The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper

Transcription

1 SOLDER PRODUCTS VALUE COUNCIL ASSOCIATION CONNECTING ELECTRONICS INDUSTRIES The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper A Research Report by the Lead Free Technical Subcommittee IPC SOLDER PRODUCTS VALUE COUNCIL

2 TABLE OF CONTENTS Mission Statement i Solder Products Value Council Members i Introduction and Statement of Problem Review of Test Program Executive Summary Void Data Summary Statistical Analysis of Void and Failure Data Metallographic Cross Section Results Summary and Conclusions Appendix A: Table of Voids and Failure Data Appendix B: Statistical Analysis of Voids and Failure Data The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper

3 IPC Solder Products Value Council Mission Statement In support of IPC s Mission Statement, IPC solder manufacturers recognize that the PCB and electronics assembly industries, comprised of the entire supply chain, must grow profitably. The IPC Solder Products Value Council (SPVC) Steering Committee s objective is to identify and execute programs designed to enhance the competitive position of solder manufacturers and their customers. Acknowledgement It is estimated that nearly $1 million was spent to conduct the round robin lead free testing program from which the data discussed in this paper was obtained. Each and all members of the IPC Solder Products Value Council contributed not only funds but also a significant amount of staff time in support of this program. However, like any program of this magnitude, the following companies and individuals have contributed to the program s success. The Council wishes to thank George Wenger and Pat Solan, Andrew Corporation; Engent AAT; Jasbir Bath, Solectron Corporation; Dongkai Shangguan, Flextronics International; Hallmark Circuits; Jean-Paul Clech; and Dean May, Crane Division-Naval Surface Warfare Center. IPC Solder Products Value Council Members AIM Inc. Henkel Technologies Nihon Superior Company Ltd. Amtech, Inc. Heraeus, Inc. P. Kay Metal Supply Inc. Avantec Indium Corporation Qualitek International Inc. Cookson Electronics Kester Senju Metal Industry Assembly Material Division Koki Company Ltd. Shenmao Technology Inc. EFD Inc. Metallic Resources Inc. Thai Solder Industry Corp Harimatec IPC Solder Product Value Council Lead Free Subcommittee Members Karl Seelig, AIM, Subcommittee Chairman Greg Munie, Kester, White Paper Editor William Gesick, Advanced Metals Technology Patrice Rollet, Avantec Paul Lotosky, Cookson Electronics John A. Vivari, EFD, Inc. Katsuji Takasu, Harimatec Douglass Dixon, Henkel Loctite Brian Bauer, Heraeus Inc. Brian Deram, Kester Masayuki Nakajima, Koki Company, Ltd. Nimal Liyanage, Metallic Resources Inc. Keith Sweatman, Nihon Superior Co. Ltd. Larry Kay, P. Kay Metal Supply Inc. Tippy Wicker, Qualitek Hiro Suzuki, Senju/Mitsui Comtek Corp. Mark Young, Shenmao Technology Inc. Somchai Vorasurayakamt, Thai Solder Industry Corporation Ltd. James Slattery, Indium Corporation i The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper

4 Introduction and Statement of Problem Due to marketing and legislative pressures in Asia and Europe, the electronics industry is moving to the adoption of lead free solders. These lead free materials are considered by some to be environmentally preferable to the current lead containing solders that dominate elec tronics manufacturing. Although the issue as to whether lead free solders are indeed environmentally preferred compared to lead containing solders is still under debate, market and legislative actions are forcing a change in materials used in electronics assembly. Accordingly, solder material suppliers are being asked to provide the electronics industry with solders that are lead free (per the accepted technical definition of that term) and yet still provide all the needed properties including ease of assembly and reliability the electron ics industry has come to expect from lead containing solders. At present, there are a large number of materials that have been proposed as replacements for Tin/Lead (SnPb) solder. Primary among these are the Tin/Silver/Copper (SAC) alloys. There are several variations of the SAC alloys that have been suggested as the preferred replacements for SnPb solders. Two are of special interest: the Japan (JEITA) adopted alloy of 96.5% Tin, 3.0% Silver and 0.5% Copper and the North American Electronics Manufacturing Initiative (NEMI) alloy of 95.5% Tin (Sn), 3.9% Silver (Ag), and 0.6% Copper (Cu). Both of these alloys have undergone significant testing. And both sponsors believe that their particular choice is the best candidate for replacement of SnPb solders. The SPVC (Solder Products Value Council) is an industry council comprised of 23 solder manufactures from around the world that is addressing issues related to solder assembly. The IPC Solder Products Value Council (SPVC) members are technically capable of providing any alloy requested by their customers. However, the IPC SPVC, as producers of solder alloys, believes it is in the best interests of the industry, from the standpoint of product consistency, quality and the conservation of natural resources to achieve a consensus on a standard lead free alloy for the electronics industry. To that end, the SPVC recently finished a 36 month, million dollar study of SAC alloys that included contributions of several organizations, including Engent Labs, NSWC Crane, Andrew Corporation, Flextronics, and Solectron. This study, which was completed in June, 2005, was designed to find a globally available, default lead free alloy. The findings of this study determined that SAC305 is the default alloy. Along with the performance of the SAC alloys, the study collected a significant amount of data on solder joint voids for the alloys. This data shows that process voids found and thermal fatigue failures seen in testing do not show a statistically significant dependence for the test vehicles and alloys examined. The electronics assembly industry generally considers voiding in BGAs as a potential defect in manufacturing. In doing so, the industry has adopted a maximum voiding specification of 25% of the ball X-ray image area. This is a debatable point since examination of the void is subject to energy levels of the X-ray, as well as beam angle. Extremes in energy levels can result in either a false pass or a false fail. Additionally, each manufacturer of X-ray inspection equipment gives a variety of ranges for the energy levels used during inspection. Unlike other test criteria where the pass/fail limits are specified by the particular piece of equipment, X-ray inspection is not specified. The type of voids is not specified, e.g. interfacial voids or voids in the bulk of the solder. However, in spite of this, most companies engaged in electronics assembly have adopted a void specification. In general, voiding seems to have more of an impact on handheld devices where high G-forces resistance is required. As a matter of fact, several papers have been written over the past few years that support the theory that voiding does not impact reliability. With advent of lead free solder, the voiding specification of 25% has been carried over to be used for lead free assemblies, as well. Lead free solder joints are known to void more than tin-lead solders, and SAC alloys void higher than other lead free alloys. 1 The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper

5 During the SPVC s analysis of thermal cycling and thermal shock failures with SAC alloys, void-to-failure ratio was studied. The results are in agreement with previous studies on tin-lead assemblies that demonstrated that there is no relationship between voids occuring in the bulk of the solder and thermal stress failures. The following paper presents data collected from the IPC SPVC study that supports the claim that voids in the bulk of the solder do not significantly impact BGA failure. It also demonstrates that failures occur at the package side of the ball bond pad, away from the solder joint. This is due to the die placement and CTE changes that occur across the package due to die layout. As no interfacial, champagne or Kirkendall voids were observed, this paper makes no inferences about the effects of those types of voids on solder joint reliability. Review of Test Program Methodology In order to determine what material is best suited to be the standard alloy, the IPC SPVC members reviewed the most likely candidates in the current list of contenders and carefully considered: What alloys are presently, through general acceptance, most likely to be used as SnPb solder replacements? What tests are applicable to make an accurate determination of the differences (if any) in the properties of the most likely candidates? Alloy Choice As was previously stated, the majority of potential standard replacement alloys are composed of Tin, Silver, and Copper with Silver varying between 3 and 4% and Tin varying between 95.5 and 96.5%. Prior to SPVC testing, the front runners were (in % of Tin/Silver/Copper) the 96.5/3.0/0.5 (JEITA) and 95.5/3.9/0.6 (NEMI) alloys. To cover that composition range represented by these alloys, the alloys chosen for testing by the IPC SPVC were: 96.5/3.0/0.5 Tin/Silver/Copper (Referred to as Alloy C in this report) 95.5/3.8/0.7 Tin/Silver/Copper (Referred to as Alloy B in this report) 95.5/4.0/0.5 Tin/Silver/Copper (Referred to as Alloy A in this report.) Elements of Testing Program To answer these questions, the IPC SPVC completed a three year round robin testing program. The elements of the program were: 1. Assembly Performance (initial screening) by council members of SAC alloys to compare basic alloy properties. 2. Dow n-select testing by Engent of SAC alloys provided by six solder manufacturers before assembly. 3. Assembly of Flextronics and Solectron test PCBs using down-selected SAC alloys. An industry standard eutectic SnPb solder was used as a control. 4. Base Line Metallographic Analysis of completed assemblies by Andrew Corporation 5. Thermal Shock and Thermal cycling conducted by NSWC Crane As noted above, this Research Paper is not intended to summarize all phases of the test program. The summaries of the first and second phases of the work, comparison of alloy properties and the comparison of assembly results, as well as the complete overview with all data on thermal testing and metallographic analysis has already been presented else where and is now available from IPC. A complete summary, including all data collected, was presented in a comprehensive third white paper also available form IPC. The intent of this work is to discuss observations made during the testing and metallographic analyses on the impact of voiding on solder joint integrity. As such a description of the phases of the testing program and an executive summary of the conclusions of this part of the study on voiding in solder interconnections are presented below. Board Assembly In support of the SPVC study, both Solectron and Flextronics populated 40 boards with each of the three SAC alloys and their incumbent Sn63/Pb37. Both companies assembled their own test vehicles using process parameters established by the SPVC Technical Lead Free Subcommittee. These two test vehicles are shown in Figures 1 and 2. 2 The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper

6 Time Above Liquidus 45 to 75 seconds is recommended. Time above liquidus may range from 30 to 90 seconds. Total Profile Length Time from ambient to peak temperature should be 3 to 4 minutes. Figure 1: Solectron Test Vehicle At Solectron, the production reflow oven used had 10 heating zones, with forced convec tion and a Nitrogen (<100 ppm O2) rich atmosphere. The lead free reflow profile was in the ranges specified. For the lead free SnAgCu solder pastes at the largest QFP 256 component on the board, solder joint peak temperature was 241 C and for one of the smallest components on the board (lead free 0.5mm CSP) solder joint peak temperature was 247 C with time over 217 C of 75 to 82 seconds. A 2D-Xray system was used for initial X-ray inspection after assembly. Voiding greater than 25% of the area was observed with all three lead free solder paste assembled alloy boards specifically for the 0.5mm CSP lead free components. For the tin-lead assembled boards, there was evidence of some voiding but much less than 25% void area on the 0.5mm tin-lead assembled CSP components. It should be noted that the SnPb solder paste used by both companies was their standard production solder paste and the time-temperature profile used for SnPb assembly was their standard reflow profile optimized for their SnPb solder paste. No rework was performed on the devices with voids, per agreement with the IPC SPVC technical committee members. Figure 2: Flextronics Test Vehicle Of primary concern was the reflow profile. A single common time-temperature profile was used for the three chosen SAC alloys. Target time-temperature values where chosen to ensure best possible wetting given the alloys under testing. It is important to note, that to preserve anonymity of the solder pastes used, there was no optimization of the profile. As a result, it is likely that an optimized profile could reduce the level of voiding. O2 ppm level of 1000 or less Ramp Rate to 1.5 C per second is optimal. The assembly compa nies agree the target would be 1-2 C. Peak Temperature 235 to 245 C is recommended. Peak tem peratures may range from 230 to 265 C. At Flextronics, the reflow oven used had 9 heating zones with a nitrogen (<1000 ppm O2) rich atmosphere. The reflow peak temperature was 240 C-248 C for the lead free solder and 217 C-222 C for the eutectic Sn-Pb solder. It also should be noted that the SnPb solder paste used by Flextronics was their standard production solder paste and the time-temperature profile used for SnPb assembly was their standard reflow profile optimized for their SnPb solder paste. An X-ray system was used for micro-focus, real-time, nondestructive inspection of the sol der joints. Voiding greater than 25% was observed with all three lead free alloys. Devices PBGA196, C-CSP224, and CSP8 all showed voids greater than 25% on almost all packages, but no voids were discovered on the LCC24 and BCC24. X-ray inspection on the eutectic Sn/Pb solder boards showed 3 The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper

7 no voids. No rework was performed on the devices with voids, per agreement with the IPC SPVC technical committee members. Test Methodology The test regime consisted of conventional industry accepted thermal cycle and thermal shock exposures. Environmental exposures were conducted on both sets of test boards with functional monitoring during the exposure. The test vehicle sets included assemblies from both Solectron and Flextronics. Each company provided four (4) groups of forty (40) test panels representing the three SAC solder alloy compositions as well as a baseline eutectic (tin/lead) solder composition. One board from each set was used for destructive metallographic analysis and not included in the thermal cycling study. The specifics relating to the thermal test events are outlined in the following paragraphs along with specifics on the test equipment, test profiles, test configurations, functional monitoring, and test schedule. Test Equipment The thermal cycling equipment incorporates the use of a BEMCO FW100 thermal chamber. This chamber is capable of cycling, when empty, from 0 C to 100 C in approximately 10 minutes. Similarly, this chamber can cycle, when empty, from -55 C to 125 C in 20 minutes. The Thermotron, model ATS 320 H 15 15, is the thermal shock chamber. This chamber provides for temperature cycling from -55 C to 125 C through physical movement of the test article within the chambers in less than one minute. The temperature stabilization would take longer and would be a function of the thermal mass of the test article. Typical stabilization times for these temperature ranges were 5 minutes. Test Approach Due to the sizing of the equipment, the payload of the thermal shock equipment was maximized with a combination of Flextronics and Solectron test vehicles. The remaining balance of test vehicles was allocated for thermal cycling. The initial estimate for this distribution consisted of approximately 24 of each vendors test vehicle subjected to thermal shock with the remaining balance of 132 of each vendor test vehicle subjected to thermal cycling. This approach allowed for 6 of each solder composition to be exposed to thermal shock with the balance subjected to thermal cycling. The size of the BEMCO chamber accommodated this large number of assemblies. The thermal cycle profile reflects the IPC test regimen and consists of a low temperature soak (0 C) for ten (10) minutes with a temperature increase ramp up to 100 C with a high temperature soak of ten (10) minutes prior to a ramp down to the low temperature. The total cycle is typically takes around sixty (60) minutes. The cycle time is a function of the chamber time to temperature and the related temperature stabilization of the test article. The thermal shock test profile is very similar to the JEDEC prescribed exposure. It consisted of a low temperature (-55 C) soak for five (5) minutes, followed by a transition to the high temp (125 C) with a high temperature soak for five (5) minutes, with a final transition back to the low temperature. This cycle would was repeated continuously. The total cycle time was approximately twenty (20) minutes. Functional monitoring was provided using Fluke NetDaq Model 2640A data acquisition units. The test provided 2- wire resistance monitoring for 700 signals based on thirtyfive (35) NetDaq units with twenty (20) channels each. In addition to the functional monitoring, metallographic analysis at every 500 thermal cycles was done at Andrew Corporation on representative samples from the two sets of tests vehicles. When comparing the results of X-ray analysis for voids, failures in thermal cycling and thermal shock and the metallographic examination of both failed and functional solder joints after thermal exposure, it is obvious that voids had little or no influence on solder joint integrity. Follow up statistical analysis, presented here, confirms that belief. 4 The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper

8 Executive Summary Section As was previously mentioned, the IPC Solder Products Value Council Lead Free Technical Subcommittee chose, because of its widespread use, the tin, silver, copper family of lead free alloys. The council assumed that although the high content silver alloys (3.8 % silver or greater) were being promoted as an alloy of choice, it appeared that the lower silver (96.5/3.0/0.5 SnAgCu commonly called SAC 305) lead free alloy would perform equally as well at lower cost. Standard tin-lead near-eutectic solder (SnPb) solder, as a part of this study, was used as a control. However, the members of the technical committee did not intend for the test program to be a head to head comparison between lead free and SnPb solder but an analysis of the SAC alloy family. The committee then, working with the appropriate company or organization, chose the testing protocol and reviewed each step of the testing program. The results of each phase of the six-phase test program can be summarized as follows: Assembly performance screening to compare alloys: No statistically significant difference was found in alloy performance when data from participating locations was compared. Experimentally the alloy properties of melt temperature (DSC), time to reach zero and maximum force in wetting balance testing and solder spread as determined by area and diameter were found to not be statistically different. In some cases a specific location found differences but when the data was averaged between locations for the same alloy no statistical difference could be found. Down selection of the solder pastes for assembly: No difference was found between alloys for the pastes tested for assembly performance. Assembly of test vehicles using SAC alloys with SnPb eutectic solder as a control: No difference was found in process ability or defect rate between the alloys as assembled at two separate test locations using two separate test vehicles. Although the materials performance was distinguishable from SnPb eutectic solder there was no difference between the lead free SAC alloys studied. Baseline metallographic analysis of the assembled test vehicles: No metallurgical difference was found between the SAC alloy solder joints after assembly and before thermal cycling. Thermal cycling testing: All three SAC alloys showed similar failure rates for similar packages. These rates were distinguishable from the behavior of SnPb solder but were not distinguishable from one another. When the data collected in this study was compared to data collected in previous studies using the NEMI 95.5/3.9/0.6 SAC alloy the results of the different studies were not distinguishable by alloy. Metallographic analysis as a function of thermal cycling: Metallographic analysis was performed at every 500 thermal cycles. Results showed there was no significant difference between SAC alloy structures with thermal cycling. Statistical analysis of the relationship between voids and solder interconnection reliability: A key by-product of this testing program was the data gathered on the much debated issue of solder joint voiding. Based on comparison of number and size of solder joint voids to thermal cycle interconnection failure data collected in this study, there is no evidence that the type of process-related solder joint voiding that was observed in the SAC alloy solder joints has any significant impact on solder joint reliability. The results of the testing done over all phases of this study indicate that SAC 305 (Sn96.5/Ag3.0/Cu0.5) should be the default alloy for use in SAC lead free applications involving reflow assembly. The presence of process related voids in the interconnections formed using the SAC alloys has been found to have no statistically significant effect on solder interconnection reliability as tested by accepted thermal cycling methods. 5 The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper

9 Void Data Summary Overview The data presented here represents the compilation of electrical failures, transmission x-ray imaging and metallographic analysis of the test vehicles described above. The test regimen lasted over 6000 thermal cycles (as per IPC-9701 test conditions.) Voiding was noted before the start of the thermal testing in both the Flextronics and Solectron assemblies. The amount of voiding in the SAC alloys was considerably greater than in the near-eutectic SnPb solder joints. In the Solectron assemblies voiding greater than 25% of the area was observed with all three lead free SAC alloy solder paste assembled boards and specifically for the 0.5mm pitch CSP84 lead free components on the boards. For the SnPb assembled boards, there was evidence of voiding but much less than 25% void area on the 0.5mm pitch SnPb assembled CSP84 components. On the Flextronics assembly, voiding greater than 25% was also observed with all three lead free SAC alloys. Devices PBGA196, C-CSP224, and wafer-level CSP8 on the assembled boards all showed voids greater than 25% on almost all packages. For both sets of assemblies no rework was performed on the devices with voids, per agreement with the IPC SPVC technical committee members. All void locations, along with other defects that were repaired, were noted with a red inspection arrow at the component location. It was hoped that thermal shock and thermal cycle testing would provide data on the correlation (if any) between the location and magnitude of the voids and attachment reliability. Shown in Figures 1-4 some typical examples of voids detected on assemblies not yet subjected to thermal cycling, i.e. the baseline metallographic analysis. Transmission X-ray images and photomicrographs of solder joints from the Flextronics solder test vehicle are shown in Figures 1, 2, and 3. These figures along with many others are from the final SPVC white paper (released separately and available from IPC). There is also less solder joint voiding in the SnPb solder joints than there is in the SAC alloy solder joints. An example of voids detected during the cross sectioning process for SAC405 alloy A is shown in Figures 3. Note that these images were made prior to any thermal cycling. Transmission x-ray imaging was performed on each component on every board that was removed each 500 cycles. Although the number of x-ray images is too large to incorporate in a report, the images did reveal that solder joint voiding was more extensive in the SAC alloy solder joints than in the SnPb solder joints. In particular the solder joint voiding, both in number as well as size, in the CSP84 package solder joints was considerably more extensive than the other area array packages. Comparison of the voiding with cross sections of temperature-cycled packages did not show any obvious correlation of voiding to interconnect failure. For example, the cross-sectioned SAC305 solder joints of the Solectron Board C11, U313 CSP84 package presented in Figure 4 shows very large voids but no indication that these voids are contributing to interconnection failure even though this package was subjected to 4500 temperature cycles. The final SPVC white paper confirms that there were enough temperature cycle induced creep-fatigue solder joint failures of the 0.8mm pitch 84 I/O CSP packages on the Solectron Pb-Free to obtain 2-parameter Weibull slope (Beta) and characteristic life (Eta) values. A Weibull plot of the failure distributions for the CSP84 packages is presented in Figure 5. The Weibull distributions show that the SAC alloy solder joints had a longer characteristic life than the SnPb solder joints (4713 to 6810 cycles for SAC alloy compared to 1595 cycles for SnPb). However, the transmission x-ray images and cross sections that were done on the nonmonitored boards every 500 cycles showed considerably more and larger voids in the SAC alloy solder joints than the SnPb solder joints. Because of this it was decided to x-ray each and every CSP84 package upon completion of the 6000 temperature cycles and attempt to correlate voiding with cycles to failure. It needs to be emphasized that although the voiding in the SAC alloy solder joints was greater than the SnPb solder joints, the voiding is believed to be due to the use of a non-optimized assembly process for the SAC alloy boards. As indicated earlier, the 6 The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper

10 Figure 1: Transmission X-Ray comparison of U1 PBGA196 showing voids bigger with Pb-free sac soldering Figure 3: X-Ray and cross section of Flextronics Board A9 U2 SAC405 assembled C-CSP224 Figure 2: Transmission X-Ray comparison of U43 WaferLevel CSP8 showing voids bigger with Pb-free sac soldering IPC SPVC defined a common reflow profile to be used for the SAC alloy assembly. The SnPb solder paste used by Flextronics and Solectron was the standard production solder paste used at each company and the reflow profile used was their standard optimized production profile and not the IPC SPVC common reflow profile. To validate that the voids in the SAC alloy solder joints were due to the use of a non-optimized assembly process, Solectron provided boards that they assembled using SAC396 with an optimized assembly process. The transmission x-ray images presented in Figure 6 and 7 compare the solder joint voiding in CSP84 packages assembled using SAC396 with an optimized assembly process voiding to CSP84 packages assembled using SAC387 with a non-optimized assembly process. Figure 4: X-Ray and cross section of Solectron Board C11 SAC305 assembled U313 CSP84 Row 12 Figure 5: Weibull 2-P Distributions for the CSP84 packages on the Solectron Boards 7 The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper

11 Figure 6: X-Ray comparison showing process effect on solder joint voids Figure 7: Another X-Ray comparison showing process effect on solder joint voids The x-rays in Figure 6 and 7 clearly show larger voids in the SAC alloy solder joints that were made using a non-optimized assembly process. The other interesting point to note in these figures is the Weibull Beta and Eta values. These statistics are based on the 24 CSP84 packages that were part of the IPC SPVC reliability test and the 60 CSP84 packages that were part of the Solectron reliability test. Although there are large voids in the SAC387 solder joints assembled using the common reflow process, there is no statistically significant difference in the characteristic life of the solder joints. In fact, the average value of the characteristic life of the solder joint with large voids is greater. To better quantify the solder joint voiding, each of the CSP84 package x-ray images was magnified and the number of solder balls with voids for each package was counted. The number of solder joints with voids greater than 25% of the PCB pad area was also counted. The magnified images presented in Figure 8 shows the voids were counted. The red numbers indicate the solder joints with voids greater than 25%. It is interesting to note that although there are approximately similar numbers of solder joints with voids in both figures, only the solder joints made using the non-optimized assembly process have voids greater than 25%. Although the CSP84 packages were monitored during temperature cycling and the number of cycles to failure for each package is known, it was impossible to remove each package at the moment it failed and cross section the solder joints to determine which solder joints failed first. However, if there is an influence of voids on failure mode this effect should be detectable graphically. For example, plots of the distribution of voids > 25% for failing and non-failing packages at 6000 cycles should be distinctly different. Figure 8: Magnifi ed X-Ray comparison showing process effect on solder joint voids 8 The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper

12 To present the voiding data in a slightly different manner Figure 10 is a summary of the voiding data showing the relation between size and number of voids in the interconnection and interconnection cycles to failure Figure 9: Distributions for comparing failed and non-failed CSP84 packages show no void effect Two data sets are shown: interconnections with voids greater than 25% area and interconnections with total voids regardless of size. As above, if voids had a significant effect the >25% voids would be expected to cluster in failures early on. However, both sets of data maintain a common scatter, typical of scatter in thermal cycling failures, across the entire span of the testing. Note the early failures at the bottom left hand area of the plot. The black circles represent the SnPb assembled CSP84 package failures. Although these failed earliest in the testing they had essentially no large voids! The total number of voids in the SAC alloy assembled CSP84 packages denoted by the blue triangles in Figure 10 are fewer than the total number of voids in the SnPb assembled packages. The SAC alloy packages cycles to failure, however, were considerably greater. Figure 10: CSP84 voiding verses cycle to failure A comparison of the distributions for failed and non-failed packages is presented in Figure 9. If the influence of voids was a negative one on interconnect reliability the two sets of distributions would diverge, i.e. fewer and smaller voids would be tracked by the red triangle distribution (across the bottom of the graph) while the blue circle distribution of interconnections with more and larger voids would climb steeply from left to right. However, both sets of distributions track each other within the expected scatter of such a plot. This would tend to imply that in this test there is no obvious effect of voids on interconnection reliability. Note that if voids impacted the attachment reliability of the packages then the occurrence of large voids/many voids would result in a high failure rate at a low number of cycles. However, the plot of size and occurrence versus cycles to failure is essentially flat within the scatter of the data for the three SAC alloys and SnPb. This implies that voids have, for this set of test conditions, no impact on attachment reliability. Neither the size nor the number of voids in a solder joint as observed in this study appears to have any effect on attachment reliability. As noted previously in this paper, the study does not discount the possibility that other types of voids not observed in the study Kirkendall voids may have an impact on solder joint reliability. 9 The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper

13 Metallographic Cross Section Results Many of the monitored packages on the Pb-Free test vehicles were cross sections after completing 6000 temperature cycling. The cross section confirmed the large voids observed in the transmission x-rays made prior to sectioning. The photomicrographs presented in Figure 20 shows a transmission x-ray and cross sectional comparison of CSP84 SnPb and SAC396 solder joints. These packages cycles. The original failure of this package occurred during temperature cycling at 3075 cycles. Although there are large voids in five of the 10 Row 10 solder joints of the SAC305 package that was assembled using the non-optimized IPC SPVC common profile, none of the Row 10 solder joints are completely cracked at the interface to the CSP84 package after the 6000 temperature cycles. The original failure on this package occurred during temperature cycling at 4196 cycles. Figure 20: Typical Cross sections of CSP84 Packagea after 6000 Temperature Cycles had been assembled using a reflow process optimized for their respective solder pastes. The photomicrographs are of the solder joints at Row 10 that is immediately under the edge of the die where the local CTE mismatch would be greatest. Also presented in Figure 20 is the transmission x-ray and cross section of SAC305 solder joints that were made using the non-optimized IPC SPVC common reflow profile. As can be seen there are large voids in many of the SAC305 solder joints. Magnified image comparisons of the individual solder joints for the three packages are presented in Figures 21a, 21b, and 21c. All of the Row 10 solder joints on the SnPb package assembled with an optimized process have completely cracked at the interface to the CSP84 package after the 6000 temperature cycles. The original failure of this package occurred during temperature cycling at 1413 cycles. Eight of the 10 (solder joint M10 is not shown for any of the three packages) Row 10 solder joints on the SAC396 package assembled with and optimized process have completely cracked at the interface to the CSP84 package after the 6000 temperature Figure 21a: CSP84 Solder Joints After 600 Cycles Figure 21b: CSP84 Solder Joints After 600 Cycles Figure 21c: CSP84 Solder Joints After 600 Cycles 10 The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper

14 Statistical Analysis of Void and Failure Data Listed in Appendix A is a table of voids and failure data. This data was used in the statistical analysis of voids and failures shown in Appendix B. In the table, packages outlined in green did not fail during the 6000 temperature cycle testing. Those packages that failed are highlighted in pink. Appendix B, Figures 11 through 19 features a graphical analysis of the void and failure data using the Minitab Software package available from SBTI Inc. Nine different methods of statistical analysis were used comparing failure rates of packages with and without voids. Summary and Conclusions Data from the IPC Solder Products Value Council reliability study on SAC alloys has been used in a comparison of voids in SAC interconnections and thermal cycles to failure. Nine separate methods of statistical analysis comparing cycles to failure looking at both voids greater than 25% of the interconnection area and total voids have been done. Absolutely no correlation between voids and failures under thermal cycling has been demonstrated. Based on comparison of the number and size of solder joint voids to interconnection failure in our thermal cycling data there is no evidence that the type of solder joint voiding observed in the SAC alloy solder joints has any significant impact on solder joint reliability. 11 The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper

15 Appendix A: Table of Voids and Failure Data At 6,000 Cycles: Green Did not fail; Pink Failure 12 The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper

16 Appendix B: Statistical Analysis of Void and Failure Data Figure 11 shows a boxplot of the cycles to failure for all CSP84 package solder joints versus number of voids greater than 25% of the area of interconnection. hand plot at the end of the line of cycles to failure versus board ID.) However, there is no correlation between total number of voids and failure boards with less than 10 Note that cycles to failure do not track the number of Figure 11: Voids Greater than 25% Area and Cycles to Failure voids. In addition it should be noted that assemblies with Sn/Pb solder paste (red boxplot) failed at approximately 1200 cycles even though they had zero voids above 25%. Figure 13: Main Effects Plot total voids failed before components with more then 39 voids. Figure 14 is a boxplot of cycles to failure for all alloys. Figure 12 shows an analysis of mean cycles to failure for all alloys. Assemblies with Sn/Pb solder paste (red square outlier) Figure 14: Cycles to Failure All Alloys Note that only SnPb stands out. No correlation between voids and cycling is apparent for SAC alloys. Figure 12: Analysis of Means of Cycles to Failure are statistically different than all the SAC alloys. However, the SAC alloys show no correlation between mean cycles to failure and number of voids greater than 25% of the interconnection area. Figure 13 is a main effects plot of cycles to failure. No correlation is shown between the number of large voids or the total number of total voids and the cycles to failure except for alloy. While SAC alloys all perform well, SnPb alloys fail earlier in cycling. (The points in the upper left 13 The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper

17 Figure 15 is a matrix plot of cycles to failure and number of voids greater than 25% of the interconnection area. Figure 17: One Way Analysis of Mean Cycles to Failure Figure 15: Matrix Plot of Cycles to Failure for Voids Greater than 25% Area There is no correlation between void size and cycle failures by SAC alloy. Figure 18 is a main effects plot of cycles to failure for all alloys. Note that SnPb is the only statistically significant stand out in effects. Voids have no effect on cycles to failure. All the Sn/Pb assemblies did however fail with very few cycles and with very few voids. Figure 18: One Way Analysis of Mean Cycles to Failure for All Alloys Figure 16: Mean Cycles to Failure All Alloys Figure 16 is a one-way analysis of mean cycles to failure for all alloys. SAC Alloy composition is not a main effect with regard to cycles to failures. Figure 19 shows a similar analysis to Figure 18. However, here all voids, including those less than 25% are examined. There is a statistically significant difference between alloys and cycles to failure but only for the comparison of SnPb to all SAC alloys, i.e. no void effect is noted for SAC while SnPb shows a greater failure rate in the absence of voids. Figure 17 is a one-way analysis of means for all alloys comparing voids greater than 25% area versus cycles to failure. There is no statically difference between SAC alloys the number of voids greater than 25% and failures. The first data point, which shows a statistical difference, is Sn/Pb solder paste. Figure 19: Main Effects Plot of Cycles to Failure for All Alloys No statistical difference between SnPb and SAC alloys concerning cycles to failure and total void count is found. 14 The Effect of Voiding in Solder Interconnections Formed from Lead Free Solder Pastes with Alloys of Tin, Silver and Copper