Engineering Materials and Their Applications
Evaluation of Bonding Strength and Interfacial Resistance of Diffusion-Bonded Ag/Si Interfaces
2021 年 62 巻 4 号 p. 539-543
詳細
2021 年 62 巻 4 号 p. 539-543
Diffusion bonding of silicon and pure silver was performed and the bonding strength and electrical resistivity were measured. Diffusion bonding was performed under a uniaxial pressure (10, 20 MPa) at 1103 K in an argon atmosphere and bonding status was observed using a scanning electron microscope. It was found that the strength of a joint increases as the bonding time increases until 90 min for both the bonding pressures. There is no clear trend in resistivity variation with bonding time. In addition, the resistance increased by about 0.3% compared to that of silicon only. Silver is a good candidate for an electrode material for silicon because it is bonded to silicon well and there is no insulating layer formed such as an intermediate compound.
Global environmental problems such as the depletion of fossil fuels and other resources have attracted much attentions. Therefore, measures have been taken for reduction of energy consumption and use of renewable energies. Thermoelectric materials have an excellent ability to convert thermal energy to electrical energy and vice versa. Noting that more than 60% of the energy produced in the world is disposed of without being used as waste heat, in recent years, expectations for thermoelectric conversion as a waste heat recovery technology or an energy acquisition technology from unused heat have been glowing.1)
In addition, in order to operate huge number of sensors in the advanced IoT-based human society in the future, there are high expectations for thermoelectric conversion technology, which can be used as maintenance-free and independent power sources.2) For these reasons, there is an urgent need to develop thermoelectric conversion devices that can be mass-produced at low cost, have high efficiency, and do not contain harmful elements.
Silicon (Si) is non-toxic and can behave as p-type or n-type semiconductor by adding a small amount of impurity element. Si has tended to be left out of candidates of materials for thermoelectric conversion modules due to its high thermal conductivity so far. However, due to its abundance and nontoxicity, it is attracting much attention for thermoelectric use for the future society. Recent studies have shown that nanostructuring is effective for lowering thermal conductivity3) or the effective heat transfer coefficient can be improved by making Si porous to compensate for the disadvantage of high thermal conductivity,4) hence use of Si as thermoelectric material could be realized in the not too distant future. Therefore, it is necessary to prepare data for the use of Si in thermoelectric devices. In order to fabricate thermoelectric conversion devices, the bonding of thermoelectric material and metal electrode is essential. In particular, silver (Ag) is one of candidates for metal electrodes because of its high electrical conductivity and the absence of intermediate compounds with Si.5,6) The bonding properties between Si and Ag at the atomic scale and the shear strength have been reported in the past.7–10) However, more data such as bonding strength and contact resistance are needed for fabrication of thermoelectric devices using porous Si and their use. The present study investigates the diffusion bonding between Si and Ag. The strength of a joint and interfacial electrical resistivity have been evaluated.
Bonding between Si and Ag were executed using a diffusion bonding technique. A schematic diagram of the diffusion bonding is shown in Fig. 1. For preparation of Si/Ag diffusion-bonded samples, phosphorus-doped Si (carrier density 2.69 × 1018 cm−3, 11.5 mm thick) and Ag sheet (99.998% purity, 0.5 mm thick) were used. A Si chunk was cut into a rectangular parallelepiped with an approximately 7 mm × 7 mm. square section by a wire electric discharge machine (HS-300, Brother Industries, Ltd.). The surface of the Si sample that was going to be bonded was grinded using a series of sandpapers and was polished finally with diamond slurry solution (9 µm and then 1 µm) on a plastic-metal composite disc. Ag sheet was cut into a square shape with 7 mm on a side. The Si and Ag samples were placed in a carbon die with a 27 mm diameter hole. Diffusion bonding was performed under a uniaxial pressure (SPD4000-I, Dai-ichi Kiden Co., Ltd.) at a high temperature in an Ar atmosphere. Table 1 shows the detailed conditions used in the bonding process.
Schematic diagram of the diffusion bonding.
The diffusion-bonded samples were then subject to strength of a joint measurements followed by fracture surface observations and electrical resistance measurements.
For tensile tests, a stainless-steel ring was attached to each end member of diffusion bonded samples using a high-strength adhesive. Stainless steel wires were threaded through the ring, pulled at a rate of 1 mm·min−1 in a tensile tester (Instron Model 4400 Universal Testing Systems, Instron Corporation) until fracture. Measurements were performed three times for each sample. Fracture strength per unit area was evaluated from the force at the fracture and the cross-sectional area of the sample, which was 7 mm × 7 mm.
Microstructural observations were made to examine the bonding status. In order to observe the center, a diamond saw was used to cut perpendicular to the joint surface, and their surface were polished with a series of water-resistant abrasive papers, followed by alumina paste slurry and colloidal silica on a buff. The bonding interface was observed using a scanning electron microscopy (SEM, S-4800, Hitachi High-Tech Corp.). Besides, for the sample after the tensile test, the fraction of the area where Ag and Si were microscopically well-bonded were determined. 10 randomly selected Si surface bonds in the samples (30 min, 120 min, 180 min annealed samples) were observed under a microscope using SEM. The contact area of the fracture Si surface was calculated by using image analysis software Image J (ver. 1.52a, National Institutes of Health and the Laboratory for Optical and Computational Instrumentation). The fraction of the area where Ag and Si were microscopically well-bonded were determined color shading on the surface after the fracture.
The electrical resistances of bonded samples were measured using a thermoelectric measurement and evaluation system (ZEM-3, ADVANCE RIKO, Inc.). In the measurements, Ag and Si in each samples were arranged in series in the electrical circuits, where the lengths of Ag and Si are 0.5 mm silver sandwiched between 5.75 mm Si, respectively. Samples were cut into rectangular parallelpiped with cross sections of 3 mm × 3 mm squares using a wire electric discharge machine. The resistivity of Si and Ag which were also used in the diffusion-bonded samples were measured to evaluate the interface resistances of diffusion-bonded samples. The same sample was measured a total of six times while inverting the top and bottom.
Result of tensile tests are shown in Fig. 2. The strength of a joint is shown for various sample preparation conditions as functions of annealing time. The strength of a joint here refers to the apparent fracture strength of the bonded samples in tensile tests. The data points include either the fracture strength at the Si/Ag bonding interface or the fracture strength for cases where fracture occurred originating with cracking in Si. The error bars shown for the points for 10 MPa slow cooling mean two standard deviations from the mean values. The error bars for other conditions are not shown since their range of errors are not very different from those for 10 MPa slow cooling. It is found that the strength of a joint increases as the bonding time increases until 90 min for all bonding pressures while it does not vary significantly after 90 min. It is also found that higher bonding pressure (20 MPa) gives higher strength of a joint than lower bonding pressure (10 MPa). In addition, the strength of a joint is higher for the slow cooling (10 K·min−1) than for the air cooling (∼15 K·min−1 in average).
Strength of a joint between Si and Ag under various bonding conditions shown as functions of bonding time.
Si fracture surfaces after tensile tests are shown in Fig. 3. For some air-cooled samples, fracture occurred in Si close to the bonding interface instead of the bonding interface. This trend is more remarkable for a long bonding time more than 120 min. The cause of the Si fracture could be attributed to the formation of cracks (Fig. 3(c)) in Si caused by the internal stress due to the difference in the thermal expansion coefficients of Si (2.7 × 10−6 K−1)11) and Ag (1.9 × 10−5 K−1).8,12) Therefore, it can be presumed that samples that fractured from the crack have a higher bonding strength at the bonding interface than the strength of a joint obtained in this study. In order to prevent the fracture due to the cracks, it is necessary to relieve internal stress by performing diffusion bonding at lower temperatures or slow cooling right after diffusion bonding.13,14) Actually, it is found in Figs. 3 and 4 that slowly cooled samples exhibit higher strength than air-cooled samples and tend to fracture at the bonding interface.
Si fracture surfaces after tensile tests of samples prepared under (a) 10 MPa pressure and slow cooling, and (b) 10 MPa pressure and air cooling. The numbers on the top of the images indicate the bonding time. (c) A side view of a sample (prepared by 300 min bonding followed by air cooling).
SEM images of the center (a) and the edge (b) of the bonding interface between Si and Ag (10 MPa pressure, bonding time 120 min). Figure (c) shows a macroscopic appearances and microstructures of the fracture surface on the Si side of the sample that fractured at the bonding interface by tensile tests (10 MPa pressure, bonding time 30, 120, 180 min).
Recent studies suggest that porous silicon, which has higher specific surface area, shows an increased effective heat transfer coefficient, which could compensate for a disadvantage of the high thermal conductivity of silicon for application for thermoelectric conversion devices.15–17) Such devices could be used for waste heat recovery utilizing hot wastewater from industry or hot springs. Regarding to this type of devices, the bonding strength between porous silicon plates and metallic electrodes presumably need to withstand water pressure. The water pressure standard for industrial water supply in Japan is considered to be at most 0.5 MPa.18) The results of the present work shows that the strength of a joint is high enough to withstand water pressure. It should be mentioned, however, that further studies are desired for more detailed distribution of stresses due to water pressure, difference in thermal expansion coefficients of Si/metallic electrodes and Si/thermal insulating materials.
3.2 Observations of bonded and fracture surfacesSEM images of the interface between Si and Ag (10 MPa pressure, bonding time 120 min, slow cooling) are shown in Fig. 4(a) and (b). It is found that Si and Ag were bonded uniformly. No voids are observed at the interface. Small bumps of Ag (∼ 3 µm) are observed sticking into Si. These are thought to be formed by plastic flow of Ag into the scratches of Si formed in the preparation of Si surfaces by polishing. There are no compounds formed at the interface between Si and Ag, which is consistent with the reported phase diagram.5)
Macroscopic appearances and microstructures of the fracture surface on the Si side of the sample that fractured at the bonding interface by tensile tests (10 MPa pressure, bonding time 30, 120, 180 min, slow cooling) are shown in Fig. 4(c). The dark phase in the SEM images is Si and the bright phase is Ag remaining on the surface of Si. Here, the area fraction of Ag residue on the fracture surface was evaluated by image analysis on the SEM images using an Image J software as shown in Table 2. It is found that the Ag residue on the fracture surface increases with increasing bonding time until 180 min. while, the increase from 120 to 180 minutes was small (about 3%), suggesting that the bonding strength at the interface is close to its upper limit. There is a trend that the bonding strength increases with increasing the area fraction of Ag residue. From this result, it is likely that the Ag residue is formed because the bonding between Ag and Si is strong and hence Ag is fractured in such regions. On the other hand, in the regions where Si is observed on the fracture surface, the bonding between Ag and Si is thought to be weak and Si is exposed as a result. From the fact that the fraction of Ag residue depends on bonding time, the area fraction of bonding can be controlled by bonding time.
Thus, there are factors that increases and decreases the strength of a joint: increase in bonding strength at the interface due to the increase in bonding area which is observed as the increase in the area fraction of Ag residue for longer bonding time or higher pressure, and decrease in apparent bonding strength, i.e. strength of a joint, due to cracking in Si, which is mention in 3.1.
3.3 Resistivity measurementsThe results of electrical resistance measurements of Si/Ag bonded samples (10 MPa pressure, slow cooling) are shown as functions of bonding time in Fig. 5. The error bars mean two standard deviations from the mean value. The resistance of the bonded samples (10 MPa, 30 min, slow cooling) is the same order of magnitude as the summation of resistances of Si and Ag pieces with the same dimensions as those in the bonded samples evaluated based on resistivity measurements of Si and Ag. The increase in the resistance corresponds to an interfacial resistance. There is no clear trend in the interface resistance variation with bonding time. The following equation was used as the error in interface resistance.
| \begin{equation*} \sigma_{\textit{Interface}} = \sqrt{(\sigma_{\textit{Si}})^{2} + (\sigma_{\textit{Bond}})^{2}} \end{equation*} |
Resistance of the Si/Ag bonded samples (10 MPa pressure, bonding time 120 min, slow cooling) measured perpendicularly to the bonding interface as functions of bonding time.
It is evident from these results that Ag is effective for Si bonding because there is no insulating layer such as an intermediate compound between Si and Ag.
The present work studied diffusion bonding between Si and Ag as a fundamental technology to fabricate thermoelectric modules using silicon as thermoelectric material and silver as metallic electrodes. The diffusion bonding was successfully made at 1103 K under unidirectional compression load. No intermediate phase was observed consistently with the reported Ag–Si equilibrium phase diagram. The strength of a joint increases with increasing bonding time up to 90 min while it does not vary around 8 MPa with bonding time longer than 90 min for the bonding with a 20 MPa bonding pressure followed by slow cooling. Furthermore, it was found to satisfy the strength of a joint required for thermoelectric devices using porous Si. On the other hand, in this measurement, no significant variation in electrical resistance with bonding time was observed. However, the interfacial resistivity is in the order of 10−10 Ω m2. Thus, Ag is useful as metallic electrodes for Si.
It is hoped that the findings that have been presented in this paper will contribute to a better understanding of the future of thermoelectric conversion devices.
This work was supported by JSPS KAKENHI Grant-in-Aid for Scientific Research (B) (Grant Number 17H03399).
