Electrochromic properties of epitaxial WO₃ thin films grown on sapphire substrates

Tungsten trioxide (WO₃) is a typical electrochromic (EC) material¹⁾ and has been widely studied for such applications as smart windows,^1–3) display devices,⁴⁾ catalysis,^5,6) pH sensors,⁷⁾ gas sensors,^8,9) batteries,¹⁰⁾ and resistive switching memories.^11,12) In addition, this material attracts considerable attention for application to novel iontronic FETs.^13–17) Owing to the wide application field and functionality, experimental^18,19) and theoretical^20,21) studies have been reported for the EC mechanism by many researchers.

WO₃ crystals are based on a pseudo-perovskite structure with the elemental unit of a WO₆ octahedron, and many polymorphs can be formed by tilting and distorting the corner- and edge-sharing octahedra.²²⁾ Therefore, it has been difficult to obtain single crystalline films, and most experimental studies were based on amorphous or polycrystalline WO₃ films grown by sputtering,^3,7,12,16) vacuum evaporation,⁹⁾ electrochemical deposition,²⁾ and sol–gel^8,10,17) methods except in the case of single-crystalline nanostructures.^4,6) Recently, however, the emerging interest in iontronic FETs has provoked a rapid progress in WO₃ epitaxial growth using molecular beam epitaxy (MBE)^5,23,24) and pulsed laser deposition (PLD).^15,25)

In this study, protonation-induced EC properties of MBE-grown epitaxial WO₃ thin films are reported by focusing on their relationship with crystallographic structural change. Compared with polycrystalline films, epitaxial films are expected to provide much simpler experimental conditions for structural analysis, although, to our knowledge, only Nishihaya et al.¹⁵⁾ have reported the structural change by intercalating Li ions to PLD-grown epitaxial WO₃ films. The structural change under the restriction by the substrate is another interest of this study considering its application to iontronic FETs.^13,15)

The WO₃ films used for this experiment were grown using an EpiQuest RC2100YS-02 MBE system equipped with an rf-excited radical cell for the oxygen source and a Knudsen-type cell for the WO₃ compound source. The growth was conducted on r-plane sapphire substrates using a two-step growth method, where the initial 15-nm-thick WO₃ layer was grown at 700 °C followed by the rest WO₃ growth at 500 °C as reported previously.²⁴⁾ The epitaxially grown films were monoclinic crystals (mono-WO₃; a = 0.7284 nm, b = 0.7507 nm, c = 0.7674 nm, and β = 90.62°)²²⁾ with (001) planes parallel to the r-plane sapphire substrates. In this epitaxial relationship, lattice mismatches in the [100] and [010] directions of mono-WO₃ films are as large as 4.6 and 8.0%, respectively.²⁶⁾ It has been confirmed that the lattice constants in both lateral and growth directions of the films are just close to those of bulk WO₃, indicating the complete relaxation of interface strain to accommodate the large lattice mismatches.

The bandgap energy of the WO₃ films determined from the Tauc plot of optical transmittance data assuming indirect transition was 3.0 eV, and the conduction type estimated by Hall effect measurement using the van der Pauw method was n with a carrier concentration of (5–10) × 10¹⁶ cm⁻³ and a mobility of 5–10 cm² V⁻¹ s⁻¹ at room temperature.

Considering the convenience for Hall effect measurement, WO₃ films of about 5 mm² area on sapphire substrates were fabricated into planar EC devices by forming metal (In) electrodes covered with epoxy resin at respective four corners. For the redox operation of EC devices, the four electrodes were bundled and connected to the negative terminal of a power supply to apply a certain voltage against the Pt counter electrode in 10 mM aqueous solution of sulfuric acid; i.e., the current was fed into the WO₃ film from the lateral direction.

Since the contact resistance of the electrodes was on the order of 10 kΩ, considerable part of the applied voltage will be devoted to the potential drop at the electrodes when the film resistance is decreased by protonation. Therefore, the applied voltage in this study was set at the optimal point within the potential window of electrolysis by measuring the current–voltage characteristics in advance, and no bubble generation was observed during the following redox operations.

Figure 1 shows the optical transmittance change of an EC device with a 300-nm-thick WO₃ film by the protonation operation at a constant current of +0.66 mA cm⁻² for 120 s. The solid and dashed lines depict the in situ-measured transmittance decreases at 1070 and 650 nm as the representatives of near-infrared (NIR) and visible (VIS) regions, respectively; the wavelength λ of which was chosen in the transparency windows of the aqueous electrolyte solution. By the operation, the color of the WO₃ film was changed from transparent (bleached state) to dark blue (colored state) for the naked eye, and coloration was performed uniformly in the whole EC plane from the beginning.

Zoom In Zoom Out Reset image size

The insets in Fig. 1 show the transmittance spectra measured in air before and immediately after protonation, indicating that coloration is dominated by the transmittance decrease at λ ≥ 500 nm. This transmittance decrease was due to the increases in absorbance and reflectance in the VIS and NIR regions, respectively. The reflectance in the VIS and NIR regions was around 20% before protonation, and only that in the NIR region of λ ≥ 820 nm increased after protonation with rough saturation at around 60% for λ ≥ 900 nm.

The coloration efficiency defined as log(T_b/T_c)/(Q/S) is estimated to be 150 and 63 cm² C⁻¹ with response times of 16 and 34 s at 1070 and 650 nm, respectively, where T_b is the transmittance of the bleached state, T_c is that of the colored state at T_b/10, Q is the electric charge required for coloration, and S is the area of the WO₃ film exposed to the aqueous electrolyte solution. Roughly, the coloration efficiency is high and the response time is average compared with the typical values reported for protonation by sulfuric acid.^27,28) It is also shown that the transmittance decrease in the NIR region starts before that in the VIS region.

When we applied a constant voltage of +4 V to the EC device, coloration fully saturated within 10 min for both VIS and NIR regions. After applying a constant voltage of −4 V for 10 min to the colored state, the decreased transmittance in the VIS region recovered to the initial bleached state, but that in the NIR region still remained. To recover the decreased transmittance in the NIR region, it was effective to anneal the WO₃ film in air at 100 °C for 10 min or to keep it in air at room temperature for a long time as described later.

This transmittance change was reproducibly observed for the cyclic protonation/deprotonation operations indeed, but the WO₃ film gradually exhibited many cracks that finally removed the film from the substrate as small particles, which is presumably due to the stress cycles under the operations. This degradation was suppressed by decreasing the WO₃ film thickness, and the film was found to endure cracking for more than scores of cycles when its thickness was not more than 50 nm. Therefore, we employed EC devices with 50-nm-thick WO₃ films for the following experiments.

In Fig. 2(a), solid lines show the optical transmittances of an EC device with a 50-nm-thick WO₃ film before, after the application of +4 V for 10 min, after the application of −4 V for 10 min, and after annealing in air at 100 °C for 10 min. Owing to the reduced WO₃ film thickness, the degree of transmittance change is much smaller, but the response to the redox operation is similar to that in the case presented in Fig. 1. The solid lines in Fig. 2(b) depict changes in carrier concentration and mobility under the operations. One can see marked increases in carrier concentration and mobility induced by the positive bias application. The increased mobility was reduced by the negative bias application, while the high carrier concentration was retained. The high carrier concentration was reduced to the initial level of the bleached state by annealing.

Zoom In Zoom Out Reset image size

The remaining of the decreased transmittance in the NIR region after the negative bias application has been similarly reported by other researchers^3,29) as the characteristic of crystalline WO₃. We also confirmed the crystallinity dependence by fabricating an EC device with a 50-nm-thick amorphous WO₃ film which was grown at room temperature by MBE and showed no X-ray diffraction (XRD) peak. The experimental results are given by the dashed lines in Figs. 2(a) and 2(b) where the degrees of increases in carrier concentration and mobility induced by the positive bias application are smaller than those in the case of epitaxial WO₃, and both increases in carrier concentration and mobility as well as the decrease in the VIS and NIR transmittances are completely recovered to the bleached state by the negative bias application.

Considering the reports by other researchers, it is suggested that the increase in absorbance in the VIS to NIR region is presumably due to the charge transfer between W⁶⁺ and W⁵⁺ and/or the polaron formation in tungsten bronzes (H_xWO₃) by the injected electrons from the metal electrodes to neutralize protonation.^1,3,29,30) The increase in reflectance in the NIR region of the epitaxial EC device, on the other hand, might be due to the Drude-like reflection^3,4,29,31) by considering the observed high free carrier concentration over 10²⁰ cm⁻³. Since the reflectance increase in colored state is observed for λ ≥ 820 nm, the crossing of the solid and dashed lines at λ ≈ 700 nm in Fig. 2(a) after +4 V application is due to the larger absorbance of the protonated epitaxial film at 700 ≤ λ ≤ 820 nm, suggesting the presence of a larger absorption band in this NIR region.

The crystal structure of the WO₃ films was analyzed by XRD using a Rigaku SmartLab system with Cu-Kα₁ radiation (λ = 0.154059 nm). The XRD patterns in the 2θ/ω configuration are given in Fig. 3 for the epitaxial WO₃ film at respective stages in Figs. 2(a) and 2(b). The peak position at 23.12° (d = 0.3844 nm) before protonation is the 002 reflection from mono-WO₃. The peak position at 23.42° (d = 0.3795 nm) after the positive bias application closely agrees with the 002 reflection from hexagonal tungsten bronze with x = 0.34 (hex-H_0.34WO₃; a = 0.7374 nm and c = 0.7586 nm).²²⁾ The peak position at 22.94° (d = 0.3874 nm) after the negative bias application approximately agrees with the 002 reflection from orthorhombic tungsten hydrate with the coordinated water mole fraction of 0.33 (ortho-WO₃ 0.33H₂O; a = 0.7325 nm, b = 1.256 nm, and c = 0.7721 nm).²²⁾

Zoom In Zoom Out Reset image size

We tentatively assign the crystal structure after the positive bias application to a hex-H_0.34WO₃ phase by considering the lattice spacing in the growth direction and the coloration in the VIS region. Also, although the lattice spacing in the growth direction is 0.3% larger, we assign the crystal structure after the negative bias application to an ortho-WO₃ 0.33H₂O phase by considering the transparency in the VIS region and the high carrier concentration.⁴⁾ After annealing, the peak position returned nearly close to that of the initial mono-WO₃ phase. These assignments are not conclusive, and other metastable structures might be probable under the restriction by the substrate. Therefore, further analysis must be carried out to clarify the crystal structures, for example, using high-resolution wide-range reciprocal space XRD mapping³²⁾ and/or Raman scattering.

In the next step, we studied the sequential change in the crystal structure mentioned above in more detail. For the protonation operation, we applied +4 V with poses and repetitions by measuring the XRD patterns at every step. For the deprotonation operation, we measured the XRD patterns repeatedly at certain intervals by keeping the protonated EC device in air at a relative humidity of 20% and 27 °C.

Figure 4(a) shows the representative XRD patterns measured at certain lapse times after positive bias application. Immediately after the application, the 002 reflection from the mono-WO₃ phase moved to the lower-angle side and temporarily stayed for several seconds at around 22.94°, indicating the appearance of the ortho-WO₃ 0.33H₂O phase. Then, the peak shifted again toward the lower-angle side, and at around 20 s, it reached 22.79° (d = 0.3899 nm) which closely agrees with the 002 reflection from hexagonal WO₃ (hex-WO₃; a = 0.7299 nm and c = 0.7796 nm).²²⁾ Subsequently, another peak from the hex-H_0.34WO₃ phase developed at 23.42°. This peak increased with the lapse time with the decrease in the contribution from the hex-WO₃ phase, and finally dominated the diffraction pattern by subjecting a shoulder peak from the ortho-WO₃ 0.33H₂O phase.

Zoom In Zoom Out Reset image size

Figure 4(b) displays the change of the XRD pattern after keeping the protonated EC device in air at room temperature. It is indicated that the shift of the pattern during the deprotonation period is thoroughly in the reverse sequence of that during the protonation period. After keeping the device in air for approximately 1 d, the initially observed peak from the hex-H_0.34WO₃ phase was replaced by that from the hex-WO₃ phase with a shoulder peak from the ortho-WO₃ 0.33H₂O phase, and then, the peak from the hex-WO₃ phase gradually moved toward the higher-angle side. Subsequently, the pattern was predominated by the ortho-WO₃ 0.33H₂O phase after 7 d, and finally it became that of the initial mono-WO₃ phase after 60 d. The XRD pattern after 7 d in Fig. 4(b) is close to that obtained immediately after the negative bias application in Fig. 3.

Considering the peak evolution shown in Figs. 4(a) and 4(b), the XRD pattern of the protonated state in Fig. 3 is composed of a major peak from the hex-H_0.34WO₃ phase with a shoulder peak from the ortho-WO₃ 0.33H₂O phase. The XRD pattern after the negative bias application is predominated by the reflection from the ortho-WO₃ 0.33H₂O phase. The sequential change from the hydrate to the tungsten bronze with the progress in protonation agrees with the time-dependent transmittance decrease shown in Fig. 1, where the decrease in the VIS region follows that in the NIR region with a delay time. In addition, the phase transition sequence during the protonation and deprotonation periods, mono-WO₃ ↔ ortho-WO₃ 0.33H₂O ↔ hex-WO₃ ↔ hex-H_0.34WO₃, is consistent with the high similarity of these crystal structures,^4,6,33) and the widely known fact that ortho-WO₃ 0.33H₂O is a precursor of hex-WO₃.^6,22,33)

Note that respective XRD patterns in Fig. 3 have interference fringes. Since the appearance of interference fringes is evidence of the surface flatness and uniformity of the films, the disappearance during the phase transition periods in Figs. 4(a) and 4(b) is as expected, but the reproducible and repeatable appearance at the respective stationary states with nearly the same lattice constants as bulk crystals is surprising, considering the restriction by the substrate. Needless to say, further study must be performed to clarify the three-dimensional crystal structures; however, such a flexible structural change might be related to the recent reports by other researchers on epitaxial WO₃ growth referring to the easy deformation of the crystal structure with the help of WO₆ octahedron distortion and in-plane dislocation lines.^23,25)

In summary, we studied the EC properties of epitaxial mono-WO₃ films grown on sapphire substrates. Although the films were restricted by the substrate, the crystal structure was found to transform at respective protonation stages. It is revealed that the coloration in the VIS to NIR region is due to the formation of the hex-H_0.34WO₃ phase and the reflection in the NIR region is due to the formation of the ortho-WO₃ 0.33H₂O phase. It is also found that the transition between the mono-WO₃ phase and the hex-H_0.34WO₃ phase is reversibly and repeatedly conducted by passing through the ortho-WO₃ 0.33H₂O phase.

Acknowledgement