Pivotal Role of Holes in Photocatalytic CO2 Reduction on TiO2.

Evidence is provided that in a gas-solid photocatalytic reaction the removal of photogenerated holes from a titania (TiO2 ) photocatalyst is always detrimental for photocatalytic CO2 reduction. The coupling of the reaction to a sacrificial oxidation reaction hinders or entirely prohibits the formation of CH4 as a reduction product. This agrees with earlier work in which the detrimental effect of oxygen-evolving cocatalysts was demonstrated. Photocatalytic alcohol oxidation or even overall water splitting proceeds in these reaction systems, but carbon-containing products from CO2 reduction are no longer observed. H2 addition is also detrimental, either because it scavenges holes or because it is not an efficient proton donor on TiO2 . The results are discussed in light of previously suggested reaction mechanisms for photocatalytic CO2 reduction. The formation of CH4 from CO2 is likely not a linear sequence of reduction steps but includes oxidative elementary steps. Furthermore, new hypotheses on the origin of the required protons are suggested.

Introduction

More than forty years have passed since the discovery of photocatalytic and photoelectrochemical reduction of CO2 on TiO2 and other semiconductors. Just like overall water splitting, this reaction promises the formation of a storable fuel with only (sun)light as energy input. In both reactions, water splitting and CO2 reduction, the desired product is a result of the reduction reaction only, but the stoichiometry necessarily requires forming an oxidation product, too. The straightforward oxidation product is gaseous O2 resulting from the oxidation of H2O. It is well known that it is the more challenging of the two half reactions, particularly on TiO2. For scientific purposes, it is possible to replace the challenging H2O oxidation reaction with the oxidation of a sacrificial donor, but since the overall reaction then usually becomes thermodynamically favorable (ΔG<0), energy storage is no longer achieved, and in addition the reaction has no more economic potential. The fate of the photogenerated holes as oxidizing equivalents must thus be understood and controlled, so that either O2 can be evolved, or another industrially relevant oxidation reaction can be performed and coupled to CO2 reduction.

A previous work by some of us has shown that in a gas‐solid photocatalytic process TiO2 is able to react with some yet unknown type of oxygen species liberated in the course of CO2 reduction, thereby stoichiometrically consuming the oxidation equivalents. Consequently, both reduction and oxidation reactions can only proceed until the TiO2 is saturated with those oxygen species. In an uninterrupted reduction reaction of CO2 in dilute concentration in a continuous flow, this saturation process takes more than 12 h due to the small yields of the reduction reaction. This may explain why many scientific articles – including some of our own works – have previously been published that do not present any oxidation product. Our attempts to facilitate O2 evolution with Ir or Co oxide cocatalysts on TiO2 was only partially successful: An active photocatalyst for overall splitting of gaseous H2O was obtained, but not a trace of any carbon‐containing reduction product was observed with this system. This can either be rationalized by a rapid reoxidation of the formed hydrocarbons by the O2 now liberated into the gas phase, or by a reaction mechanism previously suggested by Shkrob et al. This so‐called “glyoxal mechanism” does not involve a “linear” series of successive reduction steps, but also includes oxidation reactions such as the oxidation of acetaldehyde to an acetyl radical and then to a methyl radical through decarbonylation.[ , ] If this mechanism is in operation under our reaction conditions – and previous results indicate that this is the case – then CH4 formation is no longer possible when the holes are consumed for another reaction. Based on our previous results with the oxygen evolving cocatalysts, it was not possible to distinguish between the two explanations of rapid hydrocarbon oxidation on the one hand, or a pivotal role of holes in the reaction mechanism from CO2 to CH4 on the other hand because it requires removing the photogenerated holes without at the same time producing O2.

Here, results of such an attempt to reduce CO2 but circumvent O2 evolution are presented. Two different approaches are followed: Firstly, the coupling of CO2 reduction to the oxidation of an alcohol is attempted. Alternatively, H2 is added to the reaction mixture, to remove the holes and the formed oxygen species by the formation of H2O. It will be shown that any hole scavenger reduces or even entirely inhibits the formation of carbon‐based reaction products. It is concluded that the holes have a pivotal role in the reaction of CO2 to CH4, so that a full separation of oxidation and reduction reaction on bare TiO2 is likely impossible.

Results and Discussion

All measurements were performed with commercial TiO2 P25 (Evonik Industries, anatase:rutile=∼80 : 20) in our high‐purity gas‐phase photoreactor presented elsewhere[ , ] and briefly described in the Supporting Information (Figure S1 in Supporting Information). Photocatalytic CO2 reduction on P25 has been performed extensively in our previous works. It has been found reproducibly that under our standard conditions (1.5 % CO2 and 6,000 ppm H2O in He, batch reaction) CH4 is the main reaction product, and the concentration in the reactor amounts to ∼80 ppm after six hours. Aside from CH4, roughly 20 ppm CO and ∼15 ppm of H2 are formed.

Before presenting the results of the current study, we would like to briefly summarize the results of a previous study conducted under reaction conditions of continuous flow. On bare P25 (70 mg in powder form), CH4 was formed for a total of ∼14 h, with the production rate steadily declining after ∼8 h. No O2 formation was observed, but the stoichiometry of the reaction would have required the formation of 78 nmol of O2 in those ∼14 hours. When P25 was modified with Ir oxide (IrOx/TiO2), the resultant material did not reduce CO2 at all, but performed overall H2O splitting in presence of CO2. In a pure H2O splitting experiment, stoichiometric production of H2 and O2 in the expected 2 : 1 ratio was observed, but in the early stages of the experiment, the formation of O2 was substoichiometric. Certain amounts of O2 appeared to be missing in the gas phase compared to the expected stoichiometric ratio. Integration of the curve yielded an amount of ∼103 nmol of vanished O2. Considering the differing calcination temperatures of the two materials (P25 and IrOx/TiO2) and the relatively large error of this estimation, a relatively good fit of these two values can be assumed. It was concluded that the used 70 mg of TiO2 can store O2 in an order of magnitude of 100 nmol, and CO2 reduction runs only until TiO2 is saturated. In the previous publication, it was not possible to give a conclusive explanation on the fate of the holes, or why the IrOx/TiO2 sample was unable to reduce CO2 to CH4 or any other carbon‐containing product.

In the current study, oxidation reactions other than the challenging oxidation reaction of H2O were attempted, which at the same time prevented the formation of the byproduct O2. Figure 1a displays the results of a co‐reaction of CO2 with small amounts of ethanol (EtOH) on P25 under batch conditions. Please note that small amounts of H2O are also present in this reaction. The gas chromatograms of the product mixtures reveal a concentration of ∼15 ppm CO, ∼10 ppm H2 and ∼5 ppm CH4 after 6 h of reaction time. When the same reaction is performed in absence of CO2 and the amounts of the carbon‐containing products are compared (Figure 1b), they turned out to be identical. Therefore, it becomes obvious that none of these products is formed from CO2, but the entire amount of the formed CO and CH4 solely results from a decomposition reaction of EtOH. Thus, the presence of EtOH entirely suppresses the reduction reaction of CO2. It also becomes obvious that EtOH is likely not selectively oxidized under the given reaction conditions but is entirely fragmented into C1 compounds. The gas chromatograph (GC) used for these experiments (Shimadzu TRACERA equipped with barrier discharge ionization detector, BID) was calibrated for CO, CH4, CO2, H2 and H2O, and would additionally have detected C2 and C3 hydrocarbons, which were, however, not observed.

Figure 1

(a) Photocatalytic CO2 reduction (1.5 % CO2 in He) in batch mode over P25 impregnated with aqueous EtOH solution; (b) comparison of the amounts of CO and CH4 formed in reaction of EtOH on P25 with and without 1.5 % CO2 present in the gas phase.

Additional experiments have been performed using isopropanol (iPrOH) instead of EtOH, and another GC was used (Shimadzu TRACERA with BID and additional flame ionization detector, FID) which can detect hydrocarbon products up to a chain length of C14. For these experiments, 100 μl of iPrOH were added dropwise onto the surface of the P25 pellet sample in the photoreactor using a microliter pipette. Figure 2 shows the main products of the photocatalytic reaction in the presence and absence of CO2. Although a variety of other products such as ethylene (C2H4) and some yet unidentified compounds were formed (Figure S2 in Supporting Information), the main products comprise ethane (C2H6), CH4 and CO. Here, the presence of CO2 even reduces the amount of formed carbon‐containing products, hinting to a direct competition of CO2 and alcohol molecules for the photogenerated holes.

Figure 2

Comparison of the amounts of C2H6, CH4 and CO formed in reaction of iPrOH on P25 with and without 1.5 % CO2 present in the gas phase.

Furthermore, the intense blue color of the TiO2 pellet after the reaction (Figure S3 in Supporting Information) indicates that significantly more holes have been consumed than electrons because the blue color is commonly explained with a reduction of the material, or in other words, an accumulation of excess electrons, which leads to the formation of Ti3+. The blue color of the pellet was interestingly observed also on its non‐directly irradiated surface as well as in the cross‐section of the pellet. As the P25 pellet was not directly in contact with the reaction chamber bottom but was placed on top of a quartz plate, the reflected light can explain the blue color of the bottom side of the pellet. An oxygen vacancy migration effect can be assumed for the blue color of the cross‐section, as the iPrOH deposited on the surface (top side) of the pellet was also soaked into the bulk of the P25 pellet. After the end of the iPrOH oxidation experiment the pellet was removed from the photoreactor and its optical properties were examined (Figure S4 in Supporting Information). No change in the optical band‐gap (Eg) could be identified. In contact with the atmospheric air, the pellet became pale blue, but it did not completely return to its original white color as it can be seen from the flattening of the wide absorption band in the visible range. Reoxidation to full white color of the pellet was only achieved by heating it up in air to 150 °C.

Two additional P25 pellets were pressed, and the effect of TiO2 pre‐reduction was further studied. Each of the two pellets was used for the oxidation of iPrOH, leading to a blue‐colored pellet. Additionally, degradation products of iPrOH have likely been accumulated on the surface. Subsequently, one of the two blue pellets was used as a photocatalyst for a CO2 reduction experiment (addition of CO2 and H2O), whereas the other was used as a reference photocatalyst to which only He and H2O were offered. These experiments in presence or absence of CO2 provide an indirect proof of the origin of the identified products.

The formed products with and without CO2 are presented in Figure 3. The concentrations of CH4 and C2H6 appear to be similar regardless of the presence of CO2. This confirms the origin of these products from the further degradation of iPrOH ‐derived intermediates on the surface, as opposed to their origin from CO2. For CO, a small decrease can be observed when CO2 is introduced in the reactor. This may indicate a small influence of CO2 on the specific mechanisms of the surface processes involving the iPrOH degradation products. As in the case of EtOH, it is evident that all identified products originate from the decomposition of iPrOH and not from CO2. So, in spite of reduced (oxygen vacancy‐rich) TiO2 having been suggested as superior photocatalyst, CO2 reduction does not proceed if the photocatalyst is simultaneously degrading organic surface species, i. e. performs an oxidation reaction.

Figure 3

Amounts of major products formed in a photocatalytic CO2 reduction experiment (1.5 % CO2 in He, 6,000 ppm H2O, Hg/Xe lamp, light intensity 200 mW cm−2) using P25 pellets that had previously been used for iPrOH oxidation. A similar reaction without the presence of CO2 is used as comparison.

Next, the effect of H2O and H2 in the gas phase on the CH4 formation was evaluated. Please note that the results presented in the following have been obtained in a series of consecutive experiments using the same P25 sample (50 mg in powder form) without any cleaning steps in‐between to remove the adsorbates on the surface of the material, such as reaction intermediates and/or products. Figure 4 shows the effect of a variation of the H2O content under batch reaction conditions. Reacting CO2 on P25 without adding H2O resulted in a CH4 concentration of 120 ppm after 6 h (Figure 4, blue line and triangles), which was already more than under the standard reaction conditions with 6,000 ppm H2O in the gas phase. This corresponds well to similar experiments conducted previously under flow conditions. It is well known in the literature that under ambient conditions, the oxide surfaces can be covered by up to 20 layers of H2O. This allows TiO2 to perform the photoreduction of CO2 without the need for the addition of extra amounts of H2O in the gas phase. In a consecutive run without adding H2O, CH4 formation was even more enhanced, in which the 120 ppm were already formed in less than 4 h (Figure 4, grey circles and line). This may be attributable to the three following explanations: (i) The first measurement may have dried the TiO2 even further because some of the water was consumed in the reaction, and/or (ii) in a second reaction, a carbon pool was already present on the TiO2 surface, which could also contribute to product formation in a second run, and/or (iii) if less H2O is present, less holes are consumed, which are then available for the CO2 reduction pathway.

Figure 4

Consecutive CO2 reduction experiments with variation of the H2O content in the gas phase.

A drier surface according to explanation (i) can result in more CO2 molecules being adsorbed on TiO2 which in turn can lead to higher product yields. On TiO2, the adsorption of H2O is more favorable than the adsorption of CO2, so in presence of larger amounts of H2O, it may block CO2 from adsorbing in sufficient amounts. The explanation (ii) of the formation of a “carbon pool”, previously suggested for TiO2/SiO2 and TiO2, is supported by the consecutive measurement in Figure 4, in which H2O was added again to the gas phase, and product formation was even more enhanced (Figure 4, black squares and line). Although water is now present in the gas phase again, the formed carbon pool makes the reaction conditions relatively carbon‐rich, particularly near the surface. The carbon pool can consist either of products not able to desorb from the material under the selected experimental conditions, or of intermediate states along the reaction pathway. When H2O is added, the latter might eventually be converted to CH4. In this respect, in another consecutive experiment, CH4 formation was even higher, allowing to reach a concentration of more than 320 ppm in the reactor. This can be converted to a yield of 0.1 % gcat −1 h−1, which is the highest yield so far obtained with TiO2 in our gas‐solid photoreactor when only CO2 and H2O were used as reactants. With the estimated reactor volume of 25 ml, this yield, obtained with 50 mg TiO2, can be converted to ∼1.1 μmol gcat −1 h−1.

It is also remarkable about the obtained results in Figure 4 that after a short induction period, the increase in CH4 concentration in all four experiments is linear and does not start to level off as previously observed. This indicates that the CO2 reduction reaction can run unperturbed for longer periods of time under relatively carbon‐rich as opposed to water‐rich conditions. In this respect, it should also be noted that the total amount of CH4 formed in all four experiments together amounts to ∼0.80 μmol, which is much more than the 39 nmol previously observed under conditions of continuous flow in a comparable time frame and with an almost comparable sample amount (70 mg). While a detailed discussion of the implications of these observations with respect to reaction engineering is beyond the scope of this study, these results may indicate that a higher residence time is beneficial for the very slow reaction, or that the helium purging conducted in between the experiments removes some inhibiting species, such as oxygen‐related byproducts.

Altogether, the following conclusions can be drawn: If only low amounts of adsorbed water are present, relatively more CO2 will be adsorbed on the surface, thereby enhancing product formation. However, once the surface becomes too dry, not sufficient protons are still available to finish the reaction sequence to CH4, so the surface will be covered by partially hydrogenated intermediates (“carbon pool”). Then, water dosing is beneficial, indicating that water is indeed involved in the overall reaction from CO2 to CH4 as proton source, but without evolving oxygen, as we have previously suggested from experiments under continuous flow. It can also be ruled out that the primary role of water is hole scavenging because then the dosage of alcohols and water should influence product formation from photocatalytic CO2 reduction in a similar manner, which it clearly did not do: The addition of alcohols completely inhibited CO2 reduction.

This leads to the question whether another potential source of protons, or, more generally speaking, hydrogen atoms, can improve product formation when partially hydrogenated intermediates are already present on the surface. So, after the reaction sequence displayed in Figure 4, the addition of H2 was studied. Here, too, consecutive reaction cycles were run so that the sample became more and more depleted of H2O (Figure 5). The first measurement in Figure 5 (gray squares and line), conducted after the fourth measurement in Figure 4 (red circles and line), featured almost the same reaction conditions, with the sole difference that 580 ppm of H2 were additionally present in the gas phase. Surprisingly, the yield of CH4 was lower than in the corresponding measurement without H2. Removing the H2O from the gas phase (second measurement in Figure 5, red circles and line) had a slightly detrimental effect, although it was beneficial in the absence of H2. Slightly decreasing (Figure 5, blue triangles and line) or increasing (Figure 5, green triangles and line) the concentration of H2 hardly affected CH4 formation, but certainly did not promote it. In all measurements in Figure 5, a total of ∼0.84 μmol of CH4 were formed, again indicating that a cyclic operation, with purging steps in between batch experiments, may be the preferred mode of reactor operation. While in all measurements presented in Figures 4 and 5 CH4 was the main product, traces of ethane (<3 ppm) were also identified. The formation of ethane can hint that a C2 mechanism takes place in the CO2‐to‐CH4 photoconversion.

Figure 5

Effect of the addition of H2 on photocatalytic CO2 reduction under batch reaction conditions.

The detrimental, or at best inert, role of H2 in the reaction indicates that it is not an effective proton source. Potentially, TiO2 does not feature sites for H2 splitting, or H2 is not an efficient hole trap on TiO2. The latter would also explain why H2 suppresses CO2 reduction only partially, and not completely. To ensure that the tested photocatalysts remain stable during the respective experiments an XRD analysis was performed for the P25 used in iPrOH oxidation and the one used in the study of the effect of H2O and H2 in CO2 photoreduction. As it can be seen in Figure S5 (Supporting Information), the X‐ray diffractograms reveal that the photocatalysts were not structurally affected by the applied experimental conditions.

Together with our previous results on IrOx/TiO2 (see above), the obtained results evidence that the removal of holes from TiO2 suppresses the CO2 reduction reaction to CH4 as a final product. Since O2 is not formed in any of the reactions studied here, a rapid reoxidation of the formed hydrocarbons cannot explain the observations. Instead, a direct participation of holes in the reaction pathway is the more likely scenario.

At a first glance, these results seem implausible. CO2 contains the C atom in the highest oxidation state, hence further oxidation is impossible. Based on extensive EPR experiments, Shkrob et al. have suggested that in the course of CO2 reduction, various organic molecules are formed and reoxidized on the surface, whereby some stable C2 intermediates, namely glyoxal, glycolaldehyde and acetaldehyde are also being formed. Only the latter are eventually converted to liberate CH4, which cannot be obtained on any pathway involving only C1 intermediates. The final reaction steps of this sequence involve the reaction of acetaldehyde with a photogenerated hole, to liberate CO and a methyl radical. The latter then abstracts a H atom from a suitable donor, to eventually generate CH4. Alternatively, glycolaldehyde may react with a hole to CO and a hydroxymethylene radical, whereby the latter first disproportionates to methanol (MeOH) and formaldehyde. MeOH then reacts with formate to methylformate that can also liberate a methyl radical which eventually ends up in CH4. Both pathways involve holes, so the more efficiently the holes are removed from TiO2, the more efficient CH4 formation will be prevented. The reaction sequence by Shkrob et al. has been evaluated for (frozen) acidic aqueous solution of TiO2, but evidence for a similar mechanism in the gas‐solid process has previously been obtained by us, in particular with respect to the involved reaction of acetaldehyde. Another possibility may be an involvement of carbonates or bicarbonates, which have previously been found to compete with H2O for the photogenerated holes. Although we observed in our previous study a negative influence of a stabilization of carbonates on the surface of TiO2 by Na doping, we also clearly saw a change in the coordination of the carbonate and carboxylate species to the surface. It is, thus, possible that formerly active intermediates in the reaction pathway have been transformed to excessively stable inactive spectator species by the surface doping with Na.

The consecutive reactions under carbon pool formation (Figures 4 and 5) require additional consideration. If gaseous H2O is not fed to the reactor, the adsorbed water will be consumed, but not be replaced. So, the surface will become depleted of protons, which eventually also inhibits CH4 formation. Hence, an addition of H2O after carbon pool formation has taken place has a beneficial effect. H2 addition cannot compensate for this loss of protons (Figure 5), but affects the reaction negatively, if it does so at all.

All in all, the results suggest that H2O is still the best co‐reactant (“reductant”) for CO2 from those studied here (H2O, alcohol, H2), although yields remain low. H2O should also be a hole scavenger, but it is known that this reaction is unfavorable on TiO2, most obviously evidenced by the inability of TiO2 to evolve O2. In the light of the observations made here, the slow and unfavorable reaction of water with photogenerated holes may be the primary reason why it is the only effective co‐reactant. This also indicates that H2O is possibly not actively participating in the photoreaction but is instead deprotonated by a suitably reactive (i. e. Brønsted basic) surface intermediate. Alternatively, hydroxyl groups on the surface of TiO2 might be the only efficient proton donors, and the dissociative adsorption of H2O at oxygen vacancies may recreate some of them. This would also be in accordance with the overall function of TiO2 as oxygen‐scavenging reaction partner as suggested previously.

Further work, particularly involving in situ optical and vibrational spectroscopy, may clarify the open questions with regard to the sequence of surface intermediates and the location of the holes, but the economic sense of such studies must be put to question. It becomes more and more evident that overall photocatalytic CO2 reduction cannot be realized with TiO2, so research on alternative photocatalysts appears more promising in the future.

Conclusion

It has been shown that the addition of alcohols as hole scavengers to TiO2 does not improve photocatalytic CO2 reduction in a high‐purity gas‐solid reaction system. Instead, CO2 reduction is entirely prevented. The blue color of the catalyst after alcohol oxidation confirms the net reduction of TiO2, but even a subsequent CO2 reduction reaction on this blue titania does not lead to the observation of CO2 reduction products. Together with our earlier observations that enforcing O2 evolution by Ir oxide cocatalysts inhibits CO2 reduction, the results are a strong indication that holes have a pivotal role in the CO2 reduction mechanism to CH4, possibly by oxidizing intermediates such as acetaldehyde and glycolaldehyde. In addition, it has been confirmed that CH4 yields are higher in the absence of gaseous H2O, but after the formation of a carbon pool on the surface H2O is required as a proton donor. H2 is a less efficient proton donor, most probably because its splitting is not facile on TiO2. The results also indicate that H2O is potentially not an active participant in the reaction, for example, by scavenging holes, but is possibly deprotonated by a sufficiently reactive carbon‐based surface intermediate.

Experimental Section

Two different generations of our home‐made high‐purity photoreactor setups were used for the experiments as explained in Ref. [7]. A brief description is provided in the Supporting Information. In all studies, the samples (in powder or in pellet form) were irradiated by a 200 W Hg‐Xe arc lamp (Newport) for a total duration of 6 h. A water‐based IR filter was utilized in the light pathway to avoid excessively heating up of the sample. Reactions were run in batch mode with an initial pressure of 1.5 bar inside the reactor. Gaseous samples were collected every 45 min and were analyzed using gas chromatography.

To study the reaction of EtOH on bare TiO2 (P25), the oxide in powder form (70 mg) was first impregnated with a few drops of a highly dilute aqueous EtOH solution (0.5 ml EtOH in 500 ml H2O), similar to the procedure with MeOH described in Ref. [4b]. After drying in air, the sample was introduced into the photoreactor. Subsequently, the photoreaction was either carried out in a pure He atmosphere, or in the presence of 1.5 % CO2.

For the study of the effect of iPrOH on TiO2, 1 g of P25 was pressed at 2 bar for 15 min to form a dense pellet (3 cm in diameter). This pellet was inserted into the reaction chamber of the high‐purity photoreactor and 100 μl of iPrOH (99.5 %, extra dry over molecular sieve, Acros Organics) were added dropwise onto the pellet. The reaction chamber was then flushed with the appropriate gas mixture (pure He or 1.5 % CO2 in He) for 2 h until no O2 could be detected. Afterwards the reactor was filled with the gas mixture up to a final pressure of 1.5 bar and was further operated under batch conditions. In the study of H2 and/or H2O dosing in CO2 photoreduction, 50 mg of TiO2 (P25) where spread on the bottom of a specially designed quartz plate and placed inside the reaction chamber. In measurements where H2O needed to be present, the selected gas mixture (pure He, 1.5 % CO2 in He, or 1 % H2) passed through a saturator set at 5 °C (calculated H2O concentration: 6,000 ppm) before entering the reaction chamber.

Conflict of interest

The authors declare no conflict of interest.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Click here for additional data file.