Study of the Thermal Annealing on Structural and Morphological Properties of High-Porosity A-WO3 Films Synthesized by HFCVD

High-porosity nanostructured amorphous tungsten OXIDE (a-WO3) films were synthesized by a Hot Filament Chemical Vapor Deposition technique (HFCVD) and then transformed into a crystalline WO3 by simple thermal annealing. The a-WO3 films were annealed at 100, 300, and 500 °C for 10 min in an air environment. The films were characterized by scanning electron microscopy (SEM), X-ray diffraction (XRD), micro-Raman spectroscopy, high-resolution transmission electron microscopy (HR-TEM), and UV–vis spectroscopy. Results revealed that the a-WO3 films were highly porous, composed of cauliflower-like structures made of nanoparticles with average sizes of 12 nm. It was shown that the effect of annealing on the morphology of the a-WO3 films leads to a sintering process. However, the morphology is conserved. It was found that at annealing temperatures of 100 °C, the a-WO3 films are of an amorphous nature, while at 300 °C, the films crystallize in the monoclinic phase of WO3. The calculated bandgap for the a-WO3 was 3.09 eV, and 2.53 eV for the film annealed at 500 °C. Finally, the results show that porous WO3 films preserve the morphology and maintain the porosity, even after the annealing at 500 °C.


Introduction
One of the main characteristics that nanomaterials offer is the larger specific surface area (SSA) value than that in their bulk shape. This characteristic is very important because it has opened a wide range of applications like: Adsorptive removal and photocatalytic degradation of pollutants; detection of toxic, flammable, or explosive gases; and electrochromic devices [1][2][3][4][5].
For this reason, the exploration of synthesis techniques has taken the special attention of many authors, for example: Wang et al. fabricated dispersed WO 3 nanoparticles by the solvothermal method, and they reported the porous nanostructure with a SSA of 11.57 to 18.92 m 2 /gr [6]; Jain et al. synthesized porous tungsten oxide (PWO) by Reactive Spray Deposition Technology, and they calculated that the SSA equals 46 m 2 /gr [7]. Other methods to synthesize PWO films are sparking [8], sputtering [9], and templating [10]. Some of these techniques require sophisticated equipment or complicated After synthesis, each substrate was cleaved into 4 pieces (each of 25 mm 2 ), and three of them were annealed in an open room for 10 min at different temperatures: 100, 300, and 500 °C, respectively. The experimental conditions are summarized in Table 1. The samples were analyzed using a Scanning Electron Microscope (TESCAN Vega TS513SB, with an acceleration voltage of 10 kV) to explore morphology and thickness of porous tungsten oxide films. An X-Ray Diffractometer (X'PERT PRO MRD of PANalytical) was used to obtain the structural information, carrying the measurements out with a step size [2θ] of 0.01° and scan step time of 48.19 s. A Micro-Raman Spectrometer (Horiba Jobin Yvon HR800) was used to explore the vibrational modes in the tungsten oxide films. It used a laser with a nominal power of 10.7 mW, wavelength of 633 nm, and 5 s of exposition time to avoid damaging the sample. A Transmission Electron Microscope (JEOL JEM-ARM200CF) was used to obtain the particle size distribution (PSD) and to explore the change from amorphous to the monoclinic phase of tungsten oxide. Diffuse reflectance UV-vis spectroscopy (DRUV-vis) was employed to calculate the bandgap of PWO films. Spectra were recorded from 200 to 800 using a Cary 100 UV-visible spectrophotometer. XRD, Raman and TEM equipments are located at the Centro de Nanociencias y Micro-Nanotecnologías of the IPN, After synthesis, each substrate was cleaved into 4 pieces (each of 25 mm 2 ), and three of them were annealed in an open room for 10 min at different temperatures: 100, 300, and 500 • C, respectively. The experimental conditions are summarized in Table 1. Table 1. Deposition time and annealing temperature for the samples.

Sample Deposition Time (Min)
Annealing Temperature (10 Min) The samples were analyzed using a Scanning Electron Microscope (TESCAN Vega TS513SB, with an acceleration voltage of 10 kV) to explore morphology and thickness of porous tungsten oxide films. An X-Ray Diffractometer (X'PERT PRO MRD of PANalytical) was used to obtain the structural information, carrying the measurements out with a step size [2θ] of 0.01 • and scan step time of 48.19 s. A Micro-Raman Spectrometer (Horiba Jobin Yvon HR800) was used to explore the vibrational modes in the tungsten oxide films. It used a laser with a nominal power of 10.7 mW, wavelength of 633 nm, and 5 s of exposition time to avoid damaging the sample. A Transmission Electron Microscope (JEOL JEM-ARM200CF) was used to obtain the particle size distribution (PSD) and to explore the change from amorphous to the monoclinic phase of tungsten oxide. Diffuse reflectance UV-vis spectroscopy (DRUV-vis) was employed to calculate the bandgap of PWO films. Spectra were recorded from 200 to 800 using a Cary 100 UV-visible spectrophotometer. XRD, Raman and TEM equipments are located at the Centro de Nanociencias y Micro-Nanotecnologías of the IPN, México. SEM and UV-vis equipments are located at the Benemérita Universidad Autónoma de Puebla, México.

Morphology
To study the effect of the annealing process on the porous films, first, the thickness of each sample was measured (column a: s5m, column b: s10m, and column c: s15m). The cross-sectional views are shown in Figure 2. From this figure, it can be observed that the thickness of the films depends on the growth time, i.e., the longer the deposition time, the thicker the film, corresponding to 6.9, 7.2, and 8.1 µm for 5, 10, and 15 min of deposition, respectively (row 1; columns a, b, c). In a previous work, it was reported that film thickness could grow up to 30 µm after 30 min of deposition, as well as that the thickness depends on the substrate temperature [41]. In Figure 2, it can be observed that the transversal morphology of the films is highly porous, composed of nanoparticles arranged in columnar structures. Additionally, it is observed that near the substrate, nanoparticles are randomly distributed, whereas near the film surface, nanoparticles are compacted, forming cauliflower-like structures. The thickness dependence on the deposition time can be attributed to a high growth ratio in the first minutes of the process, so the tungsten filament after 5 min should be almost oxidized. SEM images of Figure 3 show the surface morphology of sample s10m, before and after the annealing. It can be seen that the film exhibits a highly porous surface formed by the nanoparticle agglomerations. The nanoparticle diameter increases as the annealing temperature does, starting at 12 nm as it is shown in the inset of Figure 3a for the as-synthesized sample. For the sample annealed at 100 °C, the size of nanoparticles was around 18 nm (Figure 3b). Samples annealed at 300 and 500   Figure 2 show the cross-sectioned samples annealed at 100, 300, and 500 • C, respectively. In all the cases, a reduction in thickness of the films was observed as the annealing temperature increased. From Figure 2, it can be observed that when samples were annealed at 500 • C, their respective thickness was reduced by 40% in comparison with the as-synthesized one. This can be attributed to a sintering process promoted by the thermal annealing [42]. Furthermore, it is observed that samples annealed at 500 • C exhibit mainly columnar structures as opposed to randomly distributed nanoparticles. However, the surface morphology seems to be conserved; thus, sample s10m was characterized with scanning electron microscopy (SEM) to explore it.
SEM images of Figure 3 show the surface morphology of sample s10m, before and after the annealing. It can be seen that the film exhibits a highly porous surface formed by the nanoparticle agglomerations. The nanoparticle diameter increases as the annealing temperature does, starting at 12 nm as it is shown in the inset of Figure 3a for the as-synthesized sample. For the sample annealed at 100 • C, the size of nanoparticles was around 18 nm ( Figure 3b). Samples annealed at 300 and 500 • C exhibited particles of 21 and 50 nm, respectively (see insets of Figure 3c,d).
Agglomerated particles give place to cauliflower-like structures with diameters from 0.5 to 1.0 µm. These structures are separated from each other, forming cracks that were broader as annealing temperature increased.
The above results are in good agreement with those reported by Chávez et al., who, measuring through a Small-Angle X-ray Scattering technique (SAXS), observed that annealing WO 3 powders under similar experimental conditions increased both the cauliflower-like agglomerations and the crystal size [43].   Figure 4 shows the X-ray diffraction patterns in the range of 20 to 60° for samples described in Table 1. The diffraction patterns indicate that the films maintain their amorphous phase for annealing temperatures below 300 °C, and above this temperature, the films begin to crystallize, completing their transformation to monoclinic tungsten oxide phase at 500 °C. It can also be seen that the diffraction peaks are more intense as the annealing temperature increases, being sharper in the thickest sample (Figure 4c).  Figure 4 shows the X-ray diffraction patterns in the range of 20 to 60 • for samples described in Table 1. The diffraction patterns indicate that the films maintain their amorphous phase for annealing temperatures below 300 • C, and above this temperature, the films begin to crystallize, completing their transformation to monoclinic tungsten oxide phase at 500 • C. It can also be seen that the diffraction peaks are more intense as the annealing temperature increases, being sharper in the thickest sample ( Figure 4c).  Samples annealed at 300 °C and 500 °C exhibit three main peaks at 23.1, 23.7, and 24.3°, corresponding to the (002), (020), and (200) planes, respectively, of the γ-WO3 monoclinic phase P21/n (JCPDS 83-0950 card). As it was expected, the thermal annealing improves the crystal quality of the tungsten oxide films. Tn this case, the crystallization process is because of the reduction of structural defects from 300 °C.

Crystalline Structure
Both the deposition time of the a-WO3 films and the annealing process play an important role in the final crystalline quality of the porous film. However, although the film thicknesses seem to be of the same order of magnitude (e.g., 2.5 to 3.5 µm for samples annealed at 500 °C), the thickest film showed better crystalline quality than the thinnest film, indicating that (1) the intensity of the peaks depends on the thickness of the films, and (2) thicker films are composed of bigger crystallites. A similar behavior has been reported elsewhere for NiO thin films [44]. Samples annealed at 300 • C and 500 • C exhibit three main peaks at 23.1, 23.7, and 24.3 • , corresponding to the (002), (020), and (200) planes, respectively, of the γ-WO 3 monoclinic phase P21/n (JCPDS 83-0950 card). As it was expected, the thermal annealing improves the crystal quality of the tungsten oxide films. Tn this case, the crystallization process is because of the reduction of structural defects from 300 • C.
Both the deposition time of the a-WO 3 films and the annealing process play an important role in the final crystalline quality of the porous film. However, although the film thicknesses seem to be of the same order of magnitude (e.g., 2.5 to 3.5 µm for samples annealed at 500 • C), the thickest film showed better crystalline quality than the thinnest film, indicating that (1) the intensity of the peaks depends on the thickness of the films, and (2) thicker films are composed of bigger crystallites. A similar behavior has been reported elsewhere for NiO thin films [44].

Specific Surface Area
The average crystallite size was calculated from X-ray diffraction patterns using the Debye-Scherrer formula, given by d = (kλ/β hkl cosθ), where d is the diameter of the crystallite, λ(CuKα) represents the wavelength of the radiation (1.54051 Å), β is the peak width of the diffraction peak profile at half maximum, k is the Scherrer constant (=0.9, assuming spherical particles), and θ is the Bragg's angle. A common way to evaluate the effective grain size as well as the degree of crystallinity of the monoclinic WO 3 layers is by evaluating the triplet peaks in the range 20 < 2θ < 30 [45][46][47]. Notice that the annealing process conduces to an increment in the crystallite size, due to the high temperature, providing nanoparticles enough energy to coalesce into larger particles promoted by the atomic diffusion [48].
As mentioned above, the specific surface area (SSA) is a crucial parameter to evaluating the potential of the porous tungsten oxide films in the field of heterogeneous catalysis, adsorption processes, and gas sensing applications. Based on the X-ray diffraction (XRD) results, the SSA was calculated for each annealing condition by the following formula [49]: where SA part represents surface area of the particle, V part represents the volume of the particle (considering spherical approximation for the crystallite size), and using the value of density for WO 3 equals 7.16 g/cm 3 . The calculated crystallite sizes and SSA for samples s5m, s10m, and s15m are summarized in Table 2. The SSA calculated for the sample s10m at 300 • C was 64.59 m 2 /gr, and at 500 • C was reduced to 29.25 m 2 /gr ( Table 2). In fact, this trend was observed for each sample. It shows that when the annealing temperature increases, the SSA decreases slightly. The SSA values reported here are comparable with earlier results reported elsewhere (see Table 3). Based on the experimental observation, we consider that the oxidation of tungsten filament is caused by the residual water, according to the following reaction [43]: The tungsten filament is heated at high temperature (~1800 • C) in an air vacuum atmosphere (~10 −3 Torr). The tungsten oxide is formed at the filament and readily desorbed due to its high vapor pressure. The desorbed molecules are collected in the substrate surface as small nanoparticles, which imitate the cauliflower-like structures. Figure 5 shows the micro-Raman spectra for the as-synthesized and annealed samples in the range of 100 to 1000 cm −1 . The signal around 521 cm −1 corresponds to the silicon substrate. The Raman spectra for both the as-synthesized samples and the samples annealed at 100 • C do not show any peaks, indicating that the porous films are amorphous, which is in accordance with the XRD results. These characteristic lines in Raman spectra have been reported earlier in [53], where authors observed Raman peaks only when samples were annealed above 400 • C.

Micro-Raman Analysis
For samples annealed at 300 and 500 • C, the spectra show two broad bands: A low-frequency band in the range of 100-300 cm −1 , associated with the O-W-O bending modes, and a high-frequency band in the range of 700-900 cm −1 , associated with the W-O stretching modes. Such bands can be associated with the monoclinic phase of WO 3 [54] . The peak located at 134 cm -1 is assigned to the lattice modes of tungsten oxide [55]. The peak around 272 cm −1 corresponds to bending vibration modes of the δ (O-W-O) bonds. The more intense peaks, observed around 714 cm −1 and 807 cm −1 , correspond to the stretching modes of the ν (W-O) bonds [56].
Raman spectra of samples annealed at 500 • C show slightly narrower and more intense peaks than those annealed at 300 • C, indicating that the crystallinity of the films increases gradually with the increase of the annealing temperature. The peak at 807 cm −1 observed for monoclinic tungsten oxide does not change as a function of annealing temperature, indicating the obtaining of highly stable crystalline porous films. A mix of γ and δ phases is discarded from this analysis mainly because it has been reported earlier that it occurs at sintering temperatures from 600 to 1000 • C [60].
The slight shift towards lower wavenumbers of peaks at 272 and 719 cm −1 could be due to the growth occurring in a non-equilibrium process, so the synthesized films are obtained with high stress (compressive or tensile) mainly because of the gradient temperature between filament and substrate. This shift can also be attributed to the increment in the particle size [53,61,62]. Figure 5 shows the micro-Raman spectra for the as-synthesized and annealed samples in the range of 100 to 1000 cm −1 . The signal around 521 cm −1 corresponds to the silicon substrate. The Raman spectra for both the as-synthesized samples and the samples annealed at 100 °C do not show any peaks, indicating that the porous films are amorphous, which is in accordance with the XRD results. These characteristic lines in Raman spectra have been reported earlier in [53], where authors observed Raman peaks only when samples were annealed above 400 °C.

Transmition Electron Microscopy (TEM) Analysis
Each sample for TEM analysis was prepared as follows: (1) The film (over silicon substrate) was immersed in 10 mL of propanol; (2) solution was sonicated for 15 min to detach the nanoparticles from the substrate; (3) with a micropipette, a drop of the solution was taken; and (4) the drop was deposited onto a TEM grid. Figure 6 shows the TEM images, as well as the particle size distribution (PSD), for the s10m sample before and after the annealing process. The PSD was determined by using an image-processing program known as ImageJ. From the TEM images, it can be observed that the higher the annealing temperature is, the bigger the nanoparticle size, due to a sintering process promoted by the thermal annealing, which is also the reason for the film's thickness reduction.
The graph of Figure 6a shows the PSD for the as-synthesized sample with sizes mainly from 4 to 7 nm. For the sample annealed at 100 • C, the PSD increased to 10 nm (graph of Figure 6b). Finally, the samples annealed at 300 and 500 • C presented particle sizes mostly from 12 to 22 nm and from 30 to 50 nm, respectively (graphs of Figure 6c,d). It is important to note that as the annealing temperature increases, so does the PSD, being broader (20 to 80 nm) when the sample was annealed at 500 • C.
(PSD), for the s10m sample before and after the annealing process. The PSD was determined by using an image-processing program known as ImageJ. From the TEM images, it can be observed that the higher the annealing temperature is, the bigger the nanoparticle size, due to a sintering process promoted by the thermal annealing, which is also the reason for the film's thickness reduction.
The graph of Figure 6a shows the PSD for the as-synthesized sample with sizes mainly from 4 to 7 nm. For the sample annealed at 100 °C, the PSD increased to 10 nm (graph of Figure 6b). Finally, the samples annealed at 300 and 500 °C presented particle sizes mostly from 12 to 22 nm and from 30 to 50 nm, respectively (graphs of Figure 6c,d). It is important to note that as the annealing temperature increases, so does the PSD, being broader (20 to 80 nm) when the sample was annealed at 500 °C.  HR-TEM images (Figure 7) show that there is no evidence of crystalline domains for the sample annealed at 100 • C, nor for the as-synthesized one, which indicates the amorphous nature of the films. On the other hand, samples annealed at 300 and 500 • C exhibit crystalline nanoparticles of 15 and 30 nanometers, respectively, with interplanar distance in both cases of 0.38 nanometers along the direction (002) of the monoclinic tungsten oxide, according to the JCPDS 83-0950 card. These results are in good agreement with those previously determined with the Scherrer's formula, summarized in Table 2.   Figure 8a shows the diffuse reflectance UV-vis spectra (DRUV-vis) recorded from 200 to 800 nm for s10m and as-synthesized samples. The black and red lines correspond to the as-synthesized samples annealed at 500 °C, respectively. The spectra revealed that the annealing increases the absorption border from 320 to 360 nm and increases the reflectance above 600 nm (see Figure 8a).  Figure 8a shows the diffuse reflectance UV-vis spectra (DRUV-vis) recorded from 200 to 800 nm for s10m and as-synthesized samples. The black and red lines correspond to the as-synthesized samples annealed at 500 • C, respectively. The spectra revealed that the annealing increases the absorption border from 320 to 360 nm and increases the reflectance above 600 nm (see Figure 8a). Figure 8a shows the diffuse reflectance UV-vis spectra (DRUV-vis) recorded from 200 to 800 nm for s10m and as-synthesized samples. The black and red lines correspond to the as-synthesized samples annealed at 500 °C, respectively. The spectra revealed that the annealing increases the absorption border from 320 to 360 nm and increases the reflectance above 600 nm (see Figure 8a). The optical bandgap (Eg) of both samples was extracted using the Kubelka-Munk theory, based on the following formula [63]:

UV-Vis Diffuse Reflectance and Bandgap
It is assumed in the above equation that the reflectance (R) and scattering coefficients (S) of the samples are constant for the entire wavelength range; thus, the corresponding absorption coefficients are related to the diffuse reflectance spectra, and then, it is possible to extract the band gap of the samples using the Tauc et al. method by the following equation [64]: The optical bandgap (Eg) of both samples was extracted using the Kubelka-Munk theory, based on the following formula [63]: It is assumed in the above equation that the reflectance (R) and scattering coefficients (S) of the samples are constant for the entire wavelength range; thus, the corresponding absorption coefficients are related to the diffuse reflectance spectra, and then, it is possible to extract the band gap of the samples using the Tauc et al. method by the following equation [64]: Therefore, Eg can be determined by extrapolation of the linear region from the plot [F(R)hv]1/n versus photon energy (hv). For tungsten oxide, the nature of the electron transition from the valence band to the conduction band is indirect, so we use n = 2.
The optical bandgap values obtained for the as-synthesized and annealed samples corresponded to 3.09 and 2.53 eV, respectively (Figure 8b), which are in accordance with results reported for amorphous and crystalline tungsten oxide [65,66]. The decrease of the optical bandgap upon increasing the annealing temperature is the result of the recrystallization process in the film [67].

Conclusions
Amorphous tungsten oxide (a-WO 3 ) films were deposited on silicon substrates with the HFCVD technique. The a-WO 3 films were annealed to obtain crystalline WO 3 films and, during this process, a film thickness reduction of around 40% was observed, which was a consequence of the sintering process. However, both the surface morphology and the bulk porosity of the films were conserved. It can be concluded that the growth time influences the film thickness in the first minutes of the process, giving place to a heterogeneous transversal morphology-as the thickness film increased (Z-axis), the nanoparticles were more packed in cauliflower-like structures, which was observed even after the annealing process. The heterogenous transversal morphology can explain the Raman signals observed in the thinnest film, where evidence of sub-oxides was detected. Because of the annealing treatment, the crystallite sizes were increased, and that provoked a slight decrease in the specific surface area; however, it does not affect the porosity nature of the films along the surface, nor along the Z-axis. In general, it is well known that HFCVD can be used to deposit metal-oxide nanostructures whenever the oxide formed on the filament has a partial pressure higher than the metal. Following that, this technique offers the possibility to probe other metals to obtain highly porous nanostructures. Moreover, the HFCVD reactor could be slightly modified by adding filaments to cover a bigger area, which would make it a scalable technique. From the presented results, it is evident that through this simple approach, it is feasible to obtain highly porous nanostructured tungsten oxide films, suitable for applications in gas sensors or photocatalysis.