Newsroom
Sponsored by NEMI, NIST, NSF, and TMS
Thursday, February 15, 2001
8:30am – 4:30pm
New Orleans, LA
Report Summary
The worldwide movement in the electronics industry to implement lead-free solders has created a need for fundamental data that accurately describe the behavior of these alloys in solder joints and can be used to develop appropriate reliability models. The NEMI Lead-free Task Force and the National Institute of Standards and Technology (NIST) are working together to gather into a single database existing physical and mechanical property data that have been developed by researchers around the world. In New Orleans, February 2001, NEMI, NIST, NSF, and TMS sponsored a workshop to bring together modelers and experimentalists to define the state-of-the-art in reliability modeling and mechanical property measurements of lead-free solders and to determine the "necessary and sufficient" experimental data modelers need to predict the reliability of lead-free solder joints. Fifty-seven packaging scientists and engineers participated in the workshop with the ratio of industry/university participants of approximately 2:1. Industrial organizations included Amkor, IBM, Intel, Lucent, and Motorola; universities and federal research laboratories included Binghamton University, Michigan State University, Northwestern University, University of California - Berkeley, University of London (Greenwich), and NIST.
From the industrial point of view, the path is clear. There is a need for stress-strain curves both on joint sized and bulk specimens for the lead-free alloys, Sn-3.9Ag0.6Cu, Sn0.7Cu, Sn3.5Ag with comparisons to Sn-Pb under the same test conditions. Data are needed for strain rates of 10-1 to 10-6 sec-1, with data at the five decades of strain rate. The data should cover the whole temperature range of ·55° to 160°C with a minimum of 5 temperature intervals. Fully documented ATC tests are needed, including Weibull statistics and microstructures, as well as data on the packaging materials, such as board CTE. From a fundamental point of view, more data on the relationship between microstructural evolution and failure modes are needed for lead-free solders. These data are necessary to put the correct failure criteria into finite element and other models. Once the needed data are available, the modeling community felt that it can use it both for company specific problems and on some standard problems for lead-free solders that will help clarify the range of applicablity of their models.
The needed mechanical properties data will take 3-4 years to develop. During the workshop there was discussion of potential sources of funding for this work, including NSF and NIST/ATP. This workshop report describing the microelectronics community’s scientific needs will serve as a roadmap for research in the reliability of lead-free solders.
Introduction
The worldwide movement in the electronics industry to implement lead-free solders has created a need for fundamental data that accurately describe the behavior of these alloys in solder joints and can be used to develop appropriate reliability models. The NEMI Lead-free Task Force and the National Institute of Standards and Technology (NIST) are working together to gather into a single database existing physical and mechanical property data that have been developed by researchers around the world. As part of this activity, we have developed a preliminary list of missing high priority data for the Sn-Ag-Cu alloys selected by NEMI companies as the new standard alloys for reflow and wave soldering. Considerable data, particularly with respect to mechanical properties, is still needed.
The goal of this workshop was to bring together modelers and experimentalists to:
- define the state-of-the-art in reliability modeling and mechanical property measurements of lead-free solders;
- determine the "necessary and sufficient" experimental data modelers need;
- describe the changes needed to improve modeling and data, including failure criteria;
- develop a consensus on the best test methods for collecting needed data; and
- complete the assessment of missing high priority data.
We hope that this workshop report describing our community’s scientific needs will serve as a roadmap for research in the reliability of lead-free solders. Although NEMI is not a source of research funding, we are supporting funding of university and government labs in these areas by communicating with the appropriate funding agencies the relevance of this research to electronic industry needs and our willingness to collaborate in fulfilling such needs.
The workshop agenda and attendees list are attached. Fifty-seven people attended the workshop with the ratio of industry/university of approximately 2:1. Presentation materials can be found in the public section for Lead-Free Solders on the NEMI website.
NEMI, NSF, and NIST/ATP Overviews
Carol Handwerker (NIST), chair of the NEMI Alloy Team, began with a discussion of the purpose of the workshop: to bring together representatives from the modeling and mechanical properties areas from universities/research organizations and companies to develop and define the modeling and data base needs for the lead-free solders of greatest interest to the microelectronics community. The leading solder alloys to replace Sn-Pb eutectic as determined by the NEMI Task Force are Sn3.9Ag0.6Cu for surface mount, and Sn0.7Cu (first priority) and Sn3.5Ag (second priority) for wave soldering applications. (The compositions are in weight %; for example, Sn3.9Ag0.6Cu is 3.9%Ag, 0.6%Cu, balance Sn.) A preliminary list of missing high priority data for the lead-free alloys of interest had been collected by the Task Force and was presented. The aim of the workshop included examining whether these data are sufficient for modelers and experimentalists to fully characterize the reliability and material properties of lead-free soldered electronics products and determining the priorities in filling in the missing data. In order to complete these tasks, the speakers agreed to speak critically about the state of the art in solder joint modeling, and the importance of solder data to the usefulness of the models.
Ron Gedney (National Electronics Manufacturing Initiative) began the formal presentations with a description of the NEMI organization and the work that it is doing, as a consortium for the electronics industry, concentrating on the Electronics Manufacturing Technology above the silicon. Members include manufacturers and organizations across the supply chain that deal with issues of common interest. The NEMI Lead-Free Task Force began in early 1999. A preliminary assessment of the lead-free data was done in advance of the organizational meeting held in May 1999. At that meeting a group of potential partners developed a strategy for lead-free soldering. Their goals included demonstrating the capability to deliver lead-free products by 2001 with an "eye" to total lead-free elimination by 2004, based on legislatives pressures on the use of lead in Japan and Europe. They also decided to select one or two replacement alloys (one for reflow and one for wave); the preferred alloys would be patent-free and no more complex than ternary alloys.
Jasbir Bath (Solectron) presented an overview of the work that the NEMI lead-free task force is doing. There are four sub-groups in the NEMI Task Force: Alloy, Processing, Reliability and Components. The Alloy group headed by Carol Handwerker (NIST) and Alek Zubelewicz (Motorola) are developing a lead-free database of the alloys of interest, defining and developing "best practices" test procedures for obtaining materials properties for these alloys, organizing workshops to bring together modelers and experimentalists to prioritize and define the data that needed to be developed, and examining possible pathways for obtaining the needed, including funding sources.
The Reliability group headed by John Sohn (Lucent) is performing accelerated thermal cycling of the selected lead-free Sn3.9Ag0.6Cu and eutectic SnPb soldered boards with lead-free and tin-lead components (CSP, PBGA, CBGA, TSSOP, 2512 resistor). Thermal cycling is being undertaken using two thermal cycling regimes: 0°C to 100 °C and ·40°C to 125 °C. The Process group headed by Jasbir Bath (Solectron) has been charged with assembling the lead-free and tin-lead boards for the reliability test group and defining the process and equipment requirements in assembling lead-free (solder paste, printing, reflow, inspection, rework).
The Component group, headed by Rich Parker (Delphi-Delco), Nick Lycoudes (Motorola) and Srini Chada (Motorola), has been charged with assessing the impact of the higher lead-free reflow temperatures on the range of components and boards used in the industry and to develop and define new standards in this area. Work also includes assessing lead-free surface finishes used on boards and components.
Edwin Bradley (Motorola) and Rick Charbonneau (StorageTek) head the Lead-Free Task Force. The lead-free report on the NEMI program is due to be completed in August/September of this year. (For further information on the NEMI Lead-Free Project, go to http://www.nemi.org/ and click on APEX Presentations and then lead-free forum.)
Cindy Murphy (University of Texas-Austin) gave an overview of the NSF’s (National Science Foundation) recent assessment of environmentally benign manufacturing worldwide. Dr. Murphy was the member of the review committee led by Delcie Durham of NSF. This 18-month-long study of the life cycle of products established baselines and identified research opportunities in a number of markets, including electronic manufacturing. In general, they found that the U.S. is focused on materials and processing and also on toxic release legislation, the EU (European Union) is focused on products (take-back legislation and landfill minimization), and Japan is focused on applications and markets. Japan is extremely focused on ISO14000, with its implication for increased ‘green’ manufacturing, recycling and reduction in the use of energy and material resources. Japan’s applied research program is far ahead of the US and Europe. Europe is the leader in post-consumer recycling and has the best co-operative efforts between government and industry. In general the report found that the electronics industry is a leader in Life Cycle Analysis, Design for the Environment.
Japan is using ISO14000 to go halide-free and lead-free in electronics assembly. Computers are likely to follow in terms of recycling for Japan. There will be no incineration of products. Companies surveyed in Japan included Sony and Hitachi.
In Europe, the emphasis is on product take-back, lead-elimination and non-brominated PCBs. Companies surveyed in Europe included MIREC and Siemens. In the Netherlands, there is a well-developed infrastructure for recycling. Green products are considered to cost more. The Netherlands does not have the economic dominance in Europe to impose its recycling structure on the rest of Europe.
In the US, companies were seen to be responding to the lead-free push, but conversion will be dependent on the customer. The emphasis appears to be on take-back, lead-free solder and non-brominated flame-retardants. There is more of an emphasis on recycling of plastics rather than incineration. EIA (Electronics Industry Association) is trying to get a unified response from the electronics industry to the WEEE directive in Europe. The DoD, DOC (Department of Commerce) and EPA were looking into recycling of electronics. There has been and will be in the future funding available from NSF for environmental projects. The deadline for proposals is in February and October of each year. An example of NSF funding for lead-free solders was for K.N. Subramanian of Michigan State in the area of modeling of lead-free solders. The current NSF projects in this area can be found at http://www.fastlane.nsf.gov/a6/A6AwardSearch.htm
Carol Handwerker (NIST) presented an overview of the NIST Advanced Technology Program (ATP) which funds high risk, high potential payoff research programs that are of benefit to the industry. An example of a successful program was in the area of "Springback Predictability in Sheet Metal Forming"; this program for modeling of deformation during car body processing brought together OEMs (Ford, GM and Chrysler), materials suppliers (US Steel and Alcoa), parts suppliers to the auto industry, a software company, and universities to develop software (and data) for die design. This software has the potential of eliminating costly trial and error testing of dies, particularly for high strength steels and aluminum alloys. The Advanced Technology Program provides matching funding for single company projects as well as for joint ventures. (Program information can be found at: http://www.atp.nist.gov/) Under current rules, single company awards can be for as much as $2M total over up to three years; there is no funding limit for joint ventures. In the Electronics and Photonics division-funding category there had been $325 million of funding for projects in the last 5-10 years. The contact names for ATP/NIST funding are Clare Allocca (clare.allocca@nist.gov, 301-975-4359) and Michael Schen (michael.schen@nist.gov, 301-975- 6741).
Critical Assessments of the State of the Art in Solder Joint Modeling and of Data needed for such Models
Ahmer Syed (Amkor Technology) gave an overview of types of reliability models and data needs. He presented overviews of 4 topics: failure mechanisms related to the solder joint, lifetime prediction, lessons learned from tin-lead solder joint prediction and the data needs for lead-free solders. He said that causes of failures can be divided into four categories: thermal/power cycling, bending and vibration, shock and drop, and ball shear. There were more failures due to cyclic bend failure than as a result of Accelerated Temperature Cycling (ATC) and Power cycling for portable electronics applications. This was related to displacement between the component and the printed circuit board. Up through the 1980’s, most solder joint tin-lead reliability prediction models were based on Coffin-Manson relations, some of which were revised to include other effects. There was very little data obtained on real tin-lead solder joints for these models. Mechanical property data was obtained from bulk solder samples. There was no consensus on stress versus strain effects and no real data on time versus temperature behavior.
More recently, energy based models have been developed to include time, temperature and stress effects. These models are based on accurate mechanical property data. (See Shi, et al, JEP 1999). This includes determination of stress (MPa) versus strain curves with strain rates of 2.78E-1 to 2.78E-5 (5 strain rates used in this range), Young’s modulus (MPa) versus temperature graphs with the same 5 strain rates and with temperatures from ·50 °C to 150 °C (5 temperatures used in this range), ductility (Elongation-%) and strength (UTS-Ultimate Tensile Strength-MPa) versus temperature graphs at these five strain rates and temperatures. Darveaux used energy-based models with evaluation of crack growth and initiation, strain rate effect and finite element analysis. Syed used partition creep strain based prediction models with +/- 25 % prediction accuracy of mean cycle to solder joint failure. More data need to be developed using real solder joints such as the mechanical test fixture for creep testing of real joints developed by Darveaux, et al. (ref. John Lau edited book on Ball Grid Array Technology: McGraw Hill).
More failure data are being reported now, which permits broader testing of the models. There is a requirement for more real test fatigue test data from thermomechanical fatigue testing. This includes the reporting of the different cycling conditions, the test board variables, component design variables such as die thickness which affects the reliability. Lifetime prediction models require material behavior determinations, failure definition and failure data and model validation for Sn-Pb and the lead-free solders. Isothermal fatigue data is not useful for lifetime prediction model development.
Sn-3.5Ag is the most studied of the lead-free alloys, but there is still much less data than Sn-37Pb. New data is appearing, such as by Xiao, et al. (2000 International Symposium on Advanced Packaging Materials p145-151). This includes strength and ductility data on Sn3.5Ag, Sn0.7Cu and Sn4Ag0.5Cu . He feels that currently used modeling tools and methodologies are sufficient to model the reliability of lead-free assemblies, but that constitutive equations for these lead-free alloys must be developed. The lead-free alloys SnCu, SnAgCu and SnAg have similar homologous temperatures (T(experiment))/T(liquidus temperature) in K) as Sn-Pb in the ·40°C to 125°C temperature cycling range which indicates that time and temperature dependent deformation mechanisms play a key role for Pb free solders as well. The behavior of a particular solder may be different, however, depending upon its deformation mechanisms and creep resistance. For example, the data from the NCMS high temperature lead-free project indicates that SnAgCu alloys are probably more creep resistant at low stresses and less so at higher stresses.
The ATC results for PBGA and flexBGA packages with Sn/Pb and lead-free solder joints were presented. Three slightly different compositions of Sn/Ag/Cu alloy (Sn3.4Ag0.7Cu, Sn4Ag0.5Cu and Sn4Ag1Cu) were evaluated and generally the SnAgCu alloys showed similar behavior with no significant differences in reliability on OSP coated boards. The tests indicated higher acceleration factors for Pb-free SnAgCu alloy over a wider temperature cycling range. The acceleration factor for eutectic Sn/Pb was reduced by 50 % (2X) when cycling from 0/100 °C to ·40/125 °C, while the acceleration factor for Sn/4Ag/0.5Cu was reduced by 67 % (3.15X) when cycling from 0/100 °C to ·40/125 °C. This implied that Pb-free alloys (SnAgCu) have a significantly higher reliability at field level conditions (0/100°C) compared with Sn-Pb.
The focus for lifetime prediction should be on the study of time and temperature dependent material behavior and developing data on realistic solder joint samples. Creep deformation will still play a large role for temperature cycling failures whereas time independent plasticity would be more relevant for vibration and other high cycle fatigue simulation. Thermal cycling data on real lead-free components should be published and the data should not be normalized. The following discussion brought out several interesting questions: How do we handle the transition from thick joints (where bulk solder properties dominate) to thin joints (where the interfaces dominate)? When is performance determined by the bulk solder and when is it determined by the solder joint?
Leon Keer (Northwestern University) overviewed constitutive and damage modeling (Fracture Mechanics approach · Crack Initiation and Crack Growth technique) and reiterated a need for data similar to what was in the NEMI guide to modelers. He described work related to modeling of failure in Sn3.5Ag solder in which microcracks developed into a macrocrack based on damage accumulation. Thermomechanical cycling was undertaken from ·25 °C to 80 °C for the joint. The CINDAS database at Northwestern had compiled some lead-free data but not enough. There was found to be some correlation between bulk and solder joint properties. There was a need for the definition of fatigue failure to be standardized and an understanding of the failure mechanism. Test conditions needed to be used which duplicated field conditions.
Chris Bailey (University of Greenwich, London) followed with a presentation on the interaction between manufacturing and reliability models. He is trying to integrate models that take devices and board layout to calculate solder shapes and stresses that develop as the solder solidifies, with models that consider the effects of voids, microstructure, and damage on joint reliability. He emphasized the importance of considering surface tension, in determining whether the solder will wick away from the joint or form a shape (such as a neck) that has poor fatigue properties. They have found that void formation and movement in the liquid state is quite complex. Voids can originate from the paste, gaps at the vias, and outgassing of the substrate. The void movement is governed by both buoyancy forces and by Marangoni flow. The solder must then solidify to form a joint. Solidification models are available that can be applied to solder, most notably the ones developed for modeling the processing of single crystal superalloy turbine blades for engine applications. Further development of such models would permit a better understanding of microstructural evolution during solder joint solidification, and subsequent thermomechanical cycling. One area of particular interest is segregation of constituents within the solder alloy during solidification; such segregation may lead to crack formation and fillet lifting.
He described work on the modeling of 2512-chip resistor with SnAgCu versus SnPb solder done in collaboration with NPL (National Physical Laboratory) in the UK. There are constitutive equations for SnPb but none for SnAgCu. Constitutive equations for Sn3.5Ag from work by Darveaux were used in the modeling for SnAgCu. Crack initiation originated under the pad. It was calculated that the Sn3.5Ag had 10 % more damage energy accumulation than SnPb. Increasing the standoff height increased the lifetime of the joint. Models were found to identify trends which would be most applicable and beneficial in the early design stages of products.
Darrel Frear (Motorola) described his view of the materials issues (with an emphasis on microstructure) for developing reliability data. He feels that microstructure is often overlooked and recommended some different etches for lead-free alloys, specifically: 10 % HCl in methanol for 1-2 seconds to reveal the bulk microstructure and 4 parts glycerol, 1 part acetic and 1 part nitric acid for the intermetallic phases. Microstructure is an important variable in life-cycle prediction, because initially coarser structures will have shorter lives. The SnAgCu is predominantly a tin-rich matrix and there is less potential for microstructural evolution than with tin-lead.
He showed some SEM images of polished then etched solder joints, which revealed a 3-D view of the roughness of the intermetallic interfaces. He emphasized the importance of the intermetallic layer thickness in determining the life as the solder joint ages. With lead-free solders, there were nodules of intermetallic growth which were difficult to measure. With Sn0.7Cu and SnAgCu solders, there was SnCuNi intermetallic growth on Nickel-based Under Bump Metallisation (UBM). There was increased nickel metallization consumption but not significant compared with tin-lead. In the solid state there were lower intermetallic growth rates with SnAgCu solder on Cu substrates compared with SnPb solder. In the liquid state the intermetallic growth rate on copper substrates was greater for SnAgCu compared with tin-lead solder.
Solder joints must withstand shock conditions with the weakest link being the interfacial intermetallic layer. Fracture toughness tests showed that there was failure through the solder for tin-lead and pure tin whereas for Sn3.5Ag there was interfacial failure. There were different types of tests described to characterize the mechanical properties of the solder but regardless of the test the microstructure needed to be characterized in order to understand behavior of the joint. Of all the various properties, he suggested that creep data are the most important for modeling. Sn/Ag alloys had slower creep rate than Sn/Pb alloys. SnAgCu had less microstructural evolution than tin-lead during thermomechanical fatigue testing and tended to crack at the higher strain rates.
There was a preference for double-sided lap shear tests (double sided to avoid bending) over indentation tests. Some interesting images of electromigration were shown across lead-tin solder joints (150 °C) under high power cycling applications. Lead-free electromigration data was lacking. He also wants a validated database of material property data. Creep data over the range of temperatures would be most useful. Joint geometry would be part of the characterization, and it was better to do joint testing than bulk testing. The constitutive relations need to be microstructure-based. Another caution is the need for good temperature control during thermal cycling and during creep test: cycling out of desired temperature range creates inappropriate damage in the solder joint. For automotive applications ·40 °C to 150 °C was the temperature range used, but this was being increased in the industry to ·40 °C to 170 °C.
Alek Zubelewicz (Motorola) followed with a view of the modeling issues. He feels that it would be dangerous to extrapolate the behavior of tin-lead to SnAgCu. This, therefore, creates the need for materials property evaluations of the new solders, which could then be used in modeling predictions of the lifetime of the solder joints. FEM (Finite Element Modeling) produces excellent data but often was too cumbersome, too complex to interpret. He emphasized the need for simpler constitutive models that accurately represent the material behavior with the least number of parameters, so that it could be easily implemented in any commercial finite element software. He supplements FEM with various simple models, which can incorporate damage criteria. There is a need for good constitutive equations for the lead-free solders with proper damage criteria and accurate statistical analysis.
He prefers simpler models that produce slightly lower accuracy (about 80% accuracy) to speed up the modeling prediction process. There are fewer parameters involved with a better quality of fit to the experimental data. There is a requirement to understand the mechanisms of crack initiation and propagation for lead-free. There will be a requirement for fracture toughness testing in addition to obtaining thermomechanical fatigue data. For modeling, bulk and joint solder data can be used but we have to know how to scale the data.
For lead-free SnAgCu solder there is a need to develop the constitutive relationships and to establish the damage criteria and acceleration factors for fatigue. He showed that the acceleration factor models, especially at low strain rates, appear to greatly underestimate the fatigue life for Sn-Pb eutectic solders. It has recently been shown by IBM (Apex 2001, Bartello, et al.) that during ATC at very high temperatures lead-free solders seem to behave much worse than lead-based solders. The implication is that without knowing the relationship between acceleration factors and reliability, we don’t know how to correlate test results with true field conditions. The industry requires a validation of reliability requirements for various applications and an improved methodology for prediction of acceleration factors. It will take 2-3 years of work to obtain the information necessary which can be used to predict by modeling lead-free solder joint reliability.
Discussion Session on Modeling and Data Base Needs
There was agreement on defining two priority lists for data base needs, one for industrial modeling applications and one for fundamental studies, because the industrial and research attendees had differing priorities.
For industrial modeling applications, data on CTE (coefficient of thermal expansion) and on the volume change on freezing were raised from medium to high priority. Development of CTE data for the lead-free alloys is complicated because the CTE varies with crystallographic orientation. We need to determine these two properties in both the liquid- and solid-state ranges. We need Young’s modulus data as a function of microstructure and orientation and raised its rating from low to high. Data on the wetting/solderability and surface tension of the lead-free solders were moved from low to medium priority. The priority of the materials properties for lead-free for industrial modeling are shown in Appendix A. We also need mechanical property data for both the bulk and joint configurations. For a joint configuration, characterization of the surface finish, substrate metal and joint thickness would be necessary.
Most of the mechanical data cited as being "high priority" come from a series of isothermal stress-strain curves. There was a need for stress-strain curves for strain rates of 10-1 to 10-6 sec-1, with data at the five decades of strain rate. The data should cover the whole temperature range of ·55 ° to 160 °C (a minimum of 5 temperature intervals). Typical testing temperatures could be ·55 °C, ·40 °C, 0 °C, 25 °C, 70 °C, 125 °C, 160 °C. In both laboratory testing and ATC of assemblies, stress-strain behavior is dependent on the microstructure. The initial and final solder microstructures should be determined and the results compared with different microstructures seen in solder joints. The specimens used to determine stress-strain behavior should be aged for 16 hours at 125°C followed by 24 hours at room temperature for stress relief before performing tests to develop a more common starting microstructure.
Some felt that we should specify two different thermal cycles for producing some of the specimens, so that the data can bracket specific times at temperature. As a starting point for the characterization of SnAgCu solder, the Darveaux work on modeling and characterization of tin-lead should be treated as the baseline work and for which to use to compare with SnAgCu.
Once these creep data are obtained and solder joint reliability modeled there must be good quality ATC data available to compare with the model predictions. For the ATC fatigue data to be useful for comparison with modeling, considerable information must be available, including temperature cycling conditions measured at the broad level (complete profile included), component and PCB descriptions, and the non-solder materials used together with their CTE and Young’s modulus values. The layout and dimensions of the board and components and the dimensions of the solder joints would need to be included. The test data should include the raw data (or access to it), the Weibull plots with · and · values given, the sample size used, the location of the failure on the board, the identification of crack location, and the microstructural evolution from SEM views of the joint and optical and SEM cross-sections.
For fundamental studies, we need more data on the microstructure evolution (both in the bulk solder and at interface, including phase transformations), starting with the as-produced specimen (just after reflow) and for various aging conditions/cycles. We need to understand microstructural changes induced by thermomechanical stress and also those that could be related to tin pest for the high tin lead-free solders. We would then need to relate all of these parameters to the mechanical properties (stress versus strain curves). We need to learn which microstructural features are the most important, so we can add only these important ones to the property models. We need micromechanical characterization and some deformation characterization to know how structures fail. This will provide some data on damage accumulation. We need fracture toughness testing at various temperature ranges, deformations, and thermomechanical cycling ranges (to insert into the damage models). In-situ testing should be carried out as a function of time with techniques such as Moire.
Some additional thoughts were: How do we handle the transition from thick joints (where bulk solder properties dominate) to thin joints (where the interfaces dominate)? Also we need to consider emphasizing the modeling of the effect of voids and microstructure on solder joint life.
In terms of modeling, there is still disagreement over how accurate finite element models can be. The sources of the disagreement are: (1) joints display wearout with a certain distribution of failure times, in ATC and in the field, so that there is not a single time to failure for all joints of a given configuration and (2) the final joint performance is sensitive to the local environment (for example, standoff height, planarity, component alignment), including many parameters that cannot be controlled. This is the difference between calculations of solder joint reliability and larger structural components where no failure is acceptable. Finite element analyses for solders, lead-containing and lead-free, are therefore frequently used to look at trends in reliability rather than absolute numbers. Others stated that they could calculate times-to-failure for new joint shapes with reasonable accuracy.
There was general agreement that the data described above for lead-free solders are critical to success for all the modelers, regardless of their approaches. Once the highest quality data possible are available, then we have the possibility of doing some quantitative comparisons using different modeling approaches. This "round robin" comparison of calculations has been very useful in other disciplines. In the area of the magnetic behavior of small elements for magnetic data storage, the modeling community led by NIST did a "round-robin" calculation for a problem for which everyone agreed on the physics and on the equations for behavior. There were eight anonymously submitted solutions and they were all different. They have now examined the conditions under which all their solutions agree, and in doing so have learned how to improve their proprietary codes. Once the mechanical property data are available, NIST is willing to host such a group if the modeling community is interested in such comparisons.
Conclusions
From the industrial point of view, the path is clear. The needed data are:
- stress-strain curves as described above, both on joint sized specimens and bulk specimens
- acceleration factors for lead-free versus tin-lead solders
- fully documented ATC tests, including Weibull statistics and damage and failure criteria
Microstructure characterization of the joint/ bulk specimens is essential before and after testing with the role the initial and evolving microstructure plays on the material properties. In order to produce a more common starting microstructure prior to testing joint and bulk sized specimens, aging conditions of 16 hours at 125°C for Sn37Pb, Sn3.9Ag0.6Cu, Sn0.7Cu and Sn3.5Ag solders are suggested followed by 24 hours at room temperature for stress relief before performing tests. This is referred to in the NEMI lead-free test procedure document. From a fundamental point of view, more data on failure modes are needed for lead-free solders. These data are necessary to put the correct failure criteria into finite element and other models. As part of fundamental and ATC studies, microstructural characterization is needed for the solder joint and at the intermetallic, with other factors such as voiding noted since thaty have an effect on reliability.
ATC data needs to include materials properties (CTE, Modulus) of the solder, board and component and model parameters of the component (complete geometry of package; for example footprint details, package dimensions (overmold size), die size and die thickness, substrate size, die attach thickness, solder mask thickness (board and component) etc). This data will be input into their software by modellers (ref. Kim, Syed, et al, ECTC 1998, p680-684). Failure mode information is also needed. For a good analysis of crack initiation and growth during ATC testing refer to the Darveaux work (ref. ECTC 2000).
Once the needed data are available, the modeling community can use it both for company specific problems and on some standard problems for lead-free solders that will help clarify the range of applicability of their models. The first problems will involve modeling the ATC tests for which all the necessary modeling data exist.
Note: Attendees will receive soft copies of the minutes of the workshop. The minutes and soft copy presentations will be available on the NEMI website.
The list of attendees with email addresses is given in Appendix B, below:
For additional information on the NIST Solder Program, see: http://www.metallurgy.nist.gov/techactv1999/AnnualReport1999.html#packaging
Appendix A — Industrial modeling list of priorities of materials properties
|
Sn3.9Ag0.6Cu |
Sn0.7Cu |
Sn3.5Ag |
Sn37Pb |
| CTE (liq. and solid state) |
1 |
1 |
1 |
1 |
| Vol. Change on freezing (liq.and solid state) |
1 |
1 |
1 |
1 |
| Specific Heat |
3 |
3 |
3 |
3 |
| Latent Heat |
3 |
3 |
3 |
3 |
| Thermal Diffusivity |
3 |
3 |
3 |
3 |
| Thermal Conductivity |
3 |
3 |
3 |
3 |
| Electrical Conductivity |
3 |
3 |
3 |
3 |
| Electrical Resistivity |
3 |
3 |
3 |
3 |
| Surface Tension at temp of solder |
2 |
2 |
2 |
2 |
| Wetting Time at O Force as f (Temp) solder |
2 |
2 |
2 |
2 |
| Wetting Time at 2/3 Force as f (Temp) solder |
2 |
2 |
2 |
2 |
| Max. Wetting Force as f (Temp) solder |
2 |
2 |
2 |
2 |
| UTS at 25°C |
1 |
1 |
1 |
1 |
| Shear Strength (strain rates from 10-1 to 10-6 s-1) |
1 |
1 |
1 |
1 |
| Ring in Plug (strain rates from 10-1 to 10-6 s-1) |
3 |
3 |
3 |
3 |
| E (Young’s modulus) at 25°C |
1 |
1 |
1 |
1 |
| E at 50°C |
1 |
1 |
1 |
1 |
| E at 100°C |
1 |
1 |
1 |
1 |
| E at 125°C |
1 |
1 |
1 |
1 |
| Total Elongation (strain rates from 10-1 to 10-6 s-1) |
1 |
1 |
1 |
1 |
|
|
|
|
|
|
Sn3.9Ag0.6Cu |
Sn0.7Cu |
Sn3.5Ag |
Sn37Pb |
| Uniform Elongation at R° (strain rates from 10-1 to 10-6 s-1) |
1 |
1 |
1 |
1 |
| Yield Strength at R° (strain rates from 10-1 to 10-6 s-1) |
1 |
1 |
1 |
1 |
| Work Hardening Coefficient (strain rates from 10-1 to 10-6 s-1) |
1 |
1 |
1 |
1 |
| Creep Resistance(strain rates from 10-1 to 10-6 s-1) |
1 |
1 |
1 |
1 |
| Min. Creep Strain rate at Stress of 20Mpa at R° |
1 |
1 |
1 |
1 |
| Min. Creep Strain rate at Stress of 20Mpa at 125°C |
1 |
1 |
1 |
1 |
| Hardness |
3 |
3 |
3 |
3 |
| Thermomechanical Fatigue resistance(strain rates from 10-1 to 10-6 s-1) |
1 |
1 |
1 |
1 |
| Isothermal Fatigue Data (strain rates from 10-1 to 10-6 s-1) |
1 |
1 |
1 |
1 |
| Thermal Fatigue Hysteresis behaviour(strain rates from 10-1 to 10-6 s-1) |
1 |
1 |
1 |
1 |
| Constitutive Behavior(strain rates from 10-1 to 10-6 s-1) |
1 |
1 |
1 |
1 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Sn3.9Ag0.6Cu |
Sn0.7Cu |
Sn3.5Ag |
Sn37Pb |
| Stress Rupture(strain rates from 10-1 to 10-6 s-1) |
3 |
3 |
3 |
3 |
| Dynamic Acoustic Measurements |
3 |
3 |
3 |
3 |
| Fracture Toughness at R° (strain rates from 10-1 to 10-6 s-1) |
3 |
3 |
3 |
3 |
|
|
|
|
|
1=High Priority, 2=medium priority, 3=low priority
Appendix B — NEMI Workshop Modeling and Data Needs Lead Free Attendees
| Speaker |
Company Name |
Phone Number |
E-Mail Address |
| Carol Handwerker |
NIST |
301-975-6158 |
carol.handwerker@nist.gov |
| Ron Gedney |
NEMI |
703-834-2084 |
rgedney@nemi.org |
| Jasbir Bath |
Solectron |
408-957-2935 |
jasbirbath@ca.slr.com |
| Cynthia Murphy |
University of Texas – Austin |
512-475-6259 |
cfmurphy@mail.utexas.edu |
| Ahmer Syed |
Amkor |
480-821-2408 ext. 5302 |
asyed@amkor.com |
| Leon Keer |
Northwestern University |
847-491-4046 |
l-keer@northwester.edu |
| Chris Bailey |
University of Greenwich, London |
44-208-331-8660 |
c.bailey@gre.ac.uk |
| Darrel Frear |
Motorola |
480-413-6655 |
darrel.frear@motorola.com |
| Alek Zublewicz |
Motorola |
512-895-7667 |
alek.zubelewicz@motorola.com |
|
|
|
|
| Attendees: |
Company Name |
Phone Number |
E-Mail Address |
| Denny Aeschliman |
3M |
|
daeschliman@mmm.com |
| David Godlewski |
NEMI |
717-651-0522 |
dgodlewski@nemi.org |
| Iver E. Anderson |
Ames Laboratory |
|
andersoi@ameslab.gov |
| Ben Huang |
Indium |
315-853-4900 |
bhuang@indium.com |
| Dr. Sammy G. Shina, |
UMass Lowell - Mechanical Engineering |
978-934-2590 |
sammy_shina@uml.edu |
| Dan Nardone. |
FCI |
717-938-7168 |
dnardone@fciconnect.com |
| Lee Patch |
NCMS |
734-995-4972 |
leep@ncms.org |
| Srinivas Chada - |
Motorola |
954-723-5293 |
srinivas.chada@motorola.com |
| Dr. W. Kinzy Jones, Jr. |
Motorola |
|
Kinzy.Jones@motorola.com |
| Chuck Woychik |
IBM |
607-755-9506 |
woychik@us.ibm.com |
| George Thiel |
IBM |
607-757-1064 |
thiel@us.ibm.com |
| Mike Griffen, |
IBM |
845-892-2300 |
|
| Kathleen A. Stalter |
IBM |
|
|
| Tom Siewert |
NIST |
303-497-3523 |
thomas.siewert@nist.gov |
| Sundar Sethuraman |
Solectron |
408-956-6545 |
SundarSethuraman@ca.slr.com |
| Valeska Schroeder |
HP |
|
|
| Attendees: |
Company Name |
Phone Number |
E-Mail Address |
| Phil Geng |
Intel |
503-696-5532 |
phil.geng@intel.com |
| Fay Hua |
Intel |
|
Fay.hua@intel.com |
| Ken Kinsman |
Formerly Intel |
480-820-2296 |
romwall@aol.com |
| Willey Desaulnier |
|
860-745-6243 |
willydee@att.net |
| K.N. Subramanian |
EGR Michigan State University |
517-353-5397 |
subraman@egr.msu.edu |
| William (Bill) Plumbridge |
Open University |
00-44-908-652630 |
w.plumbridge@open.ac.uk |
| Yoshiharu Kariya |
Open University |
|
y.kariya@open.ac.uk |
| David Suraski, |
AIM |
401-463-5605 |
dsuraski@aimsolder.com |
| Eric Cotts, |
SUNY Binghamton |
607-777-4371 |
ecotts@binghamton.edu |
| Julie Stein |
University of California -SMART |
510-642-1896 |
jstein@uclink.berkeley.edu |
| John Manock, |
Lucent |
|
jmanock@lucent.com |
| Peter Biocca |
Loctite/Multicore |
972-238-1224 ext. 114 |
Peter.Biocca@multicore.com |
| Prof Yong-Ho Kim |
(on leave from Hanyang Univ., Korea) represent University California, Berkeley |
510-486-4836 |
yhkim@lbl.gov |
| H.G. Song |
(PhD cand, University of California, Berkeley) |
|
Same as above |
| Jeff Suhling |
Auburn University Center for Advanced Vehicle Electronics |
334-844-3332 |
jsuhling@Eng.Auburn.EDU |
| Jenq-Dah Wu (J-D Wu),. |
Advanced Semiconductor Engineering, Inc |
|
jd_wu@asek.asetwn.com.tw |
| Shao-Ping Chan |
LANL |
505-667-7346 |
sc@lanl.gov |
| Steve Kilpatrick |
IBM - Fishkill |
845-894-9482 |
skilpatr@us.ibm.com |
| Jim Lucas |
Michigan State University |
517-432-2883 |
lucas@egr.msu.edu |
| Hareesh Mavoori |
Lucent |
908-582-2558 |
hareesh@lucent.com |
| Alan Gickler |
Johnson Manufacturing |
319-289-5123 |
agickler@aol.com |
| Syamal Lahiri |
|
|
|
| C. Robert Kao |
Department of Chemical Engineering National Central University, changi: Ciay, Taiwan |
|
crkao@ncu.edu.tw |
| Polina Snugovsky |
Celestica |
416-844-5016 |
polina@celestica.com |
| Leonid Snugovsky |
University of Toronto |
416-978-0606 |
Leonid.snugovsky@utoronto.ca |
| Michael Notis |
Lehigh University |
610-758-4225 |
Mrn1@lehigh.edu |
| Bruce Cook |
Ames Lab |
515-294-9673 |
cook@ameslab.gov |
| Sukyung Ko |
Agere Systems |
610-391-3822 |
suko@agere.com |
| Charan. G |
IBM |
607-755-9653 |
charan@us.ibm.com |
| Clare McCarthy |
IBM |
845-894-2314 |
clarem@us.ibm.com |
| Zequn Mei |
Agilent Technologies |
650-485-3497 |
Zequn_mei@agilent.com |
| Jianku Shang |
Univerisity of Illinois |
217-333-9268 |
j-shang@uiuc.edu |