Oxygen vacancies in vanadium dioxide: A DFT$+V$ study

TL;DR

The paper investigates how oxygen vacancies affect the structural and electronic properties of VO$_{2}$ across the rutile R and monoclinic M1 phases using DFT$+V$, which includes a static intersite term $V$ for V–V interactions. By analyzing R, M1, and vacancy configurations with varying defect concentrations, the authors find that vacancies induce localized lattice distortions but do not destroy V–V dimerization; the M1 phase becomes metallic due to partial filling of the conduction band, consistent with a rigid-band-like doping scenario. Comparisons with a background-charge model show that vacancy effects can be captured by simple electron addition to the conduction band without triggering global de-dimerization, supporting a Peierls-like separation of electronic degrees of freedom. The results elucidate a dual mechanism for MIT suppression: a structural reduction of the transition temperature and an electronic carrier-doping effect, informing defect engineering strategies for VO$_2$ and motivating future incorporation of stronger correlation methods (e.g., DFT+DMFT) to assess vacancy-site correlation effects.

Abstract

We present a density-functional theory study of the effects of oxygen vacancies on the structural and electronic properties of vanadium dioxide (VO$_2$). Our motivation is the reported suppression of the metal-insulator transition by oxygen vacancies and the lack of a clear consensus on its origin. We use the DFT$+V$ method with a static intersite vanadium-vanadium interaction term, $V$, to calculate the properties of the oxygen-deficient metallic rutile and insulating monoclinic M1 phases of VO$_2$ on the same footing. We find that oxygen vacancies induce local distortions in the M1 phase, but do not destroy the dimerization usually associated with the insulating behavior. In spite of this, we find that the M1 phase becomes metallic as a result of the partial filling of the conduction band due to a rigid-band-like doping effect.

Figures (8)

• Figure 1: Crystal structure of the M1 phase in the M1 primitive cell (dashed line). Green and blue arrows indicate the short and long bonds (SB and LB, respectively) along the $c$ direction. V (O) atoms shown in (light) gray. Inequivalent oxygen sites in the M1 structure (O$_\text{SB}$ and O$_\text{LB}$) are indicated in green and blue, respectively.
• Figure 2: (a) Detail of the primitive cell M1-O$_\text{SB}$ vacancy WF. Yellow and blue colors indicate positive and negative phase, respectively, V (O) atoms shown in (light) gray. (b) PDOS of the primitive cell with purple (green) colored lines indicating the $a_{1g}$ ($e_g^\pi$) WF. O$_V$ shown in orange. (c) Band structure of the primitive cell M1-O$_\text{SB}$ with the O$_V$ contribution shown in orange.
• Figure 3: The trace of the intersite occupation matrix, $\text{Tr}(n_\times)$, for neighboring V--V pairs along $c$ in (a) the M1-O$_\text{SB}$ and (b) the M1-O$_\text{LB}$ structures (blue) compared to stoichiometric M1 values (yellow).
• Figure 4: Unfolded band structures and PDOS of the (a-d) M1-O$_\text{SB}$, (e-h) M1-O$_\text{LB}$, and (i-l) R-O configuration before (a, b, e, f, i, j) and (c, d, g, h, k, l) after internal relaxation. Colors in the PDOS denote orbital character: purple ($a_{1g}$), green ($e_g^{\pi}$), and orange (O$_V$). Colors in the band diagram denote spectral weight $A(\mathbf{k},E)$ of the unfolded band structure.
• Figure 5: Nearest-neighbor V--V distances (purple) in the internally relaxed (a) M1-O$_\text{SB}$, (b) M1-O$_\text{LB}$, and (c) R-O structures. The stoichiometric reference values for M1 and R are shown in yellow and green, respectively.