Photocatalytic water splitting by RuO₂-loaded metal oxides and nitrides with d⁰- and d¹⁰ -related electronic configurations

Abstract

This article reviews photocatalysis for water splitting by various kinds of metal oxides and nitrides such as ferroelectric metal oxides, different kinds of titanates with d⁰ (Ti⁴⁺) electronic configuration as a core metal ion, and various typical metal oxides with d¹⁰ (In³⁺, Ga³⁺, Ge⁴⁺, Sn⁴⁺, Sb⁵⁺) configuration, and d¹⁰ (Ga³⁺) metal nitride, together with d¹⁰s² (Pb²⁺) and d⁰f⁰ (Ce⁴⁺) metal oxides. Ferroelectric metal oxides with single domain structure showed anomalous photovoltaic effects that controlled the behavior of photoexcited electrons. Various metal oxides involving ionic alkaline metal/alkaline earth metal ions showed good correlation between photocatalytic activity and the distortion of octahedral XO₆/tetrahedral XO₄ (X = core metal ion). When the metal oxides involved covalent metal ions such as Zn²⁺ and Pb²⁺, the activity was invoked even in distortion-free structures: this is due to strong electronic effects that affect the band structure. The activity of d¹⁰ metal oxides and nitrides is associated with conduction bands of hybridized sp orbitals with large dispersion that are able to generate photoexcited electrons with large mobility. A feature of the photocatalytically active metal oxides and nitrides discovered so far, which is their closed shell electronic structures is discussed.

---

Yasunobu Inoue

Yasunobu Inoue received his Doctoral degree in Chemistry from Tokyo Institute of Technology and became a faculty member of Nagaoka University of Technology. He is currently Specially Appointed Professor in Research(Chemistry) at Nagaoka University of Technology. In 2002, he received an award for his academic achievements from the Catalysis Society of Japan. His research focuses on water photolysis by solids, the artificial control of heterogeneous catalysis using surface acoustic waves and resonance oscillation, and the design of selective catalysts for hydrogenation and partial oxidation.

---

Broader context

The potential for splitting water by a combination of photocatalysts and solar energy is receiving strong interest because of its importance to the hydrogen economy. The development of highly efficient and visible-light driven photocatalysts is therefore an urgent and important issue. Since the discovery in 1972 that TiO₂ can be used for electrochemical photolysis of water, efforts have been devoted to finding new photocatalyst materials. A number of photocatalysts active for water splitting have completely different crystal and electronic structures, so it is desirable to reveal the fundamental factors that determine their photocatalytic activity in an attempt to establish useful concepts for development of new photocatalysts. We show that remarkable photocatalytic performance can be induced for metal oxides/nitrides consisting of distorted octahdedral/tetrahedral coordination and of valence and conduction band structures with large dispersion.

1. Introduction

In view of the current importance of hydrogen as a clean chemical source, the overall splitting of water by the combination of photocatalysts and solar energy has drawn strong attention as one ideal process, and the development of highly efficient and visible-light driven photocatalysts is therefore an urgent and important issue. Since the discovery of TiO₂ by Honda–Fujishima¹ for the electrochemical photolysis of water splitting in 1972, efforts have been devoted to finding new photocatalyst materials. Fig. 1 shows the types of elements in periodic table whose metal ions have been found to constitute active metal oxides and nitrides for water splitting. There are three kinds of groups in the metal ions: (i) the transition metal ions Ti, Zr, Nb, Ta and W, (ii) the rare earth metal ion Ce, and (iii) the typical metal ions Ga, In, Ge, Sn and Sb. Fig. 2 shows the representative transition metal oxides and typical metal oxides. For the transition metal ions, there are various kinds of titanates [SrTiO₃,² A₂Ti₆O₁₃ (A = Na, K, Rb),^3,4 BaTi₄O₉,^5,6 A₂La₂Ti₃O₁₀ (A = K, Rb, Cs),^7,8 Na₂Ti₃O₇,⁹ K₂Ti₄O₉¹⁰], zirconium oxide (ZrO₂¹¹) niobates [A₄Nb₆O₁₇ (A = K, Rb),¹² Sr₂Nb₂O₇,¹³ Cs₂Nb₄O₁₁,¹⁴ Ba₅Nb₄O₁₅¹⁵], tantalates [ATaO₃ (A = Na, K),^16,17 MTa₂O₆ (M = Ca, Sr, Ba),^18,19 Sr₂Ta₂O₇,¹³ ACa₂Ta₃O₁₀ (A = H, Na, Ca),²⁰ A₂SrTa₂O₇·nH₂O (A = H, K, Rb),²¹ ALn Ta₂O₇ (A = H, Na, Rb, Cs, Ln = La, Pr, Nd, Sm),²² K₃Ta₃B₂O₁₂,²³ Ba₅Ta₄O₁₅²⁴] and tungstates (PbWO₄,^25,26 Pb_xWO₄²⁷). The core metal ions of these transition metal oxides are Ti⁴⁺, Zr⁴⁺, Nb⁵⁺, Ta⁵⁺ and W⁶⁺, and it is notable that all the metal ions have d⁰ electronic configuration in common. For the rare earth metal ions Sr-doped CeO₂ was found recently to be photocatalytically active.²⁸ The core metal ion Ce⁴⁺ has d⁰ electronic configuration. For the typical metal ions, there are various kinds of indates [MIn₂O₄ (M = Ca, Sr),^29,30 AInO₂ (A = Li, Na),³¹ LaInO₃³²], gallates [Ga₂O₃,³³ ZnGa₂O₄,³⁴ MGa₂O₄ (M = Ca, Sr)], germanate (Zn₂GeO₄),³⁵ stannates [M₂SnO₄ (M = Ca, Sr)²⁹], and antimonates [NaSbO₃,^29,36 MSb₂O₆ (M = Ca, Sr),³⁶ M₂Sb₂O₇ (M = Ca, Sr)³⁶]. The core metal ions are Ga³⁺, In³⁺, Ge⁴⁺, Sn⁴⁺ and Sb⁵⁺, and interestingly, all have d¹⁰ electronic configuration. Furthermore, d¹⁰ metal nitrides such as β-Ge₃N₄,³⁷ divalent metal ion-doped GaN^38,39 and a solid solution^40,41 of GaN with ZnO, were found to be active. Especially, the solid solution is interesting, because of its capability to split water under visible light (<550 nm). The kinds and number of d¹⁰ photocatalysts discovered in recent years are almost comparable to those of d⁰ metal oxides established in the past three decades.

Fig. 1 Elements in the Periodic Table and the metal ions used as photocatalysts for overall water splitting. Metal ions, shown in brown, were found in our study to make photocatalysts.

Fig. 2 A list of metal oxides with d⁰ and d¹⁰ electronic configuration as photocatalysts for overall water splitting. Almost all of the metal oxides are powder, onto which NiO or RuO₂ are loaded as the cocatalyst.

A variety of the photocatalysts active for water splitting have completely different crystal structures and electronic structures. It is strongly desirable to reveal fundamental factors that determine their photocatalysis. In an attempt to establish the concepts useful for the development of photocatalysts, the present review is consolidated based on not only previous results, but also our recent new findings.

2. Concept for the design of efficient photocatalysts

There are three essential steps for water splitting reaction by solid photocatalysts, as shown in Fig. 3: the first is the photoexcitation of electrons from the valence band to the conduction band, the second is the transfer of the electrons and holes to the surface without recombination, and the third is the surface reactions of electrons with H⁺ and holes with OH⁻. In most powder photocatalysts, cocatalysts such as RuO₂ and NiO are deposited in the form of fine particles on metal oxide surfaces, presented schematically in Fig. 3. The roles of the cocatalyst are to promote the separation of photoexcited charges and to accelerate the third step of surface reactions. The efficiency of the first and second steps is evidently determined by the life of charges generated, and prevention of the recombination of electrons and holes is important. With respect to this, we have taken the following simple concepts into consideration: (i) electric fields, (ii) the symmetry of metal–oxygen octahedral/tetrahedral coordination forming metal oxides, and (iii) the electronic structures of valence and conduction bands. Presented here are the effects of internal bulk electric fields of ferroelectric metal oxides, distorted metal–oxygen octahedral/tetrahedral structures, and atomic orbital hybridization and dispersion of bands on photocatalytic properties.

Fig. 3 (a) Fundamental steps for water splitting by solid-state photocatalysts and (b) important factors in cocatalyst-loaded powder photocatalysts.

3. Preparation conditions for high photocatalytic performance

There are several experimental factors for determining the performance of powder photocatalysts. The photocatalytic activity depends on preparation methods and is mainly associated with the properties of metal oxides absorbing light energy and the states of cocatalyst loaded on the metal oxides.

The case of a RuO₂-loaded CaIn₂O₄ photocatalyst³² is given as an example. The metal oxide was generally synthesized by a solid-state reaction of starting metal compounds at high temperatures. The photocatalytic water splitting reaction was carried out using pure water (distilled and ion-exchanged) in a closed gas circulating apparatus under an Ar gas atmosphere. Fig. 4 shows the photocatalytic activity of RuO₂-loaded CaIn₂O₄ as a function of calcination temperatures upon synthesis of CaIn₂O₄ from Ca(CO₃)₂ and In₂O₃.³² The activity increased considerably with increasing calcination temperatures from 1273 K, reached a maximum at around 1473 K, and decreased sharply above 1500 K. The X-ray diffraction patterns showed that the major diffraction peaks of the (320), (121), and (401) planes of CaIn₂O₄ became narrower with increasing calcination temperatures. The surface area decreased with increasing temperature. SEM images show that CaIn₂O₄ particles grew considerably in the 1323–1473 K temperature range, and markedly in the 1523–1573 K range, causing the agglomeration of the particles. Thus, with an increase in calcination temperatures, both crystallization and sintering of CaIn₂O₄ take place. Similar activity-calcination temperature patterns were observed for other metal oxides such as BaTi₄O₉,⁵ ZnGa₂O₄,³⁴ and Zn₂GeO₄.³⁵

Fig. 4 Dependence of photocatalytic activity of 1 wt% RuO₂-loaded CaIn₂O₄ on calcination temperature in the preparation of CaIn₂O₄. Photocatalyst: powder, photocatalyst amount 0.25 g; water volume, 30 ml; Ar gas pressure, 1.3 kPa; light source, Xe lamp without filter; light intensity, 400 W. Ref. 32.

Crystallization eliminates structural imperfections such as vacancies and dislocation, which frequently make the recombination sites of photoexcited charges. Thus, activity enhancement with increasing calcination temperature is associated with crystallization. Sintering of the crystals causes the extraordinary growth of CaIn₂O₄ particles and thus a sharp drop in surface area. Significant decreases in the surface area of CaIn₂O₄ are responsible for activity decreases.

The photocatalytic activity is also a function of the amount and states of cocatalyst. For RuO₂ loading, ruthenium compounds such as RuCl₃ aqueous solution, Ru₃(CO)₁₂ in THF, or Ru(C₅H₇O₂)₃ in THF were loaded on CaIn₂O₄ by impregnation or photodeposition, and then oxidized at 673–773 K to convert the ruthenium species to ruthenium oxide. Without RuO₂, little hydrogen was produced, but the loading of RuO₂ permitted significant production of both H₂ and O₂. Production increased markedly in loading range 0.25–1.0 wt%, reached a plateau at around 1–1.5 wt%, and decreased slightly with further loading to 2.0 wt%. The binding energy of the Ru 3d_5/2 level in XP spectra taken for loaded RuO₂ particles was close to that for standard RuO₂ powder, indicating that the active Ru species was Ru⁴⁺ in RuO₂ particles. Namely, increasing the amounts of RuO₂ loaded on CaIn₂O₄ raised the production of H₂ and O₂, but excess amounts conversely lowered efficiency, because of the agglomeration and growth of RuO₂ particles. Thus, highly dispersed states of RuO₂ particles are appropriate for high photocatalytic performance.

Based on these results, high efficiency for photocatalytic water splitting is achieved by the combination of well-crystallized metal oxides with highly dispersed RuO₂.

4. d⁰ metal oxides

One of the most important issues to achieve high efficiency in photocatalysis by solids is to control the behavior of photoexcited electrons and holes. The internal electric fields due to dipole moment in crystals are considered to be useful. Here, the effects of macroscopic electric fields of ferroelectric crystals and the microscopic electric fields generated in distorted metal–oxygen octahedral/tetrahedral coordination of metal oxides are presented.

4.1 Effects of polarization fields in semiconducting ferroelectric metal oxides

Ferroelectric crystals have a spontaneous polarization axis. The “poling” procedure of imposing a high dc voltage through the crystal permits the arrangement of the polarization axis to have the same direction, and enables one to obtain a single domain structure. If an outer electric field is vertically applied to the ferroelectric crystal, the surface exposes the positive plane on one side and negative plane on the opposite side, presented schematically in Fig. 5. When the single domain ferroelectric crystal is heated above the Curie temperature and then cooled down without an electric field, the single domain structure changes to randomly oriented structures of polarization axis (multi domain structure). The single domain structure gives rise to an internal field in the ferroelectric crystal. Upon light irradiation for photoexcitation, it is thought that the aligned internal field permits not only the efficient separation of photoexcited charges at a light-absorption center, but also the accumulation of opposite charges on the positive and negative surfaces.

Fig. 5 Conversion of multi-domain structure to single domain structure by the application of high dc voltage (poling) to ferroelectric materials, and the reverse process by heat treatment above Curie temperature. Single domain structure exposes (+) and (−) surface on the front and back surfaces, when the polarization axis is vertical to the surface.

Fig. 6 shows the short-circuit photocurrents of Sr-doped Pb(Zr,Ti)O₃ (denoted here as PSZT) with a single domain structure.⁴² UV light irradiation to the negative surface generated a instantaneous photocurrent due to the pyroelectric effect, immediately followed by a steady photocurrent. The photocurrent continued during light-on. With light-off, the photocurrent disappeared after a transient pyroelectric current in the opposite direction. Upon UV irradiation to the positive plane, opposite photocurrent changes occurred. Fig. 7 shows the open-circuit photovoltages. Positive voltages as high as 26 V were observed for irradiation to the negative surface. The irradiation of one plane of a non-polar semiconductor disc by light with energy larger than the band gap generates photovoltages between the front and back planes of the disc by Dember effect (due to the gradient density of photoexcited electrons and holes with different mobility). In the Dember effect, the direction of photocurrent depends on the direction of light irradiation, and the resulting voltages cannot exceed the energy gap of the semiconductor. For PSZT, however, the direction of the photocurrents was determined by the direction of the polarization axis, and photovoltages were larger, by a factor of 8, than the band gap. This is defined as the anomalous photovoltaic effect (APV effect). According to a model proposed for ferroelectric polycrystalline materials, the larger-than-band-gap photovoltages V_A, are given by the equation:⁴³

V_A = P_sSN₀/E_s{1 + E_bS/(E_s(L−2S)} (1)

where P_s is the spontaneous polarization, S is the length of the space charge region, N₀ is the number of crystal grain, E_s is the dielectric constant in the space charge region, E_b is the dielectric constant of the bulk and L is the particle size in polycrystalline materials. For the present PSZT sample, V_A was calculated to be 30 V, which was in good agreement with the experimental value of 26 V. The photovoltages generated for multi domain PSZT decreased to a very low level. These results give evidence for the role of polarization axis in the control of photoexcited charges.

Fig. 6 Photocurrents generated to PSZT by UV irradiation to either the (+) or (−) surface. The direction of photocurrent was determined by the direction of polarization axis. Ref. 42.

Fig. 7 Anomalous photovoltaic (APV) effects of PSZT with single domain and multi-domain structures. PSZT (873 K) shows that PSZT was subjected to heat treatment at a temperature (873 K) above Curie temperature (553 K) and cooling in the absence of electric fields. PSZT (873 K) has a multi-domain structure. Ref. 42.

For water splitting, two device type photocatalysts were prepared using various kinds of ferroelectric materials having the opposite direction of polarization axis, i.e. exposing either positive or negative polar surfaces. One method was the employment of single domain ferroelectrics as a photocatalyst by depositing thin film cocatalyst (Pt) on either the positive or negative surface of the ferroelectrics (denoted here as Pt/(+) ferroelectric material and Pt/(−) ferroelectric material, respectively). The other was the use of positive and negative surfaces as a substrate for thin film photocatalysts: thin films of TiO₂ and WO₃ photocatalysts were deposited on the ferroelectric substrates through a Pt thin layer (denoted as thin film photocatalyst/Pt/(+) ferroelectric material and thin film photocatalyst/Pt/(−) ferroelectric material, respectively).⁴⁴

Water splitting was examined on Pt/(+)PSZT and Pt/(−)PSZT, in which 10 nm Pt was deposited on the positive and negative surfaces of PSZT, respectively.⁴⁵ Upon UV light irradiation, Pt/(+)PSZT produced considerable H₂ with a constant rate after an induction period, whereas H₂ production was very limited with Pt/(−)PSZT. The activity of the former was 8.5 fold larger than that of the latter. Table 1 shows the results of ferroelectric LiNbO₃ and LiTaO₃ single crystals with different angles of polarization axis, and multi domain PSZT (heated at 873 K higher than the Curie temperature of 583 K). Pt/(+) LiTaO₃ showed higher activity than Pt/(−) LiTaO₃, and the difference was larger when the polarization axis was vertical to the surface. Similarly, for a multi domain PSZT that had lost the characteristics of (+) and (−) surfaces, and exposed an equivalent mixture of (+) and (−) surfaces, both surfaces had very small activity. PSZT repolarized to have a single domain structure again showed a 6-fold larger activity for (+) than for (−) surface.

Table 1 Photocatalytic activity for H₂ evolution

Ferroelectrics		Activity/µmol h^−1		Activity ratio (+)/(−)
		(+)	(−)
PSZT		0.17	0.02	8.5
PSZT (873 K)		0.01	0.01	1
z-LiTaO3		0.36	0.13	2.8
36y-LiTaO3		0.28	0.14	2
z-LiNbO3		<10^−4	<10^−4	1
128y-LiNbO3		<10^−4	<10^−4	1
x-LiNbO3		<10^−4	<10^−4	1

---

Differences in the photocatalytic activity between (+) and (−) surfaces is due to the direction of polarization axis: the internal polarization fields facilitate the transfer of photoexcited electrons to the positive surface where the reduction of H⁺ to H takes place readily through the deposited Pt metal surface. In contrast, the transfer of photoexcited electrons to the (−) surface is significantly depressed, which results in negligible H₂ production. The (−) surface would preferably produce oxygen, because of the easier access of holes to the surface, but no significant evolution of O₂ was observed during the present period of irradiation. There is a possibility that oxygen might be consumed by incorporation into the crystal lattice of PSZT or via the formation of peroxide.

In the use of ferroelectrics as a substrate, thin TiO₂ and WO₃ films were combined with (+) and (−)PSZT surface through Pt thin layer, which was defined as thin film/Pt/(+)PSZT and /(−)PSZT, respectively. As shown in Fig. 8, H₂ was remarkably produced from TiO₂/Pt/(+)PSZT, but much less for TiO₂/Pt/(−)PSZT.⁴⁶ Interestingly, the substrate polarization field affects the photocatalytic behavior of TiO₂ deposited on it. Fig. 8 shows the results for WO₃ thin film photocatalysts. Again, the H₂ production was significantly larger for WO₃/Pt/(+)PSZT, while no activity was observed for WO₃/Pt/(−)PSZT. In oxidation/reduction potential, the conduction band (vacant empty band) of WO₃ is lower than that of H⁺/H level. Thus, it is impossible for WO₃ to photocatalytically produce H₂ from H₂O. However, the fact that WO₃/Pt/(+)PSZT brought about steady H₂ production but WO₃/Pt/(−)PSZT failed indicates that the polarization fields of single domain PSZT as a substrate raise the reduction potential by light irradiation to a extent to which H⁺/H occurs. Namely, APV effects have the ability of inducing photocatalysis. Fig. 9 shows APV effects on photocatalytic activity. In the case of the single domain ferroelectric substrates, such as single domain PSZT, Mn-doped PZT(MPZT), PZT that have APV effects, H₂ production takes place significantly for the (+) surface. On the other hand, H₂ production was not observed with substrates having no APV effects such as multi-domain structure PSZT, MPZT, and PZT. Furthermore, substrates of LiNbO₃ and quartz, that have no APV effects, exhibited no activity. The APV effects due to polarization fields not only activate thin film photocatalysts on the (+) surface, but also cause differences in the activity between (+) and (−) surfaces. This enabled us to design a photocatalyst differentiating reduction from oxidation sites, as shown in Fig. 10. The highest efficiency of photocatalysis will be obtained if one uses ferroelectric semiconductor materials which have both ferroelectric and semiconducting properties in an appropriate ratio.

Fig. 8 Different production of H₂ by UV light irradiation to thin film TiO₂ (upper) and WO₃ (lower) deposited on (+) PSZT and (−) PSZT. Photocatalyst: disc, 25 mm diameter; water volume, 30 ml; Ar gas pressure, 13.3 kPa; light source, Hg–Xe lamp without filter; light intensity, 200 W.

Fig. 9 Photocatalytic activity for H₂ production from water on thin WO₃ films deposited on single domain and multi-domain ferroelectric substrates. PSZT: Sr-doped PZT, MPZT: Mn-doped PZT, PZT: lead zirconium titanate. Reaction conditions are the same as those described in Fig. 8.

Fig. 10 A model of semiconducting ferroelectric photocatalysts having separate reduction and oxidation sites.

4.2 Effects of local electric fields in alkaline metal and alkaline earth metal titanates

As described above, internal polarization fields have positive effects on the acceleration of photocatalysis by controlling the behavior of photoexcited charges. Polarization fields of ferroelectric materials are due to bulk properties through the crystal. If one asks whether or not there are crystal structures that provide local polarization fields which play an important role in photocatalysis, the distorted structures of metal–oxygen octahedron/tetrahedron would be an answer.

Among various kinds of alkaline metal and alkaline earth metal titanates available, titanates with tunnel structures are interesting from the viewpoint of local geometric structures and photocatalysis. Representative titanates are BaTi₄O₉ with a pentagonal prism tunnel structure and A₂Ti₆O₁₃ (A = Na, K) with rectangular tunnel structures.^3–5 In a series of barium titanates, the stable barium titanates are, BaTi₄O₉ (orthorhombic crystal structure, pentagonal prism tunnel structure), Ba₂Ti₉O₂₀ (triclinic, hollandite-like tunnel structure), Ba₄Ti₁₃O₃₀ (orthorhombic close-packed arrays), and Ba₆Ti₁₇O₄₀ (monoclinic close-packed arrays).⁶ The examination of RuO₂-loaded barium titanates showed that only BaTi₄O₉ was photocatalytically active for water splitting, as shown in Fig. 11. Fig. 12 shows the ESR signal of the barium titanates at 77 K in 4 kPa of He under UV irradiation. BaTi₄O₉ exhibited a strong signal with g = 2.018 and 2.004, whereas the other three barium titanates failed to show any characteristically strong signals. The BaTi₄O₉ signal was also obtained in gases, such as Ar, O₂, and H₂^47,48 and was assigned to a surface radical, O⁻, derived from a lattice oxygen O₂⁻. Active A₂Ti₆O₁₃ (A = Na, K) showed strong ESR signals at g = 2.020, 2.018, and 2.004 due to the O⁻ radical under similar experimental conditions.⁴⁹Table 2 summarises O⁻ radical formation and photocatalysis for various kinds of titanates having different crystal structures. All titanates able to generate the O⁻ radical are, without exception, photocatalytically active for water splitting.

Table 2 Photocatalytic activity and photo production of O⁻ radical^a

Titanates	Structures	Photocatalysis (with RuO2)	O ^− radical production
a : active, : inactive, s: strong, m: medium, w: weak.
BaTi4O9	Pentagonal prism tunnel	(s)	(s)
KTi3TaO9	Pentagonal prism tunnel	(w)	(w)
Na2Ti6O13	Rectangular tunnel (3P)	(s)	(s)
K2Ti6O13	Rectangular tunnel (3P)	(s)	(s)
Rb2Ti6O13	Rectangular tunnel (3P)	(m)	(m)
Na2BaTi10O22	Rectangular tunnel (2P–3P)	(w)	(w)
K2SrTi10O22	Rectangular tunnel (2P–3P)	(w)	(m)
Na2Ti3O7	Zig-zag layer	(m)	(s)
K2Ti4O9	Zig-zag layer	(w)	(s)
KTi3NbO9	Pentagonal prism tunnel
Bi2Ti4O11	Zig-zag layer
Cs2Ti6O13	Infinity layer
BaTiO3	Perovskite
NaLaTiO4	Perovskite like

---

Fig. 11 Photocatalytic activity of RuO₂-loaded various barium titanates. Photocatalyst: powder, photocatalyst amount, 0.20 g; water volume, 20 ml; Ar gas pressure, 27 kPa; light source, Xe lamp without filter; light intensity, 400 W. Ref. 6.

Fig. 12 ESR signals for (a) Ba₆Ti₁₇O₄₀, (b) Ba₄Ti₁₃O₃₀, (c) BaTi₄O₉, and (d) Ba₂Ti₉O₂₀ in a He atmosphere at 77 K. Ref. 6.

Both the photocatalytic activity and the concentration of O⁻ in BaTi₄O₉ varied by heat treatment, such as reduction in temperature 1073–1173 K or reduction followed by oxidation at 1173 K. Fig. 13 shows the correlation of photocatalytic activity on BaTi₄O₉ and the concentration of O⁻ produced at the surface by UV irradiation. The activity increases almost in proportion to the concentration of O⁻ radical,⁵⁰ which indicates that the O⁻ radical is associated with photocatalytic active sites. The O⁻ radical appears as a result of photoexcitation, and its strong intensity indicates the efficient ability of BaTi₄O₉ to generate photoexcited charges.

Fig. 13 Linear correlation between the concentration of O⁻ produced by UV irradiation on BaTi₄O₉ and the photocatalytic activity for H₂ and O₂ production of RuO₂-loaded BaTi₄O₉. Photocatalyst: powder, photocatalyst amount, 0.20 g; water volume, 20 ml; Ar gas pressure, 27 kPa; light source, Xe lamp without filter; light intensity, 400 W. Ref. 50.

Raman spectra showed that BaTi₄O₉ had an extremely strong peak at high wave number, 860 cm⁻¹. This is due to a double bond-like, Ti–O bond with short distance, indicating that octahedral TiO₆ are distorted. The characteristic peak was missing for the other barium titanates: Ba₂Ti₉O₂₀, Ba₄Ti₁₃O₃₀ and Ba₆Ti₁₇O₄₀. Fig. 14 shows a pentagonal prism tunnel structure of BaTi₄O₉ that is composed of two strongly distorted octahedral TiO₆, in which Ti⁴⁺ ions are displaced by 0.03 and 0.021 nm, respectively, from the center of gravity of the surrounding six oxygen ions.^51,52 The out-of-center Ti ions generate a dipole moment of 5.7 and 4.1 D (D = debye).⁴⁸ The dipole moment causes an internal local polarization field that promotes the formation of photoexcited charges. Thus, the unique photocatalytic properties of BaTi₄O₉ are associated with the geometric effects of distorted TiO₆ with short and long Ti–O bonds. The photocatalytically active Na₂Ti₆O₁₃ also has distorted octahedral TiO₆⁵³ with a dipole moment of 6.7, 5.8, and 5.3 D.⁴⁹ These results indicate that the distorted metal–oxygen octahedron is associated with high photocatalytic performance.

Fig. 14 A pentagonal prism tunnel structure of BaTi₄O₉ and dipole moment inside octahedral TiO₆.

4.3 Effects of electronic structures

It is important to clarify the role of electronic structures, in particular band structures, to understand photocatalysis by solids. The electronic structure of BaTi₄O₉ was calculated by the plane-wave-density function theory (DFT) method performed using the CASTEP program.⁵⁴ The valence electronic configurations were 5s² 5p⁶ 6s² for Ba, 3s² 3p⁶ 3d² 4s² for Ti, and 2s² 2p⁴ for O atoms. The unit cell size was [BaTi₄O₉]₂, and the number of occupied orbitals was 112. Fig. 15 shows the energy band dispersion diagram and density of state (DOS) for BaTi₄O₉.⁵⁵ The band gap was calculated to be 2.84 eV. The density contour maps showed that the valence band was composed of purely the O 2p orbital, whereas the orbitals of the conduction bands exclusively consisted of the Ti 3d orbital. The mixing of Ti 3d with the O 2p orbital was extremely small. Thus, the degree of hybridization in the valence and conduction bands was negligible for BaTi₄O₉, and the dispersion in k-space was quite small for the BaTi₄O₉ conduction band. This indicates that mobility of photoexcited electrons is low, suggesting that structure effects of distorted octahedral TiO₆ play an important role in photocatalysis.

Fig. 15 Energy band diagram and density of states (DOS) for BaTi₄O₉. DOS breaks down into the angular momentum of atomic orbitals (AOs). The contribution of atomic orbitals (AOs) in DOS is illustrated (purple line, s-orbital; red line, p-orbital; green line, d-orbital).

5. d¹⁰ metal oxides

If one considers that metal oxides with a vacant d orbital (= d⁰) are useful for photocatalysis, then we can assume that this will also be the case for metal oxides with a completely occupied d orbital (= d¹⁰)—simply because both d⁰ and d¹⁰ configurations behave similarly from a quantum chemistry viewpoint. Accordingly, making this assumption we have found that typical metal oxides with d¹⁰ electronic configuration form a group of efficient photocatalysts for overall water splitting when combined with RuO₂ as a cocatalyst. The photocatalytic properties of RuO₂-loaded typical metal oxides involving In³⁺, Ga³⁺, Ge⁴⁺, Sn⁴⁺, and Sb⁵⁺ ions are stated here based on the local geometric structures and band structures.

5.1 Effects of geometric structures

5.1.1 Indates, AInO₂ (A = Li, Na), MIn₂O₄ (M = Ca, Sr), LnInO₃ (Ln = La, Nd). Various indates, such as alkaline metal indates (LiInO₂, NaInO₂ and Na_0.9K_0.1InO₂), alkaline earth metal indates [CaIn₂O₄, SrIn₂O₄, Sr_{1 −} xCa_xIn₂O₄(x = 0.25, 0.50, 0.75), Sr_{1 −} xBa_xIn₂O₄ (x = 0.07)] and lanthanum indate, LaInO₃, and gadolinium indate, GdInO₃, were investigated.^29–32

CaIn₂O₄ showed threshold absorption at around 450 nm, slightly higher absorption between 420 and 310 nm, and the largest level at 305 nm.³² SrIn₂O₄ provided a similar absorption pattern. A photocatalyst of 1 wt% RuO₂-dispersed MIn₂O₄ (M = Ca, Sr) produced H₂ and O₂ from H₂O under Xe lamp illumination. Both the products increased in proportion to irradiation time, and the production was nearly the same in the repeated run, indicative of the stability of RuO₂-dispersed MIn₂O₄ (M = Ca, Sr) photocatalysts. The ratio of H₂ to O₂ was close to the stoichiometric ratio.

It is of interest to compare the photocatalytic activities of various RuO₂-loaded indates such as alkaline metal, alkaline earth metal and lanthanum indates. Under Xe lamp illumination, LiInO₂ showed negligible photocatalytic activity, and NaInO₂ had a little activity. On the other hand, remarkable activity was observed for MIn₂O₄ (M = Ca, Sr), in which CaIn₂O₄ showed higher activity than SrIn₂O₄. Sr_0.93Ba_0.07In₂O₄ exhibited similar activity to that of the latter. LaInO₃ produced a small amount of hydrogen only. Thus, activity was larger in the order CaIn₂O₄ > SrIn₂O₄, Sr_0.93Ba_0.07In₂O₄ ≫ LaInO₃ > NaInO₂ ≫ LiInO₂, as shown in Fig.16. These results illustrate that there are intrinsic differences in photocatalytic properties among the indates, even though they have a common octahedrally coordinated In³⁺ (d¹⁰) metal ion as a core ion.

Fig. 16 Photocatalytic activity of various alkaline metal and alkaline earth metal indates. Photocatalyst: powder, photocatalyst amount, 0.25 g; Ar gas pressure, 13.3 kPa; light source, Xe lamp without filter; light intensity, 400 W. Ref. 55.

Fig. 17 shows the Raman spectra³² of CaIn₂O₄, SrIn₂O₄, Sr_0.93Ba_0.07In₂O₄, and NaInO₂. The spectrum of CaIn₂O₄ consisted of five major peaks at 200, 262, 328, 406, and 545 cm⁻¹ in the wave number region 200–600 cm⁻¹. The strongest peak appearing at the highest wave number, 545 cm⁻¹, was assigned to a vibration mode of ν₁(A_1g). SrIn₂O₄ showed a similar spectrum in which the peak positions were 191, 236, 321, 395, and 542 cm⁻¹. Sr_0.93Ba_0.07In₂O₄ showed a similar Raman pattern to that of SrIn₂O₄: the major five peaks were observed at 188, 231, 317, 391, and 540 cm⁻¹. On the other hand, NaInO₂ showed two peaks at 330 and 504 cm⁻¹.

Fig. 17 Raman spectra of (a) NaInO₂, (b) CaIn₂O₄, (c) SrIn₂O₄ and (d) Sr_0.93Ba_0.07In₂O₄ measured at room temperature. Ref. 55.

SrIn₂O₄ has an orthorhombic structure with the lattice parameter of a = 0.9809, b = 1.1449, and c = 0.3265 nm,⁵⁶ whereas CaIn₂O₄ has an orthorhombic structure with a unit cell of a = 0.965, b = 1.13 and c = 0.321 nm.⁵⁷ They have nearly the same crystal structure, although the size of the unit cell is slightly smaller for CaIn₂O₄. Sr_0.93Ba_0.07In₂O₄ has an orthorhombic crystal structure with a lattice parameter of a = 0.9858, b = 1.152, and c = 0.3273 nm.⁵⁸ It is readily understood that the close similarity in the unit cell among them provided similar Raman patterns, as shown in Fig. 17. LaInO₃ consists of an orthorhombic structure with a unit cell of a = 1.1402, b = 1.1796 and c = 0.8198 nm.⁵⁹ NaInO₂ has a hexagonal structure with a unit of a = 0.3232, b = 1.639 nm, α = β = 90°, γ = 120 °nm,⁶⁰ and has a layer structure. LiInO₂ has a tetragonal structure with a unit cell of a = 0.4312 and b = 0.9342 nm.⁶¹

The positive effects of local internal fields due to dipole moment have been demonstrated for transition metal oxides with d⁰ configuration (cf. 4.2). It is interesting to see a correlation of activity and the distortion of octahedral InO₆ for various indates with d¹⁰ configuration. Based on the deviation of the position of an In³⁺ ion from the center of gravity of the six oxygens surrounding the In³⁺ ion, the dipole moment generated inside the octahedral coordination was calculated. SrIn₂O₄ has two kinds of the distorted octahedral InO₆ with dipole moment of 2.80 D (D = debye) and 1.11 D. For Sr_0.93Ba_0.07In₂O₄, the dipole moments were 1.70 and 2.58 D. On the other hand, LiInO₂ and NaInO₂ are composed of distortion-free normal InO₆ without dipole moment. Fig. 18 shows the comparison of photocatalytic activity with dipole moment. The indates consisting of distorted InO₆ having dipole moments are photocatalytically active,³² whereas the distortion-free indates exhibited negligible activity. Thus, a correlation between the photocatalytic activity and the distortion of InO₆ octahedron is true for d¹⁰ metal oxides.

Fig. 18 A correlation between photocatalytic activity and dipole moment of octahedral InO₆ in AInO₂ (A = Li, Na), Sr_0.93Ba_0.07In₂O₄, and SrIn₂O₄. Ref. 55.

5.1.2 Gallates, MGa₂O₄ (M = Mg, Ca, Sr) and ZnGa₂O₄. Alkaline earth metal gallates, MGa₂O₄ (M = Mg, Sr, Ba), were employed. The photocatalytic activity for water splitting of 0.5 and 1.0 wt% RuO₂-loaded MgGa₂O₄ was negligible under UV irradiation: only a small amount of hydrogen was produced. On the other hand, both 1 wt% RuO₂-loaded SrGa₂O₄ and BaGa₂O₄ showed remarkable evolution of H₂ and O₂. Good reproducibility was observed in repeat runs. There was a difference in photocatalytic activity with the kind of alkaline earth metals.

The absorption of light occurred at around 290 nm for M = Mg, 270 nm for M = Sr and 260 nm for M = Ba. The threshold wavelength shifted slightly toward longer wavelength in the order Mg > Sr > Ba, though the wavelength for the maximum absorption was at around 240 nm. Fig. 19 shows schematically the crystal structures of MGa₂O₄ (M = Mg, Sr, Ba). MgGa₂O₄ has a cubic crystal structure and is composed of GaO₆ octahedron.⁶² In calculating the center of gravity of the six oxygen ions surrounding the Ga³⁺ ion, MaGa₂O₄ has a GaO₆ octahedron nearly free from distortion, dipole moment zero. SrGa₂O₄ has a monoclinic crystal structure, in which GaO₄ tetrahedron is the fundamental unit.⁶³ There are three kinds of GaO₄ tetrahedral units: one has a dipole moment of 0.80 D, and the other two have 1.2 D. BaGa₂O₄ has a hexagonal crystal structure consisting of GaO₄ tetrahedron whose dipole moments are 1.70, 1.11 and 2.58 D.⁶⁴ Both SrGa₂O₄ and BaGa₂O₄ have pentagonal prism-like tunnels in the tridymite-type structures.⁶⁵Fig. 20 shows a comparison between the photocatalytic activity and dipole moment. It is notable that the distorted MGa₂O₄ (M = Sr, Ba) with dipole moments are photocatalytically active, whereas distortion-free MgGa₂O₄ exhibits negligible activity.

Fig. 19 Schematic representation of crystal structures of MGa₂O₄ (M = Mg, Sr, Ba) and the geometry of involved octahedral GaO₆ and tetrahedral GaO₄.

Fig. 20 A correlation between photocatalytic activity and dipole moment of octahedral GaO₆ and tetrahedral GaO₄ in MGa₂O₄ (M = Mg, Sr, Ba). Photocatalyst: powder, photocatalyst amount 0.25 g; Ar gas pressure, 13.3 kPa; light source, Hg–Xe lamp without filter; light intensity, 200 W.

Table 3 shows active d¹⁰ and d⁰ metal oxides having pentagonal prism-like tunnel structures.⁶⁶ As described above, BaTi₄O₉ is a representative metal oxide with d⁰ configuration, which has distorted octahedral TiO₆ with dipole moment. SrIn₂O₄ with d¹⁰ configuration has distorted octahedral InO₆ with dipole moment. MGa₂O₄ (M = Sr, Ba) with d¹⁰ configuration have distorted GaO₄ tetrahdera with dipole moment. These results indicate the important role of distorted octahedral and tetrahedral coordination for both d¹⁰ and d⁰ configurations in photocatalysis for water splitting. Several explanations can be offered for the reason why distorted units with dipole moments are effective in photocatalysis. There is the possibility that short or long metal–oxygen bonds affect photoexcitation and that the poor symmetry of octahedral and tetrahedral coordination tends to form isolated orbitals, giving rise to different photoexcitation efficiency. Another is the contribution of local internal fields due to the dipole moment inside the distorted units: the fields are considered useful for electron–hole separation upon photoexcitation.

Table 3 Photocatalysts with distorted XO₆ and XO₄ in pentagonal prism-like tunnel structures

Configuration	XO6 octahedron	XO4 tetrahedron
d^0
d^10

---

In a 1 wt% RuO₂-dispersed ZnGa₂O₄ photocatalyst,³⁴ the amounts of H₂ and O₂ produced from H₂O increased in proportion to irradiation time from a Hg–Xe lamp. The repeat run showed nearly the same production, and the H₂/O₂ ratio approached the stoichiometric value of 2.0. In ZnGa₂O₄, Ga³⁺ ions have octahedral coordination, whereas Zn²⁺ ions are coordinated tetrahedrally.⁶⁷ ZnGa₂O₄ consists of distortion-free normal GaO₆ octahedron. Thus, the high photocatalyst performance, completely different from MgGa₂O₄, is somewhat puzzling. Light absorption by ZnGa₂O₄ occurred at around 300 nm, increased steeply at around 280 nm and reached a maximum level at 250 nm. It is interesting to see that a shift to longer wavelength by about 50 nm occurred with ZnGa₂O₄, compared to those of MGa₂O₄ (M = Mg, Sr, Ba). The shift suggests the strong contribution of Zn²⁺, different from the alkaline earth metal ions, to the electronic structures. Thus, the activity is strongly due to characteristic electronic structures, which will be discussed in detail later (cf. 5.2.2.).

5.1.3 Zinc germanate, Zn₂GeO₄. Most of the alkaline metal and alkaline earth metal germinates are not stable in water upon light irradiation, and Zn₂GeO₄ is one of the representaive compounds which is suitable for photocatalysts: it has an absorption threshold at around 310 nm. A photocatalyst of 1.0 wt% RuO₂-dispersed ZnGe₂O₄ showed very high activity for H₂ and O₂ production under Hg–Xe lamp irradiation.³⁵

A willemite structure of Zn₂GeO₄ consists of tetrahedral GeO₄ and ZnO₄. As shown in Fig. 21, one GeO₄ tetrahedron and two kinds of tetrahedral ZnO₄ are combined with each other through the edge oxygen.⁶⁶ The two tetrahedral ZnO₄ are heavily distorted to have dipole moments of 1.1 and 1.2 D. Replacing a Ge ion by a Si ion provides Zn₂SiO₄ with nearly the same structure as Zn₂GeO₄: the rhombohedral cell of Zn₂GeO₄⁶⁸ has a = 0.8836 nm and α = 107.42° and that of Zn₂SiO₄ has a = 0.8626 nm and α = 107.52°. Thus, Zn₂SiO₄ has two kinds of distorted tetrahedral ZnO₄ with dipole moment of 1.1 D and 1.4 D. However, RuO₂-loaded Zn₂SiO₄ was photocatalytically inactive for water splitting, indicating that only the tetrahedral GeO₄ unit constituted the photocatalytic active site. This demonstrates that distorted tetrahedral XO₄ (X = metal ion) is associated with the formation of photocatalytically active sites.³²

Fig. 21 A schematic representation of local crystal structure of Zn₂GeO₄.

5.1.4 Stannates, M₂SnO₄ (M = Ca, Sr, Ba). The UV diffuse reflectance spectra showed that the absorption of Ca₂SnO₄ started at 270 nm, increased sharply at around 260 nm and reached the highest level at 250 nm. For Sr₂SnO₄, the absorption occurred gradually at around 270 nm, and had a maximum at 240 nm. Ba₂SnO₄ showed an absorption pattern having a maximum at 260 nm, with a tail extending to 370 nm. Fig. 22 shows the photocatalytic activity of 1 wt% RuO₂-loaded M₂SnO₄ (M = Ca, Sr, Ba) under UV light irradiation. M₂SnO₄ (M = Ca, Sr) was capable of producing H₂ and O₂, but Ba₂SnO₄ produced only a small amount of H₂.

Fig. 22 Photocatalytic activity of 1 wt% RuO₂-loaded M₂SnO₄ (M = Ca, Sr, Ba) and the geometry of the octahedral SnO₆ involved. Photocatalyst: powder, photocatalyst amount, 1.0 g; water volume, 700 ml; Ar gas pressure, 4.0 kPa; light source, Hg lamp without filter; light intensity, 450 W.

As shown in Fig. 23, Ca₂SnO₄ has a layer structure in which the SnO₆ octahedron are connected to each other by edges, forming columns along c-axis.⁶⁹ On the other hand, Sr₂SnO₄ and Ba₂SnO₄ have a perovskite layer structure in which SnO₆octahedron are connected by corner atoms, forming layer structures.^70,71 The Sr or Ba ions are located between two-dimensional SnO₆ layers. The crystal structures are different between Ca₂SnO₄ and M₂SnO₄ (M = Sr, Ba) and similar between Sr₂SnO₄ and Ba₂SnO₄. The findings that the photocatalytic activity was similarly large for both Ca₂SnO₄ and Sr₂SnO₄ but small for Ba₂SnO₄ indicate that the photocatalytic activity is independent of differences in the layer structures. The octahedral SnO₆ in Ca₂SnO₄ is composed of four Sn–O bonds of 214.63 pm and two Sn–O bonds of 201.00 pm. The bond angles of ∠O–Sn–O deviate from 90°, and hence the octahedral SnO₆ has a distorted form (Fig. 22). The bond length for octahedral SnO₆ in Sr₂SnO₄ is 201.85 pm for four Sn–O bonds and 191.71 pm for two bonds. All bond angles of ∠O–Sn–O are 90°, indicating that the octahedral SnO₆ has a compressed form (Fig. 22). On the other hand, Ba₂SnO₄ has similar bond lengths of 207.06 pm for four Sn–O bonds and 206.96 pm for two Sn–O bonds. All bond angles are close to 90°, which shows that the octahedral SnO₆ is nearly normal. Thus, both photocatalytically active Ca₂SnO₄ and Sr₂SnO₄ are composed of distorted SnO₆ unit, whereas inactive Ba₂SnO₄ has nearly distortion-free SnO₆ unit. These results also indicate that the octahedral SnO₆ of active Ca₂SnO₄ and Sr₂SnO₄ involve shorter bond length, whereas that of photocatalytically poor Ba₂SnO₄ is composed of longer Sn–O bond. The activity differences among them are due to geometric effects of the octahedral SnO₆ that are distorted by the presence of short Sn–O bonds. Thus, in the alkaline earth metal stannates, it is confirmed that deformed octahedral SnO₆ play an important role in photocatalysis.

Fig. 23 Layer structure of Ca₂SnO₄ and perovskite layer structures of Sr₂SnO₄ and Ba₂SnO₄.

Fig. 24 A schematic representation of weberite structure of M₂Sb₂O₇ (M = Ca, Sr) and two kinds of octahedral SbO₆.

5.1.5 Antimonates, M₂Sb₂O₇ (M = Ca, Sr), CaSb₂O₆ and NaSbO₃. Among antimonates, M₂Sb₂O₇ (M = Ca, Sr), CaSb₂O₆ and NaSbO₃ were found to be active when RuO₂ was loaded.³⁶ The UV diffuse reflectance spectra showed that light absorption of M₂Sb₂O₇ (M = Ca, Sr) began at around 380 nm and attained the maximum level at 280 nm. For NaSbO₃, abrupt absorption occurred at around 280 nm. For CaSb₂O₆, the absorption occurred gradually at around 400 nm, but the main absorption was similar to that of M₂Sb₂O₇ (M = Ca, Sr).

All of 1 wt% RuO₂-loaded M₂Sb₂O₇ (M = Ca, Sr), CaSb₂O₆ and NaSbO₃, had the ability to split water under Hg–Xe lamp irradiation.³⁶ M₂Sb₂O₇ has two kinds of SbO₆ in a weberite structure. As shown in Fig. 24, one type SbO₆ is surrounded by four octahedral SbO₆, and the other type SbO₆ is connected to six octahedral units.⁷² The former is a compressed octahedron formed by two shorter Sb–O bonds and four bonds with the same bond length, whereas the latter is an elongated octahedron with two longer Sb–O bonds. NaSbO₃ of ilmenite contains alternating layers of edge-sharing octahedral SbO₆,⁷³ which are distorted in such a way that the Sb ion has the out-of-center position. CaSb₂O₆ consists of infinite sheets of edge-sharing octahedral SbO₆ alternating with layers of Ca ion.⁷⁴ The octahedron is distorted by stretch in the one plane and contraction in the other direction. Photocatalytically active antimonates are composed of deformed octahedral SbO₆.

Fig. 25 Energy band diagram and density of states (DOS) for SrIn₂O₄. DOS breaks down into the angular momentum of atomic orbitals (AOs). The contribution of atomic orbitals (AOs) in DOS is illustrated (purple line, s-orbital; red line, p-orbital; green line, d-orbital).

5.2 Effects of electronic structures

The photocatalytic activity is closely associated with the behavior of photoexcited electrons and holes that is determined by the band level and dispersion. To understand the band structures, the energy band dispersion diagram and the density of state (DOS) were calculated by the DFT method.⁵⁴ The core electrons were replaced by the ultra-soft core potentials.

5.2.1 Indates, AInO₂ (A = Li, Na), MIn₂O₄ (M = Ca, Sr), LnInO₃ (Ln = La,Nd). The valence electronic configurations for SrIn₂O₄ were 4s²4p⁶5s² for Sr atom, 4d¹⁰5s²5p¹ for In atom, and 2s²2p⁴ for O atom. A unit of [SrIn₂O₄]₄ was employed as a primitive unit cell, in which the number of occupied orbitals was 120. The kinetic energy cutoff was 260 eV.

Fig. 25 shows the energy band dispersion diagram and the density of state (DOS) for SrIn₂O₄.⁵⁵ The Sr4s orbital appeared at the deepest level, followed by the O2s, Sr4p, and In4d orbitals with increasing energy. There were no bonding interactions between Sr4p–O2p and In4d–O2p orbitals. The lowest part of the valence band (#73) was formed by the O2p orbital mixed slightly with the In 5s5p orbitals. Fig. 26 and 27 show the density contour maps of the HOMO and LUMO levels, respectively. The highest energy orbital (#120) showed the p orbital lobes on the O atom, which indicated that the highest orbital (HOMO) in the valence band was purely composed of the O2p orbital. On the other hand, the bottom of conduction band (LUMO) was formed by the hybridized In5s and 5p orbitals, with a small contribution from the O2p orbital. In view of large overlap between the In5s and 5p orbitals, the LUMO had very large dispersion, and hence the photoexcited electrons had large mobility, which evidently induced high photocatalytic performance. The Sr5s and Sr4d orbitals were mixed to the In5s5p orbitals at a higher energy of the conduction band. Excitation of electrons upon illumination occurred from the O2p to the In5s5p orbitals (the band gap was calculated to be 2.09 eV).

Fig. 26 Density contour maps of orbitals at the top and bottom of the valence band for SrIn₂O₄. (a) The highest energy (HOMO) level (#120) and (b) the lowest energy level(#73) of the valence band.

Fig. 27 Density contour maps of orbitals at the lower part of the conduction band for SrIn₂O₄. (a) The lowest energy (LUMO) level (#121) and (b) slightly higher level (#129) of the conduction band.

The energy-band diagram and the density of states (DOS) for LiInO₂ and NaInO₂ were similar.³¹ The O2s and In4d orbitals were located at deep core levels for both LiInO₂ and NaInO₂. The valence bands were essentially composed of the O2p orbital in both cases, and the lower part of conduction bands consisted of In5s5p hybridized orbitals (there was an interesting difference in the calculated band gaps: LiInO₂ had the normal band gap of 2.03eV, whereas NaInO₂ the extremely small band gap of 0.51 eV. The small band gap for the latter is associated with In–In interactions by the large overlap of the In atom orbitals). The atomic orbitals forming the valence and conduction bands were similar between SrIn₂O₄ and AInO₂ (A = Li, Na). However, there were differences in the degrees of mixing with In atom orbitals at conduction bands between Sr and Li (or Na) atom orbitals. The energy difference between the top of valence band and Sr5s orbital was 5.5 eV, whereas differences between the top of valence and Li2s and Na3s orbitals were 7.2 and 10.0, respectively. Thus, there was a considerably large overlap of Sr5s orbital with In5s5p orbitals, compared to the Li2s and Na3s orbitals.⁵⁵

5.2.2 Gallates, MGa₂O₄ (M = Mg, Ca, Sr) and ZnGa₂O₄. As shown in the energy-band diagram and DOS for ZnGa₂O₄(Fig. 28), the lowest and the second lowest bands were the O2s and Ga3d orbitals, respectively.³⁴ The lower part of the broad valence band was composed of O2p, Ga4s and 4p, Zn3d, 4s and 4p orbitals. The density contour maps showed that the lower energy parts of the valence band were composed of the O2p and Zn3d orbitals, contributing to the Zn–O bond formation. In the middle, the electron density was distributed around the Ga and O atoms. The higher part of the valence band O2p and Zn 4d orbitals again contributed to the formation of Zn–O bonding. Thus, the covalent bond character existed between Zn and O atoms, and also between Ga and O atoms to a less degree. The top of the valence band (HOMO) consisted of the O2p orbital only. The bottom of conduction band (LUMO) was composed of the Zn4s4p + Ga4s4p orbitals as well as O2p orbital, which was characterized by large dispersion in the k-space. The hybridized orbitals had major electron density on Zn and Ga atoms. The O2p orbital always mixed into the Zn4s4p or Ga4s4p orbitals with out of phase, which was the counterpart of the in phasemixing at the lower energy region of the valance band.

Fig. 28 Energy band diagram and DOS for ZnGa₂O₄. Purple line, s-orbital; red line, p-orbital; green line, d-orbital.

MgGa₂O₄ showed that the valence band was composed of O2p orbital, and the conduction band consisted of Ga4s4p orbitals. No contributions of atomic orbitals of Mg atom to the valence and conduction bands were observed. The band gap was calculated to be 2.56 eV. Fig. 29 shows the band structure and DOS for SrGa₂O₄. DOS breaks down into the angular momentum of atomic orbitals (AOs) represented by different line shapes. The Sr4s, O2s, Sr4p and Ga3d orbitals in the core level appeared with increasing energy. The valence band was mainly composed of the O2p orbital, and the conduction band was formed by the hybridized Ga4s4p orbitals mixed with O2p orbital. The upper part of the conduction band involved the contribution of the Sr4d orbital. The band gap was 2.97 eV.

Fig. 29 Energy band diagram and DOS for SrGa₂O₄. The contribution of atomic orbitals (AOs) in DOS is shown: purple line, s-orbital; red line, p-orbital; green line, d-orbital.

When light absorption properties were compared between ZnGa₂O₄ and SrGa₂O₄, the former showed longer absorption wavelength by 20 nm than SrGa₂O₄. Sampath et al. showed that the hybridization of Zn3d with the O2p orbital caused the upward shifts of the valence band maximum.⁷⁵ For the II–VI semiconductors, Wei and Zunger pointed out that the p–d repulsion repelled the valence band maximum upwards without affecting the conduction minimum.⁷⁶ Thus, the narrower band gap for ZnGa₂O₄ than for SrGa₂O₄ is due to this p–d repulsion. Furthermore, since the bottom of the conduction band for ZnGa₂O₄ was composed of Ga4s4p + Zn4sp hybridized orbitals, there is a possibility that strong mixing of Zn and Ga orbitals in the conduction band induces not only large dispersion but also band gap narrowing.

The electron transfer upon illumination occurs from the O2p orbital to the hybridized Ga4s4p + Zn4s4p orbitals. The large dispersion of LUMO is able to yield excited electrons with large mobility in the conduction band. This is responsible for high photocatalytic performance. Note that this is due to the strong Zn–O and Zn–Ga interactions, as observed in Zn3d–O2p orbital repulsion in the valence band and in hybridization of Ga4s4p with Zn4s4p in the conduction band. For SrGa₂O₄, on the other hand, these interactions were missing.

5.2.3 Germanates, Zn₂GeO₄. In band calculation of Zn₂GeO₄, the valence atomic configurations were 3d¹⁰4s² for Zn, 4s²4p² for Ge and 2s²2p for the O atom. The number of occupied bands was 156 for the unit cell construction of (Zn₂GeO₄)₆. Fig. 30 shows the band dispersion and DOS involving the contribution of each orbital.³⁵ The deep occupied bands were composed of the O2s and then Ge4s orbital. The valence band consisted of hybridized Zn3d and O2p orbitals. The bottom of conduction band was composed of the Ge4p orbital with a small contribution of Zn4s4p orbitals. The band gap was estimated to be 2.0 eV. The band dispersion was small for the occupied bands, but it was remarkably large at the bottom of conduction band. The electron density contour map showed that the lower part of the valence band consisted of the Zn3d and O2p orbitals, whereas the upper part was composed of the O2p orbital. The bottom of conduction band (LUMO) consisted of the Ge4p orbital, and at a slightly higher level involved the contribution of Ge4s and Zn4s4p orbitals. The electron transfer upon illumination occurs from the O2p to the Ge4p orbital. Because of large dispersion of LUMO, it is evident that photoexcited electrons have large mobility useful for photocatalysis.

Fig. 30 Energy band diagram and DOS for Zn₂GeO₄. The contribution of atomic orbitals (AOs) in DOS is shown: purple line, s-orbital; red line, p-orbital; green line, d-orbital.

5.2.4 Stannates, M₂SnO₄ (M = Ca, Sr, Ba). Fig. 31 shows the energy band diagram and DOS involving the contribution of each orbital for Ca₂SnO₄. The Ca3s orbital appeared at a deep level, followed by the Ca3p, O2s and Sn5s orbitals with increasing energy. The valence band was mainly formed by the O2p orbital. The conduction band was mainly composed of the Sn5s orbital hybridized with the Ca3d and slightly with O2p orbital. The lower part of the valence band was hybridized with the Sn5p orbital, whereas the upper part solely consisted of the O2p orbital. Furthermore, the small contribution of the Sn5s orbital was observed at the top of the valence band (split-off state). This Sn5s orbital mixed into the O2p orbital with out of phase, which possibly raised the top of valence band. The HOMO level was composed of the O2p hybridized with Sn5s orbital. On the other hand, the lowest part of the conduction band was formed by the Sn5s orbital, and the upper part was hybridized with the Sn5p and Ca3d orbitals. The band gap was calculated to be 3.41 eV.

Fig. 31 Energy band diagram and DOS for Ca₂SnO₄. The contribution of atomic orbitals (AOs) in DOS is shown: purple line, s-orbital; red line, p-orbital; green line, d-orbital.

Sr₂SnO₄ showed an electronic structure similar to that of Ca₂SnO₄. The Sn5p orbital was hybridized with the O2p orbital in the lower part of the valence band, and the small mixing of the Sn5s orbital occurred at the upper part of valence band. The bottom of the conduction band was composed of the Sn5s orbital, and the contribution of the Sr4d orbital became larger in the upper part. The Sn5s and Sn5p orbitals overlapped much more energetically than those in Ca₂SnO₄. The band gap was estimated to be 2.96 eV. For Ba₂SnO₄, although the Sn5p was hybridized with the O2p orbital in the lower part of the valence band, the contribution was weaker, compared to that of Sr₂SnO₄ and Ca₂SnO₄. At the top of the valence band, the split-off state from the Sn5s orbital was also recognized. The bottom of conduction band was composed of the Sn5s orbital. The contribution of the Ba5d orbital was observed at a slightly higher energy region and that of the Sn5p orbital at a much higher energy region. The band gap was calculated to be 2.73 eV.

DOS for M₂SnO₄ (M = Ca, Sr, Ba) showed that the energy level of the p orbitals became higher in the order of Ca3p < Sr4p < Ba5p in relation to the size of p orbitals. In both the occupied and unoccupied bands, the mixing of Sn5s and Sn5p orbital was small for M = Ca, whereas it was larger for M = Sr and M = Ba. The Sn5p orbital was hybridized with the O2p orbital in the lower part of valence band, the energy of mixing range was relatively higher for M = Ca, and lower for M = Sr and M = Ba. Mixing magnitude in this range was large in the order of Ca > Sr > Ba. For the three stannates, the upper part of valence band was solely composed of the O2p orbital, but the split-off state of the Sn5s orbital contributed to the HOMO level, which evidently raised the top of valence band and reduced the band gap. In the conduction band, LUMO level was composed of Sn5s orbital for all three stannates. Hybridization with the Sn5p orbital and alkaline earth metal d orbital was different among the stannates. For M = Sr and M = Ba, the hybridization occurred over a whole energy range except for LUMO, but for M = Ca, it occurred at a higher energy range. The dispersion of conduction band was large for the three stannates, and there was a trend to became larger in the order of Ca < Sr < Ba.

5.2.5 Antimonates, M₂Sb₂O₇ (M = Ca, Sr), CaSb₂O₆ and NaSbO₃. Fig. 32 shows the energy band diagram and DOS for CaSb₂O₆. The Ca3p and O2s orbitals appear at core level below 20–25 eV from the HOMO level. The lower part of the valence band was composed of the O2p orbital hybridized with the Sb5s orbital, whereas the upper level had the O2p orbital. The HOMO level consisted of the O2p orbital only. The conduction band was mainly comprised of the Sb5s orbital, mixed with O2p orbital. A slightly higher level of conduction band mainly involved the contribution of the Ca 4s and slightly Sb5s + O2p orbitals. The LUMO level was composed of the Sb5s orbital and had large band dispersion. NaSbO₃ showed valence and conduction band structures similar to those of CaSb₂O₆. However, in the higher energy level of the conduction band, instead of Na3s orbital, the Sb5p orbital contributed. The dispersion of the conduction band was large. The band structure of Sr₂Sb₂O₇ was analogous to that of NaSbO₃ and CaSb₂O₆: the LUMO of the valence band was made up of the O2p orbital, whereas the HOMO was constituted by the Sb5s5p orbitals with large dispersion.

Fig. 32 Energy band diagram and DOS for CaSb₂O₆. The contribution of atomic orbitals (AOs) in DOS is shown: purple line, s-orbital; red line, p-orbital; green line, d-orbital.

Fig. 33 summarizes the characteristic electronic structures of d⁰ transition and d¹⁰ typical metal oxides. The d orbitals in the conduction band of transition metal oxides are stabilized into core level with increasing atomic number, because of increasing nucleus fields, and instead the sp band forms the conduction band in d¹⁰ typical metal oxides. Large dispersion of conduction band in d¹⁰ electronic configuration due to the broad sp orbitals is the most important advantage for photocatalysis, because it generates high mobility photoexcited electrons.

Fig. 33 A model of the electronic structures of d⁰ transition metal oxides and d¹⁰ typical metal oxides.

6. d¹⁰ metal nitrides

In aiming to develop visible light-driven photocatalysts, transition metal nitrides and oxynitrides have been used. Their valence bands are composed of N2p orbital with higher energy levels than O2p orbital. Thus, the band gaps are narrower when the conduction band levels remain unchanged. The d⁰ transition metal nitrides such as Ta₃N₅,⁷⁷ and oxynitrides such as TaON,⁷⁸ MTaO₂N (M = Ca, Sr, Ba),⁷⁹ and LaTiO₂N,⁸⁰ were employed, but little photocatalytic activity was observed for overall water splitting, although either H₂ or O₂ was produced in the presence of the sacrificial agents. Since d¹⁰ typical metal oxides have the advantage of forming large band dispersion at the conduction band to yield highly mobile photoexcited electrons, the employment of d¹⁰ typical metal nitrides to overall water splitting is an interesting idea. The DFT band calculation for β-Ge₃N₄ showed that the top of the valence bands consisted of the N2p orbital, whereas the bottom of the conduction bands was composed of Ge4s4p with large dispersion. A photocatalyst of 1 wt% RuO₂ loaded β-Ge₃N₄³⁷ was capable of splitting H₂O to H₂ and O₂. The activity gradually decreased in repeat runs, but became stable after prolonged reaction. The total amount of H₂ produced over 24 h was approximately ten times larger than the amount of β-Ge₃N₄ used in the experiment. The production ratio of H₂ to O₂ was 2, indicative of the stoichiometric production. These results show that RuO₂-dispersed β-Ge₃N₄ acts as a photocatalyst, representing the first successful example of a non-oxide photocatalyst for overall water splitting.

GaN itself was not active, but by doping divalent ions to GaN converted it to become active. For example, Mg²⁺-doped GaN showed high activity for the H₂ and O₂ production in the presence of RuO₂ as a cocatalyst.^38,39 The activity was induced by the doping of divalent ions, but not by doping tetravalent ions. The undoped GaN provided emission at 373 nm, which was close to the band gap, whereas Zn²⁺-doped GaN showed a broad band gap and strong emission at around 450 nm, assigned to electron transfer from free electron to the acceptor levels above the valence bands. This is indicative of p-type GaN formation by doping. Since the conduction band of GaN consisting of Ga4s4p has large dispersion enough to generate high mobility electrons, the improvement of behavior of holes by p-type GaN is considered to induce the photocatalytic ability.

Fig. 34 A schematic representation of the crystal structure of LiInGeO₄ and dipole moment of octahedral InO₆ and tetrahedral GeO₄.

Recently interesting results have been obtained with GaN. GaN forms a solid solution with ZnO such as (Ga_{1 −} xZn_x)(N_{1 − x}O_x). Although each GaN and ZnO had band gaps of ca. 3.0 eV, the band gap became narrower with increasing x, up to 2.4–2.8 eV. When combined with RuO₂ as a cocatalyst, (Ga_{1 −} xZn_x)(N_{1 − x}O_x) exhibited remarkably high photocatalytic activity, producing H₂ and O₂ in stoichiometric ratio.^40,41 The activity increased with increasing amount of RuO₂ and was highest at 5 wt% RuO₂ loading. The dependence of activity on wavelength of light showed that the photocatalyst was able to produce H₂ and O₂ even under visible-light, 400–500 nm.^40,41

7. Configuration mixed metal oxides

7.1 d¹⁰–d¹⁰ metal oxides

The employment of composite metal oxides with two kinds of the p-block metal ions with d¹⁰–d¹⁰ electronic configuration would be interesting for the development of efficient photocatalysts, since the resultant geometric and electronic structures are heavily changed. For the local structures, the presence of two different metal ions in the unit cell leads to geometric effects that cause large distortion in the tetrahedral and octahedral coordination of the metal oxides. For electronic structures, the hybridization of the sp-orbitals of two d¹⁰ metal ions may occur so have a significant effect on the density of states and energy dispersion in the conduction bands.

When the photocatalytic activity of d¹⁰–d¹⁰ metal oxide of AInGeO₄ (A = Li, Na) was compared with that of its d¹⁰ metal oxide, such as AInO₂ (A = Li, Na) and NaGeO₃, LiInGeO₄ had a remarkably larger activity than that of LiInO_2.⁸¹ Similarly, the activity of NaInGeO₄ was approximately 4-fold higher than that of NaInO₂. These results indicate that composite metal oxides with d¹⁰–d¹⁰ configuration have significantly larger activity than the corresponding metal oxides with d¹⁰ configuration. Fig. 34 shows the crystal structure of LiInGeO₄ which is formed by connection of the octahedral InO₆ and tetrahedral GeO₄ through corner oxygen atoms.⁸² InO₆octahedron are so heavily distorted that the InO₆ unit has a dipole moment of 3.0 D. The GeO₄ tetrahedron is also distorted, and has a dipole moment of 3.4 D. The dipole moment of InO₆ of LiInGeO₄ is slightly larger than that of SrIn₂O₄ and Sr_0.93Ba_0.07In₂O₄. Thus, the correlation between activity and dipole moment is clearly observed in LiInGeO₄ involving two d¹⁰ metal ions.

Fig. 35 Total and atomic orbital projected DOS for Li, In, Ge and O atom in the occupied bands from 0 to −10 eV (left) and the unoccupied bands from 0 to 15 eV (right). Ref. 81.

Fig. 35 shows the total density of states and atomic orbital projected density of states (AO PDOS) for Li, In, Ge, and O atoms for higher energy region (−10 to 0 eV) of the occupied bands and for the lower energy region (0–15 eV) of unoccupied bands.⁸¹ For the occupied bands, the O2p orbital existed over the region, divided into three parts with energy: the lower part was composed of the hybridized Ge4s + In5s orbitals; the middle part was the hybridized Ge4p + In5p orbitals; and, the upper part was the valence band that consisted of the O2p orbital, slightly mixed with Li2s2p orbitals. The HOMO level was formed by the O2p orbital only. For the unoccupied bands, the conduction band was mainly composed of Ge4s, In5s, and O2p orbitals and, with increasing energy the Ge4p, In5p, Li2s and Li2p orbitals contributed strongly in that order. In the contour map for the LUMO level, the polarized effects of the In5p and Ge4p indicated that LUMO was formed by Ge4s and In5s orbitals, slightly mixed with Ge4p, In5p and O2p orbitals. The conduction band showed very large band dispersion. Upon light irradiation, electron transfer occurs from the O2p orbital to the hybridized In 5s5p + Ge4s4p + O2p orbitals.

In a comparison with LiInO₂ (cf. 5.2.1), there is a similarity between LiInGeO₄ and LiInO₂ with contribution from the In5s5p, O2p and Li2s2p orbitals to the unoccupied bands. However, a clear difference is that the Li2s2p orbitals of LiInGeO₄ appears only at higher energy level of the unoccupied bands, and instead, Ge4s4p orbitals contribute to the conduction band. Thus, LUMO covers an energy range of ca 4 eV and causes large overlap leading to a large band dispersion. This is an advantage of d¹⁰–d¹⁰ configuration to the design of efficient photocatalysts.

7.2 d¹⁰s²–d⁰ metal oxides

Metal ions of Pb⁴⁺ and Bi⁵⁺ have d¹⁰ configuration, but no metal oxides involving these metal ions as a core ion have been found as photocatalysts for water splitting. This is partly because of the instability of Pb⁴⁺ and Bi⁵⁺ metal ions which are readily reduced to Pb²⁺ and Bi³⁺, respectively. However, since d¹⁰s² electronic configurations of Pb²⁺ and Bi³⁺ have closed shell characteristics, it is interesting to employ d¹⁰s² configuration metal oxides together with those of d⁰ (or d¹⁰) configuration.

Various types of tungstates such as Na₂W₄O₁₃, Bi₂W₂O₉, and Bi₂WO₆ showed H₂ production from H₂O/CH₃OH and O₂ from AgNO₃ aq. solution,^83,84 indicating that H₂ or O₂ production is possible in the presence of the sacrificial agents, but no tungsten oxides were reported to be active for overall water splitting. Recently, however, PbWO₄ was found to be promising.²⁵ This compound is a good example of d¹⁰s²(Pb²⁺)–d⁰ (W⁶⁺) configuration.

In water splitting on 1.0 wt% RuO₂-dispered PbWO₄ under Hg–Xe lamp irradiation,²⁶ the production of both H₂ and O₂ occurred starting at the onset of the reaction and proceeded photocatalytically with H₂/O₂ ratio of 2.1. On the other hand, for 1.0 wt% RuO₂-dispered CaWO₄, only a small amount of H₂ was produced without O₂ production. PbWO₄ has a scheelite-type structure with a lattice parameter of a = 0.5456 and c = 1.2020 nm, and α = β = γ = 90°⁸⁵ and is formed by a network of WO₄ tetrahedron. CaWO₄ has a structure similar to that of PbWO₄: it has also a scheelite-type structure formed by a network of WO₄ tetrahedron, and a lattice parameter of a = 0.5230 and c = 1.1384 nm, and α = β = γ = 90°.⁸⁶ In spite of a close similarity in the crystal structures, PbWO₄ was active, but CaWO₄ was inactive, which indicates that the role of Pb²⁺ in electronic structures is important.

Fig. 36 shows the energy band structure and DOS for PbWO₄ and CaWO₄. There was a clear difference in the dispersion of valence and conduction bands.²⁶ For PbWO₄, the dispersion was large at the top of the valence band and at the bottom of the conduction band. On the other hand, the dispersion for CaWO₄ was very small in both the valence and conduction bands. Fig. 37 shows AO PDOS for Pb, W and O atoms in the higher energy region (−8 to 0 eV) of occupied bands and the lower energy region (0 to 8 eV) of unoccupied bands for both PbWO₄ and CaWO₄. The O2p orbital mainly formed the valence band of PbWO₄, and there is a considerable mixing with the W5d orbital in the lower part of the O2p orbital. The Pb6s6p orbitals contributed over the entire energy region. At the top of the valence, the Pb6s orbital was largely hybridized with the O2p orbital. The bottom of the conduction band was composed of the W5d orbital hybridized with O2p and Pb6p orbitals. The narrow band gap for PbWO₄ was due to the hybridization of O2p with Pb6s orbital, causing a split-off state. These results indicate that the mixing of Pb6s with the O2p orbital contributes to the significant dispersion of the valence band of PbWO₄. For CaWO₄, however, the valence band was mainly composed of O2p, although the W5d orbital was mixed with the O2p orbital in the lower energy level. The bottom of the conduction band was composed of W5d and O2p orbitals. No contribution from Ca ion was observed in either the valence or conduction bands. This explains the small dispersion for both valence and conduction bands. Photoexcitation in PbWO₄ takes place by electron transfer from the HOMO level consisting of O2p + Pb6s orbitals to the LUMO level of W5d + Pb6p orbitals. The high capability of PbWO₄ for overall water splitting is associated with large dispersions in both the valence and conduction bands, which leads to the generation of photoexcited electrons and holes with high mobility.

Fig. 36 Band dispersion and DOS for PbWO₄ and CaWO₄. Ref. 26.

Fig. 37 Total and atomic orbital projected DOS for Pb, W, and O atom of PbWO₄ and CaWO₄. Ref. 26.

7.3 d¹⁰s²–d¹⁰ metal oxides

Mixing of d¹⁰s² and d¹⁰ configuration is interesting from a viewpoint of effects on valence and conduction bands. PbSb₂O₆ is a typical example with that configuration. As shown in Fig. 38, photocatalysts of RuO₂-dispersed NaSbO₃, CaSb₂O₆ and PbSb₂O₆, the photocatalytic activity is remarkably larger for PbSb₂O₆ than for the rest under Xe lamp irradiation. Fig. 39 shows the energy band diagram and DOS for PbSb₂O₆. The top of the valence band consists of the O2p orbital hybridized with the Pb6s orbital. Because of the contribution from the Pb6s orbital, the valence band shows a large dispersion. This is different from CaSb₂O₆ in which the top of the valence band consists of the O2p orbital only and has extremely small dispersion. The bottom of the conduction band for PbSb₂O₆ is made up of Sb5s5p orbitals hybridized with Pb6p orbital and has large dispersion. Thus, large band dispersion was induced for both the valence and conduction band by the contribution of the Pb6s orbital. This was able to generate high mobility electrons and holes and is responsible for the high photocatalytic activity.

Fig. 38 Photocatalytic activity of 1 wt% RuO₂-loaded NaSbO₃, CaSb₂O₆ and PbSb₂O₆. Photocatalyst: powder, photocatalyst amount, 0.25 g; Ar gas pressure, 13.3 kPa; light source, Xe lamp without filter; light intensity, 400 W.

Fig. 39 Energy band diagram and DOS for PbSb₂O₆. The contribution of atomic orbitals (AOs) is shown: purple line, s-orbital; red line, p-orbital; green line, d-orbital.

The employment of d¹⁰s² configuration to d⁰ and d¹⁰ configuration is a useful way to design efficient photocatalysts.

8. d⁰f⁰ metal oxides

In addition to transition and typical metal oxides, lanthanide metal oxide will be a suitable candidate in the periodic table as photocatalyst for water splitting. Ce(IV)O₂ has a [Xe] f⁰d⁰s⁰ electronic structure, which is a kind of closed shell structure of d⁰ electronic configuration. Although it was reported that CeO₂ had the ability to produce oxygen from a suspension of CeO₂ containing Ce⁴⁺ ions as the electron acceptor, no photocatalytic ability for overall water splitting was observed.⁸⁷ RuO₂-loaded undoped CeO₂ exhibited negligible photocatalytic activity under UV irradiation. Interestingly, however, the doping of Sr²⁺ to CeO₂ was found to convert CeO₂ into an active photocatalyst in the presence of RuO₂ as the cocatalyst.²⁸Sr²⁺-doped CeO₂ is the first example of a photocatalytically active lanthanide oxide for water splitting. In the DFT calculation for S²⁺-doped CeO₂, the occupied valence band was composed of O2p orbital. The unoccupied band consisted of Ce4f orbital band, and Ce5d orbital hybridized with Sr5s5p orbitals, appearing at an upper energy level. In UV-visible diffuse reflectance spectra, the absorption of undoped and Sr²⁺-doped CeO₂ occurred at around 400 nm and leveled off at 350 nm. However, no photocatalytic reaction occurred in a Pyrex glass reaction cell (available for wavelength >300 nm). The threshold absorption at around 400 nm is due to electron transfer from the O2p orbital (HOMO) to the unoccupied Ce4f orbital, but excited electrons failed to participate in the water splitting reaction. This indicates that the electrons transferred to the Ce5d orbital contribute to the reaction. A band model is shown in Fig. 40.

Fig. 40 A model of band structure and photocatalysis by Sr-doped CeO₂.

The stability of a crystal structure of metal oxides depends on the ionic size ratio of r(Mⁿ⁺)/r(O²⁻) where r(Mⁿ⁺) and r(O²⁻) are the ion size of a metal ion and an oxygen anion, respectively. The ideal ionic size ratio of r(Mⁿ⁺)/r(O²⁻) for a fluorite structure of MO₈ decahedra is 0.732, whereas the ratio of CeO₂ is 0.703. The difference is considerably large, and, to maintain the fluorite structure, a part of Ce⁴⁺ is converted to Ce³⁺ ion. The partial replacement of Ce⁴⁺ by larger and/or less positively charged cations stabilizes the fluorite structure of CeO₂ without the formation of Ce³⁺.⁸⁸ In view of its slightly larger ionic size and smaller positive charge, the doping of Sr²⁺ to CeO₂ results in the formation of completely Ce³⁺-free CeO₂. A reason why CeO₂ itself is photocatalytically inactive is explained in terms of the presence of Ce³⁺ which behaves as a strongly recombination center for photoexcited charges, and it is evident that its removal activates CeO₂ to a photocatalyst with d⁰f⁰ configuration.

Table 4 Factors for the activation of metal oxide photocatalysts

Type		Octahedral XO6, tetrahedral XO4
		Distorted	Undistorted
Electronic configuration of core metal ions	d^0	BaTi4O9, M2Ti6O13 (M = Ca, Sr), K2Ti4O9, Na2Ti3O7	PbWO4
	d^10	MIn2O4 (M = Ca, Sr), NaInO2, MGa2O4 (M = Sr, Ba), M2SnO4 (M = Ca, Sr), NaSbO3, M2Sb2O7 (M = Ca, Sr), CaSb2O6, LiInGeO4	ZnGa2O4
Interactions between core and neighbor metal ions		Ionic	Covalent
Activation		Geometric and electronic structures	Electronic structures

---

Table 5 Vacant and filled d electron configurations of photocatalysts for water splitting

d^n	n = 0	Transition metal oxides
		Lanthanide metal oxides
	n = 10	Typical metal oxides
		Typical metal nitrides
	1 < n < 9	unknown

---

Conclusions

For metal oxide and nitride photocatalysts for water splitting, the effects of distorted and unsymmetrical metal–oxygen octahedral/tetrahedral coordination are established for both d⁰ and d¹⁰ metal oxides involving alkaline metal and alkaline earth metals that have ionic properties. However, we have also shown that remarkable photocatalytic performance was induced even for metal oxides consisting of distortion-free octahdedral/tetrahedral coordination, if the metal ions having covalent bond characters such as Zn²⁺ and Pb²⁺ are involved in the metal oxides. This is due to the effects of electronic structures. This concept is summarized in Table 4. The advantage of typical metal oxides with d¹⁰ configuration is the formation of conduction bands consisting of largely dispersed broad s and p orbitals in contrast to the narrow flat d band of transition metal oxides. This provides photoexcited electrons with high mobility and is effective for photocatalytic reaction. When metal oxides are composed of d¹⁰ core metal ions, together with covalent metal ions as secondary ions, and are formed by the distorted metal–oxygen octahedral/tetrahedral coordination, they are predicted to make useful photocatalysts. As shown in Table 5, although there are a wide variety of metal oxides and nitrides, all the powder photocatalysts found so far for overall water splitting are confined to either completely vacant d orbital (d⁰, d⁰f⁰) or completely filled d orbital (d¹⁰, d¹⁰s²) electronic configurations, and no open shell structures, that is, dⁿ (9 > n > 1) electronic configuration, are known, to the best of our knowledge. The discovery of open shell structure photocatalysts is very challenging from a viewpoint of electronic structures and solving this would be greatly beneficial to the development of efficient photocatalysts for solar hydrogen.

Exactly a century ago, chemists tackled different difficult problems to find a new catalyst for the fixing of N₂ as NH₃. A large number and a wide variety of chemical compounds were examined as candidates, and finally, alumina-supported Fe + K₂O was found to be useful and has commercially been used for ammonia synthesis since 1914. Ironically, this discovery induced today's CO₂ problem, because it led to the petroleum-based society of the 20th century. One century later, an essential target to solve CO₂ problems and to obtain H₂ as an ideal energy source, is the establishment of solar-driven artificial photolysis, in particular, water splitting. Fundamental research in the last 10 years has developed not only d⁰ - but also d¹⁰-photocatalysts for water splitting. Although photocatalytic efficiency still remains at a low level to obtain solar hydrogen, progress has certainly been made in molecular-level understanding and widening of solid-state knowledge for water-splitting photocatalysts, and there is no doubt that continuing research, extending to chemical compounds having other electronic configurations or mixing of different electronic configurations, will cause a break-through in this field that will permit the establishment of a hydrogen society.

Acknowledgements

The author sincerely thanks Prof. H. Kobayashi, Department of Chemistry and Materials Technology, Kyoto Institute of Technology, for collaboration to DFT calculation. The present work was financially supported by a Grant-in-Aid for Scientific Research A (20246117) and on Priority Area (19033229) from The Ministry of Education, Culture, Science, and Technology and for Basic Research (No. 5) from The Ministry of Land, Infrastructure and Transport Government of Japan.

References

1. A. Fujishima and K. Honda, Nature, 1972, 238, 37.
2. K. Domen, A. Kudo and T. Onishi, J. Catal., 1986, 102, 92.
3. Y. Inoue, T. Kubokawa and K. Sato, J. Phys. Chem., 1991, 95, 4059.
4. S. Ogura, M. Kohno, K. Sato and Y. Inoue, Appl. Surf. Sci., 1997, 121/123, 521.
5. Y. Inoue, Y. Asai and K. Sato, J. Chem. Soc., Faraday Trans., 1994, 90, 797.
6. M. Kohno, T. Kaneko, S. Ogura, K. Sato and Y. Inoue, J. Chem. Soc., Faraday Trans., 1998, 94, 89.
7. T. Takata, Y. Furumi, K. Shinohara, A. Tanaka, M. Hara, J. N. Kondo and K. Domen, Chem. Mater., 1997, 9, 1063.
8. T. Takata, K. Shinohara, A. Tanaka, M. Hara, J. N. Kondo and K. Domen, J. Photochem. Photobiol., A, 1997, 106, 45.
9. S. Ogura, M. Kohno, K. Sato and Y. Inoue, J. Mater. Chem., 1994, 8, 2335.
10. S. Ogura, K. Sato and Y. Inoue, Phys. Chem. Chem. Phys., 2000, 2, 2449.
11. K. Sayama and H. Arakawa, J. Phys. Chem., 1993, 97, 531.
12. A. Kudo, A. Tanaka, K. Domen, K. Maruya, K. Aika and T. Onishi, J. Catal., 1998, 111, 67.
13. A. Kudo, H. Kato and S. Nakagawa, J. Phys. Chem. B, 2000, 104, 571.
14. Y. Miseki, H. Kato and A. Kudo, Chem. Lett., 2005, 34, 54.
15. Y. Miseki, H. Kato and A. Kudo, Chem. Lett., 2006, 35, 10562.
16. H. Kato and A. Kudo, Catal. Lett., 1999, 58, 153.
17. T. Ishihara, H. Nishiguchi, K. Fukamachi and Y. Takita, J. Phys. Chem. B, 1999, 103, 1.
18. H. Kato and A. Kudo, Chem. Phys. Lett., 1998, 295, 487.
19. H. Kato and A. Kudo, Chem. Lett., 1999, 1207.
20. M. Machida, T. Mitsuyama, K. Ikeue, S. Matsushima and M. Arai, J. Phys. Chem. B, 2005, 109, 7801.
21. K. Shimizu, Y. Tsuji, M. Kawakami, K. Toda, T. Kodama, M. Sato and Y. Kitayama, Chem. Lett., 2002, 31, 1158.
22. M. Machida, Int. J. Inorg. Mater., 2003, 13, 1433.
23. T. Kurihara, H. Okutomi, Y. Miseki, H. Kato and A. Kudo, Chem. Lett., 2006, 35, 274.
24. H. Otsuka, K. Kim, A. Kouzu, I. Takimoto, H. Fujimori, Y. Sakata, H. Imamura and K. Toda, Chem. Lett., 2005, 34, 822.
25. N. Saito, H. Kadowaki, H. Kobayashi, K. Ikarashi, H. Nishiyama and Y. Inoue, Chem. Lett., 2004, 133, 1452.
26. H. Kadowaki, N. Saito, H. Nishiyama, H. Koayashi, Y. Shimodaira and Y. Inoue, J. Phys. Chem. C, 2007, 111, 439.
27. H. Kadowaki, N. Saito, H. Nishiyama, Y. Shimodaira, H. Kobayashi and Y. Inoue, Chem. Lett., 2007, 36, 424.
28. H. Kadowaki, N. Saito, H. Nishiyama and Y. Inoue, Chem. Lett., 2007, 36, 440.
29. J. Sato, N. Saito, H. Nishiyama and Y. Inoue, J. Phys. Chem., 2001, 105, 6061.
30. J. Sato, N. Saito, H. Nishiyama and Y. Inoue, Chem. Lett., 2001, 868.
31. J. Sato, H. Kobayashi, N. Saito, H. Nishiyama and Y. Inoue, J. Photochem. Photobiol., A, 2003, 158, 139.
32. J. Sato, N. Saito, H. Nishiyama and Y. Inoue, J. Phys. Chem. B, 2003, 107, 7965.
33. T. Yanagida, Y. Sakata and H. Imamura, Chem. Lett., 2004, 33, 726.
34. K. Ikarashi, J. Sato, H. Kobayashi, N. Saito, H. Nishiyama and Y. Inoue, J. Phys. Chem., 2002, 106, 9048.
35. J. Sato, H. Kobayashi, K. Ikarashi, N. Saito, H. Nishiyama and Y. Inoue, J. Phys. Chem. B, 2004, 108, 4369.
36. J. Sato, N. Saito, H. Nishiyama and Y. Inoue, J. Photochem. Photobiol., A, 2002, 148, 85.
37. J. Sato, N. Saito, Y. Yamada, K. Maeda, T. Takata, J. N. Kondo, M. Hara, H. Kobayashi, K. Domen and Y. Inoue, J. Am. Chem. Soc., 2005, 127, 4150.
38. N. Arai, N. Saito, H. Nishiyama, Y. Inoue, K. Domen and K. Sato, Chem. Lett., 2006, 35, 796.
39. N. Arai, N. Saito, H. Nishiyama, K. Domen, H. Kobayashi, K. Sato and Y. Inoue, Catal. Today, 2007, 129, 407.
40. K. Maeda, T. Takata, M. Hara, N. Saito, Y. Inoue, H. Kobayashi and K. Domen, J. Am. Chem. Soc., 2005, 127, 8286.
41. K. Maeda, K. Teramura, D. Lu, T. Takata, N. Saito, Y. Inoue and K. Domen, Nature, 2006, 440, 295.
42. Y. Inoue, K. Sato and K. Sato, J. Chem. Soc., Faraday Trans. 1, 1989, 85, 1765.
43. P. S. Brody and F. Crowne, J. Electron. Mater., 1975, 4, 955.
44. Y. Inoue, M. Okamura and K. Sato, J. Phys. Chem., 1985, 89, 5184.
45. Y. Inoue, K. Sato, K. Sato and H. Miyama, J. Phys. Chem., 1986, 90, 2809.
46. Y. Inoue, K. Sato and K. Sato, Chem. Phys. Lett., 1986, 129, 79.
47. M. Kohno, S. Ogura, K. Sato and Y. Inoue, Stud. Surf. Sci. Catal. A, 1996, 101, 143.
48. M. Kohno, S. Ogura and Y. Inoue, Chem. Phys. Lett., 1997, 267, 72.
49. S. Ogura, M. Kohno, K. Sato and Y. Inoue, Appl. Surf. Sci., 1997, 121/122, 521.
50. M. Kohno, S. Ogura, K. Sato and Y. Inoue, J. Chem. Soc., Faraday Trans., 1997, 93, 2433.
51. W. Hofmeister and E. Tillmanns, Acta Crystallogr., 1984, C 40, 1510.
52. D. H. Templeton and C. H. Dauben, J. Chem. Phys., 1960, 32, 1515.
53. S. Andersson and A. D. Wadsley, Acta Crystallogr., 1962, 15, 194.
54. M. C. Payne, M. P. Teter, D. C. Allan, T. A. Arias and J. D. Joannopoulos, Rev. Mod. Phys., 1992, 64, 1045.
55. J. Sato, H. Kobayashi and Y. Inoue, J. Phys. Chem. B, 2003, 107, 7970.
56. R. von Schenck and H. Mueller-Buschbaum, Z. Anorg. Allg. Chem., 1973, 398, 24.
57. F. R. Criuckshank, D. MaK. Taylor and F. P. Glasser, J. Inorg. Nucl. Chem., 1964, 26, 937.
58. A. Lalla and H. Mueller-Buschbaum, Z. Anorg. Allg. Chem., 1990, 588, 117.
59. S. Geller, Acta Crystallogr., 1957, 10, 248.
60. E. Hubbert-Paletta, R. Hoppe and G. Kreuzburg, Z. Anorg. Allg. Chem., 1970, 379, 255.
61. H. Glaum, S. Voigt and R. Hoppe, Z. Anorg. Allg. Chem., 1991, 598, 129.
62. P. D'Arco, B. Silvi, C. Roetti and R. Orland, J. Geophys. Res. B, 1991, 96, 6107.
63. V. Kahlenbeerg, R. X. Fischer and C. S. J. Shaw, J. Solid State Chem., 2000, 153, 294.
64. V. Kahlenbeerg, R. X. Fischer and J. B. Praise, J. Solid State Chem., 2000, 154, 612.
65. V. V. Zhurov, A. A. Bush, S. A. Ivanov, S. Yu. Stefanovich and B. N. Romanov, Sov. Phys. Solid State, 1991, 33, 960.
66. Y. Inoue, Photocatalytic activity for water decomposition of RuO₂-dispersed p-block metal oxides with d¹⁰ electronic configuration in Metal Oxides Chemistry and Applications, ed. J. L. G. Fierro, Taylor & Francis, 2006, ch. 20, 623–659.
67. J. Hornstra and E. Keulen, Phillips Res. Rep., 1972, 27, 76.
68. C. Hang, M. A. Simonov and N. V. Belov, Sov. Phys. Crystallogr., 1970, 15, 387.
69. M. Troemel, Z. Anorg. Allg. Chem., 1969, 371, 237.
70. B. J. Kennedy, Aust. J. Chem., 1997, 50, 917.
71. M. A. Green, K. Prassides, P. Day and J. K. Stalick, J. Chem. Soc., Faraday Trans., 1996, 92, 2155.
72. W. A. Groen and D. J. W. Ijdo, Acta Crystallogr., 1988, C44, 782.
73. B. Wang, S. C. Chen and M. Greenblatt, J. Solid State Chem., 1994, 108, 184.
74. B. G. DeBoer, R. A. Young and A. Sakthivel, Acta Crystallogr., 1994, C50, 476.
75. S. K. Sampath, D. G. Kanhere and R. Pandey, J. Phys.: Condens. Matter, 1999, 11, 3635.
76. S. H. Wei and A. Zunger, Phys. Rev. B, 1988, 37, 8958.
77. G. Hitoki, A. Ishikawa, T. Takata, J. N. Kondo, M. Hara and K. Domen, Chem. Lett., 2002, 31, 736.
78. G. Hitoki, T. Takata, J. N. Kondo, M. Hara, H. Kobayashi and K. Domen, Chem. Commun., 2002, 1698.
79. D. Yamashita, T. Takata, M. Hara, J. N. Kondo and K. Domen, Solid State Ionics, 2004, 172, 591.
80. A. Kasahara, K. Nukumizu, G. Hitoki, T. Takata, J. N. Kondo, M. Hara, H. Kobayashi and K. Domen, J. Phys. Chem. A, 2002, 106, 6750.
81. H. Kadowaki, J. Sato, H. Kobayshi, N. Saito, H. Nishiyama, Y. Shimodaira and Y. Inoue, J. Phys. Chem., 2005, 109, 22995.
82. M. Touboul and P. Toledano, Acta Crystallogr., 1987, C 43, 2004.
83. A. Kudo and H. Kato, Chem. Lett., 1997, 421.
84. A. Kudo and H. Hijii, Chem. Lett., 1999, 1103.
85. J. M. Moreau, Ph. Galez, J. P. Peigneux and M. V. Korzhik, J. Alloys Compd., 1996, 238, 46.
86. L. G. Sillen, A. L. Nylander and Ark. Kemi, Mineral. Geol. A, 1943, 17, 1.
87. G. R. Bamwenda, K. Sayama and H. Arakawa, Chem. Lett., 1999, 1047.
88. S. Yabe, M. Shigeyoshi, M. Kazuyuki, S. Yoshida, R. Li and T. Sato, Int. Inorg. Mater., 2001, 3, 1003.

---