo is close to I, which keeps B = 1 and 0:
independent of plastic strain and the definition of failure. 29
The problem of the influence of the definition of failure on the fatigue life is
not just a problem for 60% Sn-40% Pb solder; it is a general one. This is shown
by the load drop data of Figure 14-9, which is for a Pb-rich solder (92.5% Pb-
2.5% Ag-5.0% Sn), and by the fatigue life curves that can be determined from
this data (i.e., Figure 14-10). As with the 60% Sn-40% Pb data, the fatigue life
curves are not parallel (i.e., 0: decreases with increasing the load drop definition
for failure). The same D values will normalize the curves, that is, give roughly
(at worst within a factor of 2 or 3) the same values for NiD for different load
drop definitions of failure, but tests of the sort described in the reference lO5 have
not been run, so it is not demonstrated that these D values will correlate the
laboratory data with tests on actual joints. Also in agreement with the 60% Sn-
40% Pb data, at 150°C this high Pb solder exhibited parallel straight strain life
Coffin-Manson curves, 106 but again correlations of laboratory and tests on actual
joints are lacking.
The important points to remember are

1. In tests using plastic shear strain limits, the rate of drop in the hysteresis
load required to produce a given plastic strain does not accelerate as the
load drops. This means that the fatigue life to produce a given load drop
is highly dependent on the choice of the load drop.
 INFLUENCE OF THE TEMPERATURE 423

2. This effect should be less in tests using tensile loading because the crack
growth rate will accelerate as the crack gets longer, thereby causing the
rate of load drop to accelerate. This lessens the differences between the
fatigue life defined by different load drops.
3. When the definition of failure has a large influence on the fatigue life, it
is necessary to normalize the data to reduce this effect. This was done by
treating the LCF data as if it were crack growth data. This necessitated
several simplifying assumptions, but at worst the normalization could be
done to within a factor of 2 or 3 (generally the normalization was better
than a factor of 2, especially when the 10 percent load drop data are
disregarded) .
4. The question of the influence of the definition of failure is particularly
important when considering small solder joints. Here the laboratory test
joints are often larger than the joint of interest. The combination of a larger
laboratory joint and a dependence of the fatigue life on the definition of
failure can make uncorrected laboratory fatigue data highly nonconserva-
tive.

Subrahmanyan et a1. 95 •97 have also treated LCF data in terms of crack growth,
only they use a time-based approach rather than a cycle-based approach; therefore
a time dependent constitutive equation can be used. They employ a crack length
integral that defines the total crack length to failure and the crack growth per
cycle that is defined as the damage integral. While their approach uses a different
crack growth expression (they use a tensile expression in which the crack growth
is related to the crack length and a rate constant), the final equation for the
fatigue life is similar to Equation (14-11). They relate failure to the total crack
growth (defined by the crack growth integral), divided by the crack growth per
cycle (defined by the damage integral). This is in effect what is done by Equations
(14-10) and (14-11). Subrahmanyan et al. 95 have successfully used their approach
to apply Solomon's data25 ,29 to predict the thermal fatigue results of Hall and
Sherry55 for various temperature changes.

14.3 INFLUENCE OF THE TEMPERATURE

The use of isothermal LCF data to predict the behavior of thermally cycled joints
hinges on the assumption that the strains developed by thermal cycling can be
modeled by mechanical strains developed at a constant temperature. The question
arises as to what temperature to use to model cycling over a range of temperature.
Taira9 considers two possibilities: the use of the mean temperature for the thermal
cycle or, when creep and high temperature environmental interactions are pos-
sible, the use of the highest temperature. Fortunately, for 60% Sn-40% Pb solder
and for the high Pb solder discussed here, there is not a great deal of difference
424 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

in the fatigue behavior at these different temperatures. This is shown in Figures

14-11 and 14-12 for a 50 percent load drop.24,J06 There is a temperature depen-
dence of the fatigue life, but it is not very large one nor, considering the scatter
in the data, is it a very significant one. For the 60% Sn-4O% Pb solder the fatigue
life for a plastic strain range of 5 percent (a middle range strain), the fatigue
life decreases from 500 cycles at 35°C to 200 cycles at 125°C and to about 120
at 150°C (not shown in Figure 14-11 but provided in the reference24 ). The high
Pb solder shows a somewhat lower temperature dependence of the fatigue life.
The 5 percent plastic strain life at 35°C is 700 cycles and decreases to only 300
cycles at 150°C.
Norris and Landzberg70 also provide data concerning the temperature depen-
dence of the fatigue life. The data they present is for a 95% Pb-5% Sn solder
(similar to the high Pb solder of Figure 14-12) but for a very small total strain
(0.03 percent), which for the high cycling frequency of 1800 cycles/min, will
develop even a smaller plastic strain. The fatigue life was reduced from 3.5 x
lOs cycles to lOS cycles when the temperature was increased from 35 to 150°C.
This factor of 3.5 is similar to that shown in Figure 14-12 where the ratio of
the fatigue lives is about 2.5. At - 50°C, however, Norris and Landzberg show
a factor of 200 increase in the fatigue life over that at 150°C. This may be a
manifestation of the difference in the slopes of the fatigue life curves shown in
Figure 14-12. The lower slope of the -50°C data curve means that when ex-
trapolated to very small plastic strains, the fatigue life will be greatly increased.
At large plastic strains, however, the fatigue life at - 50°C is actually lower

10 1
11'" 0.3 Hz
A'Y/
8 = 1.14
'Yf
10°

t.Yp
10- 1

Nt (cycles)
Figure 14-11 low cycle fatigue curves for a SO percent load drop definition of failure for
tests on 60% Sn-60% Pb solder run at - SO, + 35, and + 125°C. (From Solomon 24)
 INflUENCE OF HOLD TIMES AND CYCLING FREQUENCY 425

l0r-----------------------------------------~
92.5% Pb - 5.0% Sn - 2.5% Ag
(Indalloy 151)
507. LOAD DROP
z
:;:
Il::
Eo-<
C/l
0.1
U
e::
C/l
j
P-
0.01

0.001
0.1 10 100 1000
+-~I-++++ttI--t_++t+......_+_+_t_Ht__+_++;_t+___~_ __t_+_<_ftI
10000 100000
FATIGUE LIFE (CYCLES)

Figure 14-12 Low cycle fatigue curves for a 50 percent load drop definition of failure for
tests on 92.5% Pb-2.5% Ag-5.0% Sn solder run at - 50, + 35, and + 150°C. (From Solomon 106)

than that at 150°C. This points up a significant error that can develop by just
mUltiplying the Coffin-Manson fatigue law by a temperature correction factor
(as is done by Norris and Landzberg) instead of using the correct Coffin-Manson
curve for the temperature of interest.
The data of Figures 14-11 and 14-12 show that the fatigue life is a function
of the cycling temperature but that even a factor of 2 or 3 in the relative fatigue
life is not very significant with respect to the scatter in the data. The most
conservative approach would be to use the peak temperature since this predicts
the lowest fatigue life. This is also the temperature to consider when there is a
hold time at this peak temperature because as we shall see in the next section,
this will cause much of the strain to be developed at this temperature.

14.4 INFLUENCE OF HOLD TIMES AND CYCLING FREQUENCY

The previous discussion has been limited to the results of tests that used a
relatively high cycling frequency (0.25 Hz or higher). Cycling at a relatively
high frequency speeds up laboratory tests but does not duplicate the actual service
behavior of most solder joints.
Service temperature changes seldom ramp up (experience a linear temperature
increase) and then ramp down. Rather the temperature is held constant for a
period of time before changing. In the case of real service joints this hold time
may be a matter of hours, a day, or even longer. To speed up laboratory testing,
laboratory tests that incorporate a hold time typically use hold times of an hour
or less. Another laboratory approach is to use a slower ramp rate (Le., a slow
linear temperature change of strain rate) with no hold time instead of using a
426 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

rapid ramp rate and a hold time. The use of a hold time or slow ramp rate is
important because hold times or slow ramp rates reduce the fatigue life relative
to that achieved with a high ramp rate and no hold time. It is thus necessary to
correct the fatigue life measured at a rapid cycling rate when applying this data
to a design situation that incorporates a hold time or slow cycling frequency.
The question of the influence of hold times or slow cycling was first studied
in Pb ll1 ,112 in the early 1950s. This work was the basis for studies ofthe influence
of hold times and slow cycling on the high temperature fatigue of superalloys 113-116
and more recently of solders.70 A eutectic solder at 25°C is at 65 percent of its
melting point (in degrees K), which is a higher relative use temperature than
that of most superalloys. It is therefore not surprising that the fatigue of solders
must be handled by using many of the same techniques developed for superalloys
and that these techniques were first developed to describe the fatigue of Pb, for
which room temperature is also a relatively high homologous temperature.
Two different approaches have been developed to describe frequency and hold
time effects. Coffin 113,115 developed a frequency modified version of the Coffin-
Manson equation and Manson, et al. 1l7 ,IIS developed strain range partitioning
(SRP). We shall first consider Coffin's frequency-modified Coffin-Manson equa-
tion, which is given by

(14-16)

This is the same as Equation (14-5) but with the addition of a frequency v term,
V(K - 1), which modifies the fatigue life. Both ramp and hold time effects are

considered because the frequency is just the inverse of the period, which is the
sum of the ramp times (i.e., the time to load and unload) and the hold times
(i.e., hold times at the maximum and or minimum loads or strains). The effects
of the frequency are determined by the magnitude of K. For K = 1, there is no
dependence of the fatigue life with frequency variations and the frequency term
is just equal to 1. For K = 0 the fatigue life is modified by lIv. Since Nlv is
just the time to failure, this results in the development of a constant time to
failure (for a given applied plastic strain range). When K = 0, if the frequency
is halved, the number of cycles to failure is halved, to give a constant time to
failure.
Figures 14-13 through 14-15 show the fatigue life as a function of the cycling
frequency for tests on 60% Sn-40% Pb solder, run at 35 and 150°C. 27 All of
the data can reasonably be fit to two curves, one with K equal to or close to 1
and the other with K equal to or close to O. There is thus a frequency above
which there is little or no frequency effect and below which there is a large
frequency effect. This behavior is typical of what is often observed in superal-
loYS.II4-116 The existence of two regimes with different K values presents a
complication to the use of the frequency-modified fatigue life but does not prevent
its use. Over the range of about 10--4 to 1 Hz, a single K value can be used
 NFLUENCE OF HOLD TIMES AND CYCLING FREQUENCY 427

HI G1
SI G SI G PATCHES •
H I S I
I • ! K=0.84
G
I
.-
.-
.- K=0.68- •
t

1~-o.42
G

• 35'C ~'YpN1 0%
N,= '\)1'K~y/la<l>1/a

v (Hz)

Figure 14-13 Fatigue life versus cycling frequency for 60% Sn-40% Pb solder tested at 35°(
and with an applied plastic strain of 10 percent. The notations G, H, 5, etc., refer to the nature
of the fracture surface. (From Solomon 27 )

without too much error. This was the approach taken by Norris and Landzberg. 7o
Based on the lead data mentioned earlier, they used a parameter - k instead of
K - I as is used in Equation 14-16. For a 99% Pb-l% Sb alloy, at 30°C, they
found k = 0.33, which corresponds to K = 0.67. This is in good agreement
with the average value for 60% Sn-40% Pb solder, shown in Figure 14-13.

IG IG
I I I
•
• • •
-- --
•
KO
= .---
• ---
:;,..-- ..
K=0,45

••

150'C ~YpN5%
N,= '\)1.K~y/la<l>1/a

11 (Hz)

Figure 14-14 Fatigue life versus cycling frequency for 60% Sn-40% Pb solder tested at
1500 ( and with an applied plastic strain of 5 percent. The notations IG, H, etc., refer to the
nature of the fracture surface. (From Solomon 27 )
 428 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

-
, ,
~rp 10%

IG IG IGtS
~ ~ StH IG S S
• ~ ~ H
en •
.• --.
•
--
Q)

•
-.
13 •
--.
K=1
~
>- ........ -== :::I:
:ct • • x
10'

• t
StH

10~0-5 10- 4 10- 1 10°

11 (Hz)
Figure 14-15 Fatigue life versus cycling frequency for 60% Sn-40% Pb solder tested at
150·C and with an applied plastic strain of 10 percent. The notations IG, H, etc., refer to the
nature of the fracture surface. (From Solomon 27 )

1000
e

35°C
ex
f1Yp -10% x
-
(/)
Q)
100
e

-
~ e
(J

;z,-
10

1~~~~~~~~~L--L~~~_~-W~ll--L~WU~

,~& 1~4 1~3 1~2 1~ 1~

71 {H J }

Figure 14-16 Comparison ofthe 35·C fatigue life Nf versus cycling frequency for continuous
cycling and cycling with a hold time for 60% Sn-40% Pb solder. (From Solomon 28)
 NFLUENCE OF HOLD TIMES AND CYCLING FREQUENCY 429

x
1000

150°C • x
x

!1yP-5% • • •
Ii)
Q)
100
x •
(3
>- x.
~
:ct ·xe
10
.x
•

x •
1L-~LL~~-L~~~ __LJ~~L-~~~~~~LW~.
1~5 1~4 1~J 1~2 1~
1/ (H 3 )

Figure 14-17 Comparison of the 150°C fatigue life N( versus cycling frequency for contin-
uous cycling and cycling with a hold time for 60% Sn-40% Pb solder. (From Solomon 2B )

Figures 14-16 and 14-17 show results for hold time tests 28 superimposed on the
continuous cycling data of Figures 14-13 and 14-14. As can be seen, the results
are similar, with perhaps a slightly larger fatigue life for the hold time tests
(compared at the same frequencies). Vaynman lOl has shown that a high Pb solder
(96.5% Pb-3.5% Sn) exhibits similar frequency-life behavior as is shown in
Figures 14-13 through 14-17 but with the break in the curve occurring at about
0.01 Hz rather than at the lower frequency exhibited by the 60% Sn-40% Pb
solder. This solder also exhibited a more pronounced transition from transgran-
ular to intergranular fracture with increasing hold time.
Figures 14-13 through 14-17 show that the transition from K = 1 to K = 0
is temperature dependent. With a hold time, one should use the data for the
temperature of the hold to determine the appropriate K value to use. There is
also a question of how the applied strain (or strain developed during the hold
time) and the frequency effect interact. Is K or the transition frequency dependent
on the applied plastic strain? Figure 14-18 shows the 35°C strain-life data, with
two assumptions regarding the relative interaction between the plastic strain and
frequency effect.24 One set of curves assumes that K and the transition frequency
are independent of the applied plastic strain. The dotted line shows another
extreme. Here it is assumed that there is a single K = 0 curve (as seems to be
the case at 150°C,24) with the transition frequency given by the intersection of
the K = 0 and K = 1 curves. At frequencies below this transition frequency,
430 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

10 5
.004
35°C
10 4
0.02

10 3 0.04
0.06
0.08
en 0.10
~ 10 2
>-
~ 0.20
<:,-
10 1 0.40
0.60
0.80

10° Mp =1.0

II (Hz)

Figure 14-18 Fatigue life as a function of the cycling frequency and applied plastic strain
(curves established from 35°C 60% Sn-40% Pb solder data. (From Solomon 24 )

the predicted fatigue life ranges between the dotted and solid curves. At higher
frequencies it is just given by the solid curve. At low frequencies there can be
orders of magnitude between the two predicted extremes. Using the dotted curve
is an extremely conservative approach, perhaps overly so.
With a single K = 0 curve (the dotted curve of Figure 14-18) to describe the
data, the fatigue life is no longer a function of the applied strain and the Coffin-
Manson expression is no longer valid. This represents an extreme limit. In
general, however, below the transition frequency, a of the Coffin-Manson
expression will be a function of frequency. It will increase,24 but not necessarily
to infinity, as is implied by the dotted curve of Figure 14-18.
The strain range partitioning approach assumes that there are different Coffin-
Manson curves for creep and time independent plastic (noncreep) strains. The
problem then becomes one of determining the fraction of creep making up the
hysteresis. An interactive damage law is used to add up the life determined by
the fraction creep, that due to the fraction time independent plastic (noncreep)
 INFLUENCE OF HOlD TIMES AND CYCLING FREQUENCY 431

defonnation, and the fraction due to mixed cycling with creep on loading and
plastic (noncreep) on unloading or vice versa. The creep Coffin-Manson fatigue
life is usually less than that of the noncreep curve. Thus the overall fatigue life,
as detennined by the interactive damage law, is reduced when the fraction of
creep is increased.
The strain range partitioning approach can also be applied to thennal fatigue l17
by developing the stress-strain hysteresis and determining the fraction of creep
for each strain increment (each occurring at a different temperature, in the
temperature cycling range).
Reducing the cycling frequency, as in done in the data of Figures 14-13 through
14-15, results in more creep defonnation. A simple way to detennine the amount
of creep in the hysteresis is to compare the creep strain with that developed by
dynamically straining the test specimen. This comparison is made for 60% Sn-
40% Pb solder at 35 and 150°C in Figure 14-19. 26 A single stress-strain rate
curve describes the strain rate data, independent of the strain range and cycling
frequency employed. This means that the defonnation of the 60% Sn-40% Pb
solder is Non-Newtonian viscous, in that the flow stress is related to the strain
rate raised to a power other than 1. The strain rate and creep curves are roughly
parallel, meaning that the fraction creep does not change very much as the cycling
frequency is reduced. Figure 14-20 shows computer calculations26 of the per-
centage of creep developed from creep curves and strain rate data (a more accurate
approach than by just comparing the curves of Figure 14-19). At 35°C, at high
strains, there is an increase in the fraction of creep when the frequency is
decreased from 0.25 Hz to about 10-2 Hz but much less of an increase when
the frequency is reduced still further. This increase in the fraction of creep should
decrease the fatigue life, but as Figure 14-13 shows, there is no decrease in the
fatigue life when the frequency is reduced from 0.25 to 10-2 Hz. There is no

~104.-----.-----.-----.-----.------.-----.-----,
0.

.~I(\J
~t03
....
(/)-
<.n
I.LI

1=10
(/)
2
0:
<t
I.LI
::x:
(/)I01~----~----~~--~~--~~--~~--~~--~
10- 8 10-7 10- 6 10- 5 10- 4 10- 3 10- 2 10- 1
STEADY STATE CREEP RATE AND STRAIN RATE (sec-I)

Figure 14-19 Comparison of 60% 5n-40% Pb solder creep and strain rate sensitivity data.
(From 501omon 26 )
432 PREDICTING THERMAL AND MECHANICAL FATIGUE blVES

100~~~~~~~~~~~~~~~~~~~

80 150°C 60/40
~_ _ _=-~ t.rp
t.rpNIO%
60 o t3 t.rp N 5%
0
~ 40 ~~ N 2%
-2.... 20 0----- - '-J t.rp N 1%
a..
w
w 100~---------------------------------------4
a:: 80 60/40
() 60
40

2~ L--'----l-~~12~~~~~L....LLl~
10- 5 10- 3 10- 1
V (Hz)

Figure 14-20 Calculated percentage of creep in the hysteresis behavior of 60% Sn-40%
Pb solder cycled at 35 and 150°C. (From Solomon 26 )

increase in the fraction creep below about 2.5 x 10--4 Hz, but there is a large
decrease in the fatigue life. For a given applied plastic strain at 150°C there is
little or no change in the percentage of creep, even though the fatigue life shows
a marked decrease when the frequency is reduced below about 10-3 Hz.
It is clear that strain range partitioning in its standard fonn does not describe
the 60% Sn-40% Pb solder data presented here (see the reference 27 for a more
detailed comparison of SRP predictions and the test data). This failure stems
from the fundamental assumption of SRP that when the defonnation is all by
creep, a characteristic fatigue life is generated for this creep strain. Lowering
the cycling frequency (or increasing the hold time) should not therefore alter the
fatigue life, providing that the defonnation remains all by creep and that the
applied strain is not altered when this change is made. Figures 14-13 through
14-17 show that this assumption is not valid for 60% Sn-40% Pb solder in the
tests described here. SRP can, however, be applied in a slightly modified fonn
if it is assumed that when the defonnation is 100 percent creep the time to failure
and not the number of cycles to failure reaches a characteristic value tcc for the
applied plastic strain. Now, when the frequency is reduced, or the hold time
increased, the number of cycles to failure will decrease, so the time to failure
tf does not change (Le., Nfv = tfo which becomes a constant with tf = t cc ).
The data of Figures 14-21 through 14-23 show the results of Figures 14-15
through 14-17 in tenns of this time-modified SRP approach. The tenns fcc and
fpp = 1 - fcc follow the SRP nomenclature and refer to the fraction of the
applied strain that is creep or noncreep. The fraction creep is detennined from
Figure 14-20. TheNpp tenn is the fatigue life when the defonnation is all noncreep
and is approximated by the fatigue life observed when the highest frequency
 INFLUENCE OF HOlD TIMES AND CYCLING FREQUENCY 433

3S'C ~Yp-1 0%

TI ME MODIFIED SRP

Figure 14-21 Time to failure tf versus the cycling frequency v, as correlated by time-
modified SRP and frequency-modified LCF, at 35°C for a plastic strain of 10 percent. (From
Solomon 27)

(0.25 Hz) was employed. As can be seen, this approach describes the approx-
imately constant time (tf = tee) to failure that is observed at low cycling fre-
quencies or long hold times. Time-modified SRP is completely analogous to the
frequency-modified approach with K = 1 at high frequencies and K = 0 at low
frequencies (K = 0 defines a constant time to failure and K = 1 defines a
constant number of cycles to failure).27
Enke et al. 1OO have also studied the influence of frequency on the fatigue
behavior of 60% Sn-40% Pb solder, using SRP. In agreement with the results
presented here they found no influence of cycling frequency when 0.05-Hz tests
were compared to those run at 0.017 Hz. The data of Figures 14-13 and 14-16
show that the frequency would have to have been reduced to below 0.0002 Hz
before a significant influence of frequency would be noted.
While conventional SRP does not appear to explain the 60% Sn-40% Pb solder
data presented here, it may do so for high Pb solders. Figure 14-24 119 shows
the creep rate-strain rate comparison for a high Pb solder (the 92.5% Pb-2.5%
Ag-5.0% Sn solder described previously). This solder does not display viscous
434 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

150'C 11:'(,,-5%

10 6 FREQUENCY MODIFIED LCF

I I u

.--. ••
-:-+-
tcc= 5 x 104 t, tee Npp
"",. •
10 5
- tcc=1US

- 104
0
(])
en
...:r

•
TIME MODIFIED SRP
10 3
I fcc ufpp Npp=100 x
-:-+-- ( Npp=100 )
t, tee Npp Icc= 5 X 10' •

10 2
10- 5 10-4 10-3 10- 2 10- 1 10 0
11 (Hz)

Figure 14-22 Time to failure tf versus the cycling frequency v, as correlated by time-
modified SRP and frequency-modified LCF, at 150°C for a plastic strain of 5 percent. (From
Solomon 27)

behavior. Hence, there are separate curves for each plastic strain that was used
in the testing. Here, the creep curves diverge from all of the strain rate curves,
meaning that the fraction creep will increase as the cycling frequency decreases.
In agreement with SRP, the fatigue life decreases as the cycling frequency
decreases and the fraction creep increases. Work is currently going on to see if
conventional SRP indeed describes these results.
Fox, Shine, and co_workers32.120.121 have developed yet another approach to
describe the influence of cycling frequency and hold times. They correlate the
matrix creep that is developed during cycling or during a hold time in a Coffin-
Manson expression that has an a of 1.0. This means that the product of the
matrix creep and the fatigue life is a constant. This is also a version of the
Monkman-Grant model,122 in which the product of the time to failure and the
creep rate is a constant. (The number of cycles to failure times the time per cycle
is the time to failure, and the matrix creep strain per cycle divided by the time
per cycle is the matrix creep rate. Thus, if the product of the matrix creep per
cycle and the fatigue life is a constant, the product of the time to failure and the
matrix creep rate per cycle will also be a constant.)
 INFLUENCE OF HOLD TIMES AND CYCLING FREQUENCY 435

1S0'C L1Yp -10%

TIME-MODIFIED SRP
;
•• I _ fCC ufpp
---t--
t, tcc Npp

•
•

FREQUENCY-MODIFIED LCF
I I u
-:-+-
t, tcc Npp

10- 3 10- 2
11 (Hz)
Figure 14-23 Time to failure tf versus the cycling frequency v, as correlated by time-
modified SRP and frequency-modified LCF, at 150°C for a plastic strain of 10 percent. (From
Solomon 27 )

10000...-----------------------,

92.5% Pb - 5.0% Sn - 2.5% Ag (INDALLOY 151)

Cii
~
en
en
r.:I
~ 1000
E-<
en

~
en

100+-_ _+_i~.....- ...............+tt<tt__ _+_i..........- + _ 1 _ + 1_ _ __+_it_t+H+I

0.000001 0.00001 0.0001 0.001 0.01 0.1
CREEP RATE and STRAIN RATE (l/sec)

Figure 14-24 Calculated percentage of creep in the hysteresis behavior of 92.5% Pb-5.0%
Sn-2.5% Ag solder cycled at 35 and 150°C. (From Solomon" 9 )
436 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

The matrix creep was determined from the upper branch (higher stress region)
of the creep curve (see Figure 14-19). This model has been shown to describe
some 60% Sn-40% Pb and 63% Sn-37% Pb fatigue data taken at about room
temperature at relatively high strain rates. It is not clear if this approach will
work at higher temperatures or at very low strain rates. Under these conditions,
the deformation is largely by creep, and the fraction of creep (and presumably
the amount of matrix creep) does not change with changes in the frequency or
hold time, but the fatigue life does. It is further not clear why only the matrix
creep correlates with the fatigue life. The value of ex of unity is not, however,
surprising. Coffin 1l3 •115 has shown this to be the case for superalloys tested at
elevated temperatures (although the applied plastic strain was used rather than
the matrix creep strain as was done by Fox and Shine).

14.5 INFLUENCE OF THE ENVIRONMENT

Time dependent plastic flow (creep) is not the only time dependent phenomenon
that can be the cause of the observed time dependence of the fatigue life. The
environment can also produce this effect. Snowden 123 •124 has shown that oxygen
contents as little as 3 X 10-2 mm Hg will reduce the fatigue life on polycrystalline
Pb by about two orders of magnitude, compared to that measured in tests run
in a vacuum. This effect, which produced intergranular fractures, saturated at 3
X 10--2 mm Hg, with tests run with higher oxygen contents having much the
same fatigue life. The tests described in Figures 14-13 through 14-17 were run
in a nitrogen filled box, but penetrations for the loading pull rods and grips
allowed oxygen to enter, making it probable that the oxygen content was well
in excess of 3 X 10-2 mm Hg.
Neugebauer et al. 76.78 have shown that destroying the hermeticity of some
electronic devices greatly reduces their thermal fatigue life compared to that
measured in identical tests but on devices whose hermeticity was not destroyed.
Superalloys have been subjected to numerous tests of the influence of the
environment on the high temperature fatigue life. The influence of the environ-
ment in lowering the fatigue life at low cycling frequencies is well docu-
mented. 115,116,125 Oxygen and water vapor 126,127 produce a strong frequency and
hold time dependence of the fatigue life.
It is quite possible that oxygen and/or water vapor are responsible for the
frequency and hold time dependence of the fatigue life of solder. This would
explain why, even though there is a strong frequency and hold time dependence
of the fatigue life of 60% Sn-40% Pb solder (see data of Figures 14-13 through
14-17), it does not correlate well with the creep behavior. Vacuum fatigue data
will be required, however, to demonstrate that the environment is indeed the
cause of the observed frequency and hold time effects. This is an important point
 MICROSTRUCTURAL CHANGES 437

since if the environment is the cause, using hermetic joints will greatly improve
the fatigue life.

14.6 MICROSTRUCTURAL CHANGES

In assuming that isothermal results can be used to predict thermal fatigue, one
is disregarding possible differences in the microstructure that are developed
during thermal fatigue to a much greater extent than during isothermal fatigue,
even isothermal fatigue at the peak temperature of the thermal range. Micros-
tructural instability is greater for thermally cycled solder because of the local
thermal stresses developed between the Pb- and Sn-rich material. These stresses
are developed because of the difference in the thermal expansion between these
two phases. Regions of coarsening of the Sn- and Pb-rich phases have been
observed. 34 in thermally fatigued specimens of 60% Sn-4O% Pb solder. Fatigue
cracks developed in the region where this coarsening developed, and it is believed34
that this coarsening plays an important role in the fatigue process. Coarsening
is less prevalent in the thermal fatigue of high Pb content solders. 128 Coarsening
also seems to be less prevalent when isothermal cycling is used and was not
readily apparent in the initially coarse grained solder 60% Sn-40% Pb solder
employed to generate much of the data used to illustrate this article. Thus,
coarsening, while an important factor in influencing the fatigue process, is not
the cause per se.
The development of a coarsened zone is important because this zone is softer
than the noncoarsened zone and will thus become more highly strained. This
leads to the development of cracks in this zone. Recrystallization and grain
growth can eliminate some fatigue damage. Thus, interactions between metal-
lurgical changes and the fatigue process may be important. In general, the number
of thermal cycles required to develop a coarsened zone may be dependent on
any or all of the following factors: the magnitude of the temperature change,
the mean temperature, the rate of temperature change or magnitude of any hold
times that are employed, the gross thermal strain being applied, the local thermal
strain associated with the expansion difference between the Sn-rich and Pb-rich
phases, the magnitude of any prior strain that has not been removed by recrys-
tallization, and, lastly but very importantly, the size and distribution of the Sn-
and Pb-rich phases at the start of thermal cycling.
The structure of eutectic solder is unstable even at room temperature,91 with
grain coarsening and precipitation occurring even as a specimen sits at that
temperature. The question therefore arises as to which structure to correlate with
fatigue tests, a structure observed when the joint is made, after the test, after
waiting for a given period of time after producing the joint, etc.? There is even
a question as to what is the structure of an actual joint of interest, the initial
438 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

structure, one developed after a bum in, after standing prior to operation, after
operation, etc.? Correlations of the change in structure with ageing and the
influence of any changes on the fatigue behavior have not been published.
Additionally, the influence of any change will be a function of the type of test
that is performed. Softening due to grain coarsening will reduce the life of a
load controlled test or a test using the total strain as the limiting variable (because
in both cases the plastic strain will increase). Softening will, however, have
much less of an impact on plastic strain limited tests because the plastic strain
is being kept constant. This is the type of test whose results are shown in the
figures of this chapter.
One approach (as was done here) to applying isothermal data to the problem
of thermal fatigue is to use a coarse microstructure, which will be more stable
than a fine one. There is also some evidenceI29.130 that a coarser structure will
lead to a shorter fatigue life. Thus, if the isothermal data is obtained on solder
with a structure as coarse as that which may be developed during thermal fatigue,
or after some time at temperature, some of the microstructural change during
thermal fatigue or after time at temperature, may be accounted for. If, however,
the coarsening during thermal fatigue is localized, the resulting strain localization
will not be accounted for. Clearly, more work in this area is required.

14.7 DETERMINATION OF THE DISPLACEMENT AND STRAIN

DISTRIBUTION IN A SOLDER JOINT

In order to apply the isothermal fatigue data to the problem of thermal fatigue,
it is necessary to first calculate the thermal displacement and from this displace-
ment to then calculate the strain distribution in the solder joint. We have con-
sidered the idealized problem of simple solder posts. Now we shall consider
more complex cases.
There are two parts to the problem of calculating the displacement that a solder
joint experiences. First one must perforln a global analysis of the complete circuit
board, complete with components, stiffeners, heat sinks, etc. In this analysis,
the components need not be considered in detail; only their influence on the
overall compliance is considered. When the whole board is considered, factors
such as board bending and resonance can be calculated. This global analysis
will show that there are components on the circuit board whose joints experience
more displacements than do the joints of other components. These are the joints
that are most likely to fail and therefore must be examined in more detail. The
displacements determined in this analysis serve as the boundary conditions for
the calculation of the displacement that a particular joint experiences. To do this
the exact details of a specific single component must be considered.
The global problem does not consider the solder joint or component in detail
and is beyond the scope of what is to be discussed here. These problems are
 DETERMINATION OF THE DISPLACEMENT AND STRAIN DISTRIBUTION 439

also extremely specific to the exact board and component population and are
therefore not amenable to a general discussion. They are more properly the
subject of a discussion of finite element techniques.
We have already considered the simple case of solder posts attaching a chip
carrier to a PWB. Now we shall consider the more complicated case of a fillet
joint subjected to a thermal excursion. The most simplified approach would be
to use Equation (14-2) to determine the displacement that the joint experiences.
This approach does not, however, consider the global strains discussed above
or the possibility that the PWB will bend, relieving some of the strain that the
joint must absorb.
Hall and co-workers40·57.58 and others 51 have determined by both experimental
and analytical means the magnitude of this bending. The critical variables de-
termining the magnitude of this bending are the elastic properties and geometry
of the chip carrier, board, and solder joint. In addition, the temperature and the
rate at which the solder is strained are of critical importance in determining the
amount of bending. The mechanical properties of solder are determined by the
temperature and by the rate at which the solder is strained. The yield strength
is increased at low temperatures and high strain rates, and the increased stiffness,
resulting from a lack of yielding, will produce more bending. Likewise, at high
temperatures or if very low strain rates are imposed, the yield strength will be
so low that the solder will impose no forces on the board or chip carrier, and
no bending will be developed. Bending has the effect of reducing the displace-
ment that must be accommodated by the solder joint because some of the dis-
placement is taken up by board and chip carrier bending. The nature of the
displacement is also a function of the rate of solder straining. At high temperatures
or at low enough rates to suppress bending, only shear displacements are de-
veloped. At low temperatures or at higher rates, there are increasingly larger
amounts of rotational and tensile displacements, with increasing amounts of
bending. There is thus a critical interplay between solder properties and the
nature and magnitude of the displacements that a solder joint will experience.
The temperature dependence of the flow stress also gives rise to a highly
asymmetric thermal fatigue hysteresis loop. 57 For thermal cycling between two
temperatures, lower stresses are required at the higher temperature than at the
lower temperature. This asymmetry was successfully modeled by the constitutive
law developed by Stone et aI., 46 enabling one to calculate the flow stress and
strains94 that are developed in simple joints that are thermally cycled. The cal-
culated thermal strains are then used in conjunction with the techniques described
above to determine the fatigue life. More complex calculations are required when
the joints are more complex and have a nonuniform strain distribution.
The strain distribution can be calculated from the displacement. For the case
of a simple solder pillar, the strain is simply calculated by dividing the displace-
ment by the height of the solder joint, as was done in Equation (14-4). Thermally
cycling a simple joint requires a constitutive relationship of the sort mentioned
 Von M.... Strain (%) t
a

1- .CJ018-.7I .,
m
2- .79-1.58 o'"
3-1.58 -2.38 9z
4-2.38 -3.17 Cl
5-a17 -3.88 ~
8-3.88 -4.75
7-4.75 -5.54 ~
r
8-5.5 4 8.34 >
Z
9-8.34 -7.13 o
~
10-7.1 3-7.95 m
n
Min = .0018% ~
z
Max = 7.95%
~
r

~
fi
c
m
r
<:
m
4,- =.25 MILS VI

X Y
~
-Copp er Pad

Figure 14-25 Calculated von Mises strain for a lead less CC-PWB joint for a castillated chip
displace ment of 0.25 mils. (From Solomon 105)
carrier subjecte d to a
 Von Mlses Strain (%)
1- 0-1.0
2-1.0-1.5
3-1.5-2.0
4-2.0-aO o
~
m
5-3.0-4.0 ;;0
!:
8-4.0-5.0 Z
7-5.0-8.0 ~
o
z
8-8.0-7.0
9-7.0-8.0 ~
-1
J:
Copper Pad m
Min = .CJ089% o
Vi
Mu = 7.97% ~
~
m
Z
-1
:>
z
o
~
~
~ z
o
~
;;0
~ c;
C
-1
o
Z

Figure 14·25 (continued) ~

442 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

above. More complex joints require a finite element analysis (FEA) to translate
the displacement that a joint experiences into a strain field distribution. This has
been done for both leadless60 ,131 and leaded41 chip carriers. For the joints con-
taining fillets that attach a leadless chip carrier to a PWB, the highest strains are
developed under the chip carrier and at the edge of the chip carrier. This is
illustrated in Figure 14-25. This figure shows the Von Mises strain developed
in joints that experienced a shear displacement of 2.5 x 10--4 in. The calculations
were made for two different joint configurations-a bulbous joint and a thin
joint-that represent extremes in the shape of the fillet. As can be seen, the
shape of the fillet has an influence on the magnitude of the strain that is developed
from a given displacement. It was also found that increasing the thickness of
the solder between the chip carrier and the PWB reduced the strains, in accor-
dance with Equation (14-4). These calculations were made to enable the cor-
relation (discussed in the next section) of isothermal, mechanical joint cycling,
with predictions made from generic isothermal data. For thermal fatigue cy-
cling the calculations would have to account for the temperature dependence of
yielding.
The Von Mises stress distribution (and thus the strain distribution because a
linear analysis was used) found by Clatterbaugh and Charles60 is similar to that
shown in Figure 14-25 in that the highest stresses (and strains) were located
under the chip carrier, at the edge of the chip carrier. Lau et al. 41 also used a
linear analysis to calculate the Von Mises stress but performed the analysis for
a l-Ieaded chip carrier to PWB joint. They found the maximum stress (and thus
strain) to be located at the tip of the outer solder fillet radius. This was the point
where failures were observed to start in laboratory tests of this type of joint.
The use of the Von Mises strains in these calculations emphasizes an important
point. The preceding LCF law was formulated in terms of the shear strain. The
more general approach is to describe the local stress state in terms of a correlating
parameter and to use this parameter to correlate fatigue test data. For instance,
the strains can be described in terms of an equivalent or Von Mises strain, that
is, in terms of the principal strains Elo E2, and E3 according to

(14-17)

This is what was done in Figure 14-25. It is then necessary to convert the strain
used in the strain-life data to Von Mises strains. The equivalent Von Mises strain
for a test run with a shear strain range A:y is ltv3(.!l'Y) (see the reference I05 ).
For a test run with a tensile strain range AE the equivalent Von Mises strain is
equal to this tensile strain range. Correlations 134 have also been made for the
octahedral shearing strain, which is equal to Y2 times the Von Mises strain.
Other similar approaches are possible; these examples are merely to illustrate a
technique, not necessarily to describe the best one.
 PREDICTION OF THE FATIGUE LIFE OF SOlDER JOINTS 443

14.8 PREDICTION OF THE FATIGUE LIFE OF SOLDER JOINTS

Isothermal displacement controlled LCF tests were performed on leadless CC-

PWB joints, 102 and the results were compared to predictions based on isothermal
fatigue data gathered on laboratory solder joints. !Os Figure 14-25 shows the
calculated strain distribution that was employed. The calculation was done for
extreme variations in the joint geometry, and therefore an average of these
extremes was used to describe an average joint. The FEA calculation was done
in terms of the total Von Mises strain [Equation (14-5)]. It was therefore necessary
to cast the LCF data into this form. The elastic strain was added to the plastic
strain of the Coffin-Manson data to develop the total strain-life curves that are
shown in Figure 14-26. The plastic strain curves are those of Figure 14-7.
Alternatively the plastic strain in the joint could be calculated and Figure 14-7
used directly. Since the calculation is given in terms of aVon Mises strain, it
was necessary to convert it into an equivalent shear strain. This was done via
Equation (14-7), which shows that a pure shear strain is equivalent to v'3 times
the Von Mises strain. !Os
Figure 14-25 shows that there is a distribution of shear strains in the joint.
To simplify the problem, it was assumed that only three strain regions existed:
a high strain region under the chip carrier (averaging 0.182 total shear strain),
a transition region where the crack entered the fillet (with a 0.095 total shear
strain), and the fillet region, which had the lowest strain (only 0.028 total shear

1.0

~ 0.1

0.01

l:lyP

0.001~~~~~~~~~~~~~~~~~~~
0.25 1 10 100 1000 10,000 100,000
Nf (CYCLES TO HALF LOAD)
Figure 14-26 LCF curves for the plastic and total strain for failure defined by a 25, 50, and
90 percent drop in the hysteresis load. (From Solomon 105)
444 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

CCI PWB Joint Failure Process

N cycles under 1, strain for L, distance

N: cycles under 12 strain for 4 distance
N, cycles under 1, strain for L, distance
N -25% load drop
N, + N2 - -50% load drop
N, + N2 + N, -100% load drop
N, (MEASURED ON CC-PWB JOINTS)

Figure 14-27 Schematic representation of cracking in CC-PWB joints. (From Solomon 105)

strain). This is illustrated schematically in Figure 14-27. Now Figure 14-26 can
be used to calculate a fatigue life for each of these strains, providing that the
size effect is corrected by Equation (14-11). The length of region 1 was measured
at 0.040 in (1.01 mm), that of region 2 was estimated as 0.01 in (0.25 mm),
and that of region 3, at 0.015 in (0.38 mm). Table 14-1 lists the calculated
cycles to failure for each region using the total strain corrected 25, 50, and 90%
load drop curves of Figure 14-7. The D values that were used are also given in

Table 14-1 Calculation of N; for N, Defined by Different Drops in Load and

Values for D
N 1 [L 1 = 0.04 N 2 [L 2 = 0.01 N 3 [L 3 = 0.015
in V (1.01 in V (0.25 in V (0.38
mm)] mm)] mm)]
For N, (25%)
D = 0.021 in (0.53 mm) 38 26 250
For N,(50%)
D = 0.083 in (2.10 mm) 25 22 343
For N,(90%)
D = 0.30 in (7.62 mm) 15 15 300
 PREDICTION OF THE FATIGUE LIFE OF SOlDER JOINTS 445

the table. The values Nb N 2, and N3 given in the table are the number of cycles
to failure for each region. As can be seen, Equation (14-7) and the appropriate
value for D allows us to use any of the load drop curves to calculate the fatigue
life. In applying the results of Table 14-1, we are guided by the failure process
observed in the joints being considered. It was found that cracks were first
observed under the chip carrier and then grew into the fillet to failure. This first
stage should be determined by N\. The number of cycles to grow the crack into
the fillet is NJ + N 2 • We are assuming that the cracking must grow in from a
free surface, where there can be an environmental interaction so a crack is not
considered to have nucleated in the second region; rather it grows in from under
the chip carrier. Hence, N J and N2 are summed to get the life to grow the crack
into the fillet (i.e., N\ and N2 are assumed to be sequential, not concurrent). For
final failure (defined by a 90 percent drop in the hysteresis load) we use N\ +
N2 + N 3. We are neglecting concurrent cracking of the fillet but since it is a
free surface, it should be occurring. Since N3 is so much larger than N\ + N 2,
there is little difference between defining failure by just N3 or by the sum of N\
+ N2 + N 3. Figure 14-28 compares these calculated values and that measured
on actual joints subjected to the displacement used in the FEA calculation. The
three boxes represent the calculations for the three regions. The bands represent
the different results obtained with different load drop laboratory curves and

,-...
lJ... 3SoC
U O.3H 3
~ 200
«
w
lJ...
100
::E /
0
a:: 50 / /
lJ...
/
0
~
«
...J
20
:J
U
...J 10
«

-
u
'-' /
5 / /
Z / / 1.5X/
2//
/
/
/

11 2 5 10 20 50 100 200 500 1000

N, (MEASURED ON CC-PWB JOINTS)
Figure 14-28 Comparison between the predictions made from FEA and LCF data and
measurements on CC-PWB joints. (From Solomon 105 )
446 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

different joint data curves. As can be seen, the prediction and results generally
fall within a factor of 2. Some refinements could improve the agreement. For
instance, if the final failure were defined by N3 instead of by NI + N2 + N 3,
there would be some improvement between the calculation and the prediction
for this last box.
The preceding analysis takes account of the geometry of the solder joint but
does not account for such factors as the influence of the cycling frequency or
hold time. The tests on both the GC-PWB and laboratory joints were run without
hold times and at a relatively high cycling frequency. As such, the results cannot
be applied to situations where long hold times or low cycling frequencies are
employed. Corrections of the sort described in the previous sections of this
chapter must be employed before a more general life prediction can be made.
Clech, Engelmaier, and co-workers I3 3-136 have developed a "figures of merit"
approach, which systematizes the influence of many of the factors, discussed in
this chapter, that alter the fatigue life. They provide expressions for the influence
of joint height, distance between joints, differences in the coefficient of expan-
sion, temperature cycling range, dwell time, and lead compliance (for leaded
joints). These expressions give relative fatigue lives based on tests run on FR-
4-type joints, which do not have fillets of the sort shown in Figure 14-25. For
this type of joint, the techniques they outline allow one to estimate the effect of
changing the solder joint height, varying the hold time, varying the cycling
temperature range, etc. However, since the data used to determine the correlations
of their model were generated on joints that do not contain large fillets, corrections
must be made when applying it to joints where much of the life is spent growing
a fatigue crack through a fillet.

14.9 INHERENT LIMITATIONS TO FATIGUE LIFE PREDICTIONS

It is clear from the previous discussion that the calculation of the fatigue life is
a complex process, requiring a knowledge of the behavior of the total CC-PWB
population, the geometry and mechanical properties of individual joints and the
components that they join, the fatigue properties of solder (including the influ-
ences of stress state, hold time, cycling frequency, temperature, and even the
definition used to define failure), fatigue crack propagation, solder creep, and
the influence of microstructural changes. Even if one knew the influence of all
of these variables, it would still not be possible to exactly determine the number
of cycles to a fatigue failure. This is because the fatigue life must be expressed
in terms of the statistical variance associated with each of the aforementioned
processes. This is especially true for solder joints that are in effect small un-
machined castings, for which the variability in the fatigue properties is greater
than that for wrought and carefully machined specimens. The general approach
is to determine the standard deviation of the fatigue life curve, taking into account
 NECESSARY FURTHER WORK 447

all the factors that influence the fatigue life, and then design to only a fraction
of this life, say with fatigue curves drawn at the mean life minus one or more
standard deviations. The Clech, Engelmaier figures of merit approach incorpo-
rates failure rates as an integral part of their methodology.
The time or number of cycles to the first joint failure can be estimated by
identifying the joint subjected to the highest displacement. Then it can be assumed
that the joint has the least favorable geometry (i.e., one for which the strain
distribution is a maximum). The appropriate fatigue life curve or crack growth
curve, corrected for the stress state, must then be considered statistically. A
fatigue life at minus one or more standard deviations, depending upon the degree
of conservatism that is desired, should be used (or if the analysis is based upon
crack growth data, an upper limit curve should be used). More complex statistical
approaches to estimate the number of cycles to the first failure can also be
employed. 139 Finally, accelerating factors l40 ,141 such as the cycling frequency,
hold time, mean temperature, grain size, and grain growth should be considered
and used to further reduce the fatigue life.
What is meant by failure and the consequence of a joint failure must also be
considered. A fatigue crack in a solder joint does not produce a significant overall
increase in resistance until the crack is almost completely through the joint. 142.102.104
Even then, if the broken halves contact each other, the increase in resistance
may be not be detrimental to the overall circuit. The problem is that vibrations
or other forces may separate the broken halves and then produce an open. The
point at which the open is actually observed may thus be a function of more
than just the fatigue behavior. This further adds to the variability in observed
device failure times. It is also possible that the failed joint may not always be
called upon to carry current, and thus the observation of a device failure would
be further postponed until a failure to carry current was noted.

14.10 NECESSARY FURTHER WORK

Comparisons have been made between the fatigue life measured on actual joints
subjected to isothermal cycles and the life calculated from generic isothermal
fatigue data. This showed how to account for the joint shape and the size effect
that was created because the generic data was developed on joints that were
larger than the actual solder joints of interest. The influence of other variables
was also discussed but not specifically incorporated in the life prediction analysis.
Similar predictions, based upon generic data, should therefore be made for
thermally cycled joints, with cycling between different temperatures, with dif-
ferent cycling rates, hold times, etc., and comparisons made with the behavior
of actual thermally cycled joints.
A considerable amount of thermal cycle data already exists, but generally
there is insufficient data on all of the thermal displacements that were developed
448 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

(i.e., all the translations and rotations about all three axis) to enable an accurate
determination of all the strains that were developed in the joint. The calculation
of these strains requires not only better input displacement data than currently
exists but also requires a temperature and strain rate dependent finite element
analysis. With a knowledge of the joint strain state, developed by a given thermal
cycle, it should be possible to determine the fatigue life based upon general
fatigue data of the sort discussed here and compare these predicted results to
those measured on the thermally cycled joints.
The aforementioned approaches have been discussed in the context of leadless
chip carrier joints. They must also be extended to leaded joints where the joint
compliance is much larger and the stresses developed much lower.
The figures of merit technique developed by Clech, Engelmaier, and co-
workers (which is discussed at length in this volume), provides a systematized
approach to many of the factors discussed here, for both leaded and leadless
joints. It is a relative approach, however, that is best applied when doing a
sensitivity analysis of the influence of various variables on the behavior of FR-
4 joints. More work is needed to extend it to other types of joints, particularly
where large fillets playa dominate role in determining the fatigue life.
The phenomena discussed at length in this chapter are largely due to crack
propagation, but instead of measuring the crack growth rate directly, it was
inferred from LCF results. While this is a reasonable approach because LCF is
dominated by fatigue crack growth, approximations were required to account
for multiple cracking and the irregular nature of the crack front that is developed.
With specific crack nucleation and crack propagation measurements one could
use the more refined crack growth design approaches that have been developed
for the design of aircraft engines and air frames. These allow for the calculation
of the remaining life in a joint containing a fatigue crack as well as for the
prediction of the overall life.
The design and life prediction of solder joints will also have to account for
cracking of metallized layers, intermetallic layers, and the strain concentration
and subsequent cracking that may be developed in grain coarsened zones. More
work is needed to determine the extent to which these factors are life limiting.
Tests run on leadless CC-PWB joints 105 have shown that metallization cracking
was not a major factor in determining the overall life because of the much larger
number of cycles spent propagating a crack through the fillet. With joints without
fillets it is expected that metallization cracking would play a larger role in
determining the joint life.
Of all the necessary future work nothing is more important than the system-
ization of a design approach that will incorporate all the factors discussed above
in a manner that can be used by the average design engineer. Some progress
has been made in this area, but all too often successful approaches have been
used beyond their "zones of applicability," say by neglecting the influence of a
long hold time or by neglecting the fact that the joint of interest has a radically
 REFERENCES 449

different geometry from the one used to develop the design data. Much of the
data exists to develop this systemization; what is needed is a more concerted
effort to organize it, do blind predictions of actual thermal cycling tests to
compare the results of various approaches, and a standardization body to codify
the approach that does the best job. Any resulting code must also define, and
clearly state, its zone of applicability. Such future work will not only result in
more accurate life predictions, and therefore more reliable solder joints, but will
also clearly point to the directions that should be taken to improve the joint life,
either though improved solder properties or improved joint designs.

14.11 ACKNOWLEDGMENTS

The author would like to acknowledge the USAF materials laboratory for their
support (through contract no. F33615-85-C-5065) of the work associated with
the references. \02,104,\05 The author would also like to acknowledge the use of
the section on "Inherent Limitations to Fatigue Life Prediction" and parts of the
sections on "Microstructural Changes" and the "Determination of the Displace-
ment and Strain Distribution in a Solder Joint" which appeared in the author's
review paperl2 that was published by The Materials Research Society as part of
their 1989 spring meeting.

REFERENCES

1. Manson, S. S., "Thermal Stress and Low Cycle Fatigue," New York, McGraw
Hill, 1965.
2. AS ME Code case N-47, loc. cit. G. R. Halford et al. ASTM-STP 942, 1988, pp.
625-636.
3. Basquin, O. H., Proc. ASTM 10, 1910, pp. 625-630.
4. Coffin, L. F., Jr., Trans. ASME, 76, 1954, pp. 931-950.5.
5. Manson, S. S., "Behavior of Materials Under Conditions of Thermal Stress," Heat
Transfer Symposium, Univ. of Michigan, June 27-28, 1952, Univ. of Mich. Press;
also NACA TN2933, July 1953.
6. Manson, S. S., Exp. Mech., 5(7), 1965, pp. 193-226.
7. Coffin, L. F., Jr., in "Fracture," London: Chapman and Hall, 1969.
8. Morrow, J., "Cyclic Plastic Strain Energy and Fatigue of Metals," in "Internal
Friction, Damping and Cyclic Plasticity," ASTM-STP 378, 1965, pp. 45-87.
9. Taira, S., "Relationship Between Thermal Fatigue and Low-Cycle Fatigue at El-
evated Temperature," in ASTM-STP 520, 1973, pp. 80-101.
10. Spera, D. A., and D. F. Mowbray, eds. "Thermal Fatigue of Materials and Com-
ponents," ASTM-STP 612, 1976.
11. Lau, J. H., and D. W. Rice, Solid State Technology, 28(10), 1985, pp. 91-104.
12. Solomon, H. D., "Review of Critical Variables Determining Solder Fatigue Lives,"
Mat. Res. Soc. Symp. Proc., 154, 1989, pp. 441-451.
450 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

13. Wild, R. N., Welding J., 51, 1972, pp. 521s-526s.

14. _ _ ,Internepcon, October 1975 (IBMl74Z000481).
15. Jarboe, D. M., "Thennal Fatigue of Solder Alloys," NTIS-BDX-613-2341, Feb-
ruary 1980.
16. Shamash, M. B., "Development of Highly Reliable Soldered Joints for Printed
Wiring Boards," NTIS N69-25697, Baltimore, MD, Westinghouse Defense ana
Space Center, 1968.
17. Rathore, H. S., R. C. Yih, and A. R. Edenfeld, J. of Testing and Evaluation, 1(2),
1973, pp. 170-178.
18. Denlinger, M. C., R. W. Korb, and V. F. Lardenoit, "Development of Improved
Solders for Electronic Reliability," AFML-TR 77-93, June 1977.
19. Inoue, H., Y. Kurihara, and H. Hachino, IEEE Trans. CHMT-9, no. 2, 1986, pp.
190-194.
20. Vaynman, S., M. E. Fine, and D. A. Jeannotte, "Isothermal Fatigue Failure Mech-
anisms in Low Tin Lead Based Solder," in TMS-AIME Symposium on Effect of
Load and Thermal Histories on Mechanical Behavior of Materials, Denver, Feb-
ruary 1987.
21. _ _ ,37th Electronic Components Conference, May 1987, pp. 598-603.
22. _ _ , "Effect of Hold Time on Fatigue of 96.5Pb-3.5Sn Solder," in Fatigue 87.
23. Lawson, L. R., M. E. Fine, and D. A. Jeannotte, "Thennomechanical Fatigue of
Lead-Base Low Tin Solder," in Fatigue 87.
24. Solomon, H. D., IEEE Trans. CHMT-9, no. 4, 1986, pp. 423-432.
25. _ _ , in "Electronic Packaging-Materials and Processes," J. A. Sartell, Ed.,
ASM, 1986, pp. 29-47.
26. _ _ , Brazing and Soldering, no. 11, Autumn 1986, pp. 68-75.
27. _ _ , in "Low Cycle Fatigue," ASTM-STP 942, H. D. Solomon, G. R. Halford,
L. R. Kaisand, and B. N. Leis, Eds., ASTM, January 1988, pp. 342-371.
28. _ _ ,38th Electronic Components Conference, May 1988, pp. 7-13.
29. _ _ , ASME J. of Electronic Packaging, 111, 1989, pp. 75-82.
30. Engberg, G., L. E. Larsson, M. Nylen, and H. Steen, Brazing and Soldering, no.
11, Autumn 1986, pp. 62-65.
31. Shine, M. C., L. R. Fox, and L. W. Sofia, 4th International Electronics Packaging
Conference, October 1984; also in Brazing and Soldering, no. 9, Autumn 1985,
pp. 11-14.
32. Shine, M. C., and L. R. Fox, in "Low Cycle Fatigue," ASTM-STP 942, H. D.
Solomon, G. R. Halford, L. R. Kaisand, and B. N. Leis, Eds., ASTM, January
1988, pp. 588-611.
33. Frost, H. J., P. R. Lavery, S. D. Lutender, and R. T. Howard, 38th Electronic
Components Conference, May 1988, pp. 13-22.
34. Frear, D., D. Grivas, M. McConnack, D. Tribula, andJ. W. Morris, Jr., "Fatigue
and Thennal Fatigue of Pb-Sn Joints," in AlME Spring Conference Proc., 1987.
35. Frear, D., D. Grivas, and J. W. Morris, Jr., "The Effect of ClI(;Sns Whisker
Precipitates in Bulk 60Sn-40Pb Solder," to be published.
36. Frear, D., D. Grivas, L. Quan, and J. W. Morris, Jr., Mat. Res. Soc. Proc. 72,
1986, pp. 181-186.
37. Weinbel, R. C., J. K. Tien, R. A. Pollak, and S. K. Kang, J. of Materials Science
22, 1987, pp. 3901-3906.
 REFERENCES 451

38. Weinbel, R. C., E. A. Schwartzkopf, and J. K. Tien, Scripta Metallurgicia 21,

1987, pp. 1165-1168.
39. Hendrix, B. C., J. K. Tien, S. K. Kang, and T. Reiley, "Stress and Strain Controlled
Low Cycle Fatigue of Pb-Sn Solder for Electronic Packaging Applications," to be
published.
40. Hall, P. M., 37th Electronic Components Conference, May 1987, pp. 579-588.
41. Lau, J., G. Harkins, D. Rice, J. Kral, and B. Wells, 37th Electronic Components
Conference, May 1987, pp. 589-597.
42. _ _ ,IEEEIIRPS, 1987, pp. 250-259.
43. Lau, J., D. Rice, and P. A. Avery, IEEE CH2295, 1986, pp. 173-184.
44. Hagge, J. K., Proc IEPS, 1986, pp. 199-208.
45. Stone, D., H. Wilson, R. Subrahmanyan, andC.-Y. Li,36thElectronicComponents
Conference, May 1986, pp. 630-635.
46. Stone, D., S.-P. Hannula, and C.-Y. Li, 35th Electronic Components Conference,
May 1985, pp. 46-51.
47. Engelmaier, W., IEPS, October 1984, pp. 360-369.
48. _ _ , IEEE Trans. CHMT-6, no. 3, 1983, pp. 232-237.
49. _ _ , Circuit World, 11(3), 1985, pp. 61-72.
50. Hieber, H.IEEEIIRPS, 1988, pp. 234-238.
51. Desaulnier, W. E., Jr., T. E. Laflamme, W. B. Ammerman, and M. C. Binder,
Proc. Institute of Environmental Sciences, 1986, pp. 6-19.
52. Marshall, J. L., and S. R. Walter, Int. J.for Hybrid Microelectronics, 10, 1987,
pp. II-l7.
53. Tanaka, T., S. Nishijima, and M. Ichikawa, Current Japanese Materials Research,
2, 1987, pp. 235-250.
54. Schmitt-Thomas, Kh. G., and S. Wege, Brazing and Soldering, no. II, Autumn
1986, pp. 27-33.
55. Hall, P. M., and W. M. Sherry, Interconnection Technology in Electronics, Proc.
3d International Conference, February 1986, pp. 47-61.
56. Sherry, W. M., J. S. Erich, K. Bartschat, and F. B. Orinz, IEEE Trans. CHMT-
8, 1985, pp. 417-426.
57. Hall, P. M., IEEE Trans CHMT-7, 1984, pp. 314-327.
58. Hall, P. M., T. D. Dudderar, and J. F. Argyle, IEEE Trans CHMT-6, 1983, pp.
544-552.
59. O'Clock, G. D., M. S. Peters, G. A. Kleese, and R. V. Martini, IEF;E Trans
CHMT-JO, 1987, pp. 82-88.
60. Clatterbaugh, G. V., and H. K. Charles, IEEE 0569-5503, 1985, pp. 60-72.
61. Bangs, E. R., and R. E. Beal, Welding J., 54, 1975, pp. 377s-383s.
62. Dunn, B. D., "The Resistance of Space-Quality Solder Joints to Thermal Fatigue,"
ESA STM-207, September 1978.
63. Coombs, V. D., in ASTM-STP 515, 1972, pp. 3-21.
64. Olsen, D. R., and H. M. Berg, IEEE Trans CHMT-2, no. 2, 1979, pp. 257-263.
65. Korb, R. W., and D. O. Ross, IEEE Trans CHMT-6, no. 3, 1983, pp. 227-797.
66. Pankratz, J.M ., and H. R. Plumlee, IEEE Trans ED-J7, no. 9, 1970, pp. 793-
797.
67. Kong, S. K., N. D. Zommer, D. L. Feutcht, and R. W. Heckel, IEEE Trans.
PHP-13, no. 3, 1977, pp. 318-321.
452 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

68. Zomrner, N. D., D. L. Feutch, and R. W. Heckel, IEEE Trans. ED-23, no. 8,
1976, pp. 843-850.
69. Levine, E., and J. Ordonez, IEEE Trans. PHP-J3, no. 3, 1977, pp. 318-321.
70. Norris, K. c., and A. H. Landzberg, IBM J. Res. Dev., May 1969, pp. 266-271.
71. Goldmann, L. S., IBM J. Res. Dev., 13, 1969, p. 251.
72. Goldmann, L. S., R. D. Herdzik, N. G. Koopman, and V. C. Marcotte, IEEE
Trans PHP-J3, no. 3, 1977, pp. 194-198.
73. Agarwala, B. N., IEEEIIRPS, 1985, pp. 198-205.
74. Taylor, J. R., and D. J. Pedder, Int. J. Hybrid Microelectronics, 5(2), 1982, pp.
209-214.
75. Lang, G. A., B. J. Fehder, and W. D. Williams, IEEE Trans ED-17, no. 9, 1970,
pp. 787-793.
76. Burgess, J. F., R. O. Carson, H. H. Glascock, C. A. Neugebauer, and H. F.
Webster, IEEE Trans. CHMT-7, 1984, pp. 405-410.
77. Burgess, J. F., R. O. Carson, H. H. Glascock, C. A. Neugebauer, and H. F.
Webster, IEEE 34th Electronic Components Conference, May 1984, pp. 144-148.
78. Neugebauer, C. A., H. F. Webster, H. D. Solomon, and R. O. Carlson, "The
Measurement of Fatigue Suppression in Electronic Solder Joints," in the 8th Sol-
dering Technology Seminar, February 1984, Naval Weapons Center, China Lake,
CA.
79. Miyazaki, M., S. Yoshioka, and A. Aijikata, Zairyo (Materials) 30(331), 1981,
pp. 8-13.
80. Berriche, R., S. Vaynrnan, M. E. Fine, and D. A. Jeannotte, in 3d Conference on
Electronic Packaging: Materials and Processes & Corrosion in Microelectronics,
Minneapolis, MN, April 1987, pp. 169-174.
81. Homa, T. R., ibid., pp. 209-216.
82. O'Clock, G. D., M. S. Peters, G. A. Kleese, and R. V. Martini, ibid., pp. 251-
257.
83. Frost, H. J., P. R. Lavery, and S. D. Lutender, ibid., pp. 259-267.
84. Frear, D., D. Grivas, M. McCormack, D. Tirbula, 1. D. Morris, Jr., ibid., pp.
269-274.
85. Frost, H. J., R. T. Howard, P. R. Lavery, and S. D. Lutender, in 38th IEEE-ECC
Conf. Proc., May 1988, pp. 13-23.
86. Bae, K., A. F. Sprecher, Z. Guo, H. Conrad, and Y. Jung, in Micro. Electronic
Packaging Technology-Materials and Processes, ASM, 1989, pp. 109-121.
87. Frost, H. G., G. J. Stone, and R. T. Howard, ibid., pp. 121-128.
88. Hendrix, B. C., P. L. Bretz, J. K. Tien, and S. K. Kang, ibid., pp. 129-134.
89. Solomon, H. D., ibid., pp. 135-146.
90. Fine, M. E., and D. A. Jeannotte, ibid., pp. 147-158.
91. Chilton, A. C., M. A. Whitmore, and W. Hampshire, ibid., pp. 159-179.
92. Lau, J. H., and A. H. Jeans, ibid., pp. 177-188.
93. Oberin, F. W., ibid., pp. 189-202.
94. Wilcox, J. R., R. Subrahrnanyan, and C.-Y. Li, ibid., pp. 203-212.
95. Subrahmanyan, R., J. R. WIlcox, and C.-Y. Li, ibid., pp. 213-222.
96. Vaynman, S., and M. E. Fine, ibid., pp. 255-260.
 REFERENCES 453

97. Subrahmanyan, R., J. R. Wilcox, and c.-y. Li, in 39th IEEE-ECC Conf. Proc.,
May 1989, pp. 240-252; also IEEE Trans. CHMT-12, no. 4, 1989, pp. 480--492.
98. Schmidt, C. G., ibid., pp. 253-258.
99. Tien, J. K., B. C. Hendrix, P. L. Bretz, and A. I. Attarwala, ibid., pp. 259-263;
also IEEE Trans. CHMT-12, no. 4, 1989, pp. 502-506.
100. Enke, N. F., T. J. Kilinski, S. A. Schroeder, and J. R. Lesniak, ibid., pp. 264-
272; also IEEE Trans. CHMT-12, no. 4, 1989, pp. 459-469.
101. Vaynman, S., ibid., pp. 273-276; also IEEE Trans. CHMT-12 , no. 4, 1989, pp.
469-473.
102. Solomon, H. D., ibid., pp. 277-292; also IEEE Trans. CHMT-12 , no. 4, 1989,
pp. 473-480; also GE report #87CRDI85.
103. Frear, D. R., ibid., pp. 293-300, also IEEE Trans. CHMT-12, no. 4, 1989, pp.
492-502.
104. Solomon, H. D., "Influence of Temperature on the Low Cycle Fatigue of Surface
Mounted Chip Carrier/Printed Wiring Board Joints," presented at the 35th Annual
Technical Meeting, Institute of Environmental Sciences, Anaheim, CA, May 1989;
also GE report #87CRDI86.
105. Solomon, H. D., V. Brzozowski, and D. G. Thompson, "Prediction of Solder Joint
Fatigue Life," GE report #88CRDIOI; also 40th ECTC, May 1990, pp. 351-360.
106. Solomon, H. D., "High and Low Temperature Strain-Life Behavior of a Pb Rich
Solder," 89-WNEEP-33, presented at the 1989 Winter ASME Meeting, December
1989, San Francisco; ASME J. Electronic Packaging, 112, June 1990, pp. 123-
128.
107. Wilcox, J. R., R. Subrahmanyan, and C.-Y. Li, "Assembly Stiffness and Failure
Criterion Considerations in Solder Joint Fatigue," presented at the 1989 Winter
ASME Meeting, December 1989, San Francisco; to be published in ASME J.
Electronic Packaging, June 1990.
108. Gilmore, R. S., K. C. Tam, J. D. Young, and D. R. Howard, Phil. Trans. R.
Soc., London, A320, 1986, pp. 215-235.
109. Wilkening, W. W., private communication.
1l0. Rice, J. R., in Treatise on Fracture, H. Liebowitz, Ed., New York, Academic
Press, 1971, V2, Chap. 3, pp. 230--232.
Ill. Gohn, G. R., and W. C. Ellis, Proc ASTM, 51, 1951, pp. 721-744.
1l2. Eckel, J. F., Proc. ASTM, 51, 1951, pp. 745-760.
1l3. Coffin, L. F., Jr., Met. Trans., 2, 1971, pp. 3105-3113.
114. Solomon, H. D., Met. Trans., 4, 1973, pp. 341-347.
115. Coffin, L. F., Jr., ASTM-STP 520, 1973, pp. 5-34.
116. Solomon, H. D., and L. F. Coffin, Jr., ASTM-STP 520, 1973, pp. 112-122.
117. Manson, S. S., G. R. Halford, and H. M. Hirshberg, in Design for Elevated
Temperature Environment, ASME, 1971, pp. 25-28.
118. Manson, S. S., ASTM-STP 520, 1973, pp. 744-782.
1l9. Solomon, H. D., "The Creep and Strain Rate Sensitivity of a High Pb Content
Solder," presented at the AIME symposium on Deformation in Electronic Materials
and Devices, October 1989, Indianapolis, IN: proceedings to be published by AIME.
120. Knect, S., and L. R. Fox., to be published.
454 PREDICTING THERMAL AND MECHANICAL FATIGUE LIVES

121. Fox, L., this volume.

122. Monkman, F. C., and N. J. Grant, Proc. ASTM, 56, 1956, pp. 593-620.
123. Snowden, K. R., Acta Met., 12, 1964, pp. 295-303.
124. Snowden, K. V., Phil. Mag., 10, 1964, pp. 435-440.
125. Gabreilli, F., and R. M. Pelloux, Met. Trans., 13A, 1982, pp. 1083-1090.
126. Bradshaw, F. J., and C. Wheeler, Applied Materials Research,S, 1966, pp. 112-
120.
127. Bradshaw, F. J., and C. Wheeler, Inti. 1. Fracture Mechanics,S, 1969, pp. 255-
268.
128. Vaynman, S., private communication.
129. Wild, R. N., NEPCON, 1974, pp. 105-117.
130. Kinser, D. L., J. G. Vaugn, and S. M. Graff, Electronic Packaging and Production,
May 1976, pp. 61-68.
131. Brzozowski, V., et al., "FEA Analysis of Solder Joints," Man. Tech. for ADSP,
monthly report March 20, 1987, Martin Marrietta Aerospace.
132. Miller, K. J., and M. W. Brown, Eds., "Multiaxial Fatigue," ASTM-STP 853,
1985.
133. Clech, J.-P., this volume.
134. Clech, J.-P., W. Engelmaier, R. W. Kotlowitz, and J. A. Augis, "Surface Mount
Solder Attachment Reliability Figures of Merit-'Design for Reliability' Tools,"
presented at the 1989 SMART V Conference, New Orleans, LA.
135. _ _ , IEEE Trans. CHMT, 12, no. 4, 1989, pp. 449-459.
136. Clech, J.-P., M. J. Tervalon, and J. A. Augis, Proc.IEPS, San Diego, CA, 1989,
2, pp. 754-768.
137. Clech, J.-P., and J. A. Augis, Proc.IEPS, 1987, pp. 385-410.
138. _ _ , Proc. IEPS, 1988.
139. Lambert, R. G., "Mechanical Durability Prediction Methods," in Proc. 1989 Annual
Reliability and Maintainability Symposium, Atlanta, GA, 1989.
140. Howard, R. T., IEEE CHI781-4/82/()()()()/0376, 1982, pp. 376-384.
141. Kotlowitz, R. W., and W. Engelmaier, 6th Inti. Electronic Packaging Conf. Proc.,
1986, pp. 841-865.
142. Weller, D., L. Fox, and R. Hannemann, 33d ECC Proc. IEEE, 0569-5503/831000-
0534, 1983, pp. 534-545.
 15
Static and Dynamic Analyses of
Surface Mount Component Leads
and Solder Joints
John H. Lau

Surface mount technology (SMT) is currently one of the strongest trends in

electronic packaging and interconnect assembly. This technology offers oppor-
tunities for significant cost reduction and for efficient production line automation.
A surface mount assembly (SMA) is basically a composite structure consisting
of three major parts: the printed circuit board (PCB), the solder joints, and the
surface mount component (SMC). In most consumer products, computers, pe-
ripherals, and automotive electronics, the PCB is made of FR-4 epoxy-glass and
the solder joint is made of 60% Sn-40% Pb or 63% Sn-37% Pb solder alloys.
Unlike traditional through-hole technology, SMT uses smaller SMCs that are
soldered directly to the pad surface of a PCB. Consequently, the solder joint is
the only mechanical means of attaching the SMC to the PCB. Thus, the integrity
of solder joints is one of the most critical issues in SMT developments. 1-202 For
a list of solder-joint papers published before 1985, see reference 185.
Basically, there are two groups of SMCs, leadless SMCs (e.g., leadless ce-
ramic chip carriers, epoxy-glass chip carriers, chip resistors, chip capacitors,
etc.) and leaded SMCs (e.g., small outline integrated circuits (SOICs), plastic
leaded chip carrier (PLCCs), plastic quad fiat pack (PQFPs), etc.). It is generally
assumed that: (I) the limited compliance of the solder joints of leadless ceramic
chip carriers on FR-4 PCB leads to serious reliability problems, and (2) the
characteristically free-standing leads of SOICs, PLCCs, and PQFPs have suf-
ficient compliance to virtually preclude any attachment reliability hazard. Un-
fortunately, the second assumption is not always correct. It is true that the stresses
and strains acting at the solder joints are reduced substantially due to the com-
pliance of the leads. However, because of solder's low resistance to fatigue, it
is not uncommon to observe solder-joint cracking in SMT assemblies with leaded
SMCs.
455
456 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

A variety of environmental stress factors such as temperature and humidity,

shock and vibration, thermal and functional cycles, corrosion, etc., may lead to
solder-joint failure. The most commonly observed failure modes in practice are
overload and fatigue. Overload failure occurs whenever the stress or strain in
the solder joint, brought about by the imposed stress factors, is greater than the
strength or fracture toughness of the solder alloy. An example of overload failure
condition is extensive bending and twisting of a PCB with SMCs soldered to its
surface. On the other hand, fatigue failure of a solder joint takes place via the
initiation and slow propagation of a crack until it becomes unstable. 185 The stress
factors that typically cause failure by fatigue are far below the overload failure
levels. Examples of fatigue failure include the vibration of a PCB with SMCs
soldered to its surface and the functional power cycling of a SMC that is soldered
to the surface of a PCB. In this chapter, both the static (including thermal effects)
and dynamic (including shock and vibration) stress factors will be considered.
There are two parts in this chapter. In Part I, "Stiffness of Gull·Wing and
J Leads and Solder Joints for Surface Mounted Chip Carriers," two 12 x
12 stiffness matrices of the two most common leads and solder joints in SMA
are presented. One is for the gull-wing lead and solder joint in a PQFP SMA,
and the other is for the J lead and solder joint in a PLCC SMA. Deformation
of these composite structures for each imposed unit displacement (or rotation)
is also presented to provide a better understanding of their mechanical charac-
teristics. Finally, the whole-field stress distributions in the leads and solder joints
for the case of a unit displacement in the I-direction are given. This case is
similar to the global thermal expansion mismatch between the chip carriers and
the PCB.
In Part II, "Solder Joint Reliability Under Vibration Conditions," both
local and global vibrations of an SMA are considered. For local vibration of the
SMA, the frequencies of soldered and unsoldered leads of surface mount wide
and narrow SOICs, PLCCs, and PQFPs are presented. It is found that the
fundamental frequency of all the soldered leads is at least five times larger than
that of the unsoldered leads. For global vibration of the SMA, the frequency
equation of a constrained PCB with a SMC is given. With this equation, vibration
frequencies of a PCB with a SMC can be ready determined. The results presented
should be useful for theoretically determining the reliability of solder joints under
shock and vibration conditions and for developing a shock and vibration exper-
iment.

Stiffness of Gull-Wing and J Leads and Solder Joints

for Surface Mounted Chip Carriers
Based on the matrix structural analysis of solid mechanics,203-219 the stiffness
matrices of the gull-wing lead and solder joint in a PQFP SMA and the J lead
and solder joint in a PLCC SMA are calculated.
 ANAL YSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS 457

The PQFP and PLCC, Figures 15-1a and b, are package families that meet
the Joint Electronic Device Engineering Council (JEDEC) package registration
specification. The PQFP has gull-wing shaped leads that are spaced on a 20-mil
(O.02-in.) pitch. The PLCC has specially designed J-shaped leads that are spaced
on a 50-mil (O.05-in.) pitch. These leads are the electrical and mechanical
connections of the packages to the outside world that are soldered to the surface
of a PCB. Consequently, the stiffness of the leads plays a major role in trans-
mitting the stress to the solder joints. In general, the larger the lead stiffness
(less compliance), the more the joint stress. Therefore, in order to predict the
solder-joint integrity, it is necessary to characterize the gull-wing (or J) lead and
solder joint as a system.
In this chapter, the finite element method (see for examples, references 203--219)
was used to calculate the stiffness of the leads and solder joints. In order to take
all the stress factors into account, all the possible deformations of the leads and
solder joints are considered, (i.e., a total of 12 degrees of freedom have been
used). It should be noted that these stiffnesses are independent of the geometry
and material of the chip carriers and the PCB.

-BUMPERS'

(a)

J LEAD
CONTACT

~~~~~~~~';:(PLASTIC
(b) COVER

Figure 15-1 (a) Plastic quad flat pack (PQFP). (b) Plastic leaded chip carrier (PLCC).
458 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

15.1 BOUNDARY-VALUE PROBLEM

The composite structures under consideration are shown in Figure 15-2a and h.
For the PQFP, the thickness of the gull-wing lead is 0.005 in., and the radius
of the inner elbow is 0.005 in. The overall standoff height of the lead is 0.06
in., and the length of the horizontal part of the lead is 0.012 in. The minimum
height of the solder joint is 0.003 in.

11

(1,2,3,7,8,9)
FORCES or
v

J--.
DISPLACEMENTS

(4,5,6,10,11,12)
MOMENTS or
W u ROTATIONS

(a)

11

(1,2,3,7,8,9)
FORCES or
DISPLACEMENTS

(4,5,6,10,11,12)
MOMENTS or
ROTATIONS

(b)

Figure 15-2 (a) FE model and coordinate system for PQFP Gull-Wing lead and solder joint.
(b) Finite-element model and coordinate system for PLCC J-Iead and solder joint.
 BOUNDARY-VALUE PROBLEM 459

For the PLCC, the thickness of the J lead is 0.008 in., the minimum height
of the solder joint is 0.005 in., and the overall standoff height of the lead is
0.097 in. The width of the shoulder of the lead is 0.028 in., and the width of
the lower part is 0.018 in. The radius of the centroidal axis of the shoulder is
0.014 in., and the radius of the centroidal axis of the lower curved part is 0.029
in.
The partial differential equations governing the 3-D structures are220--221

a2u + -a2v + -a2w + 0 -

- =0
2fl)~72u 05-1)
axl axay axaz
a2v a2 u a2w
- +- +- +0 - 2fl)V 2 v = 0 05-2)
ayZ axay ayaz
a2w a2u a2v
- 2 +- +- +0 - 2fl)V 2W = 0 (15-3)
az axaz ayaz

where

(15-4)

In Equations 05-1) through (15-3), u, v, and w are the displacement components

in the x, y, and z directions, respectively, and fl is Poisson's ratio. The rela-
tionships between the stress components (J'x (normal stress acting in the x direc-
tion), (J' y (normal stress acting in the y direction), (J'z (normal stress acting in
the z direction), Txy (shear stress acting in xy plane), T yz (shear stress acting in
the yz plane), and Tzx (shear stress acting in zx plane), and the displacement
components u, v, and w are

(J'x =Z [0 - fl) au + fl(av + aw)]

ax ay az (15-5)

(J'y = Z[(1 - fl) avay + J.L(awaz + au)]

ax 05-6)

= Z [(1 - fl) aw + fl(au + av)] (15-7)

az ax ay
(J'z

au av)
Txy = G ( ay + ax 05-8)
460 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

'Tyz = G (avaZ + aw)

ay (15-9)

'Tzx = G (-aw + -au) (15-10)

ax az
where

E
z=--- --- (15-11)
(1 + ,...)(1 - 2,...)
E
G (15-12)
2( 1 + ,...)
and E is Young's modulus for the material.
The statement of the present boundary-value problem is to find a general
solution for u, v, and w from Equations (15-1) through (15-3), which satisfies
a given set of boundary conditions. There are 12 different sets of displacement
boundary conditions in the present study and they will be stated in the results
section. The stresses IT.. , ITy , IT" 'T..y , 'T y " and 'Tzx can then be calculated from
Equations (15-5) through (15-10) and the Mises stress intensity is determined
by the following equation

(15-13)

Obviously, closed-form solutions for the present boundary-value problem would

be very difficult to obtain. Therefore, numerical schemes such as the finite
element method203-219 were adopted.

15.2 FINITE ELEMENT METHODS

The finite element method is the principal means of solving boundary-value

problems. Part of its advantage stems from its ability to take care of irregular
boundary shapes and mixed material structures. Due to the generality and richness
of the concepts underlying this method, it has been employed with remarkable
success in solving problems in virtually all areas of engineering and mathematical
physics.
The basic concept of the finite element method is that a structure can be
decomposed into a finite number of elements. For each element, trial function
 FINITE ELEMENT METHODS 461

approximations of displacement components are used in conjunction with the

variational principle or Galerkin's approximation and matrix methods to trans-
form the boundary-value problem, Equations (15-1) through (15-10), and the
boundary conditions into a system of simultaneous algebraic equations.
In formulating the finite element method by the principle of minimum potential
energy, the displacement field is represented by interpolation functions together
with generalized displacements at a finite number of nodal points in each element.
In matrix form the assumed displacement may be written as (Fig. 15-3)

[u] = [A][s] (15-14)

where [u] is a displacement column matrix, [s] is a nodal displacement column

matrix, and [A] is a shape function matrix. In Figure 15-3, <D is the element
for one kind of material and ® is the element for the other.
The corresponding strain column matrix is

[e) = [L][u] (15-15)

where [L] is a differential operator matrix. The corresponding stress column

matrix is

(15-16)

where [D) is a material matrix, and [eo] and [<10 ] are column matrices of initial
strains and initial stresses, respectively. Equation (15-16) degenerates to Equa-
tions (15-5) through (15-10) if [eo] = [<10 ] = o.
The total potential energy functional (Vp) for any elastic solid, which is divided
into a finite number (n) of discrete elements (Vn), is given by

Vp = f
~ (-
Sn
[uf[T]dS - f
Vn
[U]T[F]dV)
(15-17)

+!~f [<1f[e]dV
2 n
Vn

where [11 is a traction column matrix acting on a surface Sm and [F] is a body
force column matrix. Substituting Equations (15-14) through (15-16) into Equa-
tion (15-17) leads to

Vp =; (- [sf[S] + ~[Sf[k][S]) (15-18)

462 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOlDER JOINTS

where

[k] = f
Vn
([L][A]f[D] ([L][A])dV (15-19)

and

[S] = f
Sn
[AY[T]dS + fVn
[AY[F]dV

+f Vn
([L][AW[D][Eo]dV

-f Vn
([L][AW[<To]dV (15-20)

are the element stiffness matrix and the nodal force column matrix, respectively.
The nodal displacements [s] for different elements are not completely indepen-
dent; a transformation is needed to relate the element nodal displacements to the
independent generalized global displacements. Then the compatibility equations
of the assembled structure are written in matrix form as (Fig. 15-3)

[s] = [a][r] (15-21)

where [a] is a compatibility transformation matrix and [r] is a global displacement

column matrix.
Substituting Equation (15-21) into Equation (15-18), yields

I
v:p = - frYeR] + 2[rY[K][r] (15-22)

where

[K] = L[aY[k][a] (15-23)

n

and

[R] = L[aY[S] (15-24)

n

are the global stiffness matrix and the global force column matrix, respectively.
 FINITE ELEMENT METHODS 463

®
[5] = [a] [Ij

[u] = [A] [5]

Where [u] = Displacement Column Matrix

[5] = Nodal Displacement Column Matrix
[A] = Shape Function Matrix
.[a] =
Compatibility Transformation Matrix
[Ij =
Global Displacement Column Matrix

Figure 15-3 Nodal points, nodal displacements, and global nodal displacements.

The principle of minimum potential energy requires that 8Vp = 0, that is,

ilV
----L 8[r] = (- [R] + [K][r])8[r] =0 (15-25)
a[r]

For arbitrary 8[r],

[K][r] = [R] (15-26)

Thus, for a given force vector [R] the corresponding displacement vector [r] can
be obtained by the inversion of [K] and the nodal displacements and the cor-
responding strains within each element may be found by Equations (15-21) and
(15-15), respectively. Most of the commercially available finite element codes
calculate the stresses at the nodal points of each element by Equation (15-16)
and average the stress values at the interface between elements. In that case, for
a composite structure special care should be noted. 203
Figure 15-2a depicts the three-dimensional finite element model used for the
determination of the stiffness of the gull-wing lead and solder joint. It consists
of 832 3D solid elements. Each element has 20 nodal points. Each nodal point
has 3 degrees of freedom. The Young's modulus used for the copper gull-wing
lead is 17,500 ksi, and that for the solder alloy is 1500 ksi. The Poisson's ratio
for the lead is 0.35, and for the solder it is 0.40.
Figure 15-2b depicts the three-dimensional finite element model used for the
determination of the stiffness of the J lead and solder joint for the PLCC SMA.
464 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

It consists of 1296 3D solid elements. The element type and material properties
are exactly the same as those for the gull-wing lead and solder joint for the PQFP
SMA.
The coordinate system for measuring forces, moments, displacements, and
rotations is the same for both the PQFP and PLCC SMAs and is also shown in
Figures 15-2a and b. This system consists of 12 coordinates that are identified
by 12 numbered arrows shown at specific directions and locations on the com-
posite structure. Single arrows (1, 2, 3, 7, 8,9) represent forces and the cor-
responding displacements, while double arrows (4, 5, 6, 10, 11, 12) represent
moments and the corresponding rotations. All the arrows point in the positive
direction of the forces and moments (right-hand rule). Coordinates 1 through 6
are specified at the center of the bottom surface of the solder joint (point A in
Fig. 15-2a and b). Coordinates 7 through 12 are specified at the center of the
end surface of the leads (point B in Fig. 15-2a and b). These two surfaces are
assumed to be rigid. There are 12 different sets of displacement boundary con-
ditions imposed at these two surfaces. Therefore, there are 12 analyses of each
model and their results are summarized in the next two sections.

15.3 STIFFNESS OF GULL-WING LEAD AND SOLDER JOINT

The 12 loading components (forces and moments) and the 12 corresponding

deformation components (displacements and rotations) of the gull-wing lead and
solder joint are related to each other through the symmetrical 12 x 12 stiffness
matrix [K] as shown in Table 15-1. In this matrix, the unit for the forces is lb/
in. and the unit for the moments is in. -lb/rad. It should be noted that this
stiffness matrix [K] is not the same as the one in Equation (15-25).
The definition of a typical stiffness coefficient is denoted by Kij' which means
the value of the action (force or moment) at restraint (coordinate) j induced by
a unit displacement (or rotation) at restraint i.209-219
For example, the elements in the first row are: 1 = 165, 2 = - 325, 3 =
0, 4 = 0, 5 = 0, 6 = - 3, 7 = -165, 8 = 325, 9 = 0, 10 = 0, 11 = 0,
and 12 = 2. This means that for a unit displacement in the 1 direction (Fig.
15-2a) with all the other directions (2 through 12) fixed, the force applied at C-
1 (from now on, C- will be the abbreviation for coordinate-) is 165 lb/in., the
force applied at C-2 is 325 lb/in. (in the opposite direction of Fig. l5-2a), the
force applied at C-3 and the moments applied at C-4 and C-5 are zero, the
moment applied at C-6 is 3 in. -lb/rad (in the opposite direction of Fig. 15-2a),
the force applied at C-7 is 165 lb/in. (in the opposite direction of Fig. 15-2a),
the force applied at C-8 is 325 lb/in., the force applied at C-9 and the moments
applied at C-I0 and C-ll are zero, and the moment applied at C-12 is 2 in. -lb/
rad.
 Table 15-1 Stiffness Matrix (Force in Ib-in.; Moment in in.llb/rad) for Gull-Wing Lead and Solder Joint
165 -325 0 0 0 -3 -165 325 0 0 0 2
-325 1648 0 0 0 -6 325 -1648 0 0 0 21
0 0 138 6.1 1.3 0 0 0 -138 2.3 2.6 0
0 0 6.1 .3 .1 0 0 0 -6.1 0 .1 0
0 0 1.3 .1 .1 0 0 0 -1.3 0 0 0
a-
"'" -3 -6 0 0 0 .2 3 6 0 0 0 .1
'"
K -165 325 0 0 0 3 165 -325 0 0 0 -2.1
325 -1648 0 0 0 6 -325 1648 0 0 0 21
0 0 -138 -6.1 -1.3 0 0 0 138 -2.3 -2.6 0
0 0 2.3 0 0 0 0 0 -2.3 .1 .1 0
0 0 2.6 .1 0 0 0 0 -2.6 .1 .1 0
2 21 0 0 0 .1 -2.1 21 \ 0 0 0 .3
 (a) (b)

(e) (d)

(e) (f)

Figure 15-4 Deformatio

n. (a) Unit displacement
2-direction.(c) Unit in l-direction. (b) Unit displacement
displacement in 3-direction in
. (d) Unit rotation in 4-d
irection. (e) Unit

4&&
 z~ z~
(g) (h)

z~
( i) (j)

z~ z~
(k) (I)

acement in 7-direction.
ion in 6-direction. (g) Unit displ
rotation in 5-direction. (f) Unit rotat in 9-dir ectio n. ij) Unit rotation in
(h) Unit displacement in 8-dir
ection. (i) Unit displacement ion.
12-d irect
ll-dir ectio n. (/) Unit rotation in
lO-direction. (k) Unit rotation in

467
468 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

The calculated deformations of the composition structure (gull-wing lead and

solder joint) for each imposed unit displacement and rotation are shown in Figures
IS-4a through 4/. The stress contours acting in the gull-wing lead and solder
joint for a unit displacement (0.0001 in.) in the C-I direction are shown in Figure
IS-Sa and b, respectively. Some significant results are summarized as follows
(the unit for forces is lb/in. and the unit for moments is in. -lb/rad):

1. Unit displacement (0.0001 in.) in the I-direction. The boundary conditions

for this case are: at point A, u = 0.0001 in., v = 0, w = 0, (aw 1 ay

°- av 1 az)/2 = 0, (au 1 az - awlax) 1 2 = 0, (avlax - au 1 ay) 1 2 =

and at point B, u = 0, v = 0, w = 0, (aw 1 ay - av 1 az) 1 2 = 0,
(au 1 az - aw 1 ax) 1 2 = 0, (av 1 ax - au 1 ay) 1 2 = 0.
The finite element results for the forces and moments are shown in the
first row of Table IS-I. The deformed shape of the composite structure
is shown in Figure IS-4a. The Mises stress contours in the gull-wing lead
and solder joint are shown in Figure IS-Sa and b, respectively. It can be
seen that the maximum stress (9218 psi) acting in the gull-wing lead
occurs at the inner elbows of the gull-wing lead due to stress concentration
and bending. Higher stress also occurs at the outer shoulder of the gull-
wing lead because of bending. Because of the residual manufacturing
stresses, it is not uncommon to observe cracks at the shoulder of the gull-
wing leads. In that case, further crack propagation in the gull-wing lead
under operation conditions is likely. The maximum Mises stress (124S
psi) acting in the solder joint occurs near the inside upper-tip of the solder
fillet, Figure IS-4b. The stresses acting on the outer toe of the solder joint
are very small. The results of this case can be used as an analog for
estimation of the thermal stresses (due to the thermal expansion mismatch
between the SMC and PCB) in the lead and solder joint.
2. Unit displacement (0.0001 in.) in the 2-direction. The boundary conditions
for this case are: at point A, u = 0, v = 0.0001 in., w = 0, (aw 1 ay
- av 1 az) 1 2 = 0, (au 1 az - aw 1 ax)/2 = 0, (av 1 ax - au 1 ay) 1 2
°
= and at point B, u = 0, v = 0, w = 0, (aw 1 ay - av 1 az) 1 2 =
0, (au 1 az - aw 1 ax) 1 2 = 0, (av 1 ax - au 1 ay) 1 2 = 0.
The finite element results are shown in the second row of the stiffness
matrix (Table IS-1). The deformed shape is shown in Figure IS-4b. The
Mises stress contours in the gull-wing lead and solder joint for cases 2
through 12 are not shown in this chapter but can be found in refer-
ences. 189 ,l90 The maximum stress (32,479 psi) occurs at the upper end-
surface of the gull-wing lead at the package. The maximum stress (1062
psi) acting at the solder joint also occurs near the inner solder fillet as in
case 1.
3. Unit displacement (0.0001 in.) in the 3-direction. The boundary conditions
for this case are: at point A, u = 0, v = 0, w = 0.0001 in., (aw 1 ay
 9217. -A
(a)
8582. = 8

7fJ47. =c

7312. =0

6676. =E

6041. =F

5406. = G

4710. =H

4135. =1

3500. = J

2864... K

2229. = L

1594. =M

958. .. N

y 323. = 0

1245 -A

1160 -8
(b) 1074 -c
988 -0

902 -E
816 =F

731 -G

645 =H

559 =1

473 -J

387 -K
302 =L

z~
216 =M

130 -N

44 =0

Figure 15-5 (a) Mises stress contours in the gull-wing. (b) Mises stress contours in the
PQFP solder joint. (Unit displacement, 0.0001 in., in 1 direction).

469
470 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

- aV 1 aZ) 1 2 = 0, (au 1 aZ - aW 1 ax) 1 2 = 0, (aV 1 ax - au 1 ay) 1

2 = 0, and at point B, u = 0, v = 0, w = 0, (aw 1 ay - av 1 az) 1 2
= 0, (aulaz - awlax)/2 = 0, (avlax - aulay)/2 = O.
The finite element results for the forces and moments are shown in the
third row of the stiffness matrix (Table IS-1). Figure IS-4c shows the
deformed shape. The maximum stress (82SS psi) occurs at both sides of
the lower part of the lead near the inner elbow. The maximum stress
(1388 psi) in the solder joint occurs at both sides near the tip of the inner
solder fillet.
4. Unit rotation (0.0001 rad) in the 4-direction. The boundary conditions
for this case are: at point A, u = 0, v = 0, w = 0, (aw 1 ay - av 1 az)
12 = 0.0001 rad, (au 1 az - aw / ax) 1 2 = 0, (av / ax - au 1 ay) / 2
= 0, and at point B, u = 0, v = 0, w = 0, (aw 1 ay - av / az) 12 =
0, (au 1 az - aw / ax) 1 2 = 0, (av 1 ax - au 1 ay) 1 2 = O.
The finite element results are shown in the fourth row of the stiffness
matrix (Table IS-I). The deformed shape of this case is shown in Figure
IS-4d. The maximum stress (4718 psi) occurs near the inner elbow of
the lower part of the lead. The maximum stress (778 psi) in the solder
joint occurs at both sides near the tip of the inner solder fillet.
S. Unit rotation (0.0001 rad) in the S-direction. The boundary conditions
for this case are: at point A, u = 0, v = 0, w = 0, (aw 1 ay - av 1 az)
/2 = 0, (au 1 az - aw / ax) / 2 = 0.0001 rad, (av 1 ax - au 1 ay) 12
= 0, and at point B, u = 0, v = 0, w = 0, (aw 1 ay - av 1 az) 1 2 =
0, (au 1 az- aw 1 ax) 1 2 = 0, (av 1 ax - au 1 ay) 12 = O.
The finite element results are shown in the fifth row of the stiffness
matrix (Table IS-I). Figure IS-4e shows the deformed shape. The max-
imum stress (1279 psi) occurs near both sides of the inner elbow of the
lower part of the lead. The maximum stress (131 psi) in the solder joint
occurs near the inner solder fillet.
6. Unit rotation (0.0001 rad) in the 6-direction. The boundary conditions
for this case are: at point A, u = 0, v = 0, w = 0, (aw 1 ay - av 1 az)
12 = 0, (au 1 az - aw 1 ax) 1 2 = 0, (av 1 ax - au / ay) 1 2 = 0.0001
rad, and at point B, u = 0, v = 0, w = 0, (aw 1 ay - av 1 az) 1 2 =
0, (au 1 az - aw 1 ax) / 2 = 0, (av 1 ax - au / ay) / 2 = O. This set of
boundary conditions means there is a unit rotation in the 6-direction
(Figure IS-2a) with all the other directions (1 through Sand 7 through
12) fixed.
The finite element results are: the forced applied at C-l is 3 lb/in. (in
the opposite direction of Figure IS-2a), the force applied at C-2 is 6lbl
in. (in the opposite direction of Figure IS-2a), the force applied at C-3
is 0, the moments applied at C-4 and C-S are 0, the moment applied at
C-6 is 0.2 in. -lb/rad, the force applied at C-7 is 3 lb/in., the force
applied at C-8 is 6 lb/in., the force applied at C-9 is 0, the moments
 STIFFNESS OF GULL-WING LEAD AND SOLDER JOINT 471

applied at C-lO and C-ll are 0, the moment applied at C-12 is 0.1 in. - lb/
rad. All of these values are shown in the sixth row of the stiffness matrix
(Table 15-1). The deformed shape is shown in Figure 15-4/. The maximum
stress (2999 psi) occurs at the inner elbow of the lower part of the lead.
The maximum stress (595 psi) in the solder joint occurs at the tip of the
inner solder fillet.
7. Unit displacement (0.0001 in.) in the 7-direction. The boundary conditions
for this case are: at point A, U = 0, v = 0, w = 0, (iJw / iJy - iJv /
iJz) / 2 = 0, (iJu / iJz - iJw / iJx) / 2 = 0, (iJv / iJx - iJu / iJy) / 2 = 0,
and at point B, u = 0.0001 in., v = 0, w = 0, (iJw / iJy - iJv / iJz) / 2
= 0, (iJu / iJz - ilw / ilx) / 2 = 0, (iJv / ilx - iJu / iJy) / 2 = 0.
The finite element results are shown in the seventh row of the stiffness
matrix (Table 15-1). The deformed shape of this case is shown in Figure
15-4g. The Mises stress contours acting at the gull-wing lead and solder
joint are the same as those for a unit displacement in I-direction.
8. Unit displacement (0.0001 in.) in the 8-direction. The boundary conditions
for this case are: at point A, u = 0, v = 0, w = 0, (ilw / iJy - ilv /
ilz) / 2 = 0, (ilu / ilz - ilw / ilx) / 2 = 0, (ilv / iJx - ilu / ily) / 2 = 0,
and at point B, u = 0, v = 0.0001 in., w = 0, (ilw / iJy - ilv / ilz) / 2
= 0, (ilu / ilz - ilw / ilx) / 2 = 0, (ilv / ilx - ilu / ily) / 2 = 0.
The finite element results are shown in the eighth row of the stiffness
matrix (Table 15-1). Figure 15-4h shows the deformed shape of the com-
posite structure. The Mises stress contours in the gull-wing lead and solder
joint are the same as those for a unit displacement in 2-direction.
9. Unit displacement (0.0001 in.) in the 9-direction. The boundary conditions
for this case are: at point A, u = 0, v = 0, w = 0, (ilw / iJy - iJv /
ilz) / 2 = 0, (ilu / ilz - ilw / ilx) / 2 = 0, (ilv / ilx - ilu / ily) / 2 = 0,
and at point B, u = 0, v = 0, w = 0.0001 in., (iJw / iJy - ilv / ilz) / 2
= 0, (ilu / ilz - ilw / iJx) / 2 = 0, (iJv / iJx - iJu / ily) / 2 = 0.
The finite element results are shown in the ninth row of the stiffness
matrix (Table 15-1). Figure 15-4i shows the deformed shape. The Mises
stress contours in the composite structure are the same as those for a unit
displacement in the 3-direction.
10. Unit rotation (0.0001 rad) in the lO-direction. The boundary conditions
for this case are: at point A, u = 0, v = 0, w = 0, (ilw / ily - ilv /
ilz) / 2 = 0, (ilu / ilz - iJw / ilx) / 2 = 0, (ilv / ilx - ilu / ily) / 2 = 0,
and at point B, u = 0, v = 0, w = 0, (ilw / ily - ilv / iJz) / 2 = 0.0001
rad., (ilu / ilz - ilw / iJx) / 2 = 0, (ilv / ilx - ilu / iJy) /2 = 0.
The finite element results are shown in the tenth row of the stiffness
matrix (Table 15-1). The deforniation shape is shown in Figure 15-4j.
The maximum stress (2562 psi) occurs near the upper end-surface. The
maximum stress (65 psi) in the solder joint occurs near the inner solder
fillet.
472 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

11. Unit rotation (0.0001 rad) in the ll-direction. The boundary conditions
for this case are: at point A, u = 0, v = 0, w = 0, (aw / ay - av /
az) / 2 = 0, (au / az - aw / ax) / 2 = 0, (av / ax - au / ay) / 2 = 0,
and at point B, U = 0, v = 0, w = 0, (aw / ay - av / az) / 2 = 0,
(au / az - aw / ax) / 2 = 0.0001 rad, (av / ax - au / ay) / 2 = 0.
The finite element results are shown in the eleventh row of the stiffness
matrix (Table 15-1). The deformed shape is shown in Figure 15-4k. The
maximum stress (1525 psi) occurs near the elbow of the lower part of
the lead and at the upper end-surface of the lead. The maximum stress
(347 psi) in the solder joint occurs at both sides of the tip of the inner
solder fillet.
12. Unit rotation (0.0001 rad) in the 12-direction. The boundary conditions
for this case are: at point A, u = 0, v = 0, w = 0, (aw / ay - av /
az) / 2 = 0, (au / az - ow / ax) / 2 = 0, (av / ax - au / ay) / 2 = 0,
and at point B, u = 0, v = 0, w = 0, (ow / ay - ov / az) / 2 = 0,
(au / az - aw / ax) / 2 = 0, (av / ax - au / oy) / 2 = 0.0001 rad.
The finite element results are shown in the twelfth row of the stiffness
matrix (Table 15-1). Figure 15-41 shows the deformed shape. The max-
imum stress (5264 psi) occurs at the upper end-surface of the lead. The
maximum stress (152 psi) in the solder joint occurs at both sides of the
tip of the inner solder fillet.

In all the twelve cases considered, the location of the maximum Mises stress
acting at the solder joint is near the tip of the inner solder fillet. The stresses in
the outer solder fillet are very small. Thus, from a failure mechanism standpoint,
solder joint fatigue cracking should initiate at the tip of the inner solder fillet.

15.4 STIFFNESS OF J LEAD AND SOLDER JOINT

Based on the same definition of stiffness, analysis procedure, material property,

boundary conditions, etc., for the gull-wing lead and solder joint, the stiffness
matrix relating the 12 loading components (forces and moments) and the 12
corresponding deformation components (displacements and rotations) of the J
lead and solder joint is determined and is shown in Table 15-2. The deformed
shapes for all 12 cases are shown in Figure 15-6a through I.

15.4.1 Unit Displacement (0.0001 in.) in the I-Direction

The finite element results for the forces and moments are shown in the first row
of Table 15-2. It can be seen that for a unit displacement in the 1 direction
(Figure 15-2b) with all the other direction fixed, the force applied at C-l is 298
 Table 15-2 Stiffness Matrix (Force in Ib/in.; Moment in in.-Ib/rad) for J-Lead and Solder Joint
298 105 0 0 0 -13 -298 -105 0 0 0 -14
105 8973 0 0 0 233 -105 8973 0 0 0 -109
0 0 665 26 -21 0 0 0 -665 39 11 0
0 0 26 1.7 -.8 0 0 0 -26 .9 .5 0
0 0 -21 -.8 .9 0 0 0 21 -1.2 -.6 0
~
-13 233 0 0 0 7 13 -233 0 0 0 -2.3
"w
K -298 -105 0 0 0 13 298 105 0 0 0 14
-105 -8973 0 0 0 -233 105 8973 0 0 0 109
0 0 -665 -26 21 0 0 0 665 -39 -11 0
0 0 39 .9 -1.2 0 0 0 -39 3 .6 0
0 0 11 .5 -.6 0 0 0 -11 .6 .4 0
-14 -109 0 0 0 -2.3 14 109 0 0 0 2.1
- ~----- - -
 z~

(b)

z~ z~
(e) (d)

z~ z~
Figure 15-6 Deformation. (a) Unit displacement in 1 direction. (b) Unit displacement in
2 direction. (c) Unit displacement in 3 direction. (d) Unit rotation in 4 direction. (e) Unit

474
 z~ z~

z~
(i)

z~

rotation in 5 direction. (f) Unit rotation in 6 direction. (g) Unit displacement in 7 direction.
(h) Unit displacement in 8 direction. (i) Unit displacement in 9 direction. (il Unit rotation in
10 direction. (k) Unit rotation in 11 direction. (/) Unit rotation in 12 direction.

475
476 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOlDER JOINTS

4193 = B

3883= C

357:2= 0

z~
3261 = E

2951 = F

2640= G

2329= H

2019 = I

1708 =J
lJ9I! = I<

1087 = L

77&= lot

4&6= N

(a)

&52 = A

608= B

5&3 = C

519 = 0

475 = E

430 = F

386 = G

97 = I

252 = J

208 = I<

1&4 = L

= lot

~
119

75 =N
(b)
Z 30 = 0

Figure 15-7 (a) Mises stress contours in the J lead. (b) Mises stress contours in the PlCC
solder joint. (Unit displacement, 0.0001 in., in 1 direction).
 STIFFNESS OF J LEAD AND SOLDER JOINT 477

lb/in., the force applied at C-2 is 105 Ib/in., the force applied at C-3 and the
moments applied at C-4 and C-5 are zero, the moment applied at C-6 is 13
in. -Ib/rad (in the opposite direction of Figure 15-2b), the force applied at C-
7 is 298 Ib/in. (in the opposite direction of Figure 15-2b), the force applied at
C-8 is 105 Ib/in. (in the opposite direction of Figure 15-2b), the force applied
at C-9 and the moments applied at C-lO and C-ll are zero, and the moment
applied at C-12 is 14 in. -lb/rad (in the opposite direction of Figure 15-2b).
The deformed shape of the composite structure is shown in Figure 15-6a. The
Mises stress contours in the J lead and solder joint are shown in Figure 15-7a
and b, respectively. It can be seen that the maximum stress (4504 psi) acting in
the J lead occurs at the inner shoulder of the J lead due to stress concentration
and bending. Just like the PQFP's lead, higher stress also occurs at the outer
shoulder of the J lead because of bending. Because of residual manufacturing
stresses, it is not uncommon to observe cracks at the shoulder of the J lead. In
that case, further crack propagation in the J lead under operation conditions is
likely. The maximum Mises stress (2083 psi) acting in the solder joint occurs
near the tip of the outer solder fillet (Fig. 15-7b). The stresses acting on the
other tip of the solder joint are very small. Thus, cracking of the solder joint
should start near the tip of the outer solder fillet.

15.4.2 Unit Displacement and Rotation in Other Directions

For the other II cases the stress contours in the J lead and solder joint are not
shown in this chapter but can be found in the reference. J9J However, the cal-
culated forces and moments are summarized in Table 15-2 and the deformed
shapes are shown in Figure 15-6b through 6/.

15.4.3 Comparison of the Stiffness Matrices between the PQFPs and

PLCCs

It is important to compare the stress contours in the PLCC solder joint (Figure
15-7b) with those in the PQFP solder joint (Figure 15-4b). It can be seen that
the values of the PLCC solder joint are larger than those of the PQFP solder
joint. Furthermore, by comparing the elements of the stiffness matrix of the J
lead and solder joint (Table 15-2) with those of the PQFP gull-wing lead and
solder joint (Table 15-1), most of the elements of the PQFP are smaller than
those of the PLCC. Consequently, the gull-wing leads are more compliant than
those of the J leads, and the stresses in the PQFP solder joints are less than
those in the PLCC. Restated, the fatigue life of the PQFP solder joints should
be longer than those of PLCC solder joints.
478 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOlDER JOINTS

Solder Joint Reliability Under Shock

and Vibration Conditions
Vibration analysis of structures consists of four major steps. First, the structure
of interest is identified, its boundary conditions are estimated, and its interfaces
with other structures are understood. Second, the natural frequencies and mode
shapes of the structure are determined by analysis or direct experimental mea-
surement. Third, the time history of the loads on· the structure are estimated.
Fourth, these loads are applied to an analytical model of the structure to determine
its response. The first two steps are called "free vibration analysis." The last
two steps are called "forced vibration analysis" and are usually executed by
modal superposition, direct integration, or frequency domain methods.222-734
Free vibration analysis is the most crucial step in structural dynamics. Many
vibration problems can be minimized by avoiding coincident resonances that can
magnify the dynamic loads and stresses in the structure.
All problems of free vibration of undamped linear systems are eigenvalue
problems, see references 222-226,231-234 for examples. In most of continuum me-
chanics the problem is to determine the natural mode of vibration for which all
points in the continuum perform simple harmonic oscillations that are in phase
with each other and that have the same frequency. Accordingly, all points in
the continuum obtain their maximum velocities simultaneously as they pass
through their neutral positions, and they all reach their extreme displacement
simultaneously. A frequency for which natural vibration is possible is an eigen-
value of the boundary-value problem and is called a natural frequency. The
number of the natural frequencies is equal to the number of degrees of freedom
of the continuum. Theoretically speaking, it is infinite.
In this part, we will first present the free vibration of soldered and unsoldered
leads of SOICs, PLCCs, and PQFPs. Then we will present the free vibration of
a constrained PCB with a SMC.

15.5 FREE VIBRATION OF SOLDERED AND UNSOLDERED LEADS

The structures under consideration are shown in Figure 15-8 for SOICs and
Figure 15-2a and b for PQFPs and PLCCs, respectively. All the leads are assumed
to be clamped (fixed) at the upper end due to the larger stiffness of the package
bodies compared with the leads. The lower end of the lead is assumed to be
clamped for a good solder joint (or soldered lead) and free to move for a bad
solder joint (or unsoldered lead). The partial differential equation governing the
3D vibration of these leads is:22°--226
 FREE VIBRATION OF SOLDERED AND UNSOLDERED LEADS 479

Figure 15-8 Finite-element model for sOle lead and solder joint.

where V = gradient operator, V2 = V . V, Q = displacement vector, Il-

Poisson's ratio, p = mass density, and E = Young's modulus.
The natural modes are characterized by the equation

..
Q =L cjAj cos (wjt - 'Yj) (15-28)
j=1

in which Ci and 'Yj are arbitrary constants, Wi is the angular frequency, and Ai(x,
y, z) is a vector field that is independent of time. Substitution of Equation (15-
28) into Equation (15-27) yields

V(V . A;) + (1 - 21l-)V2 A j + 2(1 + 1l-)(1 - 21l-) -t

pW?
Aj = 0

i = 1, 2, . . . , 00.
(15-29)

For free vibration, the equations of the boundary conditions are homogeneous. 220-
226 Consequently, Equation (15-29), in conjunction with the boundary conditions,
possesses a solution other than Ai = 0 if, and only if, Wj belongs to one of an
infinite sequence of discrete values Wt. W2, W3 • • • • These are the natural
frequencies of the structure. To any natural frequency Wi, there corresponds a
480 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

nonzero vector field A; that satisfies Equation (15-29) and the boundary condi-
tions. This vector field contains an arbitrary constant factor; that is, ciA; is also
a solution. Thus, the amplitudes of the oscillations of the particles are not
determinate, but the ~atio of the amplitudes for any two particles is fixed. Each
particle executes a simple harmonic oscillation on the straight line in which the
vector A; lies.
Exact solutions of Equation (15-29) for the structures shown in Figures 15-
2a and band 15-8 are very difficult to obtain. Therefore, the finite element
method was adopted to transform the boundary-value problem, Equation (15-
29) and the boundary conditions, into a system of simultaneous algebraic equa-
tions, that is,

[K - w2 M]A =0 (15-30)

where K = stiffness matrix, M = mass matrix, and A displacement com-

ponents matrix. The size of these matrices (K and M are square matrices, and
A is a column matrix) equals the number of degrees of freedom of the finite
element model. Equation (15-30) can be easily solved for the frequencies (w;)
and the corresponding mode shapes (A;) by a computer (i = 1, 2, ... , N,
where N equals the number of degrees of freedom).

15.5.1 Vibration Results for Wide SOlCs

Figure 15-8 shows the finite element model of the wide SOIC gull-wing lead
and solder joint. The length of the horizontal parts of the lead was 0.017 in.
The radius of the centroidal axis of the elbows was 0.005 in. and the width of
the lead was 0.016 in. The thickness of the gull-wing lead was treated as a
variable. Specifically, three different thicknesses were considered, 0.008, 0.01,
and 0.012 in. Consequently, the overall standoff heights of the lead modeled
were 0.057,0.059, and 0.061 in.

Table 15-3 Vibration Frequencies for Soldered and Unsoldered Wide

SOIC Leads
Vibration Frequency (kHz)
Wide-SOIC Lead Thickness (in.imm)
Mode .008/.203 .010/.254 .012/.305
Soldered Unsoldered Soldered Unsoldered Soldered Unsoldered
lead lead lead lead lead lead
143.6 24.6 175.8 30.7 205.8 36.7
2 379.4 122.1 440.5 147.1 488.4 169.0
3 747.6 152.4 840.9 185.6 896.8 216.5
 FREE VIBRATION OF SOLDERED AND UNSOLDERED LEADS 481

299

I~ ~
I I

-
-
~

~~
-
t.:I 169 GOO\)
== i--""""
~
..!III
-

I- -

- -

BAD JOINT
- -
1 1 _I I
9
.2 .22 .24 .26 .28 .3
Wide-SOle Lead Thickness (mm)
Figure 15·9 Frequencies for various wide-SOle lead thickness.

Table 15-3 summarizes the first three natural frequencies for all the wide SOIC
cases studied. The first (fundamental) frequency of the soldered leads (good
joints) is at least 5 times larger than that of the unsoldered leads (bad joints).
This is true for all the thicknesses under consideration, as is shown in Figure
15-9. Consequently, it is possible to identify the soldered and unsoldered leads
by examining the frequency of vibration of the leads. Figure 15-9 also shows
that the fundamental frequency does not vary significantly with the lead thickness.

15.5.2 Vibration Results for Narrow SOles

The finite element model for the narrow SOIC is basically the same as for the
wide SOIC. The lead thicknesses considered in these case were 0.006, 0.008,
and 0.01 in. The overall standoff heights of the leads were 0.036, 0.038, and
0.04 in. The length of the horizontal part of the lead was 0.016 in.
482 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

Table 15-4 Vibration Frequencies for Soldered and Unsoldered Narrow

SOIC Leads
Vibration Frequency (kHz)
Narrow SOIC Lead Thickness (in.lmm)
Mode .006/.152 .008/.203 .010/.254
Soldered Unsoldered Soldered Unsoldered Soldered Unsoldered
lead lead lead lead lead lead
191. 9 35.1 248.6 46.6 300.2 57.9
2 564.2 156.8 679.8 199.3 759.8 235.2
3 1032.0 202.5 1218.7 261.3 1250.3 314.9

3ee I I I I

- ~-
_25e ~~
to:!
~.---;lO\~~
-
tIl
.!ill - GOO -
~
";./'
- -

- -

.... -

BAD JOINT

- -
e.15 I
.17
I
.19
I
.21
I
.23
I
.25
Narrow-SOle Lead Thickness (mm)
Figure 15-10 Frequencies for various narrow-SOle lead thickness.
 FREE VIBRATION OF SOLDERED AND UNSOLDERED LEADS 483

The first three frequencies for all the cases calculated are summarized in Table
15-4. It can be seen in the table that the fundamental frequency of the narrow
sOle leads is higher than that of the wide sOle leads. This is because the
narrow sOle lead is shorter than the wide sOle lead (Le., stiffer in resisting
vibration). The fundamental frequency of the narrow sOle leads with various
lead thickness is shown in Figure 15-10. It can be seen that, just like the wide
sOle cases, the fundamental frequency does not vary much with the lead thick-
ness.

15.5.3 Vibration Results for PLCCs

Figure 15-2b shows the finite element model for the PLee J lead and solder
joint. The thicknesses of the J lead considered were 0.006, 0.008, and 0.01 in.
The overall standoff heights of the lead were 0.095,0.097, and 0.099 in. The
radius of the centroidal axis of the shoulder was 0.014 in., and the radius of the
centroidal axis of the lower curved part was 0.029 in. The width of the shoulder
of the lead was 0.028 in., and the width of the lower part was 0.018 in.
Table 15-5 shows the first three frequencies of the soldered and unsoldered
leads for three different thicknesses. It can be seen that the fundamental frequency
of the soldered lead is more than 7 times larger than that of the unsoldered lead.
Figure 15-11 shows the plot of fundamental frequency versus J lead thickness.
It is again to be noted that the effect of thickness of the lead on the fundamental
frequency is negligible and that the frequencies for the soldered lead and the
unsoldered lead are very different.

15.5.4 Vibration Results for PQFPs

The finite element model of a PQFP lead and solder joint is shown in Figure
I5-2a. The thickness of the gull-wing lead were 0.003, 0.005, and 0.007 in.
The overall standoff heights of the lead were 0.056, 0.06, and 0.064 in. The

Table 15-5 Vibration Frequencies for Soldered and Unsoldered PLCC Leads
Vibration Frequency (kHz)
PLCC Lead Thickness (in.imm)
Mode .006/.152 .008/.203 .010/.254
Soldered Unsoldered Soldered Unsoldered Soldered Unsoldered
lead lead lead lead lead lead
1 67.5 9.0 88.7 12.0 108.6 14.9
2 146.9 24.4 190.6 32.4 229.9 40.2
3 207.5 80.6 255.9 106.2 291.9 130.7
484 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

12B I T T r T

1-. -
_188
~
~
-
~

== -
~~1
,!:ill I-

~
c.l
~
81!
~
GOO
~
Q,)
;j
tj4
-

-
Q,)
~
61!
~

td
+J I- -
~

e
Q,)
48
td
'0
~
I- -
;j
~
21!
BAD JOINT
-
I I I I I
~15 .17 .19 .21 .23 .25
PLee Lead Thickness (mm)
Figure 15-11 Frequencies for various PLCC lead thickness.

length of the horizontal part of the lead was 0.012 in. The radius of the centroidal
axis of the elbow was 0.008 in.
The fundamental and the next two higher frequencies of vibration of the
soldered and unsoldered leads are shown in Figure 15-12 and Table 15-6, re-
spectively. Again, it can be seen that the frequency difference between the
soldered and unsoldered leads is very large. Consequently, a good or bad solder
joint can be detected by measuring its natural frequencies.

15.5.5 Experimental Verification

A laser Doppler vibrometer (LDV) system has been developed for measuring
the vibration frequencies of all these SMC leads and for solder joint inspection. 200
The experimental results and the finite element solutions presented herein are in
close agreement. 200
 FREE VIBRATION OF SOlDERED AND UNSOLDERED LEADS 485

7
160 I I I

-
- toOl
//
--
1:1:1 120
.!:ill
~
Co)
/Cl:'O~
00\~'t
s:l
~
I- -
::3
cr
L
-
~
lot 80
~
-/
~
+'
s:l -
e
~

~
"'d 40
s:l
::3
~

- BAD JOINT -

_L
o
.08 .1 .12 .14 .16 .18

PQFP Lead Thickness (mm)

Figure 15·12 Frequencies for various PQFP lead thickness.

Table 15-6 Vibration Frequencies for Soldered and Unsoldered PQFP Leads
Vibration Frequency (kHz)
PQFP Lead Thickness (in.lmm)
Mode .003/.076 .005/.127 .0071.178
Soldered Unsoldered Soldered Unsoldered Soldered Unsoldered
lead lead lead lead lead lead
1 69.4 10.4 114.3 17.4 157.1 24.3
2 165.0 62.6 261.7 102.3 342.7 138.8
3 335.6 76.3 521.5 123.8 666.3 167.6
486 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

15.6 FREE VIBRATION OF A CONSTRAINED PCB WITH A SMC

Figure 15-13 shows a slender PCB with L as its largest dimension. It is supported
by linear and rotational springs at both ends. A SMC is attached at the center
of the PCB.
The Bernoulli-Euler equation of motion for small amplitude, free transverse
vibrations of the structure shown in Figure 15-13 is

(15-31a)

where 8(x) is the Dirac delta function and has the properties

8(x) == 0, x =1= ° (I5-31b)

f +OO
-00 8(x)dx == 1 (15-31c)

L[8(x)] == 1 (I5-31d)

and L is the Laplace transformation. In Equation (15-31a), E is the Young's

modulus, I is the area moment of inertia of beam (PCB) cross section about the
neutral axis, EI is the flexural rigidity of the PCB, y is the deflection of the
PCB, M is the concentrated mass of a SMC, and m is the mass per unit length
of the PCB.

ik
M
E,I k

'lK K[
~~

I
I
I
I
I
1== U2 I 1== U2
I

14 ~ X
~I'"
I
I
~I

Figure 15-13 Transverse vibration of a PCB with a SMC.

 FREE VIBRAITON OF A CONSTRAINED PCB WITH A SMC 487

The boundary conditions for the structure shown in Figure 15-13 are
Atx = 0:

(15-32)

(15-33)

Atx = L = 21:

a3y
EI- = Ky (15-34)
a~

EI a2y = _kay (15-35)

ail ax
where K is the stiffness of the translational spring and k is the stiffness of the
rotational spring.
For simple harmonic motion with frequency 00, we have

y(x, t) = Y(x)e iwt (15-36)

Then the equation of motion, Equation (15-31a) becomes

cry -
dXZ - <p [ML8(x - 1) + I]Y = 0 (15-37)

where

<p4 = _
moo 2
(15-38)
EI
_ M
M = mL (15-39)

And the boundary conditions of the present problem, Equations (15-32) through
(15-35) become

Atx = 0:

(15-40)
488 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

(15-41)

At x = L = 21:
d 3y _
L3- = KY (15-42)
dx 3
d2y _ dY
L- = -k- (15-43)
dx 2 dx

where

_ KL3
K=- (15-44)
EI

,,= EI kL (15-45)

Equations (15-37) and (15-40) through (15-43) define an eigenvalue problem

that is not self-adjoint, the solution of which can be most conveniently obtained
by Laplace transforms. After some algebraic manipulation, one obtains the fol-
lowing frequency equation:

M (al2B - a22A) [~ (sinh wi - sin wi) - (cos wi + cosh WI)]

- (a2JA - allB)
- M [" ~2 (cosh wi - cos wI) +

~ (sin wi + sinh wi) ]

- M~(sinh wi - sin wi) a +2a =0 (15-46)

where

p
all = ~3 (sin wL + sinh wL) + ~3 (sinh wL - sin wL)
 FREE VIBRAlTON OF A CONSTRAINED PCB WITH A SMC 489

- 2K (COS wL + cosh wL)

K _
a12 = - 13 (sin wL + sinh wL) + ~k(sinh wL - sin wL)

- (~2 - ~~) (cos wL - cosh wL)

l(2
a22 = ~(sinh wL - sin wL) + 13 (sin wL + sinh wL)

+ 2k( cos wL + cosh wL)

A = ~K (sinh wi - sin wI) - ~4 (cosh wi + cos wI)

B = ~2k (cos wi - cosh wI) - ~3 (sin wi + sinh wI)

~ = /fiL (15-47)

It can be seen from Equation (15-46) that the frequency equation is a com-
bination of the trigonometric and hyperbolic functions. It oscillates very rapidly
and attains very large values between successive roots. It can be shown that the
function of the frequency equation is monotonic and has no points of inflection
between roots. Once the root (frequency parameter, ~ = <pL) has been obtained,
the frequency if) of the structural system can be determined by Equation (15-
38) as

(15-48)

The first two roots for a wide range of values of PCB support conditions (spring
constants) are shown in Tables 15-6 through 15-10 for M = 0, 0.01, 0.05, 0.1,
1, respectively. In 199, these tables have been used to calculate the natural
frequencies of a PCB with various SMCs. These natural frequencies are very
important for designing SMA vibration experiments and for design for reliability
of solder joints.
Corresponding to each frequency, the mode shape can be easily obtained from
the following expression
490 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

y(X) M (
y(l) = 2.:l (-a12B + [K . .
a22A ) j33(smh <pX - sm <pX) - (COS <pX + cosh <px) ]

+ (a21 A - allB{:icOSh <px - cos <px) + i(Sin <px + sinh <px) ]

+ j3.:l [sinh <p(x - l) - sin <p(x - l)]U(x - l)) (15-49)

where y(l) is the deflection at x = I, and U(x - l) = 0 for x < I, V(x - l) =

112 for x = I, and V(x - l) = 1 for x > I.

Table 15-6a Values of 13 for M = 0 (Mode 1)

k
K 0.0 0.01 0.05 0.1 0.5 1.0 100.0
0.00 4.73004 0.69950 1.04348 1.23722 1.80856 2.09598 3.11092
0.01 0.37605 0.37605 0.37605 0.37605 0.37605 0.37605 0.37606
0.05 0.56222 0.56222 0.56223 0.56223 0.56224 0.56226 0.56232
0.10 0.66846 0.66846 0.66847 0.66847 0.66851 0.66854 0.66869
0.50 0.99792 0.99793 0.99797 0.99801 0.99827 0.99850 0.99962
1.00 1.18429 1.18430 1.18438 1.18448 1.18510 1.18564 1.18830
100.00 2.87675 2.87877 2.88671 2.89636 2.96383 3.02968 3.49775

Table 15-6b Values of 13 for M = o (Mode 2)

k
K 0.0 0.01 0.05 0.1 0.5 1.0 100.0
0.00 7.85320 4.73411 4.75016 4.76972 4.90881 5.04878 6.22286
0.01 0.49492 0.73969 1.05639 1.24500 1.81100 2.09750 3.11126
0.05 0.74006 0.85688 1.10377 1.27474 1.82064 2.10353 3.11261
0.10 0.88006 0.95699 1.15553 1.30921 1.83247 2.11099 3.11430
0.50 1.31568 1.34095 1.42895 1.51833 1.91971 2.16798 3.12769
1.00 1.56415 1.57924 1.63500 1.69626 2.01415 2.23331 3.14418
100.00 4.66378 4.66379 4.66380 4.66381 4.66388 4.66396 4.66473

Table 15-7a Values of 13 for M = 0.01 (Mode 1)

k
K 0.0 0.01 0.05 0.1 0.5 1.0 100.0
0.00 4.71088 0.69950 1.04348 1.23722 1.80856 2.09598 3.11092
0.01 0.37511 0.37511 0.37511 0.37511 0.375.11 0.37511 0.37512
0.05 0.56083 0.56083 0.56083 0.56083 0.56085 0.56086 0.56092
0.10 0.66680 0.66680 0.66680 0.66681 0.66684 0.66688 0.66703
0.50 0.99542 0.99543 0.99546 0.99550 0.99577 0.99600 0.99713
1.00 1.18129 1.18131 1.18139 1.18148 1.18211 1.18266 1.18533
100.00 2.86437 2.86638 2.87431 2.88393 2.95123 3.01695 3.48488
 FREE VIBRAITON OF A CONSTRAINED PCB WITH A SMC 491

Table 15-7b Values of 13 for M = 0.01 (Mode 2)

K 0.0 0.01 0.05 0.1 0.5 1.0 100.0

0.00 11.00527 4.71491 4.73082 4.75022 4.88801 5.02654 6.17803
0.01 0.49492 0.73969 1.05639 1.24500 1.81100 2.09750 3.11126
0.05 0.74006 0.85688 1.10377 1.27474 1.82064 2.10353 3.11261
0.10 0.88006 0.95699 1.15553 1.30921 1.83247 2.11099 3.11430
0.50 1.31568 1.34095 1.42895 1.51833 1.91971 2.16798 3.12769
1.00 1.56415 1.57924 1.63500 1.69626 2.01415 2.23331 3.14418
100.00 4.66378 4.66379 4.66380 4.66381 4.66388 4.66396 4.66473

Table 15-8a Values of 13 for M = 0.05 (Mode 1)

k
K 0.0 0.01 0.05 0.1 0.5 1.0 100.0
0.00 4.58791 0.69950 1.04348 1.23722 1.80856 2.09598 3.11092
0.01 0.37148 0.37148 0.37148 0.37149 0.37149 0.37149 0.37150
0.05 0.55540 0.55540 0.55540 0.55541 0.55542 0.55544 0.55550
0.10 0.66034 0.66034 0.66035 0.66035 0.66039 0.66042 0.66058
0.50 0.98570 0.98571 0.98575 0.98579 0.98606 0.98630 0.98746
1.00 1.16964 1.16966 1.16974 1.16984 1.17048 1.17105 1.17379
100.00 2.81414 2.81612 2.82392 2.83339 2.89965 2.96437 3.42732

Table 15-8b Values of 13 for M = 0.05 (Mode 2)

k
K 0.0 0.01 0.05 0.1 0.5 1.0 100.0
0.00 5.69410 4.59143 4.60528 4.62209 4.73879 4.85026 5.37269
0.01 0.49492 0.73969 1.05639 1.24500 1.81100 2.09750 3.11126
0.05 0.74006 0.85688 1.1 0377 1.27474 1.82064 2.10353 3.11261
0.10 0.88006 0.95699 1.15553 1.30921 1.83247 2.11099 3.11430
0.50 1.31568 1.34095 1.42895 1.51833 1.91971 2.16798 3.12769
1.00 1.56415 1.57924 1.63500 1.69626 2.01415 2.23331 3.14418
100.00 4.66378 4.66379 4.66380 4.66381 4.66388 4.66396 4.66473
492 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

Table 15-9a Values of 13 for M = 0.1 (Mode 1)

k
K 0.0 0.01 0.05 0.1 0.5 1.0 100.0
0.00 4.32649 0.69950 1.04348 1.23722 1.80856 2.09598 3.11092
0.01 0.36719 0.36719 0.36719 0.36719 0.36719 0.36719 0.36720
0.05 0.54897 0.54897 0.54897 0.54898 0.54899 0.54901 0.54907
0.10 0.65269 0.65269 0.65269 0.65270 0.65274 0.65277 0.65293
0.50 0.97417 0.97418 0.97422 0.97426 0.97454 0.97479 0.97598
1.00 1.15580 1.15582 1.15590 1.15600 1.15667 1.15725 1.16008
100.00 2.75095 2.75287 2.76043 2.76961 2.83381 2.89647 3.34313

Table 15-9b Values of 13 for M = 0.1 (Mode 2)

k
K 0.0 0.01 0.05 0.1 0.5 1.0 100.0
0.00 5.17591 4.32860 4.33685 4.34668 4.41004 4.46262 4.64755
0.01 0.49492 0.73969 1.05639 1.24500 1.81100 2.09750 3.11126
0.05 0.74006 0.85688 1.1 0377 1.27474 1.82064 2.10353 3.11261
0.10 0.88006 0.95699 1.15553 1.30921 1.83247 2.11099 3.11430
0.50 1.31568 1.34095 1.42895 1.51833 1.91971 2.16798 3.12769
1.00 1.56415 1.57924 1.63500 1.69626 2.01415 2.23331 3.14418
100.00 4.66378 4.66379 4.66380 4.66381 4.66388 4.66396 4.66473

Table 15-10a Values of 13 for M = 1 (Mode 1)

k
K 0.0 0.01 0.05 0.1 0.5 1.0 100.0
0.00 3.00108 0.69950 1.04348 1.23722 1.80856 2.09598 3.07129
0.01 0.31620 0.31620 0.31620 0.31620 0.31620 0.31620 0.31621
0.05 0.47265 0.47265 0.47265 0.47265 0.47267 0.47269 0.47277
0.10 0.56181 0.56181 0.56182 0.56182 0.56187 0.56191 0.56211
0.50 0.83693 0.83694 0.83698 0.83704 0.83738 0.83768 0.83914
1.00 0.99064 0.99066 0.99076 0.99089 0.99168 0.99239 0.99582
100.00 2.05921 2.06027 2.06443 2.06947 2.10395 2.13640 2.33266
 REFERENCES 493

Table 15-10b Values of ~ for M = 1 (Mode 2)

k
K 0.0 0.01 0.05 0.1 0.5 1.0 100.0
0.00 4.90867 3.00145 3.00290 3.00463 3.01592 3.02569 3.11092
0.01 0.49492 0.73969 1.05639 1.24500 1.81100 2.09750 3.07139
0.05 0.74006 0.85688 1.10377 1.27474 1.82064 2.10353 3.07182
0.10 0.88006 0.95699 1.15553 1.30921 1.83247 2.11099 3.07236
0.50 1.31568 1.34095 1.42895 1.51833 1.91971 2.16798 3.07667
1.00 1.56415 1.57924 1.63500 1.69626 2.01415 2.23331 3.08207
100.00 3.63194 3.63261 3.63523 3.63845 3.66173 3.68580 3.89504

15.7 ACKNOWLEDGMENTS

The author wants to thank Drs. Girvin Harkins, Albert Jeans, Cathy Keely, Larry
Moresco, and Donald Rice for their contributions. He also wants to thank Dr.
Donald Rice for his support, guidance, and encouragement.

REFERENCES

1. Wilcox, J. R., R. Subrahmanyan, and C.-Y. Li, "Thermal Stress Cycles and Ine-
lastic Deformation in Solder Joints," Proceedings of ASM 2nd Electronic Materials
and Processing Congress, 1989, pp. 203-211.
2. Lee, S-M, and D. S. Stone, "Grain Boundary Sliding in Surface Mount Solders
During Thermal Cycling," Proceedings of the IEEE 40th Electronic Components
and Technology Co'lference, 1990, pp. 491-495.
3. Stone, D., H. Wilson, R. Subrahmanyan, and C.-Y. Li, "Mechanisms of Damage
Accumulation in Solders During Thermal Fatigue," Proceedings of the IEEE 36th
Electronic Components Co'lference, 1986, pp. 630-635.
4. Stone, D., S. P. Hanula, and C.-Y. Li, "The Effects of Device and Materials
Variables on the Fatigue Behavior of Solder Joints During the Thermal Cycle,"
Proceedings of the 35th IEEE Electronic Components Conference, Washington,
DC, May 1985, pp. 46-51.
5. Stone, D. S., "Creep Fatigue Interaction in Solders and Solder Joints, ASME Winter.
Annual Meeting, Paper no. 89-WAlEEP-36, San Francisco, CA, December 1989.
6. Chen, W. T., L. C. Lee, C. K. Lim, and D. P. Seraphim, "Mechanical Modelling
for Printed Circuit Boards," Circuit World, 11 (3), 1985, pp. 68-72.
7. Seraphim, D. P., R. Lasky, and C.-Y. Li, Principles of Electronic Packaging, New
York, McGraw-Hill, 1989.
8. Ingraham, A. P., J. M. McCreary, and J. A. Varcoe, "Flip-Chip Soldering to Bare
Copper Circuits," Proceedings of the IEEE 40th Electronic Components and Tech-
nology Conference, 1990, pp. 333-337.
494 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

9. Homa, T. R., "Cracking of Electronic Packaging Joints Due to Creep-Fatigue,"

Proceedings of ASM's 3d Conference on Electronic Packaging and Corrosion in
Microelectronics, 1987, pp. 209-216.
10. Khutoretsky, M., "An Evaluation of Solder Balls on Printed Wiring Boards with
Dry Film Solder Resist," Proceedings of ASM' s 3d Conference on Electronic Pack-
aging and Corrosion in Microelectronics, 1987, pp. 225-233.
11. Luke, D. A., "Electroplating Copper for Printed Circuit Manufacture," Circuit
World, 13(1), 1986, pp. 18-23.
12. ___ , "Tin-Lead Composition-Its Control and Effect on Solderability," Circuit
World, 15(4), 1989, pp. 15-18.
13. Deane, P. A., J. S. Erich, and T. F. Marinis, "A Procedure for Measuring Lead
Planarity of Surface Mount Components," Proceedings of the IEEE 37th Electronic
Components Coriference, 1987, pp. 56-63.
14. Warrior, M., "Reliability Improvements in Solder Bump Processing for Flip Chips,"
Proceedings of the IEEE 40th Electronic Components and Technology Conference,
1990, pp. 460-469.
15. Weinhold, M., "Designing for Zero-Defect Soldering," Circuit World, 12(4), 1986,
pp.5-8.
16. Gunter, I. A., "The Solderability Testing of Surface Mount Devices Using the GEC
Meniscograph Solderability Tester," Circuit World, 13(1), 1986, pp. 8-12.
17. Steen, H. A. H., "Metallographic Examination of Soldered Joints," Circuit World,
14(3), 1988, pp. 37-40.
18. Landis, R. C., "Plastic Chip Carriers-A Question of Reliability," Proceedings of
NEPCON West, 1985, pp. 756-767.
19. Hutchins, C. L., "SMT Solder Yield and Reliability Test," Texas Instruments
Technical Papers on Surface Mount Technology, 1986, pp. 39-43.
20. Tien, J. K., B. C. Hendirx, and A. I. Attarwala, "Damage Storage Measurement
in Pb-Sn Solder," Proceedings of the IEEE 40th Electronic Components and Tech-
nology Conference, 1990, pp. 477-481.
21. Tien, J. K., B. C. Hendrix, P. L. Bretz, and A. I. Attarwala, "Creep-Fatigue
Interactions in Solders," IEEE Transactions on Components, Hybrids, and Man-
ufacturing Technology, 12(4), December 1989, pp. 502-505.
22. Hendrix, B. C., P. L. Bretz, J. K. Tien, and S. K. Kang, "Methods for the
Assessment of Damage in Cast Lead-Tin Solders Subjected to Accelerated Low
Cycling Fatigue," Proceedings of the 2nd ASM International Electronic Materials
and Processing Congress, 1989, pp. 129-133.
23. Enke, N. F., and B. I. Sandor, "A Generic Approach to Predicting the Fatigue Life
of Surface Mount Interconnections," Proceedings of the NEPCON West, 1988, pp.
683-687.
24. Kilinski, T. J., D. P. Goetsch, and B. I. Sandor, "Straddle-Board Modeling of
SMT Joint Behaviors," Proceedings of the IEEE 40th Electronic Components and
Technology Conference, 1990, pp. 981-988.
25. Enke, N. F., T. J. Kilinski, S. A. Schroeder, and J. R. Lesniak, "Mechanical
Behaviors of 60/40 Tin-Lead Solder Lap Joints," IEEE Transactions on Compo-
nents, Hybrids, and Manufacturing Technology, 12(4), December 1989, pp. 459-
468.
 REFERENCES 495

26. Williams, R. L., and A. W. Noblett, "Tailored Coefficient of Thermal Expansion

Printed Wiring Boards to Improve the Solder Joint Life of Leadless Ceramic Chip
Carriers," Circuit World, 15(2), 1989, pp. 30-32.
27. Riemer, D. E., "Prediction of Temperature Cycling Life for SMT Solder Joints on
TCE-Mismatched Substrates," Proceedings ofthe IEEE 40th Electronic Components
and Technology Conference, 1990, pp. 418-425.
28. Fine, M. E., and D. A. Jeannotte, "Soft Solder-General Principles and Fatigue
Lifetime Estimation," Proceedings ofthe 2d ASM International Electronic Materials
and Processing Congress, 1989, pp. 147-157.
29. Zubelewicz, A., R. Berriche, L. M. Keer, and M. E. Fine, "Lifetime Prediction
of Solder Materials," Journal of Electronic Packaging, Transactions of ASME,
111(3), September 1989, pp. 179-182.
30. Zubelewicz, A., Q. Guo, E. C. Cutiongco, M. E. Fine, and L. M. Keer, "Micro-
mechanical Method to Predict Fatigue Life of Solder," ASME Winter Annual Meet-
ing, Paper no. 89-WNEEP-44, San Francisco, CA, December 1989.
31. Vaynman, S., "Fatigue Life Prediction of Solder Material: Effect of Ramp Time,
Hold Time and Temperature," Proceedings of the IEEE 40th Electronic Components
and Technology Conference, 1990, pp. 505-509.
32. Vaynman, S., M. E. Fine, and D. A. Jeannotte, "Prediction of Fatigue Life of
Lead-Base Low Tin Solder," Proceedings of the 37th IEEE Electronic Components
Conference, 1987, pp. 598-603.
33. Cutiongco, E. C., S. Vaynman, and M. E. Fine, "Isothermal Fatigue of 63Sn-37Pb
Solder," ASME Winter Annual Meeting, Paper no. 89-WNEEP-37, San Francisco,
CA, December 1989.
34. Vaynman, S., "Effect of Strain Rate on Fatigue of Low-Tin Lead-Base Solder,"
IEEE Transactions on Components, Hybrids, and Manufacturing Technology, 12(4),
December 1989, pp. 469-472.
35. Berriche, R., S. Vaynman, M. E. Fine, and D. A. Jeannotte, "The Effect of
Environment on Fatigue of Low Tin Lead Base Alloys," Proceedings of ASM' s 3d
Conference on Electronic Packaging and Corrosion in Microelectronics, 1987, pp.
169-174.
36. Vaynman, S., and M. E. Fine, "Fatigue of Low-Tin Lead-Based and Tin-Lead
Eutectic Solders," Proceedings of the 2nd ASM International Electronic Materials
and Processing Congress, 1989, pp. 255-259.
37. Feldin, M., C. Zevnik, and I. Kovac, "A New Approach to Outgassing Study
of Plated Through-Hole Printed Boards," Circuit World, 15(1), 1988, pp.
22-27.
38. Ellis, J. R., and G. Y. Masada, "Dynamic Behavior of SMT Chip Capacitors
During Solder Reflow," Proceedings of the IEEE International Electronic Manu-
facturing Technology Symposium, 1989, pp. 23-29.
39. Lynch, J. T., "Surface Mount-Where Are The Standards?," Proceedings of the
IEEE 37th Electronic Components Conference, 1987, pp. 64-69.
40. Geis, H., and M. Kottke, "The Role of Surface Corrosion in Degrading Solderability
of Tin Coated Copper Leads," Proceedings of the IEEE 36th Electronic Components
Conference, 1986, pp. 636-640.
41. Liljestrand, L.-G., and L.-O. Andersson, "Accelerated Thermal Fatigue Cycling
496 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOlDER JOINTS

of Surface Mounted PWB Assemblies in Telecom Equipment," Circuit World,

14(3), 1988, pp. 69-73.
42. Turnbull, R. B., "Blowholes in PCB Soldering-Their Causes and Cures," Circuit
World, 12(2), 1986, pp. 59-61.
43. Scharr, T. A., and M. D. Nagarkar, "Outer Lead and Die Bond Reliability in High
Density TAB," Proceedings of the IEEE 39th Electronic Components Conference,
1989, pp. 177-183.
44. Iannuzelli, R., "Validation of Module Assembly Physical Models," Proceedings of
the IEEE 40th Electronic Components and Technology Conference, 1990, pp. 1071-
1076.
45. Davy, J. G., and R. Skold, "Computer-Aided Solderability Testing for Receiving
Inspection," Circuit World, 12(1), 1985, pp. 34-41.
46. Caswell, G., "Vapour Phase Soldering of SMDs-Reliability Characteristics," Circuit
World, 11(4), 1985, pp. 17-23.
47. Engel, P. A., "Torque Stress Analysis for Printed Circuit Boards Carrying Periph-
erally Leaded Modules," ASME Paper no. 86-WAMlEEP-2, December 1986.
48. , "Structural Analysis for Circuit Card Systems Subjected to Bending,"
Journal of Electronic Packaging, Transactions of ASME, 112(1), March 1990, pp.
2-10.
49. Chen, W., "Analyzing the Mechanical Strength of SMT Attached Solder Joints,"
Proceedings of the IEEE International Electronic Manufacturing Technology Sym-
posium, 1989, pp. 61-69.
50. Rafanelli, A. J., "Failure Analysis of a 132-Pin Leaded. Chip Carrier Gate Array,"
ASME Paper no. 86-WAM/EEP-6, December 1986.
51. ___ , "Solder Fillet Height Criteria for Surface Mounted Chip Components,"
Journal of Electronic Packaging, Transactions of ASME, 111(4), December 1989,
pp. 299-302.
52. Brierley, C. J., and J. P. McCarthy, "The Reliability of Surface Mounted Solder
Joints Under PWB CyJic Mechanical Stresses," Circuit World, 12(3), 1986, pp.
16-19.
53. Hong, J., and M. Chason, "Gold Removal From Component Terminations Via
Solder Dipping," Proceedings of the IEEE 39th Electronic Components Conference,
1989, pp. 351-354.
54. Chason, M., and J. Hong, "Kinetic Interactions of Copper Lead and Tin on Solder
Coated PC Boards Studied Using X-Ray Diffraction," Proceedings of the IEEE
International Electronic Manufacturing Technology Symposium, 1989, pp. 294-
299.
55. Suhir, E., "Can Power Cycling Life of Solder Joint Interconnections Be Assessed
on the Basis of Temperature Cycling Test?," Journal of Electronic Packaging,
Transactions of ASME, 111(4), December 1989, pp. 310--312.
56. _ _ , "Could Compliant External Leads Reduce the Strength of a Surface-Mounted
Device?" Proceedings of the 37th IEEE Electronic Components Conference, May
1987, pp. 589-597; also, Journal of Applied Mechanics, Transactions of ASME,
55, December 1988, pp. 818-821.
57. ___ , "Twist-Off Testing of Solder Joint Interconnections," Journal of Electronic
Packaging, Transactions of ASME, 111(3), September 1989, pp. 165-171.
58. ___ , "Die Attachment Design and Its Influence on Thermal Stresses in the Die
 REFERENCES 497

and the Attachment," Proceedings of the IEEE 37th Electronic Components Con-
ference, 1987, pp. 508-517.
59. Frost, H. J., and R. T. Howard, "Creep-Fatigue Modelling for Solder-Joint Reli-
ability Predictions Including the Microstructural Evolution of the Solder," Pro-
ceedings of the IEEE 40th Electronic Components and Technology Conference,
1990, pp. 482-490.
60. Frost, H. J., R. T. Howard, P. R. Lavery, and S. D. Lutender, "Creep and Tensile
Behavior of Lead-Rich, Lead-Tin Solder Alloys," IEEE Transactions on Compo-
nents, Hybrids, and Manufacturing Technology, 11(4), December 1988, pp. 371-
379.
61. Frost, H. J., P. R. Lavery, and S. D. Lutender, "Microstructure of Mechan-
ical Properties of Lead-Tin Solder Alloys," Proceedings of ASM's 3d Confer-
ence on Electronic Packaging and Co"osion in Microelectronics, 1987, pp.
259-267.
62. Frost, H. J., G. J. Stone, and R. T. Howard, "The Effects of Thermal History on
the Microstructure and Mechanical Properties of Solder Alloys," Proceedings of
the 2d ASM International Electronic Materials and Processing Congress, 1989,
pp. 121-127.
63. Suryanarayana, D., R. Hsiao, T. P. Gall, and J. M. McCreary, "Flip-Chip Solder
Bump Fatigue Life Enhanced by Polymer Encapsulation," Proceedings of the IEEE
40th Electronic Components and Technology Conference, 1990, pp. 338-344.
64. Panousis, N. T., J. F. Prosser, A. T. Wang., and Y. N. Yong, "Strength and
Reliability of Ceramic Modules Soldered to Flexible Cables," Proceedings of the
IEEE 40th Electronic Components and Technology Conference, 1990, pp. 377-
382.
65. Sherry, W. M., and P. M. Hall, "Materials, structures, and mechanics of solder-
joints for surface-mount microelectronics technology," in Proc. Conf. Intercon-
nection Technology in Electronics, Fellbach, W. Germany, Feb. 1986, Deutscher
Verband fUr Schweisstechnik, pp. 47--61.
66. Hall,_P. M.,. "Creep and Stress Relaxation in Solder Joints in Surface-Mount
Chip Carriers," IEEE Transactions on. Components, Hybrids, and Manufacturing
Technology, CHMT-IO(4), December 1987, pp. 556-565.
67. Tong, H. M., K. R. Mok, K. R. Grebe, H. L. Yeh, K. K. Srivastava, and J. T.
Coffin, "Parylene Encapsulation of Ceramic Packages for Liquid Nitrogen Appli-
cation," Proceedings of the IEEE 40th Electronic Components and Technology
Conference, 1990, pp. 345-350.
68. Engelmaier, W., and A. I. Attarwala, "Surface-Mount Attachment Reliability of
Clip-Leaded Ceramic Chip Carriers on FR-4 Circuit Boards," IEEE Transactions
on Components, Hybrids, and Manufacturing Technology, 12(2), June 1989, pp.
284-296.
69. Oberin, F. W., "Solder Joint Cracking in Potted High Voltage Modules," Pro-
ceedings of the 2d ASM International Electronic Materials and Processing Con-
gress, 1989, pp. 189-202.
70. Bae, K., A. F. Sprecher, Z. Guo, and H. Conrad, "Effect of Compliance on the
Fatigue of Solder Joints in Surface-Mounted Electronic Packages," Proceedings of
the 2d ASM International Electronic Materials and Processing Congress, 1989,
pp. 109-119.
498 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

71. _ _ , "Effect of Composition on the Low-Cycle Fatigue of Pb Alloy Solder

Joints," Proceedings of the IEEE 40th Electronic Components and Technology
Conference, 1990, 496-504.
72. Lauer, L. D., "Dynamic Mechanical Testing of Solder and Solder Joints," Circuit
World, 13(1), 1986, pp. 13-17.
73. Andrews, O. J., "Evaluation of the Corrosion and Reliability Aspects of Printed
Circuit Boards Flow Soldered Using 'No Clean' Fluxes," Circuit World, 16(3),
1990, pp. 11-13.
74. Mirman, B., and Knecht, S., "Creep Strains in an Elongated Bond Layer," Pro-
ceedings of the 2d IEEE I-THERM, Las Vegas, NV, 1990, pp. 21-32.
75. Darveaux, R., and I. Turlik, "Shear Defonnation of Indium Solder Joints," Pro-
ceedings of the 2d IEEE I-THERM, Las Vegas, NV, 1990, pp. 40-52.
76. Saint-Martin, X., and Joly, J., "Plastic Package Moisture Absorption Hazards in
SMT Board Assembly," Proceedings of the IEEE International Electronic Manu-
facturing Technology Symposium, 1989, pp. 38-42.
77. Hagge, J. K., and O. J. Davis, "Ageing, Solder Thickness and Solder Alloy Effects
on Circuit Board Solderability," Circuit World, 11(3), 1985, pp. 8-15.
78. DeVore, J. A., "Solderability and Surface Mounting," Circuit World, 14(4), 1990,
pp.37-41.
79. Wild, R. N., "Some Factors Affecting Leadless Chip Carrier Solder Joint Fatigue
Life," Circuit World, 14(4), 1988, pp. 29-36.
80. Smeby, J. M., "Solder Joint Behavior in HCC/PWB Interconnections," IEEE Trans-
actions on Components, Hybrids, and Manufacturing Technology, CHMT-8(3) ,
September 1985, pp. 391-396.
81. Haramake, T., T. Funamoto, and T. Shida, "Characteristics of Press-Soldered Joints
by Using Resistance Heating," IEEE Transactions on Components, Hybrids, and
Manufacturing-Technology, 13(2), June 1990, pp. 452-456.
82. Lodge, K. J., and D. J. Pedder, "The Impact of Packaging on the Reliability of
Flip Chip Solder Bonded Devices," Proceedings of the IEEE 40th Electronic Com-
ponents and Technology Conference, 1990, pp. 470-476.
83. Knecht, S., and L. R. Fox, "Constitutive Relation and Creep-Fatigue Life Model
for Eutectic Tin-Lead Solder," IEEE Transactions on Components, Hybrids, and
Manufacturing Technology, 13(2), June 1990, pp. 424-433.
84. Shine, M. C., L. R. Fox, and J. W. Sofia, "A Strain Range Partitioning Procedure
for Solder Fatigue," Brazing and Soldering, no. 9, 1985, pp. 11-20.
85. Fox, L. R., J. W. Sofia, and M. C. Shine, "Investigation of Solder Fatigue Ac-
celeration Factors," IEEE Transactions on Components, Hybrids, and Manufac-
turing Technology, CHMT-8(2), 1985, pp. 275-281.
86. Spitz, D. M., B. Cobb, D. L. Yenawine, and W. M. Wolverton, "Solderability
Optimization for Vapor Phase Reflowed, Surface Mount, Electronic Assemblies,"
Proceedings of ASM's 3d Conference on Electronic Packaging and Corrosion in
Microelectronics, 1987, pp. 239-245.
87. Molverton, M., W. R. Russell, B. Waller, and P. Barry, "Improving Control of
the Steam Age Solderability Test for Component Leads," Proceedings of the 2d
ASM International Electronic Materials and Processing Congress, 1989, pp. 37-
39.
88. Harada, M., and R. Satoh, "Mechanical Characteristics of 96.5Snl3.5Ag Solder in
 REFERENCES 499

Micro-Bonding," Proceedings of the IEEE 40th Electronic Components and Tech-

nology Conference, 1990, pp. 510-517.
89. Mitomi, 0., T. Nozawa, and K. Kawano, "Effects of Solder Creep on Optical
Component Reliability," IEEE Transactions on Components, Hybrids, and Man-
ufacturing Technology, CHMT-9(3), September 1986, pp. 265-271.
90. Pinamaneni, S., and D. A. Solomon, "Rapidly Solidified Soft Solder Die-Attach
Technology," IEEE Transactions on Components, Hybrids, and Manufacturing
Technology, CHMT-9(4), December 1986, pp. 416-422.
91. Keller, H. N., "Solder Connections with a Ni Barrier," IEEE Transactions on
Components, Hybrids, and Manufacturing Technology, CHMT-9(4) , December
1986, pp. 433-439.
92. Wassink, R. J. K., Soldering in Electronics, Electrochemical Publications, Scot-
land, 1989.
93. Lea, C., and F. H. Howie, "Blowholing in PTH Solder Fillets, Part 1: Assessment
of the Problem," Circuit World, 12(4), 1986, pp. 14-19.
94. Howie, F. H., and C. Lea, "Blowholing in PTH Solder Fillets, Part 2: The Nature,
Origin and Evolution of Gas," Circuit World, 12(4), 1986, pp. 20-25.
95. Seah, M. P., F. H. Howie, and C. Lea, "Blowholing in PTH Solder Fillets, Part
3: Moisture and the PCB," Circuit World, 12(4), 1986, p. 14.
96. Lea, C., M. P. Seah, and F. H. Howie, "Blowholing in PTH Solder Fillets, Part
4: The Plated Copper Barrel," Circuit World, 13(1), 1986, pp. 28-34.
97. Lea, C., and F. H. Howie, "Blowholing in PTH Solder Fillets, Part 5: The Role
of the Electroless Copper," Circuit World, 13(1), 1986, pp. 35-42.
98. _ _ , "Blowholing in PTH Solder Fillets, Part 6: The Laminate, The Drilling
and the Hole-Wall Preparation," Circuit World, 13(1), 1986, pp. 43-50.
99. Howie, F. H., D. Tilbrook, and C. Lea, "Blowholing in PTH Solder Fillets, Part
7: Optimising the Soldering," Circuit World, 13(2), 1986, pp. 42-45.
100. Lea, C., F. H. Howie, and M. P. Seah, "Blowholing in PTH Solder Fillets, Part
8: The Scientific Framework Leading to Recommendations for Its Elimination,"
Circuit World, 13(3), 1986, pp. 11-20.
101. Lea, C., "Residues on Soldered Printed Circuit Assemblies," Circuit World, 16(1),
1989, pp. 44-46.
102. _ _ , "The Effect of Solder Mask on PCB Solderability," Circuit World, 15(1),
1988, pp. 12-21.
103. _ _ , "Solderability and Joint Strength of Plated Component Leads," Proceedings
of DVS Conference, 1986, pp. 180-185.
104. _ _ , A Scientific Guide to Surface Mount Technology, Electrochemical Publi-
cations, Scotland, 1988.
105. Lea, C., and F. H. Howie, "Controlling the Quality of Soldering of PTH Solder
Joints," Circuit World, 11(3), 1985, pp. 5-7.
106. Pawling, 1. F., Surface Mounted Assemblies, Electrochemical Publications, Scot-
land, 1987.
107. Sinnadurai, F. N., Handbook of Microelectronics Packaging and Interconnection
Technologies, Electrochemical Publications, Scotland, 1985.
108. Chilton, A. C., M. A. Whitmore, and W. Hampshire, "Isothermal Mechanical
Fatigue of a Model SMD Joint," Proceedings of the 2dASM International Electronic
Materials and Processing Congress, 1989, pp. 159-176.
500 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

109. Manko, H. H., Soldering Handbook for Printed Circuits and Surface Mounting,
New York, Van Nostrand Reinhold, 1986.
110. Johnson, C., and J. Kevra, Solder Paste Technology: Principles and Applications,
Blue Ridge Summit, PA, Tab Books, 1989.
Ill. Kevra, J., and D. Mohoric, "Solder Paste: Tack and Cure," Proceedings of the 2d
ASM International Electronic Materials and Processing Congress, 1989, pp. 313-
319.
112. Love, G. F., "Packages for High-Density Surface-Mount Applications," Circuit
World, 13(3), 1987, pp. 36-39.
113. Ainsworth, P. A., "Measurement of the Tackiness of Solder Paste," Hybrid Circuits,
no. 11, September 1986, pp. 27-29.
114. Ciulik, J., and M. R. Notis, "Phase Equilibria and Physical Properties in the Au-
Sn System," Proceedings of the 2d ASM International Electronic Materials and
Processing Congress, 1989, pp. 57-61.
115. Tummala, R. R., and E. Rymaszewski, Microelectronics Packaging Handbook,
Van Nostrand Reinhold, New York, 1989.
116. Hwang, J. S., "Consideration for Surface Mounting Solder Paste," Brazing and
Soldering, no. 14, 1988, pp. 13-21.
117. _ _ , Solder Paste in Electronics Packaging, New York, Van Nostrand Reinhold,
1989.
118. Hwang, J. S., and R. M. Vargas, "Solder Joint Reliability-Can Solder Creep?,"
Proceedings of the International Symposium on Microelectronics, 1989, pp. 513-
524.
119. Verguld, M. M. F., "Wave Soldering. The Efficiency of Debridging with a Hot-
Air Knife," Circuit World, 13(2), 1987, pp. 14-18.
120. Kojima, Y., M. Kikuchi, K. Matsunaga, and N. Yamazaki, "The Reliability of
Reftow Soldering by Hot Air Reftow," Proceedings of the IEEE International
Electronic Manufacturing Technology Symposium, 1989, pp. 43-47.
121. Yamada, S. E., 'The Inverted T-Joint for Surface Mount Connectors," IEEE Trans-
actions on Components, Hybrids, and Manufacturing Technology, 13(12), June
1990, pp. 397-400.
122. _ _ , "A Fracture Mechanics Approach to Soldered Joint Cracking," IEEE Trans-
actions on Components, Hybrids, and Manufacturing Technology, 12(no. 1) Mar.
1989, pp. 99-104.
123. _ _ , "Stress Analysis of Partially Yielded Soldered Joint for Surface Mount
Connectors," IEEE Transactions on Components, Hybrids, and Manufacturing
Technology, CHMT-I0(2), June 1987, pp. 236-241.
124. _ _ , "Time Dependent Crack Growth in Soldered Joint," Engineering Fracture
Mechanics, 28(1), 1987, pp. 85-91.
125. _ _ , "The J-Integral for Augmented Double Cantilever Beams and Its Appli-
cation to Bonded Joint," Engineering Fracture Mechanics, 29(6), 1988, pp. 673-
682.
126. _ _ , "A Mechanism for Board Warpage by Thermal Expansion of Surface
Mounted Connector," IEEE Transactions on Components, Hybrids, and Manufac-
turing Technology, CHMT-9(4), December 1986, pp. 508-512.
127. Schoenbaum, G. L., and E. Socha, "Selecting the Accept/Reject Criteria for an
 REFERENCES 501

Automated Solder Process Inspection Machine," Circuit World, 13(2), 1987, pp.
19-21.
128. Val, C., J. Rakotonarivo, and A. Gerard, "Plasma Inspection of Visible or Hidden
Solder Joints," Proceedings of the International Symposium on Microelectronics,
1989, pp. 525-536.
129. ___ , "Automatic Inspectiron of Visible or Hidden Solder Joints-A New Con-
cept," Proceedings of the IEEE 39th Electronic Components Conference, 1989, pp.
623-628.
130. O'Clock, G. D., M. S. Peters, 1. R. Pater, G. A. Kleese, and R. V. Martini, "Pb-
Sn Alloy Microstructure: Potential Reliability Indicator for Interconnects," IEEE
Transactions on Components, Hybrids, and Manufacturing Technology, CHMT·
10(1), March 1987, pp. 82-88.
131. ___ , "The Use of Microstructure as a Reliability Indicator for Solder Intercon-
nects," Proceedings of ASM's 3d Conference on Electronic Packaging and Cor-
rosion in Microelectronics, 1987, pp. 251-256.
132. Mizuishi, K., and T. Mori, "Thermal Stability of Various Ball-Lirnited-Metal Sys-
tems Under Solder Bumps," IEEE Transactions on Components, Hybrids, and
Manufacturing Technology, 11(4), December 1988, pp. 481-484.
133. Chiou, B.-S., K. C. Liu, I.-G. Duh, and P. S. Palanisamy, "Intermetallic Formation
on the Fracture of SnlPb Solder and PdJAg Conductor Interfaces," IEEE Trans-
actions on Components, Hybrids, and Manufacturing Technology, 13(2), June
1990, pp. 267-274.
134. Driels, M. R., and D. J. Nolan, "Automatic Defect Classification of Printed Wiring
Board Solder Joints," IEEE Transactions on Components, Hybrids, and Manufac-
turing Technology, 13(2), June 1990, pp. 331-340.
135. Wong, B., D. E. Helling, and R. W. Clark, "A Creep-Rupture Model for Two-
Phase Eutectic Solders," IEEE Transactions on Components, Hybrids, and Man-
ufacturing Technology, 11(3), September 1988, pp. 284-290.
136. Inaba, M., K. Yamakawa, and N. Iwase, "Solder Bump Formation Using Electroless
Plating and Ultrasonic Soldering," IEEE Transactions on Components, Hybrids,
and Manufacturing Technology, 13(1), March 1990, pp. 119-123.
137. Matsui, N., S. Sasaki, and T. Ohsaki, "VLSI Chip Interconnection Technology
Using Attached Solder Bumps," IEEE Transactions on Components, Hybrids, and
Manufacturing Technology, CHMT·I0(4), December 1987, pp. 566-570.
138. Kousaka, M., H. Yanai, T. Kobayashi, and R. K. Selwood, "A Unique Material
Solution of Fine Line Printing of Solder Paste," Proceedings of the International
Symposium on Microelectronics, 1989, pp. 537-543.
139. Lin, K. M., and F. H. Friend, "Wetting Balance Methods for Component and
Board Solderability Evaluation," Circuit World, 16(1), 1989, pp. 24-28.
140. Puttlitz, K. J., "Preparation, Structure, and Fracture Nodes of Pb-Sn and Pb-In
Terminated Flip Chips Attached to Gold-Capped Microsockets," Proceedings of
the IEEE 40th Electronic Components and Technology Conference, 1990, pp. 360-
367.
141. ___ , "Corrosion of Pb-501n Flip-Chip Interconnections Exposed to Harsh En-
vironment," IEEE Transactions on Components, Hybrids, and Manufacturing Tech-
nology, 13(1), March 1990, pp. 188-193.
502 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

142. Forshaw, M., "Application of Non-Destructive Testing to Inspection of Soldered

Joints," Circuit World, 15(3), 1989, pp. 14-17.
143. Agarwala, B. N., "Thennal Fatigue Damage in Pb-In Solder Interconnection,"
Proceedings of the 23d IEEE International Reliability Physics Symposium, 1985,
pp. 198-205.
144. Marcotte, V. C., C. Vincent, N. G. Koopman, and P. A. Totta, "Review of Flip
Chip Bonding," Proceedings of the 2d ASM International Electronic Materials and
Processing Congress, 1989, pp. 73-81.
145. Fujiwara, K., M. Asahi, S. Tsurumi, and Y. Takeuchi, "Water-Soluble Flux for
Pb-Alloy Josephson Device Packaging," IEEE Transactions on Components, Hy-
brids, and Manufacturing Technology, CHMT-I0(2), June 1987, pp. 258-262.
146. Fujiwara, K., and M. Asahi, "Characterization of Intennetallic Compound For-
mation on InIBilSn Solder Bumps Used in Pb-Alloy Josephson Chip Packaging,"
IEEE Transactions on Components, Hybrids, andManufacturing Technology, CHMT-
10(2), June 1987, pp. 263-266.
147. Evans, J. W., and J. K. Beddow, "Characterization of Particle Morphology and
Rheological Behavior in Solder Paste," IEEE Transactions on Components, Hy-
brids, and Manufacturing Technology, CHMT-I0(2), June 1987, pp. 224-231.
148. Bowen, R. C., and D. M. Peterson, "A Comparison of Rapid Solidification Cast
versus Conventional Die Attach Soft Solders," IEEE Transactions on Components,
Hybrids, and Manufacturing Technology, CHMT-I0(3), September 1987, pp. 341-
345.
149. Balde, J. W., "The IEEE Compliant Lead Task Force," IEEE Transactions on
Components, Hybrids, and Manufacturing Technology, CHMT-I0(3), September
1987, pp. 463-469.
150. Matsui, A., K. Kawai, and E. Asada, "Experimental Research on Fatigue and
Fracture of Solder," Proceedings of the International Symposium on Microelec-
tronics, 1989, pp. 544-551.
151. Tervalon, M. J., "Local Mismatched in SM Solder Joint FE Analysis," Proceedings
of the IEEE 40th Electronic Components and Technology Conference, 1990, pp.
954-957.
152. Clech, J.-P., F. M. Langennan, and J. A. Augis, "Local CTE Mismatch in SM
Leaded Packages: A Potential Reliability Concern," Proceedings of the IEEE 40th
Electronic Components and Technology Conference, 1990, pp. 368-376.
153. Clech, J.-P., and J. A. Augis, "Engineering Analysis of Thennal Cycling Accel-
erated Tests for Surface-Mount Attachment Reliability Evaluation," Proceedings of
the 7th Annual International Electronics Packaging Society Conference, Boston,
MA, November 1987, pp. 385-410.
154. Clech, J.-P., W. Engelmaier, R. W. Kotlowitz, and J. A. Augis, "Surface Mount
Solder Attachment Reliability Figures of Merit (Design for Reliability) Tools,"
Proceedings of NEPCON West, March 1989, pp. 206-224.
155. _ _ , "Reliability Figures of Merit for Surface-Soldered Leadless Chip Carriers
Compared to Leaded Packages," IEEE Transactions on Components, Hybrids, and
Manufacturing Technology, 12(4), December 1989, pp. 449-458.
156. Frear, D. R., "Microstructural Evolution During Thennomechanical Fatigue of
62Sn-36Pb-2Ag and 60Sn-40Pb Solder Joints," Proceedings of the IEEE 40th Elec-
tronic Components and Technology Conference, 1990, pp. 518-524.
 REFERENCES 503

157. ___ , "Thermomechanical Fatigue of Solder Joints: A New Comprehensive Test

Method," IEEE Transactions on Components, Hybrids, and Manufacturing Tech-
nology, 12(4), December 1989, pp. 492-501.
158. Frear, D., D. Grivas, M. McCormack, D. Tribula, and J. W. Morris, "Fatigue and
Thermal Fatigue Testing of Pb-Sn Solder Joints," Proceedings of ASM' s 3d Con·
ference on Electronic Packaging and Corrosion in Microelectronics, 1987, pp.
269-274.
159. Tribula, D., and J. W. Morris, Jr., "Creep in Shear of Experimental Solder Joints,"
ASME Winter Annual Meeting, Paper no. 89-WAlEEP-30, San Francisco, CA,
December 1989.
160. Summers, T. S. E., and J. W. Morris, Jr., "Isothermal Fatigue Behavior of Sn-Pb
Solder Joints," ASME Winter Annual Meeting, Paper no. 89-W AlEEP-39, San
Francisco, CA, December 1989.
161. Tribula, D., D. Grivas, D. R. Frear, and J. W. Morris, "Observations on the
Mechanisms of Fatigue in Eutectic Pb-Sn Solder Joints," Journal of Electronic
Packaging, Transactions of ASME, 111(2), June 1989, pp. 83-89.
162. Royce, B. S. H., "Differential Thermal Expansion in Microelectronic Systems,"
IEEE Transactions on Components, Hybrids, and Manufacturing Technology, 11(4),
December 1988, pp. 454-463.
163. Schmidt, C. G., "A Phenomenological Constitutive Model for Type Sn62 Solder
Wire," Proceedings of the IEEE 39th Electronic Components Conference, 1989,
pp. 253-258.
164. MacKay, C. A., "Some Causes of Problems with Existing SMT Solder Creams
and Possible Improvements," Circuit World, 13(3), 1987, pp. 4-7.
165. ___ , "Flue Mechanisms From Solderability Measurements," Proceedings of the
2d ASM International Electronic Materials and Processing Congress, 1989, pp.
13-16.
166. Sherman, K., and C. A. MacKay, "A Thermoanalytical Study of the Components
and Formulation of a Resin Based Flux," Proceedings of the 2d ASM International
Electronic Materials and Processing Congress, 1989, pp. 17-30.
167. Roche, N. G., and C. A. MacKay, "An Activity Number for Fluxes," Proceedings
of the 2d ASM International Electronic Materials and Processing Congress, 1989,
pp. 31-35.
168. MacKay, C. A., "Solder Creams for SMT," Proceedings of the 2d ASM Interna-
tional Electronic Materials and Processing Congress, 1989, pp. 405-417.
169. Charles, H. K., Jr., and G. V. Clatterbaugh, "Solder Joint Reliability-Design Im-
plications from Finite Element Modeling and Experimental Testing," ASME Winter
Annual Meeting, Paper no. 89-WAlEEP-22, San Francisco, CA, December 1989.
170. Pan, T. Y., and Y. H. Pao, "Stress Analysis of Multilayer Solder Assemblies,"
ASME Winter Annual Meeting, Paper no. 89-WAlEEP-3, San Francisco, CA, De-
cember 1989.
171. Shah, M., "Analysis of Parameters Influencing Stresses in the Solder Joints of
Leadless Chip Capacitors," ASME Winter Annual Meeting, Paper no. 89-WAlEEP-
31, San Francisco, CA, December 1989.
172. Solomon, H. D., V. Brzozowski, and D. G. Thompson, "Predictions of Solder
Joint Fatigue Life," Proceedings of the IEEE 40th Electronic Components and
Technology Conference, 1990, pp. 351-359.
504 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOLDER JOINTS

173. Solomon, H. D., "Fatigue of 60/40 Solder," IEEE Transactions on Components,

Hybrids, and Manufacturing Technology, CHMT·9(4), December 1986, pp. 423-
432.
174. ___ , "Strain-Life Behavior in 60/40 Solder," Journal of Electronic Packaging,
Transactions of ASME, 111, June 1989, pp. 75-82.
175. _ _ , "Room Temperature Low Cycle Fatigue of a High Pb Solder (Indalloy
151)," Proceedings of the 2d ASM International Electronic Materials and Pro-
cessing Congress, April 1989, pp. 135-146.
176. ___ , "Low Cycle Fatigue of Surface Mounted Chip Carrier/Printed Wiring Board
Joints," IEEE Transactions on Components, Hybrids, and Manufacturing Tech-
nology, 12(4), December 1989, pp. 473-479.
177. _ _ , "High and Low Temperature Strain-Life Behavior of a Pb Rich Solder,"
ASME Winter Annual Meeting, Paper no. 89-WAlEEP-33, San Francisco, CA,
December 1989.
178. _ _ , "The Influence of Hold Time and Fatigue Cycle Wave Shape on the Low
Cycle Fatigue of 60/40," Proceedings of the IEEE 38th Electronic Components
Conference, 1988, pp. 7-12.
179. Huffman, B., R. Brownfield, and B. Natarajan, "Mechanical Characterization of
J-Bend Leads and Solder Joints for Surface Mounted PLeC's," Proceedings of
IEEE International Electronic Manufacturing Technology Symposium, 1986, pp.
218-227.
180. Gee, S. A., and C. G. M. Van Kessel, "Stiffness and Yielding in the PLCC 'J'
Lead," IEEE Transactions on Components, Hybrids, and Manufacturing Technol-
ogy, CHMT·I0(3), September 1987, pp. 379-390.
181. Jahsman, W. E., and P. Jain, "Comparison of Predicted and Measured Lead Stiff-
nesses of Surface Mounted Packages," Proceedings ofIEEE Electronic Components
and Technology Conference, May 1990, pp. 926-932.
182. Pope, D. E., and K. Byrd, "Experimental Measurements of Surface Mount Com-
ponent Lead Stiffness," Proceedings of IEEE Electronic Components and Tech-
nology Conference, May 1990, pp. 934-941.
183. Alger, C. L., D. E. Pope, and P. M. Rehm, "Solderability Test Requirements for
Plastic Surface Mount Packages," Circuit World, 1989, pp. 30-37.
184. Kotlowitz, R. W., "Compliance Metrics for Surface Mount Component Lead De-
sign," Proceedings of IEEE Electronic Components and Technology Conference,
May 1990, pp. 926-932.
185. Lau, J. H., and D. W. Rice, "Solder Joint Fatigue in Surface Mount Technology:
State of the Art," Solid State Technology, 28, October 1985, pp. 91-104.
186. Lau, J. H., D. W. Rice, and P. A. Avery, "Elasto-Plastic Analysis of Surface-
Mount Solder Joints," IEEE Transactions on Components, Hybrids, and Manufac-
turing Technology, CHMT·I0(3), September 1987, pp. 346-357.
187. Lau, J. H., D. W. Rice, andG. Harkins, "Thermal Stress Analysis of TAB Packages
and Interconnections," IEEE Transactions on Components, Hybrids, and Manu-
faciuring Technology, 13(1), March 1990, pp. 183-188.
188. Lau, J. H., and D. W. Rice, "Thermal Stress/Strain Analyses of Ceramic Quad
Flat Pack Packages and Interconnections," Proceedings of the IEEE 40th Electronic
Components and Technology Conference, May 1990, pp. 824-834.
 REFERENCES 505

189. Lau, J. H., and G. Harkins, "Stiffness of PQFP 'Gull-Wing' Lead and Its Effect
on Solder Joint Reliability," Proceedings of the 5th IEEE International Electronic
Manufacturing Technology Symposium, October 1988, pp. 131-142.
190. _ _ , "Stiffness of 'Gull-Wing' Lead and Solder Joint for Plastic Quad Flat
Pack," IEEE Transactions on Components, Hybrids, and Manufacturing Technol-
ogy, 13(1), March 1990, pp. 124-130.
191. Lau, J. H., "Stiffness of PLCC 'J' Lead and Its Effect on Solder Joint Reliability ,"
ASME Winter Annual Meeting, Paper no. EEP-3, Boston, December 13-15,1987,
192. ___ , "Thermal Stress Analysis of SMT PQFP Packages and Interconnections,"
Journal of Electronic Packaging, Transactions of ASME, 111, March 1989, pp.
2-8.
193. Lau, J. H., and G. Harkins, "Thermal Stress Analysis of SOIC Packages and
Interconnections," IEEE Transactions on Components, Hybrids, and Manufacturing
Technology, CHMT-ll(4), December 1988, pp. 380-389.
194. Lau, J. H., G. Harkins, D. Rice, J. Kral, and B. Wells, "Experimental and Statistical
Analyses of Surface-Mount Technology PLCC Solder-Joint Reliability," IEEE
Transactions on Reliability, 37(5), December 1988, pp. 524-530.
195. Lau, J. H., and G. H. Barrett, "Stress and Deflection Analysis of Partially Routed
Panels for Depanelization," IEEE Transactions on Components, Hybrids, andMan-
ufacturing Technology, CHMT-I0(3), September 1987, pp. 411-419.
196. Lau, J. H., S. J. Erasmus, and D. W. Rice, "Overview of Tape Automated Bonding
Technology," Circuit World, 16(2), 1990. pp. 5-24.
197. Lau, J. H., and A. H. Jeans, "Effects of Voids on SMT Solder Joint Reliability,"
Proceedings of the 2d ASM International Electronic Materials and Processing
Congress, April 1989, pp. 177-187.
198. Lau, J. H., R. Subrahrnanyan, D. Rice, S. Erasmus, and C. Li, "Fatigue Analysis
of a Ceramic Pin Grid Array Soldered to an Orthotropic Epoxy Substrate," ASME
Winter Annual Meeting Paper no. 9O-WNEEP-47, Dallas, TX, November 1990.
199. Lau, J. H., L. Powers, J. Baker, D. Rice, and W. Shaw, "Solder Joint Reliability
of Fine Pitch Surface Mount Technology Assemblies," IEEE Transactions on Com-
ponents, Hybrids, and Manufacturing Technology, 13(3), September 1990, pp.
534-544.
200. Lau, J. H., and C. A. Keely, "Dynamic Characterization of Surface Mount Com-
ponent Leads for Solder Joint Inspection," IEEE Transactions on Components,
Hybrids, and Manufacturing Technology, 12(4), December 1989, pp. 594-602.
201. Lau, J. H., G. Harkins, D. Rice, and J. Kral, "Thermal Fatigue Reliability of SMT
Packages and Interconnections," Proceedings of IEEE 25th International Reliability
Physics Symposium, San Diego, April 1987, pp. 250-259.
202. Lau, J. H., and D. Rice, "Failure Analysis of Surface Mounted Interconnection,"
Proceedings of the 12th ASM International Symposium for Testing and Failure
Analysis, Los Angeles, October 1986, pp. 73-82.
203. Lau, J. H., "A Note on the Calculation of Thermal Stresses in Electronic Packaging
by Finite Element Methods," Journal of Electronic Packaging, Transactions of the
ASME, 111, December 1989, pp. 313-320.
204. ___ , Elasto-Plastic Finite Element Analysis with Special Reference to Reiriforced
Concrete, M.A.Sc. Thesis, University of British Columbia, 1973.
506 ANALYSES OF SURFACE MOUNT COMPONENT LEADS AND SOlDER JOINTS

205. Lau, J. H., and A. H. Jeans, "Nonlinear Analysis ofElastomeric Keyboard Domes,"
Journal of Applied Mechanics, Transactions of the ASME, 56, December 1989,
pp. 751-755.
206. Lau, J. H., and L. L. Moresco, "Mechanical Behavior of Microstrip Structures
Made from YBaCuO Superconducting Ceramics," IEEE Transactions on Compo-
nents, Hybrids, and Manufacturing Technology, 11(4), December 1988, pp. 419-
426.
207. Lau, J. H., and L. B. Lian-Mueller, "Finite Element Modeling for Optimizing
Hermetic Package Reliability," Journal of Electronic Packaging, Transactions of
ASME, 111, December 1989, pp. 255-260.
208. Lau, J. H., "Thermal Stress Analysis of Plastic Leaded Chip Carriers," Proceedings
of the IEEEIASMEIlSHMINIST InterSociety Conference on Thermal Phenomena in
Electronic Systems, May 1990, pp. 57-66.
209. Zienkiewicz, O. C., The Finite Element Method in Engineering Science, London,
McGraw-Hili, 1971.
210. Burnett, D. S., Finite Element Analysis, Reading, MA, Addison-Wesley Publishing
Company, 1987.
211. Strang, G., An Analysis of the Finite Element Method, Englewood Cliffs, NJ,
Prentice-Hall, 1973.
212. Huston, R. L., and C. E. Passerello, Finite Element Methods, New York, Marcel
Dekker, 1984.
213. Bathe, K. J., Finite Element Procedures in Engineering Analysis, Englewood Cliffs,
NJ, Prentice-Hall, 1982.
214. Eisley, J. G., Mechanics of Elastic Structures, Englewood Cliffs, NJ, Prentice-
Hall, 1989.
215. Humar, J. L., Dynamics of Structures, Englewood Cliffs, NJ, Prentice-Hall, 1990.
216. Heins, C. P., and K. N. Derucher, Structural Analysis and Design, New York,
Marcel Dekker, 1980.
217. Ross, C. T. F., Computational Methods in Structural and Continuum Mechanics,
New York, John Wiley, 1982.
218. McGuire, W., and R. H. Gallagher, Matrix Structural Analysis, New York, John
Wiley, 1979.
219. Meek, J. L., Matrix Structural Analysis, New York, McGraw-Hili, 1971.
220. Fung, Y. C., Foundations of Solid Mechanics, Englewood Cliffs, NJ, Prentice-
Hall, 1965.
221. Timoshenko, S. P., and J. N. Goodier, Theorj'of Elasticity, New York, McGraw-
Hill, 1970.
222. Harris, C. M., and C. E. Crede, Shock and Vibration Handbook, New York,
McGraw-Hili, 1976.
223. Thomson, W. T., Theory of Vibrations with Applications, Englewood Cliffs, NJ,
Prentice-Hall, 1972.
224. Meirovitch, L., Analytical Methods in Vibrations, London, Macmillan, 1967.
225. Clough, R. W., and J. Penzien, Dynamics of Structures, New York, McGraw-Hili,
1976.
226. Langhaar, H. L., Energy Methods in Applied Mechanics, New York, John Wiley,
1962.
 REFERENCES 507

227. Steinberg, D. S., Vibration Analysis for Electronic Equipment, New York, John
Wiley, 1973.
228. Steinberg, D. S., "Protecting PCBs From Shock," Machine Design, March 1983,
pp. 189-192.
229. Markstein, H. W., "Designing Electronics for High Vibration and Shock," Elec-
tronic Packaging and Production, April 1987, pp. 40-43.
230. Markstein, H. W., "Design With Vibration in Electronics," Electronic Packaging
and Production, June 1989, pp. 30-34.
231. Lau, J. H., "Vibration Frequencies of Tapered Bars With End Mass," Journal of
Applied Mechanics, Transactions of ASME, 51, March 1984, pp. 179-181.
232. _ _ , "Vibration Frequencies and Mode Shape for a Constrained Cantilever,"
Journal of Applied Mechanics, Transactions of ASME, 51, March 1984, pp. 182-
187.
233. _ _ , "Vibration Frequencies for a Non-Uniform Beam with End Mass," Journal
of Sound and Vibration, 97, December 1984, pp. 513-521.
234. Lau, J. H., and R. E. Miller, "Vibration Analysis of Orthotropic Shells of Revo-
lution," Proceedings of AlAAIASMEIASCE 22d Structures, Structural Dynamics
and Materials Conferences, 1981, pp. 122-131.
 16
Integrated Matrix Creep:
Application to Accelerated Testing
and Lifetime Prediction
Sheera Knecht and Les Fox

One of the major goals of mechanical analysis in the electronics industry is to

be able to predict, before extensive testing, the reliability of printed circuit boards
with surface mounted components under thermal or mechanical cyclic loading.
Analysis cannot be entirely substituted for testing, but if it can be used as a
guide to testing, in determining the minimum tests necessary and in understanding
the results of these tests, the cost of ensuring a reliable product will be sub-
stantially reduced. It is not surprising, therefore, that a great deal of effort is
currently being applied to mechanical modelingl of electronic packages with
surface mounted devices with the aim of understanding and improving fatigue
life of solder joints.
Practical fatigue life predictions are based on the calculation of certain failure
indicators that correlate known lifetimes under one set of conditions (e.g., ge-
ometry, loading, materials) with projected lifetimes under another set of con-
ditions. Possible failure indicators might be maximum stress, plastic strain, or
some kind of "damage integral." To calculate any of these indicators, a consti-
tutive or stress-strain relation is needed for the materials involved in the analysis.
In the case of surface mounted components, a constitutive law for solder is a
critical part of mechanical modeling.
Solder material behavior is complex. Eutectic 63% Sn-37% Pb solder (the
most commonly used solder) melts at 183°C. The ratio TITmello where the tem-
perature is expressed in degrees absolute, is known as the homologous temper-
ature; therefore the homologous temperature of solder at room temperature is

508
 INTEGRATED MATRIX CREEP 509

about 0.65. At homologous temperatures greater than about 0.5 it is well known
that metals exhibit significant creep and stress relaxation. Such behavior is easily
observed for solder. However, time dependent creep effects do not account for
all of the observed inelastic solder behavior; for the rapid cycling used in ac-
celerated fatigue tests, the noncreep plasticity can be significant. In order for
mechanical modeling to provide a link between test results and real world ap-
plications, the solder constitutive equation needs to consider all important aspects
of material behavior. For example, if short time plasticity is ignored in favor of
long-term creep, results of very fast accelerated fatigue tests may be misleading
when applied to the long-term stress relaxation experienced under service life
conditions.
Although a great deal of published work on mechanical properties of solder
is available, this broad range of solder behavior is usually not presented in a
form that is convenient for mechanical analysis or for fatigue life calculations.
Moreover, existing solder constitutive equations have often not been used to
predict experimental results other than those upon which they are based. When
it can be seen that it is possible to predict results of testing under different
conditions from those used to derive an equation, we can have more confidence
that the equation is useful for mechanical modeling.
The first purpose of this study, therefore, is to derive a stress-strain relation
for eutectic or near-eutectic Sn-Pb solder that is simple enough for practical use
and yet is complex enough to represent the range of solder behavior that is
important for electronic applications. The development of the equation is based
on the data described in reference 2 for solder joints in both steady-state creep
and constant strain rate cycling. Some parameters are difficult to extract from
this data and these are taken from other published results. The equation is then
shown to predict observed data, also from reference 2, for stress relaxation tests.
The equation is developed from a continuum mechanics and not a material
science point of view; the purpose is to determine a relation that is useful for
predicting the mechanical behavior of solder joints used in different structures
(e.g., leaded and leadless chip carriers) and under different loadings. Since the
subject of solder metallurgy is complex and not well understood, no attempt has
been made to include microstructural effects in the equation. Although micro-
structure can affect mechanical properties, it is not possible to quantify micros-
tructural effects at the present time. In accordance with our approach, the solder
joints used to generate the data used here are typical joints that would be found
in electronic devices. Their microstructure is not specially controlled and is
therefore difficult to quantify; however, we feel that using "real" joints will
generate practical results. Whenever possible, we are guided in our analysis by
knowledge of solder metallurgy, but the general approach is a continuum me-
chanics one.
An adequate constitutive law allows the calculation of stresses and strains for
arbitrary geometries and loading histories. The second part of any reliability
510 INTEGRATED MATRIX CREEP

analysis is to relate these calculated parameters or "failure indicators" to expected

lifetime. The matrix creep criterion developed in reference 2 and explained below
is proposed as a failure indicator that is simple enough to be calculated relatively
easily, yet complex enough to take into account the time dependent nature of
solder fatigue. It has already been shown in reference 2 that the matrix creep
failure criterion proposed there gives good correlation with the fatigue life of
the solder joints of reference 2. This correlation was achieved using a steady-
state creep law with different, experimentally determined, coefficients for each
sample considered; the matrix creep was then integrated over the actual measured
stress history (see reference 2). It is often necessary, however, to make fatigue
life predictions before any testing can be done; therefore the second purpose of
this study is to show that the constitutive equation developed here can be used
along with the matrix creep failure indicator to predict fatigue life adequately
for the solder joints of reference 2. The equation is also applied to fatigue tests
described by other authors, and reported lifetimes are shown to correlate with
matrix creep.
Finally, the fatigue failure indicator is itself extended to the case of leaded
solder joints where stresses are multiaxial and vary with position in the joint.
Application of the extended matrix creep failure indicator to a thermal cycle
fatigue example from the literature shows the current status of the method.

16.1 GENERAL FORM OF THE CONSTITUTIVE RELATION

In order to be able to predict the loading and deformation histories encountered

in electronic applications, the constitutive relation for eutectic or near-eutectic
Sn-Pb solder is developed considering three modes of solder deformation. The
total shear strain rate 'Y is assumed to be the sum of time independent elastic,
time independent plastic, and time dependent secondary creep strain rates:

(16-1)

It is difficult to separate fast primary creep from time independent plasticity;

therefore 'Ypl may include such effects. It is also possible to include a time
dependent (transient), slower primary creep strain rate term, but for simplicity
this is not done. Similarly, any anelastic (time dependent elastic) effects are also
ignored.
The data used to develop the constitutive law consists of stable cycles of fully
reversed loading; the initial loading cycles of the test (typically fewer than five)
are not considered. As cycling is continued beyond these first few hysteresis
loops, the stress-strain curve stabilizes; the number of stable cycles depends on
the details of the loading. As cycling continues further, the load needed to
maintain a fixed opening of the hysteresis loop decreases (probably due to fatigue
 DEVELOPMENT OF THE CONSTITUTIVE RELATION 511

cracking) and therefore the relationship between stress and strain changes; these
changes are not considered either. It is felt that the most useful information for
lifetime prediction can be obtained in the simplest manner from an understanding
of the initial stable part of the deformation history.

16.2 DEVELOPMENT OF THE CONSTITUTIVE RELATION

16.2.1 Description of Data

The data used in this study are the same as used for the work of reference 2. A
torsional apparatus was employed to obtain isothermal hysteresis loops and creep
data on a 16-1/0 solder joint array. The 7.6- by 7.6-mm chip carrier had solder
pads approximately 0.64 by 1.0 mm. The 16 pads were arranged in a square
array of four per side at an average distance from the center of 3.43 mm, which
was used to calculate the average stress on the pads due to the applied torque.
These chip carriers were surface mounted to a stainless steel plate 45 by 45 mm.
The stainless steel plate had electroplated copper footprints patterned to those
of the chip carrier. The assembly was soldered together by 63% Pb-37% Sn
solder paste; The joint thicknesses ranged·from 0.102 to 0.305 mm (0.004 to
O. 12 in.). Additional data was obtained in the same manner, but using G 10
substrate material instead of stainless steel. Further details of the experimental
procedure can be obtained in reference. 2

16.2.2 Steady-State Creep Strain Component

When the rate of shear stress 0- is zero, 'Vel and 'VPI are also zero and therefore,
according to Equation (16-1), the time dependent part of the strain rate is com-
pletely described by the steady-state creep strain rate determined from constant
stress tests. Details of the steady state creep tests are given in reference 2.
Data used here were taken from samples A, AI, A2, A3, and C (stainless
steel substrates) of reference 2. The creep rates were obtained by reanalyzing
the raw data, so some variation from reference 2 can be noted. Shear stress
versus shear strain rate at 25, 60, and 100°C are shown in Figure 16-1. Two
points can be noted:

• Several authors3.4,8 have noted a grain size dependence in steady-state creep

data for eutectic or near-eutectic solder. The joints used here are typical
solder joints that would be found in actual devices; therefore, the grain size
has not been carefully controlled and varies within each joint. Still, the
thinner joints tend to have smaller grain sizes than the thicker ones, although
the amount of this difference cannot be quantified (see reference 2). Ob-
 512 INTEGRATED MATRIX CREEP

10000

~ .. ,

.. ' ...

.....
....
~ 1000 M"I
J • ...•.
t-

...-

••...
.......•

Ad Sample A-RoolD Temp (Tbickne•• =.OO5") 00 Sample A3-6~·C

77 Sample AI-Room Temp (Thicknes.=.OO58") x x Sample C-60 C
CD Sample A2-Room Temp (Thickne•• =.OOS") _ SampleA3-IOO·C
_ S=ple A3-Room Temp (Thlckness •. OO83) " ' Sample C-IOO·C
XX Sample C-Room Temp (Thickne•••. OO6.. ) _ Eqn. (16-2) - room temp
00 Sample A-60-oC __ . Eqn. (16-2) 60'C
++ SlIlDple AI-60·C _••. _ Eqn. (16-2) 100'C
** Sample A2-60·C

Figure 16-1 Steady-state creep data from (2).

servation of Figure 16-1 shows that thickness effects are not evident
here .
• Most recent experimental work on solder,4,8-1O shows two (or possibly three)
regimes of secondary (steady-state) creep with distinctly different slopes in
the stress-strain rate curve. The range of the data from reference 2 is not
really large enough to confirm this hypothesis absolutely, but the data in
Figure 16-1 do suggest it.

In light of these points, the following approach is taken: Two regimes of steady-
state creep are assumed. For convenience, these are called "grain boundary" and
"matrix" creep as in reference 2. However, in keeping with the continuum
mechanics (rather than metallurgical) approach taken here we do not need to
determine any particular creep mechanisms for these regions. We simply assume,
 DEVELOPMENT OF THE CONSTITUTIVE RELATION 513

Table 16-1 Comparison of Creep Parameters from Literature

JlH GB
Author (eV) nGB JlH MC (eV) nMC Ref. No.
Grivas et al. 0.5 2.0 0.84 7.1 3
Kashyap et al. 0.46 1.7 1.0 11.1 4
Solomon 3.1 7.3 5
Baker 0.69 3.4 7
Lam et al. 0.45 2.4 0.9 5.7 8
Wong et al. 0.47 3.0 0.47 7.0 9

on the basis of empirical data for creep in metals, that each regime has a power
law dependence of strain rate on stress and each follows an Arrhenius model for
temperature dependence with a different activation energy.
The form of the equation for steady-state creep rate as a function of stress
and temperature is

. = 'Yer
'Y . = c1 exp (-!:JI
kT GB ) T"GB + C2 exp (-!:JI
kT MC ) T"MC (16-2)

As can be seen from Figure 16-1, the data are concentrated in the region where
both steady-state creep terms contribute significantly to the stress; it is difficult,
therefore, to isolate the two regimes and to find values of the two exponents and
two activation energies. Instead these parameters are taken from published data.
Representative values of the activation energies (!:JIGB and !:JIMd and ex-
ponents (nGB and nMd found in the literature are shown in Table 16-1. In this
study, values of !:JIGB , !:JIMC , naB, and nMC are taken from Grivas, Murty, and
Morris. 3 Each term is mUltiplied by a constant that is determined by a least
squares fit to the data in Figure 16-1. As can be seen from Table 16-1, the values
assumed for !:JIGB , !:JIMC. nGB, and nMC lie within the range of results reported
by other authors; these values also fit our data quite well (see Figure 16-1). Other
values were considered, but no improvement in fit resulted.
The values of the parameters used in Equation (16-2) are as follows:

C1 = 8.31
C2 = 1.12
!:JIGB = activation energy for grain boundary creep = 0.5 eV
!:JIMC = activation energy for matrix creep = 0.84 eV

k = Boltzmann's constant = 8.63 X 10-5 eV

K
514 INTEGRATED MATRIX CREEP

Table 16-2 Values of Parameters for Creep Equation (16-3)

Co 1.17 X 1O-5 . . !. . 5.24 X 1O-5 . . !. . 2.06 X 1O-4....!...

sec sec
TO 19.8 MPa (2870 psi) 15.0 MPa (2180 psi) 11.7 MPa (1700 psi)
This equation and the data points are shown in Figure 16-1.

T = temperature, K
T = shear stress, MPa
. h . 1
'Y = s ear stram rate, -
sec
naB = grain boundary creep exponent = 2
nMC = matrix creep exponent = 7.1
The following equivalent form of this equation is more convenient for numerical
analysis; values of Co and To for the temperatures used in the tests of reference
2 are listed in Table 16-2:

(16-3)

This equation and the data points are shown in Figure 16-1.

16.2.3 Elastic Strain Component

Considerable variation exists in published values of the elastic m~dulus of solder.

For example, reference 11 quotes a value of 14.8 GPa (2.15 Mpsi) for room
temperature tests of bulk samples of eutectic solder. In reference 12 a value of
30.0 GPa (4.35 Mpsi) is reported for bulk samples of 60% Sn-40% Pb alloy at
room temperature. In reference 13,43.4 GPa (6.3 Mpsi) is determined for elastic
modulus of 60% Sn-40% Pb alloy while reference 14 shows a range from about
12.4 GPa to 17.3 GPa (1.8 Mpsi to 2.5 Mpsi) increasing with frequency. Solomon 15
discusses in detail how structural compliance in test fixtures influences the mea-
surement of elastic properties of thin joints.
No attempt has been made here to separate the effect of apparatus and substrate
compliance of reference 2 on the overall elastic properties. Instead we note that
the experiment measures an effective shear modulus and assume the following:

1 1
-=--+--- (16-4)
G eff G so1der Gstructure
 DEVELOPMENT OF THE CONSTITUTIVE RELATION 515

Because of uncertainty involved in calculating the compliance of the fixtures

used to obtain the data analyzed here, determination of the correct value of solder
modulus would best be done by independent means. References 9 •10 present tem-
perature dependent elastic modulus of eutectic solder as E = - 0.088TeC) +
32 GPa; these values are used in Figure l6-3b.
The value of Poisson's ratio for solder at room temperature is generally taken
to be about 0.4. 11

16.2.4 Time Independent Plastic Strain Component

Hysteresis test data that give stable loops of stress versus strain under steady
cycling at different temperatures and frequencies are available for some of the
samples described in reference 2. The tests that give data with least noise are
those in which the total strain is a ramp function with time (total strain rate is
known). The magnitude of the total strain range is automatically adjusted in
successive cycles to be that necessary to maintain a particular value of inelastic
strain (opening of the hysteresis loop, see reference 2), but the samples are
observed to reach a stable loop in less than 10 cycles and not to show significant
softening for a reasonably large number of cycles. It is not possible to be more
precise since the complete history of deformation of the samples is not available.
The time independent (noncreep) part of the constitutive law is derived from
these stable ramp cycles as follows:
Using Equations (16-1) and (16-3), and knowing the stress-time history for
the hysteresis cycle, the creep strain can be subtracted from the total strain at
all stress values of the loop. In this way the value of time independent strain
(elastic plus plastic) can be calculated. Typical such curves are shown in Figure
16-2.
During unloading the time independent stress strain curve is assumed to be
purely elastic with slope Geff • In the loading regime, the form of the constitutive
law is taken to be a Ramberg-Osgood-type relation:

'Ytime-independent =~+
Geff
(2.)
Tp
n (16-5)

We determine Geff from the slope of the linear unloading regime; for the data
analyzed here, Geff varies from about 350,000 to 550,000 psi at room temperature
for the stainless steel substrates and depends on details of sample preparation.
At higher temperatures and for G10 substrates, the values of Geff are lower.
Using the proper value for Geff and a linear least squares technique for Equation
(16-5), the exponent n was found to be 2 to the nearest whole number for the
range of data available. Using n = 2, Tp is determined as a function of temperature
 516 INTEGRATED MATRIX CREEP

2000

1000

-1000

-2000~~~ ____~____~__~~__~____- L____~~~

-1.00 0.00 1.00
-y(%)

AA Original HJ3teresis Loop

u Loop with Creep Strain Remond

Figure 16-2 Hysteresis loop for sample A3 1D-min cycle at 6D°C.

by an additional least squares fit. Values of Tp versus temperature are given in

Table 16-3.
The parameter Tp does not relate to any definite deformation mode. It simply
includes all effects not covered by the steady-state creep and elastic terms. As
such, it may be considered to be something similar to a "plastic modulus," which
is consistent with its high value relative to actual solder stresses. The strong
temperature dependence of Tp may be due to the primary creep effect that it
encompasses or perhaps to anelastic effects.

Table 16-3 Values of Plasticity Parameter for Equation (16-5)

Temperature
25°e 289 MPa (41,900 psi)
60 e
0
240 MPa (34,700 psi)
100 e
0
135 MPa (19,500 psi)
 COMPARISON OF THE STEADY-STATE CREEP EQUATION TO PUBLISHED DATA 517

16.3 SUMMARY OF CONSTITUTIVE EQUATION

The above analysis leads to the following form of the constitutive equation (in
shear only) expressed in terms of strain rate:

1 = 1101 = [ _1_ + 2':' (1 + sgn(TI») ] ;-

Gsolder Tp 2

where sgn(x) = 1 for x> 0, = -1 for x < 0, and = 0 for x = O. Values

for the parameters To and Co are given in Table 16-2 and values of Tp are given
in Table 16-3.
Note that for mechanical analysis where the compliance of nonsolder com-
ponents is not separately accounted for, Gsolder will be replaced by G eff • This
approach is used here for analysis of the available test data.
Since the equation was derived only from hysteresis loops of stable cycling,
no yield criterion can be obtained.

16.4 COMPARISON OF THE STEADY-STATE CREEP EQUATION

TO PUBLISHED DATA

A great deal of published data exists on steady-state creep of solder joints and
much of it is presented in the form of Equation (16-2). Comparison of creep
exponents and activation energies has been made in Table 16-1. It is also useful
to compare some overall equations to each other and to Equation (16-2).

• Solomon5 : Solomon's results for steady-state creep rate in shear at 35°C are
shown in Figure 16-3a. A transition from a regime of slope approximately
7 to one of slope approximately 3 does occur but at a lower value of strain
rate than for the data of reference 2 .
• Grivas et aL 3: This study shows a grain size dependence for a "superplastic"
region but not for the high stress region. Figure 16-3a shows that agreement
is obtained with Equation (16-2) using an average grain size of 60-lLm. This
value is much larger than the grain sizes estimated for the samples used in
reference 3. The question of the conditions under which true superplasticity
can exist is controversiaL Here we avoid the term "superplasticity" and
approach the problem from a continuum rather than a metallurgical per-
spective.
The 6O-lLm grain size, however, is in reasonable agreement with the values
estimated in reference 2, although it is probably high. Calculation of grain
size for ordinary solder joints is difficult because of the wide variation
 518 INTEGRATED MATRIX CREEP

observed in size in a single joint. For comparison, the equation from ref-
erence 3 is also plotted for a grain size of 40 ~m .
• Wong et al. 9 : These authors have taken data from references3-5.11 and several
others and fit a relation of the same form as Equation (16-2). As is done
here, they neglect any grain size dependence since they feel that experimental
data are inconclusive on this point. Upper and lower bounds for this data
are given by Clech and Augis;1O these bounds are shown in Figure 16-3b
where E = - 0.088T(°C) + 32 GPa (4.3 Mpsi at 25°C); D = exp( - 5413/
T) with Tin K. For comparison, the data from Figure 16-1 and equation
(16-2) are also shown. Note that the data from reference 2 have been con-
verted to tensile from shear using the equivalent transformations given in
reference 10 or in Equation (16-12) of this paper.
It can be seen from Figure 16-3b that the data on which this study is
based lie within or close to the bounds given in reference 10. Since the

10000

'V,j
...=:::
to-
1000

,/
,/

100~__~~~~~U-~~~~~~~____L-~~~~
1.e-06 1.e-00 l.e-04 1.e-03
..y( l / s e c )
- Equation (16-2)
--. Solomon Data'
"'__ Grivas et al. 3 60 -~m grains
_._ Grivas et al. 3 40 -~m grains

Figure 16-3 (a) Comparison of creep data in shear (35°C).

 COMPARISON OF THE STEADY-STATE CREEP EQUATION TO PUBLISHED DATA 519

steady-state creep rates of reference 2 have been recalculated from the raw
data for the purpose of this study, the data in Figure 16-3b is slightly
different than the data attributed to Shine and Fox in,1O which was taken
from reference 2.

Table 16-1 and Figure 16-3a and b indicate the difficulties inherent in developing
a constitutive law for an arbitrary solder joint: it is evident that substantially
different results are reported for what seems to be the same test. It is possible
that some part of these differences may be attributed to testing problems such
as non-steady-state conditions, loading eccentricities or variations in test tem-
perature; 10 another reason for the different results may be microstructural dif-
ferences in the solders due to differing solder technologies. At the present time,
these differences are not understood well enough to be quantified. It is the

1000000.00

100000.00

10000.00

-I: 1000.00

.. rI·
.v·
.... a

1.00
E = -.088TCC) + 32 GPa
D = exp(-S413/T) (K)
0.10

a/E

.... ;Wong reference 2 -least squares fit to various data, upper & lower bounds
_ Equation (16-2) converted to tension
A Room Temp data from reference 2 -(converted to tension)
V 60'C data from reference 2 -(converted to tension)
c 100'C data from reference 2 -(converted to tension)

Figure 16-3 (b) Comparison of creep data in tension.

520 INTEGRATED MATRIX CREEP

contention of this work, however, that useful information can be obtained using
a constitutive law for a typical solder joint as long as all relevant modes of solder
behavior are taken into account.

16.5 APPLICATION OF CONSTITUTIVE EQUATION TO DATA

OF REFERENCE 2

16.5.1 Description of Numerical Procedures

To compare the prediction of the constitutive equation to experimental results,

Equation (16-6) will be solved for two cases. First is for a triangle wave strain
history, where 'Y is constant (equal to 'Yo) except that it changes sign. Second is
for a square wave where 'Y is zero except during transition from positive to
negative strain values.
For these cases, Equation (16-6) can be expressed by the following integral:

(16-7)

where T; is the initial value of stress for the time increment considered and Tf
is the final value.
The square wave is an idealization of an actual cycle since it is assumed that
loading and unloading are accomplished infinitely fast and, therefore, that the
value of strain is always +'Vmax or -'Vmax and strain rate is always zero (or
infinite for zero time). Any actual test will have finite rise time, but for a cycle
consisting of a fast ramp and hold when the ratio of the transition time to the
hold time is much less than 1, the cycle may be reasonably approximated by a
square wave. Equation (16-6) may also be solved for any other more complex
strain history if necessary.

Triangle Wave Cycle

Considering one half-cycle and assuming 'Yo is negative: T; = + Tmax and Tf

= -Tmax • Let 'Viol = total strain range and let tcycle = time for entire cycle

(loading and unloading). Then, 'Yo = 2'Vtot and the following equation can be
tcycle
written to solve for Tmax:
 APPLICATION OF CONSTITUTIVE EQUATION TO DATA OF REFERENCE 2 521

tcycle = (Tmax G eff dT

2 Jo 2'Y
tCY;~ + Co ~
[()2 + ()701]
~
1 2T
-+-
+ L o
Tmax

2'Yto!
Geff T;
T)2 (T )7 1] ~ (16-8)
- - CO [(- + -
0

tcycle TO TO

The equation was solved using a trapezoidal rule procedure, increasing the limit
of the integrals until the correct value of time was obtained. The first integral
is stable, but the second integral becomes infinitely large as

Co
[(ToT)2 +:;;;
(T)701] -+ 2'Ytot tcycle •

This asymptotic value of T is calculated before the numerical integration so that

it can be approached slowly and not exceeded. The value of T as a function of
time is also obtained.

Square Wave Cycle

In the case of the (idealized) square wave, the general fonn of the hysteresis
loop is shown in Figure 16-4 where 'Yo is known, and Ti and Tf must be calculated.
In regions B-C and D-A of the idealized cycle of Figure 16-4, defonnation
takes place infinitely fast, so no creep strain occurs; therefore the following
equation holds:

'Yo = Ti + Tf +(Ti)2
- (16-9)
G eff Tp

In regions A-B (and C-D) unloading occurs due to relaxation:

(16-10)

This gives two equations for the two unknowns Tj and Tf (initial stress and final
stress). There are two extreme cases: first, where there is no relaxation (tf =
522 INTEGRATED MATRIX CREEP

3000
A

2000

1000

-1000

-2000

-3000~~~
c ____~____~__~~__~____~____~__~
-1.00 0.00 1.00
-y(%)

Figure 16·4 Example of hysteresis loop for square wave.

0) and Tf = T; can be calculated from Equation (16-9); second, where there is

complete relaxation (tf very large), Tf = 0, and T; is again calculated from
Equation (16-9). Knowing the range of Tf, the two equations may be easily
solved by an iterative procedure.

16.5.2 Results

Comparison of the results of solutions to Equation (16-8) with the experimental

data for triangle wave tests shows excellent agreement over the range of tem-
peratures (25, 60 and lOO°C) and cycle times (from 1- to 20-min cycles) available.
Typical results are shown in Figure 16-5a and b where Geff = 450,000 psi and
'Vtot = 1.04 percent. Comparison of Equation (16-3) to the steady-state creep
results has already been shown in Figure 16-1.
In addition to the ramp and steady-state creep tests, another set of experimental
data is available for these solder joints, namely stress relaxation tests where the
displacement-time history approximates a square wave. This data is important
because it was not used in the development of the constitutive equation. The
test results for the square wave hysteresis loops are quite noisy, so any data
 2000

1000

-1000

-2000L-~ __~____~~~____~____~~____~~~~
0.00 200.00 400.00 600.00
Time (sec)
2000

1000

-1000

-2000L-__________ ~ __________ ~ __________ ~ ________ ~

-1.00 0.00 1.00

1'(%)
__ Calculated values W Data points

Figure 16-5 (a and b) Sample A3: 60°C-1O-min cycle.

523
 2000

1000
v v v

-1000

-2000~~~~____~~~~__~____~~~~__~~~~
0.00 100.00 200.00 300.00
Time (sec)
2000

v v

v
1000 v

o v v

-2000L-~~~__~~~~__~~~__~__-=~~ __~~~
-0.00 -0.20 0.00 0.25 0.00
y(%)
_Calculated values vv Dllla points

Figure 16-6 (a and b) Sample A: 60°C-l0-min cycle.

524
 MULTIAXIAl STRESS STATES 525

points shown are somewhat approximate. The strain history is of course not a
perfect square wave, although the ratio of the transition time to the hold time is
much less than 1; given the uncertainty in the data, it was felt that further
refinements in the numerical procedure to account for the transition period were
not warranted. Still, the agreement of the calculations with the approximate data
is good. Typical results are shown in Figures 16-6a and b; for this data, G eff =
300,000 psi and 'Ytot = 0.98 percent.
Solutions to the constitutive equation indicate that for fast cycling (less than
about 1 min) at room temperature or lower, the equation describes the solder
hysteresis loop even if the creep term is neglected. For cycling of more than a
few hours at constant strain rate at room temperature or higher, the instantaneous
plastic strain is negligible compared to the creep strain. For intermediate cases
(for example, lO-min cycles at 60°C), both creep and time independent plasticity
contribute significantly to deformation.

16.6 MULTIAXIAL STRESS STATES

16.6.1 Derivation of Constitutive Equation in Three Dimensions

Uniaxial tests to determine the relation between stress and strain in one dimension
are fairly easy to conduct. For most cases of interest, however, the actual state
of stress is much more complex. It is common in plasticity theory to determine
experimentally the behavior of a material in one dimension and then to generalize
this behavior to three dimensions using some simple assumptions.
The following discussion assumes that the material is isotropic. For thin solder
joints with relatively large grain sizes there often are only a few grains (possibly
as few as one or two) through the thickness of the joint. In this case, the
assumption of isotropy is not adequate; however, in accordance with an empirical
approach (and due to lack of data) we proceed in the traditional way.
Idealizing from experimental observation, it is assumed that the hydrostatic
pressure has no effect on the inelastic strains, and thus, that these strains depend
only on the history of the stress deviator, Sij == CTij - 1I3CT.u8ij (sum over
repeated indices). Furthermore, zero plastic volume change is assumed. The
generalization of a uniaxial relation to multiaxial states is then conventionally
given in terms of the second (J2 ) invariant of the stress tensor. Equivalently,
the Mises or "effective stress" (CT e ) can be used where:

(16-11)

Using this approach and Hooke's law for generalization of the linear term, the
complete equation for strain rate becomes
526 INTEGRATED MATRIX CREEP: APPLICATION TO ACCELERATED TESTING AND LIFETIME

+ C
~ [ ( _CT)
e_ _ e _ )6.1] s..
+ (CT (16-12)
2'To V3'To V3'To .)

where a = 1 for loading (i2 > 0), a = 0 for unloading (i2 < 0), and 8ij is
the Kronecker delta function. Note that ue = (3/2)(i2/CT e ). It is easily shown
that this equation reduces to Equation (16-6) for pure shear since in that case,
CTe = V31'TI,S12 = S21 = 'T, and'Y = 2Eh.
In the case of pure tension the only nonzero stress component is CTll = CT.
From the definition of the stress deviator, SII = 2/3CT ,S22 = S33 = iCT. Therefore,

As an example of the use of Equation (16-12), for loading in pure tension the
equation becomes

V3i =
. r.:& d (
V3"E + dt V3'Tp
(J" )2 + Co [(V3'CT)2 ((J" )7.1]
To + V3'To

As yet, no experimental evidence exists to confirm the validity of this gener-

alization. For a joint with only a few grains through its thickness, the assumption
of isotropy is almost certainly not warranted. However, a generalization of the
constitutive equation to three dimensions is necessary for the analysis of any
solder joint under thermal loading conditions where stresses are always mul-
tiaxial. The generalization presented here allows the constitutive relation to be
used with commercial general purpose finite element codes as will be seen later.

16.7 FATIGUE CALCULATIONS AND MECHANICAL SHEAR TESTS

Calculation of stresses, strains, or other parameters for solder joints is not mean-
ingful unless the values obtained can be used to predict structural reliability.
The purpose of this section is to show that when the solder constitutive equation
is solved for the different test conditions and solder joints used by various authors,
calculated parameters can be correlated with fatigue life.
 FATIGUE CALCULATIONS AND MECHANICAL SHEAR TESTS 527

The failure indicator used here is the one proposed by Shine and Fox, 2 which
correlates fatigue life with the component of creep strain due to matrix creep,

(16-13)

(per cycle) (16-14)

In reference 2, C is given as 125 percent while different values of nMC, Co, and
TO are determined experimentally for each sample. In this work, a general con-
stitutive law for eutectic or near-eutectic solder is determined; to be most useful
for mechanical modeling, fixed values of nMC, Co, and To must often be known
before testing can be done and this is the approach taken here. The value of C
will then vary depending on factors such as (perhaps) solder microstructure and
the way in which failure is determined to have occurred (e.g., electrical resistance
increase or change in structural compliance). The effects of such parameters on
fatigue life are not well understood. In this context, C can be regarded as an
experimentally determined constant. Some idea of the range of values to be
expected will be given below.
Parallels can be seen between Equation (16-13) and the Manson-Coffin fatigue
relation for materials exhibiting no time dependent effects. The expressions are
somewhat similar except that in the Manson-Coffin relation, the total plastic
strain (opening of the hysteresis loop) is raised to an empirically determined
power, which is usually about - 2. With the Shine-Fox matrix creep approach,
an inverse linear relation between fatigue life and the matrix creep part of the
strain has also been determined experimentally.
Accommodation of time dependent effects within the Manson-Coffin frame-
work remains problematic. It is generally accepted that for materials that exhibit
significant time dependence, the traditional Manson-Coffin law must be modified
to account for frequency effects; the exact nature of this frequency dependence
has not been definitely established. The premise here is that Equation (16-13)
is a good failure criterion for solder as demonstrated in reference 2. Because no
special modifications need be made to account for effects such as loading history
(frequency) or temperature, the equation is very general and may be considered
a more fundamental approach to the problem of solder fatigue.
Similarities also exist between Equation (16-13) and the Monkman-Grant
relation, which states that time to failure multiplied by steady-state creep rate
for one time loading is equal to a constant. In this sense, the constant C can be
thought of as the creep ductility exhaustion factor. Again, Equation (16-13) is
more general than the Monkman-Grant expression.
528 INTEGRATED MATRIX CREEP

16.7.1 Correlation of the Data of Reference 2

In the fatigue tests of reference 2, the solder joints were cycled at a particular
temperature with essentially a square wave strain-time history. Opening of the
hysteresis loop was kept constant by automatic adjustment of the total strain
amplitude. Failure was determined when the load to maintain this opening was
reduced by 50 percent.
In reference 2, a different steady-state creep relation in the form of Equation
(16-3) was determined for each sample. Then the matrix creep term was integrated
over the actual measured stress history to calculate the amount of matrix creep
per cycle as defined in Equation (16-14). This method was the basis for the
fatigue relation of Equation (16-13) with C determined as 125 percent.
To show that Equations (16-13) and (16-14) combined with the constitutive
law are still a useful failure indicator for cases where test results are not yet
available, the fatigue tests of reference 2 were analyzed using Equations (16-9)
and (16-10) for the square wave. A value of total strain range was chosen from
plots of the raw data, cycle time and temperature were known, and a value for
Geff was determined from a ramp test on that sample if available or from the
square wave plot if not. Input data for equation solutions are shown in Table
16-4.
Matrix creep strain is easily calculated using Equations (16-9) and (16-10).
Results are shown in Figure 16-7. Note that these results are somewhat (but not
very) different from the values of matrix creep per cycle obtained in reference
2. A least squares fit to this data of an equation of the form of Equation (16-3)
gives C equal to 170 percent, in good agreement with the value obtained by
Shine and Fox. The data show the scatter that is typical in fatigue testing.

Table 16-4 Summary of Shine-Fox Data: 60°(, Square Wave T Versus Time
Relation-Raw Data for Figure 16-7
Sample Geff Period Substrate
Name (psi) 'Ytot (%) (min) Material Nf
All 70,000 5.8 10 G.10 486
AlO 200,000 2.1 10 Gl0 526
A12 110,000 3.0 10 GlO 1200
3 125,000 1.6 30 GlO 3436
2 100,000 2.5 30 Gl0 4262
1 95,000 3.3 2 GlO 5000
Al 400,000 1.6 10 Stainless 320
Steel (SS)
A2 415,000 1.1 2 SS 880
BO 350,000 1.1 10 SS 1250
A 300,000 .98 10 SS 1870
 FATIGUE CALCULATIONS AND MECHANICAL SHEAR TESTS 529

10.00

c-!'i" 1.00
"
A
"
"
0.10
A
"
A

" A

1000 10,000
Nt

- Least squares fit to Shine-Fox data

... Data points

Figure 16-7 Cycles to failure versus matrix creep strain for data of Equation (16-2).

16.7.2 Correlation of the Data of Wild and Solomon

Wild 16 reports fatigue life of solder joints as a function of total strain varied as
a sinusoidal function of time; each data point is an average of 30 to 50 tests, so
the scatter always present in fatigue tests is minimized. Failure was determined
when electrical resistance increased by 10 percent. The joints tested were lap-
shear configurations using steel specimens to reduce bending. The joint sizes
were 2.9 by 1.1 by 0.1 mm thick (0.115 by 0.045 by 0.004 in.).
To correlate the Wild data with the matrix creep failure indicator, it is necessary
to solve the constitutive equation for each reported cycle and calculate the re-
sulting matrix creep strain. All necessary parameters are available except for
Gett; somewhat arbitrarily, but based on values from other fatigue testing of
solder joints, the value is taken to be 2.41 GPa (350 ksi) at room temperature
and 1.72 GPa (250 ksi) at 100°C. Due to the approximate nature of these
530 INTEGRATED MATRIX CREEP

100.00

•
"
10.00
" •
•
9
9
0

~ 1.00
9

~
N,=-- "
Ymc •
0.10 9

_Least squares fit to Solomon and Wild-Engelmaier Data

&& Solomon Data
n Wild-Engelmaier Data

Figure 16-8 Cycles to failure versus matrix creep strain for Equations (16-5) and (16-15).

calculations, representation of the sine wave history by a triangle wave is felt

to be adequate. For the longest cycles at high strain rates, numerical problems
were encountered and no results are available. Engelmaier17 has derived an
expression to relate total strain to cycles to failure by fitting data from reference
16. This expression was used to determine lifetime versus total strain. Input data
for data points of Figure 16-8 are given in Table 16-5.
Solomon6 presents data giving fatigue life versus plastic strain (opening of
hysteresis loop) at 35°C for rapid cycling (triangle wave at 1200 cycles/h). The
cross sections of the joints are much larger than those used by the other authors
considered here (2.4 by 13 mm and 0.13 to 0.23 mm thick). It is not clear what
effect this different geometry may have on the fatigue life. Failure is defined in
the same way as in reference 2 (50 percent load reduction). A value of Geff is
reported as approximately 1.38 GPa (200,000 psi) in reference 15. Equation (16-
 FATIGUE CALCULATIONS AND MECHANICAL SHEAR TESTS 531

Table 16-5 Wild Data: Triangle Wave 'V Versus Time Relation-Raw Data for
Figure 16-8
Ceff Temperature
(psi) 'Vtot (%) Period (0C) Nf
350,000 20 12 sec 25 25
350,000 15 12 sec 25 64
350,000 8 12 sec 25 509
250,000 8 12 sec 100 208
250,000 5 12 sec 100 804
250,000 3 12 sec 100 3500
350,000 5 15 min 25 447
350,000 3 15 min 25 1730
350,000 2 15 min 25 5070

8) can be solved for the test conditions of reference 6 using an estimated value
of total strain to obtain a value of plastic strain for a cycle. Iteration gives the
correct value of total strain for a desired plastic strain. Matrix creep strain for
the cycle can then be calculated. The input data used is given in Table 16-6.
Figure 16-8 shows the results of correlating calculated matrix creep with
experimentally determined number of cycles to failure for the Solomon and Wild
data. A least squares fit to this data of an equation of the form of Equation (16-
13) gives a value of C of 890 percent, higher than the value obtained for the
Shine-Fox data.
It can be seen from Figure 16-3a that the values of the parameters Co and To
are different for the Solomon data than they are for the Shine-Fox data and we
would expect (if it were to be calculated) that this would also hold true for Tp
due possibly to different solder microstructure or geometry. In fact, Solomon,
using the relation between steady-state creep rate and stress determined for his
samples, has calculated the amount of creep strain in his measured hysteresis
100ps. 6 Using Equation (16-8) the same calculation gives the percentage of creep
per cycle to be a factor of 2 higher. Using Solomon's creep calculations, Shine 18

Table 16-6 Solomon Data: Triangle Wave 'V Versus Time Relation-Raw Data for
Figure 16-8
Ceff Temperature
(psi) Period (Oe) !:i.'Vpl (%) Nf
200,000 3 sec 35 15 66
200,000 3 sec 35 10 131
200,000 3 sec 35 5.5 446
200,000 3 sec 35 3.0 1430
532 INTEGRATED MATRIX CREEP

has shown that the Solomon data fallon the same line as the data from reference2 •
No steady-state creep data are available in reference 16 .
Solomon6 also has experimental data for lifetime versus frequency at high
fixed values of plastic strain which show a sharp drop-off in lifetime as frequency
becomes very low. As Solomon points out,19 the matrix creep failure indicator
predicts the opposite of this effect. As cycle period becomes very long, to
maintain a fixed opening of the hysteresis loop, maximum stress goes down;
since matrix creep varies as the seventh power of stress, the amount of matrix
creep per cycle actually begins to decrease with decreasing frequency despite
the longer time available for creep strain to accumulate, thereby predicting an
increase in lifetime rather than Solomon's observed decrease. It may not be valid
to apply the equations of this chapter to directly compare Solomon's data: strain
values for available data that fall into the temperature range considered here
(25°C < T < 100°C) are much larger than those covered by this work. The
lifetimes where the drop-off occurs are generally less than 100 cycles. We infer
that effects other than those covered by this work are responsible for Solomon's
observations. The matrix creep failure indicator was based on data at the relatively
moderate temperature and strain values often found in many applications. Care
must be taken in use of the matrix creep approach with due consideration given
to its range of applicability.

16.8 ANALYSIS OF LEADED SOLDER JOINTS

This section applies the constitutive equation and fatigue criterion to the analysis
of solder joints of a leaded chip carrier. For leadless carriers, it may be possible
to assume a pure shear stress state and to use the results of simple tests to get
an estimate of fatigue life. For leaded joints, however, it is known from
experiments 20 that the compliance of the lead is one of the most important factors
influencing lifetime. The interaction of the lead and solder is difficult to simplify
and thus an in-depth stress analysis may be warranted. The leaded case differs
from the simple shear results cited previously for the following reasons:

• Stresses and strains in a leaded solder joint are multiaxial: In general there
are six components of stress and strain, three tension-compression and three
shear. Each component may be expected to contribute in some way to damage
accumulation and ultimate failure.
• The stresses and strains vary with position throughout the structure.
• The definition of lifetime in fatigue tests of actual components is, for practical
reasons, almost always the onset of electrical opens and not a percentage
load drop or resistance change as was the case in the pure shear mechanical
tests.
 ANALYSIS OF LEADED SOlDER JOINTS 533

At first glance, failure of the leaded solder joint may seem to be a candidate for
analysis by fracture mechanics. However, the problem is extremely complex
from a fracture point of view: Creep plays an important role, stresses are cyclic,
and effects of different nearby materials on the thin solder layers always introduce
a mixed-mode fracture problem. Each of these situations separately is difficult
to treat using classical fracture mechanics, and together they are more a subject
for academic research than practical design engineering.
It is, therefore, our contention that a more empirical approach will be most
effective in extending previous results to the leaded case: We will define and
then calculate certain "failure indicators." These indicators will be correlated
with experimental fatigue data in order to determine a relation to predict expected
lifetime. We will be guided in our efforts to choose failure indicators by the
following:

• Fatigue failure is known to correlate with calculated strain. For low tem-
perature fatigue calculations of more conventional structures (e.g., auto-
mobile components) some form of the empirical Manson-Coffin relation23
is often used successfully. For the frequency dependent creep case, the
relation must be modified.
• Consideration of the use of total strain as a failure indicator leads to the
following paradox: For practical cases of leaded chip carriers, total dis-
placement is imposed on the lead-joint by the deformation of the board
and chip carrier and relaxation times are very long. Therefore, for stiff or
compliant leads, final strains per cycle should be approximately the same.
However, it is known that more compliant leads result in longer fatigue
life.
• The matrix creep failure indicator described earlier provides one resolution
to this paradox. Even though total strain per cycle may be the same for both
types of leads, the strain will be accumulated at lower stress in the case of
the compliant lead resulting in less matrix creep strain than for the stiff lead
case. Therefore, we hope to find correlation between matrix creep strain and
lifetime for the leaded case as we did for the leadless case.

Engelmaiero has published solder fatigue results for thermal cycling of

100-1/0 ceramic chip carriers with two types of clip leads having significantly
different compliance. In this section, we will attempt to model this structure and
calculate failure indicators using the constitutive law and finite element analysis.
In order to confidently relate these indicators to fatigue life, much more data
should be analyzed than is possible within the scope of this work. The example
presented here is offered as a beginning of this process.
534 INTEGRATED MATRIX CREEP

16.8.1 Extension of Matrix Creep Failure Indicator to General Case

Multiaxial Stress States

The matrix-creep failure criterion, Equations (16-13) and (16-14), is given

for the case where the only stresses or strains are presumed to be pure shear in
a single direction. For practical cases, multiaxial stress-strain states will be
present. The law may be generalized to three dimensions using the effective
strain:

(16-15)

where

(16-16)

This criterion reduces to Equation (16-13) for the case of pure shear but other
than that is arbitrary; however, if the material is presumed to be isotropic, the
generalization should be in terms of a quantity that is invariant with respect to
change in choice of coordinate axes. Instead of effective strain, maximum prin-
cipal strain or another similar quantity might be substituted. It should be kept
in mind that the solder joint may not be isotropic if only a few grains are present
through the thickness of the joint. For simplicity, isotropy is assumed throughout
this chapter.

Variation of Strains with Position

For leaded joints, not only are the strains multiaxial, but they vary with position
in the structure; at some points "effective" strain may be high while at others it
may be almost zero.

1. The traditional approach to this problem is to use maximum strain values

as failure indicators. Maximum strains may be expected to correlate with
crack initiation; however, failure would be expected to correlate also with
crack growth that might be a function of the geometry and stress state of
the entire structure. Using maximum values would be expected to be a
conservative approach.
2. Another method to address this question has been suggested. 24 In this
approach the total matrix creep to be used in Equation (16-15)'is defined
 ANALYSIS OF LEADED SOLDER JOINTS 535

as a volume weighted-average of the matrix creep strain throughout the

structure:

e!,!~ = ~ Je MC (x,y,z)dV (16-17)

As before, another strain invariant could be used instead of effective strain.

The difficulty with this approach is that it may overpredict lifetimes because
it takes into account the entire structure whether or not it is needed to
maintain structural integrity. It is easy to imagine a structure with a part
of the volume under low stress and a smaller part under high stress where
a crack through the high stress region would result in failure. It may not
make sense to average the strains over the entire volume in this case;
however, in some cases volume average matrix creep may be found to
correlate with lifetime and thereby find use for practical design engineering.
3. To eliminate volume averaging over low stress regions, one might confine
the integration to regions where the stress exceeds some "trigger" level.
The form of Equation (16-6) suggests that To be considered as such a
threshold since it sets the stress above which matrix creep dominates total
creep. With this consideration, we also carry out lifetime calculations using
Mises stress > To. This approach might be expected to be invalid for
lifetime estimates in very low stress cases, where the error in neglecting
finite contributions of low stress regions becomes significant. Indeed, we
see this in the following example.

Definition of Lifetime

In the tests discussed previously, Shine and Fox2 and Solomons.6 have used
50 percent load reduction to define failure while Wild 16 has used 10 percent
resistance change. It is necessary to relate these to the electrical open criterion
(the only important one from a practical standpoint) used for the Engelmaier
tests. Solomon21 has studied this problem in depth: he concludes that a resistance
increase of 0.1 percent is observed with load drops of 60 to 90 percent. In general
it takes a load drop of about 90 percent before the first open circuit is reached.
The difference in number of cycles between 50 percent and 90 percent load drop
varies with plastic strain range and can be more than an order of magnitude in
reference 21. The difference seems to be larger at smaller values of plastic strain
range.
Here, we assume that the lifetime definition is taken into account by the
constant C in Equation (16-13). This constant may be calculated for the electrical
open case by curve fitting to experimental data as was done for the pure shear
536 INTEGRATED MATRIX CREEP

mechanical tests. The constant will also depend on which definition is used for
the matrix creep parameter.
Geometrical factors related to specific joint design, affecting crack initiation
and growth, may also be accommodated by the empirical determination of C.
It is possible microstructural effects might also be accounted for.
We suggest that, once the effective matrix creep fatigue ductility coefficient
C is determined experimentally, further lifetime calculations and sensitivity stud-
ies may be carried out on a family of similar designs without further arbitrary
adjustments. Additional experimental investigation is required to fully support
this assertion.

16.8.2 Description of Model

The finite element analysis of the Engelmaier20 tests was done using the ABAQUS 2S
general purpose finite element code. The finite element mesh is shown in Figure
16-9; elements are isoparametric eight-node bricks.
Most stress analysis is subject to limited computer resources and therefore
problems must be simplified to get meaningful results in a timely fashion. The
following simplifications were made in the finite element models used here:

• The geometry of the leads and joints is simplified. The configurations are
shown in Figure 16-9. Lead geometries are taken from Kotlowitz's analysis
of lead compliance for these tests 22 but modified somewhat. The lead is
assumed to be placed at the middle of the pad with 3 mil of solder under
it. The geometry of the joints is very approximate. Fillets in the foot-toe
direction are included but those in the lateral direction are ignored (Figure
16-9).
• Effects of the leads and solder joints on the deformation of the board and
chip carrier are ignored. Therefore, the displacements imposed on the solder
joints are due only to the lateral deflection of the board and chip carrier
caused by thermal mismatch between them. Analysis of the conservative
problem of board, chip carrier, and leads with infinitely stiff solder joints
shows this to be a reasonable assumption. Imposed displacements are those
that would be experienced by a lead on the package comer since this is the
most highly stressed lead. These displacements are shown in Figure 16-1Oa
and b where, 8y = -0.39 mil and 8x = 0.35 mil.
• The temperature-time profile has been simplified: The structure is assumed
to go instantaneously from 25 to 85°C and to hold at this higher tempera-
ture for 450 sec. The second part of the cycle where the structure cools
and then holds at low temperature is not included in the analysis since it is
assumed that the amount of creep strain accumulated there is relatively
low. Two dimensional studies of a structure similar to the type A clip lead
 ANAL YSIS OF LEADED SOLDER JOINTS 537

From Reference 20

Type A
Type B

159

'L I

Figure 16-9 Finite element meshes. Copyright © 1989 IEEE.

538 INTEGRATED MATRIX CREEP

- - - L..--r-"

- -
.......-
L-- V

\ \ ~
 TS 539
ANALYSIS OF LEADED SOlDER JOIN

H-ti"""n-
- tt rr- 1
~

~ \

tL~
I--
I--
I
540 INTEGRATED MATRIX CREEP

and solder joint were performed to try to give a simple rough estimate of
the effect of this assumption; these studies show that lifetime is decreased
by about 10 to 20 percent when this part of the cycle is included. There-
fore, to reduce computational run-time, only the warm half of the cycle
was analyzed. This approach is consistent with the determination of an
effective ductility C.
• The finite element meshes for the two cases studied by Engelmaiero are
shown in Figure 16-9. It is well known that mesh density can have a large
effect on calculated strains and that the effect of mesh refinement should
always be checked. Unfortunately, limited computer resources make this
difficult for 3D highly nonlinear problems. Mesh studies were carried out
with a 2D mesh analogous to the type A lead-joints to guide the choice used
here. Results of these 2D models show the level of mesh refinement used
for the 1-2 plane in Figure 16-9 to give a maximum strain approximately
within 20 percent of the asymptotic value that is obtained from further mesh
refinement. This is only a rough estimate for the 3D case.

16.8.3 Results

As explained above, we regard C as an experimentally determined parameter

that will vary depending on the definition of failure and on the method of
extension of the matrix creep strain in pure shear to a structure with a complex
stress-strain state. For the Engelmaiero tests, the failure event or criterion is
electrical opens. The mean fatigue life is the number of cycles when 50 percent
of the population has failed. The joints of the type A (stiff) leads show a mean
of 20,000 cycles to failure with a Weibull slope of 2 as shown in Figure 16-11.
All failures are assumed to occur at the "SJ-C" location between the lead and
the FR4 board. The test was discontinued at 100,000 cycles; at that time the
board with type B (compliant) leads had not experienced any failures.

• It is assumed that the calculated lifetime is equal to the mean population

lifetime (20,000 cycles). Using this value of fatigue life and the "volume
average matrix creep" defined above, the value of C for the type A clip
lead is determined to be 1.7 (or 170 percent), which is the same as for
the data of reference 2. Using this value of C and calculating volume av-
eraged matrix creep for the type B leads, lifetime is predicted to be 860,000
cycles. This gives a value of the ratio of lifetimes between the type Band
type A lead cases of 42. Since no failures occurred in the type B lead test
before it was stopped, actual lifetime is not available. Assuming a failure
in the next cycle after the end of the test and a Weibull slope of 2 equal
to the type A clip lead case, Engelmaier estimates the minimum lifetime
to be 600,000 cycles.
 99.9
Usinl. "volume averale matrix creep"
99
95
90
10 Ullins 8 maximum matrix creep"
70
l
en
W
60
50
j:'
II:
~ 40
...J
30
~
en 20
II:
W
a:II: 10
<C .., : Type A CLIP LEADS
0 >
c.. .5 e--: Type B CLIP LEADS z
>
r0 . "NO'FAILURES ~
V>
Vi
AT 10 5 CYCLES
2 o
.."
r-
~
o
m
2 3 5" 2 3 5.' 2 3 5 I' 2 3 o
10 3 10· 10 5 10'
r-
oo
m
'"
(5
Figure 16-11 Predicted Weibull plots for type B leads. Copyright © 1989 IEEE. Z
-i
V>

'"~
542 INTEGRATED MATRIX CREEP

Table 16-7 Summary of Results for Leaded Chip Carriers

Ratio
Nf Type A Nj Type B Type B
Method (data) (calculated) C Type A
Vol. Avg. matrix creep 20,000 860,000 1.7 42
Max. matrix creep 20,000 2,300,000 137 115
Vol. Avg. Me, (J'e ~ To 20,000 QC)
36 QC)

Engelmaier 20,000 3,500,000 175

Actual 20,000 >600,000 >30

• Using matrix creep strain at that location in the structure that experiences
the maximum value of effective matrix creep strain, C for the type A lead
case is found to be 137. It is to be expected that this number is much higher
than for the pure shear case. Stresses for the "pure shear" tests are calculated
using measured total force; for an elongated joint actual stress will not be
constant and therefore the value used to detennine C for the pure shear
mechanical tests is really an average itself. Using C = 137 and maximum
effective creep strain calculated for the joints of the type B lead, the resulting
lifetime is 2,300,000 cycles, or a ratio of 115. For comparison, Engelmaier
estimates this ratio at 175 using a ratio of the relative compliances of the
leads and a relationship between total strain and fatigue life based on the
simple shear mechanical tests of Wild. 16 The Weibull plots of fatigue failure
for the type A solder joints and projections for the type B joints are shown
in Figure 16-11.
• Using volume averaging over that volume whose stress is about To results
in a value of C = 36 for the type A clip lead. For the type B lead, calculated
lifetime is infinite since there is no region where Mises stresses exceed TO.

Each of these methods give results that are consistent with available test data.
More information is needed to choose the most appropriate failure indicator.
Results are summarized in Table 16-7.

16.9 CONCLUSIONS

A stress-strain relation for a typical solder joint has been developed from shear
tests and generalized to three dimensions. The equation accounts for elastic,
time independent plastic and steady-state creep components of solder strain.
Solutions to the equation have been developed for triangular and square wave
strain versus time profiles; these solutions have shown that the equation gives a
very good description of the Shine-Fox2 data over a range of conditions.
 REFERENCES 543

The constitutive equation has been solved for pure shear mechanical test
conditions where fatigue life has been measured. The matrix creep failure in-
dicator proposed in reference 2 (Equations 16-13 and 16-14) has been calculated
for three independent sets of such solder fatigue data; it has been shown that the
slope of the logarithm of cycles to failure versus the logarithm of matrix creep
curve is about - 1 for all data considered. The effective matrix creep ductility
varies depending on the data being considered. While it is necessary to determine
this constant experimentally, relative design comparisons can be made on this
basis before testing. Since it has been shown that the slope of the Nf versus "fMC
curve does not vary, fewer tests are necessary to predict fatigue life for a given
construction.
Finally, some possible extensions of the matrix creep failure indicator to the
case of leaded chip carriers are proposed. As an example, these criteria are
applied to the analysis of thermal tests of clip leaded ceramic chip carriers
conducted by Engelmaier. The example shows that unresolved issues exist in
the application of matrix creep to the more complex stress states of leaded solder
joints. With more data, some resolution of the questions posed can be expected.

16.10 ACKNOWLEDGMENTS

The authors wish to thank B. Mirman and M. C. Shine for their helpful dis-
cussions of this work.

REFERENCES

1. Lau, J. H., D. W. Rice, and P. A. Avery, "Elastoplastic Analysis of Surface-Mount

Solder Joints," IEEE Trans. Comp., Hybrids, Manuf. Technol., CHMT-IO, 1987,
pp. 346-357.
2. Shine, M. C., and L. R. Fox, "Fatigue of Solder Joints in Surface Mount Devices,"
Low Cycle Fatigue-ASTM Special Technical Publication 942, 1987, pp. 588-610.
3. Grivas, D., K. L. Murty, and J. W. Morris, "Deformation of Pb/Sn Eutectic Alloys
at Relatively High Strain Rates," Acta Met., 27, 1979, pp. 731-737.
4. Kashyap, B. P., and G. S. Murty, "Experimental Constitutive Relations for High
Temperature Deformation of a Pb/Sn Eutectic Alloy," Materials Science and En-
gineering, 50, 1981, pp. 205-213.
5. Solomon, H. D., "Creep, Strain Rate Sensitivity and Low Cycle Fatigue of 60/40
Solder," Brazing and Soldering, no. 11, 1986, pp. 68-75.
6. _ _ , "Low-Frequency, High Temperature, Low-Cycle Fatigue of 6OSn-40Pb
Solder," Low Cycle Fatigue-ASTM Special Technical Publication 942, 1987, pp.
342-370.
7. Baker, E., "Stress Relaxation in Tin-Lead Solders," Materials Science and Engi-
neering, 38, 1979, pp. 241-247.
544 INTEGRATED MATRIX CREEP

8. Lam, S. T., A. Arieli, and A. K. Mukherjee, "Superplastic Behavior of Pb-Sn

Eutectic Alloy," Materials Science and Engineering, 40, 1979, pp. 241-247.
9. Wong, B., D. Helling, and R. W. Clark, "A Creep-Rupture Model for Two Phase
Eutectic Solders," IEEE Trans. Comp., Hybrids, Manuf. Technol., CHMT-ll, 1988,
pp. 305-325.
10. Clech, J. P., and J. A. Augis, "Temperature-Cycling, Structural Response and
Attachment Reliability of Surface-Mounted Leaded Packages," 1988 IEPS, Dallas,
November 1988, pp. 305-325.
11. Westinghouse Defense and Space Center, "Development of Highly Reliable Soldered
Joints for Printed Circuit Boards," Aug. 1968.
12. International Tin Research Institute, "Solder Alloy Data," I.T.R.I. Publication No.
656.
13. Adams, P. J., 'Thermal Fatigue of Solder Joints in Micro-Electronic Devices," M.S.
Thesis, MIT, 1986.
14. Engelmaier, W., "Test Method Considerations of SMT Solder Joint Reliability,"
IEPS October 1984, pp. 360-369.
15. Solomon, H. D., "Low Cycle Fatigue of 60/40 Solder-Plastic Strain Limited vs
Displacement Limited Testing," Electronic Packaging: Materials and Processes, J.
A. Sartell, Eds., ASM, 1989, pp. 29-47.
16. Wild, R. N., "Some Fatigue Properties of Solders and Solder Joints," IBM Report
No. 74z000481; also INTERNEPCON, Brighton, England, October 1975.
17. Engelmaier, W., "Fatigue of Leadless Chip Carrier Solder Joints During Power
Cycling," IEEE Trans. Comp., Hybrids, Manuf. Technol., CHMT-6, 1983, pp.
232-237.
18. Shine, M. C., private communication.
19. Solomon, H. D., private communication.
20. Engelmaier, W., and A. I. Attarwala, "Surface-Mount Attachment Reliability of
Clip-Leaded Ceramic Chip Carriers on FR-4 Circuit Boards," IEEE Trans. Comp.,
Hybrids, Manuf. Technol., CHMT-12, 1989, pp. 284-296.
21. Solomon, H. D., "Low Cycle Fatigue of Surface Mounted Chip Carrier/Printed
Wiring Board Joints," 1989 ECC, pp. 277-292.
22. Kotlowitz, R. W., "Comparative Compliance of Representative Lead Designs for
Surface Mounted Components," 1989 ECC, pp. 791-831.
23. Knott, J. F., Fundamentals of Fracture Mechanics, New York, John Wiley, 1973.
24. Iannuzzelli, R. , "Validation of Module Assembly Physical Models," 40th Electronic
Components and Technology Conference, May 21-23, 1990.
25. Hibbitt, Carlsson and Sorensen Inc., ABAQUS V 4.8, 1989.
 17
Solder Attachment Reliability,
Accelerated Testing, and
Result Evaluation
Werner Enge/maier

17.1 THE RELIABILITY OF ELECTRONIC ASSEMBLIES AND

SOLDER JOINT RELIABILITY

The reliability of electronic assemblies depends on the reliability of their indi-

vidual elements--e.g., integrated circuit (ICs); active components: chip carriers
(CCs), small-outline integrated circuits (SOICs), small-outline transistors (SOTs),
dual in-line packages (DIPs), small-outline J-Ieaded devices (SOJs), etc.); passive
components (chip resistors, chip capacitors); printed wiring boards (PWBs);
connectors; etc.-and the reliability of the mechanical, thennal, and electrical
interfaces (or attachments) between these elements. These interfaces can be
connector contacts, wire-wraps, thennal shunts, bonding layers between a printed
circuit board and a heat sink, leads soldered into plated through holes (P1Us),
chip solder attachments (flip-chip or C4), or surface mount (SM) solder joints.
In this context reliability is defined as: the ability to function as expected for
an expected period of time without exceeding an expected acceptable failure
probability .
It should be noted that the chapter title refers to the reliability of solder
attachments, not the reliability of solder joints. A solder joint in isolation is
neither reliable nor unreliable; it becomes so only in the context of the individual
elements that are connected via the solder joints and the characteristics of these
elements together with the use environment, the design life, and the acceptable
failure probability for the electronic assembly.
Among the different solder connections, surface mount solder attachment is
unique since the solder joints not only provide the electrical interconnection but

545
546 SOLDER ATIACHMENT RELIABILITY, ACCElERATED TESTING

are also the sole mechanical attachment of the electronic components to the
printed wiring board. Thus, surface mount solder attachment requires that special
reliability concerns be addressed and that explicit design attention is provided
to assure long-term reliability.
An additional complicating factor is the fact that at typical operating temper-
atures solder behaves in a fashion more akin to a very viscous fluid than a typical
structural material. It creeps and stress-relaxes readily and can be thought of as
having a load-bearing capability on a strictly temporary basis only. Solder joint
strength, beyond a rudimentary minimum level easily obtained with all but
grossly defective solder joints, is not of concern for the long-term reliable func-
tioning of most surface mount solder attachments.
Surface mount solder joints subjected to cyclic thermomechanical strains and
stresses in operation, accumulate fatigue damage in the solder which, in time,
results in fatigue cracks and eventually in mechanical and electrical failure of
the solder attachment. Thus, surface mount solder joints raise the concern of
premature wear-out failure for inadequate designs.

17.1.1 #Bathtub" Reliability Curve-Hazard Rate Model

Reliability experiences with electronic components have given rise to the "bath-
tub" reliability curve as shown in Figure 17-1, where the instantaneous failure
rate (hazard rate) in FITs (failures in 109 device hours) is plotted against the
operating time in hours.' The bathtub curve has three failure regions with distinct
characteristics, which are associated with failures due to "infant mortality,"
random failures in "steady state" operation, and failures due to "wearout." Nor-
mal operation of mature product defines the low steady-state (random) failure
region. The failure rate in this region is in part dependent on the complexity of
the product. In the wearout phase the accumulated damage from normal operation
becomes large enough to cause rising failure rates, signaling the approaching
end of useful life. The wearout phase is primarily dependent on the choice of
the design parameters and the use conditions.
Infant mortality comprises the operational phase from start-up to typically no
longer than one year of operation and is characterized by an initially high but
rapidly decreasing failure rate. An early high infant-mortality failure rate is
typically associated with a small fraction of the population which consists of
"weak" or "defective" "freaks" and is dependent on the process controls during
manufacture and the quality of the resulting product. Appropriate bum-in or
environmental stress screening (ESS) procedures can reduce these "infant mor-
tality" failures. The infant mortality failure rates are characterized by1.2

(17-1)
 THE RELIABILITY OF ELECTRONIC ASSEMBLIES AND SOLDER JOINT RELIABILITY 547

•• "TYPICAL" ELECTRONIC "TYPICAL· SM

.••
•• COMPONENT SOLDER
•• / W/o ATTACHMENT
•

r··.
LEADED

INFANT ••
MORTALITY ••

f··· .•,
.

40
, YEARS
0.1
1 10
OPERATING TIME (h)

Figure 17·1 Reliability "bathtub" curves for "typical" SM solder attachments and a "typical"
electronic component. The curve for the "typical" component is based on test and field failure
data. For the solder attachments the infant-mortality and steady-state phases are "best, edu-
cated" guesses, while the wearout phases are based on the material in the section "Analytical
Model of Solder Shear Fatigue." The location of the wearout curves is highly dependent on
the choice of design parameters and the use environment to the extent that the wearout region
bands shown are not wide enough to include even all likely let alone possible wearout scenarios.

where A(t) is the instantaneous failure rate in FITs, AI the failure rate at t = 1
hour, t the operating time in hours, and - a the Weibull slope. MIL Handbook
217D3 gives for reftowed solder joints AI = 18 FITs per joint and a = 0.6. It
should be noted that these values, used to plot the infant mortality region for a
"typical" surface mount solder attachment (56 solder joints) in Figure 17-1, while
widely used, 3 are rather "soft." The low failure rates preclude experimental
determination and hard field data in adequate quantities do not exist for either
the component or solder attachment infant mortality curves.
The random steady-state region is characterized by a constant, or near constant,
failure rate of mature assemblies in normal operation. The failures are of course,
neither random nor steady state and have meaning only in a purely statistical
sense; they cannot be characterized mechanistically, but failure rates tend to be
higher for more complex components. Chip components characteristically fall
well below the indicated bathtub curve in Figure 17-1, whereas very complex,
high liD chip carriers could fail at higher rates. For solder joints no evidence is
available that shows that the "random steady-state" region even exists; it is
entirely possible that "infant mortality" and "wearout" form the entire active
failure rate history. Nevertheless, Reference 3 gives a steady-state failure rate
548 SOLDER ATIACHMENT RELIABILITY, ACCELERATED TESTING

for reflowed solder joints of).L = 0.07 FITs per joint; given the experience
with chip carriers having surface mount solder attachments consisting of 44 to
100 individual solder joints, this value for ).L appears too high by perhaps as
much as one or two orders of magnitude. Large electronic systems can contain
hundreds of thousands of solder joints; a steady-state failure rate of that magnitude
would make such systems impossible. For Figure 17-1, ).L = 0.0035 FITs per
joint was assumed.
The most important aspect illustrated in Figure 17-1, however, is the difference
in the "wearout" regions of the "typical" electronic component without solder
attachment and the "typical" surface mount solder attachment. Wearout is char-
acterized by a failure rate that increases monotonically with time. It results from
the accumulation of damage from some mechanism and represents the point in
the operational life where the amount of accumulated damage has reached levels
that cause failure. For electronic components, wearout during the intended service
life is encountered only rarely; of course, given the steady-state component failure
rate, the number of components in a system, and the years of service, system
failures caused by component failures even at low failure rates could become
rather significant. Whereas, for electronic components from reputable suppliers
and which are operated in recommended fashion "wearout" failures are not
expected during the intended operational life, this is not necessarily true for
surface mount solder attachments. Without appropriate design attention, surface
mount solder attachments, particularly of the type without compliant leads, can
easily be the cause of premature wearout failures. The wearout regions in Figure
17-1 are based on the experimental results from accelerated reliability tests that
have given rise to a solder joint reliability model described later. The appropriate
design steps which can assure that significant wearout of surface mount solder
attachments does not occur during the intended operational life of the electronic
assembly, are outlined here and in Chapter 18, "Surface Mount - Attachment
Reliability. "
Consistent good quality solder joints should result in significantly lower infant
mortality and (quasi-) steady-state failure rates. However, depending on the
choice of attachment technology (leadless versus leaded), the design parameters,
and the severity of the use environment, wearout failures can start to assert
themselves about the same time as the components reach the steady-state phase.
Depending on the rigidity of the SM attachment (leadless or leaded), the wearout
failure rates will rise to exceed the steady-state failure rate of the component.
While, to perhaps "mid-life," components might dominate systems' failures,
solder joint failures could be the determining failure mode toward the end of the
design life.
From Figure 17-2, which shows the information from Figure 17-1 in terms
of the cumulative component failare probability versus operating time, it is clear
that for the first couple of years of operation the reliability, or more accurately
any lack ofreliability, is dominated by component failures, albeit at low failure
 SOLDER JOINT LOADING CONDITIONS AND RELIABILITY 549

99.9
"TYPICAL" SM
SOLDER
ATTACHMENT
50 LEAD LESS LEADED
I-~

~
z~
w- 10
z>-
ot::
c..-J
::!;1D
0«
(.)al
wo
>cc "TYPICAL" ELECTRONIC
-0.. COMPONENT
0.1
~~
::J::J
::!;:::!
::J«
(.)LL
0.01

0.001

0.0001
1 10 10 3
OPERATING TIME (h)

Figure 17-2 Cumulative failure probabilities for a "typical" electronic component and a
"typical" surface mount solder attachment. The cumulative failure probability for solder-at-
tached components is the sum of the components and attachment curves.

probabilities of less than about 0.5 percent. If premature wearout is not prevented
by an adequate "design for reliability" that takes into account the use environment
and the intended life, then failures due to wearout of the surface mount solder
attachments can cause rapidly increasing failure probabilities signifying the end
of useful life before reaching the desired product life. The cumulative failure
probability of a surface-mount-solder-attached electronic component is the sum
of the instantaneous failure probabilities of the component without attachment
and of the surface mount solder attachment. The level of failure probability that
is acceptable depends on the criticality of function of the system and the number
of components in the system.

17.2 SOLDER JOINT LOADING CONDITIONS AND RELIABILITY

The reliability of the solder joint attachments of electronic components surface

mounted to circuit board substrates requires explicit attention in the design phase
as well as during manufacturing. Surface mount (SM) solder joints in use can
be subjected to a variety of loading conditions that can lead to premature failure
in inadequate designs. The following loading conditions could exist either singly,
sequentially, or simultaneously: (1) cyclic differential thermal expansion, (2)
vibration (transport and mobile use), (3) thermal shock (rapid temperature change
causing transient differential warpages) during cooling from soldering (reflow)
550 SOLDER ATIACHMENT RELIABILITY, ACCELERATED TESTING

operations or from severe use environments, and (4) mechanical shock (high
acceleration) from accidental misuse or severe use conditions.
While vibration, thermal shock, and mechanical shOCk do occur in practical
applications, they are the exceptions rather than the norm in terms of the main
cause of failure; the failure mechanism of primary concern for SM solder joint
reliability is cyclic differential thermal expansion that results in fatigue damage.
It should be noted that the resistance against failures induced by vibration, thermal
shock, and mechanical shock, as well as low operating temperature cycles de-
pends primarily on solder joint strength because these failures are stress-driven.
This is not the case, however, for fatigue failure resistance at typical operating
temperatures. Fatigue failures, which are primarily strain-driven, are prevented
by minimizing the cyclic fatigue damage by reducing or/and accommodating the
cyclic expansion differences.
Consequently, this chapter deals primarily with the long-term fatigue reliability
of SM solder joints. While these different loading conditions-and the associated
damage mechanisms-need to be considered separately, their cumulative damage
needs to be combined for complete reliability estimates (see the section "Multiple
Cyclic Load Histories").
SM solder joints can be subjected to considerable cyclic strains during operation
caused by different thermal expansions of:

1. The surface mounted components and the substrates to which they are
solder-attached
2. The solder and the materials (component, lead, substrate) to which the
solder is bonded

The thermal expansion differences for the first cause result from differences in
temperature due to power dissipation internal to the component in combination
with system internal or external temperature variations. These temperature vari-
ations are caused by component or system load fluctuations or on/off cycles, by
diurnal cycles, and by seasonal changes. Differences in the coefficients of thermal
expansion (CTE) of the materials involved in the solder attachment system
together with these temperature variations result in the thermal expansion dif-
ferences.
Appropriate design for reliability measures for the first cause can take one of
two forms, which are best employed in combination for improVed reliability
margins. These measures are:

1. Tailoring the CTEs of substrate or/and component

2. Increasing the attachment compliancy

The thermal expansion differences for the second cause result from the dif-
ference in the CTEs of the solder itself and the material(s) to which it is bonded.
 RELIABILITY AND ACCELERATED TESTING-OVERVIEW 551

While this cause can have a significant contributory effect to cause 1, by itself
it becomes a reliability threat only when the .:1CTE is large and the use envi-
ronment is very severe. The appropriate design for reliability measure, besides
reducing .:1CTE, is keeping the solder joint attachment (interface) dimensions
as small as possible. It should be noted that this failure cause is stress-driven
(i.e. it results from overstress) and is compounded by the fact that the highest
stresses from this cause occur at the solder joint-metallization pad interface that
typically consists of brittle intermetallic compounds of tin and the metallization
metal.
Assessing the reliability potential of a design and assuring the reliability of
an electronic assembly in use requires knowledge of the use environment(s), the
resulting loading conditions and the accumulating fatigue damage. It further
requires an analytical relationship that relates the accumulating fatigue damage
to cyclic fatigue life in general, generic terms with the option of including specific
characteristics of the design in question that might also affect reliability. This
empirical model and, especially, the non-generic design-specific characteristics
need to be verified or established, which requires a well-designed and coordinated
accelerated reliability test program. Lastly, an acceleration transform is necessary
that allows valid comparisons of results from different accelerated test conditions
and that, more importantly, allows the extrapolation of these accelerated test
results to predict the reliability of the product in use. Given the time, temperature,
and stress dependent behavior of solder, these relationships are anything but
straightforward. Furthermore, some of the reliability and accelerated testing
consequences are not intuitively obvious and can be rather subtle. In particular,
the connection between accelerated test results and field reliability prediction is
frequently misunderstood. These aspects will be discussed in this chapter.

17.3 RELIABILITY AND ACCELERATED TESTING-OVERVIEW

In Figure 17-3 the steps necessary for SM solder attachment reliability assurance
are given in the form of a flow chart.
A generic understanding of reliability and the underlying failure mechanisms
is the first step toward designing and assuring product reliability. For this, a
generic database is required, one that is based on a combination of low-accel-
eration (acceleration factors of - 10 to 20 in terms of mean-time-to-failure)
fatigue tests that closely mimic expected thermal cyclic field conditions and high-
acceleration (acceleration factors - 100 to 500) "quick and dirty" fatigue tests.
Thus, in the low-acceleration tests all effects that have an impact on long-term
product reliability should be present and only the effects of the reduced dwell
times at the cycle extremes need to be scaled to estimate product reliability from
the accelerated test results. The high-acceleration tests are a compromise between
the realities of insufficient time and limited resources that do not permit the more
552 SOlDER ATIACHMENT RELIABILITY, ACCELERATED TESTING

GENERIC TECHNOLOGY
CHARACTERIZATION
WITH TEST VEHICLES
INITIAL PROCESS
QUALIFICATION
WITH TEST VEHICLES
I
r ONGOING PROCESS
VERIFICATION
WITH COUPONS
PRODUCT
SPECIFIC
ACTIVITIES

I DEFINITION OF

I USE ENVIRONMENT
& DESIGN LIFE

I
RELIABILITY DESIGN INFORMATION

INITIAL aUALIFICA TlON

FEEDBACK

PARAMETERS I
I
I
I SHIP/REJECT I -----
I

Figure 17-3 Flow chart for surface mount solder attachment reliability assurance, indicating
the need for a generic (thin outline) technical understanding to develop the tools for a design
for reliability, and the need for product specific (heavy outline) design for reliability as well
as for ongoing process verification and product control.

accurate low-acceleration tests to be performed on all product and the need,

sometimes based more on emotional rather than technically justifiable arguments,
to have reliability data on all possible product variations. However, the test
limitations and the much higher potential for inaccurate, even misleading, results
must be clearly kept in mind.
For the accelerated tests, it is imperative that the failures occur as the result
of accelerating the proper failure mechanisms and not as the result of a failure
mechanism unique to the test conditions. Thermal shock testing, in particular,
does not accelerate the long-term fatigue mechanisms of concern for most SM
applications and can produce misleading results. In this context thermal shock
is defined as a temperature change rapid enough to cause transient unequal or
even opposite warpages of components and substrate due to the transient tem-
perature gradients caused by the rapid transfer of the test vehicles from hot to
cold test environments and vice versa. The underlying driving mechanism causing
most of the long-term solder joint fatigue damage in product in operation results
from the creep and stress relaxation behavior of solder. Therefore, accelerated
fatigue testing must include well-defined dwell times at the cycle extremes to
allow for some, albeit incomplete, creep and stress relaxation. The incomplete
cyclic creep and stress relaxation inherent in accelerated test cycles is in contrast
to actual field operational environments, where sufficient time is usually available
for essentially complete creep and stress relaxation. Thus, the significantly smaller
 THE THERMAL EXPANSION MISMATCH PROBLEM 553

fatigue damage resulting from accelerated test cycles must be accounted for with
a fatigue acceleration transform.
It also needs to be understood that the strongly temperature, time, and stress-
level dependent material behavior of many of the materials involved in electronic
packaging makes accelerated tests with temperature extremes substantially dif-
ferent from actual cyclic use conditions questionable at best. Temperatures near
the dielectric glass transition temperatures (> + l(xtc for FR-4) or low tem-
peratures where stress relaxation for solder becomes negligible « - 20°C) render
results from tests employing temperatures outside this range useless and fre-
quently misleading for most actual applications.
The generic understanding generated from the accelerated fatigue tests must
be translated into reliability design information that can be used for design for
reliability. The use environment, the expected design life, and the acceptable
probability of failure at the end of this design life must be established to set the
reliability goals for the design.

17.4 THE THERMAL EXPANSION MISMATCH PROBLEM

The reliability problem caused by the mismatch in thermal expansion of com-

ponent and substrate, given the cyclic variations in the thermal conditions during
functional use, results not so much from the cyclically recurring strains in the
solder joints but from the fact that solder responds to applied strains and stresses
by time dependent, albeit relatively rapid, plastic deformation. Metals undergoing
cyclic strains sustain cyclically accumulating fatigue damage. Fatigue damage
due to plastic strains is significantly larger than the damage from elastic strains.
Solder at typical operating temperatures is being used significantly in excess of
60 percent of its absolute melting point. Thus, solder is subject to significant
stress relaxation and creep, processes that convert elastic strains to plastic strains
but which are strongly time, temperature, and stress dependent. Thus, in all but
very rapid cycling without dwell or hold times and/or operation at low temper-
atures « - 20°C), plastic deformation dominates and determines the fatigue
life of solder joints. The cyclic fatigue damage is proportional to the area cir-
cumscribed by the cyclic hysteresis loop in a stress-strain diagram.4-9 The re-
liability hazard can be reduced by decreasing the hysteresis loop size; reducing
the total stress range, the total strain range, and the magnitude of the cyclic
plastic strains accomplishes this reduction. The importance of the various design
parameters affecting reliability to first-order and the necessary trade-offs between
sometimes interdependent and competing design parameters are shown in the
section "Analytical Model of Solder Shear Fatigue."
Methods to reduce fatigue damage are CTE-tailoring and increased attachment
compliancy. CTE-tailoring reduces the cyclic hysteresis loop primarily by re-
ducing the total cyclic strain range, while attachment compliancy primarily works
554 SOLDER AITACHMENT RELIABILITY, ACCELERATED TESTING

by reducing the cyclic stress level at the solder joint. These methods are described
below.

17.4.1 Solution 1: eTE-Tailoring to Reduce Expansion Mismatch

In CTE-tailoring, the coefficients of thennal expansion of substrate and/or com-

ponents are tailored, by deliberate design or material choice, to result in the
minimum steady-state component-substrate expansion differential resulting from
the expected temperature variations and temperature gradients. 9 This is signifi-
cantly different from CTE-matching-attempting to make the CTEs of substrate
and components equal-which was forced upon vendors in order to meet un-
realistically extreme MIL-spec temperature cycling acceptance criteria (e.g.,
- 55 to + 125°C cycling)1O and which results in less than optimum reliability
for actual use conditions. It also needs to be realized, however, that the tailored
component-substrate CTE difference can only be optimum for one component-
preferably the component representing the largest reliability hazard due to large
size, leadless attachment, large power dissipation, etc.-and has to be subop-
timum for all others. For these other components, the reliability hazard should
be lower or equal due to other design features, such as smaller size and compliant
leads. Solder joints of even perfectly CTE-tailored component-substrate com-
binations will ultimately fail in fatigue because:

1. Material variations (patch-to-patch, x-y directionality) in product produce

ranges of CTE that can result in significant CTE-mismatches.
2. CTE-tailoring does not eliminate transient cyclic strains. 6
3. The expansion mismatch between the solder (-25 ppm/DC) and the material
to which it is attached 11
4. The expansion mismatch of the a and 13 phases within the solder itself

While these causes of fatigue damage can result in sizeable damage contributions,
they are not likely to cause failures by themselves within the design life of a
typical product. This is true with the exception of very large CTE mismatches
between the solder and the attachment material in combination with severe use
environments.
For the more severe use environments (e.g., Use Category 9, Transportation-
Under Hood, see section "Testing Considerations"), it is possible that the reli-
ability threat comes from the expansion mismatch between the solder itself and
the materials to which it is bonded. The reliability problem needs to be assessed
using an interfacial stress analysisl2 and appropriate accelerated testing.
For ceramic components mounted to ceramic substrates, the steady-state CTE-
tailoring occurs virtually automatically since the CTEs for ceramic chip carriers
and ceramic substrates are typically in the ranges of 5.4 to 6.7 and 6.0 to 8.3
 THE THERMAL EXPANSION MISMATCH PROBLEM 555

ppml°C, respectively.6 Active components, whether forced air or conduction

cooled, always operate at a temperature higher than the substrate; therefore, the
substrate CTE should be somewhat higher than the component CTE. Fine CTE-
tailoring and CTE-matching, while theoretically possible, are in practice very
difficult due to the inherent CTE variations in manufactured parts. However, the
limitations of ceramic substrates-small size due to brittleness, electrical losses
due to dielectric properties-provide strong incentives for organic-based sub-
strates. 13 Using constraining layers (e.g., copper-clad Invar, copper-Clad mo-
lybdenum, copper-infused molybdenum,14 graphite fibers, Kevlar fibers), the
CTE for FR-4 epoxy-glass substrates (which can range from 16 to 24 ppml°C,
depending primarily on resin content) has been reduced to as low as 8.6 ppml°C
by substituting copper-clad Invar layers for the copper power and ground layers6.15
and 6 ppml°C by substituting Kevlar for the glass fiber reinforcements. 16
Conversely, plastic leaded components have a CTE range (20 to 23 ppml°C)
falling within the CTE range of FR-4 substrates. Therefore, like ceramic-on-
ceramic assemblies, plastic components are inherently CTE-tailored. Plastic leaded
components in addition have leads that provide varying degrees (dependent on
design and material) of compliancy16--21 to accomodate expansion mismatches.
Compliant leaded attachments are described in depth below.

17.4.2 Solution 2: Attachment Compliancy to Accommodate Expansion

Mismatch
Attachment compliancy ll.16,19,21 comes primarily from suitably designed com-
ponent leads, and to a much smaller degree can be assisted by compliant surface
pad layers on the substrate. 22 For designs requiring hermetic (ceramic) chip
carriers mounted directly to conventional (not CTE-tailored) FR-4 substrates,
the large differential expansions must be accommodated by the compliancy of
the attachment design. The solution of choice is the attachment of the ceramic
chip carrier to the substrate using edge clip-leads, but even highly compliant
lead designs are inadequate for large expansion mismatches and long cyclic
lives. 7 •11 In addition, highly compliant leads can present manufacturability prob-
lems due to easy accidental bending of the leads and functional problems due
to high electrical losses at high signal speeds from the long lead length.
Leads reduce the cyclic fatigue damage in the solder joints by reducing the
cyclic stresses in the solder. A differential expansion between component and
substrate causes strains and stresses in the substrate, the component body, the
leads, and the solder joints. The strains in the substrate, component, and leads
are primarily elastic with the elastic energies stored by warpage and bending. 23,24
In solder, the elastic strains are converted to plastic strains with time via stress
relaxation and creep, processes that are accelerated at higher temperatures and
stress levels. At typical use temperatures, the solder rapidly responds to stresses
556 SOLDER ATIACHMENT RELIABILITY, ACCELERATED TESTING

and displacements by visco-plastic flow and plastically defonns not only by

converting the elastic strains within the solder joint but also by yielding to the
elastic energies stored in substrate, component body, and leads.
The primary purpose of a compliant lead is to reduce the stress level at the
solder joint significantly below the solder yield strength to reduce the cyclic
hysteresis loop area and thus the cyclic fatigue damage. It needs to be emphasized
that compliant leads do not reduce the expansion mismatch. The compliant leads
elastically store strain energy, which over time unloads into the solder joints due
to creep and stress relaxation. Thus, the total strain at the solder joint is not
affected; the maximum stresses, however, are significantly reduced. At lower
stress levels, the visco-plastic flow slows significantly to a degree that the con-
tention has been advanced that a threshold stress level exists below which creep
and stress relaxation stop. Unfortunately, the often quoted threshold stress level
for solder creep of - 300 psi can readily be demonstrated by simple experiments
to be nonexistent. An added reliability benefit can result, however, if stress levels
are reduced to a point where the conversion of the elastically stored strain energies
to plastic defonnations in the solder, due to stress relaxation or creep, is retarded
to such an extent that stress relaxation is still significantly incomplete even for
operational cycles.

17.5 ANALYTICAL MODEL OF SOLDER SHEAR FATIGUE

Analytical fonnulations for solder joint fatigue are developed in this section.
The parameter nomenclature and units, where appropriate, are provided in con-
junction with the equations where they first occur.

17.5.1 Solder Joint Fatigue

The long-tenn reliability of a SM solder attachment is governed by the difference

between the required design life and the cyclic fatigue life of the solder joints
as determined by the combined effects of component design, assembly design,
and use environment. The cyclic fatigue life of a solder joint is determined by
the amount of cyclically accumulating fatigue damage. 6 Solder joint rupture
occurs when the total accumulated damage exceeds the capability of solder to
sustain such damage.
The solder joint response to cyclic displacements is characterized by a hys-
teresis loop in the shear stress-strain plane as shown in Figure 17-4 in stylized
fonn. The area of this hysteresis loop is representative of the visco-plastic strain
a
energy density per cycle W and thus the fatigue damage accumulating per
cycle. For leadless SM solder attachments, the hysteresis loop is limited by a
constant stress envelope-independent of the thennal expansion mismatch-
determined by the solder yield strength 'T y during the initial elastic loading and
 ANALYTICAL MODEL OF SOLDER SHEAR FATIGUE 557

SHEAR STRESS
= LEADLESS

=== LEADED
s
= FIELD USE

:::: ACCELERATED TESTING

SHEAR

STRAIN

I---A~---J

Figure 17-4 Stress-strain response (stylized) of solder joints in both lead less and leaded
surface mount attachments to identical total cyclic displacements. The areas of the cyclic
hysteresis loops are indicative of the cyclically sustained fatigue damage. Thus, the difference
between lead less and compliant leaded SM attachments as well as between field use cycles
and accelerated test cycles with incomplete creep-stress relaxation is illustrated (from references8,9).

plastic yielding and by the stress reduction lines during creep and stress relax-
ation. It should be noted that unlike the yield strengths of common engineering
metals, the yield strength of solder is highly dependent on temperature and strain
rate. For typical leaded SM solder attachments, the maximum solder joint stresses,
S are significantly below the solder yield strength and depend on the thermal
expansion mismatch. Thus, the stress reduction lines, except for some initial
elastic loading, determine the hysteresis loop in the stress direction. For both
leadless and leaded SM solder attachments, the maximum shear strains A.-y are
determined by the displacements resulting from the thermal expansion mismatch
between component and substrate. Due to stress relaxation and creep, which
allow the elastically stored strain energy in the compliant leads to unload into
the solder joints with time, there is no difference in the total displacements at
the solder joints. These hysteresis loops are illustrated in Figure 17-4. It should
be noted that symmetric hysteresis loops as shown in the figure result only if
the cyclic strains are mechanically applied at isothermal conditions or if the dwell
time at To is longer than at To + A.Te to compensate for the slower stress relaxation
at the lower temperature.
A generalized fatigue damage law for metals has been proposed on the basis
of cumulative stored visco-plastic strain energy density. 4 The cyclic shear fatigue
life Nt is related to A.W in a stabilized fatigue cycle by the equation

(17-2)
558 SOLDER ATIACHMENT RELIABILITY, ACCELERATED TESTING

where c = fatigue ductility exponent

Nf = mean fatigue life [N.t<50%)], cycles-to-failure
W} = intercept energy term, a material constant
.::lW= visco-plastic strain energy density per cycle

The exponent c is in the range of -0.5 to -0.7 for most metals. The well-
known Manson-Coffin plastic strain-fatigue life relationships can be directly
derived from-and is a special stress limited case of-this generalized fatigue
damage function and is

- - -1[.::l-Yp]!
N - c (17-3)
f - 2 2E}

where .::l-yp = cyclic plastic shear strain range

E} = fatigue ductility coefficient
For metals used in temperature ranges where time and temperature dependent
creep and stress relaxation become significant relative to the initial plastic strains
caused by yielding (typically in excess of 60 percent of the absolute melting
temperature), the determination of either .::lW or .::lyp is not straightforward. The
total plastic strains will increase with time as strain energy elastically stored in
the component-lead-solder joint-substrate structure system will be converted to
cumulative irrecoverable visco-plastic strain energy in the structure member
(solder) that creeps/stress relaxes. With sufficient time, which for solder at
operating temperatures can be quite short, virtually all the elastically stored strain
energy accumulates as visco-plastic strain energy in the solder joints. This max-
imizes the cyclically circumnavigated hysteresis loop in the shear stress-strain
plane and thus maximizes the cyclic fatigue damage to the solder joint (see Fig.
17-4).
The fatigue life of solder joints is not dependent on cyclic frequency per se;
frequency can, however, appear as a correlative variable if data from cyclic tests
with identical cycle shapes (sinusodial, saw-tooth, trapezoidal, square, etc.) are
compared, since in these cases the cycle half-period and the time available for
stress relaxation/creep are directly related.
In operational use the cycles typically consist of a ramp-up, an operational
dwell period, a ramp-down, and a dwell period, which depending on the appli-
cation can be operational or non-operational. Superimposed on the system op-
erational cycles might be cyclic temperature variations due to daily and seasonal
outside environmental conditions. The ramp-up will depend on the heat transfer
and power dissipation conditions of the system and is of reliability concern only
if it leads to temperature changes rapid enough to cause thermal shock (see the
 ANALYTICAL MODEL OF SOLDER SHEAR FATIGUE 559

section "Reliability and Accelerated Testing-Overview") or if it is slow enough

for significant stress relaxation to take place during this transient period. The
operational and non-operational dwell periods depend on the operational char-
acteristics of the system for both its duration and constancy; it can have super-
imposed on it more rapid cyclic thermal variations due to system load fluctuations.
The down-ramp is a near-exponential decay depending on the system heat transfer
parameters .
Thus, frequency is not necessarily a measure of the dwell time and using
frequency instead of dwell time in correlating accelerated test data can result in
significant errors in the analysis. Frequency also does not provide any indication
of the rate of change of the thermal conditions; solder can be quite strain-rate
sensitive at higher rates of change.
For accelerated fatigue testing, the dwell times during the cycle half-periods
are insufficient for full stress relaxation to take place (see Figure 17-4); it is the
shortened dwell times that provide the test acceleration but that also require that
the reduced cyclic fatigue damage during accelerated testing be accounted for.
Test acceleration by increasing the thermal cyclic loading is only possible to
a limited extent since the material characteristics of the solder limit both the
maximum and minimum test temperatures that can be applied without introducing
damage mechanisms that do not correspond to the field operational damage
mechanisms. For leadless SM solder attachments, analysis is relatively simple
since, unlike the response of leaded attachments, it is not complicated by inter-
acting effects between the solder and the compliant lead structures.

17.5.2 Leadless Solder Attachments

A model for the mean cyclic fatigue life for the solder joints of a leadless surface
mounted component (or a component with very stiff leads) has been derived
from isothermal experimental data25 (see Fig. 11-5). This model is expressed
by the relationship6

(17-4)

where for eutectic and 60% Sn-40% Pb solde~

2£} = 0.65 (17-5)

and

c = -0.442 - 6 x 1O- T
4 sJ + 1.74 3:0)
x 1O-2In( 1 + t (17-6)
560 SOLDER ATIACHMENT RELIABILITY, ACCELERATED TESTING

100r-----,------r-----,------~---- __----_,

~
=- = =ISOTHERMAL LAP
;;:: SHEAR FATIGUE DATA
<l EUTECTIC 63% Sn - 37% Pb
LlJ
(9 10 (R. WILD, IBM)
z
<::
a:
z
«
a:
t-
UJ
a: =- 2S"C, tD = 3 SEC
<::
LlJ
I
= 100"C, 3 SEC
UJ
...J
0.1 = 2S"C,
~
~
0.01
1 10 10 5 10 6
MEAN CYCLES TO FAILURE (fi,)

Figure 17-5 Manson-Coffin plot showing the results from reference 25 for isothermal fatigue
cycling of solder joints at different temperatures and cyclic dwell times and some extrapolated
results based on Equation (17-4).

and where Ts , Tc = steady-state operating temperature for substrate, com-

ponent (Tc > Ts for power dissipation in component)
TSJ = 1 (Tc + Ts + 2To), mean cyclic solder joint temperature
To = temperature during off half-cycle
tD = half-cycle dwell time (min), average time available for
stress relaxation at TsITc and To
fl:y = cyclic total plastic shear strain range in solder after
complete stress relaxation
E; = fatigue ductility coefficient, defined for solder in
Equation (17-5)

While Equation (17-4) has the same form as Equation (17-3), it overcomes
the impractical difficulty of determining the time and temperature dependent
plastic shear strain range !:l.yP by using the maximum available displacement
shear and accounting for the incomplete degree of stress relaxation/creep with
a time and temperature dependent exponent.
The effect of time and temperature can be clearly seen in Figure 17-6 where
the experimental results from isothermal mechanically applied strain cy-
cling 17- 19 are shown for temperatures of 23 and 81°e and cyclic dwell times
ranging from 3 to 180 sec. The cycle shape for these data was trapezoidal with
constant strain rates.
The total cyclic strain range is given by
 ANALYTICAL MODEL OF SOlDER SHEAR FATIGUE 561

TEST TEMPERATURE: 8 1°C

o
W
...J

Lt
(/2.

480-14400 4400 cld TEST FREQUENCY

1~----------~----------~~--------~
3
10 4 10
5 6 10 10
FATIGUE LIFE, CYCLES TO FAILURE
Figure 17-6 Weibull statistics plot of test results from accelerated isothermal mechanical
cycling tests of plastic leaded chip carriers employing two temperatures and near square wave
cycle shapes. The tests were carried out with three similar components indistinguishable in
their fatigue lives over cyclic dwell times ranging from 90 to 3 sec (f = 480 to 14,400 cycles/
day). The test data have similar Weibull slopes indicating the same failure mechanism and
show the effect of different degrees of the time and temperature dependent cyclic stress relax-
ation in the solder joints on the cyclic fatigue lives (from references 17- 19 ).

(17-7)

where the thermal expansion mismatch is

(17-8)

and where F empirical "nonideal" factor indicative of deviations of real

solder joints from idealizing assumptions and accounting for
secondary and frequently untractable effects such as cyclic
warpage, cyclic transients, nonideal solder joint geometries,
solder microstructure, brittle intermetallic compounds, Pb-rich
boundary layers, and solderlbonded material expansion
differences, as well as inaccuracies and uncertainties in the
562 SOLDER ATIACHMENT RELIABILITY, ACCELERATED TESTING

parameters in Equations (17-7) and (17-8); 1.5 > F > 1.0

for column-like leadless solder attachments, 1.2 > F > 0.7
for castellated and chip component leadless solder
attachments with fillets, F = 1 for solder attachments
utilizing compliant leads
h = solder joint height, for leaded attachments h == 112 of solder
paste stencil depth
2LD maximum distance between component solder joints, see
Equations (17-13) and (17-14)
as,ae coefficient of thermal expansion (CTE) for substrate,
component
tl.Te cyclic temperature swing for component
tl.Ts cyclic temperature swing for substrate (at component)
tl.(atl.T) absolute cyclic differential thermal expansion between
component and substrate, see Equation (17-8)

Some values for coefficients of thermal expansion of materials common is elec-

tronic packaging are given in Table 17-1.
The right side in Equation (17-8) can be simplified to

(17-9)
where

(17-10)

and where tl.Te equivalent cyclic temperature swing, accounting for

component power dissipation effects as well as component-
external temperature variations
absolute difference in coefficients of thermal expansion of
component and substrate, CTE-mismatch, tl.CTE

Equation (17-10), of course, simplifies in the case of no power dissipation in

the component to
(17-11)

Thus, from combining Equations (17-4), (17-7), and (17-9), the mean cyclic
life of a leadless surface mount solder attachment is obtained from

(17-12)
 ANALYTICAL MODEL OF SOLDER SHEAR FATIGUE 563

Table 17-1 Some Measured Planar Linear Thermal Expansion Coefficients

Planar Linear Thermal
Component, Substrate, * Expansion Coefficients, **
Material a (ppmrC)
Plastic chip carrier 20-23
Ceramic chip carrier 5.4-6.7
FR-4 laminate 12-24
FR-4 multilayer board 16-24
CTE-modified FR-4 MLB 7.2-14
Ceramic multilayer board 6.0-8.3
Solder 23-25
'The specific construction (number of layers, resin/glass/metal content, etc.) has a major impact on a.
"Most values were measured between room temperature and 110°C.

which contains all the essential first-order design parameters. Figure 17-7 shows
in a Manson-Coffin plot the results from three different test programs that cor-
relate well with the prediction from Equation (17-12) using F = 1.0. A value
of F = 1.27 would virtually match the model prediction and the experimental
data.

100
- TEMPERATURE CYCLING
;l (L = 650mlls, h .. 8mlls, FR-4 MLB
~ 61: = 170,150130,100, BO'C)
t.O W. M. SHERRY - AT&T-Bell Labs/AL
<l
"'0'> TEMPERATURE CYCLING
LU- = =
L 650mlls, h 18.2mlls, FR-4 MlB
CJ II T = 134"C)
Z J. FISHER - ATIoT-T/ERC
«a: PRE ~ POWER CYCLING

z 10 L.!:=t::!:~:!~SD~'~C~T bL = 950mlls, h = 4mlls, FR-4 DSR &

I ~ W~~:GRE-lt~~~,_PAT&~·_3B~1 Labs/WH
<i:
a:
I-
CI)
a:
«LU
I
CI)

1
10 1 10 2 10 3
CYCLES TO FAILURE (N1)

Figure 17-7 Manson-Coffin plot showing the results (mean and range) from three accel-
erated functional and thermal cycling test programs for lead less ceramic chip carriers of different
sizes on different substrates over widely varying test cQ.nditions. To faci Iitate comparison, the
test results are normalized to a common condition of TSI = 45°C and tD = 6.2 min (f = 58
cycles/day) using the acceleration transform [Eq. (17-18)]. The model prediction [Eq. (17-4)]
for this common condition indicates excellent agreement between model and experimental
data with a non ideal factor [1.12 .;;; F.;;; 1.27, Eq. 17-7] offset (from reference6 ).
564 SOLDER ATIACHMENT RELIABILITY, ACCELERATED TESTING

The maximum distance between solder joint centers for rectangular or square
components with a multiplicity of solder joints, (e.g., chip carriers, SOles, etc.)
is approximately given by

(17-13)

where L = longest component dimension between component edges with

solder joints
W L for square components
- for rectangular components, the shorter component dimension
between component edges with solder joints or the component
dimension between component edges without solder joints

For components with two solder joints, such as discrete chip components and
metal electrode face bonded components (MELFs) the formulation

(17-14)

can be used

where LM = distance between metallizations for solder joints

For temperature cycling between low temperatures «O°C), where stress re-
laxation and creep are insignificant, the cyclic life can be determined from
Equation (17-3). Of course, the plastic strain range A"'(p is in this case significantly
less than the total strain range [Eq. (17-7)]; the elastic strain range, determinable
from the low temperature stress-strain behavior of solder, needs to be subtracted.

17.5.3 Leaded Solder Attachments

The effects of lead compliancy on the fatigue life make the analytical determi-
nation of the cyclic life of leaded solder attachments significantly more com-
plicated and these effects had until very recently7 not been fully quantified.
Techniques to determine the cyclic strain energy density in the solder joints of
leaded SM attachments have been developed. 26 From the results of accelerated
tests that mimic functional use conditions, the cyclic life of the product in the
field has been estimated II by extrapolation using the acceleration transform in
Equation (7-18). For leadless SM attachments, the high component and substrate
stiffnesses ensure rapid and near complete stress relaxation at most use temper-
 ANALYTICAL MODEL OF SOLDER SHEAR FATIGUE 565

atures, even at the cyclic low temperature excursions (provided the temperatures
are not too low and the cycles too short). Stresses are large but are limited by
the solder yield strength. In contrast, as shown in Figure 17-4, for leaded SM
attachments with adequate lead compliancy, the stresses at the solder joints are
reduced to levels at which solder creep is significantly retarded even at the cyclic
high temperature excursions and for long operational dwell times. Thus, the
stress level effects on the solder joint creep become a determining factor for SM
solder joint reliability.
The solder joint response is characterized by a hysteresis loop in the shear
stress-strain plane between the stress reduction (combination of creep and stress
relaxation) lines at the temperature extremes. The product of the largest applied
stress and the total cyclic shear strain range is the maximum available cyclic
strain energy density ~W, which is proportional to the product S~"Y (see Fig.
17-4). The largest applied stress is proportional to

(17-15)

where A effective solder joint area (= 2/3 solder-wetted lead area

projecting to solder pad)
"diagonal" flexural stiffness of unconstrained component lead
determined by strain energy methods or finite element analysis
S = largest stress applied to a leaded attachment solder joint

The magnitude of S controls the stress reduction rates during the dwells at
constant temperature. From Equations (17-7), (17-9), and (17-15), the cyclic
strain energy density (a measure of the cyclic fatigue damage) is, for surface
mount solder attachments with compliant leads, proportional to

(17-16)

A plot of the cyclic strain energy density per cycle ~W versus mean cyclic
life Nf is shown in Figure 17_8. 8 •9 The experimental data are for five different
leaded assemblies subjected to accelerated testing in functional ll or temperature
cyclingY The slope of the regression line ~W versus Nf is lei < 0.64 [see Eq.
17-2)]. For the three assemblies with the longest lives, lowest bound estimates
for Nf had to be used since no failures had been observed at the indicated numbers
~ cycles. Fatigue testing to failure undoubtedly will result in larger values of
Nf resulting in a reduced slope in Figure 17-8 and a value for the exponent c
566 SOLDER ATIACHMENT RELIABILITY, ACCELERATED TESTING

Zc
... 10 ~--~~---1~----------~----------~--------~
~.
<~
a:=
f-3:
OO<J
g~
(.)-
>-00
(')Z
:E W 1 SOIC-GW (24 VOl I Kevl.r
::JO 2 LCCC (100110 J.Ci.p)/FR ••
:E>-
-C) 3 PLCC (a81/01/FR:.
XCI: • SOJ (21 I/OI/FR ••)
<Cw 5 LCCC (100110 S·CllpI/FR ••
:Ez
W
0.1 ~-L __ ~ __-L__ ~ __ ~ __ ~-L __ ~ __ ~ __ L-~~~

10 3 10· 10 5 10' 10 7
MEAN ACCELERATED CYCLES TO FAILURE
Figure 17-8 Plot of test results from accelerated functional and thermal cycling tests for
five different leaded surface mount component assemblies. Since for three tests no failures
occurred during testing, only the Nt (50%) lives are plotted. For tests 3 to 5 the projected lives
are calculated from Equation (17-19) under the assumption that the first failure occurs on the
very next cycle after test termination. For test 5 a life projection based on the test results from
test 2 and Equation (17-18) is also shown (from references8,9,11,27).

10'~--------'----------r--------~--------~
");,"'v..... 23°C DATA NORMALIZED TO 81"C
Q. SO"C DATA NORMALIZED TO 81"C

~
oJ
105~--------+---------~---------+------~~
W TEST TEMPERATURE: 81°C
::J
C)
o LSI
~
u; eNSC
VTI
~
~ 10·~--------4---------~~~~---+--------~

103L-________ ________-L________
~
- - - MODEL ________
PREDICTION ~ ~

10 102 103 10· 10 5

TEST FREQUENCY (cycles/day)

Figure 17-9 Plot of test results from accelerated isothermal mechanical cycling tests of
plastic leaded chip carriers employing three test temperatures and near square wave cycle
shapes. The tests were carried out with three similar components indistinguishable in their
fatigue lives over cyclic dwell times ranging from 10 min to 3 sec (f = 72 to 14,400 cycles!
day). The test data are normalized to a comparison temperature of BlOC using the acceleration
transform [Equation (17-18)]; the test results and the model prediction [Eq. (17-17)] show
excellent correlation (from references I7- 19).
 ANALYTICAL MODEL OF SOLDER SHEAR FATIGUE 567

closer to the more commonly observed values (c - -0.5 to -0.4) for solder.
This is clearly shown to be the case in Figure 17-9, where the results from the
IEEE Compliant Lead Task Force 17- 19 ,28 are plotted against the model prediction
resulting in excellent correlation between the data and the model. Thus, from
combining Equations (17-4) and (17-16) and benchmarking the proportionality
in Equation (17-16) with the results in Figure 17-8, the mean cyclic life of surface
mount solder attachments with compliant leads is given by the equation

(17-17)

which contains all the essential first-order design parameters.

17.5.4 Acceleration Transform

The differences in the cyclic hysteresis loops-and thus the cyclic fatigue dam-
age-resulting from accelerated test cycles with insufficient dwell times for
complete stress relaxation and product functional cycles resulting in essentially
complete stress relaxation are shown in Figure 17-4. This difference in fatigue
damage per cycle needs to be accounted for with an acceleration transform in
order to make product reliability estimates based on accelerated test results.
From Equations (17-2) through (17-6) an acceleration transform 6 results with
which accelerated test results, denoted by (t), can be extrapolated to use con-
ditions, denoted by (u):

- (u) c !lD u J2.

Nf = 2:1[ [2 NIt)]
- t !lD
t
c. (17-18)

where

(17-19)

from Equation (17-7) leadless attachments and

!lD = aw (17-20)

from Equation (17-16) for leaded attachments. It should be noted that (u) and
(t) can also denote different accelerated test conditions; transforming the results
obtained at conditions (t) to equivalent (u) condition results permits meaningful
comparisons of test data. In Equation (17-18), the cyclic fatigue damage is in
ratio form; thus, it is not necessary that the values of !lD, the measure of cyclic
568 SOLDER ATTACHMENT RELIABILITY, ACCELERATED TESTING

fatigue damage, be explicitly known. Only the ratio relationship between test
and use fatigue damage is needed. This makes it much easier to transform data
for surface mount solder attachments since many and in the case of functional
accelerated testing (see section "Accelerated Fatigue Reliability Testing") all
parameters going into the measure of cyclic fatigue damage flD are identical
and cancel each other in the flDulADt ratio.
The validity of this acceleration transform has been demonstrated by comparing
and correlating the results of accelerated fatigue tests carried out at different
temperatures and different cycle dwell times (see Figs. 17-5 through 17-9). In
fact, Figures 17-7 and 17-9 required the use of the acceleration transform in
Equation (17-18); without it the test data resulting from different test conditions
could not be meaningfully compared. For leadless attachments such correlations
can be found in References 6-10.15,22.29 and for leaded attachments in References
11,16-21,28. Correlations have been established over dwell times ranging from 3
sec to 16 h and temperatures from -20 to + 150°C.
It needs to be emphasized that the acceleration transform cannot be used for
results from tests that either have temperature cycles that traverse over or even
come close to the glass transition temperature of any of the organic-based ma-
terials in the components or substrates, 6 or that have low temperature « - 20°C)
extremes that do not permit significant stress relaxation of the solder. 9
Further, it has been shown that for cycles with large temperature swings AT,
where the lower temperature extreme permits no significant stress relaxation of

~100~----'------r----~------__----__----~
~
~ to ~ 20 min, T = 3.7S·C/min
W EUTECTIC 63"10 Sn • 37"10 Pb
(!) 10 t - - - + '.....--'~~=----1- (R. WILD, IBM) "10 FAILED
Z
«
a:
z
«
a:
~
a:
«
w
:::c 0.1
en
= +25 - + 100"C} FATIGUE
= · 5 0 - +25"C
--l
- • 50 - + 100"C) OVERSTRESS
~
o
~0.01~~~~~~~~~~~--~~~~~~~-J
1 10 10 6
MEAN CYCLES TO FAILURE {fiIj

Figure 17-10 Manson-Coffin plot showing the results from reference10 for thermal cycling
of solder joints. For identical. cyclic strains, fl.." cycling between - 50 to + 1OO·C produces
significantly earlier failures than does cycling between either - 50 to + 25·C or + 25 to
+ lOD·C; this indicates different, unrelated damage mechanisms. It should be noted that the
total strain range does include the effect of the temperature swing flT.
 ANALYTICAL MODEL OF SOLDER SHEAR FATIGUE 569

the solder and the upper temperature extreme is well in the temperature range
of readily stress-relaxing solder, different damage mechanisms are responsible
for failure. Figure 17-10 shows experimental results 30 of thermal cycling over
typical temperature swings (AT = 75°C) and a temperature swing (AT = 150°C)
that is not typically encountered by electronic assemblies (see section "Testing
Conditions"). The results for the AT = 75°C thermal cycling~ven from -50
to +25°C-follow the damage mechanism described by Equation (17-12), for
AT = 150°C they clearly do not and the damage mechanism is likely dependent
on overstress and stress-related recrystallisation 31 considerations.

17.5.5 Failure Statistical Considerations

Solder joint fatigue failures are the result of a wearout mechanism (Fig. 17-1)
and as such have a statistical distribution around the mean cyclic life. This
statistical distribution is illustrated in Figure 17-11 using the data from a number

TEST TEMPERATURE: 81°C 23°C

90~---------r----~~--~~~----~
75~----------+---~

50~---------+~.

25~--------~~------~~--------~

10~------~r+----~~--~---
oLSI
eNSC
5~-----+--~--~~---+---
VTI

1~~-- ____ ~~~ ______ ________

~ ~

10 3 10 4 105 108
FATIGUE LIFE, CYCLES TO FAILURE
Figure 17-11 Raw data Weibull statistics plot from accelerated isothermal mechanical
cycling tests of plastic leaded chip carriers employing near square wave cycle shapes with
dwell times of 3 sec (f = 14,400 cycles/day) at two temperatures (from references 17- 19).
570 SOLDER ATIACHMENT RELIABILITY, ACCELERATED TESTING

of fatigue experiments,17-19; these results are also included in Fig. 17-6 as linear
regression plots. These statistical distributions are Weibull distributions with
typical Weibull slopes f3 = 2 10,11,16-21 for attachments via compliant leads and
f3 = 49,10,15,29 for leadless or stiff leaded attachments,
The Weibull cumulative failure distribution can be described in terms of the
mean ~.yclic life Nf by the relationship

F(N) = 1 - 0.5 [~r (17-21)

where F(N) = acceptable cumulative failure probability for the component

under consideration after N cycles
N = expected product service life, cycles of operation
f3 = Weibull shape parameter, slope of Weibull probability
plot

Equation (17-21) gives the probability of failure for a single component; for a
system consisting of m different component types each having a population of
n(m) components, the system probability of failure, F'5,.(N), is

F~(N) =1 (17-22)
m=1

which reduces for n identical components to

n[~r
F~(N) = 1 - 0.5 (17-23)

where n = number of the component under consideration in system,

number of individually monitored components on test

In statistical reliability evaluations it should be recognized, however, that the

confidence in the reliability projections resulting from relationships such as Equa-
tions (17-21) to (17-23) depends on the confidence established for the constituent
parameters and reliability evaluations for whole systems as shown in Equation
(17-22) will have significantly lower confidence levels than are established for
the individual constituent components.
The product design life and the expected cyclic frequency in operation give
the expected number of operating cycles N. Of course, given the statistical
distribution of failures in a wearout mechanism like fatigue, ~e design mean
fatigue life NlD) has to be larger than N,8,9,16 such that
 ACCELERATED FATIGUE RELIABILITY TESTING 571

- [ n In(O.S) J~ (17-24)
NAD) =N In[l _ F(N)]

where NI-D) = design mean fatigue life [NASO%)],

cycles-to-failure

for a design with n identical components. A mix of diverse components requires

a more sophisticated statistical analysis. While the effects of the number of
components and the component mix can be accounted for in the choice of the
acceptable failure probability F(N) , the primary considerations affecting the
choice of F(N) are dictated by the criticality of function of the system and any
available redundancies.
It should be noted that Equation (17-24) deals with components, not individual
solder joints. On components with a multitude of 1I0s, only the "comer" solder
joints (eight on chip carriers, four on small outline packages, two on chip
resistors) have the highest failure probability; all other solder joints have lower
failure probabilities since they experience significantly less cyclic fatigue dam-
age. From a strict point of view, the statistical liabilities of these less strained
solder joints also need to be accounted for; however, this is a rather cumbersome
task and requires accuracy to a degree which neither the analytical modeling
efforts nor accelerated testing can provide.

17.6 ACCELERATED FATIGUE RELIABILITY TESTING

17.6.1 Testing Considerations

The general considerations necessary for valid accelerated fatigue testing of SM

solder joints have been discussed at length in the literature. 6 •10,21,32 There are
two primary purposes for accelerated testing: (1) to obtain a measure of long-
term reliability of the SM solder attachments and (2) to provide a process ver-
ification test. The latter test purpose, however, is not served by an accelerat.ed
test for long-term solder fatigue. Rather, it is a combination overstress-tensile
fatigue-shear fatigue test leading to rapid failure of substandard solder joints
with reduced strength; it assures consistent quality and verifies the manufacturing
process (see Fig. 17-3); it is a processing quality screen [e.g., environmental
stress screening (ESS)], not a reliability test. The former test purpose requires
that the tests, while carried out on an accelerated basis, have to simulate use
conditions without introducing extraneous failure mechanisms or significant non-
representative physical and/or material behavior. The closer the test conditions
and the test vehicles are to operational reality, the more accurately the test results
will represent the reliability of the product. This also implies that the higher the
test acceleration-ratio of mean-time-to-failure (MTIF) for product in the field
572 SOLDER ATIACHMENT RELIABILITY, ACCELERATED TESTING

and MITF in accelerated testing-the less accurate and detailed can be the
results from the accelerated tests; that is, one gets what one pays for.
Table 17-2 lists realistic, but worst-case, use environments that can be ex-
perienced by electronic assemblies. These use environments are put into nine
use categories representing some of the most common applications and are

Table 17-2 Realistic Worst-Case Use Environments and Appropriate Accelerated Testing for
Surface Mounted Electronics by Use Categories
Worst Use Environment Accelerated Testing
Years
Use Tmin Tmax t!.T* tD Cycles! of Tmin Tmax t!.T** tD
Category (OC) (OC) (OC) (h) year service (OC) (OC) (OC) (min)
1-Consumer 0 +60 35 12 365 1-3 +25 +100 75 15
2-Computers +15 +60 20 2 1460 -5 +25 +100 75 15
3-Telecomm -40 +85 35 12 365 7-20 +25 +100 75 15
4-Commercial -55 +95 20 2 3000 -10 0 +100 100 15
aircraft
5-lndustrial & 20 12 185
automotive-- -55 +65 &40 12 100 -10 0 +100 100 15
passenger &60 12 60
compartment &80 12 20 & "Cold"t
6--Military -55 +95 40 12 100 0 +100 100 15
ground & ship &60 12 265 -5 & "Cold"t
7-Space : leo 8760 0 +100 100 15
I -40 +85 35 5-20
I
Igeo 12 365 & "Cold"t
8-Military Ia 40 2 500
avionics :b -55 +95 60 2 500 0 +100 100 15
Ic 80 2 500 -5
~----------------------
&20 1 1000 & "Cold"t
9-Automotive-- 60 1000 0 +100 100 15
under hood -55 + 125 &100 300 -5
&140 2 40 & "Cold"t & "Large t!. r t t
& = in addition
*~T represents the maximum temperature swing but does not include power dissipation effects; for power
dissipation see Equation (17-10); power dissipation can make pure temperature cycling accelerated testing
significantly inaccurate.
"All accelerated test cycles shall have temperature ramps <20oamin and dwell times at temperature
extremes shall be 15 min measured on the test boards. This wi II give - 24 test cycles/day.
tThe failure/damage mechanism for solder changes at lower temperatures; for assemblies seeing significant
cold environment operations, additional "Cold" cycling from perhaps -40 to O°C with dwell times long
enough for temperature equilibration and for a number of cycles equal to the "Cold" operational cycles in
actual use is recommended.
ttThe failure/damage mechanism for solder is different for large cyclic temperature swings traversing the
stress-to-strain - 20 to + 200( transition region; for assemblies seeing such cycles in operation, additional
appropriate testing with cycles similar in nature and number to actual use is recommended.
 ACCELERATED FATIGUE RELIABILITY TESTING 573

ordered roughly by increasing severity. It should be noted that the cyclic tem-
perature swings !::.T are not the difference between the maximum and minimum
temperatures that can be experienced; they are significantly smaller and move
between the extremes primarily due to seasonal, and perhaps geographic, tem-
perature variations. For some use categories the severity of the cyclic environment
can change seasonally leading to multiple cyclic temperature ranges; for others,
different ranges result from multiple use conditions (see the section "Multiple
Cyclic Load Histories"). It should be noted that the listed maximum temperature
swings are for component-external causes only; they do not include any effects
due to component-internal power dissipation. These effects need to be evaluated
using Equation (17-10); depending on the values of the parameters, the effective
temperature swing !::.Te can be either smaller or larger than !::.T. Also included
in Table 17-2 are estimates of the effective half-cycle dwell times tD as well as
of the design service lives. It should be noted that the Automobile-Under Hood
(Use Category 9) service life is shorter than the life for Automobile-Passenger
Compartment (Use Category 5) electronics. It is anticipated that because of the
severity of the under-the-hood environment reliable electronic packaging for the
entire life of the automobile is not possible. Under-the-hood electronics need to
be placed either into environmentally controlled enclosures or need to be viewed
(and accordingly designed) as auto parts requiring periodic replacement. Use
Category 7, space, requires different environments because of the large differ-
ences between low earth orbit ("leo") and geo-synchronous orbit ("geo") satel-
lites; the three use environments for Use Category 8, Military Avionics, account
for differences in the severity of environmental mission profiles due to the type
of aircraft or the location on the aircraft.
Table 17-2 also lists recommendations for the test conditions for accelerated
thermal cycling. For all test cycles, dwells at the temperature extremes of 15-
min duration, measured on the test vehicles, are recommended. These test rec-
ommendations are designed to assure that failures occur in accelerated testing
as the result of the same failure mechanisms as they would for product in use.
It should be noted that for electronic assemblies seeing significant cold environ-
ment operations (Use Categories 5 through 9) or large temperature swings trav-
ersing the damage mechanism region ( - 20 to + 20°C) in which the change from
stress-to-strain control takes place, additional representative testing is recom-
mended (see section "Solder Joint Fatigue" and Fig. 17-10). The total damage
from the multiple cyclic tests can be assessed using the procedure outlined in
section "Multiple Cyclic Load Histories."
Table 17-3 lists the number of failure-free test cycles (as described in Table
17-2) required to assure that the failure risk per component is below the stated
cumulative failure probabilities at the end of the stated service lives, given that
32 components are on test. The minimum number of failure-free test cycles N(t)
 Table 17-3 Accelerated Testing Required for Surface Mounted Electronics by Use Categories to Empirically Assure Failure Risk Levels at
End of Service Life
Service Life

.".
1 year 3 years 5 years
"'"
Failure-Free Test Cycles, N(t), Required* with 32 Components on Test to Assure Cumulative Failure
Probability, F(N) in ppm at End of Service Life
Use Attachment 1 100 10,000 1 100 10,000 100 10,000
Category Type ppm ppm ppm ppm ppm ppm ppm ppm ppm
Lead less 1600 470 140 5200 1500 440 9000 2600 760
Leaded** 4500 390 33 14,700 1300 107 25,400 2200 184
Leadless 1400 420 127 4400 1300 400 7500 2300 680
2
Leaded** 950 86 25 3000 270 25 5100 460 42
Leadless 1800 530 150 6000 1700 500 10,500 3000 870
3
Leaded** 5200 430 36 17,200 1400 118 29,900 2600 210
Lead less 1800 520 154 5700 1700 500 9800 2900 850
4
Leaded** 540 47 25 1700 150 25 3000 290 25
 Leadless 1200 340 100 3900 1100 320 6800 2000 560
5
Leaded** 6500 540 44 21,200 1800 146 36,900 2800 250
Leadless 3200 890 204 10,600 3000 690 18,600 5200 1200
6
Leaded** 16,000 1300 100 53,900 4300 330 94,700 7500 590
7 Leadless 15,600 4700 1420 48,800 14,800 4500 82,900 25,100 7600
leo Leaded** 19,000 1740 159 59,500 5400 500 101,000 9,200 840
7 Lead less 1090 310 86 3700 1030 290 6400 1800 510
geo Leaded** 1650 130 25 5500 440 34 9700 770 60
8 Lead less 2600 740 210 8700 2500 700 15,200 4300 1210
~T=20&40 Leaded** 4700 370 28 15,900 1250 99 27,900 2200 173
8 Leadless 5000 1430 410 16,200 4700 1350 28,100 8100 2300
~T=20&60 Leaded** 23,500 1950 160 77,100 6400 530 134,000 11,100 920
In
In 8 Leadless 7900 2300 680 25,300 7400 2200 43,500 12,700 3800
'"
~T=20&80 Leaded** 71,200 6200 530 229,000 19,800 1700 394,000 34,200 3000
Leadless 13,600 4100 1200 42,600 12,900 3900 72,400 22,200 6600
9
Leaded** 177,000 16,200 1500 553,000 50,600 4600 940,000 90,500 7800
8700 cycles"" 1 year test time
·The required number of test cycles is obtained for an acceptable cumulative failure probability F(N) at the end of the service life and a number of components n on test from
Equation (17-25).
··These numbers apply to compliant leads (diagonal lead stiffness < 100 Ib/in.); stiffer leads will move the required test cycles towards the lead less cycle numbers.
576 SOLDER ATIACHMENT RELIABILITY, ACCELERATED TESTING

assuring a cumulative failure probability F(N) with n individually monitored

components on test, is obtained from

N
(t)
=-
N(t)
[[ In(0.5) J~:;[ln[1
In[1 - F(N)] In(O.5)
-
l/n]]]~ (17-25)

where N(t) is obtained by rearranging Equation (17-18) to

- = -1[
N(t)
2
[2N] cU IlDtJ.!.c,
-

llDu
(17-26)

and N is the product of service life and the expected number of annual cycles.
Increasing the number of components on test reduces the required number of
failure-free test cycles. However, this reduction of the test duration is limited
by both the number of components that can realistically be put on test and the
slow response of the required test cycles to even substantial increases in the
number of components on test. It should be noted that for the more severe use
environments, assuring very low failure risks (1 and 100 ppm) can take longer
than the service life, which is of course absurdly impractical. However, it il-
lustrates that because of the unique properties of the materials involved in elec-
tronics packaging (see the section "Thermal Expansion Mismatch Problem")
there are definite limitations to what can be obtained by accelerated testing.
Another notable feature brought out in Table 17-3 is the difference between
stiff leadless and compliant leaded SM solder attachments in terms of evaluating
the results of accelerated testing and designing for low acceptable failure prob-
abilities. The two attachment types have different Weibull statistics slopes J3
(see section "Failure Statistical Considerations") and different dependencies on
the cyclic temperature swings aT [see Eqs. (17-12 and (17-17)]; this results in
the values of N(t) at the different failure probabilities, service lives, and use
categories, which at first glance certainly can appear peculiar.
These points are illustrated in Figure 17-12 where the cumulative failure
probability curves resulting from the different use environments described in
Table 17-2 are shown for one specific chip carrier. It is quite obvious that this
chip carrier is not appropriate for some of the more severe use environments in
its leadless version, but to illustrate the effect of the different use environments
it is necessary to have a common basis for comparison. The nine use environments
result in reliable operating lives that are as much as a factor of 250 apart for
stiff leadless solder attachments; for compliant leaded attachments this difference
shrinks to about a factor of 15. In Figure 17-12 the effect of the different Weibull
failure statistical slopes is also evident; this together with the effect of the lead
 ACCELERATED FATIGUE RELIABILITY TESTING 577

99.9 r--------,-r-TTT7r-rr--r-r-'

50 USE CATEGORIES { 8
o LEADLESS •
[ ] LEADED
10

"TYPICAL"
ELECTRONIC
COMPONENT
WID ATTACHMENT

0.1 \
0.01
,. ...... -
0.001

0.0001 L-=~~~:;:--1L:-...L...l::.--.l.-L=--.-J~---1
10 10 3 10 4 10 5 10 6 10 7 108
OPERATING TIME (h)

Figure 17-12 Cumulative failure probabilities for a typical electronic component and for
the surface mount solder attachment of a 68-1/0 chip carrier (L D = 620 mils, Tc - Ts = 5°C)
subjected to the use environments described in Table 17-2. For stiff lead less attachment h =
3 mils, as = 8 ppmfC, ac = 6 ppmfC; for the compliant leaded attachment h = 5 mils, A
= 800 mils 2 , as = 17.5 ppmfC, ac = 20.5 ppmfC, KD = 55 Ib/in. The Use Category 7 geo
curves are not shown.

compliancy produces significantly lower failure probabilities for leaded attach-

ments.
The primary parameter to accelerate a SM reliability test is the cyclic dwell
time. In order to translate the results from an accelerated test to product reliability
information, it is essential that the effects of reduced dwells be understood and
that the resulting reduced cyclic fatigue damage due to the incomplete stress
relaxation and creep of the solder joints can be accounted for (see the section
"Analytical Model of Solder Shear Fatigue" and Fig. 17-4). For product in
operation, the dwells are measured in hours or days; for functional and thermal
accelerated cycling the dwell durations can be as low as 3 to 5 min 11,29 and for
mechanical cycling as low as 3 sec. 18 ,19 The presence of an elastic strains 34 sets
limits to the minimum dwell times that can be correlated with use environments.
Another way to accelerate tests is to increase the applied strains. Frequently,
this has been attempted by extending the applied test temperatures significantly
beyond realistic use temperatures. From section "Analytical Model of Solder
Shear Fatigue" it is clear that even large increases in ilT produce only relatively
modest test accelerations; this has led to extreme test conditions such as cycling
from - 65 to + 150°C, perhaps even in thermal shock, to force early but un-
representative failures, rather than accelerated failures. Given the time and tem-
578 SOLDER ATIACHMENT RELIABILITY, ACCELERATED TESTING

perature dependence of many of the materials involved in surface mounting, 10,32

extending the test temperature ranges significantly is not a viable approach (see
the section "Acceleration Transform"). Small extensions of the test temperature
ranges are possible (see Table 17-2). However, for assemblies designed for high
reliability (.1n - 2ppm/"C) very little cyclic fatigue damage occurs in the solder
joints during thermal cycling, where the whole assembly is always at uniform
temperature. For these assemblies valid thermal cycling tests become impract-
ically long.
There is, however, another way to increase the applied strains. For accelerated
test purposes the CTEs of component and test vehicle substrate can be deliberately
mismatched6,16,32 thereby significantly increasing the applied cyclic strain range.
Examples of such deliberate mismatches would be plastic leaded components
on epoxy-Kevlar substrates or CTE-constrained FR-4 substrates using copper-
clad Invar or molybdenum; ceramic components could be tested mounted on
FR-4 substrates. The acceleration transform given in Equation (17-18) can then
be used to obtain the equivalent cyclic lives for actual product.
With mechanical cycling, increasing the applied strains is a simple matter
since the strains are applied by mechanical deformation of the test vehicle sub-
strate or external loading of the components. However, caution needs to be
exercised to prevent a change in failure mechanism due to overstressing, high
strain rates, or changing from predominantly shear to tension fatigue.

17.6.2 Accelerated Test Conditions

In Table 17-4 typical accelerated fatigue tests are listed together with some
common test parameters, acceleration factors, and comments. It should be noted
that temperature cycling could be performed at the higher frequencies possible
with functional cycling, if specially designed facilities allowing rapid uniform
temperature changes without even localized thermal shock are used instead of
standard commercial environmental chambers. In either test well-defined dwells
at the temperature extremes are necessary and should be at least 25 percent of
the cycle period in duration. Reference should also be made to the discussion
in the section "Reliability and Accelerated Testing-Overview."
Functional accelerated testing, which combines both external temperature vari-
ations and component-internal power dissipation cycling, provides the most rep-
resentative test environment without introducing extraneous and uncontrolled
differences between accelerated testing and product functional use. In functional
accelerated testing, the conditions and test vehicles mimic the product use con-
ditions and the product as closely as possible,l1,15 with the exception of the test
dwell times and thus the test frequency. While a shortening of the dwell times
leads to less complete conversion of the elastically stored strain energies to
plastic deformations in the solder via stress relaxation and creep-and thus to
 Table 17-4 Recommended Accelerated Fatigue Test Parameters
Test Temperature Frequency Warpage Acceleration Factors* Comments
Functional Realistic <100 cycles/ day; Realistic -10to20 Realistic power dissipation
cycling t < 30°C/min; and heat transfer
defined dwells
Temperature 0 - 100°C ,,;; 24 cycles/day; Somewhat higher <1 to - 10; ,,;; - 30 Very long test durations
cycling 25 _ 100°C t < 30°C/min; than realistic with deliberate test for ACTE < 5 ppm/oC
U1

<.0
" -40 - O°C defined dwells vehicle CTE-
mismatch
Mechanical Isothermal** <10,000 cycles/day; Bend radius ± ,,;;500 Excludes all CTE-
cycling near square wave 40 to 80 in. mismatch effects;
(± 1 to 2 m) beware of overstressing;
somewhat higher tensile
strains
'Mean-time-to-failure acceleration factors for IlT (test) = IlT (use).
"At O°c, room temperature, 60°C, or BOcC.
580 SOLDER ATIACHMENT RELIABILITY, ACCELERATED TESTING

less fatigue damage per cycle and a longer fatigue life in terms of cycles-to-
failures-the total time necessary to produce failures is stilI significantly reduced.
Thermal cycling provides reliability data with significantly simpler test ar-
rangements, albeit at the price of reduced accuracy and increased uncertainty.
While in functional cycling the shape of the cyclic waveform is defined by the
thermal environment of the product, special care is necessary for thermal cycling
(temperature variations in a thermal test chamber) to achieve proper test con-
ditions. The cyclic waveform needs to have heating and cooling ramps for which
the rates of temperature change are below ~ 30°C/min to avoid thermal shock
conditions. In addition, the waveform can have a significant effect on the (ef-
fective) dwell times at temperature. The preferred waveform is trapezoidal be-
cause it is closest to actual use conditions and has well-defined dwells at the
temperature extremes, but sinusoidal and even sawtooth waveform have been
used. For sinusoidal waveforms there arises the question of what the effective
dwell time is. In Reference 6 where the solder fatigue model underlying the section
"Analytical Model of Solder Fatigue" was first developed from isothermal sin-
usoidal mechanical cycling results, 25 the assumption was made that the effective
dwell time is one-quarter of the cycle period. While for sawtooth waveforms
some stress relaxation and creep occurs during the up and down ramps, it is so
small that for all practical purposes tD = 0 for accelerated tests of this nature.
Thus, sawtooth thermal cycling is more akin to low frequency vibration than to
thermal expansion mismatch fatigue. The limits on the rate of temperature change
virtually prevent the use of dual-temperature chambers where the test vehicles
are moved between two chambers that remain at the two temperature extremes.
Without complex special arrangements it is virtually impossible to avoid at least
local thermal shock conditions and putting the entire test results in question.
For the "Cold" temperature cycling, shown in Tables 17-2 and 17-4, well-
defined dwells are not important (since stress relaxation and creep are not sig-
nificant factors at those temperatures) as long as the whole assembly reaches
temperature equilibrium. The temperature ramp rates are, however, important
both to avoid thermal shock and to stay within representative strain rate ranges.
The "quick and dirty" isothermal mechanical cycling tests provide coarse
comparative data that can, however, be "benchmarked" using the more accurate
low acceleration tests. The most extensive isothermal mechanical cycling pro-
gram, reported in References 17-19 and 28, involved the controlled deflection
of the test vehicle circuit board. The equipment developed for the IEEE Compliant
Lead Task Force is commercially available. Other mechanical cycling tests in-
volve test vehicle board twisting or some tensile arrangements. 34 ,35 It should be
noted that isothermal cycling excludes any effects of thermal cycling, such as
the thermal expansion differences between the solder and the materials to which
it is bonded and between the lead frame and component body materials. While
these are secondary effects in most applications, they clearly make important
additional cyclic strain contributions in product use. 11 It should be noted, how-
 ACCELERATED FATIGUE RELIABILITY TESTING 581

ever, that isothennal mechanical cycling with well-controlled applied strains can
give highly accurate results. There is no fundamental difference between the
primary damage mechanisms operative in thennal cycling and isothennal me-
chanical cycling. Without the isothennal data from Wild25 the solder fatigue
model described earlier could not have been developed.

17.6.3 Test Vehicle Design

The test vehicles must be designed to allow electrical monitoring for continuity
of the solder joints. This requires specially designed test boards as well as
specially prepared components. In order to capture a significant number of solder
joints, the solder joints should be connected into a continuity daisy-chain test
net containing six or more solder joints. 10 Components with solder joints on four
sides can be divided into four daisy-chains, one for each side. For components
with an odd number of solder joints on a side it is best to skip the center-most
joint in the daisy-chains; the comer joints, because they are the most distant
from the component center (neutral point) and experience the highest defor-
mations and thus the most fatigue damage, should always be included in the
daisy-chains. These center-most joints not connected into the continuity daisy-
chain test net can serve to bring power to a resistor chip simulating functional
power dissipation and to monitor test "junction" temp~ratures with thennistors. 11
The test vehicles should include provisions to monitor and record the cyclic
temperature profiles (see, e.g., Reference 11) at various locations on the test
vehicles. Accurate knowledge of the actual cyclic temperature histories is es-
sential for the evaluation of the test results and the reliability extrapolation to
use conditions.

17.6.4 Sample Statistics and Solder Joint Defects

As has been discussed in the section "Failure Statistical Considerations," the

failures in a population of SM solder attached components will follow a Weibull
statistical distribution. It is therefore important for the test plan to include pro-
visions to obtain the results from a failure population large enough to provide
this statistical infonnation. The less there is known about the assembly being
tested and the less well controlled the test setup, the larger is the number of data
points required. In order to obtain a statistically adequate number of data points
(one per daisy-chain test net) for a well-controlled and understood test setup, a
minimum of 24, but preferably 32, daisy-chains should be on test.
While it is possible, but of significant difficulty, to produce test vehicles with
intentional variations and "defects" (e.g., misalignments, solder volume, lead
nonplanarities, voids, etc), it is perhaps better to subject test vehicles to the
582 SOLDER ATIACHMENT RELIABILITY, ACCELERATED TESTING

accelerated testing in quantities that capture typical, but acceptable, product

variations and defects. These variations and defects need to be identified and
recorded prior to testing to allow correlation between any early failures that
might occur and these defects. It needs to be said, however, that all evidence
so far indicates no detrimental effects of so-called defects, (i.e., pad/lead-mis-
alignments, small solder volume, lead nonplanarities, gas-bubble or shrinkage
voids, nonshiny solder, etc.) on long-tenn reliability. Gross defects are likely
to cause yield or infant-mortality problems due to inadequate solder joint strength
to withstand the rigors of manufacturing, assembly, and transport.

17.6.5 Test Vehicle Assembly and Conditioning

The test vehicles should be assembled and processed as similarly as possible to

actual product. Product-similar test vehicle assemblies should reflect product
solder joint characteristics such as grain structure, composition, and intennetallic
compound layers and their thickness. Solder grain structure is inherently unstable;
it will coarsen with time and temperature,36,37 particularly in the presence of
cyclic strains. 37- 4o Equivalent ageing in tenns of grain growth can be estimated
from

(17-27)

where the activation energy for grain coarsening EA = 0.42 eV and Boltzman's
constant k = 8.62 x 1O-5eV/K. 41
Fine grain structure in the solder joints will result in longer fatigue lives in
accelerated testing, which does not allow the time for these time dependent
processes to occur to the same degree as they would in product use over many
years. The changes in the structure of solder, which undergoes and accumulates
fatigue damage, appear to follow the sequence of grain structure coarsen-
ing,37-40 micro-voiding/cavitation fonnation at grain boundaries,40 micro-crack-
ing with ~50 to 70 percent of life remaining, coalescence of micro-cracks, and
fracture. Therefore, it stands to reason that accelerated test vehicles should start
with a solder grain structure that is more representative of the solder grain
structure in product after some period of operation.
After assembly, the test vehicles should be subjected to an accelerated thennal
aging (e.g., 300 h at WOOe ll ) in air to simulate a reasonable use period and to
accelerate such possible processes as solder grain growth, intennetallic compound
growth, and oxidation. Storing the test vehicles after this artificial ageing for
some additional time at room temperature before commencing with the fatigue
testing serves to further stabilize the solder structure.
 PREDICTION OF SM SOLDER JOINT RELIABILITY 583

17.6.6 Failure Definition and Detection

Solder joint fractures manifest themselves in product predominantly by inter-

mittant electrical failures during either thermal transients or dynamic loading
conditions. These intermittents can, at least initially, be of exceedingly short
duration « 1 f.Ls). Only after many cycles of the solder joint fracture surfaces
working against each other will electrical contact be broken during steady-state
conditions at one of the cyclic tempertures extremes. 10,29
For multiple solder joints daisy-chained into continuity nets, the optimum
failure definition is the first confirmed interruption of electrical continuity even
of very short duration. 10 This failure definition not only closely corresponds to
the failure indication during product field operation, but it also does not require
test interruptions with time consuming and uncertain failure searches. Since full
solder joint fractures manifest themselves-at least initially-as high electrical
resistance spikes of exceedingly short duration and almost exclusively during
the cyclic transients, the continuity nets require continuous, not multiplexed
(i.e., switched) monitoring.
For practical and economic reasons also, continuous monitoring for electrical
discontinuity appears to be the most viable option given the large number of test
nets required to produce statistically adequate databases. Equipment to monitor
definite changes in resistance is significantly more expensive than the electronics
necessary to detect essentially open nets, without providing more consistent and
more relevant data.
Some previous studies have been carried out by monitoring with the sensing
current limited to I rnA and 5 V and resistance spikes of 1 f.LS or longer de-
tectable. II ,15 For the IEEE Compliant Lead Task Force study 17-19,28 failure de-
tection was carried out on 256 continuously and simultaneously monitored chan-
nels. The sensing electronics could detect 1000-0 resistance spikes of 0.2-f.Ls
duration with a maximum sensing current at 1.25 rnA at zero net resistance
decreasing asympotically to zero current at 10 V potential across a full net open.
The argument has been advanced that even at these currents and voltages, arcing
across the short-duration gap is possible. Commercial equipment has recently
become available that can detect resistance spikes of 0.2-f.Ls duration with a 300-
o threshold limit with a sensing current maximum of 1.25 rnA at zero net
resistance declining to 0 rnA at 0.6-V maximum voltage at infinite net resis-
tance. 42

17.7 PREDICTION OF SM SOLDER JOINT RELIABILITY

17.7.1 Simple Cyclic Load Histories

Surface mount solder joint reliability prediction, within the stated caveats and
limitations in the section "Analytical Model of Solder Shear Fatigue," involves
584 SOLDER ATIACHMENT RELIABILITY, ACCELERATED TESTING

primarily combining and rearranging some of the equations in these sections.

From Equations (17-12) and (17-24) the expected cyclic life with a failure
probability of x = 100 F(N) is for leadless SM solder attachments

N (x%) = ![~ LollallTe]i[ln(1 - O.01X)]~ (17-28)

f 2 2Ej h In(O.5)

where NJ<x%) = number of operating cycles at x percent failure prob~bility

For leaded solder attachments, the cyclic life prediction

nt
N(x-Io)=- -
1[F Ko(LollallTe)2]i[ln(1 - O.OlX)].!.Ii (17-29)
f 2 2Ej (200 psi) Ah In(0.5)

results from combining Equations (17-17) and (17-24). Equation (17-29) is in

English units; the scaling coefficient for metric dimension is 1.38 MPa instead
of 200 psi in the denominator in the first bracket.
It should be noted that these equations apply only to solder joints consisting
of 60% Sn-40% Pb or eutectic solder. Further, for leaded components with lead
materials that have CTEs significantly lower than copper alloy materials, the
results from Equation (17-29) will be optimistic since the fatigue damage con-
tributions from the solder-lead material CTE-mismatch are not included.
For high-frequency applications, to < 3 sec, and low temperature applications,
To < - 20°C, for which the stress relaxation and creep in the solder joint is
substantially incomplete, if not non-existent, the direct application of the Man-
son-Coffin fatigue relationship (see Eq. (17-3) in the section "Analytical Model
of Solder Shear Fatigue" and Reference 5) is advised.

17.6.2 Multiple Cyclic Load Histories

Multiple cyclic load histories [e.g., "Cold" temperature fatigue cycles combined
with higher temperature creep-fatigue cycles (see Table 17-2) combined with
vibration] all make their contributions to the cumulative fatigue damage in solder
joints. Under the assumption that these damage contributions are linearly cu-
mulative-this assumption underlies Equation (17-1) as well-and that the si-
multaneous occurrence or the sequencing on!er of these load histories makes no
significant difference, the Palmgren-Miner's rule43 can be applied

(17-30)
 REFERENCES 585

where Ni = actually applied number of cycles at a specific cyclic load level

Nj .i = fatigue life at a specific cyclic load level alone

Equation (17-30) can be used with the allowable sum of the fatigue damage
fractions significantly less than unity to provide margins of safety.

17.8 ACKNOWLEDGMENT

Most of this work was done while the author was with AT&T Bell Laboratories.

REFERENCES

1. Klinger, D. J., Y. Nakada, and M. A. Menendez, AT&T Reliability Manual Eds.,

New York, Van Nostrand Reinhold, 1990.
2. Holcomb, D. P., and J. C. North, "An Infant Mortality and Long-Tenn Reliability
Failure Rate Model for Electronic Equipment," AT&T Technical Journal V. 64 No.
(1), 1985.
3. MIL-HDBK-217D, Military Handbook - Reliability Prediction of Electronic Equip-
ment, January 15, 1982, Department of Defense, Washington, D.C. 20301.
4. Morrow, 1. D., "Cyclic Plastic Strain Energy and Fatigue of Metals," ASTM STP
378, ASTM, Philadelphia, PA, 1964, pp. 45-87.
5. Manson, S. S., Thermal Stress and Low Cycle Fatigue, New York, McGraw-Hill,
1966.
6. Enge1maier, W., "Functional Cycling and Surface Mounting Attachment Reliability,"
ISHM Technical Monographic Series 6984-002, Int. Soc. for Hybrid Microelec-
tronics, Silver Spring, MD, October 1984, pp. 87-114; also in Circuit World, 11(3),
spring 1985, pp. 61-72.
7. _ _ , "Fatigue Life of Leadless Chip Carrier Solder Joints During Power Cycling,"
IEEE Trans. CHMT, 6(3) pp. 232-237.
8. Clech, J.-P., W. Engelmaier, R. W. Kotlowitz, and J. A. Augis, "Surface Mount
Solder Attachment Reliability Figures of Merit-'Design for Reliability' Tools,"
Proc. 1989 SMART V Corif., New Orleans, LA, January 9-12, 1989.
9. Engelmaier, W., "Surface Mount Solder Joint Long-Tenn Reliability: Design, Test-
ing, Prediction," Soldering & Surface Mount Technology, no. 1, February 1989, pp.
14-22.
10. _ _ , "Is Present-Day Accelerated Cycling Adequate for Surface Mount Attachment
Reliability Evaluation?," IPC-TP-653, Institute for Interconnecting and Packaging
Electronic Circuits, Lincolnwood, IL. September 1986.
11. Engelmaier, W., and A. I. Attarwala, "Surface Mount Attachment Reliability of
Clip-Leaded Ceramic Chip Carriers on FR-4 Circuit Boards," IEEE Trans. CHMT,
12(2), June 1989, pp. 284-296; also partially in IEPS Journal, 9(4), January 1988,
pp. 3-11.
586 SOLDER ATIACHMENT RELIABILITY, ACCELERATED TESTING

12. Suhir, E., "Axisymmetric Elastic Defonnations of a Finite Circular Cylinder with
Application to Low Temperature Strains and Stresses in Solder Joints," J. Appl .•
Mech., 56(2), June 1989, pp. 328-333.
13. "Guidelines for Surface Mounting and Interconnecting Chip Carriers," IPC-CM-78,
Institute for Interconnecting and Packaging Electronic Circuits, Lincolnwood, IL,
November 1983.
14. Klemencic, R., E. Kny, and W. Schmidt, "Multilayer Circuit Boards with Molyb-
denum-Copper Metal Cores," Circuit World, 15(4), July 1989, pp. 5-8.
15. Engelmaier, W., "Surface Mount Solder Joint Reliability of Leadless Ceramic Chip
Carriers on Different Substrates," to be published.
16. Kotlowitz, R. W., and W. Engelmaier, "Impact of Lead Compliance on the Solder
Attachment Reliability of Leaded Surface Mounted Devices," Proc.lnt. Electronics
Packaging Soc. Conj., San Diego, CA, November 17-19, 1986, pp. 841-865.
17. Engelmaier, W., "Solder Joint Reliability and Testing Considerations for Leaded
Chip Carriers," Nat. Electronic Packaging and Production Conj.,NEPCON West,
Anaheim, CA, February 26-28, 1985, pp. 884-885 (extended abstract only); Pre-
sentation available on IEPS Videotape VR 001, "Lead Compliance and Solder In-
tegrity of Surface Mount Leaded Packages," Int. Electronics Packaging Society,
Wheaton, IL.
18. _ _ , "Interim Report on the Test Results of IEEE Test Program," IEPS Videotape
VR 004, "IEEE Compliant Lead Task Force Results of Phase A," Int. Electronics
Packaging Society, Wheaton, IL.
19. _ _ , "IEEE Compliant Lead Task Force-Phase I: Correlation and Analysis of
Solder Joint Fatigue Results Obtained at Different Cyclic Frequencies and Temper-
atures," Proc. Int. Electronics Packaging Soc., Conj., Orlando, FL, October 21-
23, 1985, p. 279 (extended abstract only); also, presentation available on IEPS
Videotape VR 002, "Second Interim Report on the IEEE Complaint Lead Task
Force," Int. Electronics Packaging Society, Wheaton, IL.
20. Kotlowitz, R. W., "Comparative Compliance of Generic Lead Designs for Surface
Mounted Components," Proc. Int. Electronics Packaging Soc. Conf., Boston, MA,
November 9-11, 1987, pp. 965-984.
21. _ _ , "Comparative Compliance of Representative Lead Designs for Surface Mounted
Components," Proc. 1988IEPS Conj.. Dallas, TX, November 7-9, 1988, pp. 908-
948.
22. Liljestrand, L.-G., and L.-O. Anderson, "Accelerated Thennal Fatigue Cycling of
Surface Mounted PWB Assemblies in Telecom Equipment," Proc. Int. Packaging
Soc. Conj., Boston, MA, November9-11, 1987, pp. 965-984; also in Circuit World
14(3), March 1988, pp. 69-73.
23. Hall, P. M., "Strain Measurements During Thennal Chamber Cycling of Leadless
Ceramic Chip Carriers Soldered to Printed Boards," Proc. 34th Electronic Com-
ponents Conj., New Orleans, LA, May 1984, p. 107.
24. _ _ , "Creep and Stress Relaxation in Solder Joints in Surface-Mounted Chip
Carriers," Proc. 37th Electronic Component Conf.• Boston, MA, May 1987, p. 579.
25. Wild, R. N., "Some Fatigue Properties of Solders and Solder Joints," IBM Tech.
Rep. 73Z000421 , January 1973.
 REFERENCES 587

26. Clech, J.-P. and J. A. Augis, "Temperature Cycling, Structural Response and At-
tachment Reliability of Surface-Mounted Leaded Packages," Proc.J988JEPS Conj.,
Dallas, TX, November 7-9, 1988, pp. 305-325.
27. Hines, L. L., AT&T Bell Laboratories, Columbus, OH, private communication,
July 1988.
28. Balde, J. W. "The IEEE Compliant Lead Task Force," IEEE Trans CHMT-JO, Sept.
1987, pp. 463-469.
29. Ng, K. G., "Assessing Surface Mount Attachment Reliability of Ceramic Cylindrical
Resistors on FR-4 Circuit Boards," Proc. Expo SMT '88, Las Vegas, NV, November
1988, pp. 1-6.
30. Wild, R. N., "1974 IRAD Study-Fatigue Properties of Solder Joints," IBM Report
No. M45-74-002, Contract No. IBM 4A69, Jan. 5, 1975.
31. _ _ , IBM, Owego, NY, private communication, May 1989.
32. Engelmaier, W., "Test Method Considerations for SMT Solder Joint Reliability,"
Proc. Int. Electronics Packaging Soc. Coni., Baltimore MD, Oct. 29-31, 1984, pp.
360-369; also in Brazing & Soldering, no. 9, autumn 1985, pp. 40-43.
33. Tien, J. K., B. C. Hendrix, B. L. Bretz, and A. I. Attarwala, "Creep-Fatigue
Interactions in Solders" Proc. 39th Electronics Components Con/., Houston, TX,
May 1989, pp. 259-263.
34. Solomon, H. D., "Low Cycle Fatigue of Surface Mounted Chip Carrier/Printed
Wiring Board Joints," Proc 39th Electronics Components Coni., Houston, TX, May
1989, pp. 277-292.
35. Enke, N. F., T. J. Kilinski, S. A. Schroeder, and J. R. Lesniak, "Mechanical
Behaviors of 60/40 Tin-Lead Solder Lap Joints," Proc. 39th Electronics Components
Coni., Houston, TX, May 1989, pp. 264-272.
36. Lampe, B. T., Welding Research Supplement No. 10, 1976, p. 330.
37. Karjalainen, L. P., et al., "Relation Between Microstructure and Fracture Behavior
in Surface Mounted Joints," Brazing & Soldering, no. 15, autumn 1988, pp. 37-
42.
38. Wolverton, W. M., "The Mechanisms and Kinetics of Solder Joint Degradation,"
Brazing & Soldering, no. 13, autumn 1987, pp. 33-38.
39. Frear, D., D. Grivas, and J. W. Morris, "A Microstructural Study of the Thermal
Fatigue Failures of 6OSn-40Pb Solder Joints," J. Electron. Mater. (USA), 17(2),
March 1988, pp. 171-180.
40. Greenwood, P. J., T. C. Reiley, J. K. Tien, and V. Raman, "Cavitation in a Pb-
Low Sn Solder During Low Cycle Fatigue," Scripta Met., 22(9), September 1988,
pp. 1465-1468.
41. Evans, 1. W., "Grain Growth in Eutectic Solder: Impact on Accelerated Testing,"
Proc. 33rd Ann. JPC Meeting, Boston, MA, April 1-6, 1990.
42. Sofia, J. W., Anatech, Inc., Wakefield, MA, Private communications, November
1988.
43. Miner, M. A., "Cumulative Damage in Fatigue," J. Applied Mechanics, vol. 12,
1945.
 18
Surface Mount Attachment
Reliability and Figures of Merit for
Design for Reliability

}-P M. Clech and }. A. Augis

Surface Mount (SM) assemblies in the field are subject to complex thermal
fluctuations. The cyclic thermal expansion mismatch between SM components
and the substrates to which they are attached may lead to fatigue or wearout of
solder joints during the useful life of a product. Attachment reliability is the
ability of solder joints to provide electrical and mechanical connections under
stated conditions and for a specified duration. Packaging engineers and physical
designers thus bear the responsibility to design SM attachments that will survive
the desired product life. This is achieved by proper design of the attachment and
soldering according to recommended Workmanship Requirements (e.g., reference!).
While several companies have gathered attachment reliability data for generic
SM components, the data often remain proprietary or are not readily available
for less common or emerging packaging technologies. Reliability evaluation can
then be an involved process2 ,3 including the design and assembly of test vehicles,
accelerated testing, data, and failure mode analysis, all of which require the
appropriate expertise and adequate commitment of time and resources. This
process amounts to a 6-month to I-year effort while design cycles for SM
packages or electronic assemblies are as short as I month, sometimes less.
Clearly, design tools are needed to build in attachment reliability throughout the
design process and provide reliability information quickly, based on critical
design and assembly parameters.
Figures of Merit (FMs) have been developed4 ,5 that serve as design for reli-
ability tools in the design of solder joint interconnections and account for the
primary factors that determine attachment reliability in the field. The FM for-
mulation includes the substrate-component thermal expansion mismatch, the
attachment design parameters, the thermal environment, the product service life,

588
 MECHANICS AND FATIGUE OF SM SOLDER JOINTS 589

and the acceptable cumulative failure probability per device or attachment re-
liability target. The PMs are derived from simple solder joint theories and em-
pirical fatigue data for the attachment of SM devices. By using the PMs, pack-
aging engineers and physical designers can assess the impact of design parameters
on end-use attachment reliability and, if need be, change them to meet a stated
reliability target. The PM tools provide guidelines to ensure that no significant
wearout of solder joints occurs during the useful life of an electronic assembly
and give quantitative estimates of attachment reliability for quality solder joints.
Because of simplifications inherent to design tools and tolerances on design
parameters, reliability estimates derived from the PM tools should be considered
as order of magnitude estimates. End-use reliability for a given product and
application can only result from design for reliability, followed by quality as-
sembly.
This chapter focuses on engineering approaches to address SM attachment
reliability at the design stage. Fundamentals of solder joint fatigue for leadless
and leaded attachments are reviewed in the next section, which provides the
theoretical foundation for developing the PM tools. The PMs are then presented
and applied to several examples.

18.1 MECHANICS AND FATIGUE OF SM SOLDER JOINTS

This section reviews simplified concepts of the mechanics and fatigue of solder
joints in SM leadless and leaded components, providing insight into the theo-
retical aspects underlying the PM formulation.

18.1.1 SM Leadless Attachments

Figure 18-1 is a schematic of an SM leadless assembly showing the design

parameters that govern the solder joint response in thermal cycling:

• L (in.) is the distance from the neutral axis of the component to the center
line of the farthest solder joint; for a perimeter chip carrier, L is half the
diagonal dimension of the package.
• h (in.) is the solder joint height or component standoff above the substrate.
• aa (in.lin.°C) is the absolute mismatch between the coefficients of thermal
expansion (CTE) of the component a c and substrate as: aa = lac - asl.

When the assembly is subject to a temperature swing aTeC),* solder joints

of leadless components undergo a gross shear strain. 3 ,6,7 Assuming complete

*In power cycling, an equivalent temperature range of I1T. is used to account for component
power dissipation. 3
590 SURFACE MOUNT ATTACHMENT RELIABILITY

I'
/COMPONENT
NEUTRAL AXIS

SOLDER JOINT ~-----2L----~~

CENTER LINE

Figure 18-1 Schematic of surface mount lead less component showing the design parameters
that affect solder attachment reliability in thermal cycling.

stress relaxation for solder at the temeprature extremes, the amplitude of the
applied shear strain is

L
tJ.:y = - aaaT (18-1)
h

The solder joint response during thermal cycling is characterized by hysteresis

loops in the shear stress-strain plane (see Fig. 18-2). Such loops have been
obtained experimentally for leadless chip carriers8 ,9 and numerical simulations
have been proposed by several authors. 10,11,12,13 Their general shape gives insight
into the solder joint deformation and stress history. The initial response as the
temperature ramps up is elasto-plastic (elastic plus time independent plastic
deformations) and is limited by the solder yield strength. During the dwell at
higher temperature, Thigh, creep and stress relaxation occur along an isothermal
stress reduction line with a steep slope determined by the stiffness of the assem-
bly. In accelerated testing with short dwells, stress reduction may be incomplete.
If the dwell at Thigh is long enough, as in slow temperature cycling or low
frequency use conditions (e.g., 1 cycle/day) stress reduction proceeds to zero
stress and maximum plastic strain. The stress-strain response is reversed in the
next half-cycle with stress reduction proceeding in the opposite direction at low
temperature (T1ow )' For leadless or "stiff" SM attachments, the largest cyclic
strain energy density associated with the hysteresis loop area is obtained as

(18-2)

where C is a material constant characteristic of solder and independent of the

assembly design parameters.
 MECHANICS AND FATIGUE OF SM SOLDER JOINTS 591

SHEAR \
STRESS \
\
YIELD~ \

ELASTIC
LOADING
CREEP-STRESS
RELAXATION

-4..1---if-------....,t-+\----+
SHEAR
STRAIN

\
\
\
\

USE CONDITIONS
- OR LONG DWELLS

ACCELERATED TESTING
= OR SHORT DWELLS

Figure 18-2 Hysteresis loops characterizing the solder joint response in the shear stress-
strain plane during thermal cycling; case of SM lead less components.

The plastic part of the cyclic shear strain ll"{p leads to solder joint failures in
low cycle fatigue. This is described by Coffin-Manson relationships3,6,14 that
give the median cyclic life /iif as

(18-3)

where C 1 is a material constant and the exponent parameter m is about - 0.5,

The solder joint cyclic life is expected to decrease with increasing CTE mismatch.
For large temperature fluctuations and CTE-mismatched substrates and com-
ponents, the component size may have to be limited or the standoff height be
increased for the solder joints to survive the desired cycles of operation.

18.1.2 SM Leaded Attachments

One solution to the problem of solder fatigue in leadless assemblies is to provide

a more compliant attachment, using leaded components. Compliant leads take
up the component-substrate thermal expansion mismatch,15,16 thus reducing the
cumulative fatigue damage imparted to solder joints. Figure 18-3 is a schematic
592 SURFACE MOUNT ATTACHMENT RELIABILITY

K = DIAGONAL LEAD
STIFFNESS

SUBSTRATE •

Figure 18-3 Schematic of surface mount leaded component showing the design parameters,
other than those defined in Figure 18-1 for lead less assemblies, that affect attachment reliability.

of an SM leaded assembly. In addition to the design parameters defined in the

leadless case, the attachment reliability of leaded assemblies is determined by:

• K (lb/in.) the diagonal lead stiffness

• A (in. 2 ) the minimum load bearing area of the solder joint, subjected to a
shear force transmitted by the lead to the solder joint
• h (in.) a characteristic solder joint thickness

The diagonal lead stiffness K is the flexural spring constant of the lead in the
diagonal direction of the package where the thermal expansion mismatch is the
largest. The parameter K is obtained by classical strain energy methods 17- 19 or
finite element analysis. In the FM formulation for leaded components, the min-
imum cross-sectional area A is approximated as two-thirds the solder wetted lead
area projected to the surface mounting pad. We take the characteristic solder
joint thickness h as half the solder paste stencil thickness, a representative di-
mension for the average solder thickness under the lead foot.
The solder joint response in thermal cycling follows a hysteresis loop as in
the leadless case. The initial solder response is elastic and, if the leads are
compliant enough, stresses remain below the yield strength of solder (see Fig.
18-4). At the temperature extremes, stress reduction proceeds along isothermal
lines of much lesser slope than in the leadless case. 20 The product of S, the
largest applied stress, and !::.:y, the total cyclic shear strain range, gives the
maximum available cyclic strain energy density, a measure of cyclic fatigue
damage.
The thermal expansion mismatch Ll1al1T is transmitted to the joint as a force
K(Ll1aI1T), and the resulting shear stress

S = _K--,(_LI1_a_I1_T-,-)
(18-4)
A
 MECHANICS AND FATIGUE OF SM SOlDER JOINTS 593

SHEAR
STRESS

YIELD",,-

ELASTIC
LOADING

/).$' SHEAR
--f.oi;;;:::-----jj---t/-------===-"t-.... STRAIN

_ USE CONDITIONS
OR LONG DWELLS

Figure 18-4 Hysteresis loops of solder joint response during thermal cycling; case of a
leaded assembly with compliant leads.

detennines the initial stress reduction rate during dwells at constant temperature.
The maximum cyclic strain energy density is then obtained as

dW = Sd-y = K(Lda.dT)2 (18-5)

Ah

The median cyclic life for solder joints of leaded components has been found
to follow a generalized fatigue law*21

(18-6)

where C2 is a material constant and the exponent parameter m is in the range of

- 0.5 to - 0.7. Figure 18-5 shows a plot of d W versus N/,5 for five SM leaded
assemblies subjected to accelerated testing in functional 16 or slow temperature
cycling: 20 to 100°C, 10 cycles/day.23 The slope of the regression line, dW
versus Nt is m = -0.64. For the three assemblies with the longest lives, lower
bounds estimates of Nt were used since no failures were observed at the end of
tests. Testing to failure would give larger values of Nt and a reduced slope for
the regression line, closer to the more commonly observed values m - - 0.5
for solder.

*The Coffin-Manson relationship [Eq. (18-3)] is a particular case of and can be derived from
Morrow's generalized fatigue law.22
594 SURFACE MOUNT ATTACHMENT RELIABILITY

10 r-------------------------------~
<D. (j) SOlC24 -GW I KEVLAR, STC*
5 ................ ~ SLOPE: m = -0.64 (2) J.Cllp/FR4, POWER CYCUNG TEST
........ (3) PLCC/FR4, STC*
®. . . . . . . . . @ SOJ28/FR4, STC*
................
~W(lb - in.lin.3)
.... ® S·Cllp/FR4, POWER CYCLING TEST
........................................

0.5 @~
@ ........
@...............

0.1 ~ __ ~ ________ ~ __ ~ ________ ~ ___J

5000 10,000 50,000 500,000 10'
iii f (CYCLES)
* STC: SLOW TEMPERATURE CYCLING
20·100·C, 10 CYCLES/DAY

Figure 18·5 Plot of cyclic strain energy density versus median cycles to failure for SM
leaded assemblies subjected to accelerated thermal cycling.

Equations (18-5) and (18-6) suggest that the cyclic life can be increased by
providing more lead compliance (lower K) or increasing the lead-solder wetting
area (larger A). The lead stiffness can be reduced most effectively by increasing
the overall lead height l7- 19 or by using a thinner lead frame. 24 Often, an increase
in the width of the lead foot at the tip of the lead is beneficial to reliability since
it increases the lead-solder wetted area with no significant reduction in lead
compliance (see Example 18-3).

18.1.3 Failure Distribution

Another important feature of SM reliability studies is the statistical distribution

of solder joint failures. The wearout of solder joints follows a Weibull distribution
around the median cycles to failure, ** Nt (see Fig. 18-6). The cumulative
distribution function, or cumulative failure probability, at the current cycle n is

F(n) = 1- O.5(~Y (18-7)

**Median cycles to failure are used instead of the characteristic scaling factors of Weibull dis·
tributions. since fatigue data for solder and SM attachments are most often reported in terms of
median life.
 Cumulative Failure
Probability: F(n)

50%

~
m
n
I
)-
Reliability Z

Target: F
n
V>
)-
Z
o
"T1

Cycles: n ~
Ci
C
m
N Nd :::; N, o
"T1
V>
Operation Median Median ~
V>
Cycles Design Cycles or
om
Life to Failure
'"
'0
Figure 18-6 Weibull plot: cumulative failure probability versus cycles. The design line must z
be to the left of the wearout failure distribution for the reliability expectations to be met. vl
en
'en"'
596 SURFACE MOUNT ATTACHMENT RELIABILITY

where the Weibull slope, or shape factor 13, is about 4 for leadless attachmene
and 2 to 3 for leaded attachment. 15.16.23 To our knowledge, the difference in Ws
for leadless and leaded attachments is not well understood.
For practical purposes, reliability requirements are stated in terms of the largest
acceptable cumulative failure probability per device,t F(n) .:;:;; F, at the end of
the product life (that is, at most a fraction F of the overall population is tolerated
to fail). Through the Weibull distribution, we can relate Nj to the number of
operation cycles N during the product life. Reliability requirements will be met
if the median design life Nd is less than median cycles to failure (see Fig. 18-
6). This condition is written as
1
_ [In(0.5)]ii _
Nd = !n(1 - F)
X N.:;:;;Nj (18-8)

which will account explicitly for the number of operation cycles and the reliability
requirements in the formulation of design tools. The fatigue laws and the sta-
tistical failure distribution then imply

(18-9)

where lld = ll-y for leadless attachment and lld = llW for leaded attachment.
The quantity Q on the left-hand side of Equation (18-9) is a relative measure of
damage accumulated during the median design life; it should be less than the
maximum allowable damage given on the right-hand side of Equation (18-9).

18.2 FIGURES OF MERIT FOR ATTACHMENT RELIABILITY

The above considerations of solder joint design, fatigue, and wearout have been
packaged as design for reliability tools. These FM tools address SM attachment
reliability at several stages of the design process:

• FM (comp) for component design and/or selection

• FM (assy) for component-substrate assembly design
• FM (env) for the reliability impact of the thermal environment
• FM (reI) for end-use attachment reliability.

The FM formulas are summarized hereafter for easy reference. This formulation
has been developed for and applies only to SM attachment using 60% Sn -
40% Pb or eutectic solder and leaded assemblies with a copper alloy lead frame.

tOne solder joint failure is one device failure.

 FIGURES OF MERIT FOR ATIACHMENT RELIABILITY 597

For leadless (II) SM solder attachment:

h
FM(comp,ll) == 20 L (18-10)

FM(assy,ll) == 8 x 10- 5 eC- Ll1o.

~ 1) (18-11)

h
FM(env,ll) == 3.2 x 10- 3 A"""""A (18-12)
Lao.aT
0.44

FM(rel,ll) _
= 0.32 (cycles 0.44
) A A
h 044
[In(1 - F)JI3 (18-13)
Lao.aTN . In(0.5)

For leaded (ld) SM solder attachment:

FM(comp,ld) 7 ::2 (107psi or 6.9 x 104 MPa)

== 10 (psi) (18-14)

FM(assy,ld) == 2.5 x 10-4 (psi x °C- 2) K(::o.)2

(18-15)
(2.5 x 10-4 psi X °C- 2 or 1.72 Pa X °C- 2 )

FM(env,ld) == 0.625 (psi) K(L::l1T)2 (0.625 psi or 4.3 X 103 MPa)

(18-16)
0.44

.
FM(rel,ld) == 87.5 (pSI X cyclesO. 44 )
Ah
l1 l1 2~ 44
[In(1 - F)JI3
K(L a T ) ' In(0.5)
(87.5 psi x cyclesO. 44 or 0.604 MPa x cyclesO. 44 )
(18-17)

The FMs are dimensionless, with increasing values implying higher attachment
reliability. There are no set lower bounds for FM (comp), FM (assy) and FM
(env); however, for FM (reI), a target of 1 or more is needed to meet reliability
requirements. The scaling constants in the FM formulas were selected such as:

1. FM(rel) ~ 1 warranties reliability requirements with an acceptable failure

probability per device F.
598 SURfACE MOUNT ATTACHMENT RELIABILITY

2. 0.7 < FM(rel) .::;; 1 is a region of uncertainty whether these requirements

are met.
3. FM(rel) .::;; 0.7 means that the SM design does not meet the reliability
requirements.

If condition 1 is not met, it is recommended to check for condition 2; the SM

design may still be acceptable when it falls in the gray area. A more detailed
study, including testing and extrapolation of test data to use conditions, is then
necessary to establish attachment reliability and capture the effect of design
parameters not included in the FM tools. It is also emphasized that the FM
formulation assumes complete stress relaxation for solder. In the case of leaded
attachment with very compliant leads, stresses in the solder joints are likely to
be low and this will be a conservative assumption. For field temperature excur-
sions with incomplete stress relaxation, cycles to failure predicted from the FMs
will be conservative; a more accurate prediction is then necessary that includes
the effect of dwell times and mean solder joint temperatures. 25 Note also that
the above formulation does not address t1!e effect of parameters such as joint
shape, mechanical support (pad size, conformal coating), or joint quality. For
a lack of data, their impact on attachment reliability is better assessed through
a carefully planned test program.

18.2.1 Derivation of FM Formulas

The reliability figures of merit, FM(rel), are the inverse of the cumulative solder
joint damage Q [see Equation (18-9)], with appropriate weighting factors D J and
D2 that scale the FM (reI) metric to a minimum of 1 for reliable attachment.
The FM(rel)'s are defined from Equation (18-9) as

FM(rel,ll) = Nim x d-y for leadless attachment (18-18)

and

FM(rel,ld) = Nim x dW for leaded attachment (18-19)

The other formulas are obtained by working backward from Equations (18-
18) and (18-19) and preserving in each FM the design parameters relevant to
that step of the process. Appendix 18-A gives the details of the FM derivation
and of the selection of scaling constants to fit the FM model to empirical reliability
 FIGURES OF MERIT FOR ATIACHMENT RELIABILITY 599

data. Again, the end-use FM(rel) should be larger than one for reliability re-
quirements to be met. While the other FMs do not predict reliability in an absolute
way, they are indicative of reliability potential and are useful for comparison
purposes. For example, since FM(comp) measures the ability of a component
to provide for reliable attachment in a given application, a new component design
with a larger FM(comp) than the current design is an acceptable alternative when
the latter meets expected reliability requirements.

18.2.2 FMs for Multiple Thermal Fluctuations

A design evaluation based exclusively on worst-case thermal fluctuations can be.

too conservative or lead to rejection of SM attachments that would survive typical
operating cycles and occasional worst-case fluctuations. In Appendix 18-B, we
show that Miner's rule26 provides a simple way of extending the FM formulation
to thermal cycles of various amplitudes. A generalized formulation of FM(rel)
for complex thermal fluctuations with various amplitudes IlT; during Ni cycles
is obtained as
I

0.32 h [In (1 - F)JDli

FM(rel,ll) = Lila X Su In (0.5) for leadless attachment

(18-20)

and
I

87.5 Ah [In (l - F)]Dli

FM(rel,ld) = K(Llla)2 x Sid In (0.5) for leaded attachment

(18-21)

where

Sn = (~Nj(IlTj)a) a
I

and
Sid = (~N(IlTj)2a) a
are measures of the cumulative environmental stress imparted to the SM assembly
by multiple thermal fluctuations in the field, and a = 1 /0.44.
600 SURFACE MOUNT ATTACHMENT RELIABILITY

18.2.3 Graphical Interpretation and Reliability Charts

For reliability at an acceptable cumulative failure probability F per device,

FM(rel) must be larger than or equal to 1. In terms of FM(assy), and for multiple
thermal fluctuations, this condition is equivalent to
I

in the leadless case: FM(assy,ll) ~ 2.5 X 10- 4 [ In(0.5) J~ X S

In(1 - F) IL

(18-22)
I

in the leaded case: FM(assy,ld) ~ 2.86 X 10- 6 [ In(0.5) J~ X Sid

In(1 - F)
(18-23)

These equations signify that the SM attachment is robust enough to withstand

the damage imparted to solder joints throughout the product life. On the left-
hand sides of Equations (18-22) and (18-23), the FM(assy)'s group the design
parameters that govern attachment reliability and, as such, serve as measures of
assembly robustness. The quantities on the right-hand sides are relative measures
of the cumulative environmental stress.

c-________________________________F~= 10~

5.00 10.5
10 -4

i
E
10 ·3
10 ·2
10.1
l!
:. 1.00
c
C)

.~ 0.50
Q

~ 0.10

0.05 .........'--''''--_ _ _ _.l..-_~L..---_ _ __ I I

100 500 1000 5000
Environmental "Stress": ~,(N" ~ 7;>

Figure 18·7a Chart for attachment reliability estimates for SM leadless components.
 EXAMPLES 601

FM (asay,ld)
• DESIGN
PARAMETERS·

10,000 SO,OOO 100,000 500,000

ENVIRONMENTAL "STRESS": S. (N, !J. 1)

Figure 18-7b Chart for attachment reliability estimates for SM leaded components.

The above interpretation leads to a graphical estimate of attachment reliability

using the charts in Figure 18-7a and b. On the horizontal axis, SII or Sid combine
the cycles of operation and temperature fluctuations for complex thermal his-
tories; on the vertical axis, FM(assy) includes the important design parameters.
The lines of constant cumulative failure probability F = x are parallel in the
log-log coordinate systems: (SII, FM(assy,ll» for leadless devices, and (Sid'
FM(assy,ld» for leadless devices. Given the design parameters and use condi-
tions, a representative point above the line F = x ensures a cumulative failure
level less than x, per device, at the end of the design life.

18.3 EXAMPLES
18.3.1 Example 18·1: Chip Components on FR-4 Printed Wiring Boards

The use of FMs is illustrated here for size code 1206 and 1825 1 chip components
surface mounted on an FR-4 printed wiring board. The intended service life is
10 years with a reliability target F < 10-4 (0.01 percent) at the end of life. The
use thermal environment is described by a daily temperature fluctuation of 20°C
at the circuit board level. The relevant design data are

L = {0.047 in for the 1206s

0.075 in for the 1825s
h = 0.005 in.
an = 9 x lQ"6/oC
602 SURFACE MOUNT ATTACHMENT RELIABILITY

From equation (18-11), we get

FM(ass 11) = {0.946 for the 1206s

y, 0.593 for the 1825s

The cumulative environmental stress is described by

NJ = 3650 cycles (1 cycle/day, 10 years), aTI = 20°C

I

S 11 = (3650 X 20°·44)0.44 = 738

Figure 18-8 shows that, at 10 years, the representative points A and B are above
the line F = lQ-4 and reliability requirements are met for both the 1206 and the
1825 chips.
For a similar application with an extended service life of 20 years:

NI = 7300 cycles (1 cycle/day, 20 years), aTI = 20°C

I

S 11 = (7300 X 20°·44)0.44 = 1002

r__---------;-~___,r__----F____.:T"1= 10.6
5.00 10.5
10 -4
10 ·3
10.2
10.1
~ 1.00
~
Vl
Vl 0.50
~
::2
u. ,
, ,
!f\ !f\
0.10
, 'tlT=20·C
O 05 '10,' 20 YEARS
. ~~'---------~~--~----------~
100 500 1000 5000

Figure 18-8 Example 18·1: Attachment reliability estimates for size 1206 and 1825 discrete
chips on FR·4, using the design-reliability chart for lead less assemblies (Fig. 18·7a).
 EXAMPLES 603

The data in Figure 18-8 is shifted to the right, from A and B to C and D. Point
C for the 1206 chips shows the reliability requirements are still met at 20 years,
with a predicted cumulative failure probability of less than 10--5 (10 ppm). Point
D for the 1825s falls below the line: F = 10-4 and reliability is not proven in
by the FM tools. Checking for the gray area, we find FM(rel,ll) - 0.9, suggesting
that the SM design for the larger chips may still be acceptable. Indeed, accelerated
testing 27 shows that 180 mil long is an acceptable size for cylindrical resistors
on FR-4 in an outside installation with a slightly more severe thermal environment
than the one described here.
Finally, for a more severe environment with a daily temperature fluctuation
of 35°C for 10 years, the data in Figure 18-8 move from A and B to G and H.
Point G is above the target and the 1206s meet reliability requirements. Point
H for the 1825s is below target; in this harsher environment, the expected
cumulative failure rate at 10 years is between 0.1 and 1 percent.

18.3.2 Example 18-2: 50-Mil Pitch Ceramic Leaded Devices on FR-4

Next, we apply the FMs to 50-mil pitch ceramic leaded devices. The design
data for a 100 I/O device with J-clip leads on FR-4 are given as

L = 0.919 in. A = 800 X 10- 6 in. 2 h = 0.005 in.

K = 58Ib/in.

II
z

.
I
5000 10.000 50.000 100.000 500.000

Figure 18-9 Example 18-2: Attachment reliability estimates for S- and J-clip lead ceramic
leaded chip carrier on FR-4, using the design-reliability chart for leaded assemblies (Fig. 18-
7b).
604 SURFACE MOUNT ATTACHMENT RELIABILITY

From Equation (18-15), we get FM(assy,ld) = 0.23. A useful life of 10 years

is targeted in a thermal environment described by 3650 thermal fluctuations of
15°C and 120 fluctuations of 60°C:
2 2

Sid = (3650 X 15°·44 + 120 X 60°·44)0.44 = 3.03 X 104

The representative point on Figure 18-9 shows an unacceptable cumulative failure

level of about 1 percent at the end of the desired service life. Using a more
compliant S-clip lead (e.g., reference 16) for which K - 10 lb/in., FM(assy,ld)
= 1.31 results in a more reliable attachment with F < 10-5 at 10 years (see
Fig. 18-9).

18.3.3 Example 18-3: 25- and 50-Mil Pitch Plastic Leaded Components
on FR-4

For similar substrates and environments, the FMs are a function of package size,
lead compliance, and lead-solder wetting area, thus allowing a direct comparison
of attachment reliability within a family of devices. In this example, we use the
FM approach to compare attachment reliability for 25-mil pitch plastic quad flat
packs (PQFPs) and 50-mil pitch plastic leaded chip carriers (PLCCs) on FR-4.
The analysis illustrates further the type of information that can be obtained by
using the FM tools.
The design data are given in Table 18-1 for 68-110 PLCCs and PQFPs with
84, 132, and 244 1I0s. The structural models that were used to determine the
diagonal lead stiffnesses are shown in Figure 18-10. The PQFP gull-wing lead
is compliant with a low K - 12 lb/in., to be compared to K - 55 Ib/in. for the
68110 PLCC J lead. The thinner lead frame of the PQFP gull wings is a significant
contributor to their compliance. 24 The FM(assy,ld) data for the PQFPs cover the
design range for the lead-solder wetted area. The range of FM(assy,ld) for the
132-110 PQFP is comparable to that of the 68-110 PLCC; by design, attachment
reliability is expected to be similar for these two packages.
Two thermal environments are considered:

• A controlled ambient, central office with a typical daily fluctuation of 10°C

and an occasional surge of 30°C associated with air conditioning failure,
say at most once a month. For a 20-year system life, the environmental
stress is described by:

N! = 7300 cycles (1 cycle/day, 20 years), flT! = 10°C.

N2 = 240 cycles (1 cycle/month, 20 years), flT2 = 30°C
2 2
 EXAMPLES 605

Table 18-1 Design Data for Plastic Packages on FR-4 Circuit Boards
L K A h ~a FM
Package (in.) (Ib/in.) (10~in. 2 ) (mil) (l~/oC) (assy,ld)
68-1/0 PLCC 0.674 55 810 4 4 2.03
84-110 PQFP 0.531 12 240-290 3 4 3.32-4.02
132-110 PQFP 0.743 12 240-290 3 4 1.70-2.05
244-110 PQFP 1.146 12 240-290 3 4 0.71-0.86

• An outside installation with an averaged daily fluctuation of 30°C and a

system life of 20 years for which

NJ = 7300 cycles (1 cycle/day, 20 years), IlTJ = 30°C

2

The results for both environments are shown in Figure 18-11 where the
FM(assy,ld) data for the PQFPs are plotted as bands that cover the design range
for the lead-solder wetted area. Also shown are the FM(assy,ld) lines for the
132-1/0 PQFP when the lead-solder wetted area is reduced by 50 percent.
In the first scenario of a central office environment, the cumulative failure
probability at end of service life is predicted to be smaller than 10--5 (10 ppm)
for all packages. Even the large 244-1/0 PQFPs, over 1.6 in. on the side, have
a large reliability margin for this application. When the lead-solder wetted area
is reduced by 50 percent, due for example to lead offset or a reduced solder
joint length, the 132-1/0 PQFPs still have adequate attachment reliability. This
suggests that, by design, these devices are robust enough to tolerate minimum
assembly defects.
For the more severe environment of an outside plant installation, the cumu-
lative failure probability at 20 years is less than 10-5 (10 ppm) for the PLCCs
and 84- or 132-1/0 PQFPs. For the 244-1/0 PQFPs, we predict: 1(J4 ::::;; F ::::;;
10-3 , which is unacceptable for an application with stringent reliability require-
ments. Here, the size of fine-pitch PQFPs may have to be limited to no more
than 132 I/Os.
Last, some lead design considerations are discussed for the PQFPs as they
relate to design choices having reliability implications. The PQFP packages are
made to JEDEC standards 28 where the lead width is in the range 8 to 12 mil.
In the PQFP model shown in Figure 18-10, the width of the lead below the cross
section taper is 8 mil. For a larger width of 12 mil, the diagonal lead stiffness
increases by a trivial amount, less than 1 percent,24 while the lead-solder wetted
area increases by 25 percent. FM(assy,ld) increases since it goes as A / K, and
the larger width at the tip of the lead is beneficial to attachment reliability. A
606 SURFACE MOUNT ATTACHMENT RELIABILITY

PQFP 68 I/O PLCC

(25 MIL PITCH) (50 MIL PITCH)

~
: 6:
~
! 8!

-' @
18
T
9
~
! 6!
~ 8 ~

Figure 18-10 Structural models used to determine the diagonal stiffness of PQFP and PlCC
corner leads (Example 18-3).

5.00

1.00
-
~
.!!
0.50

:::E
L1.

OUTSIDE
PLANT,
20 YEARS

5000 10,000 50,000 100,000 500,000

s,. (N, 47)
Figure 18-11 Example 18-3: Attachment reliability estimates for PlCCs and PQFPs on FR-
4, using the design-reliability chart for leaded assemblies (Fig. 18-7b).
 FM FORMULAS AND SCALING CONSTANTS 607

unifonn, 12-mil-wide cross section would also simplify the lead frame manu-
facturing; however, the potential for solder bridging during reflow needs to be
considered before recommending a lead foot 12 mil wide.

18.4 CONCLUDING REMARKS

The high reliability levels expected of today's electronic assemblies make at-
tachment reliability a critical issue. SMT is a reliable technology, with good
design and assembly practices the key to end-use reliability. Upfront design for
reliability, using tools such as described in this chapter, can help packaging
engineers and physical designers build in attachment reliability.

18.5 ACKNOWLEDGMENTS

We would like to acknowledge the contribution of Werner Engelmaier and Robert

Kotlowitz of AT&T Bell Laboratories, Whippany, NJ, to the initial fonnulation
of figures of merit. We are also grateful to Peter M. Hall, Chi K. Ng, and
Ephraim Suhir for their time and effort in reviewing this paper.

Appendix 18-A
FM Formulas and Scaling Constants
At the component level, the parameters that affect attachment reliability and can be
influenced by the package designer are the component dimensions and to some degree
the attachment design: solder paste thickness, leadless or leaded attachment, and for
leaded packages, the lead compliance. The FMs at the component level are formulated
as:

h
FM(comp,ll) = R(comp,ll) L for leadless (I8-AI)

and

Ah
FM(comp,ld) = R(comp,ld) KL2 for leaded (18-A2)

In these and the next FM equations, the R (. ,.) factors are scaling constants that lead
to dimensionless FMs. The constants R(comp,ll) = 20 in.lin. (mrnIrnrn) and R(comp,ld)
= 107 Ib/in. 2 (6.9 X I<r MPa) are arbitrary; their values reflect the combinations of
design parameters for highly reliable SM attachments. For leadless components, h is the
component standoff height; for leaded components, h is taken as half the solder paste
thickness (e.g., h = 5 X 10-3 in. for lO-mil paste, h = 3 X 10-3 in. for fine-pitch 6-
608 SURFACE MOUNT ATTACHMENT RELIABILITY

mil paste. L is the distance from the neutral axis of the component to the center line of
the further solder joint. For a large chip carrier, L is the distance from the neutral axis
of the package to a comer joint, about half the diagonal dimension of the package. For
small, discrete or chip components, L is from the neutral axis to the center line of the
end solder joints (e.g., L = 47 mil for a l20-mil-Iong 1206-chip resistor).
At the next next stage of component-substrate assembly design, designers select a
substrate material. Each component has an FM(assy) that includes the component-substrate
CTE mismatch iln:

R(assy,ll)
FM(assy,ll) = FM(comp,ll) for leadless (18-A3)

and

R(aSSY,ld») 2
FM(assy,ld) = FM(comp,ld) ( iln for leaded (l8-A4)

withtheRconstants specified as R(assy,ll) = 4 x IQ-6/oC and R(assy,ld) = 5 x IQ-6I"C.

The impact of power dissipation or nominal temperature fluctuations (ilT) at the board
level is captured in the environment FMs:

R(env,lI)
FM(env,ll) = FM(assy,ll) ilT for leadless (l8-A5)

and

R(env ,ld)] 2
FM(env,ld) = FM(assy,ld) [ ilT for leaded (l8-A6)

with R(env,ll) = 40°C, and R(env,ld) = 50°C.

Last, the reliability FMs include the number of expected cycles of operation N and the
acceptable cumulative failure probability F per device at the end of the design life:

FM(rel) =
R(rel) [In(1 -
FM(env) ~.44 In(0.5)
F)]T 0.44

(l8-A7)

with R(rel) = 100 cyclesl!2 in the leadless case and R(re!) = 140 cyclesl!2 in the leaded
case.
The R(rel) constants were selected such that FM(re!) ;;;. I when reliability requirements
are met, at least to the specified acceptable failure probability. A plot of FM(env) versus
predicted median cycles of operations (N = N 50% for F = 50 percent) was generated
(Fig. IS-AI) using available reliability data for 13 types of leadless and leaded assemblies
for which design parameters and predicted field lives have been documented. 3 •4 ,5.16,25.27
Since FM(env) is inversely proportional to !J,:y in the leadless case and ilW in the leaded
case, Figure IS-AI bears similarity to an inverted Coffin-Manson plot. 14 The slope of
the regression lines for either the leadless or the leaded data is 0.44, a more accurate
value than the 0.5 that was previously used. 4 ,5 For applications with short dwell times,
 FM FORMULAS AND SCALING CONSTANTS 609

5.0

___ FM(env) = -1N 0.44

___ ~~

% 1.0 100
~

0.5

0.1
10 2 10 3 10 4 10 5 10 6
MEDIAN CYCLES OF OPERATION, N = N
50%
Figure 18-Al Fit of the lead less and leaded FM models to empirical data, using median
cycles to failure for 13 tested assemblies.

F= 10-2

5.0

1.0

0.5

* ,"II" = LEADLESS
X ,"'d" = LEADED

10
CYCLES OF OPERATION, N = N 0.01%, N1%, N 50%
Figure 18-A2 Fit of the FM models to the empirical data at several levels of cumulative
failure probabilities. FM (env) versus cycles of operation showing the fit of the FM model: FM
(rei) = 1, to the data at acceptable cumulative failure probabilities F = 10-4 (0.01 %), 10-2
(1 %), and 50%.
610 SURFACE MOUNT ATIACHMENT RELIABILITY

a more sophisticated model and a variable exponent should be used to account for
incomplete stress relaxation. 25 The R(rel) constants were obtained by fitting the FM model
FM(rel) = 1 at F = 50 percent, to the regression lines in Figure 18-Al. They are actually
the inverse of these line intercepts with the FM(env) axis. Figure 18-A2, where we have
plotted the data and the lines FM(rel) = 1 at F = 10-4, 10-2 , and 50 percent, shows
that the fit of the FM model to the data is preserved at all failure levels. Because of the
simplifications of the FM model and uncertainties in some of the design parameters, a
design for which FM(rel) is slightly less than 1 may still be acceptable. A decrease in
FM(rel) from 1 to 0.7 corresponds to a change in predicted life by a factor of 2 or an
order of magnitude in predicted failure level. In this gray area 0.7 < FM(rel) ... 1,
reliability requirements may still be met; however, a more detailed analysis than provided
by the FM tools, and perhaps an accelerated test program, are necessary to prove in
reliability. For FM(rel) ... 0.7, the SM attachment design does not meet reliability re-
quirements.

Appendix 18-8
FMs for Multiple Thermal Fluctuations
SM assemblies in the field are subject to complex thermal histories. As a first cut, the
thermal history is best described by series of cycles of variable amplitudes. For example,
the thermal history in a controlled environment for telecommunication equipment consists
of typical daily thermal fluctuations and occasional temperature surges that occur when
cooling equipment (fans or ambient air conditioning) fails.
Thermal cycles of amplitudes llTi use up a fraction N/NI,i of the useful fatigue life,
where Ni is the number of operation cycles at amplitude llTi' and NI .i is the number of
cycles to failure at the same amplitude llTi. Using Miner's rule,26 the number of cycles
to failure for multiple thermal cycles is then estimated from the relationship
N] N2 N3
- + - + - ... =D"'l (18-B1)
NI ,] NI,2 NI ,2

where D is the cumulative fatigue damage. The formulation of Miner's rule is based on
the net energy absorbed at failure, as is the cyclic fatigue life for solder joints. While
Miner's rule has not been proven in for solder joint fatigue, it is a reasonable first approach
at handling cumulative fatigue damage in SM attachments.
From the reliability condition FM(rel) = 1 for failure in the field, we get the cycles
to failure at an acceptable failure level F as

)4 [In(1In (0.5)
]

N (F) = (0.32 h - F)]~ for leadless attachment (18-B2)

I LllallT

and

)4 [InIn (0.5)F)]~
]

( 87.5 Ah (1 -
for leaded attachment (18-B3)
NJ<F) = K(LllallT)2

with a = 1/0.44.
 REFERENCES 611

Substituting Equations (l7-B2) and (17-B3) into Equation (l7-Bl) for cumulative
fatigue damage gives

N (!:J.T)a + N (!:J.T)a + ... = D (0.32 h)" [In (I - F)].!. (l8-B4)

I I 2 2 L!:J.a. In (0.5) p

for leadless attachment

N (!:J.T)2o + N (!:J.T)2o + ... = D (87.5 Ah)a [In (I - F)].!. (l8-B5)

I I 2 2 K(L!:J.a.)2 In (0.5) p

for leaded attachment

The generalized formulation of FM(rel) for multiple thermal loads derives from the
above combinations of N;'s and !:J.Tj's:

FM(rel,ll) = 0.32 h
L!:J.a. x SII
[In (I - F)]~
In (0.5)
for leadless attachment (18-B6)

and

FM(relld) = 56.25 Ah [In (I - F)]~ for leaded attachment (18-B7)

, K(L!:J.a.)2 x SId In (0.5)

where

and

are measures of the cumulative environmental stress imparted to the SM assembly by

multiple thermal fluctuations.

REFERENCES

I. IPC-SM-782, "Surface Mount Land Patterns (Configurations and Design Rules),"

Institute for Interconnecting and Packaging Electronic Circuits, Lincolnwood, IL,
March 1987.
2. Lau, J. H., and D. W. Rice, "Solder Joint Fatigue in Surface-Mount Technology:
State of the Art," Solid State Technology, Oct. 1985, pp. 99-100.
3. Engelmaier, W., "Fatigue Life of Leadless Chip Carriers Solder Joints During Power
612 SURFACE MOUNT AITACHMENT RELIABILITY

Cycling," IEEE Transactions on Components, Hybrids and Manufacturing Tech-

nology, CHMT-6, September 1983, pp. 232-237.
4. Clech, J.-P., W. Engelmaier, R. W. Kotlowitz, and J. A. Augis, "Surface Mount
Solder Attachment Reliability Figures of Merit-'Design for Reliability' Tools,"
Proceedings of the SMART V Conference, New Orleans, LA, Technical Paper SMT
V-48, January 9-12, 1989.
5. _ _ , "Reliability Figures of Merit for Surface-Soldered Leadless Chip Carriers
Compared to Leaded Packages," IEEE Transactions on Components, Hybrids and
Manufacturing Technology, 12(4), December 1989, pp. 449-458.
6. Norris, K. C., and A. H. Landzberg, 1969, "Reliability of Controlled Collapse
Interconnections," IBM Journal of Research and Development, May 1969, pp. 266-
271.
7. Howard, R. T., S. W. Sobeck, and C. Sanetra, "A New Package Related Failure
Mechanism for Leadless Ceramic Chip Carriers (LC-3s) Solder Attached to Alumina
Substrates," Solid State Technology, February 1983, pp. 115-122.
8. Hall, P. M., "Forces, Moments, and Displacements During Thermal Chamber Cy-
cling of Leadless Ceramic Chip Carriers Soldered to Printed Boards," IEEE Trans-
actions on Components, Hybrids and Manufacturing Technology, CHMT-7(4), De-
cember 1984, pp. 314-327.
9. _ _ , "Creep and Stress Relaxation in Solder Joints in Surface Mounted Chip
Carriers," IEEE Transactions on Components, Hybrids and Manufacturing Tech-
nology, CHMT-IO(4), December 1987, pp. 556-565.
10. Stone, D., S.-P. Hannula, and C.-Y. Li, "The Effects of Service and Materials
Variables on the Fatigue Behavior of Solder Joints During the Thermal Cycle,"
Proceedings of the 35th Electronic Components Conference, New York, 1985, pp.
46-51.
11. Hall, P. M., and W. M. Sherry, "Materials, Structures and Mechanics of Solder
Joints for Surface Mounted Microelectronics Technology," Lectures of the 3rd In-
ternational Conference on Interconnection Technology in Electronics, Fellbach, W.
Germany, Feb. 18-20, 1986, pp. 47-61.
12. Clech, J. -P., and J. A. Augis, "Engineering Analysis of Thermal Cycling Accelerated
Tests for Surface-Mount Attachment Reliability Evaluation," Proceedings of the
1987 International Electronic Packaging Conference, Boston, MA, November 9-
11, 1987, pp. 385-410.
13. Lovasco, F., "Solder Joint Modeling Using Finite Element Analysis," Proceedings
of the 1988 International Electronic Packaging Conference, Dallas, TX, November
7-10, 1988, pp. 410-415.
14. Manson, S.S., Thermal Stress and Low-Cycle Fatigue, McGraw-Hill, N.Y., 1966.
15. Balde, J., "The IEEE Compliant Lead Task Force," IEEE Transactions on Com-
ponents, Hybrids and Manufacturing Technology, CHMT-IO(3) September 1987,
pp. 463-469.
16. Engelmaier, W., "Surface Mount Attachment Reliability of Clip-Leaded Ceramic
Chip Carriers on FR-4 Circuit Boards," International Electronic Packaging Society
Journal, 9(4), January 1988, pp. 3-11.
17. Kotlowitz, R. W., "Comparative Compliance of Generic Lead Designs For Surface
Mounted Components," Proceedings of the 1987 International Electronic Packaging
Conference, Boston, MA, November 9-11, 1987, pp. 965-984.
 REFERENCES 613

18. _ _ , "Comparative Compliance of Representative Lead Designs for Surface Mounted

Components," Proceedings of the 1988 International Electronic Packaging Society
Conference, Dallas, TX, November 7-10, 1988 pp. 908-948.
19. _ _ , "Compliance of Surface Mount Component Lead Designs With Rectangular
and Circular Cross-Sections," Proceedings ofthe 1989 International Electronic Pack-
aging Conference, vol. II, San Diego, CA, September 11-13,1989, pp. 1071-1085.
20. Clech, J.-P., and J.A. Augis, "Temperature Cycling, Structural Response and At-
tachment Reliability of Surface-Mounted Leaded Packages," Proceedings of the 1988
International Electronic Packaging Conference, Dallas, TX, November 7-10, 1988,
pp. 305-325.
21. Morrow, J.D., "Cyclic Plastic Strain Energy and Fatigue of Metals," ASTM STP
378, ASTM, Philadelphia, 1964, pp. 45-87.
22. Lemaitre, J., "Evaluation of Dissipation and Damage in Metals Submitted to Dynamic
Loading," Proceedings of the International Conference on Mechanical Behavior of
Materials, vol. II (Fatigue of Metals), August 15-20, 1971, pp. 540-549.
23. Hines, L. L., 1988, AT&T Bell Laboratories, Columbus, OH, private communi-
cation.
24. Clech, J.-P., M. J. Tervalon, and J. A. Augis, "Reliability Evaluation of 25 and 50
Mil Pitch Surface Mounted Plastic Leaded Packages," Proceedings of the 1989
International Electronic Packaging Conference, vol. II, San Diego, CA, September
11-13, 1989, pp. 754-758.
25. Engelmaier, W., 1989, AT&T Bell Laboratories, Whippany, NJ, private commu-
nications.
26. Miner, M. A., "Cumulative Damage in Fatigue," Journal of Applied Mechanics,
September 1945, pp. A-159-A-184.
27. Ng, K. K., "Assessing Surface Mount Attachment Reliability of Ceramic Cylindrical
Resistors in FR-4 Circuit Boards," Technical Proceedings, Expo SMT '88, Las Vegas,
NY, November 1988, pp. 1-6.
28. JEDEC, JEDEC Registered and Standard Outlines for Solid State Products, JEDEC
Publication no. 95, Electronic Industries Association, November 1988.
 Authors' Biographies

Abbas I. Attarwala-Research Assistant at the Center Strategic Materials at

University of Texas-working on his PhD. He received his BS and MS degrees
in metallurgy/material science from Columbia University in New York. During
his junior and senior years he worked at AT&T Bell Laboratories as a student
employee studying the failure models of fatigue-cycled leadless surface-mount
components on circuit boards. He was awarded the Francis Rhode Prize from
the Henry Krumb School of Mines at Columbia University. Mr. Attarwala is a
member of Tau Beta Pi and the AIME.

Jacques A. Augis-Group Supervisor, Reliability of Surface Mount Tech-

nology for AT&T Applications-AT&T. Dr. Augis joined AT&T Bell Labo-
ratories, Columbus, Ohio, in 1968, working on the "Electromechanical Appa-
ratus" to improve the active life and reliability of Wire Spring and Reed Relays.
After several other assignments related to device quality and reliability, he was
transferred to the Interconnection Laboratory, Whippany, New Jersey, in 1982,
working on Printed Wiring Board reliability. Dr. Augis received the Ingenieur
degree from the Ecole Superieure de Physique et Chimie Industrielles, Paris, in
1959, and the PhD degree in physics from Case Western Reserve University,
Cleveland, Ohio, in 1969. Dr. Augis has published about 25 technical papers,
and holds four U.S. patents.

615
616 AUTHORS' BIOGRAPHIES

Jean-Paul M. Clech-Member of the Technical Staff-Advanced Computer

Research Institute, Lyon, France. He was previously with AT&T Bell Labora-
tories in Whippany, New Jersey. Within the Interconnection Systems and As-
sembly Reliability Department, he has responsibility for surface mount assembly
and reliability issues.
Dr. Clech graduated from Ecole Centrale de Paris where he received his
Diplome d'Ingenieur in 1981. He also holds an associate degree in economics
from the Sorbonne. He received the MS and PhD degrees in mechanical engi-
neering from Northwestern University in 1982 and 1985, respectively.
Dr. Clech has published and presented papers in the areas of engineering
mechanics, finite element modeling, thermomechanical fatigue and fracture,
surface mount attachment and assembly reliability. An active member of ASME,
IEEE and IEPS, he has chaired workshops and technical sessions at numerous
conferences.

Werner Engelmaier-President-Engelmaier Associates, Inc. Engelmaier

Associates, Inc. is a firm providing consulting services on reliability, manufac-
turing, and processing aspects of electronic packaging. This consulting company
provides services to OEMs, printed circuit board manufacturers, PWB subcon-
tract assembly companies and end users, such as telecommunications and com-
puter companies and the military, in the areas of design for manufacturability
and quality, processing effects, yield improvement, environmental stress screen-
ing, design for reliability, accelerated reliability testing, reliability prediction,
and failure mode analysis for solder joints, plated-through holes, and dynamically
used flexible circuitry.
Prior to forming his own company, Mr. Engelmaier was a Distinguished
Member of Technical Staff at AT&T Bell Laboratories, Whippany, New Jersey.
During his 24 years at Bell Labs, until his retirement in 1990, he dealt with
many of the aspects involved in electronic packaging and interconnection tech-
nology. His efforts within the Bell System! AT&T ranged from R&D at Bell
Labs to 'fire-fighting' at Western Electric, and included the development of
specifications and guidelines, as well as technology transfer to production fa-
cilities.
Born in Vienna, Austria, Mr. Engelmaier holds the Mechanical Engineering
degrees from Technologisches Gewerbe-Museum (TGM), Vienna, Austria, the
University of South Carolina, and Massachusetts Institute of Technology. He is
a member of the American Society of Mechanical Engineers, the American
Society for Testing and Materials, the IEEE Components, Hybrids, and Man-.
ufacturing Technology Society, the Institute for Interconnecting and packaging
Electronic Circuits, the International Electronics Packaging Society, the Surface
Mount Technology Association, Phi Beta Kappa, Pi Mu Epsilon, Sigma Xi, and
Tau Beta Pi.
He has conducted numerous tutorial workshops for AT&T, NEPCON, IEPS,
IPC, and ISHM. He has contributed to numerous task forces, specifications,
 AUTHORS' BIOGRAPHIES 61 7

standards, and design guides. He is on the Editorial Advisory Board of Electronic

Packaging & Production magazine and chairs committees at IPC and ASTM.
He has been the recipient of many honors and awards, including the fPC Pres-
ident's Award in 1988, the Electronic Packaging Achievement Award from
IEPS/Electronic Packaging & Production magazine in 1987, and the Distin-
guished Technical Staff Award from AT&T Bell Laboratories in 1986. For two
of his developments he received R&D 100 Awards in 1978 and 1981, from
Research & Development magazine. Mr. Englemaier is the author of over 90
technical publications in a variety of fields related to electronic packaging and
interconnection technology, and has been awarded a number of patents on these
areas.

Morris E. Fine-Walter P. Murphy Professor Emeritus-Materials Science

& Engineering, McCormick School of Engineering and Applied Science of
Northwestern University, Evanston, Illinois. Dr. Fine is currently Director of
the AlSI-sponsored Steel Resource Center at Northwestern. He helped to organize
the first Materials Science Department in the United States. He was on the
Manhattan Project at both the Chicago and Los Alamos sites, and a member of
the Technical Staff of the Bell Telephone Laboratories, Murray Hill, N.J., 1946-
54. His formal education was at the University of Minnesota, where he was
awarded the BMetEng with distinction in 1940, and a PhD in 1943. His profes-
sional society activities include Chairman of the Institute of Metals Division of
The Metallurgical Society of AIME, and member of the Board of Directors of
AIME and of The Metallurgical Society of AIME. Dr. Fine is a Fellow of five
societies: the American Society for Metals, The Metallurgical Society of AlME,
the American Physical Society, the American Ceramic Society, and the American
Association for the Advancement for Science. He was elected to the National
Academy of Engineering in 1973, where he was Chairman of the Membership
Committee in 1979. He has won a number of awards, including the ASM Gold
Medal, TMS Mathewson Medal, and the AIME James Douglas Gold Medal.
Dr. Fine is author or coauthor of over 200 papers. His research interests include
alloy development, phase transformations in solids, the the.ory of strength of
metals, alloys and composites, magnetic properties of metals and ceramics,
fatigue of metals and tribology of metal and ceramic matrix composites.

Les Fox-Digital Equipment Corporation. Since 1981, Dr. Fox has been
working on advanced development of high-density packaging technologies for
VLSI. He is especially interested in thermal phenomena, and has developed
several proprietary chip cooling technologies while at Digital. Dr. Fox is es-
pecially interested in the materials science and reliability physics of advanced
electronic packaging and interconnect systems. Dr. Fox holds the ScB in physics
from Brown University, and the PhD in low-temperature solid state physics from
MIT.
Dr. Fox is a member of IEEE-CHMT, Chairman of IEEE-CHMT Technical
618 AUTHORS' BIOGRAPHIES

Committee TC-6 on Electronic Packaging, and a member of IEPS. He has

published several papers on original investigations on the reliability of solder
joint interconnects.

Harold J. Frost-Faculty Member-Thayer School of Engineering. Dr. Frost's

primary research interests are in high-temperature deformation of metals and
alloys, the structure of defects and grain boundaries, and the evolution of mi-
crostructure through phase transformation and grain growth. He received his AB
in engineering and applied physics, as well as his PhD in materials science,
from Harvard. He has done post-doctoral work at Cambridge University, MIT,
and Harvard. Dr. Frost is a member of the Materials Research Society, the
American Society of Metals, and the Metallurgical Society of the AIME.

Dennis Grivas-Research Engineer-Center for Advanced Materials of the

Lawrence Berkeley Laboratory. Dr. Grivas received the PhD degree in material
science and engineering at the University of California at Berkeley in 1978.

Leslie A. Guth-Distinguished Member of Technical Staff-AT&T Bell

Laboratories' Engineering Research Center in the Environmental Materials and
Processes Group in Princeton, N.J. Dr. Guth's research interests include sol-
dering flux and cleaning technology. She has a BA in physics from the University
of North Carolina-Chapel Hill, and a PhD in materials science from the University
of Pennsylvania. Dr. Guth is an active memberofthe Institute for Interconnecting
and Packaging Electronic Circuits (IPC) , participating on several committees.
She chairs the IPC Electromigration Task Group and the IPC Test Monitoring
and Validation Committee for the Industry/DoDIU.S.EPA cooperative effort in
testing CFC alternatives. For her work in IPC, Dr. Guth received the 1988-
1990 IPC President's Award. She was also a member of the 1989 United Nations
Environmental Program's Soivents Technical Options Assessment Committee,
which evaluated CFC-113 alternative technologies. This report was used in
generating additional restrictions and modifications to the 1987 Montreal Protocol
phase-down. Dr. Guth holds two patents and has numerous publications.

Peter M. Hall-Distinguished Professor of Physics-Johnson C. Smith Uni-

versity, Charlotte, NC. Dr. Hall teaches physics, is in charge of a pre-engineering
program, and serves on the faculty of the Honors College. He has 31 years
experience at AT&T Bell Labs in the fields of thin films, superconductivity,
electronics packaging, and solder mechanics. Dr. Hall received his BA in physics
from Hobart College, and his PhD in physics from Iowa State University. He
is a Fellow of IEEE and a Bell Labs Fellow. He serves on the program committee
of the Electronic Components and Technology Conference and is program chair-
man of the 1990 VLSI and GaAs Packaging Workshop. He also serves as
Associate Editor of the AS ME Transactions of the Journal of Electronic Pack-
 AUTHORS' BIOGRAPHIES 619

aging. Dr. Hall has published over 45 papers, and is coauthor of Thin Film
Technology.

Bryan C. Hendrix-Research Scientist-University of Texas at Austin. In

1981, Dr. Hendrix joined CSTA at the U. S. Army Aberdeen Proving Grounds
as a junior engineering fellow. He joined the Center for Strategic Materials at
Columbia University in 1985 as a research assistant. Dr. Hendrix received his
BS, MS, and DES degrees in metallurgy and materials science from Columbia
University in New York. He was awarded the Francis Rhode Prize from the
Henry Krumb School of Mines, and the "Pop" Herrington medal from the School
of Engineering and Applied Science at Columbia University. Dr. Hendrix is a
member of MRS, ASM, TMS/AIME, and Tau Beta Pi.

Charles L. Hutchins-Independent Consultant-Dr. Hutchins is recognized

worldwide for his expertise in Surface Mount Technology. His experience in-
cludes 22 years at Texas Instruments, where he served in the capacities of
Engineering Manager, Quality Department Manager, and Product Manager. Dr.
Hutchins was most recently responsible for engineering management at TI's
SMT Center for 3 years, specifically process engineering, SPC implementation,
yield improvement, product design and development, and the Customer SMT
Training Program. Since leaving n, he has worked as a technical and managerial
consultant to several major companies and conducted numerous SMT training
sessions. He is also an Adjunct Professor at the National Training Center for
Microelectronics, Northampton Community College, Bethlehem, PA. Dr. Hutchins
holds PhD and MS degrees in electrical engineering from North Carolina State
University.
He is a past Vice President of the Surface Mount Technology Association,
with responsibility for the Technical Session of SM '89, the technical conference
for the SMTA. Dr. Hutchins is the author of over 25 technical papers, in addition
to his textbook Surface Mount Technology-How to Get Started. He also writes
a monthly column on SMT for Electronic Engineering Times and serves on the
Editorial Advisory Board of Surface Mount Technology Magazine.

Jennie S. Hwang-President-International Electronic Materials Corp. Prior

to this appointment, Dr. Hwang held senior managerial and research positions
with SCM Corp., Martin Marietta Corp., and Sherwin-Williams Company.
In conjunction with multidisciplinary trainings, Dr. Hwang's 14 years of
experience as technology/business manager covering various industries has brought
her a unique breadth and depth of knowledge of materials and processes from
both the theoretical and practical aspects.
Dr. Hwang received her PhD in materials science and metallurgical engi-
neering from the Engineering School of Case Western Reserve University, an
MS in liquid crystal chemistry from the Liquid Crystal Institute of Kent State
620 AUTHORS' BIOGRAPHIES

University, and a second Master's in inorganic chemistry from Columbia Uni-

versity.
Dr. Hwang's major professional interests are in global technological trends,
market development of the eletronics industry, and electronic materials in gen-
eral. In specific, her interests include interconnection and packaging mate-
rials/process, surface mount technology, paste technology, and solder paste.
Dr. Hwang is the top authority in the electronic paste field. She has been an
invited lecturer and speaker at numerous major international and national con-
ferences on the subject. Dr. Hwang is a member of the International Society of
Hybrid Microelectronics, International Electronics Packaging Society, American
Ceramic Society, American Chemical Society, American Welding Society,
American Society of Metals, American Society of Testing and Materials, Surface
Mount Technology Association, and the Adhesion Society. She has served as
the president of ASM Electronic materials-Cleveland and is serving on the mem-
bership committee of the American Society of Metals and the IPC committee.
In addition to many publications, she is the author of the book Solder Paste
in Electronics Packaging Technology & Applications for Surface Mount, Hybrid
Circuits, and Component Manufacturing. She holds several U. S. overseas patents
in the area of paste technology.

Thomas J. Kilinski-Research Assistant-University of Wisconsin-Madi-

son. Mr. Kilinski has worked with the Electronics Packaging and Interconnection
Program since 1987. He is involved with the development of mechanical test
procedures for determining the behaviors of SMT interconnections. He presented
his work at the IEEE Electronic Components and Technology Conferences in
1989 and 1990. He received the BS and MS degrees in engineering mechanics
from the University of Wisconsin-Madison in 1988 and 1990, respectively.

Sheera Knecht-Research Scientist-Digital Equipment Corporation. Ms.

Knecht is currently working on the mechanical analysis of high-density electronic
packaging, and on the development of analytical methods to assess the reliability
of solder joint interconnect. She joined Digital Equipment Corporation in 1986.
Previously, she worked at MIT, as a research engineer on problems of hydraulic
fracture, and at the Bird-Johnson Company. Ms. Knecht received her BSc degree
in engineering from Brown University, and her MS in solid mechanics from
Harvard University.

John H. Lau-Senior Engineer-Hewlett-Packard Company, Palo Alto, Cal-

ifornia. His research activities cover a broad range of electronic packaging and
manufacturing technology including MCM, SMT, TAB, FPT, CAE, and DFM.
Prior to joining Hewlett-Packard Laboratories in 1984, Dr. Lau worked for
Exxon Production and Research Company and Sandia National Laboratory. He
received the BE degree in civil engineering from National Taiwan University,
the MASc degree in structural engineering from the University of British Co-
 AUTHORS' BIOGRAPHIES 621

lumbia, the MS degree in engineering mechanics from the University of Wis-

consin, the MS degree in management science from Fairleigh Dickinson Uni-
versity, and the PhD degree in theoretical and applied mechanics from the
University of Illinois. He has more than 20 years of research and development
experience in heat transfer, thermal stress, vibration, and reliability analyses of
mechanical and electronic components and systems. He has published extensively
in these areas, and is an associate editor of the ASME Journal of Electronic
Packaging as well as an associate editor of the IEEE Transactions on Compo-
nents, Hybrids, and Manufacturing Technology. He has also served as a member
of technical committees, session chairman, and workshop speaker of several
IEEE, ASME, ASM, and ISHM conferences and symposiums. He is the chair-
man of the IEEE/CHMT TC-9 (Thermal Management) committee, and the tech-
nical program chairman of the 1990 IEEE/CHMT International Electronics Man-
ufacturing Technology Symposium. Dr. Lau is a registered professional engineer
in New York and California. He is a senior member of IEEE and is listed in the
American Men and Women of Science.

Jon R. Lesniak-Research Assistant-University of Wisconsin-Madison. Mr.

Lesniak is involved with the Electronics Packaging and Interconnection Program,
where he develops high-precision measurement systems. He presented his work
at the IEEE Electronic Components and Technology Conference in 1989. He
worked at John Deere Co. developing the SPATE 9000 system, a system that
measures stress patterns by thermal emissions. He received the BS in engineering
mechanics from the University of Wisconsin-Madison in 1987. He is continuing
his work towards the PhD.

Che-Yu Li-Faculty Member-Department of Materials Science and Engi-

neering, Cornell University. The emphasis of Dr. Li research has been on me-
chanical properties of materials and their application in the energy and electronics
industries. For the past decade, he has been actively engaged in electronics
packaging research, specializing in solder interconnects, metallization, and TAB.
These research activities include mechanical reliability studies and microme-
chanical testing of electronic components down to submicron dimensions. Dr.
Li received his PhD degree from Cornell University in 1960. He is the coeditor
of the book Principles of Electronic Packaging.

Colin Alan MacKay-Staff Metallurgist-MCC Packaging. As a permanent

member of technical staff, Dr. MacKay is involved with the ongoing bonding
program, and leads the amalgams development project. In addition he supplies
advice and direction on a wide range of bonding and metallurgical subjects to
both projects and shareholders. Dr. MacKay has held posts with the British
Insulated Callenders Cables as research officer, The International Tin Research
Institute (for 16 years), latterly as Director of Research and Head of Metallurgy
Department, and on moving to the U.S.A., he was Director of Research at Alpha
622 AUTHORS' BIOGRAPHIES

Metals Newark, N.J., and Director of Research at the Semi Alloys Subsidiary
of Allied-Signal Corporation. He joined MCC from Alli'ed-Signal in 1985. Dr.
MacKay earned his BSc (Honours Physics), MSc, and PhD at London University.
Dr. MacKay has served on committees for the British Standards Institute and
the European Commission Electrotechnique Internationale (CEI) and is currently
on several committees of IPC. He is a Chartered Engineer, a member of the
British Institute of Physics, the Institution of Metallurgists, the Institute of Metal
Finishing, the British Association of Brazing and Soldering, the International
Society of Hybrid Microelectronics (ISHM), and IEEE. He is currently Chairman
of the Solders and Brazing Committee of the ASM.
Dr. MacKay has published more than 70 technical papers. He also holds
patents in the fields of solder alloys, special solder creams, preforms, packaging,
and interconnect and amalgams.

James L. Marshall-Director of the Center for Materials Characterization

and Director of the Industry-University Cooperative Research Center-Univer-
sity of North Texas, Denton, Texas. Dr. Marshall has been a member of the
faculty of the Department of Chemistry at the University of North Texas (UNT),
developing a research program in analytical chemistry. In 1981, he moved on
to industry, helping to develop the Advanced Manufacturing Technology program
at Motorola, Inc., in Ft. Worth, Texas. In 1987, Dr. Marshall returned to UNT
to establish a materials program in the Center for Materials Characterization.
Dr. Marshall earned his BS in chemistry at Indiana University, and his PhD in
Chemistry at Ohio State University. He cofounded the North Texas Section of
the Materials Research Society, and has served as an officer, including President,
several times. Dr. Marshall is the author of a reference book on instrumental
organic chemistry.

J. W. MorriS-Program Leader for Structural Materials--Center for Ad-

vanced Materials at the Lawrence Berkeley Laboratory. The research of this
group concerns the design of new materials and includes topics that range from
fundamental theory to the application of advanced engineering materials.
Professor Morris holds an ScD in materials science from MIT. He is a Fellow
in ASM. He was awarded the Hardy Gold Metal of the AIME, the Bradley
Stoughton Teaching Award of the ASM, and has authored numerous technical
papers.

Bela I. Sandor-Professor-Department of Engineering Mechanics, Univer-

sity of Wisconsin-Madison. Professor Sandor is also Director of the Electronics
Packaging and Interconnection Program at this institution. He teaches courses
in the fatigue of materials and advanced laboratory techniques in the area of
mechanics of materials. He joined the Wisconsin faculty in 1968. His research
interests include creep-fatigue interactions, damage evolution, life prediction,
and rational test acceleration methods of monolithic and composite materials.
 AUTHORS' BIOGRAPHIES 623

He worked at Bell Telephone Laboratories in the early 1960s, and was a Research
Fellow of the Japan Society for the Promotion of Science in 1985. His PhD
degree is in theoretical and applied mechanics from the University of Illinois-
Urbana. He has published several books and about 70 papers on fatigue and
related sUbjects.

Harvey D. Solomon-Staff Metallurgist-GE Research and Development

Center, Schenectady NY, and adjunct faculty member at Union College, Sche-
nectady NY. Dr. Solomon joined the General Electric Corporate Research and
Development Center in 1968, with an initial assignment at G.E. 's Nuclear Energy
Division in San Jose. He returned to the R&D center at Schenectady in 1969 to
continue his research into materials problems encountered by the Nuclear Energy
Division. This work centered on the Low Cycle Fatigue of carbon steels and Zr
alloys, general physical metallurgy, and stress corrosion cracking of stainless
steels. Specific studies were performed on high-strain low-cycle fatigue crack
growth, phase transformations in duplex stainless steels, weld sensitization of
stainless steels, stress-corrosion cracking in high-temperature water, and the
development of corrosion and stress corrosion sensors. His interests have ex-
tended beyond the concerns of the Nuclear Energy Division into superalloys for
aircraft engines and gas turbines, the influence of creep and environment on
high-temperature fatigue, liquid metal surface tension measurements, welding
metallurgy, and, in recent years, to the problem of solder fatigue and electrical
component problems. This most recent work has included the testing and study
of leadless, as well as leaded, chip carrier/printed wiring board joints, as well
as the study of special laboratory joints. He currently teaches a course on Welding
Metallurgy at Union College. Dr. Solomon received a BS in metallurgical en-
gineering from New York University, and a PhD in metallurgy and materials
science from The University of Pennsylvania. Dr. Solomon is the author or co-
author of over 55 articles in technical journals and over 50 additional General
Electric Company documents. He was also the coeditor of three conference
proceedings and the chairman of the 1985 ASTM symposium on LCF. He has
recently completed a chapter on weld embrittlement for Vol. 25 of the Academic
Press series Treatise on Materials Science and Technology, and an invited review
article on solder fatigue for the Materials Research Society. Dr. Solomon has
presented conference keynote talks on Stainless Steel Metallurgy, the Physical
Metallurgy of Duplex Stainless Steels, and the fatigue of solders and solder
joints, as well as over 100 other talks. His work on Duplex Stainless Steels was
recognized with the 1979 ASTM Vilella Award (with T.M. Devine) and his
work on solder fatigue with the 1986 IEEE-CHMT Transactions Best Paper
Award.

Janet R. Sterritt-Manager, Research and Development-Hollis Automa-

tion. Hollis Automation is a capital equipment manufacturer for the printed circuit
assembly industry. Her expertise is in wave soldering, reflow, and cleaning
624 AUTHORS' BIOGRAPHIES

processes. Ms. Sterritt joined Hollis Automation in 1987, after designing and
fabricating boards for 10 years at Xerox Corporation. She holds a BSEE from
the University of Rochester. Ms. Sterritt holds the patent for the wave soldering
Accuknife. She developed the XL7 and FuturelSMT wave soldering processes.
She directed the R&D team in the development of various nozzle designs for
aqueous cleaning, leading to the development of the Hurricane Jet and the
Directed Impact Force nozzles. She has written numerous technical articles which
have been published in various trade journals.

Donald S. Stone-Assistant Professor-Materials Science and Engineering,

the University of Wisconsin at Madison, WI. At the University of Wisconsin
for 3 years, Dr. Stone currently researches gain boundary sliding in solders and
solder joints, and the mechanical properties of thin films with applications in
tribology and electronic packaging. His expertise is the mechanical properties
of solids with emphasis on theory and measurement. He received a BS in physics
from the University of Texas at Austin, and a PhD in materials science from
Cornell University. Dr. Stone is a member of TMS-AIME and MRS.

Ravi Subrahmanyan-Research Scientist-Motorola, Inc., Dr. Subrahman-

yan is currently involved in materials-related research and development in areas
such as electronics packaging, device technology, structural design, and relia-
bility. Since 1985, he has been actively involved in solder fatigue research at
Cornell University. To date, he has several publications to his credit that describe
fundamental aspects of this phenomenon. He has also been responsible for the
development of techniques for the fabrication and mechanical testing of solder
joints as small as 50 11m. Dr. Subrahmanyan received the BTech degree in
metallurgical engineering at the Indian Institute of Technology, Madras, India.
He earned his MS and PhD degrees in materials science at Cornell University.

Tammy Summers-University of California, Berkeley. Ms. Summers is

currently working toward her PhD Thesis in materials science. Her thesis work
involves a microstructural study of fatigue in Sn-Pb solders.

John K. Tien-Director, Strategic Materials R&D Laboratory-University

of Texas. Dr. Tien joined Pratt & Whitney Aircraft in 1969 to conduct research
onjet engine materials. He joined the Faculty of Metallurgy and Materials Science
at Columbia University in 1972. In 1983, he was honored by Columbia University
by becoming the Henry Marion Howe Chair Professor of Metallurgy and Ma-
terials Science. Dr. Tien received his BS degree in mechanical engineering from
Worchester Polytechnic Institute, in Worchester, Massachusetts, MA and MEng
degrees from Yale University, and MS and PhD degrees from Carnegie-Mellon
University in Pittsburgh, Pennsylvania. He has been elected an Outstanding
Young Man of America by the U.S. Chamber of Commerce for "outstanding
 AUTHORS' BIOGRAPHIES 625

ability, contributions and service to America." He was awarded the Bradley

Stoughton Best Young Professor A ward for excellence in teaching and research
contributions, by the American Society of Metals. In 1984, Dr. Tien was honored
with the K. C. Li Gold Medal for meritorious achievement. He is a member of
committees of the National Advisory Board and the Argonne National Labora-
tory's MST Review Committee. He is a member of Sigma Chi and Tau Beta
Pi. He is active on the national level in TMS/AIME and the ASM, and has
chaired various committees and many international conferences. Dr. Tien, to-
gether with his students and colleagues, has written over 200 technical papers
on a broad range of metallurgical and other topics.

Dana Tribula-Lam Research Institute, Fremont, California. Dr. Tribula

studied creep and thermal fatigue of near-eutectic Sn-Pb solders as her PhD
Thesis work in materials science at the University of California, Berkeley.

James R. Wilcox-Research Scientist-mM, Endicott, New York. Dr. Wil-

cox has most recently been studying the deformation and fatigue properties of
solder alloys. He joined IBM in 1982, where he worked on various process-
development projects for advanced circuit packaging. Previously, he worked for
several years at the GE Corporate Research and Development Center in Sche-
nectady, New York, where his research efforts included amorphous metal pro-
cessing and fatigue performance evaluation of Ni-based superalloys. Dr. Wilcox
received his BS and MS degrees in metallurgical engineering from Michigan
Technological University, Houghton, Michigan. In 1986, he was awarded an
IBM resident study fellowship, and received his PhD degree in materials science
from Cornell University.
 Index

Abietic acid, 20, 23-29 Constitutive relations, 319-322

Accelerated testing, 551-553, 571-583 development of, 511-517
conditions, 578-581 general form, 510-511
considerations, 571-578 inelastic deformation behavior and, 365-370
failure definition and detection, 583 Controlled atmosphere soldering, 73-80, 162-
sample statistics and solder joint defects, 163
581-582 Convection, 105
test vehicle assembly, 582 Conveyors, 135-136
test vehicle design, 581 Copper mirror test, 145-146
Acceleration transform, 567-569 Copper oxide (CUO), 23, 25-29
Alloys, 44-46 Core solder material, 168-169
microstructure and mechanical properties of, Creep
266--277 cyclic, 288-293
phases in, 175 deformation, 301
precipitation and dissolution in Pb-Sn, 268- extension of matrix failure indicator, 534-
276 536
ternary, 256--259 integrated matrix, 508-543
Aqueous cleaning, 84-85 in lead-tin solders, 279-304
Area of spread test, 3 mechanical properties, 386--388
steady-state equation, 517-520
"Bathtub" reliability curve, 546--549 steady-state strain component, 511-514
¥ismuth solders, 183-184 and stress relaxation, 313-318
Boundary-value problem, 458-460 erE. See Coefficients of thermal expansion
British corrosion test, 151-152 Cyclic creep, 288-293
Cyclic load histories, 583-585
Capillary rise test, 3 Cycling frequency, 425-436
Chip carriers, surface mounted, 456--477
Cleaning Damage
definition of soldering flux, 144 integration and failure criterion effects, 377-
enviromnental concerns, 143-144 380
and flux application methods, 163-166 mechanisms, 372-376
and flux materials, 155-163 rate formulation, 372-377
and non-liquid fluxes, 168-170 Deformation
post-solder considerations, 143-170 inelastic and stress calculation, 363-372
process issues and reliability, 166--168 mechanisms, 246--248
specifications, 144-155 Diffuse reflectance Fourier transform infrared
trends, 170 analysis (DRIFT), 30-31
Coefficients of thermal expansion (erE), 550- Displacement, 438-442
551,553-555 DRIFI'. See Diffuse reflectance Fourier
Coffin-Manson equation, 388-390, 392-393 transform infrared analysis
Conductive belt, 105
Conformal coating, 168 Edge dip test, 3
Constitutive equation, 517, 520-526 Elastic properties, 385-386

627
628 INDEX

Elastic strain component, 514-515 Flux

Electronic assemblies, 545-549 action from solderability measurements,
Environment, 143-144, 436-437 4-5
Eutectic microstructures, 229-246 application methods, 163-166
coarsening behavior, 242-243 definition, 144
degenerative, 230-233 effect of impurities on solderability, 6-8
effects of composition, 238-241 electromechanical mechanism, 9-11
intennetallics, 236-243 foam, 164
lamellar, 230 history, 1-2
recrystallized Pb-Sn, 241-242 low solid, 160--162, 166
solder joint, 233-235 materials and associated cleaning, 155-163
mechanistic studies for inorganic, 5
Failure mechanistic studies for rosin-based, 11-36
criterion effects, 379-380 non-liquid, 168-170
definition, 333-334, 413-423 residue, 167
distribution, 594-596 rosin-based. See Rosin-based fluxes
statistical considerations, 569-571 solderability tests, 2-4
Fatigue and solder paste, 42-44
accelerated reliability testing, 571-583 types, 5-36
calculations and mechanical shear tests, and wave soldering, 119-125
526-532 Foam fluxer, 164
damage integral methodology for thennal Fourier transform analysis. See Diffuse
and mechanical, 361-380 reflectance Fourier transform infrared
effect of frequency, 338-342 analysis
effect of hold time, 342-354 FR-4 printed wiring board, 601-607
effect of strain range, 334-338 Frequency
effect of temperature on isothennal, 355- cycling, 425-436
357 and fatigue life, 338-342
hybrid life prediction techniques, 395-398 Frequency-modified Coffin-Manson equation,
in lead-tin solders, 279-304 390-391
life prediction techniques, 388-398
low cycle. See Low cycle fatigue Globule test, 3-4
mechanical, 361-380 Gold-based solder, 177
predicting from isothennal low cycle data, Gold embrittlement, 205-212
406-449 Gull wing, 456-472
testing, 385-388
models, 388-395 Halide test, 146-147
as reliability issue, 213-222 Hazard rate model, 546-549
solder joint, 556-559, 589-596 Hold time, 293-301, 342-354, 425-436
solder shear, 556-571
and surface mount technology, 384-404 Indium solders, 180-183
thennal. See Thennal fatigue Inelastic defonnation, 363-372
Figures of merit, 596-601 Infrared soldering, 98-101, 112-113
derivation of fonnulas, 598-599 Inorganic fluxes, 5
examples, 601-607 In-plane gradients, 311
fonnulas and scaling constants, 607-611 Institute for Interconnecting and Packaging
graphical interpretation and reliability charts, Electronic Circuits (!PC), 153
600-601 Intennetallics, 175, 236-243
for multiple thennal fluctuations, 599 Ionic conductivity test, 152
Fillet fonnation, 167-168 Isopropyl alcohol (lPA), 22, 23-29
Finite element mesh, 536-540 Isothennal fatigue, 355-357
Finite element methods, 460-464 Isothennal low cycle data, 406-449
 INDEX 629

J leads, 472-477 microstructure in thermal fatigue of, 248-

comparison of stiffness matrices, 477 261
unit displacement, 472-477 PCB. See Printed circuit board
Plastic leaded chip carrier (PLCC), 455-457,
Lamellar eutectics, 230-231 459,464, 477, 478, 483
Lap joint specimens, 400, 401 Plastic quad flat pack (PQFP), 455-458, 464,
Laser soldering, 105-106 477,478, 483-484
LCF. See Low cycle fatigue PLCC. See Plastic leaded chip carrier
Leaching of substrate, 194-204 Powder. See Solder powder
Leaded solder attachments, 532-542, 564- Power dissipation, in component, 309
567,591-594 PQFP. See Plastic quad flat pack
Leadless solder attachments, 559-564, 589- Prereflow solder paste bake, 106
591 Printed circuit board (PCB), 455-457, 486-
Lead-tin solders. See Tin-lead solders 493
Lifetime, definition, 535-536
Low cycle fatigue (LCF), 411-413 Reflow soldering, 92-116
influence of definition for failure, 413-423 equipment, 98-106
influence of environment, 436-437 maximizing solder joint yield, 106-109
influence of hold time and cycling processing, 109-113
frequency, 425-436 process phases, 95-98
influence of temperature, 423-425 surface mount, 92-95, 113-116
microstructural changes, 427-438 Reliability, 194-222
prediction of life of solder joints, 443-449 and accelerated testing, 551-553, 571-
Low solid fluxes, 160-162, 166 583
"bathtub" curve, 546-549
Mechanical fatigue. See Fatigue, mechanical and cleaning process issues, 166-168
Mechanical testing, 385-388 of electronic assemblies and solder joints,
Microstructure 545-549
changes and low cycle fatigue, 427-438 fatigue, 213-222
eutectic, 229-246 figures of merit for attachment, 596-601
and mechanical properties of solder alloys, gold embrittlement, 205-212
266-277 leaching of substrate, 194-204
of Pb-Sn solder joints and thermal fatigue, solder joint loading conditions, 549-
248-261 551
Military (U .S.), 153 solder joint under shock and vibration
Model joints, 398-403 conditions, 478-493
cooling rates, 399 surface mount, 583-585, 588-596
inspection, 399, 401 Rosin-based fluxes, 11-36
lap joint specimens, 400, 401 action inferred from solderability
quality control, 398-401 measurements, 31-34
reflow process, 399 autocatalysis and chain reaction agents, 34-
straddle board specimens, 401-403 36
Mode of loading, definition, 333-334 and cleaning, 157-158
Multiaxial stress states, 525-526, 534 concentration dependent mechanism, 17
Multiple thermal fluctuations, 599 effect on impurities in solder, 17-19
formulations, 13-14
Paste. See Solder paste implications of alternative mechanisms, 36
Pb-Sn solder joints starting materials, 20-23
eutectic microstructures, 229-246 thermodynamics, 20
experimental details, 226-229 wetting-balance performance, 14-17
intermetallics, 236-243 Rotary dip test, 4
mechanical properties, 243-248 Rotating drum spray, 166
630 INDEX

Scanning electron microscopy/energy Solder properties, 318-319

dispersive x-ray analysis (SEMlEDX), Solder shear fatigue, 556-559
173 Solvent cleaning, 80-84
detection modes, 174 Specimen design, definition, 333-334
fatigue, 213-222 Spray fluxer, 164-166
general characterization, 176-186 Square wave cycle, 521-522
gold embrittlement, 205-212 Steady-state creep equation, 517-520
intermetallics, 175 Steady-state strain component, 511-514
leaching of substrate, 194-204 Stiffness, 456-477
phases in alloys, 175 Straddle board specimens, 401-403
reliability issues, 194-222 Strain
sample preparation, 174-175 distribution, 438-442
solderability issues, 186-193 range, 334-338
Shear fatigue, 556-559 time independent plastic component, 515-
Shock conditions, 478-493 516
Silver chromate paper (halide) test, 146-147 total strain equation, 390-391
Silver solder, 177 variation with position, 534-535
Small outline integrated circuit (SOIC), 455, Strain-life relationship, 390-391
456, 480-483 Strain range partitioning (SRP) procedure,
SMC. See Surface mount component 393-395, 397-398
SMT. See Surface mount technology Stress calculation, 363-372
SOIC. See Small outline integrated circuit Stress relaxation, 386-388
(SOIC) Substrate
Solderability, 70-72 gradients in, 309-311
"forcing" poor, 189, 192-193 ideal expansivity, 307-313
inadequate tin protective coatings, 189 leaching of, 194-204
maintaining, 186 solder expansivity mismatch, 312-313
Solder ball formation, 167 Surface insulation resistance test, 147-151
Soldering principle, 69-70 Surface mount component (SMC), 455-456,
Solder paste, 38-90 486-493
alloys, 44-46 Surface mount reliability, 583-585, 588-596
aqueous cleaning and aqueous cleaning Surface mount technology (SMT)
paste, 84-85 analyses of component leads and solder
chemical and physical characteristics, 40-42 joints, 455-493
controlled atmosphere soldering, 73-80 expert systems, 404
dispensing technique, 68-69 fatigue life prediction, 384-404
fine pitch, 87, 90 hybrid prediction techniques, 395-398
fluxing and fluxes, 42-44 mechanical testing, 385-388
formulation, 51-54 model joints, 398-403
material, 169-170 prediction techniques, 388-395
no-clean, 85-86 reflow soldering, 92-95, 113-116
powder. See Solder powder stiffness of gull wing, J leads, and solder
prereflow bake, 106 joints, 456-477
printing technique, 64-68 Surface tension balance test, 4
quality, 90 Synthetic activated fluxes, 159-160
rheology, 54-62
solderability, 70-72 Telecommunications, 154-155
soldering methods, 72-73 Temperature
soldering principle, 69-70 cycling, 322-326
solvent cleaning, 80-84 larger cycles, 326-329
viscosity and measurement, 63-64 small ranges, 323-326
Solder powder, 23, 27-29, 46-50 effect on isothermal fatigue, 355-357
 INDEX 631

Temperature (continued) Top-side fillet formation, 167-168

gradients, 307-309 Total strain equation, 390-391
and low cycle fatigue, 423-425 Triangle wave cycle, 520-521
shock, 311-312 Triethanolamine-hydrochloride (TEA-Hel),
Ternary alloys, 256-259 22-23, 24-25
Thermal expansion mismatch problem, 553-
556 Vapor phase soldering, 101-105, 110-112
Thermal fatigue, 361-380 Vibration conditions, 478-493
cycling, 267-268
life estimation, 377-379 Water soluble fluxes, 158-159
microstructural evolution under, 248-261 Wave fluxer, 163
microstructural mechanisms of, 252-255 Wavelength dispersive x-ray (WDX), 173-174
predicting from isothermal low cycle data, Wave soldering, 117-142
406-449 hardware, 119-136
rapidly solidified joints, 261 conveyors, 135-136
in shear, 249-251 fluxing and fluxers, 119-125
ternary alloys, 256-259 preheating and preheaters, 125-131
Thermomechanical test guidance, 302-303 solder waves, 131-135
Time independent plastic strain component, solidification, 135
515-516 overview, 117-118
Tin-lead solders, 176-177, 186,279-304 parameter optimization, 136-141
Tin protective coatings, 189 results, 141