Hydrogen in tungsten trioxide by membrane photoemission and density functional theory modeling

The measurement of hydrogen-induced changes in the electronic structure of transition metal oxides by x-ray photoelectron spectroscopy is a challenging endeavor, since no photoelectron can be unambiguously assigned to hydrogen. The H-induced electronic structure changes in tungsten trioxide have been known for more than 100 years but are still controversially debated. The controversy stems from the difficulty in disentangling effects due to hydrogenation from the effects of oxygen deficiencies. Using a membrane approach to x-ray photoelectron spectroscopy, in combination with tunable synchrotron radiation, we measure simultaneously core levels and the valence band up to a hydrogen pressure of 1000 mbar. Upon hydrogenation, the intensities of the ${\mathrm{W}}^{5+}$ core level and a state close to the Fermi level increase following the pressure-composition isotherm curve of bulk ${\mathrm{H}}_x{\mathrm{WO}}_3$. Combining experimental data and density functional theory, the description of the hydrogen-induced coloration by a proton polaron model is corroborated. Although hydrogen is the origin of the electronic structure changes near the Fermi edge, the valence band edge is now dominated by tungsten orbitals instead of oxygen as is the case for the pristine oxide, having wider implications for its use as a (photoelectrochemical) catalyst.

Figure 1

Sketch of the membrane XPS sample holder and the experimental setup. The enlarged area shows the hydrogen permeation and the hydrogen chemical potential through the membrane. At steady state, the chemical potential on the feed side of the membrane is nearly equal to that on the vacuum side allowing the measurement of hydrogen concentrations in the material corresponding to atmospheric pressures by surface science techniques [19, 20].

Figure 2

Illustration of the cubic crystal structure of ${\mathrm{WO}}_3$ (left) and of one polymorph of ${\mathrm{H}}_{0.25}{\mathrm{WO}}_3$ (right) as used for modeling. Note the distortion of the oxygen atoms (blue) induced by hydrogen (red) amounts as little as $x=0.25$. This distortion is local: As an example, two tungsten atoms (gray) exclusively surrounded by oxygen atoms are highlighted with a green corona, and one tungsten atom surrounded by five oxygens and one OH group is highlighted with a purple corona. In the local picture, the latter atoms are ${\mathrm{W}}^{5+}$, and the ones surrounded solely by oxygen are ${\mathrm{W}}^{6+}$. Moving an electron and hydrogen (arrow) requires structural rearrangements.

Figure 3

The fraction of ${\mathrm{W}}^{5+}$ in the sample, as derived from peak fitting shown in the Appendix (Fig. 7), is shown in black, with the linear interpolation of the slope shown by the dashed gray lines during the increasing hydrogen pressure and the constant hydrogen pressure phases, shown in red. The grayscale spectra show the changes in the W $4f$ spectra during the increasing hydrogen pressure phase.

Figure 4

Modeled partial DOS (PDOS) and total DOS of cubic ${\mathrm{WO}}_3$ (black line) and ${\mathrm{H}}_x{\mathrm{WO}}_3$ (red line) with hydrogen concentrations as small as $x=0.25$. $E_G=1.3$ eV is the indirect band gap in ${\mathrm{WO}}_3$ derived from the onsets of valence and conduction bands. Full band structure calculations give the direct band gap of 2.3 eV (not shown). Note the minimal contribution of hydrogen electrons (purple dash-dotted line) to the total density of states in ${\mathrm{H}}_{0.25}{\mathrm{WO}}_3$.

Figure 5

(a) Calculated partial and total DOS of ${\mathrm{H}}_{0.25}{\mathrm{WO}}_3$ (black) and ${\mathrm{H}}_{0.75}{\mathrm{WO}}_3$ (red). (b) Measured photoemission spectra of minimum and maximum hydrogen content in black and red. The hydrogen contents as labeled are derived from the peak intensities of the peak evolving with hydrogen pressure as shown in the inset (see Fig. 6 and text for more information).

Figure 6

The intensity of hydrogen-induced, conduction-band-like tungsten states (black) are shown as a function of time and with hydrogen back pressure (red). Solid symbols indicate increasing hydrogen pressure, and open ones mean constant hydrogen back pressure. The bulk pressure-composition isotherm data on Pt:${\mathrm{WO}}_3$ (blue) at 423 K are plotted on the same pressure axis [18]. Squares represent absorption data, whereas triangles represent desorption data.

Figure 7

(a) The pre- and posthydrogenation XPS survey spectra are shown in black and red, respectively. (b) The W $5p_{3/2}$ (brown) and W $4f$ (gray and violet) core levels of the electrodeposited ${\mathrm{WO}}_3$ as prepared, as described in Sec. 2a, measured by Al $\mathrm{K}\alpha$ XPS. The spectra are fitted with two pairs of doublets giving the ${\mathrm{W}}^{6+}$ and ${\mathrm{W}}^{5+}$ fractions, respectively. The peak fitting indicates the sample being almost completely oxidized, with a small fraction of ${\mathrm{W}}^{5+}$ states present ($\simeq 7.7\%$).

Figure 8

W $4f$ (a) and O $1s$ (b) core-level spectra recorded using 700 eV photon energy. The spectra were recorded in situ before and after exposure to hydrogen (black and red, respectively). The ${\mathrm{W}}^{5+}$ fraction increases from 37.6 to 51.2% and the oxygen defect fraction from 38.8 to 52.0% upon hydrogenation. The intensity ratio of W to O remains nearly constant with $I\left({\mathrm{O}}_{\mathrm{latt}}\right)/I\left(\mathrm{W}\right)=1.06\to 1.03$.

Figure 9

Double-log plot of the measured hydrogen pressure in the vacuum chamber as a function of the applied feed pressure. A fit yields a slope of 0.79.

Figure 10

Top: As an example, two different hydrogen constellations in ${\mathrm{H}}_{0.25}{\mathrm{WO}}_3$, denoted as $+1+1$ and $+1-1$, before relaxation. Bottom: DFT of nine different constellations. The right graph is an enlargement around the band gap.