Assembly & Reliability of Flip Chip-on-Laminate with Lead Free Solder
Zhenwei Hou, Guoyun Tian, Casey Hatcher and R. Wayne Johnson, Laboratory for Electronics Assembly & Packaging – Auburn University, 162 Broun Hall, ECE Dept., Auburn, AL 36489 USA, 334-844-1880, johnson@eng.auburn.edu, Erin Yaeger, Mark Konarski & Larry Crane, LºCtite Corporation, 1001 Trout Brook Crossing, RºCky Hill, CT 06067 USA, 860-571-2599, Larry_Crane@lºCtite.com
Abstract
The move to replace lead in electronic assemblies continues to gain momentum. The driving factors are potential lead banning legislation, primarily in Europe, recycling laws in Japan and global market pressures for more environmentally friendly products. After more than 10 years of research, a ‘drop-in’ solder alloy replacement for eutectic Sn-Pb has not been found. While a number of alloys are still being discussed, the current industry trend is to use the eutectic (or near eutectic) Sn-Ag-Cu alloy. The melting point of this alloy (~217ºC) is significantly higher (34ºC) than eutectic Sn-Pb (183ºC). This higher melting temperature will impact the assembly process. In addition, the lead-free alloys have a higher modulus than eutectic Sn-Pb. This may change the stress distribution during thermal cycle testing, affecting reliability and failure modes. This paper examines the assembly process for flip chip die with Sn-Ag-Cu solder bumps and on-going reliability testing.
Key Words: Flip Chip, Lead-free, Assembly, Reliability
Introduction
The elimination of lead in electronics assembly has been discussed since 1990. Initially, the driving force was a proposed legislative ban in the U.S. At the time no solder alloy replacement was identified and the legislation was dropped under strong pressure from the electronics industry. However, increasing restrictions on hazardous materials in landfills, recycling requirements and manufacturer responsibility for products from ‘cradle-to-grave’ have kept the topic of lead in the mind of manufacturers. Today, with proposed legislation in Europe and global competitive market pressures, particularly in Japan, the elimination of lead in many, if not all, electronic products appears imminent. The successful introduction of electronic products assembled with lead-free solders belies the arguments that it can not be done. However, just because some assembly types and reliability requirements can be satisfied with lead-free solder, does not translate to all products and all reliability requirements. Much work remains to be done.
One of the first challenges to the industry was the selection of a replacement solder alloy. The National Center for Manufacturing Sciences (NCMS) concluded in 1997, after a major four year research effort that there were no ‘drop-in’ replacements for eutectic Sn-Pb [1]. The International Tin Research Institute (ITRI) [2] and the National Electronics Manufacturing Initiative [3] are both recommending the Sn-Ag-Cu eutectic (or near eutectic) alloy for reflow solder applications. Momentum does appear to be building for this alloy selection. The Sn-Ag-Cu eutectic has a melting point of ~217ºC, significantly higher than eutectic Sn-Pb (m.p. 183ºC). This will require higher peak reflow temperatures that may in turn impact fluxes and flux residue. For flip chip applications, the interaction between the underfill and the flux residue may degrade thermal cycle and thermal shock reliability. The higher Young’s modulus of the Sn-Ag-Cu alloy (46 GPa versus 33 GPa for eutectic Sn-Pb) will also alter the stress distribution on the assembly, potentially impacting reliability and failure mechanisms.
This paper examines the assembly process for flip chip die with Sn-Ag-Cu alloy bumps and thermal shock testing. Die with eutectic Sn-Pb solder were used for controls in the experiments.
Assembly Study
A PB8 test vehicle was used to develop the assembly process. The PB8 die is 5.1mm x 5.1mm with solder bumps on a 204µm pitch. Die with eutectic Sn-Pb and Sn/3.5Ag/0.7Cu solder balls were used. The test board was a four-layer construction with ten die sites. The design has a trench in the solder mask to define the solderable pads for flip chip attachment. The board surface finish was electroless nickel/immersion gold.
Two commercial fluxes were used in the assembly evaluation, a tacky, no-clean flux and a low residue, no-clean flux. The tacky flux was applied using the rotating flux dipping station on the Siemens F5 pick and place system. The depth of the flux film on the rotating drum and the dip time controlled the tacky flux volume transferred to the solder bumps. Flux depth was varied in the assembly experiments from 35µm to 65µm to study the role of tacky flux volume on assembly yield. For the eutectic Sn-Pb solder bumps, a flux depth of 35µm was used.
The low residue flux was sprayed onto the board prior to die placement. The application of low residue flux was not varied in this experiment, but will be studied in a subsequent series of experiments. All die were placed with the Siemens F5.
The effect of peak reflow profile temperature on assembly yield was investigated. Thermocouples were placed under the flip chip die between the die and the board surface. Peak temperatures of 235ºC, 245ºC and 255ºC were used in the experiment. The test vehicles were reflowed in a nitrogen atmosphere (<25ppm O2) for all flux and reflow temperature combinations using a Heller 1700 oven. The optimum nitrogen reflow conditions (flux and reflow temperature) were evaluated for reflow in air.
After reflow, die were electrically tested. All test cells reflowed in nitrogen had 100% electrical yield. Samples were mounted and cross-sectioned. Figures 1 and 2 show cross sections of Sn-Ag-Cu solder joints with tacky, no-clean flux dip depths of 35µm and 65µm reflowed at 245ºC, respectively.
With a flux depth of 35µm, the solder wets the top of the PWB pad and there is evidence of some wetting of the pad sidewalls. Increasing the flux depth to 65µm, yielded good wetting to most pads at peak reflow temperatures of 245ºC and 255ºC. However, some random pads still did not completely wet (Figure 2). It appears that a minimum flux depth is required and with the variation in ball heights, some solder balls did not receive sufficient flux even with a 65µm flux depth. The degree of wetting was worse for all flux depths when reflowed with a peak of 235ºC. There were no quantifiable difference between reflow profile peaks of 245ºC and 255ºC.
The low residue, no-clean spray flux achieved good wetting with the Sn-Ag-Cu solder alloy at peak reflow temperatures of 245ºC and 255ºC (Figure 3). When reflowed at 235ºC, electrical connectivity was achieved, but wetting of the pad sidewall was not complete (Figure 4).
The electrical yield was poor with both the tacky, no-clean dip flux (65µm depth) and the low residue, no-clean spray flux with the Sn-Ag-Cu solder alloy reflowed at 245ºC in air. Thus, with the two fluxes chosen, nitrogen reflow is required with the Sn-Ag-Cu solder alloy.
Assembly for Thermal Shock Testing
For thermal shock testing, the test vehicles were fabricated using the low residue, no clean flux and the tacky, no clean dip flux (35µm for Sn-Pb and 65µm for Sn-Ag-Cu). The reflow profile peak temperature was 245ºC for all alloys and fluxes. The reflow was in a nitrogen atmosphere (<20ppm O2).
After reflow assembly, the die were underfilled with Loctite 3563, a fast flow, snap cure underfill. The die were underfilled using a Camalot 3700 dispense system. The substrate temperature was 95ºC. After dispensing, the underfill was cured in a conveyor oven with a peak temperature of 165ºC for 5 minutes. A Sonix scanning acoustic microscope was used in C-mode to inspect the underfill. (Figure 5). No voids were observed.
Thermal Shock Testing
A total of three liquid-to-liquid thermal shock tests have been conducted to evaluate the reliability of the underfilled test vehicles. In the first test, two boards (20 die) with each flux (dip and spray) and solder alloy (Sn-Ag-Cu and Sn-Pb) were tested. The cycle was from -40ºC to +125ºC with 5 minutes at each extreme and a 1-minute transition time. The resistance of the daisy chain was monitored in-situ to accurately determine cycles-to-failure. The test was terminated at 4700 cycles. The results of this first test are shown in Figure 6. The die with Sn-Ag-Cu solder balls failed somewhat earlier than the Sn-Pb. The difference between flux type is more evident with the Sn-Ag-Cu solder flip chips.
After the 4700 thermal shock cycles, the test vehicles were removed for failure analysis. Figure 7 compares the C-SAM images for the different fluxes and solder alloy combinations. There is little delamination with the Sn-Pb solder (slightly more with the spray flux compared to the dip flux). However, there is significant delamination with the Sn-Ag-Cu alloy.
Typical cross sections of failed solder joints with fatigue cracks are shown in Figures 8 and 9. There was no indication that the limited wetting of the Sn-Ag-Cu solder along the pad sidewalls degraded thermal shock performance. The observed solder cracks were not at the PWB pad-to-solder interface. Cracks in the PWB at the edge of the copper pad were observed in many cross sections. Some PWB cracks contained extruded solder.
Figures 10 and 11 are typical flat sections (the die has been polished away) showing shorts in the Sn-Pb and Sn-Ag-Cu samples. The monitoring system would not detect shorts. In Figure 12, a near short is observed in cross section with the Sn-Ag-Cu alloy. This appears to be solder extrusion into the delaminated region between the underfill and the die. As previously shown, there was significant delamination with the Sn-Ag-Cu alloy. The number of shorts after 4700 cycles as determined by flat sections was significantly higher for the Sn-Ag-Cu alloy. There was little delamination with the Sn-Pb alloy and cracks in the bulk underfill between adjacent solder balls and solder extrusion into the cracks was observed.
The coefficient of thermal expansion (CTE) of the underfill is 35ppm/ºC below the glass transition temperature (130ºC) compared to 23ppm/ºC for the solder alloys. Thus, on the cold side of the thermal shock cycle the underfill tries to contract more than the solder ball. With cycling, the underfill can delaminate (adhesion loss) or crack (cohesive failure). The modulus of elasticity of the solder alloy will impact the stress on the underfill. With the Sn-Ag-Cu alloy there was significant delamination with both fluxes. With the Sn-Pb solder there was little delamination (only localized spots) and some bulk cracking between solder balls with both fluxes.
To evaluate the shorting phenomena, a new test vehicle was designed. The routing pattern of the new board created two independent daisy chains for each test die. Shorts between the two daisy chains could be monitored. The observability provided by this design was 50%. For any solder ball, the adjacent solder ball was either part of the other daisy chain (observable short) or part of the same daisy chain (non-observable short). Two bond pad designs (trench and finger) were used (Figure 13).
Trench design test boards were assembled for the second thermal shock test (Test A) using the same assembly process previously described. The test was stopped at 1000 cycles due to an unexpected number of failures in the Sn-Ag-Cu/tacky, no clean dip flux group. The data is plotted in Figure 14 and labeled as Pb Free Dip Flux (Test A). A typical cross section of a failed solder joint with a fatigue crack propagating through the bulk of the solder joint is shown in Figure 15.
For the third assembly (Test B), the finger design was used with the same assembly process. The results are plotted in Figure 14 and labeled as Test B. This test was terminated after 3200 cycles for failure analysis.
The Test A and Test B results for the Sn-Ag-Cu alloy with the tacky, no clean dip flux show no significant difference, indicating there is little difference in the two builds. However, the results are significantly worse compared to the original test results. The results for the Sn-Ag-Cu with low residue, no clean spray flux and the Sn-Pb with both fluxes are also worse than in the original test, but not as dramatically. The cause of this difference is under investigation, but the only obvious change was in the test board itself.
Figures 16 - 19 show typical C-SAM images for each alloy and flux before thermal shock cycling and after 3200 cycles. As before, there was significantly more delamination with the Sn-Ag-Cu alloy and either flux.
Figures 20-23 are typical cross sections of failed solder joints for each flux and alloy combination after 3200 thermal shock cycles. The figures show fatigue cracks through the bulk of the solder on the die side.
Figures 24 and 25 are flat sections (the die has been polished away) showing cracks in the underfill and extrusion of solder. The solder extrusion has not caused a short at this point. When the test was terminated at 3200 cycles, no shorts had been recorded. However, the flat sections do indicate shorting was imminent.
Summary
With the tacky, no clean dip flux selected, more flux is required with the Sn-Ag-Cu solder alloy compared to the Sn-Pb eutectic. Some poor wetting joints still occur with a flux dip depth of 65µm. It is believed variation in ball height is responsible for the few poor wetting solder joints. Additional studies are required to determine the minimum low residue, no clean spray fluxed required.
The flip chip die with Sn-Ag-Cu solder bumps fail earlier in thermal shock cycling and exhibit more underfill delamination than die with Sn-Pb eutectic bumps. In this series of experiments, the reliability of all alloy and flux combinations decreased with the change in PWBs. However, the flip chip die with the Sn-Ag-Cu alloy bumps were more sensitive to the change. Further studies are underway to understand the origin of this decrease in reliability.
Finally, while underfill cracks were observed between adjacent bumps with the Sn-Pb alloy, the extrusion of solder and formation of shorts did not occur before solder joint opens were detected. For finer pitch flip chip die, shorts may occur before opens.
References
1."National Center for Manufacturing Sciences Lead-Free Solder Project Final Report," National Center for Manufacturing Sciences, Ann Arbor, MI, August, 1997.
2. Kay Nimmo, "Worldwide Environmental Issues In Electronics And The Transition To Lead-free," Proceedings of the IPC Lead-Free Summit, Minneapolis, October 1999.
3. NEMI News Release, January 24, 2000, www.nemi.org/PbFreePUBLIC/index.html
| Table
of Contents | Previous
Page | Next
Page |
| 
