Article pubs.acs.org/JPCC
Presence of Gap States at Cu/TiO2 Anatase Surfaces: Consequences for the Photocatalytic Activity Nicola Seriani,*,† Carlos Pinilla,‡,§ and Yanier Crespo†,∥ †
The Abdus Salam International Centre for Theoretical Physics, Strada Costiera 11, 34151 Trieste, Italy University College London, London WC1E 6BT, England § Departamento de Fisica, Universidad del Norte, km 5 Via Puerto Colombia, Barranquilla, Colombia ∥ International Institute of Physics, Av. Odilon Gomes de Lima, 1722, Capim Macio, CEP 59078-400, Natal RN Brazil ‡
S Supporting Information *
ABSTRACT: Copper-modified titania is a system of interest for its potential for photocatalytic applications in the production of solar fuels. Still, the role of copper in the process is unclear. In this work, small copper clusters on the (101) and (100) surfaces of anatase have been investigated by first-principles simulations based on density functional theory, to shed light on their atomic and electronic structure, and to understand their effect on the photocatalytic process. The main effects of copper on the electronic structure are to provide states above the edge of the valence band of titania and to lead to the formation of midgap states. There are two types of midgap states, respectively, associated with direct Cu−Ti bonds and to Ti3+ polarons. The latter are the result of charge donation from copper and lie in the vicinity of the surface. Moreover, the copper tetramer (Cu4) displays empty states at the bottom of the conduction band that play a key role in accommodating excess electrons. We discuss how these features should enhance the photoresponse of TiO2, contribute to increase the lifetime of the photogenerated electron−hole pairs and contribute to increase the activity of this material for CO2 reduction, a key step in the photoproduction of hydrocarbons.
■
decrease in the band gap of the material,21,22 but no equivalent information is available for supported copper nanoparticles. On the other side, the electrocatalytic conversion of CO2 to hydrocarbons seems to be dependent on the particle shape and surface morphology of copper catalysts.23 Such a dependence of catalytic activities on particle shape is typically linked to the surface chemistry of the catalyst. The oxidation state of copper plays a role in the activity of the catalyst as well, with nanocrystalline metallic copper being particularly active for the electroreduction of CO.24 Also, Liu et al. investigated the effect of different treatments during catalyst preparation for a Cu/ TiO2 photocatalyst for CO2 reduction, and found that a reducing treatment gives the best activity.25 Moreover, a high dispersion of the copper catalyst is important to achieve high conversion rates.26 Regarding the mechanistic aspects of surface chemistry of the Cu/TiO2 system, is is known that CO2 dissociates spontaneously on defective Cu(I)/TiO2 in the dark.27 It was proposed that electrons, trapped at Ti3+−VO sites, can migrate to CO2 resulting in a dissociative electron attachment and in CO adsorbed to Cu+.27 This hypothesis underlines the importance of the reduced state of the catalyst and relies on the presence of trapped electrons in the vicinity of
INTRODUCTION Titania is the material of choice in photocatalytic applications and has been the subject of intense scrutiny to understand the mechanisms of the photocatalytic processes.1−13 Notwithstanding its popularity, it still suffers from major drawbacks that hinder its widespread use in commercial applications. These disadvantages comprise a relatively large band gap, ∼3 eV, that prevents photoabsorption in the visible range, and a low overall efficiency of the photocatalytic process. For these reasons, many attempts to enhance its properties through proper catalyst modifications have been undertaken. Among these, the addition of copper to titania has recently received considerable attention, as this system is able to produce hydrocarbons from carbon dioxide and water vapor under solar illumination.14−16 TiO2 nanotubes with copper particles led to the production of carbon monoxide, hydrogen, and hydrocarbons.14−16 Hydrogen production was also observed when copper particles were embedded in nanostructured TiO2.17−20 Still, fundamental questions remain unanswered about the mechanisms of photoabsorption, charge separation and transfer, molecule adsorption, and reaction. In particular, the role of copper in each of these steps remains to be elucidated, and such understanding could lead to a more systematic improvement of the catalyst’s properties. Regarding the effect of copper on photoabsorption, the only experimental evidence regards copper doping in titania, which has been shown to lead to a © 2015 American Chemical Society
Received: January 27, 2015 Revised: March 10, 2015 Published: March 10, 2015 6696
DOI: 10.1021/acs.jpcc.5b00846 J. Phys. Chem. C 2015, 119, 6696−6702
Article
The Journal of Physical Chemistry C
performed with the Quantum-Espresso code,55,56 employing plane-wave expansion and Vanderbilt ultrasoft pseudopotentials.57 Calculations were performed with the simplified version of DFT+U by Dudarev et al.,58 as implemented in the Quantum-Espresso code.59 The values of U have been taken from the literature: a value of 4.2 eV was used for titanium42,43,53 and of 5.2 eV for copper.60 Convergence of the total energy, for bulk anatase, has been achieved using cutoffs of 40 and 480 Ry for wave functions and charge density respectively, and a grid of (4 × 4 × 2) k-points generated with the method of Monkhorst and Pack.61 The (101) surface has been modeled with a rectangular (1 × 2) surface cell of size (10.27 Å × 7.57 Å), the (100) surface with a (2 × 1) surface cell of size (7.57 Å × 9.55 Å). This ensures that the minimal distance between clusters is at least of 5 Å, avoiding direct binding between neighboring clusters. This corresponds to less than 0.5 monolayers of copper on titania. Similar fractional coverages were obtained also in experiments.36 The k-point grids were chosen to have a k-point density similar to that employed for the bulk. The TiO2 slabs have a thickness of ∼10 Å, and the vacuum between the TiO2 slab and its copy is ∼15 Å thick. Clusters have been deposited on one side of the slab. We have checked that this does not introduce any important dipolar interaction with the copies: by increasing the vacuum by 20 bohr (corresponding to ∼10.6 Å), the surface energy changes remain well below 1 meV/Å2. All atomic positions were relaxed and atomic relaxations were performed until forces were smaller than 10−3 a.u.
the surface. These trapped electrons were not identified in the experiments in the presence of copper. For pure titania it is known that the reduced phase (TiO2−x) shows a Ti3+ state located within the band gap.28−30 Regardless of whether the Ti3+ state is related to an oxygen vacancy27 or a titanium interstitial defect,31 it is generally accepted that it has an essential role in enhancing the photocatalytic performance of this material.31−34 However, whether copper has an effect on the presence of Ti3+ sites remains an open question. Moreover, it has been shown that metal nanoparticles enhance the charge separation through the formation of a Schottky barrier, and photogenerated electrons are transferred to the metal nanoparticles as their Fermi energy is typically lower than that of the oxide.27,35 Whether this is true also for copper remains to be seen. It is thus clear that a better characterization of the effects of copper on the electronic structure of the system is necessary and would contribute to shedding light on the overall mechanism of the photocatalytic process and the specific role of copper in it. Given the many variables that can influence the activity of the catayst (shape, oxidation state, mixing state with titania, ...) and the many subprocesses involved in the photocatalytic reaction (photoabsorption, charge separation and transfer, molecule adsorption, chemical reactions, product desorption), it is important to be able to consider different aspects of the system one at a time, to disentangle the complex dependencies of all these issues with one another. In this context, help can come from computer simulations from first principles. In the present work, we have investigated small copper clusters (1−4 atoms) on the (101) and (100) surfaces of anatase by density functional theory (DFT) and DFT+U methods, to shed light on the interaction between copper and titania and its effect on the electronic structure of the system. It is indeed possible to obtain atomically dispersed copper on titania surfaces.36−38 We shed light on the interaction between copper and titania and its effect on the electronic structure of the system. We find that copper can induce both a shift in the bands of the material and the appearance of midgap states, depending on its coordination. Moreover, charge donated by the copper clusters to the TiO2 surface induces the formation of Ti 3+ species in its neighborhood. We argue that the presence of the dipole formed by the cluster and the Ti3+ species has an important role in delaying the recombination of the photoinduced electron and hole, where the hole is mainly attracted to the Ti3+ site, while electrons get trapped in the Cu4 cluster. In the next section, the computational methods are described, then the results are presented and discussed. Finally, a summary closes the article.
■
RESULTS AND DISCUSSION Copper clusters deposited on two surfaces of anatase TiO2 have been considered: (101) and (100). Anatase is the most active titania phase for photocatalytic applications,62 and these are the two surfaces with the lowest surface energies.63−66 For the sake of brevity, in the main body of the article, we show the figures relative to the (101) surface and we comment about the differences between the two surfaces. The full figures for the (100) surface are reported in the Supporting Information (SI). Atomic structures of small copper clusters (with 1−4 atoms) have been generated by taking several configurations used in the literature for similar systems, that is, Pd on TiO2(101)67 and Ag on TiO2(100),68 and by fully relaxing them. This approach is justified by the fact that we are considering small clusters, of up to four atoms. In principle, extensive structure search employing global optimization techniques such as genetic algorithms can be used to find the lowest-energy structures of clusters and surfaces.69,70 Still, for such small clusters these techniques invariably yield simple planar or pyramidal geometries.71,72 Of course, the complexity of the geometries increases very quickly if one considers clusters larger than those investigated in the present work. Structural Properties. TiO2(101). The configurations of the energetically most stable clusters on defect-free (101) are shown in Figure 1. In Table 1 the adsorption energies of all the clusters are reported, calculated as
■
COMPUTATIONAL METHODS Spin-polarized density functional theory (DFT) in the generalized-gradient approximation (GGA) of Perdew, Burke, and Ernzerhof (PBE)39 has been used throughout, also with the addition of the Hubbard term (DFT+U).40 It is wellknown41−44 that simple DFT-based approaches fail to predict the presence of the Ti3+ state in the electronic structure, due to the self-interaction error inherent to the functionals used in the local density and the generalized gradient approximations. To correct this error, post-DFT methods like DFT+U and hybrid functionals have been implemented.41−54 In fact, it has been shown that DFT+U is able to improve the description of the electronic structure of 3d transition metal oxides, including that of localized Ti3+ defect states.30,42−44,46,53 All simulations were
Eads = −
ECu@slab − Eslab − N × ECu atom NCu atoms
(1)
where ECu@slab is the total energy of the system with the cluster on the slab, Eslab is the total energy of the bare surface, and ECu atom is the energy of an isolated copper atom, and NCu atoms is the number of copper atoms in the cluster. 6697
DOI: 10.1021/acs.jpcc.5b00846 J. Phys. Chem. C 2015, 119, 6696−6702
Article
The Journal of Physical Chemistry C
slightly elongated with respect to the smaller clusters (2.37− 2.38 Å). We calculated also planar geometries for the tetramer, but they lie ∼0.5 eV higher. It is to be noticed that, in vacuum, the planar geometry is the lowest-energy configuration;74 the pyramid is thus stabilized by the surface. As in the case of the Cu2 cluster, in the Cu4 there is a direct Cu−Ti bond, although with a longer distance of 2.92 Å. As we will show later, this bond will introduce changes in the electronic structure that could play an important role in the photocatalytic activity of the Cu/TiO2 compound. TiO2(100). TiO2(100) has a higher surface energy as a consequence of the higher number of dangling bonds present at the surface. In the stable configuration of the Cu monomer, the distance of copper to two oxygen atoms is 1.89 Å, and its distance to two Ti atoms is 2.82 Å (see Figure 1a of the SI). The Cu atom lies almost exactly along the internuclear O−O axis, whereas silver lies at the center of a triangle formed by three O atoms (see Figure 2a in ref 68). The difference is probably due to the different size of the Cu and Ag ions: the smaller Cu would be too far from the oxygen if put at the center of the triangle, and therefore this configuration lies 0.2 eV higher in energy. The interaction with the surface is so strong that the dimer is not stable, and the most convenient configuration for two Cu atoms is one where each Cu atom lies in a monomer configuration, sharing one oxygen with the other Cu atom. In fact, as shown in Table 1, on (100) the monomer is more stable than all other clusters considered. We take this as a sign of the higher reactivity of this surface, as shown also by the higher surface energy. The trimer prefers a triangular configuration with one copper atom pointing toward the vacuum and being coordinated only to the other two Cu atoms. The triangular configuration is the same as that found on (101). Also, the tetramer prefers to be in the tetrahedral geometry (pyramid) as on (101), probably favored by the ideal configuration of the surrounding oxygen atoms, such that each of the three Cu atoms of the basis of the pyramid are bound to the other three Cu atoms and to one oxygen atom (1.82−1.84 Å; see Figure 1d of the SI). A flat tetramer parallel to the surface lies 0.1 eV higher in energy, while flat tetramers perpendicular to the surface, such as those shown in Figure 4 of ref 68 lie at least 0.7 eV higher. In the stable trimer and tetramer there is no direct Cu−Ti bond and the trimer is slightly more stable than the tetramer on both surfaces. Electronic Properties. TiO2(101). The atomic configuration has important consequences for the electronic structure of the system, showing differences between the results obtained for the smaller clusters Cu1−3 and the Cu4 cluster, as shown in Figure 2. The extension of the edge of the valence band (VB) into the gap is mainly due to low-coordinated Cu atoms present in the Cu1−3 clusters (Figure 2a−c). Here the 3d and 4 s states of the copper atoms with coordination number smaller than 3 are those that contribute most to the top of the VB, extending it by ∼0.7 eV into the band gap of titania. The higher the coordination number is, the lower in energy the 3d and 4s states; for example, there is a shift of ∼0.6 eV between the peak of the 3d states of the three-coordinated Cu atom and the twocoordinated Cu atom in the dimer (see Figure 2 of the SI). This trend is confirmed in the tetramer where just a small contribution, from the 3d states, to the edge of the VB is observed (see Figure 3 of the SI). A second important feature in the electronic structure is the appearance of midgap states. There can be two types of such states. The first state consists of Cu-4 s and Ti-3d orbitals, is
Figure 1. Atomic structure of the stable clusters with 1−4 Cu atoms on the defect-free surface of anatase TiO2(101). Blue balls: titanium; red balls: oxygen; gray balls: copper.
Table 1. Adsorption Energies (Eads), Total Charge Donated to the Surface (Qdon(|e|)) and Change in the Polarization of the 6c-Ti+3 Atom (Δμ(μB)) Produced by Copper Clusters on Anatase Surfaces (101) and (100), as Calculated with DFT +U Cu 1 2 3 4
Eads (eV/Cu)
Qdon(|e|)
Δμ(μB)
(101)/(100) 2.30/2.83 2.08/2.55 2.48/2.57 2.34/2.55
(101)/(100) 0.66/0.69 0.26/1.34 0.73/0.73 0.76/1.27
(101)/(100) 0.98/0.97 0.01/0.96 0.97/0.95 0.95/0.98
A single copper adatom (Cu1) prefers to lie at a bridge position between two surface oxygen atoms. In this configuration, the distance of copper from the two neighboring oxygen atoms is 1.88 Å and its distance to two Ti atoms is 3.04 Å. A similar bridge position was found to be the most stable for the copper adatom also on the (110) surface of rutile.73 This structure was also the most stable for palladium on anatase TiO2(101).67 On the contrary, the case of the Cu dimer (Cu2) is different from palladium. We suspect that this is due to the different dimer length and stiffness for Pd and Cu: for palladium, lengths of 2.63−2.86 Å were calculated, while copper displays lengths in a much narrower range (2.28−2.31 Å). The most stable structure, shown in Figure 1b, is the one that can better accommodate this Cu−Cu length on the surface. The other structures reported in Figure 6 of ref 67 lie some tenths of eV higher in energy. In the stable dimer structure, the second atom is bound only to the first Cu and to a five-coordinated Ti (5cTi) atom at the surface, with a Cu−Ti distance of 2.59 Å. In fact, the dimer is the least stable among the cluster sizes considered (Cu1−4), probably because of the unfavorable geometry of this second Cu atom. The stiffness of the Cu−Cu bond is found also in the trimer (Cu3), where the bond lengths are 2.31−2.34 Å. The stable atomic configuration for the trimer has two Cu atoms building a bridge between two surface oxygens, and the third is pointing toward the vacuum (Figure 1c). The tetramer (Cu4) prefers the compact coordination of a tetrahedron (pyramid), where each Cu atom is bound to the other three. Here the distances are 6698
DOI: 10.1021/acs.jpcc.5b00846 J. Phys. Chem. C 2015, 119, 6696−6702
Article
The Journal of Physical Chemistry C
Figure 2. Density of states of the stable clusters with 1−4 Cu atoms on the surface of anatase TiO2(101), calculated with DFT+U. Upper part of each diagram: majority spin. Solid line: total DOS; red dashed line: Cu-3d states; green solid line with triangles: 4s states of the Cu atom with a Cu− Ti bond; brown solid line with circles: Ti-3d states of a 6-fold coordinated Ti atom in the vicinity if the Cu cluster (see text); dark brown solid line with diamonds: Ti-3d states of a 5-fold coordinated Ti atom with a direct Cu−Ti bond. The projections on the single states have been multiplied by 20 to put them on the same scale as the total DOS. The zero of energy has been set to the enegy of the lowest unoccupied orbital. We have chosen to set the zero of energy to the energy of the lowest unoccupied orbital because this choice makes the electronic structure of different systems easier to compare.
donated charge, Qdon, calculated using the Bader method,75,76 is between +0.66|e| and +1.34|e| for systems with polarons (and just 0.26|e| for the dimer on the (101) surface), as shown in Table 1, resulting in a net positive charge present on the copper cluster. Of the donated charge, only 0.2−0.3|e| is transferred to the Ti atom where the polaron is sitting, while the rest is delocalized on the oxygen atoms. This Ti atom is usually indicated as a Ti3+ ion, but it is known that effective atomic charges hardly ever reproduce the formal charges in transition metal compounds (see ref 77 and references therein). This is usually attributed to the idealized nature of formal oxidation states and to the fact that electronic charge is anyway delocalized over many atoms in the real material. Moreover, effective atomic charges may be built according to different principles, either from the total electronic charge distribution, as in the case of Bader charges,75,76,78,79 or from the wave functions, as in the case of Löwdin charges.80,81 In the latter case, the actual values of the effective charges may depend on the choice of basis set and atomic wave functions used for the projection. This makes an assignment of charge to atoms not univocal, and therefore the resulting values should be used with care. To make the situation more compicated to analyze, the electrons donated to these metal oxides go also to oxygen atoms around the cation.82 Still, it is accepted in the literature that a polaron in a transition metal oxide can be individuated by an investigation of atomic effective charges, of the magnetic moments and of the geometric distortion of the atomic configuration. In particular, it has also been shown that in these cases the spin localization, given by an increase in the atomic magnetic moment, is a better indicator of the presence of the Ti3+ ions than the effective charges.83−86 The change in magnetic moment for the Ti atom in question is between 0.95 and 0.98 μB, as calculated through Löwdin analysis80 (see Table
due to a direct Cu−Ti bond (Figure 2b,d), and is present both in DFT and DFT+U. A second state is present only when the system is relaxed with DFT+U, because it is due to polaron formation at a six-coordinated Ti (6c-Ti) atom lying right below the copper cluster. The important role of the Cu−Ti bond in determining the presence of states in the gap is confirmed by the analysis of the trimer on (100), where there is no direct Cu−Ti bonding. As a consequence, the midgap state due to this bond is absent (see Figure 2c). Midgap states due to the presence of metal−Ti bonds have also been seen in the case of Ag clusters.68 Below we will discuss how the roles of these two midgap states for the dynamics of excess charges should be different. TiO2(100). The projected DOS of clusters on (100) shows the same features as on (101) (see Figure 4 of the SI). The main differences are related to the different atomic configurations with respect to (101). The shortest distance between the Cu adatom and Ti atoms is 2.82 Å, much larger than the Cu−Ti bond found in the dimer on (101), and there are O atoms in between. As a consequence, the only midgap states are observed for polarons on subsurface titanium ions. Since the stable configuration with two copper atoms consists of two separately adsorbed Cu atoms, each induces its own subsurface polaron. Also for the tetramer, two polarons are found below the copper cluster. In these two cases the charge donated to the surface (Qdon) is larger (see Table 1). Polaron State. Polaron formation has been observed for all systems in both the (101) and the (100) surfaces, with the exception of the dimer on the (101) surface. In all cases we have characterized the polaron by the magnetic moment of the Ti atom, its charge and by the geometric distortion. As mentioned above, the polaron forms at a 6c-Ti atom lying right below the copper cluster. Depending on the cluster size, the 6699
DOI: 10.1021/acs.jpcc.5b00846 J. Phys. Chem. C 2015, 119, 6696−6702
Article
The Journal of Physical Chemistry C
calculations support the view that copper can induce the formation of further Ti3+ sites, beside those present near oxygen vacancies. If it is true that Ti3+ sites play a decisive role in fostering the photocatalytic activity, then one of the main positive effects of copper could be that of increasing their number in the vicinity of the surface. We note however that there is a major difference in the electronic structure between Cu clusters with less than four atoms and the Cu4 cluster. The tetramer displays Cu empty states right at the bottom of the conduction band, while these states are absent for the other clusters. These states are important because they are related to the dynamics of photogenerated electrons. Indeed, metal nanoparticles on TiO2 can typically trap photogenerated electrons and make them available for reduction of adsorbed molecules.27,35 This is not the case for clusters containing less than four atoms, as the available free states of Cu are high in energy in the CB (Figure 2a−c). On the contrary, the Cu4 cluster shows a localized empty state at the bottom of the CB, which is the minority spin channel of the state shown in Figure 3b. This is a state where photogenerated electrons can be trapped, in agreement with what is usually expected for metal clusters at the surface of titania, promoting charge separation and transfer to the reactants.27,93 This situation is analogous to that encountered at macroscopic metal−semiconductor interfaces, where a Schottky barrier forms,35 retarding the recombination of a e−−h+ pair and therefore increasing its lifetime.94,95 To investigate the fate of electrons and holes in this system, we have considered the electronic structure of charged cells with the copper tetramer on (101), a cell with one additional electron and a cell with a hole (Figure 4). The excess electron occupies the minority-spin state at the Cu−Ti bond (Figure 4a); this state is spatially similar to its majority-spin counterpart depicted in Figure 3b and was the lowest lying unoccupied
1). Bader and Löwdin analysis give similar results when calculating the magnetic moment of an ion.86 The Ti−O distances are also increased and amounts to 2.16 and 2.14 Å for the azimuthal O atoms, and to 2.07, 2.18, 1.98, and 1.97 Å for the basal O atoms (here we show data just for the Cu4 cluster on (101)). For comparison, the other 6-fold coordinated Ti atom under the Cu cluster has distances of 2.00 and 2.05 (azimuthal) and 1.89, 1.92, 2.03, and 2.05 Å (basal). The polaronic state is visualized in real space in Figure 3a. This state
Figure 3. Isodensity surfaces for single midgap states for the system with the tetramer on (101), as relaxed with DFT+U: (a) polaronic state consisting mainly of t2g states of subsurface six-coordinated titanium atoms (see text for a detailed description); (b) state consisting mainly of s and p states of two of the copper atoms. These states are those of the majority spin and are occupied.
is pushed down into the gap and is clearly visible in the DOS (Figure 2d). Its position is over 1 eV below the edge of the CB, close to the position of the F-centers created by oxygen vacancies at the (110) surface of rutile, which experimentally lie 0.7−0.9 eV below the edge.53,87−89 We have checked the possibility of polaron formation on other Ti ions, and found that the polaron can exist as a metastable state in other positions, with energy differences smaller than 0.2 eV in most cases. We compare this with the situation of polarons induced by VO in rutile (110) surface, where activations barriers were ∼0.1 eV,49 allowing for electron hopping at room temperature.53 Therefore, we expect that at room temperature a polaronic state will form in the 6c-Ti atom below the Cu cluster, as shown in Figure 3a, and will remain in the nearsurface region. We think that this property is also important for the enhancement of the photocatalytic performance of the Cu/ TiO2 system as argued below. As a final note, the existence of such Ti3+ sites in the presence of copper clusters could be tested by electron paramagnetic resonance.90 Consequences for the Photocatalytic Activity of Cu/ TiO2. We have shown that the effect of the Cu clusters on the electronic structure of titania is 2-fold. First, Cu atoms with a coordination number lower than three contribute to the extension of the edge of the valence band (VB) into the gap. Second, Cu clusters induce the appearance of midgap states of two kinds: one state is due to the presence of a Cu−Ti bond, while the other is a Ti3+ polaron state generated by the localization of the charge coming from the copper on a 6c-Ti atom right below the cluster. In pure titania, it has been suggested that storage of electrons at Ti3+ sites determines the photocatalytic activity,4,27,91 maybe by increasing the photoabsorption.92 Indeed, it is known that polarons easily migrate to the surface in titania, and that electrons from Ti3+ sites are involved in dissociative electron attachment of CO2 even in the dark.27 In pure titania, Ti3+ ions have been observed in correspondence of oxygen vacancies (VO-Ti3+ sites).96 Our
Figure 4. Density of states for the charged cluster with four Cu atoms system on the surface of anatase TiO2(101). Upper part of each diagram: majority spin. (a) System with an excess electron; (b) System with an excess hole. Here we have used the same lines and color codes of Figure 2. The zero of energy has been set to the energy of the lowest unoccupied orbital. 6700
DOI: 10.1021/acs.jpcc.5b00846 J. Phys. Chem. C 2015, 119, 6696−6702
The Journal of Physical Chemistry C
■
orbital in the neutral system. The polaron state remains unchanged. On the contrary, in the system with a hole, the Ti3+ polaron state disappears from the DOS, while the other features remain unchanged (Figure 4b). Thus, these results support the idea that a copper cluster at the surface of titania is very important for charge separation, with holes migrating toward Ti3+ polaron sites and electrons moving toward the copper cluster. Additionally, the presence of Ti3+ near the Cu−Ti interface could contribute to the creation of active sites for the CO2 reduction near this interface, even in the dark. Indeed, in pure reduced titania it was observed that Ti3+ sites lead to dissociative electron attachment to CO2, with production of a CO adsorbed to Cu+.27 This hypothesis will be addressed in future studies.
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS Computational resources were provided by CINECA under IscrC-Copsol and by The Abdus Salam ICTP. Y.C.’s position was partly sponsored by the “ERC Advanced Grant 320796 − MODPHYSFRICT”.
■
■
REFERENCES
(1) Fujishima, A.; Honda, K. Nature 1972, 238, 5358. (2) Asahi, R.; Morikawa, T.; Ohwaki, T.; Aoki, K.; Taga, Y. Science 2001, 293, 269. (3) Serpone, N.; Emeline, A. V. J. Phys. Chem. Lett. 2012, 3, 673. (4) Kamat, P. V. J. Phys. Chem. C 2012, 116, 11849. (5) Kamat, P. V. J. Phys. Chem. Lett. 2012, 3, 663. (6) Maeda, K.; Domen, K. J. Phys. Chem. Lett. 2010, 1, 2655. (7) Teoh, W. Y.; Scott, J. A.; Amal, R. J. Phys. Chem. Lett. 2012, 3, 629. (8) Girish Kumar, S.; Devi, L. G. J. Phys. Chem. A 2011, 115, 13211. (9) Chen, X.-B.; Liu, L.; Yu, P. Y.; Mao, S. S. Science 2011, 331, 746. (10) Tao, J.-G.; Luttrell, T.; Batzill, M. Nat. Chem. 2011, 3, 296. (11) Linic, S.; Christopher, P.; Ingram, D. B. Nat. Mater. 2011, 10, 911. (12) Lee, J. H.; Hevia, D. F.; Selloni, A. Phys. Rev. Lett. 2013, 110, 016101. (13) Seriani, N.; Pinilla, C.; Cereda, S.; De Vita, A.; Scandolo, S. J. Phys. Chem. C 2012, 116, 11062. (14) Varghese, O. K.; Paulose, M.; LaTempa, T. J.; Grimes, C. A. Nano Lett. 2009, 9, 731. (15) Roy, S. C.; Varghese, O. K.; Paulose, M.; Grimes, C. A. ACS Nano 2010, 4, 1259. (16) Shankar, K.; Basham, J. I.; Allam, N. K.; Varghese, O. K.; Mor, G. K.; Feng, X.; Paulose, M.; Seabold, J. A.; Choi, K.-S.; Grimes, C. A. J. Phys. Chem. C 2009, 113, 6327. (17) Montini, T.; Gombac, V.; Sordelli, L.; Delgado, J. J.; Chen, X.W.; Adami, G.; Fornasiero, P. ChemCatChem 2011, 3, 574. (18) Cargnello, M.; Gasparotto, A.; Gombac, V.; Montini, T.; Barreca, D.; Fornasiero, P. Eur. J. Inorg. Chem. 2011, 28, 4309. (19) Barreca, D.; Carraro, G.; Gombac, V.; Gasparotto, A.; Maccato, C.; Fornasiero, P.; Tondello, E. Adv. Funct. Mater. 2011, 21, 2611. (20) Gombac, V.; Sordelli, L.; Montini, T.; Delgado, J. J.; Adamski, A.; Adami, G.; Cargnello, M.; Bernal, S.; Fornasiero, P. J. Phys. Chem. A 2010, 114, 3916. (21) Yoong, L. S.; Chong, F. K.; Dutta, B. K. Energy 2009, 34, 1652. (22) Lopez, R.; Gomez, R.; Llanos, M. E. Catal. Today 2009, 148, 103. (23) Tang, W.; Peterson, A. A.; Varela, A. S.; Jovanov, Z. P.; Bech, L.; Durand, W. J.; Dahl, S.; Nørskov, J. K.; Chorkendorff, I. Phys. Chem. Chem. Phys. 2012, 14, 76. (24) Li, C. W.; Ciston, J.; Kanan, M. W. Nature 2014, 508, 504. (25) Liu, L.; Gao, F.; Zhao, H.; Li, Y. Appl. Catal. B: Environ. 2013, 134−135, 349. (26) Liu, D.; Fernandez, Y.; Ola, O.; Mackintosh, S.; Maroto-Valer, M.; Parlett, C. M. A.; Lee, A. F.; Wu, J. C. S. Catal. Commun. 2012, 25, 78. (27) Liu, L.; Li, Y. Aerosol Air Qual. Res. 2013, 14, 453. (28) Henderson, M. A.; Epling, W. S.; Peden, C. H. F.; Perkins, C. L. J. Phys. Chem. B 2003, 107, 534. (29) Kurtz, R. L.; Stock-Bauer, R.; Msdey, T. E.; Roman, E.; De Segovia, J. L. Surf. Sci. 1989, 218, 178. (30) Setvin, M.; Franchini, C.; Hao, X.; Schmid, M.; Janotti, A.; Kaltak, M.; van de Walle, C. G.; Kresse, G.; Diebold, U. Phys. Rev. Lett. 2014, 113, 086402.
CONCLUSIONS In this work, the atomic and electronic structure of small copper clusters (1−4 atoms) on the (101) and (100) surfaces of anatase have been investigated by DFT and DFT+U. The main conclusions are as follows: (a) On the more stable (101), copper tends to form clusters; on (100) the Cu-TiO2 interaction is stronger and, for example, favors single adatoms over dimers. (b) The effect of copper on the electronic structure is 2-fold: first, Cu atoms with small coordination numbers, that is,
