Modifying Precursor Solutions to Obtain Screen-Printable Inks for Tungsten Oxides Electrochromic Film Preparation

: Tungsten trioxide (WO 3 ) is used to prepare the important electrochromic layer of the electrochromic device as a wide bandgap semiconductor material. In this study, WO 3 electrochromic ﬁlm was successfully prepared by screen printing. To modify the thixotropy and wettability of the ink, polyvinyl alcohol (PVA) and 2-perﬂuoroalkyl ethanol (FSO) were added in the ammonium meta-tungstate (AMT) solution. We found that the PVA additive could improve the dynamic viscosity of the solution and modify the uniformity of the ﬁlm. 2-Perﬂuoroalkyl ethanol (FSO) could lower the surface tension and increase the wettability of the AMT solution on the substrate. By observing the morphology of the printed ﬁlms, the ink formulas for screen printing were selected. We found the annealing process could help remove PVA. Through characterization of electrochromic performance, it was found that the best performing device had 42.57% modulation and 93.25 cm 2 · C − 1 coloration efﬁciency (CE) for 600 nm light. This study showed great potential in the preparation of WO 3 electrochromic devices by a low-cost screen-printing method.


Introduction
Nowadays, electrochromism has had applications in many fields such as energysaving smart windows [1][2][3][4][5], batteries [6], sensors [7], next-generation displays and wearable electronics [8]. Tungsten trioxide (WO 3 ) was the most widely studied and used electrochromic material. Not only does the raw material have high cost-performance, but also WO 3 film has excellent electrochromic properties such as high optical contrast and low energy consumption [9]. There are many methods to prepare electrochromic films such as magnetron sputtering [10][11][12], ion plating and sol-gel [13][14][15][16]. Recently, there have been studies on the electrochromic film preparation by the hydrothermal method [17,18]. So far, as the mainstream industrial preparation method, magnetron sputtering requires complex equipment and process, which makes it hard to reduce the cost and further industrialize. To solve the cost problem, screen printing and inkjet printing have become the research emphasis of electrochromic film preparation because of their high cost-performance, simple process and low environmental requirements. Up to now, screen printing has been used in many fields. It is commonly used to prepare film in thermoelectric devices [19,20], flexible transparent electrode [21,22], organic semiconductor device [23]
The mesh count of the printing screen was 450 cm −2 . The distance between the screen and printing substrate was about 4 mm. The precursor solution was printed on a clean ITO substrate (2 cm × 2 cm) through the pattern (1 cm × 1 cm) by screen printing. The film was printed twice with a single solution volume of 50 µL and dried naturally for 30 s after screen printing.
It has been reported that the residues of PVA after thermal decomposition at 500 • C was about 5% [31] and decomposition of AMT at no lower than 500 • C vaporized most NH 3 and H 2 O, producing WO 3 [32,33]. Considering the thermal deformation of ITO glass, the films were annealed at 500 • C for an hour. The annealed films were packaged using the electrolyte solution (mixture of 50% volume fraction for both 1 mol/L propylene carbonate solution of LiClO 4 and UV curing adhesive, the type of UV-curing adhesive was Valigoo8500). The electrolyte solution was solidified under UV light (395 nm) irradiation to prevent it from outflowing from the device.
The surface tension and the contact angle of AMT solution and the precursor solutions were measured by Attension Theta (Biolin Scientific, TL200, Gothenburg, Sweden). The dynamic viscosity of the AMT solution and the precursor solutions was measured using a rotational rheometer (Thermo Fisher Scientific, HAAKE MARS 40, Waltham, MA, USA). The morphology of the printed films was measured by the optical microscope (MSHOT, MJ30, Guangzhou, China). Fourier transform infrared spectroscopy (FT-IR) spectra of the unannealed and annealed films were recorded by an FT-IR spectrophotometer (SHI- MADZU IR Prestige-21, SHIMADZU, Tokyo, Japan) with an ITO glass substrate acting as a blank. The crystals of the unannealed and annealed films were distinguished by X-ray diffraction using Cu Kα radiation (XRD, PANalytucal Empyrean DY1577, PANalytical, Almelo, The Netherlands), and the XRD patterns were analyzed using the software Jade 6.0. The thickness of all the annealed films were measured by a probe surface profiler (Veeco Dektak 150, Veeco, NY, USA). The transmittance of the unannealed and annealed films were measured by an ultraviolet spectrophotometer (SHIMADZE UV2600, SHIMADZU, Tokyo, Japan), with air acting as a blank. The current of the electrochromic test and the relationship between the change of transmittance and time were recorded by an electrochromic workstation (CH Instruments CHI 660E, CH Instruments, Shanghai, China) and a micro-spectrometer (Morpho PG2000, Morpho, Shanghai, China) respectively.

Results and Discussion
3.1. Effect of Polyvinyl Alcohol (PVA) and 2-Perfluoroalkyl Ethanol (FSO) on Precursor Solutions and Film Formation Figure 1 shows the dynamic viscosity of all the precursor solutions. The measured dynamic viscosity of AMT solution was 1.08 mPa·s. It is shown in Figure 1 that the dynamic viscosities of the precursor solutions without FSO were all higher than that of the AMT solution, which illustrated that the addition of PVA could increase the dynamic viscosity. The dynamic viscosity of the precursor solutions without FSO had a non-linear rise from 1.70 to 103.27 mPa·s with the increase in the mass fraction of PVA. Some studies indicated that the critical micelle concentration (CMC) of PVA-related material was in the range of 0.125-4.47 mg/L [34,35], so it could be speculated that for all the precursor solutions, PVA macromolecules had intermolecular forces and formed hydrogen bonds with each other because the concentration exceeded its CMC. Therefore, the reason for the non-linear rise of the dynamic viscosity was due to more intermolecular forces and hydrogen bonds between PVA macromolecules. It was found in Figure 1 that the dynamic viscosity of the precursor solutions with FSO increased from 2.16 to 163.61 mPa·s with the increase in the mass fraction of PVA. The increasing tendency in the dynamic viscosity of the precursor solutions with FSO was similar to that of the precursor solutions without FSO, which illustrated that the dynamic viscosity was mainly affected by PVA. Figure 1 also showed that at the same PVA mass fraction, the dynamic viscosity of the precursor solutions with FSO was higher than that of the precursor solutions without FSO, proving that the addition of FSO could increase the dynamic viscosity. A study suggested that the increase of the dynamic viscosity might be caused by hydrogen bonds formed between the surfactant and polymer [36]. USA). The morphology of the printed films was measured by the optical microscope (MSHOT, MJ30, Guangzhou, China). Fourier transform infrared spectroscopy (FT-IR) spectra of the unannealed and annealed films were recorded by an FT-IR spectrophotometer (SHIMADZU IR Prestige-21, SHIMADZU, Tokyo, Japan) with an ITO glass substrate acting as a blank. The crystals of the unannealed and annealed films were distinguished by X-ray diffraction using Cu Kα radiation (XRD, PANalytucal Empyrean DY1577, PANalytical, Almelo, The Netherlands), and the XRD patterns were analyzed using the software Jade 6.0. The thickness of all the annealed films were measured by a probe surface profiler (Veeco Dektak 150, Veeco, NY, USA). The transmittance of the unannealed and annealed films were measured by an ultraviolet spectrophotometer (SHIMADZE UV2600, SHIMADZU, Tokyo, Japan), with air acting as a blank. The current of the electrochromic test and the relationship between the change of transmittance and time were recorded by an electrochromic workstation (CH Instruments CHI 660E, CH Instruments, Shanghai, China) and a micro-spectrometer (Morpho PG2000, Morpho, Shanghai, China) respectively. Figure 1 shows the dynamic viscosity of all the precursor solutions. The measured dynamic viscosity of AMT solution was 1.08 mPa·s. It is shown in Figure 1 that the dynamic viscosities of the precursor solutions without FSO were all higher than that of the AMT solution, which illustrated that the addition of PVA could increase the dynamic viscosity. The dynamic viscosity of the precursor solutions without FSO had a non-linear rise from 1.70 to 103.27 mPa·s with the increase in the mass fraction of PVA. Some studies indicated that the critical micelle concentration (CMC) of PVA-related material was in the range of 0.125-4.47 mg/L [34,35], so it could be speculated that for all the precursor solutions, PVA macromolecules had intermolecular forces and formed hydrogen bonds with each other because the concentration exceeded its CMC. Therefore, the reason for the nonlinear rise of the dynamic viscosity was due to more intermolecular forces and hydrogen bonds between PVA macromolecules. It was found in Figure 1 that the dynamic viscosity of the precursor solutions with FSO increased from 2.16 to 163.61 mPa·s with the increase in the mass fraction of PVA. The increasing tendency in the dynamic viscosity of the precursor solutions with FSO was similar to that of the precursor solutions without FSO, which illustrated that the dynamic viscosity was mainly affected by PVA. Figure 1 also showed that at the same PVA mass fraction, the dynamic viscosity of the precursor solutions with FSO was higher than that of the precursor solutions without FSO, proving that the addition of FSO could increase the dynamic viscosity. A study suggested that the increase of the dynamic viscosity might be caused by hydrogen bonds formed between the surfactant and polymer [36].   liquid surface because the concentration exceeded its CMC. The measured surface tension of AMT solution was 58.14 mN/m, and the surface tensions of the precursor solutions without FSO were all lower than those of AMT solution, proving that PVA had surface activity. In the precursor solution, the FSO mass fraction increased with the PVA mass fraction increase because the ratio of FSO to PVA was constant. In general, because of high surface activity, FSO always tends to gather at the liquid surface, and the surface tension will decrease with the increase of its mass fraction. However, in Figure 2a, it can be seen that the surface tension of the precursor solutions with FSO increased first, and then decreased while obtaining the maximum at 3.8 wt.% PVA. A possible explanation was that hydrogen bonds were formed between PVA and FSO at first, resulting in the surface tension increase. After hydrogen bonds reached saturation in the liquid surface, the increase of FSO mass fraction would decrease the surface tension. The surface tension of the precursor solutions without FSO was all below 23.00 mN/m while the surface tension of the precursor solutions with FSO was all over 51.00 mN/m, which illustrated that the addition of FSO could greatly reduce the surface tension of the precursor solutions.  Figure 2a showed the surface tension of all the precursor solutions. As shown in Figure 2a, the surface tension of the precursor solutions without FSO were within the range of 1.33 mN/m at around 52.96 mN/m. The surface tension was almost unchanged with the increase of PVA mass fraction, indicating that PVA might reach saturation in the liquid surface because the concentration exceeded its CMC. The measured surface tension of AMT solution was 58.14 mN/m, and the surface tensions of the precursor solutions without FSO were all lower than those of AMT solution, proving that PVA had surface activity. In the precursor solution, the FSO mass fraction increased with the PVA mass fraction increase because the ratio of FSO to PVA was constant. In general, because of high surface activity, FSO always tends to gather at the liquid surface, and the surface tension will decrease with the increase of its mass fraction. However, in Figure 2a, it can be seen that the surface tension of the precursor solutions with FSO increased first, and then decreased while obtaining the maximum at 3.8 wt.% PVA. A possible explanation was that hydrogen bonds were formed between PVA and FSO at first, resulting in the surface tension increase. After hydrogen bonds reached saturation in the liquid surface, the increase of FSO mass fraction would decrease the surface tension. The surface tension of the precursor solutions without FSO was all below 23.00 mN/m while the surface tension of the precursor solutions with FSO was all over 51.00 mN/m, which illustrated that the addition of FSO could greatly reduce the surface tension of the precursor solutions.  (1)) describes the relation between contact angle and surface tension for a liquid droplet on a smooth and uniform solid surface,

Effect of Polyvinyl Alcohol (PVA) and 2-Perfluoroalkyl Ethanol (FSO) on Precursor Solutions and Film Formation
where γsv, γsl and γlv respectively denote the surface tensions of solid/air, solid/liquid and liquid/air interfaces; θ is the Young's contact angle. The measured contact angle of AMT solution was 51.31°. From Figure 2b, the contact angle of the precursor solutions without FSO was lower than that of AMT solution when PVA was no higher than 2.9 wt.% while the contact angle was higher than that of AMT solution when PVA was no lower than 3.8 wt.%. The reason might be that when the concentration of PVA exceeded a critical value between 2.9 wt.% and 3.8 wt.%, the surface tension between substrate and solution from being decreased to being increased because PVA entanglement structure changed. From Figure 2b, the contact angle of the precursor solutions with FSO had a fluctuant decline with the PVA mass fraction increase. The reason might be that FSO affected the surface tensions of liquid/air and solid/liquid through hydrogen bonds, resulting in the fluctuation of the contact angle. It was also found in Figure 2b that at the same mass fraction of PVA, the contact angle of the precursor solutions with FSO were smaller than that of the precursor solutions without FSO, which indicated that the addition of FSO could improve the wettability of the precursor solution on the substrate by changing the surface tension. The contact angle at 20 s of all the precursor solutions is shown in Figure 2b. Young's equation (Equation (1)) describes the relation between contact angle and surface tension for a liquid droplet on a smooth and uniform solid surface, where γ sv , γ sl and γ lv respectively denote the surface tensions of solid/air, solid/liquid and liquid/air interfaces; θ is the Young's contact angle. The measured contact angle of AMT solution was 51.31 • . From Figure 2b, the contact angle of the precursor solutions without FSO was lower than that of AMT solution when PVA was no higher than 2.9 wt.% while the contact angle was higher than that of AMT solution when PVA was no lower than 3.8 wt.%. The reason might be that when the concentration of PVA exceeded a critical value between 2.9 wt.% and 3.8 wt.%, the surface tension between substrate and solution from being decreased to being increased because PVA entanglement structure changed. From Figure 2b, the contact angle of the precursor solutions with FSO had a fluctuant decline with the PVA mass fraction increase. The reason might be that FSO affected the surface tensions of liquid/air and solid/liquid through hydrogen bonds, resulting in the fluctuation of the contact angle. It was also found in Figure 2b that at the same mass fraction of PVA, the contact angle of the precursor solutions with FSO were smaller than that of the precursor solutions without FSO, which indicated that the addition of FSO could improve the wettability of the precursor solution on the substrate by changing the surface tension. The relation of contact angle with the time of all the precursor solutions is shown in Figure 3. From Figure 3a, the contact angle of the precursor solutions without FSO dropped all less than 6.21 • in 30 s and became stable at 30 s. From Figure 3b, the contact angles of  Figure 3a, it was found that at the same mass fraction of PVA, the contact angle of the precursor solutions with FSO had faster descent than that of the precursor solutions without FSO, which illustrated the addition of FSO could accelerate the decreased spread of the contact angle by improving wettability. It could be speculated that the final stable contact angle of the precursor solutions with FSO would be much smaller than that of the precursor solutions without FSO, which might make the film more uniform. The relation of contact angle with the time of all the precursor solutions is shown in Figure 3. From Figure 3a, the contact angle of the precursor solutions without FSO dropped all less than 6.21° in 30 s and became stable at 30 s. From Figure 3b, the contact angles of the precursor solutions with FSO all dropped more than 12.65° in 30 s and continued to decrease at 30 s. Comparing Figure 3b with Figure 3a, it was found that at the same mass fraction of PVA, the contact angle of the precursor solutions with FSO had faster descent than that of the precursor solutions without FSO, which illustrated the addition of FSO could accelerate the decreased spread of the contact angle by improving wettability. It could be speculated that the final stable contact angle of the precursor solutions with FSO would be much smaller than that of the precursor solutions without FSO, which might make the film more uniform. Films printed using AMT solution and the precursor solutions without FSO were observed under an optical microscope at 10× magnification in Figure 4a-h, which showed the effect of the dynamic viscosity on film formation. AMT had high solubility in water, but there was polycrystal of the films in Figure 4a and the polycrystal was ammonium para-tungstate (APT) according to the XRD analysis ( Figure 6a). It was found by comparing Figure 4b-d and Figure 4a that the addition of PVA could improve the uniformity of the film. One possible reason was that the increase in PVA molecules separated the APT molecules in space. Another possible reason given by a previous study was that hydrogen bonds of NH4 + ions with −OH groups of PVA [37] separated APT, which was beneficial to improve the uniformity of the film. The films in Figure   Films printed using AMT solution and the precursor solutions without FSO were observed under an optical microscope at 10× magnification in Figure 4a-h, which showed the effect of the dynamic viscosity on film formation. AMT had high solubility in water, but there was polycrystal of the films in Figure 4a and the polycrystal was ammonium para-tungstate (APT) according to the XRD analysis (Figure 6a). It was found by comparing Figure 4a-d and that the addition of PVA could improve the uniformity of the film. One possible reason was that the increase in PVA molecules separated the APT molecules in space. Another possible reason given by a previous study was that hydrogen bonds of NH 4 + ions with −OH groups of PVA [37] separated APT, which was beneficial to improve the uniformity of the film. The films in Figure 4a  The precursor solutions for the films shown in Figure 4i-l were suitable for screen printing, so for convenient description, the term 'ink formulas' was used to refer to the precursor solutions adding 3.8 wt.%, 4.7 wt.%, 5.6 wt.% and 6.5 wt.% PVA with, respectively, 0.013 wt.%, 0.016 wt.%, 0.019 wt.% and 0.022 wt.% FSO in the next part of this study. The ink formulas are universal in some areas, which means that some other types of film can be screen-printed by replacing AMT with other material and then adjusting it slightly.

Effect of Annealing on Composition and Transmittance of Films
Having defined what was meant by 'ink formulas', the effect of annealing on the films will be discussed. The FT-IR spectra of unannealed films printed with the ink formulas are shown in Figure 5a. In Figure 5a, the infrared absorption at 3240 cm −1 , 2947 cm −1 and 1087 cm −1 can be, respectively, assigned to tensile vibration mode of −OH, −CH2 and C-O in PVA [38]. There was still a little water in unannealed films because the absorption at 1630 cm −1 in Figure 5a was caused by the tensile vibration mode of H-O-H [39]. The infrared absorption at 3427 cm −1 corresponded to the characteristic N-H stretching [40]. According to the XRD analysis in Figure 6a, N-H stretching existed in APT. The FT-IR spectra of annealed films printed with the ink formulas are shown in Figure 5b. Comparing Figure 5b with Figure 5a, it is found that PVA and water in the films were greatly reduced due to annealing. Figure 5c was an enlargement in the 1050-800 cm −1 wavenumber range of Figure 5b. From Figure 5c, the bands at 900-600 cm −1 and 990-960 cm −1 are, respectively, attributed to W-O-W stretching vibration and W=O stretching vibration, which illustrated all the annealed films contained WO3 [41]. The precursor solutions for the films shown in Figure 4i-l were suitable for screen printing, so for convenient description, the term 'ink formulas' was used to refer to the precursor solutions adding 3.8 wt.%, 4.7 wt.%, 5.6 wt.% and 6.5 wt.% PVA with, respectively, 0.013 wt.%, 0.016 wt.%, 0.019 wt.% and 0.022 wt.% FSO in the next part of this study. The ink formulas are universal in some areas, which means that some other types of film can be screen-printed by replacing AMT with other material and then adjusting it slightly.

Effect of Annealing on Composition and Transmittance of Films
Having defined what was meant by 'ink formulas', the effect of annealing on the films will be discussed. The FT-IR spectra of unannealed films printed with the ink formulas are shown in Figure 5a. In Figure 5a, the infrared absorption at 3240 cm −1 , 2947 cm −1 and 1087 cm −1 can be, respectively, assigned to tensile vibration mode of −OH, −CH 2 and C-O in PVA [38]. There was still a little water in unannealed films because the absorption at 1630 cm −1 in Figure 5a was caused by the tensile vibration mode of H-O-H [39]. The infrared absorption at 3427 cm −1 corresponded to the characteristic N-H stretching [40]. According to the XRD analysis in Figure 6a, N-H stretching existed in APT. The FT-IR spectra of annealed films printed with the ink formulas are shown in Figure 5b. Comparing Figure 5b with Figure 5a, it is found that PVA and water in the films were greatly reduced due to annealing. Figure 5c was an enlargement in the 1050-800 cm −1 wavenumber range of Figure 5b. From Figure 5c, the bands at 900-600 cm −1 and 990-960 cm −1 are, respectively, attributed to W-O-W stretching vibration and W=O stretching vibration, which illustrated all the annealed films contained WO 3 [41]. The XRD patterns in Figure 6 were analyzed using the software Jade 6.0, PDF #18-0126(5(NH4)2O·12WO3·7H2O) and PDF #32-1395(triclinic WO3). The XRD patterns of unannealed films printed with the ink formulas are shown in Figure 6a. From Figure 6a, the polycrystal in all the unannealed films was 5(NH4)2O·12WO3·7H2O, which is a kind of ammonium para-tungstate (APT) crystal. A probable explanation was that AMT was converted into APT during the printing and drying process. The XRD patterns of annealed films printed with the ink formulas are shown in Figure 6b. From Figure 6b, triclinic WO3 was found in all the annealed films. It can also be qualitatively observed in Figure 6b that the crystallinity of triclinic WO3 increased and then decreased with the increase of PVA mass fraction while obtaining the highest crystallinity at 5.6 wt.% PVA. The reason might be that the mass fraction and entanglement of PVA affected the thermal decomposition of APT in the annealing process. Comparing Figure 6a with Figure 6b, it can be illustrated that WO3 is generated from APT decomposition in the annealing process.   The XRD patterns in Figure 6 were analyzed using the software Jade 6.0, PDF #18-0126(5(NH4)2O·12WO3·7H2O) and PDF #32-1395(triclinic WO3). The XRD patterns of unannealed films printed with the ink formulas are shown in Figure 6a. From Figure 6a, the polycrystal in all the unannealed films was 5(NH4)2O·12WO3·7H2O, which is a kind of ammonium para-tungstate (APT) crystal. A probable explanation was that AMT was converted into APT during the printing and drying process. The XRD patterns of annealed films printed with the ink formulas are shown in Figure 6b. From Figure 6b, triclinic WO3 was found in all the annealed films. It can also be qualitatively observed in Figure 6b that the crystallinity of triclinic WO3 increased and then decreased with the increase of PVA mass fraction while obtaining the highest crystallinity at 5.6 wt.% PVA. The reason might be that the mass fraction and entanglement of PVA affected the thermal decomposition of APT in the annealing process. Comparing Figure 6a with Figure 6b, it can be illustrated that WO3 is generated from APT decomposition in the annealing process.  The XRD patterns in Figure 6 were analyzed using the software Jade 6.0, PDF #18-0126(5(NH 4 ) 2 O·12WO 3 ·7H 2 O) and PDF #32-1395(triclinic WO 3 ). The XRD patterns of unannealed films printed with the ink formulas are shown in Figure 6a. From Figure 6a, the polycrystal in all the unannealed films was 5(NH 4 ) 2 O·12WO 3 ·7H 2 O, which is a kind of ammonium para-tungstate (APT) crystal. A probable explanation was that AMT was converted into APT during the printing and drying process. The XRD patterns of annealed films printed with the ink formulas are shown in Figure 6b. From Figure 6b, triclinic WO 3 was found in all the annealed films. It can also be qualitatively observed in Figure 6b that the crystallinity of triclinic WO 3 increased and then decreased with the increase of PVA mass fraction while obtaining the highest crystallinity at 5.6 wt.% PVA. The reason might be that the mass fraction and entanglement of PVA affected the thermal decomposition of APT in the annealing process. Comparing Figure 6a with Figure 6b, it can be illustrated that WO 3 is generated from APT decomposition in the annealing process.
It can be found in Figures 5 and 6 that annealing removed PVA and produced WO 3 , proving that annealing was an effective post-treatment for the films screen-printed with the ink formulas. The unannealed film containing 4.7 wt.% PVA was special. From Figure 5a, there was a very wide absorption peak around 2027 cm −2 , which was assigned to stretching vibration of C≡N and C≡C. From Figure 6a, the film containing 4.7 wt.% PVA had little polycrystal. A possible explanation was that PVA formed a special physical structure by accidental entanglement, which led to the unexpected infrared absorption peak and polycrystal inhibition. Figure 7a shows the transmittance of unannealed films printed with the ink formulas. It was found in Figure 7a that transmittance increased with the increased mass fraction of PVA. A possible explanation is that the PVA molecules tended to aggregate with the increase of PVA mass fraction because the concentration exceeded the CMC, making the film more transparent. Figure 7b shows the transmittance of annealed films printed with the ink formulas. The thickness of the annealed films at 3.8 wt.%, 4.7 wt.%, 5.6 wt.% and 6.5 wt.% PVA were, respectively, 486.2 nm, 458.8 nm, 580.7 nm and 426.8 nm. For the annealed films, the transmittance in Figure 7b decreased with the increased film thickness. Comparing Figure 7b with Figure 7a, it was found that at the same fraction of PVA, the transmittance of the annealed films was overall higher than that of unannealed films because PVA and water were removed. It can be found in Figures 5 and 6 that annealing removed PVA and produced WO3, proving that annealing was an effective post-treatment for the films screen-printed with the ink formulas. The unannealed film containing 4.7 wt.% PVA was special. From Figure  5a, there was a very wide absorption peak around 2027 cm −2 , which was assigned to stretching vibration of C≡N and C≡C. From Figure 6a, the film containing 4.7 wt.% PVA had little polycrystal. A possible explanation was that PVA formed a special physical structure by accidental entanglement, which led to the unexpected infrared absorption peak and polycrystal inhibition. Figure 7a shows the transmittance of unannealed films printed with the ink formulas. It was found in Figure 7a that transmittance increased with the increased mass fraction of PVA. A possible explanation is that the PVA molecules tended to aggregate with the increase of PVA mass fraction because the concentration exceeded the CMC, making the film more transparent. Figure 7b shows the transmittance of annealed films printed with the ink formulas. The thickness of the annealed films at 3.8 wt.%, 4.7 wt.%, 5.6 wt.% and 6.5 wt.% PVA were, respectively, 486.2 nm, 458.8 nm, 580.7 nm and 426.8 nm. For the annealed films, the transmittance in Figure 7b decreased with the increased film thickness. Comparing Figure 7b with Figure 7a, it was found that at the same fraction of PVA, the transmittance of the annealed films was overall higher than that of unannealed films because PVA and water were removed.
where α is the absorption coefficient of the films; h is the Planck constant; ν is light frequency; A is a constant; n = 2 because tungsten oxide belongs to the indirect bandgap and  where α is the absorption coefficient of the films; h is the Planck constant; ν is light frequency; A is a constant; n = 2 because tungsten oxide belongs to the indirect bandgap and Eg is the optical bandgap which is the value of the fitting line horizontal intercept of (αhν) 1/2 versus the photon energy hν. It can be obtained from Figure 7c that the band gaps of the annealed films printed with the ink formulas at 3.8 wt.%, 4.7 wt.%, 5.6 wt.% and 6.5 wt.% PVA were 3.18 eV, 3.21 eV, 3.12 eV and 3.15 eV, respectively, which were all in the normal range of the WO 3 bandgap.

Electrochromic Performance
For the performance of Figure 8, the applied voltage cycle consisted of −3 V for 15 s and 3 V for 15 s. Figure 8a shows the relation of current with the time of the annealed films printed with the ink formulas. The injecting and extracting charges per unit area (∆Q) were calculated based on the current of a cycle of electrochromic test. It was found in Figure 8a that ∆Q first decreased and then increased with the increase PVA mass fraction while obtaining the minimum at 5.6 wt.% PVA. From Figure 8b, the crystallinity of triclinic WO3 increased and then decreased with the increase PVA mass fraction while obtaining the highest crystallinity at 5.6 wt.% PVA. We all know that the electrochromic mechanism refers to a process of electron and Li + migration [43], so Li + migration can affect the ∆Q. Therefore, a possible explanation provided by several studies for the relation of the crystallinity and ∆Q was that the increase of crystallinity made it more difficult for Li + to disengage and embed [44][45][46]. Eg is the optical bandgap which is the value of the fitting line horizontal intercept of (αhν) 1/2 versus the photon energy hν. It can be obtained from Figure 7c that the band gaps of the annealed films printed with the ink formulas at 3.8 wt.%, 4.7 wt.%, 5.6 wt.% and 6.5 wt.% PVA were 3.18 eV, 3.21 eV, 3.12 eV and 3.15 eV, respectively, which were all in the normal range of the WO3 bandgap.

Electrochromic Performance
For the performance of Figure 8, the applied voltage cycle consisted of −3 V for 15 s and 3 V for 15 s. Figure 8a shows the relation of current with the time of the annealed films printed with the ink formulas. The injecting and extracting charges per unit area (∆Q) were calculated based on the current of a cycle of electrochromic test. It was found in Figure 8a that ∆Q first decreased and then increased with the increase PVA mass fraction while obtaining the minimum at 5.6 wt.% PVA. From Figure 8b, the crystallinity of triclinic WO3 increased and then decreased with the increase PVA mass fraction while obtaining the highest crystallinity at 5.6 wt.% PVA. We all know that the electrochromic mechanism refers to a process of electron and Li + migration [43], so Li + migration can affect the ∆Q. Therefore, a possible explanation provided by several studies for the relation of the crystallinity and ∆Q was that the increase of crystallinity made it more difficult for Li + to disengage and embed [44][45][46].  Figure 8b shows the optical transmittance spectra at 600 nm of the annealed films printed with the ink formulas. The optical modulation of the device was defined as the difference between the highest and lowest transmittances. It is shown in Figure 8b that optical modulation of the annealed films decreased one by 1 at 3.8 wt.%, 4.7 wt.% and 5.6 wt.% PVA while there was little difference on the optical modulation of the annealed films at 5.6 wt.% and 6.5 wt.% PVA. One reason was that crystallinity affected Li + migration,  Figure 8b shows the optical transmittance spectra at 600 nm of the annealed films printed with the ink formulas. The optical modulation of the device was defined as the difference between the highest and lowest transmittances. It is shown in Figure 8b that optical modulation of the annealed films decreased one by 1 at 3.8 wt.%, 4.7 wt.% and 5.6 wt.% PVA while there was little difference on the optical modulation of the annealed films at 5.6 wt.% and 6.5 wt.% PVA. One reason was that crystallinity affected Li + migration, resulting in the decrease of the optical modulation. The other possible reason was that the residues from PVA thermal decomposition affected the number of electrons involved in the electrochromic reaction, which led to the electron injection of the film at 5.6 wt.% PVA being smaller than that of the film at 6.5 wt.% PVA but there was little difference in the optical modulations.
The time required for the device from the highest optical transmittance to the transmittance dropping by 90% of the optical modulation is recorded as the coloring response time while the time required from the lowest optical transmittance to the transmittance increasing by 90% of the optical modulation is recorded as the fading response time. Figure 8c showed the coloring and bleaching response time in the same cycle of the annealed films printed with the ink formulas. It was found in Figure 8c that the coloring response time first increased and then decreased with the increase in PVA mass fraction while obtaining a minimum at 3.8 wt.% PVA. It might be explained that that crystallinity affected Li + migration, resulting in the change of the response time. Figure 8c also illustrates that the bleaching was almost complete within 3 s for all the annealed films. This implied that these films showed a relatively fast bleaching response, which might be due to the pores introduced by PVA thermal decomposition [47]. Figure 8d shows the colored and bleached states of the electrochromic devices, demonstrating that the annealed films screen-printed with the ink formulas could complete the electrochromic function. Figure 9 indicates the electrochromic performance of the annealed film printed with the ink formula at 3.8 wt.% PVA had the best optical modulation and the shortest coloring response time. The transmittance of the initial state, colored state, and bleached state of the film is shown in Figure 9a. In Figure 9a, it is shown that the blue color of the device in the colored state was due to the good absorption of long wavelengths in the visible light region. It can also be proved in Figure 9a that the device had stronger absorption in the infrared region, which stated that the film could absorb heat.
It can be seen in Figure 8b that the transmittance continued to decrease in the coloring process. To make the transmittance of the coloring device reach a stable value, the measurement conditions were changed. The applied voltage of response time of the film at 600 nm in Figure 9b was −3 V for 240 s and then 3 V for 240 s for one cycle. From Figure 9b, the coloring response time was 87.5 s and the bleaching response time was 18.5 s. It can be seen in Figure 9b that the optical transmittance of the film still continued to decrease if the charge was injected for a long time, which proved that the film could have large optical modulation with sufficient charge injection. The long coloring response time might be caused by the low electrical conductivity of WO 3 , which could be solved by embedding silver nanowire [48].
The optical density (OD) in Figure 9c is characterized as under 1.01 × 10 −5 A·cm −2 constant current density. Coloration efficiency (CE) is also an important parameter to evaluate electrochromic performance, which is defined by the following Equations (3) and (4) [49,50]: In the Equation (2), ∆OD is the optical density of the change between two optical states at 600 nm wavelength; Tb and Tc were the points from the optical transmittance curve; the injecting and extracting charge density (Q) was calculated by integrating the current injected and A is 1 cm −2 as the electrochromic total area. It was found in Figure 9c that the CE of the film could attain 93.25 cm −2 ·C −1 . Figure 9 indicates the electrochromic performance of the annealed film printed with the ink formula at 3.8 wt.% PVA had the best optical modulation and the shortest coloring response time. The transmittance of the initial state, colored state, and bleached state of the film is shown in Figure 9a. In Figure 9a, it is shown that the blue color of the device in the colored state was due to the good absorption of long wavelengths in the visible light region. It can also be proved in Figure 9a that the device had stronger absorption in the infrared region, which stated that the film could absorb heat. Figure 9. Properties of the annealed film using 3.8 wt.% PVA ink formulas: (a) optical transmittance spectra in initial, bleached and colored state, (b) response time, (c) optical density, (d) transmittance at the start and after 60 cycles (at 600 nm wavelength). The optical transmittance spectra at the start and after 60 cycles at 600 nm in Figure 9d under applied voltage cycle consisted of −3 V for 15 s and 3 V for 15 s. it is shown in Figure 9d that the stability of the film was not very good because the optical modulation had a 25.80% decline after 60 cycles. One reason might be that the annealed films were composed of polycrystals ( Figure 6b) but generally WO 3 crystal films had better stability because of the stable structure [51]. Another possible reason was that PVA thermal decomposition produced residues and pores (Figure 8d), which might affect significantly the stability of the films under the applied voltage cycles. The stability might be improved by modifying the precursor solutions and changing the annealing method.

Conclusions
In this study, WO 3 electrochromic film was successfully prepared by screen printing. Polyvinyl alcohol (PVA) and 2-perfluoroalkyl ethanol (FSO) were added to modify the precursor solutions. The addition of PVA improved the dynamic viscosity of the precursor solutions and modify the uniformity of the films. FSO improved the wettability of the precursor solutions by decreasing the surface tension. The precursor solutions with dynamic viscosity in the range of 28.40 mPa·s to 163.61 mPa·s and surface tension in the range of 16.17 mN/m to 22.90 mN/m formed film well on the ITO substrate. The ink formulas that were the precursor solutions adding 3.8 wt.%, 4.7 wt.%, 5.6 wt.% and 6.5 wt.% PVA with, respectively, 0.013 wt.%, 0.016 wt.%, 0.019 wt.% and 0.022 wt.% FSO are universal in some areas, which makes sense in the preparation of other types of functional films by screen printing. The annealing removed PVA, produced WO 3 and improved the transmittance of the film. The optical modulation of the film was affected by crystallinity and residues of PVA thermal decomposition. It was found that the ink formulas with 3.8 wt.% PVA and 0.013 wt.% FSO had the best performance in this work. The coloration efficiency of the film could attain 93.25 cm −2 ·C −1 but the stability of the film required improvement.
In this study, the screen-printing inks were prepared and WO 3 electrochromic films were obtained through annealing, which had positive significance in the preparation of WO 3 electrochromic devices by the low-cost screen-printing method.