Warpage, heat, and brittleness can cause huge reliability problems for expensive designs.
Key Takeaways:
Low-temperature solders are becoming increasingly attractive in the chiplet era because they promise substantially reduced package warpage while enabling the use of temperature-sensitive components such as silicon photonics, LED modules, and flex circuits.
These solders are mostly used today in mobile devices, wearables, camera modules, and for thin printed circuit boards, where warpage is a significant problem. Leading-edge HPC/AI applications that must withstand high current densities and significant thermal gradients are likely to stick with the tried-and-true SAC305 solder. Nonetheless, SAC305’s high thermal budget (235 to 250 °C reflow) is becoming increasingly incompatible with large, thin, heterogeneous packages with complex stack-ups.
Low-temperature solders based on tin-bismuth (Sn-Bi) alloys also reduce the carbon footprint. Offering a 150 °C reflow temperature — 70 °C lower than that of SAC305 solder — a switch to low-temperature solder can save an SMT line 57 tons of CO2 emissions annually, suggesting that the industry could prevent 35,000 to 50,000 tons of CO2 from entering the atmosphere each year. [1]
Part of the reason for the recent surge in interest in low-temperature solders is growing concern around thermomigration and electromigration in dense multi-chiplet packages. These problems feed off one another and often manifest in the solder joints, which are the weakest link in the interconnect chain.
Electromigration is the mass transport of metal atoms caused by the electron wind from current flowing through a conductor. Solder bumps are alloys containing a mixture of two or more metals. When current density is high enough, metal will diffuse in the direction of current flow, creating tiny hillocks downstream and leaving behind vacancies or voids. With enough electromigration, failures occur due to severe line thinning, which can cause opens. In addition, the hillocks that bridge adjacent lines can cause short circuits.
Thermomigration, on the other hand, is the movement of material driven by a temperature difference between two points in a material, typically causing atoms to migrate from the warmer to the cooler region (though there are exceptions). This is why hot spots in confined areas become much bigger problems than they would be without the backside PDN, such as between front-end devices and backside power delivery networks. In flip-chip BGAs of XPUs, high current densities cause Joule heating on the silicon side, leading to thermomigration in the underlying solder bumps. Multiple chip manufacturers have reported head-in-pillow, bridging, and non-wet open failures in FCBGAs of processors and memory chips. Low-temperature solder may hold the answer.
“One of the greatest advantages of LTS is its lower melting point, which results in reduced stress in semiconductor devices during and after assembly. The lower melting point also leads to significant energy and cost savings,” stated Nokibul Islam of STATS ChipPAC in a recent paper. [1] “Lower temperatures reduce the risk of warping sensitive components during soldering, thereby enhancing product integrity.”
Recent work testing the reliability of low-temperature solder reveals:
The industry’s path to low-temperature solders began over two decades ago when it set out to replace toxic lead in tin-lead-based solders.
It began with lead-free solders
The European Union’s RoHS (Restriction of Hazardous Substances) directive of 2006 drove the elimination of lead from solder joints in electronic assemblies worldwide (with high-reliability aerospace and military exceptions). The lead-free solder of choice became SAC305, a tin alloy with small concentrations of silver and copper (3% Ag, 0.5% Cu, by weight). Because SAC305’s melting temperature of 217 to 221 °C is much higher than the 183 °C melting point of lead-tin solder of the past, the solder replacement forced the adoption of higher temperature-resistant PCB and other materials.
Interestingly, tin-bismuth (Sn-Bi) based solders, with a reflow temperature of 150 °C, were considered at that time as a key contender. However, when Bi-Sn solders are contaminated with even a small concentration of lead, they form a stable ternary metal alloy. Because much of the equipment at the time had lead-based surface finishes, the risk of contamination was simply too high, and the Sn-Bi solder was abandoned as a possibility. [2]
How SAC305 and Sn-58Bi compare
An iNEMI (International Electronic Manufacturing Initiative) Consortium, comprising experts from IBM, Intel, Shinko Electric, Indium, and others, identified Sn58-Bi42 alloys as the most popular low-temperature, lead-free solder today. [2] Since 2015, this working group has been evaluating several material candidates to identify a low-temperature drop-in replacement for SAC305 (Sn with 3%Ag, 0.05%Cu) solder that offers the mechanical, metallurgical, and reliability properties needed for high-volume assembly operations. Bi-Sn solders already have one advantage — they are used at the PCB level for thin PCBs and for PCB rework, sparing all components from repeated exposure to higher reflow temperatures.
However, the industry hasn’t fully settled on a drop-in replacement. The biggest drawback of near-eutectic Sn-Bi-based solders is their tendency toward brittleness, which poses reliability risks. [A eutectic point is the lowest possible temperature at which a solid metal alloy melts, referring to that specific composition.] As a result of the brittle behavior, engineers are exploring small additions of other metals (silver, copper, nickel, antimony) to improve metallurgical properties and the robustness of solder bumps after they’ve undergone full processing and reliability testing, including thermal cycling, high-temperature storage, and drop-shock tests.
The iNEMI team performed an apples-to-apples comparison of Sn-Bi and SAC305 solder electromigration behavior, which showed SAC305 initiates electromigration at a resistance two orders of magnitude higher than that of Bi-Sn. [2] The team’s test vehicles represent bottom-terminated components with copper landing pads on the component and PCB (bottom). Solder bumps between the pads were reflowed at their respective temperatures; some joints were subjected to three EM current and temperature levels, while the remaining joints were measured for the effects of aging on the stressed joints.
When electromigration in the solder bump initiates, the electrical resistance increases in response to exposure to high currents and temperatures. Voids begin to nucleate and grow at the cathode interface (PCB side) of the solder joint. In the case of the Bi-Sn solder, bismuth atoms are driven from the cathode to the anode. Once a large void forms, the resistance rises more rapidly, precipitating in an electrical failure.
The electromigration behavior of SAC305 alloy is fairly well understood. There is a gradual rise in resistance during early EM, as Cu6Sn5 and Cu3Sn intermetallic compounds migrate from the cathode toward the anode. During electrical stress, because copper diffuses rapidly in tin, the cathode quickly becomes more tin-rich, while the anode becomes more copper-rich, forming intermetallic compounds and increasing the overall resistance of the solder. Voids form at the cathode and expand along the intermetallic/solder interface, merging to form a long crack. Resistance rises, and the crack causes an electrical open.
Whereas EM in SAC305 is dominated by the movement of copper atoms, electromigration in Bi-Sn solder is dominated by the migration of bismuth atoms. The first step involves the coarsening or growth of bismuth and tin islands, resulting in an initial dip in resistance. Next, bismuth migrates to the anode, forming a continuous layer that thickens as bismuth atoms substitute for tin atoms. Resistance rises linearly and then tapers off when no more bismuth is available to migrate.
Testing of Bi-Sn with additions of silver (Ag), copper (Cu), nickel (Ni), and antimony (Sb) was performed on both near-eutectic Bi-Sn and hypoeutectic, containing 40% bismuth and 60wt% tin. [3] The iNEMI group chose a planar geometry this time, testing solder behavior at three temperatures and three different current levels.
Fig. 1: Alloys tested for electromigration behavior. An Arrhenius plot using this data can be used to predict solder lifetimes for any field condition. Source: IBM/iNEMI
Interestingly, the hypoeutectic solders (Sn60Bi40Cu0.5Sb0.03 or Sn60Bi40Cu0.5Ni0.3Sb0.5) exhibited better electromigration behavior, but not due to alloying effects. “The hypoeutectic Sn-Bi alloys have measurably lower electromigration rates compared to the eutectic Sn58Bi alloy. The alloying additions had negligible effects on electromigration. The hypoeutectic solders had lower electromigration rates because they had less Bi content, Bi being the element dominating the electromigration,” stated the authors.
Higher reliability with an alternative soldering method
A growing reliability problem with BGA packages is solder cracking. Caused by the combination of significant differences in the coefficient of thermal expansion (CTE) between the BGA package and the PCB, and large package sizes, cracking can result from repeated thermal and voltage stress. A group from Samsung Electronics, led by Myeonng-Hyeon Yu, compared a low-temperature solder with a melting point of 131 °C to standard SAC305, processed either in a standard reflow oven or illuminated by intense pulsed light (IPL). “When light is irradiated on the package, heat is directly generated from the package through the photothermal effect. Due to these mechanisms, preheating of the equipment is not required, and soldering can be performed in a short time. Also, power efficiency is high because there is less heat loss compared to reflow,” the authors stated.
The Samsung group compared not only solder ball materials, but also low-temperature solder paste versus SAC305 paste. After reflow, the samples were subjected to repeated thermal cycling from 0° to 125 °C for 40 minutes per cycle, with SEM cross-sections taken to assess damage. In the case of the SAC305 solder and paste, the IPL joints failed after 1,900 cycles, whereas the reflowed joints failed after 1,200 cycles. The IPL joints featured longer crack paths, which took longer to break. “The size of the Ag3Sn precipitates in the solder joint of the IPL soldering was smaller than that of reflow soldering. It is believed that the TC (thermal cycling) reliability has been improved through the dispersion strengthening effect that hinders the progression of cracks as Ag3Sn precipitates are densely formed. In the case of IPL soldering, the time above the melting point is three times shorter, and the initial cooling rate is twice as fast, so Ag3Sn precipitates are densely formed. Therefore, the IPL soldering shows higher TC reliability compared to reflow.”
The low-temperature solder paste with the SAC305 joints (hybrid joints) also performed better with IPL soldering. By increasing the peak temperature in these hybrid joints, bismuth diffuses into the SAC305 balls, thereby increasing grain density and reliability. Compared with the full low-temperature solder joints, the CTE mismatch in the hybrid joint is concentrated in the lower paste area, whereas in the full LTS joint, deformation is distributed throughout the solder joint. The IPL joints feature smaller bismuth precipitates in the IPL-soldered samples than in the reflowed joints, which apparently reinforce the solder joint structure.
Selecting a solder for photonics applications
Photonics applications, fueled by growth in AI data center buildouts, are especially sensitive to high process temperatures. In particular, the peak reflow temperature of solder plays a key role in the performance of silicon photonics devices.
A recent reliability study by engineers at STATS ChipPAC, JCET, and Jiashan Fudan Institute, led by Nokibul Islam, compared the microstructure and reliability behavior of Sn-Bi-Ag and Sn-Bi-Ag-Ni solders on copper and gold pads to identify a low-temperature solder suitable for photonics applications. “The selection criteria for LTS include mechanical reliability, workability, environmental impact, supply chain considerations, and cost,” stated the authors.
“Photonics devices must withstand temperature cycling of -40 °C to 85 °C, resulting in coefficient of thermal expansion (CTE)-induced stress on all component interconnections.”
The STATS ChipPAC team investigated the formation of intermetallic compounds in Sn-BiAg and SnBiAgNi solders during the flip-chip process, thermal cycling, and high-temperature storage reliability testing. They benchmarked against SAC305 processing. The test vehicle consisted of a chip-to-chip chip-scale package, where the top die has the plated Cu pad as UBM/surface finish, and the bottom die has attached solder balls. Thermal cycling was performed on four top and bottom unbumped metal/surface finish combinations, CuSnAg/CuNiAu and Cu/CuNiAu, using a slow cooling rate to minimize CTE differences.
Results showed that IMCs grow more slowly on the Cu/NiAu interface than on the copper interface due to nickel suppression. Compared to SnBiAgNi, SnBiAg exhibits a slightly more homogenous distribution of bismuth across all process flows (chip-chip solder, high temperature storage, and thermal cycling). The low-temperature solder produces significantly thinner IMC compared to SAC305. All test vehicles passed reliability testing. The authors concluded that intermetallic compound growth is a crucial factor in assessing solder joint reliability.
Conclusion
While SAC305 solder is clearly the established workhorse for the industry, low-temperature solder based on tin-bismuth alloys is being widely considered as a strategic alternative when SAC305 failures cannot be overcome. These include flip-chip BGA failures such as head-in-pillow, bridging, and non-wet open failures, which are becoming increasingly problematic as package sizes grow, CTE differences between the package and the PCB increase, and the number of connections now exceeds 1,000.
Low-temperature solders, especially SnBiAg alloys, offer dramatically lower reflow temperatures, reducing stress on the silicon, reflow costs, and environmental footprint. The solders can enable the assembly of thinner, larger, and more thermally sensitive packages, while reducing package warpage, another significant cause of failures.
References
Related Stories
https://semiengineering.com/novel-assembly-approaches-for-3d-device-stacks/
https://semiengineering.com/challenges-grow-for-creating-smaller-bumps-for-flip-chips/
 |
 |
 |
 |
 |
 |
 |  |
 |
 |
 |
Leave a Reply