Finding merit in dividing neighbors

See allHide authors and affiliations

Science  08 Jan 2016:
Vol. 351, Issue 6269, pp. 124
DOI: 10.1126/science.aad8688

A large fraction of energy consumed by humankind is wasted as heat. Some of this energy can be recycled with thermoelectric devices that convert a thermal gradient into electricity. However, their wide adoption will require development of materials with high thermoelectric figure of merit (ZT) that lack rare or harmful elements. On page 141 of this issue, Zhao et al. (1) report on p-doped tin selenide (SnSe), which helps meet these goals.

Packing matters.

The IV-VI salts like SnSe and column-V elements like black phosphorus (P) crystallize in three different structures, according to their size and ionicity. Both distortions of the symmetric cubic rocksalt structure favor three out of the six nearest neighbors. Solid black lines represent bonds strengthened by the shift of neighbors toward each other. In the orthorhombic structure of SnSe, strong bonds are absent across layers (compare bonds between upper and middle layers of the inset figures), which generates substantial anisotropy (3, 4).


Favorable increases in ZT can be achieved by modifying a material (usually a semiconductor) to increase its Seebeck coefficient (a measure of the thermopower, or change in voltage induced by temperature) or electric conductivity or lower its thermal conductivity. However, this combination can be difficult to achieve—for example, good electrical conductors are often good heat conductors—and only a handful of materials meet all three requirements. Unfortunately, two of the most prominent, Bi2Te3 (the best near room temperature) and PbTe (the best near 800 K), contain rare tellurium or toxic lead.

Last year, this same research group identified SnSe as a thermoelectric material with a large ZT near 900 K, thanks to its remarkably low thermal conductivity (2). This latest study reports on p-doping of the SnSe single crystals (junctions require both n- and p-doped materials) and an increase in ZT, which, along one of the crystal axes, exceeds unity between 400 and 800 K. Thus, SnSe competes with both Bi2T3 and PbTe over a wide temperature window.

The newcomer to this arena is a member of a family of binary salts that contain elements of column IV and column VI of the periodic table. The IV-VI family and column V elements crystallize in one of the three varieties derived from the cubic rock-salt structure (see the figure) (3). Like black phosphorus, SnSe adopts the orthorhombic crystal structure, which imparts its anisotropic electrical conductivity.

The complexity emerging in the presence of only one or two types of atoms has intrigued condensed-matter physicists for decades (35). Why should atoms undergoing crystallization opt for anything less symmetric than cubic? The rock-salt structure can be seen as a network of interpenetrating octahedra with an atom at the center of each octahedron and its six nearest neighbors at the vertices. What drives lower symmetry is the problem of placing 10 electrons along three perpendicular axes with six equivalent neighbors. Rhombohedral and orthorhombic distortions divide the six immediate neighbors in two distinct sets of first neighbors and second neighbors.

Heavier and ionic salts in the bottom-right corner of the figure, like PbTe, avoid the distortion. Cubic symmetry is preserved in the presence of strong spin-orbit interaction and large ionicity because the six p and the four s orbitals become energetically distinct. With only p electrons to consider, it is much easier to keep six neighbors along three perpendicular axes. Moreover, ionicity, by distinguishing between partner atoms, impedes Peierels dimerization in each of the three perpendicular chains (5). In short, covalency favors rhombohedral distortion, whereas s-p hybridization leads to orthorhombicity and a layered anisotropic structure.

There are at least two links between this structural competition and the remarkable ZT in SnSe and PbTe. The first concerns the unusually low lattice thermal conductivity. Its magnitude at 800 K in both PbTe and SnSe implies that the phonon mean-free-path becomes as short as the interatomic distance (6), the result of strong phonon-phonon scattering caused by anharmonicity in proximity of the structural instability (7).

The second link is specific to SnSe in which ZT is large along one crystalline axis. Charge flows easily along the b axis, yet the Seebeck coefficient is as large as along the other axes. The anisotropy of conductivity (as in the case of black phosphorus) results from the layered structure of the orthorhombic crystal. However, the Seebeck coefficient remains almost isotropic because in a simple anisotropic Fermi liquid, the Fermi radius and the thermal fuzziness of the Fermi surface share the same anisotropy, and the Seebeck coefficient is set by the ratio of these two (6). Indeed, in numerous layered conductors, the Seebeck coefficient remains isotropic.

The present work of Zhao et al. will certainly inspire many other studies. Are other anisotropic conductors of the family as interesting? How different are the Fermi surface topologies among the family members? According to band calculations by Zhao et al., at a carrier concentration of 4 × 1019 cm−3, the Fermi surface of p-doped SnSe consists of several anisotropic pockets at low-symmetry points of the Brillouin zone. This structure remains to be checked by experimental study of quantum oscillations, as in the better-documented case of PbTe (8).


View Abstract

Navigate This Article