High thermoelectric figure of merit of Mg2Si0.55Sn0.4Ge0.05 materials doped with Bi and Sb

Available online at www.sciencedirect.com

Scripta Materialia 69 (2013) 606–609 www.elsevier.com/locate/scriptamat

High thermoelectric figure of merit of Mg2Si0.55Sn0.4Ge0.05 materials doped with Bi and Sb A.U. Khan, N. Vlachos and Th. Kyratsi⇑ Department of Mechanical & Manufacturing Engineering, University of Cyprus, 1678 Nicosia, Cyprus Received 6 June 2013; revised 3 July 2013; accepted 6 July 2013 Available online 16 July 2013

Thermoelectric properties of new Bi- and Sb-doped Mg2Si0.55Sn0.4Ge0.05 compounds prepared by powder methods were studied in the temperature range 300–823 K. The materials exhibited compositional inhomogeneites consisting of Sn-rich and Sn-poor areas. Doping with Bi or Sb had a very strong influence on the thermoelectric properties. A high figure of merit was obtained, with a value 1.4 for Bi members and 1.2 for Sb members at high temperatures. These values are the highest reported on this system. Ó 2013 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Thermoelectric materials; Silicides; Sintering; Mapping

The efficiency of a thermoelectric material is estimated by the so-called figure of merit (ZT), defined by the formula ZT = rS2T/j, where S is the Seebeck coefficient, T stands for temperature, r represents electrical conductivity, and j is the thermal conductivity. This formula is separated into two main parts: (i) rS2, which is known as the power factor; and (ii) the thermal conductivity, where all these parameters are dependent on each other [1]. However, it is important to have a high power factor and a low thermal conductivity to achieve a high ZT, and this is already reported in various materials [2]. Mg2Si-based materials are promising for thermoelectric conversion in the middle temperature range because of their attractive ZT, low density, low cost, non-toxicity and high abundance of constituents in the Earth’s crust. In the 1960s, it was shown [3] that Mg2X (X = Si, Ge, Sn) are promising candidates for thermoelectrics. Nikitin et al. [4] also presented partial phase diagrams of the Mg2(Si,Sn) pseudo-binary system, which present a miscibility gap, while Labotz et al. [5] reported a continuous formation of solid solution for Mg2(Si,Ge) series. Mg2(Si,Sn)- and Mg2(Si,Ge)-based thermoelectric materials have already shown a high ZT [6–13], but quaternary Mg2(Si,Sn,Ge) systems have not attracted much attention [14]. In Mg2Si-based compounds, Sb and Bi are used as n-type dopants [6–13], and Li and Ag as p-type dopants [15].

⇑ Corresponding author. Tel.: +357 22892267; e-mail: [email protected]

Among the Mg2Si1xSnx series, the Sn-rich members (i.e., x P 0.6) exhibit high thermoelectric ZT when doped with Sb. The highest ZT achieved [10] for Mg2Si0.4Sn0.6 was 1.1 at 800 K. Extensive work has been done recently addressing the convergence of the conduction bands and the enhancement of the thermoelectric properties at x = 0.65–0.70 [9]. The member with x = 0.6 (i.e., Mg2Si0.4Sn0.6) is situated in the boundary between the region of the formation of solid solutions and the miscibility gap that appears at Sn concentration of 0.4 6 x 6 0.6. The Mg2Si1xSnx series, therefore, presents an unusual microstructure that seems to contribute to the best, up to now, ZT value of 1.3 [16]. This paper reports the synthesis and properties of a new compound based on the Si-rich-side Mg2(Si,Sn) series doped with Bi or Sb. This compound is situated on the other side of the miscibility gap (Mg2Si0.6Sn0.4) compared with the compound studied in previous work (Mg2Si0.4Sn0.6) [16]. In addition, a small amount of Ge was added, aimed at introducing additional atoms in the lattice, which may increase complexity in the system and affect the thermoelectric properties. Samples were prepared by mixing the elemental powders with purity >99.9% (Alfa Johnson Matthey GmbH, Germany). The synthesis of the materials was carried out in three heating steps: The mixed powders were cold pressed and heated at temperatures up to 973 K to achieve partial reactivity. The materials obtained were subsequently ball-milled. The ball-milled powders were cold pressed again and annealed at temperatures up to

1359-6462/$ - see front matter Ó 2013 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.scriptamat.2013.07.008

A. U. Khan et al. / Scripta Materialia 69 (2013) 606–609

973 K. Finally, the powders were uniaxially hot pressed under an argon atmosphere at 1043 K and 80 MPa pressure. The samples were characterized by X-ray powder diffraction. The morphological characterization was carried out by scanning electron microscopy (SEM; Jeol) and energy dispersive X-ray spectrometry (EDX; Bruker). Thermal conductivity was measured by the laser flash technique at temperatures up to 823 K (LFA457, Netzsch). The electrical resistivity and Seebeck coefficient were measured simultaneously using ULVACZEM3 for all the samples from room temperature up to 823 K. Bi- and Sb-doped Mg2Si0.55Sn0.4Ge0.05 materials were synthesized for different x values in addition to x = 0. The xBi values were 0.0175 and 0.02 for Bi members, and the xSb values were 0.0075 and 0.0125 for Sb members. X-ray powder diffraction of all the samples was performed, and Figure 1 shows typical patterns of the material after each heating step for the Sb-doped member (Mg2Si0.55xSn0.4Ge0.05Sbx, x = 0.0125). It is clear that, during the first heating, most of the starting materials have reacted, and only a very small amount of residuals exists. The major phase is that of Mg2(Si,Sn,Ge), but there is not a single composition, as concluded from the wide/split peaks that correspond to this phase. After the second heating step, the Mg2(Si,Sn,Ge) phase peaks are less wide/split, suggesting better mixing and the formation of a more homogeneous material that corresponds to a narrower composition range. Finally, after the third heating step (hot pressing), the peaks are even narrower and closer to the composition of Mg2Si0.6Sn0.4. Although the main phase was always of the CaF2-type, a small amount of MgO was observed in all samples. Many research groups working with Mg2Si-based thermoelectric materials have already reported the presence of MgO in their samples [6,7,13,17]. In order to study the morphology of the Mg2Si0.55Sn0.4Ge0.05 materials, the hot pressed samples were examined by SEM/EDX (see Fig. 2). Based on the backscattered images, it is clear that the material presents compositional inhomogeneities. Owing to these compositional variations, EDX mapping was carried out for the different elements, i.e., Si, Sn and Ge. Figure 2b, c and d shows the distribution of Si, Sn and Ge, respectively. It is observed that there are Sn-poor (Si-rich) and Sn-rich (Si-poor) areas. These areas correspond to darker and brighter regions in the backscattered image, respectively. Where Si is more evident,

after Hot Press ° Mg2(Si/Sn/Ge)

°

°

after Hot Press

♦MgO

° ♦

°

°

°

°

°

°

° nd

after 2 heating

Intensity

♦

°

nd

after 2 heating

° Mg2(Si/Sn/Ge) ♦MgO

°

°

°°

°

°

Intensity

°

°

st

°

° ∗ ∗ ∗

° 10

20

after 1 heating ° Mg2(Si/Sn/Ge)

30

♦

°

°

°

40

50 o

2θ ( )

°° 60

st

after 1 heating

∗ Mg ♦MgO

°

70

° 80

30

40

50

o

2θ ( )

Figure 1. XRD patterns of Mg2Si0.55xSn0.4Ge0.05Sbx (x = 0.0125) material after each heating step during synthesis.

(a)

607 (b)

Increasing concentration

(c)

(d)

Figure 2. Backscattered SEM image of the hot-pressed Mg2Si0.55xSn0.4Ge0.05Sbx, xSb = 0.0125, material and the elemental mapping images of (b) Sn, (c) Si and (d) Ge. The colors are qualitatively adjusted based on the concentration of each element separately.

less Sn exists, and this behavior is also evident in the Sn-rich members [16] of the Mg2Si1xSnx series (i.e., x = 0.6), where a Si-rich second phase as well as the complementary Si/Sn mapping is also observed. Ge distribution seems to be more uniform, although Ge-rich inclusions also exist. All these inclusions enhance the inhomogeneities and complexity of the microstructure and may affect the properties. The electrical resistivity and Seebeck coefficient were simultaneously measured at temperatures up to 823 K (see Fig. 3a and b). The Seebeck coefficient was negative throughout the investigated temperature range, showing the n-type character of these materials, as expected from the Si+4/(Bi,Sb)+3 substitution. When the Bi concentration increases, the Seebeck coefficient decreases (absolute values) from 395 lV K1 for xBi = 0 to 94 lV K1 for xBi = 0.0175 and xBi = 0.02. For Sb members, the Seebeck coefficient decreased again from 395 lV K1 for xSb = 0 to 133 lV K1 for xSb = 0.0075 and to 105 lV K1 for xSb = 0.0125 at room temperature. Moreover, the temperature dependence of all doped (Bi/Sb containing) materials is different compared with the undoped materials, owing to their higher doping level (i.e., higher carrier concentration). For the undoped material, the Seebeck coefficient increases initially, but soon reaches its maximum at 150 °C and then starts decreasing (see Fig. 3a), in agreement with the literature [14]. The Seebeck coefficient values are similar to those [9,10] of Mg2Si1xSnx for x = 0.4. In contrast, the electrical conductivity of the undoped material (x = 0) is low and increases with temperature, which is typical behavior for this semiconducting compound (see Fig. 3b). Doping has a significant influence on the electrical conductivity, and even a very low amount of dopant (xSb = 0.0075) changes the temperature dependence as a result of the increase in carrier concentration, in agreement with the Seebeck coefficient data. Moreover, the electrical conductivity increases with increasing Bi/Sb concentration. The temperature dependence for the Bi/Sb-containing materials is in agreement with the highly doped semiconductors. The highest electrical conductivity obtained in the present

608

A. U. Khan et al. / Scripta Materialia 69 (2013) 606–609

(a)

(b) 2000

Mg2Si0.55Sn0.4Ge0.05

0

Mg2Si0.55-xSn0.4Ge0.05Bix, x=0.0175 Mg2Si0.55-xSn0.4Ge0.05Sbx, x=0.0075

-100

Mg2Si0.55Sn0.4Ge0.05 Mg2Si0.55-xSn0.4Ge0.05Bix, x=0.0175

1800

Mg2Si0.55-xSn0.4Ge0.05Bix, x=0.02 Mg2Si0.55-xSn0.4Ge0.05Sbx,x=0.0125

-150 -200 -250 -300 -350

Mg2Si0.55-xSn0.4Ge0.05Bix, x=0.02

1600

Electrical Conductivity (S/cm)

Seebeck Coefficient (μV/K)

-50

Mg2Si0.55-xSn0.4Ge0.05Sbx, x=0.0075 Mg2Si0.55-xSn0.4Ge0.05Sbx,x=0.0125

1400 1200 1000 800 600 400

-400

200 0

-450 300

400

500

600

700

300

800

400

500

(c) 6

600

700

800

Temperature (K)

Temperature (K)

(d)

3

5

4

1.4 2

Mg2Si0.55Sn0.4Ge0.05 1.2

Mg2Si0.55-xSn0.4Ge0.05Bix, x=0.0175 Mg2Si0.55-xSn0.4Ge0.05Bix, x=0.02

1

1.0

Mg2Si0.55-xSn0.4Ge0.05Bix, x=0.0175 Mg2Si0.55-xSn0.4Ge0.05Bix, x=0.02 Mg2Si0.55-xSn0.4Ge0.05Sbx, x=0.0075

ZT

Lattice Thermal Conductivity (W/m/K)

Thermal Conductivity (W/m/K)

1.6

Mg2Si0.55-xSn0.4Ge0.05Sbx,x=0.0125

0.8

Mg2Si0.55-xSn0.4Ge0.05Sbx, x=0.0075 Mg2Si0.55-xSn0.4Ge0.05Sbx,x=0.0125

0

3

300

400

500

600

700

800

0.6

Temperature (K)

0.4 2

0.2 0.0

1 300

400

500

600

700

800

300

400

500

600

700

800

Temperature (K)

Temperature (K)

Figure 3. Temperature dependence of (a) Seebeck coefficient, (b) electrical conductivity, (c) thermal conductivity and (d) ZT of Mg2Si0.55xSn0.4Ge0.05(Bi,Sb)x material for xBi values of 0.0175 and 0.02 and xSb values of 0.0075 and 0.0125.

study corresponds to the Bi member, and is 1700 S at room temperature. The thermal conductivity of all the samples was measured for temperatures up to 823 K, and the results are shown in Figure 3c. These values correspond to the sum of electronic and lattice contribution and are considerably lower than those reported [10] on the Sb-doped Mg2Si1xSnx system for x = 0.4. Considering the small difference in the electrical conductivity between this work and the work of Zaitsev et al. [10], these results indicate a similar electronic contribution of thermal conductivity and thus significantly lower lattice thermal conductivity. The present values are closer to those from Liu et al. [9] for the Sb-doped Mg2Si1xSnx materials prepared with powder metallurgy techniques. Regarding the Bi-doped material, to the best of the present authors’ knowledge, there is no previous work on the Mg2Si0.6Sn0.4 member. For a better understanding of the influence of Sb and Bi in thermal conductivity, the electronic contribution (je) was subtracted from the total thermal conductivity (j), using the Wiedemann–Franz law, where je = LrT, using the measured values for electrical conductivity (r). The Lorentz number (L) was estimated, based on the Seebeck coefficient measurements as well as assuming scattering from acoustic phonons, following equations from the Fermi–Dirac statistics:   k B 2F 1 ðgÞ g S¼ e F 0 ðgÞ  2   kB 3F 0 ðgÞF 2 ðgÞ  4F 21 ðgÞ L¼ e F 20 ðgÞ

F i ðgÞ ¼

Z

1 0

xi dx 1 þ expðx  gÞ

where g is the reduced Fermi energy (EF/kBT), F i ðgÞ is the Fermi–Dirac integrals, and kB is the Boltzman constant. The remaining part j–je, corresponds to the lattice contribution in the thermal conductivity, since the bipolar contribution in the doped materials is eliminated. The values of the lattice thermal conductivity are low, and this can be attributed to their uncommon microstructural features owing to the coexistence of multicompositional areas (i.e., Sn-rich and Sn-poor areas and Ge-rich inclusions). The incorporation of Ge may enhance these special features. Interestingly, the Bidoped materials present lower lattice thermal conductivity than the Sb-doped materials. This can be attributed both to the higher amount of Bi that is introduced in the materials and to the heavier Bi compared with Sb, which causes stronger mass fluctuations and, as expected, stronger scattering of the phonons. The temperature dependence of ZT was estimated based on the above measurements, and is shown in Figure 3d. The maximum ZT is 1.4 at 800 K. This is the highest ZT ever achieved on the Mg2Si–Mg2Sn solid solution system compare to 1.3 for Sn-rich members [9] and 1.1 for Si-rich members [9–11]. For Sb-members, extensive work has been done on the Si-rich phase (i.e., x = 0.6), since this material exhibits remarkable properties. The ZT of Si –rich members prepared by direct melting of the components exceeded 1.1 at temperatures of 800–850 K. In addition, the Sirich members of the Mg2Si1xSnx series prepared with powder methods exhibited ZT of 1 at 800 K. In the present work, a ZT of 1.2 is presented and the variations compared with the work of Zaitsev et al. [10] are

A. U. Khan et al. / Scripta Materialia 69 (2013) 606–609

explained by the significant difference in thermal conductivity, while Liu et al. [9] presented a lower power factor. Interestingly, not much work has been done in the literature using Bi as the dopant. In previous work on the Mg2Si1xSnx system, Bi was used as the dopant [18] on the member x = 0.42 prepared by melting, and the outcome included higher Seebeck coefficient values, lower electrical conductivities and maximum ZT of 0.65. The thermal conductivities of the present work were higher than those presented by Du et al. [18], as expected based on the higher electronic contribution in the heat conduction. The Bi-doped member with x = 0.2, prepared by solid state reaction [19], had a ZT of 1.17 at 850 K. Mg2Si0.6Sn0.4-based materials are situated on the borderline of Mg2Si1xSnx solid solutions and the Si-rich side of the miscibility gap that exist in the Mg2Si–Mg2Sn phase diagram. Bi- and Sb-doped Mg2Si0.55Sn0.4Ge0.05 materials were successfully synthesized at temperatures up to 973 K, using powder techniques. The hot-pressed pellets exhibited microstructures consisting of multicompositional Sn-rich and Sn-poor regions. The small amount of Ge that was introduced seemed to increase the complexity in the microstructure of the materials through the formation of Ge-rich inclusions. ZT was estimated based on measurements of thermal conductivity, electrical conductivity and Seebeck coefficient. High values of 1.4 were found for xBi = 0.02, being the highest ZT reported on the Mg2Si–Mg2Sn system. These results show that Mg2(Si,Sn,Ge)-type solid solutions have great potential for thermoelectric devices, and the role of Ge needs to be further investigated to understand the origin of this high ZT. Furthermore, bismuth seems to be a more efficient dopant compared with antimony in terms of ZT, and the reason behind this is under investigation. Optimization of the composition in such quinary system is required and is in progress. This work is supported by the ThermoMag Project, which is co-funded by the European Commission in the 7th Framework Programme (contract NMP4-SL-2011-263207), by the European Space Agency and by the individual partner organizations. EDX mapping was done on equipment supported by

609

funding from the European Regional Development Fund and the Republic of Cyprus, through the Research Promotion Foundation (Project normal/0308/17). [1] G.S. Nolas, J. Sharp, A.H.J. Goldsmid, Thermoelectrics: Basic Principles and New Materials Developments, Springer, New York, 2001. [2] G.S. Nolas, J. Poon, M. Kanatzidis, MRS Bull. 31 (2006) 199. [3] E.N. Nikitin, V.G. Bazanov, V.I. Tarasov, Soviet Phys. Solid State 3 (1961) 2648–2651. [4] E.N. Nikitin, R.N. Tkalenko, V.K. Zaitsev, A.I. Zaslavskii, A.K. Kuznetsov, Inorg. Mater. 4 (1968) 1656–1659. [5] R.J. Labotz, D.R. Mason, D.F. O’Kane, J. Electrochem. Soc. 110 (1963) 127–134. [6] Q. Zhang, J. He, T.J. Zhu, S.N. Zhang, X.B. Zhao, T.M. Tritt, Appl. Phys. Lett. 93 (2008) 102109. [7] W. Liu, Q. Zhang, X. Tang, H. Li, J. Sharp, J. Electron. Mater. 40 (2011) 1062–1066. [8] M. Akasaka, T. Iida, K. Nishio, Y. Takanashi, Thin Solid Films 515 (2007) 8237–8241. [9] W. Liu, X. Tan, K. Yin, H. Liu, X. Tang, J. Shi, Q. Zhang, C. Uher, Phys. Rev. Lett. 108 (2012), 166601–1–5. [10] V.K. Zaitsev, M.I. Fedorov, E.A. Gurieva, I.S. Eremin, P.P. Konstantinov, A.Yu. Samunin, M.V. Vedernikov, Phys. Rev. B 74 (2006), 045207–1–5. [11] M.I. Fedorov, V.K. Zaitsev, G.N. Isachenko, Solid State Phenom. 170 (2011) 286–292. [12] H. Zhi-Ming, Z. Xin, L. Qing-Mei, Z. Jiu-Xing, Z. FeiPeng, J. Inorg. Mater. 27 (2012) 822–826. [13] X. Zhou, G. Wang, H. Chi, X. Su, J.R. Salvador, W. Liu, X. Tang, C. Uher, J. Electron. Mater. 41 (2012) 1589–1594. [14] (a) H. Yin, X.-B. Zhao, Q. Zhang, T.-J. Zhu, Int. J. Miner. Metall. Mater. 16 (2009) 564; (b) W. Wang, N. Mingo, Appl. Phys. Lett. 94 (2009) 203109. [15] Y. Isoda, S. Tada, T. Nagai, H. Fujiu, Y. Shinohara, J. Electron. Mater. 39 (2010) 1531–1535. [16] W. Liu, X. Tang, H. Li, K. Yin, J. Sharp, X. Zhou, C. Uher, J. Mater. Chem. 22 (2012) 13653. [17] G.S. Nolas, D. Wang, M. Beekman, Phys. Rev. B 76 (235204) (2007) 1–6. [18] Z. Du, T. Zhu, X. Zhao, Mater. Lett. 66 (2012) 76–78. [19] W. Luo, M. Yang, F. Chen, Q. Shen, H. Jiang, L. Zhang, Mater. Trans. 51 (2010) 288.