Lead-Free BGA Assembly
Editor's note: This article was adapted from a paper that was originally presented at APEX, which was held Feb. 24-26, 2004, in Anaheim, CA.
Though development work on lead-free components is increasing, research on lead-free components with high I/0 counts has been fairly limited.
Could a flip-chip ball-grid array (FCBGA) with 780 I/O be successfully assembled with lead-free solder at a peak reflow temperature below 260 C? How would lead-free solder affect joint reliability? How would it affect rework? We conducted a study to find out.
We mounted five FCBGAs to a circuit board designed for reliability testing. The FCBGA was 29 millimeters square with a two-piece copper heat sink. The solder balls, which measured 0.6 millimeter in diameter, were made of a tin-silver-copper alloy. The board was made of high-temperature FR4. It was 14 inches long, 4 inches wide and 93 mils thick. The surface finish was organic solder preservative, and the pads were 0.55 millimeter in diameter.
To solder the FCBGAs, we used a Type 3 lead-free, no-clean solder paste composed of 95.8 percent tin, 3.5 percent silver and 0.7 percent copper. The paste was applied using a standard printer with metal squeegee blades. The print speed was 20 millimeters per minute. The metal stencil was 6 mils thick. The stencil apertures were 22 mils in diameter, and the ratio of pad size to aperture size was 1-to-1.
The FCBGAs were put onto the boards using standard fine-pitch placement equipment.
The boards were reflowed in a 10-zone convection oven, in either air or nitrogen. Three reflow profiles were tested, with peak reflow temperatures of 225 C, 235 C and 245 C. The time that the boards were exposed to temperatures over 217 C ranged from 50 to 70 seconds. The difference between the peak temperature at the solder joints and the peak temperature at the top of the component was typically 2 to 3 C.
After reflow, we ran all the boards through visual inspection, X-ray inspection and electrical continuity testing. In addition, certain boards were selected for cross-sectional microstructure analysis of the soldered joints. Based on the results of these tests, we decided to limit first-pass assembly and reliability testing primarily to those boards that were reflowed with a peak temperature of 235 C and a nitrogen atmosphere.
Solder joint reliability was assessed by running the boards through an accelerated temperature cycling test, according to the specifications outlined in IPC-9701. Twice per hour, the boards were alternately exposed to 10 minutes at 0 C and 10 minutes at 100 C. The boards were heated or cooled at a rate of approximately 12 C per minute.
Thermal cycling identifies joint failures caused by mismatches in the coefficients of thermal expansion of the solder joint, component substrate and circuit board. Failures occur in the solder joint at points of maximum stress.
Prior to thermal cycling, the FCBGA assemblies were tested for electrical continuity. Because the devices were mounted to a specially designed circuit board that contains half of the daisy chain link, a string of inverters can be used to complete the daisy chain. A clock signal can then be sent and monitored by a PC or logic analyzer.
As a result, electrical continuity testing can be performed on all the components before-and during-thermal stress testing. The exact time to failure can be measured, and the failure location can be quickly and easily identified.
Reworking Lead-Free BGAs
Some of the boards that were reflowed at a peak temperature of 235 C were selected to test rework procedures.
An inherent problem with reworking lead-free BGAs is the high temperature gradient from the top of the package to the center of the solder joints during reflow. The high melting point of the lead-free alloy adds to the difficulty of removing and mounting these components. Too high a temperature will damage the device.
The package volume of our test component was more than 350 cubic millimeters, which is considered large by the JEDEC 020B standard. Rework profiles were developed to minimize the temperature gradient, while maintaining a peak reflow temperature of 235 C.
A standard convection rework machine for BGAs was used. A nozzle applied localized convection heat directly to the top of the package. The bottom of the board was also heated. Heating the package from both the top and the bottom provides better thermal control. However, even with dual heating, the temperature difference between the solder joints and the top of the package was still more than 20 C. Thus, a rework profile with a peak solder-joint temperature of 235 C would require the package to be heated to more than 255 C.
We designed a special nozzle to overcome the high temperature gradient between the solder joints and component. It differed from a standard rework nozzle by using a baffle to prevent direct heating of the top of the component. Instead, the heat flows toward the outer edges and sides of the FCBGA.
Thanks to our custom-made nozzle, we developed a reflow profile that produced peak temperatures of 244 C at the top of the package and 229 to 233 C at the center of the solder joints. The temperature difference between top of the component and solder joints ranged from 11 to 15 C.
After establishing a reflow profile, we reworked several of our FCBGA boards in a nitrogen atmosphere using the same solder paste as we used for initial assembly.
Once the component was removed, the pads were redressed prior to placement of a new FCBGA. We used a soldering iron, a copper wick and a no-clean liquid flux dispensed from a pen to remove residual solder and provide a smooth, level solder finish on all pads. The redressed pads were cleaned with isopropyl alcohol to remove excess flux. Lead-free solder paste was printed onto the board using a miniature stencil. The new component was then placed onto the pads, and the paste was reflowed.
After rework, the FCBGAs underwent visual inspection, X-ray inspection, electrical continuity testing and thermal testing, just as the first-pass boards were.
Visual inspection of the FCBGAs revealed that all solder joints appeared very similar, regardless of peak soldering temperature or atmosphere type. An FCBGA assembled with lead-free solder, at a peak temperature of 235 C, in a nitrogen atmosphere, will have a mixture of dull and more uniformly shiny solder joints. In contrast, the joints on BGAs assembled with tin-lead solder are typically more shiny.
X-ray inspection showed that lead-free solder did not increase the chances of bridging or voiding, regardless of peak reflow temperature or atmosphere. And, all our test boards passed electrical continuity tests.
Cross sections were taken of several boards prior to reliability testing. Sections were studied from three locations along a row of solder balls: the corner of the component, the edge of the silicon die, and the center point of the component.
Energy-dispersive X-ray analyses confirmed that a copper-tin intermetallic compound formed on the board side and a tin-copper-nickel intermetallic compound formed on the component side. The thicknesses of the intermetallic compound on the board side were similar for all cross-sections, ranging from 1.9 to 2.5 microns. On the component side, the thickness of the intermetallic compound ranged from 0.9 to 2.53 microns.
Visual and X-ray analysis did not show an advantage of using one reflow profile over another. Macro images of the cross sections also did not show any significant distinction, nor did the thermal cycling tests. All boards were failure-free at up to 3,500 cycles.
In the end, we decided to reflow most of our test boards in a nitrogen atmosphere at a peak reflow temperature of 235 C, because of surface-mount manufacturability considerations. This temperature makes the most of process improvements that can be gained by reflowing the boards in a nitrogen atmosphere. In addition, it's the lowest reflow temperature that is far enough away from the melting point of the tin-silver-copper solder paste and balls (217 C) to not cause process parameter concerns.
For rework, lead-free FCBGAs were reflowed at temperatures of 230 C and 235 C, in a nitrogen atmosphere.
Visual inspection revealed no measurable differences in appearance between the solder balls on reworked components and solder balls on new assemblies. However, reworked solder joints on lead-free components may have a more cratered surface appearance than solder joints on new, lead-free FCBGAs.
X-ray inspection indicated that rework with lead-free solder paste did not increase the risks of bridging or voiding, compared with new assemblies. No anomalies were detected.
All the reworked, lead-free components passed electrical continuity tests.
A reworked board was cross-sectioned prior to thermal testing. At approximately 3 microns, the thickness of the intermetallic compound on the board side was similar to that of new assemblies. Energy-dispersive X-ray analysis indicates that the intermetallic compound consists of copper, nickel and tin.
All reworked components passed identical thermal cycling tests as new assemblies.