Six-Micro-Meter Microbumps

In order to study different EM failure modes using Kelvin bump structures, 6-μm microbumps and 10-μm microbumps were applied in this study. The 6-μm microbumps are introduced in section 2.2 while the 10-μm microbumps are described in section 2.3 to be introduced. The measurements of “6 μm” and “10 μm” indicate the solder height in the microbumps. In the 6-μm microbump, the height of solder ranged from 6 μm to 8 μm; in the 10μm microbump, the height of solder ranged from 10 μm to 12 μm.

2.2.1. Sample Structure

Microbumps of Sn2.5Ag with Cu/Ni UBM were selected for EM tests. Since the size of a microbump was much smaller than that of a flip-chip bump, precise alignment during bonding was required. Therefore, the chip-on-chip (COC) technology was used. The Si of a smaller area and on top of the microbump was still called “Si chip” or “chip,” while the bottom Si of a larger area was called “Si interposer” or “interposer.” The thickness of both Si chip and Si interposer were 760.0 μm. The area of the Si chip is 4.5 mm×4.5 mm; that of the Si interposer is 16.0 mm×16.0 mm. The UBM comprised a 100-nm thick Ti adhesion layer, a 300-nm thick Cu seed layer, a 5.0-μm thick electroplated Cu layer and a 3.0-μm thick electroplated Ni layer on both Si chip and Si interposer sides as shown schematically in Figure 2-3 (a). The Al traces deposited by sputtering on both sides were 10 μm wide and 0.8 μm thick. The diameter of the microbump and Al pads were 18 μm and 20 μm, respectively; and the pitch was 60 μm. The diameter of Cu UBM was slightly smaller than that of Ni UBM due to the undercut during the fabrication process. On the Si chip side, Sn2.5Ag solder was electroplated on the Ni UBM, and the chip was bonded

-37-

to a Si interposer by thermo-compression at 260ºC.

As shown in Figure 2-3 (b), the fabricated microbump has a solder height of 6.2μm, and the Ni3Sn4 IMC of about 1.0 μm thick is formed on both top and bottom interfaces between the UBM and the solder. Furthermore, some small precipitated Ag3Sn particles can be found dispersed in the Sn2.5Ag solder. The total thickness of the UBM layers is approximately 16.0 μm; thus, the volume of the UBM is larger than that of the solder by over 2.5 times. This difference between a traditional flip chip solder joint and a microbump is expected to affect significantly EM behavior in the microbump significantly. In particular, in this 6-μm microbump, the low bump height caused the joint to be much more easily transformed into IMC joint. The Ni3Sn4 IMC is a kind of material that contains better EM resistance than the solder. As a result, the 6-μm microbump shows totally different failure modes during EM test.

-38-

Figure 2-3 (a) Schematic illustration and (b) cross-sectional SEM image of a 6-μm microbump.

-39-

2.2.2. Kelvin Bump Structures and Experimental Procedures

Figure 2-4 is the plane view of Kelvin bump structures in the low-bump-height microbump samples. The solid lines represent the layout of Al traces on the Si chip, and the gray-colored regions represent the traces on the Si interposer. In this structure, 16 nodes were marked as n₁, n₂ … to n₁₆, and 9 microbumps were marked as b₁, b₂ ...

to b9. There were also three traces forming a direct links between b4 and b7 and marked as t_4-5, t_5-6, and t_6-7. With this Kelvin bump structure, the bump resistance of microbumps, b5 and b6, can be in-situ monitored during various kinds of reliability tests. In addition, the bump resistances at different angles of b6 can also be obtained.

If a current was applied from n10 to n7, the electron flows through n7, b5, t5-6, b6, and finally n10. The direction of electron flow is upward (interposer to chip) in b5 and downward (chip to interposer) in b6. The methods for measuring the microbump resistances of b5 and b6 are listed in Table 2 3. The basics of measurement for the 6-μm microbump sample were quite similar to those for the flip-chip bump sample, but the bump resistance at different angles provided more information about the behavior of current crowding in the microbumps. In these samples, the 0º position is defined as the location that the trace-conducting current contacts the Al pad. The angle becomes larger with increasing distance of the edge of Al pad from the entrance of electron flow. That is, the 180º position is located at the opposite position across the Al pad.

Two different currents, 0.12 A and 0.24 A, were applied to a pair of microbumps placed on a hot plate maintained at 150ºC. The corresponding current density were 4.6

 10⁴ A/cm² and 9.2  10⁴ A/cm². The bump resistance of b₅ and b₆ at 0º and 180º were simultaneously monitored throughout the test. After the resistance reached some certain values, current stressing was terminated and the samples were polished for

-40-

microstructure analysis using a scanning electron microscope (SEM). Compositional analysis was performed by energy dispersive spectroscopy (EDS). However, before applying a high current density in the EM test, a small current, 2 mA, was employed to measure the bump resistance at all different angles. The corresponding current density is 7.86 A/cm², and the test with small current was also conducted on a 150ºC hot plate. The small current could prevent the occurrence of Joule heating. In this case, all seven resistances were simultaneously monitored (two in b₅, four in b₆, and one in t5-6). Each resistance was monitored every 20 seconds for 10 minutes and the obtained values were then averaged to eliminate the influence of temperature fluctuation.

Though this measurement with small current could not produce any microstructure evolution, it helped a lot in analyzing the current crowding behavior in microbumps.

-41-

Figure 2-4 Kelvin bump structures in a 6-μm microbump.

b₅ (e^- ↑) t5-6

b₆ (e^- ↓) Angle ( ° ) 0 180 0 60 120 180

V⁺ (node) 3 5 14 11 2 15 12 V^- (node) 6 9 3 14 1 16 13

Table 2-3 Nodes used for measuring resistances of a 6-μm microbump with current applied from n10 to n7.

-42-