Photocatalytic and photoelectrocatalytic reduction of CO₂ using heterogeneous catalysts with controlled nanostructures

Abstract

The development of efficient artificial photocatalysts and photoelectrocatalysts for the reduction of CO₂ with H₂O to fuels and chemicals has attracted much attention in recent years. Although the state-of-the-art for the production of fuels or chemicals from CO₂ using solar energy is still far from practical consideration, rich knowledge has been accumulated to understand the key factors that determine the catalytic performances. This Feature article highlights recent advances in the photocatalytic and photoelectrocatalytic reduction of CO₂ with H₂O using heterogeneous semiconductor-based catalysts. The effects of structural aspects of semiconductors, such as crystalline phases, particle sizes, morphologies, exposed facets and heterojunctions, on their catalytic behaviours are discussed. The roles of different types of cocatalysts and the impact of their nanostructures on surface CO₂ chemisorption and reduction are also analysed. The present article aims to provide insights into the rational design of efficient heterogeneous catalysts with controlled nanostructures for the photocatalytic and photoelectrocatalytic reduction of CO₂ with H₂O.

---

Shunji Xie

Shunji Xie received his BS and MSc degrees from Hunan University of China in 2008 and 2011, and obtained his PhD degree from Xiamen University in 2014. He is currently a 2011-iChEM Fellow in the Collaborative Innovation Center of Chemistry for Energy Materials of Xiamen University. He focuses on photocatalysis and photoelectrocatalysis.

Qinghong Zhang

Qinghong Zhang received her BS and MSc degrees from Nanjing University of China in 1989 and 1992, and obtained her PhD degree from Hiroshima University of Japan in 2002. She joined Xiamen University in 2002, and was promoted to a full professor in 2010. Her research interests include the synthesis and characterization of novel catalytic materials.

Guodong Liu

Guodong Liu received his MSc degree from Fuzhou University of China in 2012, and obtained his PhD degree from Xiamen University in 2015 under the supervision of Prof. Ye Wang. His main research interests include the fabrication of novel photocatalysts for CO2 reduction.

Ye Wang

Ye Wang obtained his PhD degree in 1996 from Tokyo Institute of Technology. He then worked at Tokyo Institute of Technology, Tohoku University and Hiroshima University during 1996–2000. He was promoted to associate professor at Hiroshima University in 2001. He became a professor of Xiamen University in the August of 2001. He is currently the director of State Key Laboratory of Physical Chemistry of Solid Surfaces and the director of Institute of Catalysis Science and Technology of Xiamen University. The group of Professor Wang works on catalysis for efficient utilization of carbon resources including the selective transformation of methane, syngas, biomass and CO2 into fuels and chemicals.

---

1. Introduction

The diminishing of fossil resources and the growing emission of CO₂ have stimulated research activities towards the utilization of CO₂ as a carbon source for the production of fuels and chemicals.¹ However, CO₂ is one of the most stable molecules. Thermodynamically, the conversion of CO₂ is typically an endothermic process, which requires a significant energy input and/or a high-energy reactant. The dissociation energy of the C=O bond in CO₂ is ∼750 kJ mol⁻¹, higher than those of many other chemical bonds such as C–H (∼430 kJ mol⁻¹) and C–C (∼336 kJ mol⁻¹) bonds. Generally, a high activation barrier should be overcome for the transformation of CO₂. The thermodynamic stability and kinetic inertness of CO₂ make the transformation of CO₂ a highly challenging research theme in chemical science. The transformation of CO₂ can typically proceed via the following routes: (1) the co-feeding of a high-energy reactant such as H₂, unsaturated compounds, small-membered ring compounds (e.g., epoxides) and organometallic compounds; and (2) the supply of external energy such as solar energy or electrical energy.

The hydrogenation of CO₂ to methanol and formic acid has been intensively studied using heterogeneous and homogeneous catalysts.² CO₂ can be used for the production of dimethyl carbonate and cyclic carbonate through the reaction with methanol or epoxides.³ However, these reactions are thermodynamically limited and the equilibrium conversions are low under mild conditions. The use of CO₂ for the reformation of CH₄ to produce syngas (H₂/CO) could also proceed, but high temperatures and large energy-inputs are required.⁴

In nature, the photosynthesis of green plants transforms CO₂ with H₂O under sunlight to carbohydrates and O₂ at room temperature. Inspired by this natural process, a lot of research activities have been devoted to developing artificial or synthetic photocatalysts for the reduction of CO₂ with H₂O to organic compounds as well as CO. Some homogeneous or molecular photocatalysts, in particular metal complexes such as Re and Ru complexes, have been reported for the activation and reduction of CO₂, providing CO and formate as the major products.⁵ Fujishima, Honda and their co-workers reported a pioneering work for the photocatalytic reduction of CO₂ in water in the presence of heterogeneous semiconductor powders suspended in water.⁶ Since then, a large number of semiconductors such as TiO₂, BaLa₄Ti₄O₁₅, SrTiO₃, WO₃ nanosheets, NaNbO₄, KNbO₄, Sr₂Nb₂O₇, Zn₂GeO₄, Zn₂GaO₄, Zn₂SnO₄ and some metal sulphides have been reported to be capable of catalyzing the photoreduction of CO₂ with H₂O, and the pace has increased enormously in the recent five years.^7–21

The electrochemical reduction of CO₂ is another important route for the transformation of CO₂ to chemicals and fuels. Recently, encouraging results have been achieved for the electrochemical reduction of CO₂ to hydrocarbons (e.g., CH₄ and C₂H₄) and alcohols (e.g., CH₃OH and CH₃CH₂OH), which can be used as fuels or chemicals, in addition to CO and formate.^22–26 However, the catalytic activity, product selectivity and catalyst stability are still much far from the requirements for commercial consideration. Furthermore, a large amount of electricity must be supplied because of the high overpotentials for CO₂ reduction.

An alternative approach for CO₂ transformation is the photoelectrocatalysis, which integrates photocatalysis and electrocatalysis. There are some distinct advantages of photoelectrocatalytic reduction of CO₂ with H₂O. First, the use of solar energy can significantly lower the applied voltage, thus decreasing the electricity consumption. Second, the imposition of an external bias voltage can drive the separation of photogenerated electrons and holes, which is one of the most important steps determining the photocatalytic efficiency. Furthermore, the employment of separated half cells can avoid the re-oxidation of the reactive products such as methanol, which is known to be oxidized by photogenerated holes more easily than H₂O in a conventional photocatalytic system. The photoelectrocatalytic reduction of CO₂ was first reported by Halmann,²⁷ who used a p-type GaP as a photocathode and observed the formation of HCOOH, HCHO and CH₃OH in the electrolyte solution after the irradiation of GaP in combination with an applied bias voltage. This research direction has been attracting more and more attention in recent years.^28,29

In short, the photocatalysis and photoelectrocatalysis are two highly attractive routes for the reduction of CO₂ with H₂O to fuels and chemicals using solar energy. Although many research articles and review papers have already appeared,^7–21,28,29 the progress in these areas is largely behind that in the photocatalytic and photoelectrocatalytic splitting of H₂O. In most cases, the strategies typically adopted for H₂O splitting have also been exploited for CO₂ reduction. However, the activation of CO₂ is more difficult than that of H₂O. The reduction of CO₂ by photogenerated electrons may be a competitive process with the reduction of H₂O, thus making the catalyst requirement for the two processes quite different. Moreover, the products in the reduction of CO₂ are much more complicated than in the reduction of H₂O, which produces only H₂. A variety of products (e.g., CO, HCOOH, HCHO, CH₃OH, CH₄, higher hydrocarbons and higher alcohols) may be formed in the reduction of CO₂ with H₂O. Fundamental insights for the photoreduction of CO₂ in the presence of H₂O have not been well documented in the review papers reported to date. For example, many articles reported the synthesis of nanostructured semiconductors for the photocatalytic reduction of CO₂, but the knowledge on how the structure (such as the crystalline phase, particle size, morphology and preferentially exposed facet) determines the photocatalytic activity and selectivity is still lacking. Co-catalysts are known to play crucial roles in a photocatalytic process.³⁰ However, the roles of co-catalysts in enhancing the reduction of CO₂ or H₂O and the formation of different products are still ambiguous. The effect of nanostructures of dual co-catalysts on their functions in the photocatalytic reduction of CO₂ needs deeper elucidation. Furthermore, so far much attention has been paid to the energy-band engineering of semiconductors, which is helpful for enhancing the light harvesting and the electron–hole separation. The knowledge on how to manipulate the surface and interface structures to tune the CO₂/H₂O chemisorption and activation is relatively poorly accumulated.

This Feature article will highlight recent advances in the fundamental understanding of the effects of structures of both semiconductors and co-catalysts on the photocatalytic and photoelectrocatalytic reduction of CO₂ with H₂O and the formation of different products. Besides the efficiency of photogenerated electrons used for reduction reactions, we will pay particular attention to discuss the factors determining the selectivity of the reacted electrons for CO₂ reduction and the selectivity of reduction products.

2. Photocatalysis

2.1 Fundamental aspects for photocatalytic reduction of CO₂

It is generally accepted that the semiconductor-based photocatalysis for the reduction of CO₂ with H₂O involves three main steps (Fig. 1). In the first step, electron–hole pairs are generated when a semiconductor photocatalyst is illuminated by an appropriate light source with its energy equal or greater than the band-gap energy (E_g) of the semiconductor. Then, the generated electrons and holes migrate to the surface of the semiconductor or a co-catalyst in contact with the semiconductor in the second step (Fig. 1iia). It should be mentioned that only a fraction of carriers can reach the surface of the semiconductor or co-catalyst. A large fraction of electron–hole pairs recombine together (Fig. 1iib), with the energy being released in the form of heat or photons. In the third step, the photogenerated electrons reduce CO₂, which is adsorbed on catalyst surfaces, into CO, HCOOH, CH₃OH or CH₄, whereas the holes oxidize H₂O to O₂. The first and second steps are the same as those in the splitting of H₂O. The third step is peculiar to the reduction of CO₂. The reduction of H₂O could also proceed in competition with that of CO₂.

Fig. 1 Schematic illustration of the basic mechanism of photocatalytic reduction of CO₂ with H₂O on a semiconductor photocatalyst.

Similar to that for most photocatalytic reactions including H₂O splitting, the photocatalytic efficiency for CO₂ reduction is mainly determined by the efficiencies of the light harvesting, the charge separation and the surface reaction. In particular, because the charge recombination (∼10⁻⁹ s) is usually a much faster process as compared to the reaction process (∼10⁻³–10⁻⁸ s), the acceleration of electron–hole separation is of paramount importance for almost all the photocatalytic reactions. This is also a key issue in the photocatalytic reduction of CO₂. The present article will also highlight some useful strategies that can enhance CO₂ reduction by facilitating the electron–hole separation. In particular, the construction of semiconductors with specific nanostructures, the creation of heterojunctions and the use of suitable co-catalysts will be emphasized.

Some possible reactions related to the reduction of CO₂ and the corresponding redox potentials are summarized in Table 1. HCOOH, CO, HCHO, CH₃OH and CH₄ are all possible products from CO₂. Fig. 2 displays the band-edge positions of some typical semiconductors. Thermodynamically, the semiconductor, which is capable of catalyzing the reduction of CO₂ with H₂O, should possess the conduction-band edge higher or more negative than the redox potential for CO₂ reduction. At the same time, the valence-band edge should be lower or more positive than the redox potential for the oxidation of H₂O to O₂. The photogenerated electrons can thus be used for the reduction of CO₂ and the holes for the oxidation of H₂O to O₂. The formation of CH₄ and CH₃OH, which are eight- and six-electron reduction, are thermodynamically more feasible because of their less negative redox potentials. It should be noted that CH₃OH is known to be a much better hole scavenger than H₂O. Thus, CH₃OH cannot be obtained with considerable amounts if the reduction and oxidation take place on the same catalyst or in the same reactor. In our opinion, the photocatalytic reduction of CO₂ to CH₄ should be a more promising target if one-compartment reactor is employed.

Table 1 Some possible reactions related to photocatalytic conversion of CO₂ with H₂O

Reaction	E ^0 (V) vs. NHE (pH = 7)
CO2 + 2H^+ + 2e^− → HCOOH	−0.61
CO2 + 2H^+ + 2e^− → CO + H2O	−0.53
CO2 + 4H^+ + 4e^− → HCHO + H2O	−0.48
CO2 + 6H^+ + 6e^− → CH3OH + H2O	−0.38
CO2 + 8H^+ + 8e^− → CH4 + 2H2O	−0.24
2H^+ + 2e^− → H2	−0.41
H2O + 2h^+ → 1/2O2 + 2H^+	0.82

---

Fig. 2 Band-edge positions of some typical semiconductor photocatalysts relative to the energy levels of the redox couples involved in the reduction of CO₂.

Because the redox potentials for CO₂ reduction are close to that for the reduction of H₂O to H₂ (Fig. 2), the photocatalyst, which works for the reduction of CO₂, may also catalyse the reduction of H₂O (or H⁺) to H₂. In addition, the activation of H₂O is generally much easier than that of CO₂. Moreover, if CH₄ is the main target, the formation of CH₄ needs eight electrons, which would be more difficult than the formation of H₂ from H₂O, a two-electron reduction. Thus, the reduction of H₂O to H₂ is kinetically more favourable and is a strong competitive reaction with the reduction of CO₂. To evaluate the efficiency of CO₂ reduction, we can calculate the selectivity of the reacted electrons for CO₂ reduction as follows:

Selectivity for CO₂ reduction (%) = (number of electrons reacted for CO₂ reduction/number of electrons involved in all reduction reactions) × 100%.

Therefore, the surface manipulation is crucial to enhancing the photocatalytic reduction of CO₂ in the presence of H₂O. To obtain a high efficiency for CO₂ reduction, the catalyst surface should possess a high ability to chemisorb and activate CO₂ molecules in the presence of H₂O. The present article will also discuss useful strategies such as the modification of the semiconductor surface and the fabrication of suitable co-catalysts to enhance surface conversion of CO₂ kinetically.

Similar to the pioneering work reported by Fujishima and Honda,⁶ most of the reported studies adopted a solid–liquid interface reaction mode for photocatalytic reduction of CO₂.¹³ In such a mode (Fig. 3A), the photocatalyst particles were dispersed or suspended in the aqueous solution, where CO₂ was dissolved. The solid–liquid interface reactions occurred. However, the direct contact of the catalyst with liquid H₂O and the limited solubility of CO₂ in H₂O would cause the preferential adsorption of H₂O on catalyst surfaces, limiting the reduction of CO₂ by the photogenerated electrons. To increase the solubility of CO₂ in the aqueous phase, an alkaline medium has been employed for photocatalytic reduction of CO₂.^13,31 However, the formed CO₃²⁻ or HCO₃⁻ is typically more difficult to reduce than CO₂, and CO₃²⁻ is known as a good hole quencher.³¹ The use of a solid–gas or solid–vapour interface reaction mode (Fig. 3B) may overcome these limitations and may increase the rate of CO₂ reduction. For example, when a TiO₂ (P25) or Pt–TiO₂ catalyst was placed on a catalyst holder surrounded by CO₂ and H₂O vapour, the rate of CH₄ formation increased for more than 3 times as compared to that by dispersing the catalyst in water (Table 2).³² At the same time, the rate of H₂ formation from the reduction of H₂O decreased by using the solid–vapour reaction mode. The selectivity of reacted electrons for CO₂ reduction increased significantly from 11–19% to 40–56% (Table 2). This indicates that the solid–vapour reaction mode is more suitable for the preferential reduction of CO₂ in the presence of H₂O. It is of interest that the rate of photogenerated electrons used for all the reduction reactions did not significantly change by changing the reaction mode (Table 2). The use of solid–liquid or solid–vapour reaction mode mainly determined the selectivity of the reacted electrons for H₂O reduction or for CO₂ reduction.

Fig. 3 Two typical reaction modes for photocatalytic reduction of CO₂ in the presence of H₂O.³² Reproduced from ref. 32 with permission from American Chemical Society. (A) Solid–liquid interface reaction mode. (B) Solid–vapour interface reaction mode.

Table 2 Influence of reaction mode on photocatalytic performances of TiO₂ and 0.5 wt% Pt–TiO₂ for reduction of CO₂ in the presence of H₂O^a 32

Reaction mode	Catalyst	Formation rate (μmol g^−1 h^−1)			R (electron) (μmol g^−1 h^−1)	Select. for CO2 red. (%)
		CO	CH4	H2
a Reaction conditions: catalyst, 0.020 g; CO2 pressure, 0.2 MPa; H2O, 4.0 mL; irradiation time, 10 h. Reproduced from ref. 32 with permission from American Chemical Society.
Solid–liquid	TiO2	0.80	0.11	5.3	13	19
Solid–vapour	TiO2	1.2	0.38	2.1	10	56
Solid–liquid	Pt–TiO2	0.76	1.4	55	123	11
Solid–vapour	Pt–TiO2	1.1	5.2	33	111	40

---

2.2 Design and construction of semiconductors with controlled nanostructures for photocatalytic reduction of CO₂ with H₂O

The nanostructure of a semiconductor plays crucial roles in determining its photocatalytic performance. Many structural parameters can affect the kinetics of the photocatalytic processes including photo-absorption, electron–hole separation and surface reaction. Here, we will analyse the effects of structural aspects of semiconductors, including the crystalline phases, sizes of nanoparticles, morphologies, exposed facets and heterojunctions on the photocatalytic reduction of CO₂ with H₂O. Some results of the related photocatalytic systems are summarized in Table 3.

Table 3 Photocatalytic systems demonstrating the effect of structural aspects of semiconductors on the performance for photocatalytic reduction of CO₂

Photocatalyst	Formation rate (μmol g^−1 h^−1)	Light source	Ref.
Effect of crystalline phase
Brookite TiO2 (He treated at 493 K)	CO: 2.8, CH4: 0.32	Xe lamp	37
Pt-cubic NaNbO3	CH4: 4.9	Xe lamp	38
Pt-orthorhombic NaNbO3	CH4: 2.5		39
TiO2 (anatase + rutile)	CH4: 34	Hg lamp	40
P25	CH4: 14
TiO2 (73% anatase + 27% brookite)	CH3OH: 0.59	Xe lamp (UV-vis)	43
P25	CH3OH: 0.18
TiO2 (75% anatase + 25% brookite)	CO: 2.1	Solar simulator (Oriel)	44
P25	CO: 1.3
Effect of particle size
Anatase TiO2 (optimized size: 14 nm)	CH4: 0.17 μmol m^−2	Hg lamp (254 nm)	47
Effect of morphology
Pt–Cu/N-doped TiO2 nanotube arrays	Hydrocarbons: 111 ppm cm^−2 h^−1, H2: 160 ppm cm^−2 h^−1	AM 1.5	48
1% RuO2 and 1% Pt co-loaded Zn2GeO4 nanoribbons	CH4: 6.7	Xe lamp	49
Hexagonal Zn2GeO4 nanorods	CO: 17.9 ppm h^−1, CH4: 3.5 ppm h^−1	Xe lamp (UV-vis)	50
1% RuO2 and 1% Pt co-loaded Zn1.7GeN1.8O nanosheaves	CH4: 4.4	Xe lamp (>420 nm)	51
1% RuO2 loaded Cd2Ge2O6 nanowires	CH4: 0.72	Xe arc lamp (UV-vis)	52
1% Pt loaded In2Ge2O7 nanowires	CO: 0.51	Xe lamp	53
1% RuO2 + 1% Pt co-loaded Na2V6O16 nanoribbons	CH4: 0.18	Xe lamp (>420 nm)	54
Fe2V4O13 nanoribbons	CH4: 0.7 (UV-vis), CH4: 0.2 (Vis)	Xe lamp	55
Ultrathin W18O49 nanowires	CH4: 667 ppm g^−1 h^−1	Xe lamp (>420 nm)	56
G-Ti0.91O2 hollow spheres	CH4: 1.2, CO: 9.0	Xe lamp	57
P25	CH4: 0.8, CO: 0.2
0.5% G/TiO2 nanosheets	CH4: 9.3 μmol m^−2 h^−1	365 nm	58
ZnGa2O4 nanosheet microspheres	CH4: 6.9 ppm h^−1	Xe lamp	59
1% RuO2–1% Pt co-loaded Zn2SnO4 nanosheet micro-octahedrons	CH4: 36 ppm g^−1 h^−1	Xe lamp	60
HNb3O8 nanobelts	CH4: 3.85	Xe lamp	61
P25	CH4: 0.37
KTaO3 nanoflakes	CO: 62 ppm g^−1 h^−1, H2: 1323 ppm g^−1 h^−1	Xe lamp	62
SrNb2O6 nanoplates	CH4: 3.3, CO: 16.6, H2: 6.5	Xe lamp (320–780 nm)	63
P25	CH4: 0.42, CO: 1.4, H2: 2.4
WO3 nanosheets (thickness: ∼4–5 nm)	CH4: 1.1	Xe lamp (>420 nm)	64
Rectangular sheet-like WO3 with dominant {002}	CH4: 0.34	Xe lamp	65
Bi2WO6 nanoplates with dominant {001}	CH4: 1.1	Xe lamp (>420 nm)	66
Effect of exposed facet
1% Pt/TiO2 rods with dominant {010}	CH4: 2.5	Xe lamp	72
1% Pt/P25	CH4: 2.0
Hollow anatase TiO2 mesocrystals with dominant {101} facets	CH4: 1.2	Xe lamp	73
Anatase TiO2 nanosheets exposed with 95% {100} facets	CH4: 5.8 ppm g^−1 h^−1	UV-vis light (>300 nm)	74
Anatase TiO2-{010}	CH4: 1.2	Hg lamp	75
Anatase TiO2-{101}	CH4: 0.74
Anatase TiO2-{001}	CH4: 0.19
Fluorinated anatase TiO2 nanosheets {001} : {101} = 72 : 28	CH4: 0.20, CH3OH: 0.18, H2: 0.10	Hg lamp	76
Anatase TiO2 with {001} : {101} = 58 : 42	CH4: 1.35	Xe lamp	77
P25	CH4: 0.38
Hexahedron-prism anchored octahedron CeO2{100} + {111}	CH4: 1.12	Xe lamp	78
Effect of heterojunction
PbS–Cu/TiO2	CO + CH4 + C2H6: 1.71	Xe lamp (>420 nm)	79
Pt-g-C3N4/NaNbO3	CH4: 6.4	Xe lamp (>420 nm)	80
ZnO–TiO2	CH4: 55	Xe lamp	81
CuO–TiO2−xNx	CH4: 41 ppm g^−1 h^−1	AM 1.5	82
Ag3PO4/g-C3N4	CO: 44, CH3OH: 9	Xe lamp	83

---

2.2.1 Effect of crystalline phases of semiconductors. Different crystalline phases of a semiconductor may have different energy-band and electronic structures. Thus, the crystalline phases may affect the efficiencies of light absorption and photogenerated-carrier separation. Moreover, the difference in surface structures of different crystalline phases may also result in different surface reactivities. Many studies have investigated the influence of the crystalline phase of a semiconductor (e.g., TiO₂ or BiVO₄) on its photocatalytic behaviours in water splitting and degradation of organic pollutants.^19,33–36 However, such studies on the photocatalytic reduction of CO₂ are limited.

Li and co-workers recently compared photocatalytic performances of TiO₂ nanocrystals with three different phases, i.e., anatase, rutile and brookite, which were synthesized by hydrolysis followed by the hydrothermal method, for the reduction of CO₂ with H₂O vapour.³⁷ CO was produced as a major product together with a minor amount of CH₄ over all these TiO₂ catalysts. The CO formation rate was in the order of anatase > brookite > rutile if the catalysts were only calcined in air (Fig. 4). The difference in catalytic activities for the samples with different crystalline phases may arise from a complicated interplay of light harvesting, electron–hole separation and surface reaction. Interestingly, after these catalysts were further pretreated in He flow at 493 K for 1.5 h, their activities for (CO + CH₄) formation were all enhanced. The enhancement was particularly significant for brookite and anatase, and the activity sequence changed to brookite > anatase > rutile after heat pretreatment in He (Fig. 4). This enhancement could be attributed to the creation of surface defective sites including Ti³⁺ and oxygen vacancies. These surface defective sites were proposed to be responsible for the activation of CO₂. The defective site could be generated more easily on brookite and anatase, and thus these two catalysts exhibited much higher reaction rates after He pretreatment. The pretreated brookite and anatase possessed similar densities of defective sites, but their activities were different. Further in situ FT-IR studies clarified that CO₂⁻ was formed as an intermediate on both brookite and anatase surfaces, and the further reduction of CO₂⁻ in the presence of H⁺ resulted in CO. CO₂⁻ might react with H⁺ more facilely on brookite surfaces, forming HCOOH as another reaction intermediate and providing another reaction channel. Thus, the difference in the surface reactivity of TiO₂ nanocrystals with different crystalline phases mainly determined the observed difference in activity for CO₂ reduction.

Fig. 4 Production of CH₄ and CO on TiO₂ samples with different crystalline phases with and without heat treatment under He flow.

Another example of the crystalline-phase effect was reported by Ye and co-workers. They synthesized NaNbO₃ with cubic and orthorhombic phases.^38,39 UV-vis spectroscopic measurements indicated a relatively lower band-gap energy for the cubic phase (3.29 eV versus 3.45 eV for the orthorhombic phase). CH₄ could be formed over these two catalysts after the loading of the Pt co-catalyst. The rate of CH₄ formation over Pt-promoted cubic NaNbO₃ was about twice higher than that over Pt-promoted orthorhombic NaNbO₃. Besides band-gap energy, the difference in electronic structures of the two phases may also influence the photocatalytic performance. It seems that the cubic phase favours the electron transfer.

The semiconductor containing mixed phases has been reported to show better photocatalytic performances than the pure-phase material. The interface between different phases, which possess different band-edge positions, may form a phase junction, and such a junction can promote the separation of photogenerated charge carriers (Fig. 5).^40–42 For example, it has been demonstrated that photogenerated electrons migrate from rutile to anatase, which possesses a lower conduction-band edge, in the anatase–rutile mixed TiO₂.^40,41 As compared with pure anatase and P25, such a mixed-phase TiO₂ with optimized interfaces showed much higher CH₄ formation activity in the photocatalytic reduction of CO₂ with H₂O in a solid–liquid reaction mode.⁴⁰ Recently, anatase–brookite mixed TiO₂ nanoparticles were successfully synthesized through the hydrothermal method.^43,44 In a solid–liquid mode photocatalytic reduction of CO₂, the analysis of CH₃OH formed in the liquid phase showed that the anatase–brookite mixed TiO₂ showed better CH₃OH formation activity than anatase and rutile alone.⁴³ The formation of CO and a trace amount of CH₄ were observed in the solid–vapour mode reaction using mixed-phase TiO₂ with different anatase/brookite ratios.⁴⁴ The catalyst composed of 75% anatase and 25% brookite showed the highest CO formation rate (2.1 μmol g⁻¹ h⁻¹), which was ∼3 and 1.75 times larger than those for pure brookite and anatase, respectively. The activity of this catalyst was also higher than that of P25 (1.3 μmol g⁻¹ h⁻¹), an anatase–rutile mixed TiO₂. The interface between the two phases should play a crucial role in accelerating the electron–hole separation. Therefore, an optimized photocatalytic activity for CO₂ reduction could be expected by tuning the interface of mixed phases of a semiconductor.

Fig. 5 Illustration of phase heterojunctions for promoting the charge separation for the photocatalytic reduction of CO₂ with H₂O.

2.2.2 Effect of sizes of semiconductor nanoparticles. The size of semiconductor particles also influences the photocatalytic behaviours. Optimum particle sizes of TiO₂ have been observed for several different photocatalytic reactions because of the balance among competing effects of surface area, charge-carrier dynamics and light-harvesting efficiency.^45,46 Kočí et al. investigated the catalytic behaviours of TiO₂ with particle sizes ranging from 4.5 to 29 nm for the photocatalytic reduction of CO₂ with H₂O using a solid–liquid reaction mode.⁴⁷ CH₄ was formed as a major product with a small amount of CH₃OH. An optimum size of TiO₂ particles was found for the formation of CH₄ and CH₃OH. TiO₂ particles with an average particle size of 14 nm afforded the highest rates of CH₄ (9.5 μmol g⁻¹ in 24 h) and CH₃OH (1.15 μmol g⁻¹ in 24 h) formation (Fig. 6).⁴⁷ The smaller-sized TiO₂ has a higher surface area and larger density of surface sites, which are favourable for the adsorption and conversion of CO₂. However, the decrease in TiO₂ particle sizes may cause changes in the electronic and optical structures. For example, the decrease in the particle size may increase the band-gap energy, decreasing the efficiency of light harvesting.

Fig. 6 Effect of TiO₂ particle sizes on the formation of CH₄ and CH₃OH in the photocatalytic reduction of CO₂ with H₂O.⁴⁷

2.2.3 Effect of morphology of semiconductors. Semiconductors with different morphologies can significantly affect their photocatalytic behaviours. The morphology may not only influence the electronic and surface structures, thus controlling the energy-band position and surface reactivity, but also may affect the transfer of photogenerated carriers.

One-dimensional (1D) semiconductors have been synthesized and exploited for the photocatalytic reduction of CO₂. For example, nitrogen-doped TiO₂ nanotube arrays catalysed the photocatalytic reduction of CO₂ with H₂O vapour in the presence of Pt and/or Cu co-catalysts under outdoor sunlight irradiation.⁴⁸ H₂, CO and hydrocarbons (CH₄ as well as small amounts of higher paraffins and olefins) were formed. The rate of hydrocarbon formation reached 150 μL g⁻¹ h⁻¹ under outdoor sunlight, even higher than those reported for Pt/TiO₂ under UV illumination. The doping of nitrogen could extend the light absorption to the visible region. It was proposed that the thinner wall of the nanotube morphology could facilitate the transfer of the photogenerated carriers to the surface.

Zou and co-workers reported a series of 1D composite metal oxide semiconductors for the photocatalytic reduction of CO₂ with H₂O.^49–55 For example, they synthesized Zn₂GeO₄ nanoribbons and nanorods using a solvothermal method in ethylenediamine/water and a low-temperature solution-phase methods, respectively.^49,50 These Zn₂GeO₄ samples with 1D morphology showed higher activities for CO and CH₄ formation during the photocatalytic reduction of CO₂ under UV-vis irradiation than that prepared by the conventional high-temperature solid-state reaction (SSR) method. The catalytic performance could be enhanced by further loading Pt and/or RuO₂ as co-catalysts. A nitrogen-doped Zn₂GeO₄, i.e., Zn_1.7GeN_1.8O₄, with nanosheave morphology, which contained numerous closely packed nanorods, was demonstrated to be capable of catalysing the formation of CH₄ from CO₂ under visible-light irradiation after loading Pt and RuO₂ co-catalysts.⁵¹ The nitrogen doping enhanced the light absorption in the visible-light region. The apparent quantum yield of CH₄ evolution for this catalyst at a wavelength of 420 ± 15 nm was measured to be 0.024%.⁵¹ Similarly, Cd₂Ge₂O₆ and In₂Ge₂O₇ nanowires in the presence of Pt and/or RuO₂ co-catalysts also showed better catalytic performances for the photocatalytic reduction of CO₂ to CH₄ under UV-vis irradiation than their counterparts prepared by the SSR method.^52,53 The higher photocatalytic activity of 1D semiconductors as compared to that of the SSR sample could be attributed to the following reasons: (1) the ultra-long dimensions in one direction may provide a sufficiently spacious transport channel for charge separation; (2) the 1D geometry may allow charge carriers to move rapidly from the interior to the surface to participate in surface reactions; (3) the decrease of the lateral dimensions to the nanoscale offers a high specific surface area and a high surface site density.

Zou and co-workers further succeeded in synthesizing vanadate-based visible-light responsive semiconductors, i.e., Na₂V₆O₁₆ and Fe₂V₄O₁₃, with nano-ribbon morphology.^54,55 The band-gap energies of Na₂V₆O₁₆ and Fe₂V₄O₁₃ nanoribbons were 1.93 and 1.78 eV, respectively. These semiconductors possessed conduction-band edges more negative than the redox potential for the reduction of CO₂ to CH₄, thus being capable of catalysing the reduction of CO₂ with H₂O under visible-light illumination (Fig. 7). The rate of CH₄ formation was 0.2–0.5 μmol g⁻¹ h⁻¹ over Pt- and RuO₂-co-loaded Na₂V₆O₁₆ or Pt-loaded Fe₂V₄O₁₃ catalysts under visible light (λ ≥ 420 nm).^54,55

Fig. 7 Band-edge positions of Na₂V₆O₁₆ and Fe₂V₄O₁₃ nanoribbons, two visible-light responsive photocatalysts for reduction of CO₂ in the presence of H₂O.^54,55

Ye and co-workers synthesized ultra-thin W₁₈O₄₉ nanowires with diameters of ∼0.9 nm and found that this 1D material was efficient for the photocatalytic reduction of CO₂ to CH₄ under visible light (λ ≥ 420 nm) irradiation, whereas the commercial WO₃ sample was completely inactive.⁵⁶ The W₁₈O₄₉ nanowires showed strong absorption in visible and near infrared regions due to the oxygen vacancies. It was demonstrated that the activity decreased significantly when the density of oxygen vacancies was lowered by H₂O₂ treatment. Although the mechanism is still unclear, the presence of large amounts of oxygen vacancies should play crucial roles.

Two-dimensional (2D) semiconductors also showed unique catalytic behaviours for the photocatalytic reduction of CO₂ with H₂O. Ti_0.91O₂ nanosheets with lateral dimensions of 0.1–1 μm and a thickness of ∼0.75 nm have been synthesized for the photocatalytic reduction of CO₂.⁵⁷ The 2D Ti_0.91O₂ nanosheets provided CH₄ as an exclusive product in the photoreduction of CO₂. The rate of CH₄ formation was 1.41 μmol g⁻¹ h⁻¹ under UV-vis irradiation on 2D Ti_0.91O₂. P25 provided CH₄ and CO with rates of 0.69 and 0.16 μmol g⁻¹ h⁻¹, respectively, under the same conditions. The higher activity of Ti_0.91O₂ nanosheets probably arose from the more facile migration of photogenerated charge carriers onto the catalyst surface. The fabrication of hollow spheres consisting of molecular-scale alternating Ti_0.91O₂ nanosheets and graphene nanosheets by combining a layer-by-layer assembly technique and a microwave irradiation technique could further improve the photocatalytic performance. CO became a major product, and the rates of CH₄ and CO formations were 1.14 and 8.91 μmol g⁻¹ h⁻¹, respectively, over the Ti_0.91O₂–graphene composite. The compact stacking of ultrathin Ti_0.91O₂ and graphene nanosheets may enable the transfer of the photogenerated electrons from Ti_0.91O₂ nanosheets to graphene rapidly, further enhancing the separation of charge carriers. Furthermore, the hollow structure might act as a photon trap to allow the multi-scattering of incident light, thus enhancing the light absorption.

Liang et al. investigated the effect of dimensionality of carbon nanomaterials on photocatalytic performances of nanocarbon–TiO₂ composites for the reduction of CO₂ with H₂O.⁵⁸ The comparative study revealed that the composite of 2D graphene and 2D TiO₂ nanosheets showed better CH₄ formation activity than the 1D–2D CNT–TiO₂ nanosheet composite under UV illumination. This is likely attributable to the more intimate contact and superior electronic coupling in the 2D–2D graphene–TiO₂ composite. On the other hand, 1D CNTs were demonstrated to be more effective photosensitizers, leading to better CH₄ formation activity of the 1D–2D CNT–TiO₂ nanosheet under visible-light illumination.⁵⁸

Several studies have succeeded in synthesizing hierarchical structures consisting of semiconductor nanosheets or nanoplates for the photocatalytic reduction of CO₂ with H₂O. For example, uniform hierarchical ZnGa₂O₄ microspheres scaffolded from nanosheets with a thickness of ∼6 nm were found to be more efficient for CH₄ formation than the mesoporous ZnGa₂O₄.⁵⁹ A hexagonal nanoplate-textured micro-octahedron Zn₂SnO₄ showed higher CH₄ formation activity than Zn₂SnO₄ atactic particles or micro-octahedra without hierarchical structure.⁶⁰

Alkali and alkaline earth metal niobates and tantalates are attractive for the photocatalytic reduction of CO₂ with H₂O, since these semiconductors are not only stable and non-toxic but also possess excellent CO₂ chemisorption abilities. Recent studies disclosed that these semiconductors also displayed significant morphological effects.^61–63 For instance, KTaO₃ nanoflakes synthesized by a solvothermal method in the hexane–H₂O mixture were found to exhibit higher rates of CO and H₂ formation in the photoreduction of CO₂ with H₂O than the cubic-like KTaO₃ synthesized by the SSR method or by a similar solvothermal method but in a pure ethanol mixture.⁶² The higher surface area and the shorter charge-transfer pathway are believed to be responsible for the higher activity of the KTaO₃ nanoflakes. We succeeded in synthesizing SrNb₂O₆ nanoplates with a thickness of ∼12 nm by a simple hydrothermal method.⁶³ 2D SrNb₂O₆ with nanoplate morphology showed higher CO and CH₄ formation rates than SrNb₂O₆ with nanoparticulate morphology but with a similar specific surface area. The increased capability of electron–hole separation for the nanoplate was evidenced by the higher photocurrent density in the transient photocurrent response measurements. It was demonstrated that the SrNb₂O₆ nanoplate without any co-catalysts showed a higher rate for the production of (CO + CH₄) (19.9 μmol g⁻¹ h⁻¹) than a 0.5 wt% Pt–TiO₂ reference catalyst (6.9 μmol g⁻¹ h⁻¹) under UV-vis irradiation (λ = 300–780 nm).⁶³ Moreover, the selectivity of the reacted electrons used for CO₂ reduction with the SrNb₂O₆ nanoplate was much higher (82%) than that with the 0.5 wt% Pt–TiO₂ (40%). This is probably because of the higher chemisorption ability of the SrNb₂O₆ nanoplate toward CO₂.

Another interesting example of the morphological effect is the unique catalytic behaviour of 2D WO₃ nanocrystals. Normally, the photogenerated electrons in the conduction band of WO₃ could not be used for the reduction of CO₂ or H₂O due to the lower (more positive) band-edge position (Fig. 2). Almost no CH₄ was formed during the photocatalytic reduction of CO₂ with H₂O over commercial WO₃ microcrystals under visible-light irradiation.⁶⁴ However, the ultra-thin 2D WO₃ nanosheets with a thickness of 4–5 nm synthesized using a solid–liquid phase arc discharge method in an aqueous solution could catalyze the photoreduction of CO₂ with H₂O into CH₄. The amount of CH₄ was 16 μmol g⁻¹ over the WO₃ nanosheets under visible-light irradiation (λ > 420 nm) for 14 h.⁶⁴ The WO₃ nanosheets showed a significant blue shift in the UV-vis absorption spectrum as compared to the commercial WO₃ probably due to the quantum-size effect. This increased the band-gap energy from 2.63 eV for the commercial WO₃ microcrystals to 2.79 eV for the WO₃ nanosheets. The edge of the conduction band of WO₃ nanosheets was estimated to be −0.42 V (versus NHE), becoming more negative than the CO₂/CH₄ redox potential (−0.24 V) (Fig. 8). Thus, the reduction of CO₂ to CH₄ by the photogenerated electrons in the conduction band of WO₃ nanosheets became feasible. The same phenomenon was observed by Xie et al., who successfully synthesized a quasi-cubic-like WO₃ crystal with a nearly equal percentage of {010}, {200} and {020} facets, and a rectangular sheet-like WO₃ crystal with a predominant {002} facet by controlling acidic hydrolysis of crystalline tungsten boride.⁶⁵ The cubic-like WO₃ was inactive for the photocatalytic reduction of CO₂, whereas the sheet-like WO₃ nanocrystals provided a CH₄ formation rate of 0.34 μmol g⁻¹ h⁻¹ under visible-light (λ > 400 nm) irradiation.

Fig. 8 Effect of morphology of WO₃ on the band-edge positions and the corresponding rates of CH₄ formation for photocatalytic reduction of CO₂ with H₂O.^64,65aRef. 64. ^bRef. 65.

The elevated conduction-band edge was also proposed to contribute to CO₂ reduction for the sheet-like WO₃ nanocrystals (Fig. 8).⁶⁵

Similarly, 2D Bi₂WO₆ could also be exploited for the photocatalytic reduction of CO₂.⁶⁶ Uniform Bi₂WO₆ square nanoplates with a thickness of ∼9.5 nm were synthesized using a hydrothermal method in the presence of oleylamine. This Bi₂WO₆ nanoplate possesses a band-gap energy of 2.75 eV with the conduction-band edge at −0.34 V (versus NHE), which is more negative than the CO₂/CH₄ redox potential (−0.24 V). The rate of CH₄ formation was ∼1.1 μmol g⁻¹ h⁻¹ under visible-light illumination. Under the same conditions, Bi₂WO₆ synthesized by the SSR method was inactive.⁶⁶

2.2.4 Effect of exposed facets of semiconductor nanocrystals. The morphology of a nanocrystal controls its preferentially exposed facets. Many of the morphological effects in catalysis could be interpreted by the facet-dependent reactivity or structure sensitivity.^67,68 In the previous section, we mainly discussed the effect of the morphology of nanocrystals without touching the detailed exposed crystal facets. Here, we highlight the research advances that clearly correlate the exposed facets with the activity in the photocatalytic reduction of CO₂.

Recent studies have suggested that the controlled fabrication of semiconductor nanocrystals with preferentially exposed facets is an effective strategy to enhance the photocatalytic performance. For example, as mentioned above, the 2D Bi₂WO₆ nanoplate with dominant {001} facets exposed could catalyze the photocatalytic reduction of CO₂, whereas Bi₂WO₆ synthesized by the SSR method was inactive.⁶⁶ So far, more effort has been devoted to TiO₂ nanocrystals to elucidate the effect of exposed facets on photocatalytic behaviours. Earlier work suggested that the {001} facet, the high-energy-surface of anatase, contributed more to photocatalysis.^69–71 Cheng and co-workers found that 1D anatase TiO₂ nanorods with dominant {010} facets exhibited a better photocatalytic performance for the reduction of CO₂ into CH₄ than P25.⁷² It has been clarified that the conduction-band edge is higher (more negative) for the {010}-dominant TiO₂ nanorods. Moreover, the interactions of CO₂ and H₂O with {010} facets are stronger than {001} and {101} facets. These may contribute to the better catalytic performance of the {010} facet. In a subsequent work, the same group reported that hollow anatase TiO₂ single crystals and mesocrystals with dominant {101} facets could efficiently catalyze the formation of CH₄ from CO₂ than solid TiO₂ single crystals.⁷³ It seems that the hollow structure plays a pivotal role in the photocatalytic reduction of CO₂ probably because of the shortened bulk diffusion length of charge carriers.⁷³ Ye and co-workers claimed that the TiO₂ ultra-thin nanosheets, which has a thickness of ∼2 nm and exposed 95% {100} facets, showed better catalytic performance for the reduction of CO₂ to CH₄ than TiO₂ cuboids with 53% {100} facets exposed.⁷⁴ These discrepancies suggest that the direct comparison of catalysts with different preferentially exposed facets should be careful because the parameter other than the exposed facet may be the determining factor.

A recent study compared the photocatalytic activities of anatase TiO₂ nanocrystals with different exposed low-index facets, i.e., {001} (nanoplate), {101} (nano-octahedron) and {010} (nanorod), for the reduction of CO₂ with H₂O to CH₄ in the vapour phase.⁷⁵ The percentage of each kind of facets in the synthesized TiO₂ nanocrystals reached 90%. The activity for photoreduction of CO₂ to CH₄ was found to decrease in the order of {010} > {101} > {001} (Fig. 9). The result that the 1D nanorod TiO₂ exhibited the highest CH₄ formation rate is in agreement with that reported by Cheng and co-workers.⁷² The order, however, was found to be completely different from that for the degradation of rhodamine B in water.⁷⁵ The photoluminescence (PL) spectroscopic study clarified that the ability of electron–hole separation decreased in the order of {010} < {101} < {001}. Thus, although the nanoplate TiO₂ with 90% {001} facets favours the electron–hole separation, this sample is less efficient for photocatalytic reduction of CO₂. Further analyses revealed that the conduction-band edge position became more negative in the order of {010} > {101} > {001}, suggesting that the reduction power of the photogenerated electrons is the highest for the {010} facet. Moreover, the {010} facet was more active for CO₂ chemisorption in the presence of H₂O. Thus, it seems that the reduction power of electrons and the CO₂ chemisorption capability rather than the electron–hole separation ability mainly determine the catalytic performance for CO₂ reduction.

Fig. 9 Comparison of photocatalytic activity of TiO₂ nanocrystals with different preferentially exposed facets for photocatalytic reduction of CO₂ with H₂O to CH₄.⁷⁵

Recently, the co-existence of {001} and {101} facets with a appropriate ratio in a TiO₂ nanocrystal has been proposed to favour the photocatalytic reduction of CO₂ with H₂O.^76,77 Anatase TiO₂ nanocrystals have typically truncated octahedral bipyramid morphology, which comprises eight {101} facets on sides and two {001} facets on the top and bottom (see the inset in Fig. 10).^69,77 By an addition of HF into the synthetic system, the ratio of {001}/{101} or the fraction of {001} facets increased and the morphology changed gradually to the nanoplate due to the stabilizing role of F⁻.^69,76,77 By studying the photocatalytic reduction of CO₂ with H₂O in a solid–liquid reaction mode containing NaOH in aqueous solution, He et al. found that the TiO₂ nanocrystal with a {001}/{101} ratio of ∼72 : 28 showed the highest activity.⁷⁶ Yu and co-workers performed a detailed comparison of a series of TiO₂ nanocrystals with different {001} fractions for the photocatalytic reduction of CO₂ with H₂O vapour.⁷⁷ The results showed that the rate of CH₄ formation increased significantly upon increasing the fraction of {001} facets from 11% to 58% and then decreased with a further increase in the fraction of {001} facets (Fig. 10A). The highest CH₄ formation rate over the nanocrystal with a {001}/{101} ratio of 58/42 was 1.35 μmol h⁻¹ g⁻¹. To explain the existence of an optimum ratio of {001}/{101}, Yu and co-workers proposed a “surface or facet junction” concept.⁷⁷ On the basis of DFT calculations, the conduction-band and valance-band edges of {001} and {101} facets are slightly different (Fig. 10B), and a facet junction could be formed because the {001} and {101} facets contact each other in the nanocrystal. The photogenerated electrons and holes can thus migrate to {101} and {001} facets, respectively, during a photocatalytic process. Therefore, the {101} facet plays a role in the reduction of CO₂ to CH₄, while the {001} facet plays a role in the oxidation of H₂O to O₂. The facet junction would thus enhance both the separation of electron–hole pairs and the reduction–oxidation reactions. It is easy to understand that the co-existence of {001} and {101} facets plays a crucial role in the photocatalytic reduction of CO₂ and there is an optimum ratio of the two facets because of the key role of the facet junction between the two facets.

Fig. 10 (A) Effect of fraction of {001} facets in anatase TiO₂ nanocrystals with octahedral bipyramid morphology on the photocatalytic conversion of CO₂ with H₂O to CH₄.⁷⁷ (B) Illustration of the facet junction between {001} and {101} facets.⁷⁷

Very recently, CeO₂ with co-exposed {100} and {111} facets has been demonstrated to favour the photocatalytic CO₂ reduction.⁷⁸ Hexahedron-prism anchored CeO₂ octahedra with {100} facets on the prism and {111} facets on the octahedron (Fig. 11A) were successfully synthesized by Zou and co-workers.⁷⁸ CeO₂ octahedra without hexahedron-prism anchored were inactive possibly because of the rapid recombination of photogenerated electrons and holes on the {111} facet. The anchoring of CeO₂ hexahedron-prisms significantly increased the photocatalytic activity for the reduction of CO₂ with H₂O to CH₄. The highest rate of CH₄ formation over Pt- and MnO_x-co-loaded hexahedron-prism anchored octahedron CeO₂ was 1.12 μmol g⁻¹ h⁻¹, and the molar ratio of CH₄/O₂ was close to the stoichiometric ratio (1/2). Computational studies showed slight differences in the positions of the conduction-band and valance-band edges between {100} and {111} facets, suggesting the formation of a facet junction (Fig. 11B). This would accelerate the separation of electron–hole pairs, and thus enhance the photocatalytic performance for CO₂ reduction.

Fig. 11 (A) Illustration of a hexahedron-prism anchored CeO₂ octahedron with co-exposed {100} and {111} facets.⁷⁸ (B) The facet junction between {100} and {111} facets of CeO₂.⁷⁸

2.2.5 Effect of heterojunctions. The construction of heterojunctions is believed to be capable of enhancing the photocatalytic performance mainly because the heterojunctions may accelerate the separation of electron–hole pairs. There are several different types of heterojunctions. The semiconductor that contains different crystalline phases in contact with each other may form phase junctions. For example, TiO₂ containing anatase–rutile or anatase–brookite phases showed higher photocatalytic activity for CO₂ reduction with H₂O than the anatase, rutile or brookite alone.^40,43,44 The differences in the conduction-band and valance-band edges between different phases can accelerate the electron–hole separation (Fig. 5). The facet junction may be responsible for the better catalytic performance of a single nanocrystal with different facets exposed.^77,78 The adjustment of the phase-junction or facet-junction is an effective method to optimize the photocatalytic activity of a semiconductor for CO₂ reduction.

The composite semiconductor could form heterojunctions between different components at the interface and such a heterojunction may result in significant synergistic effects mainly because of the enhanced charge separation. Typically, the charge transfer between different components in a composite may proceed via three mechanisms, i.e., sensitization, p–n junction and Z-scheme mechanisms (Fig. 12). In the sensitization mechanism (Fig. 12A), the photogenerated electrons are transferred from the conduction-band of a narrow bandgap semiconductor (e.g., PbS or C₃N₄) to that of a wide-bandgap semiconductor (e.g., TiO₂ or NaNbO₃). This is an effective way to realize the photoreduction of CO₂ under visible light irradiation.^79,80 The formation of a p–n junction between two components with proper band-edge positions could significantly promote the charge separation (Fig. 12B). Many composites such as ZnO–TiO₂, NiO–TiO₂ and CuO–TiO_2−xN_x with such a type of heterojunction have been demonstrated to be efficient for the photocatalytic reduction of CO₂ to CH₄ with H₂O.^81,82 Photocatalytic systems with Z-scheme heterojunctions (Fig. 12C), which mimic the natural photosynthetic process, have also been constructed for photoreduction of CO₂ with H₂O. For example, Fan and co-workers fabricated Ag₃PO₄/g-C₃N₄ composites, which were believed to function with the Z-scheme mechanism for photocatalytic reduction of CO₂.⁸³ The catalyst with an Ag₃PO₄/g-C₃N₄ molar ratio of 3/10 showed a CO₂ conversion rate of 57.5 μmol g⁻¹ h⁻¹, which were 6.1 and 10.4 times larger than pure g-C₃N₄ and P25, respectively.⁸³

Fig. 12 Schematic illustrations of three kinds of charge-transfer mechanisms in composite semiconductors. (A) Sensitization mechanism. (B) p–n junction mechanism. (C) Z-scheme mechanism.

2.3 Effect of noble or coinage metal or metal oxide co-catalysts on photocatalytic reduction of CO₂ with H₂O

2.3.1 Roles of noble or coinage metal co-catalysts. Noble and coinage metals (e.g., Pt, Pd, Au, Ag and Cu) have frequently been exploited as co-catalysts for photocatalytic reduction of CO₂ with H₂O. Similar to that proposed for the photocatalytic splitting of water, one major role of the noble or coinage metal cocatalysts is considered to extract the photogenerated electrons for the reduction of CO₂. Table 4 highlights some photocatalytic systems for CO₂ reduction with different co-catalysts.

Table 4 Some photocatalytic systems demonstrating the effect of cocatalysts on photocatalytic reduction of CO₂

Photocatalyst	Formation rate (μmol g^−1 h^−1)	Light source	Ref.
Effect of noble or coinage metal or metal oxide cocatalysts
Ag/BaLa4Ti4O15 (Ag: <10 nm)	CO: 63	Hg lamp	84
1.0 nm Pt/TiO2 thin film	CH4: 1361	Xe lamp	87
Cu0.33–Pt0.67/TiO2	Hydrocarbons: 3.73 mL g^−1 h^−1	UV lamp (365 nm)	88
Pt@Cu2O/TiO2	CH4: 33, CO: 8.3, H2: 25	Xe lamp (320–780 nm)	89
1 wt% Cu2O/TiO2-{001}	CH4: 8.7	Xe lamp	90
Ni@NiO/N-doped InTaO4	CH3OH: 160	Xe lamp (390–770 nm)	91
Au0.33–Cu0.67/TiO2 thin film	CH4: 2200, H2: 286	Xe lamp	92
Au0.75–Cu0.25/SrTiO3/TiO2 NT	CH4: 1.9 ppm cm^−2 h^−1	Xe lamp	93
Au0.25/Pt0.75/TiO2 NFs	CH4: 114	Xe lamp (UV-vis)	94
Effect of cocatalysts enhancing CO2 adsorption
MgO–TiO2 microspheres	CO: 21 (150 °C), 2 (50 °C)	Xe lamp (200–1000 nm)	97 and 98
0.5% Pt–1.0% MgO/TiO2	CH4: 11, CO: 0.03, H2: 11	Xe lamp (320–780 nm)	32 and 99
P25	CH4: 0.38, CO: 1.2, H2: 2.1
3% NaOH modified TiO2	CH4: 8.7, H2: 19	Xe lamp	100
NH2-MIL-125(Ti)	HCOOH: 16	Xe lamp (420–800 nm)	101
ZIF-8/Zn2GeO4 nanorods	CH3OH: 2.3	Xe lamp	102
MEA-functionalized TiO2	CH4: 8.6 ppm h^−1, CO: 67 ppm h^−1	Xe lamp (UV-vis)	103
Pt–0.85% PANI–TiO2	CH4: 50, H2: 320	Xe lamp (UV-vis)	104
Pt–TiO2	CH4: 15, H2: 113

---

We have compared the catalytic performances of TiO₂ doped with different noble or coinage metal co-catalysts, including Pt, Pd, Rh. Au and Ag, for the photocatalytic reduction of CO₂ with H₂O vapour.³² These co-catalysts accelerated the formation of CH₄. The formation of CO was, however, not significantly enhanced. The rate of CH₄ formation increased in the order of Ag < Au < Rh < Pd < Pt. This sequence is the same with that of the work function of metals, suggesting that the metal co-catalyst mainly functions for the extraction of electrons from TiO₂. This was further supported by the transient photocurrent response measurements. Thus, Pt was the most efficient co-catalyst to enhance the CH₄ formation owing to its highest ability to separate the photogenerated electron–hole pairs. In addition to CH₄, the formation of H₂ was also accelerated in the presence of Pt or other noble metal cocatalysts. Generally, the selectivity of the reacted electrons for CO₂ reduction was rather decreased due to the presence of a noble or coinage metal cocatalyst.

For some semiconductor photocatalysts with CO as the major reduction product, different trends were observed for the effect of noble or coinage metal co-catalysts. For example, Kudo and co-workers found that Ag was the most active co-catalyst for the reduction of CO₂ to CO over a BaLa₄Ti₄O₁₅ semiconductor, although the amount of the reacted electrons decreased in the order of Au > Rh > Cu > Ag.⁸⁴ The addition of Ag onto KTaO₃ nanoflakes was also found to accelerate the photocatalytic reduction of CO₂ to CO.⁶² At the same time, the rate of H₂ formation was decreased, and thus the selectivity of the reacted electrons for CO₂ reduction increased by using Ag as a co-catalyst. On the other hand, the addition of Pt rather decreased the formation of CO over the KTaO₃ catalyst. Pt is not an efficient co-catalyst for CO formation because of the poisoning effect of chemisorbed CO. Studies on electrochemical reduction of CO₂ showed that Pt was not a suitable electrocatalyst for CO formation because CO is bound to Pt too tightly, whereas Ag and Au were efficient electrocatalysts for the reduction of CO₂ to CO.^85,86 Thus, the major role of Ag may be its function in providing active sites to catalyze the following reaction:

CO₂ + 2H⁺ + 2e⁻ → CO + H₂O

2.3.2 Effects of sizes and location of noble or coinage metal or metal oxide co-catalysts. There are several studies to elucidate the size effect of metal cocatalysts for the photocatalytic reduction of CO₂ with H₂O. We found that that the decrease in the size of Pt nanoparticles loaded on TiO₂ from 6.8 to ∼4 nm significantly increased the amount of electrons for reduction reactions and the rate of CH₄ formation.³² The photoluminescence spectroscopic result suggested the increased ability to extract the electrons by smaller Pt nanoparticles from TiO₂. Biswas and co-workers performed a more detailed study on the size effect of Pt nanoclusters deposited on TiO₂ nanostructured films by a gas-phase deposition technique.⁸⁷ The mean sizes of Pt nanoclusters with diameters ranging from 0.5–2 nm were obtained by changing the deposition time from 5 to 60 s. The activity of these catalysts for the photoreduction of CO₂ with H₂O vapour in a continuous flow reactor under UV-light (λ = 250–388 nm) irradiation depended on the size of Pt nanoclusters (Fig. 13A). The rate of CH₄ formation reached a maximum value of 1361 μmol g⁻¹ h⁻¹ at a mean Pt cluster size of ∼1 nm. The quantum yield was calculated to be 2.41% under the same conditions. Both larger and smaller Pt nanoclusters were unfavourable for the photocatalytic reduction of CO₂ to CH₄. Why smaller Pt clusters rather have lower activity has been interpreted by the quantum size effect, which may become dominant for clusters smaller than ∼1 nm. In other words, when the Pt cluster size is smaller than 1 nm, the energy level of Pt clusters might increase to a level higher than the conduction-band edge of TiO₂ (Fig. 13B), thus hindering the transfer of photogenerated electrons from TiO₂ to Pt clusters.⁸⁷

Fig. 13 Effect of size of Pt nanoclusters loaded on TiO₂ nanofilms on the rate of CH₄ formation from CO₂ and the proposed energy levels.⁸⁷

Kudo and co-workers found that the activity of Ag/BaLa₄Ti₄O₁₅ for photocatalytic reduction of CO₂ to CO depended on the method to load Ag nanoparticles.⁸⁴ The work not only confirmed the importance of the size of Ag particles but also demonstrated the role of unique location of Ag nanoparticles on the working catalyst. BaLa₄Ti₄O₁₅ had a plate morphology with ∼100 nm thickness and ∼1 μm width. During the photodeposition process, Ag particles with sizes of 30–40 nm were predominantly loaded on the edge of the plate. On the other hand, Ag particles could be loaded on both edge and basal planes of BaLa₄Ti₄O₁₅ by a chemical reduction method. It is of interest that the Ag particles on the basal plane can be dissolved by photogenerated holes and re-photodeposited again on the edge during the photocatalytic reaction (Fig. 14). The sizes of re-photodeposited Ag particles on the edge were smaller than 10 nm and were more uniform than the Ag particles directly photodeposited from the aqueous solution with high concentration of Ag⁺. The re-photodeposition was also observed in the case of the sample prepared by impregnation followed by H₂ reduction, and the sizes of Ag re-photodeposited were 10–20 nm. The sizes of Ag nanoparticles on the edge plane of BaLa₄Ti₄O₁₅ under working conditions were in the sequence of liquid-phase reduction < impregnation and H₂ reduction < photodeposition, whereas the CO formation rate changed in an opposite sequence. Thus, the smaller size of Ag nanoparticles on the working catalyst showed higher activity of CO formation. The reduction of CO₂ to CO mainly occurred on the Ag nanoparticles loaded on the edge of the BaLa₄Ti₄O₁₅ plate, whereas the oxidation of H₂O to O₂ occurred on the basal plane (Fig. 14). The unique location of Ag nanoparticles on the working catalyst led to the separation of the reduction and oxidation reactions on different planes (Fig. 14), and thus would increase the activity for the photocatalytic CO₂ reduction with H₂O.

Fig. 14 The re-deposition of Ag nanoparticles on the edge of BaLa₄Ti₄O₁₅ during the photocatalytic reduction of CO₂ with H₂O.⁸⁴ Photogenerated electrons mainly go to the edge plane, while the photogenerated holes mainly go to the basal plane. The Ag nanoparticles on the basal plane were dissolved by the holes.

2.3.3 Effect of binary noble or coinage metal or metal oxide co-catalysts. As mentioned above, the insights obtained from the electrochemical reduction of CO₂ would be helpful for the design of efficient photocatalysts for CO₂ reduction. Cu is known to be an efficient electrocatalyst for the reduction of CO₂ to hydrocarbons.^23,24,85,86 However, the ability of Cu to extract electrons from a semiconductor is not strong because of its relatively low work function as compared to other noble metals. Thus, the combination of Cu with Pt may enhance the photocatalytic reduction of CO₂.

A few groups have investigated the catalytic behaviours of TiO₂-promoted with Pt and Cu binary co-catalysts.^48,88,89 For example, Grimes and co-workers found that Cu nanoparticles loaded on nitrogen-doped TiO₂ nanotube (NT) arrays efficiently catalyzed the formation of hydrocarbons (mainly CH₄ as well as small amounts of C₂–C₆ paraffins and olefins) and CO from CO₂ and H₂O under sunlight illumination.⁴⁸ Pt-promoted NTs could also catalyze the formation of hydrocarbons, but more H₂ was formed on Pt-NT. The co-loading of Cu and Pt on NTs in different places caused the simultaneous formation of hydrocarbons and H₂ with large amounts, although the rate of hydrocarbon formation was not significantly accelerated. These results indicated that Pt and Cu nanoparticles on NTs were the active centers for the reduction of H₂O and CO₂, respectively.

Bimetallic Cu and Pt cocatalysts loaded on periodically modulated double-walled TiO₂ nanotube (PMTiNT) arrays were investigated for the photocatalytic reduction of CO₂ with H₂O under illumination using a solar simulator.⁸⁸ CH₄, C₂H₄ and C₂H₆ were obtained as the major hydrocarbon products and the formation rate reached 3.73 mL g⁻¹ h⁻¹ for the reduction of diluted CO₂ (0.998% CO₂) over a Cu_0.33–Pt_0.67/PMTiNT catalyst. In addition to the role of the increased surface area and improved light absorption due to PMTiNT, the bimetallic co-catalysts played key roles in increasing the catalyst activity. However, the nature of the synergistic effect between Pt and Cu was not discussed in depth.

We found that the addition of Pt onto TiO₂ (P25) could enhance the formation of H₂ more significantly than that of CH₄, and thus the selectivity of the reacted electrons for CO₂ reduction decreased.³² On the other hand, the use of Cu as a co-catalyst increased the rate of CH₄ formation more than that of H₂ formation, although the total amount of reacted electrons was lower than that in the case of Pt. It was unexpected that the Pt–Cu/TiO₂ catalyst, which was prepared by a chemical reduction method using hydrazine as a reductant and was composed of Pt and Cu₂O nanoparticles, did not show any synergistic effects for CH₄ formation.⁸⁹ On the other hand, the Cu/Pt/TiO₂ catalyst, which was prepared by a stepwise photodeposition technique, showed a significantly enhanced CH₄ formation rate. The characterization using XPS, AES and high-sensitivity low-energy ion scattering (HS-LEIS) spectroscopy, which is known to be a powerful technique to detect the outermost composition of surfaces, suggested the formation of a core–shell structure of Pt@Cu₂O in the Cu/Pt/TiO₂ catalyst with a proper Cu content. The Cu content in the catalyst could be regulated by varying the irradiation time during the photodeposition of Cu onto Pt/TiO₂. The increase in the Cu content deposited onto 0.9 wt% Pt/TiO₂ to 1.7 wt% significantly increased the rates of CH₄ and CO formation and decreased that of H₂ formation (Fig. 15A). Thus, the selectivity of reacted electrons for CO₂ reduction increased significantly at the same time. The characterization revealed that the Pt@Cu₂O core–shell structure was formed when the Cu content reached 1.7 wt%, and a further rise in the Cu content increased the thickness of the Cu₂O shell. The maximum rates of CH₄ and CO formation were obtained over the Cu/Pt/TiO₂ catalyst with a Cu content of 1.7 wt%, being 33 and 8.3 μmol g⁻¹ h⁻¹, respectively. These results indicate that the Cu₂O shell provides sites for the activation and reduction of CO₂, whereas the Pt core mainly works for the extraction of photogenerated electrons from TiO₂ (Fig. 15B). A recent study investigated the catalytic behaviour of Cu₂O photodeposited on TiO₂ {001} facets.⁹⁰ The result confirmed that Cu₂O was the active phase for the chemisorption of CO₂ and the formation of CH₄ through the photogenerated electrons and H species. In our case, the Cu₂O shell on the Pt core could also suppress the reduction of H₂O by the photogenerated electrons (Fig. 15B), which can occur facilely on Pt surfaces. Furthermore, the transient photocurrent response measurements revealed that the ability of Pt to separate electron–hole pairs was not significantly affected by covering it with the Cu₂O shell.⁸⁹

Fig. 15 (A) Changes of product formation rates with Cu content.⁸⁹ (B) Proposed functioning mechanisms of Cu₂O and Pt co-catalysts on TiO₂ surfaces for conversion of CO₂ into CH₄ and CO.⁸⁹ Adapted from ref. 89 with permission from Wiley.

A Ni@NiO core–shell structured co-catalyst loaded on N-doped InTaO₄ (InTaO₄-N) probably worked in a similar way to the Pt@Cu₂O cocatalyst.⁹¹ The metallic Ni in the core may be responsible for trapping the electrons, while NiO on the shell worked as the active site for CO₂ activation and reduction. The formation of CH₃OH was detected over the Ni@NiO/InTaO₄-N catalyst during the photocatalytic reduction of CO₂ with H₂O in the liquid phase under visible light (λ = 390–770 nm) irradiation.

Two subsequent studies demonstrated that Au–Cu alloy particles were also efficient binary co-catalysts for the photocatalytic reduction of CO₂.^92,93 Au–Cu alloy nanoparticles loaded on TiO₂ thin films showed high activity for the photocatalytic reduction of CO₂ with H₂O under sun-simulated light. The rate of CH₄ formation could exceed 2 mmol g⁻¹ h⁻¹.⁹² The selectivity of the reacted electrons for CH₄ formation reached 97%. On the other hand, the Au/TiO₂ and Cu/TiO₂ catalysts showed either lower selectivity or lower activity. The alloy nanoparticles were proposed to account for the synergistic effect observed for CH₄ formation. Although the in-depth functioning mechanism of the alloy is still unclear, it seems that the presence of Cu contributes to the high CO₂ selectivity, whereas the surface plasmon band of Au may induce the visible-light photo-response and thus accelerate the activity under sun-simulated light. The Au–Cu binary alloy was also an efficient co-catalyst for the photocatalytic reduction of CO₂ with hydrous hydrazine over SrTiO₃/TiO₂ coaxial nanotube arrays.⁹³ Under the irradiation of UV-vis light, the reduction of diluted CO₂ (33.3% in Ar) with hydrous hydrazine provided CO and hydrocarbons with rates of 3.77 and 0.725 mmol g⁻¹ h⁻¹, respectively, over a Au₃Cu alloy-promoted SrTiO₃/TiO₂ coaxial nanotube array.⁹³

Several other studies have demonstrated Au- or Ag-containing bimetallic co-catalysts for photocatalytic reduction of CO₂ with H₂O.^94,95 Au and Ag are known to be capable of absorbing visible light due to their surface plasmon resonance (SPR) effect. For example, Zhang et al. loaded both Au and Pt nanoparticles onto TiO₂ nanofibers by an electrospinning technique.⁹⁴ The Au_0.25/Pt_0.75/TiO₂ nanofiber catalyst exhibited higher activity for CH₄ formation (0.57 μmol h⁻¹) than the Pt/TiO₂ (0.42 μmol h⁻¹) and Au/TiO₂ (0.31 μmol h⁻¹) catalysts under UV-vis irradiation. The enhanced photocatalytic behaviour could be attributable to both the electron-trapping effect of Pt nanoparticles and the SPR effect of Au nanoparticles (Fig. 16). Gupta and co-workers found a significant enhancement in the photocatalytic reduction of CO₂ with H₂O to CH₄ by co-loading Ag–Pt bimetallic nanoparticles and core–shell Ag@SiO₂ particles onto TiO₂.⁹⁵ Ag@SiO₂ improved the activity of TiO₂ primarily by the SPR effect, while the Ag–Pt co-catalysts should contribute to trapping of electrons and accelerating the transfer of electrons to CO₂ for reduction.

Fig. 16 Photocatalytic reduction of CO₂ with H₂O over Au- and Pt-co-loaded TiO₂ nanofibers possibly with the surface plasmon resonance effect.⁹⁴

The combination of reduction and oxidation co-catalysts has been demonstrated to be helpful in enhancing the water splitting reactions.⁹⁶ Noble metals such as Pt and Pd have mainly been employed as reduction cocatalysts, while metal oxides such as RuO₂ CoO_x and MnO_x have been exploited as oxidation co-catalysts. A few groups have studied the role of the co-existence of reduction and oxidation co-catalysts in the photocatalytic reduction of CO₂ with H₂O. For instance, Zou and co-workers reported that the Pt and RuO₂ co-catalysts, which were simultaneously loaded on the Zn_1.7GeN_1.8O, Cd₂Ge₂O₇, In₂GeO₇ or Zn₂SnO₄ semiconductor, exhibited synergistic effects for photocatalytic reduction of CO₂ with H₂O to CH₄.^51–53,60 Pt and RuO₂ were proposed to trap the electrons and holes, and to provide active sites for the reduction of CO₂ and oxidation of H₂O, respectively.

2.4 Effect of co-catalysts enhancing CO₂ adsorption on photocatalytic reduction of CO₂ with H₂O

2.4.1 Effect of basic metal oxide co-catalysts. The adsorption of CO₂ onto catalyst surfaces is a key step for the photocatalytic reduction of CO₂. Basic metal oxides may enhance the chemisorption of CO₂, and thus may favour the photocatalytic reduction of CO₂. Li and co-workers investigated the catalytic behaviours of MgO–TiO₂ microspheres, which were prepared by ultrasonic spray pyrolysis using P25 and MgO as precursors, for the photocatalytic reduction of CO₂.⁹⁷ The presence of MgO significantly increased the rate of CO formation at 393–423 K. Only a few studies have conducted photocatalytic reduction of CO₂ with H₂O vapour at temperatures >373 K. Although a higher temperature may be beneficial to this reaction because of the enhanced desorption of products, the adsorption of reactants is suppressed at high temperatures. The use of MgO could facilitate CO₂ adsorption on the catalyst, compensating the weakened CO₂ adsorption ability at a moderately high temperature (393–423 K). The rate of CO formation was found to reach a maximum at 423 K. The dynamic equilibrium between reactant adsorption and intermediate desorption was proposed to be a key factor governing the catalytic performance. A subsequent study clarified that the dispersion of MgO on TiO₂ affected the catalytic behaviours.⁹⁸ The MgO–TiO₂ microspheres prepared by a one-step spray pyrolysis possessed aggregated bulk-like MgO showing better CO formation rate than the catalyst prepared by a wet-impregnation method and possessed MgO uniformly dispersed on TiO₂ microspheres.

We found that the addition of MgO onto TiO₂ could enhance the rate of CO formation during the photoreduction of CO₂ with H₂O vapour at ambient temperature, but the degree of enhancement was very limited.⁹⁹ However, a significant enhancement in CH₄ formation was observed by addition of MgO into Pt–TiO₂. In addition to MgO, many basic metal oxides exhibited positive effects in promoting the rate of CH₄ formation. A recent study showed that the addition of NaOH or Na₂CO₃ onto TiO₂ also enhanced the rate of CH₄ formation even in the absence of a noble metal co-catalyst such as Pt.¹⁰⁰ We observed a linear relationship between the rate of CH₄ formation and the amount of CO₂ chemisorption on the surfaces of a series of basic metal oxide-promoted Pt–TiO₂ catalysts (Fig. 17). This clearly demonstrates that the basicity, i.e., the ability to chemisorb CO₂, plays crucial roles in the photocatalytic reduction of CO₂ with H₂O vapour to CH₄ and that the co-existence of Pt and a basic metal oxide is a key.

Fig. 17 Effect of basic metal oxide co-catalysts on the catalytic behaviours of Pt–TiO₂ catalysts for the photocatalytic reduction of CO₂ with H₂O vapour.⁹⁹ Reproduced from ref. 99 with permission from Royal Society of Chemistry.

Further studies on the trinary MgO–Pt–TiO₂ catalysts clarified that MgO existed as an amorphous layer on TiO₂ and Pt was mainly loaded on the interface between MgO and TiO₂. The optimum content of MgO for CH₄ formation was 1.0 wt% for the MgO–Pt–TiO₂ catalyst with a Pt content of 0.5 wt% (Fig. 18A).³² A larger content of MgO decreased the rate of CH₄ formation. It is of interest that the rate of H₂ formation decreased significantly upon increasing the content of MgO. Thus, the selectivity of the reacted electrons for CO₂ reduction increased from ∼40% for the Pt–TiO₂ to ∼80% for the 1.0 wt% MgO–0.5 wt% Pt–TiO₂ catalyst (Fig. 18B). The transient photocurrent response measurements clarified that the photocurrent density decreased by the addition of MgO into Pt–TiO₂ and such a decrease in photocurrent density became more significant when the content of MgO was >1.0 wt% (Fig. 18B). This suggests that the excess amount of MgO on TiO₂ may act as an insulating layer and retard the transfer of photogenerated electrons to Pt nanoparticles, thus hindering the separation of electron–hole pairs. Therefore, we propose that, compared to Pt–TiO₂, on which the reduction of H₂O proceeded more facilely (Fig. 18C), the presence of MgO can work for enhancing the chemisorption of CO₂ onto catalyst surfaces and the chemisorbed CO₂ can be reduced by the electron trapped on the nearby Pt nanoparticles (Fig. 18D). The interface among Pt, MgO and TiO₂ is the key to the synergistic effect.

Fig. 18 (A) Effect of MgO content on the product formation rates. (B) Effect of MgO content on the selectivity for CO₂ reduction and the current density. (C) Proposed functioning mechanism of the Pt–TiO₂ catalyst. (D) Proposed functioning mechanism of the MgO–Pt–TiO₂ catalyst.^32,99

2.4.2 Effect of other co-catalysts capable of enhancing CO₂ adsorption. Metal–organic frameworks (MOFs) are considered as promising materials for CO₂ capture and storage because of their high surface areas, tunable porous properties and controllable introduction of different active sites (reactive functional groups). An amine-functionalized titanium MOF, i.e., NH₂-MIL-125(Ti), was reported to catalyze the photocatalytic reduction of CO₂ in acetonitrile with triethanolamine as a sacrificial reagent under visible-light irradiation (λ = 420–800 nm), forming formate ions, i.e., HCOO⁻.¹⁰¹ Some MOFs such as zeolitic imidazolate frameworks (ZIFs) are stable in water and can be used for the photocatalytic reduction of CO₂ in water. Liu et al. prepared Zn₂GeO₄/ZIF-8 nanocomposites by growing ZIF-8 nanoparticles on Zn₂GeO₄ nanorods for the photocatalytic reduction of CO₂ in water.¹⁰² The uptake of CO₂ was increased by combining ZIF nanoparticles with Zn₂GeO₄ nanorods. CH₃OH was mainly formed over the Zn₂GeO₄/ZIF-8 nanocomposite during the photocatalytic reduction of CO₂ in H₂O by using Na₂SO₃ as a sacrificial reagent. The rate of CH₃OH formation was 0.22 μmol g⁻¹ h⁻¹ over the Zn₂GeO₄/ZIF-8 nanocomposite, which was ∼1.6 times of that over the Zn₂GeO₄ nanorod alone.

The modification of semiconductors with basic organic amines is another useful strategy to enhance the chemisorption of CO₂. Xue and co-workers prepared an amine-functionalized TiO₂ through a solvothermal process by performing a reaction between TiCl₄ and monoethanolamine (MEA) in ethanol at 473 K.¹⁰³ The amount of CO₂ adsorption on the MEA-functionalized TiO₂ (MEA–TiO₂) was much higher than that on TiO₂ or hydroxyl-functionalized TiO₂ (EG–TiO₂). In the photocatalytic reduction of CO₂ with H₂O vapour, MEA–TiO₂ provided CO and CH₄ with rates of 67 and 8.6 ppm h⁻¹, respectively, which were significantly larger than those on TiO₂ and EG–TiO₂.

Polyaniline (PANI), a conducting polymer composed of benzenoid and quinonoid units, is attractive as a co-catalyst for photocatalysis because of its high mobility of charge carriers, good stability, non-toxicity and ease of synthesis. PANI is expected to play a unique role in CO₂ photoreduction, since the combination of PANI with a semiconductor may not only enhance the transfer of photogenerated carriers but also may increase the chemisorption of CO₂ due to the nitrogen-containing functional groups in PANI. We recently confirmed the promoting effect of PANI on the catalytic performance of TiO₂ (P25) for the photocatalytic reduction of CO₂ with H₂O vapour.¹⁰⁴ The PANI–TiO₂ nanocomposite was synthesized by an in situ oxidative polymerization method. The characterization revealed that PANI was located on the TiO₂ surface as amorphous film and the thickness of the PANI film increased from 2.5 to 8.0 nm, corresponding to 5–16 single molecular layers, upon increasing the PANI content from 0.17 to 4.1 wt%. Pt nanoparticles with a content of 0.5 wt% were also loaded onto the PANI–TiO₂ nanocomposite. CH₄, CO and H₂ were mainly formed over the PANI–TiO₂ and Pt–PANI–TiO₂ nanocomposites. The introduction of PANI with a proper content significantly enhanced the rates of product formation. CH₄ and CO were not formed or only formed with a trace amount by using N₂ instead of CO₂, indicating that CH₄ and CO were formed predominantly by the reduction of CO₂. H₂ was a major reduction product in the absence of CO₂. It is of interest that the presence of CO₂ significantly enhanced the rate of H₂ formation over the PANI–TiO₂ and Pt–PANI–TiO₂ nanocomposites, whereas the rate of H₂ formation was almost the same in the presence and the absence of CO₂ over TiO₂ and Pt–TiO₂ without PANI modification (Fig. 19). In other words, the presence of CO₂ not only results in the formation of CH₄ and/or CO but also accelerates the reduction of H₂O to H₂ uniquely over the catalyst containing PANI.

Fig. 19 Comparison of rates of CH₄, CO and H₂ formations under CO₂ and N₂ atmospheres.¹⁰⁴ (A) TiO₂ and 0.85% PANI–TiO₂. (B) Pt–TiO₂ and Pt–0.85% PANI–TiO₂. Reaction conditions: catalyst, 0.010 g; CO₂ or N₂ pressure, 0.20 MPa; light source, 320–780 nm; irradiation time, 4 h. Reproduced from ref. 104 with permission from Royal Society of Chemistry.

Deep studies were performed to understand the unique phenomenon that the presence of CO₂ accelerates the reduction of H₂O to H₂ for the PANI-containing catalyst. The addition of an acid such as HCl is known to be capable of increasing the conductivity of PANI dramatically.¹⁰⁵ It can thus be considered that the conductivity of the PANI-containing catalysts may also be enhanced through the interaction with CO₂, an acidic molecule, thus favouring the electron transfer and electron–hole separation. Further studies clarified that the presence of CO₂ increased the current density in the transient photocurrent response measurement for the PANI-containing catalyst, whereas such an enhancement was not observed for the catalyst without PANI. Electrochemical impedance spectroscopy showed that the electron-transfer resistance was significantly decreased for the PANI-containing catalysts under a CO₂ atmosphere. Photoluminescence spectroscopy further confirmed that the luminescence band arising from the electron–hole recombination was suppressed for the PANI-containing catalyst by switching the gas atmosphere from N₂ to CO₂. Mott–Schottky measurements indicated that TiO₂ under both N₂ and CO₂ atmospheres had the same LUMO level or conduction-band edge of −0.18 V (vs. RHE, reversible hydrogen electrode), whereas the LUMO level of PANI changed from −0.42 V (vs. RHE) to −0.15 V (vs. RHE) by switching the atmosphere from N₂ to CO₂. Thus, the transfer of electrons from the conduction band of TiO₂ to PANI becomes feasible under a CO₂ atmosphere owing to the change in the LUMO level of PANI (Fig. 20). This would contribute to the enhancement in the separation of photogenerated electron–hole pairs in the PANI-containing catalyst under CO₂, thus resulting in higher photocatalytic performances for the formation of CH₄ and H₂ from CO₂ and H₂O.

Fig. 20 Mechanism for CO₂-enhanced photocatalytic reduction of CO₂ and H₂O to CH₄ and H₂via enhanced electron transfer.¹⁰⁴ Reproduced from ref. 104 with permission from Royal Society of Chemistry.

3. Photoelectrocatalysis

Besides the photocatalytic reduction of CO₂, electrocatalysis and photoelectrocatalysis are also promising methods to achieve the reduction of CO₂ into fuels and chemicals such as hydrocarbons, oxygenates or CO. Electrocatalysis requires a large amount of electricity to overcome the high energy barrier for CO₂ reduction. Photoelectrocatalysis is similar to electrocatalysis with regard to the experimental setup. However, photoelectrocatalysis integrates photocatalysis with electrocatalysis and typically exploits semiconductor electrodes instead of normal conducting electrodes used in electrocatalysis. Photoelectrocatalytic reduction of CO₂ would reduce electricity consumption as compared to the electrocatalytic reduction of CO₂ because of the introduction of solar energy. On the other hand, as compared to photocatalysis, photoelectrocatalysis may achieve higher efficiency because the applied external bias voltage can drive the separation of photogenerated electrons and holes, which is the most crucial step in limiting the photocatalytic efficiency.

In a typical photoelectrocatalytic setup, a semiconductor absorbs the light to promote the reactions on the surfaces of electrodes. Fig. 21 shows schematic illustrations of three typical two-compartment photoelectrocatalytic cells separated by proton exchange membranes. Semiconductors may be used as either photocathodes (Fig. 21A) or photoanodes (Fig. 21B). Both electrodes can also employ semiconductor photocatalysts (Fig. 21C). Take Fig. 21C as an example. Both electrodes absorb light to generate electrons and holes. The holes generated at the photoanode (typically an n-type semiconductor) may oxidize H₂O to O₂, while the photogenerated electrons at the photocathode (typically a p-type semiconductor) may work for the reduction of CO₂ to CO, HCOOH, methanol or hydrocarbons in the presence or absence of a co-catalyst. Fig. 22 displays a schematic energy-band edge position requirement for the two semiconductors to perform the reduction and oxidation reactions. The photogenerated electrons may be transferred through an external wire from the photoanode to the photocathode, favouring the efficiencies of both reduction and oxidation reactions. The development of efficient photocathodes and/or photoanodes with proper energy-band edges and ability to activate CO₂ molecules is the key to obtaining high efficiencies for the photoelectrocatalytic reduction of CO₂. Table 5 summarizes some recently reported systems with the three different types of photoelectrocatalytic cells for the reduction of CO₂.

Fig. 21 Schematic illustrations of three possible two-compartment photoelectrocatalytic cells separated by proton-exchange membranes for the reduction of CO₂. (A) Semiconductors as photocathodes. (B) Semiconductors as photoanodes. (C) Semiconductors as both photocathodes and photoanodes.

Fig. 22 Scheme of a full photoelectrocatalytic cell coupling H₂O oxidation (photoanode semiconductor) and CO₂ reduction (photocathode semiconductor with a co-catalyst).

Table 5 Some photoelectrocatalytic systems reported for the reduction of CO₂

Electrode	Efficiency^a	Conditions^b	Ref.
a Including reduction product amounts, concentrations or rates and Faradaic efficiency (FE). b Including electrolyte, applied potential, reaction time and light source.
Type A: photocathode + anode
(1) No co-catalyst
Photocathode: p-GaP Anode: carbon rod	HCOOH: 50 mM HCHO: 0.28 mM CH3OH: 0.81 mM	K2HPO4/KH2PO4, −1.0 V vs. SCE, 90 h, Hg lamp	27
(2) Metal complex co-catalyst
Photocathode: Ru complex modified Zn-doped p-InP Anode: glassy carbon	HCOOH: 0.17 mM FE: 62%	Pure water, −0.6 V vs. Ag/AgCl, 3 h, visible light	107
Photocathode: Ru complex polymer modified Cu2ZnSnS4 Anode: glassy carbon	HCOOH: 0.49 mM FE: 80%	Pure water, −0.4 V vs. Ag/AgCl, 3 h, visible light	108
Photocathode: p-Si + Re(bipy-tBu)(CO)3Cl Anode: Pt electrode	H2 + CO (H2/CO = 2 : 1, j = 5.6 mA cm^−2); FE: 102%	Acetonitrile/water mixtures, −1.9 V vs. Fc/Fc^+, 2.5 h, 661 nm	109
Photocathode: Re(bipy-tBu)(CO)3Cl + TiO2-protected Cu2O Anode: Pt electrode	CO (j = 1.5 mA cm^−2); FE: 100%	Methanol, −1.73 V vs. Fc/Fc^+, 5.5 h, AM 1.5	110
Photocathode: Re complex + zinc porphyrin + p-type NiO Anode: Pt electrode	CO: 0.93 μmol FE: 6.2%	Bu4NBF4 in DMF, Ag/AgNO3 reference, 430 nm	112
Photocathode: Ru(II)–Re(I)/p-NiO Anode: Pt electrode	CO: 255 nmol, FE: 98%	Et4NBF4 in DMF–triethanolamine, −1.2 V vs. Ag/AgCl, 5 h, >460 nm	113
(3) One- or two-dimensional nanostructure photoelectrode
Photocathode: polypyrrole-coated p-ZnTe Anode: carbon rod	HCOOH: 131, CO: 50, H2: 108 nmol h^−1 cm^−2	KHCO3, −0.3 V vs. SCE, 6 h, >420 nm	114
Photocathode: ZnTe/ZnO nanowire/Zn substrate Anode: Pt electrode	CO: 68 μmol cm^−2 FE: 23% (CO), 61% (H2)	KHCO3, −0.7 V vs. RHE, 1 h, >420 nm	117
Photocathode: CuO nanowires/p-Cu2O/Cu Anode: Pt electrode	HCOOH + CO + H2, j = 1 mA cm^−2	Na2SO4 (pH = 10), +0.25 V vs. RHE, 600 s, AM 1.5	119
Photocathode: Pb/CuO/Cu2O film Anode: carbon rod	CH3OH: 0.1, HCOOH: 0.52, CO: 0.24 μmol h^−1 cm^−2	KHCO3, −0.4 V vs. SCE, 1 h, visible light	120
Photocathode: Cu/Cu2O electrode Anode: Pt electrode	CH3OH: 178 ppm, HCHO: 10 ppm	Na2CO3/NaHCO3, +0.2 V vs. Ag/AgCl, 2 h, UV-vis light	121
Photocathode: Cu nanoparticles-doped Co3O4 nanotube arrays Anode: Pt electrode	HCOOH: 6.8 mmol L^−1 cm^−2; FE: ∼100%	Na2SO4, −0.9 V vs. SCE, 8 h, visible light	123
Photocathode: CuFeO2/CuO Anode: Pt electrode	HCOOH: 5 μmol h^−1; solar to chemical energy efficiency: 1%	KHCO3, no external bias potential, 24 h, AM 1.5	124
Type B: cathode + photoanode
Cathode: Pt-RGO, photoanode: Pt/TiO2 nanotubes	HCOOH + CH3OH + CH3COOH + C2H5OH: 1.5 μmol h^−1 cm^−2 FE: ∼81%	NaCl + NaHCO3, constant potential: 2 V, 8 h, Xe arc lamp	126
Cathode: Cu Photoanode: WO3	CH4: 0.7 μmol cm^−2; FE: 67%	KHCO3, +0.75 V vs. RHE, 1 h, >420 nm	128
Type C: photocathode + photoanode
Photocathode: p-InP/hybrid Ru complex polymer Photoanode: Pt/TiO2	HCOOH: 5.2 μmol cm^−2 Solar to chemical energy efficiency: 0.03%	NaHCO3, no applied bias, 24 h, AM 1.5	129
Photocathode: InP/hybrid Ru complex polymer Photoanode: reduced SrTiO3	HCOOH: 1.5 μmol, solar to chemical energy efficiency: 0.14%	NaHCO3, no applied bias, 3 h, photocathode: >420 nm, photoanode: AM 1.5	130

---

3.1 Photocathodes in combination with dark anodes

Most of the studies on the photoelectrocatalytic reduction of CO₂ reported to date have employed a type A (photocathode-dark anode) cell. It should be noted that many studies used one-compartment cell without the separation of reduction and oxidation reactions as in Fig. 21A. Halmann reported a first study on the photoelectrocatalytic reduction of CO₂ using this type of cell in 1978.²⁷ An p-GaP semiconductor, carbon and a buffered aqueous solution were used as the photocathode, counter anode and electrolyte, respectively. When p-GaP was irradiated using an Hg lamp and a voltage bias was applied, current was detected. At the same time, HCOOH, HCHO and CH₃OH were formed in the electrolyte solution (Table 5). Since then, many groups investigated different semiconductor photocathodes to enhance the efficiency. Considering the band-bending mode, p-type semiconductors are typically used as the photocathodes. Besides p-GaP, p-GaAs, p-InP, p-Si, N-doped Ta₂O₅ (N-Ta₂O₅), Cu₂O, p-NiO, p-Co₃O₄ and ZnTe have been exploited as photocathodes.^28,29 Pt is usually used as the counter anode. Many challenges still remain. For instance, the selectivity for CO₂ reduction is low in many cases because the reduction of H₂O may occur preferentially over the surface of a p-type semiconductor photocathode. Moreover, a large bias is usually required because the valence-band of a single p-type semiconductor may be not positive enough to oxidize H₂O.

In many cases, the p-type semiconductor photocathode may just work as a light harvester to generate electrons and holes but does not act as a true catalyst for the activation of inert CO₂ molecules. Thus, the combination of a co-catalyst, which is capable of activating CO₂ molecules, with the photocathode would be a useful strategy to increase the efficiency. Metal complexes that can activate CO₂ have attracted much attention for this purpose.²⁹ The most essential requirement for the metal complex is that, its LUMO should be more positive than the conduction-band edge of the semiconductor and more negative than the redox potential of CO₂ to a specific product. In addition, the interface interaction between the semiconductor and the metal complex plays a key role in the electron transfer and thus the efficiency for CO₂ reduction. Sato and co-workers combined N-Ta₂O₅, a p-type semiconductor, with a Ru complex co-catalyst, e.g., [Ru-(dcbpy)₂(CO)₂]²⁺ (dcbpy: 4,4′-dicarboxy-2,2′-bipyridin), and succeeded in transforming CO₂ to HCOOH in an acetonitrile/triethanolamine solution under irradiation with visible light (λ = 410–750 nm).¹⁰⁶ The turnover number (TON) based on the Ru complex reached 89 after 8 h of reaction. It was clarified that triethanolamine worked as the electron donor and the proton source. In a subsequent study, the same group developed a system capable of catalyzing the reduction of CO₂ using H₂O to HCOOH by combining a ruthenium-complex polymer (RCP) catalyst, i.e., [Ru(L–L)(CO)₂]_n, in which L–L is a diimine ligand, with a Zn-doped p-type InP (p-InP-Zn).¹⁰⁷ At an applied bias of −0.6 V and under visible-light (λ = 410–750 nm) irradiation, HCOOH could be formed with a concentration of 0.17 mM after 3 h with an Faradaic efficiency of 62% (Table 5). A sulphide semiconductor, Cu₂ZnSnS₄, modified with Ru complex polymers could also catalyze the photoelectrocatalytic reduction of CO₂ with H₂O to formate.¹⁰⁸

Kubiak and co-workers reported an interesting formation of syngas (H₂ + CO) by combining a heterogeneous H₂-evolution photocathode, i.e., p-Si, with a Re complex co-catalyst, i.e., Re(bipy-tBu)(CO)₃Cl (where bipy-tBu = 4,4′-di-tert-butyl-2,2′-bipyridine), which was capable of catalyzing the reduction of CO₂ to CO.¹⁰⁹ The co-generation process had an Faradaic efficiency of nearly 100%. Under a monochromatic light (λ = 661 nm) illumination with an intensity of 95 mW cm⁻² on the photocathode surface and a catalytic current density of 5.6 mA cm⁻², the total light-to-chemical energy conversion efficiency for the cathodic half cell reaction was 4.6%. Furthermore, the ratio of H₂/CO could be tuned from 0 to 2 by addition of H₂O into acetonitrile and changing the concentration of the Ru complex. CO could also be produced with high efficiency by using a cheaper p-type semiconductor, Cu₂O, instead of p-Si.¹¹⁰ However, the protection of Cu₂O was needed because of its low stability under reductive conditions. The protection of Cu₂O by amorphous TiO₂ layers is a useful strategy.¹¹¹ The combination of the TiO₂-protected Cu₂O photocathode with the Re(bipy-tBu)(CO)₃Cl complex in CO₂-saturated acetonitrile under simulated sunlight could provide a sustained cathodic current density of 1.5 mA cm⁻².¹¹⁰ 100% Faradaic efficiency could be achieved for CO formation by using a protic additive, which was proposed to eliminate the charge transfer limitations from the surface of TiO₂-protected Cu₂O to the Re complex catalyst. The photocurrent corresponded to selective CO evolution could be sustained over several hours. A Re bipyridyl complex was also used in combination with the p-type NiO photocathode for the photoelectrocatalytic reduction of CO₂ to CO in DMF solution.¹¹² In this work, a metal complex dyad composed of a Zn porphyrin as a light-harvesting sensitizer and a Re complex as the catalytic site were used instead of the Re complex alone. An electrocatalytic cell composed of the dyad-adsorbed p-NiO particle layers on an FTO photocathode and a Pt counter anode in CO₂-saturated DMF solution showed a constant cathodic photocurrent upon visible-light irradiation (λ = 420 nm) and provided CO as a reduction product. A hybrid photocathode composed of a Ru(II)–Re(I) supramolecular metal complex immobilized on a p-NiO electrode was developed by Ishitani and co-workers.¹¹³ This system allowed the catalytic reduction of CO₂ to CO with high Faradaic efficiency and relatively high durability.

Recently, instead of metal complex co-catalysts, conducting polymers such as polypyrrole (PPy) have also been exploited to increase the efficiency of CO₂ reduction over the p-type semiconductor photocathode. Woo and co-workers investigated the PPy-coated p-ZnTe photocathode for photoelectrocatalytic reduction of CO₂ in KHCO₃ aqueous solution under visible-light illumination.¹¹⁴ ZnTe is expected to be a promising photocathode because it possesses the most negative conduction band-edge position (−1.63 V vs. the reversible hydrogen electrode (RHE)) among all the known p-type semiconductors, which may lead to the largest driving force for the transfer of electrons from the semiconductor to an acceptor. The PPy/ZnTe photocathode demonstrated a Faradaic efficiency of 51% and had twice the production rate of HCOOH and CO than those of the bare ZnTe photocathode at −0.2 V. This could be attributed to the enhanced selectivity of HCOOH production, the remarkably suppressed H₂ evolution and the increased active sites for CO₂ reduction after the deposition of PPy onto ZnTe.

The fabrication of 1D semiconductors as photocathodes is another strategy to increase the photoelectrocatalytic efficiency, since 1D materials have many advantages such as large surface areas, more active sites, excellent charge transportation and potentially enhanced light absorption.¹¹⁵ A number of studies have been devoted to the fabrication of 1D nanomaterials as photoelectrodes for water splitting.¹¹⁶ However, only a few papers have been published for the use of semiconductors with 1D nanostructures for the photoelectrocatalytic reduction of CO₂. Lee and co-workers reported a fabrication of a 1D ZnTe/ZnO/Zn photocathode through a microwave hydrothermal reaction of ZnO nanowire arrays on a Zn metal substrate with an aqueous solution of sodium tellurite and hydrazine monohydrate in an aqueous solution.¹¹⁷ The formation of CO and hydrogen was observed at −0.7 to −0.2 V versus RHE in aqueous KHCO₃ with high incident-photon-to-current-conversion efficiencies (∼60% and ∼85% at −0.7 and −0.8 V vs. RHE, respectively). The 1D ZnTe/ZnO/Zn photocathode also showed a stable photocurrent and CO formation even at −0.2 V versus RHE. The selectivity for CO₂ reduction (instead of H₂O reduction) or the Faradaic efficiency for CO formation and the stability against corrosion should be overcome in the future research. CuO–Cu₂O hybrid nanorod arrays, which were fabricated on a Cu substrate by a two-step approach consisting of the thermal growth of CuO nanorods followed by controlled electrodeposition of p-type Cu₂O crystallites on their walls, were also applied to photoelectrocatalytic reduction of CO₂ in aqueous solution.¹¹⁸ Under a simulated sunlight illumination in a 0.1 M Na₂SO₄ solution saturated with CO₂, methanol was formed at an applied voltage of −0.2 V vs. SHE. It is of interest that, in the CO₂ saturated electrolyte, the photocurrents for CuO–Cu₂O nanorods reached ∼3 times higher than in the N₂-saturated counterpart solution. This indicates that CO₂ can be reduced by the photogenerated electrons more selectively than H₂O on Cu₂O surfaces. This is in essence the same with the phenomenon observed in photocatalysis.⁸⁹ Cu₂O with and without CuO nanowires grown on a Cu substrate was also efficient for the photoelectrocatalytic reduction of CO₂ in aqueous solutions under simulated sunlight irradiation, producing CO and HCOOH together with H₂.¹¹⁹

The CuO–Cu₂O thin film fabricated by oxidation of Cu foil deposited with some metal particles were also used for the photoelectrocatalytic reduction of CO₂ in a KHCO₃ aqueous solution under visible-light (λ = 400–800 nm) illumination.¹²⁰ The CuO/Cu₂O film doped with Pb particles exhibited a prominent CO₂ reduction ability with ∼40% Faradaic efficiency at −0.16 V (vs. SHE). A Cu/Cu₂O film photocathode was also used for the photocatalytic reduction of CO₂ in aqueous solution under UV-vis light irradiation.¹²¹

Co₃O₄, a p-type semiconductor with a band-gap energy of 2.07 eV, has also been used as a visible-light driven photocathode for the reduction of CO₂, but the performance of Co₃O₄ alone is quite low due to fast recombination of photogenerated electron–hole pairs.¹²² The fabrication of 1D Co₃O₄ nanotube arrays with further loading of Cu nanoparticles was demonstrated to offer a highly efficient photocathode for the reduction of CO₂ to formate in Na₂SO₄ aqueous solution under visible-light irradiation (Fig. 23).¹²³ The selectivity was nearly 100%, and noticeably, the concentration of formate reached 6.75 mmol L⁻¹ cm⁻² in an 8 h photoelectrocatalytic reaction at an applied voltage of −0.9 V. This rate was 56% higher than that over the 1D Co₃O₄ nanotube arrays without Cu particles and was ∼100 times higher than that over Cu nanoparticle-modified glassy carbon electrode. This is probably one of the highest yields among those reported to date in the literature. Deeper studies indicated that the enhanced production of formate could be attributed to the self-supported Co₃O₄ nanotube/Co structure and the interface structure between Co₃O₄ nanotubes and metallic Cu nanoparticles.¹²³ The ordered 1D structure may not only be beneficial to the fast transportation of electrons and holes but also may be in favour of light harvesting.

Fig. 23 Schematic illustration of the Cu nanoparticle-decorated 1D Co₃O₄ nanotube array as the photocathode for the photoelectrocatalytic reduction of CO₂ to formate with high efficiency.¹²³

Only a few studies have succeeded in performing the reduction of CO₂ in the photoelectrocatalytic cell without applying external bias. Very recently, Park and co-workers reported a significant progress. They found that the reduction of CO₂ to formate could proceed without any external bias by using a CuFeO₂ and CuO mixed p-type photocathode, which was fabricated via electroplating of cupric and ferric ions followed by annealing under atmospheric air.¹²⁴ The composite electrodes exhibited onset potentials at +0.9 V vs. RHE in CO₂-purged bicarbonate solution and converted CO₂ to formate with >90% selectivity under simulated solar light. The CuFeO₂/CuO photocathode and Pt anode couples connected by an external wire could produce formate for over 1 week at a solar-to formate energy conversion efficiency of ∼1% with O₂ evolved from H₂O without any external bias applied. Further elucidation of the functioning mechanism of the composite semiconductor is definitely needed in the future.

3.2 Photoanodes in combination with dark cathodes

Although many p-type semiconductor photocathodes have been reported for the reduction of CO₂ in combination with an O₂-evoling anode in the dark, the efficiency is still low. Over the p-type semiconductor, the 2-electron reduced products, i.e., HCOOH and CO, were usually obtained along with H₂, the H₂O reduction product. Moreover, the p-type semiconductor is usually expensive or unstable during the reaction. Therefore, the use of an n-type semiconductor such as TiO₂, BiVO₄ or WO₃, which is cheap and stable, as the photoanode for H₂O oxidation and an electrocatalyst that is active for CO₂ reduction as the cathode to construct a type B photoelectrocatalytic cell (Fig. 21B) would also be an attractive alternative.

Cheng et al. reported a combination of the Pt-modified TiO₂ nanotube (Pt-TNT) photoanode and the Pt-modified reduced graphene oxide (Pt-RGO) cathode for the reduction of CO₂ under UV-vis irradiation.¹²⁵ Although the applied voltage was much high (+2 V), they obtained liquid products including HCOOH, CH₃OH, CH₃COOH and C₂H₅OH. The rate for the formation of these liquid products reached 0.6 μmol h⁻¹ cm⁻².¹²⁵ The optimization of reaction conditions further increased the rate of product formation to 1.5 μmol h⁻¹ cm⁻¹.¹²⁶ The exploitation of N-doped TiO₂ instead of TiO₂ could allow the use of visible light. The combination of the N-doped TiO₂ photoanode with the Cu cathode produced CH₄ and CH₃OH in addition to HCOOH, although the applied bias was high (2.8 V) and the Faradaic efficiencies was low.¹²⁷

Lee and co-workers recently reported an interesting work using visible-light responsive WO₃ as the photoanode and Cu or Sn/SnO_x, which was known to be an efficient electrocatalyst for CO₂ reduction, as the cathode for CO₂ reduction under visible-light irradiation.¹²⁸ The system could work with a relatively lower applied bias as compared to other photoanode–cathode systems reported. When the Cu cathode was used, CH₄ was mainly formed along with C₂H₄ and CO. The increase in the applied voltage from +0.55 V to +0.75V (vs. RHE), the Faradaic efficiencies for CH₄, C₂H₄ and CO changed from 42.3%, 4.0% and 5.1% to 67.0%, 0.6% and 2.7%, respectively. On the other hand, when Sn/SnO_x was employed as the cathode, HCOOH and CO were formed as the major reduction products besides H₂. The Faradaic efficiencies for HCOOH and CO at an applied bias of 0.7 V were 27.5% and 15.8%, respectively.

3.3 Combination of photocathodes and photoanodes

It is highly desirable to perform the photoelectrocatalytic reduction of CO₂ using H₂O both as the electron donor and as the proton source under applied electric bias-free conditions. The use of only the photocathode or only the photoanode in combination with the counter dark anode or cathode is difficult to achieve this goal, since the valance-band edge of the photocathode and the conduction-band edge of the photoanode are typically not suited for the oxidation of H₂O and the reduction of CO₂, respectively. The combination of a photocathode for CO₂ reduction with a photoanode for H₂O oxidation is a useful strategy to realize the reduction of CO₂ with H₂O without external applied electric bias. For this purpose, the conduction-band edge of the photoanode for H₂O oxidation must be more negative than the valance-band edge of the photocathode for CO₂ reduction to guarantee the electron transfer from the photoanode to the photocathode through the external wire (Fig. 22).

Sato et al. constructed a two-compartment Pyrex cell separated using a proton-exchanging membrane (Fig. 21C) using InP/[MCE2-A + MCE4] (MCE2-A = [Ru(4,4′-diphosphate ethyl-2,2′-bipyridine)(CO)₂Cl₂]), MCE4 = [Ru{4,4′-di(1H-pyrrolyl-3-propyl carbonate-2,2′-bipyridine}(CO)₂(MeCN)Cl₂)] as the photocathode and Pt/TiO₂ as the photoanode (Fig. 24A).¹²⁹ It should be noted that the p-type semiconductor InP alone as the photocathode was almost inactive for the reduction of CO₂. The Ru complexes served as the co-catalysts for electron transfer and CO₂ reduction. MCE-2A possesses an anchor ligand, i.e., 4,4′-diphosphate ethyl-2,2′-bipyridine, so that it can be linked tightly with the surface of InP. Further deeper studies indicated that MCE-2A played an important role in electron transfer, while MCE4 worked as a true active species for the reduction of CO₂ to HCOO⁻. By using a solar simulator as the light source, the two-compartment photoelectrocatalytic cell could be run just by wiring the two photoelectrodes without external applied electrical bias. The TON for HCOOH was >17 in 24 h. Although H₂ and CO were also formed, the Faradaic efficiency for HCOOH reached ∼70%. It is noteworthy that the natural photosynthesis also proceeds via the Z-scheme, which utilizes two photosystems, i.e., PS-I and PS-II. Thus, the two-compartment photocathode and photoanode system has some similarity to the natural photosynthetic system. The efficiency of the conversion of solar energy to chemical energy was estimated to be ∼0.03%, lower than the solar conversion efficiency (0.2%) for switchgrass, a crop for biofuels.

Fig. 24 Schematic illustration of Z-scheme systems for photoelectrocatalytic reduction of CO₂.^129,130 (A) TiO₂–InP. (B) Reduced SrTiO₃–InP.

A further optimization of the photoanode semiconductor could increase the efficiency for CO₂ reduction. Arai et al. employed the same InP/[RuCP] (RuCP = Ru complex polymer) as the photocathode for CO₂ reduction but changed the photoanode semiconductor from TiO₂ to reduced SrTiO₃, which had almost the same band-gap energy but more negative conduction-band and valance-band edges as compared to TiO₂.¹³⁰ Thus, the difference between the conduction-band edge of the photoanode and the valance-band edge of the photocathode became larger (Fig. 24B). This would facilitate the transfer of electrons from the photoanode to the photocathode, thus being beneficial to the Z-scheme reaction. The two-compartment cell using InP/[RuCP] as the photocathode and reduced SrTiO₃ as the photoanode without the external electrical bias provided a stable photocurrent of 140 μA cm⁻² under visible-light and UV-light irradiation on the photocathode and the photoanode, respectively. HCOOH was formed as a major product along with CO and H₂. The amount of formate was 1.45 μmol after 3 h of reaction, larger than that (0.33 μmol) obtained using TiO₂ as the photoanode. The Faradaic efficiency for formate formation was 71.5%. The efficiency of the conversion of solar energy to chemical energy increased to 0.14%, approaching the value for switchgrass.

Another feature of the reduced SrTiO₃ is that, this semiconductor is less active for the degradation of HCOOH and other organic molecules. This feature has been harnessed by Arai et al. to construct a wireless one-compartment cell without using the proton-exchange membrane.¹³⁰ HCOOH was obtained with an amount of 0.83 μmol in 3 h under irradiation on both semiconductors. Although the energy efficiency for the conversion of solar energy to chemical energy is still low (0.08%), the one-compartment cell would largely decrease the cost of the photoelectrocatalytic reactions and will open the door for developing more cheaper and efficient photoelectrocatalytic systems for the reduction of CO₂ with H₂O.

4. Conclusion and outlook

Significant progress has been achieved in both photocatalytic and photoelectrocatalytic reduction of CO₂ with H₂O based on semiconductor catalysts. Rich knowledge has been accumulated for the design and fabrication of efficient photocatalysts and photoelectrocatalysts. Generally, light harvesting, photogenerated electron–hole separation or photo-carrier transfer and surface reaction are three key steps that should be mainly considered for the design of an efficient semiconductor-based photocatalyst or photoelectrocatalyst. These are common to both H₂O splitting and CO₂ reduction, the two biggest challenging reactions using solar energy. The present article has highlighted some key issues that are crucial to CO₂ reduction. In particular, more attention should be paid to the adsorption ability and reactivity of catalyst surfaces toward CO₂ molecules. In addition to the efficiency of photogenerated electrons used for reduction, the selectivity toward CO₂ reduction is a determining factor, because the reduction of H₂O to H₂ may occur faster on photocatalyst or photoelectrode surfaces due to the lower reactivity of CO₂. The selectivity of products from CO₂ reduction is another key parameter that makes the situation complicated. CO, HCOOH, CH₃OH or CH₄ is the typical major product depending on the catalytic system, although some C₂ products such as C₂H₄ and CH₃CH₂OH have also been observed in a few systems.

The structure of a semiconductor plays crucial roles in CO₂ reduction through affecting either the separation of photogenerated electron–hole pairs or the reactivity of catalyst surfaces or both. For instance, the crystalline phase of TiO₂ affected its photocatalytic activity for CO₂ reduction not only because of the different crystalline and electronic structures but also due to the different concentration of surface defective sites (Ti³⁺ sites and oxygen vacancies), which may be responsible for CO₂ activation. These defective sites could be easily generated on brookite TiO₂ surfaces, and thus brookite after heat treatment in He showed much higher activity for CO₂ reduction.³⁷ TiO₂ nanoparticles with medium particle sizes were found to be beneficial to CO₂ reduction because of the balance between the number of surface sites for CO₂ activation and the electronic/optical structure for light harvesting.⁴⁷ The higher CO₂ reduction activities of many 1D or 2D semiconductor nanowires/nanorods or nanosheets arise both from their larger concentration of surface sites and from the enhanced transfer of photogenerated electrons and holes to surfaces. The enhanced activity for a specifically exposed facet may mainly owe to the higher reactivity of the facet. The {010} facet of TiO₂ was found to be active for the reduction of CO₂ to CH₄ because of its higher conduction-band edge position and thus the larger reduction power of the photogenerated electrons as well as higher CO₂ adsorption ability, although the electron–hole separation ability of the {010} facet was rather lower than those of {001} and {101}.^72,75 Furthermore, the fabrication of a semiconductor with a specific morphology and peculiar exposed facets could allow the occurrence of CO₂ reduction that cannot occur on this semiconductor without such a specific morphology, because of the changed conduction-band or valance-band edge positions. WO₃ is an example. The conduction-band edge position of WO₃ does not allow the occurrence of the reduction of CO₂ to CH₄. WO₃ nanosheets possess more negative conduction-band edge, and thus the reduction of CO₂ becomes feasible on WO₃ nanosheets.^64,65 The concept of heterojunctions can also be harnessed for the design of efficient photocatalysts for CO₂ reduction. The creations of phase junction and facet junction have been demonstrated to be useful strategies to facilitate the electron–hole separation.^44,77,78

The noble or coinage metal or metal oxide co-catalysts can function as the active sites for CO₂ reduction in addition to extracting the photogenerated electrons or holes. For the purpose of extracting electrons, Pt is the best co-catalyst. However, the selectivity for CO₂ reduction would be decreased because Pt would enhance the reduction of H₂O to H₂ to a larger extent. Early electrocatalytic studies indicated that Ag surfaces catalysed the reduction of CO₂ to CO, whereas Cu surfaces favoured the formation of hydrocarbons from CO₂. Ag was actually an efficient co-catalyst for CO formation over several semiconductors.^62,84 Cu₂O or Cu was an efficient co-catalyst for the formation of CH₄ as well as higher hydrocarbons by providing active sites for CO₂ activation.^89,90 Bimetallic Pt@Cu₂O co-catalysts with Pt as the core and Cu₂O as the shell could both efficiently extract the electrons and selectively catalyse the conversion of CO₂ to CH₄, showing promising performances for CH₄ formation.⁸⁹ Core–shell structured Ni@NiO co-catalysts had similar synergistic effects for the reduction of CO₂ to CH₃OH. The synergistic effect was also observed for Au–Cu and Au–Pt alloy co-catalysts, where the plasmon resonance effect of Au may contribute to the increase in the visible-light harvesting.^92–94

The present article has also shown that the harnessing of a co-catalyst that can enhance the chemisorption of CO₂ is an effective strategy to increase the CO₂ reduction activity. Basic metal oxides, MOFs and PANI with nitrogen-containing basic functional groups are good candidates for this purpose.^{97–100,102–104} Significant synergistic effects have been observed between a basic metal oxide (e.g., MgO) and Pt over TiO₂ for the reduction of CO₂ to CH₄.^32,99 The chemisorbed CO₂ on MgO could be reduced by the electrons trapped on the neighbouring Pt particles. A unique CO₂-enhanced photosynthesis of CH₄ and H₂ from CO₂ and H₂O was observed over the Pt-doped PANI–TiO₂ nanocomposites. On the PANI–TiO₂ nanocomposites with or without Pt, the presence of CO₂ not only caused the formation of CH₄ and/or CO but also significantly enhanced the rate of H₂ formation.

Photoelectrocatalysis offers further opportunities to increase the efficiency for the reduction of CO₂ with H₂O by combining the advantages of both photocatalysis and electrocatalysis. The separation of CO₂ reduction and H₂O oxidation into two compartments using two electrodes would increase the flexibility to design suitable photoelectrocatalysts or electrocatalysts for each half reaction, making the optimization of each reaction more efficiently. Studies in this field have mainly focused on the development of photocathodes based on p-type semiconductors for CO₂ reduction in combination with a dark electroanode for H₂O oxidation. Semiconductors such as p-GaP, p-GaAs, p-InP, p-Si, N-doped Ta₂O₅, Cu₂O, p-NiO, p-Co₃O₄ and ZnTe were utilized as photocathodes along with a Pt or carbon anode. The reduction of H₂O to H₂ may proceed faster than the reduction of CO₂ on photocathode surfaces. Co-catalysts, in particular Ru or Re complexes, which could activate CO₂ and/or facilitate the transfer of electrons, showed a significant enhancing effect on the reduction of CO₂. Typically, CO and HCOOH were mainly formed in these systems. Synthesis gas with different H₂/CO ratios could be generated by changing the conditions on a Re-complex-modified p-Si photocathode.¹⁰⁹ The fabrication of a 1D or 2D semiconductor photocathode could accelerate the transfer of charge carriers and thus enhance the CO₂ reduction activity. For example, the 1D Co₃O₄ photocathode decorated with Cu nanoparticles exhibited an outstanding HCOOH formation rate.¹²³ A couple of studies utilized n-type semiconductors as photoanodes for H₂O oxidation and electrocatalysts for CO₂ reduction. Not only HCOOH but also CH₃OH and CH₄ could be produced by choosing proper electrocatalysts. By selecting two semiconductors with proper conduction-band and valance-band edge positions as both photocathodes and photoanodes, a few studies have succeeded in the reduction of CO₂ with H₂O without external applied bias. For example, the use of a Ru complex polymer-modified InP and reduced SrTiO₃ as the photocathode and the photoanode, respectively, HCOOH could be formed with a reasonably high efficiency, which approached that for switchgrass, an natural photosynthetic system.¹³⁰

It is undoubted that many challenges remain to be solved in the future. Some of these challenges, including the increase of light harvesting efficiency and the enhancement of the separation of photogenerated electron–hole pairs, are common for photocatalysis and photoelectrocatalysis. The development of efficient visible-light responsive semiconductors with rationally designed nanostructures should be paid more attention. In more detail, the 1D or 2D morphology can accelerate the transfer of photogenerated carriers. The fabrication of semiconductors with an optimized phase junction or facet junction would largely enhance the separation of electron and hole pairs. The materials in which the photogenerated electrons and holes can move to differently exposed facets are highly attractive because of their intrinsic ability to separate electron–hole pairs and the separated regions for the reduction and oxidation reactions.

We strengthen the future need to develop efficient co-catalysts for the activation and selective reduction of CO₂. Although some studies have pointed out the effects of basic metal oxides, nitrogen-containing organic polymers and Ru or Re complexes as co-catalysts, more effort should be put into the design of co-catalysts that are able to enhance the chemisorption of CO₂ from the mixture of CO₂ and H₂O, to activate CO₂ molecules preferentially and to catalyze the product formation with controllable selectivity. Many advances have been achieved in the thermal catalysis for the hydrogenation of CO₂ to various products. The knowledge about the active sites and active phases in those systems can be referenced in developing more efficient co-catalysts for the photocatalytic and photoelectrocatalytic reduction of CO₂ with H₂O.

Acknowledgements

This work was supported by the National Basic Research Program of China (2013CB933102), the National Natural Science Foundation of China (21433008 and 21503176), the China Postdoctoral Science Foundation (No. 2015M570555) and the Program for Innovative Research Team in Chinese Universities (No. IRT_14R31).

Notes and references

1. (a) M. He, Y. Sun and B. Han, Angew. Chem., Int. Ed., 2013, 52, 9620–9633; (b) G. Centi, E. A. Quadrelli and S. Perathoner, Energy Environ. Sci., 2013, 6, 1711–1731.
2. W. Wang, S. Wang, X. Ma and J. Gong, Chem. Soc. Rev., 2011, 40, 3703–3727.
3. T. Sakakura, J. C. Choi and H. Yasuda, Chem. Rev., 2007, 107, 2365–2387.
4. M. S. Fan, A. Z. Abdullah and S. Bhatia, ChemCatChem, 2009, 2, 192–208.
5. A. J. Morris, G. J. Meyer and E. Fujita, Acc. Chem. Res., 2009, 42, 1983–1994.
6. T. Inoue, A. Fujishima, S. Konishi and K. Honda, Nature, 1979, 277, 637–638.
7. V. P. Indrakanti, J. D. Kubicki and H. H. Schobert, Energy Environ. Sci., 2009, 2, 745–758.
8. S. C. Roy, O. K. Varghese, M. Paulose and C. A. Grimes, ACS Nano, 2010, 4, 1259–1278.
9. K. Mori, H. Yamashita and M. Anpo, RSC Adv., 2012, 2, 3165–3172.
10. A. Dhakshinamoorthy, S. Navalon, A. Corma and H. Garcia, Energy Environ. Sci., 2012, 5, 9217–9233.
11. (a) A. D. Handoko, K. Li and J. Tang, Curr. Opin. Chem. Eng., 2013, 2, 200–206; (b) K. Li, X. An, K. H. Park, M. Khraisheh and J. Tang, Catal. Today, 2014, 224, 3–12.
12. W. Fan, Q. Zhang and Y. Wang, Phys. Chem. Chem. Phys., 2013, 15, 2632–2649.
13. S. N. Habisreutinger, L. Schmidt-Mende and J. K. Stolarczyk, Angew. Chem., Int. Ed., 2013, 52, 7372–7408.
14. J. Mao, K. Li and T. Peng, Catal. Sci. Technol., 2013, 3, 2481–2498.
15. S. Navalón, A. Dhakshinamoorthy, M. Álvaro and H. Garcia, ChemSusChem, 2013, 6, 562–577.
16. Y. Izumi, Coord. Chem. Rev., 2013, 257, 171–186.
17. S. Das and W. M. A. Wan Daud, RSC Adv., 2014, 4, 20856–20893.
18. W. Tu, Y. Zhou and Z. Zou, Adv. Mater., 2014, 26, 4607–4626.
19. Y. Ma, X. Wang, Y. Jia, X. Chen, H. Han and C. Li, Chem. Rev., 2014, 114, 9987–10043.
20. N. Zhang, R. Ciriminna, M. Pagliaro and Y. J. Xu, Chem. Soc. Rev., 2014, 43, 5276–5287.
21. D. Chen, X. Zhang and A. F. Lee, J. Mater. Chem. A, 2015, 3, 14487–14516.
22. D. T. Whipple and P. J. A. Kenis, J. Phys. Chem. Lett., 2010, 1, 3451–3458.
23. K. P. Kuhl, E. R. Cave, D. N. Abram and T. F. Jaramillo, Energy Environ. Sci., 2012, 5, 7050–7059.
24. H. M. Jhong, S. Ma and P. J. A. Kenis, Curr. Opin. Chem. Eng., 2013, 2, 191–199.
25. C. Costentin, M. Robert and J. M. Savéant, Chem. Soc. Rev., 2013, 42, 2423–2436.
26. J. Qiao, Y. Liu, F. Hong and J. Zhang, Chem. Soc. Rev., 2014, 43, 631–675.
27. M. Halmann, Nature, 1978, 275, 115–116.
28. B. Kumar, M. Llorente, J. Froehlich, T. Dang, A. Sathrum and C. P. Kubiak, Annu. Rev. Phys. Chem., 2012, 63, 541–569.
29. J. Zhao, X. Wang, Z. Xu and J. S. C. Loo, J. Mater. Chem. A, 2014, 2, 15228–15233.
30. J. Yang, D. Wang, H. Han and C. Li, Acc. Chem. Res., 2013, 46, 1900–1909.
31. A. Corma and H. Garcia, J. Catal., 2013, 308, 168–175.
32. S. Xie, Y. Wang, Q. Zhang, W. Deng and Y. Wang, ACS Catal., 2014, 4, 3644–3653.
33. A. Fujishima, X. Zhang and D. A. Tryk, Surf. Sci. Rep., 2008, 63, 515–582.
34. Q. Tay, X. Liu, Y. Tang, Z. Jiang, T. C. Sum and Z. Chen, J. Phys. Chem. C, 2013, 117, 14973–14982.
35. A. Mattsson and L. Österlund, J. Phys. Chem. C, 2010, 114, 14121–14132.
36. A. Kudo, K. Omori and H. Kato, J. Am. Chem. Soc., 1999, 121, 11459–11467.
37. L. Liu, H. Zhao, J. M. Andino and Y. Li, ACS Catal., 2012, 2, 1817–1828.
38. P. Li, S. Ouyang, G. Xi, T. Kako and J. Ye, J. Phys. Chem. C, 2012, 116, 7621–7628.
39. P. Li, S. Ouyang, Y. Zhang, T. Kako and J. Ye, J. Mater. Chem. A, 2013, 1, 1185–1191.
40. G. Li, S. Ciston, Z. V. Saponjic, L. Chen, N. M. Dimitrijevic, T. Rajh and K. A. Gray, J. Catal., 2008, 253, 105–110.
41. J. Zhang, Q. Xu, Z. Feng, M. Li and C. Li, Angew. Chem., Int. Ed., 2008, 47, 1766–1769.
42. X. Wang, Q. Xu, M. Li, S. Shen, X. Wang, Y. Wang, Z. Feng, J. Shi, H. Han and C. Li, Angew. Chem., Int. Ed., 2012, 51, 13089–13092.
43. Q. D. Truong, T. H. Le, J. Y. Liu, C. C. Chung and Y. C. Ling, Appl. Catal., A, 2012, 437–438, 28–35.
44. H. Zhao, L. Liu, J. M. Andino and Y. Li, J. Mater. Chem. A, 2013, 1, 8209–8216.
45. A. J. Maira, K. L. Yeung, C. Y. Lee, P. L. Yue and C. K. Chan, J. Catal., 2000, 192, 185–196.
46. C. B. Almquist and P. Biswas, J. Catal., 2002, 212, 145–156.
47. K. Kočí, L. Obalová, L. Matějová, D. Plachá, Z. Lacný, J. Jirkovský and O. Šolcová, Appl. Catal., B, 2009, 89, 494–502.
48. O. K. Varghese, M. Paulose, T. J. LaTempa and C. A. Grimes, Nano Lett., 2009, 9, 731–737.
49. Q. Liu, Y. Zhou, J. Kou, X. Chen, Z. Tian, J. Gao, S. Yan and Z. Zou, J. Am. Chem. Soc., 2010, 132, 14385–14387.
50. S. Yan, L. Wan, Z. Li and Z. Zou, Chem. Commun., 2011, 47, 5632–5634.
51. Q. Liu, Y. Zhou, Z. Tian, X. Chen, J. Gao and Z. Zou, J. Mater. Chem., 2012, 22, 2033–2038.
52. Q. Liu, Y. Zhou, W. Tu, S. Yan and Z. Zou, Inorg. Chem., 2014, 53, 359–364.
53. Q. Liu, Y. Zhou, Y. Ma and Z. Zou, RSC Adv., 2012, 2, 3247–3250.
54. S. Feng, X. Chen, Y. Zhou, W. Tu, P. Li, H. Li and Z. Zou, Nanoscale, 2014, 6, 1896–1900.
55. P. Li, Y. Zhou, W. Tu, Q. Liu, S. Yan and Z. Zou, ChemPlusChem, 2013, 78, 274–278.
56. G. Xi, S. Ouyang, P. Li, J. Ye, Q. Ma, N. Su, H. Bai and C. Wang, Angew. Chem., Int. Ed., 2012, 51, 2395–2399.
57. W. Tu, Y. Zhou, Q. Liu, Z. Tian, J. Gao, X. Chen, H. Zhang, J. Liu and Z. Zou, Adv. Funct. Mater., 2012, 22, 1215–1221.
58. Y. T. Liang, B. K. Vijayan, O. Lyandres, K. A. Gray and M. C. Hersam, J. Phys. Chem. Lett., 2012, 3, 1760–1765.
59. Q. Liu, D. Wu, Y. Zhou, H. Su, R. Wang, C. Zhang, S. Yan, M. Xiao and Z. Zou, ACS Appl. Mater. Interfaces, 2014, 6, 2356–2361.
60. Z. Li, Y. Zhou, J. Zhang, W. Tu, Q. Liu, T. Yu and Z. Zou, Cryst. Growth Des., 2012, 12, 1476–1481.
61. X. Li, H. Pan, W. Li and Z. Zhuang, Appl. Catal., A, 2012, 413–414, 103–108.
62. K. Li, A. D. Handoko, M. Khraisheh and J. Tang, Nanoscale, 2014, 6, 9767–9773.
63. S. Xie, Y. Wang, Q. Zhang, W. Deng and Y. Wang, Chem. Commun., 2015, 51, 3430–3433.
64. X. Chen, Y. Zhou, Q. Liu, Z. Li, J. Liu and Z. Zou, ACS Appl. Mater. Interfaces, 2012, 4, 3372–3377.
65. Y. Xie, G. Liu, L. Yin and H. Cheng, J. Mater. Chem., 2012, 22, 6746–6751.
66. Y. Zhou, Z. Tian, Z. Zhao, Q. Liu, J. Kou, X. Chen, J. Gao, S. Yan and Z. Zou, ACS Appl. Mater. Interfaces, 2011, 3, 3594–3601.
67. C. Burda, X. Chen, R. Narayanan and M. A. El-Sayed, Chem. Rev., 2005, 105, 1025–1102.
68. Y. Li and W. Shen, Chem. Soc. Rev., 2014, 43, 1543–1574.
69. H. G. Yang, C. H. Sun, S. Z. Qiao, J. Zou, G. Liu, S. C. Smith, H. M. Cheng and G. Q. Lu, Nature, 2008, 453, 638–641.
70. H. Yang, G. Liu, S. Qiao, C. Sun, Y. Jin, S. Smith, J. Zou, H. Cheng and G. Lu, J. Am. Chem. Soc., 2009, 131, 4078–4083.
71. X. Han, Q. Kuang, M. Jin, Z. Xie and L. Zheng, J. Am. Chem. Soc., 2009, 131, 3152–3153.
72. J. Pan, X. Wu, L. Wang, G. Liu, G. Lu and H. Cheng, Chem. Commun., 2011, 47, 8361–8363.
73. W. Jiao, L. Wang, G. Liu, G. Lu and H. Cheng, ACS Catal., 2012, 2, 1854–1859.
74. H. Xu, S. Ouyang, P. Li, T. Kako and J. Ye, ACS Appl. Mater. Interfaces, 2013, 5, 1348–1354.
75. L. Ye, J. Mao, T. Peng, L. Zan and Y. Zhang, Phys. Chem. Chem. Phys., 2014, 16, 15675–15680.
76. Z. He, L. Wen, D. Wang, Y. Xue, Q. Lu, C. Wu, J. Chen and S. Song, Energy Fuels, 2014, 28, 3982–3993.
77. J. Yu, J. Low, W. Xiao, P. Zhou and M. Jaroniec, J. Am. Chem. Soc., 2014, 136, 8839–8842.
78. P. Li, Y. Zhou, Z. Zhao, Q. Xu, X. Wang, M. Xiao and Z. Zou, J. Am. Chem. Soc., 2015, 137, 9547–9550.
79. C. Wang, R. L. Thompson, P. Ohodnicki, J. Baltrus and C. Matranga, J. Mater. Chem., 2011, 21, 13452–13457.
80. H. Shi, G. Chen, C. Zhang and Z. Zou, ACS Catal., 2014, 4, 3637–3643.
81. G. Xi, S. Ouyang and J. Ye, Chem. – Eur. J., 2011, 17, 9057–9061.
82. S. I. In, D. D. Vaughn and R. E. Schaak, Angew. Chem., Int. Ed., 2012, 51, 3915–3918.
83. Y. He, L. Zhang, B. Teng and M. Fan, Environ. Sci. Technol., 2015, 49, 649–656.
84. K. Iizuka, T. Wato, Y. Miseki, K. Saito and A. Kudo, J. Am. Chem. Soc., 2011, 133, 20863–20868.
85. Y. Hori, H. Wakebe, T. Tsukamoto and O. Koga, Electrochim. Acta, 1994, 39, 1833–1839.
86. A. A. Peterson and J. K. Nørskov, J. Phys. Chem. Lett., 2012, 3, 251–258.
87. W. Wang, W. An, B. Ramalingam, S. Mukherjee, D. Niedzwiedzki, S. Gangopadhyay and P. Biswas, J. Am. Chem. Soc., 2012, 134, 11276–11281.
88. X. Zhang, F. Han, B. Shi, S. Farsinezhad, G. P. Dechaine and K. Shankar, Angew. Chem., Int. Ed., 2012, 51, 12732–12735.
89. Q. Zhai, S. Xie, W. Fan, Q. Zhang, Y. Wang, W. Deng and Y. Wang, Angew. Chem., Int. Ed., 2013, 52, 5776–5779.
90. S. Zhu, S. Liang, Y. Tong, X. An, J. Long, X. Fu and X. Wang, Phys. Chem. Chem. Phys., 2015, 17, 9761–9770.
91. C. Tsai, H. Chen, R. Liu, K. Asakura and T. Chan, J. Phys. Chem. C, 2011, 115, 10180–10186.
92. S. Neatu, J. A. Maciá-Agulló, P. Concepción and H. Garcia, J. Am. Chem. Soc., 2014, 136, 15969–15976.
93. Q. Kang, T. Wang, P. Li, L. Liu, K. Chang, M. Li and J. Ye, Angew. Chem., Int. Ed., 2015, 54, 841–845.
94. Z. Zhang, Z. Wang, S. Cao and C. Xue, J. Phys. Chem. C, 2013, 117, 25939–25947.
95. B. D. Mankidy, B. Joseph and V. K. Gupta, Nanotechnology, 2013, 24, 405402.
96. J. Yang, D. Wang, H. Han and C. Li, Acc. Chem. Res., 2013, 46, 1900–1909.
97. L. Liu, C. Zhao, H. Zhao, D. Pitts and Y. Li, Chem. Commun., 2013, 49, 3664–3666.
98. L. Liu, C. Zhao, D. Pitts and Y. Li, Catal. Sci. Technol., 2014, 4, 1539–1546.
99. S. Xie, Y. Wang, Q. Zhang, W. Fan, W. Deng and Y. Wang, Chem. Commun., 2013, 49, 2451–2453.
100. X. Meng, S. Ouyang, T. Kako, P. Li, Q. Liu, T. Wang and J. Ye, Chem. Commun., 2014, 50, 11517–11519.
101. Y. Fu, D. Sun, Y. Chen, R. Huang, Z. Ding, X. Fu and Z. Li, Angew. Chem., Int. Ed., 2012, 51, 3364–3367.
102. Q. Liu, Z. X. Low, L. Li, A. Razmjou, K. Wang, J. Yao and H. Wang, J. Mater. Chem. A, 2013, 1, 11563–11569.
103. Y. Liao, S. W. Cao, Y. Yuan, Q. Gu, Z. Zhang and C. Xue, Chem. – Eur. J., 2014, 20, 10220–10222.
104. G. Liu, S. Xie, Q. Zhang, Z. Tian and Y. Wang, Chem. Commun., 2015, 51, 13654–13657.
105. A. G. MacDiarmid and A. J. Epstein, Faraday Discuss. Chem. Soc., 1989, 88, 317–332.
106. S. Sato, T. Morikawa, S. Saeki, T. Kajino and T. Motohiro, Angew. Chem., Int. Ed., 2010, 49, 5101–5105.
107. T. Arai, S. Sato, K. Uemura, T. Morikawa, T. Kajino and T. Motohiro, Chem. Commun., 2010, 46, 6944–6946.
108. T. Arai, S. Tajima, S. Sato, K. Uemura, T. Morikawa and T. Kajino, Chem. Commun., 2011, 47, 12664–12666.
109. B. Kumar, J. M. Smieja, A. F. Sasayama and C. P. Kubiak, Chem. Commun., 2012, 48, 272–274.
110. M. Schreier, P. Gao, M. T. Mayer, J. Luo, T. Moehl, M. K. Nazeeruddin, S. D. Tilley and M. Grätzel, Energy Environ. Sci., 2015, 8, 855–861.
111. A. Paracchino, V. Laporte, K. Sivula, M. M. Grätzel and E. Thimsen, Nat. Mater., 2011, 10, 456–461.
112. Y. Kou, S. Nakatani, G. Sunagawa, Y. Tachikawa, D. Masui, T. Shimada, S. Takagi, D. A. Tryk, Y. Nabetani, H. Tachibana and H. Inoue, J. Catal., 2014, 310, 57–66.
113. G. Sahara, R. Abe, M. Higashi, T. Morikawa, K. Maeda, Y. Ueda and O. Ishitani, Chem. Commun., 2015, 51, 10722–10725.
114. D. H. Won, J. Chung, S. H. Park, E. H. Kim and S. I. Woo, J. Mater. Chem. A, 2015, 3, 1089–1095.
115. Y. Xia, P. Yang, Y. Sun, Y. Wu, B. Mayers, B. Gates, Y. Yin, F. Kim and H. Yan, Adv. Mater., 2003, 15, 353–389.
116. Y. Lin, G. Yuan, R. Liu, S. Zhou, S. W. Sheehan and D. Wang, Chem. Phys. Lett., 2011, 507, 209–215.
117. J. Jang, S. Cho, G. Magesh, Y. Jang, J. Kim, W. Kim, J. Seo, S. Kim, K. Lee and J. Lee, Angew. Chem., Int. Ed., 2014, 53, 5852–5857.
118. G. Ghadimkhani, N. R. de Tacconi, W. Chanmanee, C. Janaky and K. Rajeshwar, Chem. Commun., 2013, 49, 1297–1299.
119. A. T. Garcia-Esparza, K. Limkrailassiri, F. Leroy, S. Rasul, W. Yu, L. Lin and K. Takanabe, J. Mater. Chem. A, 2014, 2, 7389–7401.
120. D. Won, C. Choi, J. Chung and S. Woo, Appl. Catal., B, 2014, 158–159, 217–223.
121. J. Brito, A. Araujo, K. Rajeshwa and M. Zanoni, Chem. Eng. J., 2015, 264, 302–309.
122. X. Huang, T. Cao, M. Liu and G. Zhao, J. Phys. Chem. C, 2013, 117, 26432–26440.
123. Q. Shen, Z. Chen, X. Huang, M. Liu and G. Zhao, Environ. Sci. Technol., 2015, 49, 5828–5835.
124. U. Kang, S. K. Choi, D. J. Ham, S. M. Ji, W. Choi, D. S. Han, A. Abdel-Wahab and H. Park, Energy Environ. Sci., 2015, 8, 2638–2643.
125. J. Cheng, M. Zhang, G. Wu, X. Wang, J. Zhu and K. Cen, Environ. Sci. Technol., 2014, 48, 7076–7084.
126. J. Cheng, M. Zhang, G. Wu, X. Wang, J. Zhou and K. Cen, Sol. Energy Mater. Sol. Cells, 2015, 132, 606–614.
127. Y. P. Peng, Y. T. Yeh, S. I. Shah and C. P. Huang, Appl. Catal., B, 2012, 123–124, 414–423.
128. G. Magesh, E. S. Kim, H. J. Kang, M. Banu, J. Y. Kim, J. H. Kim and J. S. Lee, J. Mater. Chem. A, 2014, 2, 2044–2049.
129. S. Sato, T. Arai, T. Morokawa, K. Uemura, T. M. Suzuki, H. Tanaka and T. Kajino, J. Am. Chem. Soc., 2011, 133, 15240–15243.
130. T. Arai, S. Sato, T. Kajino and T. Morikawa, Energy Environ. Sci., 2013, 6, 1274–1282.

---