NCMS Lead-free Solder Project

Estimated reading time: 15 minutes

This article summarizes the results, conclusions and recommendations of a four-year-long study of alternative solder materials.

By the National Center for Manufacturing Sciences Lead-free Solder Project

In 1997, the National Center for Manufacturing Sciences (NCMS) Lead-free Solder Project, carried out by a consortium of 11 industrial corporations, academic institutions and national laboratories, completed its four-year program to identify and evaluate alternatives to eutectic tin/lead solder. The goal of the project was to determine whether safe, reliable, nontoxic and cost-effective substitutes exist for lead-bearing solders in electronics manufacturing. Finding a lead alternative requires gaining knowledge equivalent to the knowledge base already in existence for lead-bearing solders. The production of durable, reliable, safe and affordable electronic products with lead-free solders requires the manufacturer to understand material properties, manufacturing processes and equipment, toxicological effects, alloy cost and long-term availability, and reliability. This article summarizes the major results of this multi-year research program, including manufacturing and reliability trials, and studies of the root causes of "fillet lifting" in plated through-hole joints.

Introduction

This project was initiated in response to potential national and international legislation banning or restricting lead use in microelectronics. Given the competitiveness of the industry and the fact that all equipment and manufacturing requirements have been based on the properties of eutectic tin/lead solder, the costs of conversion to a lead-free solder and its associated effect on reliability must be clearly understood. The results of this study should help the industry make informed decisions on the alternative alloys, manufacturing processes and infrastructure changes necessary to convert to lead-free solders.

A comprehensive report was issued1 that summarized the major results of this multi-year research program. The results include:

- Recommendation of lead-free solders for certain applications.

- Practical guidelines for alloy composition limits based on expected cost and continued availability of constituents.

- Comprehensive database on seven down-selected lead-free alloys and comparable data for Sn/Pb eutectic solder.

- Limited materials property data for more than 70 additional candidate alloys.

- Manufacturing process conditions adequate for assembly soldering the various test vehicles. These data provide valuable insight for any further process development work.

- Interactions between various printed circuit board (PCB) and surface finishes and selected solder alloy compositions, as determined by metallographic evaluation and component shear force testing.

- End-of-life prediction software for four selected lead-free alloys and one Sn/Pb eutectic alloy, based on thermal cycling test results.

- Assessment of the toxicity of possible replacement elements for lead-free solder alloys.

A long-term benefit of this project to the electronics industry is the resulting materials database for candidate replacements of lead-bearing solders, including information regarding electronics manufacturing, product performance and reliability. In four years, this project has established a lead-free solder alloy database, particularly with respect to reliability, which is a complement to eutectic and near-eutectic lead-containing solders.

Scope of Project

Selecting suitable alternatives to tin/lead solders required careful evaluation of candidate alloys from many different perspectives (Figure 1). First, project participants identified three storage and operating temperature ranges for a wide range of applications of interest to the participants:

- -55° to 100°C for consumer electronics and telecommunications

- -55° to 125°C for military electronics

- -55° to 180°C for aerospace and automotive electronics.

From these storage and operating temperature ranges and typical assembly process temperatures, material property limits were established for candidate solder alloys.

Summary of Results

The results of these first four screenings are as follows:

Toxicology - Cadmium is toxic in many forms and is severely restricted in Europe; therefore, it was eliminated as a potential alloying element. The data on all other elements were incomplete and sometimes contradictory. Given the uncertainty in the toxicology of other possible alloying elements, no other elements were eliminated. The elements remaining for consideration were: Sn, Ag, Bi, In, Cu, Sb (as a minor constituent) and Zn. Small additions (less than 1 percent) of other common elements were also not believed to pose a risk to environmental or occupational health.

Economics and availability - To ensure that an adequate worldwide supply of a potential alloying element exists, and that the costs of an alloy using this element not exceed $10.00 per lb, it was found that a potential lead-free solder alloy should contain no more than 1.5 percent In or 20 percent Bi. Eutectic Sn/Bi was retained as a baseline alloy. The material cost increase alone associated with converting to one of the remaining lead-free alloys was estimated to be $140,000,000 to $900,000,000, depending on the alloy composition. The cost of solder paste for In-free alloys is expected to increase on the order of 5 percent over eutectic Sn/Pb because the cost of the metal is a small part of the cost of the paste. On the other hand, the cost of bar stock for wave soldering is expected to be a strong function of composition, even for In-free alloys.

Manufacturing- and reliability-related properties - Critical pass/fail down-selection criteria were established based on liquidus temperature, pasty range (the difference between liquidus and solidus temperature), wettability and mechanical properties (Table 1).

The down-selection criteria of liquidus temperature less than 225°C and a pasty range of less than 30°C severely limits the range of compositions available for use. This is demonstrated using the Sn/Ag/Bi phase diagram in Figure 2. The shaded regions indicate where the pasty range is less than 30°C, which are close to the binary and ternary eutectic temperatures. The limit of 225°C for the liquidus temperature further restricts the range of compositions. This means that compositions will be close to binary or ternary eutectics.

For the 10 lead-free alloys passing these pass/fail tests, the properties of these alloys were compared with eutectic Sn/Pb using a quantitative decision matrix that included liquidus temperature, pasty range, wettability and thermomechanical fatigue resistance. As a result of the down-selection process and further deliberations, seven alloys (plus the baseline eutectic Sn/Pb) of the original 79 alloys were carried into full-scale manufacturing and reliability assessments; the alloy compositions and liquidus temperatures are listed in Table 2.

The Sn/Pb eutectic was included as a baseline alloy. Sn/Ag eutectic was included as a baseline alloy and a candidate alloy. The Sn/Bi eutectic was also included as a baseline alloy and as a candidate alloy even though it failed the availability pass/fail criterion. The project participants were concerned that the other lead-free candidate alloys had liquidus temperatures in excess of 210°C, too high for some assemblies currently being manufactured. The Sn/Bi eutectic was used to represent the family of low-melting point, In-free alloys. Sn/3Ag/2Bi is in the Sn primary phase field in the Sn/Ag/Bi phase diagram, while Sn/3.4Ag/4.8 Bi is in the Ag/3Sn primary phase field. The primary phase field indicates the first phase to form at equilibrium. It was expected that the microstructures and properties of these two alloys would be measurably different. The alloy Sn/2.6Ag/ 0.8Cu/0.5Sb is a commercial alloy, available in both paste and bar stock and was used to represent the Sn/Ag/Cu family of alloys, and for comparison with Sn/3.5Ag and Sn/3.5Ag/0.5Cu/1Zn. Sn/2.8Ag/20In is a commercial alloy. This alloy failed the pass/fail down-selection criteria for cost, availability and pasty range (as a result of an observed large non-equilibrium pasty range), but was included for testing because the liquidus temperature is close to eutectic Sn/Pb; it was necessary to include one alloy that failed the physical properties down-selection criteria to test the ability of those criteria to discriminate among solder alloys.

The manufacturing assessment was performed using the NCMS soldering test vehicle (STV), a PCB having both surface mount and through-hole components. This was a pass/fail assembly test, performed so that alloys that could not be assembled with reasonable yield using normal electronics assembly operations (e.g., stencil printing, solder reflow, wave soldering) could be eliminated from reliability assessment. In all evaluations, eutectic Sn/Pb was included as a baseline for comparison.

No lead-free solder wet as well as eutectic Sn/Pb for four of the five PCB surface finishes examined (imidazole, immersion Sn, Ni/Pd and Pd); wetting with a Ni/Au surface finish was comparable to Sn/Pb for all lead-free solder alloys examined. The imidazole-finished copper provided the least favorable surface for solder wetting. The metallic-based surface finishes (Sn, Pd or Au) enhanced the wetting/spreading of all solders, most dramatic for the Sn/58Bi solder, which exhibited the poorest wetting on copper.

This assessment indicated that modification in board design and process parameters will be required for establishing a manufacturability of lead-free solders that is equal to or better than what currently exists for tin/lead solders. All lead-free solders exhibited adequate surface mount and bottomside through-hole joint fillets but poor topside through-hole fillets. Careful temperature profiling was required to avoid reflow of the topside surface mount components during wave soldering. These two problems observed during assembly of the soldering and reliability test vehicles are indicative of the narrow process window for these lead-free solders.

Cross sections of STV through-hole joints showed that the solder fillet had separated from the PCB land in many of the high-tin alloys on cooling from the reflow temperature. A series of designed experiments, complemented by finite element modeling, was performed to understand the cause of this "fillet lifting." Fillet lifting appears to be a strong function of composition, with the tendency for fillet lifting increasing as the pasty range increases. It is not clear from these results what effect fillet lifting will ultimately have on the long-term reliability of through-hole joints.

Reliability testing was performed on both through-hole assemblies and surface mount boards with two thermal cycles: cycling between -55° and 125°C represented the requirements of the military and automotive sectors; and cycling between 0° and 100°C represented the requirements of the telecommunications and consumer electronics industry. Short descriptions of the surface mount test vehicle, methods and results are provided here. Full details can be found in the NCMS final report.1

The surface mount reliability test vehicles (RTV-SM) were designed for double-sided assembly, both to mimic use conditions and maximize the sample size. The layout was simply repeated on the second side, resulting in a component being placed back to back with a similar-type component. Double-sided assemblies were used for electrical testing to maximize sample size, while single-sided assemblies were used for cross sectioning to facilitate sample preparation. All components and boards had lead-free surface finishes. Multilayer FR-4 epoxy glass boards were chosen for the RTV-SM to match product requirements. Components were distributed across the board so that each component type occupied both edge and central positions, because components placed near the board edges often fail earlier than those located more centrally. Components were connected in series to allow electrical continuity testing during thermal cycling. The half-moon structure allowed isolation of every pair of joints so that the defects or failures could be bypass-jumpered to continue the test after some components failed. A total of 10,320 joints were tested per alloy per test condition. The thermal cycling conditions are listed in Table 3.

Solder fatigue failures occurred for the leadless ceramic chip carrier (LCCC) and 1206 resistors only. For LCCCs for both thermal cycling conditions, all lead-free alloys performed as well or better than eutectic Sn/Pb.

No solder fatigue failures occurred for the bumpered quad flat packs (BQFP), plastic leaded chip carriers (PLCC) or 1206 chip capacitors for these experiments. Alloy performance for the 1206 chip resistors was evaluated primarily through comparisons of total number of failures in the test, cycles to first failure, cycles to 10 percent failure and microstructures, with limited use of Weibull statistics. The Weibull plots are shown in Figure 3, with the data described as follows:

The 0° to 100°C thermal cycle. The failure levels ranged from 0 to 11 percent at the end of 6,673 cycles, as seen in Figure 3a.

1.F17 was the only alloy that was failure free for the test duration and was ranked first.

2.A6 was the next best performer with less than 1 percent failure for the test duration.

3.A1, A4 and E4 were grouped together as the next best performers with approximately 2 to 5 percent failures for the test duration.

4.F2, F21 and F27 were grouped together and ranked the lowest with about 8 to 11 percent failures.

These alloy rankings were supported by cross-section photomicrographs of chip resistors that have been exposed to approximately 5,700 cycles of the 0° to 100°C temperature range.

The -55° to 125°C thermal cycle. The failure levels varied from 4 to 57 percent at the end of 5,000 cycles, as shown in Figure 3b.

1.A6 had the best performance with approximately 4 percent failed.

2.A1 and F17 are grouped together as the next best performers with about 13 and 16 percent failures, respectively.

3.The remaining five alloys were tightly grouped, with failure levels ranging from 54 to 57 percent. Listed by the 10 percent failure level, A4, E4, F2 and F27 performed essentially the same, while F21 had slightly worse fatigue life.

This general progression of alloy rankings are supported by the cross-section photomicrographs of chip resistors that have been exposed to approximately 4,300 cycles of the -55° to 125°C temperature range.

From thermal cycling of the different surface mount components in this study, it is clear that alloy ranking depends on the thermal stresses imposed on the solder by the component and PCB geometry and the thermal cycling environment. The suitability of an alloy for a particular application will be determined, therefore, by the failure of the first component on the board. From the data on the 1206 resistors, it appears that the high-Sn alloys, with the exception of Sn/3.4Ag/ 4.8Bi (F17), had poorer performance than eutectic Sn/Pb or eutectic Sn/Bi. For plastic quad flat packs (PQFP), PLCCs or 1206 capacitors, no meaningful electrical failures were observed for any of the alloys, suggesting that all lead-free candidate alloys might be acceptable as substitutes for eutectic Sn/Pb under those states of stress.

Using the test results, the participants of the NCMS project came up with three recommended alloys. It is expected that the Sn/58Bi eutectic, Sn/3.4Ag/4.8Bi, and Sn/3.5Ag eutectic solders might perform substantially better than eutectic Sn/Pb in certain applications, as summarized in Table 4. The Sn/58Bi eutectic and Sn/3.5Ag/4.8Bi had fatigue lives comparable to or better than eutectic Sn/Pb in both 0° to 100°C and -55° to 125°C thermal cycling environments. Cross sections of 1206 resistors and capacitors showed that both alloys exhibited only minor cracking in surface mount joints after 0° to 100°C thermal cycling, compared with more extensive cracking seen for all other down-selected solder alloys and for eutectic Sn/Pb. For 1206 resistors at 0° to 100°C, alloy Sn/3.5Ag/4.8Bi was the only alloy to have no electrical failures up to 6,673 cycles, surpassing the performance of eutectic Sn/Pb by at least 2,000 cycles. For 1206 resistors at -55° to 125°C, the number of cycles to first failure for Sn/3.5Ag/4.8Bi was equivalent to eutectic Sn/Pb. For 1206 resistors, Sn/58Bi showed fewer total failures by the end of the tests than eutectic Sn/Pb.

The major conclusion of the project was that no drop-in replacement was found for eutectic Sn/Pb solder; promising alternatives exist, but more detailed information is needed before implementing any wholesale changes to present industry practice. Of particular concern were the changes needed in the materials and processing infrastructure. Over the past 30 years, the electronics industry has built its packaging infrastructure around eutectic Sn/Pb. Components, board materials and fluxes were all designed for an assembly process that operated well when materials reached peak temperatures around 220°C, not 240°C. If high-Sn solders are to be used, current board, component and process materials must be evaluated to determine if they can tolerate exposure to higher temperatures during processing. For example, many integrated circuit (IC) packages may not survive higher reflow temperatures. The industry`s standard test method for rating these devices is IPC/JEDEC J-STD-020, "Moisture/Reflow Sensitivity Classification for Plastic Integrated Circuit Surface Mount Devices." A process temperature of 215° to 219°C is used in this test method in which a rating is established to effectively designate the amount of time a device can be exposed to a factory environment after a moisture removal bake and still survive a reflow soldering process. Higher process temperatures will have a severe, negative impact on component performance, and therefore on the component ratings. If these materials are not stable and use of lead-free solders is desired or required, alternative materials will be needed to allow the products to survive the manufacturing process.

REFERENCE

1 NCMS Lead-free Solder Project Final Report, NCMS, Ann Arbor, Mich., August 1997. To obtain a copy of the report and the CD-ROM archive from NCMS, contact Barbara Johnson at (734) 995-4938 or visit www.ncms.org/3portfolio/1ProjectPortfolio/pubs.htm.

The National Center for Manufacturing Sciences Lead-free Solder Project is made up of the following individuals: IRIS ARTAKI and DONNA NOCTOR may be contacted at Lucent Technologies, Engineering Research Center, Princeton, NJ 08542; E-mail: artaki@nist.gov. CHARLES DESANTIS and WILLIE DESAULNIER may be contacted at Hamilton Sundstrand, One Hamilton Road, Windsor Locks, CT 06096-1010; E-mail: desanti@hsd.utc.com. LAWRENCE FELTON and MARK PALMER may be contacted at the Center for Integrated Electronics and Electronics Manufacturing, Rensselaer Polytechnic Institute, Troy, NY 12180-3590. JOE FELTY may be contacted at Texas Instruments Inc., 2501 W. University, M/S 8059, McKinney, TX 75070; E-mail: J-Felty@ti.com. JOHN GREAVES may be contacted at EMPF, 714 N. Senate Ave., Indianapolis, IN 46202-3112. CAROL HANDWERKER may be contacted at the National Institute of Standards and Technology, Materials Science and Engineering Laboratory, Gaithersburg, MD 20899-8550; E-mail: carol. handwerker@nist.gov. JOHN MATHER and SCOTT SCHROEDER may be contacted at Rockwell International, 400 Collins Road NE, MS 106-132, Cedar Rapids, IA 52498; E-mail: jcmather@collins.rockwell.com. DUANE NAPP may be contacted at 111 Hazeltine Drive, Georgetown, TX 78628; E-mail: duane.napp@ ncms.org. TSUNG-YU PAN may be contacted at Ford Motor Co., 20,000 Rotunda Drive, Bldg R, Dearborn, MI 48121-2053; E-mail: tpan@ ford.com. JERALD ROSSER may be contacted at GM-Hughes Aircraft Co., Tucson, AZ 85734-1337. PAUL VIANCO may be contacted at Sandia National Laboratories, 1515 Eubank SE, Albuquerque, NM 87185-1411; E-mail: ptvianc@ccsmtp.sandia.gov. GORDON WHITTEN and YUN ZHU may be contacted at Delco Electronics, 700 East Firmin Street, Kokomo, IN 46904-9005; E-mail: gcwhitte@mail.delcoelect.com.

Figure 1. Program plan schematic.

Figure 2. Ternary phase diagram of Sn/Ag/Bi system with shading indicating composition ranges with less than 30°C pasty range.

Figure 3. Three-parameter Weibull cumulative failure predictions for 1206 resistors on FR-4.