High-Throughput Characterization of Structural and Photoelectrochemical Properties of a Bi–Mo–W–O Thin-Film Materials Library

1 Introduction

Research on photoelectrocatalytically active materials for solar water splitting is important [1] to discover new light-absorbing semiconductors with suitable band gaps in the range from 1.6 to 2.4 eV [2] as a basis for using a large range of the solar spectrum. In addition, the band edge positions need to match the envisaged redox processes of photoelectrochemical (PEC) water splitting and parasitic reactions, e.g. charge carrier recombination, should be minimal [3], [4], [5], [6]. Furthermore, the materials should be non-toxic, abundant, low cost, and stable [4]. Due to these requirements, it is challenging to discover suitable semiconductors as photoabsorbers which, in combination with adapted electrocatalysts, exhibit a high conversion of incident photons to electrons for product formation. Bi₂WO₆ [6], [7], [8], [9] and Bi₂MoO₆ [10], [11], [12] fulfil some of the requirements and were here selected as starting materials for a combinatorial search of new materials suitable for photoelectrocatalytic energy conversion. Bi₂WO₆ exhibits band edges with suitable positions of 1.77 eV for the valence band (VB) and −1.03 eV for the conduction band (CB), which makes Bi₂WO₆ a candidate material for photoelectrocatalytic water splitting without application of an external bias voltage [4], [8], [13]. Bi₂MoO₆ was selected as a starting compound since its absorption edge is red-shifted to a wavelength of about 560 nm, which is a basis for absorbing a wider range of the incident solar spectrum as compared with Bi₂WO₆. However, Bi₂MoO₆ showed substantially lower photoelectrocatalytic activity than Bi₂WO₆ [14].

Here, we aim to combine the properties of both material systems by fabricating Bi–W–Mo–O compositions-spread materials libraries (MLs) which are subsequently systematically analysed by high-throughput methods [15], [16] with respect to the effects of Mo addition to the Bi–W–O system and its suitability for PEC applications, aiming to reveal correlations of composition, crystal structure and microstructure on the photoelectrochemical properties (PEC) as well as the influence on the bandgap.

2 Experimental and methods

3 Results and discussion

Since Bi₂WO₆ exhibits the higher photocurrent of the two base compounds [13], the ML was designed such that it included Bi₂WO₆ for comparison. The results of the EDX composition analysis of the annealed ML is shown in a ternary diagram (Figure 1), where compositions are provided as ternary Bi/W/Mo atomic ratios. The oxygen amount in the ML is shown color-coded. The atomic ratio of oxygen was calculated according to (Oat.%Mat.%+Oat.%∗100). The compositions of the sputtered metals (M) range from 8 to 30 at.% Mo, from 25 to 74 at.% W, and from 24 to 70 at.% Bi. The oxygen amount determined by means of EDX gives only a rough estimate of the O content in the thin-film library as it was sputtered on FTO substrates and thus contributions to the amount of O from the underlying substrate cannot be excluded. According to the measured O amount, a significant uptake of O occurs in Mo-rich regions which is consistent with the thickness gradient with the thinnest layer being deposited in Bi-rich regions. The O amount is stoichiometric in the Mo-rich and thickest areas, and under-stoichiometric in the thinnest and Bi-rich area of the ML.

Fig. 1:

Color-coded ternary Bi–W–Mo–O composition diagram of the thin-film materials library after heat treatment in air at 500 °C for 4 h.

Figure 2a shows selected XRD patterns of measurements areas with W-rich to Mo-rich compositions. The XRD patterns reveal that the Aurivillius (Bi₂O₂)²⁺ (A_M−1B_MO_3M+1)²⁻ phase is present in the entire ML. The corresponding phase is (Bi₂O₂)²⁺ (Bi_M(W_1−NMo_N)_M−1O_3M+1)²⁻/(Bi₂O₂)²⁺ (Bi_MW_M−1O_3M+1)²⁻ [9], [12]. For comparison, a diffraction pattern of the Bi₂WO₆ phase is displayed at the bottom of Figure 2a. One of the characteristic peaks (2θ = 28° ≙ [131]) [8] shifts to smaller diffraction angles with increasing Mo content in the ML indicating lattice distortion [1]. Additionally, a sharp peak at 2θ = 32° ≙ [002] is observed with increasing amount of Mo.

Fig. 2:

(a) XRD patterns of selected measurement areas of the Bi₂Mo_xW_1−xO_z ML for compositions with x from 0.08 to 0.53. (b) SEM images of three regions exhibiting different Mo content with x = 0.08, 0.25 and 0.53.

The increasing intensity of these peaks indicates increasing crystallinity of the thin films with higher Mo-content. In addition, the diffraction peak at 2θ = 28° is getting sharper with higher amount of Mo in the thin film (FWHM minimum in the Mo-rich region: 0.11, FWHM maximum in Bi-rich region: 1.39) which implies that the crystallite domain size is getting larger if Mo is added [1]. Figure 2b shows different morphologies depending on the amount of Mo in the thin film. A high amount of Mo shows small linked grains, a medium amount shows plates and also small linked grains, while the lowest amount of Mo exhibts some large plates, but many more smaller non interlinked plates.

Figure 3 shows color-coded ternary diagrams for (a) the peak width (FWHM), (b) the peak center, and (c) the calculated crystallite domain sizes. The peak width is smaller with either high amount of Bi or high amount of Mo in the thin film. By considering only the quasi-binary cut with a constant Bi amount of 47% no change in the peak width can be seen. Although with low amount of Bi (25%), decreasing amount of W and increasing amount of Mo a change in the peak width can be observed. Hence, it can be concluded that the peak width is increasing with decreasing amount of Mo and is decreasing with an increasing amount of Mo indicating again an increasing crystallinity and crystallite domain size. Furthermore, the peak center shifts to smaller angles with increasing amount of Mo if the amount of Bi is <45% indicating a lattice distortion at higher Mo content. The crystallite domain size is correlated with the peak width, as calculated using the Scherrer equation in which the peak width and the peak center are the only variables [18]. The smallest crystallite domain size is visible at a small amount of Bi and Mo and a high amount of W. The biggest crystallite domain sizes were identified for a small amount of Bi and high amounts of W and Mo. The color-coded ternary map of the crystallite domain size supports the conclusion that Mo was successfully incorporated.

Fig. 3:

Color-coded ternary maps derived from the XRD patterns of the MLs for the [131] plane 2θ = 28: (a) peak width, (b) peak center, (c) crystallite domain sizes calculated by applying the Scherrer equation for the most prominent peak [131].

Optical transmission measurements were performed at each measurement area of the ML to determine the bandgap. Bandgaps were calculated by semi-automated Tauc-plot analysis (see Figure 4a). The smallest bandgap of 2.0 eV was calculated for a composition area of Bi₃₁Mo_11–23W_46–58O_z. Another area with comparatively small bandgaps is the Bi-rich region with low Mo and low W amount in the thin film. A high bandgap (2.6–2.8 eV) was found in the region with medium amount of Bi (45–50%), low amount of W (25–35%) and high amount of Mo (22–29%). The area in between shows bandgaps around 2.2–2.4 eV. The difference in the bandgaps seems to be influenced by two parameters, namely (i) the degree of lattice distortion due to substitution of Mo, and ii) the quantity of the Mo 4d to W 5d orbital involved in the conduction band. A similar correlation for the influence of the Mo amount was shown before [1]. Comparing literature values for the bandgap of Bi₂WO₆ (2.7–2.8 eV [1], [4]) it can be noticed that by addition of Mo, the bandgap was systematically decreased for the complete ML.

Fig. 4:

(a) Semi-automated Tauc-plot analysis for indirect bandgap extrapolation.

(b) Color-coded ternary map of the calculated bandgaps.

Figure 5a shows the correlation of the indirect bandgap against the concentration of Mo in the thin film and Figure 5b against the ratio of W:Mo in the ML. It can be seen, that the indirect bandgap is increasing with increasing amount of Mo. With increasing amount of W and decreasing amount of Mo, the bandgap is decreasing. This supports the hypothesis that the bandgap seems to be influenced by the degree of lattice distortion due to substitution of Mo and the quantity of the Mo 4d to W 5d orbital involved in the conduction band.

Fig. 5:

(a) XY-diagram of the indirect bandgap vs. the concentration of Mo in the thin film

(b) XY- diagram of the indirect bandgap vs. the ratio of W:Mo.

Figure 6a shows a plot of the color-coded photocurrent density in a partial ternary diagram. The highest photocurrent density was measured in the region with a Bi-content of 28–31%, a Mo content of 6–11% and a W content of 61–66%. The corresponding XRD pattern of the region with highest photocurrent density can be seen in Figure 2a with a Mo content of x = 0.08–0.14. The measurement area of highest photocurrent density (hit region) has a composition of Bi₃₁Mo₇W₆₂O_z and a photocurrent density of 31 μA/cm². Another area with relatively high photocurrent density is the region with high Bi amount (>50%) and low amount of Mo (<12%) as well as a low amount of W (≈25%). This region has the composition Bi₆₇Mo₉W₂₇O_z (Bi_1.7 Mo_0.25W_0.75O_z). This correlates with the stoichiometric Bi₂ Mo_0.25W_0.75O₆ composition of the identified crystal structure. The region in between scarcely shows any photocurrent density.

Fig. 6:

(a) Color-coded map of the photocurrent density measured with the OSDC in 0.1 M Na₂SO₄, pH = 5.8 at an applied bias volatge of 1.23 V vs. RHE at an illumination of 100 mW/cm². (b) Measured IPCE values in the region of highest photocurrent density with low Mo-concentration in comparison with Bi₂WO₆ (sputtered reference ML).

The IPCE values in the region of highest photocurrent (Bi₃₁Mo₇W₆₂O_z) and a Bi₂WO₆ sample were compared in Figure 6b. Bi_0.45Mo_0.11W_0.89O_z is photoelectrochemically active up to a wavelength of 460 nm while Bi₂WO₆ is photoelectrochemically active just up to a wavelength of 420 nm. By comparing Bi₂WO₆ and Bi_0.45Mo_0.11W_0.89O_z it can be concluded that the bandgap was decreased due to the insertion of the Mo in the lattice which is supported by the calculated bandgap values presented in Figure 4b.

Focusing on a quasi-binary cut, for example from 12–24% Mo, 25% Bi and 55–65% W (Figure 6a), it can be supposed that a high concentration of Mo (4d element) decreases the photocatalytic efficiency for water oxidation which is most likely due to the exchange of W (5d element) by Mo which leads to a change in electronegativity. The effective electronegativity of a 4d element is higher and one can anticipate that it forms more covalent bonds with oxygen and will thus lead to a possibly improved delocalization of charge carriers. Furthermore, the change in electronegativity leads to a smaller curvature of the conduction band (CB). The smaller curvarture because of the more covalent bonds with oxygen are not effective for the movement of photogenerated electron-hole pairs and will thus lead to a decreasing photocatalytic efficiency [19]. On the other hand, just a minute content of Mo in the lattice resulted in a decrease of the bandgap, shifting the absorption spectra to higher wavelength, and also improving the IPCE values as compared to Bi₂WO₆ (cf. Figure 6b) suggesting an improved electron transport due to lattice distortion analogous to BiVO₄ [20].

4 Conclusions

A Bi₂Mo_xW_1−xO₆ ML was synthesized using reactive magnetron co-sputtering. The ML shows maximum photocurrent density and IPCE values in the composition region Bi₂Mo_0.07W_0.93O₆ as compared with Bi₂WO₆. Addition of Mo decreases the bandgap to a value of around 2.2 eV providing the basis for light absorption to higher wavelengths. Moreover, Mo is increasing the crystallite domain size if the Bi amount in the thin film is low. It is shown that the crystallite domain sizes influence the photocurrent density such that an increased crystallite domain size leads to a lower photocurrent density for the case of low Bi-concentrations. The decreased photocurrent density is probably due to a smaller space charge layer, an increased travelling length for the electrons to reach the surface within their lifetime, which also correlates with the increased grain size [21].