Electronic thermoelectric power factor and metal-insulator transition in FeSb${}_2$

We show that synthesis-induced metal-insulator transition (MIT) for electronic transport along the orthorombic $c$ axis of FeSb${}_2$ single crystals has greatly enhanced electrical conductivity while keeping the thermopower at a relatively high level. By this means, the thermoelectric power factor is enhanced to a new record high S${}^2$$\sigma$ $\sim$ 8000 $\mu$W K${}^{-2}$ cm${}^{-1}$ at 28 K. We find that the large thermopower in FeSb${}_2$ can be rationalized within the correlated electron model with two bands having large quasiparaticle disparity, whereas MIT is induced by subtle structural differences. The results in this work testify that correlated electrons can produce extreme power factor values.

Figure 1

(a) Comparison of $c$-axis resistivity for two crystals. In 9 T $\rho \left(T\right)$ in crystal 1 above 80 K is consistent with semiconducting thermally activated transport with an activation gap of 120 K. (b),(c) Temperature regime(s) where $\rho \left(T\right)$ is dominated by thermally activated transport. The arrow in (b) marks the MIT in crystal 1, where $d\left[\text{ln}\right(\rho \left)\right]/dT$ changes sign. Broken lines in both panels are linear fits to activation-type behavior.

Figure 2

The $\rho$${}_{xy}$ (a) and Hall constant (b) for crystal 2. Hall mobility (c) where the hole (electron) carriers are denoted by $+$ ($-$) and carrier concentrations of high and low mobility carriers (d) as a function of temperature for both crystals.

Figure 3

(a) $\left|S\right(T\left)\right|$ for FeSb${}_2$ crystals 1 (red) and 2 (blue squares). Red and blue balls show fits to the two-band noninteracting semiconductor model. Fits to the correlated electron model for crystal 1 are shown by red, purple, and brown (crystal 1) and dark blue, light blue, and violet (crystal 2) solid lines that correspond to increasing values of fixed energy gap $\Delta$ (see text). The ratio of individual band mobilities in a two-carrier model is shown by the red/orange (crystal 1) and dark/light blue squares (crystal 2).

Figure 4

The low resistivity around MIT leads to a record high TPF. Crystal 1 has two orders of magnitude higher TPF between 8 and 100 K.

Figure 5

Structural unit of FeSb${}_2$ (a) shows $2\times 2\times 2$ Pnnm unit cell as seen from the top of the $c$ axis (left), and from the top of the $a$ axis (right). Shaded rectangles denote short Fe-Sb distances within the FeSb${}_6$ octahedra. Intrachain and interchain Fe-Sb-Fe angles are denoted as $\beta$${}_1$ and $\beta$${}_2$, respectively. Next-nearest-neighbor interchain Fe-Fe distance is indicated by the double arrow. (b) Temperature dependence of ${\beta }_1$ and ${\beta }_2$ for crystal 1 (red) and crystal 2 (blue). (c) ADP factors for Fe (anisotropic) and Sb (isotropic) for crystal 1 (red) and crystal 2 (blue). Vertical dashed lines in (b) and (c) at 40 and 100 K indicate temperatures of insulator-metal transition and change of the nature of the charge carriers, respectively.

Figure 6

Experimental neutron RDF data (inset) at 20 K reveals no detectable difference, suggesting that the disorder observed in x-ray experiment probably originates from the charge sector.