Photoinduced activation of CO₂ on Ti-based heterogeneous catalysts: Current state, chemical physics-based insights and outlook

Abstract

This article is a review of the current knowledge of the chemical physics of carbon dioxide (CO₂) conversion to fuels using light energy and water (CO₂ photoreduction) on titania (TiO₂)-based catalysts and Ti-species in porous materials. Fairly comprehensive literature reviews of CO₂ photoreduction are available already. However, this article is focused on CO₂ photoreduction on Ti-based catalysts, and incorporates fundamental aspects of CO₂ photoreduction, knowledge from surface science studies of TiO₂ and the surface chemistry of CO₂. Firstly, the current state of development of this field is briefly reviewed, followed by a description of and insights from surface state and surface site approaches. Using examples such as metal-doping of TiO₂, dye-sensitization, oxygen vacancies in TiO₂ and isolated-Ti centers in microporous/mesoporous materials, the utility of these approaches to understand photoinduced reactions involved in CO₂activation is examined. Finally, challenges and prospects for further development of this field are presented. Enhanced understanding of the CO₂ : TiO₂ system, with a combination of computational and experimental studies is required to develop catalysts exhibiting higher activity towards CO₂ photoreduction.

---

Venkata Pradeep Indrakanti

Pradeep Indrakanti will graduate with a PhD in Energy and Geo-Environmental Engineering from The Pennsylvania State University in May 2009. He has received a B.Tech. in Chemical Engineering from the National Institute of Technology, Tiruchirapalli, India. His research interests include the application of industrial ecology and green chemistry principles to clean energy technologies, carbon dioxide chemistry and semiconductor photocatalysis. He has process experience in cement and fertilizer production, and co-authors an energy blog.

James D. Kubicki

Kubicki is Associate Professor of Geosciences at The Pennsylvania State University and affiliated with the Earth and Environmental Systems Institute. He has received a B.S. in Geology from the California State University, Fullerton and PhD in Geochemistry from Yale University. Recent research has focused on mineral surface reactions with particular emphasis on the surface chemistry of TiO2 phases.

Harold H. Schobert

Harold Schobert has been involved in coal research for over thirty years, beginning with a project on making hydrogen sulfide from coal, at Deepsea Ventures Inc. He subsequently was involved in research and development on low-rank coals for the University of North Dakota Energy Research Center and its predecessor organizations, in Grand Forks ND, where he managed the Coal Science Division. He moved to Penn State in 1986, where a major research activity has been the coal-based jet fuel program, now entering commercialization. Another research interest is carbon dioxide chemistry for mineral carbonation and CO2 utilization.

---

Broader context

The conversion of carbon dioxide (CO₂) to produce fuels using light energy has the potential to reduce anthropogenic CO₂ emissions, and lower the consumption of fossil fuels. Although there have been interesting developments in the photoinduced reduction of CO₂ (CO₂ photoreduction) on various Ti-based catalysts, a detailed discussion of the physico-chemical aspects of the light-induced activation of CO₂ has been lacking. This article is a review of current knowledge of the chemical physics of CO₂ photoreduction on titania (TiO₂)-based catalysts and Ti-species in porous materials. Firstly, a comprehensive review of current literature on CO₂ photoreduction on Ti-based materials is presented, and comparisons are made between reduction of CO₂ using solar hydrogen generation and the direct photoreduction of CO₂. Various advances in this field are discussed using the surface site and surface state approaches, and future research needs for improving the efficiency of carbon dioxide conversion with sunlight are outlined.

Introduction

CO₂ utilization

There is a growing need to mitigate CO₂ emissions. Some of the strategies to mitigate CO₂ emissions are energy conservation, carbon capture and storage and using CO₂ as a raw material in chemical processes as well as enhanced oil recovery (EOR) (CO₂ utilization). Catalytic reactions involving CO₂ and producing value-added products are reviewed by Xiaoding and Moulijn.¹ More recent reviews by Song,² Sakakura et al.³ and Aresta and Dibenedetto⁴ detail additional opportunities to use CO₂ as a feedstock. Reactions involving CO₂ typically require energy input and/or a high energy substrate. As pointed out by Song,² processes involving CO₂ could be endergonic (or endothermic) and can still be feasible, provided the products are valuable enough. Moreover, as carbon legislation becomes imminent, a price for CO₂ (either as a carbon tax, or as a CO₂ offset in a cap-and-trade system) would benefit processes having a positive net CO₂ balance.

What is the potential for utilizing CO₂ by the use of various renewable energy technologies? Current global industrial consumption of CO₂ (approximately 115 million metric tons (Mt) year⁻¹)⁵ is insignificant compared to anthropogenic CO₂ emissions (∼26 giga tons (Gt) of CO₂ year⁻¹). However, we note that U.S. CO₂ consumption for enhanced oil recovery as well as chemical synthesis (∼40 Mt year⁻¹)⁶ is similar in scale to other projected means of U.S. greenhouse gas (GHG) reduction (data from a McKinsey report⁷), such as using solar PV electricity, wind energy and cellulosic biofuels, as shown in Fig. 1. With the caveat that most processes utilizing CO₂ would need reducing power and/or a source of energy which is not reflected in the dotted lines shown in Fig. 1, we note that comparatively larger amounts of GHG mitigation could be possible if CO₂ is converted into fuels. This conversion will lead to a reduction in CO₂ emissions only if the underlying energy infrastructure is not based on primary fossil fuels.⁵

Fig. 1 Comparison of the greenhouse gas (GHG) abatement potential (in million metric tonnes of CO₂ equivalents (MtCO₂e) year⁻¹) for selected technology options. The size of the circle approximately denotes the potential for abatement; negative costs represent net-savings from reducing CO₂ emissions. The data (from a McKinsey report, showing the US mid-range abatement potentials and costs) do not include a price for carbon. The estimated potential for CO₂ abatement using nuclear, wind, and cellulosic biofuel technologies is of the same order of magnitude as current CO₂ utilization. Aresta and Dibenedetto⁸ project that direct solar conversion of CO₂ to fuels would mitigate 300–700 Mt CO₂ year⁻¹.

For example: the quantity of U.S. CO₂ emissions mitigated by solar PV technologies (50 Mt CO₂e year⁻¹ by 2030) represents approximately 50% of the CO₂ currently utilized by the global chemical industry. The data shown in Fig. 1 for solar photovoltaic (PV) technologies could be considered as representing the upper limits for converting light energy to electricity. Solar fuel production could be thought of as involving solar PV electricity production, followed by generation of hydrogen or reduction of CO₂ using the electrical energy. If we assume (simplistically) that the efficiency of conversion dictates the potential size of the market, using 10% efficiency for this solar-electricity-chemicals conversion results in a 5 Mt CO₂ year⁻¹ potential for the abatement of greenhouse gas emissions using solar photoreduction (of CO₂ or water).† Given reasonable economies of scale, and conversion efficiencies, the costs of producing fuels from sunlight and CO₂ could be offset by CO₂ emission credits (under a carbon cap-and-trade framework) and the sale of the hydrocarbons produced. On the other hand, Aresta and Dibenedetto⁸ point that global CO₂ utilization to make chemicals and fuels could reach 300–700 Mt CO₂ year⁻¹ globally in the near- to long-term. Therefore, the potential for the indirect utilization of CO₂ by using renewable electricity/hydrogen would be less than the direct solar conversion of CO₂ and water to chemicals or fuels.

Although CO₂ utilization may not make an impact on directly reducing emissions, it may provide a means to limit the use of fossil fuels and thereby contribute indirectly. As noted by Aresta and Dibenedetto,⁹ a life cycle analysis (LCA) approach should be used to determine whether a given process indeed decreases fossil fuel use and the carbon intensity.

CO₂ photoreduction

CO₂ photoreduction refers to the conversion of CO₂ to reduced C₁ and C₂ compounds using light-induced reactions. Since CO₂ does not absorb either visible or UV radiation in the wavelengths 200–900 nm, this process requires suitable photosensitizers. Both metal complexes and semiconductors have been utilized to absorb visible/UV radiation and transfer this energy to CO₂.¹⁰ Additionally, various Ti species in silicate-based micro/mesoporous materials are also active towards CO₂ photoreduction. These include isolated, tetrahedral Ti species substituting framework-Si in mesoporous silicates,^11,12 TiO²⁺ species prepared via ion-exchange in microporous aluminosilicates,¹³ isolated^14,15 and bimetallic¹⁶Zr⁴⁺- and Ti⁴⁺-containing species grafted onto the pore surface of mesoporous materials.

Whereas the semiconductor-mediated reduction of CO₂ involves the generation of electron-hole pairs and their subsequent transfer to CO₂ and a reductant respectively, the metal complexes undergo local charge transfer transitions leading to electron transfer to CO₂. Analogously, the isolated and bimetallic Ti-centers in micro/mesoporous silicates also undergo local ligand-to-metal charge transfer (LMCT) and metal-to-metal charge transfer (MMCT) transitions respectively. Although the metal complex photocatalysts exhibit comparatively higher yields and turnovers towards CO₂ photoreduction, a sacrificial electron donor (typically an amine) is required to regenerate the photocatalyst. On the other hand, a semiconductor which has its conduction and valence bands placed suitably to transfer electrons to CO₂ and oxidize inexpensive reductants such as water, can be designed. Among various semiconductors, TiO₂ is widely used in many photoinduced processes because of its comparatively low cost, low toxicity¹⁷ and its ability to resist photocorrosion.¹⁸ TiO₂ is a wide-band gap semiconductor (E_g ∼ 3.2 eV). It occurs as three polymorphs: anatase, rutile and brookite. Among these, anatase is obtained via “soft” chemical syntheses, and transforms upon heating (to temperatures greater than 700–800 K) to the rutile phase. Surface science aspects of TiO₂ have been described in detail by Diebold.¹⁹Anatase is the more photocatalytically active form of TiO₂, though not for all applications. It also has a larger band gap compared to rutile. However, surface science studies of anatase single crystals have only begun in the past eight years,^20–28 whereas the rutile surfaces have been studied more extensively.^29–33

A semiconductor photocatalyst mediating CO₂reduction and water oxidation needs to absorb light energy, generate electron hole pairs, spatially separate them, transfer them to redox active species across the interface and minimize electron hole recombination. This requires the semiconductor to have its conduction band electrons higher energy compared to the CO₂reduction potential while the holes in the valence band to be able to oxidize water to O₂. A single semiconductor does not usually satisfy these requirements. Moreover, current CO₂ photoreduction catalysts do not perform as effectively as current solar hydrogen generation catalysts. Therefore, to develop better Ti-based CO₂ photoreduction catalysts, one needs to understand interactions between CO₂ on ground and excited state TiO₂ surfaces or Ti-based catalyst sites.

The reduction of CO₂ is an endergonic process requiring a hydrogen source and/or energy. The most abundant sources of energy and hydrogen are sunlight and water, respectively. Solar photocatalytic reduction to produce fuels using water as the hydrogen source thus has the potential to be a means to store intermittent solar energy and to recycle CO₂ while decreasing the use of fossil fuels. It is not clear if the state of the TiO₂ surface is restored during the photoreaction reduction of CO₂ for the photoreduction to be photocatalytic.³⁴ However, in commonality with other photoinduced surface reactions such as the photogeneration of hydrogen from water,³⁵ we can expect that both bulk as well as local phenomena influence the reactivity. Firstly, the location of the frontier orbitals of CO₂ with respect to the valence and conduction bands of TiO₂ determines the feasibility and direction of charge transfer. Secondly, local structure at the surface, involving coordinatively unsaturated atoms as well as surface/bulk defects such as cationic and anionic vacancies and interstitials can also influence reactivity to CO₂.

The organization of this article is as follows: initially, the current state of development in CO₂ photoreduction is compared to other photosynthetic processes such as solar H₂ production. Thereafter, the thermodynamics of CO₂ conversion to useful chemicals is presented. Plausible initial steps of CO₂activation are identified. Subsequently, the CO₂ surface state energy levels and the influence of doping on the work function of TiO₂ are discussed using surface state representation of the CO₂ : TiO₂ interface. Thereafter, the influence of local factors such as surface specificity, anion vacancies and coordination of the Ti atom are discussed using the surface site approach. Using examples such as metal-doping of TiO₂, dye-sensitization, oxygen vacancies in TiO₂ and isolated-Ti centers in microporous/mesoporous materials, the utility of these approaches to understand photoinduced reactions involved in CO₂activation is examined. Some of the outstanding challenges for CO₂ photoreduction on TiO₂ are identified, and are briefly discussed. This review will focus mainly on the CO₂activation at the gas/vapor–TiO₂ interface (Fig. 2).

Fig. 2 Scope of the aspects of CO₂ photoreduction discussed in this review: (a): Surface state models indicating the location of the CO₂⁻/CO₂ energy levels in comparison with the conduction and valence bands of TiO₂, surface site models of CO₂ on Ti-based catalysts, (b): CO₂ adsorption on doped/undoped TiO₂ surfaces, (c): CO₂ interacting with isolated model Ti sites in meso/microporous materials such as titanosilicates.

Current state of CO₂ photoreduction on TiO₂-based catalysts

Titania (TiO₂) based photocatalysts have been used to convert CO₂ to useful compounds, both in gas and aqueous phase photoreactions. The conversion of CO₂ and water to simple C products such as formic acid and formaldehyde using irradiated aqueous suspensions of titania was first demonstrated by Inoue et al.⁵³ Researchers have also used homogeneous metal complexes in solution to reduce CO₂.⁵⁴ A summary of recent literature on CO₂ photoreduction involving isolated Ti-species as well as bulk TiO₂ materials is given in Table 1 and Table 2. Some recent developments in this field have been moved towards rational photocatalyst design, the use of highly active isolated Ti-species in mesoporous and microporous materials,^{11,14,15,36–38,55,56} metal-doping of TiO₂, development of catalysts active at longer wavelengths than can be achieved with commercially available titania, and the elucidation of reaction mechanisms through in situ spectroscopic studies.^14,57 A comparison of the CO₂ conversion rates from various studies is shown in Fig. 3. Although it would be more instructive to compare the conversion efficiencies and quantum yields of various TiO₂-based catalysts, often such data is not readily available. The data in Fig. 3 illustrate typical rates of CO₂ photoreduction, and are dependent upon many variables such as metal doping, CO₂ : H₂O ratios, dispersion of Ti, coordination of Ti species, light intensity, type of lamp used, etc., which are already discussed in detail in the literature (see for example, Usubharatana et al.,¹⁰ and references therein). In references (B)–(C), (F) and (G) in Fig. 3, Anpo et al. used Ti-species in microporous or mesoporous silicates as catalysts whereas the rest of the works cited used doped/undoped TiO₂.

Table 1 Summary of results from recent literature on isolated Ti-species used as photocatalysts mediating CO₂reduction

Catalyst	Reaction system used	T/K and P/Pa	CO2 : H2O mole ratio	Light source used	Notes	References
Pt impregnation of (a) Ti-ion-exchanged zeolites and, (b) framework-Si substituting Ti-MCM-48.	Gas–solid system. Flat-bottomed quartz cell connected to conventional vacuum system (1 × 10^−4 Pa), 88 ml	328 K, pCO2 = 734 Pa	0.2	75 W high-pressure Hg lamp, λ > 280 nm	Pt addition increased selectivity towards CH4 compared to CH3OH, and quenched photoluminescence yields.	Anpo et al.,^13 Yamashita et al.,^36 Anpo et al.^11
Fluorination of Ti-FSM-16	do	323 K	0.2	100 W high-pressure Hg lamp, λ > 250 nm	Fluorination (hydrophobicity) increased selectivity towards CH3OH over CH4.	Ikeue et al.^37
Ti-containing porous silica thin film photocatalyst	do	323 K	0.2	100 W high-pressure Hg lamp, λ > 250 nm	Thin film photocatalyst exhibited high quantum yield (0.28%).	Ikeue et al.^38
Framework-Si substituting Ti-MCM-41	Miniature infrared vacuum cell, 3.4 ml	298 K, P = 0.1 MPa	61.67	266 nm	CO, O2 detected as products of single photon, two-electron transfer. H2O confirmed as a stoichiometric electron donor.	Lin et al.^15
Ti(IV)–Sn(II) bimetallic redox centers in MCM-41	Infrared vacuum cell	—	—	Third harmonic emission of a Nd:YAG laser, 355 nm	Ti^3+ centers formed under 355 nm and visible light irradiation due to Sn(II) → Ti(IV) MMCT.	Lin and Frei^16

---

Table 2 Summary of recent literature on doped-TiO₂ used as photocatalyst mediating CO₂reduction

Catalyst modification	Reaction system used	T, (K) and P, (Pa)	CO2 : H2O	Light source	Notes	References
Cu(II) impregnation of JRC-TiO-4	Gas–solid system. Flat- bottomed quartz cell connected to conventional vacuum system (1.33 × 10^−4 Pa), 60 ml	275 K	0.33	75 W high-pressure Hg lamp, λ > 290 nm	Increased selectivity towards CH3OH over CH4.	Yamashita et al.^39
Cu(I) doping of TiO2, sol–gel synthesis	Slurry photoreactor, 300 ml	∼323 K, P = 0.1 MPa	—	8 W Hg lamp, UV-C, λ = 254 nm	Increased selectivity towards CH3OH over CH4, Cu(I) active species.	Tseng et al.^40
Cu(I) doping of TiO2, sol–gel	Gas–solid continuous optical fiber photoreactor (OFPR), 120 optical fibers	348 K, pCO2 = 0.1 MPa	50	λ = 365 nm	Cu2O clusters influenced CH3OH formation.	Wu et al.^41
Nanoscale Ag-coated TiO2 particles embedded in Nafion membrane	High-pressure optical cell, stacked catalyst films. 2 ml H2O added after irradiation	pCO2 = 13.7 MPa	—	990 W Xe arc source, water filter.	Small TiO2nanoparticles (4–6 nm) embedded in ionomer. CH3OH main product after Ag-coating of TiO2. Catalyst films could be reused and were chemically stable.	Pathak et al.^42
Dye ([Ru(Bpy)3]^2+, perylene diimide based)-sensitized, Pt-promoted, thin and thick film-TiO2 photocatalysts	Glass chamber connected to a vacuum line	298 K, pCO2 = 0.09 MPa	34.5	75 W visible daylight lamp, 150 W UV A lamp	Pt-impregnation resulted in higher CH4 yields under UV irradiation compared to sol–gel synthesized Pt/TiO2. Organic sensitizers decreased catalytic activity under UV light and resulted in CH4 evolution under visible light irradiation.	Ozcan et al.^43
TiO2-coated Pyrex glass pellets, H2, H2O as reductants	Packed bed, circulated photocatalytic reactor	316 K	1, H2 : CO2 = 90	4 near-UV lamps, λ = 365 nm	CH4, CO and C2H6 formed as reaction products.	Lo et al.^44
Cu, Fe substitution in TiO2–SiO2, sol–gel synthesis	Same as above, 216 fibers	348 K	—	150 W high-pressure Hg lamp, λ = 320–500 nm, concentrated sunlight, λ = 500–800 nm	Fe substitution resulted in full visible light absorption. Cu–Fe/TiO2 favors C2H2 formation, Cu–Fe/TiO2–SiO2 favoredCH4 formation (quantum yield: 0.0235%).	Nguyen et al.^45,46
Ruthenium dye (N3)-sensitized Cu–Fe/TiO2catalysts	Same as above, 216 ml	348 K	—	Same as above, I^solar = 20 mW cm^−2, I^lamp = 225 mW cm^−2	Dye sensitization resulted in higher CH4 yields under concentrated sunlight due to stronger absorption bands in the visible spectrum. N3 dye stable after 6 h photoreactions.	Nguyen et al.^47
Mixed phase TiO2 nanocomposite prepared by TiCl4 hydrolysis in the presence of anatase powder	Annular glass reactor, isopropanol hole scavenger, 1280 ml	293–298 K	—	450 W, medium-pressure Hg lamp, UV-visible irradiation	Anatase-rutile nanocomposite exhibited higher CH4 evolution rates compared to Degussa P-25 TiO2 and pure anatase catalysts. Effective charge separation between anatase and rutile phases postulated to explain the higher photoactivity.	Li et al.^48

---

Fig. 3 The highest specific rates of CO₂ photoreduction (µmol CO₂ converted g⁻¹ TiO₂ h⁻¹) obtained using various doped and undoped Ti-based catalysts in selected recent articles. The values are presented here to illustrate typical rates of CO₂ photoreduction, and are dependent upon many variables such as metal doping, CO₂ : H₂O ratios, dispersion of Ti, coordination of Ti species, light intensity, type of lamp used, etc., which are already discussed in detail in the literature (see for example, Usubharatana et al.¹⁰ and references therein). (A): Yamashita et al., (1994),⁴⁹ (B): Anpo et al.,¹³ (C): Yamashita et al.,³⁶ (D): Kaneco et al.,⁵⁰ (E): Kaneco et al.,⁵¹ (F): Ikeue et al.,³⁷ (G): Ikeue et al.,³⁸ (H): Tseng et al.,⁴⁰ (I): Wuet al.,⁴¹ (J): Nguyen et al.^6,52

Comparisons of CO₂ photoreduction rates to solar H₂ generation

Although CO₂ photoreduction has been studied since the 1970s, to the authors' knowledge, there are no studies on its performance vs. alternative approaches. Comparing the rates of CO₂ photoreduction with other photosynthetic processes, such as solar hydrogen generation presents another means to assess the viability of CO₂ photoreduction to limit GHG emissions. To better understand the current state of CO₂ photoreduction using Ti-based catalysts, we converted the specific weights in Fig. 3 to units of micromoles CO₂ converted h⁻¹, using the catalyst weights, and titania weight percentages. Because most of the studies in Fig. 3 used electric lamps as the light source (for UV light), we illustrated the potential for improvements in CO₂ photoreduction as the ratio of the amount of CO₂ converted to fuels to the amount of CO₂ emitted due to the consumption of electrical energy (assuming coal-fired power generation). While we realize that the ultimate goal of CO₂ photoreduction would be to use sunlight to convert CO₂ to fuels, we think that this ratio would be a good metric for comparing future advances in the case of UV light-irradiated systems, because quantitative values for quantum efficiency are difficult to obtain in the case of powder photocatalysts. The results (Fig. 4) indicate that current Ti-based photoinduced CO₂ conversion processes emit more CO₂ than that converted. As a caveat, we note that only minimal efforts were likely made to improve the light absorption in the references cited in Fig. 4. This might be one reason why the overall efficiency (kg CO₂ converted kg⁻¹ CO₂ emitted by lamp) is low. However, significant improvements, both in the specific rate of CO₂reduction as well as the quantum yield of the photoreaction are needed to make this process economically feasible.

Fig. 4 Amount of CO₂ converted to reduced-C species per unit of CO₂ evolved from lamp operation (assuming coal-fired power generation), from recent literature data on CO₂ photoreduction using Ti-based catalysts. Current Ti-based photoinduced CO₂ conversion processes emit more CO₂ than they convert. Significant improvements, both in the specific rate of CO₂reduction as well as the quantum yield of the photoreaction are needed to make this process feasible. A lamp power of 150 W was assumed in Wuet al. (2005).⁴¹ All references except Tseng et al.⁴⁰ used lamp wattages ≥75 W. References are the same as those in Fig. 3.

There are two conceptual routes to producing renewable C-containing fuels using solar energy. The first, the topic of this review, is the direct photoreduction of CO₂ using water as a reductant, whereas the second route involves the photolysis of water to generate hydrogen and further reaction of this hydrogen with carbon dioxide forming C₁–C₂fuels. Both the efficiency of the conversion of solar energy to useful chemical energy as well as the rates of photoreduction of hydrogen/reduced C₁–C₂ compounds are pertinent parameters which determine economic feasibility of such processes. Assuming that the efficiency of solar energy conversion is almost the same in the two routes, the choice between direct and indirect hydrogenation of CO₂ is determined by the rates of photoreduction.

Doped metal oxide photoelectrodes for solar hydrogen generation have typical hydrogen evolution rates of 400–700 µmol h⁻¹ under visible irradiation.⁵⁸ Recently, Grimes et al.⁵⁹ demonstrated that nanostructured TiO₂ photoanodes in photoelectrochemical cells (under voltage bias) resulted in hydrogen generation rates of ∼7000 µmol h⁻¹ under AM 1.5 radiation. On a “sunny” day, using a single unit photoelectrocatalysis system without applied voltage bias, Ichikawa and Doi⁶⁰ found a hydrogen evolution rate of 0.42 l h⁻¹ m⁻² (approximately 4 µmol H₂ h⁻¹ cm⁻²-irradiated photocatalyst area).

On the other hand, typical rates of CO₂ photoreduction in aqueous systems containing suspended metal oxides (TiO₂, BaTiO₃etc.) are much lower, approximately 1 µmol h⁻¹ C₁/C₂ product h⁻¹ under UV radiation.⁵⁵ Using copper-doped TiO₂ and by conducting the photoreduction under a pH bias, Tseng and Wu⁶¹ obtained a slightly higher methanol formation rate of 12 µmol h⁻¹. In gas–solid systems, using isolated Ti-sites in micro/mesoporous materials, Anpo et al.^11–13 obtained CH₄/CH₃OH formation rates of approximately 0.05 µmol h⁻¹. Although the intensity of the light used in these various studies is not the same, the above numbers indicate that current CO₂ photoreduction catalysts do not perform well in comparison with solar-hydrogen generating catalysts, both in terms of product formation rates, and the wavelengths of light needed to sustain reactions. This is because both water oxidation as well as fuels formation from CO₂ are multi-electron transfer reactions. Whereas oxygen evolution viawater oxidation is the major limiting factor in solar hydrogen production, both CO₂ photoreduction as well as water oxidation have to occur simultaneously in a photoreactor using water as the reductant. In this connection, we also note that a recent design for a 100 T year⁻¹ pilot plant producing methanol⁶² from CO₂ involves solar photolysis of water to produce H₂ and subsequent hydrogenation of CO₂ using a heterogeneous catalyst and not the direct photoreduction of CO₂. Additionally, Zeman and Keith propose producing carbon-neutral hydrocarbons from atmospheric CO₂ and renewable hydrogen.⁶³

Although the kinetics of CO₂ photoreduction are two–three orders of magnitude slower than hydrogen photoreduction, one advantage of the direct photoreduction of CO₂ compared to hydrogen evolution is the lower energy needed/electron transferred to convert CO₂ to CH₄ or CH₃OH, Additionally, the use of carbon-based solar-derived fuels is likely to not need significant investments in infrastructure, unlike solar hydrogen generation and further reactions with CO₂. The thermodynamics of CO₂ photoreduction are discussed in the following section.

Thermodynamics and initial steps of CO₂activation and further conversion

“Activation” refers to the transformation of a relatively inert species (such as CO₂, CH₄) to more reactive forms using some form of energy input and/or an activating agent. The activating agent may or may not be regenerated after the process. Because of its significance in steam reforming and methanol production processes, CO₂activation on the surfaces of metals such as Pt, Fe, Nietc. has been studied in detail, both through in situ experiments on well defined surfaces⁶⁴ as well as by computational modeling (see for example, Wang et al.,⁶⁵ and references therein). Metals belonging to the platinum-group are however either expensive or will be in limited supply in the future. Nevertheless, metals play an important role in industrial catalytic processes where CO₂ is either a reactant or a product. For example: Ni-based supported catalysts are used in CO₂reforming/steam reforming of methane. The similarities between homogeneous metal complexes of CO₂ and heterogeneous metal–CO₂ surface species have been noted by Gibson in a review.⁶⁶ As an example, a study by Wovchko and Yates demonstrated the activation of CO₂ by a metal complex (Rh^I(CO)₂) immobilized on an alumina support.⁵⁷ Similar studies should lead to the rational design of photocatalysts mediating CO₂reduction. The surface chemistry of CO₂ on metals and oxide surfaces was reviewed by Freund and Roberts.⁶⁷

CO₂ is one of the most stable compounds of carbon. Gaseous CO₂ is linear and does not have a dipole moment. However, the oxygen atoms, each have a lone pair of electrons and can donate these electrons to surface Lewis acid centers. The carbon atom could also gain electrons from Lewis base centers such as oxide ions, forming carbonate-like species. Additionally, the (C–O) π electrons can also participate in reactions with electron centers. Therefore, reactions of CO₂ involving Lewis acid and base centers and the π orbitals constitute the initial steps of its activation. The simplest reduction products that could be produced from CO₂reduction are CO and HCOO⁻. One, two, four, six and eight electron reduction potentials (vs.NHE) for CO₂reduction and H₂O oxidation at pH 7 and 25 °C assuming unit activities for all gaseous and aqueous species are given in Scheme 1.^7,8

Scheme 1 One, two, six and eight electron reduction potentials (vs.NHE) of some reactions involved in CO₂ photoreduction at pH 7 and unit activity.

From Scheme 1 it is clear that CO₂ photoreduction is not a single-step reaction. Additionally, single electron transfer to CO₂ is highly endergonic, because of the negative adiabatic electron affinity of CO₂.⁶⁷ CO₂activation involves the formation of a negatively charged CO₂^δ˙⁻ species.^67,68 The first plausible step in CO₂ photoreduction on TiO₂ surfaces is therefore its activation to form a CO₂^δ˙⁻ species. CO₂˙⁻_(g) is a 23 electron radical anion. It is significantly distorted from the linear D_∞h geometry due to the repulsion among the two lone electron pairs on the oxygen atoms and the unpaired electron on the carbon atom. Therefore, the lower the O–C–O bond angle, the higher the charge of the CO₂^δ˙⁻ moiety. CO₂^δ˙⁻ can be detected via infra red spectroscopy⁶⁹ as well as electron paramagnetic resonance (EPR) spectroscopy.⁷⁰

The initial step in the photocatalytic reduction of CO₂ is the generation of electron-hole pairs upon absorption of photons of energy greater than or equal to the band gap of the photocatalyst. The time scale of this electron-hole recombination is two to three orders of magnitude faster than other electron transfer processes. Therefore, any process which inhibits electron-hole recombination would greatly increase the efficiency and improve the rates of CO₂ photoreduction. The kinetics of CO₂ photoreduction are also dependent upon many other factors such as incident light intensity, fraction of the incident light absorbed by the photocatalyst, the specific surface area of the photocatalyst absorbing the light, etc. In this article, we examined four such factors: (a) the use of titania-supported noble metals, (b) the use of isolated Ti-tetrahedral centers in micro/mesoporous materials, (c) the use of dye-sensitized TiO₂ to shift the absorption spectrum to the visible region and (d) the role of oxygen vacancies in electron trapping and activating CO₂

Charge transfer from excited TiO₂ surface to CO₂ is influenced by both bulk and local interactions. Two different approaches for understanding electron transfer to CO₂ are using surface sites and surface states. A surface site is a collection of atoms on the surface which is reactive in some manner. It may be a coordinatively under-saturated metal atom, oxygen vacancy or a combination of various other surface features resulting in an orbital having unusual electron affinity. A surface state is a localized energy level at the surface. The location of the surface state with respect to the Fermi energy (E_F) of the solid determines its occupation. The surface site and surface state approaches to CO₂ photoreduction on TiO₂ will be described in detail in the following section.

Surface state description of CO₂activation on TiO₂

The main questions addressed by a surface state approach are:

1. Is electron transfer from the TiO₂ conduction band to CO₂ feasible upon UV illumination?

2. How do surface states (created, for example: by metal doping) affect the reactivity towards CO₂ photoreduction?

A knowledge of surface state energy levels for the CO₂/CO₂˙⁻ couple at the TiO₂ surface is required to determine the feasibility of electron transfer to CO₂ from the conduction band of TiO₂ A methodology to determine the energy levels (of the surface states) at the solid–gas interface has been proposed by Morrison.⁷¹ Accordingly, for the case of a non-interacting redox couple on a solid surface, the location of the standard redox potential in the absolute vacuum scale (AVS) units with respect to the edges of the conduction and valence bands of the solid at the point-of-zero-charge (pzc) determines the potential for electron transfer. Because this article concerns the gas–solid photoreduction of CO₂, the relevant states in question are those at the TiO₂–CO₂ interface. However, it is helpful to compare the energy required to reduce CO₂ in gas–liquid systems to that in gas–solid systems. The location of the energy levels associated with the CO₂/CO₂˙⁻redox couple in various media (gaseous, aqueous solutions) are shown in comparison to the location of the conduction band edge of TiO₂ at pzc in Fig. 5. The standard reduction potential of the CO₂/CO₂˙⁻redox couple vs. the standard hydrogen electrode (SHE) is ∼1.9 V.⁷² On the absolute vacuum scale (AVS), this potential is 2.6 V, which corresponds to a redox state with an energy level at −2.6 eV. The main contributions to this energy are the electron affinity of CO₂ (−0.6 eV) and the solvation of CO₂˙⁻ in water (−3.2 eV). The high free energy of solvation indicates that solvent reorganization energy will be significant. Therefore, a fluctuating energy level mechanism is needed to describe energy levels of CO₂ and CO₂˙⁻. Nevertheless, without any specific interaction with the catalyst/electrode surface, it is clear from Fig. 5 that the reduction of CO₂ on TiO₂ surfaces in aqueous solutions is unlikely because the energy level associated with the CO₂/CO₂˙⁻redox couple is higher in energy compared to the conduction band of TiO₂.

Fig. 5 Location of the conduction band of TiO₂ (rutile) particles at the point-of-zero charge (pzc) with respect to the standard energy states associated with the CO₂/CO₂˙⁻ couple. Photoinduced electron transfer from stoichiometric TiO₂ to CO₂ is not possible because the surface state is not lower than the Fermi energy (E_F) of the electrons in photoexcited TiO₂. The bands are shown at the flat band condition. The exact locations of the Fermi energy levels for electrons and holes (nE_F^* and pE_F^*) are not to scale.

Compared to the solid–liquid interface, measurement of band edges and surface states at the solid–gas interface is relatively more difficult. The exact location of the CO₂˙⁻_(surface)/CO_2(g) surface state on TiO₂ is currently not known. One way to estimate this value is using the an approach analogous to that used by Koppenol.⁷² CO₂˙⁻ adsorption on metal oxide surfaces can be conceptually thought of as the capture of an electron by CO₂ molecule in vacuum/gas-phase forming the anion radical, followed by its adsorption on the metal oxide surface. (In practice, CO₂ adsorbed on a hydrated/hydroxylated metal oxide surface captures a trapped electron either from the defect sites or trapping sites.)

In Scheme 2, EA is the adiabatic electron affinity of gaseous CO₂ (in kJ mol⁻¹) and ΔG_ad⁰ is the heat of chemisorption of CO₂˙⁻ on the surface (in kJ mol⁻¹). F is the Faraday's constant (the charge of a mole of electrons). The overall free energy change of the reaction is ΔG⁰ = [ΔG_ad⁰(CO₂˙⁻) − EA]. In the absolute vacuum scale (AVS), the energy level associated with such a conversion will be E⁰_vac,energy = ΔG⁰/F (eV). In other words, the effect of interactions of CO₂ with the surface is to decrease the energy required for CO₂reduction.

Scheme 2 Thermodynamic cycle for the formation of CO₂˙⁻ at the TiO₂ : CO₂ interface. The surface state energy level of CO₂ can be expressed as a function of its free energy of interaction with the surface and the electron affinity of gaseous CO₂.

Typical values for the heat of chemisorption of CO₂˙⁻ on metal surfaces (Ni) are of the order of ∼50 of kJ mol⁻¹.⁶⁵ If the heats of chemisorption CO₂˙⁻ on TiO₂ surfaces were approximately 100 kJ mol⁻¹ CO₂, the surface state associated with the CO₂/CO₂˙⁻ couple will be lowered approximately 1.3 eV from its gas-phase value, resulting in values of E_energy,AVS⁰(CO_2(g)/CO₂˙⁻_(ad)) of −0.7 eV, which is much more negative than that in aqueous solution. This simplistic analysis is not fully accurate, because the heat of chemisorption (arising out of local interactions of CO₂˙⁻ with the TiO₂ surface) is comparable to the gas-phase electron affinity of CO₂. Moreover, the electrostatic contribution (due to the Madelung potential of TiO₂) is ignored. To understand this further, one needs to study the exact bonding between the CO₂/CO₂˙⁻ couple and the surface. CO₂ is plausibly bonded to the TiO₂ surface differently compared to CO₂˙⁻. Therefore, the so called “Frank–Condon splitting” of surface state energy levels also needs to be taken into account, which requires some description of the interaction of CO₂˙⁻ with the surface atoms. One example of this is a recent study by Bonapasta et al.⁷³ where the energy levels of neutral and electron-attached states of O₂ adsorbed on (100) anatase TiO₂ surface were calculated using a surface molecule description of adsorbates in combination with periodic density functional theory calculations.

Experimentally, such surface states may be identified by spectroscopic studies of CO₂ adsorbed on well-defined rutile TiO₂ surfaces. Using ultraviolet photoemission spectroscopy (UPS) and metastable impact electron spectroscopy (MIES), Krischok et al.⁷⁴ found that CO₂ interacted with rutile TiO₂ single crystals weakly, forming linearly adsorbed species. This is also supported by the results of an ab initio periodic study of CO₂ adsorption on rutile.⁷⁵ CO₂ therefore does not form a surface state upon linear adsorption (involving Lewis acid–base interactions) in the ground-electronic state (without irradiation). Irradiation with light of energy greater than the band gap creates electron-hole pairs in TiO₂, with the electrons populating the conduction band. Using the previous analysis, the energy of conduction band electrons in TiO₂ is likely to be less than that required for CO₂reduction to CO₂˙^δ−.

In conclusion, the above analysis indicates that the major factors controlling the surface state energy levels for CO₂˙⁻_(ad) are the electron affinity of CO₂ (in the gas phase) and the local interaction of CO₂/CO₂˙⁻redox couple with the TiO₂ surface. It is clear that a description of the energy levels of CO₂/CO₂˙⁻ couple on TiO₂ surfaces requires a combination of local and collective approaches. Therefore, we describe the surface site model of CO₂activation on TiO₂ in the next section.

Surface site description of CO₂activation on TiO₂

In recent years, knowledge of surface photocatalytic processes has progressed from a surface state approach to a more detailed understanding of specific surface adsorbate structures which are involved in the charge transfer. An example of this shift has been noted by Nowotny et al.³⁵ Compared to the understanding of the initial steps of H₂O oxidation on TiO₂ surface, not much is known about the initial steps of CO₂activation during the photocatalytic reduction. As noted in the earlier section, to better understand the nature of this charge transfer, one also needs to understand local chemistry at surface sites where activated species are formed. CO₂ can be adsorbed as a carbonate and a linear species on TiO₂ surfaces. Various ground-state adsorbate configurations of CO₂ on TiO₂ surfaces have been studied.^75–78 CO₂ acts as a net donor of electrons when it adsorbs with the oxygen end on Ti Lewis acid sites, and a net acceptor of electrons when the electrophilic C atom interacts with surface electron centers or Lewis basic sites (surface O atoms).⁶⁷ A limiting case of the electrophilic interaction between the C atom of CO₂ and the surface oxygen atoms is the formation of surface carbonate and bicarbonate species. On the other hand, if CO₂ does not interact strongly with TiO₂ in the ground state, the LUMO of adsorbed CO₂ may not be lowered enough to ensure electron transfer from the TiO₂ surface. The primary questions addressed by a surface site approach are:

1. Upon illumination of the surface, how does electron transfer from the TiO₂ conduction band to CO₂ occur?

2. What specific sites and specific CO₂ configurations on the TiO₂ surface promote this electron transfer? Can these sites be engineered?

The mechanisms proposed for CO₂ photoreduction over TiO₂ or Ti-based catalysts assume that CO₂ gains electrons from the conduction band of TiO₂.¹⁰ This indicates a strong interaction between CO₂ and the TiO₂ surface. Photoexcitation of TiO₂ produces electron-hole (Ti³⁺–O⁻) centers in TiO₂. In the next step, these Ti³⁺ centers are proposed to interact with CO₂ forming CO₂^δ˙⁻ species adsorbed on the surface. These carbon dioxide radical anions then undergo further reactions to form C radicals and CH_xO_y end products. Electron transfer from Ti³⁺ surface electron centers to CO₂ requires the presence of suitable unoccupied molecular orbitals on CO₂. Electronic interaction between the surface and CO₂ is necessary for CO₂ to gain electrons. Freund and Roberts⁶⁷ indicate that the lowest unoccupied molecular orbital (LUMO) of CO₂ decreases with the O–C–O bond angle. This variation of the LUMO of CO₂ with the O–C–O bond angle is shown in Fig. 6. One role of the surface might be to interact with CO₂ reducing the bond angle. In the limiting case, this corresponds to the formation of carbonate-like species on the surface. This principle is not limited to CO₂activation on irradiated metal oxides alone. Similar interactions of CO₂ with the electrocatalyst are supposed to decrease the overpotential for CO₂reduction⁷⁹ during electrochemical reduction of CO₂.

Fig. 6 The variation in the energy of the LUMO of gaseous CO₂ with the O–C–O bond angle, calculated using constrained quantum chemical calculations at the B3LYP/6 − 31 + G(d) level of theory. Decreasing the O–C–O bond angle (via surface interactions) may facilitate charge transfer to CO₂ by lowering its LUMO.

Techniques such as Fourier transform infra red (FT-IR) spectroscopy have been used to study CO₂ adsorbed on surface electron centers (Ti³⁺) created due to oxygen vacancies.⁷⁶ CO₂ adsorption on reduced TiO₂ surfaces generated IR signals corresponding to CO₂^δ˙⁻ species.

In the following section, we will examine the use of these two (surface state, surface site) approaches to understand photoinduced reactivity of Ti-based catalysts towards CO₂

Utility of surface state and surface site approaches in understanding the elementary reactions involved in CO₂ photoreduction on Ti-based catalysts

Effects of alkali and noble metal doping on CO₂ photoreduction

Some approaches to make the thermodynamics of CO₂reduction on TiO₂ more favorable are the use of promoters such as potassium (K) or through impregnation with noble metals. Alkali promotion of CO₂activation on TiO₂ has been well studied.³³ The added alkali atom donates an electron to CO₂, forming CO₂˙⁻ species. This reacts with the surface oxygen to form a surface carbonate species. Such promoters are stoichiometrically consumed upon charge transfer to CO₂ because they usually cannot be regenerated (photo)catalytically in an environment containing water vapor and oxygen. On the other hand, the modification of TiO₂ surface with metals such as Pt, Rhetc. may affect both the thermodynamics as well as kinetics of CO₂activation and further conversion.

In a study of photoinduced CO₂ dissociation on titania-supported noble metals, it was found that the activity towards CO dissociation correlated with the work function of the polycrystalline metal for titania-supported Pt, Rh and Ir/TiO₂.⁷⁸ The following mechanism was proposed to account for photoinduced CO₂activation on titania-supported noble metal catalysts: Chemisorption of CO₂ on polycrystalline metal islands results in a bent CO₂^δ˙⁻ species formed by back donation between the metal d-orbitals and the (C–O) π^* orbitals. This electron transfer results in the “depopulation” of the metal surface state. Photogenerated electrons generated upon band gap illumination would be transferred from the conduction band of TiO₂ to the metal island (because the work function of these metals is higher than that of reduced TiO₂). Therefore, illumination of TiO₂-supported noble metal catalysts in a CO₂ atmosphere increased the intensity of signals corresponding to CO₂^δ˙⁻ species. We note that such activation reactions require a concomitant oxidation reaction to be photocatalytic. Additionally, the study by Raskó⁷⁸ demonstrates that an understanding of both surface states as well as surface sites is needed to understand CO₂activation on such TiO₂-supported metal photocatalysts. The case of Rh/TiO₂ is especially interesting, from a catalytic viewpoint if not from a practical perspective, rhodium being one of the most expensive metals in the world currently.

Recently, CO oxidation on gold clusters supported on TiO₂ has been studied extensively, both by experimental⁸⁰ and modeling^81,82 studies. We note that CO oxidation also involves the formation of a CO₂^δ˙⁻ intermediate. Therefore, a good CO oxidation catalyst may also promote CO₂activation. Most studies have been performed on metallic clusters of gold on TiO₂. In these cases, density functional theory (DFT) calculations have shown that the binding strength of low-coordinated Au clusters with gaseous CO is a crucial parameter in the oxidation of CO.⁸³ Moreover, promotion effects may also contribute to the observed low temperature oxidation of CO to CO₂ in the presence of gaseous oxygen. In contrast, a Mars–van Krevelen mechanism was proposed by Chrétien and Metiu⁸⁴ to account for the activity of gold doped TiO₂, where the gold atom is positively charged and substitutes a five coordinate Ti atom on the rutile (110) surface.

In addition to activating CO₂, noble metal impregnation may also decrease the photogenerated electron-hole recombination rates, leading to higher product yields. Using femtosecond time resolved measurements of the intensity dependence of electron hole recombination in TiO₂ nanoclusters, Colombo and Bowman⁸⁵ found that more than half of the conduction band electrons underwent recombination within 20 picoseconds (ps). The remaining electrons were trapped at surface sites resulting in a longer lived species. Therefore, from an overall system perspective, the greatest gain in efficiency in CO₂ photoreduction may result by slowing down the rates of electron-hole recombination. Using surface photovoltage spectroscopy (SPS) measurements on Pd and Pd/RuO₂ doped TiO₂, Wang et al.⁸⁶ showed that the rates of CO₂ photoreduction to formate in aqueous solutions correlated with the surface photovoltage measured in air. Two pertinent conclusions could be drawn from their study. Firstly, surface modification of TiO₂ with noble metals could lead to the formation of surface states, increasing the photogenerated electron-hole pair lifetimes. Secondly, the fact that photovoltages observed in air correlated with the yields in aqueous solutions indicates that electron-hole pair separation appears to be the primary control over the reactivity.

Visible light performance of dye-sensitized TiO₂catalysts mediating CO₂ photoreduction

Another means to decrease electron-hole recombination rates and use higher wavelengths of the solar spectrum is the dye-sensitization of TiO₂. Closely related to this is the dye sensitized solar cell (DSSC), (see McConnell⁸⁷ and references therein for details about operation and comparisons). Briefly, the DSSC is a solar electric device where the light absorption and charge transport are physically separated. The organometallic dye attached to porous TiO₂ absorbs sunlight, and injects electrons into the conduction band of TiO₂, which flow through an external circuit, creating electricity. The oxidized dye is reduced by a mediator which in turn, is reduced at the metallic cathode of the solar cell. This concept has been applied to the photosensitization of TiO₂ anodes in photoelectrochemical (PEC) cells. A brief description of the operation of a dye-sensitized photoanode is provided by Bak et al.⁸⁸ However, in the case of dye-sensitized TiO₂ used for photocatalytic CO₂reduction applications,^43,47 it is less clear how the oxidized dye would be regenerated when the photoelectrochemical cell is short-circuited and no defined pathway for electron transport exists (RERITE). Sacrificial photooxidation of the dye or other reagents probably occurs unless the oxidized state of dye has a surface state level (Dye_(surf)⁺/Dye_(surf)) lower than that for water oxidation (O_2(surf)/H₂O_(surf)).

Nevertheless, Nguyen et al.⁴⁷ found substantial improvements in the photoactivity of (Cu,Fe)-TiO₂ photocatalysts towards CO₂ photoreduction in the presence of water and concentrated sunlight in a optical fiber photoreactor (OFPR). The higher activity was attributed to full visible light absorption, whereas TiO₂ only absorbs a fraction of the radiation above 400 nm. Only CH₄ was produced under sunlight, and the photocatalyst was reported to be stable for 6 h. Whereas dye-sensitization of TiO₂ enables the use of visible light, we note that future studies should consider ¹³CO₂-isotope testing to ensure that the reduced C-products are due to CO₂reduction, and not the sacrificial oxidation of the dye.

To summarize, both metal doping as well as dye-sensitization reduce the rates of electron-hole recombination, but the mechanisms involved are different. In the case of metal-doped TiO₂, the electrons are transferred from the conduction band of TiO₂ to the metal, whereas on a dye-sensitized photocatalyst, the dye injects electrons into the conduction band of TiO₂.

Isolated Ti-species acting as active sites for CO₂ photoreduction in micro/mesoporous materials

As discussed in Table 1, isolated tetrahedral Ti⁺⁴ species in zeolites or mesoporous sieves were used as catalysts by Anpo and coworkers^13,36,37,38 as well as Frei et al.^14,15 The location of the Ti atoms in mesoporous silica depend on the synthetic technique used.⁸⁹ Framework Ti-substituted MCM-41 is prepared when the Ti precursors are included in the MCM-41 synthesis gel (Lin et al.¹⁵). On the other hand, Ti-grafted MCM-41 is prepared by post-synthesis grafting (ex: Lin and Frei¹⁶). In such Ti-grafted MCM-41, the Ti atoms are located at the surface of the silicate wall and are not a part of the framework. They are therefore more accessible to reactants. However, this might also make them more susceptible to leaching in aqueous solutions. Moreover, depending on the titanium content, a variety of configurations for Ti species are possible.⁹⁰

Using in situX-ray fluorescence spectroscopy (XAFS) studies to probe the local environment around an isolated Ti atom, Yamashita et al.⁵⁶ studied the effects of the hydrophobic–hydrophilic properties of Ti-β zeolites in the photocatalytic reduction of CO₂ with H₂O. Adsorption of CO₂ did not change the XANES spectra of Ti-β zeolite catalysts. On the other hand, the addition of water increased the coordination number of Ti atom from 4 to 6. Therefore, in the ground state, H₂O interacted more strongly with Ti-β zeolites compared to CO₂. In situphotoluminescence (PL) spectroscopy measurements on irradiated Ti-β zeolites indicated that the lifetime of the charge transfer excited state decreased in the presence of CO₂ and/or H₂O molecules. Therefore, while CO₂ interacted with the excited state of the Ti-β zeolite, water interacted with the Ti-centers in both excited and ground electronic states. Additionally, using in situFT-IR spectroscopy to study the initial steps of CO₂ photoreduction with methanol as a sacrificial oxidant in a Ti silicalite molecular sieve, Ulagappan and Frei⁹¹ recorded the formation of formic acid as the primary electron transfer product. They proposed that the initial step in this reaction involved electron transfer from photoexcited LMCT excited state of the catalyst to CO₂ concurrent with methanol oxidation to form CO₂⁻ and CH₃OH⁺ species which react further to produce formic acid and formaldehyde.

Although the above isolated tetrahedral Ti-centers in zeolites/mesoporous materials are active towards CO₂ photoreduction, the absorption of these materials lies in the deep-UV (UV-C) region of the electromagnetic spectrum. To activate CO₂ using visible light, Frei's research group demonstrated metal–metal charge transfer (MMCT), involving two different metal ions grafted in a mesoporous material.¹⁶ Examples include Ti–O–Cu(I) and Ti–O–Sn(II) as well as Zr–O–Cu(I) grafted in Mobile crystalline material (MCM)-41 molecular sieve. Irradiation of the binuclear Zr–O–Cu(I)/MCM-41 catalyst under a CO₂ atmosphere resulted in the growth of IR signals corresponding to CO and H₂O.⁹² Only Zr–O–Cu(I) complexes showed CO formation. CO₂ dissociation was not observed on Ti–O–Cu(I) complexes. Lin and Frei attributed this to the higher reducing power of the Zr³⁺ transient ion. Similar studies using various other redox couples (ex: Ti⁴⁺/Fe²⁺, Ti⁴⁺/Mn²⁺) could reveal visible light-absorbing catalysts active towards CO₂ photoreduction.

Although the primary charge transfer mechanism in isolated-Ti species is localized, overall reactivity is also affected by the bulk properties. For example, Yamashita et al.³⁶ attributed higher C₁ photoproduct yields using Ti-MCM-48 compared to Ti-MCM-41 or TS-1 to the large pore size and the three-dimensional channel structure of the MCM-48 catalyst. Additionally, electrostatic fields in the zeolite/mesporous material could stabilize the CO₂˙⁻ radical ion and thereby affect the product yield and selectivities.⁹³

Effect of oxygen vacancies created by cation doping or thermal treatment of TiO₂ on the activity towards CO₂

As mentioned earlier, CO₂ does not interact with stoichiometric rutile TiO₂ surfaces. The band gaps of rutile and anatase are not widely different; therefore, this might also be expected to apply to anatase TiO₂. Defects such as oxygen vacancies control most of the chemistry at many metal oxide surfaces. These defects could be created either by doping TiO₂ with other anions or cations or by thermal treatments of stoichiometric TiO₂.

One example of this defect creation due to thermal treatments is the work of Yamashita et al.,⁴⁹ who studied the photoreduction of CO₂ with H₂O on rutile TiO₂ (100) and TiO₂ (110) single crystals. Structure-dependent reactivity was observed, with the TiO₂ (100) surface being more active towards CO₂ photoreduction compared to the TiO₂ (110) surface. Yamashita et al.⁴⁹ explained this by stating that the Ti : O atomic ratio on the (100) surface was greater than the (110) surface. A pretreatment procedure involving degassing at 725 K, followed by a 4 h reoxidation in O₂ at 725 K, followed by subsequent catalyst degassing at 725 K for 4 h at ∼1 × 10⁻⁶ Torr was used in the study. We note that this pretreatment procedure might create oxygen vacancies on the TiO₂ surface. For example: Thompson et al.⁹⁴ note that surface oxygen vacancies could be created with annealing temperatures as low as 600 K with rutile TiO₂ under ultra high vacuum (UHV) conditions. Additionally, stoichiometric reactions between CO₂ and water on oxygen-deficient rutile surfaces to yield reduced C₁–C₂ products are well known in the literature.⁹⁵ In such cases, in situ surface characterization of the catalyst under light irradiation would provide information on the oxidation state of the surface, because the surface-Ti⁴⁺ and Ti³⁺ can be differentiated using XPS 2p core-level spectra.⁹⁶

In addition to creating oxygen vacancies, transition metal doping of TiO₂ also may create specific surface structures affecting the CO₂ photoreduction activity. Tseng et al.⁴⁰ and Wu and Lin⁹⁷ found that Cu(I) doped TiO₂ promoted the formation of CH₃OH compared to undoped TiO₂. Pathak et al.⁴² also observed selectivity for CH₃OH formation over HCOOH using Ag-coated nanoscale TiO₂ immobilized in Nafion ionomer membranes. Apart from influencing the electron-hole recombination rates (as noted in the discussion on surface states) and affecting product specificity and yields, transition metal doping of TiO₂ also results in red-shifting of the absorption spectrum of TiO₂. This apparent bandgap-narrowing is attributed to transitions from localized surface states above the conduction band of TiO₂ and not due to bulk band-gap modification.⁹⁸

Discussion

One of the major factors limiting CO₂ photoreduction is the electron-hole recombination in semiconducting materials under band-gap illumination. The use of isolated Ti-species partly resolves this issue, but most Ti–O LMCT transitions occur at UV-C wavelengths (∼270 nm). Both local as well as bulk factors influence the activity of catalysts mediating CO₂ photoreduction. For example: cation doping of TiO₂ not only creates oxygen vacancies, but also creates localized intra-bandgap states which contribute to charge transfer transitions by absorbing light wavelengths below the bandgap.

Both collective and local approaches are required to understand such light-induced electron transfer reactions. The collective, surface state approach indicates that pure TiO₂ is likely to be a poor photocatalyst for CO₂activation. On the other hand, localized surface sites such as oxygen vacancies have been implicated in CO₂activation.

Using the surface state and surface site approaches, we have shown that both bulk and local variables affect the photoreduction of CO₂. Fundamental understanding of the initial steps of CO₂activation on TiO₂ surfaces and a detailed knowledge of how these bulk and local phenomena affect photoreactivity is required to design better catalysts.

Challenges for CO₂ photoreduction on Ti-based catalysts

As discussed in the earlier sections, significant improvements in the quantum yields and selectivities of Ti-based catalysts is required to bring CO₂ photoreduction using these materials closer to commercialization. Materials not based on Ti have been used to mediate CO₂ photoreduction. For example, metal chalcogenides (CdS, CdSe, ZnS) and other metal oxides (ZnO, MgO) have also been used as photocatalysts (see Usubharatana et al.¹⁰ and references therein). In most cases, the improvements in efficiency are incremental at best. On the other hand, higher quantum efficiencies are possible with homogeneous metal complexes, but a sacrificial reagent is often required to regenerate the photosensitizer.⁵⁴ Breakthroughs in related fields such as solar-electric conversion, solar production of hydrogen using water and catalytic CO₂activation will enable the further development of this field.

Essentially, the main challenges for CO₂ photoreduction using Ti-based catalysts involve finding suitable photosensitizers that can absorb visible light to produce electron-hole pairs and transfer the electrons to CO₂ with minimal losses. Additionally, the photosensitizer has to be regenerated with a relatively inexpensive reductant such as water. Choi et al.⁹⁹ noted that interfacial electron/hole transfer involves charged pair generation, followed “trapping” of the charged particles, their subsequent release and migration to the surface prior to the transfer. Electron-hole recombination occurs parallel to the interfacial charge transfer, at all stages of this process. As pointed out in a report on solar energy utilization,¹⁰⁰ potential breakthroughs in CO₂ photoreduction can result from significantly increasing the lifetime of the charge-separated state, designing materials which can work under visible light, catalyzing proton-coupled multi-electron transfer to CO₂. This requires a combination of experimental and computational studies

Conclusions

In this article we provided a brief review of and insights into the mechanistic aspects of CO₂ photoreduction using TiO₂-based metal oxides or Ti-species in micro/mesoporous materials. CO₂ utilization has a significant role in mitigating CO₂ emissions, especially in comparison to other means of CO₂ abatement. The conversion of CO₂ to fuels using solar energy is one of the “grand challenges” for a variety of scientific disciplines. Significantly increased specific rates (∼10 s of millimoles CO₂ converted g⁻¹ TiO₂ h⁻¹) under visible light irradiation are required to make this process economically feasible. Paradigm shifts in TiO₂ photocatalysis such as the use of bimetallic redox species in mesoporous materials, novel reactor design configurations (OFPR), the use of catalysts absorbing visible radiation, decreasing electron-hole recombination rates, surface modifications via doping or other chemical treatments, will continue to be required to drive innovation in this field.

Acknowledgements

VPI thanks Penn State Institutes of Energy and the Environment (PSIEE) for financial support. Computational support was in part provided by the National Science Foundation under grant no. CHE-0431328 for the Center for Environmental Kinetics Analysis (CEKA) at The Pennsylvania State University. VPI also thanks Dr Mercedes Maroto-Valer for her comments on a section of this review.

References

1. X. Xu and J. A. Moulijn, Energy Fuels, 1996, 10, 305–325.
2. C. S. Song, Catal. Today, 2006, 115, 2–32.
3. T. Sakakura, J. C. Choi and H. Yasuda, Chem. Rev., 2007, 107, 2365–2387.
4. M. Aresta and A. Dibenedetto, Dalton Trans., 2007, 2975–2992.
5. M. Mazzotti, J. C. Abanades, R. Allam, K. S. Lackner, F. Meunier, E. Rubin, J. C. Sanchez, K. Yogo and R. Zevenhoven, in IPCC 2005: IPCC Special Report on Carbon dioxide Capture and Storage. Prepared by Working Group III of the Intergovernmental Panel on Climate Change., ed. B. Metz, O. Davidson, H. C. de Coninck, M. Loos and L. A. Meyer, Cambridge University Press, Cambridge, United Kingdom and New York, USA, 2005, p. 333.
6. H. J. Herzog and E. M. Drake, Annu. Rev. Energy Environ., 1996, 21, 145–166.
7. McKinsey, Reducing U.S. Greenhouse Gas Emissions: How Much at What Cost?http://www.mckinsey.com/clientservice/ccsi/pdf/US_ghg_final_report.pdf, accessed 11/24/2008.
8. M. Aresta and A. Dibenedetto, Catal. Today, 2004, 98, 455.
9. M. Aresta, E. Quaranta and I. Tommasi, Photochem. Convers. Storage Sol. Energy, Proc. Int. Conf., 8th, 1991, 517–550.
10. P. Usubharatana, D. McMartin, A. Veawab and P. Tontiwachwuthikul, Ind. Eng. Chem. Res., 2006, 45, 2558–2568.
11. M. Anpo, H. Yamashita, K. Ikeue, Y. Fujii, S. G. Zhang, Y. Ichihashi, D. R. Park, Y. Suzuki, K. Koyano and T. Tatsumi, Catal. Today, 1998, 44, 327–332.
12. S. G. Zhang, Y. Fujii, K. Yamashita, K. Koyano, T. Tatsumi and M. Anpo, Chem. Lett., 1997, 659–660.
13. M. Anpo, H. Yamashita, Y. Ichihashi, Y. Fujii and M. Honda, J. Phys. Chem. B, 1997, 101, 2632–2636.
14. W. Lin and H. Frei, Stud. Surf. Sci. Catal., 2004, 153, 283–288.
15. W. Y. Lin, H. X. Han and H. Frei, J. Phys. Chem. B, 2004, 108, 18269–18273.
16. W. Lin and H. Frei, C. R. Chim., 2006, 9, 207–213.
17. O. Carp, C. L. Huisman and A. Reller, Prog. Solid State Chem., 2004, 32, 33–177.
18. Y. Xu and M. A. A. Schoonen, Am. Mineral., 2000, 85, 543–556.
19. U. Diebold, Surf. Sci. Rep., 2003, 48, 53–229.
20. X.-Q. Gong, A. Selloni, M. Batzill and U. Diebold, Nat. Mater., 2006, 5, 665.
21. X. Q. Gong, A. Selloni, O. Dulub, P. Jacobson and U. Diebold, J. Am. Chem. Soc., 2008, 130, 370–381.
22. X. Q. Gong, A. Selloni and A. Vittadini, J. Phys. Chem. B, 2006, 110, 2804–2811.
23. X. Q. Gong and A. Selloni, J. Phys. Chem. B, 2005, 109, 19560–19562.
24. U. Diebold, N. Ruzycki, G. S. Herman and A. Selloni, Catal. Today, 2003, 85, 93–100.
25. N. Ruzycki, G. S. Herman, L. A. Boatner and U. Diebold, Surf. Sci., 2003, 529, L239–L244.
26. M. Lazzeri and A. Selloni, Phys. Rev. Lett., 2001, 87.
27. Y. Liang, S. P. Gan, S. A. Chambers and E. I. Altman, Phys. Rev. B, 2001, 63.
28. R. Hengerer, B. Bolliger, M. Erbudak and M. Gratzel, Surf. Sci., 2000, 460, 162–169.
29. W. Gopel, G. Rocker and R. Feierabend, Phys. Rev. B: Condens. Matter Mater. Phys., 1983, 28, 3427–3438.
30. J. Nerlov, S. V. Christensen, S. Weichel, E. H. Pedersen and P. J. Moller, Surf. Sci., 1997, 371, 321–336.
31. A. Gutiérrez-Sosa, J. F. Walsh, R. Lindsay, P. L. Wincott and G. Thornton, Surf. Sci., 1999, 433–435, 538.
32. S. Krischok, O. Höfft and V. Kempter, Surf. Sci., 2002, 507–510, 69.
33. M. Brause and V. Kempter, Surf. Sci., 2001, 476, 78.
34. A. V. Emeline, G. V. Kataeva, A. V. Panasuk, V. K. Ryabchuk, N. V. Sheremetyeva and N. Serpone, J. Phys. Chem. B, 2005, 109, 5175–5185.
35. J. Nowotny, T. Bak, M. K. Nowotny and L. R. Sheppard, Int. J. Hydrogen Energy, 2007, 32, 2651–2659.
36. H. Yamashita, Y. Fujii, Y. Ichihashi, S. G. Zhang, K. Ikeue, D. R. Park, K. Koyano, T. Tatsumi and M. Anpo, Catal. Today, 1998, 45(1–4), 221.
37. K. Ikeue, H. Yamashita, M. Anpo and T. Takewaki, J. Phys. Chem. B, 2001, 105, 8350–8355.
38. K. Ikeue, S. Nozaki, M. Ogawa and M. Anpo, Cata. Lett., 2002, 80, 111.
39. H. Yamashita, H. Nishiguchi, N. Kamada, M. Anpo, Y. Teraoka, H. Hatano, S. Ehara, K. Kikui, L. Palmisano, A. Sclafani, M. Schiavello and M. A. Fox, Res. Chem. Intermed., 1994, 20, 815.
40. I. H. Tseng, W.-C. Chang and J. C. S. Wu, Appl. Catal., B, 2002, 37, 37–48.
41. J. C. S. Wu, H.-M. Lin and C.-L. Lai, Appl. Catal., A, 2005, 296, 194.
42. P. Pathak, M. J. Meziani, L. Castillo and Y.-P. Sun, Green Chem., 2005, 7, 667–670.
43. O. Ozcan, F. Yukruk, E. Akkaya and D. Uner, Top. Catal., 2007, 44, 523.
44. C.-C. Lo, C.-H. Hung, C.-S. Yuan and J.-F. Wu, Sol. Energy Mater. Sol. Cells, 2007, 91, 1765–1774.
45. T.-V. Nguyen and J. C. S. Wu, Appl. Catal., A, 2008, 335, 112–120.
46. T.-V. Nguyen and J. C. S. Wu, Sol. Energy Mater. Sol. Cells, 2008, 92, 864–872.
47. T.-V. Nguyen, J. C. S. Wu and C.-H. Chiou, Catal. Commun., 2008, 9, 2073–2076.
48. G. Li, S. Ciston, Z. V. Saponjic, L. Chen, N. M. Dimitrijevic, T. Rajh and K. A. Gray, J. Catal., 2008, 253, 105–110.
49. H. Yamashita, N. Kamada, H. He, K. Tanaka, S. Ehara and M. Anpo, Chem. Lett., 1994, 855–858.
50. S. Kaneco, Y. Shimizu, K. Ohta and T. Mizuno, J. Photochem. Photobiol., A, 1998, 115, 223.
51. S. Kaneco, H. Kurimoto, Y. Shimizu, K. Ohta and T. Mizuno, Energy, 1999, 24, 21.
52. T.-V. Nguyen and J. C. S. Wu, Sol. Energy Mater., 2008, 92, 864.
53. T. Inoue, A. Fujishima, S. Konishi and K. Honda, Nature (London, United Kingdom), 1979, 277, 637–638.
54. E. Fujita, Coord. Chem. Rev., 1999, 186, 373–384.
55. M. Anpo and H. Yamashita, in Hetergeneous Photocatalysis, ed. M. Schiavello, John Wiley & Sons, Chichester, UK, 1997, p. 136.
56. H. Yamashita, K. Ikeue, T. Takewaki and M. Anpo, Top. Catal., 2002, 18, 95.
57. E. A. Wovchko and J. T. Yates Jr., J. Am. Chem. Soc., 1998, 120, 7544–7550.
58. V. M. Aroutiounian, V. M. Arakelyan and G. E. Shahnazaryan, Sol. Energy, 2005, 78, 581–592.
59. M. Paulose, G. K. Mor, O. K. Varghese, K. Shankar and C. A. Grimes, J. Photochem. Photobiol., A, 2006, 178, 8.
60. S. Ichikawa and R. Doi, Catal. Today, 1996, 27, 271.
61. I. H. Tseng and J. C. S. Wu, Catal. Today, 2004, 97, 113–119.
62. MCI, Mitsui Chemicals to Establish a Pilot Facility to Study a Methanol Synthesis Process from CO2, http://www.mitsuichem.com/release/2008/080825e.htm, accessed 01/09/2008.
63. F. S. Zeman and D. W. Keith, Philos. Trans. R. Soc. London, Ser. A, 2008, 366(1882), 3901–3918.
64. F. Solymosi, J. Mol. Catal., 1991, 65, 337.
65. S. G. Wang, D. B. Cao, Y. W. Li, J. G. Wang and H. J. Jiao, J. Phys. Chem. B, 2005, 109, 18956–18963.
66. D. H. Gibson, Coord. Chem. Rev., 1999, 185–186, 335–355.
67. H. J. Freund and M. W. Roberts, Surf. Sci. Rep., 1996, 25, 227–273.
68. J. Rasko and F. Solymosi, J. Phys. Chem., 1994, 98, 7147–7152.
69. K. O. Hartman and I. C. Hisatsune, J. Chem. Phys., 1966, 44, 1913–1918.
70. L. L. Perissinotti, M. A. Brusa and M. A. Grela, Langmuir, 2001, 17, 8422–8427.
71. R. S. Morrison, in The Chemical Physics of Surfaces, 1990.
72. W. H. Koppenol and J. D. Rush, J. Phys. Chem., 1987, 91, 4429–4430.
73. A. A. Bonapasta and F. Filippone, Surf. Sci., 2005, 577, 59–68.
74. S. Krischok, O. Hofft and V. Kempter, Surf. Sci., 2002, 507–510, 69–73.
75. A. Markovits, A. Fahmi and C. Minot, Theochem, 1996, 371, 219–235.
76. G. Busca and V. Lorenzelli, Mater. Chem., 1982, 7, 89–126.
77. G. Ramis, G. Busca and V. Lorenzelli, Mater. Chem. Phys., 1991, 29, 425–435.
78. J. Raskó, Catal. Lett., 1998, 56, 11–15.
79. M. Schmidt, in Carbon dioxide chemistry: Environmental issues, ed. J. P. Paul, Claire-Marie, RSC, Cambridge, 1994, pp. 22–30.
80. M. S. Chen and D. W. Goodman, Top. Catal., 2007, 44, 41–47.
81. R. G. S. Pala and F. Liu, J. Chem. Phys., 2006, 125, 144714.
82. S. P. Raj Ganesh and L. Feng, J. Chem. Phys., 2006, 125, 144714.
83. I. N. Remediakis, N. Lopez and J. K. Nørskov, Angew. Chem., Int. Ed., 2005, 44, 1824–1826.
84. S. Chrétien and H. Metiu, Catal. Lett., 2006, 107, 143.
85. D. P. Colombo, Jr. and R. M. Bowman, J. Phys. Chem., 1996, 100, 18445–18449.
86. T.-f. Xie, D.-j. Wang, L.-j. Zhu, T.-j. Li and Y.-j. Xu, Mater. Chem. Phys., 2001, 70, 103.
87. R. D. McConnell, Renewable and Sustainable Energy Reviews, 2002, 6, 271.
88. T. Bak, J. Nowotny, M. Rekas and C. C. Sorrell, Int. J. Hydrogen Energy, 2002, 27, 991.
89. L. Y. Chen, G. K. Chuah and S. Jaenicke, Catal. Lett., 1998, 50, 107–114.
90. Q. Yuan, A. Hagen and F. Roessner, Appl. Catal., A, 2006, 303, 81.
91. N. Ulagappan and H. Frei, J. Phys. Chem. A, 2000, 104, 7834–7839.
92. W. Y. Lin and H. Frei, J. Am. Chem. Soc., 2005, 127, 1610–1611.
93. K. Sogabe, A. Hasegawa, Y. Yamada and M. Miura, Bull. Chem. Soc. Jpn., 1972, 45(11), 3362–3366.
94. T. L. Thompson, O. Diwald and J. T. Yates Jr., J. Phys. Chem. B, 2003, 107, 11700–11704.
95. J. N. Wilson, S. D. Senanayake and H. Idriss, Surf. Sci., 2004, 562, L231–L237.
96. P. M. Kumar, S. Badrinarayanan and M. Sastry, Thin Solid Films, 2000, 358, 122.
97. J. C. S. Wu and H.-M. Lin, International Journal of Photoenergy, 2005, 7, 115–119.
98. N. Serpone, J. Phys. Chem. B, 2006, 110, 24287–24293.
99. W. Y. Choi, A. Termin and M. R. Hoffmann, J. Phys. Chem., 1994, 98, 13669–13679.
100. N. Lewis and G. Crabtree, Report on the Basic Energy Sciences Workshop on Solar Energy Utilization, http://www.sc.doe.gov/bes/reports/files/SEU_rpt_print.pdf, accessed 11/30/2008.

---

Footnote

† If we assume that the solar PV electricity displaces coal-fired power generation, a reduction of 50 Mt CO₂ year⁻¹ would imply a solar PV generation of 50 GWh(electric) year⁻¹. Given the energy required for methanol synthesis (∼702 kJ mole⁻¹ CO₂), with a 10% efficiency for electricity-to-chemical energy conversion, we estimate the potential for CO₂ conversion to methanol using solar PV electricity to be ∼1.1 Mt CO₂ year⁻¹.

---