Fabrication and Optical Characterization of VO2-Based Thin Films Deposited on Practical Float Glass by Magnetron Sputtering and Professional Annealing

In this paper, VO2 thin films with good optical properties are fabricated on practical float glass by magnetron sputtering and a professional annealing method. The near-infrared switching efficiency (NIRSE) of the prepared film reaches 39% (@2000 nm), and its near-infrared energy modulation ability (ΔTir) reaches 10.9% (780–2500 nm). Further, the highest integral visible transmittance Tlum is 63%. The proposed method exhibits good reproducibility and does not cause any heat damage to the magnetron sputtering machine. The crystalline structure of the VO2 film is characterized by X-ray diffraction (XRD). The lattice planes (011) and (−211) grow preferentially (JCPDS 65-2358), and a large number of NaV2O5 crystals are detected simultaneously. The microstructures are characterized by scanning electron microscopy (SEM), and a large number of long sheet crystals are identified. The phase transition temperature is significantly reduced by an appropriate W doping concentration (Tc = 29 °C), whereas excessive W doping causes distortion of the thermal hysteresis loop and a reduction in the NIRSE. Oxygen vacancies are created by low pressure annealing, due to which the phase transition temperature of VO2 film decreases by 8 °C. The addition of an intermediate SiO2 layer can prevent the diffusion of Na+ ions and affect the preparation process of the VO2 thin film.

Magnetron sputtering [10,11], chemical vapor deposition (CVD) [12], sol-gel processing [4,13], pulsed laser deposition [14], and molecular beam epitaxy [15] are often used to prepare VO 2 films. Among these methods, magnetron sputtering is the most widely used method owing to its advantages of controllable film thickness, high speed, low temperature, good adhesion, high production efficiency, environmental friendliness, and great industrial potential, and it can be used to prepare compact, homogeneous, and stable films. However, it faces the problems of complex process, difficult operation, and poor reproducibility in preparing VO 2 thin films. Therefore, Gurvitch et al. [16] and Luo et al. [17] firstly used magnetron sputtering to prepare metal V film and then utilized in situ oxidative annealing to prepare VO 2 thin films, which exhibited good repeatability. However, the in-situ annealing operation at high temperature (above 300-500 • C) can damage the magnetron sputtering equipment and shorten its working life. Liu et al. [18], Xu et al. [19], and Ba et al. [20] deposited V metal films on glass and sapphire by magnetron sputtering or ion-assisted sputtering technology and then obtained VO 2 films by oxidative annealing in pure oxygen atmosphere or air atmosphere in a furnace. This two-step technique was beneficial to improve the reproducibility and protect the sputtering equipment, but they did no further works in studies of substrates, visible transmittance improvement, doping, oxygen vacancies, and so on. Silicon [21], quartz [10,22], sapphire [10,23], and other materials are often used as substrates for depositing VO 2 film to achieve better phase transition performance. However, these substrates are expensive, difficult to obtain, and inconsistent with the actual glass used in windows and doors. Float glass (Na-Ca-Si ternary glass, Na~13.8%, Ca~8.6%, Si~72.8%, others~4.8%) is the most widely used practical glass in modern buildings, so it must be explored as a substrate for VO 2 to be applied in smart windows. Presently, the most vital problem of float glass is that small Na + ions can easily diffuse into the VO 2 film, which results in poor thermochromic properties [24]. VO 2 thin films have been prepared on Si 3 N 4 [25], TiO 2 [26], ZnO [27], SnO 2 [28], and other buffer layers, but the preparation methods and properties were quite different. For application in smart windows, the VO 2 film's visible transmittance should be above 60%, and the key performance of the transition temperature must be reduced to a practical range of 25-30 • C. Tungsten doping and oxygen vacancy production have been considered to be effective methods [29][30][31], but the mechanisms of tungsten doping for reducing the phase transition temperature and the equivalent effect of oxygen vacancy are still not clear.
In this study, a two-step method of magnetron sputtering and professional annealing with good reproducibility and no heat damage to the sputtering machine was successfully utilized to deposit VO 2 -based thermochromic thin films on a practical float glass, including intrinsic VO 2 thin films, VO 2 thin films with intermediate SiO 2 layers, and W-doped VO 2 thin films. Further, the annealing control, visible transmittance adjustment, the effect of 100-nm-thick intermediate SiO 2 layer on preventing the diffusion of Na + ions, and the equivalent action of W dopant and O vacancy were investigated.

Preparation of V-Based Metal Films by Magnetron Sputtering
The practical float glass used in this study was SAIL BRAND 7101 with a thickness of 1 mm. The magnetron sputtering machine used was ADV-ZKS-0350 (Aitmet Vacuum Machinery Co., Ltd., Foshan, Guangdong, China). The float glass (hereafter referred to as glass) was cut into a size of 25.4 mm × 38.1 mm, then ultrasonically cleaned in acetone and anhydrous ethanol for 30 min, and subsequently dried off for later use. The sputtering vacuum chamber was thoroughly cleaned, and then the V metal target (ϕ: 56 mm, 99.9% purity) and glass substrate were installed. The chamber was vacuumized to 3.0 × 10 −3 Pa, and then Ar gas (99.99% purity) was passed through it to achieve a pressure of 1 Pa. DC magnetron sputtering was utilized to deposit the V-based metal films at room temperature (20 • C). The impurity atoms on the target surface were removed by pre-sputtering for 10 min, and then the baffle was opened to deposit films at a power of 77 W. The sputtering time was controlled to 5, 7.5, and 10 min to obtain V metal films with thicknesses of 30, 45, and 60 nm (calculated by statistical average deposition rate), respectively. The main process parameters are shown in Table 1, and the quartz glass is a contrast substrate.  Figure 1) and DC magnetron sputtering technology were applied to execute W doping in an appropriate ratio (0-1.0%, the ratio of the area of W bars to the total area of the target). According to the sputtering yield, sputtering area, and atomic concentration, the W dopant ratio by mol can be calculated according to the area ratio of W bars. The W dopant's ratios by area of 0%, 0.5%, and 1% correspond to the ratios by mol of 0%, 2.83%, and 5.53%, respectively. For convenience, the W dopant ratio in this paper represents the area ratio unless otherwise specified. The sputtering time for the W:V metal film was 7.5 min, and the thickness was approximately 45 nm. The main process parameters are shown in Table 1.   Figure 1) and DC magnetron sputtering technology were applied to execute W doping in an appropriate ratio (0-1.0%, the ratio of the area of W bars to the total area of the target). According to the sputtering yield, sputtering area, and atomic concentration, the W dopant ratio by mol can be calculated according to the area ratio of W bars. The W dopant's ratios by area of 0%, 0.5%, and 1% correspond to the ratios by mol of 0%, 2.83%, and 5.53%, respectively. For convenience, the W dopant ratio in this paper represents the area ratio unless otherwise specified. The sputtering time for the W:V metal film was 7.5 min, and the thickness was approximately 45 nm. The main process parameters are shown in Table 1. Figure 1. Schematic of the tungsten-vanadium inlaid target (when W bars are fully embedded, doping ratio-max: 7%). (The blue bar is W. One 1.8° W bar represents 0.5% doping ratio by area, and one 3.6° W bar represents 1.0% doping ratio by area; Increase the number of W bars can get different W doping concentrations, and the maximum can reach 7%.). RF magnetron sputtering technology was used to deposit the SiO2 layer on the glass for blocking the sodium ion diffusion. The main process parameters are shown in Table  2. Then, 30 nm V metal film was spread on the intermediate SiO2 layer (Table 1) RF magnetron sputtering technology was used to deposit the SiO 2 layer on the glass for blocking the sodium ion diffusion. The main process parameters are shown in Table 2. Then, 30 nm V metal film was spread on the intermediate SiO 2 layer (Table 1).

Preparation of VO 2 -Based Films by Professional Annealing
The deposited V-based metal films (V, W:V, V/SiO 2 ) were placed in a vacuum tube furnace (ECFK- [6][7][8][9][10][11][12][13][14], vacuumized to a needed low pressure, heated to 400 • C at 5 • C/min, annealed in low pressure air atmosphere for 1 h, and then naturally cooled to room temperature to obtain VO 2 -based films. The main process parameters are shown in Table 3, and the quartz glass is a contrast substrate.

Testing and Characterization
The transmittance of the film was examined using ultraviolet-visible-infrared spectrophotometry (UV-3600 MPC 3100, Shimadsu, Kyoto, Japan), where the testing wavelength range was 250-2500 nm, the step length was 1 nm, and the testing temperature was 20 and 90 • C. The heating attachment was a small numerical control furnace with a temperature error of ±1 • C. The near-infrared switching efficiency (NIRSE) is defined as the difference between the transmittance at low temperature and the transmittance at high temperature at 2000 nm. The thermal hysteresis loop was drawn using the transmittance curves at different temperatures in the range of 20-90 • C with an interval of 5 • C both in the heating and cooling process. Further, each temperature point was held for at least 5 min before obtaining the transmittance curve. Moreover, a standard VO 2 film sample (quartz glass substrate, NIRSE 50%, i.e., the VO 2 film by "30 nm V/quartz glass" in Table 3) was used to complete the transmittance measurement in whole wavelength 250-2500 nm at any two temperature points, e.g., 20 • C and 40 • C, to check the sameness and uniformity of the transmittance curve before every loop test. To obtain the thermal hysteresis loop of the VO 2 film, the transmittance values at 2000 nm in each curve were considered as the Y-axis, and temperature points were taken for the X-axis.

Data Treatment
According to the transmittance curve and Equation (1), the visible spectral transmittance can be obtained [32], which is also known as the integrated visual photopic transmission (IVPT, denoted by T lum ). This kind of transmittance takes the people's visual perception into account and is more suitable for the daylighting evaluation of green buildings. It is calculated as follows: where D λ is the relative spectral distribution of illuminant D65, V(λ) is the spectral luminous efficiency for photopic vision, and T(λ) is the visible transmittance. Further, the values of D λ ·V(λ)·∆λ were obtained from the international standard ISO 9095-2003.
According to the transmission curve and Equation (2), the near-infrared spectral transmittance (T ir ) can be calculated [32]. The regulation ability to near-infrared light (∆T ir ) is expressed as the difference between T ir at high temperature and T ir at low temperature. T ir is calculated as follows: where S λ is the relative spectral distribution of solar radiation, T(λ) is the near-infrared transmittance, and the values of S λ ·∆λ were obtained from the international standard ISO 9095-2003. Figure 2a shows that at the annealing pressure of 450 Pa, the NIRSE (39%) of the VO 2 film prepared by the 45 nm V metal film is higher than that prepared by the 30 nm, 60 nm V metal films (20%, 26%). This indicates that the optimal thickness of the V film under this condition is nearly 45 nm. Furthermore, Figure 2b shows that for the 45 nm V metal film, the NIRSE (39%) of VO 2 film prepared at the annealing pressure of 450 Pa is higher than that at 200 and 700 Pa (16%, 30%). This suggests that the optimal annealing pressure of the V film under this condition is approximately 450 Pa. When the thickness and annealing pressure of the VO 2 film are optimized, the NIRSE is enhanced, and the values of T ir and T lum change accordingly. Generally, as NIRSE increases, ∆T ir increases. When the thickness of the V metal film decreases or the annealing pressure increases, T lum increases, and vice versa, as illustrated in Table 4.  The XRD pattern in Figure 3a shows that the phase composition of VO2 films pre-  The XRD pattern in Figure 3a shows that the phase composition of VO 2 films prepared by annealing of 30, 45, and 60 nm V film at 450 Pa. With the thickness increasing of V film, the concentration ratio of NaV 2 O 5 : VO 2 decreases. Because of the low concentration and micro size of the VO 2 particles, the VO 2 film from 30 nm V presents nearly no corresponding diffraction peaks. The VO 2 film prepared by annealing of 45 and 60 nm V film at 450 Pa includes not only a VO 2 , but also a NaV 2 O 5 diffraction peak. The (011) and (−211) lattice planes of VO 2 grow preferentially, and their diffraction peaks are stronger. The crystalline phase (M, monoclinic) of the VO 2 film corresponds to the standard card JCPDS 65-2358. Moreover, the particle sizes of the VO 2 films by 45 and 60 nm V are both about 15 nm, as calculated by Scherrer formula. The SEM image in Figure 3b suggests that the microstructure of VO 2 films prepared by annealing of 30, 45, and 60 nm V film at 450 Pa is more or less the same. Take the VO 2 film from 45 nm V for instance, where the microstructure is composed of long sheet crystals (length: 210 nm, width: 50 nm, thickness: 20 nm). It can be inferred that the long sheets are NaV 2 O 5 crystals, which present obvious oriented growth. The same morphology of NaV 2 O 5 crystals was reported by Mjejri et al. [33]. Smaller VO 2 crystalline particles are scattered in the interspace of NaV 2 O 5 crystals, which are difficult to observe due to the influence of the depth of field of the long sheet NaV 2 O 5 crystals.  The XRD pattern in Figure 4a shows the phase composition of VO2 films prepared by the annealing of 45nm V film at pressures 200, 450, 700 Pa. Compared to 200 Pa, the 450 and 700 Pa result in an obvious higher concentration of NaV2O5, and the 450 Pa leads to a higher ratio of VO2: NaV2O5 than the 700 Pa. Moreover, the VO2 film by 450 Pa performs crystalline phase (M) corresponding to the standard card JCPDS 65-2358. And the particle sizes of the VO2 films by 45 nm V at pressures 200, 450, and 700 Pa are about 12, 15, and 11 nm respectively, calculated by the Scherrer formula. The SEM image in Figure 4b suggests that the microstructures of VO2 films prepared by the annealing of 45 nm V film at pressures 200, 450, and 700 Pa are quite different. With the pressure increasing, the film microstructure presents bigger and bigger single crystals with an obvious oriented growth and a long sheet morphology, which are inferred to be NaV2O5 likewise. VO2 crystalline particles are scattered in the single crystals with depth of field and are difficult to observe. The XRD pattern in Figure 4a shows the phase composition of VO 2 films prepared by the annealing of 45 nm V film at pressures 200, 450, 700 Pa. Compared to 200 Pa, the 450 and 700 Pa result in an obvious higher concentration of NaV 2 O 5 , and the 450 Pa leads to a higher ratio of VO 2 : NaV 2 O 5 than the 700 Pa. Moreover, the VO 2 film by 450 Pa performs crystalline phase (M) corresponding to the standard card JCPDS 65-2358. And the particle sizes of the VO 2 films by 45 nm V at pressures 200, 450, and 700 Pa are about 12, 15, and 11 nm respectively, calculated by the Scherrer formula. The SEM image in Figure 4b suggests that the microstructures of VO 2 films prepared by the annealing of 45 nm V film at pressures 200, 450, and 700 Pa are quite different. With the pressure increasing, the film microstructure presents bigger and bigger single crystals with an obvious oriented growth and a long sheet morphology, which are inferred to be NaV 2 O 5 likewise. VO 2 crystalline particles are scattered in the single crystals with depth of field and are difficult to observe.  When the thickness of V film is reduced from 45 nm to 30 nm, the optimal annealing pressure is reduced to 100 Pa, as shown in Figure 5. In general, thicker V metal films require higher optimal annealing pressure to obtain large NIRSE, and thinner V metal films require lower optimal annealing pressure to obtain large NIRSE. The optical properties of the reduced VO2 thin films are shown in Table 5.   When the thickness of V film is reduced from 45 nm to 30 nm, the optimal annealing pressure is reduced to 100 Pa, as shown in Figure 5. In general, thicker V metal films require higher optimal annealing pressure to obtain large NIRSE, and thinner V metal films require lower optimal annealing pressure to obtain large NIRSE. The optical properties of the reduced VO 2 thin films are shown in Table 5.  When the thickness of V film is reduced from 45 nm to 30 nm, the optimal an pressure is reduced to 100 Pa, as shown in Figure 5. In general, thicker V metal f quire higher optimal annealing pressure to obtain large NIRSE, and thinner V met require lower optimal annealing pressure to obtain large NIRSE. The optical prop the reduced VO2 thin films are shown in Table 5.   According to the transmittance curves in Figure 6a and the thermal hysteresis loops in Figure 6b of the W-doped VO 2 films by 45 nm W:V at 450 Pa, the phase transition temperature (T c ) of 0.5%W-doped VO 2 thin film is significantly decreased (T c = 29 • C). However, the NIRSE remains high (27%), and the shape of the thermal hysteresis loop is not distorted seriously. With the increase in the concentration of W dopant to 1% (i.e., 1.0%W-doped VO 2 film), the phase transition temperature does not decline continuously but rises, especially above the temperature T d or T d (33 • C or 28 • C), due to which the average T c decreases to 39 • C. In addition, the NIRSE is negatively affected and reduces to 13%. At the same time, the shape of the thermal hysteresis loop is seriously distorted: the height decreases (13%, while for 0.5%W: 27%), the gradient becomes more gentle, and the width increases considerably (16 • C, while for 0.5%W: 7 • C). According to the transmittance curves in Figure 6a and the thermal hysteresis loops in Figure 6b of the W-doped VO2 films by 45 nm W:V at 450 Pa, the phase transition temperature (Tc) of 0.5%W-doped VO2 thin film is significantly decreased (Tc = 29 °C). However, the NIRSE remains high (27%), and the shape of the thermal hysteresis loop is not distorted seriously. With the increase in the concentration of W dopant to 1% (i.e., 1.0%Wdoped VO2 film), the phase transition temperature does not decline continuously but rises, especially above the temperature Td or Td' (33 °C or 28 °C), due to which the average Tc decreases to 39 °C. In addition, the NIRSE is negatively affected and reduces to 13%. At the same time, the shape of the thermal hysteresis loop is seriously distorted: the height decreases (13%, while for 0.5%W: 27%), the gradient becomes more gentle, and the width increases considerably (16 °C, while for 0.5%W: 7 °C). The XRD pattern in Figure 7a shows that the peak intensities of the crystal NaV2O5 and the crystalline VO2 both decrease with the increase of W dopant. The SEM image in Figure 7b suggests that the size of the long sheet NaV2O5 crystal declines gradually with the increase of W dopant, which corresponds to the result of XRD. The W-doped VO2 crystallites cannot be seen in the SEM image for the depth of field of the long sheet NaV2O5 crystals, but the particle size calculated from XRD pattern by the Scherrer formula presents a decline trend too, with 15, 13, and 11 nm, corresponding to 0%, 0.5%, and 1.0%W respectively. W doping makes local T-VO2 phase increase and M-VO2 phase decrease, which results in tighter lattice structure and greater difficulty for Na diffusion into the The XRD pattern in Figure 7a shows that the peak intensities of the crystal NaV 2 O 5 and the crystalline VO 2 both decrease with the increase of W dopant. The SEM image in Figure 7b suggests that the size of the long sheet NaV 2 O 5 crystal declines gradually with the increase of W dopant, which corresponds to the result of XRD. The W-doped VO 2 crystallites cannot be seen in the SEM image for the depth of field of the long sheet NaV 2 O 5 crystals, but the particle size calculated from XRD pattern by the Scherrer formula presents a decline trend too, with 15, 13, and 11 nm, corresponding to 0%, 0.5%, and 1.0%W respectively. W doping makes local T-VO 2 phase increase and M-VO 2 phase decrease, which results in tighter lattice structure and greater difficulty for Na diffusion into the film. In addition, W atom has more free electrons, which leads to a greater reducibility of the doped V film and higher consumption of oxygen. In conclusion, the above two reasons result in the decrement of the crystal NaV 2 O 5 and the crystalline VO 2 .

Transmittance and Microstructure of VO 2 Film Based on the Optimization of Thickness and Pressure
film. In addition, W atom has more free electrons, which leads to a greater reducibility the doped V film and higher consumption of oxygen. In conclusion, the above two reaso result in the decrement of the crystal NaV2O5 and the crystalline VO2.

Performance and Micro Morphology of VO2 Film on Intermediate SiO2 Layer
The thickness of the intermediate SiO2 layer prepared by RF magnetron sputtering approximately 100 nm (Figure 8a). Then, 30 nm V film is deposited on it by DC magnetr sputtering (Table 1). Further, VO2 films are prepared by professional annealing of the metal films at 200, 450, and 600 Pa (Table 3). According to the SEM image of the VO2 fi prepared at the optimized pressure 450 Pa (Figure 8c), although the SiO2 layer is par punctured, the number of long sheet crystals is significantly reduced. Moreover, it comes easier to observe the VO2 crystallites (size about 14 nm by Scherrer Formula) sc tering in the film (size about 14 nm too by SEM and software Nano Measurer 1.2.5). Co pared with the VO2 film without SiO2 layer in Figure 3b-30 nm V, it indicates that the Si layer plays a crucial role in blocking the diffusion of Na + ions, which leads to an increa of the ratio of VO2: Na-V-O according to the XRD patterns in Figures 3a and 8d, 30 nm Consequently, the NIRSE rises (22% at 450 Pa, Figure 8b; original: 20%, Figure 2a), a the near-infrared energy modulation ability △Tir is also enhanced (8.8%; original: 7.0% Meanwhile, the XRD pattern (Figure 8d) demonstrates that the 100 nm intermediate Si layer cannot completely prevent the diffusion of Na + ions, and a large number of NaV2 crystals are still generated due to pinholes in the SiO2 and diffusion of Na through it.

Performance and Micro Morphology of VO 2 Film on Intermediate SiO 2 Layer
The thickness of the intermediate SiO 2 layer prepared by RF magnetron sputtering is approximately 100 nm (Figure 8a). Then, 30 nm V film is deposited on it by DC magnetron sputtering (Table 1). Further, VO 2 films are prepared by professional annealing of the V metal films at 200, 450, and 600 Pa (Table 3). According to the SEM image of the VO 2 film prepared at the optimized pressure 450 Pa (Figure 8c), although the SiO 2 layer is partly punctured, the number of long sheet crystals is significantly reduced. Moreover, it becomes easier to observe the VO 2 crystallites (size about 14 nm by Scherrer Formula) scattering in the film (size about 14 nm too by SEM and software Nano Measurer 1.2.5). Compared with the VO 2 film without SiO 2 layer in Figure 3b-30 nm V, it indicates that the SiO 2 layer plays a crucial role in blocking the diffusion of Na + ions, which leads to an increase of the ratio of VO 2 : Na-V-O according to the XRD patterns in Figures 3a and 8d, 30 nmV. Consequently, the NIRSE rises (22% at 450 Pa, Figure 8b; original: 20%, Figure 2a), and the near-infrared energy modulation ability T ir is also enhanced (8.8%; original: 7.0%). Meanwhile, the XRD pattern (Figure 8d) demonstrates that the 100 nm intermediate SiO 2 layer cannot completely prevent the diffusion of Na + ions, and a large number of NaV 2 O 5 crystals are still generated due to pinholes in the SiO 2 and diffusion of Na through it.

Necessity of Two Equipments and Two Steps
The two-step method of magnetron sputtering-professional annealing used in this study not only has good repeatability, but also causes no heat damage to the magnetron sputtering equipment, which can prolong the service life of the equipment. Crystallization using special high-temperature equipment can promote a better phase transition performance. The method can be easily optimized to adjust the doping concentration, oxygen vacancies, content of high-valence NaV2O5, film thickness, etc., which are very useful to regulate the phase transition temperature, visible transmittance, and other thermochromic properties.

Effect of Substrate on the Preparation and Property of VO2 Thin Film
Generally, the NIRSE of VO2 film on the glass substrate is lower than that on the quartz glass substrate (Figure 9a). This is because a part of the V atoms react with Na atoms from glass, O atoms from glass, and air to form NaV2O5, which results in low production of crystalline VO2, as shown in the XRD pattern (Figure 9b). In terms of the preparation technique, the VO2 film prepared on the quartz glass by annealing 30 nm V film at 1500 Pa exhibits a NIRSE value of 50% (Figure 9a). However, the film prepared on the glass by annealing 30 nm V film at 1500 Pa shows a NIRSE value of 0%. Nevertheless,

Necessity of Two Equipments and Two Steps
The two-step method of magnetron sputtering-professional annealing used in this study not only has good repeatability, but also causes no heat damage to the magnetron sputtering equipment, which can prolong the service life of the equipment. Crystallization using special high-temperature equipment can promote a better phase transition performance. The method can be easily optimized to adjust the doping concentration, oxygen vacancies, content of high-valence NaV 2 O 5 , film thickness, etc., which are very useful to regulate the phase transition temperature, visible transmittance, and other thermochromic properties.

Effect of Substrate on the Preparation and Property of VO 2 Thin Film
Generally, the NIRSE of VO 2 film on the glass substrate is lower than that on the quartz glass substrate (Figure 9a). This is because a part of the V atoms react with Na atoms from glass, O atoms from glass, and air to form NaV 2 O 5 , which results in low production of crystalline VO 2 , as shown in the XRD pattern (Figure 9b). In terms of the preparation technique, the VO 2 film prepared on the quartz glass by annealing 30 nm V film at 1500 Pa exhibits a NIRSE value of 50% (Figure 9a). However, the film prepared on the glass by annealing 30 nm V film at 1500 Pa shows a NIRSE value of 0%. Nevertheless, when the annealing pressure is reduced to 100 Pa, a high NIRSE value (35%, Figure 9a) is achieved. The surface of VO 2 film on quartz glass is flat and compact (Figure 9c), and like a passivation coating, can stand up to the high oxidation pressure 1500 Pa. Meanwhile, the surface of VO 2 film on glass is a loose structure consisted of disordered long sheet NaV 2 O 5 crystals and pinholes, which results in greater easiness of oxygen indiffusion from the atmosphere and easiness of oxidization of V to high valence NaV 2 O 5 under a high pressure. In addition, there are more free non-bridged oxygen atoms in the glass substrate. Despite the much greater difficulty of O 2− ion outdiffusion than Na + for its strong Coulomb force, there still is the possibility of O emigrating to the surface and oxidizing the V metal film during the annealing. However, in quartz glass, O atoms are firmly bonded with Si atoms by the covalent bonds and are very difficult to move. So, in conclusion, when on glass substrate, the oxidation pressure is much lower, because of the loose surface of the film majorly, as well as the O outdiffusion from the substrate minorly; and when on quartz glass, the oxidation pressure could be much higher, because of the compact surface of the film. achieved. The surface of VO2 film on quartz glass is flat and compact (Figure 9c), and like a passivation coating, can stand up to the high oxidation pressure 1500 Pa. Meanwhile, the surface of VO2 film on glass is a loose structure consisted of disordered long sheet NaV2O5 crystals and pinholes, which results in greater easiness of oxygen indiffusion from the atmosphere and easiness of oxidization of V to high valence NaV2O5 under a high pressure. In addition, there are more free non-bridged oxygen atoms in the glass substrate. Despite the much greater difficulty of O 2− ion outdiffusion than Na + for its strong Coulomb force, there still is the possibility of O emigrating to the surface and oxidizing the V metal film during the annealing. However, in quartz glass, O atoms are firmly bonded with Si atoms by the covalent bonds and are very difficult to move. So, in conclusion, when on glass substrate, the oxidation pressure is much lower, because of the loose surface of the film majorly, as well as the O outdiffusion from the substrate minorly; and when on quartz glass, the oxidation pressure could be much higher, because of the compact surface of the film.

Choice between NIRSE and Tlum
The primary goal of VO2 film preparation is to maximize the value of NIRSE, which means that the content and purity of VO2 in the film should reach the highest level. However, for smart windows, both NIRSE and visible transmittance Tlum should be considered. The minimum daylighting rate of green building glass is generally required to be above 60%. Therefore, it is useful to sacrifice a part of NIRSE to improve the Tlum, while ensuring an appropriate high value of NIRSE. The common method to improve the visible transmittance is to deposit anti-reflection films. However, two other simple approaches were used in this study: thinning the V metal film and increasing the annealing pressure. For instance, when the V metal film is thinned from 45 nm to 30 nm, the NIRSE decreases from 39% to 20%, but the Tlum increases from 44.5% to 63%, as shown in Figure 2a and Table 4. When the annealing pressure of 45 nm V film increases from 450 Pa to 700 Pa, the NIRSE decreases from 39% to 30%, but the Tlum increases from 44.5% to 61%, as shown in Figure   500 1000

Choice between NIRSE and T lum
The primary goal of VO 2 film preparation is to maximize the value of NIRSE, which means that the content and purity of VO 2 in the film should reach the highest level. However, for smart windows, both NIRSE and visible transmittance T lum should be considered. The minimum daylighting rate of green building glass is generally required to be above 60%. Therefore, it is useful to sacrifice a part of NIRSE to improve the T lum , while ensuring an appropriate high value of NIRSE. The common method to improve the visible transmittance is to deposit anti-reflection films. However, two other simple approaches were used in this study: thinning the V metal film and increasing the annealing pressure. For instance, when the V metal film is thinned from 45 nm to 30 nm, the NIRSE decreases from 39% to 20%, but the T lum increases from 44.5% to 63%, as shown in Figure 2a and Table 4. When the annealing pressure of 45 nm V film increases from 450 Pa to 700 Pa, the NIRSE decreases from 39% to 30%, but the T lum increases from 44.5% to 61%, as shown in Figure 2b and Table 4. The two methods' mechanism is: at the same high enough oxidation pressure, thinner V metal films will be more easily oxidized into high-valence vanadium; and for the V metal films with same thickness, higher oxidation pressure will result in more high-valence vanadium likewise. The content of low-valence vanadium (mainly crystalline VO 2 ) in the film decreases, and the content of high-valence vanadium (mainly crystal NaV 2 O 5 ) increases, which leads to an increase in the optical band gap, and results in the increase of the visible transmittance of the film. Moreover, there is another for the thinner film reason, namely lower light absorption.

Mechanism of W Doping and Equivalent Effect of O Vacancy
There are two mechanisms through which W doping reduces the phase transition temperature (T c ) of VO 2 : Peierls mechanism and Mott mechanism. Among them, the Peierls mechanism is described by the energy band theory [34][35][36], where structural change is considered as the driving force. After larger W atoms replace the smaller V atoms, the lattice expands and the V-V dimerization pairs are destroyed, which results in the change of energy band position. The band gap between V3d // and π* becomes narrower, the electronic migration becomes easier, and the phase transition barrier is reduced, so the T c decreases. In addition, according to the energy band theory, T c continues to decline with the increase in the concentration of W dopant, which disagrees with our experimental results (decreases first and then rises). Therefore, this paper supports the second theory of the Mott mechanism. The Mott mechanism is described by the charge-doping theory [37][38][39], where electron correlation is the driving force. After V 4+ ions are replaced by W 6+ ions, the excess electrons enter the empty π* band, and then local conduction bands and local metallic phases are formed. Therefore, the transformation into tetragonal phase becomes easier, and thus the transition temperature T c declines. However, excessive W doping can cause superabundant electrons to enter the empty π* band, and the interaction of electrons is enhanced, which results in the splitting of π* band and the formation of a new band gap. The phase transition barrier becomes larger, and thus the T c rises again. The mechanism of W doping to reduce the phase transition temperature of VO 2 is still under debate, just like the debate between Peierls mechanism and Mott mechanism for the intrinsic phase transition of pure VO 2 [40,41]. Presently, there is no unified theory that can fully describe the effect of W doping.
Although the phase transition temperature is reduced by W doping, the thermochromic properties are adversely affected, and the most significant change is that the NIRSE declines significantly. The local conduction bands are formed after W doping, and the electronic transition becomes easier. At low temperature, the ability of reflecting infrared radiation is enhanced, and the infrared transmittance is reduced, and thus the NIRSE is clearly reduced. With the increase of W dopant, the thermal hysteresis loop becomes gentle (Figure 6b) and is no longer steep. This is directly related to the reduction of phase transition temperature at low temperature (below T d or T d' ), the rise of phase transition temperature at high temperature (above T d or T d' ), and the decline of NIRSE. At low temperature (below T d or T d' ), due to the increase of W dopant, local conduction bands in empty π* band are increased, and the monoclinic-tetragonal phase transition can be easily realized, which results in the reduction of the phase transition temperature. At high temperature (above T d or T d' ), due to the gradual increase of tetragonal metallic phase, more and more electrons enter the empty π* band, and the interaction becomes increasingly stronger, which results in worse fragmentation of the π* band, a wider band gap, larger phase transition barrier, and thus the increase of the phase transition temperature.
Similar to W doping, the oxygen vacancies can lower the phase transition temperature of VO 2 , which were created by lowering the annealing pressure. For example, the T c of VO 2 film fabricated by annealing 45 nm V film at 200 Pa is 8 • C lower than that of VO 2 film fabricated by annealing 45 nm V film at 450 Pa, as shown in Figure 10a. The oxygen vacancies lead to the transformation of V 4+ into V 3+ and generation of excess electrons, which enter the empty π* band and form a local conduction band. The growth of the charge carrier concentration and the enhancement of the metal phase results in the decline of phase transition temperature, which is similar to the effect of W doping at low concentration. However, it should be noted that when W dopant is used for lowering the phase transition temperature of VO 2 , and the concentration has reached an optimal value, i.e., a level (such as 0.5%W) that can cause T c to rise rather than continually decline, the preparation of oxygen vacancies has an opposite effect. For instance, the T c of the VO 2 film prepared by annealing 45 nm 0.5%W/V film at 300 Pa is 10 • C higher than that of the VO 2 film prepared by annealing 45 nm 0.5%W/V film at 450 Pa, as shown in Figure 10a. The formation of oxygen vacancies is indicated by the increase of the ratio of V 3+ : V 4+ in the V2p 3/2 XPS spectra (Figure 10b). The three fitting peaks correspond to V 3+ (515.2 eV), V 4+ (516.0 eV), and V 5+ (517.2 eV), respectively [42]. It is found that the V 3+ : V 4+ ratio increases from 0.86 to 0.98 when the annealing pressure decreases from 450 Pa to 200 Pa for 45 nm V, and increases from 0.59 to 0.75 when the annealing pressure decreases from 450 Pa to 300 Pa for 45 nm 0.5%W/V. which enter the empty π* band and form a local conduction band. The growth of the charge carrier concentration and the enhancement of the metal phase results in the decline of phase transition temperature, which is similar to the effect of W doping at low concentration. However, it should be noted that when W dopant is used for lowering the phase transition temperature of VO2, and the concentration has reached an optimal value, i.e., a level (such as 0.5%W) that can cause Tc to rise rather than continually decline, the preparation of oxygen vacancies has an opposite effect. For instance, the Tc of the VO2 film prepared by annealing 45 nm 0.5%W/V film at 300 Pa is 10 °C higher than that of the VO2 film prepared by annealing 45 nm 0.5%W/V film at 450 Pa, as shown in Figure 10a. The formation of oxygen vacancies is indicated by the increase of the ratio of V 3+ : V 4+ in the V2p3/2 XPS spectra (Figure 10b). The three fitting peaks correspond to V 3+ (515.2 eV), V 4+ (516.0 eV), and V 5+ (517.2 eV), respectively [42]. It is found that the V 3+ : V 4+ ratio increases from 0.86 to 0.98 when the annealing pressure decreases from 450 Pa to 200 Pa for 45 nm V, and increases from 0.59 to 0.75 when the annealing pressure decreases from 450 Pa to 300 Pa for 45 nm 0.5%W/V. In addition, the causes of the decrement of Tc should probably include the strain in the film, as studied by Elia [43], Paez [44], and so on [45]. The substrate used her is glass, so there would not be induction for strain generation by certain lattice planes such as TiO2(001) [43][44][45]. However, in spite of this, there would be some strain inevitably existing in the interface between glass and W/VO2 film, grain boundaries, lattice deformation, and so on. So, the strain in the film after W doping would be another cause for Tc reduction, though not the main cause.

Effect of 100 nm Intermediate SiO2 Layer
As described in Section 3.3, a 100-nm-thick intermediate SiO2 layer pre-deposited on the glass plays an obvious role in blocking the diffusion of Na + ions. However, it does not stop the diffusion completely, and many NaV2O5 crystals are still generated. As a next step, increasing the thickness of the intermediate SiO2 layer or replacing it with a better layer (compact, hard, narrow lattice plane, etc.) can completely block the diffusion of Na + ions. Meanwhile, the addition of the intermediate SiO2 layer changes the technical parameters for the preparation of VO2 thin film, and the optimal oxidation pressure increases In addition, the causes of the decrement of T c should probably include the strain in the film, as studied by Elia [43], Paez [44], and so on [45]. The substrate used her is glass, so there would not be induction for strain generation by certain lattice planes such as TiO 2 (001) [43][44][45]. However, in spite of this, there would be some strain inevitably existing in the interface between glass and W/VO 2 film, grain boundaries, lattice deformation, and so on. So, the strain in the film after W doping would be another cause for T c reduction, though not the main cause.

Effect of 100 nm Intermediate SiO 2 Layer
As described in Section 3.3, a 100-nm-thick intermediate SiO 2 layer pre-deposited on the glass plays an obvious role in blocking the diffusion of Na + ions. However, it does not stop the diffusion completely, and many NaV 2 O 5 crystals are still generated. As a next step, increasing the thickness of the intermediate SiO 2 layer or replacing it with a better layer (compact, hard, narrow lattice plane, etc.) can completely block the diffusion of Na + ions. Meanwhile, the addition of the intermediate SiO 2 layer changes the technical parameters for the preparation of VO 2 thin film, and the optimal oxidation pressure increases from 100 Pa to 450 Pa (Figures 5 and 8b). Similar to the mechanism described in Section 4.2, the intermediate SiO 2 layer causes fewer long sheet NaV 2 O 5 crystals and a flatter, more compact surface than that on glass substrate (Figure 3b-30 nm V, Figure 8c), which results in the difficulty of oxygen indiffusion from the atmosphere. Besides, the intermediate SiO 2 blocks the migration of non-bridged oxygen atoms from the glass into the film. In conclusion, the demand of oxidation annealing pressure rises, mainly because of the compacter surface of the film, and minorly because of the block of O immigration into film from glass.

Conclusions
Magnetron sputtering and the professional annealing method have the advantages of simple operation, convenient optimization, good repeatability, and no heat damage to the sputtering machine. Therefore, the represent an appropriate method for the fabrication and systematical examination of VO 2 thin films. To obtain the largest NIRSE value is the primary objective. The thickness must match with the optimal pressure. Otherwise, the value of NIRSE will decline. Under the fixed annealing condition of 400 • C (1 h), a thicker V-based metal film requires a higher optimal annealing pressure, and the thinner film requires a lower optimal annealing pressure. When a glass substrate is used, mainly because of the loose surface structure of disordered long sheet NaV 2 O 5 and pinholes, the oxidation annealing pressure reduces sharply. In this study, it was verified that compared with the quartz glass substrate, the annealing pressure of the 30 nm V metal film on the glass substrate decreased by more than 90%. Increasing the annealing pressure or decreasing the thickness of V metal film was demonstrated as a new effective method to improve the visible transmittance T lum . The 100 nm intermediate SiO 2 layer played an obvious role in preventing the diffusion of Na + ions, though not completely. Thicker or new intermediate layers need to be further studied. This study supports the charge-doping theory. A proper concentration of W dopant can lead to the formation of local conduction bands in the π* band, which can reduce the T c of VO 2 thin film. However, excessive W doping can lead to the splitting of π* band, which results in the formation of a new band gap and an increase in the phase transition barrier, and thus the T c increases. In terms of lowering the T c of VO 2 film, oxygen vacancy and W dopant have an equivalent effect. Further, when the W doping concentration is optimal, no more oxygen vacancies should be added. Otherwise, an opposite effect can occur.

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