Journal of Materials Science & Technology  2020 , 40 (0): 113-118 https://doi.org/10.1016/j.jmst.2019.08.046

Low contact resistivity and long-term thermal stability of Nb0.8Ti0.2FeSb/Ti thermoelectric junction

Zhijie Huanga, Li Yinb, Chaoliang Hua, Jiajun Shena, Tiejun Zhua*, Qian Zhangb, Kaiyang Xiaa, Jiazhan Xina, Xinbing Zhaoa

a State Key Laboratory of Silicon Materials and School of Materials Science and Engineering, Zhejiang University, Hangzhou 310027, China
b Department of Materials Science and Engineering, Institute of Materials Genome & Big Data, Harbin Institute of Technology, Shenzhen 518055, China

Corresponding authors:   *Corresponding author.E-mail address: zhutj@zju.edu.cn (T. Zhu).*Corresponding author.E-mail address: zhutj@zju.edu.cn (T. Zhu).

Received: 2019-08-4

Revised:  2019-08-23

Accepted:  2019-08-26

Online:  2020-03-01

Copyright:  2020 Editorial board of Journal of Materials Science & Technology Copyright reserved, Editorial board of Journal of Materials Science & Technology

More

Abstract

Although half-Heusler compounds are quite promising for thermoelectric power generation, there is only limited research on the interfacial structure between metal electrode and half-Heusler compounds for device applications. This work reports on the characteristics of Nb0.8Ti0.2FeSb/Ti junction and its evolution behavior during 973 K. The Nb0.8Ti0.2FeSb/Ti interface consists of one Ti0.9Fe0.1 layer and one Fe-poor layer. There is an Ohmic contact and a low contact resistivity (0.15 μΩ cm-2) in this junction, on account of the matching of working functions between Nb0.8Ti0.2FeSb and Ti0.9Fe0.1 interlayer. The high doping of Ti high carrier concentration in NbFeSb matrix leads to a high carrier concentration, which results in inducing a large tunneling current at this interface. After aging treatment at 973 K, the Fe-poor layer and the Ti0.9Fe0.1 layer continues to expand, resulting in the increase of the thickness of the interfacial layer and the contact resistivity. The interfacial electrical is only 1.9 μΩ cm-2 after 25 days’ aging. The thickness of the interface layer has a good linear relation with the square root of aging time, which firmly indicates that the growth of the layer is determined by mutual diffusion of Fe and Ti atoms across the interface. The low contract resistivity and long-time thermal stability demonstrate the great potential of Nb0.8Ti0.2FeSb/Ti thermoelectric junction in high efficiency half-Heusler TE devices.

Keywords: Thermoelectric materials ; Half-Heusler alloys ; Electrode ; Contact resistivity ; Thermal stability

0

PDF (2075KB) Metadata Metrics Related articles

Cite this article Export EndNote Ris Bibtex

Zhijie Huang, Li Yin, Chaoliang Hu, Jiajun Shen, Tiejun Zhu, Qian Zhang, Kaiyang Xia, Jiazhan Xin, Xinbing Zhao. Low contact resistivity and long-term thermal stability of Nb0.8Ti0.2FeSb/Ti thermoelectric junction[J]. Journal of Materials Science & Technology, 2020, 40(0): 113-118 https://doi.org/10.1016/j.jmst.2019.08.046

1. Introduction

Thermoelectric (TE) materials can directly convert heat into electricity, providing an environmentally friendly approach to recycle waste heat. The conversion efficiency of a TE device mainly depends on the materials’ figure of merit zT [1,2]. In the past decades, the TE figure-of-merit of many TE materials have been obviously improved, such as half-Heusler compounds and skutterudites [[3], [4], [5]]. On the other hand, in order to make a practical TE device, which has a high conversion efficiency and good reliability, there are many other crucial issues remaining to be resolved, such as the contact resistivity between the TE material and electrode. Its exact impact on the performance of TE devices could be expressed by the following formula: zTD = zTML/(L + 2ρcσ), where the zTD, zTM, ρc, σ and L denote the effective zT of the TE device, the average zT of TE materials, the contact resistivity between the TE material and electrode, the electrical conductivity of TE material and the length of the TE leg, respectively [6]. In order to reduce the negative effect of the contact resistivity to a negligible level, it should be controlled to less than 1 μΩ cm-2 [6]. Nowadays, some good junctions have been used in TE power generators or refrigerators [7].

When choosing an electrode, the match of coefficient of thermal expansion (CTE) between electrode and thermoelectric material has to be considered. The CTE mismatch might result in many cracks at the interface and make a TE device break down. Also, for the interface between a semiconductor and a metal, the work function should match to form an Ohmic contact other than a Schottky contact [9,10]. Due to the fact that the change of the interface mainly accounts for the formation of contact resistivity, the electrode should inactively react with the TE material [11].

Half-Heusler compounds are one of the most promising TE materials, due to their high zTs and outstanding mechanical properties [2,[12], [13], [14], [15]]. Among them, NbFeSb is the best p-type TE materials. NbFeSb with 20%Ti dopant has a high zT of 1.1 at 1050 K [16]. Hf doping can further boost zT, due to the stronger point scattering to reduce the lattice thermal conductivity [17]. Alloying Ta at Nb site can significantly reduce the lattice thermal conductivity but keep the electrical properties almost unchanged because of the similar atomic size between Nb and Ta, and the zT eventually rises to 1.6 at 1200 K [18]. Recently, it is reported that TaFeSb-based Half-Heusler compounds also show a promising TE performance [3]. By optimizing the thermal and electrical properties, Ta0.74V0.1Ti0.16FeSb finally reaches a peak zT of 1.52 at 973 K and also gains a quite high average zT of 0.93 between 300 and 973 K, surpassing other sorts of p-type Half-Heusler compounds [3]. Additionally, DFT calculations have recently been used to explore new stable 18 valence electrons Half-Heusler compounds with intrinsic cation vacancies [19]. All these works demonstrate that Half-Heusler compounds are quite promising for TE power generation. However, only limited efforts have been devoted to investigate the interfacial behavior of half-Heusler compounds and electrodes [20,21]. A recent study shows that a TE module, which is made of n-type ZrNiSn-based alloys and p-type FeNbSb compounds, exhibits a conversion efficiency of 6.2% and a power output of 8.9 W under a temperature difference of 655 K, higher than 4.5% for the commercial half-Heusler compounds module based on p-type ZrCoSb-based alloys and n-type ZrNiSn-based alloys [17]. However, the contact resistivity accounts for about 3.2% efficiency loss, which indicates that more effort is required to reduce the contact resistivity in order to further boost the conversion efficiency [17]. In addition, the junction between Hf0.5Zr0.5CoSn0.2Sb0.8 and Ag has a value of contact resistivity of 89 μΩ cm-2 at the temperature of 773 K [22], implying that Ag is not a good electrode for ZrCoSb based half-Heusler alloys. Recently, the Nb0.8Ti0.2FeSb/Mo TE junction has been reported to exhibit a good contact performance [23]. The contact resistivity between the Nb0.8Ti0.2FeSb and Mo is smaller than 1 μΩ cm-2, satisfying preliminary demand due to the match of work functions between Nb0.8Ti0.2FeSb and FeMo interlayer in this junction [23]. However, during a longtime aging treatment at 1073 K, the FeMo alloy layer transforms into a FeSb2 alloy layer and the appearance of Nb3Ti phase, and the contact resistivity reaches 18.4 μΩ cm-2 after 32 days’ aging treatment [23].

In this work, Ti is chosen as the metal electrode for Nb0.8Ti0.2FeSb TE material, due to its CTE matching with Nb0.8Ti0.2FeSb. The interfacial characteristics and evolution behavior of Nb0.8Ti0.2FeSb and Ti are investigated. An Ohmic contact and a low contact resistivity (0.15 μΩ cm-2) are realized and the contact resistivity is up to 1.9 μΩ cm-2 after 25 days’ aging, which is smaller compared with that of Nb0.8Ti0.2FeSb/Mo junction, verifying that the Nb0.8Ti0.2FeSb/Ti TE junction has a very low contact resistivity and good thermal stability with a great potential for making high efficiency half-Heusler TE devices.

2. Materials and methods

2.1. Preparation

The starting materials, Nb (foil, 99.98%), Ti (rod, 99.99%), Fe (piece, 99.97%), and Sb (block, 99.999%) were weighed according to the nominal composition of Nb0.8Ti0.2FeSb. Ingots were prepared by levitation melting under argon (99.999%) atmosphere. The details can be found elsewhere [24]. Each ingot is about 20 g, and these ingots were re-melted for three times to ensure homogeneity. After that they were mechanically milled (Mixer Mill MM200, Retsch) for an hour under argon protection. These powders were loaded into a cylinder-shaped graphite die and sintered by spark plasma sintering (SPS-1050, Sumitomo Coal Mining Co.) at 1073 K for 20 min under 65 MPa in the vacuum. The relative densities of all these samples exceed 95%. These samples were annealed at 1023 K for 3 days. Then their surfaces were polished carefully. Finally, the Nb0.8Ti0.2FeSb bulk and the Ti powders were loaded into the graphite die and sintered together by spark plasma sintering (SPS) at 1023 K for 10 min under 65 MPa in the vacuum to obtain the Nb0.8Ti0.2FeSb/Ti junction (Fig. 1).

Fig. 1.   The sketch map of the final sintering process to obtain the Nb0.8Ti0.2FeSb/Ti junction.

2.2. Characterizations and measurements

The element mapping was observed by energy disperse spectroscopy (EDS). Phase structures of the Nb0.8Ti0.2FeSb bulk were investigated by X-ray diffraction (XRD) on a RigakuD/MAX-2550PC diffractometer using Cu Kα radiation (λ0 = 1.5406 Å). The back scattering electron image and elemental composition were obtained by EPMA (JOEL, JXA-8100). The I-V characteristic of the interface between Nb0.8Ti0.2FeSb and Ti was measured by using a semiconductor parameter analyzer (Agilent E5270B). The contact resistivity was measured by four-probe contact resistivity tester. The work function of Nb0.8Ti0.2FeSb was measured by ultraviolet photoelectron spectroscopy (Thermo ESCALAB 250).

3. Results

3.1. Phase and microstructure characterization

The XRD patterns of Nb0.8Ti0.2FeSb bulk after annealing in Fig. 2 show that all the diffraction peaks can be indexed to the cubic MgAgAs-type crystal structure, and no obvious impurity phase is found in these samples.

Fig. 2.   XRD pattern of Nb0.8Ti0.2FeSb.

The microstructure and elemental mapping of the Nb0.8Ti0.2FeSb/Ti junction are showed in Fig. 3.The interface between Nb0.8Ti0.2FeSb and Ti mainly consists of two layers, which are Fe-poor layer and Ti0.9Fe0.1 layer, and the thickness of the interface is about 9 μm. The chemical composition of Ti0.9Fe0.1 layer is 90 at.% Fe and 10 at.% Ti. The chemical composition of Fe-poor layer is 51 at.% Ti, 27 at.% Sb and 22 at.% Nb. The Fe atoms in Nb0.8Ti0.2FeSb bulk seem to diffuse more quickly into Ti electrode than other kinds of atoms, therefore Fe-poor layer with a thickness of 1 μm forms at the side of Nb0.8Ti0.2FeSb bulk. Furthermore, the Fe atoms which diffuse into the Ti electrode and form Ti0.9Fe0.1 layer, according to the chemical composition, about 8 μm at the side of Ti electrode. Due to the interface reactions between Nb0.8Ti0.2FeSb and Ti, the junction has a very strong connection, which is beneficial for the mechanical properties of this junction.

Fig. 3.   (a) EPMA line scanning of the interface of the Nb0.8Ti0.2FeSb/Ti junction, (b) Back-scattering electron image, and (c) EDS mapping of the detail view of the interface of the Nb0.8Ti0.2FeSb/Ti junction.

3.2. Contact type and contact resistivity

The contact resistivity is partly determined by the barrier height of the depletion layer at the interface of metal-semiconductor contact [8]. For the interface between p-type semiconductor and metal, when the work function of semiconductor is greater than that of metal (ΦS > ΦM), a blocking layer in the semiconductor near the interface comes into being and leads to a Schottky contact and a large contact resistivity [8]. In contrast, when the work function of metal is greater than semiconductor (ΦM > ΦS), an anti-blocking layer in the semiconductor near the interface comes into being and leads to an Ohmic contact and a small contact resistivity [8].

The work functions of Nb0.8Ti0.2FeSb and Ti0.9Fe0.1 are measured to identify the contact type of Nb0.8Ti0.2FeSb/ Ti0.9Fe0.1 interface. The calculating formula of work function is as followed, Φ = - (ECutoff - EFermi), where ECutoff, EFermi, h and ν respectively represent the energy of Cutting edge, the energy of Fermi edge, Plank constant and the frequency of photon. The energy of the photons in this measurement is 21.22 eV. As shown in Fig. 4(a) and (b), the energy of Cutting edge of Nb0.8Ti0.2FeSb is 16.72 eV. Therefore, the work function of Nb0.8Ti0.2FeSb is 4.50 eV. Similarly, the work function of Ti0.9Fe0.1 is 4.58 eV. Because of ΦM > ΦS, the interface of Nb0.8Ti0.2FeSb/ Ti0.9Fe0.1 is an Ohmic contact.

Fig. 4.   (a) UPS spectrum of Nb0.8Ti0.2FeSb and Ti0.9Fe0.1, (b) The binding energy of the intensity of Nb0.8Ti0.2FeSb and Ti0.9Fe0.1. (c) I-V curve of the Nb0.8Ti0.2FeSb/Ti interface (d) Contact resistivity plot of the Nb0.8Ti0.2FeSb/Ti interface.

In order to further affirm the contact type at this interface, a measurement of electric current versus voltage (I-V) is conducted. As is shown in Fig. 4(c), for both of the forward and reverse currents, no obvious rectification characteristic occurs, which verifies the interface is an Ohmic contact.

The character energy E00 is a useful criterion in determining the dominant carrier transport mechanism at the junction of semiconductor and metal, which is defined as: [8]

(1)

where q, N, m*and εs are the electron charge, carrier concentration, effective mass and permittivity of a semiconductor. When the doping level in the semiconductor is low, which means that E00 << kBT, the thermionic emission dominates the carrier transport. When the doping level in the semiconductor is medium, which means that E00 ~ kBT, both the thermally excited carriers and tunneling carriers exist. When the doping level in the semiconductor is high, which means that E00 >> kBT, the field emission dominates the carrier transport, where the carriers traverse the potential barrier all by tunneling effect. Nb0.8Ti0.2FeSb compound is a heavily doped semiconductor with Ti doping content up to 20%. The carrier concentration in Nb0.8Ti0.2FeSb is 2.2 × 1021 cm-3, the effective mass of the carrier is 1.6me and the relative permittivity is 80 [16,23]. The calculated E00/kBT = 20 indicates that the field emission should dominate the interface carrier transport between Ti0.9Fe0.1 and Nb0.8Ti0.2FeSb.

As discussed above, due to the Ohmic contact at the FeNb0.8Ti0.2Sb/Ti interface, the contact resistivity would be small. The measured contact resistivity is presented in Fig. 4(d). By linearly fitting the electrical resistance curve, the electrical conductivity is determined to be 6.2 × 105 S/m for Nb0.8Ti0.2FeSb and 2.4 × 106 S/m for Ti, which agrees well with the measured electrical conductivity for them, indicating that the measurement result is valid. There is not an abrupt change in the slope of interface layer, and the contact resistivity is about 0.15 μΩ cm-2. This contact resistivity is extremely small (<1 μΩ cm-2) and preliminarily meets the demand for a TE device.

3.3. Thermal stability

A TE device needs to serve for a long time in the situation where there exists a temperature difference. Therefore, a series of aging treatments at 973 K was conducted to evaluate the stability of the Nb0.8Ti0.2FeSb/Ti junction. Fig. 5 shows that the back-scattering electron image and the EPMA line scanning of the Nb0.8Ti0.2FeSb/Ti junction for different aging periods. The Fe-poor layer increased to 5 μm thick after aging of 5 days and the Ti0.9Fe0.1 layer extended to about 30 μm. After 10 days’ aging, the Fe-poor layer extended to 10 μm and the Ti0.9Fe0.1 layer increases to about 55 μm due to the diffusion of Fe atoms from the side of Nb0.8Ti0.2FeSb to Ti side. From 10 days to 25 days, there was almost no change in the thickness of the Fe-poor lay, whereas the Ti0.9Fe0.1 layer increased to 80 μm continuously. There are not many obvious holes in the Nb0.8Ti0.2FeSb/Ti interface even after 25 days’ aging, which is due to the good match of the CTE between Nb0.8Ti0.2FeSb and Ti. The CTE of Nb0.8Ti0.2FeSb is 10.6 × 10-6/K and that of Ti is 8.6 × 10-6/K. Finally, it can be concluded that the formation of interface layer structures is mainly due to the mutual diffusion of Fe and Ti elements through the interface.

Fig. 5.   Back-scattering electron image of the Nb0.8Ti0.2FeSb/Ti junction after an aging treatment of (a) 5 days, (b) 10 days, (c) 15 days, (d) 20 days (e) 25 days at 973 K. The EPMA line scanning of the Nb0.8Ti0.2FeSb/Ti junction after an aging time of (f) 20 days and (g) 25 days at 973 K. (h) The thickness of interface layer versus aging time.

The thickness of a reaction layer can be expressed as: [25]

Y = Y0 + (Dt)n(2)

where Y, Y0, D, t and n represent the thickness of interface layer at time t, the original thickness of interface layer, the growth rate and time exponent, respectively. This equation has been used in explaining the thickness of interface layer versus aging time in the CoSb3/Ti/Mo-Cu TE joint [25]. During the aging treatment, the average thickness of interface layer is determined by the diffusion of Fe and Ti atoms through the interface of Nb0.8Ti0.2FeSb/Ti. So the time exponent is 0.5, which means that the thickness of the interface layer must observe the square root time law: Y = Y0 + (Dt)0.5.

As is shown in Fig. 5(h), the average thickness of interface layer has a pretty good linear relation with the aging time for each sample in this experiment, which verifies that the growth of the interface layer is diffusion-controlled process in Nb0.8Ti0.2FeSb/Ti interface. Therefore, the interfacial evolution could be briefly summarized to 3 processes. Firstly, the Fe-poor layer and Ti0.9Fe0.1 layer come into being during SPS simultaneously. Then, the Fe-poor layer expands to its peak at 10 μm and Ti0.9Fe0.1 layer increases continuously to 30 μm. Finally, the Fe-poor layer stops expanding, but the Ti0.9Fe0.1 alloy layer increases gradually, forming the thickest part of interface layers. All these processes involve the diffusion of Fe and Ti atoms. Although the whole thickness of interface layer lays is only about 90 μm, further work could be done to restrain such diffusion to improve the stability of the Nb0.8Ti0.2FeSb/Ti junction.

The curve of aging time versus electrical resistance for Nb0.8Ti0.2FeSb/Ti interface is shown in Fig. 6. The contact resistivity increases with the aging time. Before 10 days’ aging, the contact resistivity slightly increases due to the increase of both Fe-poor layer and Ti0.9Fe0.1 layer, reaching 0.2 μΩ cm-2 after 5 days’ aging. However, the contact resistivity shows an obvious rise after 10 days’ aging, and finally reaches 1.9 μΩ cm-2 after 25 days’ aging generally due to the increase of Ti0.9Fe0.1 layer. This growth rate for Nb0.8Ti0.2FeSb/Ti is smaller than that of Nb0.8Ti0.2FeSb/Mo, whose contact resistivity is 2.6 μΩ cm-2 after 8 days’ aging and reaches 13.9 μΩ cm-2 after 24 days’ aging. So Nb0.8Ti0.2FeSb/Ti is a more promising junction to make a half-Heusler module in the future.

Fig. 6.   (a) Curve of aging time versus electrical resistance at room temperature for Nb0.8Ti0.2FeSb/Ti interface aged at 973 K. (b) Contact resistivity plot of the Nb0.8Ti0.2FeSb/Ti interface for different aging times at 973 K.

4. Conclusion

The contact resistivity and interfacial evolution of the Nb0.8Ti0.2FeSb/Ti junction have been systematically investigated. A Fe-poor layer and a Ti0.9Fe0.1 layer were formed at the interface during SPS. The work function of Nb0.8Ti0.2FeSb is lower than that of the Ti0.9Fe0.1 interlayer and they form an Ohmic contact. The contact resistivity is about 0.15 μΩ cm-2. The field emission was found to be the dominant carrier transport mechanism in this interface layer. After a long-time aging treatment at 973 K, the Ti0.9Fe0.1 layer and Fe-poor layer increasingly expand. Finally, the contact resistivity reaches 1.9 μΩ cm-2 after 25 days’ aging. The thickness of interface layers rises linearly with square root of aging time, indicating that the growth of interface layer layers is determined by the mutual diffusion of Fe and Ti atoms.

Acknowledgements

We would like to thank Mr Siqin Li, School of Materials Science and Engineering, Zhejiang University, for the measurement of I-V curve. This work was supported by the National Key Research and Development Program of China (2018YFB0703604), the National Science Fund for Distinguished Young Scholars (No. 51725102), and the Natural Science Foundation of China (Nos. 51761135127, 61534001 and 11574267).


/