(a) (b)
(c) (d)
Fig. 4.9： Surface morphology of unpolished AlN surface after etching for (a) 0 min, (b) 30 min, (c) 60 min, and (d) 120 min.
Fig.4.10： Variation in AlN surface roughness with etching time.
4μm
(a) (b)
(c) (d)
Fig.4.11： Surface morphologies of electroless Ni plated on polished AlN substrates for (a) 0.5 min, (b) 1 min, (c) 4.5 min, and (d) 26.5 min.
4-2.2 Surface morphology of electroless-plated metal layers
Figures 4.11 and 4.12 show the surface morphologies of the EN plated on polished and unpolished AlN substrates etched for 30 min, respectively. It can be clearly seen that the EN films plated on the polished surface possess a finer grain structure than those on the unpolished surface, however, the EN films on the unpolished substrate are rugged and
1 µm
1 µm 1 µm
1 µm
void-embedded. In both types of AlN substrates, the Ni grains should initiate from the Pd seeds formed during the surface activation. On the polished AlN substrate, the Ni granules grew and soon coalesced with each other to form the fine grain structure as shown in Fig.
4.11 [see illustration in Fig. 4.13(a)]. On the unpolished AlN substrate, even though the rougher surface might offer more nucleation sites as predicted by the heterogeneous nucleation theory, our experimental observation indicated that this could not be applied to electroless plating of the unpolished AlN substrate. As shown in Fig. 4.14(b), cross-sectional view of the Ni/AlN interface reveals that in fact the Ni does not fill up the etched holes in an unpolished AlN substrate. Probably the capillarity effects prevented the activation solution from reaching the bottom of the holes that metallization could only proceed on the top region of the unpolished surface [see illustration in Fig. 4.13(b)].
Another possible cause of this was the gaseous bubbles generated during plating, which inhibited the full coverage of metallization. Since the Ni granules located at the tops of the column-like grains must grow to a certain size to coalescence with each other, the Ni layer thus exhibits a coarse grain structure as shown in Fig. 4.12. Furthermore, if the coalescence fails, voids form in the Ni layer, as shown in Fig. 4.12(d).
(a) (b)
(c) (d)
Fig.4.12：Surface morphologies of electroless Ni plated on unpolished AlN substrates for (a) 0.5 min, (b) 1 min, (c) 4 min, and (d) 30 min.
(a)
(b)
Fig. 4.13： Illustration of grain growth on (a) polished and (b) unpolished AlN substrates.
1 µm
2 µm 8 µm
1 µm
AlN AlN
Nuclei Nuclei
Ni
Ni
(a)
(b)
Fig. 4.14： Cross-sectional view of EN layer plated for 3 min on (a) polished and (b) unpolished AlN substrates.
(a) (b)
Fig. 4.15： Surface morphologies of electroless Cu plated on (a) polished and (b) unpolished AlN substrates plating for 90 min.
Figures 4.15(a) and 4.15(b) show the surface morphologies of the electroless Cu plated on polished and unpolished AlN substrates, respectively. The differences in surface
AlN 8 µm
8 µm
8 µm 8 µm
roughness and grain size clearly indicate that the surface roughness of the AlN substrate affects the morphology of the subsequently plated film. The flat AlN substrate can generate a flat Cu layer with a fine grain structure.
4-2.3 Resistivity measurement
Figures 4.16(a) and 4.16(b) depict the resistivities of Ni and Cu layers on the polished and unpolished AlN substrates, respectively. In both cases, the resistivities of deposited films on the polished AlN surface are always lower than those on the unpolished AlN surface. Apparently, the rougher, unpolished AlN substrate surface induced defects in the film interface during grain clustering and thereby deteriorated the electrical conductivity of the deposited layers.
The specimens were also annealed at 250°C for 1 h under inert ambient. Even though a decrease in resistivity was observed, subsequent XRD analysis revealed that no marked structure change occurred in the deposited layers. The decrease in resistivity might be a result of the annihilation of point defects and dislocations, rather than of the recrystallization and grain growth phenomena in the Ni and Cu layers. There was no obvious change in the microstructures of metal films subjected to a 250°C˘1 h annealing.
(a)
(b)
Fig. 4.16： Resistivities of electroless (a) Ni and (b) Cu deposited on polished and unpolished AlN substrates.
4-2.4 Structure and composition analyses of plated Ni and Cu layers
The amorphous feature of EN films indicates that they are in a non-equilibrium state.
[57] Figures 4.17(a) and 4.17(b) show the XRD patterns of EN films before and after 250°C˘1 h annealing. It can be seen that the thin EN film was amorphous and, with the increase in plating time, the crystallinity of the EN film gradually improved. Nevertheless, the EN film plated over 15 min (or film thickness > 5µm) still has a wide peak distributed over the diffracted angle range of 37° to 55°. This is due to the fact that during plating, the phosphorus atoms were incorporated in the Ni deposition reducing its crystallization.
Figures 4.17(a) and 4.17(b) further indicate that the structure of the EN films did not change markedly with the 250°C˘1 h annealing. Figures 4.18(a) and 4.18(b) show the XRD patterns of Cu films before and after the 250°C˘1 h annealing. The Cu films were polycrystalline, evidenced by the sharp (111) and (200) diffraction peaks appearing at 43.3° and 50.4°, respectively.
(a)
(a)
(b)
Fig. 4.18： XRD pattern of Cu/Ni/AlN (a) before and (b) after 250°C˘1 h annealing.
( : AlN; :Cu)
Tables 4.3 and 4.4 show the composition of Ni on the polished and unpolished AlN substrates analyzed by SEM/EDS. It was found that after 2 min of deposition, the AlN substrates were completely covered by Ni film whose thickness was approximately equal to 0.7µm. As the plating time increases to 5 min, the content of phosphorous in Ni reached a steady value of 9.7 wt%. The composition variations of Ni films were similar regardless of the difference in roughness of AlN substrates. Previous studies[7,59] pointed out that when the phosphorous content of Ni is 7~10 wt%, its structure is a mixture of amorphous and microcrystalline grains. When the phosphorous content of Ni is higher than 10 wt%, it becomes amorphous. The secondary-ion-mass spectrometry (SIMS) depth profile shown in Fig. 4.19 revealed that the phosphorous content is the highest at the Ni/AlN interface. This implies that the phosphorous content might exceed 10 wt% and the Ni film in the vicinity of Ni/AlN interface is amorphous. This provides the EN film as a good barrier since it contains no grain boundary for fast diffusion.
Table 4.3. Compositions of electroless Ni on polished AlN surface. (unit : wt%)
Plating time Al P Ni
15 sec 77.58 6.14 16.29
30 sec 52.96 7.78 39.24
45 sec 36.33 8.95 54.72
120 sec 0 9.81 90.19
300 sec 0 9.69 90.31
Table 4.4. Compositions of electroless Ni on unpolished AlN surface. (unit : wt%)
Plating time Al P Ni
15 sec 54.29 6.76 38.95
30 sec 38.56 7.33 54.11
45 sec 18.71 8.15 73.14
120 sec 0 9.78 90.22
300 sec 0 9.72 90.28
Fig. 4.19： SIMS depth profile of EN layer plated on polished AlN substrate for 2 min.
4-2.5 Pull-off and shear tests
Table 4.5 shows the results of pull-off test for EN layers deposited on polished AlN substrates. During the test, it was found that all specimens broke at the stud/epoxy resin interface, rather than at the Ni/AlN interface. Hence the test results shown in Table 4.5 do not represent the true adhesion strength of the EN/AlN interface. The high adhesion strength of EN film on AlN should result from the mechanical interlock effects on rough substrate surface generated by the etching using the 4 wt% NaOH solution for 30 min.
Table 4.5. Pull-off test results of electroless Ni on polished AlN substrate.
Pull-off stress (kg/cm2) Location of breakage
889.20 Stud/epoxy resin interface
892.81 Stud/epoxy resin interface
923.07 Stud/epoxy resin interface
846.95 Stud/epoxy resin interface
913.14 Stud/epoxy resin interface
Since we also plated Cu on the Ni film and then carried out the pull-off test, the similar result was obtained compared to those of Ni film (see Table 4.6). This result not only indicates that the electroless Cu is able to form a relatively strong bonding with Ni,
but also ensures that the adhesion property of the Cu/Ni interconnection formed on the AlN substrate fulfills the mechanical demands of subsequent processing.
Table 4.6 Pull-off test results of electroless Ni/Cu on polished AlN substrate.
Stress (kg/cm2) Location of breakage
761.57 Stud/epoxy resin interface
937.78 Stud/epoxy resin interface
932.68 Stud/epoxy resin interface
833.22 Stud/epoxy resin interface
924.42 Stud/epoxy resin interface
The shear test results indicated that the 30-µm-high Ni lug deposited on AlN surface had a higher shear strength than the machine test limit. The upper thrust limitation of the shear tester is 100 g, so we could only conclude that the shear strength of EN on AlN was higher than 100 g.
The above results indicate that etching on the surface of AlN substrates provides a strong mechanical interlocking effect to ensure good adhesion of EN films. However, the etching must be appropriate otherwise the rough surface would induce voids in the metal overlayer and hence deteriorate its electrical property. From the measured results of surface roughness, resistivity, and adhesion tests, we concluded that a satisfactory substrate
could be obtained by etching on polished AlN surface using 4 wt% NaOH solution for approximately 30 min. The EN film plated on such a substrate surface not only possesses a compact and nearly void-free structure, it also exhibits a low electrical resistivity that meets the requirement of subsequent flip chip interconnection.
4-2.6. Flip chip bumping
In this study, we also investigated the flip chip bonding of GaAs PHEMT devices on the metallized AlN substrates, as shown in Fig. 4.20. From the SEM observation shown in Fig. 4.21, it was found that the thickness of intermetallic compounds (IMCs) increased with the time of thermal treatment at 250°C. The IMCs mainly consisted of Ni3Sn4 and Cu6Sn5, as revealed by XRD and SEM/EDS analyses.
In Fig. 4.21(b), the hemispherical scallop-like IMCs[60] formed after 2 h annealing can be clearly seen. The SEM/EDS revealed that the IMCs possess the composition of Cu:Sn = 54.5:45.5 so they should consist of the Cu6Sn5 phase. Finally, the EDS element linear scanning analysis, shown in Fig. 3-15, reveals rather weak Sn signal amplitude inside the Ni layer. This is consistent with previous studies that the amorphous EN film was a good diffusion barrier layer for UBM construction.
Fig. 4.20： Flip chip bonding of GaAs PHEMT IC on AlN substrate.
(a)
(b)
Fig. 4.21： SEM cross-sectional view of EN/AlN interfaces and corresponding line scanning EDS analysis (a) before 250°C annealing and (b) after 250°C 2 h annealing.