Search for articles:
Yeongseon Kim
, Sang Hyun Park
Corresponding authors:
Received: 2018-03-26
Accepted: 2018-08-15
Online: 2019-05-10
Copyright: 2019 Editorial board of Journal of Materials Science & Technology Copyright reserved, Editorial board of Journal of Materials Science & Technology
More
Abstract
SnSe is a promising thermoelectric material with a high figure of merit in single crystal form, which has stimulated continuous research on polycrystalline SnSe. In this study, we investigated a metallization techniques for polycrystalline SnSe to achieve highly efficient and practical SnSe thermoelectric modules. The Ag/Ni metallization layers were formed on pristine polycrystalline SnSe using various deposition technique: sputter coating Ni, powder Ni and foil Ni by spark plasma sintering. Structural analysis demonstrated that the microstructure and contact resistance could be different according to the metallization process, despite using the same metals. The Ag/Ni metallization layer using foil Ni acted as an effective diffusion barrier and minimized electrical contact resistance (2.3×10-4 Ω cm2). A power loss in the thermoelectric module of only 5% was demonstrated using finite element simulation.
Keywords：
Thermoelectric power generation involves direct conversion of heat to electricity. This technology has been researched consistently for several decades, with particular focus on waste heat recovery applications. For several decades, the power conversion efficiency of Bi2Te3-based thermoelectric modules remained less than 5% with material figure-of-merit (ZT) lower than 1 [[1], [2], [3], [4]] in the temperature range from room temperature to 400 K. Highly improved thermoelectric material characteristics based on the nano-structuring principle started to be reported from around year 2000 and resulted in highly efficient thermoelectric devices for power generation in a range of intermediate temperature (600-900 K) [[5], [6], [7]]. N-type skutterudites have been reported with a high ZT value of 1.9 at 835 K [8]. The ZT value of p-type PbTe has reached 1.8 at 810 K [9]. Mg2Si thermoelectric material with Sb0.5%-Zn0.5% doping showed a ZT value of nearly 1 at 880 K [10]. The world’s highest ZT result (2.6 at 923 K) was reported for a single-crystal SnSe thermoelectric material [11]. The main matrix elements of this substance (Sn and Se) are also relatively economical compared to other material cases utilizing Te, Co or Ag. With its exceptionally high ZT and cost effectiveness, SnSe is gaining significant attention from many research groups to determine how much the energy conversion efficiency of such devices could be enhanced, using this novel thermoelectric material.
Despite the continuous development of thermoelectric materials, there have as yet been no practical applications, mainly due to the lack of an effective technology for fabrication of thermoelectric modules. For the development of a highly efficient thermoelectric module, the metallization on the thermoelectric materials is one of the most significant of the key factors. Effective metallization is essential to avoid parasitic electrical and thermal contact losses and to prevent elemental inter-diffusion between the thermoelectric materials and metallization layer at operating temperatures. The optimum metallization layer would act as an electrical contact and diffusion barrier, and should be developed in relation to the thermoelectric materials used. There have been a few studies on SnSe metallization reported in recent years. The Wang group reported first principle calculation results of the geometries and electronic properties of SnSe and the metal contacts [12]. Our group also previously investigated SnSe metallization layers with Ag/Ni and Ti/Co/Ag as a first step toward a SnSe thermoelectric module [13,14]. In the case of Ag/Ni metallization layer, the Ni metallization layer was used as a diffusion barrier for SnSe, which is widely used as a metallization layer for both low and high temperature thermoelectric materials (BiTe, PbTe, etc.) [[15], [16], [17]]. However, in the case of SnSe, the Ni metallization layer was resulted in severe cracks because it easily reacts with SnSe forming Ni-Sn alloy. In order to alleviate interfacial cracks, the Ag was used as a secondary metallization layer together with Ni since the use of Ag enabled to control the interfacial reaction [13].
In this current work, the Ag/Ni metallization layers on SnSe were systematically investigated employing three different fabrication techniques. The Ni/SnSe interfaces were investigated using three kinds of Ni deposition methods (powder-, foil- and sputtered-Ni) to optimize interfacial properties. The microstructure and electrical contact resistance of the interfaces between Ni and SnSe were analyzed and compared. The feasibility of fabricating thermoelectric modules with SnSe and Ag/Ni metallization layers was evaluated using finite element simulation.
The Sn and Se elements were prepared and heated to 1223 K for 24 h for the synthesis. The obtained polycrystalline SnSe ingot was crushed and ground into 45 μm particles for sintering to get denser SnSe ingots. The sintering was conducted by spark plasma sintering (SPS, Fuji Electronic Industrial Co., Ltd. SPS-211LX) under a uniaxial pressure of 40 MPa, and at 823 K for 10 min.
The electrical conductivity and Seebeck coefficient were measured at the same time under argon gas from room temperature to 823 K on a Netzsch SBA 458 Nemesis instrument. Thermal diffusivity (D) as an index of the thermal transport properties was directly measured under argon atmosphere from room temperature to ∼823 K using a flash diffusivity method with a Netzsch LFA 457 MicroFlash instrument. The thermal conductivity (κ) was calculated from the equation κ =DCPρ, where ρ is the density of the sample and CP is the heat capacity of the sample. In this work, the CP was used from the literature [11].
The metallization processes were conducted in two different ways. One was the SPS co-sintering technique, which was used to sinter together the SnSe and metal at the same time. The 200-mesh Ag powder (99.99%, Sigma-Aldrich), 325-mesh powder Ni (>99.8%, Alfa Aesar) and foil Ni (>99.99%, Alfa Aesar) were prepared for the metallization. The SnSe powder was loaded into a graphite mold with silver powder and nickel to form a Ag/Ni metallization layer. In case of Ni, powder and foil were used to investigate the interfacial structure according to the metal form. The other use of Ni was by sputtering, which deposited a very thin Ni layer on the SnSe. First, the SnSe ingot was sintered by SPS and then the Ni was formed by RF magnetron sputtering (SORONA, SRN-110). The RF power and sputtering time was 200 W and 5 h to make several micrometer thickness of Ni. Ar flow rate was 20 sccm, and substrate temperature was room temperature. After that, a Ag layer was added using SPS at 723 K for 10 min.
The microstructure of the interfaces between the metal layers and SnSe thermoelectric material was investigated using SEM (JEOL, JSM-7000 F). Energy dispersive X-ray spectroscopy (EDS) line scanning was also conducted to investigate any intermetallic compounds and the diffusion property of the element composing the samples. The intermetallic compounds formed during the SPS processes were investigated in detail using a field emission transmission electron microscope (TEM, JEOL, JEM-2100 F). The cross-sectioned TEM specimen was prepared using a focused ion beam (FEI, Quanta 3D FEG) to analyze the interfaces between SnSe and each Ni layer. The electrical contact resistance between the metal contact and thermoelectric material was measured to verify the metallization properties using an in-house electrical contact resistance evaluation system [18]. The specific contact resistance was calculated from the resistance change that occurred at the interface, multiplied by the contact area. The effect of the electrical contact resistance and the expected power output of the SnSe modules were investigated using the three-dimensional finite-element simulation provided by COMSOL Multiphysics.
The polycrystalline p-type SnSe powder was sintered using SPS to make dense pellets. The thermoelectric properties were measured perpendicular to the press direction. Fig. 1 shows the temperature dependence of the Seebeck coefficient, electrical conductivity (σ), thermal conductivity (κ), and ZT of both polycrystalline SnSe and single-crystal SnSe [11]. The polycrystalline SnSe has the same anisotropic characteristic as single-crystal SnSe because of the layered structure of SnSe and the SPS process, which applied high pressure and temperature. Thus, the Seebeck coefficient and electrical conductivity of the polycrystalline SnSe had trends similar to those of single-crystal SnSe. The Seebeck coefficient and electrical conductivity maintained their values up to 573 K; then these changed abruptly, which indicated SnSe phase transition from Pnma to Cmcm. Because the thermal conductivity of polycrystalline SnSe was higher than that of the single-crystal SnSe which has strong intrinsic anharmonicity, the ZT of polycrystalline p-type SnSe was only 0.62 at 823 K. This ZT value of polycrystalline SnSe lies between those of a-axis and c-axis in single-crystal SnSe. With doping or tuning of the phase transition temperature, polycrystalline SnSe could be enhanced largely, and many researchers are working on it [[19], [20], [21], [22]]. In this study, rather than focusing on material enhancement, we focused on the metallization of the polycrystalline SnSe which is first step toward fabricating a thermoelectric module.
Fig. 1 Thermoelectric properties of polycrystalline p-type SnSe (red circle) and single crystal SnSe along the c-axis (black triangle) [
Metallization is usually conducted using techniques such as electroplating, sputtering, or hot-pressing. Because the microstructure of a metallized interface differs according to the metallization process, the properties of the metallized interface could be different despite being made of the same metals. Previous work demonstrated that SnSe with a Ag/Ni metallization layer prevents elemental diffusion at the interface and has low electrical contact resistance [13]. The Ni was used as a diffusion barrier of SnSe and Ag was used as a secondary metallization layer which can alleviate interfacial cracks occurred when using single Ni layer. This metallization layer was formed by hot-pressing, which requires exposure to high temperature for a long time, and which could deteriorate the properties of the thermoelectric materials. To form an effective diffusion barrier on SnSe with this Ag/Ni metallization layer, three different processes of Ni formation (SPS co-sintering using both Ni powder and foil and sputtering) were investigated in this study.
The Ni metal layer formed by the sputtering process had good adhesion and the thermoelectric materials were not exposed to high temperature. However, it is not easy to make a thick metal layer this way more than several micro-meters thick, and it could be non-uniform over its entire area. In contrast, the SPS co-sintering processes allowed the sintering and metallization to proceed simultaneously using the SPS system.
Fig. 2 shows the EDS line scan images of the Ag/Ni metallization layers using the three processes. The thicknesses of the sintered-powder, sputtered, and foil-Ni were 100 μm, 5 μm, and 50 μm, respectively. These Ni layers effectively prevented Se diffusion and the Ag layer prevented Sn diffusion into the Ag layer. As shown in Fig. 2(a), sintered-powder-Ni and SnSe made two kinds of intermetallic compounds that were mainly composed of Sn-Se-Ni and Sn-Ni, respectively. The sputtered-Ni and foil-Ni showed different behavior comparing with sintered-powder-Ni in EDS line scan images, which showed only Sn-Ni alloy with a thin IMC layer. This demonstrated that different interface characteristics could be generated according to the deposition processes despite being of the same metals.
Fig. 2 Cross-sectional EDS line scan images of Ag/Ni/SnSe metallized interfaces: (a) sintered-powder-Ni, (b) sputtered-Ni, and (c) foil-Ni.
The in-depth study on the interfacial microstructures of three different Ni processes (powder and foil SPS co-sintering metallization and thin film sputtering metallization) were investigated using the transmission electron microscopy. Fig. 3 shows the TEM image together with EDX mapping analysis used to investigate the phase and chemical composition of the Ni metallization layer after the powder sintering. Three regions: Se- (denoted with A), Sn- (denoted with B), and Ag-enriched (denoted with C) were observed, while Ni content was invariant over the entire image except for the Ag-enriched part (region C). The region C could be identified as Ag metal. To unravel the phase information of the other regions of the metallization layer, fast Fourier transform (FFT) patterns from the high resolution TEM image of each region were obtained and are presented in Fig. 3. The obtained diffraction patterns were indexed using the CrysTBox program [23]. The FFT patterns of bright and dark regions (region A and B) were indexed to Ni3Sn2 and Ni5.63SnSe2, respectively. The microstructural feature of the Ni metallization layer from powder sintering agrees well with the Rietveld refinement results in a previous report [13].
Fig. 3 (a-e) EDX mapping for Ni powder specimen and FFT patterns of (f) A and (g) B regions.
The TEM image and EDX mapping results of the Ni metallization layer from sputtering are shown in Fig. 4 to enable comparison with the Ni layer prepared by powder sintering. The EDX mapping and dark field TEM results show four distinctive regions (denoted A′ - D′) and the corresponding FFT patterns of each region are presented in Fig. 4. It is remarkable that Ni3Sn2 and Ag2Se phases are the majority crystalline compounds with only a small amount of unidentified crystalline phases at the Ni/SnSe interface, whereas Ni5.63SnSe2 phase is not detected. This observation suggests that Ni5.63SnSe2 phase can only be formed when the amount of Ni is sufficient, because the thickness of the Ni layer (4 μm) deposited by sputtering is much thinner than the Ni layer from powder sintering (100 μm). In addition, the formation of Ni3Sn2 and Ag2Se phases in region C′ and D′ suggests that the reaction between SnSe and Ag cannot be avoided due to the lack of Ni compared to other Ni layers.
Fig. 4 (a-e) EDX mapping for Ni sputter specimen and FFT patterns of (f) A′, (g) B′, (h) C′, and (j) D′ regions.
The TEM image and EDX mapping results of the Ni metallization layer made using Ni foil are shown in Fig. 5 to enable comparison with the Ni layer prepared by powder sintering. The EDX mapping and dark field TEM results show four distinctive regions (denoted with A’’ - D’’). Crystalline Ni3Sn2, Ni5.63SnSe2, and Ag2Se phases were found together with Ni-Sn-Se amorphous phase. Because Ni reacts vigorously with SnSe during the SPS process, Se diffusion into the Ni foil occurs; so, Ag2Se is observed at the Ni/SnSe interface.
Fig. 5 (a-e) EDX mapping for Ni foil specimen and FFT patterns of (f) A″, (g) B″, (h) C″, and (j) D″ regions.
One of the most important characteristics of metallization is contact resistance. The metallization layer should have minimum contact resistance with the thermoelectric materials to minimize parasitic loss. Fig. 6 shows the specific contact resistance of the Ag/Ni/SnSe thermoelectric legs with different Ni layers. The specific contact resistances of the foil, powder, and sputtered-Ni interface were 0.23 mΩ cm2, 0.79 mΩ cm2 and 2.55 mΩ cm2, respectively. Because the electrical resistivity of SnSe (1.6 × 10-2 Ω cm) is much higher than Ni (6.9 × 10-6 Ω cm), the intermetallic compounds should be formed at the interface between Ni and SnSe to fill the resistivity gap, in other words, to minimize the electrical contact resistance. The sputtered-Ni only formed Ni3Sn2 intermetallic compounds with thin layer. On the other hand, powder-Ni and foil-Ni formed Ni3Sn2 and Ni5.63SnSe2 intermetallic compounds with thick layers. These thick intermetallic compounds layers of both powder-Ni and foil-Ni made the specific contact resistance lower than that of sputtered-Ni. The specific contact resistance of foil-Ni interface was less than that of the powder-Ni interface because the foil-Ni made a uniform interface over the entire contact area. As a result, the SnSe metallization using the SPS co-sintering process with foil-Ni exhibited the lowest specific contact resistance.
Fig. 6 Contact resistance measurement results of Ag/Ni/SnSe samples for form of Ni metallization: powder, sputtered and foil-Ni.
The specific contact resistance of Ag/Ni/SnSe thermoelectric legs were 10-4 Ω·cm2 which was a high value for a metallization layer. Therefore, in order to verify the effect of contact resistance on the power output of the SnSe thermoelectric module, a simulation using the finite element method was conducted. The characteristics of an 8-couple SnSe module were investigated using COMSOL Multiphysics with measured thermoelectric properties (shown in Fig. 1). The cross-section area of a SnSe leg was 4 mm × 4 mm and the height was 6 mm. The thermoelectric properties of n-type SnSe were assumed to be 80% of those for p-type polycrystalline SnSe.
The resistance distribution and power output according to the contact resistance are shown in Fig. 7. The total resistance of the thermoelectric leg could be significantly different according to the specific contact resistance as shown in Fig. 7(a). The Cu electrode and Ag/Ni metallization layer had very low electrical resistance compared to SnSe, indicating negligible increases of resistance in this region. The resistance of Ag/Ni/SnSe thermoelectric leg with sputtered-Ni is around 0.23 Ω which is 15% larger than that of the foil-Ni thermoelectric leg. In the case of the Ag/Ni/SnSe with foil-Ni, the total resistance is almost same as the resistance of SnSe which is about 0.20 Ω. There was little effect on the increase in total resistance despite the high value of 10-4 Ω cm2, compared to the other thermoelectric legs.
Fig. 7 (a) Simulation result of total resistance of thermoelectric legs for each of the metallization processes, (b) the power output simulation result according to the contact resistance for each metallization processes.
The power output of the simulated 8-couple SnSe thermoelectric module for each metallization process is shown in Fig. 7(b). If this module had no contact resistance, the maximum power output could reach ∼1.1 W. The thermoelectric module with powder-Ni and sputtered-Ni had only 80% and 54% power output, respectively, due to contact resistance. The thermoelectric module with foil-Ni which specific contact resistance was 0.23 mΩ cm2 had only 5% power output loss. The reason why there was a little effect on the power output despite the high value of specific contact resistance is the low electrical conductivity of SnSe materials. Polycrystalline SnSe has lower electrical conductivity (60 S/cm at 823 K) than other thermoelectric materials used in intermediate temperature such as skutterudite (∼1600 S/cm at 776 K), half-Heusler (∼1700 S/cm at 776 K) [24,25]. Therefore, the relative ratio of the contact resistance to the resistance of thermoelectric materials was low despite the high contact resistance in the case of SnSe. As a result, SnSe material research to enhance electrical conductivity and research about minimizing contact resistance should both be considered to fabricate highly efficient SnSe thermoelectric module with high power output.
In this study, a suitable process to form a Ag/Ni metallization layer on SnSe thermoelectric materials was introduced and the interfacial characteristics were analyzed. The SPS co-sintering process induced rigorous reaction between Ni and SnSe because these were treated at high temperature and high pressure. With SPS co-sintering, thick Ni layers led to formation of Ni3Sn2, and Ni5.63SnSe2 intermetallic compounds that could act as a diffusion barrier and electrical contact. The SPS co-sintering process with foil-Ni could make effective diffusion barrier and minimize the specific contact resistance with SnSe. The Ag/Ni/SnSe thermoelectric module with foil-Ni, which had specific contact resistance of 0.23 mΩ cm2, was demonstrated to have a power loss of only 5% in the simulation results. However, SnSe had the weakness of low electrical conductivity compared to the other materials which resulted in low power output with high load matching. To make practical SnSe thermoelectric modules, it will be necessary to enhance the SnSe thermoelectric material as well as develop techniques for fabricating thermoelectric modules with SnSe.
This work was supported by the Korea Institute of Energy Technology Evaluation and Planning (KETEP) and the Ministry of Trade, Industry & Energy (MOTIE) of the Republic of Korea (No. 20172010000830). This research was also supported by the National Research Foundation of Korea (NRF) Grant funded by the Korean Government (MSIP) (NRF-2015R1A5A1036133).
The authors have declared that no competing interests exist.
WeChat
/
| 〈 |
|
〉 |
