Local structure elucidation of tungsten-substituted vanadium dioxide (V[Formula: see text]W[Formula: see text]O[Formula: see text])

Abstract

Initially, vanadium dioxide seems to be an ideal first-order phase transition case study due to its deceptively simple structure and composition, but upon closer inspection there are nuances to the driving mechanism of the metal-insulator transition (MIT) that are still unexplained. In this study, a local structure analysis across a bulk powder tungsten-substitution series is utilized to tease out the nuances of this first-order phase transition. A comparison of the average structure to the local structure using synchrotron x-ray diffraction and total scattering pair-distribution function methods, respectively, is discussed as well as comparison to bright field transmission electron microscopy imaging through a similar temperature-series as the local structure characterization. Extended x-ray absorption fine structure fitting of thin film data across the substitution-series is also presented and compared to bulk. Machine learning technique, non-negative matrix factorization, is applied to analyze the total scattering data. The bulk MIT is probed through magnetic susceptibility as well as differential scanning calorimetry. The findings indicate the local transition temperature ([Formula: see text]) is less than the average [Formula: see text] supporting the Peierls-Mott MIT mechanism, and demonstrate that in bulk powder and thin-films, increasing tungsten-substitution instigates local V-oxidation through the phase pathway VO[Formula: see text] V[Formula: see text]O[Formula: see text] V[Formula: see text]O[Formula: see text].

Figure 1

Structural changes across the MIT illustrating V–V dimerization changes accompanying structure transformation from (a) $P{4}_2/mnm$ to (b) $P{2}_1/c$ upon cooling with $T_c$ occuring at 68 °C.

Figure 2

Local approximate strain is greater than long-range strain from PDF and Rietveld refinements. Both local and long-range strain are greater than theoretical strain calculated from effective ionic radii differences between W and V.

Figure 3

Comparison of the goodness-of-fit parameter, $R_{\mathit{wp}}$, from both XRD (> 30 Å) and PDF data (1.5–30 Å) illustrates how the local structure begins to transition from the monoclinic to the tetragonal structure before the average structure based on tungsten substitution amount. This exemplifies the need for greater local structure studies for predicting and explaining the properties if substituted VO${}_2$ is to be engineered and implemented in a functional capacity.

Figure 4

In-situ TEM heating and cooling experiment resulted in structural changes and BF contrast differences allowing for direct observation of the SPT of (a) VO${}_2$ and (b) W${}_{0.008}$V${}_{0.911}$O${}_2$ upon cooling with $T_c$ occurring at 68 °C. The contrast boundary movement, highlighted using red arrows, was used to calculate approximate rates of transformation to compare to in-situ PDF data.

Figure 5

In-situ total scattering distribution refinements elucidated the local phase diagram for the W-substituted VO${}_2$ system. These observations were supported through NMF analysis of both the heating and cooling data sets. Linear regressions of each data set, agree with each other and show a slower local progression of the SPT compared to the MIT.

Figure 6

DSC and SQUID magnetization experiments determined the MIT for the tungsten substitution series. The two data sets agree with each other in the rates of $T_c$ depression despite DSC containing a partial data set compared to SQUID. This $T_c$ depression rate also agrees with the structure phase transformation rate from in-situ PDF during heating and cooling.

Figure 7

V oxidation is correlated with W-substitution as supported by (a) PDF simulations of varying amounts of V${}_6$O${}_{13}$ (C2/m) to either VO${}_2$ ($P{2}_1/c$) or VO${}_2$ ($P{4}_2/mnm$) capture the local features of the PDF upon increasing W substitution amount and (Alliance of Bioversity International and CIAT—Multifunctional Landscapes) V K-edge EXAFS fitting to paths from the following structures: VO${}_2$ ($P{2}_1/c$), V${}_6$O${}_{13}$ (C2/m), V${}_2$O${}_5$ (Pnma), and in the case of the W-substituted samples VO${}_2$ ($P{2}_1/c$), V${}_6$O${}_{13}$ (C2/m), V${}_2$O${}_5$ (Pnma), and VO${}_2$ ($P{4}_2/mnm$).