Oxyhydride Nature of Rare-Earth-Based Photochromic Thin Films.

Thin films of rare-earth (RE)-oxygen-hydrogen compounds prepared by reactive magnetron sputtering show a unique color-neutral photochromic effect at ambient conditions. While their optical properties have been studied extensively, the understanding of the relationship between photochromism, chemical composition, and structure is limited. Here we establish a ternary RE-O-H composition-phase diagram based on chemical composition analysis by a combination of Rutherford backscattering and elastic recoil detection. The photochromic films are identified as oxyhydrides with a wide composition range described by the formula REO xH3-2 x where 0.5 ≤ x ≤ 1.5. We propose an anion-disordered structure model based on the face-centered cubic unit cell where the O2- and H- anions occupy tetrahedral and octahedral interstices. The optical band gap varies continuously with the anion ratio, demonstrating the potential of band gap tuning for reversible optical switching applications.

While the structure–property relationships of single-anion materials, such as metal oxides, are to a large extent determined by variations in the cation chemistry, multianion compounds offer another dimension of control of material properties. This is due to the wide range of different anion characteristics, including electronegativity, polarizability, and ionic radius.[1] Perhaps the least studied group of multianion compounds are oxyhydrides where both oxide (O2–) and hydride (H–) ions are present. Since the successful synthesis of the transition metal (TM) oxyhydride LaSrCoO3H0.7 by Hayward et al.,[2] this material class has been extended to TM = Sc, Ti, V, Cr, Mn, Co based on the perovskite and the layered Ruddlesden–Popper crystal structure types.[3] A recent report on hydride ion (H–) conductivity in La2–SrLi1–HO3– makes oxyhydrides promising materials for energy storage and conversion applications.[4] Two synthesis routes for TM oxyhydrides have been established to date. Topochemical synthesis exploits the O2–/H– anion exchange between the parent oxide and a hydride (mostly CaH2) at moderate temperatures of 300–600 °C.[3] Alternatively, some TM oxyhydrides have been prepared by high-pressure (few GPa) direct synthesis from parent oxide and hydride powder mixtures at 1000–1300 °C.[3] These methods have also been employed to synthesize powders of lanthanide (Ln) oxyhydrides LnOH with Ln = La,[5] Nd,[6] Sm,[7,8] Gd–Er[7−9], the second established class of oxyhydride materials. The stoichiometric LnOH are reported to crystallize either in the anion-disordered face-centered cubic (fcc) fluorite type structure (Fm3̅m)[7−9] or in an anion-ordered superstructure with tetragonal (P4/nmm) symmetry.[5,6] To date, the electronic properties of the lanthanide oxyhydrides remain largely unexplored. A recent study by Ueda et al. suggests that Tb3+-doped GdOH is a promising candidate for phosphor applications.[9]

A remarkable photochromic effect at ambient conditions was discovered in reactive magnetron sputtered (MS) YOH thin films.[10] Initially, these semiconducting materials (band gap E ≈ 2.6 eV) were referred to as oxygen-containing yttrium hydride.[10,11] It was speculated that their properties can be explained in analogy to the transparent γ-YH3 phase stabilized in the fcc structure by incorporation of oxygen into the lattice.[11] A later study on sputtered YOH composition gradient thin films, which covers the transition from opaque to transparent state, reports atomic ratios of up to H/Y ≈ 3 as measured by heavy-ion elastic recoil detection (ERD) and nuclear reaction analysis (NRA).[12] However, this result seems rather questionable considering the low H2 partial pressures during reactive MS deposition. Follow-up studies by Montero et al.[13] and Nafezarefi et al.[14] clarified that the transparent YOH materials are in fact formed by air oxidation of as-deposited absorbing metallic β-YH2 films. Further, we reported that LnOH thin films with Ln = Gd, Dy, Er exhibit the same photochromic effect and fcc crystal structure as YOH.[14] On the basis of these findings and our preliminary chemical composition analysis experiments, we then adopted the term rare-earth (RE) oxyhydrides for this group of photochromic materials. However, systematic experimental evidence for the presence of H– ions in these REOH is scarce. Although a recent study by Moldarev et al. supports the concept of H– in their photochromic YOH films, they also had to assume multiple cation charge states and/or the presence of OH– complex ions in order to obtain charge balance.[15] Moreover, the relationship of the photochromic REOH thin films to the established group of stoichiometric LnOH bulk materials in terms of chemical composition and crystal structure remained unclear.

Here, we address these open questions by a combination of chemical composition analysis via ion beam methods, charge neutrality arguments, and optical characterization. We find that the photochromic REOH films are indeed oxyhydrides—clearly distinct from hydroxides—having a wide composition range along the MH3–M2O3 axis. This allows us to explain their properties in the framework of O2–/H– anion-disorder, linking the known (anion-ordered) structures of REH3 trihydrides, LnOH, and the RE2O3 sesquioxides.

Recently, we reported that metallic β-MH1.9+δ dihydride thin films with M = Y, Er, Dy, Gd can be prepared by direct current (DC) reactive MS of metal targets in an Ar/H2 atmosphere.[14] Above a certain material-dependent critical deposition pressure (p*), the films air-oxidize at room temperature (RT) to form stable semiconducting transparent photochromic MOH. In order to obtain samples with a wide range of chemical compositions, we have extended our standard synthesis procedure as follows: (i) Sc is sputtered as the RE element with the smallest ionic radius. (ii) A variable amount of O2 is added to the Ar/H2/O2 process gas (5N purity) resulting in a mixture containing (12.5–11.9) % of H2 and (0–4.7) % of O2. (iii) Pulsed DC plasma excitation (50 kHz, 90% duty cycle) is used to avoid arcing. (iv) Al capping layers (20 nm) were sputtered onto selected samples directly or after timed air exposure to prevent further oxidation. All samples were grown on unheated UV-grade fused silica (f-SiO2) and polished glassy carbon substrates (HTW Germany). Structural and optical properties were investigated by a combination of X-ray diffraction (XRD, Bruker D8 Discover) and photospectrometry (PerkinElmer Lambda 900). A custom-built optical-fiber-based in situ spectrometer (range: 230–1150 nm) with attached 385 nm LED light source and a time resolution of ∼1 s was employed to test the photochromic properties.

The MOH chemical composition is determined by a combination of Rutherford backscattering spectrometry (RBS) and ERD analysis at the 2 MV Van-de-Graaff accelerator at Helmholtz-Zentrum Dresden-Rossendorf (Dresden, Germany). RBS (ERD) measurements were performed with a 1.7 MeV 4He+ beam at 0° (70°) incidence and 160° (30°) scattering angle using semiconductor detectors with a solid angle of 3.3 msr (5.6 msr) and an energy resolution of ∼15 keV. The ERD detector was covered by a 6.6 μm Al stopper foil to discriminate recoiled H from forward scattered He. A special set of Sc, Y, and Gd dihydride reference samples was prepared by hydrogenating Pd-capped (20 nm) metal layers (150 nm) in a pressure cell at 1 bar of H2 at RT. Employing the hydrogenography method,[16] the change in optical transmittance is used to verify that all films were initially loaded to H/M ratios above the dihydride to trihydride phase transition, followed by unloading to dihydride in air after opening the pressure cell. Because of the tensile strain induced by hydrogen desorption,[17] we expect that these dehydrogenated layers have a H/M ratio corresponding to the lower end of the β-MH existence range, i.e., ScH1.68,[18] YH1.90,[17] and GdH1.80 (see Figures S1–S3 for XRD characterization). These dihydride samples turned out to be crucial to accurately calibrate the solid angle of the ERD detector. For each sample, the RBS and ERD spectra were fitted self-consistently using the SIMNRA[19] program and the SRIM2013 stopping power database.[20]

The result of this analysis is illustrated in Figure , showing the experimental ion beam spectra and corresponding simulations of a set of YOH films with different O/H ratios. The areal densities (atoms/cm2) of Y and O are obtained from RBS data, taking into account the Al or Pd capping layers. The overall fit accuracy benefits from the nearly background-free oxygen signal with good counting statistics, which is achieved by the use of carbon instead of f-SiO2 substrates. The H areal density is obtained from the ERD signal taking into account the energy loss and straggling of H recoils in the Al stopper foil as well as geometric straggling[21] due to the variation of the scattering angle across the detector area. The combined RBS and ERD analysis allows for a quantification of the element concentrations (in atom %) with an uncertainty below 2 atom %. The RBS spectrum of Pd-capped hydrogenated Y shows that the film contains no oxygen (O detection limit 1 atom %) except for a thin surface layer within the C substrate - probably a result of surface polishing. It is important to note here, that no other light elements, in particular F, could be detected in the MOH films discussed in this work. Together with the previously observed characteristic transmittance window (Figure ) and lattice constants,[14] this increases the confidence that our RE hydride samples are nearly ideal β-MH reference materials. Besides oxygen, fluorine is a common impurity in rare-earth metals - especially in Y.[22] In the past, we observed high concentrations of O (up to 7 atom %) and F (up to 13 atom %) in many commercial Y metal sputter targets of nominal 99.9% purity. Such high F concentrations complicate the chemical analysis and obscure subsequent interpretation of charge balance and electronic properties. Hence, we use high-purity Y targets supplied by Stanford Advanced Materials (United States) to avoid these problems.

Figure 1

Comparison of (a) RBS and (b) ERD experimental spectra (open circles) and corresponding SIMNRA simulations (red lines) obtained by fitting of each combined data set. A series of samples (150 nm thickness) with increasing oxygen content is shown: Pd-capped Y dihydride (black), Al-capped (blue), and uncapped Y oxyhydride (green). The broadening of the low-energy edge in the ERD spectrum is caused by thickness variations of the Al stopper foil.

Figure 5

Dependence of the MOH optical band gap on the hydrogen concentration where M = Sc, Y, Gd (colored circles). Reported band gap values of the binary compounds Y2O3,[24] YH2.7+δ,[28] and metallic YH1.9+δ are shown as colored hexagons. The metal–insulator transition upon oxidation of M2+ to M3+ with the hypothetical end point at MO0.5H2 is represented by the hatched region. Dashed lines are a guide to the eye and correspond to the charge-neutrality lines describing M3+ oxyhydrides and hydroxides in Figure . The fundamental electronic transitions of the binary end members of the Y oxyhydrides are illustrated.

The RBS and ERD simulations shown in Figure confirm that the MOH film composition is homogeneous throughout the depth of the films. We have observed chemical gradients in a few samples prepared at deposition pressures far above the critical values of p* = 0.3 Pa (Sc), 0.5 Pa (Y), and 0.7 Pa (Gd).[14] However, the interpretation of the composition–property relationships of such graded (and likely porous) films is rather ambiguous. Therefore, we have excluded these samples from further analysis.

Figure shows the results of the ion beam chemical composition analysis of our (Sc,Y,Gd)OH thin films in a generalized ternary M–O–H composition-phase diagram. This construction is based on the similar properties of the binary RE metal oxides and hydrides. At ambient conditions, the RE metals form stable M2O3 sesquioxides (except CeO2) with a cubic bixbyite (Ia3̅) structure (except La, Pr, and Nd).[23] All RE metals readily dissolve hydrogen, forming a random interstitial α-MH alloy where H occupies a fraction of the tetrahedral sites in the hexagonal close-packed (hcp) RE lattice. Upon hydrogenation, a phase transition to the metallic fcc (Fm3̅m) β-MH2 occurs for all REs. Further hydrogenation leads to a metal to insulator transition near H/M = 3. For most REs, this is accompanied by a structural phase transition to hcp γ-MH3 (except Sc, La, Pr, Nd). Moreover, the REs form hcp M(OH)3 hydroxides[25] and various MOOH oxy-hydroxide phases.

Figure 2

Ternary M–O–H chemical composition and phase diagram where M = Sc, Y ,La, Sm–Lu. Thick dashed lines indicate chemical compositions with the same charge state of cations (M2+, M3+, H+) and anions (O2–, H–). Hexagons mark the (ideal) stoichiometric compounds that have been reported earlier: (i) binary compounds: MH2 dihydride and MH3 trihydride in blue as well as M2O3 sesquioxides[23,24] and MO2 in red; (ii) ternary compounds: MOH oxyhydrides,[5−7] MOOH oxy-hydroxide, M(OH)3 hydroxide,[25] and the hypothetical MO0.5H2 in black. Colored circles show the chemical compositions of MOH thin films obtained from ion beam analysis. Note that the diameter of the data points corresponds to a composition uncertainty of approximately ±1 atom %. The region of photochromic materials is highlighted in gray.

Our results show that the chemical composition of the Sc, Y, and Gd-based thin films follow the same general trends in the M–O–H diagram. Two material groups can be clearly distinguished by considering the formal valencies and demanding charge neutrality (indicated by dashed lines). The RE oxyhydrides with a composition range of M3+O2–H3–2– (0.5 ≤ x ≤ 1.5) are found on the line connecting MH3 and M2O3. In particular, at x = 1 this includes the stoichiometric LnOH bulk compounds mentioned earlier. At x = 0.5, the data points deviate from the M3+O2–H3–2– line toward MH2 (hatched area). This is consistent with our previous work showing that as-deposited films sputtered in Ar/H2 are metallic dihydrides that oxidize in air to form transparent photochromic films.[14] We obtain gray opaque films with compositions between MH2 and MO0.5H2 by sputtering at pressures near p*. Their composition range in the ternary M–O–H diagram suggests that initially the air-oxidation proceeds via insertion of oxide ions into the MH2 lattice accompanied by oxidation of M2+ cations to M3+. The opaque character of these films is characteristic for intermixed metallic and dielectric phases indicating nucleation of semiconducting MO0.5H2 in β-MH2. Note that MO0.5H2 (x = 0.5) marks the (ideal) composition where the conduction band is fully depleted of electrons (all cations are in the M3+ state) resulting in a metal–insulator transition that is observed in terms of the appearance of an optical band gap. Increasing the pressure above p* or adding small amounts of O2 during deposition leads to the formation of transparent semiconducting oxyhydrides with compositions between MO0.5H2 and M2O3. In contrast, the RE hydroxides with the composition range M3+O2–H2+ (1.5 ≤ x ≤ 3) are located on the line connecting M2O3 and M(OH)3. These transparent films were obtained by further increasing the O2 fraction in the Ar/H2/O2 mixture during reactive sputtering.

XRD analysis confirms that all RE oxyhydride films in Figure have fcc unit cell symmetry, where the lattice constant is expanded by (1.0–2.8)% relative to the dihydride because of oxygen incorporation (see XRD analysis in the Supporting Information). Moreover, the lattice constants of sputtered dihydrides, oxyhydrides, and oxides show a systematic variation with the RE ion radius according to the lanthanide contraction effect.[14] By combining the structural similarities of the known stoichiometric compounds and the chemical composition results, we propose a generalized simplified structure model for the RE oxyhydrides as shown in Figure . It is based on the fcc (Fm3̅m) structure where the 4 lattice sites are occupied by the RE cations and the compounds differ only in the average anion occupation of the 8 tetrahedral and 4 octahedral interstices per unit cell. In this framework, the M2O3 oxides can be described by a random 6/8 occupation of tetrahedral sites by O2–, where anion-ordering would lead to a distortion of the fcc symmetry resulting in the bixbyite-M2O3 structure. The idealized structure of stoichiometric MH2 is an fcc (Fm3̅m) lattice where all 8 tetrahedral sites are occupied by H– ions. Additional filling of the 4 octahedral sites by H– leads to the idealized MH3 structure. While the trihydrides of La, Ce, Pr, and Nd remain in the fcc structure up to H/M ≈ 3, the most common hcp structure of the RE γ-MH3 can be interpreted as a distorted fcc lattice where the cubic (111) axis is parallel to the hcp c-axis. The structure of the stochiometric LnOH powder materials was previously described as a fcc (Fm3̅m) where the tetrahedral sites are randomly occupied by O2– and H– with a 1:1 ratio.[7] The corresponding anion-ordered structure is tetragonal (P4/nmm).[5]

Figure 3

Generalized crystal structure evolution of rare-earth oxyhydrides with different H–/M and O2–/M ratios based on the fcc-fluorite (Fm3̅m) structure motif. Anion sublattice disorder is visualized by multicolored spheres in terms of partial occupancy of tetrahedral (yellow polygon) and octahedral (green polygon) sites. The corresponding anion-ordered crystal structures of fcc-MH2, hcp-MH3, tetragonal MOH,[5,6] and bixbyite-M2O3[24] are given in parentheses. The arrows indicate metal–insulator transitions.

The resulting generalized picture of anion-disordered fcc symmetry as shown in Figure illustrates that in the M3+O2–H3–2–oxyhydrides the cation-to-anion ratio changes from 1:1.5 in M2O3 to 1:3 in MH3. This implies that the MOH composition (x = 1) marks a transition point regarding the occupation of tetrahedral and octahedral interstices. For x > 1, the octahedral sites are empty and additional structural tetrahedral vacancies are forming with increasing oxygen content up to a maximum amount of 25% in the M2O3 structure. In the case of x < 1, all tetrahedral sites are occupied and the octahedral sites are successively filled with hydride ions with increasing hydrogen content.

We observe photochromism of the oxyhydrides over a wide composition range for both x < 1 and x > 1, as shown by the highlighted gray area in Figure . However, the exact composition boundaries for photochromic MOH3–2 remain unclear. Regarding the lower boundary, we did not obtain any oxyhydrides with x < 0.5 by either postoxidation of metallic β-MH2 films or direct growth using oxygen-poor Ar/H2/O2 gas mixtures. This suggests that, in contrast to what was reported by You et al.,[12] H/M ratios above 2 in RE oxyhydride films cannot be achieved by reactive MS. This is because the typical H2 partial pressures of a few 10 mPa during sputtering are far below the equilibrium pressure of the MH2 → MH3 phase transition (pH ≈ 0.1 Pa)[26,27] and the addition of O2 to the process gas favors further dehydrogenation. Therefore, we prepared a γ-YH2.7+δ (x = 0) thin film by hydrogenation of a Pd-capped sputtered Y metal film at pH = 3.2 kPa in a vacuum cell mounted into our in situ spectrometer. This trihydride did not show photochromism at RT, whereas air-oxidized RE dihydride films with x ≈ 0.5 are photochromic. Hence, this value is marked as the O-poor boundary for photochromic RE oxyhydrides in Figure .

In order to discuss the O-rich boundary for photochromism, the effect of the anion ratio on optical properties has to be considered. Transmittance spectra of YOH films are presented in Figure . The dense YH film prepared at p = 0.3 Pa (i.e., far below the critical deposition pressure for Y) shows a transmittance window centered around 700 nm characteristic of metallic β-YH1.9+δ dihydride, which is caused by a combination of weak interband and free-electron absorption.[28] The RE oxyhydride and hydroxide films are transparent semiconductors. The optical band gap of Y oxyhydride increases continuously with increasing O2– content from (2.5 ± 0.1) eV at x ≈ 0.7 up to (4.9 ± 0.2) eV at x ≈ 1.4 (Figure ). An extrapolation of this nonlinear trend (dashed line) leads to good agreement with the band gaps of Y2O3 (5.6 eV) and YH3 (2.6 eV), which are determined by the O 2p → Y 3d and H 1s → Y 3d interband transition, respectively.[29] Therefore, the upper valence band of the RE oxyhydrides is likely formed by a mixture of occupied H 1s and O 2p states where the valence band maximum (VBM) shifts to lower energies with increasing O2– concentration because of the higher electronegativity of oxygen (χO = 3.44) compared to hydrogen (χH = 2.20). The resulting band gap widening of the Y oxyhydrides does not depend linearly on the anion composition. This “band gap bowing” effect also occurs in many other semiconducting multianion compounds such as oxysulfides,[30,31] oxynitrides,[32] and III–V materials.[33] Its origin can be traced back to the mismatch in atomic orbital energy and (an)ion size, which leads to coupling of electronic states affecting the relative positions of the VBM and the conduction band minimum (CBM) in random alloys.[34] It seems plausible that these established principles also apply to the electronic structure of the oxyhydrides. However, comprehensive studies on this topic are still lacking.

Figure 4

Transmittance spectra of Y-based ternary M–O–H films corresponding to Figure with different compositions illustrating the metal–insulator transition from the dihydride YH1.9+δ to the oxyhydride MOH3–2 phase followed by band gap widening with increasing O2–/H– ratio. The transmittance of a hydroxide-like film with a large H+ concentration of ∼30 atom % is shown for comparison. The optical band gap values as obtained from Tauc plots are given. The black dashed line is the transmittance of the bare fused silica substrate.

Figure shows that the band gaps of the Sc- and Gd-based oxyhydrides follow the same trend as YOH3–2, which is reasonable considering the similarity in band gaps of the oxides: Sc2O3 (∼5.8 eV)[35] and Gd2O3 (5.4 eV)[36] as well as the trihydride GdH3 (2.5 eV).[37] Slightly lower Eg values (−0.3 eV) for M = Sc, Gd are observed in oxyhydrides with low oxygen content (x ≈ 0.5). This is in agreement with our work[14] on MOH with M = Y, Gd, Dy, Er and indicates that the effect of different RE cations on Eg is relatively weak compared to anion-alloying. Moreover, all RE hydroxide thin films have large optical band gaps (Eg ≈ 5.5 eV) which are independent of the anion composition and similar in value to the sesquioxides, suggesting that the unoccupied H 1s orbitals do not interfere with the states at the VBM and CBM. Given the chemical similarity of the RE elements, we suppose that the anion-composition dependence of the optical band gap of most lanthanide oxyhydrides closely resembles the trend shown in Figure . Notable exceptions are expected for M = Ce, Pr, Eu, Tb where the M 4f orbitals are located within the O 2p → M 5d forbidden gap, resulting in a reduction of the M2O3 optical band gaps.[36,38]

Our ongoing photodarkening experiments with wavelength-dependent excitation indicate that photon energies larger than Eg are required to trigger photochromism in RE oxyhydrides. Using UV excitation by Hg vapor lamps with λ = 254 nm, we observe weak photochromism in YOH3–2oxyhydrides with band gaps up to 4.2 eV. According to Figure this corresponds to x = 1.3 (cH ≈ 15 atom %) which is marked as the oxygen-rich boundary for photochromism in Figure .

In conclusion, we have established a ternary RE–O–H composition-phase diagram demonstrating that (i) the previously reported photochromic Y-based thin films[10,14] are O2–/H– multianion compounds which are members of the REOH3–2oxyhydrides material class including Sc and (most of) the lanthanides and that (ii) an anion-disordered fcc lattice model can be used to describe the crystal structure of the RE oxyhydrides linking the known REH3, RE2O3, and LnOH[5−7] phases. These oxyhydride films are photochromic over nearly their entire composition range of 0.5 ≤ x ≤ 1.5, which implies that the presence of both oxide and hydride ions is crucial for the photochromic effect at ambient conditions. We notice a general trend toward decreased photochromic contrast and faster bleaching kinetics with increasing O2–/H– ratio—in agreement with the report of Moldarev et al.[15] On the basis of our results, we expect that the stoichiometric bulk LnOH compounds (x = 1) should be photochromic as well. Moreover, the RE oxyhydride optical band gap can be adjusted over a wide range by controlling the O2–/H– anion ratio in order to adapt the photochromic response for applications such as energy-saving smart windows and adaptive eyewear. Further systematic investigation of the electronic structure and defect formation is required to clarify the origin and physical limitations of the photochromic effect in the REOH materials.