Modulation of Structure and Optical Property of Nitrogen-Incorporated VO2 (M1) Thin Films by Polyvinyl Pyrrolidone

VO2, as a promising material for smart windows, has attracted much attention, and researchers have been continuously striving to optimize the performance of VO2-based materials. Herein, nitrogen-incorporated VO2 (M1) thin films, using a polyvinylpyrrolidone (PVP)-assisted sol–gel method followed by heat treatment in NH3 atmosphere, were synthesized, which exhibited a good solar modulation efficiency (ΔTsol) of 4.99% and modulation efficiency of 37.6% at 2000 nm (ΔT2000 nm), while their visible integrated transmittance (Tlum) ranged from 52.19% to 56.79% after the phase transition. The crystallization, microstructure, and thickness of the film could be regulated by varying PVP concentrations. XPS results showed that, in addition to the NH3 atmosphere-N doped into VO2 lattice, the pyrrolidone-N introduced N-containing groups with N–N, N–O, or N–H bonds into the vicinity of the surface or void of the film in the form of molecular adsorption or atom (N, O, and H) filling. According to the Tauc plot, the estimated bandgap of N-incorporated VO2 thin films related to metal-to-insulator transition (Eg1) was 0.16–0.26 eV, while that associated with the visible transparency (Eg2) was 1.31–1.45 eV. The calculated Eg1 and Eg2 from the first-principles theory were 0.1–0.5 eV and 1.4–1.6 eV, respectively. The Tauc plot estimation and theoretical calculations suggested that the combined effect of N-doping and N-adsorption with the extra atom (H, N, and O) decreased the critical temperature (τc) due to the reduction in Eg1.


Introduction
Monoclinic phase vanadium dioxide [VO 2 (M1)] has attracted great attention for smart coating applications because of its promising thermochromic properties [1][2][3][4]. VO 2 (M1) material undergoes a reversible metal-to-insulator transition (MIT) at a critical temperature (τ c ) of around 65 • C. It exhibits an insulating monoclinic (M1) phase at room temperature and exhibits metal characteristics as a rutile (R) phase while above the critical temperature. However, the high phase transition temperature (τ c ), low luminous transmittance (T lum ), not remarkable enough thermochromic modulation efficiency for solar energy (∆T sol ) or near-infrared modulation efficiency (∆T NIR ), and the less desirable color of the intrinsic VO 2 (M1) thin films constrain the practical application of VO 2 materials. The thermochromic performance of VO 2 is closely related to the phase composition and the microstructure, which are largely dependent on the synthesis method and growth control. Furthermore, changing the microstructure and the thickness of the VO 2 (M1) thin film can significantly regulate its physical properties. of VO 2 affected by different bonding models between N and additional elements (H, N, O, and C).

Sample Preparation
The VO 2 thin films were synthesized using the polymer-assisted chemical solution method. Vanadium (V) pentoxide (V 2 O 5 , 99.9%,) and hydrazine dihydrochloride (N 2 H 4 ·2HCl, 99.9%,) were the raw materials for the preparation of vanadium-containing solutions. PVP (K88-96, Aladdin reagent Co., Ltd., Shanghai, China) was employed as the film-forming promoter. A yellow or orange aqueous suspension (~50 mL) containing 1.820 g of V 2 O 5 was kept at 60 • C with continuously stirring. Then, the suspension color turned dark to yellowish brown, as 0.525 g of N 2 H 4 ·2HCl was added. The suspension color continued to change from dark green to blue through the slow addition of a concentrated HCl solution (38%, 4 mL). Finally, a clear blue solution of VOCl 2 was formed after heating and stirring for an appropriate period. The concentration of VOCl 2 solution was then adjusted to 0.1 mol/L, and different contents of PVP (0 wt.%, 3 wt.%, 6 wt.%, and 9 wt.%) were added. Accordingly, the film samples prepared from the VOCl 2 -PVP precursors with different PVP contents were labeled as P1 (or 0%), P2 (or 3%), P3 (or 6%), and P4 (or 9%).
All the precursor films were spin-coated on quartz glass (25 mm × 25 mm) at 500 rpm for 9 s, followed by 3000 rpm for 30 s. Subsequently, the wet sol films were dried at 70 • C for 10 min. The coating and drying procedure was repeated once. The annealing procedure was implemented in a closed vacuum tube furnace. The amount of NH 4 HCO 3 deposited in the furnace was 0.1 g. Meanwhile, 1 g of CaO powder was also placed to establish a dry NH 3 atmosphere during heat treatment. The vacuum pump was turned off when the base pressure reached 1800 ± 100 Pa. All film samples were annealed in the furnace at 500 • C with a heating rate of 10 • C/min and then kept at this temperature for 30 min. More details about the preparation of vanadium-containing sol and dry gel film can be found in [27,28].

Characterization
The crystal phase of the thin films was ensured using glance incident X-ray diffraction (GIXRD, Empyrean, Panalytical Inc., Almelo, The Netherlands) at room temperature with a grazing angle of 0.5 • . Raman spectra were measured through a Raman microscope spectrometer (Invia, Renishaw, Wotton-under-Edge, UK) using a laser emission wavelength of 632.8 nm with an output power of 5 mW. According to the reproducibility correction of the 520 cm −1 peak position of monocrystalline silicon, the error was less than 0.03 cm −1 . The micro-morphologies of the film samples were observed by field-emission scanning electron microscopy (FESEM, ULTPAPLUS-43-13, Zeiss, Jena, Germany). The FESEM sample preparation method was applied to cut the VO 2 film-coated quartz glass into small pieces of about 2 × 2 × 1 mm, to glue the back side of the glass (film side up) to the sample table with conductive tape, and to spray carbon before electron microscope observation. The X-ray photoemission spectroscopy (XPS) measurements were recorded using a THERMO PHI Quantum 2000 system. Since V on the surface of the film is easily oxidized to a high valence state, the chemical binding energy of elements inside the sample was studied by sputtering etching with an Ar + ion beam at a depth of about 1 nm/min. The peak positions of elements in the test results were corrected through the 284.6 eV peak of C1s or 530.0 eV peak of O1s using XPSPEAK4.1. Optical performance tests were carried out on a UV/Vis/NIR spectrometer (UV-3600, Shimadzu, Kyoto, Japan) in the wavelength range of 300-2500 nm. Transmittance spectra before and after phase transition were recorded at room temperature (around 20 • C) and heated temperature (set at 90 • C), respectively. The details of the optical performance test and dimming performance evaluation methods of thin films can be found in [27,28].

Ab Initio Calculations
On the basis of the experimental results of XPS, density functional theory (DFT) in the Vienna Ab initio Simulation Package (VASP) [29] was used to ab initio calculate the electronic band structures of various N-containing VO 2 systems. All structures were completely relaxed and highly accurate. The GGA + U method [30] was employed, where the effective Coulomb repulsion potential U eff = U − J was set to 3.4 eV [31,32]. The cutoff energy was set as 600 eV, and a 5 × 5 × 5 Monkhorst-Pack grid of k points was selected for all structures. The total energy convergence criterion for electron self-consistent rings was 1 × 10 −5 eV. The geometry of the system was completely relaxed to a maximum force of less than 0.01 eV/Å.

Phase Determination of VO 2 (M1) Thin Films
The phase composition of vanadium oxide films prepared by mixing different mass fractions of PVP in the precursor was determined from the glance incident XRD patterns and Raman spectra ( Figure 1). As shown in Figure 1a, the broad peaks at 2θ of 15-25 • were attributed to the amorphous quartz glass substrates. There were very weak diffraction peaks corresponding to the VO 2 (M1) phase in sample P1 (0%) and sample P2 (3%) with low PVP content. As the content of PVP was increased to 6-9 wt.%, the peaks corresponding to the VO 2 (M1) phase appeared, although very small amounts of impure phases of V 2 O 5 and vanadium oxide Magnéli phases (V n O 2n−1 ) emerged. The main peaks were indexed to the JCPDS Card No. 82-0661. It is well known that the chemical valence of V is variable due to the coexistence of many nonstoichiometric VO 2 phases [33]. Nevertheless, this issue can be resolved by annealing in an appropriate concentration of NH 3 and air [27,28]. However, the increase in PVP content seems to inhibit the preferred growth orientation of VO 2 crystal on the (011) crystal plane, while diffraction peaks of other crystal planes go higher. The Raman scattering spectra (Figure 1b) of these four samples tested at room temperature were consistent with the XRD results. The tested Raman modes at 142 (B1g), 191 (Ag), 222 (Ag), 304 (Ag), 389 (Ag), 496 (Ag), and 612 (A1g) cm −1 were well marched with the pure VO 2 (M1) phase [16,27,28,34]. With the increase in PVP content, it can be detected that the crystallization of the corresponding VO 2 (M1) phase in the film improved. According to the results above, we can see that PVP is a good film-forming agent, which can form a network structure by complexing metal group VO 2+ through crosslinking polymerization. The increasing content of PVP was helpful to improve the viscosity of the precursor sol. Under the same spinning speed, the VO 2 film became thicker with the increase in the viscosity of precursor sol, resulting in the growth of VO 2 crystals in certain lattice planes.

Microstructures of VO2 (M1) Thin Films
The microstructure and thickness of VO2 films were observed from the top view and

Microstructures of VO 2 (M1) Thin Films
The microstructure and thickness of VO 2 films were observed from the top view and cross-sectional view by scanning electron microscopy (SEM), as shown in Figures 2 and 3, respectively. Figure 2 displays that this group of film samples consisted of crystal sheets or tetragonal prism crystal particles with a large number of pores. Unlike the spherical or tetragonal morphology of the particles in [27,28], the crystals in Figure 2 exhibited short bar or prismatic shapes. All characteristic XRD peaks in Figure 1a were not very strong, indicating that the films were not well crystallized. Accordingly, we can see a disordered arrangement of crystal particles on the surface of the film. This result reveals that the change in the film sample morphology was closely related to the concentration of vanadium ion complexed with polymers in the precursor. The porosity of the films increased when the concentration of vanadium decreased. Due to the addition of PVP, the crystal particles in the films were packed from loose texture to tight arrangement. Crystal boundaries became obvious, consistent with the rise in peak intensity of some feature peaks of VO 2 and V 2 O 5 in Figure 1a. Furthermore, the crystal particle shape and size were more uniform, and the crystal angular morphology became smoother. Thus, a certain content of PVP could regulate the crystal particle morphology and the pore structure of VO 2 films due to the space hindrance. With the increase in PVP content, the crystal particles in the film sample became denser and more closely packed. The pores in the film reduced in number, and the crystal particle size distribution became more uniform. Otherwise, due to the increase in PVP content, the viscosity of the prepared sol increased. As a result, the thickness of the prepared film, as well as the adhesion strength of the film on the substrate, increased. As the film thicknesses could also be controlled by varying the spin-coating parameters, the spinning speed and coating times were fixed. The cross-section morphology of sample P1 prepared without PVP can be observed in Figure 3a, where only a small quantity of particles were attached to the glass substrate. The film formation of this sample was poor, and its adhesion to the substrate was weak in the absence of PVP. Most of the VO2 film was detached from the substrate during sample preparation. As can be seen from the cross-section, the film thickness of the sample increased from a few tens of nanometers (P2) to around 100 nm (P3), and the thickness of P4 increased to approximately 120-200 With the increase in PVP content, the crystal particles in the film sample became denser and more closely packed. The pores in the film reduced in number, and the crystal particle size distribution became more uniform. Otherwise, due to the increase in PVP content, the viscosity of the prepared sol increased. As a result, the thickness of the prepared film, as well as the adhesion strength of the film on the substrate, increased. As the film thicknesses could also be controlled by varying the spin-coating parameters, the spinning speed and coating times were fixed. The cross-section morphology of sample P1 prepared without PVP can be observed in Figure 3a, where only a small quantity of particles were attached to the glass substrate. The film formation of this sample was poor, and its adhesion to the substrate was weak in the absence of PVP. Most of the VO 2 film was detached from the substrate during sample preparation. As can be seen from the cross-section, the film thickness of the sample increased from a few tens of nanometers (P2) to around 100 nm (P3), and the thickness of P4 increased to approximately 120-200 nm ( Figure 3d). However, the edge and surface of the P4 film were rough, which could be ascribed to the increased viscosity of the precursor solution. On the one hand, the homogeneous deposition of sol on the substrate was hindered by the high viscosity; on the other hand, the crystalline process was hindered by the viscous resistance. Therefore, PVP from a high-viscosity precursor reduced the flatness of the VO 2 film.

Chemical Analysis of VO2 (M1) Thin Films
The chemical composition of the vanadium oxide films was determined by XPS testing. To eliminate the effect of surface oxidation and contamination, a survey scan and high-resolution XPS measurements were performed after Ar + ion etching. The XPS fullspectrum scan of the VO2 film sample showed that the film consisted of O, V, C, and N elements. Cl element appeared in the precursor gel volatilized by pyrolysis during the subsequent heat treatment. The chemical valence states of O 1s, V 2p, N 1s, and C 1s elements inside the film surfaces are shown in The ΔBE values (the difference between O 1s and V 2p3/2 core levels) for V 5+ , V 4+ , and V 3+ were 12.8 eV, 14.16 eV, and 14.71 eV, respectively. Accordingly, the oxidation states of vanadium corresponding to V 2p core level energies were V 5+ (517.2 eV), V 4+ (515.84 eV), and V 3+ (515.29 eV) [35][36][37]. The V valence states inside the VO2 thin-film samples were V 4+ , while the V valence states near the film surface increased in valence to between V 4+ and V 5+ . This indicates that the VO2 film exposed to air was gradually oxidized to V2O5 from the surface to the center. The bulges produced on the higher BE shoulder of O 1s around 532 eV were due to the collective effects of satellite [37][38][39]. Si element was derived from the quartz substrate, while H, C, and N

Chemical Analysis of VO 2 (M1) Thin Films
The chemical composition of the vanadium oxide films was determined by XPS testing. To eliminate the effect of surface oxidation and contamination, a survey scan and highresolution XPS measurements were performed after Ar + ion etching. The XPS full-spectrum scan of the VO 2 film sample showed that the film consisted of O, V, C, and N elements. Cl element appeared in the precursor gel volatilized by pyrolysis during the subsequent heat treatment. The chemical valence states of O 1s, V 2p, N 1s, and C 1s elements inside the film surfaces are shown in Accordingly, the oxidation states of vanadium corresponding to V 2p core level energies were V 5+ (517.2 eV), V 4+ (515.84 eV), and V 3+ (515.29 eV) [35][36][37]. The V valence states inside the VO 2 thin-film samples were V 4+ , while the V valence states near the film surface increased in valence to between V 4+ and V 5+ . This indicates that the VO 2 film exposed to Materials 2023, 16, 208 7 of 16 air was gradually oxidized to V 2 O 5 from the surface to the center. The bulges produced on the higher BE shoulder of O 1s around 532 eV were due to the collective effects of satellite peaks of V 4+ 2p 1/2 , V 5+ 2p 3/2 , and V 3+ 2p 1/2 , as well as the influence of O-Si, O-H, O-C, or O-N bonds [37][38][39]. Si element was derived from the quartz substrate, while H, C, and N elements could have resulted from the pyrolysis of PVP and the NH 3 atmosphere. The C 1s curves (Figure 4c,f) and N 1s curves (Figure 4b,e) were calibrated according to the C 1s core level at 284.6 eV. The C 1s peaks of all samples located around 284.6 eV were mainly ascribed to the sp 2 binding of the C-C bond, C-H bond, and C-O bond [40][41][42][43]. The C 1s peak intensity of the film sample weakened with the etching of the Ar + ion. Therefore, these detected C 1s peaks may have been derived from the elemental carbon from the XPS instrument or the absorbed carbonate species from the environment. It is worth noting that the C atom was not doped into the VO2 crystal lattice, since no characteristic peak of the V-C bond (around 282 eV) was detected.
The XPS spectra of N 1s peaks probed at depths of about 0.5 nm and 5.5 nm of VO2 thin-film samples are presented in Figure 4b,e, respectively. All broad N 1s peaks could be resolved into two or three peaks with locations around 401 eV (molecularly adsorbed N2 or NOx) [44], 398.7-398.9 eV (N-H bond [27,45,46], N-O bond [47], or N-C bond) and 396.6 eV (V-N bond [27]). Thus, in addition to the N atoms doped into VO2 crystal lattice in the form of substitution, there were interstitial N atoms that could trap other nonmetallic atoms such as H, O, N, and C into voids of VO2 bulk phase through bonding. Compared with samples fabricated from PVP-containing precursors, the VO2 film sample (P1) prepared without PVP showed that the N source from the NH3 atmosphere mainly existed in the form of interstitial filling and substitutional doping. When PVP was added, adsorbed N species (N2 or NOx molecules) were also detected in all film samples (P2, P3, and P4). Thus, the N introduced by PVP mainly appeared in the form of adsorption in VO2 film, while the N derived from the NH3 atmosphere promoted the substituted N-doping mode and formed V-N bonds as previously verified [27].
The estimates of the concentration of N species in VO2 thin film could be extracted from the XPS data [48]. According to the different N 1s BE locations, N 1s peaks were divided into adsorbed N (Nads) around 401 eV, substituted N (Nsub) close to 396.6 eV, and interstitial N (Nint) around 399 eV. The concentrations of different N species inside VO2 thin-film samples are listed in Table 1. For further intuition, the correlations between the detected amount of nitrogen content in VO2 thin films and the PVP concentration in The C 1s curves (Figure 4c,f) and N 1s curves (Figure 4b,e) were calibrated according to the C 1s core level at 284.6 eV. The C 1s peaks of all samples located around 284.6 eV were mainly ascribed to the sp 2 binding of the C-C bond, C-H bond, and C-O bond [40][41][42][43]. The C 1s peak intensity of the film sample weakened with the etching of the Ar + ion. Therefore, these detected C 1s peaks may have been derived from the elemental carbon from the XPS instrument or the absorbed carbonate species from the environment. It is worth noting that the C atom was not doped into the VO 2 crystal lattice, since no characteristic peak of the V-C bond (around 282 eV) was detected.
The XPS spectra of N 1s peaks probed at depths of about 0.5 nm and 5.5 nm of VO 2 thin-film samples are presented in Figure 4b,e, respectively. All broad N 1s peaks could be resolved into two or three peaks with locations around 401 eV (molecularly adsorbed N 2 or NO x ) [44], 398.7-398.9 eV (N-H bond [27,45,46], N-O bond [47], or N-C bond) and 396.6 eV (V-N bond [27]). Thus, in addition to the N atoms doped into VO 2 crystal lattice in the form of substitution, there were interstitial N atoms that could trap other nonmetallic atoms such as H, O, N, and C into voids of VO 2 bulk phase through bonding. Compared with samples fabricated from PVP-containing precursors, the VO 2 film sample (P1) prepared without PVP showed that the N source from the NH 3 atmosphere mainly existed in the form of interstitial filling and substitutional doping. When PVP was added, adsorbed N species (N 2 or NO x molecules) were also detected in all film samples (P2, P3, and P4). Thus, the N introduced by PVP mainly appeared in the form of adsorption in VO 2 film, while the N derived from the NH 3 atmosphere promoted the substituted N-doping mode and formed V-N bonds as previously verified [27]. The estimates of the concentration of N species in VO 2 thin film could be extracted from the XPS data [48]. According to the different N 1s BE locations, N 1s peaks were divided into adsorbed N (N ads ) around 401 eV, substituted N (N sub ) close to 396.6 eV, and interstitial N (N int ) around 399 eV. The concentrations of different N species inside VO 2 thin-film samples are listed in Table 1. For further intuition, the correlations between the detected amount of nitrogen content in VO 2 thin films and the PVP concentration in vanadium precursor are illustrated in Figure 5.  In summary, it was deduced from the VO2 film fabrication process that an N source could be introduced from pyrrolidone-N of PVP into a vanadium precursor. The N source originating from the NH3 atmosphere during heat treatment played different roles and led to different products. The NH3 atmosphere during heat treatment mainly played a role in promoting the substitution doping of nitrogen for oxygen in VO2. This is consistent with the The contents of the N element and its different species in VO 2 thin films are exhibited in Figure 5a,c. Both graphs show that the total N content and adsorbed N species (N ads ) content gradually increased with the increase in PVP content in the vanadium precursor. As the etching time was prolonged, the probe depth increased, and the detected total N content was enhanced. In particular, because the V-N bond could be produced by Ar + ion sputtering [49], the N sub content (V-N bond) increased with the probe depth, while the variation of N int content was not significant. Meanwhile, the content of N ads in samples containing 3% to 9% PVP slightly decreased with the etching time. Furthermore, it is noticeable that the N ads proportion with the same depth in VO 2 thin film increased as the PVP content increased. Interestingly, this is the opposite of the influence rule of NH 3 concentration on N ads content in samples during heat treatment presented in a previous study [27].
The relative proportion of each N species in total N is illustrated in Figure 5b,d. It can be seen from the graphs that the increase in PVP from 3% to 9% content in the precursor did not affect the relative proportion of the three N species (N ads , N sub, and N int ) in thin-film samples. In the vicinity of the sample surfaces, all the N sub proportions of prepared VO 2 thin films are in the range of 10% to 20% whether the PVP was added or not. In comparison, the N int proportion varied from 40% to 50%, and the N ads proportion varied from 30% to 40% when the PVP was appended to the precursor. However, no N ads species were detected in the sample without the addition of PVP in the precursor. Underneath the film surface, it was found that N ads species, whose proportion was maintained around 20%, only existed in the samples with the addition of PVP in the precursor.
In summary, it was deduced from the VO 2 film fabrication process that an N source could be introduced from pyrrolidone-N of PVP into a vanadium precursor. The N source originating from the NH 3 atmosphere during heat treatment played different roles and led to different products. The NH 3 atmosphere during heat treatment mainly played a role in promoting the substitution doping of nitrogen for oxygen in VO 2 . This is consistent with the conclusion verified in previous work [27]. The pyrrolidone-N may have introduced N-containing groups with N-H or N-C bonds into the vicinity of the surface or void of the film in the form of molecular adsorption or atom (N, O, H, and C) filling. Figure 6 shows the optical transmission and absorption profiles of VO 2 thin films prepared with different PVP contents in vanadium precursors. Table 2 illustrates the luminous transmittance (T lum ), regulation efficiency of solar transmittance (∆T sol ), and modulation efficiency at 2000 nm (∆T 2000 nm ) between ambient temperature (20 • C) and heated state (90 • C). It can be observed that the T lum of the film decreased successively, while the values of ∆T sol , ∆T NIR , and ∆T 2000 nm first increased with PVP concentration from 0 to 6 wt.%, and then decreased when the PVP concentration increased to 9 wt.%. The optical properties of different samples were affected by the phase composition and the thickness of the films, which varied with the content of PVP. Numerous well-crystallized VO 2 (M1) crystal particles in films yielded a good thermochromic modulation effect, corresponding to high ∆T sol and ∆T NIR . However, the thick film was harmful to the transmittance of visible light, corresponding to lower T lum . Meanwhile, the increase in film thickness also led to the enhancement of light absorption by the film samples, as shown in Figure 6b. The optical absorption edges (λ*) [50] of VO 2 thin films differed, resulting in the different colors of the films observed.

Optical Properties of N-Incorporated VO 2 (M1) Thin Films
corresponding to high ΔTsol and ΔTNIR. However, the thick film was harmful to the transmittance of visible light, corresponding to lower Tlum. Meanwhile, the increase in film thickness also led to the enhancement of light absorption by the film samples, as shown in Figure 6b. The optical absorption edges (λ*) [50] of VO2 thin films differed, resulting in the different colors of the films observed. Figure 6. The UV/Vis/NIR transmittance (a) and absorptance (b) spectra of VO2 thin films prepared using different contents of PVP in precursor. The transmittance spectra were measured at 20 °C and 90 °C, respectively; the absorptance spectra were measured at 20 °C.

Band Structure of N-Incorporated VO2 (M1)
The bandgap energy (Eg) of N-incorporated vanadium oxide thin films at room temperature can be estimated from the Tauc formula [8,51,52]: where hν is the photon energy, A is a constant coefficient, and n was chosen to be 1/2 (most authors assume that M1 phase VO2 follows an indirect allowed transition). The optical absorption coefficients α can be obtained according to the following equation: where d is the film thickness, T is the transmission, and R is the reflection. The transmission spectra and reflection spectra corresponding to 300-2500 nm could be measured directly, and the film thickness was averaged according to the observed sample cross-section. According to the Tauc plot results of samples P1 to P4 in Figure 7a, two kinds of different energy bandgaps were sketched. Figure 6. The UV/Vis/NIR transmittance (a) and absorptance (b) spectra of VO 2 thin films prepared using different contents of PVP in precursor. The transmittance spectra were measured at 20 • C and 90 • C, respectively; the absorptance spectra were measured at 20 • C.

Band Structure of N-Incorporated VO 2 (M1)
The bandgap energy (E g ) of N-incorporated vanadium oxide thin films at room temperature can be estimated from the Tauc formula [8,51,52]: where hν is the photon energy, A is a constant coefficient, and n was chosen to be 1/2 (most authors assume that M1 phase VO 2 follows an indirect allowed transition). The optical absorption coefficients α can be obtained according to the following equation: where d is the film thickness, T is the transmission, and R is the reflection. The transmission spectra and reflection spectra corresponding to 300-2500 nm could be measured directly, and the film thickness was averaged according to the observed sample cross-section. According to the Tauc plot results of samples P1 to P4 in Figure 7a, two kinds of different energy bandgaps were sketched. According to octahedral crystal field theory, a bandgap exists between the antibongding π orbitals (π*) band at a higher level and the d band at a lower level in the vicinity of the Fermi energy level (E F ) in the insulating phase VO 2 . This bandgap is called E g1 . Another bandgap that corresponds to the bandgap between bonding π orbitals and π* is E g2 . A schematic of the energy band structure of E g1 and E g2 is distinguished in Figure 7b [53][54][55]. E g1 is related to the MIT phase transition, whereas E g2 correlates with the visible-light transmittance and color. According to octahedral crystal field theory, a bandgap exists between the antibongding π orbitals (π*) band at a higher level and the dǁ band at a lower level in the vicinity of the Fermi energy level (EF) in the insulating phase VO2. This bandgap is called Eg1. Another bandgap that corresponds to the bandgap between bonding π orbitals and π* is Eg2. A schematic of the energy band structure of Eg1 and Eg2 is distinguished in Figure 7b [ [53][54][55]. Eg1 is related to the MIT phase transition, whereas Eg2 correlates with the visiblelight transmittance and color.
For samples P1 and P2, only the energy gap Eg2 appeared with the same value of approximately 1.90 eV. Both Eg1 and Eg2 were found in sample P3 and sample P4, of which Eg2 increased from 1.31 to 1.45 eV. However, Eg1 decreased from 0.26 eV to 0.16 eV with the increase in PVP content from 6 wt.% to 9 wt.%. Combining the changing trend of energy gaps in different PVP contents with the XPS results and optical performance, it could be conjectured that the adsorption of N species (Nads) on the surface and pores of the VO2 film and N species bonded with H, N, O, or C (Nint) may have altered the values of Eg1 and Eg2, thereby regulating the phase transition of VO2 and the film color. However, the hypothesis of N adsorption and filling of elements such as H, N, O, or C inferred from XPS analysis on the surface and in the voids of the VO2 thin film needs further experimental verification.
To reveal the possible fine-tuning effects of N species on optical properties (both τc and color) of N-incorporated VO2 (M1) thin films, DFT calculations were performed using VASP. According to the XPS results of sample P4, the doping ratio of substitutional N atoms, interstitial N atoms, and adsorbent N relative to O atoms were all about 3%. Therefore, the chosen doping amount of each N species in the N-incorporated VO2 (M1) model established was about 3.13% (2.75 wt.%), in which one N atom was substituted for one of 32 O atoms. These calculations took substitutional N-doping, interstitial N-doping, and N-adsorption into account (Figure 8) to be in line with the experimental investigation. On the basis of previous calculation work [27] on pure VO2 (M1) and N-doped VO2 (M1), models of VO2 (M1) doped with 3.13% substitutional N atom, doped with 3.13% interstitial N atom, and bonded with H, N, O, or C atom were used. Nint-N, Nint-H, and Nint-O could also be regarded as the adsorption of N2, NHx, and NOx, respectively. For samples P1 and P2, only the energy gap E g2 appeared with the same value of approximately 1.90 eV. Both E g1 and E g2 were found in sample P3 and sample P4, of which E g2 increased from 1.31 to 1.45 eV. However, E g1 decreased from 0.26 eV to 0.16 eV with the increase in PVP content from 6 wt.% to 9 wt.%. Combining the changing trend of energy gaps in different PVP contents with the XPS results and optical performance, it could be conjectured that the adsorption of N species (N ads ) on the surface and pores of the VO 2 film and N species bonded with H, N, O, or C (N int ) may have altered the values of E g1 and E g2 , thereby regulating the phase transition of VO 2 and the film color. However, the hypothesis of N adsorption and filling of elements such as H, N, O, or C inferred from XPS analysis on the surface and in the voids of the VO 2 thin film needs further experimental verification.
To reveal the possible fine-tuning effects of N species on optical properties (both τ c and color) of N-incorporated VO 2 (M1) thin films, DFT calculations were performed using VASP. According to the XPS results of sample P4, the doping ratio of substitutional N atoms, interstitial N atoms, and adsorbent N relative to O atoms were all about 3%. Therefore, the chosen doping amount of each N species in the N-incorporated VO 2 (M1) model established was about 3.13% (2.75 wt.%), in which one N atom was substituted for one of 32 O atoms. These calculations took substitutional N-doping, interstitial N-doping, and N-adsorption into account (Figure 8) to be in line with the experimental investigation. On the basis of previous calculation work [27] on pure VO 2 (M1) and N-doped VO 2 (M1), models of VO 2 (M1) doped with 3.13% substitutional N atom, doped with 3.13% interstitial N atom, and bonded with H, N, O, or C atom were used. N int -N, N int -H, and N int -O could also be regarded as the adsorption of N 2 , NH x , and NO x , respectively.
The simulated band structures for the N-doped VO 2 (M1) absorbing H, N, O, and C atoms are illustrated in Figure 9. The various energy levels of N-containing VO 2 (M1) calculated from Figure 9 are summarized in Table 3. Concerning previous work [27] for pure VO 2 (M1), E g1 was taken as 0.62 eV, and E g2 was taken as 1.44 eV. Considering the combined effect of N sub and N int with additional atoms (C, H, O, or N), a comparison of calculation results and estimated experimental results is also provided in Table 3. The calculated E g1 values of N-incorporated VO 2 (M1) ranged from 0.1 eV to 0.5 eV, and the values of E g2 ranged from 1.4 eV to 1.6 eV. The theoretical calculation results of E g1 were generally smaller than those of pure VO 2 (M1), indicating that the introduction of N, including doping and adsorption, decreased the phase transition temperature (τ c ) of VO 2 . Furthermore, the existence of N int -H could sharply decrease the E g1 , in good agreement with the absence of E g1 in P1 and P2 samples. Similarly, the appearance of E g1 in P3 and P4 could be attributed to the existence of N-N (N 2 molecule adsorption) and N-O bonds. Unlike the N-H, N-N, and N-O cases in N-incorporated VO 2 (M1), the band structure of the case containing the N-C bond showed metal properties. The slight increase in E g2 could have theoretically caused a decrease in the near-infrared absorption of VO 2 , thereby enhancing the T lum of VO 2 film and diluting the film color. However, in contrast to the theoretical calculation results, the experimental values of E g2 varied from 1.31 eV to 1.90 eV, which could have been mainly affected by the crystal phase composition and film thickness.  Table 3. Concerning previous work [27] for pure VO2 (M1), Eg1 was taken as 0.62 eV, and Eg2 was taken as 1.44 eV. Considering the combined effect of Nsub and Nint with additional atoms (C, H, O, or N), a comparison of calculation results and estimated experimental results is also provided in Table 3. The calculated Eg1 values of N-incorporated VO2 (M1) ranged from 0.1 eV to 0.5 eV, and the values of Eg2 ranged from 1.4 eV to 1.6 eV. The theoretical calculation results of Eg1 were generally smaller than those of pure VO2 (M1), indicating that the introduction of N, including doping and adsorption, decreased the phase transition temperature (τc) of VO2. Furthermore, the existence of Nint-H could sharply decrease the Eg1, in good agreement with the absence of Eg1 in P1 and P2 samples. Similarly, the appearance of Eg1 in P3 and P4 could be attributed to the existence of N-N (N2 molecule adsorption) and N-O bonds. Unlike the N-H, N-N, and N-O cases in N-incorporated VO2 (M1), the band structure of the case containing the N-C bond showed metal properties. The slight increase in Eg2 could have theoretically caused a decrease in the near-infrared absorption of VO2, thereby enhancing the Tlum of VO2 film and diluting the film color. However, in contrast to the theoretical calculation results, the experimental values of Eg2 varied from 1.31 eV to 1.90 eV, which could have been mainly affected by the crystal phase composition and film thickness.

Conclusions
N-incorporated VO2 (M1) thin films were synthesized through a PVP-assisted solgel route followed by heat treatment in vacuum containing a small amount of NH3 gas. The microporous structure, crystallization, film thickness, and optical properties of VO2 (M1) film could be effectively regulated by varying the PVP concentration in the vanadium precursor solution. XPS analysis demonstrated that the N element introduced by the pyrrolidone-N in PVP mainly existed in the form of N2 or NOx molecular adsorption in VO2 film or combined atom (N, O, and H) filling in the vicinity of the surface or void of the film. The subsequent annealing in NH3 was attributed to the generation of substitutional N-V bonds in the VO2 crystals. As the PVP concentration increased, Tlum of the VO2 thin film decreased due to the increase in film thickness, while ΔTsol and ΔTNIR first increased and then decreased. The best results of thermochromic film in this work showed a Tlum of 52.19-56.79%, a ΔTsol of 4.99%, a ΔTNIR of 7.39%, and a ΔT2000 nm of 37.6%. Both the Tauc plot estimation and the theoretical calculation suggested that the combined effect of substitutional N atom, interstitial N-X species (X = H, N, or O), and adsorbed Ncontaining molecules decreased the τc due to the reduction in Eg1. According to the calculated value of Eg2, the introduction of elements such as N, H, or O had little effect on Figure 9. The calculated band structures of N-incorporated VO 2 (M1) according to density functional theory (DFT) calculations. All calculations were based on the corresponding models in Figure 8.

Conclusions
N-incorporated VO 2 (M1) thin films were synthesized through a PVP-assisted sol-gel route followed by heat treatment in vacuum containing a small amount of NH 3 gas. The microporous structure, crystallization, film thickness, and optical properties of VO 2 (M1) film could be effectively regulated by varying the PVP concentration in the vanadium precursor solution. XPS analysis demonstrated that the N element introduced by the pyrrolidone-N in PVP mainly existed in the form of N 2 or NO x molecular adsorption in VO 2 film or combined atom (N, O, and H) filling in the vicinity of the surface or void of the film. The subsequent annealing in NH 3 was attributed to the generation of substitutional N-V bonds in the VO 2 crystals. As the PVP concentration increased, T lum of the VO 2 thin film decreased due to the increase in film thickness, while ∆T sol and ∆T NIR first increased and then decreased. The best results of thermochromic film in this work showed a T lum of 52.19-56.79%, a ∆T sol of 4.99%, a ∆T NIR of 7.39%, and a ∆T 2000 nm of 37.6%. Both the Tauc plot estimation and the theoretical calculation suggested that the combined effect of substitutional N atom, interstitial N-X species (X = H, N, or O), and adsorbed N-containing molecules decreased the τ c due to the reduction in Eg 1 . According to the calculated value of Eg 2 , the introduction of elements such as N, H, or O had little effect on the bandgap between π and π* in the insulating VO 2 (M1) phase. Actually, the film thickness and the micro-morphology had significant impacts on the visible-light transmittance and color of the films. This work presents an approach to adjust the optical properties of VO 2 (M1) film, which is expected to be applied to smart windows or other VO 2 -based devices.  Acknowledgments: The authors thank the Analytical and Testing Center of WUT for the help with carrying out GIXRD, Raman, and FESEM analyses, as well as the State Key Laboratory of Silicate Materials for Architectures for the help with performing the UV/Vis/NIR test.

Conflicts of Interest:
The authors declare no conflict of interest. bandgap energy related to metal-to-insulator transition E g2 : bandgap energy related to the visible transparency GIXRD:

Abbreviations
glance incident X-ray diffraction XPS: X-ray photoemission spectroscopy FESEM: field-emission scanning electron microscopy UV/Vis/NIR: ultraviolet/visible/near-infrared light DFT: density functional theory VASP: Vienna Ab initio Simulation Package N ads : adsorbed N N sub : substituted N N int : interstitial N