Characterization of Low Melting Temperature, Low-Ag, Bi-Containing, Pb-Free Solder Alloys

Transcription

1 Characterization of Low Melting Temperature, Low-Ag, Bi-Containing, Pb-Free Solder Alloys by Eva Kosiba A thesis submitted in conformity with the requirements for the degree of Masters of Applied Science Materials Science and Engineering University of Toronto Copyright by Eva Kosiba 2016

2 Characterization of Low Melting Temperature, Low-Ag, Bi- Containing, Pb-Free Solder Alloys Abstract Eva Kosiba Masters of Applied Science Materials Science and Engineering University of Toronto 2016 Restrictions of lead in solder lead to adoption of SAC305 in consumer products. While high reliability applications use SnPb, supply constraints are driving the adoption of a replacement. SAC305 has reliability concerns related to elevated process temperatures and the formation of Ag 3 Sn. Reliability performance of three low-ag, Bi-containing, low melting temperature alloys were compared to SAC305. All three alloys under test performed as well or better for consumer applications. Drop testing and accelerated thermal cycling revealed no differences that would preclude use of these alloys in production. They allow for the use of lower T g printed wire boards materials, which have been shown reliable. These alloys show promise for high reliability applications. In accelerated thermal cycling, all alloys outperformed the circuit boards. Bi precipitation resulted in less degradation to the bulk microstructure. Bi did not impact the IMC formation or growth, a small amount of Ag mitigated growth of Cu 3 Sn. ii

3 Acknowledgments I would like to express my gratitude to Professor Doug Perovic, Dr Polina Snugovsky and John McMahon for their guidance, generosity and support. Their knowledge of various aspect of this research was an inspiration. I would also like to thank Lorna Devereux and Marianne Romansky for reading through the many revisions and offering clarity to my often confusing prose. The following people provided much laboratory assistance and support for which I am grateful: Russell Brush, Subramaniam Suthakaran, Michael Thomson, Hissan Syed, André Delhaise and Salvatore Boccia. I would also like to thank Professor Erb and Professor Wang for agreeing to participate on the review committee on short notice. Finally I would like to acknowledge the support of the Quality and Reliability Laboratory at Celestica and the Department of Materials Science and Engineering at the University of Toronto. iii

4 Table of Contents Abstract... ii Acknowledgments... iii List of Tables... vii List of Figures... ix List of Acronyms... xiii Chapter 1 Background Introduction Need for New Low Melt Solder Legislative Changes SAC305 and Current Issues with Existing Alloys Search for Replacement Solder Binary Alloys Ternary and higher order alloy options Celestica s Low Melt Solder Research Program Phase 1: Alloy Selection Requirement Formulation Literature Search and Phase Diagram Analysis Preliminary Alloy Selection Metallurgical Analysis: DSC and Microstructural Evaluation Phase 2: Manufacturing Feasibility Study Phase 3: Screening Experiments Aerospace and Defense, Screening Experiment Structure Microstructural Assessment Accelerated Thermal Cycling (ATC) Vibration Summary of Findings Objective of Thesis: Low Melt Solders for Consumer Sector References Chapter 2 Solder Joints after Reflow (As Manufactured) Introduction Bulk Solder Microstructure Solidification Process Cu Dissolution in Molten Solder Bulk Solder Solidification Undercooling During Solidification Formation of Facetted IMCs in Bulk Solder Bi in Solution and as a Precipitate Sn-Cu Reaction and Formation of Interfacial IMC Cu-Ni-Sn Interface iv

5 5 Experimental Set Up Test Vehicle Assembly Test Matrix Test Method Microstructural Evaluation Comparison of Bulk Microstructure Bulk Microstructure of QFP Solder Joints Bulk Microstructure of BGAs Comparison of Interfacial IMC Layers Summary of Findings References Chapter 3 Accelerated Thermal Cycling Accelerated Testing for Reliability Analysis Microstructural Evolution Changes to Bulk Solder Changes to Interfacial IMC during Accelerated Thermal Cycling Effects of Bi Experimental Setup Materials Test Vehicle Test Strategy C Thermal Cycling Harsh Environment (-55 C to 125 C) Thermal Cycling Post ATC Evaluation and Failure Analysis Results Reliability and Failure Analysis Results C to 100 C Accelerated Thermal Cycling C to 125 C Accelerated Thermal Cycling Microstructure Evaluation Bulk Microstructure IMC Growth during Thermal Cycling Summary of Findings and Conclusions Findings Based on Reliability Data Findings Based on Microstructural Observations References Chapter 4 Tin Whisker Testing Introduction Whisker Growth Kinetics Sources of Compressive Stress Morphology of Sn Whiskers The Effects of Bi in Solder on Whisker Formation Experimental Set Up Materials v

6 2.2 High Temperature High Humidity Thermal Shock Post Exposure Evaluation Results High Temperature High Humidity Results Thermal Shock Results Summary of Findings and Conclusions References Chapter 5 Mechanical (Drop) Shock Testing Introduction Experimental Set Up Materials Test Vehicle Assembly Test Strategy Reliability Results Failure Analysis and Microstructural Evaluation Dye and Pry Procedure Failure Isolation Procedure Evaluation of High T g Board after Drop Testing Evaluation of Normal Tg Board after Drop Testing Summary of Findings and Conclusion References Chapter 6 Summary of Findings and Conclusions Chapter 7 Future Work Celestica/Indium Sponsored Whisker Resistant Solder Paste ReMap M1: Lower Temperature Soldering Alloys with Improved Mechanical and Thermal Fatigue Reliability ReMap M2: Alloys, Board and Component Surface Finish Interactions with Reduced Propensity for Whisker Growth ReMap M3: Aging Effect of New Lead-Free Materials on Reliability References vi

7 List of Tables Table 1: Sn-based eutectic alloys... 8 Table 2: Preliminary alloy selection 10, Table 3: DSC analysis of proposed alloys with 75% SAC Table 4: Alloys for screening experiments Table 5: Latin square of paste and finish, and board materials Table 6: Build matrix showing number of assemblies Table 7: Resultant BGA interconnect composition Table 8: ATC failures on 170 C T g board material Table 9: Reliability results on 170 C T g boards after ATC Table 10: ATC failures on 150 C T g board material Table 11: Reliability results on 150 C T g boards after ATC Table 12: Vibration failures after 2G testing Table 13: Vibration failures after 5G testing Table 14: Consumer alloys under test Table 15: Build matrix for as assembled analysis Table 16: Table of components evaluated by cross sectioning Table 17: Composition of QFPs (wt%) Table 18: Experimental composition of BGA solder joint Table 19: Theoretical composition of BGA solder joint Table 20: Results of ANOVA test for equal variance and compare means of IMC thickness of the BGA IMC layer Table 21: Results of ANOVA test for equal variance and compare means of IMC thickness of the QFP IMC layer Table 22: IMC type Table 23: Sample of temperature cycling requirements Table 4-1 in IPC-9701A Table 24: Worst case use environments of SMT Table 25: Build matrix for ATC testing Table 26: Monitored components for ATC testing Table 27: Test matrix for 0 C to 100 C ATC Table 28: Test matrix for C ATC Table 29: Summary of QFP failures after 0 C to 100 C ATC Table 30: Summary of BGA failures after 0 C to 100 C ATC Table 31: Summary of QFP failures after -55 C to 125 C ATC Table 32: Summary of BGA failures after -55 C to 125 C ATC Table 33: Results of Levine-Test to compare the variance ( ) of IMC measurement and 2-sided t-test to compare the means (µ) of IMC measurements Table 34: Results of Levine-Test to compare the variance ( ) of Cu 3 Sn measurement and 2-sided t-test to compare the means (µ) of Cu 3 Sn measurements Table 35: Results of ANOVA test to equal variance and compare means Table 36: CTE values for common materials in solder joints Table 37: JESD22A test conditions Table 38: Alloys screened for whisker growth vii

8 Table 39: Summary of whisker growth after 1610 cycles thermal shock Table 40: Results of ANOVA test for equal variance and compare means of IMC thickness of the QFP IMC layer after 1610 cycles thermal shock Table 41: Build matrix for drop shock testing Table 42: Monitored components for drop testing Table 43: Drop test results Table 44: Failures of BGA on High T g boards Table 45: Failures of BGA on Normal Tg boards Table 46: Solder paste test matrix Table 47: ReMap M2 test matrix Table 48: ReMap M3 test matrix viii

9 List of Figures Figure 1: Schematic of a BGA solder joint... 1 Figure 2: Image of a QFP solder joint Figure 3: Pad crater defect seen in a) cross section indicated by arrows, and b) viewed from above... 4 Figure 4: Ag 3 Sn platelets seen in a) cross sectioned solder joint and b) SEM image of solder joint after removal of Sn phase... 5 Figure 5: Sn Whisker growing from SAC305 solder... 6 Figure 6: Sn-Cu phase diagram... 7 Figure 7: a) SAC ternary phase diagram and b) Sn-rich corner of SAC ternary phase diagram Figure 8: Head in pillow defect Figure 9: a) gull wing leaded solder joint, b) BGA solder joint Figure 10: Solder joint formed with leadless component Figure 11: SnBi phase diagram Figure 12: a) SnAgBi ternary phase diagram and b) Sn-rich corner of SnAgBi phase diagram Figure 13: a) SnCuBi ternary phase diagram and b) Sn-rich corner of SnCuBi phase diagram Figure 14: Example of ternary phase diagram Figure 16: DSC scan of Sn20%Bi 10, Figure 17: Microstructure of solder joint formed with SAC305 solder ball and a no-ag solder paste alloy Figure 18: Honeywell test vehicle, medium complexity board Figure 19: Example of wetting on OSP, QFP Figure 22: New approach proposing to use solder with no or low-ag content Figure 23: Typical reflow profile for SAC Figure 24: Copper dissolution of BGA solder joint after a) 1 replacement b) 3 replacements and c) 5 replacements Figure 25: SnAgCu phase diagram with SAC305 equilibrium solidification path Figure 26: A 3D phase diagram of Sn rich portion of Sn-Ag-Cu ternary system Figure 27: SAC solder joint viewed a) in cross section and b) after selective electrochemical etching to remove Sn phase Figure 28: Solder joint between Sn-based solder and Cu a) rod shaped Cu6Sn5 ( ) Figure 29: SnPb solder jointed to Cu substrate a) showing Kirkendall voids b) showing Pb phase pooling at IMC/solder interface (cross section by Zohreh Bagheri). 51 Figure 30: Typical morphologies of Cu6Sn5 grains formed on single crystal Figure 31: IMC formed between SAC solder and Ni(P)Au Figure 32: Test vehicle Figure 33: Area for EDX compositional analysis of a) bulk BGA solder joint, Figure 34: Optical images QFP a) SAC305 b) Senju M42 c) Sunrise and d) Sunflower. 59 Figure 35: Bi present in a) eutectic colonies in Sunrise and b) throughout bulk solder (eutectic colonies and Sn dendrite arms) of Sunflower ix

10 Figure 36: QFP solder joint as seen a) optically b) SE SEM and c) BSE SEM Figure 38: SEM image of Sunflower BGA solder joint Figure 39: EDX mapping of intermetallic in Sunflower showing a) image Figure 42: Polarized light images BGA a) SAC305 b) Senju M Figure 44: IMC measurements of QFP solder joints after reflow (U2) Figure 45: IMC bond layers formed on a) QFP components and b) BGA components Figure 46: IMC formed on BGA at component side Figure 47: Typical morphology of a) Cu 6 Sn 5 IMC as formed on a QFP solder joint with Sunflower b) and c) (Cu,Ni) 6 Sn 5 and Ni 23 Cu 33 Sn 44 IMCs respectively both formed on a BGA with SAC305 solder paste Figure 49: Hysteresis loop for thermal cycle Figure 50: SnPb solder a) before testing and b) after 3000 cycles -55 C to 125 C Figure 51: SAC305 solder a) before testing and b) after 3000 cycles -55 C to 125 C Figure 52: SAC305 after ATC shown with a)polarized light and b)ebsd mapping Figure 53: Test vehicle with monitored components Figure 54: Card set up in thermal cycling chamber Figure 55: Chamber profile for 0 C to 100 C thermal cycling Figure 57: Weibull plots of SAC305 QFP solder joints after 6010 Cycles 0 to 100 C comparing High T g to Normal T g boards Figure 58: Probability Density Function for SAC305 QFP solder joints after 6010 Cycles 0 to 100 C comparing High Tg to Normal Tg boards Figure 59: QFP fracture of SAC305 on High T g board a) optically and b) cross section 97 Figure 60: Fracture surface of QFP solder joint with SAC305 on High T g board after 6010 cycles Figure 61: Fracture initiation in QFP a) Sunflower b) Senju M42 after 6010 cycles Figure 62: Weibull plots of BGA solder joints on Normal T g boards after 6010 Cycles 0 to 100 C cycles comparing Four Alloys Figure 63: BGA failures after 0 to 100 C ATC a) failure in board material by via plating crack and b) partial failure through bulk solder (SAC305) near component side IMC Figure 64: Weibull plots of QFP failures on High T g boards after 1000 cycles C 103 Figure 65: Probability Density Function for QFP failures on High T g boards after 1000 cycles C Figure 66: Weibull plots of QFP failures on Normal T g boards after 1000 Cycles C Figure 67: QFP176 fractures in a) Sunflower and b) Senju M42 after 1000 cycles Harsh testing Figure 68: BGA failures after -55 to 125 C ATC failures in board material by via plating crack in a) boards built with Sunrise and b) boards built with Sunflower Figure 70: Weibull plots of BGA solder joints after 1000 Cycles -55 to 125C cycles comparing two board materials Figure 71: Weibull plot for QFP solder joints on Normal Tg boards comparing two test conditions x

11 Figure 72: Probability Density Function for QFP failures on Normal Tg boards comparing two test conditions Figure 73: SAC305 at Time 0 a) optically and b) SEM image and after b) 3148 cycles and Figure 74: Senju M42 at 1000x after a) 3148 cycles and b) 6010 cycles Figure 75: Sunrise after 6010 cycles C at a) 500x and b) 1000x Figure 76: Sunflower at a) Time 0 and b) after 6010 cycles C at 500x Figure 77: IMC layers formed on QFP176 solder joints between the board side Cu layer and solder paste a) sunrise and b) SAC305 after 438 cycles of Harsh thermal cycling. Location 1 shows the Cu 3 Sn layer, location 2 shows the Cu 6 Sn 5 layer 112 Figure 78: IMC growth at the board side of QFP during thermal cycling Figure 80: Interval plot of IMC thickness at the board side of QFP after ATC Figure 81: Interval plot of Cu 3 Sn thickness at board side of QFP after ATC Figure 83: Main effects plot of IMC thickness at the board side QFP during C. 119 Figure 85: Main effects plot of Cu3Sn thickness board side of QFP during C Figure 86: Interaction plot of IMC thickness at the board side of QFP after ATC Figure 88: Pb "cap" on whisker from SnPb component finish Figure 89: a) Schematic for a typical solder joint of a leaded component using SnPb solder, b) cross section showing solder joint formed with Pb-free solder Figure 90: Cyclic Dynamic Recrystallization resulting in whisker formation Figure 91: Source of compressive stress contributing to whisker growth Figure 92: Ag3Sn oxide zone with whisker Figure 93: Whisker morphology a) long, thin whiskers and b) short, kinked whiskers Figure 94: Cross sections of plating surface made with SEM FIB of a) Sn and b) SnPb Figure 95: Samples in HTHH chamber Figure 96: Two stage, air to air, chamber for thermal shock testing Figure 97: Thermal shock temperature profile Figure 98: Schematic showing locations on lead where whiskers formed Figure 99: Two adjacent leads with Sunflower solder paste after 1610 thermal shocks 141 Figure 100: Whisker growth on SAC305 after 1610 thermal shocks. a) and b) whisker growth in location 4 c) massive deformation and d) whisker growth in location Figure 101: Whisker growth on Senju M42 after 1610 thermal shocks. a) and b) location 4 with short, thick whisker surrounded by many, very short whisker nucleation sites c) and d) short, thick whiskers growing at site of contamination in location Figure 102: Whisker growth on Sunrise after 1610 thermal shocks. a) and b) location 2 with short, thick whisker c) location 4 and d) location 1 with many, very short whisker nucleation sites short Figure 103: Whisker growth on Sunflower after 1610 thermal shocks. a) and b) location 1 with short, thick whisker c) some, very short whisker nucleation sites at location 4 and d) longest whisker observed at location xi

12 Figure 104: Whisker growing from Sunflower at grain boundary of recrystallized grains 146 Figure 105: Hillock growing from Sunrise after thermal shock Figure 106: Senju M42 bulk recrystallization Figure 107: IMC measurements of QFP solder joints after 1610 cycles thermal shock (U1) Figure 108: Interval plot of IMC thickness at the lead side of QFP after thermal shock 149 Figure 109: IMC layer at lead with a) SAC305 and b) Sunrise after 1610 cycles of thermal shock Figure 111: Bi content at various locations of a Sunrise solder joint Figure 112: Cross section of Sunflower showing Bi accumulating at grain boundaries 152 Figure 113: Test vehicle with monitored components Figure 114: Solder mask defined vs. non-solder mask defined Figure 115: Example of accelerometer secured with RTV silicone Figure 116: Test set up Figure 117: Target pulse shock defined by JESD22-B110 service condition B Figure 118: Sample of pulse shock achieved during test Figure 119: Individual value plot of drops to fail, BGA (SAC305 + alloy) (U205) Figure 120: Individual value plot of drops to fail, QFP (U2) Figure 121: Failure modes of solder joint as defined by IPC/JEDEC Figure 122: BGA failure by pad cratering, Sunflower on High T g boards Figure 123: D&P mapping of Senju M42 on High T g board, with images of Figure 124: Schematic of Quad-Flat-Package (QFP) Figure 125: QFP failures in Sunflower on High T g boards Figure 126: D&P mapping of Sunrise on Normal Tg board, with images of Figure 127: BGA failure in Sunflower on Normal T g boards Figure 128: QFP failures in Sunflower on Normal T g boards Figure 129: QFP failure in Sunrise on Normal T g board Figure 130: Orchid at a) Time 0 and b) after 3000 cycles ATC Figure 131: Sunrise solder joint after 1000 hrs HTHH followed by 2 years ambient storage xii

13 List of Acronyms AF: Accelerating Factor Ag: Silver ALT: Accelerated Life Testing ANOVA: analysis of variance ATC: Accelerated Thermal Cycling ATHH: Ambient Temperature High Humidity Au: Gold BGA: Ball Grid Array Bi: Bismuth BSE: Backscatter Electron CBGA: Ceramic Ball Grid Array Cd: Cadmium CTE: Coefficient of Thermal Expantion Cu: Copper D&P: Dye and Pry DI: De-Ionized DRX: Dynamic Recrystallization DSC: Differential Scanning Calorimeter EDX: Energy Dispersive X-Ray Spectroscopy ENEPIG: Electroless Nickel Electroless Palladium Immersion Gold ENIG: Electronless Nickel Immersion Gold FA: Failure Analysis FR4: flame retardant level H 0 : null hypothesis H a : alternate hypothesis HiP: Head in Pillow HTHH: High Temperature High Humidity ICT: In Circuit Testing IMC: Intermetallic Compound In: Indium inemi: International Electronics Manufacturing Initiative JEDEC: Joint Electron Device Engineering Council LQFP: Low Profile Quad Flat Pack MLF: Micro Lead Frame MTTF: Mean Time To Failure Ni: Nickel NTC: Number of Thermal Cycles OSP: Organic Solderability Preservatives Pb: Lead Pb-free: Lead free PBGA: Plastic Ball Grid Array PDF: probability density function PWB: Printed Wire Board xiii

14 QFP: Quad Flat Package RoHS: Restriction of Hazardous Substances SAC: Tin-Silver-Copper (SnAgCu) SE: Secondary Electron SEM: Scanning Electron Microscope SiC: Silicon Carbide SMT: Surface Mount Technology Sn: Tin SnPb: Tin Lead (usually eutectic Tin Lead) T g : Glass Transition Temperature T h : Homologous Temperature T m : melting temperature V acc : Acceleration Voltage Zn: Zinc xiv

15 Chapter 1 Background 1 Introduction A solder joint is formed when two metal conductors with high melting temperatures, usually copper (Cu) which melts at 1085 C, are joined with another metal having a low melting temperature. Tin (Sn) comprises the majority of materials used as solder because of its low melting temperature (231.9 C), and its unique ability to form intermetallics with many different joining substrates. 1 In the case of electronic surface mount technology (SMT), which is the focus of this study, these bonds act as electrical, mechanical and thermal bonds. The solder joint is comprised of two interfaces between the copper and solder as well as the bulk solder, which is filler material between the two interfaces. The interfaces are chemical bonds between the Sn from the solder and the Cu from the connecting conductors. 2 In some cases, an intermediary nickel (Ni) layer will also be present. The intermetallic compound (IMC) layer, which forms during the reflow and solidification processes, tends to be more brittle than either the Cu or the bulk solder. Figure 1 is a schematic of a typical ball grid array (BGA) type of component solder joint and Figure 2 is an image of a solder joint formed using a quad flat package (QFP) type of leaded component with gull winged shaped leads. Both are attached to the printed wire board (PWB) by the formation of a solder joint. The solder joints formed using these two types of components represent two of the main ways in which solder is used to form electrical contacts. Figure 1: Schematic of a BGA solder joint 3 1

16 Figure 2: Image of a QFP solder joint 3 2 Need for New Low Melt Solder 2.1 Legislative Changes In 2006 The Restriction of Hazardous Substances Directive or RoHS (2002/95/EC) took effect for electronic products put onto the market across the European Union. 4 Each of the six substances restricted within this legislation (with some exemptions) pose some degree of challenge to the electronic manufacturing industry. However, it is the ban of lead (Pb) which presents the greatest disruption to traditional manufacturing processes. Some 50 years of manufacturing knowledge based on eutectic tin-lead (SnPb) solder requires revision, update and change. Due to the global nature of the electronic manufacturing sector, this change has impacted all electronic manufacturing whether the products are destined for European markets or elsewhere. Furthermore, the European Union revised the legislation in 2011, as part of the RoHS recast, 5 and expressed an ongoing commitment to review legislation regularly, in particular material exemptions. Any replacement to SnPb solder therefore needs to consider the toxicity and downstream environmental impacts of the new materials. 2

17 2.2 SAC305 and Current Issues with Existing Alloys The search for a replacement alloy for electronic assemblies has lead to the current industry standard of tin-silver-copper alloys, (Sn-Ag-Cu or SAC); the most widely used are SAC305 and SAC405, both contain 0.5wt% Cu and 3wt% and 4wt% silver (Ag) respectively. Although these Sn-based alloys are in wide use, there remain a number of technical and environmental limitations, which necessitate the continued search for a more ideal replacement to SnPb. Some of the key issues with SAC305 and SAC405 are summarized below. 6 Both SAC305 and SAC405 are non-eutectic alloys with melting ranges of C and C respectively; considerably higher than the 183 C melting point of eutectic SnPb. Process temperatures for SAC alloys currently run at 240 C, which is higher than the typical process temperatures of SnPb (220 C). This increase in process temperature results in higher energy usage of reflow ovens, and consequently increased CO 2 emissions, negating some of the environmental benefits of a less toxic substance. The process temperature employed for manufacturing with SAC alloys is much closer to the melting range, narrowing the process window and thus requiring greater process control. While SnPb is routinely run at approximately 40 C above the solder melting temperature, such a superheat temperature would prove damaging to the overall product when applied to products using SAC alloys. Both the PWB laminate material and some of the more temperature sensitive components would not be able to withstand these higher temperatures. Kelly et al. 7 showed that in standard SAC reflow processes, the component body temperature reached as high as C. It is expected that smaller components may reach temperatures as high as 250 C. Rework processes, or any wave soldering, may require even higher temperatures, up to 260 C, for compatibility with the current Pb-free processes. 8 Component suppliers have indicated that moisture sensitivity of components may change in response to these high temperature requirements and may necessitate preconditioning before reflow. All additional processes increase production cost and potentially compromise the overall component reliability. 3

18 Moreover, the laminate materials typically used in conjunction with SnPb solders have been replaced by new materials which have a higher glass transition temperature (T g ) in order to withstand the higher processing temperatures needed for SAC alloys. The development of these new materials involved changing the epoxy resin and curing agents for additional crosslinking, as well as the inclusion of ceramic particle fillers. 9 The unintended consequence of these changes has been the introduction of a new failure mode known as pad cratering. The harder and stiffer laminate materials fail as they separate from the conductive copper traces within the PWB (Figure 3). These laminates are also more susceptible to warpage, delamination and may require additional baking processes. Figure 3: Pad crater defect seen in a) cross section indicated by arrows, and b) viewed from above This failure mode is dangerous as it may not be readily identifiable during in- circuittesting (ICT) within the factory, but may lead to reliability concerns in the field as the fracture propagates, potentially severing the copper trace and causing eventual electrical failure. This is primarily a concern for products required to withstand damage by drop shock, such as consumer electronics (e.g. mobile phones), which are prone to being repeatedly dropped over the course of useful life. It would therefore be preferable for any new alloy to be compatible with older laminate materials similar to those used with traditional SnPb solders, in which this failure mode was not prevalent. 4

19 The mechanical properties of SAC solder alloys also have disadvantages. In addition to requiring a laminate material with a higher elastic modulus (E), the alloy itself is stiffer than SnPb. 10. SnPb has an E value of 40GPa at 25 C and SAC305 has a value 50GPa at 25 C. 11 Finally, there is potential for the formation of undesirable intermetallic compounds (IMC), i.e. Ag 3 Sn (Figure 4). This intermetallic may in some cases be present as a large, sharp platelet which can potentially behave as a stress raiser or provide a path along which cracks may more easily propagate, thereby negatively impacting the materials resistance to drop shock. 6 In addition to the aforementioned thermo-mechanical issues, the high cost of Ag makes its inclusion as an alloying element undesirable for the cost-sensitive market of consumer electronics. Figure 4: Ag 3 Sn platelets seen in a) cross sectioned solder joint and b) SEM image of solder joint after removal of Sn phase 12 Finally, Sn whiskers discussed in detail in Chapter Chapter 4, present a reliability concern for all Sn-based solders. Although the presence of Pb in high concentration does not eliminate the potential for growth of Sn whiskers completely, Pb does mitigate the formation of long whiskers, and therefore is not considered a reliability risk. The move to high Sn solder compositions coincides with the further miniaturization of components, solder joints and solder joint spacing. These factors all increase the risk of whisker growth to a length at which bridging and hence short-circuiting of solder joints would be 5

20 a concern. While this risk presents a reliability concern for all products using Pb-free solders, it is of particular concern to the aerospace and defense sectors, for which these reliability concerns present an unacceptable level of risk. 10 Figure 5 shows a Sn whisker growing from SAC305 solder on the surface of a solder joint. Figure 5: Sn Whisker growing from SAC305 solder In spite of these concerns, at the time of RoHS implementation in 2006, near-eutectic SAC alloys provided the best compromise in a supply chain where Pb containing components used in printed circuit boards were still common. 2.3 Search for Replacement Solder Binary Alloys Prior to implementing SAC as an alternative to SnPb, an attempt to find an optimal replacement began with the exploration of binary alloy options. A number of binary alloys were initially considered (ideally near the eutectic point) but all had associated issues, which disqualified them as viable alternatives. 13,14 Tin (Sn) was selected as one of the two binary elements in all cases due to its low toxicity, relatively low cost, resistance to oxidation and contribution to the melting point or range of an alloy. Much of this analysis was performed by reviewing the equilibrium phase diagrams of various Sn-based alloys. Information pertaining to the temperature at 6

21 the eutectic point, as well as IMCs that may form, was weighed against a number of factors associated with the individual alloying elements. Figure 6: Sn-Cu phase diagram 15 SnCu has a eutectic point at 0.7% Cu close to the melting temperature of pure tin at approximately 227 C, (Figure 6) which is still more than 40 C higher than that the SnPb eutectic temperature. 16 While this alloy presents some advantages, the melting temperature is still too high to be considered a viable binary compound alternative. Regardless, the SnCu alloy must be considered in the development of a replacement alloy due to the interaction of solder with a copper pad on the PWB. In some cases, there will be a barrier between the copper pad and the Sn-based solder, for example a nickel (Ni) barrier layer. In other cases, the Sn-based solder will be exposed directly to a solid copper substrate, which would represent the worst case scenario in terms of copper dissolution. Whichever alloy is selected to replace SnPb, it is likely that the Sn-Cu reaction will play a large role. This will be discussed further in Chapter Chapter 2. 7

22 Other binary alloys formed with Sn were also considered and dismissed for several reasons. A summary of Sn-based eutectic alloys, their melting temperatures and their eutectic compositions is provided (Table 1). Table 1: Sn-based eutectic alloys 17 System Eutectic Temperature ( C) Eutectic Composition (wt%) Sn-Cu Sn-Ag Sn-Au Sn-Zn Sn-Pb Sn-Cd Sn-Bi Sn-In SnIn, SnAg and SnAu all present very expensive options. The alloying elements are comparatively scarce and therefore do not provide attractive alternatives to SnPb. Further, the eutectic composition of SnAg and SnAu, as well as that of SnCu is predominately Sn. These systems will therefore form intermetallic compounds within the Sn matrix, introducing inhomogeneity into the microstructure. 17 Additionally, Ag 3 Sn precipitates, which are primarily present as a network of small, rod like structures, may also form as an undesirable platelet structure as already described above and illustrated in Figure 4. SnZn and SnCd both have melting temperatures close to that of SnPb, however both have been disqualified as viable options. SnZn readily oxidizes; 10 this alloy would require processing in an inert gas environment and the addition of an excessive amount of flux. Since the oxide forms so easily, it would also potentially form a very weak solder joint. The main issue with SnCd is the toxicity of Cd; any replacement to SnPb should not present the same environmental concerns as the original alloy. 8

23 The binary SnBi eutectic alloy was initially rejected as a viable replacement to SnPb because of the possibility of forming a dangerously low melting intermetallic compound when combined with Pb. Moon et al. 19 showed that even a 0.1% Pb contamination content was sufficient to reduce the melting temperature by between 38 and 55 C. During the initial transition to lead free solders, it was anticipated that this contamination could be introduced when using SnPb tinned leaded parts within a lead-free assembly. The potential for this contamination within the supply chain was deemed to be too great and therefore all bismuth (Bi) containing alloys were avoided. As the legislation has now been in place for almost a decade, this contamination has been deemed to be less likely to occur and therefore Bi is being reconsidered a suitable alloying element. 20 Research on eutectic SnBi alloy continues. It has also been has been adopted in high volume production for special applications; flat screen televisions are built using a combination of SAC and eutectic SnBi. However, there remains concern due to the brittle nature of the eutectic and its resulting poor performance in drop shock testing; 21 this study will instead consider Bi as an alloying element within a predominately Sn-based alloy system Ternary and higher order alloy options Once it was determined that binary eutectic Sn-based alloys would not meet the needs of a SnPb replacement, ternary alloys were explored. The International Electronics Manufacturing Initiative (inemi) proposed the following criteria for the selection of a replacement alloy: Have melting point as close to Sn-Pb eutectic as possible Be eutectic or very close to eutectic Contain no more than three elements (ternary composition) Avoid using existing patents, if possible (for ease of implementation) Have the potential for reliability equal to or better than SnPb eutectic. 22 Further criteria were added for narrowing down a suitable ternary alloy. These include 22 : Liquidus temperature as close as possible to 183 C Solidus temperature as close as possible to liquidus temperature (small pasty range) 9

24 Solidus temperature significantly higher than the solder joints maximum operating temperature SnAgCu, SnBiCu, and SnAgBi were all considered. Based on the possibility of forming very low melting temperature eutectics in the presence of Pb (BiPb eutectic at C, SnPbBi ternary eutectic at 98 C), Bi was ruled out during the initial transition period and SnAgCu alloys became standard. However, now that the supply chain has been largely purged of SnPb components, Bi containing alloys are being reconsidered. Bi-rich alloys continue to be considered for some specialty applications, for example Sn-88%Bi and Sn- 59%Bi-1.2%Ag. However, as they have a melting temperature closer to 138 C, they fail the third criteria listed above for applications which may be exposed to harsh operating environments of up to 125 C. The SnAgCu eutectic system is of the composition Sn-3.5Ag-0.9Cu by wt% with a eutectic melting temperature of 217.4±0.8 C. The system forms a ternary eutectic made up of faceted Cu 6 Sn 5 and non-faceted Ag 3 Sn internetallics within a Sn matrix. The melting temperature of this eutectic is 10 C lower than that of eutectic SnCu. Near eutectic compounds of SnAgCu are currently the preferred solders in use today, specifically SAC305, SAC405 and to a lesser extent SAC105. Where cost is a driving force (i.e. consumer products) the preference is for a lower quantity of Ag. At the other end of the spectrum, Ag is limited due to the potential formation of undesirable Ag 3 Sn platelets. Figure 7 shows the phase diagram of SAC alloys and specifically the Sn-rich region, which highlights the formation of a Sn - Sn+Ag 3 Sn-Sn - Cu 6 Sn 5 ternary compound. 10

25 a. b. Figure 7: a) SAC ternary phase diagram and b) Sn-rich corner of SAC ternary phase diagram 23 SAC105 has been considered as an alternative to SAC305 or SAC405 as it reduces the amount of expensive Ag, and also reducing the possible formation of the dangerous Ag 3 Sn platelets. However, the reduction of Ag also coincides with an increase in liquidus temperature to 227 C. 6 The corresponding increase in required reflow, rework and other processing temperatures offsets any potential benefit gained by this slightly less stiff, higher fracture toughness alloy. The introduction of this alloy into the ball grid array (BGA) supply chain, where it was combined with a paste alloy of SAC305 or SnPb (in the case of mixed metallurgy builds early in the transition to Pb-Free) lead to issues of incomplete melting of the solder ball. This incomplete melting resulted in a Head-in- Pillow (HiP) defect (Figure 8), since the manufacturing reflow temperatures were generally set to accommodate the SAC305 paste. 11

26 Figure 8: Head in pillow defect Celestica s Low Melt Solder Research Program Celestica, in partnership with the University of Toronto, set up a low melt research program in 2009 with the main goal of introducing a solder alloy which would allow for the use of conventional flame-retardant level 4 (FR4) laminate printed wire board (PWB) material with SnPb solder. This combination would require a process temperature of approximately 10 C lower than SAC305 and potentially reduce failures via pad cratering. This new solder paste would need to be compatible with both leaded and discrete components - where the solder paste makes up the majority of the solder joint, and with ball grid array components (BGA) where the paste is mixed with industry standard SAC305 or SAC105 to form a solder joint which has the composition made up of approximately 25% by volume paste, and approximately 75% by volume BGA alloy 10 (Figure 9). 12

27 Figure 9: a) gull wing leaded solder joint, b) BGA solder joint Celestica s program was divided into a number of phases as follows: Alloy Selection Manufacturing Feasibility Screening Experiments: o Aerospace and Defense Sector o Consumer Sector o Telecommunications Sector Reliability Testing This work explored the alloys selected for both the Consumer and Telecommunications sector portions of the screening experiments. The Aerospace and Defense sector program is described here in order to provide context for the current project. The remainder of this chapter summarizes the prior work and provides context for the work in this thesis Phase 1: Alloy Selection The initial work, described in Snugovsky et al. 6, outlines the process by which various alloys were selected for further study. This alloy selection phase of the project was divided further into the following stages: 13

28 Requirement Formulation Literature Search and Phase Diagram Analysis Preliminary Alloy Selection Metallurgical Analysis: Differential Scanning Calorimetry (DSC) and Microstructural Evaluation of: o Alloy o Alloy plus Cu simulate solder on board with leaded or leadless component o Alloy plus Cu plus SAC305/SAC105 simulate solder on board with BGA solder ball Requirement Formulation The following requirements were established: C process temperature, at least 10 C lower than SAC305 Approximately C pasty range Compatible with SAC305/105 solder balls as well as with leaded and discrete components Low Ag content 6 Equal or better thermo-mechanical properties than SAC305. As described in section 2.2, an ideal replacement to SAC solders needs to be compatible with temperature sensitive components. It should also be compatible with laminate PWB material used with SnPb solder, which is less expensive than newer, high temperature laminate material, requires less baking and processing prior to build and reduced adverse mechanical properties associated with these new materials; pad cratering, de-lamination and warpage. The pasty range is also important as a narrow range ensures good wettability due to less oxide formation and reduces the possibility of forming HiP defects if the paste does fully 14

29 melt and mix with the BGA solder ball. HiP may also result from excessive warpage in the laminate material. A small pasty range also prevents composition gradients and unexpected concentrations driven by thermal gradients in the cooling cycle. Finally, any new low melt solder should form an interconnection which exhibits equal or better properties in mechanical and thermo-mechanical testing than SAC305 and comparable to SnPb solder. The interconnection includes the solder paste, the board material (copper pad and any surface finish material), and either the BGA solder ball, the lead material of a leaded component (Figure 9), or a terminal finish of a leadless component (e.g. a capacitor) (Figure 10). Figure 10: Solder joint formed with leadless component 25 The reduction of Ag will reduce the cost of the alloy and also may improve mechanical properties, particularly in drop shock testing. It should also be noted that the ideal Ag content for drop shock performance may not be optimal in thermal cycling. 26 It is therefore necessary to balance these requirements Literature Search and Phase Diagram Analysis 6 Bi containing alloys are the focus of this study. It has been shown that the addition of Bi will improve thermal and thermo-mechanical properties over conventional SAC alloys. 27,28 These benefits can be attributed to solid solution and precipitate strengthening which will be described further in Chapter Chapter 2. Additionally, Bi may reduce the propensity for whisker growth in a Sn rich alloy. 6 The Bi addition was intended to depress the melting temperature but not to enter the very low melting temperature regime around the SnBi eutectic (Figure 11). The Bi content is tuned in conjunction with altering 15

30 the amount of Ag and/or Cu in order to maximize the temperature depressing characteristic of the Bi, without introducing the very low melting temperature (~ C) ternary eutectics (Figure 12 and Figure 13). Figure 11: SnBi phase diagram 29 Since alloys of higher order than ternary are difficult to illustrate using phase diagrams, the combined knowledge of the existing ternary and binary eutectic temperatures and compositions can be compared to determine an ideal range for the proposed alloys. Since a melting temperature of approximately 138 C is well below the outlined criteria, alloys which do not have any liquid component around this temperature are viable candidates. A slice along the z-axis, which represents temperature, of a 3D ternary phase diagram (Figure 14) can be used to explore the eutectic region of a particular alloy system. The solubility limit of Bi in Sn is 21%, however this is reduced with the addition of other alloying elements, namely Ag and/or Cu. An examination of the Sn-Ag-Bi, Sn-Cu- Bi and Sn-Ag-Cu systems is required. 16

31 a. b. Figure 12: a) SnAgBi ternary phase diagram and b) Sn-rich corner of SnAgBi phase diagram 30 From Figure 12 and Equations 1-1 and 1-2, a ternary eutectic (Point E ) of Sn-Ag-Bi exists at C, at a composition of Sn-0.68%Ag-43.47%Bi, and a binary eutectic (Point A ) at C with a composition of Sn-3.73%Ag is seen. L -> Ag 3 Sn + (Bi) + (Sn) 2-1 L -> Ag 3 Sn + (Sn) 2-2 Upon cooling along the binary eutectic line from point A, highlighted in blue in Figure 12, both Sn and Ag 3 Sn are formed simultaneously. At point E (the ternary eutectic point defined in Equation 1-1) all remaining liquid forms simultaneously into Sn, Ag 3 Sn and Bi. 31 Compositions along this line were explored in the development of proposed new alloys. 17

32 a. b. Figure 13: a) SnCuBi ternary phase diagram and b) Sn-rich corner of SnCuBi phase diagram 32 Similarly, from Figure 13 and Equation 1-3, we can see that a ternary eutectic exists at C at a composition of Sn-0.01%Cu-43.09%Bi and a binary eutectic at C with a composition of Sn-0.89%Cu. L -> Cu 6 Sn 5 + (Sn) + (Bi) 2-3 L -> Cu 6 Sn 5 + (Sn) 2-4 Exploring the binary eutectic line following cooling towards the ternary eutectic from point A in Figure 13 to point E for the SnCuBi system gives a starting point for determining ideal alloys. Figure 15 is an isothermal projection of the SnBiAg phase diagram at C. The area highlighted in yellow represents an area where ternary eutectic is present at this temperature. Compositions from within these triangles should be avoided in order to avoid melting in the low temperature range. Additionally, the ternary eutectic of the Sn- Ag-Bi system should be avoided because Bi is present in the eutectic as a primary precipitate. These primary particles may be large. It is preferable to have small Bi 18

33 particles, the type which will likely precipitate out of solid solution after initial solidification. Finally, it should also be noted that current production processes such as reflow ovens and hand solder rework, provide rapid cooling environments and do not allow for equilibrium solidification. Other possible effects of cooling, such as Bi segregation, should be accounted for in the development of any new alloy. Figure 14: Example of ternary phase diagram 31 19

34 Figure 15: SnAgBi ternary phase diagram, isothermal section Preliminary Alloy Selection 6 Based on examination of phase diagrams as described in section , 23 Bi-containing alloys were selected for further evaluation including those listed in Table 2. The first two alloys in the table, J.Hwang and Senju M42, are currently available in industry. Of these four industry available alloys, none fully meet the criteria outlined in section but were included for study because they met most criteria and for purposes of comparison. 20

35 Table 2: Preliminary alloy selection 10,6 Alloy Composition (wt%) Published Melting Temperatures ( C) Min Max Pasty Range Experimental Melting Temperatures ( C) Min Max Pasty Range J. Hwang 34 Sn3.1%Ag0.5%Cu3.1%Bi Senju M42 35 Sn2%Ag0.75%Cu3%Bi Paul 27,28 Sn3.4%Ag4.8%Bi Orchid Sn2%Ag7%Bi Violet Sn2.25%Ag0.5%Cu6%Bi Sunflower Sn0.7%Cu7%Bi Cornflower Sn0.7%Cu10%Bi Sunrise Sn1%Ag0.7%Cu7%Bi Metallurgical Analysis: DSC and Microstructural Evaluation 6 Alloys from Table 2 were evaluated using the thermoanalytical technique, differential scanning calorimetery (DSC). DSC measurements of the alloys were made in two forms: alloy alone representing the solder paste used in the production of a typical leaded or leadless solder joint, and the alloy combined with SAC305 in approximately 25%-75% ratio to represent the solder joint formed when the paste is used with a BGA component (Figure 9). DSC measures the difference in energy input into a cell containing the sample within a sample holder vs. a reference cell, usually the sample holder (minus the sample). 36 By plotting the heat flow (J/g*min) vs. temperature ( C) it is possible to determine the onset of a phase transformation by the decrease in heat flow, the temperature range over which this transition occurs and the total enthalpy of the reaction. In the case of solder in this experiment, the onset of melting and the pasty range could be 21

36 determined. Table 2 summarizes both the published values of melting temperature and range as well as values determined in this study. DSC can also be used to determine if there are any additional, undesirable phase changes that occur in the alloys. As the selected alloys were close to the eutectic compositions, only one peak is desirable. Figure 16 is a good example of an alloy with an undesirable hypoeutectic composition. Note that the saturation point of Bi in Sn is at 21wt% Bi. The eutectic composition is approximately 57wt% Bi. As the alloy is heated from room temperature, the first peak is encountered at approximately 139 C (point A), which is the solid solubility limit for Bi in Sn as well as the initiation of melting. This is characterized on the DSC curve by a sharp peak indicating an invariant phase change. Continued heating of this alloy shows that it passes through a melting range characterized on the DSC curve by a secondary peak, at point B (liquidus). 37 Figure 16: DSC scan of Sn20%Bi 10,38 Table 3 summarizes the DSC analysis performed on the proposed alloys combined with 75% SAC305 to simulate the composition of the solder paste mixed with a SAC305 solder ball of a BGA. 22

37 Alloy Table 3: DSC analysis of proposed alloys with 75% SAC305 Composition (wt%) Experimental Melting Temperatures ( C) Min Max Pasty Range J. Hwang 39 Sn3.1%Ag0.5%Cu3.1%Bi Senju M42 40 Sn2%Ag0.75%Cu3%Bi Paul 27,28 Sn3.4%Ag4.8%Bi Orchid Sn2%Ag7%Bi Violet Sn2.25%Ag0.5%Cu6%Bi Sunflower Sn0.7%Cu7%Bi Sunrise Sn1%Ag0.7%Cu7%Bi Microstructural evaluation was performed by cross sectioning the alloyed samples and examining the microstructure using a scanning electron microscope (SEM) and energydispersive X-ray spectroscopy (EDX). This evaluation was done using three forms of the proposed alloys: Alloy Alloy reflowed onto a copper foil Alloy reflowed onto a copper foil with SAC305 The second and third scenarios represent the alloy reflowed onto a copper pad found on a PWB and a BGA mixed with the alloy onto a copper pad respectively. Using the above three scenarios, each of the alloys were evaluated based on the following microstructural factors: intermetallic thickness and morphology, microstructural uniformity and coarseness of the overall microstructure. The intermetallic formed between the solder alloy and the copper pad creates both the mechanical and electrical bond of the solder joint. The formation is discussed further in Chapter Chapter 2. Both Cu 6 Sn 5 and Cu 3 Sn are possible intermetallics that form during solder reflow, and are substantially more brittle than Sn or Cu alone and have facetted 23

38 morphologies. During a typical reflow of Sn based solder onto a copper foil, it is typically the Cu 6 Sn 5 IMC layer in a scalloped morphology which forms the interconnect. A continued solid-state reaction between the Sn in the solder, the Cu 6 Sn 5 layer and the Cu may persist at elevated temperatures (including those present in service conditions) causing the Cu 6 Sn 5 layer to grow and possibly resulting in the formation of a Cu 3 Sn layer between the Cu 6 Sn 5 and Cu over time. In this study, the initial IMC formation processes were compared. There should be a sufficient, continuous IMC formed with no interruptions to ensure a proper bond has formed. Conversely, the IMC layer thickness should be minimized as it is the most brittle portion of the solder joint and prone to fracture. Thicknesses will vary based on solder composition, PWB surface finish, reflow profile (time and temperature) and post reflow exposure. An optimal thickness definition does not currently exist in industry. Microstructural uniformity and grain coarseness were also examined. The microstructure plays a role in the materials ability to resist creep by allowing the induced load to be uniformly distributed over the specimen thereby reducing strain-induced grain growth and crack initiation. 41 Solder joints can readily reach temperatures of approximately of the melting temperature (T m ) during normal operation. This temperature range may result in both the coarsening of the primary grain structure, in this case the Sn matrix grains, as well as growth of Ag 3 Sn platelets. 42 In this study, the alloy itself, as well as the alloy mixed with SAC305 in a 1:3 ratio were examined to simulate a solder joint formed with the proposed alloy and a SAC305 BGA solder ball, as shown in Figure

39 Figure 17: Microstructure of solder joint formed with SAC305 solder ball and a no- Ag solder paste alloy Phase 2: Manufacturing Feasibility Study 43 Manufacturing feasibility, or solder paste performance, was tested at Celestica. The pastes were tested for printability, wettability and performance in the screening process. In general, the performance of a solder paste is driven largely by the size and morphology of the solder powder, the composition and volume of flux and the overall rheology of the paste. These variables are largely alloy independent and therefore not considered in detail during this study. This study concluded that the paste variables have been sufficiently addressed and optimized in order to move to further stages of evaluation, which would better distinguish between alloys Phase 3: Screening Experiments From the alloys described in 2.4.1, it was determined that no one alloy would meet the needs of the three market segments of interest. It was therefore proposed that a number of screening experiments be performed in order to down-select the number of alloys which would undergo full reliability studies. Table 4 summarizes the alloys selected for each of the screening experiments for different market segments. 25

40 Market Segment Aerospace and Defense Consumer Telecommunications Table 4: Alloys for screening experiments Alloy Composition (wt%) Process Temperature ( C) Pasty Range ( C) Paul Sn3.4%Ag4.8%Bi Violet Sn2.25%Ag0.5%Cu6%Bi Orchid Sn2%Ag7%Bi Sunflower Sn0.7%Cu7%Bi Sunrise Sn1%Ag0.7%Cu7%Bi Senju M42 Sn2%Ag0.75%Cu3%Bi Violet Sn2.25%Ag0.5%Cu6%Bi Sunflower Sn0.7%Cu7%Bi Sunrise Sn1%Ag0.7%Cu7%Bi Senju M42 Sn2%Ag0.75%Cu3%Bi Aerospace and Defense 44, 45 The screening experiments for the three alloys selected for the aerospace and defense sector, as listed in Table 4, were performed between 2011 and 2013 by Celestica in partnership with Honeywell. In these screening experiments, three alloys were compared against baselines of both SAC305 and SnPb. Both BGA, using SAC305 solder balls, except with SnPb solder in which case SnPb solder balls were used, and leaded components typical of those used in aerospace and defense assemblies were investigated using a medium complexity board designed by Honeywell, as shown in Figure 18. Figure 18: Honeywell test vehicle, medium complexity board 26

41 Screening Experiment Structure A Latin squares approach was used in this screening experiment. 46 This approach allowed for the most information to be obtained without testing every possible combination, which would be cost prohibitive. Three factors were identified: board material, board finish: Organic Solderability Preservative (OSP), Electroless Nickel Immersion Gold (ENIG) and Electroless Nickel Electroless Palladium Immersion Gold (ENEPIG) and paste (Violet, Orchid and Paul), in nine combinations rather than the full 27 combinations. Since only two T g levels were of interest, instead of three, as with the other factors, the higher T g was repeated, as summarized in Table 5. Boards with T g of 150 C and 170 C were used. The overall build matrix is summarized in Table 6. No clean solder paste of the experimental alloys was prepared by a major solder paste supplier for the purposes of this experimental work (Table 6). Boards were built at Celestica. Two different reflow temperatures were used in the manufacturing process. One with a peak temperature was 240 C, was used with the SAC305 alloy and one with a peak temperature of 222 C for all other alloys. Both had a time above liquidus of approximately seconds. Table 5: Latin square of paste and finish, and board materials Alloy Finish OSP ENIG ENEPIG Paul Violet Orchid

42 Table 6: Build matrix showing number of assemblies PWB Material Normal T g (150 C) High T g (170 C) Alloy Surface Finish ENEPIG ENIG OSP SnPb 2 3 SAC Paul 4 Violet 4 Orchid 4 SnPb 2 3 SAC Paul 4 4 Violet 4 4 Orchid Microstructural Assessment A detailed discussion on the microstructural evaluation can be found in Snugovsky et al. 44 and in Juarez et al. 45. All of the assemblies were inspected optically and with x-ray radiography, to ensure that no major anomalies or concerns existed prior to further testing. Figure 19 illustrates wetting of the solder paste on quad flat pack (QFP) leaded components on OSP board finish. SnPb exhibited the best wetting properties, while SAC305 showed the worst. All three experimental alloys showed some degree of improved wetting over SAC305. Figure 20 shows the cross section evaluation performed on solder joints made with all solder pastes on OSP board material. All BGAs showed good mixing characteristics between the solder paste and the SAC305 solder ball. Figure 19: Example of wetting on OSP, QFP240 28

43 Figure 20: BGA and QFP solder joints on OSP Table 7 summarizes the composition of the resulting interconnect BGA solder ball after assembly. This final BGA solder ball composition is a mixture of the paste alloy and the original component SAC305 solder ball. The compositional measurements were made using the semi-quantitative method of SEM-EDX and are provided here for comparative purposes only. Table 7: Resultant BGA interconnect composition Paste Surface Ball Composition (wt%) Alloy Finish Sn Ag Cu Bi Pb SnPb OSP SAC305 OSP Paul OSP Violet OSP Orchid OSP SAC305 ENIG Paul ENIG Violet ENIG Orchid ENIG Paul ENEPIG Violet ENEPIG Orchid ENEPIG Accelerated Thermal Cycling (ATC) 17 assemblies of various paste, board finish, and board material combinations, were exposed to upwards of 3000 cycles of harsh temperature thermal cycling in accordance 29

44 with IPC-9701A: Performance Test Methods and Qualification Requirements for Surface Mount Solder Attachments 47 using Test Condition 1 (-55 C to 125 C) for 3000 cycles. Table 8 and Table 10 summarize the number of failures over the course of the entire test. Mean Time to Failure (MTTF) is defined by Equation 1-5 and the failure rate ( ) is defined by Equation MTTF N 1 MTTF i N T i i 1 N i N T i i Table 9 and Table 11 attempts to quantify the time to failure over the course of the 3,000 plus cycles in spite of the lack of a full statistical set of test data. Table 8: ATC failures on 170 C T g board material Paste type SnPb SAC305 Paul Violet Orchid ENEPI G OSP ENEPI G OSP ENEPI G ENI G ENEP IG Finish 352 BGA 2/2 2/2 240 QFP OSP ENIG OSP 30

45 Table 9: Reliability results on 170 C T g boards after ATC Alloy Finish Component MTTF R(3039) =Failure/cycle SnPb ENEPIG 352 BGA QFP OSP 352 BGA QFP SAC305 ENEPIG 352 BGA QFP OSP 352 BGA QFP Paul ENEPIG 352 BGA QFP ENIG 352 BGA QFP Violet ENEPIG 352 BGA QFP OSP 352 BGA QFP Orchid ENIG 352 BGA QFP OSP 352 BGA QFP Table 10: ATC failures on 150 C T g board material Solder Alloy SnPb SAC305 Paul Violet Orchid Finish ENEPIG OSP ENEPIG OSP OSP ENIG ENEPIG 352 BGA 2/2 2/2 2/2 2/2 1/2 240 QFP 1/2 2/2 2/2 2/2 1/2 1/2 31

46 Table 11: Reliability results on 150 C T g boards after ATC Alloy Finish Component MTTF R(3039) =Failure/cycle SnPb ENEPIG 352 BGA QFP OSP 352 BGA QFP SAC305 ENEPIG 352 BGA QFP OSP 352 BGA QFP Paul OSP 352 BGA QFP Violet ENIG 352 BGA QFP Orchid ENEPIG 352 BGA QFP In the above ATC testing, there are only two replicates of each condition. Even if both fail, there is insufficient data to plot failure rates using Weibull, or any other type of statistical distribution. Accordingly, MTTF and failure rate ( ) were used to quantify the data from this screening experiment and to describe the length of time for which each condition survived. All combinations using the 170 C T g boards survived beyond the 1000 cycles required by the Aerospace and Defense sector. SnPb and Violet BGAs both experienced failures with a MTTF of 1611 and 2271 respectively on OSP boards. As OSP is not a finish that is favored by this sector, the results on the ENIG and ENEPIG finish boards, which survived in all cases, show promise. Combinations on the 150 C T g boards experienced more failures on all board finishes. Only the SAC305 combinations on OSP show MTTF below the required 1000 cycles: 551 for BGAs and 802 for QFPs. Again, the boards with OSP finish failed earlier than boards with other finishes. Upon further investigation, these failures were found to be in the board material, in the form of barrel cracks, rather than in the solder joint interconnect. 32

47 % Failure Vibration Seventeen (17) assemblies of various paste, board finish and board material combinations were exposed to two force levels of vibration testing. The test procedure, a discussion of the failure modes, as well as a detailed discussion of the reliability results, can be found in Juarez et al Chart of % Failure Alloy Vibration Level (G) Violet Paul 2 Orchid Violet SnPb Paul 5 Orchid SAC305 Figure 21: Failures in vibration testing Figure 21 provides a summary of the number of failures that occurred during vibration testing, at two different levels of acceleration, 2G and 5G respectively. The table illustrates that all three experimental alloys outperform SAC305, with Violet even outperforming SnPb. Table 12 and Table 13 summarize the failures that occurred based on varying conditions and test combinations. A more comprehensive discussion of the survival times can be found in the original work referenced above. Table 12: Vibration failures after 2G testing Alloy Paul Violet Orchid Finish OSP ENIG ENEPIG OSP ENIG ENEPIG OSP ENIG ENEPIG Board Material BGA 240 QFP 2/2 2/2 2/2 2/2 1/2 33

48 Table 13: Vibration failures after 5G testing Alloy Paul Violet Orchid OSP ENIG ENEPIG OSP ENIG 34 ENEPI G OSP ENIG ENEPIG Finish Board Material BGA 2/2 2/2 1/2 2/2 2/2 2/2 2/2 1/2 240 QFP 2/2 2/2 2/2 2/2 2/2 2/2 2/2 2/2 Alloy SAC305 SnPb Finish OSP OSP ENEPIG ENEPIG OSP OSP ENEPIG ENEPIG Board Material BGA 2/2 2/2 2/2 2/2 2/2 2/2 240 QFP 2/2 2/2 2/2 2/2 2/2 2/2 2/ Summary of Findings The three alloys compared against SAC305 and SnPb in the Aerospace and Defense Sector screening experiments showed acceptable manufacturability characteristics with improved wetting and voiding over SAC305. Two of the alloys, Paul and Orchid, did not form proper intermetallics with ENIG or ENEPIG board finishes while Violet did. This is finding is attributed to the lack of Cu present in these alloys. All three experimental alloys met the Aerospace qualification of 1000 cycles of harsh thermal cycling. Only SAC305 on 150 C T g boards with OSP finish failed to meet the criteria, however, the failures were determined to be via failures within the board. This good performance is attributed to the fact that Bi particles evenly precipitate into the Sn matrix during thermocycling, reducing grain coarsening and microstructural degradation. In vibration testing, all experimental alloys showed an improved performance over SAC305, both Violet and Paul performed equally or better than SnPb. 3 Objective of Thesis: Low Melt Solders for Consumer Sector The screening experiments for the three alloys selected for the Consumer Sector, as listed in Table 4 and Table 14, are the focus of this thesis. While SAC305 has been widely

49 adopted by the industry during the initial transition to Pb-free, desired improvements in mechanical and thermal performance, as well as cost reductions, continue to drive research into new alloys. Figure 22 illustrates the process for developing requirements for a new replacement alloy for the consumer sector. The figure outlines the main shortcomings of the current market available Pb-free solution based on SAC alloys, namely the high cost and poor drop/shock performance. These shortcomings, in all probability could be overcome by a new alloy, which is both less stiff and has a lower process temperature. The lower process temperature would allow for normal T g board material that which has been used with SnPb solders for many years to again be used, eliminating collateral failure modes such as pad cratering and an increased susceptibility to moisture delamination. New Pb-Free Solder Less Stiff Lower Process Temperature Normal T g Board Material No Pad Cratering No Moisture Delamination/Cracks Improved Drop/Shock Performance Less Expensive Figure 22: New approach proposing to use solder with no or low-ag content 6 Any replacement alloy for the consumer sector should perform better than SAC305 with respect to drop/shock performance and thermal cycling. Both tests are representative of service conditions that would likely be experienced by any consumer electronic product. 35

50 The following accelerated test conditions are accepted by industry as representative of life conditions for these products: ATC in accordance with IPC-9701A Performance Test Methods and Qualification Requirements for Surface Mount Solder Attachments 47 Drop/Shock Testing in accordance with the Joint Electron Device Engineering Council (JEDEC) Standard JESD22-B110A Subassembly Mechanical Shock Test 49 Table 14: Consumer alloys under test Alloy Composition Assembly Temperature SAC305 Sn 3%Ag 0.5%Cu 240 C Senju M42 Sn 2%Ag 0.75%Cu 3%Bi 224 C Sunrise Sn 1%Ag 0.7%Cu 7%Bi 222 C Sunflower Sn 0.7%Cu 7%Bi 226 C This thesis presents the results of both ATC and Drop Shock testing of the alloys selected for consumer electronics applications (Table 4 and Table 14). These alloys were also tested for a lower probability of Sn whisker formation according to JESD22A

51 4 References 1 K. Sweatman, J. Read, T. Nishimura, and K. Nogita The Effect of Microalloy Additions on the Morphology and Growth of Interfacial Intermetallic in Low-Ag and No-Ag Pb-Free Solders, presented at SMTAI, Chicago, Il, K-N. Tu Copper Tin Reactions in Bulk Samples in Solder Joint Technology : Materials, Properties and Reliability New York, Springer, 2007, ch. 1, pp IPC-A-610E Acceptability of Electronic Assemblies April Directive 2002/95/EC of the European Parliament and of the Council of 27 January 2003 on the restriction of the use of certain hazardous substances in electrical and electronic equipment 5 Directive 2011/65/EUof the European Parliament and of the Council of 8 June 2011 on the restriction of the use of certain hazardous substances in electrical and electronic equipment 6 P. Snugovsky, S. Bagheri, M. Romansky, D. Perovic, L. Snugovsky, J. Rutter New Generation of Pb-Free Solder Alloys: Possible Solution to Solve Current Issues with Main Stream Pb-Free Soldering, SMTA Journal Volume 25 Issue 3, M. Kelly, D. Colnago, V. Sirtori, C. Grosskopf, K. Lyjak, C. Ravenelle, E. Kobeda, J. Bath, S.K. Tan, and L.H. Teo Component Temperature Study on Tin-Lead and Lead-Free Assemblies, Journal of SMT, Volume 15 Issue 4, IPC/JEDEC J-STD-020D.1 Moisture/Reflow Sensitivity Classification for Nonhermetic Solid State Surface Mount Devices March B.Gray Correlation of Printed Circuit Board Properties to Pad-Crate Defects Under Monotonic Spherical Bend, M.A.Sc. thesis, Dept. Mech. Eng., Ryerson University, Toronto, Ontario P. Snugovsky Low Melt Pb-Free Solder to Solve Pb-Free Transition Challenges 11 T. Sawamura and T. Igarashi Difference between Various Sn/Ag/Cu Solder Compositions, Almit Ltd. June K.S. Kim, S.H. Huh, K. Suganuma Effects of intermetallic compound on properties of Sn-Ag-Cu lead free solder joints Journal of Alloys and Compounds 1 (2002) 13 L. Turbini. Process and material issues related to Lead-free soldering, J Mater Sci: Mater Electron (2007) 18: U. Kattner, Phase Diagrams for Lead-Free Solder Alloys JOM December 2002 pp

52 16 C. Handwerker. Fundamental Properties of Pb-Free Solder Alloys, in Pb-Free Soldering, Springer 2007, ch.2, pp K-N. Tu Introduction in Solder Joint Technology : Materials, Properties and Reliability New York, Springer, 2007, sec , pp J. Dutkiewicz, L.A. Zabdyr, Z. Moser and J. Salawa. ASM Handbook Volume 3 Alloy Phase Diagrams Materials Park, OH: ASM International 1992 pp K.-W. Moon, W.J. Boettinger, U.R. Kattner, C.A. Handwerker, and D-J. Lee. The Effect of Pb Contamination on the Solidification Behavior of Sn-Bi Solders, Journal of Electronic Materials, Vol. 30, No. 1, M. Ribas S. Chegudi, A. Kumar, S. Mukherjee, S. Sarkar, R. Pandher, R. Raut, and B. Singh. Low Temperature Alloy Development For Electronics Assembly Part II 21 P. Vianco, J. Rejent and R. Grant Development of Sn-Based, Low Melting Temperature Pb-Free Solder Alloys Materials Transactions, Vol 45. No. 3 (2004) pp C. Handwerker. Fundamental Properties of Pb-Free Solder Alloys, in Pb-Free Soldering, Springer 2007, ch.2, pp Acceptability of Electronic Assemblies IPC-A-610 Revision D February 2005 pp D.A. Shnawah, M.F.M. Sabri, I.A. Badruddin A review on thermal cycling and drop impact reliability of SAC solder joint in portable electronics products, Department of Mechanical Engineering, University of Malaya, Kuala Lumpur, Microelectronics Reliability 52 (2012) pp P.T. Vianco, J.A. Rejent Properties of Ternary Sn-Ag-Bi Solder Alloys: Part 1 Thermal Properties and Microstructural Analysis, Journal of Electronic Materials, Vol.28, No.10, pp , P.T. Vianco, J.A. Rejent Properties of Ternary Sn-Ag-Bi Solder Alloys: Part 2 Wettability and Mechanical Properties Analysis, Journal of Electronic Materials, Vol.28, No.10, pp , D.A. Porter and K.E.Easterling, Thermodynamics and Phase Diagrams in Phase Transformations in Metals and Alloys, 2 nd ed. UK: Chapman & Hall, 1992, ch. 1, pp

53 J.S. Hwang A Strong Lead-free Candidate: the Sn/Ag/Cu/Bi System, SMT Magazine, August 1, S.W. Chen, C-C Lin, C. Chen Determination of the Melting and Solidification Characteristics of Solders Using Differential Scanning Calorimetry, Metallurgical and Materials Transactions A, Volume 29A, July 1998 pp L. Rycerz. Practical remarks concerning phase diagrams determination on the basis of differential scanning calorimetry measurements J Therm Anal Calorim, 113, 2013 pp M. H. Kaye, K.M. Jaansalu and W.T. Thompson Condensed Phases in Inorganic Materials: Metallic Systems in Measurement of the Thermodynamic Properties of Multiple Phases, R.D. Weir and Th.W. deloos, Ed. San Diego, Elsevier Inc., 2005, pp J.S. Hwang A Strong Lead-free Candidate: the Sn/Ag/Cu/Bi System, SMT Magazine, August 1, J.W. Evans Introduction to Solder Alloys and Their Properties in A Guide to Leadfree Solders: Physical Metallurgy and Reliability, Silver Spring, MD: Springer, 2005, pp J.W. Evans Microstructural Instability in Solders in A Guide to Lead-free Solders: Physical Metallurgy and Reliability, Silver Spring, MD: Springer, 2005, pp E. Kosiba S. Bagheri, Z. Bagheri, P. Snugovsky, and D. Perovic Assembly Feasibility and Property Evaluation of Low Ag, Bi-Containing Solder Alloys in ICSR SMTA Conference, Toronto, ON, P. Snugovsky, E. Kosiba, J. Kennedy, Z. Bagheri, M. Romansky, M. Robinson, J.M. Juarez, Jr., J.Heebink Manufacturability and Reliability Screening of Lower Melting Point Pb-free Alloys Containing Bi, in IPC APEX EXPO Conference, San Deigo, CA, J. Juarez Jr., M. Robinson, J. Heebink, P. Snugovsky, E. Kosiba, J. Kennedy, Z. Bagheri, S. Suthakaran, M. Romansky Reliability Screening of Lower Melting Point Pb-Free Alloys Containing Bi, in IPC APEX EXPO Conference, Las Vegas, NV, D.C. Montgomery. Design and Analysis of Experiments, 8 th Edition. Jonh Wiley & Sons,

54 47 IPC-9701A Performance Test Methods and Qualification Requirements for Surface Mount Solder Attachments, February J. Bentley Introduction to Reliability and Quality Engineering, 2 nd Ed, Essex, England, Pearson Education Limited, 1999, ch. 2, pp Subassembly Mechanical Shock, JESD22-B110A, November

55 Chapter 2 Solder Joints after Reflow (As Manufactured) 1 Introduction The formation of a solder joint is influenced by the conditions at which the chemical bond is formed: primarily, the maximum temperature of the reflow, the time above solidus and the cooling rate. Therefore, the melting temperature and pasty range are primary characteristics which differentiate solder alloys. The solder joint microstructure further evolves when exposed to elevated temperatures for a prolonged period of time or if exposed to other forms of stress, due to solid-state diffusion and other physical effects described below. The microstructural properties described in this chapter, which result from the solidification of liquid solders, will have an influence on the mechanical and thermo-mechanical properties of the solder joint and therefore impact the reliability in field conditions. 2 Bulk Solder Microstructure Solidification Process Slight changes to the alloy composition may result in different solidification regimes within the solder joint and therefore significantly change properties. It is therefore of interest to examine the factors which govern this process. 2.1 Cu Dissolution in Molten Solder As the molten solder comes into contact with the solid Cu pad layer on the PWB board, Cu will dissolve into the liquid solder. For this reason, the overall concentration of Cu within the molten solder will increase at a rate that is dependant on the time spent above the solidus temperature. SAC305, as an example, has an initial Cu concentration of 0.5wt% however, by the time cooling is initiated, the Cu concentration may be as high as 1wt%. Figure 23 shows a typical SAC305 reflow profile in which a particular solder joint experienced 71 seconds above 217 C, the solidus temperature for SAC305. This Cu dissolution during the liquid phase of solder joint formation will continue, given enough 41

56 time, until the liquids solubility limit is reached. This reaction has been characterized by the Nernst-Brunner equation (Equation 2-1): X X s 0 S X X KAt V ln 2-1 where X S represents the solubility limit, X 0 represents the initial concentration, K is the temperature-dependent Nernst-Brunner dissolution rate constant, A is the contact area at the interface, V is the total volume of solder, t is time. 1 Figure 23: Typical reflow profile for SAC305 If dissolution kinetic conditions are satisfied (for example if a very high temperature is maintained over a sufficient length of time), the Cu present in the board material can be dissolved into the solder up to a very high concentration. Potentially the entire copper pad region on the PWB can be consumed by the intermetallic and bulk solder. Figure 24 shows the Cu pad being progressively consumed by the IMC layer after successive rework/replacement cycles. In this example, the solder joint is repeatedly heated above the solder alloy liquidus temperature. 42

57 Figure 24: Copper dissolution of BGA solder joint after a) 1 replacement b) 3 replacements and c) 5 replacements 2 It has been shown that an increase in copper concentration within a Sn-based solder alloy from 0.5%Cu to 0.7%Cu decreases the amount of copper dissolution from the PWB material, which is attributed to the reduction in the concentration gradient. 3 This is the primary reason for the presence of Cu in the SAC system of alloys and in the alloys proposed in this study. The Cu dissolution described above typically results in a concentration within the hypereutectic region of the Cu-Sn phase diagram (Figure 6). As the solder cools Cu 6 Sn 5 particles will begin to nucleate within the liquid. As the Cu dissolution into the liquid solder occurs, a competing process is taking place in which the Cu and liquid solder react to form a chemical bond at the interface, which takes the form of an intermetallic compound the interfacial IMC. This reaction is discussed in Bulk Solder Solidification Within the bulk solder, the equilibrium solidification of SAC305 follows the path outlined in red, in Figure 25, where non-faceted Sn grains begin to form a dendritic structure. As the liquid continues to be consumed by the new Sn structure, the remaining liquid composition moves along the red line from A towards B. At B, a eutectic of Sn + Ag 3 Sn forms in the interdendritic spaces of the already solidified Sn dendrite. This eutectic solidification continues, and the liquid compositions following the eutectic valley from B towards E. At this point, any remaining liquid is now at the ternary 43

58 equilibrium composition (point E) and will solidify as a ternary eutectic of Sn + Ag 3 Sn + Cu 6 Sn 5. 4 Figure 25: SnAgCu phase diagram with SAC305 equilibrium solidification path 5 The solidification path described above does not account for Cu dissolving into the bulk solder from the board side Cu pad and, in the case of a leaded component, from the lead itself. The Cu concentration in the bulk liquid is therefore increasing; the blue line in Figure 25 indicates a concentration of approximately 1wt% Cu. In this case, the Cu 6 Sn 5 begins to solidify before the Sn. The Cu 6 Sn 5 continues to form until the concentration reaches the eutectic line and a binary eutectic of Sn + Cu 6 Sn 5 forms. Additionally, Cu 6 Sn 5 IMC, which has already formed at the Cu-liquid solder interface, and will be described in section 3, may break off and move through the bulk solder. As the Sn dendritic arms grow, the composition of the liquid at the Sn/liquid interface changes and becomes Cu rich. In the Cu rich regions, a binary of Sn + Cu 6 Sn 5 will begin to nucleate and grow. The remaining liquid at the Sn + Cu 6 Sn 5 eutectic/liquid interface will be Ag rich. At this point, the remaining liquid will either solidify as a 44

59 eutectic of Sn + Ag 3 Sn to form the double binary eutectic 6, or as a ternary eutectic of Sn + Cu 6 Sn 5 + Ag 3 Sn. 2.3 Undercooling During Solidification The solidification path described above assumes equilibrium cooling conditions, in which sufficient time is allowed for all processes to occur. In reality, cooling rates are usually faster than equilibrium conditions resulting in some effects of undercooling. A number of different structural morphologies can be expected with different cooling rates. For example, a quick quench after melting of a SAC alloy resulted in a fine Sn dendrite structure with very fine, non-faceted ternary eutectic structure in the interdendritic spaces. 7 Figure 23 shows that in a typical SMT reflow process, the solder alloy cools relatively quickly after reaching maximum temperature. This cooling profile will induce microstructures more typical of undercooling conditions than of equilibrium cooling as described above. In the case of a SAC alloy, in which Sn, Cu 6 Sn 5 and Ag 3 Sn all form, cooling below the eutectic temperature of 217 C would represent an undercooling condition for all three of the solid phases. As one of these phases begins to nucleate and grow, the composition of the remaining liquid changes. A move away from the eutectic point would represent a condition of even further undercooling of one of the phases, as, in all cases the liquidus surface increases away from the eutectic point. As the second phase nucleates, the remaining liquid again shifts in concentration until the nucleation of the third and final phase occurs. This is referred to as constitutional undercooling, describing the changing conditions of the liquid at the solidification front. As described above, it is expected that a SAC305 solder, cooled quickly below the ternary eutectic point of 217 C, would initially form a Sn phase, followed by Sn + Ag 3 Sn 5 eutectic and finally the ternary Sn + Ag 3 Sn 5 + Cu 6 Sn 5 eutectic. As the concentration of Cu in the molten solder is increased, Cu 6 Sn 5 is favored to solidify first. This implies that the primary phase to solidify will be the one with the highest liquidus temperature, however there are a number of other factors to consider including: 45

60 nucleation from undercooled liquid the growth kinetics of facetted vs. non-facetted phases the availability of Cu and Ag within the volume to form corresponding phases. 6 Figure 26 shows a 3D phase diagram in the Sn rich portion of a SAC alloy. The liquidus surfaces are extended below the eutectic point to help visualize the solidification of the three phases in undercooling situations. The red line shows the equilibrium solidification of a particular Ag rich alloy, while the white line shows the solidification path of the same alloy in non-equilibrium, undercooling conditions. In this scenario, Ag 3 Sn continues to form until the composition reaches the eutectic of Ag 3 Sn + Cu 6 Sn 5. The undercooling of SAC305 with further dissolved Cu (Cu concentration of ~1wt%) would result in a solidification path along the Cu 6 Sn 5 surface below the eutectic point. As the temperature is significantly lower than the liquidus temperature for this phase, Cu 6 Sn 5 will nucleate and grow. The composition of the liquid will then change and move to another area on the phase diagram. The next phase to solidify will be the one with the highest liquidus temperature, and therefore the greatest degree of undercooling. This will continue until all liquid has solidified. Figure 26: A 3D phase diagram of Sn rich portion of Sn-Ag-Cu ternary system 8 46

61 The growth kinetics of the three phases also plays a role in the solidification process. While this process is beyond the scope of this work, it should be considered that both Ag 3 Sn and Cu 6 Sn 5 have facetted structures and are therefore restricted to grow in specific crystallographic directions, while the non-facetted Sn can grow through the fluid in a more unrestricted fashion. 6 Finally, the volume fraction of elements in the liquid needs to be considered. As one phase solidifies, its growth is further restricted by the availability of atoms of the correct type. The final volume fraction of the Ag 3 Sn and Cu 6 Sn 5 phase is therefore limited by the availability of Ag and Cu atoms respectively within the liquid. Ag 3 Sn requires a localized concentration of 73.2wt% Ag within a total volume with a concentration of only 3wt% in SAC305 and even lower in the other alloys considered in this study. This would require significant diffusion of Ag through the liquid, which may not occur in undercooling conditions. Similarly Cu 6 Sn 5 requires a local concentration of 39.1wt% Cu within a liquid with only approximately 1wt% Cu. Sn, by contrast, will likely have a ready supply of Sn at the solidification fronts of the other two phases since it makes up the vast majority of the volume fraction. 2.4 Formation of Facetted IMCs in Bulk Solder The presence of large, faceted IMCs in the bulk solder, Ag 3 Sn and Cu 6 Sn 5, may present a reliability concern as they can act as stress risers and paths along which cracks propagate. 9 Large, primary Cu 6 Sn 5 particles form when the total concentration of Cu in the liquid is high and Cu 6 Sn 5 is the first phase to solidify (Figure 25). Some undercooling is also required, however the undercooling required to nucleate Cu 6 Sn 5 is less than that required to nucleate Sn. 10 This phase has a facetted structure of hexagonal morphology, often hollow on the inside. The structure then grows in a rod, forming a structure similar to a pencil a hexagonal tube with a facetted center (Figure 27). As a primary phase solidifying from liquid, this facetted rod structure will be surrounded by a liquid whose composition moving along the blue line shown in Figure 25 until a eutectic of Sn + Cu 6 Sn 5 forms together. Since the base phase (Cu 6 Sn 5 ) is a more complex crystal 47

62 structure, it dictates the final shape of the eutectic colony. In this case, the Cu 6 Sn 5 portion of the eutectic may form as branches of the original rod like structure. 11 Ag 3 Sn IMCs form within the bulk solder. Similar to the Cu 6 Sn 5 structures, Ag 3 Sn may form as primary phases, or as part of a binary or ternary eutectic solidification regime depending on the alloy composition and cooling rate. At compositions below approximately 3.7wt% Ag, Ag 3 Sn particles are not expected to form as a primary phase. It is more likely that the Ag 3 Sn will form in the interdendritic spaces of the Sn and/or Cu 6 Sn 5 phases and as part of a eutectic with one of both of the other two phases. In a typical cross-section view, the Ag 3 Sn IMC will appear as small particles. In fact, these particles have a fine, fibrous structure (Figure 27), which can be seen after selectively etching away the Sn phase. The network of fibers typically appears to surround the Sn dendrite arms, indicating that they form in the interdendritic spaces during eutectic solidification. 12 Figure 27: SAC solder joint viewed a) in cross section and b) after selective electrochemical etching to remove Sn phase Bi in Solution and as a Precipitate As discussed in Section , the three experimental alloys shown in Table 15 were selected from outside the ternary eutectic triangles of the ternary Sn-Ag-Bi and Sn-Bi-Cu alloys so as not to form any ternary eutectic phase with Bi. This is both to avoid the 48

63 formation of a low melting temperature (~138 C) phase, as well as to avoid the formation of primary Bi particles, which form directly from the liquid state. Depending on the amount of Bi in the overall composition, favourable Bi particles may be produced in the final solder joint microstructure; however these particles are the result of solid solution precipitation from within the Sn. Bi precipitation from solid solution has been confirmed by the DSC curve for each studied alloy. In all cases, only one main peak appears, indicating that all liquid was consumed prior to the ternary eutectic solidification which could have resulted in a primary Bi phase. Bi, as an alloying element in a Sn based solder, differs from Ag and Cu in two significant ways: Bi has high solubility in Sn (up to 21wt%), and Bi does not form any intermetallics with Sn. It should also be noted that Sn is not soluble in Bi; any Bi particles that form will therefore be 100% Bi. 3 Sn-Cu Reaction and Formation of Interfacial IMC The chemical bond which forms between Sn and Cu provides a strong interaction between the solder and the substrate. As described earlier, this bond provides a chemical, mechanical and electrical bond between the two materials. Both Cu 6 Sn 5 ( ) and Cu 3 Sn ( ) phases can potentially form during solidification at the interface. It has been shown that the formation of the IMC layer increases with increased temperature. 14 This generally occurs up to a thickness of approximately 2.5µm at which point the IMC appears to form as more needle like faceted rods, initially at the interface, and then through the bulk solder (Figure 28). The phase was not shown to form at initial solidification, but rather occurred over time as an effect of aging. This will be discussed further in Chapter Chapter 3. Many studies looking at the Sn-Cu diffusion couples considered two solid substrates in contact while exposed to heat. When considering a solid Cu with a molten Sn coupling to form a molten-sn/solid- /solid-cu interface, two rate dependent steps need to be considered: 14 49

64 Diffusion of Sn through to solid Cu interface Reaction of Cu and Sn at this interface Figure 28: Solder joint between Sn-based solder and Cu a) rod shaped Cu6Sn5 ( ) b) scallop shaped Cu6Sn5 ( ) IMC layer SAC305 has solidus and liquidus temperatures of 217 C and 220 C respectively. A typical reflow profile (Figure 23) will have a time above solidus of anywhere from 60 to 90 seconds and a maximum temperature of approximately 240 C. The initial formation of the interfacial IMC layer between the Cu and phase is governed by time, temperature and contact area. While the contact area is fixed for a particular build, both the time and temperature can be varied by altering the solder reflow profile or solder alloy composition. Although covered in Chapter Chapter 3, it is worth noting that the interfacial IMC will continue to grow at rates determined by time and temperature. At lower aging temperatures (ambient temperature to about 70 C), the layer grows preferentially indicating that the Cu diffuses at these temperatures, while at higher storage temperatures (above 135 C) both the and layer will form and grow, indicating that Sn diffusion begins to dominate at these increased temperatures. 15 Another possibility is that the Sn supply between the phase and the copper interface is inadequate for the formation of the phase. 16 The morphology of the /solder interface in a typical solder joint tends to be scalloped in shape (Figure 28b) but may flatten to a more planar interface as it grows 50

65 upon aging. The phase, which is sandwiched between the and the Cu (resulting in a / /Cu interfaces), tends to form a more planar morphology. 16,19 It has also been observed that microvoids may form between the copper and interface 17. These Kirkendall voids 18 form as the Cu diffuses into the layer leaving behind voids (Figure 29a). These present a long-term reliability concern as they weaken the interface. Another consideration is the morphology of the interface that is formed. In the SnPb solder system joined to a copper substrate, the Sn is consumed from the area close to the solid copper forming the Cu 6 Sn 5 intermetallic compound. This can leave a localized region of Pb in direct contact with the intermetallic (Figure 29b). This effect may continue under aging conditions as the Cu 6 Sn 5 continues to grow, and the Sn in the immediate vicinity is depleted. 19 This presents a potential reliability concern; the Cu 6 Sn 5 being brittle and the segregated Pb providing a convenient path for crack propagation and eventual fracture. This combination has been observed to fracture predominately along the /solder interface. 19 Figure 29: SnPb solder jointed to Cu substrate a) showing Kirkendall voids 20 b) showing Pb phase pooling at IMC/solder interface (cross section by Zohreh Bagheri) In this manner, Bi has been found to behave in a similar way as Pb. SnBi eutectic solders may therefore have the same reliability concerns related to the pooling of one phase, Bi, along the Cu 6 Sn 5 interfacial IMC. The lower level of Bi used in this study, between 3-7wt%, is not expected to result in this condition. 51

66 The scallop shaped morphology, typically associated with the Cu 6 Sn 5 IMC bond layer, is a result of the solidification of the molten solder and dissolved Cu 6 Sn 5 at the interface. A fast, non-equilibrium cooling rate typical in a solder reflow profile will result in a scalloped morphology. A slower rate will allow for the interfacial IMC layer to form in a smoother morphology and a faster rate will result in a more complicated dendritic, or coral shaped structure. It has also been shown that the structure of the base Cu is a contributing factor to the morphology of the interfacial IMC layer. 21 It was found that Cu 6 Sn 5 grows in a prism-type morphology from single crystal Cu with (001) or (111) grain orientations (Figure 30). Other crystallographic orientations result in a scallop-type structure similar to that found in a polycrystalline Cu substrate. Since the board side Cu pad is polycrystalline and not single crystal, it will form a scallop shaped IMC layer. However, in localized areas where a Cu grain may have a (001) or (111) orientation, a prism structure may form amongst smooth scallops. Figure 30: Typical morphologies of Cu6Sn5 grains formed on single crystal a) (001) Cu and b) (011) Cu21 Growth of the IMC bonding layers can also occur by solid-state diffusion during aging conditions. This will be considered in Chapter Chapter 3. 52

67 4 Cu-Ni-Sn Interface Because of the tendency for Cu to oxidize, protective layers on the Cu surfaces of PWBs are used. For this project an OSP finish was used which is dissolved by the flux in the solder paste during the reflow, and does not need to be considered when evaluating the resulting solder joint. Other board finishes utilize layers of Ni, Au, Pd or other inert metals as protective barriers. These are usually present in very thin layers, often in combination. For example, an ENIG finish is made up of a thin Au layer (0.05µm) over an electroless Ni layer (3-6µm). Additionally, it is common to find an electrolytic Ni barrier layer used between the Cu pad on a BGA component and the attached solder ball; it is for this reason that the Cu-Ni-Sn interaction is important in solder joints formed in this study. Ni also has a much slower diffusion rate than Cu in molten solder therefore it serves as a diffusion barrier to Cu. With a SnPb solder, it was found that a scallop shaped Ni 3 Sn 4 intermetallic formed, however with SAC alloys this phase appeared to be suppressed in favour of a (Cu,Ni) 6 Sn 5 IMC bonding layer. It is the Cu from within the solder which is present in the final IMC composition. 22 This is verified by examining a similar IMC formed using a SnPb solder, in which no Cu is present; it forms a Ni 6 Sn 5 IMC with no Cu present. The (Cu,Ni) 6 Sn 5 phase is made up of many needle-like, hexagonal cylinders with pointy tips or, when seen as a layer has a coral type morphology (Figure 31). 23,24 Snugovsky et al. 25 showed that in the Sn rich corner of the Cu-Ni-Sn system, two quasiperitectic reactions exist: L+(Ni,Cu) 3 Sn 4 Ni 23 Cu 33 Sn 44 +Sn at C 4-1 L+Ni 23 Cu 33 Sn 44 (Cu,Ni) 6 Sn 5 +Sn at C 4-2 Therefore the IMC bond layer formed at the component side of the BGA, which is formed from the Ni barrier layer, and both Sn and Cu from the SAC305 alloy will likely be one of the compounds listed in Equation 2-2 and 2-3. Additionally, the IMC which forms towards the board side of the BGA component will be formed primarily from the Cu layer and the Sn from within the SAC305 solder ball and paste. Ni, which has 53

68 diffused through the molten solder, may also be present. The Ni 23 Cu 33 Sn 44 intermetallic has a smooth, cellular morphology. 26 The formation of Kirkendall voids is also a concern at the (Cu,Ni) 6 Sn 5 Ni(P) interface. Figure 31: IMC formed between SAC solder and Ni(P)Au 23 5 Experimental Set Up 5.1 Test Vehicle Assembly Celestica s RIA3 test vehicle (Figure 32) was used for the work described in this paper. It was originally designed to simulate a typical, medium complexity assembly. It is an 8 x10 surface made up of 12 Cu layers for a total thickness of with an OSP finish. This board is one which is often used to test new, lead-free solders and other material parameters. LQFP176, PBGA256, CBGA64 and MLF20 components were populated, two of each on each board. The BGA components all employed SAC305 ball alloy. The board and SMT components were attached with solder pastes of the alloys of interest. The solder paste alloys were experimental formulations of the alloy and paste flux suitable for the process temperatures. The SAC305 and Senju M42 pastes were commercially formulated by Indium Corp. The reflow was performed in an air environment in a 10 zone oven. One board from each combination listed in Table 15 was held for evaluation of the As Assembled microstructure. 54

69 Figure 32: Test vehicle Table 15: Build matrix for as assembled analysis Alloy Composition Assembly Temperature SAC305 Sn 3%Ag 0.5%Cu 240 C Senju M42 Sn 2%Ag 0.75%Cu 3%Bi 224 C Sunrise Sn 1%Ag 0.7%Cu 7%Bi 222 C Sunflower Sn 0.7%Cu 7%Bi 226 C 5.2 Test Matrix Each of the four components listed in Table 16 was cross sectioned to evaluate the solder mixing, shape of the solder joint, degree of voiding and any possible anomalies. Table 16: Table of components evaluated by cross sectioning Component BGA-256 BGA-256 QFP-176 QFP-176 Reference Designator U204 U205 U1 U2 55

70 5.3 Test Method Components were cut from the test vehicle shown in Figure 32 and mounted in epoxy for cross-sectioning. Each cross-section was ground and polished through the following sequence: 500 and 1200 grade SiC paper, polishing with 6 µm and 1 µm DiaPro diamond suspensions (Struers), and an oxide polish (Struers OP-S). Optical microscopy was performed using a Nikon Measurescope MM-11. Prior to SEM analysis, the samples were carbon coated using an Emitech K950X. SEM microscopy was performed using a Hitachi S-4500 and Hitachi S-3000N with the following EDX systems: Oxford and ThermoScientific respectively. 6 Microstructural Evaluation 6.1 Comparison of Bulk Microstructure EDX was used to evaluate both the overall composition of the bulk material in the solder joint and the composition of Sn grains within the solder joint (Figure 33). Although EDX is a semi-quantitative approach, it was deemed sufficient for purposes of comparison. Scans of the bulk solder joints were taken of cross sectioned samples as seen in Figure 33 a) and b) as well as of the Sn grain (Figure 33c) for each alloy. The results are tabulated in Table 17 and Table 18. Each value represents an average of 3-5 measurements. Figure 33: Area for EDX compositional analysis of a) bulk BGA solder joint, b) bulk QFP solder joint and c) Sn grain 56

71 6.1.1 Bulk Microstructure of QFP Solder Joints Table 17: Composition of QFPs (wt%) Paste Alloy Composition of Paste Alloy Bulk Solder (wt%) Sn Grain (wt%) (wt%) Sn Ag Cu Bi Sn Bi SAC305 Sn 3%Ag 0.5%Cu Senju M42 Sn 2%Ag 0.75%Cu 3%Bi Sunrise Sn 1%Ag 0.7%Cu 7%Bi Sunflower Sn 0.7%Cu 7%Bi The bulk composition of the QFP solder joints is close to that of the original paste composition with the exception of increased Cu concentration and a very small increase in the Sn content. The increase in Cu concentration is expected due to Cu dissolution from the board side Cu pad and from the lead. The leads are plated with a thin layer of Sn, which melts during the reflow process and combines with the bulk solder. It is also expected that the concentration of Bi within the Sn grain would be lower than in the overall composition as can be seen in two of the three Bi-containing alloys in Table 17, Senju M42 and Sunrise. This is a result of the off-eutectic solidification in which primary Sn dendrite arms form prior to the final eutectic solidification of the remaining liquid. The primary Sn grain will continue to grow, driving the Bi content of the liquid at the solid/liquid interface to increase. Finally, the remaining liquid solidifies as a eutectic structure within the interdendritic spaces. The Bi concentration of the eutectic portion of the Sn will therefore be greater than that of the primary Sn dendrite arms. Over time and with the precipitation of Bi from solid solution, it is expected that this concentration gradient will diminish. The Sunflower alloy has a composition which lies on the eutectic line on the Sn-Bi-Cu phase diagram (Figure 13). In this case, a eutectic of Sn + Cu 6 Sn 5 will solidify, instead of the nucleation and growth of Sn grains prior to solidification of other phases. The Bi content of the Sn dendritic arms in Sunflower is similar to the overall final composition of the solder joint, as seen in Table 17. Figure 34a and b show the microstructures of SAC305 and Senju M42 solder joints respectively. Both have very similar microstructures with fine Sn dendritic arms 57

72 surrounded by fine Ag 3 Sn and Cu 6 Sn 5 IMCs solidified as either binary or ternary eutectics with Sn in the interdendritic spaces. In the case of Senju M42, all Bi appears to be in solid solution with Sn, with no visible Bi precipitates. Figure 34c and d show cross sectional images of Sunrise and Sunflower respectively. Both have larger Sn dendritic arms and Cu 6 Sn 5 IMCs. Sunrise shows a small amount of Ag 3 Sn particles, while Sunflower does not have any due to the lack of Ag in the paste composition. In Sunrise, while Bi particles are visible and believed to have precipitated from solid solution of the interdendritic Sn, which exists as either a binary eutectic with Sn + Ag 3 Sn, Sn + Cu 6 Sn 5, or a ternary eutectic of Sn + Ag 3 Sn + Cu 6 Sn 5. As described above, this interdentritic, eutectic Sn, rather than that found in the primary Sn dendrite arms, has a higher Bi content resulting from the solidification of the off eutectic alloy (Table 17). Sunrise shows an overall composition with 5.6wt% Bi, and only 4.0wt% in the primary Sn dendrite arms. The bulk of the Bi, therefore is in the Sn within the interdendritic, eutectic colonies (Figure 35a). Sunflower on the other hand shows the presence of Bi in both interdendritic and within the Sn dendritic arms (Figure 35b). 58

73 Figure 34: Optical images QFP a) SAC305 b) Senju M42 c) Sunrise and d) Sunflower Figure 35: Bi present in a) eutectic colonies in Sunrise and b) throughout bulk solder (eutectic colonies and Sn dendrite arms) of Sunflower 59

74 The presence of Bi, as well as the distinction between Cu 6 Sn 5 and Ag 3 Sn, are visible optically (Figure 36a) however are better distinguished in scanning electron images, particularly in backscatter electron (BSE) images (Figure 36c) in which differences in atomic numbers are more visible. In Figure 36b and c, the Bi is shown in light colour corresponding to a higher atomic number (83) than the surrounding Sn (50). In the secondary electron (SE) image (Figure 36b), Cu 6 Sn 5 and Ag 3 Sn appear indistinguishable, however in BSE (Figure 36c), Cu 6 Sn 5 appears significantly darker while Ag 3 Sn becomes difficult to distinguish from the Sn. Figure 36: QFP solder joint as seen a) optically b) SE SEM and c) BSE SEM The Bi precipitating from the interdendritic Sn appears in both Sunrise and Sunflower as both large particles and as fine dispersions (Figure 37). It is expected that the Bi atoms within the Sn matrix will diffuse together to form a small volume of Bi precipitate, and then will rearrange themselves into the Bi crystal structure. 27 Bi, unlike Cu 6 Sn 5 and Ag 3 Sn, has a non-faceted structure. 60

75 Figure 37: BSE images of a) Sunrise QFP solder joint and b) Sunflower QFP solder joint Bulk Microstructure of BGAs The final composition of a BGA solder joint is made up of both the screened solder paste as well as the component solder ball. The BGA-256, which was examined in this study, is made up of SAC305 solder spheres with a diameter of 30 mil and solder paste of each of the four alloys screened using a 5 mil stencil and a 30 mil aperture. This results in a volume composition of approximately 87% SAC305 solder sphere and 13% solder paste. Table 19 provides a summary of the expected final composition of the solder joint formed from the original solder sphere and the screen solder paste. Table 18 provides a summary of the composition as measured using EDX as shown in Figure 33a. As with the QFPs, the composition is close to the expected composition with the exception of an increased Cu concentration. This is the expected result due to Cu dissolution from the board side Cu pad. The Cu content increase in SAC305 is greater than that of the other three alloys, which can be attributed to the higher process temperature. Cu dissolution from the component side of the solder joint would be mitigated by the Ni barrier layer described in Section 4. Table 22 indicated that some Ni was present in the board side IMC. This would indicate that Ni from the component side would diffuse through the solder during liquid phase mixing and may be present in the final bulk solder composition. Ni was not found to be present and may therefore be present in trace amounts only. 61

76 Table 18: Experimental composition of BGA solder joint Paste Alloy Composition of Paste Alloy Bulk Solder (wt%) Sn Grain (wt%) (wt%) Sn Ag Cu Bi Sn Bi SAC305 Sn 3%Ag 0.5%Cu Senju Sn 2%Ag 0.75%Cu M42 3%Bi Sunrise Sn 1%Ag 0.7%Cu 7%Bi Sunflower Sn 0.7%Cu 7%Bi Table 19: Theoretical composition of BGA solder joint Paste Alloy Theoretical Composition Composition of of Bulk Solder Paste Alloy (wt%) (wt%) Sn Ag Cu Bi SAC305 Sn 3%Ag 0.5%Cu Senju M42 Sn 2%Ag 0.75%Cu 3%Bi Sunrise Sn 1%Ag 0.7%Cu 7%Bi Sunflower Sn 0.7%Cu 7%Bi Table 19 gives the expected final composition of BGA solder joint based on the volume of paste versus the SAC305 solder ball. This calculation does not take into account Cu dissolution. Table 18 is the observed composition measured by EDX. Of note is the increase in Cu content. Senju M42, Sunrise and Sunflower all have similar Cu concentrations between wt%, which is also similar to that found in the QFPs. The solder joint formed fully of SAC305 has almost tripled the Cu content at 1.8wt%. This is likely due to the increase in process temperature; the boards built with SAC305 solder paste reached a maximum temperature of 240 C where as the other solder pastes reached maximum process temperatures between 222 and 226 C. Good mixing of these two constituent materials is demonstrated in Figure 38 and Figure 39. These images show a portion of a BGA solder joint, formed between SAC305 solder 62

77 ball and Sunflower solder paste, close to the board side IMC. The EDX mapping shows the presence of Ag at the lower end of the solder joint. As the screened solder paste, Sunflower, contributed no Ag to the composition, this indicates good mixing from the solder ball. Figure 38: SEM image of Sunflower BGA solder joint 63

78 Figure 39: EDX mapping of intermetallic in Sunflower showing a) image b) Sn c) Ag and d) Cu In all solder joints shown in Figure 40, there are no visible precipitates of Bi found in any of the BGAs solder joints formed. With an overall composition of up to 1wt% Bi, it is not expected that Bi would precipitate from solid solution under equilibrium conditions. BGA solder joints formed between SAC305 solder balls and each of the four alloys under investigation all exhibited Cu 6 Sn 5 flower shaped structures, which were a complex dendritic structure of the Cu 6 Sn 5 pencil shaped rods with branches of the same shape (Figure 41). These are primary IMCs that form directly from liquid, indicating some undercooling conditions. The equilibrium phase diagram of Sn-Ag-Cu (Figure 25) indicates that Cu 6 Sn 5 would be the first phase to solidify and continue for a short time before the solidification of Sn + Cu 6 Sn 5 eutectic. The presence of large and highly 64

79 branched Cu 6 Sn 5 faceted structures indicates some degree of undercooling during the solidification process. In this case, the Cu 6 Sn 5 would continue to grow for an extended period prior to the formation of other phases. These Cu 6 Sn 5 flowers appeared in solder joints formed with all four examined solder pastes. Figure 40: Optical images BGA formed with SAC305 solder balls and paste with the following alloys a) SAC305 b) Senju M42 c) Sunrise and d) Sunflower 65

80 Figure 41: Cu 6 Sn 5 flower structures seen in BGA solder joints made up of SAC305 and a) SAC305 and b) Sunflower solder paste Figure 42 shows the cross sectional image of a number of BGA solder joints under polarized light. Polarized light is used to reveal the grain structure of the solder joint. As is expected from this size of solder joint, a small number of grains comprised the entire solder joint. While Figure 42b appears to show many small grains, it is more likely to be interlaced dendrite arms from two to three grains. 28 If cross-sectioned through another plane, this solder joint would likely look similar to the other 3 alloys. The addition of Bi to the solder does not appear to have a noticeable impact on the number of grains. Further study is needed to understand the impact of Bi on the size and structure of Sn grains. 66

81 Figure 42: Polarized light images BGA a) SAC305 b) Senju M42 c) Sunrise and d) Sunflower 6.2 Comparison of Interfacial IMC Layers IMC bond layer thicknesses are difficult to measure because of their uneven morphologies; scalloped in the case of Cu 6 Sn 5 and a needle shaped coral structure in the case of (Cu,Ni) 6 Sn 5. In order to determine a representative thickness, the following method was used. For BGA solder joints, a cross-section was evaluated using the SEM. Three solder joints were selected, one from each side and one from the middle of the component. For each joint, three locations were selected, again one from each side and one in the middle of the solder joint. For these nine locations, an image was recorded at 3000x magnification. This image was divided along the horizontal length into six equal segments. Five measurements were then taken between the six segments. A total of 45 measurements were taken, representing various locations across the component. This methodology allows for any difference in IMC thickness, due to thermal differentials 67

82 across a component, to be captured. Also, it ensures random locations are selected as opposed to exclusively selecting the high or low points for measurement. This same process was used for determining the IMC thickness of the QFP solder joints, however only two leads were analyzed, resulting in a total of 30 measurements. There is no statistically significant difference in the thickness measurements of the IMC layers formed using each of the four alloys on the board side of the BGA. Each solder paste formed a board side interfacial IMC layer with a normally distributed thickness of approximately equal variance. The variation in thickness (Figure 43) is large, which can be attributed to the scalloped shape of the (Cu,Ni) 6 Sn 5 IMC bond layer, which forms at the solder/board interface (Figure 45b). Table 20 and Table 21 provide a summary of test for equal variance ( ) and analysis of variance (ANOVA) testing of the means (µ) of the various alloys for BGA and QFP solder joints respectively. These tables provide probability (p-values) for each test. In all cases, the null hypothesis (H 0 ) and alternate hypotheses (H a ) are as follows: H 0 : µ SAC305 =µ Senju M42 =µ Sunrise =µ Sunflower and SAC305 = Senju M42 = Sunrise = Sunflower H a : at least one µ is different and at least one is different Values in Table 20 and Table 21 which are bold italicized represent points with a p-value less than This indicates that, within a 95% CI, the H 0 is rejected and the H a is assumed to be valid. In the case of the board side IMC layer of BGA solder joints, the H 0 is valid, the IMC layers are statistically similar. In the case of the component side IMC bond layer, there is a statistical difference in the thickness measurements: H 0 is rejected due to the very low probability. SAC305 and Sunrise appear to have a greater mean thickness than Senju M42 and Sunflower, however there are many outliers within the distribution, corresponding to the random, long, needle-shaped shapes which occur in the (Cu,Ni) 6 Sn 5 phase (Figure 46). These points were removed and the analysis was performed a second time. The new distribution was 68

83 IMC Thickness (µm) also found to that the mean IMC thicknesses where not equal however this time, it was only Sunflower exhibited a greater mean thickness. The new analysis also included many new outliers. The jagged nature of these IMC layers makes it very difficult to compare mean thicknesses of the component side IMC; the analysis was deemed to be inconclusive SAC305 Senju M42 Sunrise Board Side Sunflower SAC305 Senju M42 Sunrise Component Side Sunflower Figure 43: IMC measurements of BGA solder joints after reflow (U204) Table 20: Results of ANOVA test for equal variance and compare means of IMC Location Variable / p value thickness of the BGA IMC layer 69 Equal variance Total IMC ANOVA Board Side 0.36* Component Side 0.09** 0.07** (with outliers) (without outliers) *Bartlett s Test for equal variance used since distributions where found to be Normal. **Levene s Test for equal variance used since distributions was found to be non-normal.

84 IMC Thickness (µm) The mean thicknesses of the IMC bond layer formed in the QFP solder joints are similar for all four alloys in the as manufactured state (Figure 44). It should be noted that SAC305 has a thicker IMC layer than the other three alloys at t = 0 if a 93% CI is used. This is expected since the process temperature is higher than for the other three alloys SAC305 Senju M42 Sunrise Board Side Sunflower SAC305 Senju M42 Lead Side Sunrise Sunflower Figure 44: IMC measurements of QFP solder joints after reflow (U2) Table 21: Results of ANOVA test for equal variance and compare means of IMC thickness of the QFP IMC layer Total IMC Variable / p value Equal variance ANOVA Board Side Location Component Side

85 Figure 45: IMC bond layers formed on a) QFP components and b) BGA components The composition of the IMC bond layers was determined using EDX analysis. Although EDX provides only a semi-quantitative analysis of the composition, it has been deemed sufficient for the purpose of identifying phases. The Sn portion of this IMC layer is contributed from the solder, all four solder alloys contain between 91.3 and 96.5 wt% Sn. The Cu portion of the IMC is contributed from both the solder (all four solder alloys contain between 0.5 to 0.75 wt% Cu), and the joining material (lead or board side Cu pad). Table 22: IMC type BGA Paste Alloy Component Board Side Side SAC305 (Cu,Ni) 6 Sn 5 Ni 23 Cu 33 Sn 44, (Cu,Ni) 6 Sn 5 Senju M42 (Cu,Ni) 6 Sn 5 Ni 23 Cu 33 Sn 44, (Cu,Ni) 6 Sn 5 Sunrise (Cu,Ni) 6 Sn 5 Ni 23 Cu 33 Sn 44, (Cu,Ni) 6 Sn 5 Sunflower (Cu,Ni) 6 Sn 5 Ni 23 Cu 33 Sn 44, (Cu,Ni) 6 Sn 5 QFP Board Side Lead Side Cu 6 Sn 5 Cu 6 Sn 5 Cu 6 Sn 5 Cu 6 Sn 5 Cu 6 Sn 5 Cu 6 Sn 5 Cu 6 Sn 5 Cu 6 Sn 5 The IMC bond layers which form on the BGA components are a combination of the solder paste (one of the four tested alloys), SAC305 from the component solder ball, and 71

86 the material finish at both the component interface and the board side interface (Figure 45b). The component side interface consists of a Cu pad with a Ni finish barrier/combined solder (paste alloy + ball alloy). On the board side, the interface consists of a Cu pad/combined solder. EDX analysis shows that the IMC formed at the component side is some combination of Ni 23 Cu 33 Sn 44 and (Cu,Ni) 6 Sn 5 (Figure 46). The morphology of these needle-like, coral shaped IMCs accounts for the large spread in thickness measurements (component side of Figure 43) and the number of large values found outside of the 75% quartile denoted as * in Figure 43. Figure 46: IMC formed on BGA at component side spectrum 1) Ni23Cu33Sn44 and spectrum 2) (Cu,Ni)6Sn5 Figure 47: Typical morphology of a) Cu 6 Sn 5 IMC as formed on a QFP solder joint with Sunflower b) and c) (Cu,Ni) 6 Sn 5 and Ni 23 Cu 33 Sn 44 IMCs respectively both formed on a BGA with SAC305 solder paste 72

87 7 Summary of Findings Each of the four alloys under investigation produced solder joints of good shape and acceptable degree of mixing on both QFPs and BGAs. The Bi content of the three Bicontaining alloys allowed for melting temperatures of C lower than that of SAC305 without negatively impacting the resulting solder joint formation. Differences in microstructure have been observed related to the decrease in Ag content, addition of Bi and varying amounts of Cu dissolution resulting from different process temperatures. The QFP solder joints formed with Senju M42 were very similar to those formed with SAC305, with all of the Bi present in the Sn phase. This dissolved Bi may act as a solid solution strengthener, the effect of which needs to be verified in mechanical testing. Both Sunflower and Sunrise showed some degree of Bi precipitating out of solid solution. Both fine and medium sized particles were observed. This may achieve some precipitation hardening within the solder joints, but again will need to be verified by further testing. The final composition of BGAs was similar to that of SAC305, which was the main contributing alloy to the final composition, and in all cases having a final Bi concentration of 1wt% or less. All of the Bi was present in solution within the Sn phase. The presence of many large, complex Cu 6 Sn 5 flower-shaped imtermetallics indicates some degree of undercooling in solder joints formed in the BGAs using SAC305 as the ball material and each of the four alloys as the paste material. The QFP interfacial IMCs at the board interface and lead interface were similar for all four alloys. SAC305 results in a thicker IMC layer forming within a 93%CI. In all cases Cu 6 Sn 5 was identified. The BGA also showed no difference in the type of interfacial IMCs that formed at both the board side and at the component side. There appears to be no statistically significant difference in the layer thicknesses of the IMC formed on BGA, although this is harder to evaluate with the Ni 23 Cu 33 Sn 44 layer that forms at the component side of the BGA due to the variable nature of the IMC shape. 73

88 8 References 1 A. Zbrzezny Characterization and Modeling of Microstructural Evolution of Near- Eutectic Sn-Ag-Cu Solder Joints Ph.D. Thesis, Dept. MSE, Univ. Toronto, Toronto, Canada, L. Nie M. Osterman and M. Pecht Copper Pad Dissolution and Microstructure Analysis of Reworked Plastic Grid Array Assemblies, at IPC APEX EXPO Conference, Las Vegas, NV, M. Kelly, C. Hamilton and P. Snugovsky Have High Cu Dissolution Rates of SAC305/405 Alloys Forced a Change in the Lead Free Alloy Used During PTH Processes, SMTA Proc. Pan Pacific Microelectronic Jan D.A. Porter and K.E.Easterling, Thermodynamics and Phase Diagrams in Phase Transformations in Metals and Alloys, 2 nd ed. UK: Chapman & Hall, 1992, ch. 1, pp L.Snugovsky, P. Snugovsky, D. Perovic; T. Sack; J. W. Rutter Some aspects of nucleation and growth in Pb free Sn Ag Cu solder, Mater. Sci. Technol. 21 (2005) A.R. Zbrzezny Microstructure Characterization of Sn-Ag-Cu Lead-Free Solder Solidified at Different Cooling Speeds, Microsc. Microanal. 8 (Suppl. 2), T-K. Lee, T. Bieler, C-U Kim, H. Ma Phase Equilibria and Microstructure of Sn Ag Cu Alloys in Fundamentals of Lead-Free Solder Interconnect Technology, New York, Springer, 2015 ch 3, pp Y Takamatsu, H Esaka, K Shinozuka Formation Mechanism of Eutectic Cu 6 sn 5 and Ag 3 Sn after Growth of Primary Sn in SN-Ag-Cu Alloy, Materials Transactions Vol. 52, No. 2, pp , J.W. Elmer, E.D. Specht, M. Kumar Microstructure and In Situ Observations of Undercooling for Nucleation of -Sn Relevant to Lead-Free Solder Alloys, Journal of Electronic Materials, Vol. 39, No.3, P. Snugovsky, Z. Bagheri, C. Hamilton Microstructure and Reliability Comparison of Different Pb-Free Alloys Used for Wave Soldering and Rework, Journal of Electronic Materials, Vol. 38, No. 12, T. Hurtony A Bonyár, P Gordon, G Harsányi Investigation of intermetallic compounds (IMCs) in electrochemically stripped solder joints with SEM Microelectronics Reliability 52,

89 13 T. Hurtony, A Bonyár, P Gordon, G Harsányi. Investigation of intermetallic compounds (IMCs) in electrochemically stripped solder joints with SEM Microelectronics Reliability 52, A. So and Y.C. Chan Reliability Studies of Surface Mount Solder Joints Effect of Cu-Sn Intermetallic Compounds IEEE Transactions on Components, Packaging, and Manufacturing Technology Part B, Vol 19, No 3, August Z Mei, A.J. Sunwoo, J.W. Morris Analysis of Low-Temperature Intermetallic Growth in Copper-Tin Diffusion Couples. Metallurgical Transactions A, Volume 23A, March K.H. Prakash, T Sritharan Interface Reaction Between Copper and Molten Tin-Lead Solder, Acta Mater. 49 (2001) S. Kumar, C.A. Handwerker and M.A. Dayananda Intrinsic and Interdiffusion in Cu- Sn Systems, Journal of Phase Equilibria and Diffusion Vol. 32 No. 4, K-N. Tu Copper Tin Reactions in Bulk Samples in Solder Joint Technology : Materials, Properties and Reliability New York, Springer, 2007, ch. 2, sec , pp A.J. Sunwoo, J.W. Morris, G.K. Lucey, The Growth of Cu-Sn Intermetallics at a Pretinned Copper-Solder Interface, Metallurgical Transactions A, Volume 23A, April S. Kumar, J. Smetana, D. Love, J. Watkowski, R.Parker, C.A. Handwerker, Microvoid Formation at Electrodeposited Copper-Solder Interface During Annealing: A Systematic Study of Root Cause, presented at SMTAI, Chicago, Il, H.F. Zou, H.J. Yang, Z.F. Zhang A Study on the Orientation Relationship Between the Scallop-Type Cu 6 Sn 5 Grains and (011) Cu Substrate using Electron Backscattered Diffraction Journal of Applied Physics, 106, (2009) 22 K-N. Tu Solder Reactions on Nickel, Palladium, and Gold in Solder Joint Technology: Materials, Properties and Reliability New York, Springer, 2007, ch. 7, pp J-Y Tsai and J. Gaida The Interaction between SnAgCu Solder and Ni(P)Au, Ni(P)PdAu UBMS 24 K Sweatman, J. Read, T. Nishimura and k. Nogita The Effect of Microalloy Additions on the Morphology and Growth of Interfacial Intermetallic in Low-Ag and No-Ag Pb-Free Solders, presented at SMTAI, Chicago, Il, L. Snugovsky, P. Snugovsky, D.D. Perovic, J. W. Rutter Phase Equilibria in Sn Rich Corner of Cu-Ni-Sn System Materials Science and Technology, 2006, Vol 22, No 8. pp

90 26 P. Snugovsky, E. Kosiba, J. Kennedy, Z. Bagheri, M. Romansky, M. Robinson, J.M. Juarez, Jr., J.Heebink Manufacturability and Reliability Screening of Lower Melting Point Pb-free Alloys Containing Bi, in IPC APEX EXPO Conference, San Deigo, CA, D.A. Porter and K.E.Easterling, Diffusional Transformation in Solids in Phase Transformations in Metals and Alloys, 2 nd ed. UK: Chapman & Hall, 1992, ch. 5, pp B. Arfaei, N. Kim and E.J. Cotts Dependence of Sn Grain Morphology of Sn-Ag-Cu Solder on Solidification Temperature, Journal of Electronic Materials, Vol. 41, No

91 Chapter 3 Accelerated Thermal Cycling 1 Accelerated Testing for Reliability Analysis Reliability can be defined as the probability of a product to meet specifications over a given time period while subjected to determined environmental conditions. 1 The reliability of any electronic assembly depends on the reliability of each individual element. An assembly is likely to fail due to component failure in the short term and due to solder attach failure in the long term. 2 The solder attach, which include the solder joint and the material to which it adheres, particularly in the case of surface mount devices, presents a complicated situation in that it acts as both the electrical contact as well as the mechanical attachment. 3 In field operation, a solder attach will experience loading conditions in the form of: (i) mechanical load, (ii) vibration, (iii) thermal shock and (iv) differential thermal expansion. The accumulated damage caused by one or more of these stress conditions will eventually lead to wear-out failure. Therefore, determination of the useful life of an electronic product is contingent upon the understanding of the failure rate of the surface mount solder attach. Accelerated testing is used to obtain more information than would be practical, or even possible, under normal field conditions where years may pass before a sufficient number of failures occur to determine the reliability of a product. Accelerated testing is also essential in determining the effects of changes to product design through its design cycle. Accelerated conditions may include a test environment in which conditions are more severe than that experienced during normal equipment use or by increasing the frequency in which a condition is applied to a product. In both cases, care should be taken to avoid the introduction of failure mechanisms, which would not be encountered in field operation of a product. 4 If failure mechanisms are maintained, the accelerated data can be used to extrapolate the expected failure rates in field conditions. 2 Of the four loading conditions described above, thermomechanical fatigue resulting from cyclic differential thermal expansion is the primary failure mechanism of surface mount 77

92 solder joints and the focus of this chapter. 3,5 This form of fatigue is the combination of plastic and creep deformation which occurs as a solder attach is exposed to cyclical heating and cooling through powering on and off, as well as exposure to environmental conditions. Accelerated Thermal Cycling (ATC) is used to test the response of a product to rapidly repeating changes in temperature to a level higher than normally experienced by a product in field conditions. In this case, the applied cyclical stress is in the form of differential thermal expansion and contraction between the various materials within a circuit board. These differences are defined by the coefficient of thermal expansion (CTE) of the differing materials that make up the system. In a circuit board, there are two levels of recognized differential thermal expansion: 1. Global: thermal expansion mismatch between components and substrate, as illustrated in Figure Local: thermal expansion mismatch between solder and material to which it is bonded 6 Figure 48: Representation of stress generated in circuit due to CTE mismatch 7 Stress-strain hysterisis loops, such as the one illustrated in Figure 49, are often used to describe the cumulative stress history experienced by a solder attach during ATC. As the temperature increases, the differential expansion of the various materials results in the 78

93 initial applied stress, which leads to plastic deformation. During the hot dwell (or hold) period, which generates stresses lower than the yield strength of the solder, creep and stress relaxation within the solder material occur. During accelerated testing, there may or may not be enough time allotted for full stress relaxation as would likely exist in field conditions. Stress is then applied in the opposite direction the stress axis representing an absolute value during the cold cycle. As the cycles are repeated, the solder joint fails due to low-cycle fatigue and can be described by the Coffin-Manson relationship given in Equation 3-1: 1( p 1 m N f C ) 1-1 where N f is the average number of cycles to failure, C 1 and m are material constants. 8 Greater temperature fluctuations or greater differences in CTE mismatch will result in higher values of plastic shear strain ( p ). This value will also be impacted by component geometry. For example leaded components will result in lower shear plastic strain than leadless components. Ramp rate and dwell time, at the temperature extremes, will also impact the shape of the hysteresis loop and the strain level. The Coffin-Manson model assumes that plastic strain is the main deformation mechanism. This simplified model has since been revised to include the more dominant creep effect, as well as cyclical parameters. These models are presented in IPC-SM-785 and take into account such factors as component and solder fillet geometry, potential cyclic fatigue damage at complete stress relaxation, and fatigue ductility coefficients. These models are also presented with numerous caveats indicating which conditions are required for each model and what limitations may exist. 2 Modeling of complex creep-fatigue in solder attach, particularly for new solder alloys, is a continuing field of research. 79

94 Figure 49: Hysteresis loop for thermal cycle 9 In order to relate the results of ATC testing, or any Accelerated Life Testing (ALT), to life condition, data from at least two conditions need to be used to establish a trend line in order to extrapolate to field condition. Low-level and a high-level acceleration condition should be chosen. Three or more conditions would allow for determination of non-linear relationships. Low-level acceleration should produce a mean time-to-failure (MTTF) of about times shorter than the actual field life. High-level acceleration should produce a MTTF of approximately times shorter than actual field life. 2 Acceleration Factors (AF) are then calculated as the ratio between the two conditions and then extrapolated from the chosen conditions to field conditions. ATC qualification test conditions for various applications have been determined and compiled in specification IPC-9701A 6. These tests have been devised as minimum criteria for accelerated testing for qualification purposes. While Appendix 1 of this specification does provide guidance and examples for determining AF, it is not necessary to determine these for every test. Rather, the minimum requirements per industry have been established. This specification also states at least 32 samples per condition should be tested and the test should run for the required duration or at least until 63.2% failure in order to characterize the failure distribution. Tests that are stopped at the end of the 80

95 required Number of Thermal Cycle (NTC) level with insufficient (or no) failures will be characterized by failure analysis (FA) alone. FA is to be performed on a minimum of three randomly chosen samples. Table 23 provides a summary of the test conditions laid out in this standard and Table 24 provides some examples of worst-case scenarios for various products in field conditions. Table 23: Sample of temperature cycling requirements Table 4-1 in IPC-9701A 6 Low High Test Mandated Temperature Test Duration Temperature Temperature Condition Condition Ramp Rate Dwell Dwell NTC-E TC1 0 C to 100 C 6,000 cycles (preferred for TC1) NTC-C 10 minutes (+0 C/-5 C) 10 minutes (+5 C/-0 C) 20 C/minute TC4-55 C to 125 C 1,000 cycles (preferred for TC4) Table 24: Worst case use environments of SMT 6 Application T min T max T ( C) ( C) ( C) Typical service Acceptable Cycles/year life (years) failure risk (%) Consumer Military Aircraft a) b) c)

96 Note in Table 24, T is not simply the difference between T max and T min representing the absolute maximum and minimum temperatures experienced in the given environment, but rather the expected worse-case scenarios which occur in a given service condition. 2 Solder in ATC testing has been shown to fail according to a creep-fatigue model resulting from the applied thermomechanical fatigue stresses. The mechanical properties of solder are temperature dependent and therefore change continuously over the course of a fatigue cycle, both in the bulk microstructure of the solder as well as in the IMC layer formed between the solder and the Cu substrate. There are also cumulative effects of repeated cycles. 10 The total strain ( T ) can be expressed by Equation 3-2 where e and p are the elastic and plastic strains respectively and c represents creep strain. T 1-2 e p c During the high temperature dwell period of the thermal cycle, creep and stress relaxation are the dominant strain evolution mechanisms. During ramp up and down, plastic and elastic strain deformations are dominant unless this transition is slow enough to allow for full stress relaxation, which is not typically the case in accelerated testing. It is the repeated plastic and elastic strain which constitutes fatigue fracture. 2 Microstructural Evolution 2.1 Changes to Bulk Solder Notably, many of the tests and models described above have been developed for SnPb solders, which have been tested over decades. It has been shown that the microstructural response to creep and fatigue behavior of Sn based SAC alloys is very different than that of SnPb. 11 Therefore, it is possible the models and even the applicable test conditions need to be updated. In SnPb solder joints, the initial response to applied thermomechanical stress is grain coarsening in an attempt to reduce the internal energy in the smaller grains. Both the applied shear strain which results from CTE mismatch and the elevated temperature aid in grain coarsening. In eutectic SnPb, the Sn rich regions and 82

97 Pb regions will coalesce into larger islands. During coarsening, it should be noted that the IMC attach layer also thickens a Cu 3 Sn region may form between the Cu substrate and the Cu 6 Sn 5. Immediately adjacent to the IMC a layer of Pb-rich phase will form as the local Sn has been consumed in the IMC growth process. Pb also tends to pool around the grain boundaries of the Sn rich phase. Grain coarsening is accomplished via grain boundary sliding and grain boundary diffusion-induced migration. 5 Microvoids begin to appear along grain boundaries, interphase boundaries and interfaces with IMCs. Microvoids continue to propagate along grain boundaries or other interfaces until they connect to form microcracks and eventually catastrophic macro level cracks. Figure 50 illustrates a typical SnPb solder attach a) before any accelerated testing and b) after 3000 cycles of harsh ATC. These images, taken at the same magnification, illustrate the degree of grain coarsening, IMC growth and phase coalescence that occurs during repeated stress cycles. While this joint did not fail catastrophically after 3000 cycles, a large crack is propagating through the bulk solder along a Pb-rich region and, if testing had continued, would have lead to a full electrical and mechanical fail. Figure 50: SnPb solder a) before testing and b) after 3000 cycles -55 C to 125 C 12 The microstructure of solder attach formed with SAC type alloys are found to behave differently from that of eutectic SnPb. As described in Chapter Chapter 2, these solders are made up of up to 98% Sn with small amounts of Ag and Cu present as intermetallics. The Sn matrix, as well as the Ag 3 Sn and Cu 6 Sn 5 precipitates all influence the creep response of the solder attach. As the strain increases, an initial coarsening of the IMC particles occurs. Recrystallization of the Sn grains takes place in the area of high strain, 83

98 usually close to the bulk solder interface with the attach IMCs. The grain boundaries formed between these new grains, as well as some large precipitates, provide sites along which voids can form (at the triple points) and fractures can easily propagate. 10,13 Additionally, it is during the hot dwell cycle that the brittle IMC attach layer may grow. 7 Figure 51 shows a solder attach made up of SAC305 a) before testing and b) after 3000 cycles of harsh testing. From Figure 51a) we can see the dendritic arms of the Sn appear relatively small and evenly spaced with small dispersiods of Ag 3 Sn in the interdendritic spacing. After ATC, the Ag 3 Sn particles appear to have coalesced into fewer, larger particles. Additionally, the IMC layers formed between the lead and solder as well as between the solder and copper pad have increased in size and appear to have two phases, Cu 3 Sn and Cu 6 Sn5 as seen in Figure 51b). A Cu 3 Sn layer was not detectable at zero time. Finally, the dendritic structure does not appear to be well defined after thermal cycling. Polarized light or electron backscatter diffraction (EBSD) is needed to properly view the grain structure. Figure 52 shows an example of a BGA solder joint that failed during thermal cycling; recrystallized grains appear in the region where high shear strength is experienced, and where the final crack propagated. Using EBSD, it has been proposed that this recrystallization occurs ahead of the propagating crack. 10 Fractures can easily propagate along the regions created under creep conditions, notably between grain boundaries and along the bulk solder/imc interface. Figure 51: SAC305 solder a) before testing and b) after 3000 cycles -55 C to 125 C 12 84

99 Figure 52: SAC305 after ATC shown with a)polarized light and b)ebsd mapping Changes to Interfacial IMC during Accelerated Thermal Cycling A good interfacial IMC layer is required for a properly formed solder joint, as it provides the mechanical, electrical and thermal connection between the bulk solder and the Cu substrate. This layer, however tends to be brittle and if a very thick layer forms or develops, it may become a reliability concern. Interface layers also tend to be preferential sites for crack formation and propagation. Additionally, the IMC layer in solder joints may be a reliability concern because it consumes Cu from the substrate, Cu being the dominant diffusing species. 14 The consumption of Cu may result in a weakened Cu pad as described in 2.1. Cu 3 Sn was not detected upon initial solidification (Chapter Chapter 2); it forms as a result of solid-state reaction during thermal aging at temperatures above 60 C. 17 The Cu 3 Sn growth occurs between the Cu 6 Sn 5 and the Cu substrate and is governed by the availability of Cu from the substrate and Sn either from the already present Cu 6 Sn 5 as described in equation 3-3 or Sn from the bulk solder diffusing through the Cu 6 Sn 5 substrate as described in equation 3-4. The initial formation of Cu 6 Sn 5, and its continued growth are dependent on Sn from the bulk solder and Cu from the substrate. The Cu is either present during the initial solidification reaction, or diffuses through the Cu 3 Sn layer. Both are described by equation

100 Cu 6 Sn 5 + 9Cu 5Cu 3 Sn 2-1 3Cu + Sn Cu 3 Sn 2-2 6Cu + 5Sn Cu 6 Sn The Cu 3 Sn layer is of concern for three reasons: it is a brittle layer more so than Cu 6 Sn 5 or bulk solder, it is associated with large volume shrinkage, and the growth of a Cu 3 Sn layer next to Cu is often accompanied by Kirkendall voids along the interface of the two phases. 16 These small voids may form a weak interface. Both Ag and Cu additives to the solder have been found to be beneficial in suppressing the growth of the Cu 3 Sn layer. 17 Ag does not participate in the interfacial IMC layer at all and can therefore only act as an influence by retarding the diffusivity of Sn through the bulk solder towards the IMC layer. Other alloying elements (e.g. Ni, Zn) have also been found to retard the growth of Cu 3 Sn and in some instances Cu 6 Sn 5. In these cases, trace amounts of the element are found in, or at the interface of the IMC layer and therefore are thought to participate directly in the reaction or provide a diffusion barrier between the layers. 18 The change in thickness of the intermetallic layers can be described by the Arrhenius equation (3-6) where x is the thickness, x 0 is the initial thickness in m, t is time in s, T is temperature in K, R is the universal gas constant (8.314 kj/mol K) and A, Q and n are material constants. 20 n x x0 At exp( Q ) 2-4 RT 2.3 Effects of Bi Bi as an alloying element is expected to have an overall influence on the reliability of a solder joint in two ways. The first is the reduced process temperature due to the addition of Bi, which may result in a reduced thickness of the interfacial IMC layer as well as 86

101 reducing some of the deleterious effects of high temperature on both board material and components. 19 The second involves modification of the bulk solder microstructure. Sn phase is considered to have most potential for modification with alloying as it makes up most of the volume within a SAC solder joint. Alloys with Bi have been explored as a means to improve thermomechanical properties through solid solution strengthening and precipitate hardening. 20 In a previous study, for example, after 3000 cycles harsh environment ATC (-55 C to 125 C), Bi which was previously present in solid solution with Sn precipitated out in an evenly distributed manner, reducing microstructural degradation. 12 Other studies have shown that Bi segregates at the grain boundaries, potentially reducing creep resistance. 21 The influence of Bi on the formation of interfacial IMC layers has also been explored. It has been shown that Bi does not participate in the formation of either intermetallic compound layers (Cu 6 Sn 5 or Cu 3 Sn), but there is a higher concentration within the Sn close to the Cu 6 Sn 5 suggesting that Bi is rejected from the Sn as Sn is contributed to the formation of Cu 6 Sn In this thesis, the behavior of Bi containing alloys is investigated. Also, the impact of reducing the amount of Ag in SAC-based solders is also examined. 3 Experimental Setup The work in this section forms a screening experiment in which the performance of selected alloys was evaluated during thermal cycling under two different conditions. The intention was to provide data for determining the suitability of these solders for more statistically relevant reliability testing, which is too costly to perform on all combinations of solder, board material and temperature conditions. These tests were performed in accordance with test conditions outlined in IPC-9701A 6 and IPC-SM The statistical requirements of these specifications were not adhered to due to limited resources. 87

102 ATC testing assumes that solder joints were properly formed and no wetting issues exist, which would result in infant mortality due to defective solder joints. 6 The quality of solder joints used in this study was demonstrated to be satisfactory in Chapter Chapter 2. In addition to testing a statistically relevant number of samples, IPC-9701A states that the duration of testing would ideally generate at least 63.2% failures within that sample set in order to properly evaluate reliability and possibly determine acceleration factors. This specification states that if an insufficient number of failures, or no failures at all, are generated at the end of the number of thermal cycles (NTC) level, then failure analysis should be performed on a random selection of parts (a minimum of three per test variation) to verify that no failure occurred. This failure analysis can also be useful in determining whether or not any microstructural changes occurred during the testing. Finally, as this testing is intended to address solder joint reliability, only those failures determined to be the result of thermomechanical component/board interaction will be considered. Board level failures such as via cracks, delamination, will not be included in the data set for any solder joint reliability calculations Materials Three Bi-containing alloys, Senju M42 containing 3%Bi and Sunrise and Sunflower each containing 7wt% Bi, with low or no Ag, were tested against SAC305. Two board materials were tested; a normal T g material one which is typically used for SnPb applications around 150 C, and a high T g material one which is now required for most Pb-free applications at around 170 C. Table 25 summarizes the combinations of solder and board material tested. As described in 2.2, the lower T g board materials that had been previously used with SnPb solders had a lower elastic modulus and were less susceptible to pad cratering, delamination and warpage failures. 88

103 Table 25: Build matrix for ATC testing Alloy Composition Board Material Assembly Temperature SAC305 Sn 3%Ag 0.5%Cu 240 C Senju M42 Sn 2%Ag 0.75%Cu 3%Bi High T g 224 C Sunrise Sn 1%Ag 0.7%Cu 7%Bi (170 C) 222 C Sunflower Sn 0.7%Cu 7%Bi 226 C SAC305 Sn 3%Ag 0.5%Cu 240 C Senju M42 Sn 2%Ag 0.75%Cu 3%Bi Normal T g 224 C Sunrise Sn 1%Ag 0.7%Cu 7%Bi (150 C) 222 C Sunflower Sn 0.7%Cu 7%Bi 226 C 3.2 Test Vehicle Celestica s RIA3 test vehicle, shown in Figure 53, was selected as the test vehicle. It was designed to simulate a typical, medium complexity SMT assembly. It is an 8 x10 surface made up of 12 copper layers for a total thickness of with an Organic Solderability Preservative (OSP) finish. This board is often used to test new, lead-free solders and other material parameters. LQFP176, PBGA256, CBGA64 and MLF20s were populated, two of each per board. The BGA components all had SAC305 ball alloy and all component groups were reflowed in an air environment. Figure 53 and Table 26 shows the test vehicle and highlights the monitored components. Figure 53: Test vehicle with monitored components 89

104 3.3 Test Strategy Table 26: Monitored components for ATC testing Component Reference Designator A B R400-R R420-R439 CSP64 U208 CSP64 U209 MLF20 U5 MLF20 U6 PBGA256 U204 PBGA256 U205 QFP176 U1 QFP176 U2 All components were monitored using electrical data loggers; failures were defined as a 20% increase in nominal resistance over five consecutive scans. Failures were cut from the boards, and then the rest of the components were returned to the chamber for further testing. Each of the two thermal profiles was allowed to run for a preset number of cycles. Figure 54: Card set up in thermal cycling chamber 90

105 C Thermal Cycling This portion of the testing was done in accordance with IPC-9701A Test Condition 1 (TC1) of 0 C (+0/-5 C) to 100 C (+5/-0 C) for Number of Thermal Cycles (NTC-E), which is preferred for TC1, of 6000 cycles. The boards under test, listed in Table 27 were set up in a chamber using a racking system which allowed for airflow around the boards (Figure 54). The final thermal cycle profile is shown in Figure 55. Table 27: Test matrix for 0 C to 100 C ATC Alloy Composition Board Material # of Boards Tested SAC305 Sn 3%Ag 0.5%Cu High T g (170 C) 4 Normal T g (150 C) 3 Senju M42 Sn 2%Ag 0.75%Cu High T g (170 C) 6 3%Bi Normal T g (150 C) 5 Sunrise Sn 1%Ag 0.7%Cu 7%Bi High T g (170 C) 5 Normal T g (150 C) 6 Sunflower Sn 0.7%Cu 7%Bi High T g (170 C) 5 Normal T g (150 C) 5 Figure 55: Chamber profile for 0 C to 100 C thermal cycling 91

106 3.3.2 Harsh Environment (-55 C to 125 C) Thermal Cycling This portion of testing was done in accordance with IPC-9701A Test Condition 1 (TC4) of -55 C (+0/-5 C) to 125 C (+5/-0 C) for NTC-C of 1000 cycles. Notably, the maximum temperature selected for the thermal cycle profile should be 25 C lower than the T g of the board material. 6 In this case the Normal T g board material is 150 C, while it is exactly 25 C lower than the T g, does not allow for the 10 C leeway on the hot end of the cycle making the board material vulnerable to damage during testing. The boards under test are listed in Table 28. The final thermal cycle profile is shown in Figure 56. Table 28: Test matrix for C ATC Alloy Composition Board Material # of Boards Tested SAC305 Sn 3%Ag 0.5%Cu High T g (170 C) 3 Normal T g (150 C) 3 Senju M42 Sn 2%Ag 0.75%Cu 3%Bi High T g (170 C) 5 Normal T g (150 C) 5 Sunrise Sn 1%Ag 0.7%Cu 7%Bi High T g (170 C) 5 Normal T g (150 C) 6 Sunflower Sn 0.7%Cu 7%Bi High T g (170 C) 5 Normal T g (150 C) 5 92

107 Figure 56: Profile for -55 C to 125 C thermal cycling 3.4 Post ATC Evaluation and Failure Analysis Samples from groups exposed to each of the two ATC profiles were evaluated to determine both the failure modes and to examine the changing microstructure over the course of the test. Samples were selected at time of failure, at the halfway point of each test and at the completion of each test. Each sample was mounted in epoxy, and then ground and polished through the following sequence: 500 and 1200 grade SiC paper, polishing with 6 µm and 1 µm DiaPro diamond suspensions (Struers), and an oxide polish (Struers OP-S). Optical microscopy was performed using a Nikon Measurescope MM-11. Prior to SEM analysis, the samples were carbon coated using an Emitech K950X. SEM microscopy was performed using a Hitachi S-4500 and Hitachi S-3000N with Oxford and ThermoScientific EDX systems respectively. 93

108 4 Results 4.1 Reliability and Failure Analysis Results Using IPC-9701A as a guideline, this test focuses on solder joint reliability; other modes of failure, for example those attributed to the laminate material, were removed from consideration in the reliability calculations. Therefore any identified failures that were removed from the test prior to completion and later found to be attributed to something other than solder joint fatigue should be identified as censored. Finally, testing was suspended after a predetermined amount of time whether or not failures were identified. Parts that did not fail were identified as censored. This type of data censoring is referred to as Type 1, or time censored data in which a test is concluded after a predetermined amount of time, in spite of the fact that failures may not have occurred C to 100 C Accelerated Thermal Cycling A total of 6010 cycles of ATC from C were completed. The test was periodically stopped and the failures, which have been recorded using a data logger, were removed and verified by manual resistance measurement. If the failure was determined to be within the test vehicle, as opposed to a cabling issue, the component was cut out of the board for future failure analysis. The remaining components on the test vehicle where returned to the chamber and the testing continued. Because no failures had been identified at the halfway point of the test, four boards, one for each alloy, were removed for metallurgical analysis. This occurred after 3157 cycles had been completed. These components were identified as censored. Table 29 summarizes the number of cycles to first failure of the QFP176s monitored during the 6010 cycles of the test. There are not enough failures at the end of 6010 cycles to plot failure distribution charts for each alloy. Only SAC305 experienced failures when built on High T g (170 C) board material and those occurred towards the end of the test. Figure 57 shows the distribution of failures of solder joints built on High T g boards 94

109 compared with those built on Normal T g boards. A two-parameter Weibull distribution is used. The shape parameter determines the shape of the probability distribution and can be viewed by plotting the probability density function (PDF). At values between 3 and 4, the shape parameter approximates a normal distribution. Shape parameter values higher than 4, as shown in Figure 58 results in a near-normal distribution with a left tail. The scale parameter (or slope on the Weibull distribution) described the range of the distribution, or the rate of at which the failure rate increases. It can be seen that the High T g boards are expected to survive much longer than those built with Normal T g when using SAC305 as the solder paste. Table 29: Summary of QFP failures after 0 C to 100 C ATC Component type QFP176 Failure Cycle SAC305 Senju M42 High T g Sunrise Sunflower SAC305 Normal T g Senju M42 Sunrise Sunflower 2/6 0/10 0/7 0/8 6/6 0/10 1/12 0/ % % %

110 Figure 57: Weibull plots of SAC305 QFP solder joints after 6010 Cycles 0 to 100 C comparing High T g to Normal T g boards Figure 58: Probability Density Function for SAC305 QFP solder joints after 6010 Cycles 0 to 100 C comparing High Tg to Normal Tg boards Figure 59 shows the failure occurred as a result of fracture initiating in the bulk solder and propagating along the lead side IMC layer. This is further illustrated in Figure 60 which shows the fracture surface of a similar solder joint made up of SAC305 on High T g 96

111 board material. The area highlighted in red shows the crack initiation. The area highlighted in green illustrates the point at which the fracture moves along the IMC layer. Within the green area, EDX scans were taken of area 1, which was shown to be pure Sn indicating bulk solder and area 2: which was shown to be Cu 6 Sn 5 indicating the IMC layer. Finally, the area highlighted in blue shows a ductile fracture, which likely occurred as a result of overload on the remaining area. Figure 59: QFP fracture of SAC305 on High T g board a) optically and b) cross section 97

112 Figure 60: Fracture surface of QFP solder joint with SAC305 on High T g board after 6010 cycles Figure 61 shows that fractures had initiated in solder joints formed with the other three alloys on high T g board material, but by 6010 cycles had not yet propagated far enough into the solder joint to cause an electrical failure. It is expected that these solder joints would ultimately fail in the same manner as shown in Figure 59 and Figure 60 if the testing had continued. 98

113 Figure 61: Fracture initiation in QFP a) Sunflower b) Senju M42 after 6010 cycles As described in 6.1.2, the BGA solder joints were made up of paste from the experimental alloy and SAC305 solder balls. This resulted in a solder joint composition closer to SAC305 and with no more than 0.9wt% Bi. No failures occurred within the 6010 cycles of the test when high T g board material was used. A number of failures occurred for each of the paste alloys when normal T g board material was used as outlined in Table 30. Using resistance probing, these were all found to be the result of failure within the component-solder-board system; either in the board circuit pattern beneath the component, the solder connecting the component to the board or within the component itself. It is not possible to determine where within this system the failure occurred without destructive techniques. Each component was therefore cut from the board and cross-sectioned to find the point of failure. Table 30: Summary of BGA failures after 0 C to 100 C ATC Component type Normal T g SAC305 Senju M42 Sunrise Sunflower Fraction of components failed 5/6 9/10 8/12 6/10 % of components failed 83.3% 90.0% 66.7% 60.0% Cycles to Failure

114 Figure 62 illustrates that there is no significant difference in the reliability performance of the solder joints formed using paste of the four alloys on Normal T g boards. All BGA failures were found to result from board failures, specifically via cracks within the laminate material. The probability distributions for each alloy in this test exhibit similar shape and scale parameters, which is to be expected since all failures are characteristics of the board material and construction, rather than the solder alloy. Figure 62: Weibull plots of BGA solder joints on Normal T g boards after 6010 Cycles 0 to 100 C cycles comparing Four Alloys Figure 63 illustrates a) a typical via barrel failure and b) a bulk solder fracture although incomplete failure near the component side IMC. 100

115 Figure 63: BGA failures after 0 to 100 C ATC a) failure in board material by via plating crack and b) partial failure through bulk solder (SAC305) near component side IMC C to 125 C Accelerated Thermal Cycling A total of 1000 cycles of ATC from -55 C to 125 C, or Harsh Environment ATC, were completed. The test was periodically stopped and the failures, recorded using a data logger, were removed and verified by manual resistance measurement. If, during the course of testing, the failure was determined to be within the test vehicle, as opposed to a cabling issue, the component was cut out of the board for future failure analysis. The remaining components on the test vehicle were returned to the chamber and the testing continued. Four boards, one for each alloy, were removed for metallurgical analysis near the halfway point of the test, after 438 cycles completed. Table 31 summarizes the number of cycles to first failure of the QFP176s monitored during the 1000 cycles of the test. 101

116 Table 31: Summary of QFP failures after -55 C to 125 C ATC Component type Fraction of components failed % of components failed Cycles to Failure SAC305 Senju M42 High T g Sunrise Sunflower SAC305 Normal T g Senju M42 Sunrise Sunflower 3/4 6/10 4/9 4/10 4/5 6/8 10/12 5/ % 60.0% 44.4% 40.0% 80.0% 75.0 % 83.3% 50.0% Figure 64 and Figure 65 show probability distribution plots and the probability density functions of the four alloys on High T g board materials respectively. Sunrise showed some early fails, resulting in a different shape of the PDF. Figure 66 shows probability distribution plots of the four alloys on Normal T g board materials respectively. In all cases, the distribution of failures occurred within overlapping confidence intervals; there is insufficient data to distinguish between the alloys after 1000 cycles of testing on Normal T g boards. 102

117 Figure 64: Weibull plots of QFP failures on High T g boards after 1000 cycles C Figure 65: Probability Density Function for QFP failures on High T g boards after 1000 cycles C 103

118 Figure 66: Weibull plots of QFP failures on Normal T g boards after 1000 Cycles C Just as in the 0 to 100 C cycling test, all QFPs experienced solder fracture initiating in the bulk solder and propagating along the lead side IMC (Figure 67). These however did not result in complete fracture, and would not account for the electrical failure. Further testing, by probing various test points in the test vehicle, determined that the likely cause of electrical failure was the copper trace separating from the board material. Copper trace fractures result in electrical open circuits. 104

119 Figure 67: QFP176 fractures in a) Sunflower and b) Senju M42 after 1000 cycles Harsh testing The BGA components that underwent harsh thermal cycling experienced many failures (Table 32). Upon cross sectional failure analysis however, these failures were found to be the result of board damage, as in the C test, specifically via cracks within the laminate material (Figure 68). While there are enough data points to plot probability distributions, this did not provide a means of distinguishing between alloys. Table 32: Summary of BGA failures after -55 C to 125 C ATC Component type Fraction of components failed % of components failed Cycles to Failure SAC305 Senju M42 High T g Sunrise Sunflower SAC305 Normal T g Senju M42 Sunrise Sunflower 5/6 10/10 6/6 9/10 4/6 10/10 12/12 10/ % 100.0% 100.0% 90.0% 66.7% 100.0% 100.0% 100.0%

120 Figure 68: BGA failures after -55 to 125 C ATC failures in board material by via plating crack in a) boards built with Sunrise and b) boards built with Sunflower Further examination of the solder joints revealed some crack initiation in the solder joints towards the component side of the BGA solder joint. These crack initiation sites (Figure 69) were less significant than the damage observed after 6010 cycles of C thermal cycling and would not account for a failure reading. Electrical failures, just as in the case of C thermal cycling, occurred as a result of via cracks in the board material (Figure 68). A probability plot comparing all failures on the two different board materials (Figure 70) illustrates that the board material was a more significant distinguishing factor than the various solder alloys. Figure 69: BGA solder crack initiation after -55 to 125 C ATC in solder joints formed with SAC305 solder balls and a) Sunflower and b) Sunrise 106

121 Figure 70: Weibull plots of BGA solder joints after 1000 Cycles -55 to 125C cycles comparing two board materials QFP solder joint failures on Normal T g board material was used to compare the two test conditions, C for 6010 cycles and -55 to 125 C for 1000 cycles. The Harsh environment test (-55 to 125 C) produced many failures over the 1000 cycles, including a few early failures. This resulted in a Weibull distribution with a shape parameter of 1.26, a PDF with a right tail. The C was less severe and therefore resulted in fewer failures and no early failures. The shape parameter of 5.59 indicates that the failure rate increases over time. 107

122 Figure 71: Weibull plot for QFP solder joints on Normal Tg boards comparing two test conditions Figure 72: Probability Density Function for QFP failures on Normal Tg boards comparing two test conditions 108

123 4.2 Microstructure Evaluation Bulk Microstructure SAC305, before thermal cycling (Figure 73a and 73b) appears as small dendrite arms surrounded by small Ag 3 Sn and Cu 6 Sn 5 particles. After thermal cycling from 0 to 100 C for 3148 and 6010 cycles (Figure 73c and 73d respectively) the intermetallic particles coalesced into fewer, larger particles. Senju M42 behaved in the same manner. In both cases, the bulk of the microstructural transition occurred between 0 and 3148 cycles, with little change between 3148 and 6010 cycles. With 3% Bi in Senju M42, there does not appear to be any significant Bi precipitation from the solid solution (Figure 74). Figure 73: SAC305 at Time 0 a) optically and b) SEM image and after b) 3148 cycles and d) 6010 cycles of C ATC 109

124 Figure 74: Senju M42 at 1000x after a) 3148 cycles and b) 6010 cycles Sunrise and Sunflower solder paste alloys, which both contain 7wt% Bi, showed a significant amount of Bi precipitation after ATC. In both cases large, uneven particles of Bi were present upon solidification, as seen in Chapter 2. After ATC, Sunrise showed many, very small particles precipitated throughout the bulk, evenly dispersed from the Sn grains (Figure 75). Sunflower also exhibited some Bi accumulation along the grain boundaries (Figure 76). In both cases, there did not appear to be any significant accumulation of Bi along the interface with the interfacial IMC layer. Figure 75: Sunrise after 6010 cycles C at a) 500x and b) 1000x 110

125 Figure 76: Sunflower at a) Time 0 and b) after 6010 cycles C at 500x IMC Growth during Thermal Cycling The interfacial IMC layers were measured as described in section 6.2. A comparison of the board side IMC layer of a QFP176 component follows. This IMC layer formed during the reflow process in which a bond was formed between the board side copper pad and the molten solder. The Time 0 results reference this IMC layer after reflow but before any thermal cycling. As described in section 6.2, only Cu 6 Sn 5 was identified at Time 0. Measurements were made after approximately the halfway point and again at the completion of the two thermal cycling regimes. For the C test, measurements were completed after 3157 cycles and then again after 6010 cycles. For the harsh environment test (-55 C to 125 C) measurements were taken at 438 cycles and then again at 1000 cycles. After exposure to thermal cycling, both Cu 6 Sn 5 and Cu 3 Sn species were identified (Figure 77). 111

126 Figure 77: IMC layers formed on QFP176 solder joints between the board side Cu layer and solder paste a) sunrise and b) SAC305 after 438 cycles of Harsh thermal cycling. Location 1 shows the Cu 3 Sn layer, location 2 shows the Cu 6 Sn 5 layer Figure 78 provides a comparison of the overall IMC thickness at the various test intervals; Figure 79 provides a comparison of the Cu 3 Sn layer within the IMC at various test intervals. Table 33 and Table 34 provide a summary of results from two-sided T-test to compare the mean values and Levine-test to compare the variance of the IMC and Cu 3 Sn thicknesses respectively. These tables are provided in order to examine the changes in thickness of the overall IMC and the Cu 3 Sn layers, for each alloy, over the course of the two thermal cycling conditions. The tables provide probability (p-values) for each test. In all cases, the null and alternate hypotheses are as follows: H 0 : µ 1 =µ 2 and 1 = 2 H a : µ 1 µ 2 and 1 2 Values in Table 33 and Table 34 which are bold italicized represent points with a p-value less than This indicates that H 0 should be rejected at a 95% confidence level and the H a is assumed to be valid. In this case, it assumes that the mean value has changed. These results show that between 0 and 1000 cycles of Harsh testing, the IMC layers for each alloy increased significantly. For Senju M42 and Sunrise, this occurred primarily within the first 438 cycles. For SAC305 and Sunflower the increase continued throughout 112

127 the test. In C ATC testing, the IMC layer for all alloys except SAC305 increased throughout the test. The IMC thickness for SAC305 increased most significantly during the first half of the test. All t-tests of the µ in Table 34 result in rejecting the H 0. This indicates that the thickness of the Cu 3 Sn layer continues to increase throughout the course of the two tests. Table 35 provides a summary of test for equal variance and ANOVA testing of the various alloys. This test compares the IMC and Cu 3 Sn thicknesses of all four alloys, at a given point during ATC. This table provides probability for each test. In all cases, the null and alternate hypotheses are as follows: H 0 : µ SAC305 =µ Senju M42 =µ Sunrise =µ Sunflower and SAC305 = Senju M42 = Sunrise = Sunflower H a : at least one µ is different and at least one is different The result of ANOVA testing, indicated in Table 35: Results of ANOVA test to equal variance and compare means, show that while the total IMC thickness increased relatively consistently among the four alloys, the increase in the thickness of the Cu 3 Sn layer was different in at least one alloy. Figure 79 clearly shows that Sunflower exhibited a significantly thicker Cu 3 Sn layer after Time 0 in both test conditions. 113

128 Figure 78: IMC growth at the board side of QFP during thermal cycling Figure 79: Cu 3 Sn (portion of IMC) growth at board side of QFP during thermal cycling 114

129 Table 33: Results of Levine-Test to compare the variance ( ) of IMC measurement and 2-sided t-test to compare the means (µ) of IMC measurements Test HARSH Condition (-55 to 125 C) (0 to 100 C) Comparison 0 to to to to to to cycles cycles cycles cycles cycles 6010 cycles Variable / p value µ µ µ µ µ µ SAC Senju M Sunrise Sunflower Table 34: Results of Levine-Test to compare the variance ( ) of Cu 3 Sn measurement and 2-sided t-test to compare the means (µ) of Cu 3 Sn measurements Test HARSH Condition (-55 to 125 C) (0 to 100 C) Comparison 438 to 1000 cycles 3157 to 6010 cycles variable µ µ SAC Senju M Sunrise Sunflower

130 Test Condition Table 35: Results of ANOVA test to equal variance and compare means Variable / p value Equal variance ( ) Total IMC ANOVA (µ) Equal variance ( ) Cu 3 Sn ANOVA (µ) Time cycles HARSH (-55 to 125 C) 1000 cycles HARSH (-55 to 125 C) 3157 cycles (0 to 100 C) 6010 cycles (0 to 100 C) Figure 80 through Figure 87 provide details of the ANOVA analysis. The interval plot Figure 80 and Figure 81 show the 95% confidence intervals for the mean overall IMC layer thickness and the mean Cu 3 Sn layer thickness respectively. While it appears that SAC305 has a thicker overall IMC layer at Time 0, the interval plot shows substantial overlap between the 95% CIs of all four alloys and therefore indicates that the difference in thickness is not statistically significant. Further analysis indicates that there is a difference in Time 0 thicknesses within a 93% confidence level. The thicker IMC layer of SAC305 is a result of the higher processing temperature (240 C). The intervals for the Cu 3 Sn layer thickness of Sunflower do not overlap with any of the 95% CI of the other alloys, thereby indicating a significant difference in thickness. 116

131 Figure 80: Interval plot of IMC thickness at the board side of QFP after ATC Figure 81: Interval plot of Cu 3 Sn thickness at board side of QFP after ATC 117

132 The main effects plots in Figure 82 to Figure 85 are used to de-couple the main factors in the measurements to better understand their individual contributions. Figure 82 and Figure 83 show the condition, or time at exposure to thermal cycling, has a greater influence on the overall IMC thickness than do the particular alloys. SAC305 has a Time 0 IMC thickness greater than the mean of the other three alloys, which all have mean values closer to the grand mean. Figure 84 and Figure 85 show that the condition is also a greater influence on the mean overall thickness of the mean Cu 3 Sn layer than the alloys. The difference between Cu 3 Sn thicknesses amongst the alloys shows the same pattern: SAC305 and Senju M42 have mean thicknesses close to the grand mean, Sunrise has a mean lower than the grand mean and Sunflower is significantly greater than the grand mean. Figure 82: Main effects plot of IMC thickness at the board side QFP during HARSH (-55 to 125 C) ATC 118

133 Figure 83: Main effects plot of IMC thickness at the board side QFP during C Figure 84: Main effects plot of Cu3Sn thickness at board side of QFP during HARSH (-55 to 125 C) ATC 119

134 Figure 85: Main effects plot of Cu3Sn thickness board side of QFP during C Interaction plots found in Figure 86 and Figure 87 tests the factors to identify any interactions. Parallel, or near parallel lines, as seen in both cases indicate there are no clear interactions; the influence of one factor is not dependent on the other. The mean thickness of the IMC layer for SAC305 at Time 0 is greater than that of the other alloys, however over the course of ACT testing under two test conditions, this distinction seems to disappear, the mean IMC thicknesses are no longer distinguishable between alloys. 120

135 Figure 86: Interaction plot of IMC thickness at the board side of QFP after ATC Figure 87: Interaction Plot of Cu3Sn Thickness at Board Side of QFP after ATC 121

136 5 Summary of Findings and Conclusions 5.1 Findings Based on Reliability Data During two different thermal cycling regimes the only failures which could be attributed to solder joint fatigue failures occurred in SAC305 during 0 to 100 C cycling; one sample of Sunrise, which failed but only after 5416 cycles. All other electrical failures were attributed to failures within the board material. In the case of BGAs, this occurred due to cracks within via barrels; in QFPs it was the result of copper trace fractures. Thermal cycling did not significantly distinguish between the reliability performances of the four alloys. High T g board material appeared to outperform Normal T g board material. It can be concluded that the three lower process temperature solders performed as well as, or better than SAC305 during accelerate thermal cycling. All solders survived longer than the board materials. 5.2 Findings Based on Microstructural Observations Bi had a beneficial impact on the bulk microstructure. It precipitated from the bulk solder evenly in very small particles. This favorable influence of Bi only appeared when the concentration was 7wt%; the 3wt% Bi of Senju M42 did not appear to have the same impact. Senju M42 appeared to have the same aging characteristics as SAC305. There was some degree of Bi segregation along the grain boundary observed in the Sunflower alloy; however there did not appear to be full segregation of the phases. The interfacial IMC layer ( ) Cu 6 Sn 5, formed during solidification, was similar in thickness for all alloys. During thermal cycling, the total thickness of the interfacial IMC continued to increase, as well as ( ) Cu 3 Sn layer formation. While the overall thickness increased similarly for all four alloys, the Cu 3 Sn layer of the Sunflower alloy was much thicker than the other alloys. This alloy has no Ag, which indicates that Ag plays a role in suppressing Cu 3 Sn growth. This is in spite of the fact that Ag does not participate in the IMC formation; no Ag-containing species exists at the interfacial IMC. It has previously been shown that eutectic SnAg solder exhibited a lower layer-growth 122

137 coefficient for Cu 6 Sn 5 than for Cu 3 Sn. 24 This was believed to be the result of the Sn diffusion in SnAg being slower than the Sn diffusion in, for example eutectic SnPb solder, in turn favoring the growth of Cu 3 Sn. It appears, through this study, that even small amounts of Ag in the solder (1%, 2% and 3%) are sufficient to suppress the diffusion of Sn to the growing interfacial IMC layers. Bi alloying did not significantly change the growth rates of either intermetallic compound. The increased growth of the Cu 3 Sn phase in the Sunflower alloy did not correspond to any decrease in reliability during ATC. SAC305, which was processed at a higher temperature, initially had a larger interfacial IMC layer. This effect however did not continue during thermal cycling where the IMC layers of all four alloys formed and increased to similar overall thicknesses. 123

138 6 References 1 J. Bentley Introduction to Reliability and Quality Engineering, 2 nd Ed, Essex, England, Pearson Education Limited, 1999, ch. 2, pp IPC-SM-785 Guidelines for Accelerated Reliability Testing of Surface Mount Solder Attachments, November W. Engelmaier Solder Attachment Reliability, Accelerated Testing, and Results Evaluation in Solder Joint Reliability J.H. Lau, Ed. New York, Springer 1991, ch.17. pp Blueprints for Product Reliability, The Reliability Information Analysis Center (RIAC), &Deskref=blueprint1#3point4 5 J.W.Evans Thermomechanical Fatigue in A Guide to Lead-free Solder: Physical Metallurgy and Reliability, W. Engelmaier, Ed. London, UK, Springer 2007, ch.7, pp IPC-9701A: Performance Test Methods and Qualification Requirements for Surface Mount Solder Attachments, February A. MacDiarmid Thermal Cycling Failure Part One of Two in The Journal of the Reliability Information Analysis Center, January J-P.M. Clech and J.A. Augis Surface Mount Attachment Reliability and Figures of Merit for Design for Reliability in Solder Joint Reliability J.H. Lau, Ed. New York, Springer 1991, ch.18. pp J.W.Evans Introduction to Solder Alloys and Their Properties in A Guide to Lead-free Solder: Physical Metallurgy and Reliability, Ed. London, UK, Springer 2007, ch.1, pp A. Zbrzezny Characterization and Modeling of Microstructural Evolution of Near- Eutectic Sn-Ag-Cu Solder Joints Ph.D. Thesis, Dept. MSE, Univ. Toronto, Toronto, Canada, R. Coyle, R. Parker, M. Osterman, S. Longgood, K. Sweatman, E. Benedetto, A. Allen, E. George, J. Smetana, K. Howell, J. Arnold inemi Pb-Free Alloy Characterization Project Report: Part V The Effect of Dwell Time on Thermal Fatigue Reliability presented at SMTAI, Chicago, Il, J. Juarez Jr., M. Robinson, J. Heebink, P. Snugovsky, E. Kosiba, J. Kennedy, Z. Bagheri, S. Suthakaran, M. Romansky Reliability Screening of Lower Melting Point Pb-Free Alloys Containing Bi, in IPC APEX EXPO Conference, Las Vegas, NV,

139 13 B. Arfaei, M. Anselm, S. Joshi, S. Mahin-Shirazi, P. Borgesen, E. Cotts, J. Wilcox, and R. Coyle Effect of Sn Grain Morphology on Failure Mechanism and Reliability of Lead-Free Solder Joints in Thermal Cycling Tests, presented at SMTAI, Chicago, Il, K-N. Tu Copper Tin Reactions in Thin-Film Samples in Solder Joint Technology : Materials, Properties and Reliability New York, Springer, 2007, ch. 3, pp N. Mookam and K. Kanlayasiri Evolution of Intermetallic Compounds between Sn- 0.3Ag-0.7Cu Low-Ag Lead-free Solder and Cu Substrate during Thermal Aging J. Mater. Sci. Technol., 2012, 28(1), C. Yu, Y.Yang, P. Li, J. Chen, H. Lu Suppression of Cu 3 Sn and Kirkendall voids at Cu/Sn-3.5Ag solder joints by adding a small amount of Ge, J.Mater Sci: Mater Electron (2012) 23: C. Yu Chen, Kai-Yun Wang, Jing-Qing Chen, Hao Lu Suppression effect of Cu and Ag on Cu 3 Sn layer in solder joints, J Mater Sci: Mater Electron (2013) 24: G.C. Moon, S.K. Kang, D-Y Shin and H.M. Lee Effects of Minor Additions of Zn on Interfacial Reactions of Sn-Ag-Cu and Sn-Cu Solders with Various Cu Substrates during Thermal Aging Journal of Electronic Materials, Vol. 36, no. 11, P. Snugovsky, E. Kosiba, J. Kennedy, Z. Bagheri, M. Romansky, M. Robinson, J.M. Juarez, Jr., J.Heebink Manufacturability and Reliability Screening of Lower Melting Point Pb-free Alloys Containing Bi, in IPC APEX EXPO Conference, San Deigo, CA, P. Vianco, and J.A. Rejent Properties of Ternary Sn-Ag-Bi Solder Alloys: Part 1 Thermal Properties and Microstructural Analysis, Journal of Electronic Materials, Vol.28, No.10, pp , D.Witkin Creep Behavior of Bi-Containing Lead-Free Solder Alloys Journal of Electronic Materials, Vol. 41, No. 2, P. Vianco, and J.A. Rejent Properties of Ternary Sn-Ag-Bi Solder Alloys: Part 1 Thermal Properties and Microstructural Analysis, Journal of Electronic Materials, Vol.28, No.10, pp , CRE Primer, Chapter IX: Data Collection, p. IX-5, Quality Council of Indiana D.Kwon Intermetallic Formation and Growth in A Guide to Lead-free Solders: Physical Metallurgy and Reliability, Silver Spring, MD: Springer, 2005, pp

140 Chapter 4 Tin Whisker Testing 1 Introduction Whiskers, spontaneous columnar or cylindrical filaments which emanate from a surface, are found to form from various metals including Ag, Zn, Cd and Sn. 1,2 The danger posed within an electronic system is that whiskers are both conductive and grow spontaneously. It is therefore challenging to predict where they could form, and if they would grow long enough to cause potential electrical bridging issues. Electrical bridges can form by whiskers growing long enough to cross the minimum gap, or by whiskers broken off from the surface and falling freely around the electronic assembly. Furthermore, there is a potential for two whiskers to grow sufficiently close together to allow for electrical arcing between them. As the demand for increasingly small electrical components continues, and the minimum lead-to-lead gap shrinks, the risk whiskers pose increases with miniaturization. Sn whiskers tend to have the following characteristics: growth over time, thicknesses varying from sub-micron to just above a few microns, tough and electrically conductive. 3 Although Sn whiskers were first observed shortly after WWII in telephone transmission line channel filters 1, they were not extensively studied since Pb in Sn (at least 1wt%) was found to be an effective form of mitigation. Although whiskers can grow from SnPb solder, they were observed as short and often topped with a Pb-rich cap (Figure 88), which appeared to limit their growth. 126

141 Figure 88: Pb "cap" on whisker from SnPb component finish 2 Initial investigations of whisker formation focused on Sn plating of electronic components. As whiskers are a surface, rather than a bulk, phenomenon, a thin coating of matte Sn found on most Pb-free components, was initially thought to be the main area of concern. Correspondingly, most standards developed for testing, measuring and qualifying new materials focused on component surface finishes. 4,5 During assembly using a SnPb solder, the solder is expected to wet a portion of the lead and form a fillet (represented by E in Figure 89a) while the rest of the lead maintains its original plating material. Due to the higher reflow temperatures used to assemble SAC solders, the electroplated Sn on the lead frame melts. The solder is therefore likely to completely wet and cover the lead (Figure 89b). 6 With SnPb solder assemblies, using matte Sn plated components; the risk of whisker formation exists on the exposed Sn area of the lead. Originally it was thought that whiskers grow primarily from thin film coatings, however it has been shown they also grow from bulk solder; whiskers have been observed growing from bulk SAC305 and SAC This indicates the risk of whisker growth now spans the entire solder joint surface on a Pb-free assembly. It is believed however, that areas where a thin layer of solder exists are of greatest concern. 127

142 Figure 89: a) Schematic for a typical solder joint of a leaded component using SnPb solder 8, b) cross section showing solder joint formed with Pb-free solder 1.1 Whisker Growth Kinetics Currently there is no widely accepted explanation for the mechanism causing Sn whiskers; however it is widely agreed that the driving force for whisker nucleation and growth involves local compressive stresses. It is accepted that Sn whiskers are one of the forms of stress relaxation within the localized region. 6 It is also understood that whiskers are a surface stress relief mechanism; when relaxation occurs within the bulk microstructure, whiskers will not grow. 9 Therefore areas with thin solder coating, or thin plating material appear to be more susceptible to whisker formation. For example, on an assembly of Sn plated components built with SAC305 solder, whiskers grew from areas where the solder was less than 25µm thick. 10 Complicating matters, it has been shown that the total length of the whisker does not correlate to the applied stress, thereby indicating the mechanism is not one of bulk diffusion. Additionally, whiskers appear to originate from newly formed grains at the surface. For these reasons, one of the currently accepted models, proposed by Vianco, describes whisker growth as a process of cyclic dynamic recrystallization (DRX). 2,11 This model suggests that whiskers grow under specific conditions requiring both a cyclic DRX, in order to nucleate a new grain at the surface of a thin film, and a mass transport 128

143 mechanism to facilitate the supply of Sn. In both case the driving force is applied compressive stress. This model also proposes that stress relaxation will occur by means of recrystallization rather than recovery when the homologous temperature (T h ) is approximately 0.6. Sn, even at room temperature, has a homologous temperature of The formation of whisker growth is further dependent on strain rate where a slower strain rate will favor whisker formation over a high strain rate. Finally if the applied stress is too high, even at a low strain rate, whiskers will not likely form. As compressive stress is applied to a surface, strain within the localized system accumulates, usually in the form of dislocations at pre-existing grain boundaries. Once this strain energy pile-up exceeds a certain limit, a new grain will nucleate and grow in a process known as DRX. This process differs from a dynamic recovery process in which the dislocations will either annihilate each other or be absorbed within the grain boundaries instead of forming new grains. The growth of grains in the form of whiskers represents a surface phenomenon rather than grain growth within the bulk of a material. Strain rate, temperature and grain size all factor into the cyclic DRX model and subsequently whisker growth. 2 Compressive Stress Creep Deformation Work Hardening Dynamic Stress Relaxation Recrystallization Recovery T h = ~0.6 (DRX) no whiskers T h = Slow strain rate Cyclic DRX Continuous DRX no whiskers Fast strain rate X < σ < 5 MPa σ > 5 MPa no whiskers Figure 90: Cyclic Dynamic Recrystallization resulting in whisker formation

144 1.2 Sources of Compressive Stress Sn whisker growth is initiated by self generating compressive stress within the material and can be mechanical, thermal or chemical in nature. 9 An IMC layer growing between the lead frame material and the plating material, or solder, is usually irregular, or scallop shaped (Chapter 3). The growth of this layer and any IMC particles, which subsequently break off into the bulk material, introduce compressive stresses into the bulk material. In the case of Sn plated leads, the Sn layer is often found to have a columnar shape, with IMC particles accumulating along the grain boundaries. Figure 91: Source of compressive stress contributing to whisker growth 13 When exposed to various temperature cycling environments, stress induced into the Sn plating or solder is attributed to the mismatch in the Coefficient of Thermal Expansion (CTE) between the Sn and the lead frame material. Localized compressive stresses may occur and increase the whisker propensity where thin layers of solder exist. Also, the uneven interfacial IMCs, which continue to grow during the heating portion of the thermal cycle, may introduce compressive stresses into the solder. There may not be sufficient time to relieve stress before the cold portion of the cycle occurs. Previous work found that CTE mismatch and corresponding whisker growth was greater for Alloy 42 lead frame materials than for Cu. 10,13 (Table 36). 130

145 Table 36: CTE values for common materials in solder joints Material Composition Coefficient of Thermal Expansion (CTE) ppm/ C Lead Finish Sn Sn 22.0 Solder Paste SnPb Sn - 37Pb 21.6 Materials SAC305 Sn - 3Ag0.5Cu 21.6 Alloy 42 Fe - 42Ni 5.8 Cu Fe Cu194 Lead Frame 0.15P Zn 17.5 Materials Cu151 Cu - 0.1Zr 17.7 Cu7025 Cu Ni Si Mg 17.3 The formation of oxides on the surface of a solder joint or plated lead has also been associated with whisker growth. 9,14 In this case, the compressive stresses are related to the volume expansion caused by the formation of oxides, which can lead to a volume expansion of approximately 29 to 34% as observed in many studies testing samples in humid conditions. 15 Oxidation can also be associated with the presence of IMC particles. For example, when exposed to humidity and chloride contamination, Ag 3 Sn IMC particles found at the surface of a SAC305 solder joint are found to exhibit corrosion within the interdendritic spaces (Figure 92). This is attributed to the differing chemical potentials of the IMC and the surrounding bulk solder, which facilitate galvanic corrosion. 6,16 Sn-based solders are particularly susceptible due to the roughness of the surface; shrinkage voids resulting from the dendritic solidification of these solders leaves a very rough surface which can easily entrap contaminates during the manufacturing process or during field use. Cu-based lead frame materials have been shown to be more susceptible to these types of compressive stresses than Alloy

146 Figure 92: Ag3Sn oxide zone with whisker 6 Mechanically-induced residual stresses also result in whisker formation. For example, the compressive stresses generated in the Sn plating of a leaded component have been attributed to the bending or stretching processes used to form leads into their final shape after plating. 17 Studies show that whiskers growing as a result of these mechanically induced stresses tend to have a higher growth rate than those resulting from room temperature IMC internal stresses Morphology of Sn Whiskers As stated in 1.1, a correlation between applied stress and resulting whisker length has not been found. Whisker growth is, to date, unpredictable as is the final morphology of the whisker. Long, thin, straight whiskers are often found alongside short, kinked whiskers of various thicknesses (Figure 93). The long, fine whiskers have been found to contain only Sn, usually of a single crystal structure. Shorter, thicker whiskers, hillocks and shell shaped protrusions are made up of mainly Sn but may contain Ag 3 Sn and Cu 6 Sn 5 IMC particles. 7 It is the long, thin, straight whiskers which are of most concern as they have 132

147 the highest potential of bridging a gap between electrical contacts. However, it is not currently possible to predict the form or total length of whiskers. Figure 93: Whisker morphology a) long, thin whiskers and b) short, kinked whiskers The Effects of Bi in Solder on Whisker Formation SnPb solder and plating finish has long been used in electronics without major reliability concerns related to Sn whiskers. Eutectic SnPb plating over a Cu surface forms an equiaxed grain structure rather than the columnar grain structure typical of Sn plating (Figure 94). This equiaxed structure is believed to allow stress relaxation to occur by incorporating displaced Sn grains more readily at grain boundaries, which lie parallel to the surface. This would allow for the material to relieve the compressive stresses in a uniform creep. 19,20 Jadhav et al. 19 showed that both the grain structure and the concentration of Bi had an affect on the overall stress relaxation of a thin film over Cu. Figure 94: Cross sections of plating surface made with SEM FIB of a) Sn and b) SnPb

148 Small amounts of Pb, as small as 1wt%, result in grain refinement and subsequently the suppression of whisker growth. It was found that hillocks grow instead and that these hillocks are of limited length. 18 As Pb is no longer available as an alloying agent in solder, other alloying elements are being explored for the same grain refinement, creep resistance and ultimately whisker suppression characteristics. In this work, the focus is on Bi alloying. Section 1.2 identified the growth of an uneven, scalloped interfacial IMC layer as one of the sources of compressive stress within a solder joint, particularly in areas with thin solder coverage. Jo found that the IMC layer, which continued to grow during room temperature storage and high temperature high humidity conditions, became more uniform (i.e. less scalloped) with increasing concentrations of Bi. 18 Annealing has also been proposed as a means of creating a more uniform IMC layer. If the IMC layer were to become more uniform in shape, it would introduce less compressive stress into the system, thereby eliminating one possible source of stress. The surface area of the IMC would also be reduced, which in turn would reduce the diffusion rate of Cu into the bulk solder. The Cu 6 Sn 5 IMC particles within the bulk solder are another source of internal compressive stress on the localized system. Additionally, Jo found that as little as 0.5wt% Bi reduced whisker propensity. Other researchers have indicated that as much as 3-5wt% Bi is required for whisker suppression. 1, 21 The three main mechanisms by which Bi is thought to mitigate against whisker growth are: Refining the grain size Altering the grain structure from a typical Sn columnar structure to an equiaxed structure more similar to that found in SnPb Reducing growth and irregularity of interfacial IMC

149 2 Experimental Set Up The work describes in this section formed a screening experiment in which the selected alloys were screened for whisker mitigation properties. The results were intended to determine which, if any, of these alloys would be good candidates for further testing. Table 37 provides a list of test conditions recommended by JEDEC for qualifying plating finishes on components. It should be noted that there is no current test or qualification guideline for assessing the propensity of solder alloys in an assembly to grow whiskers. Further, it has been found that suspension of some tests, particularly those involving humidity storage, will impact the total whisker growth and lead to under-reporting of whisker length. When tests are restarted, new whiskers nucleate and grow rather than existing whiskers continuing to grow. Therefore, in order to understand the time dependence of whisker length, each successive inspection interval needs to be made after an increased amount of uninterrupted exposure. 6 The guidelines set out in Table 37 were used as the basis for developing this works test conditions, however the conditions were also augmented by further study. In high temperature/humidity storage conditions for example, it was found that whisker nucleation occurred faster at 85 C than at C. 22 Table 37: JESD22A test conditions 4 Stress Type Temperature Cycling Ambient Temperature/Humidity Storage High Temperature/Humidity Storage Ref. Spec. JESD22 -A104 Test Conditions Min Temperature -55 to -40 (+0/-10) C Max Temperature +85 (+10/-10) C air to air: 5 to 10 minute soak; ~3 cycles/hour Recommendations Inspection Minimum Interval Duration 500 cycles 1000 cycles 30 2 C and 60 3%RH 1000 hours 3000 hours 60 2 C and 87 +3/-2%RH 1000 hours 3000 hours 135

150 2.1 Materials Two alloys with low Ag (Senju M42 and Sunrise), and one alloy with no Ag (Sunflower), and varying amounts of Bi, were tested against SAC305 (Table 38). The alloy selection method was described in Chapter Chapter 1. QFP components (U1) were cut from the test vehicle described in 5.1. Table 38: Alloys screened for whisker growth Alloy Composition Assembly Temperature SAC305 Sn 3%Ag 0.5%Cu 240 C Senju M42 Sn 2%Ag 0.75%Cu 3%Bi 224 C Sunrise Sn 1%Ag 0.7%Cu 7%Bi 222 C Sunflower Sn 0.7%Cu 7%Bi 226 C 2.2 High Temperature High Humidity Two samples of each alloy, consisting of 176 leads each, were placed in a humidity chamber. Teflon cabling was used to suspend the samples (Figure 95) and de-ionized (DI) water was used to generate the humidity conditions so as not to introduce possible contaminates into the system. The samples were exposed to 85 C/85%RH for 1000 hours. Figure 95: Samples in HTHH chamber 136

151 2.3 Thermal Shock Two samples of each alloy, consisting of 176 leads each, were placed in a two stage, airto-air, thermal shock chamber (Figure 96). The samples were exposed to -55 C to 85 C thermal shocks using the profile shown in Figure 97 for a total of 1610 shock cycles. Figure 96: Two stage, air to air, chamber for thermal shock testing 137

152 Temperature in C :09:04 8:11:04 8:13:04 8:15:04 8:17:04 8:19:04 8:21:04 8:23:04 8:25:04 8:27:04 8:29:04 8:31:04 8:33:04 Cold Chamber Hot Chamber Sample Figure 97: Thermal shock temperature profile 2.4 Post Exposure Evaluation 8:35:04 8:37:04 8:39:04 8:41:04 8:43:04 8:45:04 8:47:04 8:49:04 8:51:04 8:53:04 8:55:04 8:57:04 8:59:04 9:01:04 9:03:04 9:05:04 9:07:04 9:09:04 9:11:05 9:13:05 After each of the two exposures, components were inspected for whiskers using a variable pressure SEM (Hitachi S-3000N) at 15 kv acceleration voltage (V acc ), and 25MPa vacuum pressure. All leads on the components where examined at low magnifications, between 100x and 250x. Areas which showed signs of irregularity, nucleation and/or whiskers growth, where then further examined at higher magnifications of 1000 to 5000x. The samples exposed to thermal shock, which subsequently showed whisker growth, were then further evaluated by cross section. Each sample was first mounted in epoxy, and then ground and polished through the following sequence: 500 and 1200 grade SiC paper, polishing with 6 µm and 1 µm DiaPro diamond suspensions (Struers), and an oxide polish (Struers OP-S). Optical microscopy was performed using a Nikon Measurescope MM-11. Prior to SEM analysis, the samples were carbon coated using an 138

153 Emitech K950X. SEM microscopy on cross-sectioned samples was performed using a Hitachi S-4500 and Hitachi S-3000N with the following EDX systems: Oxford and ThermoScientific respectively. 3 Results 3.1 High Temperature High Humidity Results No whiskers were found on any of the four alloys tested after 1000 hours exposure to 85 C/85% RH. Further exposure, for at least an additional 3000 hours in the same conditions is recommended to distinguish between the whisker growth propensities amongst the test alloys. This would be closer to the conditions set out in JESD22A121.01, however due to resource constraints was not completed as part of this work. 3.2 Thermal Shock Results The results of whisker inspection are summarized in Table 39 including the location of whiskers, as described in Figure 98 and a summary of the Bi content found in the location of whisker growth. All alloys were found to form whiskers under the thermal shock conditions. All whiskers (or hillocks) where short, less than 10µm and would therefore not be considered to fail JEDEC Standard No. 201A for class 2 components, which requires that no whisker to exceed 45µm after 1000 cycles of thermal shock. 5 While this specification is intended for component finishes rather than soldered components, it is the only currently available guideline by which to make a comparison. 139

154 Table 39: Summary of whisker growth after 1610 cycles thermal shock Alloy Composition Whisker Location Morphology wt% Bi at Location SAC305 Sn 3%Ag 0.5%Cu Yes 1,4 few hillocks 0 Senju M42 Sn 2%Ag 0.75%Cu 3%Bi Yes 1,4 few, hillocks 1,4 many, very small, whisker nucleation sites 1.3, few, hillocks 3.7 Sunrise Sn 1%Ag 0.7%Cu 7%Bi Yes 1,4 many, very small whisker nucleation sites 3.3, 3.3 1,2,4 few, hillocks 2.9, 4.1, 3.3 Sunflower Sn 0.7%Cu 7%Bi Yes 4 some, very small whisker nucleation sites 3.3 Figure 98: Schematic showing locations on lead where whiskers formed 140

155 Along with the observed whisker growth, other competing stress relaxation mechanisms were found in all solder alloys. Figure 99 shows two adjacent leads, each with very different surface morphologies after 1610 thermal cycles. The lead on the left shows massive, bulk deformation and volume recrystallization. The lead on the right exhibits very little deformation and appears unaffected by the local stresses. Massive eruptions and protrusions of bulk IMC particles were also observed. Figure 99: Two adjacent leads with Sunflower solder paste after 1610 thermal shocks SAC305 solder joints exhibited a small amount of whisker growth after 1610 cycles (Figure 100), however there were no whiskers longer than 10µm. Short, thick whiskers which can be classified as hillocks, were found in locations 1 and 4 (Figure 98). Location 1 is of greater concern as it has a short bridging distance with the adjacent lead. Both locations are known to have thin solder coverage; a thin solder layer is more susceptible to whisker growth as compared to an area with a larger volume of bulk solder which may relieve internal stresses by competing mechanisms. Unlike the Bi containing alloys, SAC305 did not exhibit any area with many, small leads or fields of very short and thin whiskers. All three of the Bi containing alloys exhibited some degree of hillocks growing amongst fields of very short, very thin whisker nucleation sites. This was most often found in locations 1 and 4 (Figure 101 to Figure 103). Sunflower exhibited the least area of these 141

156 fields but had the two single longest whiskers seen in this study. Both however were still shorter than 10µm. Figure 100: Whisker growth on SAC305 after 1610 thermal shocks. a) and b) whisker growth in location 4 c) massive deformation and d) whisker growth in location 1 142

157 Figure 101: Whisker growth on Senju M42 after 1610 thermal shocks. a) and b) location 4 with short, thick whisker surrounded by many, very short whisker nucleation sites c) and d) short, thick whiskers growing at site of contamination in location 4 143

158 Figure 102: Whisker growth on Sunrise after 1610 thermal shocks. a) and b) location 2 with short, thick whisker c) location 4 and d) location 1 with many, very short whisker nucleation sites short 144

159 Figure 103: Whisker growth on Sunflower after 1610 thermal shocks. a) and b) location 1 with short, thick whisker c) some, very short whisker nucleation sites at location 4 and d) longest whisker observed at location 1 Figure 104 shows a whisker growing from Sunflower solder after 1610 cycles of thermal shock. The base of the new whisker grain is located at the intersection of three surface grains. This finding aligns with the dynamic recrystallization model. A cross section of a hillock formed from Sunrise (Figure 105), shows that the hillock forms from a new grain or grains. Figure 106 shows mass bulk recrystallization of the Senju M42 alloy in location 2, where the volume of solder is greater and more susceptible to bulk deformation. Whiskers, being a surface phenomena, will not likely form in these regions. The stress relaxation will likely occur as a result of one of the competing mechanisms. 145

160 Figure 104: Whisker growing from Sunflower at grain boundary of recrystallized grains Figure 105: Hillock growing from Sunrise after thermal shock 146

161 Figure 106: Senju M42 bulk recrystallization Interfacial IMC layer growth has been identified as a potential source of compressive stresses (section 1.2). The IMC may continue to grow during testing, particularly during the high temperature portion of the cycle. If this growth continues in an irregular, scallop shape, it will continue to drive compressive stress into the solder. Figure 107 provides a comparison of IMC thicknesses after exposure. The IMC layers, which form and grow at the board side of the solder joint, do not show significant distinction between the solders. Table 40 provides a summary of test for equal variance and ANOVA testing of the various alloys. This table provides probability (p-values) for each test. In all cases, the null (H 0 ) and alternate hypotheses (H a ) are as follows: H 0 : µ SAC305 =µ Senju M42 =µ Sunrise =µ Sunflower and SAC305 = Senju M42 = Sunrise = Sunflower H a : at least one µ is different and at least one is different There is no statistically significant difference in the thickness measurements of the IMC layers formed using each of the four alloys on the board side of the QFP after 1610 cycles of Thermal Shock. However there is a difference in the thickness of the layer that at the component side of the solder joint after Thermal Shock. In the case of the board side IMC 147

162 IMC Thickness (µm) layer, the H 0 is assumed to be valid while in the case of the component side, it is rejected. The difference in variance of measurements for each alloy is not significant as determined by a p-value The difference in the mean (µ) thickness measurement at the component side, in which SAC305 and Senju M42 have similar distributions, Sunrise and Sunflower have similar, and smaller thickness distributions (Figure 108). This suggests that the Bi content has some impact on the growth of the IMC layer under thermal shock conditions. At the lead side, the mean of the IMC layer height is lower in Sunrise and Sunflower than in SAC305 and Senju M42, which have roughly the same overall height. The lead side IMC is of more interest than the board side as the whisker growth has been observed mainly in locations 1, 3 and 4 (Figure 98) where a thin layer of solder and the IMC layer interact potentially creating ideal conditions for whisker growth. The board side IMC interacts with a larger volume of solder, at location 2. This allows for more bulk recrystallization and massive deformation and may not drive the growth of whiskers SAC305 Senju M42 Sunrise Board Side Sunflower SAC305 Senju M42 Lead Side Sunrise Sunflower Figure 107: IMC measurements of QFP solder joints after 1610 cycles thermal shock (U1) 148

163 IMC Thickness (µm) Table 40: Results of ANOVA test for equal variance and compare means of IMC thickness of the QFP IMC layer after 1610 cycles thermal shock. Total IMC Variable / p value Equal variance ANOVA Board Side Location Lead Side % CI for the Mean - Lead Side SAC305 Senju M42 Sunflower Sunrise The pooled standard deviation was used to calculate the intervals. Figure 108: Interval plot of IMC thickness at the lead side of QFP after thermal shock 149

164 Figure 109: IMC layer at lead with a) SAC305 and b) Sunrise after 1610 cycles of thermal shock Figure 110 shows a whisker growing from SAC305 at location 4. The image on the right shows that a thin layer of solder, approximately 10µm where the relative influence of the IMC layer, which ranges from 1-4µm, may be great. Figure 110: Whisker growing from SAC305 after thermal shock An examination of the solder surface as well as cross sections revealed that the Bi is not uniformly distributed through the solder joint. Figure 111 shows the Bi concentration at various locations along the solder joint formed using Sunrise solder paste after 1610 cycles of thermal shock. The Bi content is lowest in the thin areas of the lead, those most susceptible to whisker formation. 150

165 Figure 111: Bi content at various locations of a Sunrise solder joint In both Sunrise and Sunflower, where the concentration of Bi is 7wt%, the Bi precipitates, seen as the lighter areas in Figure 112, have been found around the grain boundaries. Bi is also seen precipitating out of the primary Sn dendritic structure and can be seen as fine particles, particularly visible in the top two images of Figure

166 Figure 112: Cross section of Sunflower showing Bi accumulating at grain boundaries 4 Summary of Findings and Conclusions The whisker testing executed comprised a screening experiment. A systemic count and measuring scheme, similar to those carried out in other studies 10,13,16 is required to provide statistical conclusions. In this study, two samples of each alloy, consisting of 176 leads each were tested in each environmental condition. This provides a large enough sample size, however it was found that the length of test, specifically the HTHH produced an insufficient number of whiskers and no whiskers of excessive length, which would be of concern. The current required length of test is 3000 hours; in this experiment, only 1000 hours was completed. From previous work 6, it is known that interruption of test corresponds to the interruption of whisker growth. It is therefore recommended that the samples be reexamined after an additional 3000 hours of HTHH cycles of Thermal Shock testing is sufficient to meet the requirements for whisker testing. In this case, all four alloys passed the requirements set out in JEDEC Standard 152

167 No. 201A for class 2 components. It is believed that the internal, compressive stresses induced by this type of thermal shock are very high (Figure 90) and may favor another form of stress relaxation over whisker growth, like massive bulk deformation. 10 The whiskers which formed on the surface of the three Bi-containing alloys all differed from those seen on SAC305. While all four alloys showed a small amount of hillock growth, particularly in locations 1, 2 and 4, the Bi containing alloys also showed a significant amount of small whisker nucleation sites. Further study is needed to quantify this behavior, however it is believed that stress relief through many, short whiskers presents less of an overall reliability concern. If sufficient stress is relieved through many short whiskers, it is believed that the likelihood of one whisker continuing to grow to a catastrophic length may be reduced. Another observation of this study worthy of further consideration, is the local presence of Bi at the site of whisker growth. It was found that the final composition of the solder, which is a combination of the solder paste, the tin plating on the component and any diffused Cu from the lead and board, is not uniform along the entire length of the lead. In areas which have the greatest risk of whisker growth, those along the upper lead where the solder is thin, also have the lowest overall composition of Bi. In solder joints formed with Senju M42, which has an initial concentration of 3wt% Bi, the Bi concentration at areas of whisker growth was experimentally found to be between 1.3 and 1.8wt%. Solder joints formed from Sunflower and Sunrise, which have an initial concentration of 7wt% Bi had between 2.9 and 4.1 wt% Bi in areas of whisker growth. Finally, in Sunrise and Sunflower, each of which have 7wt% Bi, the Bi was found to have precipitated from the Sn, likely from the interdendritic eutectic region. This process was shown, in section 6.1.1, to have started during solidification. Over the course of the thermal cycling, the Bi particles have been observed along the grain boundaries. Additionally, very fine precipitates appear to have formed during the course of thermal cycling. Therefore it is not known what the overall impact of Bi content in the solders is on the overall propensity for whisker growth. 153

168 5 References 1 JP002: Current Tin Whisker Theory and Mitigation Practices Guideline, March P.T. Vianco Dynamic Recystallization (DRX) as the Mechanism for Sn Whisker Development. Part I: A Model in Journal of Electronic Materials, Vol. 38, No. 9, L. Panashchenko Evaluation of Environmental Tests For Tin Whisker Assessment M.Sc. Thesis, Dept. Mech. Eng., University of Maryland, College Park, Maryland JESD22-A121A: Test Method for Measuring Whisker Growth on Tin and Tin Alloy Surface Finishes, July JESD201A: Environmental Acceptance Requirements for Tin Whisker Susceptibility of Tin and Tin Alloy Surface Finishes, September P.Snugovsky, S. Meschter, Z. Bagheri, E.Kosiba, M.Romansky, J. Kennedy Whisker Formation Induced by Component and Assembly Ionic Contamination in the Journal of Electronic Materials, Vol. 41, No. 2, P.Snugovsky, Z. Bagheri, M. Romansky Whisker Growth on SAC Solder Joints: Microstructure Analysis in ICSR SMTA Conference, Toronto, ON, IPC-A-610E-2010: Acceptability of Electronic Assemblies 9 K-N. Tu Spontaneous Tin Whisker Growth: Mechanism and Prevention in Solder Joint Technology : Materials, Properties and Reliability New York, Springer, 2007, ch. 6, pp S.J. Meschter, P. Snugovsky, J. Kennedy, Z. Bagheri, and E.Kosiba Strategic Environmental Research and Development Program (SERDP) Tin Whisker Testing and Modeling: Thermal Cycling Testing presented at International Conference on Solder Reliability, Toronto, Ontario, Canada, P.T. Vianco Dynamic Recystallization (DRX) as the Mechanism for Sn Whisker Development. Part II: Experimental Study in Journal of Electronic Materials, Vol. 38, No. 9, P.Snugovsky, S. Meschter, Z. Bagheri, E. Kosiba, M. Romansky, J. Kennedy Whisker Formation on SAC305 Assemblies in TMS Conference, San Diego CA, S.J. Meschter, P. Snugovsky, J. Kennedy, Z. Bagheri, and E. Kosiba SERDP Tin Whisker Testing: Low Stress Conditions presented at International Conference on Solder Reliability, Toronto, Ontario, Canada,

169 14 D.Kwon Packaging Architecture and Assembly Technology in A Guide to Lead-free Solders: Physical Metallurgy and Reliability, Silver Spring, MD: Springer, 2005, pp A.Baated, K-S. Kim, K. Suganuma, S. Huang, B. Jurcik, S. Nozawa, B. Stone, M. Ueshima. Effects of Reflow Atmosphere and Flux on Tin Whisker Growth of Sn-Ag-Cu Solder presented at SMTAI, Chicago, Il, P.Snugovsky, E. Kosiba, S. Meschter, Z. Bagheri, J. Kennedy Assembly Cleanliness and Whisker Formation presented at IPC APEX EXPO Conference, San Diego, CA, M.Osterrnan Mitigation Strategies for Tin Whiskers prepared for CALCE Working Group, J-L Jo Tin Whisker Growth Mechanism and Mitigation for Lead-Free Electronics Ph.D. Thesis, Dept. of Adaptive Machine System, Osaka University, Japan N.Jadhav, M. Williams, F.Pei, G. Stafford, E. Chason. Altering the Mechanical Properties of Sn Films by Alloying with Bi: Mimicking the Effect of Pb to Suppress Whiskers in Journal of Electronic Materials, Vol. 42, No. 2, W.J. Bottinger, C.E. Johnson, L.A. Bendersky, K.-W. Moon, M.E. Williams, G.R. Stafford Whisker and Hillock Formation on Sn, Sn-Cu and Sn-Pb electrodeposits in Acta Materialia, Vol. 53 pp , GEIA-HB : Standard for Mitigating the Effects of Tin Whiskers in Aerospace and High Performance Electronic Systems, January S.Meschter, P. Snugovsky, J. Kennedy, S. McKeown and E. Kosiba Tin Whisker Testing and Risk Modeling Project SMTA Journal Volume 24, Issue 3 (2011) 155

170 1 Introduction Chapter 5 Mechanical (Drop) Shock Testing The proliferation of hand held devices and complex portable electronic devices occurred at the same time as the requirement of Pb-free solders took effect. 1 This confluence of usage and regulation has introduced some significant issues related to the drop shock response of new Pb-free solder alloys. These devices are particularly susceptible to accidental drops over the course of their life time. At the same time SAC305 solder, which is currently the favorite amongst commercial product manufacturers, is significantly stiffer than traditional SnPb solder, and therefore tends to perform poorly in response to drop shock. As described in Chapter Chapter 1, new alloys are being explored with the intent of improving upon SAC305. Within a SAC alloy, the Sn phase exhibits the lowest elastic modulus and yield strength, and is therefore the most ductile phase within the SAC solder. The IMCs which form within the bulk solder act to increase the overall strength and reduce the ductility of the solder. In drop shock conditions, this higher strength and lower acoustic impedance allow for the stress to more readily transfer to the soldercopper interface, the more brittle IMC layer. 2 Brittle fracture along the IMC is the typical failure mode in drop shock conditions of a SAC solder joint as opposed to a more ductile type fracture of SnPb solder joints, which typically failed within the bulk solder or a combination of bulk solder and interface fracture. 3 Additionally, an increased amount of Ag can lead to the formation of platelet shaped IMC, as opposed to the finer particles found with low concentrations of Ag. These platelets may act as stress concentrators, reducing the overall performance of the solder joint in drop shock testing. It has been shown that improved drop shock performance can be attributed to lowering of the Ag component of an alloy, i.e. from SAC405 to SAC105. This transition to better performance occurs around the 3% Ag level, alloys with Ag below 3% perform better 156

171 than those with Ag above 3%. 4 Lowering the Ag content will reduces the yield strength and consequently increases the bulk solder joints ability to dissipate high plastic energy. 1 Finally the higher process temperature required for proper formation of SAC solder joints presents two additional problems, which impact the solder joints resistance to drop shock: increased exposure to higher temperatures may lead to a thickening of the brittle IMC interface layer abdrequiring higher T g board materials to withstand the higher temperature. These new board materials are more prone to a failure mode which has not been typically seen with SnPb alloys pad cratering. Any new alternative alloy will therefore need to address both failure modes. Bi as an alloying element has typically been used to reduce the melting temperature of the main alloy. Bi may also suppress the formation of Ag 3 Sn platelets and reduce intermetallic compound (IMC) growth. Furthermore, Bi acts to refine the grain structure of the bulk solder when added to a SAC solder joint. 2 This work considers a number of low (or no) Ag Pb-free alloys with differing degrees of Bi. All of these alloys have melting temperatures between 10 and 18 C lower than that of SAC305 allowing for the use of a lower T g board material. It is believed that the improved properties of the new, low melt alloys, along with the lower T g board material will have a combined effect of improving the mechanical strength of the overall solder joint. In this screening experiment the drop shock testing was performed on boards with an OSP finish. It should be noted that the board finish, and subsequent interconnection IMC, may greatly influence the performance of an alloy. Future testing should therefore be performed on other surface finishes. 2 Experimental Set Up The work presented in this section forms a screening experiment in which the selected alloys were screened for drop shock response; the results are intended to determine which, if any, of these alloys would be good candidates for further testing. 157

172 2.1 Materials Two alloys with low Ag (Senju M42 and Sunrise), and one alloy with no Ag (Sunflower), and varying amounts of Bi, were tested against SAC305. The alloy selection method was described in Chapter Chapter 1. The board material was also a factor in these screening experiments. Two board materials were tested; a normal T g material one which is typically used for SnPb applications around 150 C, and a high T g material one which is now required for most Pb-free applications at around 170 C. Table 41: Build matrix for drop shock testing Alloy Composition Board Material Assembly Temperature SAC305 Sn 3%Ag 0.5%Cu 240 C Senju M42 Sn 2%Ag 0.75%Cu 3%Bi High T g 224 C Sunrise Sn 1%Ag 0.7%Cu 7%Bi (170 C) 222 C Sunflower Sn 0.7%Cu 7%Bi 226 C Sunrise Sn 1%Ag 0.7%Cu 7%Bi Normal T g 222 C Sunflower Sn 0.7%Cu 7%Bi (150 C) 226 C 2.2 Test Vehicle Celestica s RIA3 test vehicle, shown in Figure 113, was used. It was originally designed to simulate a typical, medium complexity assembly. It is an 8 x10 PWB made up of 12 copper layers for a total thickness of with an Organic Solderability Preservative (OSP) finish. This board is one which is often used to test new, lead-free solders and other material parameters. LQFP176, PBGA256, CBGA64 and MLF20s were populated, two of each on each board. The BGA components all had SAC305 ball alloy and the reflow was performed in an air environment. While this test vehicle was not originally designed for evaluating consumer electronics, it served the needs of the screening experiment in that it provided for monitoring a variety of component types. 158

173 Figure 113: Test vehicle with monitored components Pad cratering has been identified as the main failure mode for Pb-free solders in mechanical testing. This is due to the changes in laminate materials required to survive a higher reflow temperature. PWB manufacturers have had to change the epoxy resins as well as increase the amount of ceramic particle filler material in order to reduce the coefficient of thermal expansion (CTE). 5 To mitigate this failure mode in the drop shock portion of testing, Solder Mask Defined (SMD) BGA pads were used. Figure 114 illustrates how the SMD boards introduce a sharp corner to the solder joints. It was intended that this corner act as a stress concentrator within the solder joint and drive any failure into the solder as opposed to within the laminate material, thereby allowing for a comparison of the different solder alloys. SMD pads are also used in many commercial applications such as cell phones and are therefore representative of commercial products where drop strength would be a requirement. 159

174 2.3 Assembly Figure 114: Solder mask defined vs. non-solder mask defined 6 The assembly process utilized surface mount technology (SMT) parts, as described above, placed and then reflowed with target reflow profiles outlined in Table 41. Primary SMT was performed with No Clean flux in a ten-zone oven. 42 RIA3 boards were built with an OSP finish and solder mask defined (SMD) copper pads. The RIA3 card was cut in half in order to better represent the smaller board sizes used in commercial applications (i.e. smart phones). One half of the board (RIA3-1) had five monitored components, while the other half (RIA3-2) had only one monitored component as per Table 42 and Figure Test Strategy Five boards of each combination were exposed to drop shock testing, one was used for Time 0 analysis, and one was held back as a do nothing board for possible verification purposes which may arise in the future. Table 42: Monitored components for drop testing Board Component Reference Designator RIA3-1 BGA-256 U204 RIA3-1 BGA-256 U205 RIA3-1 QFP-176 U1 RIA3-2 QFP-176 U2 RIA3-1 CSP-64 U208 RIA3-1 CSP-64 U

175 In order to test the mechanical strength of the solder joints, board level drop testing, based on JESD22-B110A 7, was performed. This test method aims to evaluate a subassembly s ability to withstand moderately severe shocks as a result of suddenly applied forces or abrupt change in motion. Within a subassembly there are four basic types of failure which typically resulting from drop shock: 1) Fracture, or permanent deformation, caused by high applied stress 2) Chatter created by high acceleration levels, e.g. causing bolts to loosen 3) Impact between adjacent objects, caused by high displacement 4) Momentary electrical failure associated with a shock pulse 8 In this test, condition 1 is the most likely and desired failure mode. Conditions 2 and 3 are mitigated by the test set up in which all resistance cables, strain gages and accelerometers are secured using room temperature vulcanization (RTV) silicone, as seen in Figure 115, and loaded within a fixture allowing for sufficient spacing to avoid impact with other objects as seen in Figure 116. Condition 4 presents a problem as a momentary electrical failure, which recovers after a shock pulse, and may be the result of an electrical response. This is mitigated by using an event detector to monitor failures, which require an increase of 300 for at least 200ns. An electrical pulse less than this would not be recorded as a failure. 161

176 Table Accelerometer Figure 115: Example of accelerometer secured with RTV silicone Figure 116: Test set up As per JESD22-B110, the subassembly was supported in a manner which is representative of field conditions, in this case allowing the PWB to flex during the drop shock. The parameters are a peak acceleration of 1500G with a 0.5ms duration defined in 162

177 JESD22-B110 Service Condition B in a half-sine waveform. Figure 117 and Figure 118 show the target and the achieved half-sine waveform respectively. In Figure 118 the red line represents the acceleration experienced by the drop table, monitored by an accelerometer attached directly to the table. A limited amount of rebound was measured immediately after the drop. This may be attributed to the movement of the accelerometer relative to the board (or table) to which it is affixed, which should be minimized by use of RTV silicone or a noise in the collected signal. It may also be the result of poor dampening. Compared to the initial drop shock signal, this rebound is relatively small and therefore not considered a significant contributing factor. The green line in Figure 118 represents the acceleration experienced by the PWB itself and is monitored via an accelerometer attached directly to the board, generally directly opposite the monitored component shown in Figure 113. As the card is secured to the fixture by posts at the four corners of the board, it was allowed to freely flex back and forth after the initial shock impulse. Per JESD22-B110, this back and forth flexure is representative of a subassembly life condition and therefore a desired part of the test set up. An Analysis Tech STD-256 event detector was used to monitor the resistance threshold of the components under test during the mechanical shock. A failure was recorded when the channel resistance increased by 300 or more for at least 200ns. The testing was stopped as soon as an electrical failure was detected (first failure). Figure 117: Target pulse shock defined by JESD22-B110 service condition B 163

178 Figure 118: Sample of pulse shock achieved during test While the method outlined for this test can be used to calculate the change in velocity and the displacement of the monitored items, in this project it was used to compare the response of the various alloys. Specifically, the number of drops required to produce an electrical failure was measured, while all other parameters were monitored to ensure consistency of test conditions. 3 Reliability Results Drop shock testing as described above was performed on five boards of each combination described in Table 41. Each of the two halves was tested separately. It was initially intended that all components would be monitored to failure allowing for 10 failures to be recorded for each component type. This would provide a small sample size, but one sufficient to produce basic Weibull plots. A number of events occurred during the testing which reduced the number of components tested. These events are outlined below: Based on similar tests performed in the past, it was deemed that additional weights were needed to increase the strain and help induce failure in the solder joints. Two 220g weights were added to the center of each board. 9 After testing five cards, it was determined the weight was not needed to induce failure. The resulting data 164

179 associated with RIA3-1 portion of five boards was not used in the analysis. Testing then continued with no additional weight added to the remaining boards. The process of attaching the weights described above involved drilling into the laminate material to securely attach the additional weight. The resulting hole was located very close to U1 on RIA3-1. It was therefore decided not to use any data obtained from the monitoring of U1 as it could not be determined whether the failure was due to the drilling process or the drop shock, or a combination of the two. No drilling was performed on RIA3-2 and therefore all data from U2 was used in this evaluation. The CSP components, as well as the BGA, did not fail. These components were located towards the edge and corners of the test vehicle and would therefore not experience the same degree of strain as those located in the centre of the test vehicle. Therefore, data from a total of 25 out of 30 BGAs and from all 30 QFPs was available for evaluation. The results are given in Table 43 and in Figure 119 and Figure

180 Table 43: Drop test results Board Material Drops to Failure Alloy Board Number (T g ) BGA (U205) QFP (U2) 22 --* SAC Senju M High Tg (170 C) * 45 Sunrise * * 57 Sunflower Sunrise Normal T g (150 C) Sunflower * 84 * Samples removed for reasons provided on page

181 Figure 119: Individual value plot of drops to fail, BGA (SAC305 + alloy) (U205) Figure 120: Individual value plot of drops to fail, QFP (U2) 167

182 The results of testing, albeit with a small sample size, indicate the board material was a more significant factor in drop shock performance rather than the alloy. Normal T g boards outperformed High T g boards in all cases. The data also indicates that SAC305 outperformed the other alloys when comparing the High T g board materials. This is particularly significant in the QFPs, likely due to the fact that the QFP solder joints are made up of 100% test alloy, the BGA solder joints are made up of 13% test alloy and 87% SAC305 from the component solder ball. There is therefore less of a difference in the overall composition of the BGA solder joints. Among the three test alloys, it appears that Sunflower performed the best; however there is not enough data for a conclusive determination. A larger sample size and perhaps a more appropriate test vehicle would be able to provide a greater distinction between the alloys. It is important to note that the limited reliability data provides only a cursory view, providing information only on which solder joints failed during the drop testing. For a better understanding of what occurred, failure analysis of the solder joints was required to form a clearer picture of the drop shock response. Both reliability and failure analysis need to be assessed together for a complete understanding. 4 Failure Analysis and Microstructural Evaluation Failure analysis of the drop shock test boards was performed by one of two methods: Dye and Pry (D&P) Testing: This method provides an overall image of the extent of the failures across an entire component, as well as allowing for the determination of the failure mode Cross Sectional Analysis: This method provides a clearer observation of the failure mode in the solder joint; however, it limits the number of solder joints to be examined. 168

183 4.1 Dye and Pry Procedure Dye and Pry testing was performed per Celestica s procedure DOC Dye and Pry Testing 10. The boards were immersed in red dye and subjected to a vacuum in order to force the dye into any pre-existing cracks, which occurred during drop testing. The dye was cured prior to component removal. The fracture surfaces were then inspected for the presence of dye a percentage of dye penetration for each solder joint was determined, as was the mode of failure (refer to Figure 121). Figure 121: Failure modes of solder joint as defined by IPC/JEDEC Failure Isolation Procedure The location of failures was isolated by probing the daisy chain pattern using vias on the bottom side of the PWB for BGAs, or between leads for QFPs. Cross sections were then prepared at locations where electrical failures were found to have occurred. 4.3 Evaluation of High T g Board after Drop Testing Evaluation performed using D&P and cross sectioning showed that BGAs on high T g boards failed exclusively by pad cratering. Figure 122 illustrates a pad crater from a cross sectional perspective. The solder joint remains completely intact; no cracking was seen along the IMC or within the bulk solder is observed, furthermore there was no separation between the IMC and the copper pad. All stress was dissipated through the laminate material, which failed between the epoxy and the layers of glass weave within the board material. 169

184 Figure 122: BGA failure by pad cratering, Sunflower on High T g boards Dye and pry results allow for an overall view of the failures within a component, both the number of failed solder joints and the extent to which the fracture penetrated through the solder joint. It is important to note that pad cratering may not result in an electrical failure. Partial fractures, as indicated by the various shades of yellow/orange in the D&P mapping, allow for some electrical contact to be retained. Even complete fractures, indicated by red, may still have intact copper traces associated with them. These types of failures are particularly insidious because they are not readily identifiable but do present a real reliability threat as the soft copper trace continues to be stressed. Figure 123 shows the D&P mapping of one component with accompanying images of representative failures in specific solder joints. The number within each cell of the map corresponds to a failure mode described in Figure 121. Table 44 provides a summary of the failures, as seen by D&P of all the alloys tested on High T g board material. % Fracture corresponds to the total area of the fracture surface, which was penetrated by dye. This indicates the degree to which the fracture penetrated through the solder joint prior to the pry process. The next four columns show the number of solder joints, as a percentage of the solder joints on the component, which showed some degree of dye penetration. From this we see that Senju M42 showed the least amount of damage. All of the components selected for D&P failed between 8 and 10 drops and all showed complete dye penetration within many of the solder joints on the outer edge of the component closest to the center of the 170

185 board. All solder joints which did not show any dye penetration resisted fracture up to the point of first failure. Figure 123: D&P mapping of Senju M42 on High T g board, with images of board side and b) component side of the fracture surface Table 44: Failures of BGA on High T g boards % Fractured High T g Board SAC305 Senju M42 Sunrise Sunflower % 7% 4% 7% 6% 50-75% 5% 1% 3% 4% 25-50% 7% 5% 9% 7% 5-25% 5% 1% 4% 5% Total: 24% 11% 23% 21% The QFPs on high T g boards did not exhibit any pad cratering. This was likely due to the structure of the part, as illustrated in Figure 124, which is much more compliant than a BGA. With copper leads on all four sides, the component itself is able to efficiently dissipate stress. 12 Electrical failures were mainly associated with damaged leads, as seen in Figure 125a, which after a number of drops failed at the knee where large repetitive 171

186 strain would have occurred. Although no complete mechanical failure through the bulk solder was observed on QFPs, the beginning of IMC failure was observed in a number of the solder joints. The crack would begin on the outside of the solder fillet and move to the copper lead where it would continue to propagate along the brittle IMC layer as seen in Figure 125b. Figure 124: Schematic of Quad-Flat-Package (QFP) 13 Figure 125: QFP failures in Sunflower on High T g boards a) fractured lead b) solder fillet fracture 172

187 4.4 Evaluation of Normal Tg Board after Drop Testing Sunrise and Sunflower alloys were tested on Normal Tg boards, as seen in Figure 119 and Figure 120. Both alloys survived, on average, more drops to first failure than on High T g boards. Failure analysis also revealed differences in the failure mode and the degree of damage occurring within the component and within each individual solder joint. Figure 126 shows the D&P mapping of Sunrise on a Normal T g board. While a small amount of pad cratering was visible on one edge of the component, the dominant failure mode was between the IMC interfaces with the bulk solder material towards the board side of the component. Figure 127 shows this same failure mode, between the IMC and bulk solder, as seen in a cross section of a BGA solder joint formed between SAC305 (BGA ball) and Sunflower (paste). Finally, Table 45 summarizes the degree of failure which occurred in BGAs using these two alloys. While the amount of failure appeared relatively similar between the two alloys, it is significantly less damage than was observed on the High T g boards using the same alloys. Figure 126: D&P mapping of Sunrise on Normal Tg board, with images of board side and b) component side of the fracture surface 173

188 Table 45: Failures of BGA on Normal Tg boards % Fractured Normal T g Board Sunrise Sunflower % 1% 1% 50-75% 2% 1% 25-50% 1% 0% 5-25% 4% 4% Total: 8% 6% Figure 127: BGA failure in Sunflower on Normal T g boards The QFPs, as in the case of High T g boards, failed via lead fracture as seen in Figure 128a. No pad cratering was observed. Fracture, initiating in the solder fillet and propagating along the IMC was observed, however no complete fracture of this type was observed. 174

189 Figure 128: QFP failures in Sunflower on Normal T g boards a) fractured lead b) solder fillet fracture The solder alloy may be a factor in failures caused by fractured leads; the cracks appear to originate in the solder material and then propagate through the lead at a point of high mechanical strain. 14 This is illustrated in Figure 129. Figure 129: QFP failure in Sunrise on Normal T g board 5 Summary of Findings and Conclusion The drop testing carried out comprised a screening experiment. While the sample size may not be statistically significant, it does show some promising results. Many of the results relate to the board material rather than to the solder material. This is significant as 175

190 lower T g board materials are a viable option for lower melting temperature solders, SAC305 often requires higher T g board material. Normal T g boards appear to outperform High T g boards independent of solder alloy. High T g boards failed exclusively by pad cratering, which is a latent defect and therefore not readily detectable. Pad craters present a reliability concern because they will propagate through the laminate and eventually sever the copper trace. Pad craters will not be detectable until the copper trace is broken leading to electrical failure. Even if the laminate defect could be detected at manufacturing, it is not readily repairable. Normal T g boards survived longer. While the solder joints built on Normal T g boards exhibited mixed failure modes (sometimes failing by pad cratering and other times by solder fracture at the IMC) there were also significantly less failed solder joints, and they primarily failed at the IMC interface. Additionally, there were some differences among the solder alloys used in drop shock testing. On the High T g board material, there is not enough data to properly distinguish between the performances of each alloys in terms of drops to first fail, however Senju M42 does appear to have sustained the least amount of damage. Furthermore, when comparing Sunflower to Sunrise on both High and Normal T g board materials, it appears that Sunflower slightly outperforms Sunrise. The following are recommendations for further testing based on this screening experiment: The ideal test vehicle would have one component at the centre of the board, which could be monitored to failure. Each solder/board material combination should include 32 repeats, of which at least 63% are tested to failure in order to provide sufficient data for reliability plots Using both SAC305 and SnPb as a baseline for comparison would provide a better understanding of how low-melt, Bi-containing alloys perform against current industry standards. 176

191 6 References 1 D.A. Shnawah, M.Sabri, I.A. Badruddin A review on thermal cycling and drop impact reliability of SAC solder joint in portable electronic products, Microelectronics Reliability, 52 (2012) pp P. Ranjit and T. Lawlor., Effects of Silver in common lead-free alloys 3 T. A. Woodrow and J. Bath Lead-Free Reliability in Aerospace/Military Environments in Lead-Free Solder Process Development, New York, Wiley, G. Henshall, R. Healey, R. Pandher, K. Sweatman, K. Howell, R. Coyle, T. Sack, P. Snugovsky, S. Tisdale, F. Hua. inemi Pb-Free Alloy Alternatives Project Report: State of the Industry, proceedings SMTAI, p. 109 (2008). 5 B.Gray Correlation of Printed Circuit Board Properties to Pad-Crate Defects Under Monotonic Spherical Bend, M.A.Sc. thesis, Dept. Mech. Eng., Ryerson University, Toronto, Ontario PCB Layout Recommendations for BGA Packages, Lattice Semiconductor Corp. Oct Subassembly Mechanical Shock, JESD22-B110A, November D.S. Steinberg Designing Electronics for Shock Environment, in Vibration Analysis for Electronic Equipment, 3 rd ed. New York, Wiley, ch. 11, pp W.Liu, N-C, Lee, S. Bagheri, p. Snugovsky, R.Brush, J.Bragg, B.Harper. Drop Test Performance of BGA Assembly using SAC105Ti Solder Spheres in IPC APEX EXPO Proceedings, Dye and Pry Testing, Celestica DOC , Monotonic Bend Characterization of Board-Level Interconnects, IPC/JEDEC-9702, June P. Snugovsky, J. Bragg, E. Kosiba, M. Thomson, B. Lee, R. Brush, S. Subramaniam, M. Romansky, A. Ganster, W. Russell, J. P. Tucker, C.A. Handwerker, D.D. Fritz. Drop Test Performance of A Medium Complexity Lead-Free Board After Assembly and Rework in the IPC APEX EXPO Conference Proceedings, J. Juarez Jr., M. Robinson, J. Heebink, P. Snugovsky, E. Kosiba, J. Kennedy, Z. Bagheri, S. Suthakaran, M. Romansky Reliability Screening of Lower Melting Point Pb-Free Alloys Containing Bi, in IPC APEX EXPO Conference, Las Vegas, NV,

192 Chapter 6 Summary of Findings and Conclusions This work explored the reliability performance and microstructural differences of three low melting temperature, low-ag, Bi-containing alloys compared with SAC305. From the current work, the following conclusions can be made: The three low melting temperature, low-ag, Bi-containing alloys under investigation all produced good solder joints with QFP type components. BGA components with SAC305 solder balls also formed good solder joints when built with these pastes; all examined solder joints had an acceptable degree mixing. Differences in microstructure have been observed related to the addition of Bi, decrease in Agcontent, and varying amounts of Cu dissolution resulting from different process temperatures. Solder joints with final Bi concentrations of less the ~3wt% showed no detectable Bi precipitates. This includes QFP solder joints formed with Senju M42 and all BGA solder joints formed with a combination of paste and SAC305 solder balls. In these cases, all the Bi was present within the Sn phase upon solidification. Bi was observed to precipitate out of solid solution from as assembled solder joints with Bi compositions of greater than 7wt%. This is the case with QFP solder joints formed using Sunrise and Sunflower solder pastes. Bi also remains present in the Sn phase, observed through EDX analysis. It is therefore likely that the Bi precipitates are formed from the binary, ternary or double binary eutectic regions within the interdendritic spaces of the Sn phase where the concentration of Bi in Sn is highest. Both fine and medium sized particles were observed. In Sunflower, the concentration of Bi within the Sn dendrites and within the Sn portion of the binary eutectic ( Sn + Cu 6 Sn 5 ) within the interdendritic spaces appeared to be approximately equal. Bi concentrations greater then 7wt% had a beneficial impact on the changes to bulk microstructure, which occurred during accelerated thermal cycling. The Bi precipitated from the bulk solder evenly in very small particles. This favorable 178

193 influence did not appear when the Bi concentration was 3wt%. There was some degree of Bi segregate along the grain boundary observed in the Sunflower alloy; however there did not appear to be full segregation of the phases. After exposure to thermal shock, Bi particles were observed along the grain boundaries in solder joints formed with alloys containing 7wt% Bi. Additionally, very fine precipitates appear to have formed during the course of thermal shock. Cu 6 Sn 5 ( ) interfacial IMC layers that form between the studied solders and the Cu pad of the board or the lead material of the QFP solder joint during solidification were not statistically distinguishable in terms of thickness. No other species was detected at the solder/cu interface. In the BGA solder joint, the interfacial IMC layer that formed between the combined solder paste/sac305 solder ball and the Cu pad of the board was found to be (Cu,Ni) 6 Sn 5. The Ni is contributed to this system from the component side IMC and diffuses through the bulk solder prior to solidification. Using a 93% CI, SAC305 appeared to have a thicker IMC layer upon solidification. This can be attributed to the higher process temperature. The interfacial IMC layer that formed between the Cu/Ni/(paste/SAC305 solder ball) was found to be (Cu,Ni) 6 Sn 5 and Ni 23 Cu 33 Sn 44. Due to the irregular shape of this IMC layer, it was difficult to compare the mean thicknesses of the IMC layer that formed when using different solder paste alloys. During thermal cycling, the total thickness of the interfacial IMC continued to increase, and a ( ) Cu 3 Sn layer formed. While the overall thickness increased similarly for all four alloys, the Cu 3 Sn layer of the Sunflower alloy was much thicker than the other alloys. This alloy has no Ag, which indicates that Ag plays a role in suppressing Cu 3 Sn growth. Bi concentration did not appear to significantly change the growth rates of either intermetallic compound. The increased growth of the Cu 3 Sn phase in the Sunflower alloy did not correspond to any decrease in reliability during ATC. 179

194 SAC305, which was processed at a higher temperature, initially had a slightly thicker interfacial IMC layer. During thermal cycling, the IMC layers formed in solder joints using the various alloys increased to similar overall thicknesses. The three lower process temperature solders performed as well as, or better than SAC305 during accelerate thermal cycling. During two different thermal cycling regimes (0 to 100 C and -55 to 125 C) the only failures which could be attributed to solder joint fatigue failures occurred in SAC305 during 0 to 100 C cycling and one sample of Sunrise, which failed towards the end of the test. All other electrical failures were attributed to failures within the board material. High T g board material appeared to outperform Normal T g board material during ATC. High Temperature, High Humidity conditions (85 C, 85%RH) produced an insufficient number of whiskers and no whiskers which would fail the requirements of JEDEC Standard No. 201A for class 2 components. This test was run for 1000 hours, while current required length of test is 3000 hours. It would therefore be worthwhile to continue testing cycles of Thermal Shock testing is sufficient to meet the requirements for whisker testing. In this case, all four alloys passed the requirements set out in JEDEC Standard No. 201A for class 2 components. The whiskers which formed on the surface of the three Bi containing alloys during thermal shock all differed from those seen on SAC305. All alloys showed a small amount of hillock growth, the Bi containing alloys also showed a significant amount of small whisker nucleation sites. This may prove beneficial as many short whiskers may effectively reduce the stress in the system and decrease the likelihood of one whisker continuing to grow to a catastrophic length. The final composition of the solder, formed between the solder paste and the dissolving Sn finish, is not uniform along the entire length of the lead on a QFP component. In areas which have the greatest risk of whisker growth, those along the upper lead where the solder is thin, also have the lowest overall concentration of Bi. 180

195 Normal T g boards appear to outperform High T g boards in drop testing, independent of solder alloy. This is significant as lower T g board materials are a viable option for lower melting temperature solders where as SAC305 often requires higher T g board material. The manner in which boards fail varies according to board T g : High T g boards failed exclusively by pad cratering, Normal T g boards, which survived longer, than High T g boards, exhibited mixed failure mode of pad cratering and/or brittle fracture at the IMC. While there was not enough data to distinguish the different solder paste alloys based on number of drops to failure, there was some interesting information gleaned during failure analysis regarding the degree or number of solder joints which failed per component. When analysing High T g boards, Senju M42 appears to have sustained the least amount of damage. The other three alloys were not distinguishable. Furthermore, when comparing Sunflower to Sunrise on both High and Normal T g board materials, it appears that Sunflower slightly outperforms Sunrise. Significantly fewer solder joints per component failed when using Normal T g boards. Based on comparisons to the mainstream consumer alloy SAC305, which has a minimum reflow temperature of 223 C and a Ag concentration of 3wt%, the new, low melting temperature, low-ag, Bi-containing alloys may be recommended as viable replacements. 181

196 Chapter 7 Future Work The work in this thesis, as well as that in the proposed future work, is a continuation of the research collaboration between Celestica Inc and U of T into the development of a reliable solder alloys, that will conform with the requirements of their respective end use environments while meeting the environmental lead (Pb)-free constraints imposed on the commercial supply base. 1 Celestica/Indium Sponsored Whisker Resistant Solder Paste It has been shown that component and assembly cleanliness affect the propensity of a solder joint to grow whiskers, particularly in high humidity environments. It was found that contamination impacted the degree to which corrosion formed at the surface, and consequently the extent of internal compressive stresses applied at the surface of the solder, which is a required element in whisker nucleation and growth. 1 In the production process, a major source of contamination is the flux residue, which may remain on the product after the build. The role of flux is to provide an oxide-free surface for the solder to bond with a base metal, usually copper. It is intended that the flux should be fully consumed, or burned off, by the end of the reflow process. In practice however, it is possible that some residue remains after production. This flux residue now behaves as a contaminate. Various flux types have been formulated to reduce the corrosive impact of the flux residue depending on the end use of the electronic product. In addition to RoHS legislation, the electronics industry is undertaking further environmental initiatives, including the reduction of halogens from electronic products. 2,3 This initiative restricts the use of chlorides and bromides throughout the electronic supply chain including the solder paste flux formulations. Halides have been used in fluxes in order to rapidly reduce the metal oxide and allow for the formation of a robust solder joint. Any remaining halide may, however, behave as a corroding agent if not fully 182

197 utilized in the reflow process. In a halogen-free solder paste, it is still unclear whether the alternative agents will be more, or less corrosive. In developing solder pastes for low melt applications with three to four elements (Sn, Cu, Bi and sometimes Ag), Indium and Celestica are working to optimize the solder flux as well as the alloy. It is therefore important to identify the corrosion mechanism(s) which may affect the various elements and compounds within a particular solder. For example, it is hoped that this work will determine whether the dominant form of corrosion is intergranular, galvanic or some other form. 23 flux/paste combinations will be tested according to the matrix outlined in Table 46 with the goal of better understanding the interaction of the flux/paste combination. Table 46: Solder paste test matrix Alloy Flux SAC305 Sn 3%Ag 0.5%Cu Violet Sn 2.25%Ag 0.5%Cu 6%Bi Sunflower Sn 0.7%Cu 7%Bi Base BaA1 BaA2 BaA3 Resistively 1 R1A1 R1A2 R1A3 Resistively 2 R2A1 MPU 1 M1A1 M1A2 M1A3 MPU 2 M2A1 Corrosion 1 C1A1 C1A2 C1A3 Corrosion 2 C2A1 ph (1) P1A1 ph (2) P2A1 P2A2 P2A3 ph (3) P3A1 P3A2 P3A3 ph (4) P4A1 183

198 2 ReMap M1: Lower Temperature Soldering Alloys with Improved Mechanical and Thermal Fatigue Reliability The ReMap M1 project has two main goals: to develop a test protocol for testing alloys in a combined thermal mechanical reliability test, and to evaluate Bi containing alloys according to this protocol. Currently, reliability tests exist for thermally induced stress (i.e. thermal cycling) and for mechanical stress (e.g. vibration testing, drop testing) separately. There is currently no existing industry standard test protocol for examining the interaction of these stress types. Electronic assemblies, particularly those intended for high reliability applications, are often exposed to a various stresses and often at the same time. It is believed that the material behavior resulting from thermal stress will impact the material response to mechanical stress. The mechanical behavior at various temperatures is therefore thought to be significantly different. Also, the behavior of the material as it cycles through different thermal regimes may also affect the response to mechanical stresses. It is believed that the inclusion of Bi in these alloys will result in material strengthening by one or more of the following mechanisms: solid-solution strengthening or precipitation hardening. The mechanism will likely dependent on the amount of Bi present in the alloy. 4 The aim of this project is to correlate reliability data with the metallurgical findings in order to better understand the behavior of various alloys under combined stress conditions. 3 ReMap M2: Alloys, Board and Component Surface Finish Interactions with Reduced Propensity for Whisker Growth The current work presented a preliminary assessment of the whisker mitigation properties of a number of Bi containing alloys. This will be further developed in the ReMap M2 184

199 project, in which a number of Bi containing alloys will be tested (Table 47), and assessed according to statistically accepted guidelines, in a variety of conditions, for whisker growth characteristics. These findings will also be compared to the microstructural properties of the alloys, as they have evolved in the listed conditions. In previous work, it has been shown that internal stresses are created from a variety of sources. High Temperature High Humidity (HTHH) and Ambient Temperature High Humidity (ATHH) where selected to evaluate how internal, compressive stresses are created at the surface of a solder joint by oxidation. This has been shown to affect Cu leaded solder joints more than Alloy 42. ATHH with contamination is an extension of the evaluation by oxidation, which is further enhanced by an ion-contaminated environment. Thermal cycling, at two different levels, will also be evaluated in order to study the affects of internal compressive stresses created from a mismatch of CTE amongst the various materials within a solder joint. It has been shown that Alloy 42 is affected more than Cu based lead frame alloys due to the larger difference in CTE. 5,6,7,8,9,10 It is believed that Bi mitigates the nucleation and growth of whiskers in a variety of ways. Even a small amount of Bi has been shown to lead to a more equiaxed grain structure. This is thought to elevate some of the internal compressive stresses on specific, columnar grain boundaries, which have been observed in plating material. Bi also acts as a grain refiner, by providing more grain boundaries, there is the possibility to annihilate dislocations and therefore absorb stress internally. Finally, with the lower process temperatures of Bi containing alloys, the scallop shaped interfacial IMC layer is likely to be suppressed. Due to its irregular shape, this IMC layer is a major contributor to the internal compressive stress of a thin film or a thin layer of solder. 11,12 185

200 Surface Finish ENIG OSP Table 47: ReMap M2 test matrix Alloy HTHH ATHH (85 C/85% RH) (25 C/85% RH) Test Condition ATHH Thermal Thermal (25 C/85% RH) + contamination Cycling Cycling (-55 C to (-20 C to 80 C) 100 C) SAC305 3x 3x 3x 3x 3x Senju M42 3x 3x 3x 3x 3x Sunflower 3x 3x 3x 3x 3x Violet 3x 3x 3x 3x 3x SAC305 3x 3x 3x 3x 3x Senju M42 3x 3x 3x 3x 3x Sunflower 3x 3x 3x 3x 3x Violet 3x 3x 3x 3x 3x 4 ReMap M3: Aging Effect of New Lead-Free Materials on Reliability The current work has focused on understanding the properties of Bi containing alloys in the short term, with respect to time from manufacture. In reality, all of the testing took place over 4 years and started approximately one year after the initial build, during which time all samples were stored at ambient conditions. The effects of aging in this study are therefore not well understood. It has been shown in a number of studies that both SnPb alloys and SAC based alloys exhibit significant deterioration of mechanical properties upon aging. This has been found to occur at both room temperature and elevated storage conditions, with SAC based solders exhibiting a deterioration of mechanical properties in the order of 25 times greater than that experienced by SnPb. 13 It is believed that the microstructural coarsening is the main mechanism responsible for the reduction in mechanical properties; coarsening 186

201 occurs on both SnPb and SAC based alloys (Figure 50,Figure 51) 14. In previous work, it appears that Bi containing alloys may actually improve the mechanical properties by reducing the amount of microstructural coarsening and through either solution hardening (in the case of low concentrations of Bi) or precipitation hardening (in the case of higher Bi content). 15 Figure 130 illustrates the microstructural change that occurred in Orchid (Sn2%Ag7%Bi) after 3000 cycles of harsh environment ATC (-55 C to 125 C). Significant growth of the Sn dendrites was not observed, and while the other IMC particles did appear to coalesce to some degree, the main observed change is that the Bi precipitated out of the Sn, likely from the interdendritic eutectic first, and remained as small, evenly spaced precipitates. Figure 130: Orchid at a) Time 0 and b) after 3000 cycles ATC Lower melting point alloys, specifically those containing Bi, may also improve the mechanical properties of a solder joint by suppressing the growth of the interfacial IMC layer. The increased reflow temperatures associated with SAC alloys, as well as the continuous growth of the interfacial IMC over time, has been identified as a major contributing factor to the reduced reliability of these alloys. 16 While these initial findings are promising, more study is required. Potential pitfalls of Bi precipitation also exist. If the Bi begins to coalesce into larger particles, it will no longer present the benefits associated with a small evenly spaced particles. Also, there is a danger that the Bi will preferentially locate along grain boundaries or along the interfacial IMCs, similar to the known issue with SnPb solders discussed in Section

202 During the course of this study, unexpected behavior was observed. Approximately two years after the High Temperature High Humidity exposure described in Section 2.2, samples, which had been stored at ambient conditions, where re-examined for further whisker growth. While a small amount of whisker nucleation was observed on the sides of the leads, a more interesting observation was that the Bi appeared to be precipitating out of the solder (Figure 131). Figure 131: Sunrise solder joint after 1000 hrs HTHH followed by 2 years ambient storage Through the ReMap M3 project, the team will investigate the evolution of three Bi containing alloys through a number of different aging conditions (Table 48). The microstructure and hardness of each alloy, exposed to each different test condition will be evaluated. The samples will then be down selected for reliability study. A subset of alloys and aging conditions will be tested in both ATC and vibration conditions, and possibly a combined environment as described in Section