Synergy between Defects, Photoexcited Electrons ... - ACS Publications

Subscriber access provided by Kaohsiung Medical University

Article

Synergy between Defects, Photoexcited Electrons, and Supported Single Atom Catalysts for CO Reduction 2

Junbo Chen, Satish Kumar Iyemperumal, Thomas Fenton, Alexander D Carl, Ronald L. Grimm, Gonghu Li, and Nathaniel Aaron Deskins ACS Catal., Just Accepted Manuscript • DOI: 10.1021/acscatal.8b02372 • Publication Date (Web): 01 Oct 2018 Downloaded from http://pubs.acs.org on October 1, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

Synergy between Defects, Photoexcited Electrons, and Supported Single Atom Catalysts for CO2 Reduction Junbo Chen,† Satish Kumar Iyemperumal,† Thomas Fenton,‡ Alexander Carl,¶ Ronald Grimm,¶ Gonghu Li,‡ and N. Aaron Deskins∗,† †Department of Chemical Engineering, Worcester Polytechnic Institute, Worcester, MA 01609 ‡Department of Chemistry, University of New Hampshire, Durham, NH 03824 ¶Department of Chemistry, Worcester Polytechnic Institute, Worcester, MA 01609 E-mail: [email protected]

Abstract

Experimental results corroborate these theoretical calculations and show that activation of CO2 occurs most readily for TiO2 catalysts with dispersed Cu and Ov . CO2 photoreduction also occurs most readily for TiO2 catalysts with dispersed Cu and Ov , compared to TiO2 or Cu over stoichiometric TiO2 catalysts. We also modeled atomic Pt to understand how metals besides Cu may behave. We found that Pt over TiO2 also activates CO2 but that dissociation of CO2 over Pt with Ov does not occur as readily as for Cu with Ov . Our results show that tailoring TiO2 surfaces with defects in conjunction with specific atomic catalysts like Cu may lead to fast desirable photoreduction of CO2 .

Dispersed atomic catalysts can achieve high catalytic efficiency and have the potential to enable chemical transformation of inert molecules like CO2 . The effect of surface defects on photocatalytic reduction of CO2 using supported single atom catalysts however requires clarification. Using density functional theory and experimental techniques, we have investigated the role of surface oxygen vacancies (Ov ) and photoexcited electrons on supported single atom Cu catalysts and CO2 reduction. Adsorption of Cu was strong to the TiO2 surface, and charges of the Cu atoms was highly dependent on whether surface defects were present. Cu atoms with Ov aided in the adsorption of activated bent CO2 , which is key to CO2 reduction. Our results also show that CO2 dissociation (CO2 * → CO* + O*), which is a proposed initial step of CO2 reduction to hydrocarbon products, occurs very readily for a single Cu atom in an Ov , with barriers ∼0.19 eV. Such low barriers do not occur with Cu over a stoichiometric surface. Furthermore, the presence of a photoexcited electron leads to a substantial increase in reaction rate for Cu over a stoichiometric surface; the Cu/TiO2 surface is largely inert in the absence of photoexcited electrons.

Keywords Photocatalysis, Single Atom Catalysts, CO2 Reduction, Photoexcitation, Surface Defects, Oxygen Vacancy, TiO2 anatase, CO2 Dissociation

1

Introduction

Small clusters and single-atom catalysts can have potentially increased reactivity compared to larger clusters or nanoparticles. Such cata-

ACS Paragon Plus Environment

1

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

lysts have a single atom (or a few atoms) embedded in a matrix or on a support. 1–8 For instance, recently DeRita et al. discovered that an isolated single Pt atom on TiO2 had a twofold turnover frequency increase and lower activation barrier than a 1 nm Pt nanocluster for the CO oxidation reaction; 9 other work also showed single atom Pt can catalyze CO oxidation in ceria 10 and FeOx 11 support. Of particular importance is the CO2 reduction reaction, which can curtail this greenhouse gas. Several reports describe the use of supported single atom catalysts for CO2 reduction. 12,13 Reduction of CO2 can produce products such as methane or methanol. 14–19 Supported Cu is a promising catalyst, 20–24 especially small Cu clusters or atoms on TiO2 . 25–28 Other work highlights how sub-nanometer catalysts can be active for CO2 reduction, such as supported sub-nanometer size, Pt 29,30 Mn, 28 Pd 28 and Ir 31 on TiO2 , graphene 32 and metal-organic framework. 33 Of note, experimental work indicates Cu clusters on TiO2 with surface oxygen vacancies can enable CO2 dissociation, even without light. 20,34 Surface features, such as oxygen vacancies, may play an important role in in CO2 reduction over metal oxide catalysts. 35–45 Density functional theory (DFT) has been used to model supported sub-nanometer metal catalysts. 46–58 For instance modeling CO2 reduction over Pt or Ag clusters on TiO2 has been shown to enable CO2 activation. 59–61 In our previous work we modeled adsorbed Cu atoms on TiO2 surfaces as photocatalysts. 62,63 We found Cu sites likely contribute to the improved photocatalysis by stabilizing surface adsorption of CO2 on TiO2 . Other authors have also modeled surfaces with oxygen vacancies and concluded that surface defects can enhance the CO2 adsorption properties. 37,38,42,64–66 Still, how oxygen vacancies and photoexcited electrons may interact with supported metal atoms is not fully clear. Atomic-level details on the interactions between support, single atoms, excited electrons, and defects for catalytic reactions, such as CO2 reduction, need further elucidation since understanding the role of these various features may guide better catalyst design. In this work we have modeled using DFT ad-

Page 2 of 25

sorbed Cu atoms on perfect TiO2 surfaces and surfaces with oxygen vacancies. We also modeled CO2 activation and reactivity over these Cu/TiO2 surfaces. We further determined the effect of photoexcited electrons on Cu adsorption and CO2 reduction. Our experimental results also compare Cu/TiO2 catalysts with different type of adsorption surface (stoichiometric and defective) to show the important role of Cu and Ov in activating CO2 . Finally, we briefly compare our results to calculations of supported single Pt atoms to determine how metals besides Cu may behave as photocatalysts for CO2 reduction. These results show how single metal atoms, surface defects, and photoexcited electrons may work together to enable facile CO2 photoreduction.

2 2.1

Methodology Computational Methodology

We ran all DFT calculations with the CP2K code. 67,68 CP2K uses a mixed method involving Gaussian functions and plane waves. 69 The plane wave representation was expanded to a cutoff of 300 Ry, and the valence electrons were represented by double ζ basis sets. 70 Core electrons were treated by Goedecker-Teter-Hutter pseudopotentials. 71 All calculations were spinpolarized and used the Perdew Burke Ernzerhof (PBE) exchange correlation functional. 72 We also included dispersion corrections through Grimme’s D3 method with Becke-Johnson damping. 73,74 Due to self-interaction errors of DFT we utilized the DFT+U correction 75,76 on Ti d orbitals with a U value of 2.5 eV. We chose this value since it gives gap states for a surface with an oxygen vacancy (from projected density of states calculations) about 0.7 eV below the conduction band, similar to the work of Haa and Alexandrova 43 (U = 3.6 eV) and Sel¸cuk and Selloni 77 (U = 2.5 eV). This U value of 2.5 eV has also been suggested by Hu and Metiu 78 for modeling defective TiO2 . Similar settings have been used in our previous work. 63 All calculations were sampled using a single k-point at gamma. We performed charge


2

Page 3 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

analysis using the DDEC6 method. 79,80 In our previous work 63 we compared this method to the popular Bader method 81 of charge analysis and found DDEC6 to be comparable.

CO* + O*) between initial and final states. We used at least 10 images for the nudged elastic band calculations. The NEB calculations were converged when the maximum force on the atoms was smaller than 0.05 eV/˚ A, similar to 77 Sel¸cuk and Selloni. We determined vibrational frequencies for adsorbed species (i.e. CO and CO2 ) using a finite difference approach to calculate the Hessian matrix. For these calculations we used a higher cutoff of 600 Ry to obtain better accuracy, and took previous geometries and reoptimized them at the higher cutoff energy. Because these calculations were computationally expensive we limited the the number of atoms that were displaced during the vibrational frequency calculations. Only atoms in the vicinity of the adsorbate (∼ 90 atoms), or within a 9 ˚ A radius from the C atom of CO2 or CO were displaced. This approach is similar to our previous work. 63 The step size used to construct the Hessian was 0.001 bohr.

Figure 1: The supercell used in this work, being a (2x4) representation of the anatase (101) surface. The slab had 288 (fully stoichiometric) or 287 (one oxygen vacancy) atoms. Indicated are different atom types and an oxygen vacancy (Ov ). The subscripts indicate the coordination number of the atoms, e.g., O2c is a two-coordinated oxygen atom. Ov indicates an oxygen vacancy.

2.2

P25 TiO2 was obtained from Evonik and used as received. Cu/TiO2 was synthesized via surface adsorption. In a typical synthesis, 500 mg of dried TiO2 was dispersed in 75 mL Milli-Q water and sonicated for 10 minutes. 100 mg of CuCl2 (Sigma-Aldrich, 99.995%) was then added to the TiO2 suspension under constant stirring. The resulting mixture was stirred at room temperature for 12 hours before Cu/TiO2 was recovered via centrifugation and was washed with Milli-Q water for three times. The Cu/TiO2 sample was then dried at 90o C for 2 hours. Thermal treatment of powder samples was conducted in a Harrick Praying Mantis diffuse reflectance accessory in a Thermo Nicolet 6700 FTIR spectrometer. A small amount of TiO2 or Cu/TiO2 was placed in the sample holder of the accessory, purged with Ar at room temperature, and then annealed at 300o C for 1 hour under 5% H2 (95% Ar) or O2 prior to further CO2 adsorption and photochemical studies. The synthesized Cu/TiO2 sample was examined using transmission electron microscopy

TiO2 anatase (101) surfaces were modeled using periodic slabs with three TiO2 bilayers, or a total of six individual layers, as shown in Figure 1. The supercell had ∼ 20 ˚ A of vacuum between slabs. The surface lattice vectors were 20.6 ˚ A and 15.1 ˚ A to give a 2x4 supercell (288 atoms). The atoms in the bottom bilayer were fixed at their bulk positions. We modeled two different surfaces, a stoichiometric surface and a surface with one oxygen vacancy. Adsorption energies of species M were found by the following: ∆Eads−M = EM/T iOx − ET iOx − EM .

Experimental Methodology

(1)

In this equation EM is the energy of species M (e.g. Cu atom or CO2 ) in vacuum, ET iOx is the energy of a titania surface (either stoichiometric or with an oxygen vacancy), and EM/T iOx is the energy of the system with species M adsorbed. We performed climbing image nudged elastic band (CI-NEB) calculations 82,83 to search the minimum energy pathway and determine activation barriers in CO2 dissociation (CO∗2 →


3

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(TEM, Zeiss/LEO 922 Omega). The loading of Cu on TiO2 was quantified using a Varian Vista AX induced coupled plasma atomic emission spectrometer. X-ray photoelectron spectroscopy (XPS) analysis was conducted on a PHI5600 system. 84 UV-visible spectra of powder samples were obtained on a Cary 50 Bio spectrophotometer. A Barrelino diffuse reflectance probe was used to collect UV-visible spectra of powder samples using BaSO4 as a standard. For studies with XPS and UV-visible spectroscopy, thermal treatment was carried out in a STF 1200 tube furnace at 300o C under 5% H2 (95% Ar) or O2 .

3 3.1

Page 4 of 25

used to help guide our choice of Cu adsorption sites. We found the most stable adsorption site over the reduced surface occurred when Cu filled the oxygen vacancy (Figure 2b), with an adsorption energy of -2.23 eV. A slightly lower adsorption energy (-2.12 eV) was found when the Cu atom adsorbed near the vacancy but did not completely occupy it (Figure 2c). When the Cu atom was far from the oxygen vacancy, it interacted with the surface in a similar manner as with the stoichiometric surface (Figure 2d), giving a comparable adsorption energy (-2.26 eV). Our results show that the Cu atoms became positively charged when interacting with O2c atoms (stoichiometric and far from the oxygen vacancy) but negatively charged when in or near the oxygen vacancy. The region around the vacancy is negatively charged due to unpaired electrons left in the surface on nearby Ti atoms upon O removal, and the surface donates this charge to the adsorbed Cu atom. Negative metal atoms adsorbed in an Ov have been reported previously, 50,57,86 as well as negative copper clusters. 87 All adsorption energies were close to each other (-2.12 to -2.30 eV), indicating that all geometries could occur upon Cu adsorption over TiO2 and may be catalytically relevant. We thus considered all four geometries when modeling CO2 adsorption and reduction over Cu/TiO2 surfaces. Our surface model (lone Cu atoms on the TiO2 surface) is commensurate with the Cu/TiO2 materials synthesized in our experiments. We have shown in previous work 62 that atomic Cu species can be deposited on the TiO2 surface, giving catalysts with CO2 photocatalytic activity. In other work, 63 we demonstrated Cu dimer formation is an endothermic process, which would limit the growth of single Cu atoms. Furthermore, we showed that Cu atoms have high diffusion barriers on the TiO2 surface, which may also limit the formation of dimers and larger clusters. 63 The thermodynamic and kinetic limitations for dimer formation favor single Cu atoms on the surface. In Figure 3 is shown the density of states (DOS) for Cu adsorbed on the various surfaces. DOS for the stoichiometric (Figure 3a) and reduced TiO2 surfaces (Figure 3b) are also given

Results and Discussion Adsorption of Cu Atoms

We first modeled a Cu atom adsorbed over stoichiometric and reduced (101) surfaces, as summarized in Figure 2. We considered several different initial sites, as discussed in the Supporting Information, but the most stable sites are shown. Over the stoichiometric surface the Cu atom preferred a bridge geometry between two O2c atoms, similar to previous work, 53,58,63 with Cu-O2c distances of 1.88 ˚ A. The adsorption energy for a Cu atom over the stoichiometric surface (-2.30 eV) agreed very well with the work by Seriani et al. (-2.30 eV) 53 and Alghannam et al. (-2.26 eV). 58 The calculated charge of the adsorbed Cu atom over the stoichiometric surface (+0.5) was also in agreement with the work of Alghannam et al. (+0.64). 58 We tested the effect of +U correction on the adsorption energy of Cu (-2.59 eV without +U correction and -2.30 eV with +U correction), and it agreed with our previously reported value (-2.56 eV without +U correction). 63 For a surface with an oxygen vacancy, we found three stable adsorption modes: Cu in the vacancy, near the vacancy, and far from the vacancy (Figures 2b-d). Several initial sites were modeled, as discussed in the Supporting Information. Previous results on metal atom adsorption over reduced TiO2 surfaces with other transition metals, such as Pt and Au, 49,50,85 were


4

Page 5 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

Figure 2: Cu adsorption over TiO2 anatase (101) surfaces. Shown is Cu adsorption over a (a) a stoichiometric surface, (b) in an oxygen vacancy, (c) near an oxygen vacancy, and (d) far from the oxygen vacancy. The Cu atom is depicted in orange. Given are adsorption energies while numbers in parentheses are calculated Cu charges. for comparison. A gap state for a surface with Ov was observed due to electrons localizing on Ti atoms, and is a well-known phenomenon. For Cu over TiO2 (Figure 3c), a gap state created by Cu was also observed. Similar gap states have been seen by other authors. 57,59 For instance, Yang et al. 59 observed gap states for Ag (one row below Cu) clusters in Ov to be near -1.2 eV and -1.8 eV below the edge of the conduction band. For Cu in an Ov a similar gap state due to Cu occurs (near -1.5 eV), but there is also a gap state on TiO2 due to the Ov near -0.7 eV which is composed of Ti electrons. Overlap between the Cu and TiO2 gap states suggest strong interactions, such as Cu in an Ov (∼ -0.7 eV) and Cu near an Ov (between -0.4 eV to 0 eV). Indeed electron transfer occurs from the Ov to Cu, with Cu having a -0.35 charge when in the Ov . When an Ov forms two unpaired electrons are left in TiO2 (O2− → Ov + 12 O2 + 2e− ) which may reduce two Ti atoms. The two reduced Ti atoms near the Ov had charges of 1.93 e− and 1.83 e− , compared to the average charge of 2.42 e− for the other Ti atoms. After Cu adsorbed in an Ov , one of the reduced Ti atoms had a charge of 2.19 e− while the other reduced Ti atom had a charge of 1.73 e− , indicating a net transfer of electrons to the Cu atom from the two Ti atoms. After the charge transfer occurred from TiO2 to Cu, the Ov gap states hybridized with the Cu states, as shown in Figure 3d and e. For Cu far from an Ov both gap states for Cu and TiO2 are seen, but since the Cu does not interact with the Ov , these states are largely independent of each other and have little overlap.

Figure 3: Projected density of states plots for adsorbed Cu. The zero energy level is set at the conduction band edge. The filled curves represent occupied states, while unfilled curves represent unoccupied states. An important consideration is the effect of adsorbed Cu atoms on the reduction of the TiO2 surface, for instance through O2c removal. It is possible that adsorbed Cu may facilitate or influence the formation of Ov . Experimental work for instance indicates that Rh atoms may enhance Ov formation. 88 The formation energy of an Ov (O2c → Ov + 12 O2 ) for TiO2 was found to be 4.43 eV, similar to the value by Cheng et al. (4.34 eV). 36 However, we calculated the Ov formation energy in the presence of Cu (or starting with a Cu/TiO2 surface) to be between 4.47 and 4.61 eV, depending on whether the Cu atom in the final state was in the Ov , near the Ov , or far from the Ov . These energies indicate that Cu is not likely to help increase the number of Ov . Rather Ov on the surface could form


5

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

by other means, such as annealing, 89 doping, 90 other treatments, 91,92 or synthesis conditions. 93

3.2

Page 6 of 25

for bent CO2 . Figure S3 shows the adsorbed linear CO2 geometries , while Figure S4 shows other less stable bent configurations that we found. Over stoichiometric TiO2 (configurations S1 and S2), linear CO2 adsorption was more stable than bent CO2 , regardless of whether Cu is present. The linear and bent CO2 energies on TiO2 (-0.20 eV for bent and -0.39 eV for linear) and Cu/TiO2 (-0.28 eV for bent and -0.37 eV linear) are in agreement with our previous work. 63 Linear CO2 had charges near 0 e− (see Figure 4b) while bent CO2 had negative charges near -0.3 e− over stoichiometric TiO2 . Stoichiometric TiO2 has no free electrons available to readily transfer to CO2 which could explain why neutral, linear CO2 is more stable than the bent, anionic CO2 over the stoichiometric surface. Electron transfer from the Cu/TiO2 surface to CO2 was also apparently difficult, at least compared to the reduced surfaces as discussed below. This leads to linear CO2 being more stable than bent CO2 over the two stoichiometric surfaces we modeled. We note that the bent CO2 configuration over the stoichiometric surface (see configuration S1 in Figure 5) resembled a carbonate-like structure, which was also observed in previous work. 37,38,63 We next analyzed the effect of surface vacancies on CO2 adsorption. Bent CO2 is markedly stabilized over the surfaces with Ov present, as Figure 4 and Figure 5 show. Over TiO2 , the Ov leads to a very strong bent CO2 structure, as the CO2 molecule fills the Ov (see R1 in Figure 5). The adsorption energy of bent CO2 in an Ov was -1.28 eV, compared to linear CO2 having an adsorption energy of -0.85 eV. We also compared our adsorption energies to literature papers that modeled the TiO2 anatase (101) with Ov and adsorbed bent CO2 . Sorescu et al. 38 reported an adsorption energy of -1.35 eV for the R1 case, which is similar to our adsorption energy of -1.28 eV. He et al. 37 reported an adsorption energy of -1.09 eV for the R1 configuration at the PBE level, similar to our calculated PBE value of -0.92 eV. He et al. used a (2 x 1) TiO2 surface cell whereas our work using a larger (2 x 4) surface cell, which might explain the slight differences. Further details

Adsorption of CO2

Figure 4: Summary of CO2 adsorption over over TiO2 and Cu/TiO2 . (a) The adsorption energies of CO2 of the most stable linear and bent configurations are given. (b) The calculated charges of the most stable adsorbed CO2 are also shown. Geometries for the most stable bent configurations are given in Figure 5. We modeled CO2 adsorption over the various TiO2 and Cu/TiO2 surfaces. Formation of bent CO2 is a key step in CO2 activation in order to further reduce the molecule. 37,94 Linear CO2 is inert and resistant to reactivity. We have summarized results for both linear and bent CO2 adsorption in Figure 4, giving both adsorption energies and calculated charges of CO2 . We modeled CO2 adsorption over stoichiometric surfaces (configurations S1 and S2), and reduced surfaces (configurations R1-R4). Figure 5 shows the final optimized geometries


6

Page 7 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

Figure 5: Most stable bent CO2 adsorption geometries over TiO2 and Cu/TiO2 surfaces. Shown is CO2 adsorption over several surfaces: a stoichiometric surface (S1), a stoichiometric surface with a Cu adatom (S2), a reduced surface with oxygen vacancy (R1), a reduced surface with a Cu adatom in an oxygen vacancy (R2), a surface with Cu near an oxygen vacancy (R3), and a surface with Cu far from the oxygen vacancy (R4). The CO2 adsorption energies are given. on the effect of the +U correction can be found in the Supporting Information. When Cu is present on a reduced surface, we observed strong stabilization of bent CO2 , with bent CO2 adsorption energies of -0.85 eV (R2), -1.30 eV (R3), and -1.29 eV (R4). Interestingly, the bent CO2 adsorption energies over a surface with Ov (R1) and Cu/Ov (R3) are comparable. In all cases over reduced surfaces bent CO2 is preferred over linear CO2 , but for the R2, R3 and R4 states this preference is very pronounced, with energy differences of -0.34, 1.03 eV and -0.80 eV, respectively, between the linear and bent CO2 structures. Such significant differences in adsorption energies between linear and bent CO2 suggest high selectivity for bent CO2 formation over surfaces with both Cu and Ov present, and hence potentially larger CO2 reduction activity. Bent CO2 had O-C-O angles near 130◦ (see Table S1) while linear CO2 angles were all near 180◦ , indicating the activated nature of CO2 . In the R1 configuration, CO2 had all three atoms (O, C, O) bonding with the Ti near the oxygen vacancy. The R4 configuration had essentially the same CO2 bent adsorption energy and structure as R1, suggesting that the Cu adatom in R4 was far enough from the Ov so that it did not interact with the adsorbed CO2 or Ov significantly. In the R2 configuration, the Cu adatom bound to the Ov strongly and blocked the CO2

molecule from filling the Ov . The CO2 bent as it formed bonds (Cu-C and Ti-O) with the surface. In the R3 configuration, the optimized bent CO2 structures appeared to be very similar to R2 configuration (see Figure 5). CO2 only had two atoms (O and C) bonding with the surface in the R2 and R3 configurations: one O atom of the CO2 bound to the Ti atom near the Ov and the C atom interacted with two Ti atoms. The CO2 charge in the R2 configuration was -0.56 e− for instance, while in the R3 configuration the CO2 charge was -0.63 e− . Their adsorption energies were however different, with the R3 configuration being 0.45 eV more stable than the R2 geometry. The Cu atom in the R3 configuration was further from the Ov which allowed Cu interactions with the surface O2c atom. In the bent CO2 R2 configuration, the Cu atom was 1.23 ˚ A from the Ov site, while in the R3 configuration the Cu atom was 1.48 ˚ A from the Ov site. This slightly different configuration could lead to the R3 configuration being more stable than R2 configuration. We further examined the stability of the R2 site by perturbing the Cu atom away from the Ov and allowing geometry optimization. We found that with sufficient perturbation the R2 configuration relaxed to the R3 configuration, which suggests that R2 may be a local minimum. Ov should be highly reactive due to undercoordinated surface atoms, but may also generate excess electrons which may transfer to the


7

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

CO2 molecule, as shown in previous work. 95,96 Figure 4b shows that all the bent CO2 became very negative when Ov was present, being -0.28 e− (S1) and -0.35 e− (S2) over stoichiometric surfaces while -0.74 e− (R1), -0.56 e− (R2), 0.63 e− (R3), and -0.74 e− (R4) over the reduced surfaces. All the linear CO2 had similar charges, being slightly cationic, near 0 e− . The most negatively charged CO2 occurred in the R1 and R4 configurations(-0.74 e− ), where both O atoms of CO2 formed bonds with nearby Ti atoms. In both the R1 and R4 configurations, one reduced Ti atom (adjacent to the adsorbed CO2 ) lost 0.36 e− and the other gained 0.1 e− upon CO2 adsorption. Likewise, CO2 gained 0.37 e− while the C atom was nearly neutral after adsorption. When only one O atom and the C atom of CO2 interacted with surface Ti and Cu atoms, as observed as in the R2 and R3 cases, the calculated charges on CO2 were -0.56 e− and -0.63 e− , respectively. In R3 for example, the C had a charge of +0.28 e− , while the O atoms had charges of -0.35 e− (O away from the surface) and -0.57 e− (O close to the surface) after CO2 adsorption. However, despite having less CO2 charge, the R3 configuration had almost the same adsorption energy as R1 and R4, which had more negative CO2 charge, suggesting that the presence of the Cu adatom in R3 had a substantial effect in stabilizing bent CO2 . Projected density of states details for adsorbed bent CO2 are shown in Figure 6. For the S1 configuration, the majority of CO2 bands are near the edge of the valence band (-2 eV below the conduction band), which is 2 eV higher than the CO2 bands in the gas phase which has a characteristic peak near -4 eV below the conduction band (shown in Figure S6). For the S2 configuration, the Cu state overlapped with the TiO2 states, while overlap of Cu and CO2 states were minimal, suggesting stronger Cu/TiO2 and weaker Cu-CO2 interactions. This agrees with the previous work by Iyemperumal and Deskins. 63 For the R1 configuration, the highest CO2 energy level was near -1.7 eV. The gap state created by the oxygen vacancy (shown in Figure 3b) also shifted to near -1.7 eV in the R1 configuration (Fig-

Page 8 of 25

Figure 6: Projected density of states plots for adsorbed bent CO2 . The zero energy level is set at the conduction band edge for each plot. The filled areas represent occupied energy states. ure 6c). Charge transfer from the reduced Ti atoms to CO2 and interactions between the CO2 molecule and nearby Ti atoms occurred. Charge analysis showed that the one of the Ti3+ atoms lost 0.26 e− and the other gained 0.10 e− after CO2 adsorption to have charges of 2.19 and 1.73, which brought one of the Ti atoms to a Ti4+ state (near 2.31 e− ). The one Ti3+ state shifted to near -1.7 eV and completely overlapped with the CO2 band. Despite a net charge transfer from the Ti3+ to bent CO2 , other surface atoms also donated charge to CO2 (leading to a final CO2 charge of -0.74 e− ). In the R2 and R3 cases (Cu in and near Ov ), Cu bands were in the valence band region, while a portion of the CO2 states were at higher energies near the edge of the conduction band. These CO2 bands hybridized very well with gap states of TiO2 and Cu states, indicating strong interactions between CO2 and Cu/Ov surface.

3.3

CO2 Reactivity

We already demonstrated how activation of CO2 to its bent structure occurs when Cu atoms and Ov are present. We now discuss the reactivity of CO2 over the various TiO2 surfaces. Using the CI-NEB method we calculated several barriers for CO2 dissociation over select surfaces. Formation of CO from CO2 is one of


8

Page 9 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

the proposed early steps of CO2 reduction, 42,64 with subsequent hydrogenation steps forming various intermediates and products. In Table 1 we summarize the calculated reaction energies and activation barriers for the reaction CO2 * → CO* + O*. We started with bent CO2 in all cases, and details on CO and O adsorption can be found in the Supporting Information. For the stoichiometric surface (S1) the CO2 dissociation energy was found to be 1.50 eV. Since the reaction energy was so large, we expect the activation barrier to be at least 1.50 eV, and did not calculate the activation barrier. Over the Cu/TiO2 surface (S2) the dissociation energy was 0.58 eV. We calculated the activation barrier to be 0.83 eV for the Cu/TiO2 case (S2). Thus, it appears that CO2 dissociation over these stoichiometric surfaces is likely to proceed at a very slow rate since the activation barriers prevent rapid CO2 dissociation and reduction. The dissociation path for the S2 configuration can be found in Figure S8a in the Supporting Information. We also modeled CO2 dissociation over reduced surfaces. The calculated dissociation energy of CO2 in an Ov (configuration R1) was -0.40 eV, and the barrier was 0.95 eV. Sorescu et al. 38 reported an activation barrier of 0.90 eV for CO2 dissociation over Ov /TiO2 , while Ji and Luo 42 found a barrier of 0.73 eV. Our calculated activation barrier was comparable to these literature results. We expect that for Cu far from the Ov (R4) the activation barrier will be similar to R1, given their similar adsorption energies, geometries, and dissociation energies. The effect of single metal atoms and Ov on CO2 reactivity is unknown, and we applied the CINEB method to calculate activation barriers for CO2 dissociation starting with the R2 and R3 configurations. Figure 7 shows the two pathways for CO2 dissociation over these surfaces. In both cases, the CO2 dissociates to fill the Ov and pushes the Cu out or away from the Ov . The barriers are very low in both cases. For the R2 configuration, the process is energetically downhill with a small activation barrier of 0.1 eV (energy from the local minimum in image 3 to image 4). If the energy is not fully dissipated to the surrounding surface as the CO2 molecule

Figure 7: The calculated climbing image nudged elastic band pathways for CO2 dissociation over (a) Cu in an Ov (R2) and (b) Cu near an Ov (R3). transitions from image 1 to the lower energy image 3 state, then CO2 dissociation may be barrier-less. The R3 configuration has an activation barrier of only 0.19 eV. Both the R2 and R3 dissociation pathways are similar. But the R2 configuration, as we discussed in Section 3.2, was less stable than the R3 configuration. This corresponds to a dip in the energy path from image 1 to 3 (shown in Figure 7a) where the CO2 molecule reaches a state similar to the R3 configuration. The slightly different geometry contributes to the difference in the activation barrier of 0.09 eV between the R2 and R3 cases. These results show that a single Cu atom and Ov lowers the barrier for CO2 reactivity sub-


9

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 25

Table 1: Summary of reaction energies and select activation barriers for bent CO2 dissociation, CO2 * → CO* + O*. All numbers are in eV.

Site Reaction Energy Reaction Barrier Reaction Energy Reaction Barrier Reaction Energy Reaction Barrier

S1

S2 Cu 1.50 0.58 >1.50 eV 0.83 Cu + e1.39 -0.50 >1.39 eV 0.57 Pt -0.21 1.50 >1.50 eV 0.85

R1

R2

R3

R4

-0.40 -1.71 -1.36 -0.30 0.95 0.10 0.19 0.95 -0.42 -1.44 -1.17 -0.39 0.97 0.19 0.27 0.97 -0.40 -1.08 -1.13 -0.39 0.95 0.43 0.46 0.95

0.57 e− , while the other O atom had a charge of -0.35 e− . The C atom in the R3 case had a charge of +0.28 e− . Furthermore the CO2 charge distributions in R1 and R3 also mirrored the C-O bond distances. The CO2 in R1 was symmetric with C-O bond distances of 1.26 ˚ A, while the CO2 in R3 had C-O bond distances of 1.22 ˚ A and 1.33 ˚ A (summarized in Table S1). The O atom that interacted with Ti in R3 had an elongated C-O bond compared to R1, indicating that this molecule could be in a preactivated state. Indeed, the activation barrier to break this bond is much lower for R3 compared to R1, as our CI-NEB calculations show.

stantially compared to just stoichiometric or reduced TiO2 surfaces or Cu/TiO2 . Importantly, our work confirms experimental work, where it was shown that Cu and Ov could lead to spontaneous CO2 dissociation even under dark conditions. Liu et al. 20,34 observed experimentally that CO2 species spontaneously dissociated into CO even in the dark on a reduced Cu/TiO2 surface. In their proposed CO2 dissociation mechanism CO2 dissociated and healed the oxygen vacancy and formed a Cu-CO complex. Our CI-NEB results confirm this mechanism as we have shown the activation barrier is very low and promotes CO2 dissociation significantly. Our results show that CO2 dissociation in an Ov has a high barrier (0.95 eV) , while dissociation in or near Cu/Ov is much smaller (< 0.2 eV). The role of Cu can be further understood by analyzing the CO2 charges and geometries. In the R1 case, where CO2 filled the Ov , the surface excess electrons transferred to both O atoms. The O atoms in CO2 had charges of -0.37 e− , while the carbon remained neutral. Analysis of the TiO2 charges indicates that these electrons on O came predominantly from reduced Ti atoms. However, in the case of Cu near Ov (R3) the O atom in CO2 that interacted with the surface had a charge of -

3.4

The Effect of Photoexcited Electrons

In order to analyze the possible effect of photoexcited electrons, we modeled both Cu and CO2 adsorption in the presence of an unpaired electron, similar to the method used by He et al. 37 A hydrogen atom was placed on the surface, away from the CO2 adsorption site, and served to donate an electron to the surface. Further details on the hydrogen atom position can be found in the Supporting Information. This extra electron could transfer to adsorbates and


10

Page 11 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

tron donor, performed similarly in modeling the effect of a photoexcited electron. 37 Rodriguez et al. for example studied CO2 reduction on the brookite TiO2 (210) surface after introducing an extra electron. 97 Figure 8a shows adsorption energies of Cu atoms when such an electron was present compared to the case of no extra electron present, and Figure 8b shows similar results for adsorption of bent CO2 . For Cu on the stoichiometric surface or away from the Ov , the adsorption energy either did not change or was slightly destabilized by 0.1 eV. In the presence of an extra electron the Cu binding to TiO2 with Ov was strengthened by 0.43 eV (R2) and 0.44 eV (R3). Since the Cu atom was more stabilized in an Ov when an excited electron was present, this may favor the migration or adsorption of Cu atoms into Ov sites under photocatalytic conditions. The charges of the Cu atoms in Figure 8a can be found in Table S2 in the Supporting Information. We also modeled the effect of photoexcited electrons on linear CO2 adsorption as summarized in Figure S10. We found the adsorption trends were the same compared to the no extra electron case, in that bent CO2 was preferred over the reduced surfaces and linear CO2 was preferred stoichiometric surfaces. The linear CO2 adsorption energy for Cu in an Ov (R2) and for Cu near an Ov became more exothermic by 0.31 eV and 0.07 eV, respectively, in the presence of an extra electron. Other linear CO2 adsorption energies did not change in the presence of the extra electron. We also determined the effect of the hydrogen atom position on the bent CO2 adsorption energy by placing the hydrogen atom at four different positions relative to the adsorbed CO2 . All adsorption energies were within ∼ 0.1 eV of each other, which indicates that the hydrogen position has little effect when donating an electron for this model. We discuss more details on this issue in the Supporting Information. For bent CO2 adsorption in every case (except over stoichiometric TiO2 ) the adsorption energies became stronger when an extra electron was present (see Figure 8b). The adsorption energies became more negative by 0.05 to 0.28 eV. We calculated the charges of CO2 over

Figure 8: Adsorption energies of (a) Cu adatoms and (b) bent CO2 over various surfaces in the presence of an extra electron. The extra electron was modeled using a hydrogen atom, similar to the work of He et al. 37 mimicked a photoexcited electron. This model of course does not fully simulate a photoexcited electron, and we would expect photoexcited electrons to be in higher energy states. Any charge transfer from a photoexcited electron is therefore likely to be more energetic compared to our approach. Rather this approach is a quick method to determine any potential effect of such electrons, such as if adsorbates will accept photoexcited electrons. More complicated methods, such as time-dependent DFT, can better describe photoexcited electrons, but at much greater computational cost. Yang et al. recently used this photoexcited electron model with hydrogen atoms on the surface to mimic the injection of electrons into TiO2 . 61 We note that surface models with an net -1 charge (essentially one extra electron in the system), rather than an added hydrogen atom as an elec-


11

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

the surfaces with an unpaired electron present, and summarize them in Table S3. We found that the charges of CO2 did not change significantly when the electron was present. Since the charges of bent CO2 were the same regardless of whether an unpaired electron was present, we attribute the stronger adsorption energies to the ease of electron transfer from the surface to CO2 . There is an energy cost for removing charge from the surface to the CO2 molecule, and when an extra electron is present, this cost is diminished. The extra electron more readily transfers to CO2 compared to the surface without an electron present. These results suggest that photoexcited electrons, along with Cu atoms and Ov , may further stabilize activated, bent CO2 and lead to efficient CO2 photoreduction. We also considered the effect of photoexcited holes as previous work 98,99 showed cases where photogenerated holes could promote organic molecule dissociation on TiO2 surfaces. Wanbayor et al. 100 also found CO oxidation could occur in the presence of holes. We placed an electron-withdrawing group, a hydroxyl, on the surface to mimic a hole, similar to previous DFT work. 99,101,102 Figure S12 shows the positions of hydroxyls on the surface. The adsorption energies of bent CO2 over the stoichiometric surfaces became 0.10 and 0.06 eV more positive for S1 and S2, respectively, when the hydroxyl was present. For the R1 case, the adsorption energy of bent CO2 became endothermic in the presence of a hydroxyl group (1.06 eV). These results indicate that the photoexcited hole did not contribute to stabilizing activated CO2 on both the stoichiometric and reduced TiO2 surfaces. Intuitively these results make sense since CO2 activation and dissociation proceeds through reduction (gaining electrons) rather than an oxidation process. We also modeled CO2 dissociation in the presence of the unpaired electron. Results are shown in Table 1. Over stoichiometric TiO2 (S1) the presence of the electron had a slight effect and the reaction energy was more exothermic by 0.11 eV, but overall the reaction was still quite endothermic at 1.39 eV. In the case of Cu/TiO2 (S2) the effect was quite dramatic.

Page 12 of 25

The reaction energy dropped by 1.08 eV and becomes -0.50 eV, which is reasonably exothermic. Further NEB modeling of this reaction gave a barrier of 0.49 eV (see Figure S8 for NEB pathways), which indicates that photoexcited electrons may very strongly enable CO2 adsorption and reactivity over Cu/TiO2 . For the reduced surfaces where Cu did not interact with the CO2 molecule (R1 and R4), the extra electron did not change the reaction energies significantly. The calculated reaction barriers with an photoexcited electron for Cu in an Ov (R2) and Cu near an Ov (R3) were 0.09 eV and 0.08 eV higher than barriers with no extra electron, indicating this extra electron did not have a positive role in dissociating CO2 like for other cases (e.g. S2 or Cu/TiO2 ). The photoexcited electron therefore had a more substantial effect for CO2 dissociation on Cu/TiO2 (S2) than on Ov (R1) and Cu/Ov (R2,R3). The reduced surfaces already had unpaired electrons, so further unpaired electrons did not improve CO2 dissociation kinetics.

3.5

Experimental Results on Cu atoms and Ov

Highly dispersed Cu sites on TiO2 were prepared via a simple adsorption method, in which Cu2+ cations were adsorbed onto TiO2 nanoparticles suspended in an aqueous solution of CuCl2 . The synthesized sample was thermally treated under H2 (to produce oxygen vacancies) or under O2 . See the Methodology for further details. Quantification with elemental analysis indicated a loading of 20.5 µmol Cu per gram TiO2 , corresponding to a coverage of 1 Cu atom per 5 nm2 surface area on TiO2 . 62 Examination with microscopy confirmed the absence of Cu aggregates or nanoparticles in the synthesized Cu/TiO2 material (Figure S13). Thermal treatment with H2 resulted in the formation of oxygen vacancies in Cu/TiO2 , as shown by its optical spectrum (Figure S14, trace b). Further characterization with X-ray photoelectron spectroscopy revealed the presence of both Cu2+ and Cu+ on the synthesized Cu/TiO2 that was not thermally treatment, but only Cu+ on the same sample thermally treated in the presence


12

Page 13 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

Figure 9: FTIR spectra of species formed upon CO2 adsorption on (a) H2 -treated TiO2 , (b) H2 treated Cu/TiO2 , and (c) O2 -treated Cu/TiO2 in the spectral region between 1500 cm−1 and 1800 cm−1 . The spectra are fitted to individual bands with Lorentzian line-shapes. The fitted peak at 1638 cm−1 , corresponding to bent CO2 , is indicated by distinct coloring for each sample. of H2 or O2 (Figure S15). A substrate overlayer model interprets the Cu 2p and Ti 2p photoelectron peak areas in terms of a fractional coverage of copper atoms on the TiO2 substrate. Detailed in the supplementary information section, the model determines a coverage of 2.6% of a copper monolayer from the Cu 2p3/2 -to-Ti 2p3/2 peak area ratio of 0.053 for the experimentally representative data in Figure S16. Considering that this coverage is well below that of a single monolayer, we interpret the results to imply that the surface is dominated by single copper atoms rather than dimers or small clusters. Moreover, theoretical results indicate Cu dimer formation over TiO2 is unfavorable, 63 reinforcing this view. Further investigation of the copper adlayer morphology remains the subject of ongoing study. In our study using Fourier transform infrared (FTIR) spectroscopy, the formation of bent CO2 molecules adsorbed on TiO2 was observed in the presence of oxygen vacancies and/or surface Cu sites. TiO2 or Cu/TiO2 placed in an in situ infrared cell 103 were annealed at 300 o C prior to CO2 adsorption and light irradiation. Figure 9 shows spectra of species formed upon CO2 adsorption on different TiO2 samples. In addition to multiple carbonate peaks, a band at 1638 cm−1 associated with surface-adsorbed bent CO2 molecules 104 is clearly distinguished

in the spectrum of H2 -treated TiO2 (Figure 9a). This peak is absent from the spectrum of untreated TiO2 , which does not contain oxygen vacancies, after CO2 adsorption (Figure S16).

Figure 10: Difference FTIR spectra of (a) H2 treated TiO2 , (b) H2 -treated Cu/TiO2 , and (c) O2 -treated Cu/TiO2 after photochemical CO2 reduction for 1 h. The peak at 1638 cm−1 is also present, with greater intensity, in the spectrum of H2 -treated Cu/TiO2 (Figure 9b). In comparison, this peak


13

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 14 of 25

bare TiO2 surface may occur, so that the ∼1800 cm−1 peak could appear. Both experiment and theory indicated that bent CO2 was not formed over pure TiO2 , but that bent CO2 formed over surfaces with Cu atoms and Ov . In our photochemical studies, thermally treated TiO2 or Cu/TiO2 was subject to irradiation with UV light in the presence of gaseous CO2 . Difference FTIR spectra were obtained by subtracting infrared spectra after light irradiation from corresponding spectra before light irradiation. A peak at 2111 cm−1 is clearly seen in the difference spectrum of H2 treated Cu/TiO2 , indicating the formation of CO adsorbed on Cu+ sites upon light irradiation (Figure 10b). This peak is also observed in the spectrum of O2 -treated Cu/TiO2 but with much lower intensity (Figure 10c). Since a much greater amount of bent CO2 molecules were produced by H2 -treated Cu/TiO2 than O2 -treated Cu/TiO2 (Figure 9), the comparison between the spectra shown in Figure 10b and 10c suggests that the increased number of bent CO2 molecules promoted photoinduced dissociation of CO2 on the TiO2 materials. CO2 dissociation may also have been facilitated by lower barriers, as our calculations show, over H2 -treated Cu/TiO2 which would presumably have more Ov . Formation of CO by photoinduced dissociation of CO2 likely occurred on H2 -treated TiO2 as well, but was not observed in the spectrum shown in Figure 10a due to the absence of surface sites (such as Cu+ ) for strong CO adsorption. These results, again in agreement with our DFT results, indicate that Ov and Cu may increase CO2 reduction activity. We also analyzed CO formation in the dark over our catalysts, but no adsorbed CO was detected, unlike previous work. 64 Our Cu loadings were small, which could limit CO2 dissociation in the absence of light. We also calculated CO stretching vibrational frequencies using DFT to compare with our experimental FTIR data. In the Supporting Information we show optimized structures for adsorbed CO in Figure S17. Our calculated gas phase vibrational frequency of CO was 2161 cm−1 . CO was weakly bound to the pure TiO2 surface (S1 case) with an adsorption energy of -

is much less intense in the spectrum of O2 treated Cu/TiO2 (Figure 9c). The presence of bent CO2 molecules on O2 -treated Cu/TiO2 is likely due to the fact that surface Cu sites stabilize the formation of bent CO2 , as indicated in our prior work 62 and current work. We conclude therefore based these experimental spectra and peak intensities that bent CO2 forms with the following intensities: Cu/Ov TiO2 > Ov TiO2 > Cu/TiO2 > TiO2 . This is in agreement with our DFT calculations. In order to further validate the formation of activated bent CO2 we calculated vibrational frequencies of bent CO2 on various TiO2 surfaces using DFT. Our calculated vibrational frequencies for the relevant stretching mode of adsorbed bent CO2 were 1834 (S1), 1670 (S2), 1696 (R1), 1665 (R2), 1670 (R3), and 1676 (R4) cm−1 . The calculated vibrational frequency of S2 compares well with our previous reported value of 1685 cm−1 . 63 Our calculated vibrational frequency for S1 (1834 cm−1 ) and R1 (1696 cm−1 ) also agree well with the DFT results by Sorescu et al.: 38 1796 cm−1 (S1) and 1700 cm−1 (R1). He et al. 37 also obtained similar magnitudes: 1776 cm−1 (S1) and 1704 cm−1 (R1). He et al. attributed the large difference in vibrational frequency between the S1 and R1 cases to the large differences in bonding and geometry between the two sites. Our calculated vibrational frequencies for most adsorption modes (1665 to 1696 cm−1 ) are in good agreement with our experimental FTIR value of 1638 cm−1 for bent CO2 , indicating that bent CO2 indeed formed on these various surfaces. The DFT frequencies for bent CO2 are slightly higher than the experimental value, but it is known that DFT frequencies are often higher than experimental frequencies, and scaling factors may be used to get better agreement such as in He et al. 37 Bent CO2 over the stoichiometric surface (S1) however had a calculated vibrational frequency (∼ 1800 cm−1 ) not observed in the experimental FTIR data. Bent CO2 interacting with the TiO2 surface alone does not bind strongly, as our DFT results show, and accordingly this peak was not observed in the experimental results. At higher pressures and lower temperatures, binding of bent CO2 to the


14

Page 15 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

0.02 eV and the CO vibrational was 2178 cm−1 , only a small shift from the gas phase value. On the other hand, CO adsorbed on Cu (S2) or Cu with oxygen vacancy (R2/R3) resulted in a CO stretch frequency of 2089 (S2) and 2065 (R2/R3) cm−1 , as well as corresponding strong binding energies of -0.91 eV (S2) and 0.58 eV (R2/R3). CO adsorbed in an oxygen vacancy (R1) resulted in much stronger shifted value of 1929 cm−1 and very strong adsorption energy of -1.72 eV. Our calculations indicate that in the presence of Cu atoms and/or oxygen vacancies, the CO vibrational frequencies are shifted compared to the gas phase CO stretching frequency of 2161 cm−1 . Our experimental results displayed a peak at 2111 cm−1 for the Cu/TiO2 and Cu/Ov /TiO2 catalysts, attributed to CO adsorbed on Cu (see Figure 10). No vibrational peaks for CO in an oxygen vacancy were observed, likely due to the filling of oxygen vacancies upon CO2 dissociation. The DFT frequency for CO over Cu (2089 cm−1 ) is within ∼ 20 cm−1 of the experimental value, or in good general agreement. Our theoretical and experimental results demonstrate that CO2 photodissociation occurs predominantly in the presence of Cu and/or oxygen vacancies on the surface. We mention that our CO stretch frequencies of 2178 (S1) and 1929 cm−1 (R1) are comparable with the values of Lustemberg and Scherlis at 2178 (S1) and 1936 cm−1 (R1). 105 Other DFT work also provides similar values for the S1 CO vibrational frequency: Scaranto and Giorgianni 106 reported 2192 cm−1 , Mino et al. 107 reported 2185 cm−1 and Setvin et al. 108 reported 2182 cm−1 . An important consideration for the CO2 reduction cycle is the fate and regeneration of Ov . Our results show that Ov are likely healed as CO2 dissociates to fill the Ov . For a catalytic cycle involving Ov to occur, oxygen vacancies must form after being healed. After exposure of CO2 to the H2 -treated Cu/TiO2 catalyst and irradiation, adsorbed CO was observed in the difference FTIR spectrum (see Figure 10). We then purged the catalyst, and again exposed the catalyst to CO2 and UV light. We again detected CO, but the appropriate FTIR peak was greatly diminished, suggesting that Ov in

our catalysts had healed. Creating catalysts with defects is a challenge and ongoing area of research. 91,109 Viable strategies for Ov creation or regeneration include hydrogen treatment, thermal treatment, doping, and possibly other methods. We have demonstrated that atomic Cu and Ov can work together to enable CO2 photoreduction, and future work may concentrate on modifying the TiO2 nanoparticles, such as changing particle size or doping the particles, so that Ov can be readily regenerated during a catalytic cycle.

3.6

Pt/TiO2 , and CO2 Reduction

While we have focused on Cu/TiO2 catalysts, we also briefly consider Pt atoms adsorbed on TiO2 . Pt is a prototypical catalyst and has been studied for CO2 reduction using small clusters. 29,59–61,110 Our analysis could provide insight on how different metals may perform as supported single atom catalysts for CO2 reduction. In Figure 11a, the adsorption energies of Pt are shown over stoichiometric and reduced surfaces, as well as for comparison results of Cu. Corresponding geometries can be found in Figure S17. Pt atoms bound substantially stronger to all of the four sites compared to Cu. The Pt adsorption energy over the stoichiometric surface was -3.3 eV, while the adsorption energy was -5.4 eV for Pt in an Ov and Pt near the Ov , and -3.1 eV for Pt away from an Ov . Similar results were obtained by Gong et al. 50 who reported an adsorption energy of -4.71 eV for Pt in an Ov . Our calculations were at the PBE+D3+U level, while Gong et al. used PBE only, which could explain the slight difference in adsorption energies. In any case, Pt binds much stronger to the surface than Cu. Since Pt interacts differently than Cu with the TiO2 surface, it could influence the formation of Ov differently than Cu. We found Cu to have no real effect on Ov formation (see Section 3.1). In contrast, the Ov formation energy decreased by 2.13 eV and 2.24 eV when a Pt atom was present in/near Ov respectively, and increased by 0.13 eV when Pt was away from Ov (similar to the case of Cu away from the Ov ). This indicates that Pt could potentially increase the


15

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

number of Ov , which may have a strong effect on surface reactivity. Experimental work has shown for instance that Rh atoms can enable Ov formation. 88

Page 16 of 25

Cu/reduced TiO2 case. However, comparing Pt and Cu, bent CO2 in the R2 and R3 configurations was more strongly bound over Cu/TiO2 compared to Pt/TiO2 . This difference could be a result of the Pt charges (negative), where Pt was -0.51 e− and Cu was +0.02 e− for the R2 case. The two anions (Pt and CO2 ) could have more repulsion compared to the Cu/Ov /TiO2 case, since Cu was positively charged, leading to stronger CO2 adsorption over Cu/Ov /TiO2 . We also considered CO2 reactivity over Pt/TiO2 surfaces. Table 1 contains a summary of CO2 dissociation energies for Pt over TiO2 . The reaction energy for dissociation of CO2 was exothermic for the Pt/TiO2 (S2) case compared to an endothermic reaction energy for Cu/TiO2 . However, the activation energies over Pt and Cu were nearly the same over these catalysts, being 0.83 eV (Cu-S2) and 0.85 (PtS2). When Pt was in the Ov (R2) the CO2 dissociation energy was -1.08 eV, which was less exothermic than the Cu-R2 case (-1.71 eV ). The dissociation barrier was 0.43 eV, which is about twice the barrier for the Cu case (0.19 eV). The Pt-R2 and Pt-R3 configurations had similar reaction energies and activation barriers due to the similarity in geometries. These results show that Pt over stoichiometric surfaces may be a more promising catalyst compared to Cu, but that the strong synergy between metal atom and Ov is largely absent for Pt, while such strong synergy occurs for Cu. In this work, we did not model Pt/TiO2 with photoexcited electrons. However, in recent work by Yang et al., 61 a Pt tetramer with multiple photoexcited electrons on a TiO2 surface was shown to have a lowered barrier for CO2 reduction. We therefore could expect that photoexcited electrons could have a stabilizing effect for single Pt atoms, just like for Cu atoms.

Figure 11: Results from modeling Pt/TiO2 surfaces. (a) Adsorption energies of Pt adatoms compared to Cu adatoms. (b) Adsorption energies of bent and linear CO2 over Pt/TiO2 surfaces. In Figure 11b, we give the adsorption energies of linear and bent CO2 . In contrast to Cu, over the stoichiometric surface with Pt (S2) bent CO2 was preferred over linear CO2 by 0.21 eV. Notice also that for Pt, the S2 configuration has a close adsorption energy (near -0.64 eV) to that of the R2 and R3 configurations, indicating that Pt/Ov did not promote CO2 activation as well as Cu/Ov did. Over the Pt/reduced TiO2 surfaces, the bent CO2 binding geometries to the surface (R2, R3, R4 configuration) are nearly identical to the Cu case (see Figure S18). The bent CO2 adsorbed more strongly (in the range of 0.38 eV to 0.53 eV) than linear CO2 , which was the same trend we observed for the

4

Conclusions

In this work, we systematically investigated the effect of defects and excited electrons on CO2 reduction using TiO2 supported single atom metal photocatalysts. We first found using DFT that Cu binds strongly to both the sto-


16

Page 17 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

Figure 12: A summary of possible CO2 dissociation pathways as calculated by DFT. Given are reaction and activation energies. The dotted line suggests a ”cut-off” activation energy (< 0.6 eV), where activation energies below this line are expected to proceed reasonably fast. Several surfaces are good catalysts for CO2 reduction: Cu/TiO2 with photoexcited electron, Pt/Ov /TiO2 , Cu/Ov /TiO2 , and Cu/Ov /TiO2 with photoexcited electron. ichiometric and reduced TiO2 surfaces, indicating that single site Cu catalysts maybe stable over TiO2 . Activation of CO2 occurs as bent CO2 forms from the linear CO2 structure. Cu atoms in/near an oxygen vacancy were highly selective towards bent CO2 CO2 formation, in contrast to Cu over stoichiometric TiO2 . Furthermore, the Cu/Ov complex can lower the reaction barrier for CO2 dissociation (CO2 * → CO* + O*) to very low values (0.10 - 0.19 eV), which gives high catalytic production of CO. This spontaneous dissociation of CO2 also agrees with experimental results. 20,34 Our own experimental work showed that bent CO2 formed over Cu/TiO2 samples with measured adsorption intensities of Cu/Ov /TiO2 > Ov /TiO2 > Cu/TiO2 > TiO2 , in agreement with our theoretical calculations. Our photochemical studies also indicate CO2 dissociation most readily occurs on Cu/Ov /TiO2 surfaces. We also confirmed that a photoexcited electron may be beneficial in stabilizing bent CO2 on both Cu/TiO2 and Cu/Ov surfaces. Significant decreases in the reaction energy (∼1 eV) and barrier (0.26 eV) were calculated when an electron was present for Cu over stoichio-

metric TiO2 . Such an effect was not observed over reduced surfaces since such surfaces already had unpaired electron(s) present. These results imply that photocatalytic activity over activity over Cu/TiO2 should be higher than thermal catalytic activity, and details the direct role of photoexcited electrons in reducing CO2 . We also modeled Pt as an alternative to Cu, and found that Pt bound much stronger to TiO2 than Cu. Pt/Ov also showed selectivity to adsorbed bent CO2 over linear CO2 , which was similar to Cu/Ov . Reaction barriers for CO2 dissociation were low for the Pt/Ov surface (0.43 eV), although the strong synergy for Cu/Ov (reaction barriers 0.1 to 0.19 eV) was not observed. We have summarized our calculated reaction results in Figure 12. This figure shows that several catalysts may enable facile reduction of CO2 , and that Ov and photoexcited electrons may assist single atoms significantly in this process. Our work on single atomic catalysts for CO2 reduction highlights promising ways to tailor the catalyst, such as controlling surface defects or increasing photoexcitation yield, and further motivates research on a variety of supported single atom


17

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

photocatalysts.

Supporting Information Available

Page 18 of 25

(7) Fako, E.; Lodziana, Z.; López, N. Comparative single atom heterogeneous catalysts (SAHCs) on different platforms: a theoretical approach. Catal. Sci. Technol. 2017, 7, 4285–4293. (8) Liu, L.; Corma, A. Metal Catalysts for Heterogeneous Catalysis: From Single Atoms to Nanoclusters and Nanoparticles. Chem. Rev. 2018, 118, 4981–5079.

Additional details on the theoretical and experimental work. Acknowledgement The authors thank XSEDE for providing computational resources (TG-CTS170005). The authors also thank the WPI high-performance computing staff, especially Spencer Pruitt, for providing support and computational resources. This work was supported by the National Science Foundation through grants #CBET-1511672 and #CBET1510810.

(9) DeRita, L.; Dai, S.; Lopez-Zepeda, K.; Pham, N.; Graham, G. W.; Pan, X.; Christopher, P. Catalyst Architecture for Stable Single Atom Dispersion Enables Site-Specific Spectroscopic and Reactivity Measurements of CO Adsorbed to Pt Atoms, Oxidized Pt Clusters, and Metallic Pt Clusters on TiO2 . J. Am. Chem. Soc. 2017, 139, 14150–14165.

References

(10) Jones, J.; Xiong, H.; DeLaRiva, A. T.; Peterson, E. J.; Pham, H.; Challa, S. R.; Qi, G.; Oh, S.; Wiebenga, M. H.; Hernández, X. I. P. Thermally stable single-atom platinum-on-ceria catalysts via atom trapping. Science 2016, 353, 150–154.

(1) Flytzani-Stephanopoulos, M.; Gates, B. C. Atomically Dispersed Supported Metal Catalysts. Annu. Rev. Chem. Biomol. Eng. 2012, 3, 545–574. (2) Yang, X.-F.; Wang, A.; Qiao, B.; Li, J.; Liu, J.; Zhang, T. Single-Atom Catalysts: A New Frontier in Heterogeneous Catalysis. Acc. Chem. Res. 2013, 46, 1740–1748.

(11) Qiao, B.; Wang, A.; Yang, X.; Allard, L. F.; Jiang, Z.; Cui, Y.; Liu, J.; Li, J.; Zhang, T. Single-atom catalysis of CO oxidation using Pt1 /FeOx . Nat. Chem. 2011, 3, 634–641.

(3) Vajda, S.; White, M. G. Catalysis Applications of Size-Selected Cluster Deposition. ACS Catal. 2015, 5, 7152–7176.

(12) Matsubu, J. C.; Yang, V. N.; Christopher, P. Isolated Metal Active Site Concentration and Stability Control Catalytic CO2 Reduction Selectivity. J. Am. Chem. Soc. 2015, 137, 3076–3084.

(4) Peng, C.; Reid, G.; Wang, H.; Hu, P. Perspective: Photocatalytic reduction of CO2 to solar fuels over semiconductors. J. Chem. Phys. 2017, 147, 030901.

(13) Kwak, J. H.; Kovarik, L.; Szanyi, J. Heterogeneous Catalysis on Atomically Dispersed Supported Metals: CO2 Reduction on Multifunctional Pd Catalysts. ACS Catal. 2013, 3, 2094–2100.

(5) Liu, J. Catalysis by Supported Single Metal Atoms. ACS Catal. 2017, 7, 34– 59. (6) Chen, Y.; Huang, Z.; Ma, Z.; Chen, J.; Tang, X. Fabrication, characterization, and stability of supported single-atom catalysts. Catal. Sci. Technol. 2017, 7, 4250–4258.

(14) Inoue, T.; Fujishima, A.; Konishi, S.; Honda, K. Photoelectrocatalytic reduction of carbon dioxide in aqueous suspensions of semiconductor powders. Nature 1979, 277, 637–638.


18

Page 19 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

(15) Habisreutinger, S. N.; SchmidtMende, L.; Stolarczyk, J. K. Photocatalytic reduction of CO2 on TiO2 and other semiconductors. Angew. Chem. Int. Ed. 2013, 52, 7372–7408.

(23) Yan, Y.; Yu, Y.; Cao, C.; Huang, S.; Yang, Y.; Yang, X.; Cao, Y. Enhanced photocatalytic activity of TiO2 –Cu/C with regulation and matching of energy levels by carbon and copper for photoreduction of CO2 into CH4 . CrystEngComm 2016, 18, 2956–2964.

(16) Cheng, D.; Negreiros, F. R.; Aprà, E.; Fortunelli, A. Computational Approaches to the Chemical Conversion of Carbon Dioxide. ChemSusChem 2013, 6, 944–965.

(24) Zhao, H.; Pan, F.; Li, Y. A review on the effects of TiO2 surface point defects on CO2 photoreduction with H2 O. J. Materiomics 2017, 3, 17–32.

(17) Ma, Y.; Wang, X.; Jia, Y.; Chen, X.; Han, H.; Li, C. Titanium Dioxide-Based Nanomaterials for Photocatalytic Fuel Generations. Chem. Rev. 2014, 114, 9987–10043.

(25) Liu, C.; Yang, B.; Tyo, E.; Seifert, S.; DeBartolo, J.; von Issendorff, B.; Zapol, P.; Vajda, S.; Curtiss, L. A. Carbon dioxide conversion to methanol over size-selected Cu4 clusters at low pressures. J. Am. Chem. Soc. 2015, 137, 8676–8679.

(18) White, J. L.; Baruch, M. F.; Pander, J. E.; Hu, Y.; Fortmeyer, I. C.; Park, J. E.; Zhang, T.; Liao, K.; Gu, J.; Yan, Y.; Shaw, T. W.; Abelev, E.; Bocarsly, A. B. Light-Driven Heterogeneous Reduction of Carbon Dioxide: Photocatalysts and Photoelectrodes. Chem. Rev. 2015, 115, 12888–12935.

(26) Sharma, P. K.; Cortes, M. A. L. R. M.; Hamilton, J. W. J.; Han, Y.; Byrne, J. A.; Nolan, M. Surface modification of TiO2 with copper clusters for band gap narrowing. Catal. Today 2017, In Press.

(19) Chang, X.; Wang, T.; Gong, J. CO2 photo-reduction: insights into CO2 activation and reaction on surfaces of photocatalysts. Energy Environ. Sci. 2016, 9, 2177–2196.

(27) Yang, B.; Liu, C.; Halder, A.; Tyo, E. C.; Martinson, A. B. F.; Seifert, S.; Zapol, P.; Curtiss, L. A.; Vajda, S. Copper Cluster Size Effect in Methanol Synthesis from CO2 . J. Phys. Chem. C 2017, 121, 10406–10412.

(20) Liu, L.; Zhao, C.; Li, Y. Spontaneous Dissociation of CO2 to CO on Defective Surface of Cu(I)/TiO2−x Nanoparticles at Room Temperature. J. Phys. Chem. C 2012, 116, 7904–7912.

(28) Yan, Y.; Yu, Y.; Huang, S.; Yang, Y.; Yang, X.; Yin, S.; Cao, Y. Adjustment and Matching of Energy Band of TiO2 Based Photocatalysts by Metal Ions (Pd, Cu, Mn) for Photoreduction of CO2 into CH4 . J. Phys. Chem. C 2017, 121, 1089– 1098.

(21) Liu, C.; He, H.; Zapol, P.; A. Curtiss, L. Computational studies of electrochemical CO2 reduction on subnanometer transition metal clusters. Phys. Chem. Chem. Phys. 2014, 16, 26584–26599.

(29) Gao, G.; Jiao, Y.; Waclawik, E. R.; Du, A. Single Atom (Pd/Pt) Supported on Graphitic Carbon Nitride as an Efficient Photocatalyst for Visible-Light Reduction of Carbon Dioxide. J. Am. Chem. Soc. 2016, 138, 6292–6297.

(22) Park, M.; Kwak, B. S.; Jo, S. W.; Kang, M. Effective CH4 production from CO2 photoreduction using TiO2 /x mol% Cu-TiO2 double-layered films. Energy Convers. Manage. 2015, 103, 431–438.

(30) Li, H.; Wang, L.; Dai, Y.; Pu, Z.; Lao, Z.; Chen, Y.; Wang, M.; Zheng, X.; Zhu, J.;


19

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Zhang, W.; Si, R.; Ma, C.; Zeng, J. Synergetic interaction between neighbouring platinum monomers in CO2 hydrogenation. Nat. Nanotechnol. 2018, 13, 411– 417.

Page 20 of 25

(37) He, H.; Zapol, P.; Curtiss, L. A. A Theoretical Study of CO2 Anions on Anatase (101) Surface. J. Phys. Chem. C 2010, 114, 21474–21481. (38) Sorescu, D. C.; Al-Saidi, W. A.; Jordan, K. D. CO2 adsorption on TiO2 (101) anatase: A dispersion-corrected density functional theory study. J. Chem. Phys. 2011, 135, 124701.

(31) Chen, X.; Su, X.; Su, H.-Y.; Liu, X.; Miao, S.; Zhao, Y.; Sun, K.; Huang, Y.; Zhang, T. Theoretical Insights and the Corresponding Construction of Supported Metal Catalysts for Highly Selective CO2 to CO Conversion. ACS Catal. 2017, 7, 4613–4620.

(39) He, H.; Zapol, P.; Curtiss, L. A. Computational screening of dopants for photocatalytic two-electron reduction of CO2 on anatase (101) surfaces. Energy Environ. Sci. 2012, 5, 6196–6205.

(32) Gao, C.; Chen, S.; Wang, Y.; Wang, J.; Zheng, X.; Zhu, J.; Song, L.; Zhang, W.; Xiong, Y. Heterogeneous Single-Atom Catalyst for Visible-Light-Driven HighTurnover CO2 Reduction: The Role of Electron Transfer. Adv. Mater. 2018, 30, 1704624.

(40) Cheng, Z.; Sherman, B. J.; Lo, C. S. Carbon dioxide activation and dissociation on ceria (110): A density functional theory study. J. Chem. Phys. 2013, 138, 014702.

(33) Zhang, H.; Wei, J.; Dong, J.; Liu, G.; Shi, L.; An, P.; Zhao, G.; Kong, J.; Wang, X.; Meng, X.; Zhang, J.; Ye, J. Efficient Visible-Light-Driven Carbon Dioxide Reduction by a SingleAtom Implanted Metal–Organic Framework. Angew. Chem. 2016, 128, 14522– 14526.

(41) Fronzi, M.; Daly, W.; Nolan, M. Reactivity of metal oxide nanocluster modified rutile and anatase TiO2 : Oxygen vacancy formation and CO2 interaction. Appl. Catal., A 2016, 521, 240–249. (42) Ji, Y.; Luo, Y. New Mechanism for Photocatalytic Reduction of CO2 on the Anatase TiO2 (101) Surface: The Essential Role of Oxygen Vacancy. J. Am. Chem. Soc. 2016, 138, 15896–15902.

(34) Liu, L.; Gao, F.; Zhao, H.; Li, Y. Tailoring Cu valence and oxygen vacancy in Cu/TiO2 catalysts for enhanced CO2 photoreduction efficiency. Appl. Catal., B 2013, 134, 349–358.

(43) Haa, M.-A.; Alexandrova, A. N. Oxygen Vacancies of Anatase (101): Extreme Sensitivity to the Density Functional Theory Method. J. Chem. Theory Comput. 2016, 12, 2889–2895.

(35) Nolan, M.; Elliott, S. D.; Mulley, J. S.; Bennett, R. A.; Basham, M.; Mulheran, P. Electronic structure of point defects in controlled self-doping of the TiO2 (110) surface: Combined photoemission spectroscopy and density functional theory study. Phys. Rev. B 2008, 77, 235424.

(44) Yin, W.-J.; Wen, B.; Bandaru, S.; Krack, M.; Lau, M.; Liu, L.-M. The Effect of Excess Electron and hole on CO2 Adsorption and Activation on Rutile (110) surface. Sci. Rep. 2016, 6, 23298.

(36) Cheng, H.; Selloni, A. Energetics and diffusion of intrinsic surface and subsurface defects on anatase TiO2 (101). J. Chem. Phys. 2009, 131, 054703.

(45) Billo, T.; Fu, F.-Y.; Raghunath, P.; Shown, I.; Chen, W.-F.; Lien, H.-T.; Shen, T.-H.; Lee, J.-F.; Chan, T.-S.; Huang, K.-Y. Ni-Nanocluster Modified


20

Page 21 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

Black TiO2 with Dual Active Sites for Selective Photocatalytic CO2 Reduction. Small 2018, 14 .

(54) Ma, S.; Song, W.; Liu, B.; Zhong, W.; Deng, J.; Zheng, H.; Liu, J.; Gong, X.-Q.; Zhao, Z. Facet-dependent photocatalytic performance of TiO2 : A DFT study. Appl. Catal., B 2016, 198, 1–8.

(46) Giordano, L.; Pacchioni, G.; Bredow, T.; Sanz, J. F. Cu, Ag, and Au atoms adsorbed on TiO2 (110): cluster and periodic calculations. Surf. Sci. 2001, 471, 21–31.

(55) Ma, S.; Song, W.; Liu, B.; Zheng, H.; Deng, J.; Zhong, W.; Liu, J.; Gong, X.Q.; Zhao, Z. Elucidation of the high CO2 reduction selectivity of isolated Rh supported on TiO2 : a DFT study. Catal. Sci. Technol. 2016, 6, 6128–6136.

(47) Lopez, N.; Nørskov, J. K. Theoretical study of the Au/TiO2 (110) interface. Surf. Sci. 2002, 515, 175–186.

(56) Schlexer, P.; Chen, H.-Y. T.; Pacchioni, G. CO2 Activation and Hydrogenation: A Comparative DFT Study of Ru10/TiO2 and Cu10/TiO2 Model Catalysts. Catal. Lett. 2017, 147, 1871–1881.

(48) Campbell, C. T.; Starr, D. E. Metal Adsorption and Adhesion Energies on MgO (100). J. Am. Chem. Soc. 2002, 124, 9212–9218. (49) Han, Y.; Liu, C.-j.; Ge, Q. Effect of Surface Oxygen Vacancy on Pt Cluster Adsorption and Growth on the Defective Anatase TiO2 (101) Surface. J. Phys. Chem. C 2007, 111, 16397–16404.

(57) Wang, H.; An, T.; Selloni, A. Effect of reducible oxide–metal cluster charge transfer on the structure and reactivity of adsorbed Au and Pt atoms and clusters on anatase TiO2 . J. Chem. Phys. 2017, 146, 184703.

(50) Gong, X.-Q.; Selloni, A.; Dulub, O.; Jacobson, P.; Diebold, U. Small Au and Pt Clusters at the Anatase TiO2 (101) Surface: Behavior at Terraces, Steps, and Surface Oxygen Vacancies. J. Am. Chem. Soc. 2008, 130, 370–381.

(58) Alghannam, A.; Muhich, C. L.; Musgrave, C. B. Adatom surface diffusion of catalytic metals on the anatase TiO2 (101) surface. Phys. Chem. Chem. Phys. 2017, 19, 4541–4552.

(51) Cui, L.; Tang, Y.; Zhang, H.; Hector, L. G.; Ouyang, C.; Shi, S.; Li, H.; Chen, L. First-principles investigation of transition metal atom M (M = Cu, Ag, Au) adsorption on CeO2 (110). Phys. Chem. Chem. Phys. 2012, 14, 1923– 1933.

(59) Yang, C.-T.; Balakrishnan, N.; Bhethanabotla, V. R.; Joseph, B. Interplay between Subnanometer Ag and Pt Clusters and Anatase TiO2 (101) Surface: Implications for Catalysis and Photocatalysis. J. Phys. Chem. C 2014, 118, 4702–4714.

(52) Ong, S. V.; Khanna, S. N. Theoretical Studies of the Stability and Oxidation of Pdn (n = 1–7) Clusters on Rutile TiO2 (110): Adsorption on the Stoichiometric Surface. J. Phys. Chem. C 2012, 116, 3105–3111.

(60) Yang, C.-T.; Wood, B. C.; Bhethanabotla, V. R.; Joseph, B. CO2 Adsorption on Anatase TiO2 (101) Surfaces in the Presence of Subnanometer Ag/Pt Clusters: Implications for CO2 Photoreduction. J. Phys. Chem. C 2014, 118, 26236–26248.

(53) Seriani, N.; Pinilla, C.; Crespo, Y. Presence of Gap States at Cu/TiO2 Anatase Surfaces: Consequences for the Photocatalytic Activity. J. Phys. Chem. C 2015, 119, 6696–6702.

(61) Yang, C.-T.; Wood, B. C.; Bhethanabotla, V. R.; Joseph, B.


21

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Electron injection study of photoexcitation effects on supported subnanometer Pt clusters for CO2 photoreduction. Phys. Chem. Chem. Phys. 2018, 20, 15926–15938.

Page 22 of 25

a mixed Gaussian and plane waves approach. Comput. Phys. Commun. 2005, 167, 103–128. (70) VandeVondele, J.; Hutter, J. Gaussian basis sets for accurate calculations on molecular systems in gas and condensed phases. J. Chem. Phys. 2007, 127, 114105.

(62) Liu, C.; Iyemperumal, S. K.; Deskins, N. A.; Li, G. Photocatalytic CO2 reduction by highly dispersed Cu sites on TiO2 . J. Photonics Energy 2016, 7, 012004–012004.

(71) Goedecker, S.; Teter, M.; Hutter, J. Separable dual-space Gaussian pseudopotentials. Phys. Rev. B 1996, 54, 1703.

(63) Iyemperumal, S. K.; Aaron Deskins, N. Activation of CO2 by supported Cu clusters. Phys. Chem. Chem. Phys. 2017, 19, 28788–28807.

(72) Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized gradient approximation made simple. Phys. Rev. Lett. 1996, 77, 3865.

(64) Liu, L.; Zhao, H.; Andino, J. M.; Li, Y. Photocatalytic CO2 Reduction with H2 O on TiO2 Nanocrystals: Comparison of Anatase, Rutile, and Brookite Polymorphs and Exploration of Surface Chemistry. ACS Catal. 2012, 2, 1817– 1828.

(73) Grimme, S.; Antony, J.; Ehrlich, S.; Krieg, H. A consistent and accurate ab initio parametrization of density functional dispersion correction (DFT-D) for the 94 elements H-Pu. J. Chem. Phys. 2010, 132, 154104.

(65) Pipornpong, W.; Wanbayor, R.; Ruangpornvisuti, V. Adsorption CO2 on the perfect and oxygen vacancy defect surfaces of anatase TiO2 and its photocatalytic mechanism of conversion to CO. Appl. Surf. Sci. 2011, 257, 10322–10328.

(74) Grimme, S.; Ehrlich, S.; Goerigk, L. Effect of the damping function in dispersion corrected density functional theory. J. Comput. Chem. 2011, 32, 1456–1465.

(66) Huygh, S.; Bogaerts, A.; Neyts, E. C. How Oxygen Vacancies Activate CO2 Dissociation on TiO2 Anatase (001). J. Phys. Chem. C 2016, 120, 21659–21669.

(75) Dudarev, S. L.; Botton, G. A.; Savrasov, S. Y.; Humphreys, C. J.; Sutton, A. P. Electron-energy-loss spectra and the structural stability of nickel oxide: An LSDA+ U study. Phys. Rev. B 1998, 57, 1505.

(67) Hutter, J.; Iannuzzi, M.; Schiffmann, F.; VandeVondele, J. cp2k: atomistic simulations of condensed matter systems. WIREs Comput. Mol. Sci. 2014, 4, 15– 25.

(76) Dudarev, S. L.; Manh, D. N.; Sutton, A. P. Effect of Mott-Hubbard correlations on the electronic structure and structural stability of uranium dioxide. Philos. Mag. B 1997, 75, 613–628.

(68) CP2K is freely available https://www.cp2k.org/. 2018.

(77) Sel¸cuk, S.; Selloni, A. Influence of external electric fields on oxygen vacancies at the anatase (101) surface. J. Chem. Phys. 2014, 141, 084705.

from:

(69) VandeVondele, J.; Krack, M.; Mohamed, F.; Parrinello, M.; Chassaing, T.; Hutter, J. Quickstep: Fast and accurate density functional calculations using

(78) Hu, Z.; Metiu, H. Choice of U for DFT+U Calculations for Titanium Oxides. J. Phys. Chem. C 2011, 115, 5841– 5845.


22

Page 23 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

(79) Manz, T. A.; Limas, N. G. Introducing DDEC6 atomic population analysis: part 1. Charge partitioning theory and methodology. RSC Adv. 2016, 6, 47771– 47801.

of anionic copper clusters and copper cluster monocarbonyls. J. Chem. Phys. 2000, 112, 1713–1720. (88) Guan, H.; Lin, J.; Qiao, B.; Miao, S.; Wang, A.-Q.; Wang, X.; Zhang, T. Enhanced performance of Rh1/TiO2 catalyst without methanation in water-gas shift reaction. AIChE J. 2017, 63, 2081– 2088.

(80) Gabaldon Limas, N.; A. Manz, T. Introducing DDEC6 atomic population analysis: part 2. Computed results for a wide range of periodic and nonperiodic materials. RSC Adv. 2016, 6, 45727–45747.

(89) Hou, L.; Zhang, M.; Guan, Z.; Li, Q.; Yang, J. Effect of annealing ambience on the formation of surface/bulk oxygen vacancies in TiO2 for photocatalytic hydrogen evolution. Appl. Surf. Sci. 2018, 428, 640 – 647.

(81) Bader, R. F. Atoms in molecules. Acc. Chem. Res. 1985, 18, 9–15. (82) Henkelman, G.; Uberuaga, B. P.; Jónsson, H. A climbing image nudged elastic band method for finding saddle points and minimum energy paths. J. Chem. Phys. 2000, 113, 9901–9904.

(90) Bharti, B.; Kumar, S.; Lee, H.-N.; Kumar, R. Formation of oxygen vacancies and Ti3+ state in TiO2 thin film and enhanced optical properties by air plasma treatment. Sci. Rep. 2016, 6, 32355.

(83) Henkelman, G.; Jónsson, H. Improved tangent estimate in the nudged elastic band method for finding minimum energy paths and saddle points. J. Chem. Phys. 2000, 113, 9978–9985.

(91) Wang, G.; Yang, Y.; Han, D.; Li, Y. Oxygen defective metal oxides for energy conversion and storage. Nano Today 2017, 13, 23–39.

(84) Carl, A. D.; Kalan, R. E.; Obayemi, J. D.; Zebaze Kana, M. G.; Soboyejo, W. O.; Grimm, R. L. Synthesis and Characterization of AlkylamineFunctionalized Si(111) for Perovskite Adhesion With Minimal Interfacial Oxidation or Electronic Defects. ACS Appl. Mater. Interfaces 2017, 9, 34377–34388.

(92) Ou, G.; Xu, Y.; Wen, B.; Lin, R.; Ge, B.; Tang, Y.; Liang, Y.; Yang, C.; Huang, K.; Zu, D.; Yu, R.; Chen, W.; Li, J.; Wu, H.; Liu, L.-M.; Li, Y. Tuning defects in oxides at room temperature by lithium reduction. Nat. Commun. 2018, 9, 1302.

(85) Panta, R.; Ruangpornvisuti, V. Adsorption of hydrogen molecule on noble metal doped on oxygen-vacancy defect of anatase TiO2 (101) surface: Periodic DFT study. Int. J. Hydrogen Energy 2017, 42, 19106–19113.

(93) Zuo, F.; Wang, L.; Wu, T.; Zhang, Z.; Borchardt, D.; Feng, P. Self-Doped Ti3+ Enhanced Photocatalyst for Hydrogen Production under Visible Light. J. Am. Chem. Soc. 2010, 132, 11856–11857.

(86) Chen, H.-Y. T.; Tosoni, S.; Pacchioni, G. Adsorption of Ruthenium Atoms and Clusters on Anatase TiO2 and Tetragonal ZrO2 (101) Surfaces: A Comparative DFT Study. J. Phys. Chem. C 2015, 119, 10856–10868.

(94) Freund, H.-J.; Roberts, M. W. Surface chemistry of carbon dioxide. Surf. Sci. Rep. 1996, 25, 225–273. (95) Deskins, N. A.; Rousseau, R.; Dupuis, M. Defining the Role of Excess Electrons in the Surface Chemistry of TiO2 . J. Phys. Chem. C 2010, 114, 5891–5897.

(87) Spasov, V. A.; Lee, T.-H.; Ervin, K. M. Threshold collision-induced dissociation


23

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(96) Garcia, J. C.; Deskins, N. A. Detailing Ionosorption over TiO2 , ZrO2 , and HfO2 from First Principles. J. Phys. Chem. C 2012, 116, 16573–16581.

Page 24 of 25

(104) Rasko, J.; Solymosi, F. Infrared Spectroscopic Study of the Photoinduced Activation of CO2 on TiO2 and Rh/TiO2 Catalysts. J. Phys. Chem. 1994, 98, 7147– 7152.

(97) Rodriguez, M. M.; Peng, X.; Liu, L.; Li, Y.; Andino, J. M. A Density Functional Theory and Experimental Study of CO2 Interaction with Brookite TiO2 . J. Phys. Chem. C 2012, 116, 19755–19764.

(105) Lustemberg, P. G.; Scherlis, D. A. Monoxide carbon frequency shift as a tool for the characterization of TiO2 surfaces: Insights from first principles spectroscopy. J. Chem. Phys. 2013, 138, 124702.

(98) Di Valentin, C.; Fittipaldi, D. Hole Scavenging by Organic Adsorbates on the TiO2 Surface: A DFT Model Study. J. Phys. Chem. Lett. 2013, 4, 1901–1906.

(106) Scaranto, J.; Giorgianni, S. DFT calculations of carbon monoxide adsorbed on anatase TiO2 (101) and (001) surfaces: correlation between the binding energy and the CO stretching frequency. Mol. Simul. 2013, 39, 245–249.

(99) Zhang, J.; Peng, C.; Wang, H.; Hu, P. Identifying the Role of Photogenerated Holes in Photocatalytic Methanol Dissociation on Rutile TiO2 (110). ACS Catal. 2017, 7, 2374–2380.

(107) Mino, L.; Ferrari, A. M.; Lacivita, V.; Spoto, G.; Bordiga, S.; Zecchina, A. CO Adsorption on Anatase Nanocrystals: A Combined Experimental and Periodic DFT Study. J. Phys. Chem. C 2011, 115, 7694–7700.

(100) Wanbayor, R.; Deák, P.; Frauenheim, T.; Ruangpornvisuti, V. First principles theoretical study of the hole-assisted conversion of CO to CO2 on the anatase TiO2 (101) surface. J. Chem. Phys. 2011, 134, 104701.

(108) Setvin, M.; Franchini, C.; Hao, X.; Schmid, M.; Janotti, A.; Kaltak, M.; Van de Walle, C. G.; Kresse, G.; Diebold, U. Direct View at Excess Electrons in TiO2 Rutile and Anatase. Phys. Rev. Lett. 2014, 113, 086402.

(101) Li, Y.-F.; Selloni, A. Theoretical Study of Interfacial Electron Transfer from Reduced Anatase TiO2 (101) to Adsorbed O2 . J. Am. Chem. Soc. 2013, 135, 9195– 9199.

(109) Pan, X.; Yang, M.-Q.; Fu, X.; Zhang, N.; Xu, Y.-J. Defective TiO2 with oxygen vacancies: synthesis, properties and photocatalytic applications. Nanoscale 2013, 5, 3601–3614.

(102) Wang, D.; Wang, H.; Hu, P. Identifying the distinct features of geometric structures for hole trapping to generate radicals on rutile TiO2 (110) in photooxidation using density functional theory calculations with hybrid functional. Phys. Chem. Chem. Phys. 2015, 17, 1549– 1555.

(110) Umezawa, N.; Kristoffersen, H. H.; Vilhelmsen, L. B.; Hammer, B. r. Reduction of CO2 with Water on Pt-Loaded Rutile TiO2 (110) Modeled with Density Functional Theory. J. Phys. Chem. C 2016, 120, 9160–9164.

(103) Dubois, K. D.; Petushkov, A.; Garcia Cardona, E.; Larsen, S. C.; Li, G. Adsorption and Photochemical Properties of a Molecular CO2 Reduction Catalyst in Hierarchical Mesoporous ZSM-5: An In Situ FTIR Study. J. Phys. Chem. Lett. 2012, 3, 486–492.


24

Page 25 of 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

Graphical TOC Entry


25

Synergy between Defects, Photoexcited Electrons ... - ACS Publications

Recommend Documents