Global energy demands and climate change have stimulated intense research on the sustainable transformation of greenhouse gas CO2 to chemicals and fuels. Among various approaches that enable these CO2 conversions, of particular interest is the gas‐phase solar powered heterogeneous catalytic reduction of CO2 to CO, dubbed the photocatalytic reverse water gas shift reaction. This process can enable very high CO2 conversion rates, and is compatible with existing chemical and petrochemical industry infrastructure, implying that it can be easily scaled, integrated, and commercialized. Key to achieving the practical implementation of this promising renewable‐energy technology is to develop efficient, robust, and scalable photocatalysts for the hydrogenation of CO2. Despite tremendous efforts, it remains a great challenge since the CO2 molecule is fully oxidized and extremely stable.
Recently, semiconducting metal oxides have been widely investigated as photocatalysts for CO2 hydrogenation mainly because of their photochemical stability. However, limited by their performance, much effort has been devoted to enhancing their photocatalytic activity. In this context, Indium oxide is one of the most widely used, n‐type transparent conducting metal oxides in electronic, optoelectronic, and optical devices due to its large optical bandgap, low electrical resistivity, and excellent photostability. Our group has recently reported that indium oxide, In2O3−x(OH)y, with engineered oxygen vacancy (electron trapping) and hydroxide (hole trapping) defects, is active for the photocatalytic hydrogenation of CO2 to CO. The CO2 conversion rate to CO however is at best 33.3 µmol g cat−1 h−1, which is too low for its practical implementation in a CO2 refinery.
Herein, we show that heterostructure engineering of In2O3−x(OH)y with other metal oxide semiconductor can dramatically improve its photocatalytic activity. By this means, the CO2 to CO conversion rate was boosted from 33.3 µmol g cat−1 h−1 for pristine In2O3−x(OH)y to as high as 1400 µmol g cat−1 h−1 for heterostructure engineered In2O3−x(OH)y, a remarkable enhancement factor of 44‐fold. Such a dramatic enhancement in conversion rate is a rare event. Furthermore, the heterostructure engineered In2O3−x(OH)y exhibited excellent long‐term, operational photochemical durability.
The heterostructure between the two semiconductors In2O3−x(OH)y and Nb2O5 nanorods (Figure S1, Supporting Information) was constructed by a facile, two‐step process (Figure 1a). First, the In(OH)3@Nb2O5 precursor was prepared by growth of In(OH)3 on Nb2O5 in an ethanol/water mixture. Second, the desired In2O3−x(OH)y@Nb2O5 heterostructure was formed by thermally induced dehydroxylation of In(OH)3@Nb2O5 at 250 °C in air for a predetermined time. To understand the physical and chemical properties of heterostructure engineered In2O3−x(OH)y, we prepared pristine In2O3−x(OH)y as a control reference sample. The synthetic procedure for making pristine In2O3−x(OH)y is similar to In2O3−x(OH)y@Nb2O5, except there is no Nb2O5 involved in the preparation process.
Enlarge this image.a) Illustration of the synthesis of In2O3−x(OH)y nanocrystals grown on the surface of Nb2O5 nanorods. b) PXRD patterns of pristine In2O3−x(OH)y and Nb2O5, a chemical heterostructure In2O3−x(OH)y@Nb2O5. c) XPS and curve fitting for the In 3d5/2 and In 3d3/2 core level, spin–orbit split binding energies of pristine In2O3−x(OH)y, a physical mixture In2O3−x(OH)y/Nb2O5 and a representative chemical heterostructure In2O3−x(OH)y@Nb2O5, S3 sample. d) O 1s XPS spectrum of S3 and physical mixture sample. e) Deconvolution of the high‐resolution O1s XPS spectrum of the S3 sample. The main peak at 530.3 eV is attributed to indium oxide. The two additional shoulder peaks at 531.7 and 532.5 eV are attributed to oxygen vacancies in the structure and surface OH groups, respectively.
Thermogravimetric analysis showed that the behavior of supported In(OH)3 in In(OH)3@Nb2O5 is similar to In(OH)3 alone (Figures S2 and S3, Supporting Information). One observes a sharp weight loss after the temperature reached ≈200 °C, accompanied by an endothermic peak at ≈240 °C, which corresponds to the conversion of cubic In(OH)3 into cubic, defect‐laden bixbyite In2O3−x(OH)y through controlled dehydroxylation of the number of hydroxyl groups with concomitant loss of water. A set of three different In2O3−x(OH)y @Nb2O5 heterostructure samples were fabricated from the corresponding precursor In(OH)3@Nb2O5 (Figure S4, Supporting Information). These samples are denoted S1–S4, where the label 1–4 refers to samples with gradually increasing In2O3−x(OH)y content. The amount of In2O3−x(OH)y in S1–S4 were 9.79%, 15.49%, 23.21%, and 30.7%, respectively, as quantified by inductively coupled plasma‐atomic emission spectroscopy, For comparison, a physically blended mixture of nanocrystalline In2O3−x(OH)y and Nb2O5 nanorods was also prepared with the content of In2O3−x(OH)y in the mixture adjusted to be identical with that in S3.
Powder X‐ray diffraction (PXRD) peaks of In2O3−x(OH)y@Nb2O5, pristine In2O3−x(OH)y, and Nb2O5 are shown in Figure b, all of the peaks of In2O3−x(OH)y in S1–S4 were matched well with pristine In2O3−x(OH)y, indicating that the In2O3−x(OH)y@Nb2O5 are successfully prepared after calcining the precursor of In(OH)3@Nb2O5, at the right temperature for the appropriate time. To explore the effect of the heterostructure on the electronic structure of In2O3−x(OH)y, X‐ray photoelectron spectroscopy (XPS) of the S3, pristine In2O3−x(OH)y, and the physical mixture were conducted (Figure c, and Figure S5, Supporting Information). The In 3d core level spectra of pristine In2O3−x(OH)y and the physical mixture show two peaks located at 442.8 and 450.3 eV, which are attributed to the characteristic spin–orbit split 3d5/2 and 3d3/2, respectively. Interestingly, the 3d5/2 and 3d3/2 peaks of In2O3−x(OH)y in the S3 sample are shifted toward higher binding energy, located at 443.3 and 450.9 eV, respectively (Figure c). This indicates an increase in effective positive charge of the In sites as a result of the heterostructure engineering. The increased population of coordinately unsaturated In sites arising from the larger population of oxygen vacancies seen in the O1s photoemission spectra in In2O3−x(OH)y can be related to the enhanced binding energy. This point was further supported by XPS analysis of O species in these samples. Considering the content of In2O3−x(OH)y in S3 and physical mixture samples is identical, we compared the oxygen vacancy [O] concentrations in the two samples. As shown in Figure d,e, the population of [O] in the S3 was 20.9%, which is higher than that in physical mixture sample (16.6%) (Figure S6, Supporting Information). Later in this article, we demonstrate that such an increase of oxygen vacancies in heterostructure engineered In2O3−x(OH)y enhances the population and lifetime of photoexcited electron–hole pairs and as a result provides a boost to the photocatalytic activity.
We took S3 sample as representative to examine its morphology and chemical heterostructure using high resolution transmission electron microscopy (HRTEM). From Figure 2a, In2O3−x(OH)y nanocrystals with a diameter of ≈4 nm, uniformly distributed on the surface of Nb2O5, are observable. It is noted that the use of ammonia rather than urea as the base hydrolysis source in the synthesis, proved unable to produce heterostructures of In2O3−x(OH)y and Nb2O5. Instead, only two separate normal phases of In2O3−x(OH)y and Nb2O5 resulted (Figure S7, Supporting Information). This observation demonstrates the importance of the controlled nucleation and growth process for making heterostructured In2O3−x(OH)y@Nb2O5 by using the strategy of slow release of base by the decomposition of urea at 80 °C. The HRTEM data show well‐defined heterojunctions between nanocrystalline In2O3−x(OH)y and Nb2O5 nanorods (Figure b). Analysis of the In2O3−x(OH)y nanocrystals on the Nb2O5 nanorods in the HRTEM image reveals a lattice spacing of 4.03 Å corresponding to the (211) plane. Figure c–f shows the scanning transmission electron microscopy (STEM) image and corresponding elemental mappings of Nb, In, and O, respectively. It can be clearly seen that In2O3−x(OH)y was uniformly dispersed on the surface of Nb2O5 nanorod, a favorable requirement for optimized catalytic activity.
Enlarge this image.a) Low‐resolution TEM image of S3. Yellow circle indicates that In2O3−x(OH)y nanocrystal with a diameter of ≈4 nm was grown on the surface of Nb2O5 nanorod. b) HRTEM images of S3. c–f) STEM image and corresponding EFTEM elemental (Nb, In, O) mapping.
All samples exhibited a mesoporous structure (Figure S8, Supporting Information), and the Brunauer–Emmett–Teller (BET) specific surface areas of Nb2O5, S1, S2, S3, and S4 are 84.6, 63.7, 60, 88.8, and 96.6 m2 g−1, respectively, were lower than that of pristine In2O3−x(OH)y (117 m3 g−1) (Figure S9 and Table S1, Supporting Information). Noted also, is that the BET surface areas of S1 and S2 are lower than that of Nb2O5. This can best be explained by the fact that nanocrystalline In2O3−x(OH)y grown within the mesopores of Nb2O5 nanorods, decreases the BET surface areas of S1 and S2 compared to Nb2O5. With further increase of the content of In2O3−x(OH)y in S3, the mesoporous structure of In2O3−x(OH)y improves their surface areas. This result, together with those from the TEM observations can explain the growth process of In2O3−x(OH)y nanocrystal on the surface of Nb2O5 to form heterostructures. First, the indium chloride and urea were adsorbed in the mesopores of Nb2O5. Second, In(OH)3 precursor forms and anchors to the surface of Nb2O5 concomitant with the decomposition of urea to NH3 at 80 °C. Finally, In2O3−x(OH)y nanocrystals form on the surface of Nb2O5 nanorods during the dehydroxylation process at 250 °C.
The mesoporous structure of the as‐synthesized photocatalysts (Figure 3a) is advantageous for fast diffusion of CO2 to catalytic centers thereby improving their performance. Photocatalytic activity was measured using the as‐synthesized samples deposited on borosilicate filter films (Figure b). To ensure the products did not originate from adventitious carbon residues in our samples, isotope labeled 13CO2 authenticated the origin of the products of the reduction reaction.
Enlarge this image.a) Digital photographs of as‐prepared pristine In2O3−x(OH)y, S1‐S4, and Nb2O5. b) Digital photograph of S3 sample deposited on borosilicate filter film for photocatalytic measurement. c) Photocatalytic performance of the listed materials. d) Long‐term photocatalyst stability measurements of the S3 sample.
We observed a remarkable enhancement of the hydrogenation rate of CO2 to CO on passing from S1 to S3 (Figure c), reaching as high as 1400 µmol g cat−1 h−1 for S3, which then decreases for S4. This can be explained by the fact that there is more isolated In2O3−x(OH)y in S4 rather than heterostructures of In2O3−x(OH)y/Nb2O5. The product is completely 13C labeled 13CO without impurities (Figure S10, Supporting Information), demonstrating high selectivity of our photocatalysts for solar powered hydrogenation of CO2 to CO.
Note that Nb2O5 is highly inactive under these photocatalytic‐operating conditions, showing no activity toward CO2 to CO reduction. We also showed there is no CO produced in the dark, confirming the photocatalytic nature of the samples. It is observed that the hydrogenation rate of CO2 to CO over pristine In2O3−x(OH)y and their physical mixture are only 33 and 21 µmol gcat−1 h−1 respectively under the same conditions, much lower than that of the heterostructured samples. These observations strongly support the proposition that heterostructure engineering of photocatalysts described herein provides a new and effective strategy for enhancing the performance of gas‐phase heterogeneous CO2 hydrogenation.
The key for practical applications is the long‐term stability of a photocatalyst. As shown in Figure b, stability testing of S3 for photocatalytic hydrogenation of CO2 to CO showed no decay even after 10 cycles, which involved continuous operations for 40 h. This result confirms that heterostructured In2O3−x(OH)y is stable under operating conditions employed in this study.
In metal oxides, both bulk and surface oxygen vacancies can act as traps for photoexcited electrons, while surface hydroxyl groups can function as traps for photoexcited holes. Efficient electron–hole pair separation is a prerequisite for high performance photocatalysis, and a reaction must have sufficient time to occur before recombination to the electronic ground state. To delve more deeply into the effect of heterostructure engineering on electron–hole separation efficiency, transient state surface photovoltage (TSPV) measurements were undertaken. Details of these TSPV experiments are included in the Supporting Information. We used two different excitation energies corresponding to above bandgap (355 nm) and near‐bandgap (532 nm) excitation to understand how photoexcited charge carriers with different energies interact with the defect states available within the samples. The TSPV spectra measured in air are in Figure 4. They depict the photovoltage responses to be positive for all samples under pulsed laser excitation at 355 nm (Figure a) and 532 nm (Figure S11, Supporting Information).
Enlarge this image.a) Transient state surface photovoltage responses of S3, pristine In2O3−x(OH)y, and a physical mixture of In2O3−x(OH)y and Nb2O5 nanorods, using 355 nm laser excitation. b) Fluorescence decay temporal profiles measured and fit for S3 and pristine In2O3−x(OH)y.
The results obtained under ambient conditions indicate that adsorbed O2 on the surface of the samples capture photogenerated electrons, while the photogenerated holes preferentially diffuse to the collector electrode surface. Noticeably, S3 exhibits a much higher photovoltage compared to the other samples, indicating that more holes accumulate on the surface of In2O3−x(OH)y@Nb2O5 with a noticeably longer lifetime of ≈3 ms. As mentioned earlier, XPS results for the core level oxygen demonstrate that heterostructured In2O3−x(OH)y@Nb2O5 creates more oxygen vacancies than pristine In2O3−x(OH)y. Thus, the larger population of oxygen vacancies in In2O3−x(OH)y for S3 capture the photogenerated electrons, while the holes are captured by the hydroxide, thus enhancing electron–hole separation efficiency and photocatalytic activity.
Additional evidence for the enhanced electron–hole separation efficiency of S3 seen by transient state surface photovoltage measurements derives from time‐resolved fluorescence spectroscopy measurements of S3, also shown in Figure b. The peak of the steady‐state fluorescence of S3 is located at 410 nm (Figure S12, Supporting Information). This emission is associated with the recombination of the photogenerated hole trapped at mid‐gap hydroxide defects with the photogenerated electron trapped at sub‐bandgap O vacancy defects in In2O3−x(OH)y. The radiative lifetime was extracted by exponential fitting using the Levenberg–Marquardt method: y (t) = y0 + A1exp (t/τ1) + A2exp (t/τ2), where τ1 and τ2 are the decay times, of the time‐resolved fluorescence decay curves for S3 (τ1 and τ2 are 0.38 and 4.05 ns, respectively), which are longer than in pristine In2O3−x(OH)y (τ1 and τ2 are 0.35 and 3.12 ns, respectively). These results provide additional evidence for enhanced electron–hole separation in the heterostructure In2O3−x(OH)y@Nb2O5.
At the interface between the In2O3−x(OH)y and Nb2O5 semiconductors, one expects charge‐transfer and the formation of a space charge region to occur as the Fermi level is equilibrated across the sample. This electronic structure model is substantiated by measurement of the Fermi levels of S3, pristine In2O3−x(OH)y and Nb2O5 nanorods at −4.03, −3.84, and −4.46 eV, respectively, by means of ultraviolet photoelectron spectroscopy, UPS (Figure S13, Supporting Information). The UPS results provide evidence for an electronic band model in which electrons flow from Nb2O5 to In2O3−x(OH)y across the heterostructure in S3. The heterojunction alignment whereby electrons flow from Nb2O5 to In2O3−x(OH)y is also confirmed through band alignment calculations (Figure S14, Supporting Information). Oxygen vacancies formed in the heterostructures of In2O3−x(OH)y@Nb2O5 in S3 serve as reservoirs for photoexcited electrons, while the hydroxides act as reservoirs for photoexcited holes. The real space separation of the charge carriers stabilizes the photoexcited state, resulting in longer carrier lifetimes in In2O3−x(OH)y@Nb2O5 that provide a significant boost in photocatalytic activity for the solar powered reverse water gas shift reaction CO2 + H2 → CO + H2O.
From a scientific perspective, challenges remain in the quest to understand how heterostructures between metal oxide semiconductors control photocatalytic performance. These challenges include disentangling the effects of heterostructure interfaces and interfacial charge transfer on defect populations; how these affect the lifetimes of photoexcited electron–hole pairs; the effect on the acidity and basicity of surface frustrated Lewis pairs, and their corresponding activity toward the hydrogenation of carbon dioxide. In terms of practical applications, significant improvements in conversion and energy efficiency together with a techno‐economic life cycle assessment will be required before the class of photocatalysts described in this paper is ready for engineering a pilot scale solar fuels process.
Acknowledgements
G.A.O. acknowledges the financial support of the following agencies: Ontario Ministry of Research and Innovation (MRI); Ministry of Economic Development, Employment and Infrastructure (MEDI); Ministry of the Environment and Climate Change (MOECC); Best in Science (BIS); Ministry of Research Innovation and Science (MRIS) Low Carbon Innovation Fund (LCIF); Ontario Centre of Excellence Solutions 2030 Challenge Fund; Alexander von Humboldt Foundation; Imperial Oil; University of Toronto Connaught Innovation Fund (CIF); Connaught Global Challenge (CGC) Fund; Natural Sciences and Engineering Research Council of Canada (NSERC). D.D.P. is the Celestica Chair in Materials for Microelectronics. H.W. acknowledges the financial support from the Nankai University and National Science Foundation of China (Nos. 21875119 and 51802208). This research used equipment funded by the ARC LIEF Grant No. LE120100104 and located at the UOW Electron Microscopy Centre.
Conflict of Interest
The authors declare no conflict of interest.
© 2019. This work is published under http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Details
1 Key Laboratory of Functional Polymer Materials of the Ministry of Education, Institute of Polymer Chemistry, College of Chemistry, Nankai University, Tianjin, P. R. China
2 Materials Chemistry and Nanochemistry Research Group, Solar Fuels Cluster, Departments of Chemistry, University of Toronto, Toronto, ON, Canada; Department of Materials Science and Engineering, University of Toronto, Toronto, ON, Canada
3 Materials Chemistry and Nanochemistry Research Group, Solar Fuels Cluster, Departments of Chemistry, University of Toronto, Toronto, ON, Canada; Institute of Functional Nano & Soft Materials (FUNSOM), Jiangsu Key Laboratory for Carbon‐Based Functional Materials & Devices, Soochow University, Suzhou, Jiangsu, P. R. China
4 SciML, Scientific Computing Department, Rutherford Appleton Laboratory, Didcot, UK
5 Materials Chemistry and Nanochemistry Research Group, Solar Fuels Cluster, Departments of Chemistry, University of Toronto, Toronto, ON, Canada
6 UOW Electron Microscopy Centre, University of Wollongong, Wollongong, New South Wales, Australia
7 Institute of Functional Nano & Soft Materials (FUNSOM), Jiangsu Key Laboratory for Carbon‐Based Functional Materials & Devices, Soochow University, Suzhou, Jiangsu, P. R. China
8 Department of Materials Science and Engineering, University of Toronto, Toronto, ON, Canada
9 Key Laboratory of Functional Inorganic Material Chemistry, Ministry of Education School of Chemistry and Materials Science, International Joint Research Center for Catalytic Technology, Heilongjiang University, Harbin, P. R. China
10 Department of Materials, Imperial College London, London, UK; Department of Materials Science and Engineering, Yonsei University, Seoul, Korea
11 Institute for Micro Process Engineering, Karlsruhe Institute of Technology, Eggenstein‐Leopoldshafen, Germany