Theoretical design of highly active SrTiO₃-based photocatalysts by a codoping scheme towards solar energy utilization for hydrogen production

Abstract

SrTiO₃ is a promising photocatalyst for the production of hydrogen from water splitting under solar light. Cr doping is an effective treatment for adjusting its absorption edge to the visible-light range, although the performance of Cr-doped SrTiO₃ is strongly affected by the oxidation number of the Cr ions. In this study, we theoretically predict that elevating the Fermi level, i.e., n-type carrier doping in SrTiO₃, can stabilize the desirable oxidation number of chromium (Cr³⁺), contributing to a higher activity for H₂ evolution. Our computational results, based on hybrid density-functional calculations, reveal that such an n-type condition is realized by substituting group-V metals (Ta, Sb, and Nb), group-III metals (La and Y), and fluorine atoms for the Ti, Sr, and O sites in SrTiO₃, respectively. From our systematic study of the capability of each dopant, we conclude that La is the most effective donor for stabilizing Cr³⁺. This prediction is successfully evidenced by experiments showing that the La and Cr codoped SrTiO₃ dramatically increases the amount of H₂ gas evolved from water under visible-light irradiation, which demonstrates that our guiding principle based on Fermi level tuning by the codoping scheme is valid for the design of advanced photocatalysts.

---

Introduction

There has recently been much interest in the potential use of photocatalytic materials in the production of clean-energy fuels and the reduction of environmental pollutants.¹ Since the discovery of the Honda–Fujishima effect,² TiO₂ has been extensively studied and applied to various commercial products, mainly for decomposing harmful chemicals or antifouling coatings. However, its application to fuel production is still limited due to the insufficient efficiency of overall water splitting for H₂ evolution.^2,3 It is therefore a challenge to find a highly active photocatalyst to meet today's demands for alternative energy resources. The perovskite strontium titanate (SrTiO₃ or STO) is a promising photocatalyst because of several advantages, such as high chemical stability and abundance of the constituent elements. The valence-band and conduction-band edges of STO are favorably situated for overall water splitting:^4–7 the valence-band maximum (VBM) is much lower than the oxidation potential of water, and the conduction band minimum (CBM) is ∼0.8 eV (at pH = 0) higher than the standard hydrogen electrode potential (SHE).⁷ The sufficiently large band offset of the CBM with respect to SHE, which is much larger than that of anatase TiO₂ (0.1–0.2 eV), is a great benefit for the chemical reactions of photoexcited electrons with protons on the surface: 2H⁺ + 2e⁻ → H₂. Nevertheless, the photoabsorption range of STO is restricted to ultra-violet light because of its relatively wide band gap (3.25 eV), and the majority of sunlight, i.e. visible light is hardly utilized in photocatalytic reactions on pristine STO. One approach to overcome this problem is to induce a red-shift of the absorption edge while keeping the position of the CBM unchanged, in order to maintain its potential for proton reduction. Chromium (Cr) is of particular interest as an effective dopant for activating visible-light absorption in wide-gap semiconductor oxides.^8–14 In fact, Cr-doped STO can work continuously and stably under visible light for more than 500 hours, and therefore possesses great potential for future solar hydrogen production.^10,12 Importantly, Cr-doped STO is active only when the Cr ions possess a lower oxidation number (Cr³⁺), while it is inactive for higher oxidation states of Cr.^11,12 Thus, the stabilization of Cr³⁺ is a key strategy for enhancing the photocatalytic performance of Cr-doped photocatalysts.

Recently, we have theoretically shown that the desirable oxidation state Cr³⁺ predominates in n-type STO,¹⁵ namely the condition at which the Fermi level (ε_F) is located near the CBM. The negatively charged substitutional Cr at the titanium site (Cr⁻_Ti), which was identified as the source of Cr³⁺ in Cr-doped STO, gives rise to fully occupied states above the VBM. Fig. 1(a) shows schematically the electronic structures associated with substitutional Cr at a Ti site with two different charge states (Cr⁰_Ti and Cr⁻_Ti). In a low ε_F condition, the neutral charge state Cr⁰_Ti, which corresponds to a high oxidation state of the Cr ion, is stabilized and photo-exited electrons are likely to be trapped in unoccupied gap states, degrading photocatalytic reactions. By contrast, shifting ε_F towards the conduction band eliminates the trapping centers due to the realization of the desirable charge state Cr⁻_Ti. Our computational results revealed that the occupied gap states associated with Cr⁻_Ti play an important role both in the photoresponse under visible light and in the elimination of electron trapping centers.¹⁵

Fig. 1 (a) Schematic illustration of the effect of shifting ε_F towards the conduction band on the stabilization of a low oxidation state of the Cr ion associated with Cr⁻_Ti in Cr-doped STO. The energy levels of the in-gap Cr d states are shown. The arrows indicate the electron occupation corresponding to the associated charge state. (b) and (c) Illustrations of the conditions under which various Cr-related defects are stable, derived from the formation energy, with respect to the chemical potentials of Sr and Ti for (b) ε_F = 0.3 eV (p-type STO) and (c) ε_F = 3.0 eV (n-type STO). For the growth of STO under thermal equilibrium, μ_Sr, μ_Ti, and μ_O must lie in the dark-shaded region.

Previously, some experimental reports showed that the photocatalytic performance of Cr-doped STO is improved by codoping with antimony (Sb),¹⁰ tantalum (Ta),¹¹ or niobium (Nb),¹¹ or by pretreatment in a reducing H₂ atmosphere.¹¹ In all of these treatments, the concentration of Cr³⁺ is expected to increase because of the electron carrier doping upon incorporation of these elements into STO. Although the effect of Cr³⁺ has been clearly addressed in our previous work,¹⁵ knowledge of the role played by codopants is still lacking. Since STO is also a promising semiconductor for oxide electronics,^16,17 carrier doping of this technologically important material is of great interest to a wide range of researchers. In particular, n-type doping of this wide band-gap semiconductor is a key target in many different fields of research,^18–21 and the most suitable donor for STO remains to be identified.

Here, we systematically investigate the effects of doping STO with the group-V metals Ta, Sb, Nb, and vanadium (V), the group-III metals lanthanum (La) and yttrium (Y), and the halogen fluorine (F), in order to theoretically identify the most effective donor in STO and the best candidate codopant for stabilizing Cr³⁺ in Cr-doped STO.

The above-mentioned metal dopants can be readily incorporated into the STO lattice thanks to the similarity of their ionic radii to those of the host elements. For instance, the ionic radii of La³⁺ (1.16 Å) and Nb⁵⁺ (0.64 Å) are very close to those of Sr²⁺ (1.26 Å) and Ti⁴⁺ (0.61 Å), respectively, that is, the group-V metals preferentially occupy the Ti site rather than the Sr site, while the group-III metals tend to occupy the Sr sites. Our theory, based on hybrid density-functional calculations, predicts that La is the best donor among the candidate elements in terms of electron doping ability and the ability to stabilize Cr³⁺. The prediction is successfully validated by our experiments, which clearly show that La and Cr codoped STO exhibits the best performance for hydrogen evolution from water splitting under visible-light irradiation among the candidate materials examined in this work.

Methods of calculation

Our calculations were based on density-functional theory with the hybrid-functional of Heyd, Scuseria, and Ernzerhof (HSE),²² which was implemented in the Vienna Ab initio Simulation Package (VASP) code.²³ In the HSE approach, the Coulomb potential in the exchange energy of the Perdew, Burke, and Ernzerhof (PBE)²⁴ functional is divided into short-range and long-range parts with a screening length of 10 Å. The long-range part and correlation potential are represented by the PBE functional, while the short-range part is mixed with the non-local Hartree–Fock (HF) exchange. We reproduced the experimental band gap of STO (3.25 eV) by using the HF mixing parameter of 28%. The HSE functional has been shown to improve the band gap of semiconductors and give a good description of the electronic structure and formation energies of defects and impurities.^25,26 The cutoff energy of 400 eV was used for the plane-wave basis set. The defect calculations were performed using a (3 × 3 × 2) 90-atom supercell and a 2 × 2 × 2 grid of Monkhorst-Pack special k points for the Brillouin zone integration. The effects of spin polarization were included in all calculations involving unpaired electrons.

The solubility of a donor-type impurity must be sufficiently high under a given sample growth condition. The formation energy is the key quantity that must be calculated to investigate the stability of a dopant, and is given by the following equation:

E^f(X^q_Ti) = E_tot(X^q_Ti) − E_tot(STO) + μ_Ti − μ_X + qε_F (1)

Here, E_tot(X^q_Ti) is the total energy of a STO supercell containing an impurity X (X = Ta, Sb, Nb, or V) at the Ti site with q charge state, and was obtained by our density-functional calculations with a 90-atom supercell in the same way as that in case of Cr doping.¹⁵E_tot(STO) is the total energy of perfect STO, obtained using the same supercell. ε_F is the Fermi level referenced to the bulk VBM, which corresponds to the energy of the electron reservoir for charged impurities. A correction was applied to ε_F to account for the difference in electrostatic potential between the bulk and the defect-supercell.²⁷ The impurity X was taken from a reservoir with energy μ_X, and the substituted Ti atom was placed in a reservoir with energy μ_Ti. The energies μ_X and μ_Ti thus represent the chemical potentials, and depend on the experimental growth or annealing conditions. The procedure for estimating μ_X will be addressed at great length later. When STO is grown under equilibrium conditions, the chemical potentials of the constituent atoms, which are referenced to the values of their elemental forms, must satisfy the equilibrium condition μ_Sr + μ_Ti + 3μ_O = ΔH_f(STO) = −16.2 eV, where ΔH_f(STO) is the formation enthalpy of STO. The formation of unwanted compounds, such as SrO, can be avoided by imposing the condition μ_Sr + μ_O < μ_SrO. Similar bounds were imposed for SrO₂, TiO, TiO₂, and Ti₂O₃. Analogous forms of eqn (1) can be written for the substitutional La_Sr, Y_Sr, and F_O. The above-mentioned conditions yield the range of chemical potentials that allow the stable growth of STO, which can be represented by a domain in the two-dimensional space defined by μ_Sr and μ_Ti (dark-shaded areas in Fig. 1(b) and 1(c)).

Results and discussion

Here we initially examine the effect of an upward shift of ε_F on the stabilization of Cr⁻_Ti. Fig. 1(b) and 1(c) illustrate the Cr impurities that give the lowest formation energy at each chemical potential condition of (μ_Sr, μ_Ti):Cr_Ti, Cr substituted for Sr (Cr_Sr), and Cr at an interstitial site (Cr_i). These are shown in Fig. 1(b) for ε_F = 0.3 eV (p-type STO) and in Fig. 1(c) for ε_F = 3.0 eV (n-type STO). Here, μ_O is uniquely determined at each point (μ_Sr, μ_Ti) through the above-mentioned equilibrium condition. For p-type STO, Cr²⁺_Ti and Cr⁴⁺_i are the dominant species under the stable growth condition, as illustrated by the dark-shaded domain in Fig. 1(b). By contrast, for n-type STO where ε_F is located near the CBM, the dark-shaded domain is associated only with Cr⁻_Ti (Fig. 1(c)). This suggests that a higher concentration of Cr⁻_Ti, which is advantageous for photocatalytic activity, can be obtained under n-type conditions. However, such an n-type condition is not achieved without external doping, as addressed in the following. In order to determine realistic values of the chemical potentials, μ_O was estimated by including contributions from enthalpy and entropy.²⁸μ_O is expressed as

μ_O(T, P₀) = {(H₀ + ΔH(T)) + T(S₀ + ΔS(T))}/2 (2)

where the tabulated values for standard O₂ enthalpy at T₀ = 298 K and P₀ = 1 atm (H₀ = 8.7 kJ mol⁻¹) and entropy (S₀ = 205 J mol⁻¹) are used, ΔH(T) = C_P(T − T₀) and ΔS(T) = C_Pln(T/T₀). Using the ideal gas law for T ≥ 298 K, C_P = 3.5k_B is used for the heat-capacity per diatomic molecule. The experimental growth conditions of STO described in ref. 12, at a calcination temperature of 1100 °C and a pressure of 1 atm, yields μ_O = −1.57 eV. The equilibrium growth condition is now given as μ_Sr + μ_Ti = −16.2 + 3 × 1.57 = −11.50 eV, which is represented by the dotted lines in Fig. 1(b) and 1(c). The chemical potentials μ_Ti = −6.05 eV and μ_Sr = −5.45 eV (represented by solid circles in Fig. 1(b) and 1(c)) then give the least favorable condition for Cr_Ti, at the boundary of the shaded domain associated with the stable growth condition.

Here, we address how to estimate the chemical potential of the dopants μ_X, which is an important factor in determining the impurity solubility. Fig. 2 illustrates the thermodynamically stable phases for the relevant transition metal (TM) compounds over the range of possible growth conditions between the O-poor limit (μ_O = −5.27 eV) and the extreme O-rich limit (μ_O = 0). μ_x can be derived from the formation enthalpy of its limiting phase which is subject to an upper bound. For instance, for the realistic growth condition μ_O = −1.57 eV (dashed-dotted line in Fig. 2), μ_Cr is determined by the formation enthalpy of Cr₂O₃, that is, 2μ_Cr + 3μ_O ≤ ΔH_f(Cr₂O₃), where the upper bound of μ_Cr = (μ_Cr2O3 − 3μ_O)/2: if μ_Cr is pushed higher than the upper bound, Cr₂O₃ will form in addition to the Cr incorporated in the STO lattice. Using a similar procedure, the upper bound for μ_X where X = Ta, Sb, Nb, V, La, and Y can be determined.

Fig. 2 The relationship between the transition metal (TM) chemical potentials μ_TM and the O chemical potential μ_O, as derived from the equilibrium growth of the host: xμ_TM + yμ_O = ΔH_f(TM_xO_y). Only the μ_TM that gives the upper bound (the lowest μ_TM) for the respective μ_O are shown. The dotted-, dashed-, and dashed-dotted vertical lines denote the O-poor (μ_O = −5.42 eV), O-rich (μ_O = 0), and realistic O-rich conditions (μ_O = −1.57 eV), respectively.

Fig. 3 shows the calculated formation energies of the Cr-related defects for these realistic chemical potentials, indicating that Cr_Ti is a more stable form of chromium than the other forms (Cr_Sr or Cr_i) for a wide Fermi level range, as we discussed in our previous paper.¹⁵ Under a thermodynamic equilibrium condition, the charge neutrality must be satisfied when the concentrations of the positively charged Cr⁺_Ti and the negatively charged Cr⁻_Ti are balanced. This determines the pinning level of ε_F, shown as an arrow in Fig. 3. Here, the neutral Cr⁰_Ti, which is likely to act as an electron trapping center,¹⁵ is the predominant charge state at the pinned ε_F. To stabilize the favorable charge state, Cr⁻_Ti, the pinning level must be elevated. This motivated us to seek donor dopants that can effectively shift ε_F upwards and stabilize Cr⁻_Ti without the formation of electron-trapping centers.

Fig. 3 Calculated formation energy as a function of ε_F for Cr-related defects under a realistic growth condition (μ_O = −1.57 eV). The arrow indicates the estimated pinned ε_F. The slope of each line indicates the charge state. The transition levels between different charge states are denoted as kinks in each plot.

First, we explore the group-V elements (V, Nb, and Ta) and another pentavalent element Sb (we include Sb in “group-V” in the following discussion). As the ionic radii of all four dopants are closer to that of Ti than that of Sr, preferential substitution at the Ti site is expected. This is supported, for example, by the fact that the calculated formation energy of Ta_Sr is ∼7 eV higher than that of Ta_Ti. We found that Sb_Ti, Ta_Ti, and Nb_Ti do not introduce in-gap states when positively charged (Sb⁺_Ti, Ta⁺_Ti, and Nb⁺_Ti). For the neutral charge state (Sb⁰_Ti, Ta⁰_Ti, and Nb⁰_Ti), an extra electron occupies the bottom of the conduction band, which consists of an extended state resulting from the perturbation of the Ti-d states of the host band induced by the impurity. Fig. 4(a) shows the formation energies of the substitutional group-V impurities, along with that of Cr_Ti. We found that the positive charge state is energetically stable for Sb_Ti, Ta_Ti, and Nb_Ti over the entire range of ε_F positions in the band gap. The transition levels at which the stable charge state changes from 1+ to 0, ε(+/0), were found to be located just above the CBM for Sb_Ti, Ta_Ti, and Nb_Ti (not shown in Fig. 4), indicating that they act as shallow donors in STO. By contrast, V_Ti creates deep levels in the band gap as characterized by the transition levels ε(+/0), ε(0/−), and ε(−/2−). The in-gap states originate from the vanadium t_2g orbitals and are likely to give rise to carrier-trapping centers.

Fig. 4 Calculated formation energies as a function of ε_F for (a) group-V codopants (Nb, Ta, Sb, and V), (b) group-III codopants (La and Y), and (c) F, as well as Cr_Ti in each case. All formation energies were calculated for μ_O = −1.57 eV.

For the group-III codopants, we focused on the Sr site for the substitution, because the calculated formation energies showed that La preferably substitutes Sr rather than Ti. The formation energies of La_Sr and Y_Sr shown in Fig. 4(b) indicate that they act as shallow donors. Without forming any impurity states in the band gap, La_Sr and Y_Sr are unlikely to act as trapping centers for photoexcited electrons. Fig. 4(c) also shows that F_O is a shallow donor in STO. The ionic radius of F⁻ (1.33 Å) is slightly smaller than that of O²⁻ (1.40 Å), suggesting that O²⁻ is readily substituted by F⁻. In a similar fashion to the other shallow donors, the extra electron of F⁰_O was found to occupy the extended state located at the bottom of the conduction band, whereas F⁺_O does not form any in-gap states. We note that interstitial F (F_i) is energetically unstable, having a much higher formation energy than F_O. The density of states (DOS) for Sb_Ti, Ta_Ti, Nb_Ti, La_Sr, Y_Sr, and F_O in the positive charge state, showing that they do not introduce in-gap states, are shown in the ESI.†

We now discuss the capability of the codopants to stabilize Cr³⁺ and enhance the photocatalytic activity of STO. The first factor to consider is that the codopants should not cause trapping centers for photoexcited electrons. This is true for the Ta, Nb, Sb, Y, La, and F impurities that are incorporated into STO (based on the assumption that the dominant forms of the Ta, Nb, Sb, Y, La, and F impurities are Ta_Ti, Nb_Ti, Sb_Ti, Y_Sr, La_Sr, and F_O, respectively). By contrast, the incorporated V impurity creates deep states in the gap and should therefore be excluded from the candidate codopants. The second factor to consider when choosing suitable codopants is their solubility. The donor impurities are activated only when their solubility in the host material is sufficiently high. From Fig. 4, it is clear that La_Sr has the lowest formation energy (highest solubility) among the examined donors under the realistic growth condition corresponding to μ_O = −1.57 eV, which determines the upper bound of the chemical potential of each dopant as discussed before; Ta_Ti, Nb_Ti, V_Ti, Sb_Ti, Y_Sr, and La_Sr are limited by the formation of the binary metal oxides Ta₂O₅, Nb₂O₅, V₂O₅, Sb₂O₃, Y₂O₃, and La₂O₃, respectively (see Fig. 2). The solubility of F_O is limited by the formation of SrF₂. In other words, the chemical potential of a dopant must be carefully estimated from the limiting phase, since it significantly affects the solubility of the dopant.

In Cr-doped STO, the pinned Fermi level (ε_F = 0.92 eV) is relatively low, and the majority of the Cr_Ti is present in the neutral charge state (Cr⁰_Ti), as mentioned above. Under this condition, the photocatalytic activity of Cr-doped STO is suppressed.¹⁵ Because the photocatalytic activity is triggered when Cr⁻_Ti is stabilized, an upward shift of the pinned ε_F should improve the performance of the material. This is made feasible by incorporating a shallow donor that has a sufficiently low formation energy in the positive charge state (for example, La⁺_Sr), thus suppressing the formation of Cr⁰_Ti. Here, ε_F becomes pinned at a higher level (ε_F = 1.95 eV for La_Sr) corresponding to the intersection of the lines for La⁺_Sr (positive slope) and Cr⁻_Ti (negative slope) in Fig. 4(b), as a consequence of the charge balance. The same method was used to determine the pinning levels of ε_F for STO codoped with Cr⁻_Ti and the other dopants considered here. The donor impurity with a lower formation energy gives a higher position of the pinning level. The position of the pinning level with respect to VBM significantly affects the relative concentrations of Cr⁰_Ti and Cr⁻_Ti, as shown in Fig. 5.

Fig. 5 Calculated relative concentration of Cr_Ti defects, derived from hypothetical formation energies, as a function of Fermi level (ε_F). The pinned ε_F results from the charge-neutrality condition for codoped (circles) and non-codoped (square) Cr-doped STO.

Under thermal equilibrium, the concentration of Cr^q_Ti can be expressed as N_sitesexp(−E^f(Cr^q_Ti)/k_BT), where N_sites (∼1.67 × 10²² cm⁻³) is the number of available Ti sites in STO that can be substituted by Cr atoms. The negative formation energy of Cr_Ti, shown in Fig. 3, gives a concentration that is too high for an impurity, indicating that the actual μ_Cr is much lower than the upper bound estimated by our calculations; to obtain a positive value of E^f(Cr_Ti), the actual μ_Cr should be at least 0.9 eV lower than the value of μ_Cr taken from Cr₂O₃. Therefore, in order to estimate the relative concentrations of Cr⁰_Ti and Cr⁻_Ti for different values of the pinned ε_F (Fig. 5), we consistently added a constant value to the formation energies of the impurities shown in Fig. 4 such that E^f(Cr_Ti) became positive over the entire range of ε_F. Without the incorporation of the codopant, Cr⁰_Ti remains the predominant species (solid square in Fig. 5). Codoping successfully shifts the pinning level towards the CBM, and the majority of Cr_Ti is replaced by the favorable charge state Cr⁻_Ti at the ε_F for Sb_Ti substitution. The concentration of Cr⁻_Ti scales exponentially as a function of the pinned ε_F; we found the best and worst dopants for increasing the concentration of Cr⁻_Ti to be La and Y, respectively.

We also investigated the possibility of the formation of a neutral defect-complex by coupling Cr⁻_Ti with a codopant, which can be computed from the binding energy,

E_b = E^f(Cr⁻_Ti) + E^f(D⁺) − E^f((Cr_Ti–D)⁰). (3)

Here, E^f(D⁺) is the formation energy of the donor impurity, such as Sb⁺_Ti, and E^f((Cr_Ti–D)⁰) is the formation energy of the Cr_Ti–D complex, in which Cr⁻_Ti and D⁺ occupy adjacent sites. A positive value of E_b indicates that the Cr_Ti–D complex is lower in energy than the isolated Cr⁻_Ti and D⁺ constituents. We found that the calculated binding energies for all the neutral complexes were relatively small (the highest value was ∼0.3 eV). For instance, E_b((Cr_Ti–La_Sr)⁰) = 0.15 eV, indicating that Cr⁻_Ti and La⁺_Sr experience only a weak driving force to bind to each other. Once Cr and La are incorporated into a sample, under thermal equilibrium (high temperature) they will be confined to their energetically favorable sites (the Ti and Sr sites, respectively); only a small proportion of these dopants will form a complex. This can be clarified as follows. Under equilibrium growth conditions, we can estimate the proportion of the formed Cr_Ti–La_Sr complex by the following detailed balance:where [Cr⁻_Ti], [La⁺_Sr], and [(Cr_Ti–La_Sr)⁰] are the concentrations of Cr⁻_Ti, La⁺_Sr, and (Cr_Ti–La_Sr)⁰, respectively. N_config is the number of configurations of the complex that can be formed, which is 8 for the Cr_Ti–La_Sr complex. Using the temperature of 1373 K and [La⁺_Sr] ∼ 2.9 × 10¹⁷ cm⁻³, determined from the charge-neutrality condition described above, we obtained [(Cr_Ti–La_Sr)⁰]/[Cr⁻_Ti] ∼ 5.0 × 10⁻⁴, which means that only 0.05% of Cr⁻_Ti participates in the formation of the complex. During the cooling process, Cr⁻_Ti and La⁺_Sr remain on their sites if they are assumed to be immobile, and thus the proportion of dopants forming a stable complex remains small. Notably, the calculated binding energies of the complexes between the dopants in STO are much smaller than those in TiO₂.²⁹ Our results show that the incorporated donor impurities in Cr-doped STO are unlikely to form complexes with Cr_Ti, and thus can be considered isolated. The photoabsorption properties of the codoped samples can hence still mainly be attributed to Cr⁻_Ti.

Previous experimental results have shown that codoping Cr-doped STO with Ta,¹¹ Sb,¹⁰ or Nb,¹¹ significantly enhances the photocatalytic activity, whereas codoping with V¹¹ decreases the activity, in good agreement with our calculations. It has also been found that Cr³⁺ is stabilized under O-deficient conditions,³⁰ which enhances the n-type condition. Our results are not only applicable to the stabilization of active Cr ions in the Cr-doped STO photocatalyst, but can also provide a route towards the fabrication of n-type STO. Our study clearly shows that n-type STO can be realized by doping with Nb or La, both of which have high solubility under the realistic growth condition discussed above. Indeed, it has long been realized that doping STO with Nb or La induces a transition from an insulator to an n-type semiconductor.^31–33 The dopants Ta, Sb, and Y also act as shallow donors, although their solubility in STO is limited due to their relatively high formation energies compared to those of Nb and La (see Fig. 4). As in other oxides, such as ZnO,³⁴ F_O is a shallow donor in STO and is highly soluble under O-deficient conditions. By contrast, F_i is a deep acceptor with a high formation energy, and thus cannot contribute to p-type conductivity in STO. Theoretical studies of electron-doped STO have also been carried out and are consistent with our results.^19,35

Now we demonstrate the photocatalytic performance of Cr-doped STO codoped with La, Nb, Sb, and Y, which are theoretically predicted to be the best to the worst codopants for increasing Cr⁻_Ti, in that order. The detail of experiments can be found in the ESI.† We found that all codoped samples can absorb light in the visible region, as shown in Fig. S1.†Fig. 6 shows the photocatalytic activities of Cr-doped STO codoped with La, Nb, Sb, and Y for H₂ evolution from water splitting under visible light irradiation. Cr-doped STO codoped with La (La–Cr:STO) showed the highest activity, in excellent agreement with our theoretical prediction. However, the H₂ evolution activity of Nb–Cr:STO is the lowest, which is inconsistent with our prediction. This discrepancy might result from the presence of other types of Nb-related defects, such as Nb interstitials or its complexes, and the carrier concentration of doped electrons is not sufficiently high to shift the Fermi level up to the expected position under our growth condition. However, the identification of the most stable Nb-related defect in STO is beyond the scope of this work.

Fig. 6 Photocatalytic H₂ evolution from water splitting of Cr and D codoped SrTiO₃ under visible light irradiation, where D = La, Sb, Y, and Nb. All codoped samples were prepared using 5 mol% of Cr and 5 mol% of D.

In contrast, although the H₂ evolution activity of Sb–Cr:STO was much lower than that of La–Cr:STO, it is still higher than that of Nb–Cr:STO. This could be attributed to the fact that Sb does not have partially filled d states, and thus its related defects, if formed, are unlikely to cause carrier recombination. Y–Cr:STO exhibits H₂ evolution activity comparable to that of the Sb–Cr:STO sample at the beginning of the photochemical reaction. However, the activity gradually decreases with longer irradiation time (Fig. 6). This is qualitatively consistent with our prediction that the activity of Y–Cr:STO is lower than that of Sb–Cr:STO. As mentioned above, the La–Cr:STO possesses the highest amount of Cr⁻_Ti compared to the other codoped samples and thus is a highly active photocatalyst under visible-light irradiation; recently, via a synthesis improvement and a reaction-environment modulation, La–Cr:STO was found to exhibit a high quantum yield of 25.6% (λ = 425 ± 12 nm) for H₂ evolution,³⁶ which lends strong support to the findings of our study herein. Next we discuss the effect of donor impurities on the oxidation state of Cr in the codoped samples.

X-ray photoelectron spectroscopy (XPS) measurements were performed to investigate the oxidation state of doped Cr. Fig. 7 shows the XPS spectra of Cr 2p of the codoped samples. All the codoped samples showed a peak at about 576.4 eV, which is very close to the peak assigned to Cr³⁺ at 576.7 eV, however, some of our samples showed the peak assigned to Cr⁶⁺ at 579.2 eV.¹⁰ The relative intensity of XPS allows us to understand the effect of the codopant on the oxidation state of doped Cr as follows. La–Cr:STO and Nb–Cr:STO showed a high peak intensity of Cr³⁺, while the peaks assigned to Cr⁶⁺ are almost negligible, in agreement with our calculations that codoping with either La or Nb should efficiently increase the concentration of Cr³⁺. However, based on our experimental result, codoping with Nb does not enhance the photocatalytic activity even though the concentration of Cr³⁺ is high, indicating that codoping with Nb possibly leads to the formation of other types of defects which cause carrier recombination. Sb–Cr:STO showed a relatively low peak intensity of Cr³⁺, while a small shoulder of the peak assigned to Cr⁶⁺ was observed. Y–Cr:STO clearly showed a high peak intensity of Cr⁶⁺ compared to that of Cr³⁺, in agreement with our theory (Fig. 5) and experiment (Fig. 6).

Fig. 7 X-ray photoelectron spectra of Cr 2p of (a) La–Cr, (b) Sb–Cr, (c) Y–Cr, and (d) Nb–Cr codoped SrTiO₃. The peaks assigned to Cr³⁺ and Cr⁶⁺ are indicated by dashed and dotted lines, respectively.

Conclusions

In summary, we have systematically studied the capability of several dopants (V, Nb, Ta, La, Y, and F) for the enhancement of the photocatalytic performance of Cr-doped SrTiO₃. Based on our density-functional calculations for the stability of the donor-type impurities, La substituted for Sr (La_Sr) was predicted to be the best donor for shifting the Fermi level towards the conduction band of SrTiO₃. The n-type condition is advantageous for increasing the concentration of negatively charged Cr⁻_Ti (Cr³⁺), which is the key oxidation state for activating photocatalytic reactions. The prediction from theory has been successfully evidenced by our experiments showing that La and Cr codoped SrTiO₃ shows markedly high activity for H₂ evolution from water splitting under visible-light irradiation compared to the other candidate dopants (Nb–Cr, Sb–Cr, or Y–Cr). Furthermore, our computational results show that the other substitutional defects, Sb_Ti, Ta_Ti, Nb_Ti, Y_Sr, and F_O, are also shallow donors, although they are less soluble in SrTiO₃ than La_Sr is, whereas V_Ti creates deep states in the band gap and is unable to produce the n-type condition. By combining theoretical and experimental investigations, our guiding principle of Fermi level tuning by a codoping scheme was found to be reliable for the design of Cr-doped SrTiO₃, and represents a significant step forward in the development of advanced photocatalysts.

Acknowledgements

This work was partially supported by the Japan Science and Technology Agency (JST) Precursory Research for Embryonic Science and Technology (PRESTO) program, and the World Premier International Research Center Initiative on Materials Nanoarchitectonics (MANA), MEXT. P.R. and N.U. thank the High-Performance Computing Center of the California NanoSystems Institute (CNSI, UCSB) and the Synchrotron Light Research Institute (SLRI, Thailand) for their hospitality. P.R. thanks S. Limpijumnong for the advice.

Notes and references

1. H. Tong, S. Ouyang, Y. Bi, N. Umezawa, M. Oshikiri and J. Ye, Adv. Mater., 2012, 24, 229–251.
2. A. Fujishima and K. Honda, Nature, 1972, 238, 37–38.
3. A. L. Linsebigler, G. Lu and J. T. Yates, Chem. Rev., 1995, 95, 735–758.
4. M. S. Wrighton, A. B. Ellis, P. T. Wolczanski, D. L. Morse, H. B. Abrahamson and D. S. Ginley, J. Am. Chem. Soc., 1976, 98, 2774–2779.
5. A. K. Ghosh and H. P. Maruska, J. Electrochem. Soc., 1977, 124, 1516–1522.
6. K. Domen, S. Naito, T. Onishi, K. Tamaru and M. Soma, J. Phys. Chem., 1982, 86, 3657–3661.
7. Y. Xu and M. A. A. Schoonen, Am. Mineral., 2000, 85, 543–556.
8. R. U. E. ’t Lam, L. G. J. de Haart, A. W. Wiersma, G. Blasse, A. H. A. Tinnemans and A. Mackor, Mater. Res. Bull., 1981, 16, 1593–1600.
9. M. Matsumura, M. Hiramoto and H. Tsubomura, J. Electrochem. Soc., 1983, 130, 326–330.
10. H. Kato and A. Kudo, J. Phys. Chem. B, 2002, 106, 5029–5034.
11. T. Ishii, H. Kato and A. Kudo, J. Photochem. Photobiol., A, 2004, 163, 181–186.
12. D. Wang, J. Ye, T. Kako and T. Kimura, J. Phys. Chem. B, 2006, 110, 15824–15830.
13. J. W. Liu, G. Chen, Z. H. Li and Z. G. Zhang, J. Solid State Chem., 2006, 179, 3704–3708.
14. W. Wei, Y. Dai, H. Jin and B. Huang, J. Phys. D: Appl. Phys., 2009, 42, 055401.
15. P. Reunchan, N. Umezawa, S. Ouyang and J. Ye, Phys. Chem. Chem. Phys., 2012, 14, 1876–1880.
16. A. Ohtomo and H. Y. Hwang, Nature, 2004, 427, 423–426.
17. D. Kan, T. Terashima, R. Kanda, A. Masuno, K. Tanaka, S. Chu, H. Kan, A. Ishizumi, Y. Kanemitsu, Y. Shimakawa and M. Takano, Nat. Mater., 2005, 4, 816–819.
18. R. Astala and P. D. Bristowe, J. Phys.: Condens. Matter, 2002, 14, L149.
19. W. Luo, W. Duan, S. G. Louie and M. L. Cohen, Phys. Rev. B: Condens. Matter Mater. Phys., 2004, 70, 214109.
20. A. Kinaci, C. Sevik and T. Çağin, Phys. Rev. B: Condens. Matter Mater. Phys., 2010, 82, 155114.
21. R. K. Astala and P. D. Bristowe, Modell. Simul. Mater. Sci. Eng., 2004, 12, 79.
22. J. Heyd, G. Scuseria and M. Ernzerhof, J. Chem. Phys., 2003, 118, 8207–8215.
23. G. Kresse and J. Furthmüller, Comput. Mater. Sci., 1996, 6, 15–50.
24. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865.
25. A. Janotti, J. B. Varley, P. Rinke, N. Umezawa, G. Kresse and C. G. Van de Walle, Phys. Rev. B: Condens. Matter Mater. Phys., 2010, 81, 085212.
26. A. Stroppa, G. Kresse and A. Continenza, Phys. Rev. B: Condens. Matter Mater. Phys., 2011, 83, 085201.
27. C. G. Van de Walle and J. Neugebauer, J. Appl. Phys., 2004, 95, 3851–3879.
28. J. Osorio-Guillén, S. Lany, S. V. Barabash and A. Zunger, Phys. Rev. Lett., 2006, 96, 107203.
29. Y. Gai, J. Li, S.-S. Li, J.-B. Xia and S.-H. Wei, Phys. Rev. Lett., 2009, 102, 036402.
30. S. F. Alvarado, F. La Mattina and J. G. Bednorz, Appl. Phys. A: Mater. Sci. Process., 2007, 89, 85–89.
31. O. N. Tufte and P. W. Chapman, Phys. Rev., 1967, 155, 796.
32. T. Tomio, H. Miki, H. Tabata, T. Kawai and S. Kawai, J. Appl. Phys., 1994, 76, 5886–5890.
33. R. Moos, A. Gnudi and K. Härdtl, J. Appl. Phys., 1995, 78, 5042.
34. A. Janotti, E. Snow and C. G. Van de Walle, Appl. Phys. Lett., 2009, 95, 172109.
35. N. Shanthi and D. D. Sarma, Phys. Rev. B: Condens. Matter Mater. Phys., 1998, 57, 2153.
36. S. Ouyang, H. Tong, N. Umezawa, J. Cao, P. Li, Y. Bi, Y. Zhang and J. Ye, J. Am. Chem. Soc., 2012, 134, 1974–1977.

---

Footnote

† Electronic supplementary information (ESI) available: details of sample synthesis and experimental details for photocatalytic H₂ evolution, XPS measurements, and optical absorption spectra for the selected codoped SrTiO₃. The calculated density of states for the supercell containing Cr defects and the supercell containing donor-type codopant defects. See DOI: 10.1039/c2ta00450j

---