Semiconductor-based photocatalytic CO₂ conversion

Abstract

Climate change and its impact on the Earth and Society has been recently reassessed by the International Panel on Climate Change. The panel estimates that the greenhouse gas emissions should be reduced by half by 2030 to mitigate climate change. Photocatalytic CO₂ conversion is one of the promising technologies that can help with this modest goal. This review discusses the theoretical and practical aspects of CO₂ conversion over semiconducting photocatalysts and overviews the recently reported CO₂ conversion photocatalysts. A spectrum of photocatalysts reviewed in this work includes titania and its composites with metal oxides, metals, and advanced carbon allotropes; other solid photocatalysts, mostly based on germanium, gallium, tungsten, and niobium; graphitic carbon nitride; silver–silver halide plasmonic systems; photocatalytically active metal–organic frameworks; and graphene-based systems. Finally, a summary of the current state and an outlook for the future are provided.

---

Michal Marszewski

Michal Marszewski received his MSc degree in the field of chemistry in 2011 from the Faculty of New Technologies & Chemistry at the Military University of Technology (Warsaw, Poland). Currently, he is a PhD candidate in the Department of Chemistry & Biochemistry at Kent State University (Kent, OH, USA). His research interests include the design, synthesis, and characterization of nanostructured materials for environmental and energy-related applications. In his doctoral research, he mainly focuses on titania materials; however, he is also involved in studying silica, carbon, alumina, and metal–organic framework materials; for more information see Michal@LinkedIn.com/in/MichalMarszewski.

Shaowen Cao

Shaowen Cao received his BS in Geochemistry in 2005 from the University of Science and Technology of China, and his PhD in Materials Chemistry & Physics in 2010 from the Shanghai Institute of Ceramics, Chinese Academy of Sciences. Then he worked as a Research Fellow at the School of Materials Science and Engineering, Nanyang Technological University until Feb 2014. He is now a Professor at the Wuhan University of Technology. His current research interests include the design and fabrication of photocatalytic materials for energy and environmental applications. See more details on http://www.researcherid.com/rid/I-8050-2013.

Jiaguo Yu

Jiaguo Yu received his BS and MS in chemistry from Central China Normal University and Xi'an Jiaotong University, respectively, and his PhD in Materials Science in 2000 from the Wuhan University of Technology. In 2000, he became a Professor at the Wuhan University of Technology. He was a postdoctoral fellow at the Chinese University of Hong Kong from 2001 to 2004, a visiting scientist from 2005 to 2006 at the University of Bristol, a visiting scholar from 2007 to 2008 at the University of Texas at Austin. His current research interests include semiconductor photocatalysis, photocatalytic hydrogen production, CO2 reduction to hydrocarbon fuels, and so on. See more details on http://www.researcherid.com/rid/G-4317-2010.

Mietek Jaroniec

Mietek Jaroniec received his MS and PhD from M. Curie-Sklodowska University (Poland) in 1972 and 1976; afterward, he was appointed as a faculty at the same University. Since 1991, he has been Professor of Chemistry at Kent State University, Kent, Ohio (USA). His research interests include interfacial chemistry and chemistry of materials, especially adsorption at the gas/solid and liquid/solid interfaces and nanoporous materials. At Kent State he has established a vigorous research program in the area of nanomaterials such as ordered mesoporous silicas, organosilicas, inorganic oxides, carbon nanostructures, and nanostructured catalysts/photocatalysts, focusing on their synthesis, characterization and environmental and energy-related applications.

---

1 Introduction

The International Panel on Climate Change (IPCC) has just released the 5th version of the Assessment Report,^1–3 which states beyond doubt the existence of climate change. This report states that the human population and the resulting anthropogenic greenhouse gas (GHG) emissions contribute most to the warming effect.¹ Unfortunately, the emissions have been on the rise since the industrial revolution, despite the ongoing reduction efforts. As a result, the ongoing climate change has already had an impact on human health, agriculture, water resources, and ocean and land ecosystems.² The main GHG sources invariably remain fossil fuel combustion and industrial processes. The IPCC estimates that the effective mitigation of climate change will require cutting the GHG emissions by half by 2030.³ The technologies proposed to achieve this modest goal are renewable and nuclear energy, carbon capture and storage, and broadly defined improvement in energy efficiency. CO₂ conversion is one of the future technologies that fits these criteria.

CO₂ conversion is a process of chemical reduction of carbon dioxide.^4,5 The possible products are CO, HCHO, CH₃OH, and CH₄, depending on the number of transferred electrons (2e⁻, 4e⁻, 6e⁻, and 8e⁻, respectively). Although reported,^6–11 higher hydrocarbons are rarely the product. Naturally, the conversion is an endergonic process (ΔG > 0), considering that a very stable molecule is transformed into high-energy hydrocarbons. The process can be driven by various forms of energy, resulting in different conversion variants such as direct or catalyzed chemical conversion, thermochemical conversion, electrochemical conversion, and photochemical conversion.^12,13 The last variant is the most attractive option because it utilizes the sun's radiation as the energy source. In contrast, thermochemical and electrochemical processes require heat or electricity to drive the process. Of course, both of these can be supplied from green and renewable sources; however, the vision of direct conversion of solar energy into useful fuels is so much more appealing.

Photochemical CO₂ conversion is a “kill two birds with one stone” approach. The use of CO₂ as the feedstock can help with the required emission cuts, while the hydrocarbon products can serve as a green and renewable energy source (solar fuels). The process can consume either atmospheric or man-supplied CO₂ to help with its reduction. In the latter case, CO₂ feedstock could be supplied from the capture and storage strategy (CCS), yielding even more powerful results. Instead of much less appealing geological storage, CO₂ could be reused in the form of solar fuels.¹⁴ The latter could render the CCS as a more economical and attractive solution, in addition to solving the CO₂ storage problem. Secondly, the produced hydrocarbons have potential to replace the currently used fossil fuels, using the available infrastructure. Particularly, methane is widely used for electricity generation (natural gas power plants), heating, transportation, and as a chemical feedstock. Methanol, on the other hand, is already being tested in flex-fuel cars as a gasoline substitute. Thus, methane and methanol are the most desired products from the economic viewpoint.

Photochemical conversion of CO₂ is mostly carried out using solid photocatalysts, which are the sole focus of this review. By far, the most studied materials are semiconducting metal oxides and sulfides; however, a number of other materials have been studied for this purpose, including: graphitic carbon nitride, carbon nanotubes, graphene oxide, plasmonic nanoparticles, and metal–organic frameworks. Table 1 lists the recently reported CO₂ conversion photocatalysts, along with the product yield and experimental details. The following section discusses theoretical and practical aspects concerning every solid photocatalyst, before the discussion of specific photocatalytic systems in the third section. Finally, the last section provides a summary of the current state and an outlook for the future. In the intent to keep this review concise, yet informative and leading-edge, the reference selection comprises mostly reports published within the last three years.

Table 1 Selected photocatalysts reported for CO₂ conversion

Material description	Material efficiency (μmol g^−1 h^−1)	Reference efficiency (μmol g^−1 h^−1)	Normalized efficiency	Experimental details	Ref.
a Values reported by the authors (no reference data available for calculation of the normalized efficiency).
Titania–metal oxide composites
Pt–MgO/TiO2	11 (CH4)	0.35 (P25)	31	10 hours, 100 W Xe lamp, 320–780 nm, CO2 at 2.0 MPa	15
Porous microspheres of MgO-patched TiO2	2.1–21.3 (CO)	2.8–9.9 (P25)	0.75 (50 °C), 2.2 (150 °C)	5 hours, temp. range 50–150 °C	16
Magnesium-modified TiO2	1.0 (CH4)	0.22 (pristine TiO2)	4.5	Solar light simulator with an AM1.5 filter, 100 mW cm^−2	17
Cu2O@Pt core–shell loaded onto TiO2 (Degussa P25)	33 (CH4), 8.3 (CO), 25 (H2)	1.2 (CH4, P25), 2.5 (CO, P25), 4.9 (H2, P25)	28 (CH4), 3.3 (CO), 5.1 (H2)	200 W Xe lamp (320–780 nm), wet CO2 at 0.2 MPa	18
Cu2O/TiO2	28.4 (CH4, ppm g^−1 h^−1)	3.7 (P25, ppm g^−1 h^−1)	7.7	Static experiment, 10 hours, 300 W Xe lamp, wet CO2 at 80 kPa	19
Hybrid CuO–TiO2−xNx hollow nanocubes	41.3 (CH4, ppm g^−1 h^−1)	16.2 (P25, ppm g^−1 h^−1)	2.5	300 W Xe lamp with an AM1.5 filter, 100 mW cm^−2	20
Ordered mesoporous CeO2–TiO2 composites	1800–2100 (CH4), 12 000–13 100 (CO)	220 (CH4, P25), 5500 (CO, P25)	8.2–9.5 (CH4), 2.2–2.4 (CO)	300 W Xe arc lamp	21
Montmorillonite modified TiO2 nanocomposites	442 (CH4), 103 (CO)	—	12 (CH4)^a, — (CO)	500 W Hg lamp, UV light, at 100 °C	22
Ni^2+–TiO2 photocatalysts (nickel in NiO form)	7.7 (CH4), 19.5 (CH3OH)	—	—	UV light, 365 nm, 200 W Hg lamp, 33.42 mW cm^−2	23
Titania–metal composites
Ultrafine Pt nanoparticles on TiO2 single crystals	1360 (CH4)	11 (P25)	124	Continuous flow, wet CO2, 400 W Xe lamp, 19.6 mW cm^−2 in UV range	24
Cu/TiO2 (P25) nanoparticle catalysts	0.46–0.65 (CH4), 12–24 (CO)	—	189 (CH4, P25)^a, 10 (CO, P25)^a	6.5 hours, continuous CO2 flow, 150 W solar simulator, 90 mW cm^−2	25
Ag-doped TiO2	0.44 (CH4 + CH3OH)	0.16 (P25)	2.8	24 hours, 4750 mW cm^−2	26
Cu and Pt loaded, periodically modulated multiwalled TiO2 nanotube array catalyst	2.6 (CH4, mL g^−1 h^−1), 3.5 (total CxHy, mL g^−1 h^−1)	0.4 (CH4, pristine titania nanotube array, mL g^−1 h^−1), 0.5 (total CxHy, pristine titania nanotube array, mL g^−1 h^−1)	6.5 (CH4), 7 (total CxHy)	1 sun AM 1.5 simulated sunlight, wet 99.9% CO2 at 4900 Torr	6
Au/Pt nanoparticle-decorated TiO2 electrospun nanofibers	114 (CH4), 17 (CO)	6 (CH4, pristine TiO2 nanofibers), 12 (CO, pristine TiO2 nanofibers)	19 (CH4), 1.4 (CO)	500 W Xe lamp, wet CO2	27
Titania–carbon composites
TiO2–graphene 2D sandwich-like hybrid nanosheets	4.6–10.1 (CH4), 7.2–16.8 (C2H6)	0.69 (CH4, P25), 0 (C2H6, P25)	6.7–14.6 (CH4) N/A (C2H6)	300 W Xe arc lamp	7
Hollow spheres consisting of alternating titania nanosheets and graphene nanosheets	1.1 (CH4), 8.9 (CO)	0.69 (CH4, P25), 0.16 (CO, P25)	1.6 (CH4), 56 (CO)	300 W Xe arc lamp	28
MWCNT/TiO2 core–shell nanocomposites	0.17 (CH4)	0.03 (pristine anatase TiO2)	5.7	Continuous CO2 flow, atmospheric pressure, 15 W light bulb	29
Solvent-exfoliated-graphene loaded Ti nanosheets	9.3 (CH4, μmol m^−2 h^−1)	2.7 (pristine TiNS, μmol m^−2 h^−1)	3.4	100 W Hg vapor lamp, 365 nm, 110 W m^−2, H2O-saturated CO2, 3 hours	30
Single-walled carbon nanotube loaded Ti nanosheets	5.3 (CH4, μmol m^−2 h^−1)	2.7 (pristine TiNS, μmol m^−2 h^−1)	2	100 W Hg vapor lamp, 365 nm, 110 W m^−2, H2O-saturated CO2, 3 hours	30
Pure titania materials
Electrospun mesoporous TiO2 nanofibers after solvothermal treatment	3.08–19.55	0.78 (P25)	4–25	Ambient temperature and atmospheric pressure, 300 W Xe arc lamp with UV and visible irradiation, CO2 and H2O produced by reacting HCl with NaHCO3	31
Anatase TiO2 with coexposed {001} and {101} facets	1.35 (CH4)	0.38 (P25)	3.55	300 W simulated solar Xe arc lamp, CO2 and H2O generated in situ by reaction of NaHCO3 and HCl	32
Bicrystalline TiO2 with controllable anatase–brookite phase content	1.2 (100% A, CO), 2.1 (75 : 25 A : B, CO), 0.7 (100% B, CO)	1.3 (P25)	0.92 (100% A), 1.6 (75 : 25 A : B), 0.54 (100% B)	Continuous wet CO2 flow, 150 W solar simulator, 69.6 mW cm^−2 (200–1000 nm) and 12.5 mW cm^−2 (200–400 nm)	33
Hollow anatase TiO2 single crystals and mesocrystals with dominant {101} facets	1.6 (CH4)	0.3 (pristine anatase TiO2)	5.3	300 W Xe lamp, wet CO2 at 0.06 MPa	34
Anatase TiO2 nanosheets with exposed 95% {100} facets	6 (CH4, ppm g^−1 h^−1)	1.2 (cuboid shaped TiO2 with 53% of {100}, ppm g^−1 h^−1)	5	300 W Xe arc lamp, wet CO2 at 80 kPa	35
Non-titania photocatalysts
Mesoporous 4.5(ZnGa2O4):(Zn2GeO4)	3.8 (CH4), 22 (O2)	—	33 (CH4, ZnGa2O4)^a, 4.9 (O2, ZnGa2O4)^a	UV-enhanced (200 to 350 nm), 300 W Xe arc lamp, wet CO2	36
Single-crystalline, ultrathin ZnGa2O4 nanosheets	6.9 (CH4, ppm h^−1)	0.3 (bulk ZnGa2O4, ppm h^−1)	23	Ultraviolet enhanced 300 W Xe arc lamp with an IR cut filter + gas pump to accelerate gas diffusion, wet CO2	37
ZnGa2O4 nanocubes with exposed {100} facets	1.2 (CH4), 12 (O2)	0.18 (CH4, pristine mesoporous ZnGa2O4), 5 (O2, pristine mesoporous ZnGa2O4)	6.7 (CH4), 2.4 (O2)	UV-enhanced (200 to 350 nm), 300 W Xe arc lamp, wet CO2	38
ZnAl2O4-modified ZnGaNO (Pt loaded)	4.7 (CH4)	0.9 (pristine ZnGaNO, Pt loaded)	5.2	300 W Xe arc lamp (>420 nm), wet CO2	39
Porous Ga2O3	0.21 (CH4), 1.5 (CO)	—	4 (CH4, reference Ga2O3)^a	300 W Xe lamp with an IR cutting filter, 500 mW cm^−2, wet CO2	40
Zn-doped, Ag-modified Ga2O3	117 (CO), 16.9 (H2), 70.1 (O2)	—	—	1 wt% Ag as co-catalyst, 0.1 M NaHCO3 aqueous solution, CO2 bubbled through the solution at 30 mL min^−1, 400 W high-pressure mercury lamp	41
Mesoporous zinc germanium oxynitride (Pt loaded)	2.7 (CH4, ppm g^−1 h^−1)	1.1 (Pt-loaded bulk material, ppm g^−1 h^−1)	2.5	300 W Xe arc lamp (>400 nm), wet CO2 at 80 kPa	42
Cd2Ge2O6 nanowires (RuO2 loaded)	0.72 (CH4)	—	2.2 (unloaded Cd2Ge2O6)^a	UV-enhanced 300 W Xe arc lamp with a water filter + gas pump to accelerate gas diffusion, wet CO2	43
Sheaf-like, hyperbranched Zn1.7GeN1.8O nanoarchitectures (Pt & RuO2 loaded)	4.4 (CH4)	—	—	300 W Xe arc lamp (>420 nm), gas pump to accelerate gas diffusion, wet CO2	44
Bi2WO6 hollow microspheres	16.3 (CH3OH)	0.64 (pristine Bi2WO6)	25.5	In aqueous solution at 0 °C, 300 W Xe arc lamp (>420 nm)	45
Ultrathin, single-crystal WO3 nanosheets	1.2 (CH4)	0.09 (commercial WO3)	13	Wet CO2, 300 W Xe arc lamp (>420 nm)	46
Bare rectangular sheet-like WO3 crystals	0.34 (CH4)	—	—	300 W Xe lamp (>400 nm), wet CO2 at 0.06 MPa	47
SiO2-pillared HNb3O8	3.8 (CH4)	1.7 (anatase)	2.2	350 W Xe arc lamp, 34.8 mW cm^−2, 60 °C	48
NaNbO3 photocatalyst	2.45–5.7 (CH4)	—	—	CO2 at 80 kPa, 300 W Xe arc lamp	49
NaNbO3 photocatalyst	1.7–2.9 (CH4)	—	—	CO2 at 80 kPa, 300 W Xe arc lamp	50
Na2V6O16·H2O nanoribbons co-loaded with Pt and RuO2	0.19 (CH4)	0.0068 (pristine nanoribbons)	28	300 W Xe arc lamp (>420 nm), wet CO2	51
Cu2O/RuOx heterojunction photocatalyst	0.29 (CO)	0.05 (pristine Cu2O)	5.8	150 W Xe arc lamp, wet CO2, 0.7 M Na2SO3 as a hole scavenger	52
Ordered mesoporous Fe-doped CeO2 catalysts	2.9 (CH4), 11 (CO)	1.2 (CH4, pristine mesoporous CeO2), 6.3 (CO, pristine mesoporous CeO2)	2.4 (CH4), 1.7 (CO)	300 W Xe arc lamp with sun-like spectrum, wet CO2 at 30 °C and 110 kPa	53
In2O3/g-C3N4 hybrid photocatalysts	1990 (with Pt, ppm h^−1 g^−1), 958 (without Pt, ppm h^−1 g^−1)	—	—	500 W Xe lamp with 1200 mW cm^−2 light intensity, wet CO2	54
Red phosphor/g-C3N4 heterojunction photocatalyst (Pt loaded)	295 (CH4)	145 (r-P), 107 (g-C3N4)	2 (r-P), 2.8 (g-C3N4)	500 W Xe arc lamp, wet CO2	55
Ag-modified BaLa4Ti4O15	1.0 (HCOOH), 14.3 (CO), 33.3 (H2), 23.3 (O2)	—	—	1.0 wt% Ag, aqueous conditions, CO2 bubbled through the solution at 15 mL min^−1, 400 W high-pressure mercury lamp	56
Layered double hydroxides	0.5–3.7 (CO), 0.4–6.4 (O2)	—	—	Aqueous conditions, over 10 h, 200 W Hg–Xe lap equipped with fiber optics, a collective lens, and a mirror	57
Silver–silver halide systems
Plasmonic shaped AgX:Ag	38 (AgCl:Ag, CH3OH), 22 (AgBr:Ag, CH3OH)	—	—	300 W Xe arc lamp (>400 nm), in water solution, 5 hours	58
Ag@AgBr/AgCl heterostructured nanocashews	45 (CH3OH), 66 (CH3CH2OH)	—	—	Visible light (>400 nm), in water solution at 0 °C, 5 hours	8
Red Ag/AgCl photocatalyst	29 (CH3OH), 45 (CH3CH2OH)	—	—	500 W Xe arc lamp (>420 nm), in water solution, 5 hours	9
Ag@AgBr/CNT nanocomposites	28.45 (CH4), 17.21 (CH3OH), 2.94 (CO), 7.11 (CH3CH2OH)	—	—	150 W Xe lamp with a UV cutoff filter (λ > 420 nm), measured light intensity 22.4–25 mW cm^−2, 0.2 M KHCO3 solution, agitated with magnetic stirring, vessel pressurized with CO2 to 7.5 MPa	59
Ternary alloyed AgClxBr1−x nanocrystals	36 (CH3OH), 72 (CH3CH2OH)	—	—	500 W Xe arc lamp (>420 nm), in water solution, 5 hours	10
Metal–organic frameworks
Copper porphyrin based metal–organic framework	263 (CH3OH, ppm g^−1 h^−1)	37.5 (unmodified MOF, ppm g^−1 h^−1)	7	Visible light, aqueous solution	60
Amine-functionalized zirconium metal–organic framework NH2-Uio-66(Zr)	26.4 (HCOOH)	16.3 (NH2-MIL-125(Ti))	1.6	500 W Xe lamp, 420–800 nm, in acetonitrile with triethanolamine	61
Amine-functionalized titanium metal–organic framework NH2-MIL-125(Ti)	16 (HCOOH)	—	—	500 W Xe lamp, 420–800 nm, in acetonitrile with triethanolamine	62
Other inorganic–organic hybrid systems
ZIF-8/Zn2GeO4nanorod composite (Pt loaded)	0.28 (CH3OH)	0.14 (pristine Zn2GeO4 nanorods)	2	500 W Xe arc lamp, in 0.1 M Na2SO3 solution, thermostatted	63
Polymeric-semiconductor–metal-complex hybrid photocatalyst:mesoporous C3N4 loaded with Ru complex (cis,trans-[Ru{4,4′-(CH2PO3H2)2-2,2′-bipyridine}(CO)2Cl2]	375–483 (HCOOH), 36–75 (CO), 32–43 (H2)	—	—	20 vol.% triethanolamine in acetonitrile, 450 W high-pressure Hg lamp (>400 nm)	64
Graphene-based systems
Cu2O/reduced graphene oxide composites	75 (CO, ppm g^−1 h^−1)	5.1 (P25, ppm g^−1 h^−1)	14	150 W Xe lamp, wet CO2, Na2S as a hole scavenger	65
Graphene oxide	0.172 (CH3OH)	0.03 (P25)	5.7	Continuous gas flow, 25 °C, 300 W halogen light, wet CO2	66
Reduced graphene oxide–CdS nanorods	2.5 (CH4)	0.21 (pure CdS nanorods)	12	300 W Xe arc lamp (>420 nm), 150 mW cm^−2	67
Porphyrin and graphene containing photocatalyst	170 (C2H2 + CH4, μmol m^−2 h^−1)	—	—	Wet CO2, solar light simulator, 100 mW cm^−2, at 45 °C	11

---

2 Theoretical and practical considerations

This section discusses the theoretical and practical aspects of photochemical CO₂ conversion over a solid photocatalyst. The process requires a chain of specific steps that must occur to complete the conversion. The process starts with adsorption of a CO₂ molecule on the photocatalyst surface, where the molecule is susceptible to reduction by photogenerated electrons. These electrons are generated through light absorption, excitation, and charge separation. First, a bound electron in the photocatalyst absorbs a quantum of electromagnetic radiation, provided its energy is equal to or greater than a gap between the valence band (VB) and conduction band (CB). The consumed energy excites the electron from the VB to the CB, leaving a deficit of negative charge in the VB. The latter is referred to as a “hole” and serves as a positive charge carrier. Both charge carriers travel through the photocatalyst, in their respective bands, until they reach the surface. Then the electrons can serve as a reducing agent for the adsorbed CO₂ molecules, while the holes can drive a charge-balancing oxidation reaction. After the reduction process, the product molecule (e.g., methane) has to desorb and diffuse to the gas phase to accomplish the whole conversion process. Notably, all these subsequent steps will impact the overall product yield (a measure of photoactivity); thus, understanding their mechanism is crucial for improving the photoactivity.

2.1 Adsorption

A seemingly trivial adsorption step is not only the necessity but often the foundation toward improving the photoactivity of solid photocatalysts. If more molecules are adsorbed on the surface, then more quickly electrons can reach and reduce them. The latter results in lower surface recombination (less time spent on the surface) and higher product yield (more electrons consumed per unit time). At equilibrium, the number of adsorbed molecules is determined by equilibrium uptake. During the conversion process, however, CO₂ adsorption is not at the equilibrium due to the constant conversion of CO₂. Thus, the actual uptake will depend on the equilibrium uptake (driving force) and how fast the molecules can be replenished (gas diffusion and adsorption kinetics). According to adsorption thermodynamics, the equilibrium uptake is a function of: (1) partial pressure of an adsorbate, (2) temperature, (3) surface area, and (4) surface energetics. Clearly, by increasing the CO₂ pressure and lowering the temperature one can achieve higher CO₂ uptake but these parameters are purely experimental and can be controlled. From the material viewpoint, the surface area and surface energetics are the important aspects.

Surface area. Notably, the surface area of a photocatalyst is an important quantity. A large surface area can accommodate more CO₂ molecules and also more photogenerated electrons. The latter relates to quicker reduction and higher product yield. In addition, the large surface area relates to lower charge carrier density, which results in smaller recombination probability. Over time, scientists established various strategies to develop materials with large surface areas, which can be achieved by exposing more of either external or internal surface. In the former case, one or more of the material's dimensions are reduced to nanosize, resulting in: nanoparticles (0D),⁶⁸ nanorods (1D),^69,70 and nanosheets (2D).⁷¹ In the latter case, an internal surface area is developed by introduction of porosity, which can be achieved through either templating or structuring.⁷² A detailed discussion of these strategies goes beyond the scope of this review but exhaustive information can be found in the provided references. Regardless of the method, the development of materials with large surface area should be a priority for every photocatalyst.

Surface energetics. In parallel, the surface energetics is an important factor as well. The surface energy depends on the structure, composition, and chemistry; and scientists exploit them all to improve the CO₂ uptake and photoactivity. A common strategy is modification of the surface composition and/or chemistry to enhance the basic properties. CO₂ is an acidic molecule; thus, basic surfaces show stronger interactions and higher uptake of CO₂.^{15–17,19,60–62,67} Another powerful and often used strategy is the exposure of high-energy facets on the photocatalyst's surface; for instance, {001} facets of titania⁷³ or {110} facets of ZnGa₂O₄.³⁷ High-energy facets differ in the atomic structure and bonding environment from low-energy, stable facets. In the case of titania, {101} facets (the most stable and dominant) consist of 5 and 6-coordinated Ti atoms (1 : 1 ratio), whilst {001} facets (highly energetic) consist of exclusively 5-coordinated Ti.⁴⁷ This unsaturated environment results in the improved adsorbent–adsorbate interactions and photoactivity.

Generally, the most stable, low-energy facets dominate the surface to minimize the total surface free energy; however, specific synthetic paths can afford materials with exposed high-energy facets to a high degree. Usually, the desired facets are capped with protecting molecules to deteriorate their activity and favour growth along the low-energy facets. Afterwards, the removal of capping agents results in exposing the highly energetic surface. Unfortunately, the capping molecules are rather dependent on the type of material and facet; for instance, HF is often used to obtain {001} TiO₂ facets.⁷³ To achieve the maximum exposure, the materials are structured in the form of nanosheets with the desired facets at the top and the bottom. This arrangement affords as much as 95% of high-energy facets, as well as enlarges the specific surface area.³⁵ So far, {001} facets of anatase have been dubbed the most energetic ones⁷³ but some authors reported varied data: Jiao et al.³⁴ reported TiO₂ with {101} facets that performed 5-times better than pristine material and Xu et al.³⁵ reported titania sheets with 95% exposed {100} facets that performed 5-times better than a material with only 53% exposed {100} facets. A similar diversity is noticeable for other compounds as well.^37,38,46,47 Although these discrepancies are apparent, the general conclusion holds true that the exposure of specific crystal facets can be used to significantly improve the photoactivity.

Mass transfer phenomena. Once the conversion starts and CO₂ is transformed into products, new molecules have to adsorb to supply the ongoing process. The proceeding adsorption process triggers a pressure gradient between the surface proximity and the bulk gas phase, which drives new molecules to move toward the surface. If the supply rate – diffusion and adsorption kinetics – is poor, the product yield will suffer because (1) less CO₂ molecules will reach the surface in the same time and (2) the pressure in close proximity to the surface will be lower than in the bulk, which will further impact the CO₂ uptake. The problem can occur in nanostructured materials with very small pores, voids, or constrictions. Although small pores feature a large surface area, they can impact gas diffusion. To avoid such problems, nanostructured photocatalysts should possess mesopores (2–50 nm), which deliver a good trade-off between the surface area and mass transfer properties. Importantly, all reactants (typically CO₂ and H₂O) and all products (of both reduction and oxidation) deserve the same consideration as their shortage or accumulation (respectively) will impact the product yield. Fortunately, mesoporosity improves mass transfer from the bulk gas to the surface proximity and vice versa, serving all the components at once, thus greatly benefiting the product yield.

2.2 Light absorption, excitation, and charge separation

On the surface of a photocatalyst, CO₂ molecules undergo reduction by photogenerated electrons and water undergoes oxidation by holes. To take part in these reactions, the charge carriers have to (1) form through a light absorption and subsequently an electron excitation, and (2) reach the surface. Consequently, these two steps will impact the performance of a photocatalyst.

Light absorption. As discussed above, electrons in a semiconductor absorb only the photons with specific energy, equal to or greater than a band gap, which depends solely on the VB and CB positions. Thus, band engineering is the primary strategy for improving light absorption and charge carrier generation. Since the products' chemical energy is independent of the charge carriers' energy, the VB and CB edges should be close to the required redox potentials, just enough to drive the process. Unfortunately, most of the currently investigated photocatalysts are wide-band gap semiconductors. TiO₂, for instance, has a band gap of 3.2 eV, which limits its absorption to <390 nm wavelengths and wastes ca. 2 eV during water oxidation. Placing the VB edge (ca. 2.9 eV) closer to the water oxidation potential (0.82 eV, pH = 7) would result in a smaller band gap and a greater light absorption. Fig. 1A shows the band edge positions of the selected photocatalysts discussed in this work along with electrochemical potentials of water oxidation and CO₂ reduction. By far, the most common strategy for band engineering is doping with either metal or non-metal heteroatoms and titania is the most extensively studied in this regard.^79,80 The provided references contain a detailed discussion of this comprehensively studied subject. Other interesting strategies of band engineering include quantum confinement,⁴⁶ crystal facet-dependent electronic effect,⁴⁷ and chemical modification.^36,43,61,62 On the other hand, light harvesting can be improved by modification with either organic or inorganic components: metal oxides and sulfides,^18–21 carbon allotropes,^29,30 dyes and other optically active molecules,^11,64,81 and plasmonic nanoparticles.^8,9,27,58

Fig. 1 (A) Band edge positions for the selected photocatalysts discussed in this work along with electrochemical potentials for water oxidation and CO₂ reduction; the values of CV and VB positions adapted from ref. 11, 19, 24, 36, 39, 40, 46, 54, 58, 65–67 and 74–78 and (B) structure of a heterojunction with good charge separation and improved light absorption.

Bulk and surface recombination. After an excitation, electrons and holes (on their way to the surface) may recombine leading to their deactivation. In such a case, electrons de-excite from the CB to the VB and the excess energy is lost through different relaxation paths. Recombination occurs in the bulk and on the surface (bulk recombination and surface recombination, respectively). The more time a charge carrier spends in the bulk and on the surface, the higher the chance of its recombination. Time is a common enemy; however, bulk and surface recombination can be overcome differently.

The time a charge carrier spends in the bulk depends on its mobility and distance to the surface. Thus, a photocatalyst with high carrier mobility and short bulk-to-surface distance will suffer less of bulk recombination. The mobility is rather carrier and material specific but it may be negatively impacted in some cases. Crystal defects and impurities will slow the carriers' transport – either through trapping or scattering – and intensify bulk recombination. Specifically, Ohtani et al.⁸² investigated a set of amorphous-anatase TiO₂ mixtures and showed that the (1) photoactivity of amorphous TiO₂ was negligible, (2) photoactivity of the mixtures increased with increasing anatase fraction, and (3) photoactivity of anatase improved upon further calcination. Thus, crystallinity and purity are important in suppressing bulk recombination. Besides, some morphologies – such as nanoparticles, nanorods, nanosheets, and ordered porous materials – allow an easy control over a bulk-to-surface distance. Reducing their dimensions such as diameter, width, and/or pore wall thickness shortens the carriers' path in the bulk and lessens the recombination. Ideally, the path should be infinitely small; however, reduction below an average path length already yields significant improvements. Consequently, a supreme photocatalyst would be of high crystallinity and purity and had a short bulk-to-surface distance. Notably, these two priorities may stand astray as the simultaneous improvement of crystallinity and retention for a nanostructure is often difficult or impossible.⁸³

In contrast, the time spent on the surface depends on how quickly charge carriers are consumed in the redox reactions. The main improvement strategies include an increase in the reactants' uptake and surface area, which were discussed in Section 2.1. Briefly, the accumulation of reactants helps the carriers reach them more quickly; whereas larger surface area lowers the charge carrier surface densities and reduces their encounter probability.

Charge separation at the heterojunction. Spatial separation of the charge carriers is a powerful strategy to prevent bulk and surface recombination. The separation takes place between two materials that have energy levels appropriately positioned with respect to each other (see Fig. 1B). The materials create a heterojunction that only electrons or holes can pass through. As a result, one material accumulates mostly electrons while the other mostly holes. Bulk and surface recombination are greatly suppressed as there are virtually no free charges of the opposite sign in the respective materials. In addition, the separation of charge carriers results in spatial separation of redox processes and their products. The latter limits undesired back and side reactions that otherwise could affect the product yield.⁸⁴ Finally, the second material often provides additional benefits, such as improved light harvesting, surface chemistry, and structural properties. For all these reasons, heterojunction materials often show several-fold improved product yields, which is the case for a majority of TiO₂ and non-TiO₂ composites listed in Table 1.

The same effect can develop between two crystal phases of a single compound, which can explain an often-dubbed synergistic effect in mixed-phase materials. TiO₂, for instance, has three common crystal phases: anatase, brookite, and rutile that differ slightly in their band gap energy and the VB and CB positions.⁸⁵ As a result, mixed phases may yield an improved charge separation and thus an improved photoactivity.³³

2.3 Desorption

After the reaction, the newly formed products must desorb and transfer to the gas phase. A poor desorption or mass transfer will impact the product yield because (1) adsorbed products occupy a space that could adsorb fresh reactants and (2) products on the surface are susceptible to back and/or side reactions. This part of the process is virtually reversed to the transport and adsorption of the substrate molecules; thus, the same considerations apply. Ideally, the product should quickly desorb after the reaction and transfer from the surface proximity to the bulk of the gas phase.

Experimental setup. In a static experiment, the products will accumulate in the gas phase over time and later in close proximity to the surface, effectively hindering desorption and affecting the product yield. By the same reasoning, the substrates may over time deplete. Both instances may result in a decrease in the conversion efficiency with experiment time. To minimize those effects: (1) products' concentrations should be small throughout an experiment or the products should be withdrawn in a continuous manner and (2) substrates' concentrations should be large throughout an experiment or the substrates should be supplied in a continuous manner. Liquid products – methanol and methanal – have to be specially considered as their condensation within pores will result in the pore blockage and surface area reduction. In contrast, methane and CO are not condensable under the usually used experimental conditions.

2.4 Concept of the normalized yield and final remarks

Experimental conditions, reported yield, and normalized yield. Clearly, various experimental conditions will impact the overall product yield independent of the material properties. Light source, reactor construction, and photocatalyst preparation will determine the number and energy of photons reaching a photocatalyst. Temperature, pressure, experiment mode (static or continuous), and other details will affect all the steps from adsorption/desorption and mass transfer to light absorption, charge carrier generation, transport, separation, and utilization. As a result, product yields, measured in different experiments, cannot be meaningfully compared to evaluate the relative photoactivity.

To arrive at a rational comparison, we propose to use a normalized yield quantity. The normalized yield is defined as a ratio of the reported photocatalyst yield to the reference photocatalyst yield obtained under the same experimental conditions. We assume that two photocatalysts can be reasonably compared using their normalized yields, if the reference photocatalyst was the same. The widespread use of Degussa P25 as a reference photocatalyst enabled compilation of the P25-normalized yields for many of the materials listed in Table 1. Although convenient, this approach makes two major assumptions (1) photoactivity of Degussa P25 is reproducible and (2) experimental conditions equally affect the photoactivity of all materials. Number 1 implies that a difference in Degussa P25 product yields is purely due to the difference in experimental conditions. Number 2 implies that a change in experimental conditions that caused an n-fold change in the Degussa P25 product yield would cause also an n-fold change in the product yield of the photocatalyst studied. Whereas the first assumption seems plausible as P25 is a well-studied commercial material, the second assumption may not hold in some instances. Mainly, Degussa P25 is active only in UV; thus, changes outside the UV range will be misrepresented unless they equally propagate in the UV range. Still, a comparison of the normalized yields remains more reasonable than the straight comparison of absolute yields. A brief survey of Table 1 shows that the CH₄ yields span over five orders of magnitude, whereas the corresponding normalized yields span over two orders of magnitude only. Unfortunately, not all publications report the yield on P25, which is by far the most common for TiO₂-related studies. The authors hope that the outlined advantages of the normalized yield will encourage reporting of the P25 yields for all photocatalytic systems.

Finally, it should be mentioned that the catalytic activity often does not change linearly with the amount of catalyst and approaches saturation at high photocatalyst amounts. The reported product yields, expressed per gram of a photocatalyst, represent a best scenario for each photocatalyst, because most often only milligram quantities of the photocatalyst are used in the regime of the highest activity per gram.

Importance of oxidation. In most cases the studies of CO₂ conversion are exclusively focused on the reduction part, even though the conversion requires both reduction and oxidation to proceed. Oxidation is necessary to supply reduction with hydrogen and consume holes to balance the charge. A sluggish oxidation will deprive the reduction of the necessary hydrogen and increase recombination due to hole accumulation, negatively impacting the product yield. For instance, Yan et al.³⁶ showed that an improvement of oxidation positively affected the methane yield. Iizuka et al.⁵⁶ reported Ag-modified BaLa₄Ti₄O₁₅ that stoichiometrically produced O₂ during CO₂ conversion. More recently, Teramura et al.⁴¹ reported Zn-doped, Ag-modified Ga₂O₃ that consumed electrons and holes at the same rate resulting in highly efficient CO₂ conversion. These examples show that the oxidation part of CO₂ conversion has a strong effect on the photoactivity and an equal consumption of electrons and holes is beneficial for the photocatalytic process.

Concluding remarks. Section 2 presents a generic photochemical CO₂ conversion with water oxidation over a semiconducting photocatalyst. A real-life example of such a system is conversion of water-saturated CO₂ over the TiO₂ photocatalyst under UV illumination. The outlined steps and relevant parameters are readily applicable to most other photocatalytic systems. For all the rest, the essential ideas and reasoning remain valid but the specifics need adaptation; for instance, (1) diffusion and adsorption phenomena will differ between gas and liquid systems and (2) the mechanism of charge carrier generation, transport, and separation will vary for some materials (plasmonic nanoparticles and organic–inorganic photocatalysts are primary examples). A widely studied photocatalytic evolution of hydrogen from water, for instance, is different from the discussed CO₂ conversion only in the following aspects: (1) the former occurs rather in the liquid phase than in the gaseous one, (2) uses water as a single reactant for both reduction and oxidation (although, often a sacrificial agent for oxidation is used to improve the activity), and (3) requires a different reduction potential (or a suitable co-catalyst, such as Pt).

Two parameters – photon-to-charge carrier quantum efficiency and CO₂-to-product yield – were not discussed in the above example. Although they impact the product yield, their control is not well established. Properties at the heart of catalysis, such as stability, cyclability, turnover number, etc., are not discussed as well. More often than not, publications lack the discussion of these properties, mainly because of their purely exploration nature. The aim of this section was to outline the theoretical and practical aspects of the CO₂ conversion process before reviewing the recently reported photocatalytic systems.

3 Materials review

3.1 Titania–metal oxide composites

Titania is by far the most widely studied photocatalyst. Pure TiO₂, however, has some strong drawbacks such as poor light absorption. Thus, TiO₂ is often combined with other metal oxides (MOs) to afford various composite photocatalysts with superior photoactivity, thanks to improved charge separation,^15,19,23 light absorption,^19–21,23 structural properties,^21–23 and surface chemistry.^15–17,23 As discussed in Section 2.2, the charge separation is achieved at the heterojunction between two materials. TiO₂–MO composites can take advantage of this effect, if their VBs and CBs are properly matched. In addition, the charge carrier separation results in separation of redox reactions as well, which prevents back and side reactions and benefits the product yield. If incorporated MO has a smaller band gap than TiO₂, it will improve the composite's light absorption. Importantly, the VBs and the CBs must be well matched to achieve simultaneous charge separation and improved light absorption. If the incorporated MO is nanostructured, it is beneficial for structural parameters – surface area and/or porosity – of the composite. Otherwise, if either TiO₂ or MO is in a bulk form, it negatively impacts the structural properties. Finally, the surface chemistry of the incorporated MO may be more favourable than TiO₂; for instance, basic MOs can improve CO₂ adsorption properties if incorporated in the framework or deposited on the surface of TiO₂. Depending on the MO used, the composite may take advantage of one or more of these benefits. The most common MOs reported in the recent literature are MgO^15–17 and Cu_xO^18–20,86 but other MOs, such as NiO^23,87,88 and CeO,²¹ are reported as well.

Xie et al.¹⁵ investigated various metal oxides deposited on TiO₂/Pt, including: MgO, SrO, CaO, BaO, La₂O₃, and Lu₂O₃ and found that the basicity had a strong impact on the CH₄ yield (see Fig. 2A). All MO–TiO₂ composites showed an increase in the CH₄ yields; in particular, the MgO–TiO₂ composite showed the highest yield. The authors attributed this result to chemisorption of CO₂ on MgO patches that concentrated and destabilized CO₂ molecules to be more reactive. In addition, the authors discussed a possibility of charge separation between Pt and MgO and inhibition of CH₄ oxidation as a result. Liu et al.¹⁶ observed an increase in the product yield over MgO-patched TiO₂ microspheres as well (see Fig. 2B); however, the detected product was CO instead of CH₄. This difference may have been due to the lack of Pt loading or the elevated temperature used for conversion of CO₂ (50–150 °C).

Fig. 2 (A) Product yield (CH₄ and CO) plotted as a function of CO₂ uptake for a series of Pt–TiO₂ photocatalysts modified with basic metal oxides (reprinted with permission from ref. 15. Copyright The Royal Society of Chemistry 2013); (B) porous MgO-patched TiO₂ microspheres: (a and b) SEM images, (c) elemental mapping, and (d) HRTEM images (reprinted with permission from ref. 16. Copyright The Royal Society of Chemistry 2013).

Xu et al.¹⁹ prepared the Cu₂O–TiO₂ photocatalyst by loading porous TiO₂ with Cu₂O. Copper(I) oxide has a band gap of 2.0–2.2 eV, effectively absorbing the visible light, and its VB and CB are well matched to create a heterojunction with TiO₂. As a result, Cu₂O–TiO₂ composites benefit from both charge separation and better light harvesting. In addition, the authors structured the material to shorten the distance from the bulk to the surface by using porous TiO₂ and Cu₂O nanoparticles. The resulting structure had a large specific surface area (206 m² g⁻¹) and pore volume (0.29 cm³ g⁻¹). Although the surface area and pore volume increased only slightly after Cu₂O loading, the CO₂ uptake increased by more than 2-fold as compared to pure TiO₂. All these factors improved the charge separation, light absorption, CO₂ uptake, surface area and porosity, which resulted in over 7-fold higher CH₄ yield as compared to the Degussa P25 catalyst (and pure TiO₂).

Zhai et al.¹⁸ prepared a Pt and Cu co-loaded TiO₂ photocatalyst through photoreduction of H₂PtCl₆ and CuSO₄ in two subsequent steps. This unique approach resulted in the deposition of Pt nanoparticles on the P25 surface in the first step and reduction of Cu on these nanoparticles in the second step. The final material consisted of Cu₂O–Pt core–shell structures loaded on the TiO₂ surface. The material scored an impressive 28-fold higher CH₄ yield than the original Degussa P25. The authors concluded that the Cu₂O layer helped to destabilize CO₂ molecules and suppressed the concurrent H₂O reduction over Pt, while the Pt core served as a collector of electrons.

In et al.²⁰ prepared a CuO–TiO_2−xN_x photocatalyst with hollow cube morphology through a facile multi-template strategy (see Fig. 3). The authors compared it with the CuO–TiO₂ composite prepared through traditional impregnation, and with Degussa P25. Interestingly, the hollow cubes had the lowest specific surface area of all three materials: 24 m² g⁻¹ (CuO–TiO_2−xN_x), 31 m² g⁻¹ (CuO–TiO₂), and 50 m² g⁻¹ (Degussa P25). Yet, the cubes showed a 2.5-fold higher product yield compared to Degussa P25 and 1.6-fold higher compared to plain CuO–TiO₂. This improvement may be attributed to the nitrogen doping and/or hollow cube morphology.

Fig. 3 (a) Schematic illustration of the multistep template strategy that afforded CuO–TiO_2−xN_x hollow cubes (TNBT = titanium n-butoxide); (b) TEM image of Cu₃N nanocubes; (c) TEM image of TiO₂@Cu₃N; and (d) FESEM images of CuO–TiO_2−xN_x hollow nanocubes (reprinted with permission from ref. 20. Copyright Wiley-VCH 2012).

Wang et al.²¹ prepared CeO₂–TiO₂ nanostructured photocatalysts through nanocasting using ordered mesoporous silica, SBA-15, as a template. Incorporation of CeO₂ extended the absorption from the UV to the visible range and narrowed the band gap to 2.16–2.24 eV. Interestingly, even pure nanocast TiO₂ showed a narrowed band gap of 2.81 eV. Only Degussa P25 showed a close-to-usual value of 3.10 eV. In addition, due to the nanostructuring effort, the resulting materials had a large specific surface area of 140–180 m² g⁻¹ and pore volumes of 0.53–0.61 cm³ g⁻¹. The materials achieved up to 9.5-fold higher CH₄ yield and up to 2.4-fold higher CO yield as compared to Degussa P25.

3.2 Titania–metal composites

Metals can improve the photoactivity of TiO₂ as well, which is mainly achieved through charge separation^6,24–27 and/or plasmon-enhanced photocatalysis.²⁷ Metals, as good conductors, readily accept electrons, serving as an electron sink. The electron withdrawal results in the charge separation, which benefits the conversion process. In addition, some metal nanoparticles exhibit surface plasmon resonance, which may be used to improve the photocatalyst's light harvesting efficiency. The resulting plasmon-enhanced photocatalysis is typically explained by either: (1) the charge transfer mechanism or (2) local electric field enhancement; however, there seems to be no definite agreement on this issue. Ref. 89 provides a good insight into the theoretical and practical aspects of this phenomenon. Often, metals are incorporated as a necessary co-catalyst, resulting in a product that would not be produced otherwise. Among all the reported metals, the noble metals – Pt,^6,24,27 Au,²⁷ Ag,²⁶ and Cu^6,25 – seem to be employed the most often.

The size and structure of both TiO₂ material and loaded metal have a crucial effect on the photocatalytic activity. The latter was accentuated by Wang et al.²⁴ who prepared films composed of 1D TiO₂ single crystals and ultrafine Pt nanoparticles (0.5–2 nm; see Fig. 4). The films achieved extraordinary 124-fold higher yield than Degussa P25, which was attributed to the charge separation on well dispersed Pt particles and fast electron transfer within 1D single crystals. Zhang et al.⁶ prepared a periodically modulated multi-walled TiO₂ nanotube array through pulse anodization of the Ti substrate. The array loaded with Cu and Pt showed a 6.5-fold higher yield compared to the pristine nanotubes.

Fig. 4 1D single-crystal TiO₂ thin film loaded with ultrafine Pt nanoparticles: (a) FESEM image (white rectangle indicates the position of the TEM image); (b) AFM image; (c) elemental mapping (green = Ti and red = Pt); (d) TEM images (white rectangle indicates the position of the HRTEM); (e) schematic illustration of the conversion process and energy level diagram (CB = conduction band; VB = valence band; H⁺ = proton; h⁺ = hole; reprinted with permission from ref. 24. Copyright American Chemical Society 2012).

Synergistic effects are very important as well. Electrospun TiO₂ fibres decorated with Au and Pt²⁷ showed a modest 19-fold increase in CH₄ yield, which was due to the combined effect of charge separation on Pt and plasmon enhancement of Au. Liu et al.²⁵ prepared a Cu-loaded TiO₂ photocatalyst through a simple precipitation method. The precipitated material contained Cu²⁺ species and was thermally treated in He and H₂ to reduce copper, resulting in a mix of Cu⁺ and Cu⁰ species. Both these copper species can enhance the TiO₂ performance by improved light absorption, charge separation, and surface chemistry. The synergistic effect of Cu⁺ and Cu⁰ resulted in a whopping 189-fold higher CH₄ yield as compared to Degussa P25.²⁵

3.3 Titania–carbon composites

Advanced carbon allotropes (ACAs) are appealing alternatives for improving the photoactivity of TiO₂. Especially graphene^7,28,30 and carbon nanotubes^29,30 have been used to improve CO₂ conversion because of their well-developed surface area and attractive optical and electrical properties. The electronic properties depend on the specific allotrope: graphene is a conductor on a par with metals, while graphene oxide shows semiconducting properties, and carbon nanotubes are either conducting or semiconducting depending on their structure. As a result, ACAs can substitute both metals and metal oxides in TiO₂ composites due to their flexible, tunable, and potent electronic properties. In addition, their large surface area benefits the conversion process in more than one way, as discussed in detail in Section 2. Finally, the light absorption of ACAs often extends to the visible range, allowing for improved light harvesting of TiO₂–ACA composites. Thus, the ACAs improve charge generation, separation, and transport, and possess a large surface area where the substrates and charge carriers can be collected and reacted.

Liang et al.³⁰ showed a very interesting effect of ACA dimensionality on the photoactivity of TiO₂–ACA (see Fig. 5). The authors prepared two types of titania nanosheet–ACA composites: one with carbon nanotubes (1D) and the second with graphene (2D). The authors observed higher photoluminescence quenching in the graphene composite, which implies a better charge separation, presumably due to the favourable orientation and electronic coupling of the sheets. The latter resulted in superior product yield under UV light illumination. Under visible light illumination, however, the nanotube composite scored the higher result. This was because of the sensitization of TiO₂ by carbon nanotubes – which exhibited strong absorption in visible light and resulted in a greatly improved light harvesting.

Fig. 5 SEM image of (a) the SWCNT–TiNS composite and (b) the SEG–TiNS composite; schematic illustration of (c) the SWCNT–TiNS composite and (d) the SEG–TiNS composite; (e) photooxidation rate constant of acetaldehyde under ultraviolet and visible irradiation; (f) results of CO₂ conversion under ultraviolet and visible irradiation (SWCNT = single-walled carbon nanotubes; SEG = solvent-exfoliated graphene; TiNS = titania nanosheets; reprinted with permission from ref. 30. Copyright American Chemical Society 2012).

This strategy was used by Gui et al.²⁹ who prepared TiO₂/multi-walled carbon nanotube core–shell structures. The conductive nanotubes served as a sensitizer and an electron sink and the resulting photocatalyst showed an almost 6-fold higher CH₄ yield as compared to the pristine TiO₂.

Tu et al.^7,28 prepared two interesting TiO₂–graphene systems: (1) hollow spheres of alternating titania and graphene sheets and (2) sandwich-like TiO₂–graphene hybrid nanosheets. Both materials showed improvement in the photocatalytic performance over Degussa P25; however, the sandwich structure achieved better performance and efficiently converted CO₂ to CH₄ and higher alkanes, whereas the hollow spheres produced mostly CO. The improvement of up to 15-fold is most probably due to charge separation between well-coupled TiO₂ and graphene in the sandwich structure. Importantly, this sandwich-type structure is competitive in relation to many metal-loaded TiO₂ photocatalysts.

3.4 Non-titania photocatalysts

Noticeably, titania is the most widely studied photocatalyst for CO₂ conversion. Fortunately, the number of non-titania photocatalysts that are developed and studied continues to increase. Over half of the materials listed in Table 1 are photocatalysts not based on TiO₂. These materials offer new or improved properties, often unmatched or hard to match by TiO₂. The former include smaller band gap, better matched VB and/or CB, more favourable surface chemistry, and new or excelling structural, physical, and crystallographic properties. Despite these exciting properties, the performance may sometimes be limited as compared to TiO₂, which is not surprising because it took a couple of decades to perfect the latter. However, it is noteworthy that these new materials offer room to grow with a set of promising features.

A brief survey of Table 1 shows that the current efforts in the development of inorganic photocatalysts are focused on gallium, germanium, and tungsten-based materials (oxides and salts); however, niobium, vanadium, copper, and cerium compounds are reported as well. In addition, silver–silver halide plasmonic systems, photocatalytic metal–organic frameworks, organic–inorganic composite photocatalysts, and graphene-based composites are widely encountered.

Gallium and germanium-based materials. The most studied gallium and germanium compounds within the last three years are ZnGa₂O₄ and Zn₂GeO₄. These materials facilitate reduction of CO₂ to CH₄ without the need for Pt or other co-catalysts, which is due to their high CB positions. Unfortunately, both these compounds have wide band gaps, ca. 4.5 eV (ref. 36 and 37) and ca. 4.65 eV, respectively.⁴⁴ As compared to TiO₂, these materials show inferior light absorption. Thus, scientists develop strategies to narrow the band gap and improve light harvesting. Yan et al.³⁶ prepared a solid solution of Zn₂GeO₄ in ZnGa₂O₄ to investigate the effect of Ge on the band gap. The resulting mixed compound had a band gap of 4.18 eV, smaller than either of the constituents. The narrowing was a result of VB upshift and CB downshift, due to the contributions of s and p orbitals of Ge, as evidenced by UV-Vis diffusive reflectance spectroscopy, valence band XPS, and DFT calculations. In addition to the band gap narrowing, the authors emphasized the importance of oxidation reaction and its kinetics. Incorporation of Zn₂FeO₂ resulted in the enhancement of H₂O oxidation due to the improvement of hole mobility, and reflected positively on the CH₄ yield. Liu et al.⁴³ developed a generalized synthesis of metal germanate nanowires and showed that the change of the cation from zinc to cadmium narrowed the band gap to 3.9 eV. The resulting Cd₂Ge₂O₆ nanowires were further doped with cobalt, which significantly narrowed the band gap to 2.9 eV.

Alternative to incorporation of metals, band gap narrowing is achieved by incorporation of non-metals. For instance, Yan et al.³⁹ prepared ZnGaNO (E_g = 2.55 eV); Zhang et al.⁴² prepared zinc germanium oxynitride with varying stoichiometry (E_g = 2.6–2.8 eV); and Liu et al.⁴⁴ prepared sheaf-like, hyper-branched nanostructures of Zn_1.7GeN_1.8O (E_g = 2.6 eV; see Fig. 6A). In all these examples, incorporation of nitrogen resulted in considerable narrowing of the band gap. Consequently, all compounds showed visible light activity.

Fig. 6 (A) Zn₂GeO₄ sheaf-like, hyperbranched nanostructures: (a–g) FESEM images, (h) UV-Vis absorption spectra of Zn₂GeO₄ (a) and Zn_1.7GeN_1.8O (b), (i) photo images of Zn₂GeO₄ (a′) and Zn_1.7GeN_1.8O (b′) (reprinted with permission from ref. 44. Copyright The Royal Society of Chemistry 2012); (B) single-crystalline, ultrathin ZnGa₂O₄ nanosheet microspheres: (a–d) FESEM images and (e) results of CO₂ conversion for different ZnGa₂O₄ materials (reprinted with permission from ref. 37. Copyright American Chemical Society 2014).

Park et al.⁴⁰ showed the importance of porosity in the product yield. The prepared porous gallium oxide (Ga₂O₃) exhibited a 2-fold larger specific surface area, 3-fold higher CO₂ uptake, and 4-fold higher CH₄ yield, as compared to a commercially available gallium oxide. The result exemplifies the significant effect of porosity and surface area on the activity of this photocatalyst.

Similar to TiO₂, many authors pointed out the effect of surface energetics on the product yield. Liu et al.³⁷ reported synthesis of microspheres composed of single-crystalline, ultrathin ZnGa₂O₄ nanosheets with exposed {110} crystal facets (see Fig. 6B). The obtained photocatalyst scored a 23-fold higher CH₄ yield as compared to the bulk ZnGa₂O₄. This enhancement may have come partially from an 11-fold larger surface area than the bulk material. Thus, the authors compared the prepared spheres to mesoporous ZnGa₂O₄ with a 2 times larger surface area, and still noted a 35% higher yield. Even more astonishing is the result of Yan et al.³⁸ who synthesized ZnGa₂O₄ nanocubes with mainly exposed {100} crystal facets. The material achieved a 6.7-times higher CH₄ yield than that of mesoporous ZnGa₂O₄ with a 10-times larger specific surface area.

Tungsten-based materials. Tungsten-based materials are promising candidates for CO₂ conversion that exhibit visible light absorption and photoactivity. For instance, Cheng et al.⁴⁵ synthesized Bi₂WO₆ hollow microspheres, which had a specific surface area of 24 m² g⁻¹ and a band gap of 2.76 eV. The material converted CO₂ to methanol in the aqueous phase under visible light irradiation. WO₃ is another example of photocatalyst with a narrow band gap (2.6–2.7 eV). Unfortunately, its conduction band is too low to directly facilitate CO₂ reduction. Chen et al.⁴⁶ and Xie et al.⁴⁷ showed two interesting strategies to overcome this problem. Both groups synthesized WO₃ nanosheets with uplifted CB that efficiently converted CO₂ under visible light irradiation (see Fig. 7). Chen et al.⁴⁶ argued that the extremely thin nanosheets (4–5 nm, ca. six repeating unit cells) induced a quantum confinement effect and resulted in changes in the band structure. The nanosheets had a 0.16 eV wider band gap and a 0.47 eV raised CB edge, as compared to a commercial WO₃. Xie et al.⁴⁷ on the other hand, assumed that the band structure of WO₃ can be engineered by exploiting the crystal facet-dependent electronic structure effect. The authors showed that WO₃ nanosheets with exposed {002} crystal facets reduced CO₂ to CH₄, whereas WO₃ cubes with equal amounts of {200}, {020}, and {002} facets were inactive. The authors proposed that {002} crystal facets have a higher CB edge than the bulk WO₃ due to the different atomic structure.

Fig. 7 (A) Ultrathin, single-crystal WO₃ nanosheets: (a–c) TEM images, (d) crystalline model, (e) energy level diagram (reprinted with permission from ref. 46. Copyright American Chemical Society 2012); (B) WO₃ cubes and nanosheets: (a) schematic illustration of the synthesis procedure, (b and c) SEM images of H_0.23WO₃ nanocubes, (d) SEM image of cube-like WO₃ obtained by calcination of H_0.23WO₃, (e–g) SEM images of sheet-like WO₃, and (h) energy level diagram of the cube-like and sheet-like WO₃ (reprinted with permission from ref. 47. Copyright The Royal Society of Chemistry 2012).

Niobium-based materials. Niobium-based photocatalysts are gaining some attention as well. Li et al.⁴⁸ synthesized a SiO₂-pillared HNb₃O₈ photocatalyst and Li et al.^49,50 synthesized NaNbO₃-based photocatalysts, which reduced CO₂ to CH₄; however, these materials had wide band gaps (3.3–3.5 eV) and were only active in the UV range. Despite this fact, the SiO₂-pillared HNb₃O₈ photocatalyst still showed better performance than anatase TiO₂ (E_g = 3.2 eV). HNb₃O₈ is a layered compound and the authors were able to increase its interlayer spacing from 0.37 nm to 2.10 nm. This change resulted in a 30-fold increase in the specific surface area (from 6.5 m² g⁻¹ to 197 m² g⁻¹) and a 6.2-fold increase in the CH₄ yield as compared to the bulk material. The CH₄ yield was 2.2-times higher as compared to TiO₂ with a specific surface area of 152 m² g⁻¹. This interesting result was explained in terms of the layered structure that could be regarded as a stack of nanosheets. The pillaring resulted in nanosheet separation and brought all the benefits of such a structure: improved charge transport, shortened distance to the surface, and enlarged specific surface area.

Graphitic carbon nitride-based materials. Graphitic carbon nitride (g-C₃N₄) is yet another promising photocatalyst.^90–92 g-C₃N₄ has a narrow band gap of 2.7 eV and can facilitate both CO₂ reduction and H₂O oxidation. Although the loading of noble metals is not required to facilitate the reduction process, it brings significant yield improvements due to enhanced charge separation. For instance, Yu et al.⁹² prepared a series of g-C₃N₄ loaded with various amounts of Pt (0–2%) and investigated their activity for CO₂ conversion (see Fig. 8A). The obtained products were CH₄, CH₃OH, and HCHO; however, the aldehyde yield was small as compared to the other products. The optimal concentration of Pt (0.75–1%) resulted in a 4.3-fold increase in the CH₄ yield and a 2.2-fold increase in the CH₃OH yield. Yuan et al.⁵⁵ used a red phosphor (r-P) as an electron acceptor and reported a 2.8-fold increase in the CH₄ yield as compared to phosphor-free g-C₃N₄ (both materials were Pt loaded).

Fig. 8 (A) g-C₃N₄ loaded with Pt: (a) TEM image, (b) HRTEM image, (c) UV-Vis diffuse reflectance spectra (the insets show photo images of the indicated samples), (d) results of CO₂ conversion under solar simulated light for 4 hours (reprinted with permission from ref. 92. Copyright The Royal Society of Chemistry 2014); (B) red Ag/AgCl photocatalyst: (a–c) SEM images, (d) photo image, (e) band gap estimation using Kubelka–Munk theory and UV-Vis diffuse reflectance spectra, and (f) results of CO₂ conversion experiments (reprinted with permission from ref. 9. Copyright The Royal Society of Chemistry 2014).

Silver–silver halide plasmonic systems. Ag/AgX systems are an interesting family of photocatalysts. Silver halides (AgCl and AgBr) are wide-band gap semiconductors (3.25 eV and 2.67 eV, respectively)⁹³ that show photocatalytic activity. Ag nanoparticles, on the other hand, exhibit strong surface plasmon resonance. Thus, a combination of Ag nanoparticles and AgX compounds results in the plasmon-enhanced photocatalysis and visible light response for CO₂ conversion.⁵⁸ Interestingly, these systems tend to convert CO₂ to methanol and ethanol rather than CH₄. For instance, Cai et al.⁹ reported a red-colour Ag/AgCl photocatalyst, which performed better than a regular Ag/AgCl photocatalyst (see Fig. 8B). Although the red photocatalyst was not compared to Degussa P25 in CO₂ reduction, it showed a significantly higher performance in methyl orange oxidation and Cr(IV) reduction. The authors attributed the strong colour and the high light harvesting ability to the plasmonic effect and unique morphology that facilitated Mie scattering. The same group prepared AgCl, AgBr, and AgCl_xBr_1−x compounds and compared their CO₂ conversion abilities.¹⁰ The ternary compounds showed intermediate band gaps between AgCl and AgBr. Interestingly, the highest yield was achieved by the ternary compound with a band gap of 2.9 eV, rather than AgBr (ca. 2.5 eV). This material showed 2- to 4-fold higher yield as compared to AgCl and AgBr, respectively.

Metal–organic frameworks. Metal–organic frameworks (MOFs) are 3D crystalline microporous materials composed of metal–oxo clusters interconnected with organic linker molecules. Thanks to their extremely large surface area and porosity, and superior flexibility in tailoring their structure and composition (both linkers and metals), MOFs are widely employed in various applications such as adsorption, separation, and gas storage. A main drawback of these materials is weak-to-moderate thermal and hydrothermal stability, which is due to organic linkers – mainly weak organic acids – making the structure. Usually, MOFs have to be modified to achieve photochemical activity. Probably the first MOF-based photocatalytic CO₂ conversion was reported by Wang et al.⁹⁴ who doped the UiO-67 MOF with ligands containing Ir, Re, and Ru. The Re-containing MOF showed CO₂ photoreduction activity. Since then, various authors have reported new MOFs for CO₂ photoreduction. For instance, Fu et al.⁶² prepared a Ti-containing and amine-functionalized MIL-125 MOF (denoted as NH₂-MIL-125(Ti); see Fig. 9A). The NH₂-MIL-125(Ti) MOF reduced CO₂ to HCOO⁻ under visible irradiation, whereas MIL-125(Ti) (MOF without the amine functionality) showed only UV activity. In addition, the amine functionality increased the CO₂ uptake, further benefiting the conversion process. Similar results were found by Sun et al.⁶¹ who functionalized the UiO-66(Zr) MOF with the amine moiety. The NH₂-UiO-66(Zr) showed a 1.6-fold higher yield as compared to the NH₂-MIL-125(Ti).

Fig. 9 (A) Ti-containing and amine-functionalized MOF NH₂-MIL-125(Ti): (a) FESEM image, (b) UV-Vis spectra and photo images of MIL-125(Ti) (a) and NH₂-MIL-125(Ti) (b), (c) result of CO₂ conversion by NH₂-MIL-125(Ti) (a) and MIL-125(Ti) (b) under 420–800 nm irradiation (reprinted with permission from ref. 62. Copyright Wiley-VCH 2012); (B) reduced graphene oxide–CdS nanorod composite: (a) TEM image of graphene oxide, (b) TEM image of CdS nanorods, (c and d) TEM and HRTEM images of the RGO–CdS composite, (e) UV-Vis diffuse reflection spectra and the photo images of the indicated materials, and (f) CO₂ conversion results under visible light irradiation (reprinted with permission from ref. 67. Copyright The Royal Society of Chemistry 2014).

Liu et al.⁶⁰ showed the importance of CO₂ activation (destabilization of CO₂ through sorption and interaction with the framework that renders it more reactive) in the conversion process. The authors prepared two porphyrin-based MOFs: one with plain porphyrin and the second with Cu-containing porphyrin. The Cu-containing MOF showed a 2-fold higher CO₂ uptake and 7-fold higher product yield. This marked improvement was due to the CO₂ activation upon adsorption and interaction with Cu porphyrin, which was proved by FTIR studies.

Carbon-based systems. Systems based on graphene and graphene-related compounds are another subgroup of non-titania photocatalysts.⁹⁵ As discussed in the previous section, graphene and other ACAs exhibit a well-developed surface area and unique electronic and optical properties. For instance, graphene oxide (GO) is a wide-band gap semiconductor; however, its band gap depends on the carbon-to-oxygen ratio. Hsu et al.⁶⁶ exploited this property and synthesized a set of GO materials with varying oxygen content. The resulting band gaps varied from 2.9 to 4.4 eV and all materials showed better CH₃OH yield than Degussa P25 under similar irradiation conditions (the authors reported that CH₃OH was the only product detected).

Graphene and its derivatives can be used to replace noble metals as an electron sink. For instance, Yu et al.⁶⁷ and An et al.⁶⁵ synthesized reduced graphene oxide (RGO) composites with CdS and Cu₂O, respectively. The first group observed over 10-fold increase in the CH₄ yield after RGO incorporation and over 1.6-fold increase as compared to the optimized Pt-loaded CdS, which show that RGO has potential to outperform Pt as a co-catalyst (see Fig. 9B). The authors pointed a few features that could explain this observation: (1) RGO is close in its properties to graphene and performs similarly well as an electron sink, (2) CO₂ preferentially adsorbs on RGO (after RGO incorporation the surface area increased by ca. 17% whereas CO₂ uptake doubled), (3) RGO-adsorbed CO₂ undergoes activation due to strong π–π conjugation interactions, and (4) high concentration of CO₂ and accumulated electrons on the RGO surface enhance the reduction performance. The second group reported similar observations, with the RGO/Cu₂O photocatalyst achieving a 14-fold higher CO yield than that of Degussa P25 (no other products were detected).

Wu et al.¹¹ showed a photocatalytic system composed of porphyrin and graphene, where porphyrin acted as a light harvester and graphene served as an electron collector and CO₂ adsorbent. Interestingly, acetylene was the primary product, whereas CH₄ constituted only 1/3 of the total carbohydrate production.

4 Summary and future outlook

The IPCC set a modest goal of reducing the greenhouse gas emissions in the next one and a half decade by 50%. This goal may be difficult to achieve considering that despite the ongoing reduction efforts the anthropogenic GHG emissions have been on the rise and the increase rate has been the highest since 1970, when the industrial revolution started. The technologies proposed to achieve this goal, such as carbon capture and storage, may help in a short term but a more sustainable approach will be necessary in the future.

Photochemical CO₂ conversion is one promising long-term solution. It may help to achieve the required emission cuts and simultaneously serve as a new source of hydrocarbon fuels. The process can be applied as an extension of the carbon and capture strategy, where instead of geological storage CO₂ is recycled into other useful molecules. Because CO₂ is captured and recycled before it may reach the atmosphere, the desired effect can be achieved despite the increasing emissions, resulting from the ongoing growth of the industry and manufacturing. Thus, this strategy may sustain the current and future stages of development.

In addition, the hydrocarbon products, methane and methanol, can substitute for the currently used fossil fuels using the existing infrastructure. These solar fuels may be regarded as a green, renewable, and sustainable energy source because they are produced with supply of the solar energy and easily replenishable CO₂ and H₂O.

CO₂ conversion is mainly done over the solid semiconducting photocatalysts. An ideal photocatalyst should possess: (1) possibly a large surface area with favourable chemistry and energetics, (2) well-developed suitable porosity with good mass transfer properties, (3) a narrow band gap and good light harvesting properties, (4) CB and VB edges positioned properly to drive both CO₂ reduction and H₂O oxidation, (5) high purity and crystallinity, (6) proper morphology with a short bulk-to-surface distance, and (7) an efficient spatial charge separation.

In addition to these features, the experimental conditions will affect the observed product yield independent of the material's photoactivity. Thus, the product yield should always be reported along with the yield obtained for a reference photocatalyst (possibly Degussa P25) under the same experimental conditions. This approach enables a meaningful comparison of various photocatalysts tested in different experiments.

Titania is by far the most widely studied photocatalyst for CO₂ conversion and it is often used in combination with other metal oxides, metals, and advanced carbon allotropes (carbon nanotubes, graphene). Non-titania photocatalysts include germanium, gallium, tungsten, and niobium-based materials, graphitic carbon nitride, silver–silver halide plasmonic systems, photocatalytically active metal–organic frameworks, and graphene-based systems.

Numerous improvements of the TiO₂ photoactivity have been reported; however, its strongest drawbacks, the wide band-gap and the related poor light absorption, remain a problem. The non-titania photocatalysts, on the other hand, offer new or improved properties, often unmatched or hard to match by TiO₂, which include a smaller band gap, better matched VB and/or CB, more favourable surface chemistry, and new or excelling structural, physical, and crystallographic properties. Although not as well studied, these materials offer a strong start-up and potential for future improvements, often using the same strategies developed for TiO₂ throughout the years. Considering that the four decades of research – since the seminal report by Fujishima and Honda⁹⁶ – did not crown TiO₂ as a perfect photocatalyst, further development and studies of new photocatalysts should be imperative.

Several photocatalysts show better light harvesting properties than TiO₂ and are good candidates to surpass it. The former include copper oxides, graphitic carbon nitride, zinc germanium oxynitrides and zinc gallium oxynitrides, tungsten-based compounds, and advanced carbon allotropes. Care should be exercised, however, to account for the material's stability, cyclability and other features necessary for a true catalyst. The unquestionable success of commercial TiO₂, despite the aforementioned downsides, may partly be attributed to its stability, cyclability, non-toxicity, and other important properties, which every prospective photocatalyst should possess besides good light harvesting and photoactivity.

Finally, a deeper understanding of the photocatalytic process is necessary. The latter can help to optimize the surface and other material properties as well as uncover new important aspects and/or ideas. A combination of experimental and theoretical results from fundamental studies and computer simulations should be preferred as it has the strongest potential to advance the current state of knowledge. As a result, novel and superior photocatalytic CO₂ conversion systems could be designed.

Acknowledgements

J. Yu would like to acknowledge the 973 program (2013CB632402), and NSFC (51272199, 51320105001 and 21433007) and Deanship of Scientific Research (DSR) of King Abdulaziz University (90-130-35-HiCi) for support.

Notes and references

1. IPCC, Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 2013.
2. IPCC, Climate Change 2014: Impacts, Adaptation, and Vulnerability. Working Group II Contribution to the IPCC 5th Assessment Report, Final draft, 28 October 2013.
3. IPCC, Climate Change 2014: Mitigation of Climate Change. Working Group III Contribution to the IPCC 5th Assessment Report, Final draft, 17 December 2013.
4. W. Tu, Y. Zhou and Z. Zou, Adv. Mater., 2014, 26, 4607–4626.
5. K. Li, X. An, K. H. Park, M. Khraisheh and J. Tang, Catal. Today, 2014, 224, 3–12.
6. X. Zhang, F. Han, B. Shi, S. Farsinezhad, G. P. Dechaine and K. Shankar, Angew. Chem., Int. Ed., 2012, 51, 12732–12735.
7. W. Tu, Y. Zhou, Q. Liu, S. Yan, S. Bao, X. Wang, M. Xiao and Z. Zou, Adv. Funct. Mater., 2013, 23, 1743–1749.
8. C. An, J. Wang, C. Qin, W. Jiang, S. Wang, Y. Li and Q. Zhang, J. Mater. Chem., 2012, 22, 13153–13158.
9. B. Cai, J. Wang, S. Gan, D. Han, Z. Wu and L. Niu, J. Mater. Chem. A, 2014, 2, 5280–5286.
10. B. Cai, J. Wang, D. Han, S. Gan, Q. Zhang, Z. Wu and L. Niu, Nanoscale, 2013, 5, 10989–10995.
11. T. Wu, L. Zou, D. Han, F. Li, Q. Zhang and L. Niu, Green Chem., 2014, 16, 2142–2146.
12. A. Goeppert, M. Czaun, J. Jones, G. K. S. Prakash and G. A. Olah, Chem. Soc. Rev., 2014, 43, 7995–8048.
13. E. V. Kondratenko, G. Mul, J. Baltrusaitis, G. O. Larrazabal and J. Perez-Ramirez, Energy Environ. Sci., 2013, 6, 3112–3135.
14. G. Sneddon, A. Greenaway and H. H. P. Yiu, Adv. Energy Mater., 2014, 4, 1301873.
15. S. Xie, Y. Wang, Q. Zhang, W. Fan, W. Deng and Y. Wang, Chem. Commun., 2013, 49, 2451–2453.
16. L. Liu, C. Zhao, H. Zhao, D. Pitts and Y. Li, Chem. Commun., 2013, 49, 3664–3666.
17. M. Manzanares, C. Fabrega, J. Oriol Osso, L. F. Vega, T. Andreu and J. Ramon Morante, Appl. Catal., B, 2014, 150, 57–62.
18. Q. Zhai, S. Xie, W. Fan, Q. Zhang, Y. Wang, W. Deng and Y. Wang, Angew. Chem., Int. Ed., 2013, 52, 5776–5779.
19. H. Xu, S. Ouyang, L. Liu, D. Wang, T. Kako and J. Ye, Nanotechnology, 2014, 25, 165402.
20. S. In, D. D. Vaughn II and R. E. Schaak, Angew. Chem., Int. Ed., 2012, 51, 3915–3918.
21. Y. Wang, B. Li, C. Zhang, L. Cui, S. Kang, X. Li and L. Zhou, Appl. Catal., B, 2013, 130, 277–284.
22. M. Tahir and N. S. Amin, Appl. Catal., B, 2013, 142, 512–522.
23. O. Ola and M. M. Maroto-Valer, J. Catal., 2014, 309, 300–308.
24. W. Wang, W. An, B. Ramalingam, S. Mukherjee, D. M. Niedzwiedzki, S. Gangopadhyay and P. Biswas, J. Am. Chem. Soc., 2012, 134, 11276–11281.
25. L. Liu, F. Gao, H. Zhao and Y. Li, Appl. Catal., B, 2013, 134, 349–358.
26. S. Krejcikova, L. Matejova, K. Koci, L. Obalova, Z. Matej, L. Capek and O. Solcova, Appl. Catal., B, 2012, 111, 119–125.
27. Z. Zhang, Z. Wang, S. Cao and C. Xue, J. Phys. Chem. C, 2013, 117, 25939–25947.
28. 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.
29. M. M. Gui, S. Chai, B. Xu and A. R. Mohamed, Sol. Energy Mater. Sol. Cells, 2014, 122, 183–189.
30. Y. T. Liang, B. K. Vijayan, O. Lyandres, K. A. Gray and M. C. Hersam, J. Phys. Chem. Lett., 2012, 3, 1760–1765.
31. J. Fu, S. Cao, J. Yu, J. Low and Y. Lei, Dalton Trans., 2014, 43, 9158–9165.
32. J. Yu, J. Low, W. Xiao, P. Zhou and M. Jaroniec, J. Am. Chem. Soc., 2014, 136, 8839–8842.
33. H. Zhao, L. Liu, J. M. Andino and Y. Li, J. Mater. Chem. A, 2013, 1, 8209–8216.
34. W. Jiao, L. Wang, G. Liu, G. Q. Lu and H. Cheng, ACS Catal., 2012, 2, 1854–1859.
35. H. Xu, S. Ouyang, P. Li, T. Kako and J. Ye, ACS Appl. Mater. Interfaces, 2013, 5, 1348–1354.
36. S. Yan, J. Wang, H. Gao, N. Wang, H. Yu, Z. Li, Y. Zhou and Z. Zou, Adv. Funct. Mater., 2013, 23, 1839–1845.
37. 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.
38. S. Yan, J. Wang, H. Gao, N. Wang, H. Yu, Z. Li, Y. Zhou and Z. Zou, Adv. Funct. Mater., 2013, 23, 758–763.
39. S. Yan, H. Yu, N. Wang, Z. Li and Z. Zou, Chem. Commun., 2012, 48, 1048–1050.
40. H. Park, J. H. Choi, K. M. Choi, D. K. Lee and J. K. Kang, J. Mater. Chem., 2012, 22, 5304–5307.
41. K. Teramura, Z. Wang, S. Hosokawa, Y. Sakata and T. Tanaka, Chem.–Eur. J., 2014, 20, 9906–9909.
42. N. Zhang, S. Ouyang, T. Kako and J. Ye, Chem. Commun., 2012, 48, 1269–1271.
43. Q. Liu, Y. Zhou, W. Tu, S. Yan and Z. Zou, Inorg. Chem., 2014, 53, 359–364.
44. Q. Liu, Y. Zhou, Z. Tian, X. Chen, J. Gao and Z. Zou, J. Mater. Chem., 2012, 22, 2033–2038.
45. H. Cheng, B. Huang, Y. Liu, Z. Wang, X. Qin, X. Zhang and Y. Dai, Chem. Commun., 2012, 48, 9729–9731.
46. X. Chen, Y. Zhou, Q. Liu, Z. Li, J. Liu and Z. Zou, ACS Appl. Mater. Interfaces, 2012, 4, 3372–3377.
47. Y. P. Xie, G. Liu, L. Yin and H. Cheng, J. Mater. Chem., 2012, 22, 6746–6751.
48. X. Li, W. Li, Z. Zhuang, Y. Zhong, Q. Li and L. Wang, J. Phys. Chem. C, 2012, 116, 16047–16053.
49. P. Li, S. Ouyang, G. Xi, T. Kako and J. Ye, J. Phys. Chem. C, 2012, 116, 7621–7628.
50. P. Li, S. Ouyang, Y. Zhang, T. Kako and J. Ye, J. Mater. Chem. A, 2013, 1, 1185–1191.
51. S. Feng, X. Chen, Y. Zhou, W. Tu, P. Li, H. Li and Z. Zou, Nanoscale, 2014, 6, 1896–1900.
52. E. Pastor, F. M. Pesci, A. Reynal, A. D. Handoko, M. Guo, X. An, A. J. Cowan, D. R. Klug, J. R. Durrant and J. Tang, Phys. Chem. Chem. Phys., 2014, 16, 5922–5926.
53. Y. Wang, F. Wang, Y. Chen, D. Zhang, B. Li, S. Kang, X. Li and L. Cui, Appl. Catal., B, 2014, 147, 602–609.
54. S. Cao, X. Liu, Y. Yuan, Z. Zhang, Y. Liao, J. Fang, S. C. J. Loo, T. C. Sum and C. Xue, Appl. Catal., B, 2014, 147, 940–946.
55. Y. Yuan, S. Cao, Y. Liao, L. Yin and C. Xue, Appl. Catal., B, 2013, 140, 164–168.
56. K. Iizuka, T. Wato, Y. Miseki, K. Saito and A. Kudo, J. Am. Chem. Soc., 2011, 133, 20863–20868.
57. K. Teramura, S. Iguchi, Y. Mizuno, T. Shishido and T. Tanaka, Angew. Chem., Int. Ed., 2012, 51, 8008–8011.
58. C. An, J. Wang, W. Jiang, M. Zhang, X. Ming, S. Wang and Q. Zhang, Nanoscale, 2012, 4, 5646–5650.
59. M. Abou Asi, L. Zhu, C. He, V. K. Sharma, D. Shu, S. Li, J. Yang and Y. Xiong, Catal. Today, 2013, 216, 268–275.
60. Y. Liu, Y. Yang, Q. Sun, Z. Wang, B. Huang, Y. Dai, X. Qin and X. Zhang, ACS Appl. Mater. Interfaces, 2013, 5, 7654–7658.
61. D. Sun, Y. Fu, W. Liu, L. Ye, D. Wang, L. Yang, X. Fu and Z. Li, Chem.–Eur. J., 2013, 19, 14279–14285.
62. Y. Fu, D. Sun, Y. Chen, R. Huang, Z. Ding, X. Fu and Z. Li, Angew. Chem., Int. Ed., 2012, 51, 3364–3367.
63. Q. Liu, Z. Low, L. Li, A. Razmjou, K. Wang, J. Yao and H. Wang, J. Mater. Chem. A, 2013, 1, 11563–11569.
64. K. Maeda, K. Sekizawa and O. Ishitani, Chem. Commun., 2013, 49, 10127–10129.
65. X. An, K. Li and J. Tang, ChemSusChem, 2014, 7, 1086–1093.
66. H. Hsu, I. Shown, H. Wei, Y. Chang, H. Du, Y. Lin, C. Tseng, C. Wang, L. Chen, Y. Lin and K. Chen, Nanoscale, 2013, 5, 262–268.
67. J. Yu, J. Jin, B. Cheng and M. Jaroniec, J. Mater. Chem. A, 2014, 2, 3407–3416.
68. L. Sang, Y. Zhao and C. Burda, Chem. Rev., 2014, 114, 9283–9318.
69. X. Wang, Z. Li, J. Shi and Y. Yu, Chem. Rev., 2014, 114, 9346–9384.
70. K. Lee, A. Mazare and P. Schmuki, Chem. Rev., 2014, 114, 9385–9454.
71. L. Wang and T. Sasaki, Chem. Rev., 2014, 114, 9455–9486.
72. D. Fattakhova-Rohlfing, A. Zaleska and T. Bein, Chem. Rev., 2014, 114, 9487–9558.
73. S. Liu, J. Yu and M. Jaroniec, Chem. Mater., 2011, 23, 4085–4093.
74. F. P. Koffyberg and F. A. Benko, J. Appl. Phys., 1982, 53, 1173–1177.
75. K. Nakaoka, J. Ueyama and K. Ogura, J. Electrochem. Soc., 2004, 151, C661–C665.
76. Z. Yang, G. Huang, W. Huang, J. Wei, X. Yan, Y. Liu, C. Jiao, Z. Wan and A. Pan, J. Mater. Chem. A, 2014, 2, 1750–1756.
77. 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.
78. 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.
79. X. Chen, S. Shen, L. Guo and S. S. Mao, Chem. Rev., 2010, 110, 6503–6570.
80. A. Kubacka, M. Fernandez-Garcia and G. Colon, Chem. Rev., 2012, 112, 1555–1614.
81. N. Yazdanpour and S. Sharifnia, Sol. Energy Mater. Sol. Cells, 2013, 118, 1–8.
82. B. Ohtani, Y. Ogawa and S. Nishimoto, J. Phys. Chem. B, 1997, 101, 3746–3752.
83. M. Marszewski and M. Jaroniec, Langmuir, 2013, 29, 12549–12559.
84. M. S. Hamdy, R. Amrollahi, I. Sinev, B. Mei and G. Mul, J. Am. Chem. Soc., 2014, 136, 594–597.
85. A. Di Paola, M. Bellardita, R. Ceccato, L. Palmisano and F. Parrino, J. Phys. Chem. C, 2009, 113, 15166–15174.
86. J. Wang, G. Ji, Y. Liu, M. A. Gondal and X. Chang, Catal. Commun., 2014, 46, 17–21.
87. F. Sastre, A. Corma and H. Garcia, Angew. Chem., Int. Ed., 2013, 52, 12983–12987.
88. F. Sastre, A. V. Puga, L. Liu, A. Corma and H. Garcia, J. Am. Chem. Soc., 2014, 136, 6798–6801.
89. W. Hou and S. B. Cronin, Adv. Funct. Mater., 2013, 23, 1612–1619.
90. S. Cao and J. Yu, J. Phys. Chem. Lett., 2014, 5, 2101–2107.
91. Y. Wang, X. Wang and M. Antonietti, Angew. Chem., Int. Ed., 2012, 51, 68–89.
92. J. Yu, K. Wang, W. Xiao and B. Cheng, Phys. Chem. Chem. Phys., 2014, 16, 11492–11501.
93. P. Scop, Phys. Rev., 1965, 139, A934–A940.
94. C. Wang, Z. Xie, K. E. deKrafft and W. Lin, J. Am. Chem. Soc., 2011, 133, 13445–13454.
95. H. Sun and S. Wang, Energy Fuels, 2014, 28, 22–36.
96. A. Fujishima and K. Honda, Nature, 1972, 238, 37–38.

---