Introduction

Vanadium dioxide (VO2), a strongly correlated electron and metal-insulator transition (MIT) material, is an extremely interesting material suitable for many technological applications. The most striking features of VO2 are its abrupt first-order MIT near 68° C, exhibiting a large change in the resistivity (up to five orders of magnitude) and the infrared transmittance/reflectivity in the sub-picoscecond time scale1. Simultaneously, the crystal structure transforms from a high-temperature tetragonal rutile (R) structure characterized by a single V−V distance of 2.85 Å and linear chains of edge-shared Jahn−Teller-distorted VO6 octahedra, to a low-temperature monoclinic (M1) structure containing V4+–V4+ pairs forming a zigzag chain with alternating V−V distances of 2.65 and 3.12 Å and more distorted VO6 octahedra (see Figure 1). Moreover, different external stimuli such as thermal, electrical, optical, or magnetic field can trigger the MIT in VO22. These unique characteristics make VO2 suitable for application such as smart windows3, memory devices4, uncooled microbolometers5, electronic/optical switch devices6,7, thermal/chemical sensors8,9, etc.

Figure 1
figure 1

The crystal structures of low-temperature monoclinic (space group P21/c) and high-temperature tetragonal rutile (space group P42/mnm) phases of VO2.

The typical MIT temperature of the pure VO2 is ~68°C, which unfortunately is not ideal for practical applications. Much effort has been devoted to regulate the MIT critical temperature (TC) of VO2. An effective route for regulating the TC is doping with metal ions10,11,12, in addition to adding internal/external stress13,14 or controlling the microstructure and defects15,16,17. In the available literatures, it was shown that the TC could be decreased by doping large dopants ions with higher-valence such as W6+, Mo5+ and Nb5+, or increased by small dopants ions with lower-valence (Al3+, Cr3+ and Ga3+)12,18,19,20,21,22,23. In this framework, the still open challenge is the deeper understanding of the intrinsic mechanism for the regulation of TC by ions-doping, which have an important significance in VO2 based functional devices.

The intuitive understanding is that the regulation of the TC by metal doping inside the VO2 lattice depends on the relative size and the relative valence of the dopant ion compared to that of the V4+ ion22,23. The substitution may give rise to the changes of lattice structure and carrier density (or the conductivity) in parent VO211,19,24,25. In the case of the W6+-doping, it has been considered that the W6+-doping can donor two extra-electrons to the VO2 host if considering the charge neutralization. The increase of the electron density affects the band structure and the activation energy, facilitating the transformation to the metallic phase25,26. Moreover, each W dopant is doped substitutionally and disrupts the dimeric V4+–V4+ bond to form W6+–V3+ and V4+–V3+ pairs. This replacement destabilizes the monoclinic phase and thus lowers the energy barrier for the transition to the rutile structure26,27,28. The different contributions finally result in the reduction of TC. Other researchers focused on the influence of local structure perturbations induced by dopant ions23,28. They suggested that the TC was not affected by the carrier density variation in VO2, but by the lattice distortion induced by the dopants with different ion radius. The change trend of TC can be correlated with the relative size of the dopant ion compared to that of the V4+ ion. Booth et al28. claimed that the effect of W6+ dopants on neighboring cells would be only structural. Based on EXAFS data, they concluded that a local rutile structure was formed around W6+ dopants and a significant expansion in the [110]R and directions induced by W6+ dopants broke the dimeric homopolar V-V pairs due to the decreasing d orbital overlap, showing the lattice deformations towards the high-temperature rutile structure and thus resulting in the reduction of TC. Recently, a combination of XANES and EXAFS spectra has been used to characterize the electronic contribution and the local structure perturbations on the host VO2 upon W6+-doping, which indicates that both contributions are responsible for the reduced TC18.

The detailed mechanisms involved in the increase of Tc by doping trivalent ions: Al3+, Cr3+ and Ga3+ in VO2 are complex to describe since more phases will be involved. Indeed a substitution of Al3+ or Cr3+ dopants for V4+ in VO2 gives rise to additional insulating phases: two specific monoclinic (M2 and M3) and a triclinic (T) phases, apart from the most common monoclinic (M1) phase29,30,31.

From the literatures, considering the atomic radius and the valence states of each dopant, it is clear that the regulation of Tc by doping ions is always associated with the lattice distortion and the carrier density change. The change in the carrier density of VO2 inevitably occurs when the dopant ion is not tetravalent, which means that a donor/acceptor-type doping (i.e., charge doping) of the VO2 band structure may occur. Nevertheless, which factor plays the critical role in the TC behavior is still unclear.

In order to reveal the intrinsic mechanism for the regulation of TC by ions-doping, it is mandatory to identify the roles of the lattice distortion and the charge doping caused by dopants and which one plays the main role in this mechanism. To this purpose, considering that tetravalent ions-doping may rule out the carrier contribution of tetravalent dopant ions to neighboring vanadium ions, i.e., to minimize the charge doping effect, we choose Ti4+ ion as the dopant. In Ti4+-doped VO2 system, it is imperative to clarify the behavior of Ti4+ dopants and their influence on the host VO2 lattice. To address the above issues, the most suitable tool is the X-ray absorption fine structure (XAFS) spectroscopy, because of its specific element selectivity and the sensitivity to the local structure (2–5 Å) around the absorber atoms as well as the electronic structure32.

In this work, TixV1-xO2 nanopowders were prepared by a hydrothermal method with a subsequent Ar annealing treatment. We systematically explored the electronic and local geometric structure of both the Ti dopants and the host V atoms in TixV1-xO2 samples using XAFS spectroscopy at both Ti and V K-edges. A combined experimental and theoretical analysis of the mechanism of the regulation of Tc by Ti-doping was performed for the first time. In addition, a comparative analysis of different ion-doping systems was also performed to identify the critical factors in regulating the TC in ions-doped VO2 systems.

Results

The influence of Ti4+-doping on phase transition properties of VO2

Figure 2 shows the DSC curves of TixV1-xO2 samples with different Ti concentrations. At low Ti concentrations, as shown in Figure 2 (left), during the heating process the TC slowly decreases with increasing Ti concentration. It reaches a minimum at the Ti concentration of 2.8%. During the cooling process the starting phase transition temperature is almost unchanged and after the starting phase transition a broad exothermic peaks subsequently appear, indicating the occurrence of a non-uniform phase transition. At the higher Ti concentrations, e.g., 5.0%, 6.1% and 8.1%, as shown in Figure 2 (right), during the heating process the TC increases with increasing Ti concentration. The same trend of the TC is also observed during the cooling process. In addition, the double endothermic/exothermic peaks appear during the heating/cooling process, probably due to the non-uniform doing or the polydispersity in the size distribution19.

Figure 2
figure 2

Comparison of DSC curves of TixV1-xO2 samples and an undoped VO2 (M1) sample during heating and cooling cycles.

Generally, the endothermic peaks appearing during the heating process are used to determinate the TC of the ions-doped VO2. The TC of the TixV1-xO2 samples slowly decreases reaching a minimum and then gradually increases with increasing Ti concentration, in agreement with data of Beteille et al.22. Previous researches did not show a unique behavior of the Ti4+-doping on regulating the TC. Most of researches reported the increase of TC, however, only showed the increase of TC in high Ti concentration (>3%) due to weak ability of Ti dopants to regulate the TC33,34,35,36. In addition, it was found that the TC saturated at around 80°C when the Ti doping concentration reaches to a higher level35. From the current results, it seems that the TC decreases in initial low Ti concentration, while it increases gradually with further increasing Ti concentration in a certain concentration range.

Although Ti doping did not regulate the transition temperature significantly, Ti doping can effectively modify the thermochromic properties of VO2, such as the decrease of hysteresis sloop width of phase transition, the improvement of the temperature coefficients of resistance and the enhancement of visible transmittance (Tvis, 380–780 nm) and solar transmittance (Tsol, 240–2600 nm)33,34,35,36,37. Compared with TixV1-xO2 film, there are no lattice mismatch and thermal stress for TixV1-xO2 nanopowders. Thus TixV1-xO2 nanopowders can be used to clarify the influence of Ti4+ dopants on the VO2 lattice structure and then on the TC of VO2.

The influence of Ti-doping on the crystalline structure of VO2

Ion-doping inevitably modifies the host VO2 lattice structure, e.g., the phase transformation from a monoclinic to a rutile structure in the NxV1-xO2 systems (N = W6+, Mo5+ or Nb5+)19,24 and the formation of stabilized M2, M3, or T phases in the MxV1-xO2 systems (M = Al3+ or Cr3+)29,31. Figure 3 shows XRD patterns of TixV1-xO2 samples with different Ti concentration. No characteristic peaks of titanium oxides are observed, indicating the formation of titanium ions solid solutions with the VO2. At low Ti concentrations, the XRD patterns of the TixV1-xO2 samples (x = 0, 0.6%) match well with that of the monoclinic (M1, space group P21/c) phase (JCPDS card No. 72-0514). With increasing Ti concentration, the diffraction peaks are consistent with that of the M1 phase except for two obvious peaks in the range 63.5° < 2θ < 66.0° (showed in the gray areas). In this range, the broad diffraction peak gradually splits into two obvious peaks with increasing Ti concentration. This scenario was also observed in the WxV1-xO2 system18,26. The appearance of the two peaks was associated to a rutile phase at high W concentration. Although the Ti-doping does not change the kind of crystal lattice of TixV1-xO2 samples, a local structure phase transition from the monoclinic to the rutile structure may occur in some regions of the samples with increasing Ti concentration. In addition, several diffraction peaks gradually shift to lower diffraction angles, e.g., the peaks in Figure 3b, pointing out upon Ti4+-doping a continuous increase of the interplanar spacing due to the larger radius of the Ti4+ ion (0.0605 nm) compared to the V4+ ion (0.058 nm). The opposite trend was observed when VO2 was doped with Al3+ ions (0.054 nm)21.

Figure 3
figure 3

(a) XRD patterns measured at room temperature for the TixV1-xO2 samples with different Ti concentration. The peaks indicated with asterisk (*) clearly shift towards lower diffraction angles with increasing Ti concentration, such as a magnified view in panel (b).

The morphology and microstructure of TixV1-xO2 samples

Figure 4a–e show SEM images of TixV1-xO2 and undoped VO2 samples after the Ar annealing treatment. TixV1-xO2 samples are actually composed of nanoparticles (100–300 nm), compared with microparticles (1–10 μm) present in undoped VO2 sample. Moreover, the Ti doping had a strong effect on the morphological evolution of TixV1-xO2 samples. The TixV1-xO2 nanoparticles gradually reduce their mean particle size with increasing Ti concentration, due to the reduced crystallization ability (i.e., enhanced heterogeneous nucleation process) of TixV1-xO2 nanoparticles by Ti doping35. The same situation occurred in other ions-doped systems21.

Figure 4
figure 4

(a–e) SEM images of the TixV1-xO2 samples with different Ti concentration and an undoped VO2 sample; (f) EDS spectrum of the TixV1-xO2 sample (x = 0.6%); (g) TEM image of the TixV1-xO2 nanoparticles (x = 5.0%) and the corresponding selected area electron diffraction (SAED) pattern (h) and EDS spectrum (i).

In Figure 4f, the Energy dispersive X-ray (EDX) fluorescence spectrum obtained from SEM analysis shows the Ti, V, O and Si characteristic peaks. Actually, the Si peak appears also due to the scattering induced by Si substrate. In Figure 4i, the EDS spectrum obtained from TEM analysis shows no Si peak since a Cu grid was used for supporting the sample and clearly shows Ti, V, O and Cu peaks. The Cu peaks are clearly due to the scattering induced by the Cu grid. Therefore, both EDS spectra obtained from TEM and SEM analysis confirm that TixV1-xO2 nanoparticles contain only Ti, V and O elements.

Figure 4g shows the TEM image of TixV1-xO2 nanoparticles (x = 5.0%). The corresponding SAED pattern (Figure 4h) is in agreement with the diffraction pattern along the [211] crystal axis of the M1 phase (JCPDS No. 72-0514), showing the monoclinic single-crystalline nature of the TixV1-xO2 nanoparticles. In addition, the measured interplanar spacings are 0.10–0.12 Å larger than theoretical values, due to the lattice expansion when Ti ions were incorporated into the VO2 lattice.

XAFS analysis

To clarify the behavior of Ti dopants and their influence on the host VO2, the local structures of both Ti and V atoms in the TixV1-xO2 samples, as well as their chemical states, were systematically investigated by XAFS spectroscopy at Ti and V K-edges.

Figure 5a shows V K-edge XANES spectra of TixV1-xO2 and undoped VO2 (M1) samples. For vanadium oxides, the energy positions of the threshold, the pre-edge peak and the absorption-edge exhibit a monotonic dependence on the oxidation states of the absorber atoms according to Kunzl's law26,38,39. For TixV1-xO2 samples (x = 0.6%, 1.7%, 8.1%), the energy positions of the threshold, the pre-edge peak and the absorption-edges are almost the same and coincide with that of the undoped VO2 (M1), pointing out the tetravalent valence of V ions in the TixV1-xO2 samples. In addition, the pre-edge peak can be used to evaluate the changes in the local symmetry of V atoms, due to its sensitivity to the local coordination environment of the absorber atoms and the electron density of d states38,39,40. Pre-edge peak intensity increases with a lower local symmetry while decreases for a higher local symmetry of the absorber atoms. As an example, the pre-edge peak intensity is negligible in VO characterized by a regular octahedral symmetry (Oh) around the absorber V atoms, while increases if the local symmetry is lower than the Oh symmetry such as in VO2, V2O3 and V2O5, reaching the maximum for vanadates with a tetrahedral coordination (Td)38. At low Ti concentrations (0.6% and 1.7%), TixV1-xO2 samples show an increased pre-edge peak intensity. However, a decreased intensity occurs at high Ti concentration, e.g., at 8.0% Ti concentration and the pre-edge peak intensity decreases back to close to that of VO2 (M1). The change of the pre-edge peak intensity indicates that with increasing Ti concentration, the distortion of the VO6 octahedra around V atoms firstly increases to a maximum from the initial VO6 octahedra of VO2 (M1) and then decreases back to the VO6 octahedra of VO2 (M1).

Figure 5
figure 5

(a) V K-edge XANES spectra of the TixV1-xO2 samples (x = 0.6%, 1.7% and 8.1%) and the undoped VO2 (M1) sample. The insert shows the enlarged view of the pre-edge absorption peak. (b) V K-edge EXAFS oscillations [k3χ(k)] and (c) their Fourier transforms (FTs), along with the theoretical curves of M1 phase for reference.

V K-edge EXAFS data confirm the evolution of the local structure around V atoms. Figure 5b and c show V K-edge EXAFS oscillations [k3χ(k)] and their Fourier transforms (FTs), respectively. In Figure 5b, the 8.1% Ti sample shows similar EXAFS oscillations with the undoped VO2 (M1) sample, while 0.6% and 1.7% Ti samples show significantly different EXAFS oscillations respect to the VO2 (M1). Likewise, in Figure 5c, the FTs curves clearly demonstrate a similar local structure around V atoms for the 8.1% Ti sample and the VO2 (M1), in agreement with the theoretical spectrum of the VO2 (M1), i.e., their FTs curves exhibit the four characteristic peaks of M1 phase: the two peaks at ~1.33 and 1.75 Å´, corresponding to the first V–O coordination shell and other two at ~2.15 and 2.95 Å´ associated to the V–V1M and V–VM shells. On the contrary, the low Ti concentration samples (0.6% and 1.7%) show FTs curves significantly different with that of VO2 (M1), compatible with local structures different from the standard M1 phase structure. Therefore, the local structure around V atoms deviates from the standard M1 phase structure first in the initial Ti doping process. With the further increase of the Ti concentration, the local structure around V atoms will return back to the M1 phase structure. This is accord with the change trend of VO6 octahedra in above V K-edge XANES analysis.

To understand the intrinsic local structure change around the host V atoms within the doping process, we then focus on the local structure around Ti dopants. Figure 6a and b show Ti K-edge XANES spectra and their EXAFS oscillations [k2χ(k)] of TixV1-xO2 samples, respectively, which both depict the evolution of the local structure around Ti dopants. The 0.6% and 1.7% Ti samples show similar XANES spectra with TiO2 (A) except for the pre-edge structure (Figure 6a). When the Ti concentration increases to 8.1%, the XANES spectrum matches well with that of TiO2 (R). This remarkable and systematic evolution also appears in the Ti K-edge EXAFS oscillations. As shown in Figure 6b, the 0.6% and the 1.7% Ti samples exhibit the characteristic EXAFS oscillations of TiO2 (A) while the 8.1% Ti sample exhibits the characteristic EXAFS oscillations of TiO2 (R). Therefore, Ti K-edge XANES and EXAFS data both indicate that the local structure around Ti dopants changes from the TiO2 (A)-like to the TiO2 (R)-like structure with increasing Ti concentration (see Figure 6c). Namely, at the initial low Ti concentration, the local anatase structure around Ti dopants is formed in the host monoclinic VO2 structure and subsequently a local rutile structure around Ti dopants is gradually formed with increasing Ti concentration. The local structure dynamics of Ti dopants ought to be responsible for the local structure change of the host V atoms. Moreover, the local rutile structure around Ti dopants perfectly account for the appearance of two peaks in the range 63.5° < 2θ < 66.0° in the XRD patterns at high Ti concentrations. In addition, the energy positions of the absorption-edges of the TixV1-xO2 samples are constant and coincide with those of TiO2 (A) and TiO2 (R), pointing out the tetravalent valence of Ti ions in the TixV1-xO2 samples.

Figure 6
figure 6

(a) Ti K-edge XANES spectra of the TixV1-xO2 samples (x = 0.6%, 1.7% and 8.1%) and (b) their EXAFS oscillations [k2χ(k)], along with the reference samples of TiO2 (A) and TiO2 (R). (c) The local structure evolution of Ti dopants in the host monoclinic structure with increasing Ti concentration. The green balls in the panel denote the titanium atoms.

Discussion

Based on our XAFS results, it seems that the doping VO2 with Ti4+ ions has a position between clearly donor- and acceptor-like defects, due to the same valence between the Ti4+ dopant and the V4+ ion. To confirm this scenario, we performed electron density of states (DOS) calculations using the projector augmented wave (PAW) method implemented in the Vienna Ab-initio Simulation Package (VASP)41. The PBE form of the generalized gradient approximation (GGA) and the DFT+U scheme42 (U = 4.0 and 6.6 eV for V and Ti atoms, respectively) were used to describe the electron exchange-correlation interaction. A 2 × 2 × 1 supercell with 48 atoms containing 15 V atoms, 32 O atoms and 1 Ti atom, corresponding to ~6.25% Ti concentration was used. For comparison, we also performed the similar DOS calculations of W-doped VO2, except U = 0 eV for W atom. Figure 7 compares the total and partial DOS of undoped VO2, 6.25% Ti-doped VO2 and 6.25% W-doped VO2. It can be observed that below the Fermi level, the total DOS and the partial V-3d/O-2p DOS do not show clear differences between the undoped VO2 and the 6.25% Ti-doped VO2 (Figure 7a and 7b), indicating the negligible influence of Ti4+ dopants on the valence band structure of VO2. But for W-doped VO2, the DOS (Figure 7c) shows the Fermi level in the bottom of the conduction band, indicating the electron doping of VO2 due to a charge transfer between W6+ and V4+ ions, in accord with the detection of reduction of V4+ to V3+ ions in W-doped VO218,26. Therefore, from the above DOS calculations, we confirm that the incorporation of Ti4+ ions in VO2 basically do not induce a donor or acceptor doping of the VO2 band structure due to the lack of a charge transfer between Ti4+ and V4+ ions. This DOS calculations result can match with the constant valences of V and Ti ions from above XANES spectra and the XPS spectra reported by Chen et al.37, which also showed the unchanged valence states of V4+ and Ti4+ in Ti-doped VO2. All of these indicate that the carrier concentration experience no significant change in TixV1-xO2 system. Accordingly, the charge doping effects caused by Ti4+ doping almost can be ruled out and only the local structure perturbations can be considered to have the dominated effect on the regulation of Tc in TixV1-xO2 nanopowders samples.

Figure 7
figure 7

The DOS of (a) undoped VO2, (b) 6.25% Ti-doped VO2 and (c) 6.25% W-doped VO2 calculated by using the DFT method.

Therefore, the two trend of the local structure change around V atoms induced by Ti-doping (from above V K-edge XAFS spectra), roughly corresponds to the two observed changes of the TC. On the basis of experimental and theoretical results, we consider that the mechanism of the regulation of Tc by Ti4+ doping is mainly associated with the local structure perturbations induced by Ti4+ dopants.

Our DSC results showed that the TC slightly decreases in low Ti concentration level (within about 3% Ti concentration), while a small amount of W doping (within 3.4% W concentration) will result in a distinct Tc decreasing (Figure S1). Previous literatures also showed that the ability to regulate the TC in the TixV1-xO2 system was smaller than the MxV1-xO2 (M = W6+, Mo6+or Nb5+) systems12,19,26,33,36,43. Table 1 shows the radius of several dopant ions and the V4+ ion. Ti4+ and W6+ ions have the radius close to each other, but W6+ ions have the considerably larger ability to regulate the TC of VO2, i.e., a reduction in TC by 20 ~ 30 K/at.% W for the bulk and by ~50 K/at.% W in nanostructures11,24,43,44.

Table 1 Comparison of the radius of different doped ions with V4+ ion

In the case of W-doping, the VO6 octahedra shows a distortion trend with increasing W concentration, until the concentration of 1.7% (Figure S2). However, it cannot be suggested that the gradually decreasing of TC (Figure S1) is mainly attributed to the distortion of VO6 octahedra since the electron doping of W6+ ions in VO2 is also conducive to the reduction of the TC11,18,26. In TixV1-xO2 system, the charge doping effects caused by Ti4+ doping can be ignored, i.e., decoupling the lattice distortion and charge doping effects on the phase transition behavior of VO2.

Due to the large ion size, when W or Ti atoms occupy the V sites, the substitution doping will yield the detwisting of the nearby monoclinic VO2 lattice in the similar way especially within low doping concentration. This type of lattice detwisting includes the decreasing V–V pairs tilting, depairing of dimerizated V–V pairs and distorting the VO6 octahedra in surrounding monoclinic structure18,24,28. The distortion of VO6 octahedra induced by ions-doping can change the hybridization between V 3d and O 2p orbitals, resulting in the shift of π and π* bands near the Fermi level in the band structure of VO2, that finally changes the energy gap24.

The above comparative analysis suggested us that the charge doping (i.e., donor/acceptor doping) in VO2, plays a more fundamental role in the regulation of the TC, although the local structure perturbations induced by dopants has an inevitable influence on the TC. Actually, the smaller ability to regulate the TC in the TixV1-xO2 system is most likely due to the negligible charge doping effect upon Ti4+-doping.

In conclusion, TixV1-xO2 nanopowders exhibit two trends for the TC: the TC slightly decreases to a minimum and then increases with increasing Ti concentration. The behavior of Ti4+ dopants and their influence on the host VO2 lattice has been explored for the first time by XAFS spectroscopy and theory calculations. With increasing Ti concentration, the local structure around Ti dopants displays an evolution from a local anatase to a rutile-like structure, which induces a perturbation of the nearby monoclinic VO2 lattice. As a result, the distortion of the VO6 octahedra in the monoclinic VO2 lattice becomes more and more distinct at the initial doping stage. With the further Ti doping, the distorted VO6 octahedra will return to the initial VO6 octahedra together with the appearance of the local rutile structure around Ti dopants. The structure evolution induced by Ti doping is actually considered responsible for the observed trends of TC in the DSC tests.

Finally, we should underline that, although the current Ti-doping research shown a direct influence of the local structure perturbations induced by Ti dopants on the regulation of TC, this modulation effect of the Ti doping strategy on VO2 materials is not such pronounced, in particular if we consider the variation of the phase transition temperature. By comparison of the ability to regulate the TC in different ions-doping systems, such as W6+ dopants, we may claim that the charge doping for VO2 may play a critical role in the effective regulation of the TC in no tetravalent ions-doped VO2 systems.

Methods

Synthesis of TixV1-xO2 nanopowders

The Ti-doped VO2 (TixV1-xO2) samples were synthesized by a hydrothermal method followed by an annealing treatment. The different amount of Ti (SO4)2 aqueous solution (0.01 M) were added to VO(acac)2 aqueous solutions under vigorous stirring. The value of x in TixV1-xO2 refers to the Ti atomic percent in the feed. Each of the final solutions was transferred to the Teflon cup, which was later heated in a sealed autoclave at 200°C for 24 hours. After the hydrothermal reaction, the precipitate was collected by centrifugation, washed with copious amounts of deionized water, N, N-dimethylformamide (DMF) and ethanol and then dried in vacuum at 60° C. Finally, TixV1-xO2 samples were calcined under an argon (Ar) atmosphere at 700°C for 6 hours.

Characterization

The crystalline structure of the TixV1-xO2 samples was determined by X-ray diffraction (XRD) using a theta/theta rotating anode X-ray Diffractometer (mode: TTR-III, Cu Kα radiation). The morphology and the microstructures were characterized with a Field-emission scanning electron microscope (FE-SEM, JEOL JSM-6700F) and transmission electron microscopy (TEM, JEM-2010(HR)). The phase transition behavior was studied by differential scanning calorimetry (DSC, Q2000).

XAFS spectra measurement and analysis

XAFS spectra were measured at ambient temperature (~24°C) at the beamline 1W2B of the Beijing Synchrotron Radiation Facility (BSRF), using a Si(111) double crystal monochromator with an energy resolution (ΔE/E) of <1−3 × 10−4 @9 keV. Ti and V metal foils were respectively used for calibrating energy at the Ti K- and V K-edges. Ti K-edge XAFS spectra were collected in the fluorescence mode using a Lytle detector, while V K-edge XAFS spectra were collected in the transmission mode using ionization chambers filled with Ar/N2. V K- and Ti K-edges XAFS spectra were collected in the energy range of 5268–6251 and 4768–5419 eV, respectively. In the data processing procedure, the experimental absorption data were processed using the ATHENA module (version 0.8.054) implemented in the IFEFFIT package46,47.