Symmetry driven control of optical properties in WO₃ films

The looming environmental necessity of developing carbon-free sustainable energy generation and storage technologies dictates the critical need for fundamental understanding and functional control of photoactive materials. The use of photovoltaic, photocatalytic, or photochemical processes promises the ability to harness and store the virtually inexhaustible source of photoenergy provided by the sun. However, there are still many hurdles that must be overcome to bring the prices of energy production and storage of photocentric approaches to levels that will allow market forces to drive a transition to this cleaner model. The ability to design materials around specific application parameters of light absorption, charge generation, and charge transport is thus vital to improving efficiencies and creating novel functionalities necessary for increasing implementation of photoactive materials. Many highly photocatalytic materials, such as TiO₂ and ZnO, show poor overall performance due to their inherent optical bandgaps above 3 eV, which precludes them from absorbing a significant portion of the visible light spectrum.¹ Other semiconductors have lower bandgaps and a good match with the spectrum of sunlight, but their alignment of the valence band (VB) and conduction band (CB) is unfavorable for specific electrochemical reactions like the oxygen and hydrogen evolution reaction that are fundamental for photocatalytic water splitting.² Thus, the ability to optimize absorption spectra and finely manipulate band alignments in photoactive materials is critical to our future energy and environmental needs.

Tungsten oxide (WO₃) is one of the most well studied photoactive materials. Its low cost and versatile properties make it an attractive material for applications. These applications include electrochromic devices, solar cells, photoelectrochemical cells for water splitting and other photocatalytic reactions.^3–10 WO₃ has a ReO₃-type structure that can be thought of as a perovskite ABO₃ structure without the large A-site cation in the center of the unit cell.¹¹ Upon cooling from high temperature, WO₃ undergoes a series of symmetry lowering phase transitions as cation displacement and oxygen octahedral rotation patterns shift.^12,13 These phases show drastic differences in optical,¹⁴ ferroelectric,¹³ and electronic transport¹⁵ properties. For example, the optical bandgap (E_g) of the room temperature monoclinic $γ$-WO₃ phase (⁠$Eg∼3.9eV$⁠) is much larger than the high temperature tetragonal $α$-WO3 phase (⁠$Eg∼2.5eV$⁠).^14,16 This highlights the strong coupling between electronic and structural instabilities and indicates that modifying WO₃’s symmetry by controlling its structural degrees of freedom provides an excellent avenue to manipulating its functional properties.

Imposing structural changes through strain may thus be a promising route to optimize the band structure of WO₃ for specific applications. Strain engineering is a standard approach to tune the electronic structure of non-oxide semiconductors and complex oxides.¹⁷ However, the common approach to control thin films using biaxial in-plane (ip) strain imposed by heteroepitaxy is rather inflexible, since continuous or ex situ strain control is not feasible. In this approach, strain states are set during growth through selection of substrate; the limited availability of suitable substrates means that only a few discrete strain geometries may be applied and that these states cannot be changed post-growth. Further, single axis lattice control, which may be critical to fine manipulations of internal symmetry, is impossible with epitaxial methods due to Poisson effects.

Recently, a new route around these limitations was discovered.¹⁸ Here, low energy helium ion implantation was shown to allow independent control of out-of-plane (oop) lattice expansion in epitaxial films without changing the in-plane (ip) lattice constants. The mechanism for lattice expansion arises from the inserted interstitial He atoms applying local chemical pressure which propagates toward the impinged film’s free surface, thereby elongating the crystal along the out-of-plane axis. This uniaxial “strain doping” approach provides an unprecedented level of structural control which can be used to continuously manipulate thin film properties.^18,19 Of further importance, the implantation process is applied after film growth and can be easily integrated into common lithography processes, which means that it can be immediately applied to existing material systems and quickly scaled up using existing infrastructure widely used in the semiconductor industry.

In this work, we study the influence of strain doping on the optical absorption spectrum of epitaxial WO₃ films. We show that low doses of He implanted into the WO₃ lattice efficiently induce uniaxial strain that leads to a sizable shift of the absorption spectrum and reduction of the optical bandgap. We support our results by density functional theory calculations and attribute the large strain response to the structure of WO₃ that allows the strain to be accommodated by tilts and rotations of oxygen octahedra.

Epitaxial 20 nm thin films of WO₃ were grown on single-crystalline LaAlO₃ (100) substrates by pulsed laser deposition (PLD). LaAlO₃ (a_pc = 3.788 Å) was chosen due to the small lattice mismatch to the room temperature bulk phase of WO₃ (monoclinic $γ$-WO₃; a = 7.306 Å, b = 7.540 Å, and c = 7.692 Å).²⁰ The films were deposited at $700°$C in 100 mbar O₂ and a laser fluence of 1.2 J/cm². After the deposition, He was implanted into the film at two different parts of the sample with an energy of 4 keV and a dose of $4×1015$ He/cm² and $8×1015$ He/cm². A 15 nm thick Au capping layer was sputtered onto the WO₃ film prior to the ion implantation to prevent sputtering, normalize the distribution of He through the film thickness, neutralize incoming He ions before they enter the film, and reduce the generation of lattice defects.¹⁸ After implantation, the Au layer was mechanically removed. The Au layer deposition and removal procedures have been shown to have no significant effect on the structural or morphological properties of the underlying film.²¹ In this work, the same routine was carried out on the undosed sample in order to ensure identical treatment.

X-ray diffraction (XRD) demonstrates epitaxial single-domain growth with high structural quality. Figure 1(a) shows $2𝜃−𝜃$ curves around the (001) reflection of the undosed (black), $4×1015$ He/cm² (red), and $8×1015$ He/cm² (blue) film. With increasing He dose, the film peak is gradually shifted to the left. This shows that the oop lattice parameter c is successively elongated by He implantation. It is important to note that the structural quality of the WO₃ films is not significantly degraded by the ion implantation process, with Laue fringes still being visible even under the highest dose. The small peak broadening is caused by the presence of a small strain gradient resulting from a slight difference in He distribution across the film thickness.

Figure 1(b) shows reciprocal space maps (RSM) around the pseudocubic (103) reflections of the three films. The undosed film is fully strained to the LaAlO₃ substrate with pseudocubic ip and oop lattice parameters of a_pc = 3.788 Å and c_pc = 3.659 Å, respectively. The RSMs reveal that upon He implantation the WO₃ film peak moves closer to the LaAlO₃ substrate, but the coherency with the ip lattice is preserved. This illustrates the introduction of single-axis oop strain into the WO₃ lattice.

Raman scattering was used to further characterization of the structural changes induced by He implantation. The Raman spectra shown in Fig. 1(c) demonstrate a red shift of the A_1g and B_2g vibrational Raman modes upon strain doping.²² This is in good agreement with a softening of phonon modes due to uniaxial strain. The two Raman modes shown here are excitations of O-W-O stretching phonons and thus particularly sensitive to the unit cell dimensions. Inducing oop strain due to strain doping increases the O-W-O bond distances and weakens the mechanical coupling, thereby reducing the phonon energy. This link between bond distance and phonon energy has been observed and well described in perovskites,^23,24 and in the present work, it serves as additional proof of the oop expansion of the unit cell due to He implantation while demonstrating that the underlying lattice structure has not been significantly altered by unintended lattice defects arising from the implantation process. This is important because the semiconducting behavior of WO₃ is extremely sensitive to oxygen deficiency. Small concentrations of oxygen vacancies lead to a drastic reduction of the electric resistance or even an insulator-metal transition^25,26 and a reduced bandgap due to the creation of defect states below the conduction band.^27,28 Thus, we further confirmed a lack of oxygen vacancies in the lattice by measuring electric transport properties of the three samples. In all cases, the resistance was too high to be recorded with our measurement setup (see supplementary material).

Variable angle spectroscopic ellipsometry (VASE) has been used to determine the optical properties of the WO₃ films. The energy dependent ellipsometric angles Ψ and Δ were recorded for all three samples. The data was then fit to a simple two-layer model consisting of the substrate and film in order to determine the optical properties of the films. The substrate and the WO₃ layer are approximated by a Kramers-Kronig consistent B-spline with 16 data points over the full energy range. Here, the optical properties of the LaAlO₃ substrate are first determined independently by fitting the VASE data of a bare substrate that was annealed under the same conditions as the actual sample during the PLD process. Next, the recorded film-substrate VASE data were simulated by fitting the WO₃ optical properties and leaving the LaAlO₃ properties fixed. This two-layer model is a reasonable simplification considering that (i) the roughnesses of the substrate-film interface and film-air surface are small compared to the film thickness, (ii) the He doping level variation across the film thickness is small enough that the film can be approximated as a single layer with homogeneous optical constants, and (iii) He that passes the WO₃ film and ends up in the first few nanometers of the substrate has a negligible influence on the optical properties. The latter requirement was tested by He implanting the bare LaAlO₃ substrate and determining the optical properties. No significant change was found after implantation [see supplementary material for experimental verification of (i)–(iii)].

In Fig. 2(a), we show the $(αhν)2$ Tauc-plot of the optical absorption spectrum for all three films with varied He dose. This plot allows for the determination of direct optical bandgaps by extrapolating the linear parts of the curves to zero. The bandgap E_g of the undosed film is 4.05 eV. This agrees well with the E_g of the monoclinic phase of WO₃ phase that is stable in bulk form at room temperature.¹² Strain doping the WO₃ film shifts the absorption spectrum toward lower energies as the c-axis expands. The low He dose induces $∼$1% c-axis expansion which reduces the bandgap by 0.18 eV. Further increasing the amount of He in the WO₃ lattice induces $∼$2% c-axis expansion which reduces the bandgap to 3.71 eV - a significant change of 0.34 eV to the undosed film. This is illustrated in Fig. 2(b) which shows E_g decreases almost linearly with the oop strain induced by strain doping epitaxial WO₃. Thus, strain doping is shown to be capable of reducing optical bandgaps while providing access to never before possible fine and continuous control over the absorption spectrum. A further interesting avenue to follow would be to coherently grow films on different substrates and combine epitaxial in-plane strain with out-of-plane strain via strain doping. This may allow for an even higher tunability of bandgaps.

To better understand the mechanisms driving the structure-function relationship of the strain doping process in WO₃, we performed density functional theory (DFT) calculations using the Vienna Ab initio Simulation Package (VASP) with the projector-augmented wave (PAW) potentials. The exchange-correlation was approximated within the generalized gradient approximation (GGA) of Perdew-Burke-Ernzerhof (PBE) functionals. We assume the monoclinic $γ$-structure of WO₃ for our calculations, which has P2₁/n symmetry and an a⁻b⁺c⁻ oxygen octahedral rotation pattern. This phase is stable at room temperature in bulk, and DFT calculations have been shown to predict only slightly smaller bandgaps than the experimental observations.²⁹ The ip lattice of the $γ$-WO₃ was strained to the experimental pseudocubic lattice parameter of the LaAlO₃ substrate (a = 3.788 Å). The other unit cell and internal lattice parameters are relaxed to determine the unstrained as-grown state. Then, the oop lattice is strained from −4% to 4% and the system is relaxed within the P2₁/n symmetry. The Monkhorst-Pack k-point sampling method in the Brillouin zone with a $4×4×4$ and $12×12×12$ mesh for ionic and electronic optimization, respectively. The energy cutoff was 500 eV, and the criteria for energy and force convergence are set to be 0.1 meV and 10 meV/Å, respectively.

Figure 3 shows plots of the total density of states (DOS) for varying strain states ranging from −4% (oop compression) to +4% (oop expansion). The energy scale reference is set to the Fermi energy. Due to the well-known underestimation of DFT, the calculated bandgaps are consistently smaller than the experimental optical bandgaps. However, the influence of oop strain is in qualitative agreement with our measurements. The bandgap clearly decreases continuously from about 1.9 eV at −4% to less than 1.2 eV at +4% strain. The experimentally observed decrease in E_g is slightly higher than DFT prediction, but the calculations confirm that a substantial reduction in E_g is expected with oop lattice expansion. This large strain response can be understood as the result of the strong coupling between electronic and structural instabilities.¹³ The graphs in Fig. 3 depict the calculated changes of the octahedral rotations and tilts between +4% and −4% strain. Compared to the oop compression under −4% uniaxial strain, the +4% oop elongation of the unit cell strongly reduces octahedral rotations about both ip axes, while the rotation about the oop axis is only slightly affected. This is consistent with previous strain doping effects experimentally observed in orthorhombic SrRuO₃.¹⁹ Due to the multitude of structural distortions from the ideal cubic WO₃ structure related to oxygen octahedral rotations and distortions as well as cation displacements, the band structure of WO₃ is quite complex.^27,29 Intuitively, one would expect Jahn-Teller-type distortions to be more important under positive strain where oxygen octahedra are straightened out, while under compressive strain the changes of bond angles and bond lengths due to varying octahedral rotations should have a larger influence.

We calculated the projected DOS for the +4% and −4% strain state in order to understand the large response of WO₃. Figures 4(a) and 4(c) show the DOS projected on the W 5d and O 2p states, respectively. For both, the +4% and −4% states, the VB consists almost entirely of O 2p states, while the CB consists of W 5d states hybridized with O 2p states. This is in agreement with previous calculations on bulk WO₃.^27,29,30 It also implies that the VB maximum is only changing marginally under uniaxial strain. However, large changes in the character of the CB can be observed. Under oop expansion the CB minimum is of antibonding $π*$ character with the main contribution coming from the W 5d_xy orbital that mixes in with O 2p_x and O 2p_y orbitals [illustrated in Fig. 4(b)]. Under compressive strain, the onset of the W 5d_xy band moves to much higher energies [see the black arrow in Figs. 4(a) and 4(c)], while the W 5d_xz and W 5d_yz only shift slightly [illustrated in Fig. 4(d)]. We conclude that the large reduction of the optical bandgap under uniaxial oop strain can mainly be ascribed to the lowering of the CB due to a shift of the antibonding W 5d_xy state. This particular sensitivity of the W 5d_xy state is a result of the specific structural changes with increasing strain. Since the positions of the bonding or antibonding W 5d states are determined by the overlap with the O 2p orbitals, they are extremely sensitive to bond angle changes due to oxygen octahedral rotations. As discussed before, it is mainly the octahedral rotations about the ip axes that are reduced upon oop expansion. Thus, the W 5d_xy orbital lying in the film plane will experience the most pronounced changes.

In this work, low energy He ion implantation is shown to induce continuously tunable out-of-plane lattice expansion in epitaxial single crystal WO₃ films. We demonstrate that the optical bandgap has a significant nearly linear inverse relationship with the length of the out-of-plane lattice parameter. Density functional theory closely matches experimental observations while further suggesting that uniaxial lattice expansion can be used to induce orbital modifications allowing fine control over the conduction band. The ability to continuously modify optical bandgap and conduction bands in a single crystal material has tremendous promise in allowing new capabilities to design materials around specific application parameters of light absorption, charge generation, and charge transport. The simplicity and versatility of strain doping through low energy helium implantation means that it has the capability to make significant contributions to our fundamental understanding of structure-function relationship in materials across many material classes.

See supplementary material for more details on He implantation, a discussion of the lattice structure and topography of the films, an annealing study, and details on the procedure to determine the optical bandgaps.

This work was supported by the DOE Office of Science, Basic Energy Sciences, Materials Sciences and Engineering Division and the Office of Science Early Career Research Program. This research was in part conducted at the Center for Nanophase Materials Sciences, which is a U.S. Department of Energy (DOE), Office of Science User Facility, and used resources of the National Energy Research Scientific Computing Center, which is supported by the Office of Science of the US Department of Energy under Contract No. DE-AC02-05CH11231. The work of S.F.R. was supported from PN 16 14 03 02 implemented within the ANCSI authority.

This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public-access-plan).