Three types of samples were used in this thsis. The first type was actual high-lead solder bumps onto the eutectic SnPb pre-solder reflowed over Ball Grid Array (BGA) substrates.
The second type was actual eutectic solder bumps reflowed over Ball Grid Array (BGA) substrates and under-bump-metallization (UBM). The third type was electroplated metal/high-lead solder/metal multi-layer diffusion couples. Samples of the second type did not have to go through reflow because the as-electroplated layers were already in intimate contact.
Fist type: Composite solder flip chip
A schematic drawing of the flip-chip structure used in this study is shown in Fig. 2.1.
On the chip side, an outer under-bump-metallization (UBM) layer composed of 3 m of Ni was in direct contact with a 95Pb5Sn solder bump. The 95Pb5Sn solder was electroplated onto the Ni layer and then reflowed at 350 °C to form a spherical bump with a nominal diameter of 95 m.
The top metal layer on the soldering pads was electroplated Cu. The pre-solder was screen-printed over this Cu layer. The mass ratio of the 95Pb5Sn solder in the solder bump relative to the 37Pb63Sn content in the screen-printed solder paste was 16:5. In this mass ratio, these two solders, when completely mixed, will from in an alloy with an overall composition of 81Pb19Sn.
Reflow state
The flip-chip assembly in this study was reflowed between one and ten times. The reflow during which the 37Pb63Sn solder paste first became molten and the substrate was bonded to the chip was considered the first reflow. The reflow used to form the 95Pb5Sn solder bump was not counted. Each reflow lasted 115 s, during which time the solder was molten and the peak reflow temperature was 240(±2) °C. The nominal ramp and cooling rates were both 1.5 °C/s.
Aging state
After assembly, the samples were subjected to high-temperature storage at 100, 130, 150, and 175 °C for times as long as 4000 hr.
Second type: Eutectic SnPb flip chip
A schematic drawing of the flip-chip structure used in this study is shown in Fig. 2.2.
On the chip side, an outer under-bump-metallization (UBM) layer composed of 3 m of Ni was in direct contact with a eutectic SnPb solder. The spherical bump was a nominal diameter of 95 m. The top metal layer on the soldering pads was electroplated Cu.
Reflow state
The flip-chip assembly in this study was reflowed between one and ten times. Each reflow lasted 115 s, during which time the solder was molten and the peak reflow temperature was 240(±2) °C. The nominal ramp and cooling rates were both 1.5 °C/s.
Aging state
After assembly, the samples were subjected to high-temperature storage at 100, 130, 150, and 175 °C for times as long as 4000 hr.
Temperature cycling test
The flip-chip packages were then subjected to the temperature cycling test. The temperature profile for each cycle consisted of a 15 min ramp-up from -55 to 125 °C, a 15 min constant temperature at 125 °C, a 15 min ramp-down from 125 to -55 °C, and a 15 min constant temperature at -55 °C. Taking into account the coefficients of thermal expansion for Si (2.6 ppm/oC) and the organic substrate (15 ppm/oC) [JAN], one calculated that this temperature difference between the high dwell temperature (125 oC) and the low dwell temperature (-55 oC) generated a maximum displacement of 12.6 m between the two corner joints along the diagonal direction of the die, as illustrated in Fig. 2.2 (a). This displacement value translated into a maximum shear strain of 12.6/94=0.13 for the corner joints.
In order to observe the microstructural evolution, some of the flip-chip packages were cross-sectioned along the center of the first-row solder joints, i.e. along line aa in Fig. 1 (a).
These packages were then underwent the TCT for a specific number of cycles, and then examined under an electron microscope. After observation, these same packages were returned to TCT for further test and subsequent observation. This kind of treatment was referred as in-situ TCT in this article.
Third type: Ni/solder/Cu diffusion couples
Samples of the second type were prepared by electroplating. In this thesis, six kinds of diffusion couples were produced. There are 95Pb5Sn/Cu, 95Pb5Sn/Ni, Ni/95Pb5Sn/Cu, diffusion couple. The Cu foil was 400 m thickness and 99.9 wt% pure. The Ni foil was 400
m thickness and 99.5 wt% pure.
The 95Pb5Sn/Ni diffusion couples were prepared by electroplating a 100 and 400 m 95Pb5Sn layer over the Ni foil substrate. The 95Pb5Sn/Cu diffusion couples were prepared by electroplating a 100 and 400m 95Pb5Sn layer over the Cu foil substrate. The Ni/Sn/Cu diffusion couples were prepared by electroplating a 20 m Ni layer and two kinds of thickness Sn layers (100 and 400 m) over the Cu foil substrate. Two different solders, 95Pb5Sn and 90Pb10Sn, were used. A schematic drawing for the samples structure is shown in Fig. 2.3.
After electroplating, the foils were cut into 10 mm×10 mm diffusion couples. The samples were directly subjected to aging at 100, 130, 150, and 175 °C without going through the reflow step.
Analytical
The first reflow was monitored by a stereo-optical microscope to establish the melting process. After soldering and thermal aging, these samples were mounted in epoxy and metallographically polished to reveal the microstructure. Image analysis software was used to measure the thickness of the layer of intermetallic compounds. The thickness of each layer was defined as the total area occupied by the intermetallic compound divided by the linear length of the interface. A scanning electron microscope (SEM) was used to examine the
reaction zone for each solder joint, and a state-of-the-art field-emission electron probe microanalyzer (FE-EPMA, JEOL JXA-8500F, Tokyo, Japan) operated at 12 keV was used to identify the composition of the reaction products. During microprobe measurement, the measured x-rays were Ni K, Cu K, Pb L, and Sn L for Ni, Cu, Pb, and Sn respectively, and the standards used were pure Ni, Cu, Pb, and Sn, respectively. In microprobe analysis, the concentrations of each element was measured independently, and the total weight percentage was within 100 1% in each case. For every data point, at least four measurements were made and the average value was reported.
Fig. 2.1 Schematic drawing showing a 95Pb5Sn solder bump placed over a 37Pb63Sn pre-solder before assembly. Under the situation that the 95Pb5Sn in the solder bump and the 37Pb63Sn content in the printed solder paste are completely mixed, the overall composition will be 81Pb19Sn.
Chip
95Pb5Sn 95Pb5Sn
Ti/Cu NiNi
3m
95 95 m m Eutectic
Eutectic SnPb SnPb
Cu
BGA substrate
Fig. 2.2 Schematic drawing showing eutectic SnPb solder after assembly. The Ni UBM thickness was about 3 m.
Cu
BGA substrate
37Pb63Sn 37Pb63Sn
Chip
Cu NiNi
3m
8 mm die
8 mm 24.5 mm
24.5 mm
Y= 8.9 m
X= 8.9 m
XY= 12.6 m
@ T= 180
oC
(a)
a a
(b)
Fig. 2.3 Schematic drawing showing the structure of all diffusion couples.
400 400 m m
Cu
Ni 20 m Ni 20 m
Cu Pn5Sn
Pn5Sn or or Pb10Sn Pb10Sn
Pn5Sn Pn5Sn or or Pb10Sn Pb10Sn 100 100 m m
Cu
Pn5Sn Pn5Sn or or Pb10Sn Pb10Sn
Ni Pn5Sn
Pn5Sn or or Pb10Sn Pb10Sn 100 and 400
100 and 400 m m 100 and 400 100 and 400 m m
1. Solder/Cu diffusion couples 2. Solder/Ni diffusion couples
3. Ni/Solder/Cu diffusion couples
Chapter 3 Solding State - Composites Solder
3.1 In situ reflow observation
Figure 3.1 shows the sequence of events during the first reflow observed through a stereo microscope. The temperature and the time are shown at the upper-left-hand and upper-right-hand corners, respectively. The sample temperature was 31 °C when the reflow started, as shown in Fig. 3.1(a). The bright spheres represent the 95Pb5Sn solder bumps, and the dark regions immediately below the spheres represent the 37Pb63Sn solder paste.
Figure 3.1(b) shows the microstructure 1 s before the solder paste began to melt; the temperature at this point was 181 °C. Figure 3.1(c) shows the exact moment when the temperature reached the eutectic point and when the solder paste started to melt. Because the solidus temperature of the high-lead solder bump was higher, the solder bump remained solid.
As the solder paste became molten, its color changed from a dark color to that of shiny liquid metal. After another second, the molten 37Pb63Sn was displaced slightly to the side of the solid bumps, the standoff (i.e., the vertical distance between the chip and substrate) decreased, and the molten 37Pb63Sn climbed up the side of the solder-bump spheres, as shown in Fig. 3.1(d). The temperature at this point was 185 °C. When the peak reflow temperature was reached, the molten 37Pb63Sn climbed even higher up the side of the solder-bump spheres, as shown in Fig. 3.1(e). When the reflow ended, the standoff was the same as that at the peak reflow temperature, as shown in Fig. 3.1(f).
A closer examination of the solidified solder joints revealed that the height to which 37Pb63Sn ascended along the 95Pb5Sn-solder-bump sphere was uneven. Figure. 3.2 shows that the shape of the solidified 37Pb63Sn solder was similar to that of a calyx (i.e., a cuplike
structure) wrapped around the high-lead solder-bump sphere.
3.2 Cross-sectional observation
In this study, the flip-chip assembly had 45 solder bumps for the first outer row of solder joints. The cross-sectional microstructures from the first outer row of each package were examined. These solder bumps were numbered left to right from 1 to 45. Accordingly, solder bumps 1, 23, and 45 were the left-corner, center, and right-corner bumps, respectively.
Figure 4 shows the cross-sectional microstructures of bumps 1, 23, and 45 after 1–10 reflows.
The darker regions in Fig. 3.3 are the Sn-rich phase, and the lighter regions are the Pb-rich phase. After the first reflow, the intermixing of 95Pb5Sn and 37Pb63Sn was relatively limited, and the boundary between these two solders could still be approximately delineated as shown by the dashed curve in Fig. 3.3(a)–(c). As the intermixing increased with the increase in reflows, it became increasingly more difficult to define a clear boundary between these two solders. Nevertheless, even after 10 reflows, these two solders still did not mix completely, as shown in Fig. 3.3(p)–(r). There was still a more Sn-rich phase on the substrate side and a more Pb-rich phase on the chip side.
A clear difference can be seen between solder bumps 1, 23, and 45 when comparing the joint shapes. The solder joints for the center solder bump (bump 23) exhibited a symmetrical shape; however, the shape of the solder joints for solder bump 1 was asymmetrical and skewed slightly to the left. The solder joints for bump 45 were also asymmetrical, but they were skewed to the right. The schematic drawing shown in Fig. 3.4 helps clarify the difference in joint shapes. The solder bumps were perfectly aligned with the pre-solder on the substrate before reflow, as shown in Fig. 3.4(a). During reflow, as the oven temperature
increased, the chip and the substrate expanded. The difference in the coefficient of thermal expansion between the silicon chip ( = 2.6 ppm/°C) and the chip-carrier substrate ( = 15 ppm/°C) caused the substrate to expand more than the chip. Consequently, the soldering pads on the chip side became misaligned with the substrate, as shown in Fig. 3.4(b). When the oven temperature rose beyond the eutectic temperature, the solder paste melted and climbed up along the high-lead solder bump, as shown in Fig. 3.4(c). During the cooling stage, the molten solder solidified, and the substrate contracted and slightly deformed the solder joints.
These processes resulted in the final joint shapes shown in Fig. 3.4(d).
The flip-chip assembly had 16 solder bumps for the middle row of solder joints. The cross-sectional microstructures from the middle row were also examined. These solder bumps were numbered left to right from 1 to 16. Accordingly, solder bumps 1, 8, and 16 were the left-corner, center, and right-corner bumps, respectively. Figure 3.5 shows the cross-sectional microstructures of bumps 1, 8, and 16 in the middle row after 1–10 reflows.
The result shows that the middle row of chip and the first outer row of chip have a similar microstructure, indicating that the thermal history near the center of the packages was not much different from that near the edges of the packages.
As the number of reflows increased, gaps between the UBM and the solder developed in some of the solder joints such as the ones shown in Fig. 3.3(m) and Fig. 3.3(p). Magnified views of these gaps are shown in Fig. 3.6. This phenomenon was attributed to the dewetting caused by the molten 37Pb63Sn, which climbed high enough to reach the chip side, as illustrated schematically in Fig. 3.7. After the first reflow, the molten solder only reached about half the height of the high-lead solder bump, as illustrated in Fig. 3.7(a). As the number of reflows increased, the tip of the molten solder climbed continuously higher and eventually
continued to increase, the molten solder penetrated into the interface between the UBM and 95Pb5Sn, as illustrated in Fig. 3.7(c). During the cooling stage, the molten solder solidified, which was accompanied by a volume contraction. Because the bulk of the heat was drawn away from the substrate side, the molten solder solidified near the substrate side first. The last remaining molten solder was the solder near the UBM. The volume contraction occurred at the location of the last remaining molten solder (i.e., the tip of the calyx near the chip side), causing the formation of the gaps shown in Fig. 3.6. The formation of these gaps caused serious reliability concerns because the tip of the gap can serve as a site for the concentration of stress. Because these gaps had a tendency to develop in the samples with high reflow numbers, the application of reflows should be limited to less than five in products that are fabricated with composite solder joints.
The gravity effect should also be considered in this study. If the top side was the chip with high-lead bump and the bottom side was substrate with eutectic PbSn pre-solder. The gravity effect tended to drag down the molten solder, away from the chip/bump interface, and favor the formation of the dewetting phenomenon. On the other hand, if the top side had been the substrate and the bottom side had been the chip, then the molten 37Pb63Sn solder would tend to flow down toward the chip side. Therefore, the dewetting phenomenon was less likely to occur in this condition.
The interfacial reactions during reflow are described and discussed below. Figure 3.8 (a) shows the microstructure of the chip-side interface after one reflow. There was only a very thin layer of intermetallic compound at the interface. The thickness of this layer was too thin to be positively identified by FE-EPMA. Because the FE-EPMA detected only Ni and Sn signals, the layer was merely labeled as a Ni-Sn compound. Literature results carried out under similar situations suggest that this compound could very likely be Ni3Sn4 [HO9, HO10,
HO12, TSA2, CHE]. Figure 3.8 (b) shows that there were two intermetallic layers on the substrate side. The FE-EPMA measurement results showed that the outer layer was Cu6Sn5. The FE-EPMA did not detect a Ni signal in this Cu6Sn5 layer. The layer between Cu6Sn5 and Cu was very likely Cu3Sn according to results found in similar studies [SUN, YAN, CHO, WAN1, WAN2, WAN3].
Additional reflows did not change the interfacial reaction on the substrate side, and the results were omitted here for the sake of brevity. Additional reflow, however, did produce substantial change on the chip-side interface, as shown in Fig. 3.9. After two reflows, the reaction product was the same as that obtained from one reflow (i.e., a Ni-Sn compound), as shown in Fig. 3.9(a). After three reflows, (Cu0.6Ni0.4)6Sn5 replaced the Ni-Sn compound that formed along the outer rim of the solder joint, as shown in Fig. 3.9(b). The composition of this compound was determined by FE-EPMA measurements. In the central joint region, the Ni-Sn compound remained. After 10 reflows, the (Cu0.6Ni0.4)6Sn5 compound had completely replaced the Ni-Sn compound, as showed in Fig. 3.9(c). Figure 3.7 helps illustrate why the (Cu0.6Ni0.4)6Sn5 compound initially formed along the outer rim. Cu atoms, which were needed to form this compound, came from the Cu soldering pad on the substrate side. The Cu atoms dissolved quickly during the reflow of the molten-Sn to solid-Cu reaction system [HUN]. Therefore, the melting of the pre-solder dissolved many Cu atoms during reflow.
After multiple reflows, molten 37Pb63Sn reached the outer rim of the Ni UBM.
Consequently, Cu atoms reached the outer rim of the solder joints first. Apparently, the Cu concentration was high enough to convert the intermetallic compound from Ni-Sn to (Cu0.6Ni0.4)6Sn5. The Cu concentration required to convert the compound was about 0.5 wt.%
for lead-free solders [HO9, HO10, CHE]. The corresponding concentration for 37Pb63Sn, however, has not been established, but it should not significantly deviate from 0.6 wt.%.
Fig 3.1 In-situ observation of the first reflow process. (a) at 31
oC, (b) at 181
oC, (c) at 183
oC, (d) at 185
oC, (e) at 240
oC, and (f) at 31
oC. The number at the upper-right corner of each picture was the time (minutes and seconds) of the reflow.
(c) (d)
(a) (b)
(e) (f)
0 s 2 m 22 s
2 m 23 s 2 m 24 s
3 m 02 s 8 m 12 s
Fig. 3.2 Schematic drawing showing the calyx-shaped 37Pb63Sn presolder after the first reflow. The 37Pb63Sn wrapped around the high-lead solder bump.
Cu
Chip-carrier substrate
95Pb5Sn
Chip
NiNi
Ti/Cu
37Pb63Sn
37Pb63Sn
Fig 3.3 Cross-sectional views of the corner bumps (bump 1 and bump 45)
Fig 3.4 Schematic drawings showing the joint shapes development during the
37Pb63Sn 37Pb63Sn37Pb63Sn 37Pb63Sn37Pb63Sn
Cu
37Pb63Sn 37Pb63Sn37Pb63Sn 37Pb63Sn37Pb63Sn
Cu
BGA substrateCu Cu CuCu
BGA substrateCu
BGA substrateCu CuCu
37Pb63Sn
BGA substrateCu Cu CuCu
BGA substrateCu
BGA substrateCu CuCu
Chip
37Pb63Sn 37Pb63Sn37Pb63Sn 37Pb63Sn37Pb63Sn
Cu
37Pb63Sn 37Pb63Sn37Pb63Sn 37Pb63Sn37Pb63Sn
Cu
BGA substrateCu Cu
37Pb63Sn
37Pb63Sn 37Pb63Sn37Pb63Sn 37Pb63Sn37Pb63Sn
Cu
BGA substrateCu Cu CuCu
BGA substrateCu
BGA substrateCu CuCu
Chip
37Pb63Sn 37Pb63Sn37Pb63Sn 37Pb63Sn37Pb63Sn
Cu
37Pb63Sn 37Pb63Sn37Pb63Sn 37Pb63Sn37Pb63Sn
Cu
BGA substrateCu Cu
37Pb63Sn
37Pb63Sn 37Pb63Sn37Pb63Sn 37Pb63Sn37Pb63Sn
Cu
BGA substrateCu Cu CuCu
BGA substrateCu
BGA substrateCu CuCu
(a)
x 1 x 1
x 10 x 10 x 5 x 5
Bump 1
Bump 1 Bump 8 Bump 8 Bump 16 Bump 16
20 m
(a) (b) (c)
20 m
(d)
20 m
(e)
20 m
(f)
20 m
(h)
20 m
(i) (g)
20 m 20 m
20 m
Fig. 3.5 Cross-sectional views of the corner bumps (bump 1 and bump 16)
and the center bump (bump 8) after 1-10 reflows. These bumps are from
the middle row of a flip chip package.
10 m 10 m
(a) 5 times reflow (b) 10 times reflow
Fig. 3.6 Magnified views showing the gap in Fig. 3.3 (m) and (p).
Fig 3.7 Schematic drawings showing the emergence of the gap caused by the de-wetting phenomenon. (a)-(c) represent increasing number of reflow
.Cu
Chip
Chip--carrier substratecarrier substrate 37Pb63Sn
Chip--carrier substratecarrier substrate 37Pb63Sn
Chip--carrier substratecarrier substrate 37Pb63Sn
Chip--carrier substratecarrier substrate 37Pb63Sn
Chip--carrier substratecarrier substrate 37Pb63Sn
Chip--carrier substratecarrier substrate 37Pb63Sn
(a) As Reflow – chip side
(b) As Reflow – substrate side
Cu
6Sn
5 Cu6Sn5CuCu Cu
Cu
33SnSn
5 m10 m
NiNi
Ni-Sn
Ni-SnFig. 3.8 Cross-
Cross-sectional micrographs showing the (a) chip side and sectional micrographs showing the (a) chip side and
(b) substrate side interfaces after the first reflow.(b) substrate side interfaces after the first reflow.
5 m
(b) 3 Reflows – chip side
(Cu,Ni)
6Sn
5 (Cu,Ni)6Sn5 NiNiNi-Sn
Ni-Sn(Cu,Ni)
6Sn
5(Cu,Ni)6Sn5
5 m
(c) 10 Reflows – chip side
(Cu,Ni)
6Sn
5 (Cu,Ni)6Sn5NiNi
5 m
NiNi
Ni-Sn
Ni-Sn(a) 2 Reflows – chip side
Fig. 3..9 Cross-sectional micrographs showing the chip side interfaces
after (a) 2 reflows. (b) 3 reflows, and (c) 10 reflows.
Chapter 4 Solid State - Composites Solder
In this section, the cross-sectional microstructures from the center bump of the first outer row of each package were examined. As reported in chapter 3, after assembly the intermixing of 95Pb5Sn and 37Pb63Sn was relatively limited. There was only a very thin layer of NiSn intermetallic compound near the chip-side. The thickness of this layer was too thin to be positively identified by FE-EPMA, but this layer was likely to be Ni3Sn4. On the substrate side, an outer layer of Cu6Sn5 and an inner layer of Cu3Sn existed in the as-assembly condition.
Figure 4.1 shows the microstructure evolution for the solder joints aging at 130 °C to 175 °C for 100-2000 h. The initial small Sn-rich grains (the dark regions) coarsened into larger grains with aging, as shown in Fig. 4.1 (a)–(d). However, coarsening was not the only process at work. There was a second and parallel process that controlled the microstructure of the solder, i.e. the consumption of the Sn atoms due to the intermetallic formations at both interfaces. As shown in Fig. 4.1 (e)–(h) and (i)–(l), the amount of the Sn-rich phase decreased with aging. Accompanying this decrease of the Sn-rich phase, the intermetallic layer thickness at both interface increased with aging. At 175 °C, the Sn-rich phase had almost been completely exhausted after 3000 h of aging, as shown in Fig. 4.2.
The growth of intermetallic compounds at 175 °C during aging at both interfaces is shown in Fig. 4.3. As far as the types of the compound formed at the interface, the results shown in Fig. 4.3 are typical. The formation and microstructure at other temperatures were
The growth of intermetallic compounds at 175 °C during aging at both interfaces is shown in Fig. 4.3. As far as the types of the compound formed at the interface, the results shown in Fig. 4.3 are typical. The formation and microstructure at other temperatures were