Coloration mechanism of electrochromic NaxWO3 thin films.

The coloration mechanism of tungsten trioxide (WO3) upon insertion of alkali ions is still under debate after several decades of research. This Letter provides new insights into the reversible insertion and coloration mechanisms of Na+ ions in WO3 thin films sputter-deposited on ITO/glass substrates. A unique model based on a constrained spline approach was developed and applied to draw out ε1+iε2from spectroscopic ellipsometry data from 0.6 to 4.8 eV, whatever the state of the electrochromic active layer, i.e., as-deposited, colored, or bleached. It is shown that electrochemically intercalated sodium-tungsten trioxide, NaxWO3(x=0.1,0.2,0.35), exhibits an absorption band centered at ca. 1.14 eV in ε2 governing the coloration mechanism.

The coloration mechanism of tungsten trioxide (WO3) upon insertion of alkali ions is still under debate after several decades of research. This Letter provides new insights into the reversible insertion and coloration mechanisms of Na + ions in WO3 thin films sputterdeposited on ITO/glass substrates. A unique model based on a constrained spline approach was developed and applied to draw out ε1+iε2 from spectroscopic ellipsometry data from 0.6 to 4.8 eV whatever the state of the electrochromic active layer, i.e. as-deposited, colored or bleached. It is shown that electrochemically intercalated sodium-tungsten trioxide, NaxWO3 (x=0.1, 0.2, 0.35), exhibits an absorption band centered at ca. 1.14 eV in 2 governing the coloration mechanism. © 2018 Optical Society of America Electrochromic (EC) devices, whose optical properties change under the effect of an electrical excitation, see an increasing interest because they allow a control of the optical properties in the visible and near infrared ranges for applications from coatings for building and airplanes smart windows to printable colorchanging paper to anti-dazzling rear-view mirrors. Promoted by cathodic polarization, the insertion of small cations from an electrolyte and simultaneous injection of electrons from the back contact in tungsten oxide (WO3) are used in the fabrication of EC displays since the 1980s [1][2][3]. Schematically the corresponding electrochemical process can be expressed as follows: (1) where M + is usually a proton or alkali ion (Li + , Na + ) and 0≤x≤1. Allceramic devices based on proton or lithium conduction have received considerable interest [4][5][6][7][8]. Whereas the reversible insertion of Na + into WO3 has been scarcely studied, the recent development of highly conductive Na + ion conductors in thin film form as well as the possibility to use WO3 in sodium-ion batteries [9] is calling for a greater attention to this system. A maximum Na/W ratio around 0.35 can be obtained using electrochemical intercalation and fulfills the requirements for such an application, i.e. a film with sufficient electronic conductivity, good chemical stability, and optical properties allowing large coloration contrast [10][11][12].
A pending issue about this system and, more generally, MxWO3, is the exact origin of the blue coloration mechanism, which is of large interest, e.g. for prediction of their life-cycle performance. As reviewed by Granqvist [13], two limiting cases can be considered to illustrate the coloration mechanism: for crystalline films, electron delocalization overpowers [14] whereas for more disordered films, localized electrons dominate and polaronic or closely related models [1,[15][16][17][18][19][20][21][22] are used. The latter imply transitions between W m+ and W n+ sites with m and n being the IV, V or VI oxidation states. Additionally, for nanoparticles (NPs), localized surface plasmon resonance (LSPR) of free electrons was advanced [23][24][25], sometimes combined with polaron absorption [23,26].
In this Letter we report on the coloration mechanism of NaxWO3 thin films by means of the analysis of their dielectric functions investigated by spectroscopic ellipsometry (SE). SE is a technique of predilection to study optical properties of materials, based on the change in polarization state between incident and reflected light on a sample [27].
Sub-stoichiometric tungsten trioxide layers synthesized within this study possess mixed conducting properties and an amorphous structure allowing the intercalation of sodium cations. The intercalation level x was calculated by electrochemical investigations via coulometric integration and applying Faraday's law while taking into account the geometric characteristics of the samples and based on the following assumptions: (i) only Na + ions were reversibly intercalated (confirmed by SIMS analysis), (ii) an experimental mean WO3 density of 4.72 g/cm 3 was used (corresponding to ~30% porosity, of columnar-like type [28]) in agreement with other work [29].
The direct current deposition was performed in a 40-liter chamber using a 2 inch-diameter, 3 mm-thick tungsten target (99.95 % purity) mounted on a magnetron system and whose surface faces the substrate holder surface. The magnetron is placed off-axis with respect to the axis of the substrate-holder and the latter was put in rotation in order to minimize thickness and composition inhomogeneities. More details about the geometry of the chamber can be found in Ref. [30]. Sputtering was performed in an Ar/O2 reactive gas mixture. The gas flow rates, target current, target axis-to-substrate and target-to-substrate holder distances, and working pressure were 85 standard cubic centimeters per minute (sccm) Ar, 1.6 sccm O2, 0.2 A, 100 mm, 75 mm, and 4.5 Pa, respectively. A preliminary study showed such a pressure enables to produce an open columnar morphology of the as-deposited WO3 film maximizing the coloration contrast [11]. The working pressure was adjusted by setting the turbomolecular pump speed around 18500 rpm. The self-established voltage discharge was close to 415 V, corresponding to a power of 83 W dissipated by the target. Substrates consist of ca. 2.5 x 2.5 cm² samples of ITO-coated glass. The selected O2 flow rate allows a good transparency of the as-deposited state (labeled as "WO3"). Sodium intercalation and deintercalation were performed in a 0.1 M Na2SO4 electrolyte buffered with a ~2.7 pH unit solution (0.1 M C8H5KO4/0.1 M HCl). Chronoamperometric methods using a three electrode cell configuration (WO3 acting as working electrode, Ag/AgCl sat. KCl reference electrode, and Pt counter electrode) yielded colored ("Na0. 35WO3", corresponding to 92 mC/cm 2 of charge) or bleached ("Na0WO3") state, respectively for 180 s at -0.6 V or +0.2 V. Intermediate levels were also obtained ("Na0. 1WO3" and "Na0. 2WO3") corresponding to ca. 13 and ca. 43 mC/cm 2 of charge, respectively. Measurements were preceded by three activation cycles in the same electrolyte.
NIR to NUV (0.60-4.81eV with a 0.01 eV resolution) ellipsometric experiments were performed in reflection mode (UVISEL, Horiba Jobin Yvon) on stacks consisting of glass/ITO layer/EC active layer. parameters were measured in the spectral range of interest for incidence angles between 60° and 75° for three samples for each mentioned EC state [see Fig. 1, symbols]. All back faces of the glass substrates were roughened in order to eliminate incoherent reflection. The inversion of ellipsometric data was performed using a four-phase representative model of the sample: glass/ITO layer/ EC active layer/roughness layer/air. In this model the thickness of the roughness layer, the thickness of the active layer, and the dielectric function of the active layer were unknown, while the roughness layer was modeled by a mixture of 50% air and 50% active material according to Bruggeman effective medium approximation (BEMA) [31]. The unknown parameters were determined by minimizing the mean-square difference between generated and experimental Is and Ic data. The particular representation of the unknown dielectric function by a constrained spline approach was adopted. This method is described and illustrated in detail elsewhere [32,33]. The dedicated ellipsometric data inversion procedure neither requires precise a priori knowledge of the considered stacks nor requires the use of dispersion relations. Briefly, in this approach the spectral imaginary part of the dielectric function is represented by a collection of third order polynomials (elemental splines) over reduced spectral ranges linked by connection points while the real part is represented by the superimposition of the Kramers-Kronig derived contribution and an additional Sellmeier term to account for higher energy transitions. The connection between the elemental splines allows obtaining a continuous dielectric function over the whole considered spectrum, and is performed using particular constraints for the first derivatives at the connection points in order to obtain realistic values of slopes and avoid unphysical parasitic oscillations. In the fitting process the abscissa of the connection points are fixed while the ordinates representing the dielectric function values at the considered connection points are used as fitting parameters. In the particular present problem of the active layer, a decomposition of the dielectric function into eight parts (nine connection points) over the whole considered spectrum was used. The optical constants of both the glass substrate and the ITO layer were predetermined individually by ellipsometry. Fig. 1. Experimental (symbols) and fitted (lines) ellipsometric curves (Is, Ic) at variable angle of incidence for different states of the active layer: (a-b) WO3, (c-d) Na0. 35WO3, and (e-f) Na0WO3. Figures 1(a-f) show the typical experimental and fitted ellipsometric spectra of the stack layers for different states of the active layer, indicating the overall good match between experimental and fitted data. Additionally, morphological parameters in Table 1 reveal two points. First, the level of top roughness of a few nanometers is coherent with the RMS roughness obtained by AFM, and with reported values [34]. Secondly, the film thickness evolves upon reversible intercalation from 396 nm (WO3) to 413 nm (Na0. 35WO3) and back to 398 nm (Na0WO3), corresponding to a volume expansion of ca. 4% of the host oxide due to ion intercalation [35]. The thickness was confirmed by cross-sectional SEM images (not shown here). As an extra validation of the optical model, Figure 2 presents the generated transmission spectra of the stack which agree qualitatively with the experimental ones, yielding a transmission contrast T of ca. 60% at 550 nm. By analyzing another stack with thicker EC thickness (ca. 900 nm), a T of ca. 77% is also accordingly predicted. The corresponding dielectric functions are displayed in Fig.  3(a). The optical functions were also tabulated in terms of n+ik (see Data Files 1-3) and are found to be in accordance with the reported evolution in hydrogen-tungsten bronze [36]. On the higher energy side, the absorption onset is related to the fundamental absorption gap of the material (between 3.0 and 3.4 eV [37,38]). The low energy of 2 spectra evidences an absorption band located at ca. 1.14 eV whose amplitude increased upon x varying from 0.1 to x=0.35, that is notably different from WO3 and Na0WO3 spectra that are nearly superimposed and flat in this region. We note that the slightly non-zero absorption at ca. 1.1 eV is probably related to the sub-stoichiometry of the films [39]. As expected for amorphous films, no metallic behavior was evidenced. However, such a feature in a bulk material can be hindered in the resulting effective  in the case of LSPR. Our determined  are effective ones, and while, in principle, no NPs clusters should be considered in principle, interfaces due to the film's open morphology could generate LSPR and explain this NIRvisible feature. However, we were able to reject this hypothesis through the use of a BEMA using the DFT-calculated  values of NaxWO3 [25] and the void value.
The pronounced absorption at ca. 1.14 eV or, as depicted in Fig. 3(b), ca. 1.3 eV when termed in absorption coefficients difference (x)-(0), is now considered in the frame of a polaronic hypothesis. Indeed, such characteristic shape agrees well with several works [1,22,36,[40][41][42] in which this asymmetric feature is deconvoluted into one or more peak functions associated with small polaron transitions. Another reported option, which does not contravene with the former theory due to the large polaron coupling constant in WO3, is related to large polaron formalism as the similar ~0.75 eV peak observed for LixWO3 with very low intercalation levels [43]. Figure 3(c) also reports older (x) data [44] obtained for sputtered NaxWO3 thin films showing a global coherent tendency. Among the aforementioned approaches we used the superimposition of three Gaussians as performed in [1] on lithium-tungsten bronze (generalized site-saturation model) to further deconvolute the (x)-(0) spectra, as shown in Fig.3(b). By comparison with the reported peaks strengths in Fig.3(c) to our values, we find contributions centered around 1.3 eV, 2.43 eV and 3.7 eV assignable to W 5+ ⬄W 6+ , W 4+ ⬄W 6+ and W 4+ ⬄W 5+ polaron transitions, respectively. Transition probabilities and magnitude of the constants used are given in Fig.3(c). The tendency is quite similar to that reported for Li + [1] or H + [36] intercalations. However for H0. 35WO3 [36] the probabilities are of 87, 9 and 4% for (5-6), (4)(5)(6) and (4)(5) transitions, respectively. In our case of Na0. 35WO3 a similar level is found for (5-6) and (4-6) transitions (46 and 48%) plus a ~6% relative amplitude for the (4-5) one. Figure 3(c) shows also a decrease in strength with x for the lowest amplitude (4)(5) transition, consistent with other reports [1,40]. Note that Zhang et al. [29,45] proposed this transition is a signature for the sub-stoichiometry in as-deposited trioxide. We have tried to analyze our (0) spectra on this basis and according the Bryksin model [42] but the signal was too weak to be fitted with good confidence. So this point remains unclear.
In summary, the complex dielectric functions for different states of the studied EC material i.e. sputter-deposited tungsten trioxide, electrochemically intercalated Na-tungsten bronze NaxWO3 (x=0.1, 0.2, 0.35), and reversibly bleached oxide free of guest Na + ions, have been reported for the first time, supporting fundamental research on the coloration process via advanced controlling tool. The dedicated variable-angle spectroscopic ellipsometry with the constrained spline approach of data provided the dielectric functions of the materials. The results indicate a profound change in optical properties of NaxWO3 (x=0, 0.35). Of peculiar interest 2 features revealed a specific NIR peak at ca. 1.14 eV whatever the value of x, responsible for the colored state, supporting optical polaron transitions in an extended range i.e. at 1.3 eV, 2.43 eV and 3.6 eV when termed in absorption, corresponding to different amplitudes of W 5+ /W 6+ , W 4+ /W 6+ and W 4+ /W 5+ transitions, respectively. Free electron effects, especially plasmonic resonance, seem to be absent in these amorphous thin films. The resulting optical absorption data is compared with those reported for other tungsten bronzes MxWO3 (M=H, Li, and Na).