Tuning the Metal–Insulator Transition Properties of VO2 Thin Films with the Synergetic Combination of Oxygen Vacancies, Strain Engineering, and Tungsten Doping

Vanadium oxide (VO2) is considered a Peierls–Mott insulator with a metal–insulator transition (MIT) at Tc = 68° C. The tuning of MIT parameters is a crucial point to use VO2 within thermoelectric, electrochromic, or thermochromic applications. In this study, the effect of oxygen deficiencies, strain engineering, and metal tungsten doping are combined to tune the MIT with a low phase transition of 20 °C in the air without capsulation. Narrow hysteresis phase transition devices based on multilayer VO2, WO3, Mo0.2W0.8O3, and/or MoO3 oxide thin films deposited through a high vacuum sputtering are investigated. The deposited films are structurally, chemically, electrically, and optically characterized. Different conductivity behaviour was observed, with the highest value towards VO1.75/WO2.94 and the lowest VO1.75 on FTO glass. VO1.75/WO2.94 showed a narrow hysteresis curve with a single-phase transition. Thanks to the role of oxygen vacancies, the MIT temperature decreased to 35 °C, while the lowest value (Tc = 20 °C) was reached with Mo0.2W0.8O3/VO2/MoO3 structure. In this former sample, Mo0.2W0.8O3 was used for the first time as an anti-reflective and anti-oxidative layer. The results showed that the MoO3 bottom layer is more suitable than WO3 to enhance the electrical properties of VO2 thin films. This work is applied to fast phase transition devices.


Introduction
VO 2 is one of the simplest correlated oxide materials, with a metal-insulator transition (MIT) phase transition at 68 • C [1]. The semiconductor phase was observed at low temperatures with a monoclinic crystal structure. In contrast, at high temperatures, it is a tetragonal rutile structure [2]. VO 2 is the most intensively investigated solid-state thermochromic, metamaterials, and switching terahertz device [3][4][5]. It is still not ready for industrial or commercial application due to several problems. For instance, the transition temperature at 68 • C is too high for applications in daily life (such as intelligent window applications). the prepared monoclinic VO 2 layer, several characterizations such as X-ray photoelectron spectroscopy (XPS) are reported.

Preparation of Thin Films
The samples were produced in a VAKSİS-MİDAS-3M1T magnetron sputter system (VAKSIS, Ankara, Turkey). Figure 1 shows the schematic diagram of the multilayer structure of the device. There are two 13.56 MHz power supplies and one DC power supply in the system with three magnetrons. Multilayer thin films were fabricated from 3 inch and 99.9% purity vanadium, molybdenum, and tungsten metallic targets by reactive sputtering. Argon and oxygen gases connected to the magnetron sputter system had a purity of 99.98%. Before starting the sputtering process, the FTO substrates were cleaned with acetone and isopropyl alcohol by keeping them in an ultrasonic cleaner for 15 min. They were then rinsed with deionized water and dried with high-purity nitrogen gas. After the substrates placed in the magnetron sputter holder were placed into the system, the system started to vacuum and was reduced to a 9.3 × 10 −7 mbar pressure level. After reaching the desired pressure level, the sample holder was heated to 400 • C. It was kept at this temperature for 30 min for the temperature to stabilize on the substrate. Abbreviations of sample names and all deposition parameters are summarized in Table 1. The system pressure was increased to 6.8 × 10 −3 mbar for the reactive sputter process for all samples. Time-dependent thickness optimization was achieved by using samples that had been produced in different thicknesses and whose thickness measurements had been made before. This optimization was used in sample fabrications as shown in Table 1. By controlling the film deposition times, the thicknesses of V-O, W-O, Mo-O, and Mo-W-O films were kept constant for all samples. The Mo 0.2 W 0.8 O 3 thin film is a composite film of Mo and W, in the presence of a 12.1 sccm oxygen environment. Whereas, the Ar flow is 50 sccm. The Mo and W powers are 27 and 110 W, respectively. The deposition time was 16.7 min during the whole process for both Mo and W targets. The idea here is to introduce a small amount of Mo atoms into the WO 3 film to form a high-quality anti-refractive anti-oxidative layer. In addition, this layer provides high transmission and good electrical performance, as predicted from our previous work [15,17]. The optimization of the content of Mo into WO 3 has been covered before [17,18].

Characterization Techniques
The optical transmission (T) result was measured using UV-VIS-NIR Spectrophotometer (UV-3600i Plus, Shimadzu, Tokyo, Japan). The topography of the films was studied using Atomic Force Microscopy Park XE7 system (Park, Santa Clara, CA, USA) with a scanning area of 1 × 1 µm 2 , and a tip scan speed of 0.5 Hz, through noncontact mode. The crystal structures were analysed using X-ray diffraction GNR ADP PRO 2000 (GNR Srl, Novara, Italy) with a radiation source of CuK α (λ = 1.5405 Å) step of 0.01. These data were refined via the Rietveld method applied using Fullprof suite software (Version 5 January 2021) to determine the exact metal-oxygen concentrations of each layer and different phases. The surface morphology was recorded using field-emission scanning electron microscopy (FESEM, GeminiSEM 500, ZEISS, Cambridge, UK). The energy dispersive X-ray (EDX) analysis of the samples was carried out to characterize the quantitative analyses of the samples using an Oxford, Xmax 50 SEM (Oxford Instruments, Oxford, UK) attached with an Oxford EDS detector. Raman vibrational modes were obtained using Renishaw inVia confocal Raman microscope (Renishaw, New Mills, UK) using 532 nm laser beam with 10 mW laser power. X-ray photoelectron spectroscopy (XPS) measurements were carried out on the Thermo Scientific K-alpha XPS system (Thermo Scientific™, Waltham, MA, USA) using an AlK α source and a spot size of 400 µm. Fourier Transform Infrared Spectrometer (FTIR) through a 200-4000 nm range were collected with Thermo Scientific-Nicolet iS20 (Waltham, MA, USA). The electrical I-V phase transition measurements were taken using a 2450 Keithley Source Meter (Tektronix, Beaverton, OR, USA) and a four-probe system in which a heating stage is connected. The thicknesses of the samples were measured using the thin film analyzer system F20-UV (FILMETRICS, San Diego, MA, USA) with a ±1 nm error. The values are as follows: S1 (VO 2 ) is 50

Characterization of FTO Supported Oxide Thin Film
The samples studied in this work are based on the successive deposition of metal oxide thin film. Therefore, the preparation conditions regarding each desired layer on FTO glass have been optimized. Typically used, the Raman spectra show the successful preparation of each thin film ( Figure 2). Within the VO 2 layer (also used as sample S1), the peaks related to lattice motion involving V-V bonds (132 and 229 cm −1 ) as well as the vibrational mode of V-O bonds in the VO 2 monoclinic insulating phase (500 & 828 cm −1 ) are the most critical [19][20][21][22][23]. For the MoO 3 thin film, the most robust peaks at 820, 860, and 994 cm −1 are related to the α-MoO 3 crystal phase and the M-O stretching modes [24][25][26]. The weaker peaks at 190, 283, 336, and 352 cm −1 are based on Mo-O's bending mode [27,28], while the one at 658 cm −1 is assigned to triply coordinated oxygen Mo 3 -O stretching [29]. Concerning the WO 3 layer, the monoclinic phase is characterized by peaks at 270 and 326 cm −1 [30]. The structure is confirmed via the WO 3 lattice mode peak at 132 cm −1 [31] and the stretching mode peaks of W-O-W bonding at 711 & 810 cm −1 [30]. Finally, the Raman peaks for Mo 0.2 W 0.8 O 3 thin films are very weak. However, the prominent peaks for the WO 3 are distinguishable, suggesting a severe disruption of the WO 3 structure with the introduction of Mo. Nevertheless, Raman spectroscopy is not accurate enough for the precision of the structural properties of thin films, which strongly influence their thermochromic performance. Mo, W, and V metals are metal atoms with large coordination numbers, and the number of phases with oxygen is also large. Some of these phases are metastable, while others are stable. Synthesis conditions such as temperature, oxygen ratio, and deposition pressure are the basic parameters that determine the phase to be formed. The X-ray diffractogram has been used to determine the oxidation state and the structural parameters of the thin films (V-O, W-O, Mo-O, and Mo-W-O) as in Figure 3a, separately. These XRD patterns were obtained by GIXRD examination of thick films (>300 nm) grown on Soda-Lime Glass (SLG) under the same conditions for each layer. The examination was carried out with a grazing angle of 1 • and counting for 5 s at each angle value in 0.01 • steps. XRD patterns presented separately for each layer were used to determine the phase or phases of the thin film structure. Rietveld structure refinement was performed with the technique described in [17] for Mo-O, and Mo-W-O, W-O phases and the technique described in [32] for the VO 2 phase, using the FullProf Suite software, which yielded the structure parameters and stoichiometric ratios of metal and oxygen. The use of Rietveld refinement to characterize the structural properties and concentration of vacancies is a frequently used method, reported in numerous reported studies [17,[32][33][34][35][36][37]. However, performing structural analysis with XRD patterns that occur when XRD is taken for multilayer structures brings some errors. In the resulting multilayer structure, XRD peaks will occur in a very complex order and it will be difficult to determine which phase they belong to.
Vanadium oxide film shows a monoclinic phase with a space group of C 12 /m1 The diffraction peaks observed at 2θ = 27.9, 37.2, 42.3, 55.6, and 57.7 • refer to (011), (200), (212), (220), and (022), respectively. The main peak is observed for 2θ = 27.9 • , indicating a preferential orientation toward the (011) plan. The diffractogram corresponds roughly to the VO 2 structure. However, the oxygen defects detected via the refinement show the presence of the VO 1.75 compound. The oxygen vacancies act as electron donors with high n-type conductivity. They can alter the electron orbital occupancy and the band structure of the V-O layer [32,38]. VO 1.75 layer is a rutile metallic state [39][40][41] which is unstable due to its sensitivity to atmospheric oxygen. Tungsten oxide film shows a monoclinic phase (γ-WO 2.94 ) with a space group of P 12 /m1. This phase is also maintained for the Mo-W-O layer (Mo 0.2 W 0.8 O 2.8 ). Concerning the molybdenum oxide (MoO 3 ), the diffractogram corresponds to an orthorhombic structure with the Pnma space group. However, this layer is biphasic, consisting of 40% MoO 2.08 and 60% MoO 3.01 . Thus, an oxygen deficiency for each layer has been observed, confirming their metallic characteristic. On the other hand, the XRD patterns of each layer present the central peak at around 23 • , which corresponds to the plan (002) of the monoclinic phase and (011) of the orthorhombic structure. The atomic alignment with an angle of 90 • for both plans allows overlaying. This characteristic is also true for the plan (011) of the V-O layer. That is why the heterojunction is probably obtained following these orientations within the multilayer structure.
However, the widths of the XRD peaks give us a rough value of the crystallite size with the help of the Debye-Scherrer equation. The crystallite size was calculated from the Debye-Scherrer equation (D = 0.9λ/βcosθ). After that, the main peak positions (2θ) and their corresponding FWHM (β), crystalline size (D), dislocation density (δ, where δ = 1/D 2 ), and microstrain (ε, where ε = β/4tanθ) were estimated ( Table 2). Although this value is not exact, it is valuable in terms of speculating. Decreased peak widths indicate an increase in crystallite size. It is important to note that the XRD of WO 3 shows crystalline and amorphous phases together. As the growth temperature of the W-O film is 400 • C, this temperature is close to the W-O crystallization temperature. The W-O structure crystallizes in the range of 400-600 • C [42].
In order to calculate the FWHM for the crystallite size in the structures containing W-O, a deconvolution was performed using the PeakFit software (Version 4.12), assuming a Lorentzian-shaped peak located at about 23 degrees. The crystallite size calculated for Mo-O was found to be larger than the W-O film. It was also found to be relatively small for the Mo-W-O layer. Penetration of Mo into the W-O layer resulted in a reduction of the main peak width. With the incorporation of Mo atoms into the W-O structure, an expected increase in the crystallite size, hence in the crystallinity, occurred, and it can be said that W-O crystal structures is maintained even in the presence of Mo atoms. Thus, the crystallinity of the oxide layers is improved with the presence of Mo in contrarily to W [43]. The calculation shows that the crystallite size for the W-O film (5.

Effect of the Buffer Layer on the VO 2 Properties
The MIT properties of the multi-layer oxide structure are governed via the behaviour of the VO 2 layer. As seen above, the buffer layer (WO 3 and MoO 3 ) effect on vanadiumbased oxide has been studied (S1-S3). Raman spectroscopy analysis of these structures S2-S6 is depicted in Figure S1 (supplementary information). Similar peaks are observed for these structures, as have been reported above for the single layers. The topographical analysis of MoO 3 and WO 3 deposited only on the glass substrate is investigated as in Figure S2.
First of all, the morphology of the VO 2 thin film is strongly influenced by the buffer layer ( Figure 4). The VO 2 layer, as deposited on the FTO glass (S1), consists of small particles (about 20 nm as predicted with XRD) homogeneously and densely distributed. Table 3 provided information about the topographic data of S1-S6 samples. The mean grain area is below 9.5 × 10 −3 µm 2 , involving a very low average roughness (Ra). However, when deposited on the MoO 3 layer (S2), VO 2 grains are bigger, and aggregates are visible on the surface. The average roughness is more than 12 times bigger than for S1, according to the relief observed on the FESEM picture. Concerning the WO 3 bottom layer (S3), small particles completely disappear, and clusters with nano leaves are present on the surface of the VO 2 thin film. Even if the mean grain area reaches more than 1.77 × 10 −3 µm 2 , Ra values are only about 6 nm. This effect has been ascribed to the monoclinic crystal structure of WO 3 film, which favours electron diffusion [44]. Finally, these three samples confirm that the VO 2 thin film morphology is sensitive to any internal or external strain or stimuli [45,46]. Used as support for metal oxide layers, it is known that FTO glass is transparent at more than 70% on the UV-visible-NIR domains. However, the optical properties are different according to the metal oxide layers ( Figure S3). The transparency is reducing below 40% for the VO 2 layer. However, for the Mo-O, Mo-W-O, and W-O layers, the transparency can reach values higher than 80% up to almost 100% Mo-W-O at around 500 nm.
On the other hand, the surface bonding is also affected by the effect of the buffer layer. The vanadium binding on the XPS spectrum of the film supported with WO 3 (S3) is characteristic of highly oxidized vanadium (V 4+ /V 5+ ) at around 517 and 524 eV [47][48][49]. While peaks are related to low oxidization (V 3+ ) at around 41 and 68 eV when supported with MoO 3 (S2) (Figure 5a,b) [50,51]. In other words, the interaction between molybdenum and vanadium oxides allows for keeping an oxidation degree below 4. In contrast, the vanadium-based layer is more sensitive to oxidation when supported with WO 3 . Therefore, MoO 3 has been preferred as a buffer layer for the samples aiming to study the anti-reflective top layer effect on the thermochromic properties of vanadium oxide. In addition, the observed signals of Na, Ca, Sn, and Si stem from the thin layer of the FTO glass substrate. It is important to mention that all films were deposited with high vacuum sputtering techniques, with ultrahigh homogenous depositions. These conditions avoid the absence of covering some parts during the deposition of S2. The diffusion of some elements from the substrate into the S2 is highly expected during the film deposition. This is because the films were deposited at an approximately high temperature of 400 • C, followed by an in situ annealing at 400 • C for around 2 h. These two processes may cause direct diffusion of some elements from the substrate into the film.

Anti-Reflective Top Layer Surface State
Due to their optical properties of WO 3 (S4) and Mo 0.2 W 0.8 O 3 (S5) were coated in situ on the surface of MoO 3 supported VO 2 layer. Both surfaces contain particles that locally aggregate, inducing pores formation between them ( Figure 6). The topographic data show that the surface roughness of S4 is comparable to S2. The WO 3 top layer can fit the structure of the VO 2 thin film. However, when 20% W is substituted with Mo within the top layer (S5), its particle's grain area increases from 1.183 × 10 −3 to 1.526 × 10 −3 µm 2 , involving lower surface roughness. The former presents smaller particles and higher roughness when deposited on W doping of the VO 2 (S6). The multilayer morphological and topographical analysis demonstrates that the surface is dependent mainly on the bottom layer state and, to a lesser extent, on the upper layer properties.

FTIR Spectroscopic, Optical Properties, and Urbach Tail
Here, we investigated the optical properties of the deposited thin film with UVvisible-NIR spectroscopy and FTIR spectroscopy, as depicted in Figure S4 (supplementary information). The FTIR spectrum of S1 presents three prominent absorption peaks at around 400, 766, and 901 cm −1 ( Figure S4a). Even if they are attributed to the V-O stretching mode for the lowest wavenumber [52] and lattice vibration mode for the two others [53,54], these peaks are not Gaussian. However, the UV-visible-NIR spectroscopy shows the optical transmission as in Figure S4b. Introducing W into the VO 2 structure improves the transparency close to the support values. This evolution can probably be explained by two reasons, the electron density evolution of the local crystal strain introduction within the VO 2 layer.  The calculation of the Urbach energy (E U ) evolution provides a clue concerning the transmittance fall observed (Figure S5a,b). E U is visible in low crystalline, disordered, or amorphous materials with localized states in the standard bandgap due to exciton/phonon or electron/phonon interactions. The results show that with the modification of the VO 2 structure, the electronic transitions from band to tail and tail to tail are facilitated with probably a better order between the layers [55]. Thus, the combination Mo 0.2 W 0.8 O 3 /W-VO 2 /MoO 3 is optimum regarding the optical properties of the present materials.

Electric Properties
Due to the specificity of the correlated oxide, the evolution of the structural, optical, and electrical properties is parallel with temperature. Thus, following the electrical properties is an alternative to understanding the MIT characteristics. An Ohmic behaviour is observed at room temperature for S1 and S2. In contrast, a Schottky contact is detected for the others (S3 to S6) (Figure 7a). However, the most resistive sample is S1, and the most conductive are S3 and S4, which are the only ones containing the WO 3 layer. More than the samples' multilayer and grain structure, the influence of the intrinsic conductivity of WO 3 is the most important parameter [56]. As seen above, WO 3 favours the formation of V 2 O 5, which is the highest oxidative state among the vanadium oxides and the most resistive [57]. On the other hand, the presence of WO 3 and/or MoO 3 layers does not significantly affect samples' response time, contrarily to the samples (S5 and S6) where the Mo 0.2 W 0.8 O 3 layer is involved (Figure 7b). Whatever the combination, after roughly 9 s, the resistance of multilayer samples reaches the low values. Since Mo 0.2 W 0.8 O 3 contains similarly WO 3 , MoO 2 , and MoO 3 phases, and electrons diffusion are probably deleted within this layer. Finally, except for S1 which is purely resistive, the other samples present a semi-conductive behaviour. The VO 2 phase change from monoclinic to tetragonal crystal structure with temperature involves the decrease of the resistivity. Therefore, the MIT value (also denoted T c ) for each structure can be evaluated by following the electrical properties of the samples. When a sharp decrease is not observed in the resistivity, the measured MIT will be considered from the starting point of the decrease. In this work, the MIT for all samples is lower than the theoretical value T c = 68 • C for VO 2 (Figure 8a,f). According to the characterization above, the typical profile for the VO 2 (VO 1.73 ) layer is observed for S1. The low MIT at about 35 • C is due to the oxygen vacancies [58,59]. Nevertheless, the high resistivity values and the hysteresis observed between the heating and cooling curves show a forced transition phase. When deposited on MoO 3 (S2), the VO 2 phase transition starts at lower T c (25 • C), and the hysteresis dramatically disappears. The strain effect of MoO 3 combined with the oxygen deficiency is profitable; however, the resistivity is still high. The resistivity is significantly reduced when WO 3 is a bottom layer (S3), and a sharp resistivity decrease appears at 41 • C during heating. The T c is higher than for S1, but the theoretical abrupt decrease is only visible with this sample.
On the other hand, the lowest resistivity value is not reached at 41 • C, and the XPS results from the formation of V 2 O 5 at the film surface. Therefore, it is suggested that WO3 contributes to improving the structure's conductivity. However, probably due to its insufficient strain effect, it cannot prevent the oxidation of VO 2 to V 2 O 5 , the most stable vanadium oxide. Currently, WO 3 is used as an anti-reflective layer. Therefore, sandwich structure with MoO 3 at the bottom and WO 3 at the upper layer was investigated with S4. As expected, the MIT was reduced to T c = 30 • C, and the resistivity values were low. Unfortunately, a critical hysteresis appears. Since the presence of Mo seems to facilitate the strain effect and based on the performance demonstrated within our previous work, Mo 0.2 W 0.8 O 3 , which is a Mo doped WO 3 thin film, was tested as the upper layer (S5). The lower MIT temperature (about 20 • C) and the highest resistivity values are reached, and a critical hysteresis is observed with S5. The introduction of Mo within the WO 3 structure involves the simultaneous presence of MoO 2 , MoO 3, and WO 3 . The presence of MoO 3 possibly contributes to the strain of the VO 2 structure. However, it also disrupts the WO 3 continuity involving a drop in the conductivity. Due to its confirmed anti-reflective and anti-oxidative performance, Mo 0.2 W 0.8 O 3 was kept as the upper layer. W-doped VO 2 was designed to decrease the sample S6 resistivity. The W doping is exceedingly profitable in decreasing the hysteresis and the resistivity of the structure. However, T c increases to up to 36 • C. Therefore, thanks to the sandwich structure prepared with two doped thin films, the hysteresis and the resistivity decreased significantly by almost keeping the reference T c . Based on the literature reviews, we conclude that our study is a facile and efficient way to modulate the phase transition of VO 2 thin film as in Table 4.

Conclusions
VO 2 is one of the most investigated phase transition materials since the MIT is observed close to ambient temperature. According to the application, setting the T c is a crucial parameter. In this study, the MIT temperature decreased to 41 • C via oxygen-vacancy vanadium dioxide (VO 1.73 ) through the deposition conditions. The evolution of the bandgap and the Urbach tail demonstrate the influence of the bottom and top layers on the VO 2 layer. These interactions are probably guided by the common vibration observed in the Raman spectrum between two layers which have an interface. The results showed that the structure of the VO 2 layer is more preserved when MoO 3 is used than WO 3 . The electrical measurements show that the structure resistance is lower, and the phase transition is easier with the WO 3 bottom layer. On the other hand, WO 3 does not reduce the hysteresis between the heating and cooling curve contrarily to MoO 3 . The former decreases the T c down to 25 • C, and the hysteresis is significantly reduced. Therefore, the combination of MoO 3 bottom layer and WO 3 upper layer was investigated with the WO 3 /VO 2 /MoO 3 , in which resistance is low, T c of VO 2 is kept, but the hysteresis is still present. Meanwhile, the high microstrain on VO 1.75 film of 119.9 × 10 −3 is exerted by the monoclinic WO 2.94 buffer layer, which contributed to a fast and sharp phase transition WO 2.94 /VO 1.75 /FTO glass with a low phase transition. The transmittance of Mo-doped WO 3 (Mo 0.2 W 0.8 O 3 ) is higher than WO 3, and the dislocation density is comparable to the VO 2 layer. The potential of Mo 0.2 W 0.8 O 3 as an alternative upper layer was confirmed with Mo 0.2 W 0.8 O 3 /VO 2 /MoO 3 , where the Tc decreases to 20 • C. Nonetheless, the hysteresis and the resistance are incredibly increasing. To maintain the profit of Mo 0.2 W 0.8 O 3 and improve the conductivity of the sample, VO 2 was doped with metallic W. The combination Mo 0.2 W 0.8 O 3 /VO 2 -W/MoO 3 successfully showed low T c at 36 • C, low resistivity at tens of ohm, and no hysteresis. Thanks to this study, the synergy between oxygen deficiency, strain engineering, and metal doping was reached for improving the MIT characteristics of VO 2 thin film.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.