The following article appears in the journal JOM, 53 (6) (2001), pp. 28-32.

Lead-Free Solder: Research Summary

Pb-Free Solders for Flip-Chip Interconnects

TABLE OF CONTENTS
INTRODUCTION Pb-FREE SOLDER ALLOYS AND UBM STRUCTURES PHYSICAL METALLURGY OF Pb-FREE SOLDER ALLOYS MECHANICAL METALLURGY OF Pb-FREE SOLDER ALLOYS OPTIMAL Pb-FREE SOLDER ALLOY EXPERIMENTAL PROCEDURE ACKNOWLEDGEMENTS References
Figure 1. Chart of lead-free solder tem-peratures (dark–solidus; light–liquidus).

A variety of lead-free solder alloys were studied for use as flip-chip interconnects including Sn-3.5Ag, Sn-0.7Cu, Sn-3.8Ag-0.7Cu, and eutectic Sn-37Pb as a baseline. The reaction behavior and reliability of these solders were determined in a flip-chip configuration using a variety of under-bump metallurgies (TiW/Cu, electrolytic nickel, and electroless Ni-P/Au). The solder microstructure and intermetallic reaction products and kinetics were determined. The Sn-0.7Cu solder has a large grain structure and the Sn-3.5Ag and Sn-3.8Ag-0.7Cu have a fine lamellar two-phase structure of tin and Ag₃Sn. The intermetallic compounds were similar for all the lead-free alloys. On Ni, Ni₃Sn₄ formed and on copper, Cu₆Sn₅ Cu₃Sn formed. During reflow, the intermetallic growth rate was faster for the lead-free alloys, compared to eutectic tin-lead. In solid-state aging, however, the interfacial intermetallic compounds grew faster with the tin-lead solder than for the lead-free alloys. The reliability tests performed included shear strength and thermomechanical fatigue. The lower strength Sn-0.7Cu alloy also had the best thermomechanical fatigue behavior. Failures occurred near the solder/intermetallic interface for all the alloys except Sn-0.7Cu, which deformed by grain sliding and failed in the center of the joint. Based on this study, the optimal solder alloy for flip-chip applications is identified as eutectic Sn-0.7Cu.

INTRODUCTION

Metallic lead combined with tin forms a low melting-temperature alloy that is extensively used for electronic interconnects. However, medical studies have shown that lead is a heavy-metal toxin that can damage the kidneys, liver, blood, and central nervous system. Less than one percent per year of the global lead consumption is used in solder alloys for electronic products but electronics and electrical systems make up an increasingly larger fraction of landfills.¹ The issue of lead leaching from landfills into the water table has raised alarms as a potential source of long-term contamination of soil and ground water. International laws have recently been proposed to limit or ban the use of lead in manufactured products. The most aggressive and well-known effort is the European Union’s Waste in Electrical and Electronic Equipment directive that proposes a ban on lead in electronics by 2008. The Japanese Environmental Agency has proposed that lead-containing scrap must be disposed of in sealed landfills to prevent lead leaching. The Japanese Ministry of International Trade and Industry and the Japan Automobile Industries Association called for a 50% reduction of lead in vehicles (excluding batteries) by 2001 and a 33.3% reduction by 2003.² Electronic-equipment manufacturers have responded to these proposed bans in a variety of ways. Many companies have not taken a stance hoping that legislation will not be enacted. Other companies have aggressively pursued solutions to the proposed bans and are using lead-free products as a “green” marketing strategy. Motorola, as a corporation, has introduced Environmentally Preferred Products and, in 2001, will release new lead-free products.

Extensive research on lead-free solders has been published over the past decade. A comprehensive review of lead-free solders was performed by Glazer,^3,4 and, since that time, research has focused on carrier-to-board (surface-mount and through-hole) interconnects.^5–15 A growing area of interest is in lead-free solders for flip-chip interconnects.

Flip-chip interconnects are the electrical and mechanical connections between the semiconductor integrated circuit and the package (or board for direct-chip attach). These interconnects are formed on the periphery or in an area array on the top surface of an active die. An under bump metallurgy (UBM) forms the solderable surface on the die. The UBM also acts as a diffusion barrier between the tin and the solder and must be thick enough to withstand interactions (intermetallic formation) between the solder and UBM. Flip-chip interconnects are smaller (on the order of 100 mm diameter) than surface-mount joints and are projected to have pitches that shrink below 150 mm. Flip-chip interconnects must be able to withstand a potentially high level of strain mismatch between tin and an organic substrate. Flip-chip technology has moved from ceramic packaging with high lead solder (97.5Lead-2.5Sn) to an organic package that requires lower temperature reflow (<260ºC). This can be accomplished by bumping the die with high lead solder then joining it to an organic board with eutectic tin-lead, but this method adds costs and can be eliminated with a monolithic solder. The joints must withstand board-level reflow environments compatible with joining to organic substrates that, again, have a maximum reflow temperature of 260ºC. The lead-free solder must meet these requirements and perform at, or above, the level of performance of the eutectic tin-lead solder it is intended to replace.

One additional benefit of a lead-free flip-chip interconnect is the reduction of pb²¹⁰ -created alpha-particle radiation. Due to their proximity to active elements, the flip-chip solder interconnects should have minimal levels of alpha-particle radiation to limit soft errors in complementary metal-oxide semiconductor (CMOS) technology that become more critical as cell size on the die is reduced.^16,17 The elemental constituents of lead-free solders (tin, copper, silver, bismuth, indium, and antimony) do not radioactively decompose, so alpha-particle radiation is minimal.

Pb-FREE SOLDER ALLOYS AND UBM STRUCTURES

Lead-free solders for electronic applications are based on tin-rich compounds that fall into a melting temperature range similar to the traditional eutectic lead-tin solder alloys (183ºC). These include eutectic Sn-3.5Ag with alloying elements of bismuth, copper, antimony, indium, or zinc. Other alloys based on the tin-copper, tin-indium, tin-antimony, tin-bismuth, and tin-zinc systems have also been proposed. A small two-phase region (temperature difference between liquidus and solidus) is desired because it prevents the joint from moving and becoming disturbed during solidification. Binary or ternary near-eutectic alloys are also desired because simpler alloys reduce the potential for compositional variations that affect the behavior of the solder joint. A chart of the melting temperatures of potential solder alloys is shown in Figure 1. The alloys Sn-0.7Cu, Sn-3.5Ag, and Sn-3.8Ag-0.7Cu were immediately eliminated as promising flip-chip solder alloys based on the selection criteria. Eutectic Sn-37Pb was included as a control. The tin-antimony alloy was deemed to have too large a two-phase region and the liquidus temperature of 240ºC was too high for chip attachment to organic substrates (a process temperature of 245ºC is desired). The tin-zinc alloy had too many processing difficulties due to the rapid oxidation behavior of the zinc in the molten state and corrosion susceptibility of the alloy after solidification. The tin-silver-bismuth alloy had too large a two-phase region. The Sn-Ag-Cu-Sb was not studied because it was feared that in a flip-chip application the solder would damage the die rather than deform due to the high strength of the alloy.¹⁸ Flip-chip interconnects are part of a package that must undergo carrier-to- board reflow that occurs, at a minimum, at 220ºC. The tin-bismuth eutectic alloy melts at far too low a temperature (138ºC) to withstand this reflow. The alloys can be processed on silicon die either as solder paste stencil printed on the defined UBM pads or through direct plating on the UBMs.

The UBMs used in this study were selected based on prior experience with these structures with eutectic tin-lead alloys. The UBMs were sputtered TiW/plated copper, electroless plated Ni-P/immersion gold, and electrolytic nickel (used only for solder plating). The nickel was chosen because it reacts slowly, compared to copper or gold, with tin-rich solders. The TiW/Cu used sufficient copper to survive multiple reflows with the tin-rich alloys.

PHYSICAL METALLURGY OF Pb-FREE SOLDER ALLOYS

The microstructure of the solder alloys studied are shown in Figure 2, Figure 3, Figure 4, and Figure 5. The tin-lead eutectic has a structure of tin-rich and lead-rich phases that form as lamella (Figure 2). Similarly oriented lamella form cells, or colonies, separated by slightly coarsened cell boundaries. This structure is susceptible to heterogeneous microstructural evolution that concentrates strain at the cell boundaries and causes them to further coarsen and, eventually, be the site of failure. In the fine scale of the flip-chip interconnect, only a few cells are present in each bump but heterogeneous coarsening and failures are still observed.¹⁹

Figure 2. An optical micrograph of the microstructure of Sn-37Pb solder in a flip-chip bump on a copper UBM.		Figure 3. An optical micrograph of the microstructure of Sn-0.7Cu solder in a flip-chip bump on a copper UBM.
Figure 4. An optical micrograph of the microstructure of Sn-3.5Ag solder in a flip-chip bump on a copper UBM.		Figure 5. An optical micrograph of the microstructure of Sn-3.8Ag-0.3Cu solder in a flip-chip bump on a copper UBM.

Figure 3 shows the Sn-0.7Cu microstructure. This solder is composed of large, tin-rich grains with a fine dispersion of Cu₆Sn₅ intermetallics. The solder grains, which form and grow out from the bond pad interfaces, are large, on the order of 20–50 mm.

The Sn-3.5Ag microstructure consists of a fine structure of alternating tin-rich/Ag 3 Sn intermetallic lamella (Figure 4). Grain colonies also form in this microstructure but the boundaries are not coarsened. In addition to the fine Ag₃Sn intermetallics, large needles of Ag₃Sn are present and are typically attached to one of the bump-pad interfaces.

The Sn-3.8Ag-0.7Cu alloy has a structure of tin-rich grains with Ag₃Sn and Cu₆Sn₅ intermetallics dispersed throughout (Figure 5). The Ag₃Sn also forms as large plates attached to the interfacial intermetallics.

When the molten solders come into contact with the nickel or copper surfaces, they wet and react to form interfacial intermetallics. The intermetallics grow out into the solder as rods, or plates, and continue to grow when the solder is solid. Even though all the solders studied are tin-rich, the morphology and reaction kinetics differ between alloys.

On copper, Sn-37Pb forms a two-phase intermetallic of Cu₆Sn₅ adjacent to the solder and Cu₃Sn adjacent to the copper. The Cu₃Sn is planar with a columnar grain structure and the Cu₆Sn₅ consists of elongated nodules. On nickel, eutectic tin-lead forms irregularly shaped Ni₃Sn₄. The formation and growth of the interfacial intermetallics between copper, nickel, and eutectic tin-lead solder are well known.^18–20 The intermetallics follow parabolic growth kinetics and do not extensively spall off into the solder.

Figure 6. SEM micrographs of the lead-free solders on an electroless nickel UBM showing the morphology of the interfacial intermetallics.

The intermetallic formation for the lead-free solders on nickel after two reflows is shown in Figures 6a, 6b, and 6c. Two reflows represent typical processing for flip-chip interconnects. The first reflow represents ball formation on the wafer, while the second represents flip-chip to substrate interconnect. Figures 6a, 6b, and 6c show the structure of the interface on electroless nickel-phosphorus, but the same observations were made for the electrolytic nickel UBM. The intermetallic that forms is Ni₃Sn₄. The Ni₃Sn₄ intermetallic between Sn-0.7Cu and nickel is thin, regular, and is the most uniform of the lead-free alloys. For Sn-3.8Ag-0.7Cu, the intermetallic is thicker but has the same regular chunky morphology as for Sn-0.7Cu. TheSn-3.5Ag solder on nickel has a different intermetallic morphology that consists of nodules and chunks of Ni₃Sn₄ that spall off into the solder. This morphology has been attributed to the lack of copper in the solder. It is hypothesized that the copper acts to saturate the solder with respect to the nickel and inhibits dissolution and spalling of the intermetallic into the solder.²¹ However, this mechanism remains to be fully understood.

The growth of interfacial intermetallics while the solder is in the solid-state is of concern for flip-chip interconnects. If the intermetallic layer coarsens significantly, it can consume the UBM and cause the joint to dewet at the layer beneath the UBM. Also, because the intermetallic is brittle, if it becomes a significant fraction of the solder joint, it can act as a site for crack initiation and propagation when the joint is deformed.

Figure 7. Thickness of consumed copper for the solders on electroless Ni-P/Au UBM.

The consumption of the nickel layer by the formation of Ni₃Sn₄ intermetallic during solid-state aging for each of the solder studied is shown in Figure 7. The solders were aged for 1,000 hours at 150ºC and 170ºC. Figure 7 shows results on an electroless Ni-P/Au UBM (similar results were observed on the electrolytic nickel UBM). For all solders studied, less than 2 mm of nickel was consumed while transforming into Ni₃Sn₄. Because the nickel reacts slowly with tin-based solders, it is preferred for tin-rich UBM structures. The lead-free solders consume more nickel and form more slightly more intermetallic than tin-lead eutectic solder. but this increase is relatively small. The Ag₃Sn intermetallic plates are attached to the Ni₃Sn₄ interfacial intermetallics, similar to that observed on the copper UBM (Figure 4 and Figure 5).

Figures 8a, 8b, 8c, and 8d show SEM micrographs of the solders on a copper UBM after two reflows. The intermetallic that forms is Cu₆Sn₅ , no Cu₃Sn was found but it may have been too thin to be observed. The interfacial intermetallic has the same morphology for all the lead-free alloys as for tin-lead eutectic.The intermetallic consists of regularly spaced nodules of Cu₆Sn₅.The silver-containing solders all have large Ag₃Sn intermetallics that are attached to the Cu₆Sn₅ interface (Figure 4 and Figure 5). All three lead-free alloys also have small, discrete particles of Cu₆Sn₅ present in the bulk of the solder. For Sn-0.7Cu and Sn-3.8Ag-0.7Cu, the copper is present in the solder before joining to the UBM. For the Sn-3.5Ag solder theCu₆Sn₅ is present due to the dissolution of some of the UBM into the solder. Spalling of the interfacial intermetallics into the molten solder was not observed, probably due to the presence of copper in each of the solders during reflow. The copper inhibits growth and spalling of the intermetallic because it saturates the solder.²² Of the three lead-free alloys on copper, theSn-0.7Cu solder structure is the most uniform, and has the thinnest intermetallic structure.

One of the concerns of using tin-rich, lead-free solders is the reaction of the copper with the solders is feared to be so fast that the UBM will dissolve during reflow. A plot of the copper consumed after two reflows for tin-lead and the lead-free solders is shown in Figure 9. The lead-free solders consume only 10–20% more copper than tin-lead, which is less than 2 mm after 2 reflows. A plot of copper consumed during solid-state aging is shown in Figure10 for the solders on copper at 150ºC for 500 and 1,000 hours. In the solid state, the copper was consume data slower rate in lead-free solders than for eutectictin-lead. The lead appears to play a role in enhancing intermetallic growth, perhaps by enhancing tin diffusion to the intermetallic/solder interface.Of the alloys studied, Sn-0.7 Cu had the slowest consumption rate of the copper UBM.

Figure 8. SEM micrographs of the lead-free solders on a copper UBM.

MECHANICAL METALLURGY OF Pb-FREE SOLDER ALLOYS

The flip-chip solder joint must mechanically hold the chip to asubstrate, but the solder can not impose significant strain to the semiconductor device or the device could crack and fail. The shear strength of the solder alloys is shown in Figure11. The strength of the solder alloys is a measure of the ability of the flip-chip interconnect to be compliant to the imposition of strain. Sn-37Pb and Sn-0.7Cu solder bumps have similar values of shear strength while the Sn-3.8Ag-0.7Cu is 20%stronger.TheSn-3.5Ag solder alloy has the greatest shear strength at 25% greater than Sn-0.7Cu.The shear failure for all joints tested occurred solely through the solder. The solder joint strength and failure mode were similar to, and independent of, the UBM type. Eutectic Sn-3.5Ag, and the companion Sn-Ag-X alloys (including tin-silver-copper), have shown susceptibility to brittle interfacial delamination in tensile ¹⁸ and shear tests ²³ for surface mount interconnects. The failure occurs at the intermetallic/solder interface where one side of the fracture shows intermetallic and the other side reveals an impression of where the intermetallic grew in to the solder. It is not clear why the Sn-Ag(-X) solder/intermetallic interface is so weak. There is some evidence ¹⁸ that silver segregates to the interface and weakens it by “poisoning.” The brittle fracture is exacerbated with gold contamination.²⁴

Creep behavior is important for flip-chip interconnects because the solders deform to relax stress over time when held at a constant strain.The creep rate of a solder must be sufficiently fast so that the strain is minimized in joined bulk components.However, the creep rate must not be so fast that the components move over time. The creep behavior of solders can be summarized empirically using one of two equations:

where dg/dt is the creep rate in shear, A is a constant, a is the stress constant, s is the flow stress, n is the stress exponent, and Q is the creep activation energy. Equation 1a works well for creep mechanisms that remain constant over all test temperature. Equation 1b is the Garofalo, or sinh, creep relation that captures up to two different creep mechanisms in a single formulation. Creep tests were not performed in this study but have been investigated to some extent in the published literature.^25-28 Table I represents the fitted results of creep tests on lead-free solders using these equations.Results are presented for tin-lead using Equation 1a and Equation 1b but the Garofalo fit was much better.²⁹ The creep rate of high tin-content solders is slower than tin-lead.

Figure 9. Consumed copper thickness after two reflows for the solders on a copper UBM.		Figure 10. Bar chart of the consumed copper thickness of the solders on a copper UBM after aging for 500 hours and 1,000 hours at 150 C.		Figure 11. Shear strength of the solder flip-chip bumps.

Figure 12. Plot of thermal fatigue life as a function of applied thermal strain for the Sn- 3.5Ag, Sn-37Pb, Sn-3.8Ag-0.7Cu, and Sn- 0.7Cu.

Eutectic Sn-3.5Ag has greater strength and a higher creep resistance than eutectic tin-lead.²⁹ The creep behavior of Sn-3.8Ag-0.7Cu is similar to Sn-3.5Ag and has been found to have the highest creep resistance of the lead-free alloys.³⁰ The steady-state creep behavior for tin-copper has been found to be faster than eutectic tin-lead.The faster creep rate of tin-copper is desired for flip-chip applications because damage can be accommodated by the solder, rather than the more brittle joined components. Furthermore, a faster creep rate often translates into a longer thermomechanical fatigue lifetime.

Temperature variations encountered during use conditions, combined with the materials of differing coefficients of thermal expansion in the electronic package, result in cyclic temperature and strain on the solder joints. Glazer presented a summary of work on lead-free solders through 1994 and concluded that the thermomechanical fatigue data for low melting-point solders are scarce and appear contradictory.^3,4 The published data has primarily focused on surface mount interconnects; thermomechanical fatigue data in a flip-chip configuration is still needed.

A diagram of thermal fatigue life vs. strain range was obtained, as shown in Figure 12, using the characteristic life extracted from Weibull plots of the thermomechanical fatigue data and thermal strain range calculated from Equation A both for the 0ºC–100ºC and–40ºC–125ºC tests. This data was collected on flip-chip packages with no underfill. By eliminating the underfill, this study evaluated the relative thermal fatigue performance of different lead-free solders and characterized the failure mechanisms. Although the strain evaluated here tends to be much higher than that of packages with underfill encapsulation, the failure mode remains valid. The thermal fatigue failure data follows a power law function. In Figure 12, the thermal fatigue data from Sn-0.7Cu flip-chip joints on both NiP and TiW/Cu UBMs fall onto the same straight line. The thermal fatigue performance of Sn-0.7Cu is UBM independent and is confirmed by the failure analysis discussed below, where failures occurred solely through the solder joint. The cross sections of the failed joints shown in Figures 13a, 13b, 13c, and 13d are on the TiW/Cu UBM, but similar failure behavior was exhibited for the electrolytic nickel and electroless nickel/gold UBM structures. Sn-3.8Ag-0.7Cu and eutectic tin-lead have a similar thermomechanical fatigue life. Sn-3.5Ag has the shortest thermal fatigue lifetime.

Figure 13. Optical micrographs of solder joint cross sections of solder joints on copper UBM/ Cu pads on organic substrates after thermal cycling at 0ºC–100ºC.

EXPERIMENTAL PROCEDURE Silicon wafers with aluminum daisy chain test structures were bumped with the UBM and solder on an array with a pitch of 300mm The UBMs deposited included sputtered TiW/plated copper, electrolytically plated nickel, and electroless plated Ni-P/immersion gold. The solder was deposited as a screen-printed paste or directly plated on the UBM pads. The lead-free solder/UBM systems were characterized after reflow at 260ºC for a time of 1 minute above 220ºC.The eutectic Sn-40Pb samples under went reflow at 220ºC for a time of 1 minute above 190ºC.The multiple reflow tests followed the same profiles.The dice were assembled to organic substrates with copper bond pads. A no-clean flux was used to promote wetting and was not removed after assembly. No underfill was placed between the die and substrate in order to reduce the number of cycles to failure.The underfill acts as a mechanical reinforcement between the die and substrate so by eliminating the underfill, the solder is exposed to the imposed strains.The dice were mechanically tested in a shear orientation to measure strength.The tests were performed using a die shear test fixture in a load frame at a displacement rate of 0.25 mm/second. Two thermal cycling conditions were chosen for the evaluation, 0 –100ºC and –40 –125Cº, both at the frequency of 1 cycle/h. Based on actually measured bump distance to the neutral point (DNP) and thermal mismatch (Da) between die and substrate, thermal strain is calculated from: (A) where: g =strain on the outermost flip-chip bumps; Da =difference of chip/substrate coefficient of thermal expansion (CTE)(ppm/ºC), i.e., 15.2 ppm/º8C; DT =relative thermal excursion; H =flip-chip bump gap height (mm) after assembly; DNP = distance to the neutral point of the die (mm) . The calculated strain on the outermost bumps is 9.81% for the bumps on TiW/Cu UBM during 0 – 100ºC thermal cycling. In the –40 –125ºC thermal-cycling condition, the strain on the outermost bumps is 15.48%. This calculation assumed that the strain caused by thermal mismatch between die and substrate is fully loaded onto the solder bump without consideration of relaxation in the substrate. During thermal fatigue cycling, the parts were electrically tested every 25–50 cycles on a bench tester. The failed parts were pulled and electrically probed to determine the failure location. Then, the parts were cross-sectioned to analyze the fatigue crack profiles. The microstructure of the solder interconnects were characterized by sequentially grinding and polishing down to 0.25 mm followed by a polish/etch in a slightly acidic colloidal suspension. The structures were revealed using an etch of 10% hydrochloric acid +90% methanol for a few seconds at room temperature. Etching the solders more extensively, 4 parts glycerol +1 part acetic acid +1part nitric acid at 80ºC for a few seconds removes the solder and leaves the interfacial intermetallic interface that can be characterized in a scanning electron microscope.

The Sn-37Pb flip-chip interconnects fail by crack formation and propagation through heterogeneous coarsened bands near the UBM/solder interface as shown in Figures 13a, 13b, 13c, and 13d. The heterogeneous coarsened region forms as a result of the strain concentrating at the eutectic cell boundaries,^31–33 which are slightly coarsened and are the weakest regions of the joint. Damage accumulates at the cell boundaries and anneals at the high-temperature portion of the thermal cycle, resulting in coarsening. Coarsening continues until the tin and lead grains in the coarsened bands are so large that they can no longer slide and rotate to accommodate the strain, resulting in grain-boundary crack initiation and propagation. The UBM/solder interface is the highest strain region in the joint and is where coarsening and failures typically occur for eutectic tin-lead flip-chip interconnects. The cracks were observed to form on the outer edge of the bump and propagate to the center of the bump. The surface of the solder joint remains smooth after thermal cycling, indicating that the damage was localized to the heterogeneous coarsened band.

The eutectic Sn-0.7Cu solder exhibited a failure mode that differed from the other solder alloys studied. The initiation and propagation of fatigue cracks is through the grain boundaries in the Sn-0.7Cu solder, as shown in Figures 13a, 13b, 13c, and 13d. The fatigue cracks initiated on the highest point of strain on the bump, closer to the edge of die, then propagated across the middle of the bump toward the center of the die. After thermal cycling, the surface of the solder bumps is no longer smooth. The Sn-0.7Cu solder deforms by grain-boundary sliding. The cracks were observed to propagate at the grain boundaries, significantly removed from the UBM/bump interface, near the center of the joint. This solder is the most compliant in thermal fatigue and undergoes massive deformation before failing by crack propagation.

The failure mode in ternary Sn-3.8Ag-0.7Cu solder on TiW/Cu UBM is significantly different from that in Sn-0.7Cu. The thermal fatigue cracks initiated and propagated at the intermetallics and intermetallic/solder interface, as shown in Figures 13a, 13b, 13c, and 13d. The solder bump surface did not deform during thermal cycling nor did the microstructure evolve. The fine colony and dendritic microstructure makes this solder bump stronger and less compliant to thermal fatigue deformation than Sn-0.7Cu solder. The strain imposed during thermal cycling is accommodated at the weak link of the SnAgCu/UBM interconnect at the IMC/ solder interface. This failure mode has a shorter fatigue life for Sn-3.8Ag-0.7Cu compared to Sn-37Pb or Sn-0.7Cu.

The failure behavior of Sn-3.5Ag solder is shown in Figures 13a, 13b, 13c, and 13d. Similar to the Sn-3.8Ag-0.7Cu alloy, the thermal fatigue cracks in the Sn-3.5Ag solder cracks initiate and propagate through the intermetallics and at the intermetallics/solder interface. The surface of the solder bump also exhibited no deformation with the damage concentrated at the solder/intermetallic interface. The Sn-3.5Ag solder has a shorter thermal fatigue lifetime than Sn-3.8Ag-0.7Cu, even though the failure behavior appears to be similar. The only difference between the two alloys is the microstructure of the Sn-3.5Ag bump appears to have more Ag 3 Sn intermetallics present. It is hypothesized that the large Ag₃Sn intermetallic plates strengthen the joint, as shown in Figure 11, and further reduce the compliance of the solder, thereby shortening the thermal fatigue life.

For the solders examined in this study, the shear strength has a direct correlation to fatigue performance. The weaker, more compliant, and faster steady-state creep rate solder (Sn-0.7Cu) has a longer fatigue lifetime. Eutectic tin-lead has the same low shear strength as Sn-0.7Cu but has a shorter thermal fatigue lifetime due to heterogeneous coarsening. It is well established that heterogeneous coarsening concentrates strain in the joint to the coarsened band and accelerates crack formation and propagation.^31–33

The Sn-0.7Cu alloy undergoes massive deformation during thermal cycling. This deformation protects the semiconductor device from damage due to imposed strain. Interestingly, even with significant surface deformation, the crack formation takes longer in Sn-0.7Cu than the other solders that had no surface deformation. The Sn-0.7Cu accommodates the thermal strain by grain-boundary sliding and rotation. The self-diffusion of tin to accommodate the sliding and rotation is sufficient to delay the formation of cracks. An additional positive aspect of the Sn-0.7Cu solder is that failure is only observed in the solder joint away from the brittle intermetallic/solder interface.

OPTIMAL Pb-FREE SOLDER ALLOY

Based on the results of this study, the optimal lead-free solder alloy for flip-chip interconnects is Sn-0.7Cu. The alloy contains no silver, so it is the lowest cost alternative of the lead-free alloys. The solder can be applied in either a paste form (for screen-printed bumps) or plated. The other lead-free solder alloys can only be plated with difficulty. Sn-0.7Cu has the highest melting temperature of the lead-free alloys proposed for flip-chip applications and, thus, provides some melting temperature hierarchy for subsequent package solder processing. The intermetallic structure of the lead-free alloys is similar to that of eutectic tin-lead. The interfacial intermetallics grow at a faster rate during reflow of the lead-free alloys, but, during solid-state aging, the lead-free solders have a slower intermetallic growth rate. Sn-0.7Cu has superior thermal fatigue performance of the lead-free alloys, surpassing that of the low melting temperature standard of eutectic tin-lead. For lower cost, ease of processing and superior thermal fatigue performance, Sn-0.7Cu is the optimal lead-free solder alloy for flip-chip interconnects.

ACKNOWLEDGEMENTS

The authors would like to acknowledge the support of Dr. Li Li, Jaynal Molla, and Owen Fay for solder bumping the test wafers.

References

1. N.C. Lee, “Pb-free Soldering—Where the World is Going,” Advancing Microelectronics (Sept./Oct. 1999), p. 29.
2. A. Grusd, “Integrity of Solder Joints from Pb-Free Solder Paste,” Proc. NEPCON West ’99 (Norwalk, CT: Reed Exhibition Companies, 1999).
3. J. Glazer, “Metallurgy of Low Temperature Pb-Free Solder for Electronic Assembly,” Int. Mater. Reviews, 40 (2) (1995), pp. 65–93.
4. J. Glazer, “Microstructure and Mechanical Properties of Pb-Free Solder Alloys for Low-Cost Electronic Assembly: A Review,” J. Electron. Mater., 23 (8) (1994), pp. 693–700.
5. G. Whitten, “Pb-free Solder Implementation for Automotive Electronics,” Proc. 50th Electron. Comp. Tech. Conf. (Piscatawny, NJ: IEEE Publications, 2000), pp. 1410–1415.
6. E. Bradley, III and J. Hramisavljevic, “Characterization of the melting and wetting of Sn-Ag-X Solders,” Proc. 50th Electron. Comp. Tech. Conf. (in Ref. 5), pp. 1443–1448.
7. W.K. Choi and H.M. Lee, “Effect of Soldering and Aging Time on Interfacial Microstructure and Growth of Intermetallics Between Sn-Ag Solder and Cu substrates,” J. Electron. Mater., 29 (10) (2000), pp. 1207–1213.
8. F. Guo et al., “Effects of Reflow of Wettability, Microstructure, and Mechanical Properties in Pb-free Solders,” in Ref. 7, pp. 1241–1248.
9. T.S. Choi, K.N. Subramanian, and J.P. Lucas, “Thermomechanical Fatigue Behavior in Sn-Ag Solder Joints,” in Ref. 7, pp. 1249-57.
10. Y. Miyazawa and T. Ariga, “Microstructural Change and Hardness of Pb-free Solder Alloys,” Proc. 1st Int. Symp. On Environ. Conscious Design (Los Alamitos, CA: IEEE Comput. Soc., 1999), pp. 616–619.
11. M. Abtew and G. Selvardery, “Pb-free Solder in Microelectronics,” Mater. Sci. and Eng., 27 (2000), pp. 95–141.
12. H.K. Seelig and D. Suraski, “The Status of Pb-free Solder Alloys,” Proc. 50th Electron. Comp. Tech. Conf. (in Ref. 5), pp. 1405–1409.
13. K.G. Snowden, C.G. Tanner, and J.R. Thompson, “Pb-free Soldering Interconnects: Current Status and Future Developments,” Proc. 50th Electron. Comp. Tech. Conf. (in Ref. 5), pp. 1416–1419.
14. T.M. Korhonen et al., “Reactions of Pb-free Solders with CuNi Metallizations,” in Ref. 7, pp. 1194–1199.
15. J.C. Foley et al., “Analysis of Ring and Plug Shear Strengths for comparison of Pb-free Solders,” in Ref. 7, pp. 1258–1263.
16. M.W. Roberson et al., “Conversion between Standard and Low-alpha Pb in Solder Bumping Production Lines,” in Ref. 7, pp. 1274– 1277.
17. Z. Hasnain and A. Ditali, “Building-in Reliability: Soft Errors-A Case Study,” 30th Annual Proc. Reliability Physics (1992), pp. 276–280.
18. D.R. Frear and P.T. Vianco, “Intermetallic Growth Behavior of Low and High Melting Temperature Solder Alloys,” Metall. Trans. A, 25A (1994), pp. 1509–1523.
19. D.R. Frear et al., editors, Solder Mechanics: A State of the Art Assessment (Warrendale, PA: TMS, 1990).
20. W.J. Boettinger et al., The Mechanics of Solder Alloy Wetting & Spreading (New York: Van Nostrand Reinhold, 1993), Ch. 4.
21. H.K. Kim, K.N. Tu, and P.A. Totta, “Ripening-assisted Asymmetric Spalling of Cu-Sn Compound Spheroids in Solder Joints on Si Wafers,” Appl. Phys. Lett., 68 (16) (April 1996), pp. 2204–2206.
22. S. Chada et al., “Cu Substrate Dissolution in Eutectic Sn-Ag Solder and Its Effect on Microstructure,” in Ref. 7, pp. 1214–1221.
23. B. Roesner et al., “Thermal Fatigue of Solder Flip-Chip Assemblies,” Proc. 48th Electron. Comp. Tech. Conf. (Piscataway, NJ: IEEE Publications, 1998), pp. 872–877.
24. M. Harada and R. Satoh, “Mechanical Characteristics of 96.5 Sn/ 3.5 Ag Solder in Micro-bonding,” IEEE Trans. on Comp, Hybrids, and Manuf. Tech., 13 (4) (1990), pp. 736–742.
25. J.J. Stephens and D.R. Frear, “Time Dependent Deformation Behavior of Near Eutectic 60Sn-40Pb Solder,” Metall. Trans. A, 30A (1999), pp. 1301–1313.
26. Z. Mei and J.W. Morris, “Characterization of Eutectic Sn-Bi Solder Joints,” J. Electron. Mater., 21 (1992), pp. 599–607.
27. D.R. Frear, Constitutive Behavior of Pb-Free Solder Alloys, Sandia National Labs Report # SAND96-0037 (1997).
28. J. Liang et al., “Creep Study for Fatigue Life Assessment of Two Pb-Free High Temperature Solder Alloys,” Mater. Res. Soc. Symp. Proc., 445 (1997), pp. 307–312.
29. A. Grusd, “Integrity of Solder Joints from Pb-Free Solder Paste,” Proc. NEPCON West’99 (Norwalk, CT: Reed Exhibition Companies, 1999).
30. A. Grusd, “Lead Free Solders in Electronics,” Proc. Surface Mount Int. Conf. (Edina, MN: Surface Mount Technology Assoc., 1998), pp. 648–661.
31. D.R. Frear, D. Grivas, and J.W. Morris, Jr., “Microstructural Study of the Thermal Fatigue Failures in 60Sn-40Pb Solder Joints,” J. Electron. Mater., 17 (1988), p. 171.
32. D.R. Frear, D. Grivas, and J.W. Morris, Jr., “Parameters Affecting Thermal Fatigue Behavior of 60Sn-40Pb Solder Joints,” J. Electron. Mater., 18 (1989), pp. 671–680.
33. D.R. Frear, “Microstructural Evolution During the Thermomechanical Fatigue of Solder Joints,” The Metal Science of Joining, ed. M.J. Cieslak et al. (Warrendale, PA: TMS, 1992), pp. 191– 200.

D.R. Frear, J.W. Jang, J.K. Lin, and C. Zhang are with Interconnect Systems Laboratories at Motorola.

For more information, contact D.R. Frear, Interconnect Systems Laboratories, Motorola, Tempe, AZ 85284; (480) 413-6655; fax (480) 413-4511; e-mail darrel.frear@motorola.com.