Selective CO Production by Photoelectrochemical Methane Oxidation on TiO2.

The inertness of the C-H bond in CH4 poses significant challenges to selective CH4 oxidation, which often proceeds all the way to CO2 once activated. Selective oxidation of CH4 to high-value industrial chemicals such as CO or CH3OH remains a challenge. Presently, the main methods to activate CH4 oxidation include thermochemical, electrochemical, and photocatalytic reactions. Of them, photocatalytic reactions hold great promise for practical applications but have been poorly studied. Existing demonstrations of photocatalytic CH4 oxidation exhibit limited control over the product selectivity, with CO2 as the most common product. The yield of CO or other hydrocarbons is too low to be of any practical value. In this work, we show that highly selective production of CO by CH4 oxidation can be achieved by a photoelectrochemical (PEC) approach. Under our experimental conditions, the highest yield for CO production was 81.9%. The substrate we used was TiO2 grown by atomic layer deposition (ALD), which features high concentrations of Ti3+ species. The selectivity toward CO was found to be highly sensitive to the substrate types, with significantly lower yield on P25 or commercial anatase TiO2 substrates. Moreover, our results revealed that the selectivity toward CO also depends on the applied potentials. Based on the experimental results, we proposed a reaction mechanism that involves synergistic effects by adjacent Ti sites on TiO2. Spectroscopic characterization and computational studies provide critical evidence to support the mechanism. Furthermore, the synergistic effect was found to parallel heterogeneous CO2 reduction mechanisms. Our results not only present a new route to selective CH4 oxidation, but also highlight the importance of mechanistic understandings in advancing heterogeneous catalysis.

Introduction

Steam reforming of methane (SRM) is an important process that produces H2 and syngas, which are key feedstocks for downstream chemical processes such as Fischer–Tropsch synthesis.[1,2] Despite the scale, SRM faces critical challenges, including catalyst deactivation and high energy consumption, most of which are associated with the high temperatures needed to activate the C–H bond. To solve these issues, one may meet the energy needs for CH4 activation by electrochemical or photochemical instead of thermochemical activation.[3−11] Indeed, promising results have already been obtained. For instance, Surendranath et al. recently reported selective CH4 conversion toward methanol derivatives by electrochemically turning over a Pd-based molecular catalyst in concentrated H2SO4.[10] By comparison, photochemical CH4 oxidation is less developed. The pioneering work by Yoshida et al. succeeded in oxidizing CH4 by photocatalysts, but mostly produced CO2 instead of CO or other valuable products.[12−17] An important reason is the lack of understanding of the reaction mechanisms by these photocatalytic processes.

More broadly, poor understanding of reaction mechanisms, especially at the molecular level, is a critical weakness of researches on heterogeneous catalysis.[18,19] In recognition of this deficiency, great research efforts have been made to unravel the mechanisms that govern the selectivity of heterogeneous catalytic reactions.[20,21] Within the context of CH4 oxidation, the demonstrations of colloidal PdAu nanoparticles catalysts with >98% selectivity toward CH3OH formation highlight the importance of initial activation of CH4 by •OH.[22]In situ XANES (X-ray absorption near edge spectroscopy) and Fourier-transform infrared (FTIR) spectroscopic studies of copper-exchanged mordenite zeolite (CuMOR) have revealed that the bridging O in Cu–O–Cu is key to CH3OH production.[23] Notably, these examples were all thermochemical transformations. Recent advances in a parallel field of CO2 reduction studies also exemplify the importance of mechanistic understanding to heterogeneous catalysis.[24−30]

Inspired by these previous efforts and the importance of CH4 oxidation, we studied this reaction under photoelectrochemical (PEC) conditions, wherein the chemical reaction proceeds at the semiconductor photoelectrode/electrolyte interface under light with an applied bias. The externally applied potential creates an extra electric field to facilitate charge separation between photogenerated electrons and holes. Our goal was to enable the reaction at room temperature so as to address the critical challenges faced by conventional SRM. It was found that the product selectivity is highly sensitive to the applied potentials, and photoexcitation is critical to the selective formation of CO. Detailed mechanistic studies by spectroscopic and computational methods revealed that surface O• radical formation is key to the observed results, and that the synergistic effect between adjacent Ti sites plays a critical role. Our results shed new light on the interactions of C species with metal centers of heterogeneous catalyst substrates.[31−33] They highlight the importance of synergistic effects in defining the reaction routes.[34−36] The insights may find applications in CO2 reduction, as well.[37]

Results and Discussion

A typical representation of our experimental setup is schematically shown in Figure a, where a 3-electrode configuration was employed. A TiO2 photoelectrode (ca. 50 nm in thickness) was used as the working electrode.[38] A Pt wire served as the counter electrode, and the reference electrode was a saturated calomel electrode (see the Supporting Information, SI, for more details). In comparison with previous studies on photochemical oxidation of CH4,[7,8,12−17] we introduced two key innovations. First, atomic-layer-deposition-grown (ALD-grown) TiO2 was studied, which shows critical differences in product selectivity when compared with TiO2 prepared by other methods. Second, the separation of the oxidation (TiO2 working electrode) and reduction (Pt counter electrode) reactions permitted us to focus on CH4 oxidation without confounding effects by the reduction reactions of H2 generation. The light source used for this body of research was a UV lamp (λ = 254 nm; intensity 0.1 mW/cm2). The electrolyte was 1.0 M NaOH (pH 13.6). CH4 was bubbled into the electrolyte under standard conditions. The solubility is expected to be ca. 0.0010 mol/kg.[39,40] A representative current voltage curve is plotted in Figure S2 in the SI, where the turn-on potential was 0.2 V vs RHE (reversible hydrogen electrode). A stability test was carried out at 0.6 V vs RHE (Figure S3), and no obvious decay was observed during the first 15 h. Under our experimental conditions, the measured anodic photocurrents could be due to CH4 oxidation, H2O oxidation, or both. Care was taken to ensure that no other parasitic reactions (such as oxidation of organic impurities) contributed to the measured currents. The products of CH4 oxidation were detected using a GC-MS instrument (Shimadzu QP2010 Ultra, with a Carboxen 1010 PLOT column). The products of H2O oxidation were quantified by a Clark-type BOD oxygen electrode (Thermo Scientific 9708 DOP).[41] No production of H2O2, which was examined by iodine clock reactions, was detected (see details in the SI). The Vapp’s were limited within 0.4–1.2 V vs RHE for this body of study, so that CO oxidation was ruled out (Figure S4 in the SI). The possibility that H2 oxidation contributes to the measured photocurrent was ruled out by the measurement of ca. 100% Faradaic efficiency of H2 detection (by GC-MS, see Table S2). For a typical data set as shown in Figure b, at least 3 different samples were tested for each CO or O2 detection, and the statistical variation (the standard deviation at each Vapp was less than 5.1%) was insignificant. Taken together, the selectivity trend as shown in Figures and 3 is statistically significant.

Figure 1

(a) Schematic illustration of selective CH4 oxidation to CO on a TiO2 photoelectrode, starting with charge separation between O2– and Ti4+ to produce –•O–Ti3+ upon illumination. The separation of the redox half reactions permitted us to focus on CH4 oxidation. (b) Dependence of the CO efficiency and selectivity on the applied potentials (left axis, efficiency, %; right axis, selectivity of CO over all carbonaceous products, %). PEC bulk electrolysis was conducted on ALD TiO2 in CH4-saturated 1.0 M NaOH electrolyte at the corresponding applied potentials.

Figure 3

Product selectivity dependence on substrates. (a) Comparison of CO selectivity on three different types of TiO2 samples. PEC bulk electrolysis was conducted in CH4-saturated 1.0 M NaOH electrolyte at 0.6 V vs RHE. (b) EPR spectra showing the EPR signals consistent with the presence of Ti3+ in the three samples (the color coding of the spectra is the same as the left panel). Of them, ALD TiO2 (Sample 1) features the highest intensity (g⊥ = 1.99 and g∥ = 1.96).

For the comparison of reaction selectivity under different conditions, we present the Faradaic efficiencies in Figure b. As can be seen, the detected CO and O2 collectively account for 80–90% of the overall charges as measured by PEC bulk electrolysis. The unaccounted charges were attributed to the production of carbonates, the existence of which was confirmed by spectroscopic measurements (Figures S6 and S7; vide infra). Although the amount of carbonates was too insignificant to quantify by weighing precipitates with Ba(II) (more details in the SI), the yield is estimated based on the balance of charge transferred (Figure b). It is noted that, for this body of research, the most important quantitative information is the yield of CO. We see from Figures b and 3a the following trends. First, the yield of CO is highly sensitive to the applied potentials. At 0.4 V vs RHE, CO accounts for 81.9% of all charges other than O2 formation. Given the low concentration of CH4 (0.0010 mol/kg) under STP,[39,40] H2O oxidation is expected to compete favorably in receiving photogenerated holes, which explains the relatively high yield of O2. The yield of CH4 oxidation (such as CO) is expected to be readily improved by increasing the partial pressure of CH4.[10] By comparison, at 1.2 V vs RHE, CO only accounts for 24.7% of all charges other than O2 formation. Second, the yield is sensitive to the type of TiO2 substrate used. For ALD-grown TiO2, at 0.6 V vs RHE, CO accounts for 62.7% of the total charges other than O2 formation; the yield was <4% for commercially obtained anatase TiO2 or P25 TiO2 nanoparticles. Third, we observed no CH4 oxidation products in the absence of illumination, applied potentials, or both (see Table S1 and Scheme S1). The observation is consistent with the understanding that the activation of the first C–H bond requires significant energy input. It further highlights the importance of photoexcitation for CH4 oxidation.

Significant research efforts have been attracted to understanding the activation of CH4. For heterogeneous catalysis, homolytic C—H dissociation has been reported as one of the activation mechanisms by hydrogen (H) abstraction on an active oxygen (O–) center, forming methyl radicals (•CH3).[33] How to effectively regenerate the surface-active oxygen species for continuous CH4 activation was a major challenge. A solution was found in TiO2 under photocatalytic conditions. It has been established by previous research that photo-oxidation on TiO2 starts with charge separation between O2– and Ti4+, producing –•O—Ti3+.[42,43] The oxygen radical then attacks the C—H bond in CH4 to yield Ti—O—CH3. Subsequent reactions turn this species back to Ti4+—–OH (or Ti4+—O2– in the deprotonated form), ready for regeneration of the reactive site by photoexcitation. The oxygen radical then attacks the C—H bond in CH4 to yield surface-adsorbed methoxy (OCH3) species.[22] The understanding is summarized in Figure . The evolution of this methoxy species and surface-adsorbed CH4 upon light illumination was detected by in situ Raman spectroscopy. The applied potential facilitates electron transport away from the Ti site through the conduction band of TiO2.[44] Subsequent photoexcited electron transfers between O and Ti drive the reaction to reach a quasistable intermediate state of Ti—O=CH2, which was observed by in operando attenuated total reflection–Fourier-transform infrared spectroscopy (ATR–FTIR). To explain the dependence of CO product selectivity on the type of TiO2 substrates and the applied potentials, we hypothesize that the mechanism of Ti—O=CH2 oxidation depends on the nature of the adjacent Ti sites. In the presence of Ti3+, it favors the formation of a Ti3+—C bond, leading to selective production of CO. An increase in CO selectivity was observed at higher light intensity, since more Ti3+ sites would be induced and stabilized (see Figure S8 in the SI). Otherwise, the lack of Ti3+ sites would favor the formation of a Ti4+—O—C bond, resulting in the formation of carbonates and, hence, complete oxidation of CH4. The proposed mechanism shown in Figure is inspired by recent advances on CO2 reduction and supported by the detection of key intermediates.[26] The mechanism is also supported by computational calculations. Below, we present detailed evidence to support the proposed reaction mechanisms.

Figure 2

Proposed mechanisms of photo-oxidation of CH4 on TiO2. The species highlighted by dotted circles have been confirmed by various spectroscopic techniques. The key distinguishing step is identified at the bottom of the catalytic cycle, where the synergistic effect between two adjacent Ti sites promotes the switching of C=O—Ti to O=C—Ti, leading to selective formation of CO. Alternatively, in the absence of such a switching, CO2 is the preferred oxidation product.

Our first task was to examine the difference between different TiO2 substrates. Three prototypical TiO2 substrates, namely, TiO2 by ALD (Sample 1), anatase TiO2 (Sigma-Aldrich, 99.8% trace metal basis; Sample 2), and P25 (Evonik Industries, Aeroxide TiO2 P25; Sample 3), were compared. Electron paramagnetic resonance (EPR) spectroscopy was employed for this purpose. While the true nature of the Ti3+ sites may not be fully understood until later, a theoretical study by Selloni et al. has shed important light on the problem. Their results showed that photogenerated charges become self-trapped on Ti, reducing the site to become Ti3+. EPR detects unpaired electrons, which reports on the localized Ti3+ sites in TiO2.[45] As shown in Figure b, a significantly higher intensity of the EPR signal associated with Ti3+ was observed for Sample 1 than the other samples,[8,46,47] which is indicative of electrons trapped at the Ti3+ sites (see Table S3 for comparison of Ti3+ concentration for all three samples). It is important to note that the EPR data as shown in Figure b were obtained in an atmosphere with CH4 (97%) and H2O (3%), immediately after illumination (<120 s delay; see the SI for more details). Our results indicate a clear correlation of the Ti3+ concentration as detected by EPR with the CO selectivity as detected by GC-MS (Figure a). That is, reaction routes involving Ti3+ would be favored when Sample 1 is used as the substrate, which is the CO pathway as shown in Figure . Further research would be needed to fully explain the high concentration of unpaired electrons in the trapped sites in Sample 1. Possible reasons include high concentration of O vacancies or incomplete removal of the ligands (isopropoxide).[48] It is, nevertheless, emphasized that Sample 1 has been shown highly active and stable toward PEC water splitting, suggesting that the high concentration of Ti3+ does not undermine its optoelectronic properties.[38]

Next, we carried out a multimodal study to detect several key intermediates as shown in Figure . The techniques we employed include in situ Raman and in operando ATR–FTIR measurements. Among them, in situ Raman allowed us to detect Ti–O–CH3 (Figure a), which is a product of the initial oxidation of CH4 after photoexcitation of electrons from O 2p orbitals to Ti 3d orbitals. The Raman spectra as shown in Figure a were collected in a homemade quartz cell filled with CH4 (97%) and H2O (3%) at 1 atm. Without photoexcitation, we only observed a characteristic peak at 2917 cm–1 that is due to free CH4 molecules.[49] With photoexcitation, a shoulder peak at 2905 cm–1 appeared, which is ascribed to adsorbed CH4.[49] The red shift in comparison with free CH4 is due to a weakened bond as a result of adsorption to the TiO2 substrate. Most prominently, two satellite peaks at 2850 and 2965 cm–1 gradually evolved during the in situ photoexcitation experiments. Similar peaks have been assigned to the C–H symmetric stretching and asymmetric stretching modes of O–CH3 on TiO2, respectively, in the literature.[50] These results provide strong support for the conclusion that we have obtained CH3–O–Ti bonds upon photoexcitation under our experimental conditions. To rule out the possibility that these two satellite peaks are due to impurities, we performed Raman characterization using isotope-labeled methane (CD4). The characteristic Raman shifts of the C–D stretching modes (1800–2500 cm–1) are distinctly different from the C–H stretching modes (2800–3000 cm–1).[51] Similar satellite peaks were observed (Figure S11), providing additional strong support for our interpretation. The proposed reaction mechanism that governs the formation of CH3–O–Ti is illustrated in the top panel of Figure . Alternatively, Ti–O–CH3 may be generated by an indirect oxidation of CH4 by water oxidation intermediates, such as H2O2. A series of control experiments have been performed by adding H2O2 to the electrolyte under different conditions (more details in Table S4). The results indicate that the last suggested alternative route is unlikely.

Figure 4

In situ Raman and FTIR detection of key reaction intermediates. (a) In situ Raman spectra of samples in dark (left) and in light (right) confirm the H3C—O—Ti species as a result of the initial oxidation of CH4. No electrical potential was applied. (b) Evolution of H2C=O—Ti, Ti—CO, and HCOO(H)—Ti surface species as a function of time as detected by FTIR. The IR data were collected at Vapp = 0.3 V (vs Pt counter electrode) under illumination.

Next, we conducted in operando ATR–FTIR measurements to probe the reaction intermediates in the subsequent oxidation steps. Similar to the implementations by Hamann et al.,[52] we carried out the experiments by pressing the working electrode against the IR-transparent Si ATR crystal (Scheme S3 in the SI). The electrolyte sandwiched between TiO2 and Si was saturated with either CH4 or N2 (as a control). The IR data were collected at a constant applied potential of 0.3 V (vs Pt counter electrode) under illumination. They are presented in Figure b after correction for background absorption (more details in the Experimental Details section in the SI). We focused our attention on the region between 1700 and 2300 cm–1 because the IR absorption beyond 2700 cm–1 is cut off by a long-pass IR filter in front of the detector, and other regions (below 1600 cm–1, 2300–2700 cm–1) are interfered by broad H2O and CO2 absorptions. Within the observation window, three distinct peaks at 1735, 1763, and 2065 cm–1 were observed, but only in the presence of CH4. All three peaks exhibited an obvious dependence on the illumination duration, strongly suggesting that they are due to photo-oxidation reactions on the surface of TiO2. Of them, the most prominent peak at 1763 cm–1 is assigned to the C=O stretch due to surface-adsorbed formic acid species.[53−56] The peak at 1735 cm–1 is assigned to the C=O stretch due to surface-adsorbed formaldehyde species, where the oxygen adatoms are coordinated to Ti atoms.[53,56,57] The peak at 2065 cm–1 is assigned to the C≡O stretch due to surface-adsorbed CO species.[58,59] The assignments of these peaks are supported by literature reports. The importance of these reaction intermediates to the overall reactions is illustrated in the proposed catalytic cycle (Figure ).

The proposed mechanism is supported by DFT calculations. We found that the proposed intermediates, including CH3O, CH2O, CHO, and HCOO, are all stable on the TiO2 surface. Using the computational hydrogen electrode,[60] we calculated Gibbs free energy changes (ΔG) for both the CO and carbonate pathways (see more details in the SI). As shown in Figure , the CO pathway goes through the CHO intermediate, while the carbonate pathway proceeds via the HCOO intermediate. The Gibbs free energy changes (ΔG) for the CHO intermediate and HCOO intermediate are −2.21 and −2.70 eV, respectively. Importantly, we found that ΔG for the CO pathway is 0.31 eV, much lower than that for the carbonate pathway (1.98 eV). ΔG for all species as reported here is calculated relative to the formaldehyde species (3 in Figure ). This indicates that the CO pathway is thermodynamically more favorable than the carbonate pathway, thereby supporting our observation that much more CO is produced than carbonate at low applied potentials (Figure b).

Our investigations of the catalytic cycle highlight the importance of the Ti3+ site adjacent to Ti—O=CH2 in promoting CO formation. This understanding is consistent with the potential dependence we observed. At relatively high applied potentials (e.g., 1.2 V vs RHE), photoexcited electrons are readily removed from the Ti3+ site, which would favor the pathway leading to CO2 (or carbonate) formation. At less positive potentials (e.g., 0.4 V vs RHE), the higher concentration of Ti3+ sites facilitates the formation of Ti—C(H)=O bonds, leading to CO formation. In essence, the switching between Ti—O—C and Ti—C—O bonding appears to be critical to the selectivity for CO production. The conclusion reminds us of parallel studies on CO2 reduction by metal catalysts. It is well accepted that the reaction starts with a M—C bond (where M represents the metal catalytic center). Computational results suggest that formation of more reduced products such as CH4 or CH3OH often requires switching from M—C—O to M—O—C bonding (Figure );[24] otherwise, the reaction favors the formation of CO. While experimental evidence to explicitly verify this understanding is still missing, multiple computational studies provide strong supports for this hypothesis.[24,61,62] The switching likely proceeds through a concerted mechanism with adjacent metal sites, in a similar fashion to our proposed mechanism as shown in Figure . We suggest that insights into CO2 reduction could be drawn from studies of CH4 oxidation, given the similarities of the key steps.

Figure 5

Switching between Ti–O–C and Ti–C–O bonding is critical to the selective production of CO from CH4 (top). In parallel studies of CO2 reduction, the selective production of CH4 or CH3OH may proceed by switching between M–C–O and M–O–C bonding in a similar fashion (bottom). The key steps are highlighted in the dotted boxes.

In conclusion, we have observed highly selective CO production by CH4 photoelectrochemical oxidation. Only with a moderately positive applied potentials (between 0.4 and 1.2 V vs RHE) and on ALD-grown TiO2 did we obtain a high yield of CO production. EPR studies revealed that the key difference between ALD-grown TiO2 and commercially obtained samples is the concentration of Ti3+ sites. We propose that the adjacent Ti3+ sites to the Ti—C=O bond are key to the high selectivity for CO formation. Spectroscopic evidence from in situ Raman and FTIR spectroscopy provides strong support to the proposed mechanism. The feasibility of the proposed mechanism is supported by computational studies as well. The proposed mechanism draws similarities to heterogeneous CO2 reduction reactions. Together, the results highlight the importance of synergistic effects in promoting complex chemical reactions such as CH4 oxidation and CO2 reduction. The photo-oxidation reaction as presented here opens up new doors toward highly selective reforming of CH4 for high-value industrial chemicals such as CO.