Evidence for a Correlated Insulator to Antiferromagnetic Metal Transition in CrN

We investigate the electronic structure of chromium nitride (CrN) across the first-order magnetostructural transition at $T_N\sim 286\mathrm{K}$. Resonant photoemission spectroscopy (PES) shows a gap in the $3d$ partial density of states at the Fermi level and an on-site Coulomb energy $U\sim 4.5\mathrm{eV}$, indicating strong electron-electron correlations. Bulk-sensitive high-resolution (6 meV) laser PES reveals a clear Fermi edge indicating an antiferromagnetic metal below $T_N$. Hard x-ray Cr $2p$ core-level PES shows $T$-dependent changes across $T_N$ which originate from screening due to coherent states as substantiated by cluster model calculations using the experimentally observed $U$. Electrical resistivity confirms an insulator above $T_N$ ($E_g\sim 70\mathrm{meV}$) becoming a disordered metal below $T_N$. Thus, CrN transforms from a correlated insulator to an antiferromagnetic metal, coupled to the magnetostructural transition.

Figure 1

(a) Temperature dependent resistivity $\rho (T)$ with inset showing the magnetostructural transition. (b) Using the expression $\rho (T)={\rho }_0{e}^{-(Eg/2k_{B}T)}$, a linear fit to $\ln \rho$ Vs $1/T$ yields the band gap $E_g\sim 70\mathrm{meV}$ in the region above $T_N$. (c) VB spectra obtained using soft x rays shows a highly reduced DOS at $E_F$. (d) Comparison of the on-resonance VB with laser VB, showing a clear Fermi edge in the latter. Inset shows the laser VB fitted with the resolution-convoluted Fermi Dirac function.

Figure 2

(a) Res-PES measured across the Cr $2p_{3/2}-3d$ threshold at 25 K using incident photon energies indicated by tick marks on the Cr $L_{3,2}$ XAS curve shown in (b). The calculated XAS using the $U$ obtained from res-PES compared to the experimental Cr $L_{3,2}$ XAS is also shown. (c) N $K$-edge XAS spectrum recorded at 300 and 25 K.

Figure 3

Cr $2p$ core-level spectra measured at (a) $T=300\mathrm{K}$ and (b) across $T_N$ (shown on expanded scale). The well-screened feature at the low BE side of $2p_{3/2}$ peak is reproduced in the calculated spectra using metallic screening channel at $E_F$, while the spectrum calculated without metallic screening does not match with the experimental spectra.