ABSTRACT 
 
 
 
 
Title of Document: RELIABILITY OF LEAD-FREE HIGH 
TEMPERATURE SURFACE MOUNT 
COMPONENT ATTACHES 
  
 Timothy James Oberc, Master of Science, 2008 
  
Directed By: Associate Professor F. Patrick McCluskey 
Department of Mechanical Engineering 
 
 
This work investigates the relative reliabilities of SAC305, eutectic Au-Sn, and Ag-In 
transient liquid phase sintered (TLPS) solder joints subjected to high temperature 
passive thermal cycling.  These solder materials were monitored for electrical 
resistance, mechanical pull strength, and microstructural changes during cycling.  In 
fabricating the test assemblies, SAC305 and eutectic Au-Sn manufacturing 
parameters were gathered from the paste distributors while fabrication with Ag-In 
TLPS required in-house development.  Work with the Ag-In TLPS paste revealed that 
reducing the additive (In) particle size led to statistically significant improvements in 
the solid volume fraction of the joints.  Successful attachment of ceramic quadflat 
packs (CQFPs) to polyimide circuit boards using Ag-In TLPS demonstrated that 
surface mount joints with mechanical and electrical integrity could be manufactured 
from the material.  Au-Sn was found to be the strongest of the materials while 
cracking in the SAC305 and Ag-In samples during cycling resulted in weaker joints. 
  
 
 
 
 
 
 
 
 
 
 
 
 
RELIABILITY OF LEAD-FREE HIGH TEMPERATURE SURFACE MOUNT 
COMPONENT ATTACHES    
 
 
 
By 
 
 
Timothy James Oberc 
 
 
 
 
 
Thesis submitted to the Faculty of the Graduate School of the  
University of Maryland, College Park, in partial fulfillment 
of the requirements for the degree of 
Master of Science 
2008 
 
 
 
 
 
 
 
 
 
 
Advisory Committee: 
Dr. F. Patrick McCluskey, Chair and Advisor 
Dr. Teng Li 
Dr. Peter Sandborn 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
? Copyright by 
Timothy James Oberc 
2008 
 
 ii 
 
Dedication 
To Rachel, Scruff McGruff, my parents and sister. 
 
 iii 
 
Acknowledgements 
I would first like to thank my advisor, Dr. McCluskey, for all of his time and effort 
spent on this research.  I would also like to thank my committee members, Dr. Teng 
Li and Dr. Peter Sandborn, for their input on how to improve the work herein. 
 
I would also like to thank everyone in the McCluskey research group for making my 
time at the University of Maryland so enjoyable.  Special thanks to Pedro, Rui, 
Lauren, Adam, and Rupal for being such good friends and teachers along the way. 
 
I would also like to thank CALCE for the use of its facilities and equipment.  Thanks 
in particular to Bhanu, Anshul, and Ahmed for their assistance in the labs. 
 
Last, but certainly not least, I would like to thank Rachel and my family for their love 
and support during the creation of this work and for all the years leading up to it.  I 
don?t know where I would be today without all that you have given me. 
 
 
 
 iv 
 
Table of Contents 
 
 
Dedication .................................................................................................................ii 
Acknowledgements ..................................................................................................iii 
Table of Contents .....................................................................................................iv 
List of Tables............................................................................................................vi 
List of Figures .........................................................................................................vii 
Chapter 1: Introduction..............................................................................................1 
1.1 Electronics for High Temperature Applications..........................................2 
1.2 Surface Mount Attach for High Temperature Environments .......................4 
Chapter 2: Prior Work .............................................................................................10 
2.1 SAC305 Studies .......................................................................................10 
2.2 Eutectic Au-Sn Studies.............................................................................13 
2.3 TLPS for Surface Mount Attach...............................................................15 
Chapter 3: Initial Embodiment of Ag-In TLPS Surface Mount Component Attach...18 
3.1 Selection of Ag-In System........................................................................18 
3.2 Initial Ag-In TLPS Surface Mount Solder Joints ......................................21 
3.3 Ag-In TLPS Densification Study..............................................................23 
3.3.1 Areal Analysis using ImageJ Software..............................................24 
3.3.2 Validation of Areal Analysis using the Point-Counting Method........26 
3.3.3 Volume Fraction Trends ...................................................................29 
Chapter 4: Experimental Procedure..........................................................................37 
4.1 Test Vehicle Materials..............................................................................37 
4.1.1 Edge-card Circuit Boards..................................................................38 
4.1.2 Ceramic Quadflat Pack Components.................................................42 
4.1.3 Combined Test Vehicle ....................................................................45 
4.1.4 Solder Pastes ....................................................................................46 
4.1.5 Stencil Foils......................................................................................48 
4.2 Test Vehicle Fabrication...........................................................................50 
4.2.1 Stencil Printing.................................................................................50 
4.2.2 CQFP Placement ..............................................................................52 
4.2.3 Reflow / Temperature Processing .....................................................53 
4.3 Reliability Testing Procedures..................................................................64 
4.3.1 Passive Thermal Cycling ..................................................................66 
4.3.2 In-situ Electrical Testing...................................................................68 
4.3.3 Mechanical Pull Testing ...................................................................70 
4.3.4 Cross Sectional Sampling .................................................................72 
Chapter 5: Reliability Assessment Results and Discussion.......................................74 
5.1 Solder Joint Strain Energy Density Estimation .........................................75 
5.1.1 Diagonal Lead Stiffness....................................................................76 
5.1.2 Diagonal Length, CTE Mismatch, and ?T........................................85 
5.1.3 Minimum Load Bearing Area and Characteristic Solder Joint Height86 
5.1.4 Strain Energy Densities ....................................................................88 
5.2 Electrical Data Trends and Discussion......................................................89 
 
 v 
 
5.3 Mechanical Data Trends and Discussion ................................................104 
5.3.1 SAC305 Mechanical Results ..........................................................106 
5.3.2 Eutectic Au-Sn Mechanical Results ................................................111 
5.3.3 Ag-In TLPS Mechanical Results.....................................................115 
5.4 Cross Sectional Data Trends and Discussions.........................................119 
5.4.1 SAC305 Microstructural Results ....................................................120 
5.4.2 Eutectic Au-Sn Microstructural Results ..........................................127 
5.4.3 Ag-In TLPS Microstructural Results...............................................132 
Chapter 6: Conclusions, Contributions, and Future Work.......................................139 
6.1 Conclusions............................................................................................139 
6.2 Contributions .........................................................................................141 
6.3 Future Work...........................................................................................141 
Appendix A ? MATLAB Lead Compliance Formulas ...........................................142 
Appendix B ? MATLAB codes for Electrical Data Analysis..................................145 
Works Cited ..........................................................................................................147 
 
 
 
 
 
 
 
 
 
 
 vi 
 
List of Tables 
 
Table 1.  Common electronic product temperature ranges. .........................................2 
Table 2.  Solder alloys for HT applications. ...............................................................7 
Table 3.  Point and volume fractions of an Ag-In surface mount solder joint............28 
Table 4.  Paste formulations for densification study. ................................................33 
Table 5.  Omega furnace settings and procedures for proper reflow of SAC305 and 
eutectic Au-Sn. ................................................................................................60 
Table 6.  Ag-In TLPS processing open/close furnace door sequence. .......................63 
Table 7.  Sun Electronic Systems EC12 Controller LT and HT Programs. ...............67 
Table 8.  Average lead angles from 44-lead and 64-lead CQFPs. .............................80 
Table 9.  Estimated parameters of the inclined gull-wing structural model. ..............82 
Table 10.  Inclined gull-wing lead spring constants..................................................84 
Table 11.  Characteristic solder joint heights and minimum load bearing areas. .......88 
Table 12.  Strain energy densities. ...........................................................................89 
Table 13.  Strain range estimations. .........................................................................89 
Table 14.  Adjusted Electrical Sample Populations. ...............................................103 
Table 15.  Observed mechanical testing periods and samples. ................................ 105 
Table 16.  SAC305 ANOVA results. .....................................................................111 
Table 17.  Eutectic Au-Sn ANOVA results. ...........................................................114 
Table 18.  Ag-In TLPS ANOVA results................................................................. 117 
Table 19.  Observed cross section sampling schedule and sizes..............................120 
Table 20.  EDS point scan results of SAC intermetallic..........................................127 
Table 21.  EDS point scan results of Au-Sn bulk solder. ........................................128 
 
 
 vii 
 
List of Figures 
Figure 1.  Inclined gull-wing leaded surface mount package. .....................................5 
Figure 2.  Temperature vs. time of a sample reflow profile for Sn-Ag-Cu solder........6 
Figure 3.  Ag-In Binary Phase Diagram from Hansen and Anderko [30]..................19 
Figure 4.  ImageJ transformation from threshold adjustment....................................25 
Figure 5. Bounded joint for area fraction analysis. ...................................................25 
Figure 6.  Point-counting method grid overlay. ........................................................28 
Figure 7.  Volume fraction as a function of particle ratio (R = DAg/DIn). ..................34 
Figure 8.  Ag-In TLPS unprocessed paste ESEM/EDS images (Ag in red)...............35 
Figure 9.  Wetting comparison of P1 and P4 pastes under large package leads. ........36 
Figure 10.  Small CQFP footprint. ...........................................................................39 
Figure 11.  Large CQFP footprint. ...........................................................................39 
Figure 12.  Polyimide test board edge-card feature...................................................40 
Figure 13.  Polyimide test board design. ..................................................................41 
Figure 14.  64-lead CQFP daisy chain wiring scheme. .............................................43 
Figure 15.  44-lead CQFP daisy-chain wiring scheme. .............................................44 
Figure 16.  Desired CQFP Gull-wing geometry (not to scale). .................................45 
Figure 17.  Test vehicle metallization stack-up.........................................................46 
Figure 18.  Stencil foil design in EAGLE 4.16. ........................................................49 
Figure 19.  Stencil printing fixture. ..........................................................................51 
Figure 20.  Stencil printing orientation and direction................................................51 
Figure 21.  CQFP component placement fixture.......................................................53 
Figure 22.  Furnace temperature testing thermocouple arrangement. ........................55 
Figure 23.  Manufacturer?s recommended SAC305 reflow profile. ..........................57 
Figure 24.  SAC305 reflow profile as measured within Omega furnace....................58 
Figure 25.  Manufacturer?s recommended eutectic Au-Sn reflow profile..................59 
Figure 26.  Eutectic Au-Sn profile as measured within Omega furnace. ...................60 
Figure 27.  Ag-In TLPS profile as measured within Omega furnace.........................64 
Figure 28.  LT and HT profiles. ...............................................................................67 
Figure 29.  Electrical testing wiring. ........................................................................70 
Figure 30.  45? pull test sample orientation [46].......................................................71 
Figure 31.  Pre-cut locations on polyimide test board...............................................72 
Figure 32.  Structural model for inclined gull-wing [49]. .........................................78 
Figure 33.  Lead angle definitions............................................................................79 
Figure 34.  Geometric overlay of cross section images to estimate gull-wing structure.
........................................................................................................................81 
Figure 35.  Measurement definitions in estimating characteristic solder joint height.87 
Figure 36.  Au-Sn chain resistance and temperature as a function of passive thermal 
cycling time. ....................................................................................................92 
Figure 37.  Representative baseline resistance signature of a 44-lead CQFP, SAC305 
daisy chain.......................................................................................................94 
Figure 38.  Chain resistance degradation as a function of passive thermal cycles. ....96 
Figure 39.  Potential failure sites within electrical wiring. ........................................98 
Figure 40.  Resistance trend of a chain completed by an aluminum insert. .............100 
 
 viii 
 
Figure 41.  Observed chain similarities across varying samples and connectors. ....101 
Figure 42.  Interfacial failure between component lead and solder joint during pull 
strength testing............................................................................................... 106 
Figure 43.  Pull strength trends of SAC, 64-lead CQFP samples. ...........................108 
Figure 44.  Pull strength trends of SAC, 44-lead CQFP samples. ...........................109 
Figure 45.  Pad lift failure mode during pull strength testing. .................................112 
Figure 46.  Pull strength trends of Au-Sn, 44-lead CQFP samples..........................113 
Figure 47.  Pull strength trends of Au-Sn, 64-lead CQFP samples..........................114 
Figure 48.  Pull strength trends of Ag-In TLPS, 44-lead CQFP samples.................116 
Figure 49.  Pull strength trends of Ag-In TLPS, 64-lead CQFP samples.................116 
Figure 50.  Comparison plot of solder material pull strengths................................. 119 
Figure 51.  SAC305 64-lead CQFP joint subjected to 0 thermal cycles. .................121 
Figure 52.  SAC305 64-lead CQFP joint subjected to 800 LT thermal cycles.........122 
Figure 53.  SAC305 64-lead CQFP joint subjected to 800 HT thermal cycles. .......122 
Figure 54.  Intermetallic layer growth and crack progression in SAC 64-lead CQFP 
HT joints........................................................................................................123 
Figure 55.  SAC lead side intermetallic thickness as a function of thermal cycles...124 
Figure 56.  SAC305 44-lead CQFP HT solder joint heel imaged using light 
microscope.....................................................................................................125 
Figure 57.  EDS Map of lead side intermetallic......................................................126 
Figure 58.  EDS Point Scan locations.....................................................................126 
Figure 59.  Truncated Au-Sn binary phase diagram [56]. .......................................128 
Figure 60.  Phase distribution within eutectic Au-Sn solder joint............................129 
Figure 61.  Solder joint heel of an Au-Sn 44-lead CQFP LT sample. .....................129 
Figure 62.   Lack of fatigue and microstructural coarsening in Au-Sn 64-lead CQFP 
HT solder joints. ............................................................................................131 
Figure 63.  Ag-In TLPS 64-lead CQFP solder joint subjected to 0 thermal cycles. .132 
Figure 64.  Ag-In TLPS lead side intermetallic thickness as a function of thermal 
cycles.............................................................................................................133 
Figure 65.  Au-In Binary Phase Diagram from [30]................................................134 
Figure 66.  Crack progression in Ag-In TLPS 64-lead CQFP HT solder joints. ......135 
Figure 67.  Cracking in Ag-In TLPS 44-lead CQFP HT solder joint subjected to 400 
HT cycles.......................................................................................................136 
Figure 68.  EDS map of Ag-In particle...................................................................137 
Figure 69.  Evidence of silver diffusion from Ag-In particles within Ag-In TLPS 
matrix. ...........................................................................................................137 
 
 
 
 1 
 
Chapter 1: Introduction 
 
The world of electronics has become an expansive realm that touches our lives in 
many ways.  Electronic systems are now available for a wide array of functions 
ranging from computational aides and monitoring devices to defense systems and 
information networks.  These systems have contributed to a significant increase in the 
standard of living around the world.  Research in the field of electronics, as in any 
other, has encompassed the continual pursuit of new knowledge that can be used to 
improve existing products as well as to develop novel technologies capable of 
previously unachievable feats.  The boundaries that limit the capability of electronic 
technologies are continually expanded by these research efforts, which have 
subsequently resulted in an ever-increasing variety of electronic products. 
 
Due to the overwhelming variety of electronic products that exist, categorization of 
these devices is required to simplify discussion within the field.  At its highest level, 
this categorization can serve to qualitatively group electronic products based on their 
end use within society.  Another more useful categorization of electronic products 
serves to organize them based upon some metric that describes their capabilities or 
conditions of use.  A popular method of quantitative categorization uses the 
temperature ranges over which the technologies are designed to operate.  In 
particular, all electronic technologies can be categorized as capable of operating 
within the commercial, industrial, military, or extreme temperature ranges.  These 
temperature ranges are provided in Table 1 [1][2]. 
 
 
 2 
 
Table 1.  Common electronic product temperature ranges. 
 Temperature Range (?C) 
Commercial 0 to 70 
Industrial -40 to 85 
Military -55 to 125 
Extreme Below -55 or Above 125 
 
As the table above shows, the extreme temperature range covers any electronic 
technology capable of operating from below -55?C, commonly referred to as the low 
temperature (LT) range, or above 125?C, commonly referred to as the high 
temperature (HT) range [2].  As the title of this work indicates, this thesis 
investigation will address a portion of the packaging required for electronic products 
capable of performing within the HT range. 
 
1.1 Electronics for High Temperature Applications 
 
The push for electronic technologies capable of operating in high temperature 
environments has resulted from the need for the electronics to perform distributed 
sensing, actuating, and control functions in harsh, uncontrolled environments.  
Common applications of electronics in distributed sensing and control systems 
subjected to HT conditions can be found in the automotive, avionics, and energy 
exploration industries.  Distributed sensing and control, which refers to the 
implementation of signal processing and control capabilities local to the sensors 
interacting with the environment, improves system performance over the 
conventional centralized architecture where signal processing capabilities are housed 
away from the sensing elements.  An example of distributed control is a drill system 
in a HT down-hole environment capable of sensing pressure and adjusting the 
rotational speed without needing to send a signal back to the surface for processing; 
 
 3 
 
all sensing and computation takes place down-hole, effectively eliminating the need 
for wiring between the drill-head and surface.  The application of HT electronics in a 
particular system is often preferred over the alternative of implementing standard 
temperature electronics in these applications because they can be operated at higher 
temperatures than their normal temperature counterparts, therefore eliminating the 
need for passive or active cooling technologies that can add cost and complexity to a 
given sensing and control system. 
 
In order for electronic systems to be capable of withstanding HT conditions, all 
electronic components of the system must be able to reliably operate in the given 
environment.  Proper operation of the electronic components is dependent upon the 
temperature related behavior of the electrical devices on the die in addition to the 
materials used to package it.  Some newer devices fabricated from silicon carbide 
(SiC) have been successfully operated at temperatures of 500?C [3][4] while other 
sources have claimed operation at temperatures up to 600?C.  These temperatures are 
well above the capability of silicon (Si) devices, which cannot be used above ~300?C 
due to limits imposed by the material?s reverse breakdown voltage [5].  With new SiC 
devices able to operate at increasingly higher temperatures, the development of 
materials to package and connect them at these temperatures remains a significant 
hurdle that must be overcome.  The foremost materials-related problem addressed by 
researchers within the packaging community has been that of the first-level 
interconnection between device and substrate.  Suggested die attach technologies to 
satisfy this packaging problem have ranged from exotic, high melting point alloys to 
various sintered metals and even high temperature organic attaches.  Although this 
 
 4 
 
first level of interconnection is an important packaging task to address, it is also 
critical to research materials suitable for other levels of interconnection as well.  This 
thesis seeks to contribute to the knowledge base regarding second level 
interconnection of surface mount devices to printed wiring boards (PWB) for HT 
conditions. 
 
1.2 Surface Mount Attach for High Temperature Environments 
 
Second level interconnection of electronic packages can be achieved in a number of 
ways.  Prior to the advent of the surface mount package, 2nd level interconnection was 
performed using through-hole connection of packages, whereby leads attached to the 
electronic parts were inserted into plated through holes in the circuit board and an 
interconnection was made by filling the space around the lead in the hole with molten 
solder which was then allowed to solidify.  The shift to surface mounted components 
has largely eliminated through-hole connections, allowing for the miniaturization of 
many electronic systems by increasing utilization of the PWB area.  The shift to 
surface mounting of electronic components has spurred the invention of a wide array 
of package geometries that include the commonly used small outline (SOP), quad 
flatpack (QFP), ball grid array (BGA), and the more unique column grid array and 
leadless chip carrier packages.  All of these package designs incorporate unique 
features for use in connecting to the PWB.  In particular, devices on QFPs are 
connected to the PWB in part through surface mount leads bent into gull-wing 
geometries.  An illustration of a surface mount package featuring inclined gull-wing 
leads can be seen in Figure 1.   
 
 5 
 
 
 
Figure 1.  Inclined gull-wing leaded surface mount package. 
 
Second level interconnection relies on component attach to serve critical electrical 
and mechanical functions.  Electrically, the component attach is called upon to carry 
an electrical signal between the leads and the PWB without distortion or 
discontinuity.  With respect to its mechanical function, it is required to fasten the 
component to the PWB and resist any loading that would separate the two.  Although 
these two functions are distinct, it is clear that they are interrelated because without a 
mechanical connection, electrical continuity cannot exist.  Some surface mount 
component attaches also serve a thermal function by transmitting heat from the 
semiconductor device to the PWB. 
 
The common method of manufacturing joints between the surface mount package and 
PWB involves the screen or stencil printing of a metal paste onto the PWB and 
subsequent placement of the surface mount package on top of the printed paste pads.  
The entire assembly is then subjected to a temperature profile that includes a hold 
time in which the melting temperature of the metal paste is exceeded, followed by 
cooling to a temperature at which the joints solidify.  This temperature excursion is 
referred to as reflow and a sample profile is shown in Figure 2.  Put another way, 
 
 6 
 
reflow processing of electronic assemblies operates on the basic principle that the 
maximum temperature of reflow, Tr, must exceed the melting temperature, Tm, of the 
surface mount component attach. 
 
 
Figure 2.  Temperature vs. time of a sample reflow profile for Sn-Ag-Cu solder. 
 
In order for the surface mount solder joints to perform their intended functions at a 
given temperature they must be solid, which implies the application temperature, Ta, 
needs to remain lower than Tm.  This translates into the need for solder materials with 
higher melting temperatures when a surface mount component attach is required for 
HT conditions.  Solder materials commonly used for products subject to the 
commercial temperature range are typically not implemented as part of an electronic 
product subject to HT conditions because of their lower melting temperatures.  For 
example, the popular component attaches of 96.5 wt. % Sn 3.0 wt. % Ag 0.5 wt. % 
Cu (SAC305) and the former industry standard, 63 wt. % Sn 37 wt. % Pb (eutectic 
Sn-Pb) feature melting temperatures of 217?C and 183?C, respectively.  Although 
some HT conditions may not exceed these temperatures, the threat of time dependent 
deformation is a common reliability concern when using these solders at elevated 
temperatures because the proximity of the operating temperatures to the melting point 
 
 7 
 
of the material is known to accelerate the creep phenomenon.  Because of these 
reliability concerns, many gold-based solders and high-Pb alloys are utilized for HT 
conditions because of their higher melting temperatures and resistance to creep.  A 
few of these alloy formulations are listed in Table 2.  For those combinations that are 
off-eutectic, the temperature range is given for the melting temperature. 
 
Table 2.  Solder alloys for HT applications. 
Solder Alloy (wt. %) Solidus Temperature (?C) Liquidus Temperature (?C) 
97Au 3Si 363 363 
88Au 12Ge 356 356 
95Pb 5Sn 308 312 
80Au 20Sn 280 280 
90Pb 10Sn 275 302 
 
As the table shows, these melting temperatures are significantly higher than those of 
SAC305 and eutectic Sn-Pb.  Typical reflow profiles call for peak temperatures ~30 
to 40?C above a solder?s liquidus temperature, meaning that implementation of these 
HT solder alloys would require profiles with peak temperatures ranging from ~310 to 
400?C.  Such high temperature reflow profiles can lead to significant processing 
related damage of the assembly and component materials prior to application.  These 
damages may include charring of the PWB, delamination of the copper from the 
boards, popcorning, outgassing, and burning of the component encapsulant as well as 
increased intermetallic formation.  In addition to such damage, residual stresses built 
into the electronic assembly are proportional to the absolute difference between peak 
reflow and room or operating temperature.  As a result, the reflow profiles for HT 
solders can lead to increased residual stresses within the assemblies, which together 
with thermal damage translates into consumption of an appreciable amount of the 
assembly?s useful life prior to implementation in the field. 
 
 8 
 
 
In order to avoid the necessity of processing at such elevated temperatures to achieve 
a HT second-level interconnection, a transient liquid phase sintered (TLPS) material 
has been investigated as a possible solution.  TLPS is a variation on liquid phase 
sintering whereby a liquid is formed and consumed by diffusional homogenization 
during the sintering process [6].  In effect, the processing temperatures for TLPS 
materials remain below the melting point of at least one of the material constituents, 
so that solid and liquid phases coexist until the liquid is completely consumed by the 
diffusional process.  When the material constituents are chosen carefully, the TLPS 
process can be a very powerful tool in the manufacture of HT electronics. During this 
investigation, implementation of a TLPS material composed of a select ratio of Ag 
and In powders subject to processing temperatures more commonly associated with 
SAC305 yielded a high temperature component attach that can remain solid up to at 
least 540?C.  Although TLPS processing offers an attractive alternative to traditional 
reflow processing, the transient liquid during the process is influenced by several 
parameters that make TLPS extremely sensitive to processing conditions [6].  Prior 
work with the Ag-In TLPS system [7][8] provided the necessary predictive 
capabilities for manufacturing interconnections from the Ag-In TLPS material.  This 
work serves to investigate the feasibility of implementing the Ag-In TLPS material as 
an alternative high temperature component attach by utilizing common surface mount 
attach fabrication methods to fabricate samples and compare them to assemblies 
manufactured from commercially available solder pastes: SAC305 and 80 wt. % Au 
20 wt. % Sn (eutectic Au-Sn).  The successful application of the Ag-In TLPS alloy as 
a 2nd level interconnection not only depends on whether surface mount joints can be 
 
 9 
 
successfully fabricated from the material, but also on how reliable the subsequent 
joints are when subjected to environmental stressing conditions.  Relative reliability 
of the three component attaches was determined from HT passive thermal cycling of 
samples fabricated on identical test vehicles that were monitored for changes in 
critical mechanical and electrical metrics.   
 
 
 10 
 
Chapter 2: Prior Work 
 
This chapter outlines the results of prior work that were employed in developing the 
design and manufacturing approaches used in this study.  In the manufacturing of 
interconnections from the commercially available surface mount materials, SAC305 
and eutectic Au-Sn, an effort was made to adhere to commonly accepted 
manufacturing practices gathered from the available literature.  With respect to the 
Ag-In TLPS material, efforts were made to utilize the previous work [7][8] regarding 
implementation of this processing method and binary metallic system while 
contributing original work that serves to advance it as a HT surface mount component 
attach and encourage future investigation in this area. 
 
2.1 SAC305 Studies 
 
RoHS and WEEE legislation has served as a catalyst for significant materials related 
research aimed at identifying a suitable replacement for the former industry standard 
solder paste, eutectic Sn-Pb.  Now that these directives have been in effect for over 
two years, a number of ternary and quaternary alloys composed of select ratios of tin, 
silver, and copper have become the front runners for manufacturing of commercial 
electronic products that feature surface mount technology.  Of the ternary alloys, 
SAC305 has been heavily researched as an attach material for a wide variety of 
electronic packages and applications.  These investigations have attempted to provide 
behavior predictive data so that users of the surface mount attach possess a good 
understanding of the limitations associated with its manufacture and application 
environments.  With respect to this investigation, all recommended manufacturing 
 
 11 
 
processes regarding the SAC305 paste were readily available from the paste supplier.  
The handling, application, and temperature control of test assemblies manufactured 
with the SAC305 material were all performed according to these recommendations.  
This information included the recommended reflow profile tailored to the particular 
flux packaged with the SAC305 alloy.  Discussion of the joint manufacture for this 
solder paste is provided in detail within the experimental methods of this work.  In an 
effort to make contributions in the area of testing SAC305, investigations were 
consulted to understand the limits of past temperature related environmental testing 
on SAC assemblies.  This information, when assessed along with that collected for 
eutectic Au-Sn and Ag-In TLPS, provided important data that aided in the selection 
of the type and magnitude of environmental testing of the experimental assemblies. 
 
As a direct result of the research resources dedicated to understanding SnAgCu 
solders, solder pastes such as SAC305 are now widely used materials in electronics 
manufacturing.  Because these solder materials have been tasked with replacing 
eutectic Sn-Pb, much research has gone into uncovering how SnAgCu solders would 
perform under material and environmental conditions similar to those of the former 
Sn-Pb assemblies.  Although SnAgCu alloys have served as suitable replacements to 
leaded solders, changes to the electronics manufacturing industry were required to 
accommodate them.  The most notable of these adaptations was that made to the 
reflow profile.  Lead-free alloys like SAC305 (Tm = 217?C) require higher reflow 
temperatures on the order of ~240 to 250?C to process the pastes.  In contrast to the 
eutectic Sn-Pb melting temperature of 183?C, this higher SnAgCu melting 
temperature opens up the possibility of elevated temperature applications 
 
 12 
 
inappropriate for electronic assemblies featuring eutectic Sn-Pb attaches.  Based on a 
survey of the current temperature tests performed on SAC assemblies, the maximum 
temperature of many of these investigations has been ~150?C.  Choubey et. al. 
[9][10][11] worked to characterize the intermetallic growth between lead-free solders, 
including SAC305, and various board metallizations at aging temperatures up to 
150?C.  In one particular investigation [10], the growth constants and activation 
energies of intermetallic formation for the various lead-free solders and finish 
combinations were predicted from the collected data.  The effect of aging temperature 
and board finish on the pull strength of surface mount leads were also assessed as a 
function of time at aging temperature.  In Kim et. al. [12], lead-free and eutectic Sn-
Pb solders were used to fabricate solder joints that were subjected to temperature 
cycling from room temperature up to a maximum of 150?C.  The fatigue testing of 
these samples demonstrated that for higher ?T values, eutectic Sn-Pb solders 
exhibited higher fatigue resistance while for lower ?T values, the lead-free 
assemblies exhibited higher fatigue resistance.  An investigation performed by 
Suhling et. al. [13] featured passive thermal cycling of lead-free ternary and lead-free 
quaternary alloy samples between ?40 and 125?C as well as ?40 and 150?C.  In 
contrast to other similar studies, the samples fabricated for that investigation paired 
ceramic chip resistors with organic circuit boards in an attempt to simulate the 
material combinations of automotive electronics.  Lastly, Ma et. al. [14] assessed the 
relative reliability of SnAgCu solders during temperature aging tests at maximum 
temperatures of 150?C.  From the temperature related lead-free investigations that 
were reviewed in preparation for this study, it is clear that passive thermal cycling 
 
 13 
 
and temperature aging tests are popular environmental conditioning schemes for 
evaluating the reliability of lead-free and eutectic Sn-Pb systems.  From these studies 
[9][10][11][12][13][14], it is clear that the commonly used maximum temperature for 
testing of the lead-free assemblies has been 150?C.  In an effort to supplement these 
types of reliability studies, the maximum temperature of testing during this 
investigation will push the limits of these prior works and assess the reliability of 
SAC305 surface mount solder joints at even higher temperature extremes.  The 
resulting data could prove useful to those looking to implement a lead-free solder 
paste alloy for more demanding HT applications. 
 
2.2 Eutectic Au-Sn Studies 
 
In contrast to the wealth of studies and data available for SnAgCu, the investigations 
conducted on eutectic Au-Sn are fewer.  The comparatively smaller number of 
applications for the material combined with its significantly higher cost translates into 
fewer studies on the alloy.  Featuring a melting temperature of 280?C, eutectic Au-Sn 
is an alloy capable of surviving many demanding HT applications.  This hard solder 
is well known for its excellent fatigue and creep resistance due to its high flow stress 
[15] and high melting temperature, which make it an ideal candidate to serve as a HT 
baseline against which the performance of the SAC305 and Ag-In TLPS solders can 
be compared in this study.  Many applications of the eutectic Au-Sn alloy feature 
solder preforms shaped as ribbons, pellets, or ring frames [16].  However, as the size 
and pitch of surface mount devices continue to decrease, the feasibility of preforms as 
an option in manufacturing HT electronic assemblies decreases as well.  Early studies 
 
 14 
 
with the Au-Sn alloy demonstrated that highly reliable joints could be formed with 
eutectic Au-Sn preforms.  In studies conducted by Matijasevic and Lee [17][18], 
GaAs devices were successfully attached to alumina substrates using eutectic Au-Sn 
preforms.  In order to ensure void-free attachment, pressure was applied to the 
die/preform/substrate stack-up when subjected to the reflow profile.  Further study 
[19] demonstrated that silicon devices could be successfully attached to alumina 
substrates in the same manner.  In all three of these studies, thermal shock testing of 
the samples between ?196 and 160?C demonstrated reliability in that all test 
assemblies avoided joint or die cracking during testing.  In contrast to the 
performance of eutectic preforms, the application of eutectic Au-Sn solder paste has 
shown the tendency to form voids in joints and attachments fabricated from the paste 
material [16][20].  Fabrication with eutectic Au-Sn preforms is often flux free and 
pressure assisted, circumventing the voiding associated with flux containing solder 
pastes.  However, minimization of voids within Au-Sn bonds comes at a cost: 
manufacturing of electronic assemblies through the application of Au-Sn solder 
preforms is complicated by the need for more fixturing and features lower throughput 
when compared to conventional stencil printing methods.  In an effort to make the 
implementation of eutectic Au-Sn pastes more practical and appealing, the work by 
Ishikawa et. al. [16] addressed the voiding issues inherent in the manufacturing 
process of Au-Sn pastes.  Recommendations from the work ranged from lowering the 
surface oxidation of the metals by using larger particle sizes to increasing the 
temperature and time of the reflow profile preheat stage to making slight changes to 
the composition of the alloy by moving to ~1 to 2 wt. % richer in Sn [16].  Although 
 
 15 
 
eutectic Au-Sn is commercially available in paste form, there are few studies in 
existence that investigate this embodiment of the alloy as a surface mount component 
attach.  The work performed during this investigation will serve to augment the 
available data on the Au-Sn alloy by conducting passive thermal cycling under unique 
temperature excursions and subsequent stressing conditions for the flux containing 
solder paste version of the alloy.  Successful manufacture of Au-Sn solder joints 
without the aid of an inert atmosphere during reflow will demonstrate 
manufacturability of the paste using standard electronics manufacturing practices. 
 
2.3 TLPS for Surface Mount Attach 
 
While the SAC305 and eutectic Au-Sn alloys are well characterized, commercially 
available solder materials, the novel Ag-In TLPS material proposed as a surface 
mount attach for this investigation will require significantly more process 
development to successfully attach surface mount components to the PWB.  Although 
the principle of TLPS has been investigated and demonstrated for a number of 
material systems and applications, there has not yet been a translation of the 
technology to successfully attach surface mount components to the PWB using 
traditional manufacturing methods.  Furthermore, there has not been any work that 
serves to assess the viability of such an attach material based upon its behavior during 
environmental testing.  Prior work [21][22] with the Sn-Bi and Sn-Sb systems 
demonstrated that solder pastes mixed from elemental particles and flux were able to 
undergo a shift in melting temperature that resulted from compositional 
homogenization that occured during the TLPS process.  However, further steps to 
 
 16 
 
attach surface mount components to circuit boards with these pastes in order to assess 
their performances under environmental stresses were not undertaken.  In addition, 
the melting temperature shifts exhibited during studies of the Sn-Bi and Sn-Sb were 
within the 140 to 250?C range which are notably lower than those possible when 
using the Ag-In system.  In Palmer et. al. [23], bonding of copper strips was 
accomplished using liquid assisted sintering of an altered commercial solder paste.  
During this investigation, Sn-Ag solder paste was doped with an Sn-Bi alloy that 
served as the liquid forming constituent during temperature processing.  Joints formed 
between the copper strips using liquid assisted sintering demonstrated mechanical 
strength values comparable to Sn-Pb joints while those Sn-Ag joints sintered without 
the Sn-Bi metal present in the mixture failed to demonstrate mechanical integrity.  
Successful attachment of surface mount components using hybrid conductive 
adhesives that implemented TLPS to form internal metallic frameworks was 
accomplished by Gallagher and Shearer [24][25].  The cohesive structures formed by 
the metals having undergone TLPS were shown to have improved the mechanical 
integrity of samples fabricated with the hybrid conductive adhesive over those 
fabricated with discrete metallic particles dispersed in the adhesive matrix.  More 
recently, a multi-stage study of the Ag-In system processed through TLPS was 
completed [7][8].  This investigation, conducted at the University of Maryland, first 
worked to provide modeling tools that would allow for prediction of required 
processing related parameters during TLPS manufacture of the Ag-In system based 
on a number physical considerations that included silver particle size, constituent 
weight ratios, ramp rate during processing, isothermal temperature during processing, 
 
 17 
 
and isothermal hold time [8].  Subsequent work with the Ag-In TLPS material 
provided electrical and mechanical characteristics of the processed material as a 
function of constituent weight ratios and concluded with the demonstration of the 
material?s applicability as a die attach for HT or high power environments.  Although 
successful attachment of silicon devices to alumina substrates was demonstrated [8], a 
full reliability analysis of the Ag-In TLPS was never conducted.  In addition, the use 
of the material as a surface mount attach was not investigated.  The encouraging 
results from this work served to inspire the investigation of the Ag-In TLPS system as 
a viable HT surface mount attach as part of this study.  The work performed on the 
Ag-In TLPS system during this investigation will serve to demonstrate its 
manufacturability using common surface mount manufacturing methods.  Inherent in 
this demonstration will be the successful attachment of surface mount components 
using the Ag-In TLPS paste.   Lastly, this work will provide novel data regarding the 
material?s reliability during environmental conditioning when used as a surface 
mount solder joint. 
 
 18 
 
Chapter 3: Initial Embodiment of Ag-In TLPS Surface 
Mount Component Attach 
 
3.1 Selection of Ag-In System 
 
Before discussing the initial embodiment of the Ag-In TLPS solder attach, it is 
necessary to give some insight into why the binary system was selected as a candidate 
for applying TLPS manufacturing.  Many researchers [21][22][26][27][28][29] have 
worked on understanding and applying the TLPS process for their particular alloy of 
interest.  The phase diagrams of these selected systems feature characteristics that are 
conducive to producing a successful TLPS compact.  Systems such as Au-Sn, Al-Au, 
Fe-Ti, Fe-Al, Ag-Sn, Cu-Sn, Sn-In, Sn-Bi and Ag-In, among others, have been 
identified or used as TLPS candidates due to availability of a comparatively low 
melting temperature eutectic or additive liquid during processing.   
 
 
 19 
 
 
Figure 3.  Ag-In Binary Phase Diagram from Hansen and Anderko [30]. 
 
With respect to the Ag-In system, the phase diagram shown in Figure 3 demonstrates 
a few features that made it attractive to use for TLPS processing of HT solder joints.  
First, the additive metal, indium, has a large solid solubility in silver.  As the diagram 
shows, up to 20 wt. % indium can be accommodated within the silver phase over a 
wide temperature range while additional high melting temperature phases extend all 
the way to ~35 wt. % indium.  This solubility of indium in the base metal, silver, is an 
important characteristic as it allows for a substantial amount of liquid forming 
additive to exist during initial stages of TLPS prior to compositional homogenization.  
In prior applications of TLPS materials, the systems have generally been constrained 
by the surrounding surfaces of a mold or combination of die and substrate.  These 
constraints leave the liquid formed during processing largely responsible for the 
 
 20 
 
rearrangement and densification of the base metal prior to solidification.  However, 
during TLPS processing of surface mount solder joints, the liquid is also required to 
form the desired heel and toe geometries common to surface mount solder joints.  
Because of this additional requirement, the availability of a significant amount of 
liquid is a very desirable characteristic in the TLPS manufacturing of surface mount 
solder joints.  In addition to the availability of liquid, the melting point differential 
between silver and indium makes it an attractive system because indium rich liquid 
can be formed at temperatures significantly lower than the melting temperature of the 
final composition.  For example, during prior investigations with this system [26][27] 
a final composition of 75-80 wt. % Ag and 20-25 wt. % In was set prior to processing 
through deposition of thin films.  Subsequent processing of these thin films at ~200 to 
220?C over time (many hours) yielded a bond capable of remaining completely solid 
up to 693?C with a liquidus temperature of ~750 to 780?C.  This subsequent shift in 
melting temperature is significant because it allows for the relatively low temperature 
processing of HT materials.  Lastly, silver and indium metals are commonly used 
materials in electronics packaging.  Silver features the highest electrical conductivity 
of any element while indium is a metal commonly recognized for its exceptional 
wetting abilities.  Because of the aforementioned characteristics, Ag-In was 
considered one of the most promising binary alloy systems to manufacture surface 
mount assemblies using TLPS. 
 
 
 21 
 
3.2 Initial Ag-In TLPS Surface Mount Solder Joints 
 
Before the Ag-In TLPS material could be tested as part of any reliability study, its 
application as a surface mount attach needed to be demonstrated first.  Although Ag-
In paste had been used to successfully attach silicon devices to alumina substrates [8], 
the physical characteristics of the application were different than those of surface 
mount components.  In particular, the volume of material deposited for attachment of 
the 10 mm by 10 mm devices [8] was well over 100 times what is needed for 
attachment of a surface mount lead.  A material that performs well when applied in 
comparatively large volumes does not necessarily behave the same way when these 
amounts are significantly scaled down.  Additionally, when applied as a surface 
mount component attach, the Ag-In TLPS material will need to demonstrate 
compatibility with common PWB metallizations whereas the aforementioned silicon 
devices and alumina substrates were coated with a layer of silver to facilitate bonding. 
 
In recognition of these differences, a paste was mixed from -325 mesh silver and 
indium particles and a commercially available, high temperature, no-clean flux in 
order to demonstrate that Ag-In TLPS could be applied as a surface mount 
component attach.  The composition of the metal load was selected to be 75 wt. % Ag 
and 25 wt. % In, which was well within the compositional window (< 35 wt. % In) 
that would result in the desired melting point shift during TLPS.  For stencil printing 
purposes, the metal load of the paste was set to 91% by weight during the mixing 
process.  The resulting Ag-In TLPS paste was printed onto ENIG, polyimide PWBs 
through an 8-mil (200 ?m) thick stencil prior to surface mount component placement 
 
 22 
 
and temperature processing.  Additional details regarding these materials and 
methods are provided in the experimental procedure section of this work. 
 
The resulting Ag-In TLPS solder joints demonstrated mechanical integrity and 
electrical conductivity; however, the density and wetting behavior of the material 
needed improving.  Because the silver within the mixture does not melt during 
processing, the resulting TLPS joints are inherently porous.  In response to this, 
subsequent formulations of the Ag-In paste were mixed to a composition of 67 wt. % 
Ag and 33 wt. % In to provide additional indium rich liquid during processing for 
improved wetting and densification.  In order to further improve densification of the 
joints, the effect of indium particle size on joint volume fraction was investigated.  
The methods and results of this study will be discussed in the following section.   
 
Lastly, fabrication of the initial Ag-In joints demonstrated that paste control during 
deposition and component placement is critical, as the Ag-In TLPS paste does not 
self-align during processing like other solder pastes.  This is due to the fact that only a 
fraction of the joint melts during processing.  The excessive paste volume that 
resulted from printing with an 8-mil (200 ?m) thick stencil was addressed in 
subsequent fabrications by using a 6-mil thick (150 ?m) stencil instead.  Apertures of 
this 6-mil (150 ?m) stencil were also adjusted to deposit slightly smaller solder pad 
footprints as well.  These adjustments allowed for improved paste control during 
fabrication of the Ag-In TLPS samples for reliability testing. 
 
 
 23 
 
3.3 Ag-In TLPS Densification Study 
 
In order to quantify the porous nature of the Ag-In surface mount joints formed 
during TLPS processing, cross sections were prepared and imaged using a light 
microscope and accompanying camera.  In this investigation it was observed that as 
the grit size of the abrasive cross sectioning paper decreased, the probability and 
frequency of material scooping increased during polishing of the Ag-In samples.  In 
order to avoid removal of material from the plane of interest, sample preparation was 
adjusted accordingly.  Infiltration of the porous Ag-In joints was performed by 
pouring the liquid mixture of epoxy and hardener over the surface mount joints prior 
to the removal and potting of the joint as an individual sample.  This ensured that 
vibration during sample cutting did not cause additional cracking or damage to the 
solder joint as well as provided the liquid epoxy-hardener mixture the opportunity to 
penetrate the porous structure of the surface mount joints prior to potting.  In addition 
to infiltrating the samples prior to potting, a rougher polishing scheme was used to 
prepare cross sections to measure volumetric density of the Ag-In joints.  These 
samples were subjected to 240 and 400 grit treatments only.  In contrast, all samples 
prepared for imaging of any visual progression that occurred during passive thermal 
cycling were subjected to 240, 400, 600, 800, and 1200 grit treatments to provide a 
polished surface capable of highlighting stress-related cracking.  All grinding and 
polishing paper used for sample preparation during this investigation featured silicon 
carbide as the abrasive material. 
 
 
 24 
 
3.3.1 Areal Analysis using ImageJ Software 
 
Following the 400 grit treatment of the samples prepared for volume fraction 
measurements, images of the cross sections were collected for analysis.  The National 
Institute of Health (NIH) image processing software, ImageJ, was used to estimate the 
volume fraction of the samples using their cross section images.  First, unprocessed 
images were cropped to areas surrounding the solder joint to limit the file size.  
Following this, the black/white threshold of the image was adjusted so that areas in 
which material existed were transformed into pure black pixels whereas those areas 
void of material were transformed into pure white pixels.  It is important to note that 
care was taken to image the samples using the light microscope with a filter 
conducive to highlighting the contrast between the epoxy potting compound and the 
metals of the solder joint, copper pad, and lead foot.  This natural contrast captured by 
the original image helped reduce errors associated with changing the black/white 
threshold as it provided clear distinctions between metal/non-metal areas during the 
transformation.  An example of this transformation on one of the images can be seen 
in the following figure. 
 
 
 25 
 
 
 
Figure 4.  ImageJ transformation from threshold adjustment. 
 
After the threshold transformation was conducted, the entire periphery of the surface 
mount solder joint was bounded using the polygon selection tool of the ImageJ 
software.  With the area of interest bounded, an analysis was performed to estimate 
the volume fraction of the Ag-In solder joint.  The bounding performed with the 
polygon selection tool can be seen below in Figure 5 as the solid line surrounding the 
entire Ag-In solder joint. 
 
 
Figure 5. Bounded joint for area fraction analysis. 
 
Following its analysis of the bounded area, the ImageJ software returned the area 
fraction of non-zero, white pixels representative of joint?s empty space; therefore, 
subtracting this reported area fraction from unity yielded the Ag-In area fraction of 
interest where material existed within the joint.  It was assumed that the area fraction 
calculated from the ImageJ analysis was equivalent to the true volume fraction of the 
 
 26 
 
entire Ag-In solder joint.  This assumption has been demonstrated for opaque, two-
dimensional sections and serves as a basis for areal analysis [31]. 
 
3.3.2 Validation of Areal Analysis using the Point-Counting Method 
 
The aforementioned areal analysis was carried out on all samples in an effort to 
reduce the time required to complete volume fraction estimation over more traditional 
estimation methods like the point-counting technique described in ASTM 
Specification E562 [32].  In addition, the software offers potentially higher accuracies 
on the volume fraction estimates because of the significantly higher resolution of the 
areal analysis; the program calculates the percentage of non-zero (white) pixels 
within the bounded area to determine the area fraction [33].  The scale of this pixel-
by-pixel grid is significantly more refined than any observer would use for an areal or 
point-counting analysis and the software execution is practically instantaneous when 
compared to the time necessary to perform an analysis by hand.  Although the ImageJ 
software can be used for the area fraction calculation described above, it is important 
to ensure that the reported area fractions are within reason when compared to 
estimations performed using point-counting methods.  In an effort to demonstrate 
agreement between volume fractions estimated using the ImageJ software and the 
point-counting technique, the representative sample from Figure 4 was analyzed 
using both methods.  Point-counting involves overlaying a grid of the desired size on 
top of the image to be analyzed.  The array of points generated by the intersection of 
the orthogonal grid lines determines where the cross section will be sampled for 
material or lack thereof.  At each of these point locations the observer assesses 
 
 27 
 
whether the point does or does not lie on the phase of interest, Ag-In material in this 
case, or whether it lays on the phase boundary.  Each point that does lie on the phase 
of interest is counted as one and each that lays on a phase boundary is counted as one-
half.  When the entire grid has been assessed, the sum of points scored for material or 
boundary hits, ?Pm, is divided by the total number of points within the grid area, PT 
[31].  The resulting value is referred to as the point fraction, PP, and is assumed to be 
equivalent to the linear, LL, area, AA, and volume, VV, fractions of the given field 
[31].  These relationships can be represented as 
 
P PP L A VP m
T
L A V? ? ? ?
? .                     (1) 
 
With respect to the point-counting method, there are a number of recommendations 
that are suggested for using it in the most accurate and efficient manner [31].  As a 
general rule of thumb, the sample magnification should be high enough so that the 
observer may be able to accurately assess the location of the grid points on top of the 
sample.  However, the magnification should be low enough so that, if possible, no 
more than one grid point falls on a given phase of interest in order to preserve 
statistical accuracy of the density estimations.  For this particular investigation, it was 
observed that an area of 750 square pixels per grid point satisfied the aforementioned 
balance of magnification.  For the given image in Figure 4, this grid resolution 
translates to an area of ~5.29x10-6 in2 (3410 ?m2) per grid point.  The overlay of this 
grid on the cropped, original image of the Ag-In solder joint from Figure 4 can be 
seen in Figure 6. 
 
 
 28 
 
 
Figure 6.  Point-counting method grid overlay. 
 
The ImageJ software was used to size and randomly place a grid over the image.  
Following placement of the grid, the point-counting method was employed as 
previously described.  From the overlay, it was observed that the field within the 
solder joint area contained approximately 80 to 85 grid points, depending on the 
randomized placement of the grid over the image.  In order to achieve an average 
point fraction for the volume fraction estimation of the Ag-In solder joint, five 
randomly placed grids of previously described spacing were laid over the image and a 
point fraction was measured in each instance.  The point fraction measured for each 
of these five instances can be viewed below in Table 3. 
 
Table 3.  Point and volume fractions of an Ag-In surface mount solder joint. 
Instance ?Pm PT Point fraction, PP Volume fraction, VV 
1 59.5 85 0.700 70.0% 
2 49.5 85 0.582 58.2% 
3 48.5 80 0.606 60.6% 
4 57.0 82 0.695 69.5% 
5 52.0 80 0.650 65.0% 
AVG 53.3 82.4 0.647 64.7% 
 
An ImageJ areal analysis of the same sample using the previously described process 
yielded an area fraction of 61.5%, which agrees to within 5% of the average 
area/volume fraction of 64.7% from the point-counting method.  With the volume 
fraction estimation process using the ImageJ software substantiated by the 
measurements performed with the point-counting method, it is now appropriate to 
 
 29 
 
discuss the volume fraction trends identified within the Ag-In surface mount solder 
joints with changing particle sizes. 
 
3.3.3 Volume Fraction Trends 
 
The intent of the volume fraction study was to understand the relationship between 
the sample area fraction and the constituent particle size with the goal of increasing 
the volume fraction of the Ag-In solder joints so as to improve the mechanical, 
electrical, and thermal properties of the joints.  With respect to the mechanical 
properties, an increase in volume fraction provides more material to withstand the 
shear forces developed during passive thermal cycling.  In terms of electrical and 
thermal performance, higher material volume fractions correspond to higher joint 
cross-sectional areas to carry current and heat, therefore reducing the inherent 
electrical and thermal resistances of the solder joints. 
 
With respect to persistent liquid phase sintered and TLPS solder joints, there are a 
multitude of variables that can be altered in order to improve the final material 
volume fraction.  These include particle size, green density, amount of additive metal, 
time, temperature, heating rate, additive homogeneity, and powder porosity, among 
others [6].  In the previous list, the term additive refers to the metal responsible for 
forming the liquid phase during processing while green density refers to the density 
of the solder joint or compact prior to liquid phase or transient liquid phase sintering.  
Prior to continuing the discussion on how to improve volume fraction, it is necessary 
to relate the term to one more commonly used in liquid phase and transient liquid 
phase sintering literature.  Many times density, rather than volume fraction, is used to 
 
 30 
 
describe how solid or compact a TLPS sample is.  As described by German [6], 
density is the mass per unit volume and is often expressed as the fraction of 
theoretical density, ?t.  Porosity serves as the compliment to fractional density, and 
when summed together, the two fractions must equal unity [6].  Volume fraction and 
density can be related to one another through ?t.  By defining volume fraction as 
 
jntsolder
metal
v V
VV ?
,                     (2) 
 
and the total mass of a porous solder joint as 
 
voidsairmetaltvoidsmetaljntsolder VVmmm ?? ???? ,                     (3) 
 
a relationship between VV and the density of the solder joint can be established with 
some additional simplification.  When metals are used for the solder joint, ?t >> ?air, 
and therefore the ?airVvoids term can be dropped from equation (3).  By dividing both 
sides of the remaining expression by Vsolder jnt, one arrives at 
 
vtjntsolder V?? ? .                     (4) 
 
As equation (4) demonstrates, the term density commonly referred to in liquid phase 
and transient liquid phase sintering literature is directly proportional to the volume 
fraction through a constant given by the theoretical density of the metals used to form 
the joint.  Subsequently, any reference to an increase in the density of the joint is 
synonymous with an increase in the volume fraction and vice versa.  Prior to 
discussion of the trends in TLPS works, the concept of densification needs to be 
defined.  It can be qualitatively described as the change in porosity of a compact 
 
 31 
 
following processing normalized to its initial or green porosity [6].  Positive 
densifications therefore correspond to compacts that have eliminated a given fraction 
of pores or voids from their structures during liquid phase or transient liquid phase 
sintering.  Densification is often expressed as a percentage because it is a 
dimensionless value [6]. 
 
Although the final density of a TLPS solder joint or compact is sensitive to a large 
number of material and process related variables, there are only a few that can be 
easily altered.  In particular, the size of the constituent metals is a variable that can be 
changed by simply mixing TLPS solder pastes using particles of varying sizes.  Due 
to the commercial availability of metal spheres and in order to maintain a manageable 
experimental scope, the base metal particle size was held constant while additive 
indium spheres of varying sizes were utilized in preparing a number of Ag-In solder 
pastes in an effort to improve the volume fraction of the resulting TLPS solder joints.  
Research in the area of liquid phase and transient liquid phase sintering has 
demonstrated that smaller particle sizes of the constituent metals have correlated to 
greater densification and final compact densities [6].  In particular, finer additive 
particle sizes positively influence the densification and final density of a processed 
compact for a number of reasons [6].  First, during the liquid phase or transient liquid 
phase sintering process, the size of the additive particles determines the size of the 
pores that form at their locations prior to their melting, spreading, and interdiffusion 
with the base metal particles.  Smaller additive particle sizes therefore leave behind 
smaller pores or voids after they form the liquid constituent of the TLPS material 
during processing.  Second, smaller additive particles sizes lead to higher green 
 
 32 
 
densities that in turn correlate to higher final densities of liquid phase and transient 
liquid phase sintered compacts.  A third reason that finer additive particle sizes aid in 
densification and improve final compact density is that as the size of the additive 
particle decreases, the distribution of the additive within the mixture improves, 
therefore improving homogeneity and subsequently leading to higher densifications 
due to a uniform distribution of liquid phase within the joint during processing.  This 
trend in increasing final compact densities with decreasing additive particle size is 
common to liquid phase, transient liquid, and reactive sintering processes for a variety 
of metallic combinations [28][34][35]. 
 
As previously mentioned, the aim of using smaller indium particles was to improve 
the densification, and subsequently, the volume fraction of the Ag-In solder joints.  
For this portion of the study, Ag-In pastes were mixed in-house using commercially 
available metal powders from Alfa Aesar and Advanced Machine and Materials.  
Silver powder with a characteristic diameter of 50 microns was obtained from Alfa 
Aesar.  Indium powders from Advanced Machine and Materials with characteristic 
diameters of 45, 25, 10, and 5 microns were purchased as well.  These characteristic 
diameters correspond to mesh sizes -325, -500, -1250, and -2500, respectively.  Each 
indium powder was mixed with the 50 micron silver powder in order to achieve 
varying ratios of silver particle diameter over indium particle diameter.  These ratios, 
R, in addition to the naming convention assigned to the subsequent mixtures can be 
seen in Table 4.  The no-clean, Indium tack flux NC-SMQ51SC was mixed with these 
metal powders in order to formulate the pastes for the densification assessment.  Prior 
to mixing, all three constituents were weighed in order to achieve a metal load of 91% 
 
 33 
 
by weight.  The composition of the metal load was formulated to be 67 wt. % Ag and 
33 wt. % In.    
 
Table 4.  Paste formulations for densification study. 
Paste Composition (wt. %) Metal Load (wt. %) DAg (?m) DIn (?m) R = DAg/DIn 
P1 67Ag 33In 91 50 45 1 
P2 67Ag 33In 91 50 25 2 
P3 67Ag 33In 91 50 10 5 
P4 67Ag 33In 91 50 5 10 
 
The surface mount components, test boards, processing time, and processing 
temperature required to appropriately manufacture Ag-In solder joints from these 
pastes are all described in detail within the experimental procedure section of this 
work.  In addition, however, each solder joint was processed under compression from 
the weight of the surface mount component.  Two distinct gull-wing, surface mount 
package sizes were used to fabricate samples with each of the pastes for a total of 8 
distinct samples; small samples fabricated from pastes P1 through P4 had solder 
joints subjected to an average lead pressure of ~.23 psi (1585 Pa), while large 
package samples had solder joints subjected to an average lead pressure of ~.14 psi 
(965 Pa).  These pressures are the direct result of the weight of the packages acting 
over the surface area of the leads that come in contact with the solder pads during 
processing.  In order to assess any change in densification as a result of indium 
particle size, four distinct cross sections were collected and averaged for each paste 
and pressure combination for a total of 32 volume fractions estimated altogether.  The 
results of the densification study are summarized in Figure 7.  
 
 
 34 
 
 
Figure 7.  Volume fraction as a function of particle ratio (R = DAg/DIn). 
 
As shown in Figure 7, as the indium particle within the Ag-In paste becomes smaller, 
indicated by an increasing particle ratio, the volume fraction or density of the 
processed solder joints increases monotonically.  Statistically significant differences 
in volume fraction can be observed between the joints formed with a particle ratio of 
10 and 1 when using the 90% confidence intervals provided.  It is believed that these 
volume fraction improvements are a direct result of increasing green density and 
indium distribution homogeneity within the Ag-In TLPS paste with decreasing 
indium particle size.  The progressive improvement of these characteristics within the 
unprocessed Ag-In TLPS pastes is evident in Figure 8. 
   
 
 35 
 
 
 
Figure 8.  Ag-In TLPS unprocessed paste ESEM/EDS images (Ag in red). 
 
Due to the higher volume fractions observed for the P4 paste samples (R = 10, DAg = 
50 ?m, DIn = 5 ?m), it was decided that samples fabricated for subsequent reliability 
testing would feature the P4 paste as described in Table 4.  It is also worth noting that 
the wetting behavior exhibited by the liquid phase during TLPS of the joints seemed 
to increase as indium particles became smaller.  This qualitative assessment was 
observed at the heel and toe areas of the Ag-In TLPS solder joints.  A comparison of 
black/white images of a large package P1 and P4 Ag-In solder joint demonstrates this 
trend.  The volume fractions of the two formulations are clearly different in addition 
to the fact that the P1 paste has largely maintained its as-deposited shape while the P4 
 
 36 
 
paste has formed a heel and toe on the lead common to commercial surface mount 
solder attaches. 
 
 
Figure 9.  Wetting comparison of P1 and P4 pastes under large package leads. 
 
 37 
 
Chapter 4: Experimental Procedure 
 
This chapter describes the materials, methods, and testing procedures necessary for 
executing the comparative reliability study between the three solder materials: 
SAC305, eutectic Au-Sn, and Ag-In TLPS.  Materials and manufacturing 
considerations regarding the test assembly are discussed in depth along with the 
selection and implementation of passive thermal cycling for environmental stressing 
of the test assemblies. 
4.1 Test Vehicle Materials 
 
In order to perform the comparative reliability study mentioned earlier in this work, a 
sufficient number of samples of each of the three solder attach materials was required 
for testing.  SAC305 and eutectic Au-Sn were fabricated using best-known 
manufacturing methods collected from product literature.  Ag-In TLPS samples were 
fabricated using methods developed as part of this thesis.  All sample fabrication was 
performed in-house using the necessary materials in conjunction with the appropriate, 
available equipment at the University of Maryland, College Park.  The test vehicle 
used during this investigation was composed of a polyimide laminate circuit board in 
conjunction with ceramic, gull-wing, surface mount components.  Materials used 
during sample fabrication of the SMT samples included the following: edge-card 
circuit boards, ceramic quadflat pack (CQFP) components, solder pastes (SAC305, 
eutectic Au-Sn, and Ag-In TLPS), and stencil foils. 
 
 
 38 
 
4.1.1 Edge-card Circuit Boards 
 
For this investigation a polyimide, edge-card test board was designed in-house and 
fabricated by an outside supplier, K&F Electronics, Inc.  Polyimide was selected as 
the laminate of choice due to its relatively high Tg (~250?C) with respect to other 
popular organic laminates such as FR-4 (Tg ~ 135-175?C) or BT resin (Tg > 175?C).  
A material possessing a high Tg was selected for the following two reasons: first, the 
higher Tg ensured that the breakdown/decomposition temperature of the organic 
laminate was above 250?C, and second, board expansion with respect to temperature 
would remain a linear phenomenon during the thermal cycling of the reliability study, 
as it would not surpass this 250?C Tg limit.  This simplifies the strain estimations 
within the solder joints.  Ceramic boards processed through low temperature cofiring 
(LTCC) methods were also considered, but design constraints and their brittle nature 
ruled them out.  In summary, the robustness of organic laminates, the additional 
survivability that comes with a high Tg, and the constant CTE during the selected 
thermal cycling were all considerations that led to polyimide as the selected test board 
material.  This mirrors the fact that most applications in the 125 to 200?C range 
routinely use polyimide boards. 
 
Design of the polyimide test boards was conducted with the layout editor, EAGLE 
4.16.  This CAD program allows users to designate groups of electrically connected 
nodes within the design, referred to as ?nets,? and connect them.  Manual routing of 
the electrical nets was conducted in the study.  Each polyimide test board was 
designed to accommodate a total of 4 CQFP components.  The description of these 
 
 39 
 
ceramic package geometries will follow the discussion of the test boards.  For 
simplicity, of these 4 CQFP packages per board, 2 44-lead CQFPs were designated as 
?small? and the remaining 2 64-lead CQFPs were designated as ?large.?  These 
simple names refer to the overall package footprint, which directly translates to the 
strains imposed on the solder joints connecting them to the test boards.  A visual 
comparison of the package sizes can be seen in Figure 10 and Figure 11 below. 
 
 
Figure 10.  Small CQFP footprint. 
 
 
Figure 11.  Large CQFP footprint. 
 
 40 
 
 
In order to facilitate installation, removal, and repair of the test board assemblies 
during the reliability study, the polyimide test boards were designed to interface with 
selected edge card connectors.  By implementing an edge card connection to the test 
boards, one allows for the electrical connection of the board to the necessary 
monitoring equipment without requiring a permanent solder connection.  The edge 
card connection allows for the easy removal and reinstallation of the polyimide test 
boards during reliability testing if and when inspection of the boards is required.  A 
description of the materials and geometries of the edge card connectors will be given 
in the subsequent equipment section of this work.  A close-up CAD image of the 
male, edge-card feature of the polyimide test boards is displayed in Figure 12. 
 
 
Figure 12.  Polyimide test board edge-card feature. 
 
The test board had an overall footprint of 4? (102 mm) by 3.2? (82 mm).  For the 
given version of the EAGLE 4.16 software, this was the maximum footprint 
permissible by the program.  Although a smaller board was possible with the given 
space and trace requirements, the board was intentionally oversized in order to 
facilitate handling during the sample fabrication and testing steps.  A board thickness 
of 59 mil (1.5 mm) was selected in order to ensure robustness of the boards during 
handling in addition to proper interfacing with the selected edge card connectors. 
 
Due to the simplicity of the board design, a single layer of copper was required to 
electrically connect all of the CQFP packages to the edge card fingers.  The thickness 
 
 41 
 
of this single layer of copper was specified as 1-ounce per square foot, which 
translates to 1.4 mil (35 ?m).  All signal traces on the test boards, excluding the 
ground trace, were designed as 10 mil (255 ?m) trace and 10 mil space.  The ground 
trace, which can be identified as the trace running along the right side of the board 
design in Figure 13, was specified as 25 mil (635 ?m) trace width in order to reduce 
the total loop resistance of each of the electrical chains.  Trace and surface pad 
arrangement on the test boards was performed in order to create an electrical ?daisy-
chain? pattern that, when assembled with the CQFP components, would form 
continuous electrical chains.  A total of 8 distinct electrical chains were specified for 
each CQFP, resulting in a total of 32 distinct chains per test board.  Details regarding 
the wiring within the CQFP components will be given in the following section.  The 
surface mount pads of the polyimide board were finished with 3 to 8 micro inches 
(0.076 to .2 ?m) of immersion gold over 100 to 200 micro inches (2.54 to 5.08 ?m) of 
electroless nickel (ENIG); edge connector pads were finished with 30 micro inches 
(.76 ?m) of gold over 500 micro inches (12.7 ?m) of nickel. 
 
 
Figure 13.  Polyimide test board design. 
 
 42 
 
 
4.1.2 Ceramic Quadflat Pack Components 
 
Selection of the type of surface mount components was as critical to this study as the 
selection of the test boards.  Elevated temperatures during fabrication and thermal 
cycling eliminated epoxy-based SMT components from consideration due to their low 
glass transition temperatures.  For example, a popular molding compound used in 
many dummy components, Sumitomo 6300H, features a glass transition temperature 
of ~165?C [36] which falls within the extremes of the planned temperature cycling of 
this investigation.  Although suitable for lead-free manufacturing and subsequent 
standard temperature cycling, this type of material was unlikely to survive the 
elevated reflow temperatures required for eutectic Au-Sn assembly, the extended 
processing time required for the Ag-In TLPS assembly, or the elevated temperature 
cycling.  In light of the need to avoid common plastic encapsulated surface mount 
packages for the test assemblies, ceramic body components were researched as an 
alternative.  Ceramic surface mount components were purchased from Practical 
Components as flat ?CQFP? packages.  In contrast to typical plastic encapsulated 
quadflat packages, these ceramic packages were only available with flat leads 
connected to a common tie-bar around the perimeter of the package.  Two component 
sizes were selected for the study: ?large?, 64-lead CQFPs and ?small,? 44-lead 
CQFPs.  The desired ceramic packages were purchased as inactive, dummy 
components with internal daisy-chain patterns designed to mate with those on the 
polyimide test boards to form continuous electrical chains.  The package daisy chains, 
when connected to those on the polyimide test boards, were designed to create 8 
 
 43 
 
electrically distinct chains within each package.  The intent of each of these electrical 
chains was to allow for monitoring of their respective corner solder joints, as these 
joints are expected to be the first to fail during temperature cycling.  Schematics of 
the particular daisy-chain wiring schemes within the ?large? and ?small? CQFPs can 
be seen in Figure 14 and Figure 15, respectively.  The bold, curved lines between 
adjacent leads denote an internal electrical connection.   
 
 
Figure 14.  64-lead CQFP daisy chain wiring scheme. 
 
It is important to note that due to the odd number of leads per side on the 44-lead 
CQFP, not every lead could be utilized in creating a pattern capable of forming 8 
electrical chains when joined to the test board.  In the case of the 64-lead CQFP 
above, the even number of leads per side allows for all to be used in the continuous 
electrical chain.  Although failure at the corner joints was expected during 
temperature cycling, it is worth mentioning that the daisy-chain patterns specified for 
the ceramic packages were designed in order to detect joint failure at any of the solder 
 
 44 
 
joints within the package.  The only caveat to this was that a single interior lead on 
each side of the 44-lead CQFP could not be monitored. 
 
 
Figure 15.  44-lead CQFP daisy-chain wiring scheme. 
 
In order to singulate and bend each of the flat leads into gull wing geometries 
common in many surface mount packages, a gull wing bending tool was necessary.  
A custom designed bending and cutting tool from Excelta Corporation made it 
possible to form and trim the flat leads of the CQFP packages into the desired gull-
wing geometry displayed in Figure 16.  These dimensions are important because they, 
in conjunction with the thickness and width of the leads, determine the lead 
compliancy that affects the strain energy density within the joints during passive 
thermal cycling.  In both ?large? and ?small? CQFP packages, the distance between 
adjacent leads (lead pitch) was 50 mil (1.27 mm).  The decision to use parts with such 
a large lead pitch was influenced heavily by manufacturability concerns; when the 
pitch gets smaller, assembly limitations as well as adjacent pad shorts become 
significant hurdles to constructing experimental samples. 
 
 45 
 
 
 
Figure 16.  Desired CQFP Gull-wing geometry (not to scale). 
 
With respect to the materials that composed the CQFP components, the ceramic in the 
packages was Kyocera A-440 alumina, while the leads were Alloy 42.  The leads of 
all CQFP packages were finished with 60 micro inches (1.52 microns) of gold over 
100 micro inches (2.54 microns) of nickel.   
 
4.1.3 Combined Test Vehicle 
 
When assembled, the resulting global coefficient of thermal expansion (CTE) 
mismatch between the polyimide board and ceramic components (?polyimide- ?alumina) 
was approximately 5 ppm/?C.  This disparity in global expansion between the two 
materials is responsible for the strains imposed on the solder joints during passive 
thermal cycling.  The design of the test vehicle allowed for the selection of ideal 
board, assembly, and component materials.  The resulting metallization stack-up prior 
to processing of the test vehicle assemblies is summarized in Figure 17. 
 
 
 46 
 
 
Figure 17.  Test vehicle metallization stack-up. 
 
4.1.4 Solder Pastes 
 
The first of the three solder pastes to be discussed is SAC305.  The SAC305 solder 
paste used during this investigation was Kester?s EnviroMark 907.  This lead-free 
paste formulation featured a no-clean chemistry, meaning that post-reflow steps 
involving the cleaning of flux residues were not required during the fabrication 
sequence.  Notable characteristics of the Kester SAC305 paste include its excellent 
solderability to ENIG surface metallizations, successful stencil printing down to 
component pitches of 16-20 mil (0.4-0.5 mm), as well as the ability to be processed in 
an air or nitrogen atmosphere during reflow [37].  With respect to the metal to flux 
ratio within the SAC paste, a typical metal load of 88.5% by weight was used.  Metal 
loads up to ~90% by weight are typical of pastes intended for stencil printing, while 
those up to ~88% by weight are intended for dispense processes [38].  The size of the 
metal particles within the SAC305 paste were type 3 (-325/+500 mesh).  Particles 
classified as type 3 mesh range in diameter from 25-45 microns (~1-2 mil).  The 
recommended reflow profile of the Kester SAC305 paste was provided in the data 
 
 47 
 
sheet accompanying the product.  Further discussion of the recommended reflow 
parameters for this paste will follow in a discussion of the Omega furnace. 
 
The second of the three solder pastes investigated was eutectic Au-Sn.  This Au-Sn 
solder paste was acquired from Indium Corporation of America, where it is classified 
as Indium Alloy 182.  As indicated by the ?eutectic? labeling, the metal load of the 
paste is composed of 80 wt. % Au and 20 wt. % Sn.  The metal particles within the 
paste were again type 3 (-325/+500 mesh).  The metal load of the eutectic Au-Sn 
solder paste was 91% by weight.  The paste was formulated by mixing the previously 
described metal load with Indium flux NC-SMQ51SC.  This particular flux is 
characterized by a no-clean formulation and wide reflow processing window that 
allows it to be used in conjunction with the high melting temperature eutectic Au-Sn 
alloy [39].  The recommended reflow parameters [40] of the eutectic Au-Sn solder 
paste will be discussed in a subsequent section. 
 
The last of the solder pastes used for sample fabrication for the reliability 
investigation was the Ag-In TLPS paste.  This paste was mixed in-house using 
indium and silver powders along with a high-temperature flux formulation.  Earlier 
experimentation with the effect of particle size on densification provided guidance 
regarding the size of silver and indium particles that should be used in the final 
embodiment during the reliability testing portion of the investigation.  Silver powder 
with a characteristic diameter of 50 microns was obtained from Alfa Aesar.  Indium 
powder from Advanced Machine and Materials with a characteristic diameter of 5 
microns was purchased as well.  The no-clean, Indium tack flux NC-SMQ51SC was 
 
 48 
 
mixed with these two metal powders in order to formulate the final paste for 
reliability testing.  Prior to mixing, all three constituents were measured in order to 
achieve a metal load of 91% by weight.  The composition of the metal load was 
formulated to be 67 wt. % Ag and 33 wt. % In.  As described earlier, this composition 
provides a melting point shift during processing that results in a final melting 
temperature of 540?C. 
 
4.1.5 Stencil Foils 
 
In order to manufacture test assemblies using the test boards, CQFPs, and solder 
pastes previously described, a method of depositing the solder paste onto the boards 
was required.  In SMT manufacture, paste deposition is one of the first steps 
performed in the fabrication sequence, prior to component placement and reflow 
processing.  There are a number of ways to deposit solder paste in a controlled 
manner, some of which include screen printing, stencil printing, as well as dispensing.  
In this investigation, stencil printing was selected as the method of deposition.  This is 
a common method of SMT assembly which provides improved deposition control 
over the older screen printing methods.  As was the case with the polyimide test 
boards, the design of the stencil foil was performed in EAGLE 4.16.  A CAD image 
of the stencil foil design can be seen in Figure 18. 
 
 
 49 
 
 
Figure 18.  Stencil foil design in EAGLE 4.16. 
 
Those areas in hatched texture represent the apertures within the stencil foil that allow 
paste to be printed onto the test board solder pads when properly aligned.  In order to 
ensure proper stencil alignment, two circular stencil apertures 200 mil (5.08 mm) in 
diameter were placed on the far right side of the stencil.  Corresponding fiducials 
placed on the test board, as seen in Figure 13, were used to help properly align the 
stencil prior to printing.  The stencil foil designs were forwarded to Photostencil for 
fabrication by laser cutting on stainless steel foils.  Specification of the stencil 
apertures ensured that the aspect and area ratios were above the commonly prescribed 
minimum values [41], 1.5 and .66, respectively.  Apertures of the 64-lead CQFPs 
featured aspect ratios of 2.85 and area ratios of 1.156 while those of the 44-lead 
CQFPs featured aspect ratios of 1.85 and area ratios of .804.   In order to fabricate the 
SAC305 and eutectic Au-Sn test assemblies, an 8-mil (200 ?m) thick stencil foil with 
the above aperture design was utilized.  Initial attempts to fabricate Ag-In TLPS 
samples using this stencil thickness led to issues regarding unnecessarily large solder 
volumes.  Subsequent Ag-In TLPS samples were fabricated with a foil thickness of 6 
 
 50 
 
mil (150 ?m), which produced more consistent test assemblies with the solder 
material. 
 
4.2 Test Vehicle Fabrication 
 
Fabrication of the board and CQFP assemblies for each of the component attaches 
was performed by stencil printing the paste, placing the CQFPs, and subjecting the 
assembly to a specified profile within the Omega LMF-3550 infrared (IR) furnace.  
Recommended profiles for the SAC305 and eutectic Au-Sn solder pastes were 
gathered from the available data sheets [37][40] while the processing profile for the 
Ag-In TLPS paste was designed using an empirical model capable of predicting the 
remaining weight % indium based on a number of factors that include holding time, 
ramp rate, processing temperature, among others [8]. 
 
4.2.1 Stencil Printing 
 
Stencil printing of the SAC305, eutectic Au-Sn, and Ag-In TLPS solder pastes 
occurred by way of a manual, fixture-assisted method.  A hinged fixture to hold the 
stencil foil and polyimide test board was constructed and implemented during this 
portion of the fabrication sequence.  Hinging of the stencil foil provided a fixed 
stencil edge that helped secure the foil during the printing process as well as 
facilitated the controlled removal of the foil following paste printing.  This fixture can 
be seen in Figure 19.  All boards were cleaned with isopropyl alcohol in order to 
remove any contaminants prior to the stencil printing process. 
 
 
 51 
 
 
Figure 19.  Stencil printing fixture. 
 
Prior to printing, alignment of the stencil foil to the test board was performed by 
using the circular stencil apertures and corresponding test board fiducials mentioned 
earlier.  Once alignment was achieved, a paste roll approximately 200 mil (5.08 mm) 
in diameter was deposited along one side of the stencil foil.  Using an electroformed 
nickel squeegee blade, this paste roll was swept across the stencil in order to deposit 
the paste within the open apertures of the foil design.  The direction of the paste print 
with respect to the design of the stencil foil can be viewed in Figure 20. 
 
 
Figure 20.  Stencil printing orientation and direction. 
 
 52 
 
 
Following a successful single pass of the squeegee blade and inspection of the filled 
aperture holes, the stencil foil was carefully removed from the surface of the test 
board by opening the hinged fixture.  It is important to note that each pass of the 
squeegee was terminated before arriving at the circular stencil apertures to avoid 
depositing paste directly on the soldermask of the test boards.  With the paste 
successfully deposited on the surface mount pads of the polyimide test boards, the 
next step performed in the fabrication sequence was that of CQFP component 
placement. 
 
4.2.2 CQFP Placement 
 
Placement of the CQFP packages was also performed by way of a manual, fixture-
assisted method.  Prior to placement, all CQFP components were checked for 
coplanarity and then cleaned in a series of solvents: acetone, then methanol, and 
lastly, isopropyl alcohol.  The fixture used to assist the manual placement of these 
CQFPs was designed to house a vacuum pick tool with the freedom to move in the 
vertical direction.  The fixture was intended to facilitate the alignment and placement 
process by providing a reliable way in which the components could be suspended 
above the paste printed test boards during the alignment process.  An image of this 
fixture with the vacuum pick tool can be observed in Figure 21. 
 
 
 53 
 
 
Figure 21.  CQFP component placement fixture. 
 
Using the component placement fixture, parts were picked up, positioned over top of 
their location on the paste printed test board, lowered to an appropriate height for 
alignment, and finally placed onto the board.  Because of the coarse pitch of the 
surface mount packages, alignment could be performed visually by aligning two 
perpendicular sides of a package and checking for alignment of the remaining two.  
The design of the component placement fixture ensured that visual alignment could 
be performed by looking down onto the board: the ideal perspective for the task.  The 
last of the fabrication steps, reflow/temperature processing, followed directly after the 
completed placement of all components on the test board. 
 
4.2.3 Reflow / Temperature Processing 
 
Final processing of the assembled test vehicles required reflow temperature 
processing within the Omega LMF-3550 infrared (IR) furnace.  The specifications of 
 
 54 
 
the machine state that it is capable of reaching temperatures ranging from 50 to 
1100?C at a ramp rate of 0 to 40?C/min [42].  In order to achieve a more acceptable 
ramp rate for reflowing electronic assemblies (on the order of 60?C/min) and to 
ensure that the solder paste within the test assembly was experiencing the desired 
reflow profiles, temperature testing of the chamber was required.  Earlier work with 
the Omega furnace had demonstrated that the optimum sample placement within the 
machine to achieve maximum ramp rate was approximately half of its vertical height.  
All temperature tests and sample processing occurred at this height by way of a 
lightweight, aluminum stand fabricated for use within the furnace. 
 
Due to the internal size restrictions of the Omega furnace, only two fully populated 
test assemblies could be processed at the same time.  Because of this constraint, all 
temperature testing was performed on a dummy set of two polyimide test boards, 
without solder paste, and fully populated with CQFPs.  K-type thermocouples were 
evenly distributed among the two test boards and positioned underneath the gull wing 
leads of the CQFPs in order to mimic pads of solder paste subjected to the selected 
temperature profile.  The arrangement of the thermocouples with respect to the boards 
and CQFP components can be seen in Figure 22.  It is important to note that 
thermocouples 1 through 4 were arranged as a primary diagonal around which 
thermocouples 5 through 8 were used to collect temperature data on other 
geographical extremes of the test board arrangement. 
 
 
 55 
 
 
Figure 22.  Furnace temperature testing thermocouple arrangement. 
 
The Agilent 34970A Data Acquisition Switch Unit was used to collect temperature 
data from these prearranged K-type thermocouples.  Each thermocouple was wired 
into its respective channel on the 34901A 20-channel multiplexer module that 
interfaced directly with the 34970A data logger.  Data was gathered at a rate of 1 
reading every 2 seconds.  Further description of the data logger and switch units will 
be provided later during a discussion of the monitoring performed for reliability 
testing. 
 
In order to create a distinct profile for each of the pastes used for this study, active 
monitoring of the temperature data as received by the 34970A data-logging unit was 
required.  When necessary, adjustment to the typical furnace procedures were 
performed in order to achieve temperature profiles as recommended by the paste 
manufacturers [37][40].  Regarding these adjustments, pre-heating of the furnace to a 
given steady-state temperature was used to increase the ramp rate of processing 
 
 56 
 
performed within the furnace.  Because the furnace was not equipped with any 
cooling technologies, removal of the test assemblies to the ambient was performed to 
cool them at an acceptable rate.  Alternative cooling schemes were considered and 
attempted, but heterogeneous cooling rates in addition to significantly longer cooling 
times thought to negatively influence microstructure of the as-built joints were 
observed during their implementation.  To summarize, pre-heating and sample 
removal was implemented in all three reflow/temperature processing profiles in order 
to achieve the desired forms. 
 
4.2.3.1 SAC305 Reflow Profile 
 
Development of the SAC305 reflow profile was based upon that recommended within 
the EM907 product data sheet [37].  As this profile in Figure 23 shows, the 
manufacturer recommended a ramp of less than 2.5?C/sec to a peak temperature 
between 235 and 255?C approximately 3.5 minutes (210 seconds) into processing.  
The total time above reflow was recommended as 60-75 seconds and no more 90 
seconds altogether.  Although a cool down rate was not prescribed within the 
SAC305 documentation, a recommendation of less than 4?C/sec was found within 
eutectic Au-Sn documentation and used for all solder materials.      
 
 
 57 
 
 
Figure 23.  Manufacturer?s recommended SAC305 reflow profile. 
 
In order to achieve the SAC305 reflow profile prescribed above, the Omega furnace 
was preheated to 275?C and allowed to stabilize to ensure temperature uniformity.  
The aluminum stand populated with the test assemblies was then introduced to the 
furnace which was set to ramp at 40?C/min to a temperature well above the original 
set point of 275?C.  Following ~212 seconds within the furnace, the aluminum stand 
still populated with the test assemblies was removed to the ambient temperature of 
~24?C.  The resulting temperature data from the aforementioned series of processing 
steps can be viewed in Figure 24.  This trace is the averaged data from the collection 
of eight K-type thermocouples attached to the test assemblies.  Although not 
presented individually for the sake of simplicity, all temperature-time data collected 
from each of the thermocouples was found to be within the recommended 
specifications cited earlier.  As the plot below shows, the average ramp rate of the 
measured SAC305 reflow was approximately 0.9?C/sec while the cool down rate was 
approximately 2.5?C/sec.  An average peak temperature of 245?C was reached during 
a time above reflow of 45 seconds.  During the practical application of the profile, the 
target time above reflow of 60 seconds had to be shortened to 45 seconds due to 
 
 58 
 
considerations regarding furnace operation.  In order to ensure the maximum 
temperature of the solder joints did not exceed the recommended upper bound of 
255?C, the time above reflow needed to be cut back 15 seconds from the prescribed 
one minute because ramp rate was in effect constant during the process.  From room 
to peak temperature, the process took approximately 212 seconds. 
 
 
Figure 24.  SAC305 reflow profile as measured within Omega furnace. 
 
4.2.3.2 Eutectic Au-Sn Reflow 
 
With respect to the eutectic Au-Sn reflow profile, recommendations regarding the 
process were gathered from the related data sheet [40].  A linear ramp rate of 1-
2?C/sec was recommended to a maximum temperature of 310-330?C.  Time above 
liquidus was prescribed as ranging from 45 to 90 seconds above 280?C.  During cool 
down, a rate of no more than 4?C/sec was recommended, as too slow of a cool down 
would prematurely coarsen the microstructure of the solder joint and too fast of a cool 
down would overstress the test assembly by way of thermal shock.  A representation 
of these recommendations as taken from [40] is shown in Figure 25.  
 
 
 59 
 
 
Figure 25.  Manufacturer?s recommended eutectic Au-Sn reflow profile. 
 
Practical implementation of the eutectic Au-Sn reflow profile occurred by preheating 
the Omega furnace to 300?C and allowing temperature stabilization to occur.  
Following sample introduction into the furnace, the machine?s maximum ramp rate of 
40?C/min was programmed to reach a temperature well above the original set point of 
300?C.  After ~274 seconds from introduction to the furnace, the test assemblies were 
removed to the ambient temperature of ~24?C.  The resulting temperature data from 
the eight thermocouples monitored during the process is shown below in Figure 26.  
Again, the plot in the figure below is an average of the aforementioned thermocouple 
data.  As it shows, an average ramp rate of 1.05?C/sec was achieved during reflow 
while the test assemblies experienced a cool down rate of 3.25?C/sec.  The average 
maximum temperature of the eutectic Au-Sn solder joints was 315?C while time 
above liquidus was 55 seconds.  It is clear from this description that all parameter 
ranges recommended by the product documentation were followed in the reflow 
profile developed for use with the Omega furnace.     
 
 
 60 
 
 
Figure 26.  Eutectic Au-Sn profile as measured within Omega furnace. 
 
A summary of the furnace settings required to achieve the SAC305 and eutectic Au-
Sn reflow profiles can be viewed in Table 5.  It is important to note that the final set 
point of both reflow processes was selected in order to be an unachievable value for 
the furnace to attain during the time the samples were inside of the furnace.  This 
ensured that the furnace was continually heating while the samples were inside, 
effectively supplementing the sample ramp rate due to the preheat temperature.  This 
type of scheme was critical in attaining average ramp rates of approximately 
60?C/min (1?C/sec) necessary for proper solder paste reflow. 
 
Table 5.  Omega furnace settings and procedures for proper reflow of SAC305 and eutectic Au-
Sn. 
Solder Preheat Set point Ramp Rate Final Set point Overall time in furnace 
SAC305 275?C 40?C/min 400?C 212 seconds 
Au-Sn 300?C 40?C/min 400?C 274 seconds 
 
4.2.3.3 Ag-In TLPS Temperature Processing 
 
The final of the three pastes manufactured for this investigation required a far more 
elaborate effort in order to create a suitable profile for temperature processing.  As 
previously discussed, the Ag-In TLPS paste used to fabricate the reliability testing 
 
 61 
 
samples was characterized by a composition of 67 wt. % Ag and 33 wt. % In.  This 
paste composition in conjunction with a number of factors including silver particle 
size, processing ramp rate, processing temperature, and hold time was shown to 
influence the remaining wt. % In of the resulting solder joints [8].  Unlike 
traditionally reflowed component attaches, the TLPS paste does not require 
temperatures during processing to exceed its overall melting temperature for 
solidification to occur.  The temperature of the TLPS paste during processing was 
specified in order to melt the In and ensure that all In-rich phases remain liquid until 
diffusion of In has occurred to the point where a new, high melting temperature phase 
was formed from the Ag-In mixture.  Based on the given silver particle size, paste 
composition, and assuming a processing ramp rate of ~1?C/sec, it was determined 
that processing at 250?C for a duration of 90 minutes would be required to eliminate 
all low melting temperature phases.  250?C was selected as the processing 
temperature because it was low enough to be compatible with surface mount 
processing but high enough to permit Ag-In reaction in a reasonable amount of time.  
In this state all In-rich phases would have been diffused into the Ag matrix, thus 
achieving the melting point shift and a new melting temperature of 540?C [30]. 
 
It is important to note that temperatures during processing remain significantly lower 
than the melting point of the Ag particles, thus rendering this a liquid-assisted 
sintering process.  The melting of the low melting point constituents within the TLPS 
paste provides two favorable characteristics over traditional sintering: one, the rates 
of diffusion between liquid and solid during TLPS are much higher than those 
between solid and solid, and two, the liquid within the mixture provides wetting 
 
 62 
 
between Ag particles as well as between Ag particles and the surfaces to be joined.  
Details regarding the phase transformations and stages of diffusion of this TLPS 
material during processing are available in the prior work [8].   
 
As was the case with the SAC305 and eutectic Au-Sn reflow profiles, the temperature 
processing profile developed for the Ag-In TLPS material consisted of a preheat step 
and ended in removal of the test assemblies to ambient for cooling.  In contrast to the 
profile descriptions from earlier, additional effort on the part of the furnace operator 
was required in achieving the desired profile that involved a 1?C/sec ramp to 250?C, 
followed by a hold of 90 minutes prior to removal of the test assemblies from the 
furnace.  Much of this effort was focused on eliminating the significant temperature 
overshoot of the temperature controller when ramping to 250?C.  Prior to introduction 
of the Ag-In TLPS test assemblies, the furnace was stabilized at 275?C.  Immediately 
following insertion of the samples into the furnace the set point was changed to 
400?C at a programmed ramp rate of 40?C/min.  As the sample thermocouples 
approached 235?C, the furnace set point was reprogrammed to 250?C.  This 
reprogramming was performed to ensure that the furnace would maintain the desired 
processing temperature during the extended hold of 90 minutes.  Although the 
chamber controller performs poorly with respect to temperature overshoots resulting 
from long ramp up times, it was shown to be very reliable at maintaining a 
preprogrammed steady-state temperature.  After the samples reached the appropriate 
250?C processing temperature, a sequence of partial openings and closings of the 
furnace door was performed to exhaust the overheat produced by the furnace during 
the ramp up.  The furnace door was consistently opened to a height of ~1.5? (38 mm) 
 
 63 
 
from a fully closed position and returned to a fully closed position during the partial 
opening and closing procedure.  The timing associated with the open/close procedure 
can be viewed in Table 6.  It is important to note that following this ~19 minute 
period, the furnace was completely stabilized at 250?C at which time the furnace door 
remained closed for the remainder of the 90 minute processing profile.    
 
Table 6.  Ag-In TLPS processing open/close furnace door sequence. 
 Time (MIN:SEC)  Time (MIN:SEC) 
Furnace set to 400?C 0:00 Open 6 8:00 
Furnace reset to 250?C 3:00 Close 6 8:30 
Open 1 3:19 Open 7 9:00 
Close 1 5:05 Close 7 9:30 
Open 2 5:25 Open 8 10:00 
Close 2 5:40 Close 8 10:30 
Open 3 6:00 Open 9 12:00 
Close 3 6:15 Close 9 12:30 
Open 4 6:35 Open 10 14:00 
Close 4 7:05 Close 10 14:30 
Open 5 7:20 Open 11 18:30 
Close 5 7:40 Close 11 18:45 
     
The resulting profile generated by the previous operations can be viewed in Figure 
27.  As the temperature processing profile demonstrates, an average ramp rate of 
1?C/sec was achieved while ramping to the processing temperature of 250?C.  
Although some transients can be observed during the initial 19 minute period of the 
open/close sequence, it is clear that they are insignificant with respect to the 
processing temperature.  Removal of the samples from the furnace following the 90 
minute process resulted in a cooling rate of 1.3?C/sec. 
 
 
 64 
 
 
Figure 27.  Ag-In TLPS profile as measured within Omega furnace. 
 
Having discussed the test vehicles for the reliability study in addition to the 
fabrication of test assemblies with each of the three solder pastes it is now appropriate 
to describe the design of the testing procedures implemented during the reliability 
study portion of this investigation. 
 
4.3 Reliability Testing Procedures 
 
One of the aims of the reliability study portion of this investigation was to assess the 
suitability of the new Ag-In TLPS solder paste as a SMT material.  It is not enough to 
simply serve as a connection between two surfaces; the given solder joint must be 
capable of reliably performing its intended functions for the desired amount of time to 
be accepted as a suitable solder material.   Solder joints at the 2nd level of 
interconnection serve three important functions: electrical, mechanical, and thermal 
[15].  With respect to small signal applications associated with many SMT devices, 
the electrical and mechanical functions of the solder joints are paramount in realizing 
a successful package on the circuit board.  Electrically, the solder joint is responsible 
for transferring the signal quickly and without distortion.  With respect to its 
 
 65 
 
mechanical responsibility, the solder joint must be capable of reliably fixing a SMT 
package to a given location on the circuit board under any number of loads that 
include vibration, shock, and bending, among others.  In order to assess the electrical 
and/or mechanical reliability of a given set of solder joints, a number of tests can be 
performed that include thermal cycling, power cycling, vibration, shock, and thermal 
aging to name a few.  For this investigation, passive thermal cycling of the sample 
assemblies was selected as the test to assess their relative reliabilities.  In order to 
accurately assess the reliabilities of the solder materials, an experimental method was 
designed and implemented.  The purpose of this method was to systematically apply 
the stressing conditions provided through passive thermal cycling and monitor the 
subsequent response from the SAC305, eutectic Au-Sn, and Ag-In TLPS test 
assemblies.  Response of the solder joints was monitored through electrical testing, 
pull strength testing, and cross sectional sampling.  Electrical testing was carried out 
through in-situ monitoring while pull strength testing and cross sectional sampling 
occurred at distinct intervals during cycling.  Electrical testing of the experimental 
samples was used to monitor for and identify solder joints failures stemming from 
cracking and fatigue during passive thermal cycling.  Mechanical testing was 
intended as supporting evidence of these failures because significant cracking within 
the joints would subsequently weaken the pull strengths of the samples.  Although 
electrical and mechanical testing of solders is common practice, these methods of 
monitoring simply serve as indicators instead of representations of physical 
phenomena within the solder joint.  For this reason, repeated sampling of solder joint 
cross sections over the duration of cycling was performed in order to understand the 
 
 66 
 
physical changes within the joints that led to the electrical and mechanical behaviors 
that were observed. 
 
4.3.1 Passive Thermal Cycling 
 
During passive thermal cycling, the global CTE mismatch of 5 ppm/?C between the 
polyimide boards and the CQFPs was what caused stress to be placed on the solder 
joints and associated gull wing leads.  For this assessment, two passive thermal 
cycling profiles were implemented in separate Sun Electronic Systems EC12 
chambers.  The Sun EC12 is a nitrogen cooled, bench-top temperature cycling 
machine capable of minimum and maximum temperatures of -184?C and 320?C, 
respectively [43].  Ramp rates up to 15?C/min are possible within the chamber.  For 
simplicity, one of the profiles will be referred to as the high temperature (HT) profile 
while the other will be referred to as the low temperature (LT) profile.  The HT 
profile was designed to subject test assemblies to temperature cycling from ?55 to 
185?C with a 5-minute dwell at -55?C and a 10-minute dwell at 185?C.  This HT 
profile resulted in a ?T of 240?C characterized by a mean temperature, Tmean, of 
65?C.  The LT profile was designed to subject test assemblies to temperature cycling 
from ?55 to 150?C with a 5-minute dwell at -55?C and a 10-minute dwell at 150?C.  
This LT profile resulted in a ?T of 205?C characterized by a Tmean of 47.5?C.  Both of 
these profiles were designed to complete a full cycle every 96 minutes.  Because of 
the difference in the ?T magnitudes between the thermal profiles, the HT profile must 
heat and cool at a rate of ~6?C /min while the LT profile must heat and cool at a rate 
 
 67 
 
of ~5?C /min in order to satisfy the aforementioned cycle period.  A visual 
representation of the LT and HT profiles is illustrated in Figure 28. 
 
 
Figure 28.  LT and HT profiles. 
 
The controller programs implemented in order to carry out the above temperature 
profiles can be found in Table 7 below.  It is important to note that the counting 
scheme of the for loop used in the LT and HT programs operates from 
0,1,2?47,48,49 and immediately stops when I0 = 50.  In this way it properly 
performs the desired 50 instead of what may appear to be 51 cycles. 
 
Table 7.  Sun Electronic Systems EC12 Controller LT and HT Programs. 
LT Controller Program 
 
For I0 = 0,50 
Rate = 5 
Wait = 10 
Set = 150 
Wait = 5 
Set = -55 
Next I0 
End 
HT Controller Program 
 
For I0 = 0,50 
Rate = 6 
Wait = 10 
Set = 185 
Wait = 5 
Set = -55 
Next I0 
End 
 
 
 68 
 
As the above table suggests, passive thermal cycling was run within the EC12 
chambers 50 cycles at a time.  Practical implementation of cycling required this type 
of discrete testing schedule in order to make time for pull testing and microstructure 
sampling.  Each 50-cycle period required 80 hours (3 days 8 hrs) within the chambers 
to complete.  As a result, an average of 50 cycles per week was completed when 
accounting for the time required for testing outside of the chambers.  Experimental 
samples were arranged inside the LT and HT chambers on a vertical, multi-leveled 
rack.  Shelving for the racks was fabricated from aluminum, coarse mesh sheets to 
avoid from disturbing airflow within the chambers and skewing the internal 
temperature distribution. 
 
4.3.2 In-situ Electrical Testing 
 
Failure of the experimental solder joints was based upon their electrical behavior over 
the duration of this assessment.  Electrical behavior of the joints was monitored 
through the collection and analysis of resistance data during cycling.  Out of the three 
monitoring schemes planned for this investigation, electrical testing was the only one 
that could be conducted while the SAC305, eutectic Au-Sn, and Ag-In TLPS samples 
were under test in the temperature cycling chambers.  This was a significant benefit 
over the other planned testing because it meant that the electrical test could capture 
the immediate response of the solder joints and detect a failure if it was in response to 
the temperature extremes experienced during cycling.  As previously mentioned, each 
CQFP featured an electrical chain dedicated to monitoring one of the 8 corner solder 
joints that exist within the package.  In accordance with IPC-SM-785 [44], 32 corner 
 
 69 
 
solder joints were monitored per solder, per strain energy density condition for a total 
384 resistances to be monitored during the experiment.  This sample size 
recommendation was a statistical consideration because approximately 20 to 30 data 
points are required for properly determining the appropriate statistical fit to the failure 
data [45].  Further discussion of the stressing conditions for each of the solder 
materials will occur in the results and discussions portion of this work.  In order to 
generate a data pool of 32 corner solder joints per solder per condition, 4 polyimide 
boards with their associated CQFPs were used to fabricate samples.  Because 3 solder 
materials were the subject of this reliability study, a total of 12 boards were dedicated 
to electrical testing to satisfy the suggested sample size.  Two boards of each solder 
material were cycled within each chamber for a total of 6 electrical boards within the 
LT chamber and another 6 in the HT chamber.  In order to collect the resistance data 
from these boards, edge card connectors capable of withstanding temperatures up to 
200?C were connected to the boards and wired to Agilent 40-channel multiplexer 
modules using the screw terminals provided.  The interconnect wire used during this 
investigation also needed to withstand the temperatures within the HT and LT 
chambers.  In recognition of this, Weico wire no. 2324 with a temperature rating of    
-60 to 200?C was used.  The wired modules were then plugged into Agilent 34970A 
data acquisition switch units that were responsible for monitoring the 384 channels 
and logging the data in a useable format for analysis.  An illustration of this setup can 
be seen in Figure 29. 
 
 
 70 
 
 
Figure 29.  Electrical testing wiring. 
 
In accordance with IPC-9701, a scan interval duration of one minute or less was 
observed while monitoring the resistance of these chains [1].  As per the 
specifications of the Agilent equipment, a 1 mA test current was applied to each 
electrical path during the measurement period.  It is important to note that the 
operation of the 40-channel multiplexer modules ensures that only a single channel is 
connected at any time during the scanning process.  When the measurement of a 
given electrical path is finished, the path is opened and a subsequent path closed for 
measurement.  The failure definition of the solder joints will adhere to the suggested 
definition as provided by [1]: joints will be ruled failures when a 20 percent increase 
in nominal resistance is observed. 
 
4.3.3 Mechanical Pull Testing 
 
In addition to in-situ electrical monitoring of the solder joints, monitoring of the 
mechanical health of the assemblies was performed in order to correlate electrical 
behavior to another quantifiable metric.  In order to accomplish this, a number of 
 
 71 
 
assemblies equal to those used in electrical testing were fabricated for the sole 
purpose of mechanical testing during passive thermal cycling.  Mechanical testing of 
the assemblies was performed through pull testing of the CQFP corner solder joints.  
Pull test samples were fixed in accordance with the published IEC standard for 
surface mount pull testing [46].  The standard stipulates that the plane of the substrate 
must be oriented 45? from the vertical of the pull test hook.  This sample orientation 
is illustrated Figure 30. 
 
 
Figure 30.  45? pull test sample orientation [46]. 
 
Pull tests were conducted using the Dage 4000 series pull tester with a 10 kg load cell 
and test speed of 5 ?m/s.  The test speed selected was the slowest available on the 
Dage 4000 and was chosen in order to maximize the probability that yield of the 
solder joint would occur during testing.  Baseline, as-built pull strengths were 
collected for all three solders prior to passive thermal cycling.  Changes in these pull 
strength values were expected to be indicators of cracking within the solder joints 
tested.  To facilitate removal of the packages for testing, pre-cuts were made in the 
polyimide boards prior to fabrication of the assemblies marked for mechanical 
testing.  As Figure 31 shows, the pre-cut locations within the test boards enabled the 
 
 72 
 
removal of any of the 4 CQFPs by cutting away a small amount of material identified 
by the rings in the image.  Implementation of the pre-cuts minimized required cutting 
when the packages needed to be removed for testing while still allowing the 
mechanical boards to be printed and processed in the same way as the electrical 
boards.  
 
 
Figure 31.  Pre-cut locations on polyimide test board. 
 
4.3.4 Cross Sectional Sampling 
 
The last of the three methods of monitoring the solder joints under test was cross 
sectional sampling.  As previously mentioned, observation of the solder joint 
microstructure allows one to determine what is physically occurring to the sample, 
which can then be tied to changes in indicators such as joint resistance or pull 
strength.  For this investigation baseline cross sectional images were collected from 
the corner joints of SAC305, eutectic Au-Sn and Ag-In TLPS samples.  Following a 
desired number of temperature cycles, microstructures were again collected from 
these solders for comparison to the baseline.  All stressing conditions were accounted 
for during this sampling as cross-sections from large and small CQFP packages were 
prepared each time data was gathered.  Standard sample preparation using the 
 
 73 
 
following progression in grinding paper grit was observed: 240? 400? 600? 
800? 1200.  When required, 1 micron alumina powder was used to polish the sample 
face.  All samples were imaged using the Environmental Scanning Electron 
Microscope (ESEM) or light microscope. 
 
 74 
 
Chapter 5: Reliability Assessment Results and Discussion 
 
In order to assess the reliability of the solder materials during passive thermal cycling, 
the mechanical, electrical, and cross sectional data collected throughout this 
investigation were analyzed.  Analysis of the data not only focused on trends 
observed within the particular sampling method, but how they related to the trends of 
other methods as well.  For example, changes in microstructure during this 
investigation were closely tied to those in mechanical pull strength and vice versa.  
Discussion of the resulting data is divided among the solder joint sampling methods, 
however, relationships between the trends are identified when necessary.  The corner 
solder joints of surface mount packages are expected to be the first to fail because 
they are subject to the highest strain energy densities that result from their location 
with respect to the neutral axis of the surface mount package.  Mechanical, electrical, 
and cross sectional sampling were targeted for monitoring these corner solder joints.  
The combination of two temperature profiles with two package sizes ensured that 4 
distinct solder strain energy densities existed within the various corner solder joints of 
the test assemblies.  As a result, it is expected that the electrical, mechanical, and 
microstructural trends observed during this investigation will reflect this staggered 
damage scheme.  In order to quantify these strain energy densities, material and 
geometry data was collected from the assemblies.  These values were used as inputs 
in an equation for calculating the strain energy density of surface mount solder joints.  
The form of this equation is provided in the subsequent discussion along with the 
collection and estimation of the necessary parameters.  Following this strain energy 
 
 75 
 
density estimation, the results of the electrical, mechanical, and cross sectional 
sampling methods will be presented. 
 
5.1 Solder Joint Strain Energy Density Estimation 
 
As previously mentioned, when the test assemblies are subjected to one of the two 
selected thermal profiles, damage to the joints will occur as a result of the global CTE 
mismatch between the component and the substrate.  The corner solder joints are said 
to experience the largest overall deformation or damage because they lie farthest from 
the package?s neutral axis or center.  The imposed cyclic damage within these joints 
can be quantified as the strain energy density, which is represented by the area within 
the shear stress-shear strain hysteresis loop.  For surface mount, leaded components, 
the maximum cyclic strain energy density (?W) experienced by the solder joint can 
be estimated [47] by 
 
Ah
TLKW d ?????? 2)( ? .                     (5) 
 
The variables from the equation are as follows: Kd ? diagonal lead stiffness, A ? 
minimum load bearing area of the solder joint, h ? characteristic solder joint 
thickness, L ? distance from package?s neutral axis to the center of the corner joint, 
?? ? global CTE mismatch between component and substrate, and ?T ? magnitude of 
temperature change during thermal cycling.  From this list it is clear that many of 
these variables are easily measured from the characteristics of the test assembly (A, h, 
L, ??) or have been predefined by the passive thermal cycling test set-up (?T).  The 
 
 76 
 
diagonal lead stiffness, Kd, is the one variable that requires more sophisticated 
estimation techniques to provide an input value for calculating ?W.   
 
5.1.1 Diagonal Lead Stiffness 
 
As demonstrated above, the diagonal lead stiffness is a value that is critical in 
estimating the strain energy density experienced by solder joints during passive 
thermal cycling.  The parameter itself is exclusively dependent upon the material and 
geometry of the surface mount lead and therefore completely independent of any 
material or geometrical considerations of the solder joint itself.  Leads that are 
characterized as more compliant feature lower lead stiffness values while those that 
are less compliant have higher stiffness values.  Lead stiffness of a surface mount 
package is an important characteristic that must balance two considerations.  On the 
one hand, the surface mount leads must be compliant enough to absorb a significant 
share of the strain energy during cycling to protect the less compliant solder joints 
from early fatigue, while on the other, they must be stiff enough to tolerate the 
manufacturing process and maintain their form to ensure consistent solder joint 
formation.  Computation of the stiffness values of a number of surface mount lead 
designs has been performed in prior works [48][49][50].  The generalized equations 
developed within these papers are extremely useful as a means of comparing the 
relative compliances of differing lead geometries in design for reliability exercises.  
For this particular investigation the equations developed for the inclined gull wing 
configuration described in [49] were used to estimate the diagonal lead stiffness of 
the corner leads on the 64-lead and 44-lead CQFP packages.  All equations developed 
 
 77 
 
within the aforementioned papers were based on classical elastic strain energy 
methods.  The elastic loading assumption is one commonly made for lead frame 
materials whereas non-elastic assumptions regarding solder materials during passive 
thermal cycling are often required for accurate modeling of joint behavior.  
According to the authors, independent computation of lead stiffness values for the 
various surface mount lead designs in [48][49][50] was conducted using finite 
element structural modeling for validation purposes.  With respect to the general 
equations developed in [49], six degrees of freedom that included translation along as 
well as rotation about the x, y, and z-axis were assumed to govern motion of the 
leads.  For each of these possible motions, the lead was assumed to feature an 
independent flexural spring constant responsible for governing how it would respond.  
Due to limitations of the loading experienced during passive thermal cycling, the 
authors identified translation along the x and y axes, which together form a plane 
parallel to the surface of the circuit board, as the expected modes of lead motion 
during testing.  It is important to note that such an assumption implies that negligible 
substrate warping occurs during passive thermal cycling of the test assemblies.  The 
simplification regarding the expected loading during passive thermal cycling allowed 
the authors to compute two (of the six) flexural spring constants, Kx and Ky, to 
determine the operative stiffness of the corner gull wing lead.  For this particular 
investigation, the equations to calculate Kx and Ky from [49] were implemented in 
MATLAB for the 44-lead and 64-lead CQFP packages and can be found in Appendix 
A.  As the code shows, there are a number of geometry parameters that must be 
quantified in order to estimate the lead spring constants in the x and y directions.  
 
 78 
 
Reproduction of spring constant values provided in [49] for a particular inclined gull-
wing configuration served as a check of the aforementioned code.  An illustration 
[49] demonstrates where these values are assessed on the inclined gull-wing lead. 
 
 
Figure 32.  Structural model for inclined gull-wing [49]. 
 
As Figure 32 shows, the length values include L1, L2, L3, and L4 while R1 and R2 
quantify the radius of curvature at the bends within the gull-wing structure.  The 
angle, ?, is the value, in radians, that defines the incline of the lead from the vertical.  
If there is a transition in the cross section of the lead, this location can be described in 
the model with point T, having a cross-section defined by l2 and w2 above this point 
and l1 and w1 below this point.  For this particular investigation, there was no such 
transition within the cross section of the lead.  For this reason, the value for L1 was 
defined as the entire length between the two radii of curvature and L2 was set to zero.  
Additionally, this lack of transition meant that l2 = l1 and w2 = w1 throughout the 
entire lead length.  It is important to note that all previously mentioned geometries 
from the inclined gull-wing model are independent of one another.  The last 
geometric variable defined in Figure 32 is that of the overall lead height, H.  This 
 
 79 
 
particular variable a function of L1, L2, R1, R2, and ? and can therefore be computed 
from these values. 
 
In order to quantify the average angle of inclination for each of the packages, 16 64-
lead CQFP and 16 44-lead CQFP lead samples were measured.  As was the case in 
the densification study, ImageJ was used to make measurements necessary in 
determining the inclination angle ?.  In order to facilitate discussion of the lead angle 
measurements, Figure 33 has been provided for reference.   
 
 
Figure 33.  Lead angle definitions. 
 
Within the ImageJ software there is an angle tool available for performing such 
measurements.  The tool allows for the overlay of two vectors on top of the selected 
image.  To ensure the most accurate angle measurements, one vector was placed 
along the centerline of the lead leg between the two radii of curvature and the other 
placed along the horizontal provided by the surface mount pad of the polyimide 
board.  This angle was measured in favor of directly measuring ? because there were 
 
 80 
 
more visual cues afforded in measuring ?.  Using trigonometric relations, the lead 
inclination angle, ?, could be related to the angle measured in ImageJ, ?, through 
 
?90?? ?? .                     (6) 
 
Using the 16 measurements of ? for each of the package sizes an average inclination 
angle was computed for leads of the 64-lead and 44-lead CQFPs.  Independent 
averages were computed for the two package sizes because although the same tool 
was used to bend the leads to the geometries specified in Figure 16, the differing lead 
cross sections of the packages significantly influenced the final lead geometry 
following the trim and form process.  As Table 8 demonstrates, there is a noticeable 
difference between the average inclination angles of the two different surface mount 
packages.  Sample standard deviations, s, on the mean values have been provided for 
convenience. 
 
Table 8.  Average lead angles from 44-lead and 64-lead CQFPs. 
 ? (degree) ? (degree) s (degree) ? (radian) s (radian) 
44-lead CQFP 130.6 40.6 2.7 .71 0.048 
64-lead CQFP 125.5 35.5 1.8 .62 0.03 
 
These computed averages were then used in subsequent geometry calculations and 
ultimately for the estimation of the diagonal stiffness calculations. 
 
With the average lead inclination angle of the two package sizes, further calculation 
of the lead geometry could be performed.  The parameters, R1, R2, L1, L2, L3, and 
L4 were then the next to be quantified.  In order to measure these variables of the 
inclined gull-wing structural model, simple geometric forms were overlaid on cross 
section images to trace the centerline of the gull-wing leads; the radii of curvature 
 
 81 
 
were estimated using perfect circles while the linear variables were estimated using 
straight lines.  The size of these geometric forms therefore served as estimations of 
the aforementioned parameters.  A representative overlay of these shapes on a given 
cross section can be viewed in Figure 34. 
 
 
Figure 34.  Geometric overlay of cross section images to estimate gull-wing structure. 
 
Because of the cross section preparation process, the length associated with L3 was 
consistently destroyed by the band saw during removal of the leads from their 
respective packages.  This characteristic of the cross section samples is clear from the 
representative image provided in Figure 34.  In an effort to quantify this length, a 
measurement of the average cut width from the diamond band saw was performed.  
This measurement was then used in place of a direct measurement of L3.  Due to the 
time intensive nature of the overlay estimation method, a total of 8 cross sections 
were analyzed between the two package sizes.  The resulting averages computed from 
these samples are displayed, by package size, in Table 9 below. 
 
 
 82 
 
Table 9.  Estimated parameters of the inclined gull-wing structural model. 
 R1 (mm) R2 (mm) L1 (mm) L2 (mm) L3 (mm) L4 (mm) 
44-lead CQFP 1.20 1.20 1.75 0 0.9 0.61 
64-lead CQFP 1.15 1.15 1.70 0 0.9 0.52 
 
As Table 9 shows, the radii of curvature for each package size were assumed to be 
equal due to observations made during the initial investigation regarding estimation 
of the above parameters.  The overlay in Figure 34 serves to validate this assumption, 
as R1 = R2, while providing a good estimation to the centerline of the lead geometry.  
In order to check the averages provided in Table 9, the total curvilinear length 
provided by the estimates was calculated using  
 
)5708.1)(( 214321 ???????? RRLLLLC ,                     (7) 
 
from Kotlowitz and Taylor [49], where ? is in radians.  This equation can be easily 
established through inspection of the structural model in Figure 32 and 
implementation of trigonometric relations.  The curvilinear lengths calculated using 
the above estimates were shown to be within 5 percent of the total 220 mil (5.6 mm) 
length denoted by the original gull-wing specification as shown in Figure 16.   
 
Having determined values for all the parameters requiring estimation, the final values 
required for the code in Appendix A were those of the lead cross sections and 
material properties of the leads themselves.  This information was readily available 
through specifications and drawings provided by the CQFP manufacturer.  Leads of 
the 64-lead CQFP featured cross sectional lengths of 18 mil (0.46 mm) and widths of 
8 mil (0.2 mm) while those of the 44-lead CQFP featured cross sectional lengths of 
10 mil (0.25 mm) and widths of 11 mil (0.28 mm).  The orientation of these 
measurements can be referenced to the gull-wing structural model provided in Figure 
 
 83 
 
32.  As mentioned previously in the experimental procedures section of this work, the 
leads of the CQFP packages were composed of alloy 42 with the appropriate surface 
coatings of nickel and then gold.  From Brown et. al. [15], the elastic modulus of 
alloy 42 is 20.6*106 psi (142 GPa).  In order to estimate the shear modulus of the 
material using  
 
)1(2 ??? GE                      (8) 
 
from Timoshenko [51], Poisson?s ratio, ?, of the material was necessary.  This ratio 
was estimated as ~0.3 from the properties of Ni and Fe [52] using a simple rule of 
mixtures.  This value in conjunction with E = 20.6*106 psi (142 GPa) yielded a shear 
modulus, G, of 7.92 *106 psi (54.6 GPa) using equation (8). 
 
Having provided all necessary inputs to the generalized equations for computing Kx 
and Ky of the CQFP leads, the final estimation of these spring constants could be 
made.  With these values for Kx and Ky, it was then possible to compute the diagonal 
lead stiffness of the corner leads using  
 
221
414.1)(
yx
yx
rd
KK
KKK
?
??  from Kotlowitz [48].                     (9) 
 
The above equation for Kd is given for a square package featuring an aspect ratio of r 
= 1 only.  It results from a simplification of other equations [48] suitable for 
rectangular packages with aspect ratios other than unity.  The 44-lead and 64-lead 
CQFPs were both square packages, with side lengths of 0.65 inches (16.5 mm) and 
0.9 inches (22.9 mm), respectively.  The resulting Kd for both package sizes along 
with Kx and Ky are displayed below in Table 10. 
 
 84 
 
 
Table 10.  Inclined gull-wing lead spring constants. 
 Kx (lb/in) Ky (lb/in) Kd (lb/in) Kd (N/m) 
44-lead CQFP 34.10 6.55 9.10 1595 
64-lead CQFP 20.15 18.18 19.09 3345 
 
The trend of the results above is expected for a number of reasons.  With respect to 
the Kx values, the leads of the 44-lead CQFP should be, and are, stiffer than those of 
the 64-lead CQFP in this direction for two major reasons.  Earlier mention was made 
of the larger lead inclination angle of the 44-lead CQFP when compared to that of the 
64-lead CQFP.  As the inclination angle of the lead is increased, the Kx and Ky spring 
constants increase accordingly.  The gull-wing model provided in [49] predicts this 
stiffness value behavior with respect to inclination angle.  In addition to a higher lead 
inclination angle, the leads of the 44-lead CQFP feature a larger cross sectional width, 
11 mil (0.28 mm), than those of the 64-lead CQFPs, 8 mil (0.2 mm).  This translates 
into a higher moment of inertia that resists bending caused by forces imparted to the 
lead along the x-axis.  Table 10 reflects this reasoning as Kx for leads of the 44-lead 
CQFP is close to double that of leads from the alternative package.  With respect to 
those values of Ky, the difference can be attributed to the significant disparity 
between packages in the material available to resist bending when the lead is acted 
upon by a force along the y-axis.  As mentioned earlier, the leads of the 64-lead 
CQFP were 18 mil (0.46 mm) in cross sectional length while those of the 44-lead 
CQFP were only 10 mil (0.25 mm).  It is because of this difference in cross sectional 
lengths that the Ky spring constant for leads of the 64-lead CQFP is close to triple that 
of leads from the alternative package.  Sensitivity analysis on the x and y-direction 
stiffness values demonstrated that the cross sectional geometries of the package leads 
 
 85 
 
were significantly more influential than the given difference in lead inclination 
angles.  This behavior provides an explanation as to why the value of Ky for the 64- 
lead CQFP is approximately three times that of the 44-lead CQFP while consideration 
of the lead inclination angles would suggest the relationship should be otherwise.  
 
As a check on how reasonable the results from Table 10 are, x and y direction lead 
spring constants of commercially available packages were sought out and located 
[53].  Leads on common Intel CQFP packages were cited as featuring Kx and Ky 
values of 9.05 lb/in (1585 N/m) and 23.0 lb/in (4030 N/m), respectively, while plastic 
quadflat packages (PQFP) featured Kx and Ky values of 39.8 lb/in (6970 N/m) and 
85.9 lb/in (15050 N/m), respectively.  Although the specific geometries of the leads 
on these packages are not known, it is beneficial to know that common surface mount 
packages feature leads with directional spring constants comparable to those 
calculated for this investigation. 
 
5.1.2 Diagonal Length, CTE Mismatch, and ?T 
 
Having appropriately estimated the diagonal lead stiffness for leads of the 44 and 64-
lead CQFPs, there is one remaining term in the numerator of the strain energy density 
calculation that requires quantifying.  This term, (L???T)2, is defined by the half 
component diagonal, L, the global CTE mismatch between the polyimide board and 
ceramic body of the CQFP, ??, and the temperature swing, ?T.  As defined by the 
specifications and drawings of the CQFPs, the 64-lead CQFP features a square body 
0.9 inches (22.9 mm) along a side while the 44-lead CQFP features a square body 
0.65 inches (16.5 mm) along a side.  These dimensions translate into half component 
 
 86 
 
diagonals of L64-lead CQFP = 0.64 inches (16.25 mm) and L44-lead CQFP = 0.46 inches 
(11.7 mm).  With respect to the global mismatch, this variable has already been 
defined many times as 5 ppm/?C (?polyimide = 12 ppm/?C, ?alumina = 7 ppm/?C).  The 
temperature swings have also been previously defined and identified by the HT cycle, 
where ?T = 240?C, and the LT cycle, where ?T = 205?C.  With the entire numerator 
of the strain energy density equation defined and quantified, it is now appropriate to 
complete the estimation by quantifying the denominator. 
 
5.1.3 Minimum Load Bearing Area and Characteristic Solder Joint Height 
 
As defined [47], the minimum load bearing area of the solder joint, A, can be 
quantified as two-thirds the solder wetted lead area projected to the surface mount 
pad.  The characteristic solder joint height, h, is typically defined as half of the solder 
paste stencil thickness, but because of the available experimental samples and ease of 
estimation using the ImageJ software, it was decided that this height should be 
gathered from the solder cross sections as well.  In order to estimate h, an average 
solder joint height was calculated from the given samples by computing the cross 
sectional solder joint area directly beneath the wetted lead length and dividing it by 
the wetted lead length projected to the surface mount pad.  This calculation therefore 
yielded an estimate of the average solder joint height, h, which would provide an 
equivalent cross sectional area when multiplied by the wetted lead length projected to 
the surface mount pad.  An illustration of these values can be seen below in Figure 
35. 
 
 
 87 
 
 
Figure 35.  Measurement definitions in estimating characteristic solder joint height. 
 
In order to estimate the minimum load bearing area of the solder joint, the wetted lead 
lengths projected to the surface mount pad were multiplied by the cross sectional lead 
lengths (identified in Figure 32) and finally by a factor of 2/3 as called for in the 
definition.  As previously discussed during the estimation of the lead diagonal 
stiffness, these cross sectional lead lengths were defined as 18 mil (0.46 mm) for the 
64-lead CQFP and 10 mil (0.25 mm) for the 44-lead CQFP from the available 
manufacturer specifications.  For the eutectic Au-Sn and SAC305 samples printed 
with the 8-mil (200 ?m) thick stencil, averages for the characteristic solder joint 
height and minimum load bearing area were estimated from a total of 16 samples for 
each component size.  The Ag-In TLPS samples fabricated from printing with a 6-mil 
(150 ?m) thick stencil were initially assessed differently due to the fact that a smaller 
characteristic solder joint height was expected due to the thinner stencil, however, it 
was demonstrated that the average height and minimum load bearing areas were very 
similar to those computed for Au-Sn and SAC305.  As a result, the average solder 
joint height and minimum load bearing area for each of the component sizes were 
computed independent of the solder joint materials for ease of discussion in 
 
 88 
 
subsequent sections.  The values resulting from these computations can be found in 
Table 11.  
 
Table 11.  Characteristic solder joint heights and minimum load bearing areas. 
 h (mil) h (?m) A (sq. mil) A (?m2) 
44-lead CQFP 4.0 101.6 368 237420 
64-lead CQFP 4.6 116.8 671 432900 
 
With respect to the values shown above, it is clear that the estimate of half of the 
stencil thickness serves as a good indicator of the characteristic solder joint height 
when traditional solders are used.  In the case of the Ag-In TLPS material where full 
melting does not occur, blindly using such an estimate would have predicted a solder 
joint height of ~3 mil (76 ?m) while the true characteristic solder joint heights were 
as much as 1 to 1.5 mil (25 to 38 ?m) greater than this value.  This behavior is likely 
a result of the porosity of the joints.  Having quantified all values associated with the 
estimation of the strain energy density within a surface mount solder joint, the various 
strain energy densities resulting from the different package size / temperature swing 
interactions can be computed. 
 
5.1.4 Strain Energy Densities 
 
By design, this investigation was intended to feature four distinct strain energy 
densities that could be used to quantify the difference in damages experienced by the 
surface mount solder joints during passive thermal cycling.  The successful estimation 
of all parameters related to strain energy density through equation (5), as previously 
described, allowed for computation of these distinct strain energy densities as a 
 
 89 
 
function of package size and cyclic temperature swing.  These values are provided in 
Table 12 and will be used to simplify subsequent reliability testing related discussion.  
 
Table 12.  Strain energy densities. 
  ?W (lb/in2) ?W (N/m2) 
44-lead CQFP LT 1.4 9650 
 HT 1.9 13100 
64-lead CQFP LT 2.7 18615 
 HT 3.6 24820 
 
For convenience, the strains, ??, experienced by the solder joints under test were 
estimated as well.  These strain estimations are inherent within the calculation of the 
strain energy densities from equation (5) as ?W = S??, where S represents the 
maximum stress and ?? represents the strain range within the solder joints during 
passive thermal cycling [47].  The strain range estimations are provided in Table 13. 
 
Table 13.  Strain range estimations. 
  ?? 
44-lead CQFP LT 0.119 
 HT 0.140 
64-lead CQFP LT 0.142 
 HT 0.167 
 
5.2 Electrical Data Trends and Discussion 
 
As was discussed in the experimental procedure section of this work, a total of 384 
chain resistances were monitored during passive thermal cycling.  Following every 50 
cycles, the data files generated from the resistance readings were removed from the 
computer linked to the data logger array.  At a sampling rate of one reading every 
minute, each 50 cycle duration generated ~1.85 million resistance values across all 
samples.  In order to manage such a vast amount of data, a number of MATLAB 
 
 90 
 
programs were created to aid in the tagging and analysis process.  These codes can be 
found in Appendix B and will be discussed later. 
 
From the baseline resistance testing conducted prior to passive thermal cycling, it was 
observed that the resistances of the surface mount solder joints were negligible when 
compared to the resistances of the daisy chains of which they were a part.  This 
observation held true for the SAC305, eutectic Au-Sn, and Ag-In TLPS test 
assemblies.  A significant fraction of the monitored resistance was attributed to the 
inclined gull wings and internal wirebonds connecting these leads together in 
packages of both sizes.  In order to confirm this, unsoldered packages were sampled 
for the resistance of electrically connected lead pairs.  The schematic of which lead 
pairs were electrically connected to one another can be observed in the wirebond 
schemes provided in Figure 14 and Figure 15.  With respect to the 64-lead CQFPs, 
this lead pair resistance was found to be ~0.33 ? having removed all contributions 
from contact resistance.  In the 64-lead CQFPs, the daisy-chain pattern designed for 
this investigation featured four of these lead pairs per chain monitoring the solder 
joints on each corner.  As a result, the total resistance contribution from the 
electrically connected lead pairs was estimated to be ~1.32 ?.  When baseline data 
collected from 64-lead CQFP daisy chains were corrected to eliminate resistance due 
to PWB trace lengths, connector contact resistance, and interconnect wire lengths the 
average resistance value of just the joints plus the lead pair resistances was found to 
be ~1.36 ?.  Correction values were generated from direct resistance measurements, 
resistance estimations from measured trace lengths, and manufacturer provided data.  
These correction values were subtracted from the raw resistance data to arrive at the 
 
 91 
 
corrected resistance values.  It is clear that the resistances of the gull wing leads and 
wirebonds within the 64-lead CQFP heavily influence the corrected daisy chain 
resistance of 1.36 ? as they account for close to 97% of this value.  Similarly large 
resistance fractions were noted for 44-lead CQFPs as well.  Corrected daisy chain 
resistances for the 44-lead CQFPs were found to be ~0.83 and ~0.6 ? for those chains 
featuring three and two electrically connected lead pairs, respectively.  The total 
resistance contributions from the lead pairs in the 44-lead CQFPs were measured as 
~0.6 and ~0.4 ? for those chains featuring three and two lead pairs for an average 
contribution of 70% of the corrected resistance values.  Although these corrected 
values were useful in identifying the fraction of the resistance contributed from 
sources other than the solder joints themselves, the raw resistance values were used as 
the nominal values for analysis and failure definition instead.  This decision was 
made following observations of the raw data subjected to passive thermal cycling.  As 
one would expect, it was shown that the temperature within the chambers 
significantly affected the raw resistances of the monitored daisy chains.  Because the 
aforementioned correction values could not be independently characterized as a 
function of temperature, it was decided that use of the raw data would be less prone to 
error and therefore a more accurate metric for defining failure.  A discussion of the 
observed raw resistance temperature dependence and its impact on the assumed 
failure definition will follow.  
 
Prior to passive thermal cycling of the test assemblies, the experimental procedure 
called for a failure definition of a 20 percent increase in the nominal resistance of the 
monitored chain, as given in [1].  However, data collected from initial cycling 
 
 92 
 
revealed that the raw resistance of daisy chains surpassed this criterion during every 
heating stage of both thermal profiles.  An example of this temperature dependent 
resistance behavior is displayed for a 64-lead CQFP, Au-Sn chain in Figure 36.  As 
the temperature trend on this plot shows, the chain was subjected to the LT profile.  It 
is clear from the resistance trend that the 20% threshold is easily surpassed during a 
typical LT cycle.  In fact, the maximum chain resistance was found to be ~40% 
higher than the nominal resistance registered at room temperature. 
 
 
Figure 36.  Au-Sn chain resistance and temperature as a function of passive thermal cycling 
time. 
 
All test assemblies monitored for raw chain resistance were observed to surpass this 
prior established threshold for defining electrical failure.  For all solder materials, 
those chains of 64-lead CQFPs were found to have maximum resistances of ~70% 
and 40% above their nominal values when subjected to HT and LT profiles, 
respectively.  Those chains of 44-lead CQFPs were found to have maximum 
 
 93 
 
resistances of ~55% and 40% above their nominal values when subjected to HT and 
LT profiles, respectively.  Because of this behavior, it is necessary to consider an 
alternative definition of electrical failure.  Although industry standards [1] provide 
useful information for setting up and conducting testing, it is important to always 
consider the behavior of the samples under test prior to accepting any generic failure 
definitions provided by the literature.  Similar conclusions regarding the selection and 
acceptance of failure criteria from accepted industry standards without consideration 
of sample characteristics were reflected in other works [54].  In order to ensure that 
resistance signals interpreted as electrical failures were indeed indicative of some 
type of crack or delamination within the solder joint, redefinition of failure was 
reserved until cross section images could be used to confirm out-of-the-ordinary 
resistance behavior. 
 
In an effort to characterize the signatures of healthy solder joints to use as reference 
for subsequent failure definition, plots were generated from data gathered during the 
first 50 thermal cycles for all materials and strain energy densities.  As previously 
discussed, the large amount of data collected on a weekly basis required a minimum 
level of automation in order for the data analysis to be manageable.  In order to 
automate the plotting of chain resistance data, a MATLAB program was created.  
This automatic plotting program can be found in Appendix B.  Prior to execution, the 
program required the user to specify the location of the data to be plotted and to what 
destination the generated plots were to be saved.  After this, the program would then 
automatically plot and save the graphs in a series of labeled bitmap files.  These plots 
could then be observed in order to identify trends in the resistance behavior.  At the 
 
 94 
 
beginning of experimentation, these plots served as the baseline resistance signatures 
of healthy test assemblies against which subsequent signatures would be compared.  
A sample plot generated from the program for a 44-lead CQFP, SAC305 daisy chain 
can be viewed in Figure 37 below.  This particular chain from the plot below was 
subjected to its first 50 HT cycles and therefore representative of the chain when all 
joints were healthy and relatively undamaged. 
 
 
 
Figure 37.  Representative baseline resistance signature of a 44-lead CQFP, SAC305 daisy chain. 
 
As Figure 37 shows, the trend of the chain resistance is highly repeatable as one 
might expect.  The cyclic variations are, of course, directly related to the change in 
temperature during passive thermal cycling.  The correlation of chain resistance to 
sample temperature is clearly displayed in Figure 36.  For this particular chain, the 
resistance varied from ~2 ohms at 185?C to ~.95 ohms at -55?C.  Although other 
chains did not adhere to these absolute limits, the forms of their automatically 
generated resistance plots were very similar to that displayed above.  Those chain 
 
 95 
 
resistance plots that varied in any way from their established baselines were 
subsequently investigated as potential corner joint failures. 
 
Over the course of passive thermal cycling, a number of chain resistance signatures 
began showing signs of what were initially interpreted as corner solder joint failures.  
Generally speaking, these ?failed joint? signatures were spread evenly among the 
solder materials used to fabricate the test assemblies.  The abnormal resistance plots 
were first recognized in the population of SAC305 test assemblies around 250 to 300 
passive thermal cycles.  Many times small instabilities were evident in the early 
stages of cycling that subsequently grew into larger inconsistencies with an increasing 
number of thermal cycles.  This progressive deterioration of some of the chain 
resistance trends suggested the gradual weakening and eventual destruction of the 
surface mount solder joints in question.  With respect to the SAC305 test assemblies, 
cross sections of selected samples at 400 passive thermal cycles seemed to confirm 
the possible electrical deterioration as significant cracking was observed in corner 
joints of both package sizes subjected to HT cycling.  Although this offered a 
convincing correlation, investigation of the other materials within the study soon 
called this agreement into doubt.  During this investigation, samples fabricated from 
eutectic Au-Sn were found to be extremely reliable under the environmental 
conditions as cracks were not detected in any of the cross sections prepared nor did 
the mechanical strength data collected suggest significant degradation of the corner 
solder joint.  However, some Au-Sn chain resistance signatures were found to 
degrade in a fashion similar to those from the SAC305 and Ag-In TLPS populations.  
 
 96 
 
A chain resistance representative of the degradation observed across samples can be 
seen below for a 64-lead CQFP, Au-Sn chain subjected to HT cycling. 
 
 
Figure 38.  Chain resistance degradation as a function of passive thermal cycles. 
 
As Figure 38 shows, by 500 thermal cycles the daisy chain characterized by these 
signatures looks to have changed in a way that has caused the 500 to 550 cycles plot 
to become significantly more irregular than that of the baseline.  At various points in 
the latter of the two plots, the chain resistance more than doubles the resistance 
readings of the baseline.  As previously mentioned, all mechanical and cross sectional 
data collected for the Au-Sn samples showed no indication of joint failures during 
cycling.  This discord between the electrical, mechanical, and cross sectional data 
 
 97 
 
prompted an investigation into the causes behind the degrading electrical behavior of 
some chains in the Au-Sn, SAC305, and Ag-In TLPS test assemblies. 
 
Due to the nature of the electrical testing performed during this investigation there 
existed a number of failure points within the wiring route that might explain why the 
chain resistance signature behaved as it did for the Au-Sn daisy chain displayed in 
Figure 38.  As previously mentioned, in order for the data logger array to monitor the 
chains within the test assemblies, a continuous path from the positive to negative 
terminal of the multimeter within the data logger was required.  This continuous path 
was described in the experimental procedure section above and illustrated in Figure 
29 from earlier.  Figure 39 below highlights a total of 7 possible locations in the 
wiring scheme where the electrical signal could be affected by degradation or faulty 
connections.  These locations include the following:  
 
1. the solder joints within the daisy chain of the surface mount package,  
2. the PWB traces running to and from the package daisy chain,  
3. the press-fit connection at the edge card connector / PWB interface,  
4. the solder joint connecting the hook-up wire to the edge card connector,  
5. the hook-up wire itself,  
6. the screw terminal between the hook-up wire and multiplexer module, and 
7. the press-fit connection at the multiplexer module / data logger interface. 
   
 
 98 
 
 
Figure 39.  Potential failure sites within electrical wiring. 
 
In order to determine if any of these possible failure sites was causing the 
inconsistencies, a diagnostic approach was taken.  As a preliminary measure, those 
chains that exhibited the erratic chain resistance signature displayed in Figure 38 were 
tested at room temperature with a hand-held ohmmeter.  When all chains in question 
were checked, the probe points of contact were shifted throughout the remaining 
wiring levels to determine whether any additional resistances above the baseline 
signatures could be detected.  This room temperature probing failed to confirm any 
electrical issues at the potential sites already identified, however, this result was not 
unexpected as many times electrical failures during passive thermal cycling occur 
only during periods of actively changing temperature.  In an effort to identify the 
failure site location under passive thermal cycling, dummy PWBs were fabricated 
from 62 mil (1.6 mm) thick aluminum cut to the appropriate size for insertion into all 
edge card connectors under test.  These aluminum inserts provided the means to test 
potential failure sites 3 through 7 when subjected to passive thermal cycling.  By 
 
 99 
 
moving to this step in the diagnostic process it was assumed that failures were not 
located within the surface mount solder joints or any connective wiring on the PWB.  
This assumption was made because of the evidence provided by the healthy eutectic 
Au-Sn samples in addition to the fact that testing these sites under passive thermal 
cycling without the data logger array would have required an additional monitoring 
set-up.  Provided that failure could not be detected with this new diagnostic set-up, it 
would have suggested that the boards or solder joints were, in fact, the failure sites of 
interest. 
 
The chain resistance signature provided in Figure 40 is from the same Au-Sn chain as 
discussed in Figure 38, save for the fact that the dummy aluminum PWB was inserted 
in place of the true PWB.  It is clear from the image that this waveform is irregular as 
well.  In contrast to this behavior, those chains that exhibited predictable, cyclic 
behavior as shown in Figure 37 continued to behave as such when the dummy 
aluminum PWBs were tested.  When the data collected from passive thermal cycling 
with the Al inserts was cross-checked with the earlier sample data, it was observed 
that every degrading electrical signal was again observed in the dummy tests.  This 
result suggested that failures existed between the data logger array and edge card 
connectors and were not a function of damage in any surface mount solder joints.    
 
 
 100 
 
 
Figure 40.  Resistance trend of a chain completed by an aluminum insert. 
 
In response to the feedback indicating that some type of failure between the edge card 
connector and data logger array was responsible for the erratic behavior, repairs were 
immediately made to solder joints connecting the hook-up wire to the edge card 
connector.  These repairs were performed by applying additional high-lead solder to 
each of the eyelets connected to channels on the multiplexer module by way of hook-
up wire.  Additionally, hook-up wires were inspected for cuts or imperfections caused 
by temperature cycling and screw terminals of the multiplexer modules were 
tightened when necessary to ensure ideal electrical contact.  Spring contacts of the 
edge card connectors were also checked to ensure that stress relief of these parts was 
not causing the irregular signatures.  Subsequent testing of the connectors was again 
conducted with the Al inserts to check if the repairs had corrected the erratic 
resistance signatures.  Following this second session of testing with the dummy 
aluminum PWBs, the erratic resistance signatures were still observed, therefore 
suggesting that the source of these variations came from the multiplexer modules 
 
 101 
 
themselves or their connections to the data logger array.  Final adjustments were 
made by switching connectors among multiplexer modules to observe whether the 
erratic chains traveled with the connectors or whether they were characteristic of the 
multiplexer module in question.  When sample data from an old connector on a 
multiplexer module was compared with data collected on the same module with a 
new connector, the chain signature remained much the same.  Those modules that 
behaved predictably as a function of temperature on the old connectors did so on the 
new connectors and those modules that exhibited a chain signature that behaved 
erratically on the old connectors were shown to do the same on the new connectors.  
An example of this behavior is demonstrated in Figure 41 in which a given 
multiplexer module recorded SAC305 resistance data using edge connector B while 
during later dummy testing it was used to record aluminum insert data using an 
entirely different edge connector L.  The selected chain presented in the figure 
demonstrates how similar the resistance signatures are even though the only common 
thread between the two is the module used during the resistance monitoring process. 
 
 
Figure 41.  Observed chain similarities across varying samples and connectors. 
 
 
 102 
 
The resulting conclusion gathered from the diagnostic work performed during 
electrical testing was that as passive thermal cycling carried on, select channels of the 
multiplexer modules became corrupted and began recording distorted resistance data.  
Although it is not known what caused this corruption, or how to avoid it, steps were 
taken to mitigate its effects on the sample populations.  In other words, all corrupted 
channels on the multiplexer modules in question were identified so that the 
appropriate sample swapping among modules could take place to maximize the 
number of uncorrupted readings during electrical testing.  Within solder materials, 
those multiplexer modules that showed no channel corruption were switched with 
those that showed a significant amount of corruption, up to ~30% of channels on the 
module in some cases.  This approach in preserving the validity of the electrical data 
was necessitated by the limited number of multiplexer modules available.  
Subsequent sample swapping was extremely effective in mitigating the effects of the 
corrupted channels on the breadth of electrical data; however, it was not able to 
eliminate them altogether.  As mentioned earlier, the sample population for electrical 
testing was designed to feature 32 corner solder joints per solder material per strain 
energy condition.  Some of these original sample populations, P0, were modified 
when the corrupted channel effects could not be removed.  These adjusted 
populations, Padjusted, are displayed in Table 14. 
 
 
 
 
 
 
 
 
 
 103 
 
Table 14.  Adjusted Electrical Sample Populations. 
 Chamber Package ?W (N/m2) P0 Padjusted 
Ag-In TLPS LT 44-lead CQFP 9650 32 27 
- HT - 13100 32 28 
- LT 64-lead CQFP 18615 32 32 
- HT - 24820 32 28 
      
Au-Sn LT 44-lead CQFP 9650 32 32 
- HT - 13100 32 28 
- LT 64-lead CQFP 18615 32 32 
- HT - 24820 32 30 
      
SAC305 LT 44-lead CQFP 9650 32 32 
- HT - 13100 32 29 
- LT 64-lead CQFP 18615 32 32 
- HT - 24820 32 31 
   
At the conclusion of thermal cycling and testing, the Au-Sn and SAC305 samples had 
accrued 800 cycles while the Ag-In TLPS samples had been subjected to 400 cycles.  
The decision to conclude testing was arrived upon in response to cracking observed in 
sample cross sections in addition to the behavior of sample mechanical strengths.  
These other considerations will be discussed in more detail within subsequent 
sections.  Although significant cracking within the Ag-In TLPS and SAC305 joints 
was observed in samples taken from the cycled population, the resistance signatures 
at the conclusion of passive thermal cycling did not reflect these observations.  In 
fact, all corner solder joints from the adjusted populations featured final resistance 
signatures characteristic of their original baselines.  This was an expected outcome 
for the Au-Sn samples as cross sectional and mechanical data suggested the joints had 
not suffered significant, permanent damage during passive thermal cycling, however, 
it warrants discussion with respect to the Ag-In TLPS and SAC305 samples.  For 
these samples, it is hypothesized that the resistance sampling rate in conjunction with 
the number of solder joints within each daisy chain may be responsible for this 
 
 104 
 
behavior.  Although significant cracking was observed in the Ag-In TLPS and 
SAC305 corner solder joints, it is reasonable to suggest that compressive forces from 
adjacent joints were responsible for holding any fractured surfaces in sufficiently 
good contact with one another.  If the cracking had occurred through the entire joint, 
subsequent thermal cycling would have caused these surfaces to slide relative to one 
another, therefore causing transient resistance changes that can be recorded by 
continuously monitoring event detectors or oscilloscope setups.  Although the 
resistance data failed to characterize this cracking within the corner solder joints, the 
mechanical data collected during this investigation did correlate with these physical 
observations. 
   
5.3 Mechanical Data Trends and Discussion 
 
In contrast to the in situ measurements possible with electrical testing, the mechanical 
testing portion of this investigation required careful sample planning in order to 
balance the resolution and absolute range of the assessment.  Because the mechanical 
testing was destructive in nature, the finite number of mechanical samples prepared 
for testing needed to be carefully distributed among the planned test points during 
cycling.  Without knowledge of how the samples would behave in response to passive 
thermal cycling, mechanical sampling points were placed closely to one another 
during the initial stages of cycling.  This high resolution was planned in order to 
capture any short-term changes to the solder joints that would have affected pull 
strength.  The number of cycles between mechanical test points was then gradually 
increased in order to preserve sample quantities for longer than planned passive 
 
 105 
 
thermal cycling or unexpected changes in mechanical strength during the later stages 
of testing that may have required additional resolution.  The resulting sampling 
schedule when taking all of these considerations into account was as follows: 0, 50, 
100, 200, 400, 600, and 800 passive thermal cycles.  During testing, SAC305 
remained on schedule while mechanical samples were collected from Au-Sn 
mechanical assemblies at 700 instead of 600 passive thermal cycles and from Ag-In 
TLPS samples at 250 instead of 200 passive thermal cycles.  This deviation from the 
planned sampling schedule was driven by concerns about sample populations.  A 
summary of the true sampling periods and populations observed during this 
investigation can be seen below in Table 15. 
 
Table 15.  Observed mechanical testing periods and samples. 
Passive Thermal Cycles  
0 50 100 200 250 400 600 700 800 
Ag-In TLPS 7 4 3 - 4 7 - - - 
Au-Sn 7 4 3 4 - 3 - 4 7 
SAC305 7 4 3 4 - 3 4 - 7 
 
The number of corner solder joints tested at each of the sampling periods can be seen 
above in Table 15.  Larger sample populations were collected at the beginning and 
end of mechanical testing in an effort to reduce data variation at these points.  This 
reduction in variation enabled conclusions regarding trends in the pull strength data to 
be made with higher confidence.  Many of the sample populations cited in Table 15 
are odd in number because at these intervals, single samples were set aside for cross 
sectioning.  Proper discussion of the data collected through cross sectioning will 
occur in the sections to follow. 
 
 
 106 
 
5.3.1 SAC305 Mechanical Results 
 
 
Presentation and discussion of the mechanical testing results will be divided among 
solder materials in order to identify failure modes and mechanical strength trends 
specific to the individual solders prior to making any general comparisons between 
them.  The first of the solder materials to be discussed is SAC305.  Pull testing of 
these samples yielded consistent interfacial failures between the bottom of the surface 
mount lead and top of the solder joint as described in [46].  The location of these 
failures with respect to the surface mount assembly is illustrated in Figure 42 below. 
 
 
Figure 42.  Interfacial failure between component lead and solder joint during pull strength 
testing. 
 
This failure mode was observed throughout mechanical testing of SAC305 samples 
for both package sizes and temperature cycling conditions.  Cross sections of 
SAC305 joints demonstrated that a significant intermetallic layer existed between the 
surface mount lead the solder joint itself.  This intermetallic was likely the location of 
failure as cracking along this feature in HT SAC305 samples correlated with a 
decrease in recorded pull strengths during mechanical testing.  In order to provide the 
 
 107 
 
mechanical data in a presentable format, sample means and variances were calculated 
using the following equations, respectively: 
 
)...(1 321
__
nXXXXnX ?????                      (10) 
 
and 
 
2__
1
2 )(
1
1 XX
nS
n
j
j ??? ?
?
 from Kreyszig [55].                     (11) 
 
The normal distribution was assumed to govern the shape of all mechanical sample 
probability density functions.  Although sample standard deviations were computed 
to characterize the calculated means, they were not used directly to present the pull 
test data.  Rather, these values were used to calculate 90% confidence intervals (CIs) 
on the established means; these CIs were then used in subsequent presentations of the 
pull strength trends. 
 
With respect to the strength trends observed within the SAC305 population, there was 
a distinct difference based upon passive thermal cycling condition, with HT samples 
deteriorating at a faster rate than those subjected to the LT profile.  This deterioration 
of HT SAC samples is evident in Figure 43 as significant weakening between 200 and 
800 cycles.  LT SAC samples from this plot do not undergo any significant 
strengthening or weakening trends and are shown to maintain relatively constant pull 
strength values over the duration of cycling.  Cross sectional data gathered during 
passive thermal cycling indicated the growth of a very thick intermetallic layer at the 
surface mount lead / solder interface on HT samples between 0 and 100 cycles.  
Significant cracking between this intermetallic layer and the bulk solder joint was 
 
 108 
 
observed in a subsequent cross section collected at 400 cycles.  LT SAC samples did 
not mirror this intermetallic behavior and subsequently, did not crack during cycling.  
It is this disparity in SAC solder joint intermetallic growth and insipient cracking that 
is believed to have contributed significantly to the trend of the SAC pull strength data 
provided in Figure 43 and Figure 44. 
 
 
Figure 43.  Pull strength trends of SAC, 64-lead CQFP samples. 
 
The trend of 44-lead CQFP SAC samples provided in Figure 44 mirrors that of Figure 
43 in that the pull strength of small package HT joints drops sharply to approximately 
half its original value within 200 cycles and then continues to decrease to almost one 
third of the baseline strength by the end of cycling.  In both package sizes, the LT 
SAC samples are seen to be appreciably stronger than their HT counterparts at the 
conclusion of mechanical testing.  This behavior correlates well with the physical 
cracking in HT SAC joints and the lack thereof in the LT SAC samples.  It is also 
worth noting that the behavior of HT SAC pull strength appears to become 
 
 109 
 
asymptotic at 600 cycles for the 64-lead CQFP and 400 cycles for the 44-lead CQFP.  
This trend suggests that intermetallic formation, which is self-limiting, had reached a 
point at which little additional intermetallic was forming with time at temperature.  
As such, the joints were no longer becoming increasingly brittle, which would then 
explain the decreasing rate of pull strength degradation observed in Figure 43 and 
Figure 44.  Additionally, it is believed that severe cracking of the HT joints may have 
led to increased compliancy and resulted in an effective reduction in imposed strain 
energy density.  It is possible that the asymptotic behavior observed is a combination 
of these two considerations occurring simultaneously.  
 
 
Figure 44.  Pull strength trends of SAC, 44-lead CQFP samples. 
 
In order to confirm the observed trends within the data using statistical methods, a 
one-way analysis of variance (ANOVA) was performed between the baseline and 
800-cycle pull strengths for each of the estimated strain energy densities.  This 
statistical test was intended to determine whether the solder joints mechanically tested 
 
 110 
 
at the conclusion of passive thermal cycling were of a different population than those 
that were pull tested from the as-manufactured samples.  The ANOVA was 
performed using the MATLAB function, anova1.  Pull strength values of the baseline 
and 800-cycle solder joints were organized in separate columns of a matrix as 
required by the anova1 function.  The resulting MATLAB analysis output a p-value 
that indicated whether or not the two columns of data were of distinct populations 
based on the comparison of their means and variances; a p-value close to zero 
indicated that the means of the two data columns were significantly different than one 
another.  Significantly differing means suggested that pull strength degradation due to 
cracking and/or intermetallic formation had occurred within the sample solder joints 
during passive thermal cycling.  A common critical p-value of 0.01 was selected for 
the ANOVA of all solder materials; p-values lower than this threshold indicated that 
the sample means were significantly different while p-values higher than the critical 
value indicated similar sample means.  The results of the ANOVA for the SAC305 
samples are provided in Table 16.  As the table shows, SAC305 samples subjected to 
HT cycling were characterized by p-values equal to zero.  This result statistically 
validates the observation that significant pull strength degradation occurred within 
these samples as the ANOVA p-values reported were below the critical p-value of 
0.01.  SAC305 solder joints of the 44-lead CQFPs subjected to LT passive thermal 
cycling also featured a p-value indicative of pull strength degradation, however, this 
value suggests that the degradation was not as severe as that of the HT samples 
because it is non-zero.  ANOVA of the SAC305 64-lead CQFP LT pull strength data 
resulted in a p-value of ~0.75, which was well above the critical value of 0.01, 
 
 111 
 
meaning that pull strength degradation did not occur within these samples.  The non-
zero p-values reported for the SAC305 LT samples stood in contrast to the p-values 
of zero for the HT samples.  This behavior reflects the possibility that a temperature 
related degradation mechanism is responsible for the SAC pull strength trends. 
Table 16.  SAC305 ANOVA results. 
  ?W (lb/in2) ?W (N/m2) ?? p-value 
44-lead CQFP LT 1.4 9650 0.119 0.00052 
 HT 1.9 13100 0.140 0.00000 
64-lead CQFP LT 2.7 18615 0.142 0.75460 
 HT 3.6 24820 0.167 0.00000 
 
5.3.2 Eutectic Au-Sn Mechanical Results 
 
The next solder material to be discussed regarding mechanical testing is eutectic Au-
Sn.  Similar to the failure mode observed in SAC305 samples, Au-Sn samples 
subjected to mechanical testing often failed along the surface mount lead / solder joint 
interface depicted in Figure 42.  Although this interfacial failure was the dominant 
mode observed during testing, an alternative failure mode commonly referred to as 
pulling off the land [46] or pad lift was occasionally observed in failure of 44-CQFP 
solder joints.  An illustration of this failure mode is provided in Figure 45.  Clearly, 
when this type of failure occurs the strength of the solder joint is greater than the 
bonding strength between the copper pad and the PWB.  It is hypothesized that the 
majority of pad lifts occurred in the 44-CQFP solder joint population because the 
surface mount pads for these packages were characterized by a cross-sectional area 
two-thirds that of the surface mount pads for the 64-lead CQFPs.  This comparatively 
smaller footprint led to higher stresses at the pad / PWB interface which at times were 
high enough to exceed the adhesive strength of the pad to the PWB.  An additional 
 
 112 
 
feature causing this failure mode in Au-Sn joints may have been the higher 
temperature to which the PWBs were subjected during reflow, resulting in thermal 
damage of the board. 
 
 
Figure 45.  Pad lift failure mode during pull strength testing. 
 
Those Au-Sn joints that failed through pad lift were removed from the population of 
mechanical test data.  In particular, all Au-Sn joints on 44-lead CQFPs subjected to 
LT cycling sampled at 50 and 400 cycles failed through pad lift.  It is for this reason 
that the discontinuity exists in Figure 46.  As far as the trend of this plot is concerned, 
it is clear that the Au-Sn samples survived thermal cycling much better than the SAC 
samples from before.  The plateaus formed by the pull strength trends of the 44-lead 
CQFP packages subject to either cycling condition are evidence that microstructure of 
the joints have stabilized and they remained largely undamaged up to the conclusion 
of testing.  As expected, the joints subject to higher temperature and strain energy 
densities feature lower pull strengths than their counterparts, however, their final 
values remain very close to their baseline reading of ~800 grams force.  It is likely 
that accelerated grain growth resulting from the additional temperature of the HT 
 
 113 
 
cycle is responsible for the difference in final pull strengths.  Further discussion of 
sample microstructure will be provided in the analysis of the cross section data. 
 
 
Figure 46.  Pull strength trends of Au-Sn, 44-lead CQFP samples. 
 
The pull strength trends of Au-Sn joints on 64-lead packages can be found in Figure 
47 below.  From the plot, it is clear that a significant readjustment occurs within the 
window of 0 to 50 cycles for samples subject to both cycling conditions.  Following 
the readjustments experienced in the initial stages of cycling, both large package 
trends level out and achieve stable pull strengths over a span of 400 cycles prior to the 
conclusion of testing.  This repeated plateau provides additional evidence of the 
material?s excellent reliability with respect to the environmental testing performed 
during this investigation.  This agreement of pull test data between package sizes 
suggests that evidence of solder joint fatigue will not be found during the cross 
sectional analysis of any of the Au-Sn joints.  Furthermore, the nature of the alloy 
indicates there will be minimal interfacial intermetallic growth although 
 
 114 
 
microstructural coarsening may be detected, as there is evidence of some initial 
change within solder joints sampled for the trends of Figure 47. 
 
 
Figure 47.  Pull strength trends of Au-Sn, 64-lead CQFP samples. 
 
The ANOVA of the eutectic Au-Sn sample pull strength data was also performed as 
described by the prior SAC305 pull strength discussion.  The p-values that resulted 
from the statistical analysis are provided in Table 17.  As the table shows, all but one 
of the p-values were above the critical value of 0.01, therefore confirming the lack of 
pull strength degradation within the solder joints because the sample means of the as-
manufactured and cycled joints were found to be statistically similar. 
Table 17.  Eutectic Au-Sn ANOVA results. 
  ?W (lb/in2) ?W (N/m2) ?? p-value 
44-lead CQFP LT 1.4 9650 0.119 0.05740 
 HT 1.9 13100 0.140 0.06040 
64-lead CQFP LT 2.7 18615 0.142 0.01640 
 HT 3.6 24820 0.167 0.00580 
 
 
 115 
 
 
5.3.3 Ag-In TLPS Mechanical Results 
 
Having discussed testing of SAC305 and eutectic Au-Sn, the last of the solder 
materials to be treated within this section is Ag-In.  Mechanical pull testing of Ag-In 
TLPS samples demonstrated that this material failed at the lead / solder joint 
interface, similar to joints from the SAC and Au-Sn populations.  The location of the 
failure was likely due to the presence of brittle AuIn intermetallics formed at the 
interface by the reaction of the Au plating of the lead with In from the solder paste.  
However, at the moment of failure this material was observed to behave differently 
from the other solders.  When samples from the SAC305 and eutectic Au-Sn 
populations failed, the solder joint remained intact and attached to the PWB as one 
might expect.  In contrast to this behavior, when Ag-In TLPS joints failed, the toe of 
the solder joint would fracture and be ejected from the vicinity.  This apparent brittle 
nature of the material may very well be characteristic of the compacts formed with 
TLPS, but it may also be related to the initial pull strength of the joints and/or the 
existence of cracking or delamination at the solder joint / surface mount pad interface. 
 
The behavior of the pull strength trends in Figure 48 and Figure 49 are very similar to 
those discussed above.  In both cases the solder joints subject to higher strain energy 
densities feature the lower of the two pull strengths at the conclusion of testing.  
Preliminary cross sections of the samples confirmed that cracking along the 
intermetallic layer formed at the interface between the solder joint and the lead 
existed within samples of all strain energy densities.  It is hypothesized that this 
cracking is responsible for the pull strength trends below.  At the conclusion of 
 
 116 
 
mechanical testing at 400 thermal cycles, some Ag-In TLPS corner solder joints were 
severely cracked.  It was for this reason in conjunction with sample quantity 
considerations that passive thermal cycling of the material was concluded at this 
point.  
 
 
Figure 48.  Pull strength trends of Ag-In TLPS, 44-lead CQFP samples. 
 
 
Figure 49.  Pull strength trends of Ag-In TLPS, 64-lead CQFP samples. 
 
 117 
 
 
As the pull strength trends of the Ag-In TLPS joints show, the variance of the 
mechanical data collected for this solder material is more significant than that of the 
SAC305 and Au-Sn solders due to smaller pull strength means.  The ANOVA of the 
Ag-In TLPS was therefore an important tool in providing a quantitative assessment of 
any pull strength degradation.  The results of this statistical analysis are displayed in 
Table 18.  As the table shows, all p-values computed during the ANOVA are well 
above the critical value of 0.01.  This result suggests the baseline and cycled pull 
strength samples for all strain energy densities are statistically similar, however, it is 
believed that the inherent variance of the pull strength data strongly influenced this 
outcome.  This is most likely due to the fact that the site of crack initiation in the 
sample is the Au-In intermetallic, which is formed upon melting of the indium in the 
initial bonding step and does not grow with subsequent cycling because the Au is 
completely consumed in the initial bonding.  This intermetallic formation results in a 
joint that is weak to start with, remains weak throughout, and features a wide variance 
in pull strength. 
 
Table 18.  Ag-In TLPS ANOVA results. 
  ?W (lb/in2) ?W (N/m2) ?? p-value 
44-lead CQFP LT 1.4 9650 0.119 0.97880 
 HT 1.9 13100 0.140 0.26560 
64-lead CQFP LT 2.7 18615 0.142 0.26590 
 HT 3.6 24820 0.167 0.08360 
 
In summary, mechanical pull strength testing of the various test assemblies provided 
valuable information regarding the health of the solder joints under passive thermal 
cycling.  With respect to the SAC305 test assemblies, it was observed that HT cycling 
 
 118 
 
led to significant losses in pull strength while samples subjected to LT cycling 
maintained noticeably higher strengths throughout the duration of the investigation.  
It has been proposed that intermetallic growth at the surface mount lead / solder joint 
interface under HT conditions is responsible for this behavior.  Evidence of this 
intermetallic layer in addition to the localized cracking will be provided in the 
subsequent section.  Although slight changes in the Au-Sn sample pull strength were 
observed, the material proved to maintain a high level of mechanical integrity and 
was regularly the strongest material for all strain energy density classifications.  
Provided that additional testing to these samples could be performed, it would be 
expected that well over 1000 thermal cycles would be required to weaken the joints to 
the levels observed for the SAC and Ag-In materials.  In order to illustrate the 
relationship between the pull strengths of the various materials, Figure 50 has been 
prepared.  As it shows, eutectic Au-Sn was consistently stronger than the other 
materials with SAC305 showing the greatest drop in strength with passive thermal 
cycling.  The low strength of the Ag-In TLPS samples is believed to be due to the 
porosity of the material in conjunction with the presence of the brittle AuIn 
intermetallic layer at the lead / solder joint interface. 
 
 
 119 
 
 
Figure 50.  Comparison plot of solder material pull strengths. 
 
5.4 Cross Sectional Data Trends and Discussions 
 
Although monitoring the electrical continuity and mechanical strength trends of an 
electronics assembly is a common practice that yields valuable information about the 
health of the system, these methods still require microstructural data to correlate the 
observed behavior to physical phenomena.  For this particular investigation, cross 
sections were collected from all test assemblies in an effort to provide qualitative 
microstructural data.  As the corner solder joints of the surface mount packages were 
expected to be the first to fail, these were subjected to cross-sectioning and 
subsequent imaging.  Due to the destructive nature of this data collection method, 
samples allotted for cross sectioning were designed into the mechanical testing 
 
 120 
 
population.  The sampling schedule for this investigation can be seen in Table 19 
below. 
 
Table 19.  Observed cross section sampling schedule and sizes. 
Passive Thermal Cycles  
0 50 100 200 400 600 700 800 
Ag-In TLPS 1 - 1 - 5 - - - 
Au-Sn 1 - 1 - 1 - - 5 
SAC305 1 - 1 - 1 - - 5 
 
As the table above shows, a single solder joint per solder per strain energy condition 
was collected until the final sampling period.  The last sampling period for each 
solder material was supplemented with additional corner solder joints collected from 
the electrical test population for a total of 5 samples per solder per strain energy 
condition.  These additional samples at the conclusion of thermal cycling were 
intended to ensure that physical observations were common to all sampled joints and 
therefore could be assumed to characterize the entire population.  As expected, 
intermetallic layers at the lead / solder joint interface were found to relate to cracking 
in select SAC305 and Ag-In TLPS samples while those of Au-Sn remained largely 
unchanged by passive thermal cycling. 
 
5.4.1 SAC305 Microstructural Results 
 
With respect to the cross sections collected from SAC samples during this 
investigation, it was observed that for those test assemblies subjected to HT thermal 
cycling, a significant intermetallic layer rapidly formed between the lead and solder 
joint.  However, SAC samples subjected to LT cycling did not exhibit such extreme 
intermetallic growth and therefore remained crack free for the duration of testing.  
 
 121 
 
This behavior was predicted by the aforementioned mechanical data as HT joints 
were found to lose significant fractions of their baseline pull strengths while the LT 
joints maintained more consistent values.  In the HT SAC samples the interface 
between the thick intermetallic and bulk solder joint then became the site of crack 
initiation and propagation.  Figure 51 is an ESEM image of an as-built, 64-lead CQFP 
SAC solder joint.  From inspection is it clear that the joint is healthy and any 
intermetallics at the two interfaces are not observable at the provided magnification.    
 
 
Figure 51.  SAC305 64-lead CQFP joint subjected to 0 thermal cycles. 
 
Figure 52 is the final sample collected from the SAC 64-lead CQFP population 
subjected to 800 LT thermal cycles.  There is no evident deviation from the 
characteristics of the baseline sample shown in Figure 51.  The joint is free from 
cracks and the intermetallic layers are thin and uniform.  The image in Figure 53 of a 
SAC 64-lead CQFP joint subjected to 800 HT thermal cycles stands in contrast to the 
above observations.  In this sample, cracking through the entire joint is easily seen in 
the ESEM image.  In addition, it is clear that a thick, 2-phase intermetallic has formed 
between the lead and solder joint.  The extent of this growth and characterization of 
the intermetallic will be discussed shortly.  It is important to note that the 
 
 122 
 
observations made for SAC joints of 64-lead CQFP packages were found to repeat in 
the 44-lead CQFP solder joints as well. 
 
 
Figure 52.  SAC305 64-lead CQFP joint subjected to 800 LT thermal cycles. 
 
 
Figure 53.  SAC305 64-lead CQFP joint subjected to 800 HT thermal cycles. 
 
In both package sizes under HT conditions, growth of the intermetallic layer and 
subsequent cracking between it and the bulk solder were found to progress as a 
function of thermal cycling.  At 100 cycles, the intermetallic layer was visible at 
~100x magnification along with a number of cracks spanning the thickness of this 
layer (perpendicular to the surface of the lead).  At 400 cycles, new cracks running 
parallel to the surface of the lead can be seen to begin splitting the intermetallic from 
the bulk solder.  These parallel cracks can then be seen to complete run completely to 
 
 123 
 
the edge of the solder heel and toe at the conclusion of testing (i.e. 800 cycles).  It is 
important to note that while this cracking is occurring, the intermetallic layer is 
continually growing.  An illustration of the previously described progression can be 
viewed in Figure 54 where the heels of 64-lead CQFP SAC joints subjected to HT 
cycling have been placed side-by-side to facilitate the comparison. 
 
 
 
Figure 54.  Intermetallic layer growth and crack progression in SAC 64-lead CQFP HT joints. 
 
 
 124 
 
In an effort to characterize the apparent growth of the lead side intermetallic, its 
thickness was measured as a function of thermal cycles.  Five independent 
measurements of the intermetallic thickness were collected per strain energy density 
per cycle count.  These measurements were then averaged and plotted against the 
number of thermal cycles in Figure 55.  As the plot clearly shows, the lead-side 
intermetallic layer grew monotonically with thermal cycles for all samples.  It is of 
particular interest that the LT samples from both package sizes ended with 
intermetallic thicknesses of ~2.5 times their original value while HT samples from 
both package sizes showed an increase of ~10 times. 
 
 
Figure 55.  SAC lead side intermetallic thickness as a function of thermal cycles. 
 
The trends from the above plot are consistent with the laws of diffusion as the higher 
temperature afforded by HT thermal cycling corresponds to thicker lead side 
intermetallics regardless of package size.  In addition to tracking the growth of this 
intermetallic layer, efforts were made during this investigation to characterize its 
 
 125 
 
composition as well.  The layer was composed of two distinct compositions that 
contrast well under imaging with a standard light microscope.  An image of a SAC 
44-CQFP HT solder joint heel has been provided in Figure 56 in order to illustrate the 
difference between the layers.  As it shows, a lighter composition can be found 
closest to the bulk solder joint while a distinctly different layer exists closer to the 
lead.  In order to determine the compositional makeup of the two layers, the energy 
dispersive x-ray spectroscopy (EDS) function of the ESEM was required.  
 
 
Figure 56.  SAC305 44-lead CQFP HT solder joint heel imaged using light microscope.   
 
An EDS map of the intermetallic layers demonstrated that they were composed 
largely of tin and nickel as shown by Figure 57.  In this image the lead is located to 
the upper right while the solder can be found to the lower left of the intermetallic 
layer.  These can be easily identified due to the increased levels of iron and nickel in 
the alloy 42 lead and increased levels of tin in the SAC solder.  Next, an EDS point 
scan was conducted to quantify the ratios of the metallic constituents.  The locations 
of the EDS point scans were targeted on two sites on either side of the compositional 
dividing line as shown in Figure 58. 
 
 
 126 
 
 
Figure 57.  EDS Map of lead side intermetallic. 
 
 
Figure 58.  EDS Point Scan locations. 
 
The resulting quantification of the metallic constituents at the scan sites above is 
provided in Table 20.  As expected, the sample site located closer to the bulk solder is 
slightly more tin rich than that closer to the lead.  Additionally, site 2 contains gold 
that was most likely dissolved from the lead plating during reflow.  It is believed that 
the intermetallic represented by site 1 was the former location of the nickel 
metallization while the intermetallic of site 2 was the location of the immersion gold 
finish of the lead.  The compositions of the intermetallics at site 1 and site 2 are 
believed to be NiSn3 and (Ni,Au)Sn3, respectively. 
 
 127 
 
 
Table 20.  EDS point scan results of SAC intermetallic. 
Metal Site 1 (Atomic %) Site 2 (Atomic %) 
Tin (Sn) 68.41 75.8 
Nickel (Ni) 25.11 11.9 
Iron (Fe) 5.97 3.88 
Gold (Au) 0.51 8.43 
 
In summary, for those SAC305 test assemblies subjected to LT conditions, it was 
found that the solder joints survived testing without showing any signs of fatigue or 
excessive intermetallic growth.  However, those SAC solders that experienced HT 
cycling rapidly developed thick intermetallic layers at the lead / solder interface that 
served as crack propagation sites in all samples that were observed.  As previously 
described, the resulting cracks likely developed as perpendicular separations within 
the intermetallic that began to propagate along the bulk solder / intermetallic interface 
after initiation in a direction parallel to the bottom surface of the lead.  Because of the 
swift formation of these intermetallics and their association with the observed damage 
to the SAC solder joints, it is recommended that this solder material be avoided for 
applications around 185?C and higher for similar test assembly material systems. 
 
5.4.2 Eutectic Au-Sn Microstructural Results 
 
In contrast to a number of changes observed within the SAC joints, those samples 
fabricated from eutectic Au-Sn were notably unaffected by the thermal cycling 
conducted for this investigation.  Baseline Au-Sn solder joints were characterized by 
a lamellar structure consisting of a ?`- phase and ? - phase.  Performing EDS on 
cross-sectioned samples in the ESEM quickly identified these compounds after cross-
 
 128 
 
referencing with a Au-Sn binary phase diagram.  The results from the EDS point scan 
over the suspected phases are provided in Table 21 below. 
 
Table 21.  EDS point scan results of Au-Sn bulk solder. 
 ?` (Atomic %) ? (Atomic %) 
Gold (Au) 82.3 50.6 
Tin (Sn) 17.7 49.4 
   
The above atomic fractions reported from the EDS scan agree well with the data 
provided from the relevant portion of the Au-Sn binary phase diagram in Figure 59.  
As the diagram shows, the AuSn intermetallic, ?, lies at 50 at. % Sn while ?` lies 
between 17 and 18 at. % Sn. 
  
 
Figure 59.  Truncated Au-Sn binary phase diagram [56]. 
 
The distributions of ?` and ? can be viewed within an Au-Sn 64-lead CQFP LT solder 
joint in the image provided in Figure 60.  The ?` phase corresponds to the lighter 
shapes within the image due to its higher gold content that results in stronger 
backscatter signal within the ESEM.  This particular microstructure is shown in the 
context of the solder joint in Figure 61.  As this image demonstrates, the lamellar 
 
 129 
 
structure dominates the bulk of the solder joint while larger areas of ?`- phase are 
located along the lead and pad interfaces.  Additionally, islands of the ?`- phase were 
many times observed within the bulk of the Au-Sn solder joints, however, the 
distribution and concentration of these islands did not seem to form a pattern or trend 
over the cross-sectioned samples. 
 
 
Figure 60.  Phase distribution within eutectic Au-Sn solder joint. 
 
 
Figure 61.  Solder joint heel of an Au-Sn 44-lead CQFP LT sample. 
 
 130 
 
 
As previously mentioned, the microstructure observed in the as-built samples did not 
undergo any noticeable changes during passive thermal cycling.  In addition, there 
was no evidence of crack initiation and propagation within any of the Au-Sn solder 
joints.  In order to illustrate the exceptional health of these solder joints as a function 
of thermal cycling, the heels of the joints subjected to the highest strain energy 
density, those of the 64-lead CQFP HT test assemblies, are shown in Figure 62.  As 
the series of images demonstrates, the Au-Sn solder joints remain free of cracks and 
any significant microstructural coarsening over the duration of passive thermal 
cycling.  An additional feature from these images that warrants discussion is the 
existence of voiding within the Au-Sn solder joints.  In order to quantify the 
frequency of voiding within the joints, the entire population of cross-section samples 
was assessed.  During the assessment it was observed that approximately 50 percent 
of the Au-Sn solder joint population featured voiding.  This voiding was distributed 
rather evenly over the population in that solder joints of large packages were just as 
likely or unlikely to be voided as those of small packages.  Measurements of the void 
diameters showed that sizes ranged from the small 20 micron diameter voids typically 
located within the solder joint toe to very large 200 micron diameter voids located 
within the solder joint heel.  Although voiding was found to affect a sizeable portion 
of the eutectic Au-Sn population, it was not observed to have had any influence on 
the reliability of the solder joints. 
 
 
 131 
 
 
 
Figure 62.   Lack of fatigue and microstructural coarsening in Au-Sn 64-lead CQFP HT solder 
joints. 
 
Cross-sectional sampling of the eutectic Au-Sn solder joints demonstrated that the 
material is a highly desirable option for HT electronics as zero cracking was observed 
during passive thermal cycling.  This high fatigue resistance coupled with its stable 
microstructure makes it an ideal choice when ENIG finished assemblies require a HT 
solder material.  The observations made during the cross-sectional sampling correlate 
 
 132 
 
well with the trends of the mechanical data in that both indicate few changes to the 
physical characteristics and behavior of the Au-Sn after 800 thermal cycles.   
 
5.4.3 Ag-In TLPS Microstructural Results  
 
With respect to the last of the three solder materials tested during this investigation, 
the Ag-In TLPS solder joints were found to form a lead-side intermetallic layer that 
later negatively influenced their reliabilities.  This layer can be observed as the thin 
white line between the surface mount lead and the bulk solder joint within the as-built 
64-lead CQFP Ag-In TLPS joint shown in Figure 63. 
 
 
Figure 63.  Ag-In TLPS 64-lead CQFP solder joint subjected to 0 thermal cycles. 
 
In order to characterize the thickness of the intermetallic and its growth as a function 
of thermal cycling, the layer was monitored using the same approach previously 
described for SAC305: five thickness measurements were used to calculate an 
average thickness per strain energy per cycle count.  These averages were then plotted 
as a function of cycle count in order to establish the growth trend.  In the case of the 
Ag-In TLPS solder joint lead side intermetallic, the initial thickness formed during 
manufacturing of the joints remained largely unchanged during passive thermal 
cycling.  This behavior can be observed from the plot in Figure 64 below. 
 
 
 133 
 
 
Figure 64.  Ag-In TLPS lead side intermetallic thickness as a function of thermal cycles. 
 
As the figure above shows, 3 of the 4 package size / cycling condition combinations 
remain constant while the Ag-In TLPS 44-lead CQFP HT solder joint intermetallic 
sees a small increase of ~1 micron or approximately 16 percent of its baseline 
thickness.  The scale of these changes can be considered negligible when compared to 
those observed in the SAC samples: 1000% increase in baseline thickness for HT 
joints and a 250% increase for LT joints.  In order to understand why the growth of 
the intermetallic occurs in this particular manner, it is helpful to know what metals 
compose the layer.  EDS mapping and point scanning again determined the 
composition of the intermetallic layer, which was shown to contain the atomic 
composition of approximately 50 at. % Au and 50 at. % In.  This composition 
corresponds to the AuIn intermetallic of the Au-In binary phase diagram displayed in 
Figure 65 below. 
 
 
 134 
 
 
Figure 65.  Au-In Binary Phase Diagram from [30]. 
 
With an understanding of the composition of the intermetallic layer at the lead side 
interface, the growth trends of this intermetallic can then be understood as a behavior 
limited by surrounding resources.  In contrast to the intermetallics formed at the 
SAC305 lead side interface, those seen in the Ag-In TLPS can only be formed from 
the indium of the TLPS solder and the ~1.5 micron layer of gold coating the CQFP 
leads prior to processing.  It is clear from the growth behavior displayed in Figure 64 
that all available gold had been consumed and alloyed with indium during TLPS 
processing to form the AuIn intermetallic.  Therefore, subsequent thermal cycling of 
the Ag-In TLPS test assemblies did not result in subsequent growth or decomposition 
of this layer. 
 
 135 
 
 
As previously mentioned, passive thermal cycling of the Ag-In TLPS test assemblies 
resulted in cracking of the solder joints.  For all strain energy densities, a significant 
portion of the cracking occurring in the joints was found along the interface between 
the AuIn intermetallic and the bulk solder joint composed of the Ag-In matrix.  An 
illustration of this progressive cracking along the interface is provided in Figure 66. 
 
 
 
Figure 66.  Crack progression in Ag-In TLPS 64-lead CQFP HT solder joints.  
 
For this particular strain energy density, the heel of the solder joint has been 
completely removed from the lead at the conclusion of thermal cycling.  Although 
cracks were largely located at the interface, they were seen moving into the bulk Ag-
In matrix at times breaking the transient liquid sintered particles apart before 
returning to the interface between the intermetallic and the bulk matrix.  This 
behavior is demonstrated by the selected image from an Ag-In TLPS 44-lead CQFP 
 
 136 
 
HT solder joint (Figure 67) where the crack can be seen splitting through a small 
portion of the Ag-In bulk prior to reentering the intermetallic layer and terminating at 
the surface of the lead.  This increasing crack length with cycling is believed to be the 
source of the observed mean pull strength degradation, although that degradation is 
not statistically significant owing to the already low force needed to cause cracking in 
as bonded samples. 
 
 
Figure 67.  Cracking in Ag-In TLPS 44-lead CQFP HT solder joint subjected to 400 HT cycles. 
 
In addition to the observed cracking of the Ag-In TLPS solder joints, passive thermal 
cycling of the test assemblies revealed interesting behavior regarding the cores of 
former silver particles within the Ag-In matrix.  As passive thermal cycling 
progressed, the silver cores were seen becoming progressively roughened due to 
material loss resulting from the outward diffusion of silver into the Ag-In matrix.  
EDS maps were used to identify the former silver cores as seen in Figure 68.  As the 
image shows, the Ag-In particles are characterized by a silver rich core while 
traveling out in a radial fashion results in an alloy increasing in indium content. 
 
 
 137 
 
 
Figure 68.  EDS map of Ag-In particle. 
 
As previously mentioned, evidence of silver diffusion from the core of these Ag-In 
particles was observed as a function of passive thermal cycling.  Figure 69 
demonstrates this physical phenomenon through the pits that have formed at the 
center of the Ag-In particle subjected to 400 HT cycles.  
 
 
Figure 69.  Evidence of silver diffusion from Ag-In particles within Ag-In TLPS matrix. 
 
In summary, although cracking to the same extent as observed in the HT SAC 
samples did not occur within the Ag-In TLPS samples, the drop in already poor 
mechanical pull strength due to the aforementioned cracking within the heels for all 
package sizes combined with the material?s brittle nature suggest that further work 
 
 138 
 
with the potential HT solder and the best corresponding lead / board finishes is 
required to improve its reliability. 
 
 
 139 
 
Chapter 6: Conclusions, Contributions, and Future Work 
 
6.1 Conclusions 
 
This thesis presents the relative reliabilities of SAC305, eutectic Au-Sn, and Ag-In 
TLPS solder joints subjected to HT passive thermal cycling.  These solder materials 
were monitored for electrical resistance, mechanical pull strength, and microstructural 
changes as a function of thermal cycles. 
 
In order to provide assemblies for testing, each of the solder materials was used as a 
part of a typical surface mount manufacturing process.  SAC305 and eutectic Au-Sn 
manufacturing parameters were gleaned from the appropriate sources while 
fabrication with Ag-In TLPS required significantly more investigation and 
development.  During development of the Ag-In TLPS paste, it was demonstrated that 
reducing the additive (indium) particle size led to statistically significant 
improvements in the volume fraction of the joints.  Successful attachment of CQFPs 
to the PWB using Ag-In TLPS also demonstrated that surface mount joints with 
mechanical and electrical integrity could be manufactured from the material. 
 
Subsequent testing of all solder samples demonstrated that for the given test vehicle 
metallization and cycling conditions, eutectic Au-Sn was the most reliable solder of 
the three.  This solder material consistently exhibited the highest pull strength values 
at all strain energy densities over the duration of testing.  Additionally, the stable 
microstructure of the Au-Sn joints as well as the lack of any cracking in them 
indicates that the solder material is the best suited for the HT testing that was 
 
 140 
 
conducted.  SAC305 samples subjected to the HT cycling profile (-55 to 185?C) were 
observed to grow significant Sn-based intermetallics between the surface mount lead 
and solder joint within a short period of time.  These intermetallic layers were 
responsible for the extensive cracking observed in the SAC305 joints at the 
conclusion of cycling.  SAC305 samples subjected to the LT cycling profile (-55 to 
150?C) did not exhibit this rapid intermetallic growth and therefore survived to the 
conclusion of testing.  In both cases, the mechanical pull strength reflected the 
microstructural characteristics of the cross section samples in that HT profile SAC 
samples experienced significant drops in mechanical strength while LT profile SAC 
samples maintained consistently higher strength values during cycling.  Ag-In TLPS 
samples exhibited mechanical integrity during this investigation, however, it was 
consistently the weakest solder material of the three.  Extensive cracking along the 
AuIn intermetallic formed at the lead interface was responsible for the observed 
weakening of the Ag-In TLPS pull strength.  Experimentation with varying test 
vehicle metallizations and work on improving the brittle nature of the material are 
suggested as developmental tasks to advance this promising material. 
Although significant solder joint cracking was observed in portions of the SAC305 
and Ag-In TLPS populations, all solder samples tested for electrical resistance during 
this investigation maintained continuity throughout the duration of thermal cycling.  
This result agrees with the mechanical data because although degradation of pull 
strength was observed, all samples maintained some level of mechanical integrity at 
the conclusion of testing. 
 
 
 141 
 
6.2 Contributions 
 
The major contributions provided by this work are as follows: 
 
??The first implementation of Ag-In TLPS as a surface mount component attach 
 
??The first study of indium particle size on Ag-In TLPS solder joint density 
 
??The first characterization of Ag-In TLPS as a function of passive thermal 
cycling and subsequent comparison against a baseline HT solder material 
(eutectic Au-Sn) and commercial solder material (SAC305). 
 
??The first study of SAC305 at 185?C on alloy 42. 
 
6.3 Future Work 
 
The following are suggestions for future work: 
 
??Implementation of Ag-In TLPS on silver finished test assemblies to assess 
reliability without AuIn intermetallics observed during this study. 
 
??Further optimization of the Ag-In TLPS paste to improve joint density. 
 
 142 
 
Appendix A ? MATLAB Lead Compliance Formulas 
 
MATLAB code utilizing the Lead Compliance Formulas for the Inclined Gull-Wing 
Lead Structural Model from [49]. 
 
clear all 
clc 
 
%GEOMETRIES 
%side view 
%H = .762*(40/1000); 
L1 = 8/1000; 
L2 = 55/1000; 
L3 = 0/1000; 
L4 = 0/1000; 
R1 = 30/1000; 
R2 = 12/1000; 
%X = .549*(40/1000); 
alpha = 0;%atan(X/H); in RADIANS! 
H = (R1+R2)*(1-sin(alpha))+(L1+L2)*cos(alpha) 
Curv = L1+L2+L3+L4+(R1+R2)*(1.5708-alpha) 
%cross-section view 
l = 15/1000; 
w = 8/1000; 
 
l2 = 27/1000; 
w2 = 8/1000; 
 
%PRE-DEFINED CONSTANTS 
C1 = H-R2; 
C2 = .7854-(alpha/2)+(1/4)*sin(2*alpha); 
C3 = .7854-(alpha/2)-(1/4)*sin(2*alpha); 
C4 = L4*sin(alpha)+R1*cos(alpha); 
C5 = L4*cos(alpha)+R1*(1-sin(alpha)); 
C6 = (L1+L2)*sin(alpha)+(R1+R2)*cos(alpha)+L4; 
 
%ADDITIONAL PARAMETERS 
f = 1.20; 
A1 = l*w; 
%A2 = A1; 
A2 = l2*w2; 
 
I1 = (l*w^3)/12; 
%I2 = I1; 
I2 = (l2*w2^3)/12; 
 
 143 
 
 
I3 = (w*l^3)/12; 
%I4 = I3; 
I4 = (w2*l2^3)/12; 
 
%E = 20.6*10^6;  %(Alloy 42)  
E = 17.5*10^6; %(Cu Alloy) lbf/in^2 
%G = 7.92*10^6 %(Alloy 42)  
G = 6.6*10^6; %(Cu Alloy) lbf/in^2 
 
n1 = .65; 
beta1 = .141+.088*(l/w-1)^n1 
n2 = .47; 
beta2 = .141+.088*(l2/w2-1)^n2 
T1 = (1/12)*(1/beta1)*(l/w)^2*(E/G); 
%T2 = T1; 
T2 = (1/12)*(1/beta2)*(l2/w2)^2*(E/G); 
 
%Flexural stiffness in x-direction 
S1 = (1/3)*L1^3*(cos(alpha))^2+R1*L1*(1-sin(alpha))*(L1*cos(alpha)+R1*(1-
sin(alpha)))... 
    +R1^3*(2.3562-1.5*alpha-2*cos(alpha)+(1/4)*sin(2*alpha)); 
 
S2 = H^2*L3+(1/3)*L2^3*(cos(alpha))^2+L2*(R1*(1-
sin(alpha))+L1*cos(alpha))^2+R2*(C1^2*(1.5708-alpha)... 
    +2*C1*R2*cos(alpha)+C2*R2^2)+L2^2*cos(alpha)*(R1*(1-
sin(alpha))+L1*cos(alpha)); 
 
S11 = L4+R1*C2+L1*(sin(alpha))^2+f*(E/G)*(R1*C3+L1*(cos(alpha))^2); 
 
S12 = L3+R2*C2+L2*(sin(alpha))^2+f*(E/G)*(R2*C3+L2*(cos(alpha))^2); 
 
Kx = 1/(((1/E)*((S1/I1)+(S2/I2)))+((1/E)*((S11/A1)+(S12/A2)))) 
 
%Flexural stiffness in y-direction 
 
S3 = (1/3)*(L4^3+(C4+L1)^3-
C4^3)+R1*(R1^2*C3+R1*L4*(cos(alpha))^2+L4^2*C2)... 
    +T1*(L1*C5^2+R1*(L4^2*C3+R1*L4*(1-sin(alpha))^2)... 
    +R1^3*(1.5708-alpha-2*cos(alpha)+C2)); 
 
S4 = (1/3)*((L1+L2)^3-L1^3+(C6+L3)^3-C6^3)+C4*(C4*L2+L2^2+2*L1*L2)... 
    +R2*((C1^2+C6^2)*(.7854-alpha/2)+(1/4)*(C6^2-
C1^2)*sin(2*alpha)+C1*C6*(cos(alpha))^2)... 
    +T2*(C5^2*L2+H^2*L3+R2*((C1^2+C6^2)*(.7854-
alpha/2)+(1/4)*sin(2*alpha)*(C1^2-C6^2)... 
 
 144 
 
    -2*C6*R2*(1-sin(alpha))+C1*cos(alpha)*(2*R2-C6*cos(alpha))+R2^2*(1.5708-
alpha))); 
 
S13 = f*(E/G)*(L1+L4+R1*(1.5708-alpha)); 
 
S14 = f*(E/G)*(L2+L3+R2*(1.5708-alpha)); 
 
Ky = 1/(((1/E)*((S3/I3)+(S4/I4)))+((1/E)*((S13/A1)+(S14/A2)))) 
 
%DIAGONAL FLEXURAL STIFFNESS 
Lpack = .9; 
Wpack = .9; 
r = Lpack/Wpack; 
 
Kd = (1.414*Kx*Ky)/(sqrt(Kx^2+Ky^2)) 
 
  
 
 
 145 
 
Appendix B ? MATLAB codes for Electrical Data Analysis 
Thermal Cycle Tagging Code 
 
This code was created in order to apply cycle counts to the input data.  Raw data from 
the data logger array came only with time stamps.  Based on the collected 
temperature probe readings from the raw data, this program provided a data stamp 
with the cycle number of the resistance readings. 
 
clear all 
clc 
 
% This part of the program is used to tag the data from the loggers with a 
% the correct number of passive cycles experienced by the solder joints 
 
[NUMERIC,TXT,RAW]=XLSREAD('G:\Research\Thesis Materials\The REAL 
Deal\450 to 500 cycles AuSn and SAC305\Datalogger 3\Data INSTR 3 7_7_2008 
15_44_57 3_tabs'... 
    ,'Data INSTR 3 7_7_2008 15_44_57 ','ak45:ak5400'); 
 
[m,n]=size(NUMERIC); 
count = zeros(1,m); 
count(1)=1; 
tot_cycle=1; %this sets a cycle counter that displays the cycle BEING completed... 
             %not the cycle that HAS BEEN completed! 
thres=85; %difference that needs to be exceeded between true cycle counts (this... 
            %is the number of counts that occur between true full cycles... 
            %for example, a cycle that takes 90 min long sampled every 10 
            %seconds would require a threshold of approx. 540 
            %(90min*(60sec/1min)*(1sample/10sec)=540samples 
for i=1:m 
    [Y1,I1]=max(count); 
    if NUMERIC(i)>=45 & NUMERIC(i)<=55 & (i-I1)>thres 
        count(i)=1; 
        count(I1)=0; 
        tot_cycle=1+tot_cycle;       
    end 
    cycle(i)=tot_cycle; 
end 
 
A = [NUMERIC, cycle']; 
cycles = cycle'; 
 
num2clip(cycles) 
tot_cycle 
 
 146 
 
Automatic Resistance Signature Plotting Code 
 
This code was created in an effort to simplify the characterization of electrical chain 
resistance signatures.  The plots automatically generated and saved by this code 
provided snapshots of resistance signatures that could be quickly looked over for 
abnormal behavior.  The creation of this code greatly facilitated work with electrical 
data during this investigation. 
 
clear all 
clc 
 
[A,TXT,RAW]=XLSREAD('C:\Documents and Settings\toberc\My 
Documents\Research\The REAL Deal\400-405 cycles P4, 800-810 cycles ELEC  
Scruff\McGruff\is\the\coolest\dog\that\ever\has\lived\what\a\terrific\terrier!\ 
SAC (Data Salvaging)\DL4\Data INSTR 4 9_27_2008 16_14_46 4_tabs'... 
    ,'Data INSTR 4 9_27_2008 16_14_46','d74:bo1349'); 
 
[m n] = size(A); 
x = linspace(1,m,m); 
 
for i=1:n 
  max1 = max(A(:,i));   
  min1 = min(A(:,i)); 
  plot(x,A(:,i)) 
  hold on 
    if min1 > 0 
    axis([0 m 0 max1]) 
    hold off 
    else 
    axis([0 m min1 max1]) 
    hold off 
    end 
  % ... 
  filename=['C:\Documents and Settings\toberc\My Documents\Research\The REAL 
Deal\400-405 cycles P4, 800-810 cycles ELEC SAC (Data Salvaging)\Figures\DL4\-
0' num2str(i) '.bmp']; 
  saveas(gcf,filename,'bmp'); 
end 
  
%saveas(gcf, 'G:\Oberc\Matlab Files\Figures\'num2str(i), 'bmp') 
 
 147 
 
Works Cited 
 
[1] ?Performance Test Methods and Qualification Requirements for Surface Mount 
Solder Attachments.?  IPC-9701, Jan. 2002, pp. 13-14. 
 
[2] R. Kirschman, ?Extreme Temperature Electronics.?  October 2008. Available: 
http://www.extremetemperatureelectronics.com/index.html. 
 
[3] P. Neudeck, D. Spry, G. Beheim, and R. Okojie, ?6H-SiC Transistor Integrated 
Circuits Demonstrating Prolonged Operation at 500?C,? Proceedings of International 
Conference on High Temperature Electronics, Albuquerque, NM, May 12-15, 2008. 
 
[4] M. Levinshtein, et al., "High-temperature (up to 773K) operation of 6-kV 4H-SiC 
junction diodes." Solid-State Electronics 49: no. 7, July 2005, pp. 1228-1232. 
 
[5]  F. McCluskey, R. Grzybowski, and T.Podlesak,  ?High Temperature 
Electronics.?  CRC Press: Boca Raton, 1997,  p. 9 
 
[6] R. German, ?Liquid Phase Sintering.?  Plenum Press: NewYork, 1985, p.164. 
 
[7] US provisional patent serial number: US60/891,763; ?Ag-In Transient Liquid 
Phase High Temperature Solder Paste Attach.?  Patent filed January, 2008. 
 
[8] P. Quintero, ?Development of a Shifting Melting Point Ag-In Paste Via Transient 
Liquid Phase Sintering for High Temperature Environments.?  University of 
Maryland, PhD Thesis, 2008.  
 
[9] A. Choubey, et al., ?Lead-Free Assemblies for High Temperature Applications.?  
Proceeding of IMAPS International Conference on High Temperature Electronics 
(HITECH 2006), May 2006,  pp. 384-389. 
 
[10] A. Choubey, et al.,  ?Intermetallics Characterization of Lead-free Solder Joints 
under Isothermal Aging.?  Journal of Electronic Materials: Vol. 37, No. 8, August 
2008, pp. 1130-1138. 
 
[11] A. Choubey, et al., ?Effect of Aging on Pull Strength of SnPb, SnAgCu, and 
Mixed Solder Joints in Peripheral Surface Mount Components.? University of 
Maryland CALCE Center, Internal Research Paper. 2007.  
 
[12] I. Kim, et al.,  ?A Comparative Study of The Fatigue Behavior of SnAgCu and 
SnPb Solder Joints.?  Key Engineering Materials: Vols. 297-300, 2005, pp. 831-836. 
 
[13] J. Suhling, et al., ?Thermal cycling reliability of lead free solders for automotive 
applications.?  The Ninth Intersociety Conference on Thermal and Thermomechanical 
Phenomena in Electronic Systems. Vol.2, 2004, pp. 350-357.   
 
 148 
 
 
[14] H. Ma, et al., ? Reliability of the aging lead free solder joint.?  2006 Proceedings 
56th Electronic Components & Technology Conference.  June 2006, p. 849 ? 864 
 
[15] W. Brown and R. Ulrich, eds.  Advanced Electronic Packaging.  IEEE Press 
Series on Microelectronic Systems. A John Wiley & Sons, Inc: Piscataway, NJ. 2006. 
 
[16] M. Ishikawa, et al., ?Application of Gold-Tin Solder Paste for Fine Parts and 
Devices.?  2005 Proceedings of 55th Electronic Components and Technology 
Conference.  June 2005, p. 701-709. 
 
[17] C. Lee and G. Matijasevic.  ?Highly reliable die-attachment on polished GaAs 
surfaces using gold-tin eutectic alloy.?  IEEE Transactions on Components, Hybrids, 
and Manufacturing Technology: Vol. 12, No. 3, Sept. 1989.  
 
[18] G. Matijasevic and C. Lee, ?Void-free Au-Sn eutectic bonding of GaAs dice and 
its characterization using scanning acoustic microscopy.?  Journal of Electronic 
Materials: Vol. 18, No. 2, Mar. 1989, p. 327-337. 
 
[19] G. Matijasevic, C. Wang, and C. Lee, ?Void Free Bonding of Large Silicon Dice 
Using Gold-Tin Alloys.?  IEEE Transactions on Component, Hybrids, and 
Manufacturing Technology, Vol. 13, No. 4, Dec. 1990, p. 1128-1134. 
 
[20] K. Hungar and W. Mokwa, ?Gold/tin soldering of flexible silicon chips onto 
polymer tapes.?  Journal of Micromechanics and Microengineering: Vol. 18,  June 
2008. 
 
[21] X. Qiao and S. Corbin, ?Development of Transient Liquid Phase Sintered 
(TLPS) Sn-Bi Solder Pastes.?  Materials Science and Engineering A:  Vol. 283, 2000, 
p. 38-45. 
 
[22] S. Corbin, ?High Temperature Variable Melting Point Sn-Sb Lead-free Solder 
Pastes using Transient Liquid Phase Powder Processing.?  Journal of Electronic 
Materials: Vol. 34, 2005, p. 1016-1025.   
 
[23] M. Palmer, N. Erdman, and D. McCall,  ?Forming high temperature solder joints 
through liquid phase sintering of solder paste.?  Journal of Electronic Materials: Vol. 
28, No. 11,  1999. 
 
[24] C. Gallagher, G. Matijasevic, and J. Maguire, ?Transient Liquid Phase Sintering 
Conductive Adhesives as Solder Replacements.?  Electronic Components and 
Technology Conference: 1997. 
 
[25] C. Shearer, B. Shearer, G. Matijasevic, and P. Gandhi, ?Transient Liquid Phase 
Sintering Composites: Polymer Adhesives with Metallurgical Bonds.?  Journal of 
Electronic Materials: Vol. 28, No. 11, 1999. 
 
 149 
 
 
[26] C. Lee and W. So, ?High temperature silver-indium joints manufactured at low 
temperatures.?  Thin Solid Films: 366, 2000, p. 196-201. 
  
[27] R. Chuang and C. Lee, ?Silver-Indium Joints Produced at Low Temperature for 
High Temperature Devices.?  IEEE Transactions on Components and Packaging 
Technologies: Vol. 25, No. 3, Sept. 2002. 
 
[28] R. German and J. Dunlap, ?Processing of Iron-Titanium Powder Mixtures by 
Transient Liquid Phase Sintering.?  Metallurgical Transactions A: Vol. 17A, Feb. 
1986,  p. 205-13. 
 
[29] M. Hou and T. Eagar, ?Low Temperature Transient Liquid Phase (LTTLP) 
Bonding for Au/Cu and Cu/Cu Interconnections.?  Journal of Electronic Packaging: 
Vol. 114, Dec. 1992, p. 443-47. 
 
[30] M. Hansen and K. Anderko. Constitution of Binary Alloys, McGraw-Hill, New 
York, 1958.  
 
[31] G. Vander Voort, ?Metallography, Principles and Practice.? ASM International: 
Ohio, 1999. 
 
[32] Point-counting Technique.  ASTM Specification E562.   
 
[33] ?ImageJ Analyze Menu.?  Image J: Image Processing and Analysis in Java. 
2008. Available:  http://rsbweb.nih.gov/ij/docs/menus/analyze.html. 
 
[34] J. Johnson, J. Brezovsky, and R. German, ?Effects of Tungsten Particle Size and 
Copper Content on Densification of Liquid-Phase-Sintered W-Cu.?  Metallurgical 
and Materials Transactions A: Vol. 36A, Oct. 2005, p. 2807-14. 
 
[35] B. Rabin and R. Wright, ?Synthesis of Iron Aluminides from Elemental 
Powders: Reaction Mechanisms and Densification Behavior.?  Metallurgical 
Transactions A: Vol. 22A, Feb. 1991, p. 277-86. 
 
[36] ?Process Change Notification PCN072 Update.? Altera Corperation. Available: 
http://www.altera.com/literature/pcn/pcn0702.pdf.  p. 6. 
 
[37] ?EnviroMark 907 Lead-Free No-clean Solder Paste.?  Kester Product Data 
Sheet. 
 
[38] C. Harper. Electronic Packaging and Interconnection Handbook.  McGraw-Hill 
Professional, 2004.  ISBN 0071430482.  Chap 5, p. 26. 
 
[39] ?NC-SMQ51SC Solder Paste.?  Indium Corporation of America Data Sheet. 
 
 
 150 
 
[40] ?Eutectic Gold-Tin Solder Paste.?  Indium Corporation of America Data Sheet.  
 
[41] ?Stencil Design Guidelines.? IPC-7525: May 2000. 
 
[42] Omega LMF-3550 Product Manual.  Accessed 2008. Available: 
http://omega.com/manuals/manualpdf/M2159.pdf.  p. 43. 
 
[43] ?EC12 Product Manual.? Sun Electronics. 
 
[44]  ?Guideline for Accelerated Reliability Testing of Surface Mount Solder 
Attachments.?  IPC-SM-785: Nov. 1992, p. 33. 
 
[45] R. Abernethy, ?The New Weibull Handbook 3rd Edition.?  1998.  p. 31. 
 
[46] ?Environmental and Endurance Test Methods for Surface Mount Solder Joint: 
Pull Strength Test.? IEC 91/579/CDV:  Jan. 2006.    
  
[47] J. Clech and J. Augis, ?Surface Mount Attachment Reliability and Figures of 
Merit for Design for Reliability in Solder Joint Reliability: Theory and Applications.? 
J. Lau, Ed.  New York: Van Nostrand Reinhold, 1991, pp. 591-94. 
 
[48] R. Kotlowitz, ?Comparative Compliance of Representative Lead Designs for 
Surface Mounted Components.?  IEEE Transactions on Components, Hybrids, and 
Manufacturing Technology: Vol. 12, No. 4, Dec. 1989. 
 
[49] R. Kotlowitz and L.Taylor, ?Compliance Metrics for the Inclined Gull-Wing, 
Spider J-Bend, and Spider Gull-Wing Lead Designs for Surface Mount Components.?  
IEEE Transactions on Components, Hybrids, and Manufacturing Technology: Vol. 
14, No. 4, Dec. 1991. 
 
[50] R. Kotlowitz and A. Rizzo, ?Comparative Compliance of a Representative 
Surface Mount Leadless Solder Connection and Commercial Lead Designs.?  44th 
Proceedings of Electronic Components and Technology Conference: 1994. 
 
[51] S. Timoshenko. Strength of Materials 3rd Edition.  Malabar: Robert E. Krieger 
Publishing Company, 1958, p. 60. 
 
[52] Matweb Material Property Data. Available: http://www.matweb.com/. 
 
[53] CQFP Lead Stiffness Agreement Reference:  Performance Characteristics of IC 
Packages: Ch. 4, Intel, p. 4-32. Available: www.intel.com/design/packtech/ch_04.pdf. 
 
[54] H. Qi, et. al., ?Analysis of Solder Joint Failure Criteria and Measurement 
Techniques in the Qualification of Electronic Products.?  IEEE Transactions on 
Components and Packaging Technologies: Vol. 31, No. 2, June 2008, p. 469-477. 
 
 
 151 
 
[55] E. Kreyszig. Advanced Engineering Mathematics, 9th Edition.  John Wiley & 
Sons, Inc., 2006, p. 1053.   
 
[56] B. Yan, et. al., ?The Microstructure of Eutectic Au-Sn and In-Sn Solders on 
Au/Ti and Au/Ni Metallizations during Laser Solder Bonding Process for Optical 
Fiber Alignment.?  IEEE 2006 7th International Conference on Electronics Packaging 
Technology, 2006.