Magnetically Driven Metal-Insulator Transition in ${\mathrm{NaOsO}}_3$

The metal-insulator transition (MIT) is one of the most dramatic manifestations of electron correlations in materials. Various mechanisms producing MITs have been extensively considered, including the Mott (electron localization via Coulomb repulsion), Anderson (localization via disorder), and Peierls (localization via distortion of a periodic one-dimensional lattice) mechanisms. One additional route to a MIT proposed by Slater, in which long-range magnetic order in a three dimensional system drives the MIT, has received relatively little attention. Using neutron and x-ray scattering we show that the MIT in ${\mathrm{NaOsO}}_3$ is coincident with the onset of long-range commensurate three dimensional magnetic order. While candidate materials have been suggested, our experimental methodology allows the first definitive demonstration of the long predicted Slater MIT.

Figure 1

Slater MIT. (Left) commensurate AFM order doubles the unit cell. (Right) Simplified band structure. The opposite potential on each neighboring ion in the magnetic regime splits the electronic band, creating an insulating band gap.

Figure 2

Refined structural parameters from NPD. The vertical dashed line indicates $T_{\mathrm{MIT}}$. The insets of (d) and (e) are a representation of the crystallographic bond lengths and angles considered, with the larger spheres indicating the Os ions and the small spheres indicating the oxygen ions. The crystal structure shows an anomaly in the $a$ and $c$ lattice constants at $T_{\mathrm{MIT}}$. However, there is no indication of a structural symmetry change.

Figure 3

(a)–(b) NPD modeled with crystallographic space group $Pnma$ and $G$-type AFM ordering. Magnetic order is shown schematically inset, with only the Os ions represented for clarity. (c) Polarized NPD with polarization parallel to the scattering vector with flipper-on (triangles) and flipper-off (squares) unambiguously assigns the scattering centered around $|Q|=1.43{\AA }^{-1}$ as purely magnetic. The lower $|Q|\approx 1.35{\AA }^{-1}$ structural peak corresponds to an unknown non-magnetic impurity phase. (d) Integrated intensity of the $|Q|=1.43{\AA }^{-1}$ peak as a function of temperature. The solid line is a power law fit as described in the text.

Figure 4

Measurement of the resonant enhancement at the $L2$ and $L3$ edges. $\sigma$ ($\pi$) corresponds to polarization perpendicular (parallel) to the scattering plane. The $\sigma \mathrm{\text{−}}\pi$ scattering, sensitive to resonant magnetic scattering, is evident at the expected resonant energies, with a similar enhancement of $\sim {10}^2$ at the $L2$ and $L3$ edges. In the $\sigma \mathrm{\text{−}}\sigma$ channel no scattering is observed, showing that the resonant intensity is $\pi$ polarized, indicating the resonant enhancement is purely from magnetic scattering. The 4 pairs of solid lines in the top portion of the figure are results from FDMNES calculations. No SOC indicates no splitting of the $t_{2g}$ manifold degeneracy, SOC indicates splitting. The ratio of $L3:L2$ intensities is shown.