The Rational Design of a Single‐Component Photocatalyst for Gas‐Phase CO2 Reduction Using Both UV and Visible Light

The solar‐to‐chemical energy conversion of greenhouse gas CO2 into carbon‐based fuels is a very important research challenge, with implications for both climate change and energy security. Herein, the key attributes of hydroxides and oxygen vacancies are experimentally identified in non‐stoichiometric indium oxide nanoparticles, In2O3‐x(OH)y, that function in concert to reduce CO2 to CO under simulated solar irradiation.


DOI: 10.1002/advs.201400013
successful gas-phase photocatalysts -particularly those active in the visible region of the solar spectrum -suggesting that new approaches to materials discovery are necessary. [ 13 ] A class of materials capable of photocatalytically reducing CO 2 are oxygen defi cient metal oxides. Oxygen vacancies can function as active catalytic sites and enhance both the absorption of visible light and the photocatalytic activity of the material. [ 14,15 ] The most notable example of this is black titania, TiO 2-x H x , which exhibits a substantial increase in light absorption and photoactivity for water splitting after hydrogen treatment. [ 16,17 ] Another effective approach to increasing the photocatalytic activity of metal oxide nanomaterials is to improve the CO 2 capture capacity of the nanoparticle surface. Several groups have demonstrated that surface hydroxides can enhance the affi nity of CO 2 for a photocatalytic surface, which can have a signifi cant effect on the photocatalytic activity and CO 2 reduction rates. [18][19][20] Clearly, the surface, optical, and electronic properties of metal oxide nanoparticles must work in concert for photocatalytic reduction of CO 2 to occur; understanding this relationship is critical for the advancement towards a practical global scale solar fuels technology. [ 13,17,19,[21][22][23] Indium oxide is a material with surface, optical, and electronic properties that make it a compelling choice as a CO 2 reduction photocatalyst. For example, its conduction band (CB) and valence band (VB) positions on an energy band diagram straddle the H 2 O oxidation and CO 2 reduction half reaction energies required to drive artifi cial photosynthetic production of hydrocarbons and carbon monoxide. [ 4,24 ] Furthermore, In 2 O 3 has a direct "forbidden" band gap where the lowest-energy optical transition from the top of its VB to the bottom of its CB and vice-versa is forbidden by symmetry. [ 25 ] This "forbidden" transition has been shown in other materials to provide a built in mechanism for decreasing photo-excited electron-hole pair recombination rates and prolonging their lifetime, thereby greatly increasing their chances of carrying out useful surface chemistry. [ 26 ] In addition to these benefi cial optical and electronic properties, the surface properties of In 2 O 3 have garnered interest in the fi eld of thermally driven heterogeneous catalysis. Sun et al. have demonstrated the high activity of In 2 O 3 towards the reverse water gas shift (RWGS) reaction at high temperatures, specifi cally citing CO 2 capture as a key factor in enhancing the activity. [ 27 ] Ye et al. have suggested from computational modeling that surface oxygen vacancies could act as active sites to promote thermally driven methanol synthesis. [ 28 ] In this paper, hydroxylated indium oxide nanoparticles (In 2 O 3-x (OH) y ), populated with surface hydroxides and oxygen vacancies, are investigated as a gas-phase CO 2

Introduction
The emerging fi eld of solar fuels centers on storing radiant solar energy in the form of chemicals that can be used as an alternative to fossil fuels. A major goal in this fi eld is to realize an "artifi cial leaf" -a material that converts light energy in the form of solar photons into chemical energy -using CO 2 as a feedstock to generate useful chemical species. Enabling this technology will allow the greenhouse gas, CO 2 , emitted from energy production and manufacturing exhaust streams to be converted into valuable products (such as solar fuels or chemical feedstocks), thereby creating huge economic and environmental benefi ts by simultaneously addressing energy security and climate change issues. [1][2][3][4] While the global research effort with respect to the artifi cial leaf has focused on H 2 O splitting, the photocatalytic reduction of CO 2 remains a signifi cant challenge and thus form the focus of our work. [ 5 ] This artifi cial leaf can exist in multiple confi gurations, of which gas-phase photocatalysis has been identifi ed as the most practical and economically feasible option for large-scale CO 2 reduction. [ 5 ] Thus, the envisioned artifi cial leaf will be a multi-component system that intakes large quantities of gaseous CO 2 and pipes out large volumes of carbon-based fuels. Clearly, the key component of this artifi cial leaf system is a functional material that utilizes the energy from absorbed solar photons to drive the complex multi-electron and proton transfer reactions involved in reducing CO 2 to fuels. As a result, there is growing interest in synthesizing semiconductor nanomaterials, which have the surface, optical, and electronic properties that can enable photocatalytic reduction of gas-phase CO 2 to generate solar fuels. [5][6][7][8][9][10][11][12] However, despite the growing interest and investment in the fi eld, there are few examples of reduction photocatalyst. We use a temperature-programmed thermal dehydration reaction to make In 2 O 3-x (OH) y nanoparticles from In(OH) 3 . This simple and "green" fabrication method has numerous advantages including high atom economy, ease of scale-up, and negligible residual carbon contamination, which can block active sites and lower the overall gas-phase adsorption capacity and catalytic activity. [ 29 ] Moreover, since it has been reported that the sample calcination temperature has an effect on the incident photon-to-electron conversion effi ciency (IPCE) of In 2 O 3 fi lms for photoelectrochemical water splitting [ 30 ] as well as the photocatalytic degradation of dyes, [ 31 ] we produced, characterized and evaluated the photocatalytic performance of In 2 O 3-x (OH) y nanoparticles prepared via thermal dehydration reactions at 250 °C, 350 °C, and 450 °C, in addition to crystalline In(OH) 3 nanoparticles prepared from the same precursor.
Although minimal amounts of organics are present in our synthesis, we still took precaution by using 13 C-labelled CO 2 ( 13 CO 2 ) as a reactant while testing the photocatalytic performance of these nanoparticles for CO 2 photocatalytic activity. Light-driven CO 2 conversion rates reported in the literature are often low and the ubiquitous carbon contamination from carbon-containing precursors, organic solvents, and organic additives that are used to control the size and morphology of the nanostructure can create false positive results, calling into question the validity of previously reported photoactivity. [ 29 ] In fact, until recently few studies provided this type of evidence to support their claims, however this practice is becoming increasingly more common due to increased recognition of the importance of these tests. [ 32 ] The overall reaction between CO 2 and H 2 O is highly endergonic, with the majority of the energy consumed in splitting water; but vast improvements in H 2 O splitting systems have opened the possibilities for a sustainable and economically competitive supply of H 2 as a reactant. Reactions with H 2 and CO 2 are thermodynamically favourable relative to those between H 2 O and CO 2 . Thus, H 2 generated separately via solar-driven water splitting can be used in the subsequent photocatalytic reduction of CO 2 -such as the one described herein -to maximize the potential of the harvested sunlight. Therefore, CO 2 reduction photocatalyts that operate in a H 2 environment at reasonably elevated temperatures provide valuable insights into CO 2 reduction mechanisms and increase the opportunity for researchers to discover components for a scalable artifi cial leaf.
The majority of CO 2 reduction photocatalysts reported in the literature operate at room temperature or 80 °C for gas-and aqueous-phase reactions, respectively. [ 13,17 ] A key insight presented in this study is that although a photocatalyst may show little or no activity at these low temperatures, by slightly elevating the reaction temperature the material can be activated and function as a CO 2 reduction photocatalyst. These moderate temperatures can easily be reached by using simple solar trough concentrators, [ 33 ] meaning that no external energy input is required to heat the samples.
In this work we report gas-phase photocatalytic conversion of 13 CO 2 in the presence of H 2 to generate 13 CO at a rate as high as 0.25 µmol g cat −1 h −1 in a batch reactor at 150 °C under simulated solar illumination intensities of 2200 W m −2 on hydroxylated indium oxide nanoparticle fi lms. We then perform the isotope tracing experiments with coupled gas chromatography-mass spectroscopy (GC-MS) analysis to confi rm -with complete certainty -that the observed gaseous products originate from 13 CO 2 feedstock rather than adventitious carbon sources. [ 34 ] Furthermore, under only visible light irradiation (λ > 420 nm) we fi nd that our indium oxide nanoparticles photocatalytic reduce 13 CO 2 at a rate of 70 nmol g cat −1 h −1 at the same light intensity. Finally, by using a tubular fl ow reactor under similar conditions but with fl owing CO 2 and H 2 , the observed CO production rate can be further increased to 15 µmol g cat −1 h −1 . Our results show that by combining the favourable optical and electronic properties inherent to indium oxide with a judiciously tailored surface, In 2 O 3-x (OH) y nanoparticles can function as an active photocatalyst for gas-phase CO 2 reduction. This study provides valuable insight about key parameters for the composition selection, materials design and performance optimization of photocatalysts suited for large-scale solar fuels production.

Characterization of Hydroxylated In 2 O 3-x (OH) y via Temperature-Controlled Decomposition of In(OH) 3
Hydroxylated indium oxide nanoparticles were produced via a thermal dehydration of In(OH) 3 ( Figure 1 ). As the transition from In(OH) 3 to In 2 O 3-x (OH) y does not occur until approximately 210 °C, only samples heated above this temperature undergo dehydration to form indium oxide. [ 35,36 ] The transmission electron microscopy (TEM) images in Figure 1 illustrate the change in nanostructure morphology with increasing calcination temperature. The In(OH) 3 sample calcined at 185 °C (Figure 1 a) consists of large porous sheet-like structures. As the calcination temperature is increased to 250 °C, the sheetlike structures decompose into clusters of fused nanoparticles approximately 5 nm in diameter (Figure 1 b and 1 f). The overall clusters are similar in size to the In(OH) 3 sheets, indicating that the observed porosity is likely a result of water molecules being released from the lattice as the In(OH) 3 decomposes. As the calcination temperature increases further to 350 °C (Figure 1 c and 1 g) and 450 °C ( Figure 1 d and 1 h), the average particle size increases and overall porosity of the clusters decreases. The high angle angular dark fi eld (HAADF) high-resolution TEM (HR-TEM) images and the powder X-ray diffraction patterns in Figure 2 a confi rm that each sample consists of a single pure crystalline phase. The sample treated at 185 °C crystallizes to form pure cubic In(OH) 3 , while all other samples form pure bixbyite In 2 O 3 with no observable In(OH) 3 crystalline phases. For clarity, the series of In 2 O 3 samples prepared at the different calcination temperatures of 250 °C, 350 °C, and 450 °C will be referred to as I-250, I-350, and I-450 respectively.
The optical properties of each sample were determined from the diffuse refl ectance spectra shown in Figure 2 b. As expected, the absorption edge of In 2 O 3-x (OH) y is signifi cantly red-shifted in comparison to In(OH) 3 . These diffuse refl ectance spectra were fi tted with a modifi ed Kubelka-Munk function [ 37 ] ( Figure S1, Supporting Information) to determine the optical band gap of each sample, as indicated in Figure 2 c. By  correlating these values with the valence band maxima and Fermi energy (E F ) data obtained from X-ray photoelectron spectroscopy (XPS), we can calculate the band alignment relative to the vacuum level ( Figure 2 c), which corresponds well to what has been reported in the literature. [ 25,38 ] The position of the Fermi energy (E F ) just below the conduction band indicates that the as-prepared In 2 O 3-x (OH) y samples are n-type semiconductors [ 25,39 ] and the overall band alignment suggests that all samples may have suffi cient reducing power to photocatalytically drive gas-phase CO 2 reduction reactions. [ 13 ]

Demonstration of Photocatalytic Activity using 13 CO 2 Labeling
In order to confi rm the photocatalytic activity of the In 2 O 3-x (OH) y samples, carbon-13 labelled carbon dioxide ( 13 CO 2 ) is used as a tracer molecule to identify products from the photocatalytic reaction in the presence or absence of irradiation. This is an important step that determines whether the carbon source of the products originates from CO 2 or from adventitious carbon contamination on the sample. [ 29 ] After 16 hours of reaction at 150 °C under both light (800 W m −2 using a 1000 W metal halide bulb) and dark conditions, I-250 produced CH 4 , CO, H 2 O and trace amounts of higher chain hydrocarbons. It was found that CH 4 is produced at an average rate of 11 nmol g cat −1 hour −1 under irradiation, and was produced even in the absence of irradiation. It was also observed that the CH 4 production rate decreased with subsequent batch reactions. The product ion-fragmentation pattern obtained using GC-MS ( Figure S2, Supporting Information) shows that the intensity of the 16 AMU parent peak of 12 CH 4 is significant, while the intensity of the 17 AMU parent peak of 13 CH 4 is barely above noise level. This suggests CH 4 is produced by the decomposition or reaction of adventitious carbon on the surface and not from the CO 2 feedstock. In contrast, it was found that CO is unequivocally a product of CO 2 photocatalytic activity and is produced only under light irradiation at an average rate of 0.25 µmol g cat −1 hour −1 in batch reactors. Figure 3 a shows the relative intensities of the 28 AMU parent peak of 12 CO and the 29 AMU parent peak of 13 CO under both dark and light conditions. The absence of the 29 AMU peak in the dark and the signifi cant increase in its intensity under irradiation demonstrates that the conversion of 13 CO 2 to 13 CO is light driven. This fi nding is further confi rmed by a comparison of time dependent product formation. Figure 3 b shows that CO production increases linearly with time only under irradiation, while CH 4 production remains at near baseline levels under both dark and light conditions.

Investigating the Effects of Sample Calcination Temperature, and Reaction Temperature on Photocatalytic Activity
Previous studies have indicated that sample calcination temperature can have a strong effect on the aqueous-phase photoelectrochemical performance of indium oxide nanostructures. [ 30,31 ] In order to determine if the calcination temperature also affects the photocatalytic performance in the gas-phase, we measured the CO production rates under light and dark conditions for a set of In 2 O 3-x (OH) y sample fi lms, each loaded with 20 mg of sample and calcined at different temperatures (I-250, I-350 and I-450) as well as a 20 mg In(OH) 3 control sample fi lm. The CO production rates for these samples under 80 mW cm −1 of irradiation using a 1000 W metal halide lamp at a range of reaction temperatures are shown for each sample in Figure 3 c. In general, indium oxide samples calcined at lower temperatures produced CO at higher rates for the range of reaction temperatures studied. The photocatalytic CO 2 reduction rate at 150 °C for I-250 is roughly twice that of I-350 and approximately an order of magnitude greater than that of I-450. The In(OH) 3 control produced almost no CO at any temperature. Both I-250 and I-350 demonstrate an increase in photocatalytic CO production with increasing reaction temperature, reaching a maximum at 150 °C and decreasing in activity at 170 °C. Sample I-450, on the other hand, shows an increase in reaction activity at 170 °C relative to 150 °C, in contrast to I-250 and I-350. Due to limitations on the maximum operating temperature of the photoreactors, the reaction was not investigated at temperatures higher than 170 °C. Alternative reactor designs are currently being investigated to understand these trends at higher temperatures. The effect of irradiation on CO production rates for each sample at different reaction temperatures is also shown in Figure S3, Supporting Information. As expected, all samples produced very little CO under dark conditions, with a maximum of 2.8 nmol g cat −1 h −1 measured at 170 °C for I-250, demonstrating that the observed gas phase CO 2 reduction is a light-driven process.

Investigating the Effect of Light Intensity and Spectral Distribution on Photocatalytic Activity
In order to determine the effect of light intensity on the CO production rate, a 20 mg I-250 fi lm was irradiated with a Newport 300 W Xe Lamp fi tted with an AM1.5 fi lter to simulate the solar spectrum. The photocatalytic activity of the sample was tested at 150 °C under varying light intensities from 0.8 to 2.2 suns. Figure 3 d shows a linear increase in CO production rate with increasing light intensity, which further confi rms that the CO 2 to CO conversion is a light-driven reaction. A single sample was used for the duration of these measurements, demonstrating the robustness of this photocatalyst.
The spectral dependence of CO production was also investigated ( Figure S4, Supporting Information). A single 20 mg I-250 fi lm was irradiated with a Newport 300W Xe Lamp, fi tted with either an AM1.5 fi lter or an AM1.5 fi lter combined with either a 420 nm or a 615 nm high-pass fi lter. The light intensity was set to 1700 W m −2 , using a focusing lens to adjust the intensity and a calibrated reference cell to measure the output. When the I-250 sample was initially irradiated with the AM1.5 fi ltered light, a CO production rate of 0.20 µmol g cat −1 hour −1 was observed. The second run -with the additional 420 nm high pass fi lter that cut off all wavelengths with energy greater than 420 nm -produced CO at a rate of 70 nmol g cat −1 hour −1 . No CO was detected when a 615 nm high-pass fi lter was used. Finally, a repeat of the initial measurement using only the AM1.5 fi lter was conducted, reproducing the rate of 0.20 µmol g cat −1 hour −1 . These results demonstrate that not only is the I-250 capable of converting gaseous CO 2 to CO using only visible light, which correlates well with the diffuse refl ectance measurements, but also that In 2 O 3-x (OH) y is stable under these reaction conditions and can produce rates consistent with the initially measured values even after being irradiated continuously for 4 days.

Photocatalytic CO 2 Reduction Rates using a Flow Reactor
In an attempt to simulate more industrially relevant conditions, preliminary photocatalytic rate measurements were carried out in a tubular fi xed bed fl ow reactor irradiated with a Newport 300W Xe lamp. This investigation of the In 2 O 3-x (OH) y nanoparticles revealed that under similar conditions to those in the batch photoreactors (2200 mW cm −2 , 150 °C and 3 atm), under fl owing CO 2 and H 2 , CO is photocatalytically produced at a rate of 15 µmol g cat −1 h −1 with 24 mg (3 cm bed length) of the I-250 nanoparticle powder sample. These rates are higher than previously reported CO production rates for other single component metal oxides; for instance 1.61 µmol g cat −1 h −1 for MgO, [ 40 ] and 0.56 µmol g cat −1 h −1 for ZrO 2 . [ 41 ] This increased activity is the focus of a current on-going investigation. Additionally, in order to ensure consistency in this study, the photocatalytic rate data presented here are limited to those from identical batch photoreactors. This mitigates effects caused by variations in particle size between I-250, I-350, and I-450 and allows us to more accurately make comparisons between these samples. Differences in particle size of powders packed in a catalyst bed can result in substantial variations in pressure gradients, making comparisons between samples more diffi cult.

Characterization of Surface Hydroxides and Oxygen Vacancies
In order to further understand the effects of calcination temperature on the photocatalytic activity, XPS measurements were conducted. Figure 4 a shows that the In3d 5/2 core level peak shifts to a lower binding energy as the calcination temperature is increased, indicating an increase in charge density around the In atoms as a result of the removal of OH groups. The O1s core level spectra in Figure 4 b shows a sharp contrast between In(OH) 3

and wileyonlinelibrary.com
Adv. Sci. 2014, 1, 1400013 In 2 O 3-x (OH) y samples. There is an approximately 2.5 eV shift to lower binding energy of the main O1s peak, from 532.7 eV for In(OH) 3 to 530.2 eV for I-450. Additionally, a shoulder peak appears in the O1s core level peak of the In 2 O 3-x (OH) y samples, indicating that there is more than one chemical state of oxygen present in the structure. Indeed, the O1s peak for the In 2 O 3-x (OH) y samples can be de-convoluted into three distinct peaks: the main oxide peak at 530.3 eV and two additional peaks at 531.7 eV and 532.5 eV (Figure 4 c-e). The peak at 531.7 eV is commonly attributed to the presence of oxygen vacancies in the structure. [ 42,43 ] It is also consistent with the observed n-type position of the Fermi-levels relative to the conduction bands (Figure 2 c), which is typically a result of non-stoichiometry. [ 44 ] This vacancy peak is shifted to a higher binding energy relative to the main oxide peak. This is a result of the change in O interaction with an In centre that is more reduced in character because it is surrounded by less than six O atoms (due to the oxygen vacancies). The peak at 532.5 eV has been attributed to surface OH groups [ 42 ] and agrees well with the O1s spectra for the pure In(OH) 3 peak. From these plots in Figure 4 c-e, it is clear that the shoulder peak -with contributions from both vacancies and surface hydroxides -decreases with increasing calcination temperature. XPS measurements were also conducted on samples exposed to reaction conditions and results indicate a slight change in the shoulder peak of the O1s spectra. Current on-going research aims to investigate this material in situ to determine how the surface may change during reaction. The hydroxide content of the samples was investigated further by both Fourier transform infrared (FT-IR) spectroscopy and thermogravimetric measurements. The intensity of the OH stretches in the FT-IR spectra ( Figure S5, Supporting Information) decreases with increasing calcination temperature, illustrating that samples treated at higher temperatures have lower hydroxide contents. In order to establish the extent of hydroxide loss during synthesis, each In 2 O 3-x (OH) y sample was synthesized in situ within a thermogravimetric analyzer. The weight loss observed was attributed to hydroxide condensation to form bridging oxides. Figure S6, Supporting Information, shows the weight loss of each sample after calcination under air fl ow for 3 hours. It is clear that lower calcination temperatures correspond to less overall weight loss and result in stabilization at a higher relative weight. By comparing these plots to the theoretical maximum weight loss (when all hydroxides are converted to bridging oxides), it is apparent that I-450 should have almost no hydroxides left. I-250 and I-350, on the other hand, have additional weight above the theoretical value, which we attribute to the retained hydroxyl groups. From this data and the Brunauer-Emmett-Teller (BET) surface area described below, we estimate that the surface hydroxide coverage is on the order of 6 µmol m −2 for I-250 and 3 µmol m −2 for I-350. It is expected that the combination of both hydroxide groups and oxygen vacancies is a key feature of our functioning In 2 O 3-x (OH) y nanoparticle photocatalysts and that both of these entities are present and work in concert at active sites.

Correlation of CO 2 Capture Capacity with Observed Photocatalytic Rates
To clarify the observed trend in CO production rates between In 2 O 3-x (OH) y samples, the CO 2 capture capacity was determined for each sample at 150 °C, the reaction temperature at which the highest CO production rates for all samples were observed. Furthermore, in order to more accurately compare the CO 2 capture capacity and the photoactivity, both are normalized to the surface area of each sample, determined using the BET method. The surface areas for the In(OH) 3 , I-250, and I-350 were remarkably similar at 124.7 m 2 g −1 , 125.0 m 2 g −1 , and 129.6 m 2 g −1 , respectively. Only I-450 had a signifi cantly lower surface area at 90.0 m 2 g −1 , which is likely a result of some nanoparticle sintering at the higher calcination temperatures. Figure 5 shows the surface-area-normalized CO 2 capture capacities for each sample plotted together with their corresponding CO production rates. There is a notable strong correlation between CO 2 reduction rates and the normalized CO 2 capture capacity.

Designing a Surface for CO 2 Photocatalytic Activity
The affi nity of a photocatalyst's surface for CO 2 has been identifi ed in this study, as well as in others, [18][19][20] as a critical factor infl uencing photocatalytic activity. As Figure 5 demonstrates, the CO 2 capture capacity of the In 2 O 3-x (OH) y nanoparticles corresponds very well with photo-reactivity, indicating that CO 2 adsorption plays an important role in the light-driven reaction. Intuitively, CO 2 molecules must be able to approach and interact with the surface for a suffi cient amount of time in order for electron transfer to occur. Surface hydroxides have a known affi nity for the Lewis-acidic CO 2 . [ 45 ] This could explain the strong positive correlation between CO 2 capture capacity and hydroxide content. However, while the In(OH) 3 control sample has the highest hydroxide content and a similar surface area to that of I-250, it also has a signifi cantly lower CO 2 capture capacity. This indicates that surface hydroxides alone are not suffi cient to facilitate CO 2 capture and photocatalytic reduction of CO 2 .
In addition to hydroxides, the surface of the In 2 O 3-x (OH) y nanoparticles is also populated with oxygen vacancies. The presence of these oxygen vacancies in the In 2 O 3-x (OH) y samples is supported by both the de-convolution of the XPS O1s core level peaks (Figure 4 c-e) as well as the n-type position of the Fermi-levels relative to the conduction bands (Figure 2 c), which is typically a result of non-stoichiometry. [ 44 ] It is apparent from both fi gures that I-250 has the largest peak associated with oxygen vacancies as well as the highest Fermi energy, implying a higher abundance of vacancies compared to the other temperature-treated In 2 O 3-x (OH) y samples. The increase in oxygen vacancies for I-250 may result from the natural increase in surface defect sites as the particle size decreases. Surface oxygen vacancies may also arise from the crystal structure of In 2 O 3 . The cubic In 2 O 3-x (OH) y samples have a bixbyite structure, which can be understood as the CaF 2 -type lattice with 25% of the tetrahedral anion sites vacant. This additional space in the bixbyite structure may result in more dynamic fl exibility, especially at the nanoparticle surface, allowing for more atomic mobility in the lattice; indeed In 2 O 3 is a known solid ionic and protonic conductor. [ 46 ] Additionally, these intrinsic oxygen vacancies may increase the stability of the vacant surface sites, allowing the material to remain stable under reaction conditions. By contrast the stoichiometric In(OH) 3 with its perovskite structure does not have a signifi cant concentration of surface oxygen vacancies. The implied necessary combination of surface hydroxides and oxygen vacancies could provide an explanation for the stark difference in CO 2 capture capacity and photocatalytic activity of In(OH) 3 and I-250.
Surface oxygen vacancies may also form due to interactions between surface oxygen sites and H 2 or CO under elevated reaction temperatures. In 2 O 3 has been investigated experimentally [ 47,48 ] and theoretically [ 28,49 ] H 2 reduction of the surface under these reaction conditions may suggest the temperature dependence of the CO production rates for the In 2 O 3-x (OH) y is due to the availability of surface oxygen vacancies (Figure 3 c). As shown, very little CO is observed at 110 °C (sample I-250 is the only sample to produce a signifi cant amount of CO at 110 °C). However, at reaction temperatures above 130 °C, CO production under light irradiation is signifi cant. While the photocatalytic production increases from 110 °C to 150 °C, the activity decreases at 170 °C. The decrease in CO production at 170 °C may be due to oxidation of CO by lattice oxygen on the In 2 O 3-x (OH) y surface as shown in Equation ( 2) . [ An alternative explanation for the observed dependence on temperature trend is the adsorption and desorption of molecules at the surface. At higher temperatures, product molecules such as H 2 O, which can block active sites, may desorb enabling more turnovers at these active sites. [ 50 ] Since it is observed that In 2 O 3-x OH y samples achieve a maximum effi ciency at 150 °C, this may indicate that 150 °C is a "sweet spot," combining efficient CO 2 adsorption and effi cient CO and H 2 O desorption. It is also possible that the reaction takes place between CO 2 and the surface oxygen vacancies, as outlined in Equation ( 1) to produce CO through a surface oxidation reaction. However, pre-reducing the sample in H 2 at elevated temperature followed by a batch reaction in CO 2 yielded no trace of CO under irradiation. Thus it is believed that the observed reaction is the reverse water gas shift (RWGS) reaction as shown in Equation ( 3) . 2 2 2

CO H CO H O
While we have observed water as a product, an exact reaction stoichiometry is diffi cult to quantify to complete a mass balance for the proposed RWGS reaction due to uncertainties created by the strong interaction of water with the tubing connecting the reactors to the GC and GC-MS.

Conclusion
A functional single component CO 2 reduction photocatalyst must have surface, optical, and electronic properties working in concert for photocatalytic reduction of CO 2 to occur in the gas phase. In this study the In 2 O 3-x (OH) y nanoparticles demonstrate activity for the photocatalytic reduction of CO 2 in the presence of H 2 at temperatures as low as 130 °C using both ultraviolet and visible light. Our work strongly suggests that the observed activity of In 2 O 3-x (OH) y samples is associated with surface populations of oxygen vacancies and hydroxides, which may act in concert as active sites for CO 2 adsorption and charge transfer under simulated solar irradiation.
We have produced a series of nanostructured In 2 O 3-x (OH) y materials via a temperature controlled thermal dehydration of In(OH) 3 . Using 13 CO 2 as a tracer molecule, strong light and temperature-dependent photocatalytic reduction of gaseous 13 CO 2 to 13 CO is confi rmed in the presence of H 2 . The surface hydroxide and oxygen vacancy content strongly correlates with both an increase in 13 CO 2 capture capacity and an increase in photocatalytic activity for 13 CO production. By combining the favourable surface, electronic and optical properties of nanostructured In 2 O 3-x (OH) y with the bixybite crystal structure and its enhanced CO 2 capture capabilities, we have demonstrated a combination of key components to be considered in the discovery, optimization, and scaling of new and effi cient gas-phase CO 2 reduction photocatalysts for solar fuels production.

Experimental Section
Synthesis of In 2 O 3-x (OH) y Nanoparticles : An In(OH) 3 precursor was synthesized and subsequently dehydrated into In 2 O 3 nanoparticles following a modifi ed version of a previously published procedure. [ 51 ] All chemicals were used as received without any further purifi cation. In a typical synthesis, of indium(III) chloride (3.6 g, 16.2 mmol, Sigma Aldrich, 98%) was dissolved in a 3:1 solution (72 mL) of of anhydrous ethanol (Commercial Alcohols) and deionized, nanopure water (resistivity 18.2 MΩ cm). In a separate beaker, a 3:1 mixture of ethanol and ammonium hydroxide was prepared by combining aqueous ammonium hydroxide (18 mL, Caledon, 28-30% adjusted to 25 wt% with deionized water) and of anhydrous ethanol (54 mL). The solutions were rapidly combined, resulting in the immediate formation of a white precipitate. To control the particle size, the resulting suspension was immediately immersed in a pre-heated oil bath at 80 °C and stirred for 10 min. The suspension was then removed from the oil bath and allowed to cool to room temperature. The precipitate was separated via centrifugation and washed 3 times with deionized water. The precipitate was sonicated between washings to ensure adequate removal of any trapped impurities and then dried overnight at 80 °C in a vacuum oven. The dried precursor powder was fi nely ground with a mortar and pestle and calcined for 3 hours in air at 185 °C, 250 °C, 350 °C, and 450 °C. Sample fi lms were prepared for photocatalytic testing by drop casting 20 mg of each sample powder -suspended via sonication in deionized, nanopure water (3 ml) -onto 1" × 1" binder free borosilicate glass microfi ber fi lters (Whatman, GF/F, 0.7 µm) placed on top of a vacuum fi ltration funnel that was under very weak vacuum. This sample loading was selected to optimize CO production rates, as illustrated in Figure S7, Supporting Information.
Physical Characterization : Powder X-ray diffraction (PXRD) was performed on a Bruker D2-Phaser X-ray diffractometer, using Cu Kα radiation at 30 kV. Nitrogen Brunauer-Emmet-Teller (BET) adsorption isotherms were obtained at 77 K using a Quantachrome Autosorb-1-C. Sample morphology was determined using a JEOL-2010 high resolution transmission electron microscope (HR-TEM). Fourier transform infrared spectroscopy (FT-IR) was performed using a Perkin Elmer Spectrum-One FT-IR fi tted with a universal attenuated total refl ectance (ATR) sampling accessory with a diamond coated zinc selenide window. Diffuse refl ectance of the samples was measured using a Lambda 1050 UV/VIS/ NIR spectrometer from Perkin Elmer and an integrating sphere with a diameter of 150 mm. Sample weight loss during the calcination process was determined by placing approximately 10 mg of un-calcined indium hydroxide precursor in a TA Instruments Q500 thermogravimetric analyzer, jumping to the set temperature of either 250 °C, 350 °C or 450 °C and holding for 3 hours under a fl ow of compressed air. The sample weight was determined using the built-in ATI CAHN C-34 microbalance. The fi lm morphology and thickness was characterized by scanning electron microscopy using a QUANTA FEG 250 ESEM. The borosilicate glass microfi ber fi lters were used as a substrate to provide increased surface area as well as mechanical stability. Figure S8, Supporting Information, shows SEM micrographs of a typical In 2 O 3 sample on the fi lter. Figure S8a, Supporting Information, shows a crosssection of the fi lm, indicating its thickness is approximately 50 µm. The magnifi ed image shown in Figure S8b, Supporting Information, indicates that the as-prepared sample maintains its high porosity, an important factor for gas-phase reactions. X-ray photoelectron spectroscopy (XPS) was performed using a Perkin Elmer Phi 5500 ESCA spectrometer in an ultrahigh vacuum chamber with base pressure of 1 × 10 −9 Torr. The spectrometer uses an Al Kα X-ray source operating at 15 kV and 27 A. The samples used in XPS analyses were prepared by drop-casting aqueous dispersions onto p-doped Si(100) wafers in the case of the In 2 O 3 samples and fl uorine-doped tin oxide substrate in the case of the In(OH) 3 sample. All data analyses were carried out using the Multipak fi tting program and the binding energies were referenced to the NIST-XPS database and the Handbook of X-ray photoelectron spectroscopy. [ 52,53 ] Gas-Phase Photocatalytic Measurements : Gas-phase photocatalytic rate measurements were conducted in a custom fabricated 1.5 mL stainless steel batch reactor with a fused silica view port sealed with Viton O-rings. The reactors were evacuated using an Alcatel dry pump prior to being purged with the reactant gases H 2 (99.9995%) and CO 2 (99.999%) at a fl ow rate of 6 mL min -1 and a stoichiometry of either 4:1 (stoichiometric for Sabatier reaction) or 1:1 (stoichiometric for reverse water gas shift reaction). During purging, the reactors were sealed once they had been heated to the desired temperature. The reactor temperatures were controlled by an OMEGA CN616 6-Zone temperature controller, with a thermocouple placed in contact with the sample. The pressure inside the reactor during reaction was monitored during the reaction using an Omega PX309 pressure transducer. Reactors were irradiated with a 1000 W Hortilux Blue metal halide bulb (the spectral output is shown in Figure S9, Supporting Information) for a period of 16 hours. Product gases were analyzed by a fl ame ionization detector (FID) and thermal conductivity detector (TCD) installed in a SRI-8610 Gas Chromatograph (GC) with a 3' Mole Sieve 13a and 6'Haysep D column. Isotope tracing experiments were performed using 13 CO 2 (99.9 atomic% Sigma Aldrich). The reactors were evacuated prior to being injected with 13 CO 2 , followed by H 2 . Isotope product gases were measured using an Agilent 7890A gas chromatographic mass spectrometer (GC-MS) with a 60 m GS-Carbonplot column fed to the mass spectrometer. The spectral dependence of the photoactivity of the In 2 O 3 nanoparticles was investigated using a Newport 300 W Xe Lamp (the spectral output is shown in Figure S9, Supporting Information) fi tted with a combination of AM1.5, 420 nm high-pass and 615 nm highpass fi lters. Since each fi lter reduced the total irradiation intensity, the beam was focused using collimating lenses to maintain an irradiation intensity of 100 mW cm -2 . The spectral output was measured with a StellarNet Inc spectrophotometer. Irradiation intensity was measured by a Newport 91150V calibrated reference cell and meter. Additional photocatalytic rate measurements were carried out in a borosilicate tube (3 mm outer diameter and 2.5 mm inner diameter) reactor packed with 24 mg (3 cm bed length) of In 2 O 3-x (OH) y nanoparticle powder. Quartz wool was used to support either end. The reactor was held in a custom designed stand. Heating was supplied from a heated copper block fi xed below the fi xed catalyst bed. The top of the reactor was exposed in order to allow light irradiation from a Newport 300 W Xe Lamp (at a distance of 4 cm and a light intensity of 2.2 suns). The reactor was purged with H 2 (99.9995%) and CO 2 (99.999%) at a fl ow rate of 10 mL min -1 and a stoichiometric ratio of 1:1 (stoichiometric for reverse water gas shift reaction). The reactor temperatures were controlled by an OMEGA CN616 6-Zone temperature controller. A thermocouple was in contact with the top of the reactor so that the reactor maintained a constant temperature of 150 °C.
CO 2 Capture Capacity : The CO 2 capture capacity of each sample was measured by thermogravimetric analysis (TGA) with a TA Instruments Q500 thermogravimetric analyzer. A desorption step was fi rst carried out under N 2 fl ow at a rate of 100 mL min -1 with a temperature ramp of 10 °C min -1 up to 150 °C; the temperature was held at 150 °C for 3 hours. To measure the amount of CO 2 adsorption, the gas was then switched to 100% dry CO 2 at a fl ow rate of 100 mL min -1 ; the temperature was then maintained at 150 °C for 10 hours. The weight gain observed during this adsorption step was used to calculate the CO 2 capture capacity of the sample. Desorption of CO 2 was performed by switching the gas fl ow back to N 2 fl ow for 5 hours while keeping the temperature constant at 150 °C.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.