Solution Processing of Nanoceramic VO2 Thin Films for Application to Smart Windows

Yanfeng Gao1,2,3,, Litao Kang1, Zhang Chen1 and Hongjie Luo2 1State Key Laboratory of High Performance Ceramics and Superfine Microstructure, Shanghai institute of Ceramics, Chinese Academy of Sciences, Shanghai, 2Research Center for Industrial Ceramics, Shanghai institute of Ceramics, Chinese Academy of Sciences, Shanghai, 3Xinjiang Key Laboratory of Electronic Information Materials and Devices, Xinjiang Technical Institute of Physics & Chemistry, Chinese Academy of Sciences, Beijing Road, Urumqi, Xinjiang, China


Introduction
Energy conservation has directed a global trend towards sustainable development. Due to global warming, air conditioning systems have been widely used in daily life, thus inducing a series of problems, 1 including increases in electricity consumption and carbon dioxide emissions along with the formation of other atmospheric pollutants from the electricitygeneration process. Air conditioning in China accounts for 40-60% of a building's energy consumption (the exact figure depends on the area of the building), and overall, uses 28% of the total available primary energy. These figures will grow rapidly with urban development, as the case of China. One effective way to reduce the amount of electricity consumed by cooling is to apply solar-control coatings to glass windows, or so-called "energy-efficient windows" or "smart windows". Because lighting demands transparency, most of the smart windows are designed to intelligently control the amount of light and heat (mainly in the near infrared region) passing through in response to an external stimulus such as light (photochromic), heat (thermochromic) or electricity (electrochromic). [2][3][4][5][6][7] In this regard, the thermochromic smart window, typically based on a vanadium dioxide (VO 2 ) functional layer, has received particular interests due to two aspects. First, it can respond to environmental temperatures, making reversible structural changes from an infraredtransparent semiconductive crystalline phase to an infrared-blocking metallic crystalline phase. Second, the visible transparency remains almost unchangeable. VO 2 , which undergoes a metal-insulator transition (MIT) at a critical temperature T c (68 C for bulk VO 2 ), 8 has attracted much attention as a thermochromic material for smart windows. 9 Owing to the MIT, VO 2 transforms between the monoclinic (P2 1 /c, M 1 ) and the tetragonal (P4 2 /mnm, R) phase, inducing a severe change in optical properties. VO 2 is transparent to near infrared (NIR) light at temperatures below T c , but NIR-light reflective above T c . The MIT is simply determined by environmental temperature and occurs fast, to a difference in the solar energy transmittance across metal-insulator transition (MIT) and is used to characterize the thermochromic properties of VO 2 films. This value is usually below 10% for a single layered VO 2 film. [4][5] In the past several years, we have worked to develop a solution-based process for VO 2based thin films with a special emphasis on their preparation, thermochromic property study and application to smart windows. 43,44,[60][61][62][63][64][65][66][67][68][69][70][71][72] Polymer-assisted deposition (PAD) process was realized for the preparation of VO 2 and VO 2 -based multilayered films. The method enables us to facially control over the film thickness, morphology and optical constants. By combining clever control of the optical parameters and/or their thickness and microstructural regulation, we obtained VO 2 films with high visible transmittance (40-84%), controllable Mott phase transition temperatures and high switching efficiencies (max. 15.1%). The results show that the current solution process is a powerful competitor towards practical applications of this material.

The method
The process was started with vanadium oxides or inorganic salts, as shown in Figure 1. These raw materials were treated to make an aqueous transparent solution. To the above solution the selected, weighed soluble polymers and doping agents were added. Then the precursor VO 2 film was prepared by traditional solution methods, such as spinning coating or dip coating. After drying at 80 C in air, the precursor film was annealed at 300-600 C in N 2 . Readers can read references for details. 43,44 It is also found that precursor solution without polyvinylpyrrolidone (PVP) was unstable and a large amount of precipitates formed after aging for several days. Whereas, PVPemploying solution stayed stable for several months with only a little precipitates suspending in the solution, indicating that PVP improved stability of the precursor solution, probably because the negatively charged carbonyl groups bound with aqua vanadium ions to form a relatively stable precursor solution. To examine the interactions between PVP and aqua vanadium ions, Fourier transform infrared spectroscopy (FTIR) was employed to characterize the precursor solution with or without PVP. The results were shown in Figure 2. To compare, FTIR spectra of PVP ( Figure  2a) and PVP in the presence of H 2 SO 4 ( Figure 2b) were also included. Generally, the strong and sharp peak at 1654 cm -1 was assigned to the stretching vibration of -C=O, and the peak at 1292 cm -1 was attributed to -C-N ( Figure 2a). 73 When H 2 SO 4 was added to PVP ( Figure  2b), the stretching vibration of -C=O shifted to low wavenumber of 1635 cm -1 , which originates from the loosening of the -C=O double bond by coordinating between negatively charged carbonyl groups and H + . 74 For precursor solution with PVP (Figure 2c), the frequency of -C=O stretching vibration became lower (1620 cm -1 ) due to the influence of aqua vanadium ions. The coordination interactions between the carboxyl group of PVP and metal ions (Li, Ca, Co, Ag) were also reported in PVP-DMF-MCl n systems 75 as well as polymer/silver salt complex membranes. [76][77][78] Although the vibration at 1624 cm -1 in the spectrum for a PVP-free precursor solution was poorly identified (Figure 2d), the vibration of -C-N kept almost unchanged (Figure 2a, b and c) for all these films, indicating that there were no interactions between amine groups and aqua vanadium ions. According to above discussion, the effects of PVP on stabilizing precursor solution were due to the interactions of the negatively charged carbonyl groups in PVP with qua vanadium ions. Furthermore, it is reported that the interactions between metal ions and the carbonyl groups from different PVP molecules increased the apparent viscosity of the PVP-DMF-MCl n solution (M = Li, Ca, Co), 75 where metal ions act as cross-linking points between different PVP molecular chains, 75 improving the film formability.    The simultaneous TG, DTA, DTG and evolved gas analytical curves of the gel contain K90 PVP in a nitrogen atmosphere are presented in Figure 3 and Figure 4 (the dot line). On the basis of MS (mass spectrum) signals, the evolution of H 2 O and CO 2 was detected ( Figure 3).

www.intechopen.com
At temperatures up to 250 °C, adsorbed and chemisorbed H 2 O was released, indicating by MS and DTG changes centered at 185 -190 °C (a continuous weight loss from 100 to 250 °C in TG) along with an endothermic peak in DTA. After the evaporation of water, there is a plat stage in both MS and TG curves from 250 to 300 °C. At 300 -450 °C, both H 2 O and CO 2 were released, as observed distinct peaks in MS curves at 345 °C, implying the oxidization of PVP with a trace amount of oxygen in the nitrogen flow. CO is formed when oxygen is insufficient, although the MS signals for CO were failed to be detected due to the similar molecular weight of CO and N 2 . This conclusion also can be drawn from the DTA/DTG results, in which an endothermic peak appeared at 360°C in the DTA curve (a corresponding shoulder at 371°C in the DTG curve). In the range of 450 -650 °C, no obvious reactions occurred. Above 650 °C, a gradual weight loss in the TG curve was observed, implying that the residual carbon (which had been confirmed by Raman analysis) can further reduce VO 2 to V 2 O 3 (from XRD results). Further DTG analysis on precursor gels containing PVP of different molecular weights reveals that the decomposition sequence is related to the molecular weight of the polymer. As show in Figure 4, both DTG peaks for water desorption and polymer decomposition shifted to lower temperatures for the low molecular weight PVP (K30 vs K90). However, in both cases, heating temperatures ≥ 450 °C are needed to decompose the polymer. Therefore, the thermal analyses suggest that the appropriate annealing temperature for crystallization to thermochromic VO 2 should be in the range of 450 -650 °C. Subsequently, a series of annealing experiments were performed and the results are summarized in Figure 5. It was shown that at 450 °C a relatively longer annealing time was www.intechopen.com required to crystallize VO 2 with thermochromic properties. At 500 °C and above, the annealing time could be significantly shortened, and holding time was even unnecessary. These results are coincident with those of thermal analyses, and are very profitable for scaleup mass manufacturing. Interestingly, when K30 PVP was replaced by K90 PVP, the annealing time at 500 °C should be prolonged to exceeding 20 min to achieve the thermochromic properties. This result shows that the degradation process of polymers effectively influences the formation of thermochromic VO 2 , probably due to interactions between PVP and VO 2+ at the atomic scale. This conclusion supports the film-forming mechanism that we proposed in the previous work. 43a Furthermore, these interactions ensure the formation of homogeneous hybrid precursor films after solvent evaporation. 43 Thereby, the morphologies of final films could be easily controlled via adjusting the degradation rate of polymers. Figure 6 (a) shows the TEM image for precursor gel. PVP aggregates, inorganic clusters and their aggregates were clearly identified. Using this precursor solution, a homogeneous film was obtained, as show in Figure 6 (b). After crystallization, both PVP-free and PVPemploying films were particulate, porous, while the particle size of PVP-employing films (~100 nm in diameter) was much smaller than PVP-free ones (over ~100 nm) that showed a broad size distribution. The porosity of PVP-free films was obviously larger than PVPemploying films, typically about 63.5 % (Figure 7a   A schematic illustration of film-forming mechanism was given in Figure 8. The interactions among polymer molecules via the oppositely charged groups along with that between the carbonyl groups and the metal ions ensured the formation of cross-linked high-quality gel films after the solvent evaporation. In addition, steric entangling of the polymer chains assists in the enhancement of cross linking.   Figure 9 shows that the VO 2 film prepared by polymer-assisted deposition process had a relatively pure crystalline phase. For the film obtained by a PVP-free solution, only limited peak were detected, which were assigned to the characteristic peak of M-phase VO 2 (JCPDS Card No. 72-0514, P21/c, a = 0.5743 nm, b = 0.4517 nm, c = 0.5375, β = 122.61°), but two weak diffraction peaks at 9.36 ° and 12.24° were also detected, which were too skimpy to be determinately identified. For the film prepared by the PAD process, only M-phase VO 2 characteristic peaks (011) was detected, which suggested that the film was in a preferential orientation. The formation of M-phase VO 2 was further confirmed by Raman spectra. At an output power of 1 mW ( Figure 10, curve a) for a pure film, an almost complete set of Raman bands of M-phase VO 2 were observed. The bands agree well with references 9, 79-80 for M phase VO 2 , with centers at 192, 222, 261, 308, 337, 391, 440, 497 and 615 and 816 cm -1 . There is at least one corresponding band for every Raman band of our film and the positions of the bands are agreed well with each other, meaning that every Raman bands we collected here can be assigned to M-phase VO 2 . A Raman spectrum for a 1 at % W-doped VO 2 film ( Figure 10, curve b) at 1 mW showed only weak bands, suggesting that the film was at the phase transition point due to the reduction effect of the W-doping on the transition temperature. No Raman bands appeared at output power of 20 mW for both of the films ( Figure 10, curves c and d), indicating that complete phase transition from M-phase VO 2 to R-phase VO 2 occurred under this output power condition. The contribution of a fused quartz substrate has also been included ( Figure 10, curve e) for the sake of comparison. Although an overlap of Raman bands of fused quartz and WO 3 at 807 cm -1 makes it difficult to determine the presence of WO 3 from the enlarged Raman spectra of W-doped films, other strong Raman bands located at 716, 275 cm -1 were not observed, excluding the absence of WO 3 . Meanwhile, no obvious Raman bands for impure phases were found after the phase transition, indicating that there were no obvious impure vanadium oxides.   The 1 at % W-doped VO 2 film exhibits a hysteresis loop centered at 54 °C with a width of 16 °C (Figure 11b), implying a decrease in phase transition temperature of about 15 °C. This decreasing efficiency is less than other reports. 36,38 Significantly, the decreasing efficiency of phase transition temperature for 1 at % W-doped film was lower than the average efficiencies of other doping doses in references, which employed similar doping process to www.intechopen.com prepare VO 2 particles, 81 probably because the tungsten ion was not completely incorporated into the final VO 2 films. W4f 5/2 and W4f 7/2 XPS peaks with a binding energy of 37.7 and 35.3 eV, respectively, are clearly seen in Figure 11b inset for the W-doped thin films and the tungsten ion in these films is W 6+ according to the standard binding energy. The width narrowing of the hysteresis loop can be explained by the martensitic transformation model for VO 2 . 82 W-doping increases the density of structural defects, which is a power function of the driving force, and relatively reduces the activation energy of the coordinated jumps of V cations (phase transitions take place at a certain defect density for size-fixed crystal grains 58 ). Activation energy further influences the widths of hysteresis loops via degrees of supercooling or superheating. This deduction of martensitic transformation model is in agreement with various experiment results. 25,39,83

Optimized thermochromic hysteresis properties of single-layered films
Thermochromic properties are important parameters for the practical application of this material. These parameters include phase transition temperature, thermochromic hysteresis and energy-saving efficiency. Figure 12 shows a schematic description of phase transition temperature and thermochromic hysteresis. From the transmittance (Tr) -temperature (T) data, a plot of d(Tr)/d(T) -T is obtained, yielding one or two peaks with well-defined maxima (see Figure 12). Each of the d(Tr)/d(T) -T curves has been fitted with a Gaussian function using the peak fitting module of Originpro 7.5 software. The temperature corresponding to the maximum d(Tr)/d(T) is defined as the phase transition temperature (T c ) of the branch (T c,h and T c,c represent T c of heating and cooling branches, respectively). For cooling branches, the appearance of steps introduces two peaks in the d(Tr)/d(T) -T curves. The T c values of these branches are determined by the main peaks. The slope of the transition is expressed by the full width at half maximum (FWHM) of the peak. The width of the hysteresis loop, ΔT c , is defined as the temperature difference of T c,h -T c,c . Figure 13 shows typical SEM photos of VO 2 films heated at 600 °C for different times with a heating rate of 30 °C·min -1 . All the films are porous, consisting of irregular particles. The pore formation is attributed to the degradation of PVP and the shrinkage of the gel film during annealing. Grain boundaries change from clear (Figure 13a, b, c) to fuzzy (inset in Figure 13d) as the annealing time is prolonged. Meanwhile, the particle size varies dramatically, associated with distinct change of porosity and pore shape. To evaluate the differences of particle size among samples, the distributions of particle sizes was measured. The corresponding size distribution scatter graph as well as fitting curves with Gaussian distribution is shown in Figure 14. The scatter graphs were plotted by randomly measuring the dimensions of 300 grains in a given micrograph. The particle size distribution scatter graphs of Figure 13d failed to be obtained due to fuzzy grain boundaries. The results showed that the grain size of Sample I complied with Gaussian distribution, while the www.intechopen.com distributions for Samples II and III distinctly exhibited two maxima. The size distribution scatter graphs indicated that when annealing time increases from 5 to 20 min, the particle size increases rapidly (Figure 13a, b and Figure 14a, b). However, the VO 2 particles shrink noticeably for the long time annealing samples (Figure 13c and Figure 14c, d), resulting in few large particles and a broad distribution of particle sizes. The shrinkage of particle size (which is also observed in the 500 °C annealed films) and change of grain boundaries are tentatively attributed to the mass transport via surface diffusion during annealing. This mass transport has been reported as the main reason for morphology evolution of VO 2 films during the heat treatment of pre-deposited amorphous ones at 450 °C. 83 Fig. 14. Scatter graphs showing the particle size distributions in Sample I (a), Sample II (b) and Sample III (c). The solid curves are those after fitting by Gaussian distribution. Samples were obtained by annealing at 600 °C for different times, 5 min (a), 20 min (b) and 60 min (c) with heating rate of 30 °C·min -1 . 44 Significantly, obvious steps were observed at the cooling branches for Samples I, II and III (insets of Figure 15a, b, c, annealing for 5, 20 and 60 min, respectively). The appearance of steps suggests another loop width. The steps are resulted from a difference in temperature of inhomogeneous occurrence of phase transition in the films due to the two-humped grain size distributions (Figure 14b, c) 84 or the site-selective nucleation of product phase, as observed in literature. 85 However, there are also some contradictions. First, this deduction is difficult to explain the step appearing in Sample I, which manifests a Gaussian distribution www.intechopen.com with one maximum in grain size. Second, this deduction suggests that the temperatures of the d(Tr)/d(T) -T peaks should vary as the size distribution changes. For Sample II and III, however, although the size distribution curves are quite different (Figure 14b, c), the step appears at similar temperatures on the cooling branch of the d(Tr)/d(T) -T curves. Therefore, it seems that these experiment results don't agree to the model suggested by Klimov, V. A et al.. 84 In fact, the propagation of the phase transition through grain boundaries may possibly counteract inhomogeneous distribution of T c in the VO 2 films. Therefore, there is usually no step observed in relatively narrow hysteresis loops. [86][87] It is believed that any models related to the steps in the hysteresis loops should take the influence of grain boundary into account. For Sample IV (annealed for 180 min), the step seems to be depressed (inset of Figure 15d). This result could be explained by the coalescence of grain boundaries and the improvement of phase transition propagation through grain boundaries, which would push the step upward. The appearance of the step in only one of the two branches is associated with the asymmetry of the elementary hysteresis loops (i.e., the loops assigned to individual grains) with respect to the phase equilibrium temperature. 86 The asymmetry of the elementary hysteresis loops, in turn, is attributed to a transition temperature shift accompanying with a decrease in grain size, [88][89] or with stress at the substrate/film interface. 86,90 To further clarify the effects of grain size on the widths of hysteresis loops, films with similar morphologies and relatively homogeneous size distribution are needed. The current solution process enables us to achieve widely morphology control. In previous research, we have revealed that the precursor solution is in a state of solution rather than sol. After the solvent evaporation, the interactions among polymer molecules, along with those between the carbonyl groups and the metal ions, ensure the formation of cross-linked high quality gel films. 43a In these films, the metal ions are bonded with the polymer through electrostatic interactions, forming a uniform organic-inorganic hybrid precursor film. The formation of metal oxide occurs after degradation of polymer begins. 43, 91 Therefore, the morphologies of final films could be easily controlled via adjusting the degradation rate of polymers. Meanwhile, for crystallization in a solid-state reaction, the dependence of nucleation and growth rate on temperature is usually different. 92 Thus, it is expected that the morphologies of VO 2 films could be tailored by variations of heat treatments. Accordingly, the films have been prepared under the same synthesis conditions as Samples I -IV, but at an annealing temperature of 500 °C. And the typical SEM photos and optical transmittance spectra, along with the hysteresis loops, are shown in Figure 16. The sample obtained by annealing at 500 °C for 5 min showed a bluish color, but almost no thermochromic properties, and is excluded. Films produced at this annealing temperature have similar morphologies (Figure 16a, b, c). Compared to films obtained by annealing at 600 °C, in which many quasi-isolated grains appear, all of these films show grains tightly connected across boundaries. The grain sizes are smaller than those treated at 600 °C. The film obtained by annealing for 20 min (Sample V) consists of connected, irregular particles ( Figure 16a). Typically, the largest dimension of each particle is around 100 nm. These particles possess flat surfaces, indicating that the mass transport during annealing is feeble for a short annealing time. The film produced by annealing for 60 min (Sample VI) shows similar granular morphologies with a relatively large roughness, and the particle size is reduced notably (Figure 16b). Prolonging the annealing time to 180 min (Sample VII) results in increases in porosity and surface roughness (Figure 16c), indicating the enhancement in mass transport that changes the surface morphology of these particles. 83,92 To further verify our discussion on grain boundaries, two films (Sample I and Sample VIII) were scraped off from their substrates and observed by TEM (Figure 17). Figure 17a, b and c present TEM photos of Sample I. It is shown that the sample is assembled from irregular particles. Distinct grain boundaries are observed in high resolution TEM photos (Figure 17a  (600 °C, 5 min). Generally, spaces about 2-5 nm in width were observed between grains, and they were attributed to poorly crystallized phases and/or voids. It is widely reported that VO 2 films with second or amorphous phases at grain boundaries demonstrate a broadened hysteresis, 82,93,94 which is definitely consistent with our experiments. Sample VIII, on the other hand, consisted of smooth connected particles and interconnected pores among particles (Figure 17d, e). This morphology should be a result of mass transport and aggregation of the primary VO 2 crystals during annealing. 83 The high magnification TEM photo clearly demonstrates that the grain boundaries are very sharp (Figure 17f). In fact, it seems that the lattice fringes pass through grain boundaries. The FFT patterns ( Figure  17f insets) of the corresponding crystal grains are composed of two groups of bright diffraction spots, confirming the existence of boundaries and the high crystallinity of these crystals. The TEM photos are fully consistent with the SEM results (Figure 13a   To evaluate the optical and thermochromic properties of the samples, the integral transmittances of films annealed with heating rate of 30 °C·min -1 were calculated in the visible region (380 -780 nm), so were the integral infrared transmittance reductions before and after the S-M phase transition in the wavelength of 1500 -2500 nm. The results are shown in Figure 18. According to the theory suggested by J. Narayan and V. M. Bhosle, 57 the amplitude of the transition should deteriorate as the defect content increases. It is expected www.intechopen.com that the 500 °C annealed samples should be less crystallized and thus contain high defect content. Accordingly, they should exhibit deteriorated infrared transmittance reduction compared with the 600 °C annealed samples. This phenomenon was not obviously detected in our films, even though the hysteresis widths of the films vary considerably. In fact, it is clearly revealed that the 500 °C annealed samples show better visible transmittance than those annealed at 600 °C, while the infrared transmittance reduction remains comparable. For instance, the infrared transmittance reduction of sample VI (53.8%, 500 °C, 60 min annealed sample) is a little larger than that of sample II (51.2%, 600 °C, 20 min annealed sample). Meanwhile, the integral visible transmittance of sample VI (34.5%) is much higher than that of sample II (20.2%). The difference in the microstructure and crystallinity had effects on the optical properties of these films. Hemispherical reflectance spectra of Sample II and VI were collected from 240 to 2600 nm ( Figure 19). It was found that the integral hemispherical reflectance of Sample VI (18.7%) was 3.5% larger than that of Sample II (15.2%) in the visible region. Because both of the integral transmittance and integral reflectance for Sample II are smaller than those of Sample VI in the visible region, one can conclude that the absorption and/or scattering of Sample II are much larger. The hemispherical transmittance of the sample was recorded, in order to eliminating the influence of scattering. It was still shown that both the hemispherical transmittance and the hemispherical reflectance of Sample II are less than those of Sample VI in the visible region, indicating that the absorption of Sample II is larger than that of Sample VI.

Optimized visible transmittance of single-layered films
For the practical application of VO 2 -based smart windows, a low luminous transmittance (T lum ) and solar modulating ability (∆T sol ) are two major drawbacks. The T lum and T sol values were obtained from ,, ,  99 This study further predicts that a limit for noticeable solar energy modulation is T lum = 40%, and ∆T sol ≤ 10%. 99 We have recently confirmed experimentally the effects of composite matrix. 68 VO 2 -ZrV 2 O 7 composite films with similar thickness of about 95 nm showed decreased particle sizes and significantly enhanced luminous transmittances (from 32.3% at Zr/V=0 to 53.4% at Zr/V=0.12) with increasing Zr/V rations. However, the influence of porosity on optical properties of single-layered VO 2 films, which should have priority over the stack structure of VO 2 -based multilayers, is seldom reported. Pores with air in VO 2 films can be considered as a secondary component that should have similar effects on improving T lum and ∆T sol . In this study, calculations confirm that optical constants and film thickness have important effects on the thermochromic properties of these films. An optical-admittance recursive method was used to simulate the spectral transmittance using the optical constants of VO 2 and a fused-silica-glass substrate. The optical constant is assumed to be linearly dependent on the volume fraction. 100 Figure 20 shows the computed luminous transmittance and solar modulation of VO 2 films with different thicknesses and porosities. The results indicate that ∆T sol could be increased without decreasing T lum by increasing the porosity, which is derived from the depression of the reflection. To validate the above prediction, we prepared VO 2 films by PAD. X-ray diffraction and Raman spectra confirmed the formation of a monoclinic (M) phase with a trace amount of V 2 O 5 . Figure 21a shows a top-down SEM image of VO 2 films, which reveals that the sample consisted of interconnected VO 2 particles and irregular nano-pores. The size of particle and pore ranges from 20-70 and 15-80 nm with a mean value of 38 and 28 nm, respectively. The feature size of the film is well below the wavelength of visible and infrared light, favoring the improvement of optical quality. The nano-porous feature of the films is observable (inset of Figure 21a), which is also supported by the low n and k values of ellipsometry results compared with those in the literature (Figure 21b). 46 For example, the value of n is 2.2 for our VO 2 film, which is around 3 in the literature. 46 The result also shows that T lum in the current research is much higher (by ≥12%) than that in the literature with comparable ∆T sol due to the different optical constants (Figure 21b). 46 These results are attributed to the degradation of PVP and the shrinkage of the gel film during annealing. 43a  Figure 22 shows the transmittance and reflectance spectra of typical VO 2 films. The MIT transition is clearly observed as a dramatic infrared-transmittance change with temperature ( Figure 22a). T lum reduces steadily with increasing film thickness, which is ascribed to the strong absorption of VO 2 in this region. 44,99 The change in the infrared transmittance of VO 2 films at 90 °C with different film thicknesses shows a similar behavior. Nevertheless, the infrared transmittance at 20 °C for a 428-nm-thick film is obviously higher than that of a 215nm-thick film after 1700 nm (Figure 22a). These changes in the transmittance spectra correspond to reflectance valleys in Figure 22b, suggesting the existence of a selfantireflection effect in these thicknesses due to the destructive interference of light reflected from the film-substrate and the air-film interfaces. The change of the optical constants of VO 2 across the MIT can effectively modulate the infrared transmittance and shift the position of the reflectance valley at 20 °C, leading to a significant enhancement of the IR modulating ability at a certain wavelength. This phenomenon could be harnessed to boost the thermochromic properties of a single-layer film in selected wavelength ranges. For instance, a 215-nm film shows a transmittance change of 50% (from 61.1% to 11.1%) at 1350 nm across the MIT, the highest value at this wavelength to our knowledge. Furthermore, the enhancement of the IR modulating ability could be adjusted to longer wavelengths by a simple regulation of the film thickness. A 428nm-thick VO 2 film exhibits a IR transmittance change of 76.5% (from 83% to 6.5%) at 2500 nm, even prior to the optimized result of sputtering films (74%). 19 Another interesting phenomenon is that the changes in luminous transmittance (∆T lum ) across the MIT are thickness dependent. For thin films, the visible transmittance at 20 °C is generally lower than that at 90 °C ( Figure 22 and Table 1) and, vice versa. The visible transmittance at 20 °C for the above 100-nm-thick films exceeds that at 90 °C ( Figure 22b and Table 1). This reversion in ∆T lum is ascribed to interference effects and was also reported by Xu et al. 46 In view of the fact that solar energy is mainly in the visible region with a peak at 550 nm, the ∆T lum reversion across the MIT effectively influences ∆T sol . For instance, ∆T sol increased by only 0.6% (from 6.4% to 7.0%) as the film thickness increased from 43 nm to 77

Optimized optical properties of VO 2 -based double or multi-layered films
The current techniques used to improve visible transparency mainly include Mg doping, 95 formation of mixtures (VO 2 /SiO 2 ), 97 regulation of the thickness of VO 2 films 63 and deposition of antireflective layers. 35,52,96 Among these techniques, VO 2 -based multi-layered structures containing antireflection layers show better optical performance, especially a www.intechopen.com balance between luminous transmittance (T lum ) and switching efficiency (ΔT sol ). Moreover, the antireflective layer can protect VO 2 from oxidation and add new functions such as photocatalysis. 101,102 ΔT lum represents the improvement of T lum after antireflecion. An integrated improvement of 23 % (from 32 to 55 %) in T lum can be achieved for VO 2 /ZrO 2 double layers using ZrO 2 as an antireflective layer. 52 A TiO 2 /VO 2 /TiO 2 three layer shows ΔT sol =2.9 % and ΔT lum = 27% (increased from 31 to 58%). 35 A TiO 2 /VO 2 /TiO 2 /VO 2 /TiO 2 five layers can improve ΔT sol to 12.1% (6.7 % for the single VO 2 film). 98 All of these films were prepared by gas-phase deposition. And, it seems difficult to improve ΔT lum and ΔT sol simultaneously at a higher ΔT lum level for VO 2 films. The ΔT sol of the TiO 2 /VO 2 /TiO 2 three layer decreased from 3.9 to 2.9 %, while the T lum of the simulation, followed by all-solution preparation and investigation of the improvement in T lum and ΔT sol to confirm the computational predictions. By adding a quarter-waved optical thickness TiO 2 film on VO 2 , the integrated luminous reflectance (R lum ) of VO 2 was reduced dramatically from 31.2 % to 3.0 %, and T lum is close to that of the TiO 2 /VO 2 /TiO 2 layer films. 102 These results are comparable to the films prepared by gas-phase deposition, 4, 102 but ΔT sol is slightly higher (6.9 % vs. 6.0 %). In addition, methods to improve ΔT sol while maintaining a high ΔT lum were explored; the highest ΔT sol was 15.1 % for optimised doublelayered films, which still showed T lum = 49.5 % at 20 °C and 44.8 % at 90 °C. The simulation results in Figure 23 show that the changing trend of wavelength-dependent transmittance agrees well with the experimental results, and are also similar to the results in the reference. 98 These results suggest that although veracity is not high enough by using optical constants of gas-phase-derived VO 2 films, the simulation still can give a trend prediction on the optical properties.
To improve T lum , the value and position of reflection minima are two key factors. The simulation results show that VO 2 can be simplified as a RI-fixed dielectric film. For a double- Fig. 24. Transmittance (at 20 and 90 °C) and reflectance (20 °C) spectra before (blue, dotted lines) and after (red, solid lines) adding SiO 2 films (a, c: transmittance and reflectance spectra for sample B2, using SiO 2 as antireflection layer) or TiO 2 films (b, d: transmittance and reflectance spectra for sample B1, using TiO 2 as antireflection layer) on VO 2 films. Insets (a) and (b) are hysteresis loops. 71 layered film containing two dielectric layers, reflectance minima can be achieved by adjusting the thickness and RI. 103 Figure 24 shows the transmittance and reflectance spectra of VO 2 /SiO 2 (B2) and VO 2 /TiO 2 (B1) double-layered films. Hysteresis loops at 2000 nm before and after coating show that the antireflective coatings have little effects on the thermochromism of VO 2 films. TiO 2 films had a better antireflection efficiency than SiO 2 because the RI of TiO 2 is closer to an ideal value for VO 2 . 104 For the simplified VO 2 -based double-layered structure, the ideal n t should be 2.14 to achieve a reflection minimum, n u (2.6) is the RI of VO 2 in the middle of the visible wavelength range measured by ellipsometry, and n s (1.45) is the RI of quartz glass. Theoretical simulation results by optical admittance recursive method showed the minimum reflection at 20 °C appearing at n t =2.10 due to the influence of the extinction coefficient, whereas the minimum R lum (2.8 %, representing luminous reflection) appears at n t =2.04. For TiO 2 , n t =1.94, the minimum reflectance is 0.5 %, and R lum is 3.0 %; for SiO 2 , n t =1.43, the minimum reflectance is 8.4 %, and R lum is 13.6 %. Experiments results show a reflectance minimum of sample B2 is 7.9 %, and antireflection peaks appeared at 610 nm (20 °C) and 560 nm (90 °C infrared (780-1100 nm) regions. Visible transmittance is almost constant across MIT of VO 2 films, so the transmittance difference in the short-wave near-infrared region makes a major contribution to ΔT sol . One attempt to improve the performance difference at short-wave near-infrared wavelengths is to shift the reflectance minima toward longer wavelengths by regulation of the thickness of antireflective layers. ΔT lum is caused by the different RI of VO 2 films at 20 and 90 °C, which influences the value and position of reflectance minima. The half-quarterwaved structure has two reflection minima with one minimum at longer wavelengths, which can improve ΔT sol . The VO 2 film (sample A, 95nm) with a 73nm TiO 2 antireflective layer gave two antireflection peaks at 20 °C at about 475 nm and 685 nm ( Figure 25) and led to an increase in ΔT sol of 2.0 %. ΔT sol could be further improved by shifting antireflection peaks to longer wavelengths, but this treatment sacrifices ΔT lum (reflective minimum departure from the middle of visible wavelength range).
To simultaneously increase T lum and ΔT sol , one way is to design an antireflective layer that can form two antireflection peaks both in the visible wavelength range and the short-wave near-infrared wavelength range. For this purpose, a 3-times quarter-waved thick TiO 2 (210 nm) film was prepared. Based on the simplified model (using the optical constants of VO 2 at 560 nm at 20 °C) two antireflection peaks (correspondingly, two transmittance maxima) should appear at 1680 nm and 560 nm at 20 °C. Experimentally, because the optical constants of VO 2 films are wavelength-dependent and existence of extinction coefficient, the first antireflection peak shifted to 1250 nm (20 °C) and 2400 nm (90 °C) ( Figure 26). However, the second antireflection peak fell in the middle of the visible range (570 and 540 nm for 20 and 90 °C, respectively), and the transmittance difference in the short-wave nearinfrared is enlarged due to different ΔT lum at 20 and 90 °C. High ΔT lum values close to those of the quarter-waved samples (17.3 and 11.8 % at 20 and 90 °C) and an improvement of 3.7 % (7.2 to 10.9 %) for ΔT sol were achieved. Fig. 26. Transmittance spectra at 20 °C (1, 2) and 90 °C (1', 2') of VO 2 film (sample B3) after adding a 3 times quarter-wave thick TiO 2 (3λ/4) film (1 and 1'). 71 This work reveals an impressive improvement on the visible transmittance and switching efficiency, and is an important technical breakthrough toward the practical application of VO 2 -based smart windows. For a double-layered system, luminous transmittance and switch efficiency could be greatly improved by regulation of the RI and thickness of the films. Single quarter-waved TiO 2 films on VO 2 could reduce R lum from 31.2 to 4.2 %, improving T lum up to 21.2 % at 20 °C (from 40.3 to 61.5 %). Sample with T lum of 78.1 % was further obtained by optimising the thickness of a VO 2 film and adding an antireflective coating, which still had ΔT sol of 7.5 %. The highest ΔT sol achieved was 15.1 %. Films with balanced luminous transmittance (T lum =58.0 and 53.9 % at 20 and 90 °C, respectively) and switching efficiency (10.9 %) were prepared.

Summary remarks and outlook
This chapter introduces a solution method, polymer assisted deposition process, for the preparation of VO 2 and VO 2 -based multilayered films. The method enables us to facially control over the film thickness, morphology and optical constants. By combining the optical design in respect to the materials selection and/or their thickness and microstructural control, we obtained VO 2 films with high visible transmittance (40-84%), controllable Mott phase transition temperatures and high switching efficiencies (max. 15.1%). The results show that the current solution process is a powerful competitor towards practical applications of this material.