A Density Functional Theory and Experimental Study of CO2

Aug 16, 2012 - The interactions of CO2 with the (210) surface of brookite TiO2 were studied using first-principle calculations on cluster and periodic...
1 downloads 0 Views 4MB Size
Article pubs.acs.org/JPCC

A Density Functional Theory and Experimental Study of CO2 Interaction with Brookite TiO2 Monique M. Rodriguez,† Xihong Peng,‡ Lianjun Liu,§ Ying Li,§ and Jean M. Andino*,†,∥ †

Chemical Engineering, Arizona State University, PO Box 876106, Tempe, Arizona 85287, United States Applied Sciences and Mathematics, Arizona State University, Polytechnic Campus, Mesa, Arizona 85212, United States § Mechanical Engineering, University of WisconsinMilwaukee, 3200 North Cramer Street, Milwaukee, Wisconsin 53211, United States ∥ Civil, Environmental, and Sustainable Engineering, Arizona State University, PO Box 876106, Tempe, Arizona 85287, United States ‡

S Supporting Information *

ABSTRACT: The interactions of CO2 with the (210) surface of brookite TiO2 were studied using first-principle calculations on cluster and periodic slab systems. Charge and spin density analyses were implemented to determine if charge transfer to the CO2 molecule occurred and whether this charge transfer was comparable to that seen with the anatase TiO2 (101) surface. Although the brookite (210) surface provided energetically similar CO2 interactions as compared to the anatase (101) surface, the brookite surface had negligible charge transfer to the CO2 molecule. This result suggests that unmodified brookite is not a suitable catalyst for the reduction of CO2. However, the results also suggest that modification of the brookite surface through the creation of oxygen vacancies may lead to enhancements in CO2 reduction. The computational results were supported with laboratory data for CO2 interaction with perfect brookite and oxygen-deficient brookite. The laboratory data, generated using diffuse reflectance Fourier transform infrared spectroscopy, confirms the presence of CO2− at significant levels on the oxygen-deficient brookite.



materials.4,8,10 However, the CO2 conversion efficiency to useful fuel products currently is rather low for industrial applications.9 It is thought that the formation of the CO2− on the surface of TiO2 is the first step in CO2 reduction,19 and experimental evidence demonstrates the existence of CO2− on irradiated TiO2.20 He et al.12 performed first-principle calculations on cluster and periodic models of the anatase TiO2 (101) surface, the most exposed anatase surface, to determine the binding configurations of CO2. The results of their research show that CO2− forms on the surface of anatase TiO2. As stated previously, CO2 reduction over rutile, anatase, and mixtures of these two phases have been performed. Research has yet to be published concerning CO2 reduction over brookite TiO2, but recent studies have suggested that brookite is a good photocatalyst and may exhibit higher photocatalytic activity than both rutile and anatase.16,21,22 Specifically, density functional theory (DFT) calculations performed on the brookite (210) and anatase (101) surfaces demonstrated that the brookite (210) surface has similar structural building blocks as the anatase (101) surface.21 Both the anatase (101) and

INTRODUCTION With increasing greenhouse gas emissions and depleting energy resources, interest has arisen in the development of environmentally clean and safe processes that are capable of expanding our energy infrastructure. Because carbon dioxide has been identified as a significant greenhouse gas, it is critical to find methods for removing this compound from the atmosphere. Carbon capture and sequestration in geological formations is a current option for removing CO2, but uncertainties exist regarding permanent storage of the CO2.1 A promising solution that could simultaneously utilize CO2 and produce potential fuel products is the photocatalytic reduction of CO2. In recent years, researchers have investigated the photocatalytic reduction of CO2 over various modified and unmodified forms of titania (TiO2).2−12 TiO2 has been the photocatalyst of choice because of its stability, low cost, and nontoxicity. It exists in three naturally occurring forms: rutile, anatase, and brookite. The rutile and anatase phases are the more extensively studied, with anatase being regarded as the more photocatalytically active phase.13 The brookite phase is the least studied, partly due to past difficulties in creating pure brookite samples.14−18 When examining the CO2 reduction process, only rutile, anatase, and rutile−anatase mixtures have been explored, either as pure phase components or with metal and nonmetal modifications to the anatase or rutile © 2012 American Chemical Society

Received: March 11, 2012 Revised: August 15, 2012 Published: August 16, 2012 19755

dx.doi.org/10.1021/jp302342t | J. Phys. Chem. C 2012, 116, 19755−19764

The Journal of Physical Chemistry C

Article

photoexcited electron. The oxygen-deficient brookite (210) cluster (Ti7O27H28) was generated by removing a bridging oxygen atom in the neutral brookite cluster. Four different configurations were created for the oxygen-deficient brookite cluster, similar to the work reported by He et al.12 for the anatase system. All calculations of the cluster models were performed using DFT with the B3LYP functional.25,26 The calculations were implemented using the Gaussian 03 program package.27 Optimizations of the individual species (CO2, neutral and negatively charged Ti7O28H28, and Ti7O27H28) were performed first. The geometries of the various CO2−Ti7O28H28 configurations and CO2−Ti7O27H28 were then optimized. The adsorption energies of the various CO2−Ti7O28H28 and CO2−Ti7O27H28 configurations were calculated using the following equation:

brookite (210) surfaces have blocks consisting of exposed 5and 6-coordinated Ti atoms and 2- and 3-coordinated O atoms. The brookite surface has slightly shorter interatomic distances and a different block arrangement.21 These features assist in the generation of highly active sites and stronger adsorptions of various molecules on the brookite surface.21 Because the brookite phase has not been explored for CO2 photoreduction, yet the literature suggests that this phase has increased photocatalytic activity, it is of keen interest to see how CO2 interacts with this TiO2 polymorph. The purpose of this work was to use computational chemistry to investigate the interaction of CO2 and the formation of CO2− on the brookite (210) surface, in its neutral, negatively charged, and oxygendeficient states. Ultimately, the goal was to determine whether the brookite (210) surface could more readily facilitate charge transfer to CO2 as compared to the anatase TiO2 (101) surface. The brookite (210) surface was chosen because it is one of the most frequently exposed and stable surfaces of the brookite phase with similar structural features to the anatase (101) surface.23 Additional experimental work was performed using perfect brookite and oxygen-deficient brookite to provide supporting evidence for the computational work. Oxygen vacancies (Vo) on the TiO2 surface lead to the redistribution of two electrons, creating Ti3+ sites that are highly reactive due to their unsaturated coordination.24 Computational work performed by Pan et al.22 demonstrated that oxygen defects were more easily created in brookite than in anatase or rutile, and increased the photocatalytic ability in the visible light region. It was shown that oxygen defects on the anatase (101) surface facilitate CO2 adsorption and charge transfer to the CO2 molecule.12 Since brookite is more readily reduced and oxygen vacancies appear to increase the reactivity of Ti sites, it is of interest to see how oxygen vacancies affect the interaction of the CO2 molecule with the surface. Computational work was performed to model the interactions of CO2 on the neutral and negatively charged brookite TiO2 (210) surfaces as well as on the oxygen-deficient surface. The results were compared with previously published findings for the interaction of CO2 on the anatase TiO2 (101) surface.12 To our knowledge, this is the first computational study of the interaction of CO2 with the brookite phase. The information obtained through this work may provide new avenues for further exploration of brookite and mixed-phase brookite TiO2 as potential photocatalysts for CO2 reduction.

Eadsorption = ECO2 − Ti cluster − (ECO2 + E Ti cluster)

All hydroxyl groups were frozen during optimization to mimic the surface environment. The 6-31+G(2df,p) basis set28−31 was used for the CO2 molecule and the five surface oxygen atoms of the brookite cluster. The LanL2Dz basis set32−34 was used for the remaining oxygen atoms and all titanium and hydrogen atoms. Vibrational frequencies for all configurations were calculated and scaled by a factor of 0.9652.35 Natural bond orbital (NBO) charge analysis was performed for all cluster calculations. Calculations were performed on the optimized cluster models to check for wave function instabilities and to correct for basis set superposition errors (BSSE). Single-point energy calculations were also performed on the optimized configurations using the PBE functional but using the same basis sets as those used in the original B3LYP calculations. Finally, the clusters were also optimized using the B3LYP functional and the 6-31+G(d) basis set for the CO2 and the five surface oxygen atoms, while all other atoms were represented by the LanL2DZ basis set. Adsorption energies for the BSSE corrected calculations, PBE functional calculations, and the smaller basis calculations are presented in the Supporting Information document. In the periodic slab model, the brookite (210) surface was represented using a 2 × 1 supercell along the [001] and [12̅0] directions. The surface slab model was reduced from the brookite bulk with lattice constants of a = 9.183 Å, b = 5.455 Å, and c = 5.134 Å (lattice constants were obtained using the computational criteria listed below and are very close to the experimental values36). A slab thickness of six TiO2 trilayers (six O−Ti−O stacks) with eight Ti atoms in each layer was chosen. The bottom trilayers were fixed during optimization to mimic the bulk, while the coordinates of the remaining atoms were allowed to relax until the forces on each atom were less than 0.03 eV/Å. A vacuum layer of about 11 Å was introduced in the z direction (perpendicular to the surface) to separate the slab from its images, resulting from the periodic boundary conditions. The periodic slab calculations were performed using the Vienna Ab-initio Simulation Package (VASP).37,38 The PBE functional39 and projector augmented wave (PAW) potentials40 were used along with plane wave basis sets. The energy cutoff for the plane wave basis set was set to be 400 eV. The total energy was converged to within 0.01 meV. Reciprocal space was sampled at Γ only because the size of the supercell is relatively large (10.268 and 14.261 Å along the x and y directions, respectively). The Ti 2p, 3d, and 4s and O 2s and 2p



METHODS Computational Methods. In this study, the interactions of gas-phase CO2 on neutral, negatively charged, and oxygendeficient brookite (210) surfaces were modeled and compared with findings reported for the interaction of CO2 on the anatase TiO2 (101) surface.12 The brookite TiO2 surfaces were simulated using both clusters and periodic slab models. In the cluster model, a Ti7O28H28 cluster that was cleaved from the brookite TiO2 (210) surface was adopted to represent the brookite (210) surface. Six configurations were developed that varied according to the location of the CO2 molecule over the brookite (210) surface. Stoichiometric TiO2 in its ground state is a singlet; therefore, the cluster was cleaved to model a zero-charge singlet cluster in its ground state. The dangling bonds on the oxygen atoms were terminated with hydrogen atoms with an initial bond length of 0.96 Å. The negatively charged brookite (210) surface was developed by introducing an extra electron into the system, thereby modeling the 19756

dx.doi.org/10.1021/jp302342t | J. Phys. Chem. C 2012, 116, 19755−19764

The Journal of Physical Chemistry C

Article

Figure 1. Optimized geometry of the neutral Ti7O28H28 cluster representing the brookite (210) surface; 1A represents the front side of the cluster, and 1B shows the back side of the cluster. Symbols for the individual atoms are Ti in green, O in red, and H in white.

Figure 2. Optimized geometry configurations of CO2 adsorbed onto the neutral Ti7O28H28 cluster. Colors represent atoms accordingly: Ti in green, O in red, H in white, and C in gray. Distances and angles are in Å and degrees, respectively.

Nicolet 6700 Fourier transform infrared (FTIR) spectrometer (Thermo Electron) equipped with a liquid-nitrogen-cooled MCT detector, employing a resolution of 4 cm−1 and 32 scans. The FTIR spectrometer was equipped with a Harrick Scientific diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) system that employed a high temperature reaction chamber (HVC-DRP). The brookite samples were prepared by using an aqueous solution of titanium bis(ammonium lactate) dihydroxide (50%) and urea (7 M) in a Teflon-lined autoclave through a hydrothermal process at 160 °C for 24 h, followed by washing,

electrons were treated as valence electrons. The negatively charged brookite (210) surface, representing a photoexcited electron, was modeled by introducing an extra electron into the system with an additional neutralizing positive background charge. Atomic charges were analyzed using the Bader scheme.41



EXPERIMENTAL METHODS CO2 (99.999%, Praxair) was exposed to brookite and oxygendeficient brookite surfaces in order to explore the formation of CO2−. Infrared transmission spectra were recorded using a 19757

dx.doi.org/10.1021/jp302342t | J. Phys. Chem. C 2012, 116, 19755−19764

The Journal of Physical Chemistry C

Article

Figure 3. Optimized geometry configurations of CO2 adsorbed onto the negatively charged Ti7O28H28 cluster. Colors represent atoms accordingly: Ti in green, O in red, H in white, and C in gray. Distances and angles are in Å and degrees, respectively.

drying, and final calcination for 3 h at 400 °C. The crystal structure of the brookite was confirmed using X-ray diffraction (XRD). Oxygen-deficient brookite was created in situ inside the DRIFTS high temperature reaction chamber by purging helium (99.999%, Praxair) through the brookite sample. The methods used to create oxygen-deficient surfaces are similar to those used previously.42−45 A temperature program consisting of a ramp of 5 °C/min up to a temperature of 350 °C with a 1 h hold time at 350 °C was employed for the treatment process. Background experiments showed that the temperatures that were employed were sufficiently low enough to avoid the conversion of brookite to other titania phases.

configurations exist, and are labeled 2A−2E. For better viewing, two images of configuration 2E are provided and are labeled 2E(a) and 2E(b). Figure 3 displays the various configurations of CO2 adsorbed to the negatively charged brookite (210) surface. A total of five distinct configurations were found; they are labeled 3A−3E. Two images of configuration 3E are provided and are labeled 3E(a) and 3E(b). Additional work was undertaken to determine whether a brookite surface that contained oxygen vacancies would more favorably adsorb CO2 and facilitate charge transfer to the CO2 molecule. The bridging oxygen atom of the first 5-coordinatedTi atom was removed from the cluster system. Figure 4 displays the optimized configurations of CO2 adsorbed to the oxygendeficient brookite surface cluster. The ground states of configurations 4A, 4B, and 4D are singlet states, and the ground state of 4C configuration is a triplet state. Table 1 reports the calculated adsorption energies (in eV) of CO2 bound to the neutral and negatively charged brookite surfaces as well as the neutral and negatively charged anatase surfaces for both the cluster and periodic slab systems. Adsorption energies of CO2 adsorbed onto the oxygendeficient brookite and anatase surfaces are also provided in Table 1 for the cluster system.



RESULTS AND DISCUSSION Figure 1 displays the front (Figure 1A) and back side (Figure 1B) of the optimized neutral Ti7O28H28 brookite cluster where atoms 1 and 2 are the 5-coordinated-Ti reaction centers. Ti atom 1 is connected to a 6-coordinated-Ti atom by a 3coordinated-O atom, whereas Ti atom 2 is not connected to a 6-coordinated-Ti atom via an oxygen atom. Instead, the 6coordinated-Ti atom dangles to the side. CO2 Adsorbed to the Brookite (210) Surface. Figure 2 displays the various configurations of CO2 interacting with the neutral brookite (210) surface. A total of five distinct 19758

dx.doi.org/10.1021/jp302342t | J. Phys. Chem. C 2012, 116, 19755−19764

The Journal of Physical Chemistry C

Article

Figure 4. Optimized geometry configurations of CO2 adsorbed on the cluster system of the brookite (210) surface with an oxygen vacancy. Colors represent atoms accordingly: Ti in green, O in red, H in white, and C in gray. Distances and angles are in Å and degrees, respectively.

Table 1. Calculated Adsorption Energies (in eV) of CO2 on the Neutral, Negatively Charged, and Oxygen-Deficient Brookite (210) Surface and the Neutral, Negatively Charged, and Oxygen-Deficient Anatase (101) Surface Using Cluster and Periodic Slab Systems phase neutral (configuration 2)

negatively charged (configuration 3)

model

cluster

periodic

cluster

periodic

cluster

cluster

periodic

cluster

periodic

cluster

A B C D E F

−0.44 −0.50 −0.30 −0.72 0.33

−0.16

−0.36 −0.41 −0.51 −0.90 0.08

−0.07

−1.53 −0.92 −0.51 −1.59

−0.34

−0.20

−0.35

−0.07

−0.41

−0.14

−0.35

−0.23

−1.09 −0.36 0.08 −0.97

0.22

0.06

0.21 0.33

−0.03 0.72

state

a

anatase (101)a

brookite (210)

−0.09 0.06

−0.18

oxygen-deficient (configuration 4)

0.02

neutral

negatively charged

oxygen-deficient

Adsorption energies (in eV) of CO2 adsorbed onto neutral, negatively charged, and oxygen-deficient anatase (101) surface reported by He et al.12

As can be seen from Figures 2 and 3, CO2 interacts with both the neutral and negatively charged brookite (210) surface in similar binding configurations. The binding site for configurations A and C is located on 5-coordinated-Ti atom 1. In configurations B and D, the binding site is located on the second 5-coordinated-Ti atom with the dangling 6-coordinatedTi atom. Configuration E for both the neutral and negatively charged brookite surface has significant differences. In configuration 2E, both oxygen atoms of the CO2 bind to the 5-coordinated-Ti atoms to form a polydentate carbonate. On the negatively charged brookite surface, only one oxygen atom binds to a 5-coordinated-Ti atom. The binding site occurs on the 5-coordinated-Ti atom connected to the 6-coordinated-Ti atom. On the neutral brookite cluster, in configurations 2B and 2D, CO2 binds more closely to the surface and experiences a stronger adsorption (Table 1) than configurations 2A and 2C. It is worth noting that, on the brookite (210) surface, although 2B is lower in energy (−0.50 eV) than configuration 2A (−0.44 eV), the energetic difference (−0.06 eV) is within the errors of

the calculation, thus making the two configurations energetically competitive. Of the five configurations, configuration 2D has the lowest adsorption energy (−0.72 eV), making it the most likely binding formation to occur on the neutral brookite (210) surface. Configuration 2E, which forms a polydentate carbonate, has the highest adsorption energy (0.33 eV) and is therefore the least likely to exist. On the negatively charged brookite cluster, configurations 3C and 3D are the more energetically favorable binding forms with adsorption energies of −0.51 and −0.90 eV, respectively. Although configuration 3B binds closer to the negatively charged brookite surface and experiences a lower adsorption energy than configuration 3A, both configurations remain energetically competitive. Configuration 3E is the most energetically unfavorable with an adsorption energy of 0.08 eV. In regard to both the neutral and negatively charged cluster system, the negatively charged brookite surface appears to be more reactive. Configurations 3C, 3D, and 3E have significantly lower adsorption energies than their neutral counterparts with 19759

dx.doi.org/10.1021/jp302342t | J. Phys. Chem. C 2012, 116, 19755−19764

The Journal of Physical Chemistry C

Article

Figure 5. Optimized geometries of CO2 and CO2− using the B3LYP/6-31+G(2df, p) level. Colors represent atoms accordingly: O in red and C in gray. Distances and angles are in Å and degrees, respectively.

0.5−1.7° and 0.01−0.10 Å, respectively. Two additional configurations exist on the surface of the brookite (210) cluster, configurations 2B and 2D. On the negatively charged brookite surface, configurations 2A and 2C are structurally similar to the binding configurations calculated by He et al.12 for the negatively charged anatase (101) surface. Three new configurations exist on the brookite (210) surface, configurations 3B, 3D, and 3E. As was described previously, only one oxygen atom from the CO2 molecule of configuration 3E binds to the negatively charged brookite surface. On the negatively charged anatase surface, both oxygen atoms from the CO2 molecule bind to the two 5-coordinatedTi atom sites, forming a polydentate carbonate. No configuration similar to F exists for the brookite surface. Configuration F on the negative anatase surface had a CO2 as a bridging bidentate with the carbon atom pointing upward and the oxygen atoms bound to the two 5-coordinated-Ti atoms.12 In Table 1, the adsorption energies for the binding configurations of both neutral and negatively charged brookite (210) surfaces and the anatase (101) surfaces are presented and labeled A−E. As suggested by previous DFT calculations,21 the reactivity of the brookite (210) surface does affect the adsorptive ability of molecules and, in some cases, creates stronger binding sites. In the case of the neutral brookite and anatase surface cluster systems, configuration A has a significantly lower adsorption energy for CO2 on the brookite (210) surface as compared to the anatase (101) surface. Additionally, the brookite surface allows for the formation of two additional CO2 binding configurations (B and D) that have much lower adsorption energies than any of the configurations formed on the anatase surface. This implies that the neutral brookite (210) surface allows for stronger CO2 binding sites than the neutral anatase (101) surface. On the negatively charged brookite and anatase surfaces, configuration A is structurally and energetically similar. Although configuration C for both surfaces is structurally similar, that of brookite has a lower adsorption energy and is therefore more favorable. Configurations B and D are two additional configurations formed on the negatively charged brookite surface that exhibit much lower adsorption energies than configurations on the negatively charged anatase surface. Configurations A and B for both brookite and anatase and configuration C for anatase are energetically competitive. Configuration D is the most favorable of the structures with an adsorption energy of −0.90 eV, while configurations E and F are the most unstable with configuration F nonexistent on the brookite surface. All configurations for the brookite periodic system are energetically competitive with their respective counterparts in the anatase periodic system.

the exception of 2B, which has an adsorption energy similar to configuration 3C. In the periodic slab models, three distinct configurations were used for both the neutral and negatively charged CO2brookite (210) surface (see the Supporting Information) corresponding to configurations A, C, and E of the cluster system. The adsorption energies were less favorable than that of the cluster system (see Table 1). The adsorption energies of configurations A, C, and E are −0.16, −0.09, and 0.06 eV, respectively, for the neutral periodic system and −0.07, −0.18, and 0.02 eV, respectively, for the negatively charged periodic system. In the neutral periodic system, configurations A and C are energetically similar. This is not the case for the negatively charged system in which the formation of configuration A is less favorable than that of configuration C. As was seen in the cluster system, configuration E for both neutral and negatively charged periodic slab system was the most unstable configuration, having a positive adsorption energy. Carbon dioxide more readily adsorbs to the brookite surface with oxygen vacancies. As can be seen from Table 1, configurations 4A, 4B, 4C, and 4D have more favorable CO2 adsorption energies than both the neutral and negatively charged brookite (210) surface. Configurations 4A and 4D have comparable adsorption energies, which are significantly lower than 4B and much more favorable than the most stable configuration on both the neutral and negatively charged brookite surfaces. Configuration 4B is energetically equivalent to the most stable configuration (3D) on the negatively charged brookite surface and more favorable than the most stable configuration (2D) on the neutral brookite surface. Although configuration 4C is the least energetically favorable of the four configurations, its adsorption energy is still comparable to configurations 2A, 2B, and 3C. Additional calculations that were performed on all cluster systems using the PBE functional with the same basis sets as the original B3LYP calculations (6-31+G(2df,p) and LanL2DZ) and the B3LYP functional with the 6-31+G(p) and LanL2DZ basis sets yielded energetically similar trends as the original calculations using the B3LYP functional and the larger basis set. These data all lead to the conclusion that oxygen-deficient brookite surfaces are more favorable for CO2 adsorption over unmodified brookite surfaces. BSSE corrected calculations also resulted in the same conclusion. Results for these calculations are available in the Supporting Information. Comparison of CO2 Adsorption to the Brookite (210) and the Anatase (101) Surfaces. Configurations 2A, 2C, and 2E are structurally similar to the binding configurations calculated by He et al.12 for the neutral anatase (101) surface. The bond angles and bond lengths of the binding configurations on the brookite (210) surface vary from those of the binding configurations on the anatase (101) surface by 19760

dx.doi.org/10.1021/jp302342t | J. Phys. Chem. C 2012, 116, 19755−19764

The Journal of Physical Chemistry C

Article

Table 2. Charge and Spin Distributions of CO2 on the Neutral, Negatively Charged, and Oxygen-Deficient Brookite (210) and Anatase (101) TiO2 Surfaces Using Cluster and Periodic Slab Systems anatase (101)a

brookite (210) neutral

negatively charged

charge

charge

spin

oxygendeficient

neutral

charge

charge

oxygendeficient

negatively charged charge

cluster

periodic

cluster

periodic

cluster

periodic

cluster

cluster

periodic

cluster

periodic

cluster

A B C D E F

0.17 0.18 −0.16 −0.16 −0.03

−0.01

0.14 0.14 −0.23 −0.30 −0.11

0.00

0.00 0.01 0.00 −0.01 0.01

0.00

−0.50 −0.77 −0.41 −0.41

0.10

0.00

0.09

−0.03

0.03

0.00

−0.23

−0.04

−0.31

−0.11

0.01

0.00

−0.63 −0.91 −0.50 −0.57

−0.18

−0.11

−0.16 −0.45

−0.18 −0.65

0.00 0.87

0.00 0.81

a

0.13

0.07 0.16

0.00 0.00

cluster

charge

model

0.04

periodic

spin

Charge and spin distribution of CO2 adsorbed onto neutral, negatively charged, and oxygen-deficient anatase (101) surface reported by He et al.12

Figure 6. FTIR spectra of (a) CO2 exposed to oxygen-deficient brookite, B(Vo), and perfect brookite, B (b) CO2 exposed to helium pretreated brookite (i.e., oxygen-deficient brookite) for various times, and (c) CO2 exposed to helium pretreated rutile for various times.

Most notable is the significantly lower adsorption energy of the 4C configuration, which is 0.59 eV lower than the corresponding anatase configuration. Charge, Spin Density, and Vibrational Frequency Analyses of CO2 Adsorbed on the Brookite (210) Surface. Several reported studies have demonstrated the ability of the anatase TiO2 to reduce CO2 to fuels.2−12 The formation of the CO2− on the surface of TiO2 is considered the first step in the reduction of CO2.19 The reduction of CO2 to this anionic state is accompanied by a signature bending from its linear form (Figure 5a) to an O−C−O angle of ∼138°. This

In the oxygen-deficient brookite system, configurations 4A, 4B, 4C, and 4D have geometries similar to those on the oxygendeficient anatase (101) surface with distances and bond angles differing by 0.01−0.16 Å and 0.1−1.9°, respectively. The adsorption energies are significantly more favorable on the brookite surface than on the anatase surface. The respective adsorption energies of 4A, 4B, 4C, and 4D are −1.53, −0.92, −0.51, and −1.59 eV, whereas the corresponding adsorption energies on the anatase surface are −1.09, −0.36, 0.08, and −0.97 eV.12 This suggests that CO2 is more likely to experience stronger interactions with the oxygen-deficient brookite (210) surface than with the oxygen-deficient anatase (101) surface. 19761

dx.doi.org/10.1021/jp302342t | J. Phys. Chem. C 2012, 116, 19755−19764

The Journal of Physical Chemistry C

Article

Table 3. Calculated Vibrational Frequencies (in cm−1) of Bending (ν2), Symmetric Stretching (ν1), and Asymmetric Stretching (ν3) Modes of CO2 on the Neutral, Negatively Charged, and Oxygen-Deficient Brookite (210) Surfacesa brookite (210) surface neutral

negatively charged

oxygen-deficient

mode

A

B

C

D

E

ν2 ν1 ν3 ν2 ν1 ν3 ν2 ν1b ν3b

623, 639 1332 2347 638 1328 2337 738 891 1758

649 1326 2347 644 1327 2345 761, 742 866 1248

772 987 1764 780 1004 1729 716, 725 1275 1659

779 985 1770 788 1005 1715 744, 761 1142 1702

778 1253 1710 776 1223 1738

Vibrational frequencies were scaled by a factor of 0.9652.35 bVibrational modes ν1 and ν3 assigned to configurations A and B are for two C−O stretching modes other than symmetric and asymmetric stretching modes. a

deformation is shown in Figure 5b. Adsorbed CO2− also has two signature bands at 1640 and 1219 cm−1.20 He et al.12 demonstrated the ability of the anatase (101) surface to reduce CO2 to its anionic form. Although highly unstable, as is shown by the high adsorption energies for configuration F and only stabilized on the negatively charged anatase (101) surface, charge analysis demonstrated a net charge transfer to the CO2 and most notably to the carbon (Table 2). The calculated vibrational frequency values (1638 and 1265 cm−1)12 corresponded well with the band values for the CO2−. However, our calculations show that no such configuration exists on the perfect brookite surface. Charge, spin density, and vibrational frequency calculations on the CO2−brookite configurations confirm no CO2− formation on the perfect brookite (210) surface. In Table 2, the charge and spin densities of CO2 of the different configurations on the neutral and negatively charged brookite and anatase surfaces are reported for both cluster and periodic slab systems. Table 2 also provides the CO2 charge values for the oxygen-deficient brookite and anatase cluster surfaces. The charge values are the sum of the NBO atomic charges and the Bader charges on CO2 for the cluster system and the periodic system, respectively. In regard to the neutral and negatively charged brookite surface cluster system, configurations 2C−2E (neutral) and 3C−3E (negatively charged) have a negative charge on the CO2 molecule. Charge transfer to the CO2 molecule in the periodic slab model is negligible in both the neutral and negatively charged system. The spin density was calculated to determine if an unpaired electron exists on the CO2 molecule. From Table 2, it can be seen that the spin density located on the CO2 is minute for all configurations in both the cluster and periodic slab systems. In other words, the unpaired electron on the surface is not transferred to the CO2 molecule. Figure S3 in the Supporting Information depicts the spin densities of the various binding configurations for the negatively charged brookite surface cluster. On all configurations, it can be seen that charge transfer occurs in the direction of the second 5coordinated-Ti atom. The spin density suggests that the unpaired electron is located primarily on the second 5coordinated-Ti atom and not on the CO2 molecule. The charge on the CO2 molecule on oxygen-deficient brookite is larger than on either neutral or negatively charged brookite surfaces, confirming that oxygen-deficient brookite facilitates charge transfer to the CO2 molecule. On the oxygendeficient anatase surface, the charge transfer is only slightly

larger than that of the brookite surface. Similar to the anatase cluster system, the 4C configuration is in a triplet state with a spin located on the CO2 molecule, most notably on the carbon atom and a spin on a 5-coordinated-Ti atom (see Figure S4, Supporting Information). This confirms that oxygen vacancies aid in the formation of CO2− with a greater likelihood of its formation on the brookite surface than on the anatase surface. The vibrational frequencies were examined experimentally and computationally for each configuration to help identify the adsorbed CO2 species, including the CO2−. Figure 6a shows two DRIFT spectra of CO2 exposed to perfect brookite (B) and to brookite with oxygen vacancies (B(Vo)) in the absorption region 1000−2000 cm−1, while Table 3 presents the computationally calculated vibrational frequencies in cm−1. Figure 6b and c shows IR spectra for pretreated brookite and rutile samples in the absorption region 1000−2000 cm−1 at different CO2 exposure times. Additional background IR spectra are included in the Supporting Information document to show that the surfaces were clean, with only adsorbed H2O appearing on perfect brookite and OH groups on oxygendeficient brookite. The spectra for brookite (B) and oxygen-deficient brookite B(Vo) in Figure 6a contain peaks associated with carbonate species formation that are in good agreement with literature data.46,47 Bands are assigned to bidentate carbonates (1059, 1586 cm−1 for the perfect brookite spectrum but are not detectable in the oxygen-deficient brookite, B(Vo), spectrum in Figure 6a), bicarbonates (1222, 1416 cm−1 for the brookite spectrum and 1222, 1421 cm−1 for the B(Vo) spectrum), and carboxylates (1248, 1672 cm−1 for the B(Vo) spectrum in Figure 6a). The 1248 and 1672 cm−1 bands from the B(Vo) spectrum have been assigned to bent CO2−, and are in agreement with reported CO2− band values.20,46,47 While the perfect brookite spectrum in Figure 6a has a peak at 1673 cm−1, the lack of the second defining peak near 1248 cm−1 suggests the absence of bent CO2−. The weak 1673 cm−1 feature is similar to reported observations by Yang et al.46 on Cu(I)/ TiO2. It is unclear whether the feature at 1673 cm−1 belongs to a bicarbonate or carboxylate species. Additional IR spectra for pretreated brookite and rutile samples with increased CO2 exposure time are included in Figure 6b and c. The “0 min” curve represents the IR spectrum before any intentional CO2 exposure. No IR features were observed on either brookite or rutile at “0 min”. After CO2 exposure, the IR features appeared, and the intensities of the adsorption bands increased with exposure time. The CO2− 19762

dx.doi.org/10.1021/jp302342t | J. Phys. Chem. C 2012, 116, 19755−19764

The Journal of Physical Chemistry C

Article

band at 1248 cm−1 was observed on pretreated brookite, and its intensity increased with exposure time. By contrast, the CO2− band did not appear on the pretreated rutile sample (Figure 6c). These experimental data serve to confirm the selective formation of CO2− on a brookite surface with oxygen vacancies. The extent of CO2− formation on perfect versus oxygendeficient brookite was quantified by normalizing the peak areas for CO2− at 1248 and 1672 cm−1, respectively, by that of the CO32− peak in the 1530−1650 cm−1 range (i.e., ratio 1 = peak area of 1248 cm−1/peak area of 1530−1650 cm−1 feature and ratio 2 = peak area of 1672 cm−1/peak area of 1530−1650 cm−1 feature). The peak for the CO32− species did not change appreciably in comparing the perfect and oxygen-deficient brookite samples. Three separate samples were used to compare the areas. The data indicate that the average ratio of CO2− peaks to CO32− peaks for brookite were 0.71 (ratio 1) and 9.44 (ratio 2) for perfect brookite as compared to the average values of 4.24 (ratio 1) and 16.64 (ratio 2) for oxygendeficient brookite. These data indicate that more CO2− is formed on the oxygen-deficient brookite surface. The detailed data are found in the Supporting Information. We first examined the neutral and negatively charged systems. For configurations A and B, the calculated vibrational frequencies (Table 3) correspond well with the reported band values of adsorbed linear CO2.48,49 The values are also in good agreement with the computational findings of adsorbed linear CO2 reported by He et al.12 Configurations C−E have vibrational frequencies that are most consistent with side-on bonded CO2 on cationic sites.39 Rasko et al.20 identified two signature absorption bands (1640 and 1219 cm−1) attributed to bent CO2−. From Table 3, no configurations have vibrational frequencies that relate to both of the reported signature bands for CO2−. The calculated vibrational frequencies are different from the IR vibrational frequency data seen in Figure 6. This could be due to factors such as the brookite cluster size and surface. However, symmetric stretching modes of configurations C and D for both the neutral and negatively charged cluster are within 75 cm−1 of the 1059 cm−1 peak assigned to bidentate carbonate. On the negatively charged brookite surface, configuration E has a symmetric mode that closely reflects that of the bicarbonate band 1222 cm−1. Although configuration E for the neutral surface has a symmetric mode similar to the CO2− (1248 cm−1), the negligible charge on the CO2 molecule indicates that the CO2− is not formed, and thereby implies that this band is most likely that of a bicarbonate species. Configurations C, D, and E for the neutral and negatively charge cluster system have a net charge transfer (Table 2). However, charge analysis suggests that the net charge transfer to the carbon atom of the CO2 molecule for any of the configurations is negligible. Although configuration E for the neutral and negatively charged surfaces both exhibit the characteristic bending that is accompanied with the formation of the CO2−, the CO2 molecule lacks notable charge and spin as well as vibrational frequency values (Table 3) associated with the signature bands of CO2− as reported by Rasko et al.20 Additionally, the frequencies do not compare well with the values reported in this experimental study or in previous ones.46,47 Configurations C and D also lack vibrational frequency values characteristic of the CO2−. As suggested by Indrakanti et al.,50 configuration E can be thought of as an exception in that CO2 experiences charge transfer, although

miniscule, and undergoes bending but the carbon atom is not reduced. The lack of the characteristic bands and insignificant charge transfer to the CO2 molecule for any of the configurations on the neutral and negatively charged brookite (210) surface suggest that there was no CO2− formation. Table 3 also presents the vibrational frequency data for CO2 adsorbed onto the brookite (210) surface with oxygen vacancies. Shifts in vibrational frequencies were expected when considering the geometry and bonding of the structures. Noticeable shifts occur primarily for the ν1 vibrational modes. Configuration 4C is the most interesting structure of the four because of its resemblance to the CO2−. Additionally, the calculated vibrational frequency bands (1275 and 1659 cm−1) are similar to the experimental bands found on oxygen-deficient brookite from this study (Figure 6a, brookite(Vo)) and the other reported bands for CO2−.20,46,47 The O−C−O angle (131.3°) is not the characteristic ∼138° associated with the CO2−, but it is close. The increased charge transfer, similar vibrational frequency bands, and O−C−O angle near 138° suggest that configuration 4C may be a precursor to the CO2− formation. It is suggested that the CO2 molecule becomes easier to reduce as the O−C−O bond angle decreases from 180°.50 With increased charge transfer to the CO2 molecule and the decreased O−C−O angle, it is likely that configurations 4A, 4B, 4C, and 4D are precursors to CO2 reduction. The more favorable adsorption energies and increased charge transfer suggest that oxygen vacancies on the brookite (210) surface favor bent CO2 configurations and aid in the stabilization of negatively charged CO2−.



CONCLUSION Adsorption energies and charge analysis calculations were performed on the interactions of CO2 with perfect and oxygendeficient clusters and periodic slab systems of brookite TiO2. The intention of this research was to demonstrate that the brookite (210) surface had a similar or greater ability for charge transfer to the CO2 molecule, thus facilitating the first step in CO2 reduction. As demonstrated by previous DFT calculations with other molecules,21 the brookite (210) surface exhibited stronger adsorption of CO2 in certain configurations. However, unlike on the negatively charged anatase (101) surface, the CO2 anion did not form on the negatively charged brookite (210) surface. Although some of the CO2−brookite (210) configurations had more favorable adsorption energetics as compared to the CO2 interactions with the anatase (101) surface, the computational results suggest that an unmodified brookite (210) surface by itself does not result in charge transfer to the CO2. However, the presence of oxygen vacancies within the brookite (210) surface enhanced the interaction of CO2 with the surface and facilitated charge transfer to the CO2 molecule.



ASSOCIATED CONTENT

S Supporting Information *

Geometries of CO2 adsorbed to the neutral brookite (210) surface for the periodic slab system; geometries of CO2 adsorbed to the negatively charged brookite (210) surface for the periodic slab system; spin densities of the CO2 adsorbed to the negatively charged brookite (210) surface for the cluster system; spin density of configuration 4C; IR spectra of perfect brookite and oxygen-deficient brookite before CO2 exposure; additional computational calculations: BSSE corrections, PBE 19763

dx.doi.org/10.1021/jp302342t | J. Phys. Chem. C 2012, 116, 19755−19764

The Journal of Physical Chemistry C

Article

(23) Gong, X. Q.; Selloni, A. Phys. Rev. B 2007, 76, 235307; DOI: 10.1103/PhysRevB.76.235307. (24) Deskins, N. A.; Rousseau, R.; Dupuis, M. J. Phys. Chem. C 2010, 114, 5891−5897. (25) Becke, A. D. J. Chem. Phys. 1993, 98, 5648−5652. (26) Lee, C. T.; Yang, W. T.; Parr, R. G. Phys. Rev. B 1988, 37, 785− 789. (27) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Montgomery, J. A., Jr.; Vreven, T.; Kudin, K. N.; Burant, J. C.; et al. Gaussian 03; Gaussian, Inc.: Wallingford, CT, 2004. (28) Ditchfie., R; Hehre, W. J.; Pople, J. A. J. Chem. Phys. 1971, 54, 724−728. (29) Hehre, W. J.; Ditchfie., R; Pople, J. A. J. Chem. Phys. 1972, 56, 2257−2261. (30) Clark, T.; Chandrasekhar, J.; Spitznagel, G. W.; Schleyer, P. V. J. Comput. Chem. 1983, 4, 294−301. (31) Frisch, M. J.; Pople, J. A.; Binkley, J. S. J. Chem. Phys. 1984, 80, 3265−3269. (32) Hay, P. J.; Wadt, W. R. J. Chem. Phys. 1985, 82, 270−283. (33) Wadt, W. R.; Hay, P. J. J. Chem. Phys. 1985, 82, 284−298. (34) Hay, P. J.; Wadt, W. R. J. Chem. Phys. 1985, 82, 299−310. (35) Merrick, J. P.; Moran, D.; Radom, L. J. Phys. Chem. A 2007, 111, 11683−11700. (36) Meagher, E. P.; Lager, G. A. Can. Mineral. 1979, 17, 77−85. (37) Kresse, G.; Furthmuller, J. Phys. Rev. B 1996, 54, 11169−11186. (38) Kresse, G.; Furthmuller, J. Comput. Mater. Sci. 1996, 6, 15−50. (39) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865−3868. (40) Kresse, G.; Joubert, D. Phys. Rev. B 1999, 59, 1758−1775. (41) Sanville, E.; Kenny, S. D.; Smith, R.; Henkelman, G. J. Comput. Chem. 2007, 28, 899−908. (42) Salari, M.; Konstantinov, K.; Liu, H. K. J. Mater. Chem. 2011, 21, 5128−5133. (43) Macak, J. M.; Zollfrank, C.; Rodriguez, B. J.; Tsuchiya, H.; Alexe, M.; Greil, P.; Schmuki, P. Adv. Mater. 2009, 21, 3121−3125. (44) Macak, J. M.; Gong, B. G.; Hueppe, M.; Schmuki, P. Adv. Mater. 2007, 19, 3027−3031. (45) Liu, L. J.; Zhao, C. Y.; Li, Y. J. Phys. Chem. C 2012, 116, 7904− 7912. (46) Yang, C. H.; Yu, Y. H.; van der Linden, B.; Wu, J. C. S.; Mul, G. J. Am. Chem. Soc. 2010, 132, 8398−8406. (47) Su, W. G.; Zhang, J.; Feng, Z. C.; Chen, T.; Ying, P. L.; Li, C. J. Phys. Chem. C 2008, 112, 7710−7716. (48) Henderson, M. A. Surf. Sci. 1998, 400, 203−219. (49) Ramis, G.; Busca, G.; Lorenzelli, V. Mater. Chem. Phys. 1991, 29, 425−435. (50) Indrakanti, V. P.; Kubicki, J. D.; Schobert, H. H. Energy Fuels 2008, 22, 2611−2618.

functional, and smaller basis set. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS



REFERENCES

This research was supported by the Chemical, Bioengineering, Environmental and Transport Systems (CBET) division of the National Science Foundation (NSF), grants CBET-1067340 and CBET-1067233 and by the Western Alliance to Expand Student Opportunities Bridge to the Doctorate (WAESOBD), Louis Stokes Alliance for Minority Participation Bridge to the Doctorate (LSAMPBD) NSF grant HRD-1025879. The authors also acknowledge the Arizona State University Advanced Computing Center for providing computational resources on the Saguaro cluster.

(1) White, C. M.; Strazisar, B. R.; Granite, E. J.; Hoffman, J. S.; Pennline, H. W. J. Air Waste Manage. Assoc. 2003, 53, 645−715. (2) Linsebigler, A. L.; Lu, G. Q.; Yates, J. T. Chem. Rev. 1995, 95, 735−758. (3) Anpo, M.; Yamashita, H.; Ichihashi, Y.; Ehara, S. J. Electroanal. Chem. 1995, 396, 21−26. (4) Tseng, I. H.; Wu, J. C. S. Catal. Today 2004, 97, 113−119. (5) Kitano, M.; Matsuoka, M.; Ueshima, M.; Anpo, M. Appl. Catal., A 2007, 325, 1−14. (6) Koci, K.; Obalova, L.; Lacny, Z. Chem. Pap. 2008, 62, 1−9. (7) Varghese, O. K.; Paulose, M.; LaTempa, T. J.; Grimes, C. A. Nano Lett. 2009, 9, 731−737. Correction to original article published in: Varghese, O. K.; Paulose, M.; LaTempa, T. J.; Grimes, C. A. Nano Lett. 2010, 10, 750. (8) Yang, H. C.; Lin, H. Y.; Chien, Y. S.; Wu, J. C. S.; Wu, H. H. Catal. Lett. 2009, 131, 381−387. (9) Indrakanti, V. P.; Kubicki, J. D.; Schobert, H. H. Energy Environ. Sci. 2009, 2, 745−758. (10) Li, Y.; Wang, W. N.; Zhan, Z. L.; Woo, M. H.; Wu, C. Y.; Biswas, P. Appl. Catal., B 2010, 100, 386−392. (11) Koci, K.; Mateju, K.; Obalova, L.; Krejcikova, S.; Lacny, Z.; Placha, D.; Capek, L.; Hospodkova, A.; Solcova, O. Appl. Catal., B 2010, 96, 239−244. (12) He, H. Y.; Zapol, P.; Curtiss, L. A. J. Phys. Chem. C 2010, 114, 21474−21481. (13) Hurum, D. C.; Agrios, A. G.; Gray, K. A.; Rajh, T.; Thurnauer, M. C. J. Phys. Chem. B 2003, 107, 4545−4549. (14) Addamo, M.; Bellardita, M.; Di Paola, A.; Palmisano, L. Chem. Commun. 2006, 47, 4943−4945. (15) Dambournet, D.; Belharouak, I.; Amine, K. Chem. Mater. 2010, 22, 1173−1179. (16) Park, J. Y.; Lee, C.; Jung, K. W.; Jung, D. Bull. Korean Chem. Soc. 2009, 30, 402−404. (17) Kandiel, T. A.; Feldhoff, A.; Robben, L.; Dillert, R.; Bahnemann, D. W. Chem. Mater. 2010, 22, 2050−2060. (18) Xie, J. M.; Lu, X. M.; Liu, J.; Shu, H. M. Pure Appl. Chem. 2009, 81, 2407−2415. (19) Centi, G.; Perathoner, S. Catal. Today 2009, 148, 191−205. (20) Rasko, J.; Solymosi, F. J. Phys. Chem. 1994, 98, 7147−7152. (21) Li, W. K.; Gong, X. Q.; Lu, G.; Selloni, A. J. Phys. Chem. C 2008, 112, 6594−6596. (22) Pan, H.; Gu, B. H.; Zhang, Z. Y. J. Chem. Theory Comput. 2009, 5, 3074−3078. 19764

dx.doi.org/10.1021/jp302342t | J. Phys. Chem. C 2012, 116, 19755−19764