Switchover of Reaction Mechanism for the Catalytic Decomposition of

The convergence tolerance for SCF was 1.0 × 10-5 Ha. Those of geometry optimization for energy and maximum force were 2.0 × 10-5 Ha and 0.04 Ha/nm, ...
0 downloads 0 Views 659KB Size
J. Phys. Chem. C 2007, 111, 16379-16386

16379

Switchover of Reaction Mechanism for the Catalytic Decomposition of HCOOH on a TiO2(110) Surface Yohei Uemura,† Toshiaki Taniike,† Mizuki Tada,† Yoshitada Morikawa,‡ and Yasuhiro Iwasawa*,† Department of Chemistry, Graduate School of Science, The UniVersity of Tokyo, Hongo, Bunkyo-ku, Tokyo 113-0033, Japan, and The Institute of Scientific and Industrial Research, Osaka UniVersity, Mihogaoka, Ibaraki, Osaka 567-0047, Japan ReceiVed: June 12, 2007; In Final Form: August 19, 2007

Catalytic decomposition of HCOOH on a TiO2(110) surface switches over between unimolecular dehydration (HCOOH f H2O + CO) and bimolecular dehydrogenation (HCOOH f H2 + CO2), depending on the reaction conditions. As the dehydration and dehydrogenation reactions proceed on acidic and basic oxide catalyst surfaces, respectively, the aspect observed on the same single crystal surface seems to be not compatible with the conventional acid-base catalysis concept. To clarify the origin of the switchover of the acid-base catalysis, the reaction mechanism of the HCOOH dehydrogenation was studied by density functional theory (DFT) calculations. It was concluded from the DFT calculations together with the rate equation and experimentally determined activation energy that the bimolecular dehydrogenation proceeds between a strongly adsorbed bridging formate anion and a weakly adsorbed HCOOH molecule by cooperative catalysis of three adjacent surface Ti4+ ions as Lewis acidic sites on the surface. This mechanism is entirely different from the previous dehydration mechanism that the dehydration occurs on an oxygen point defect (basic character) formed in situ by H2O desorption from two OH under the catalytic dehydration reaction conditions. Thus, the TiO2(110) surface provides two kinds of active sites for the HCOOH decomposition in a manner different from the traditional acid-base catalysis concept.

1. Introduction Despite the success of a wide range of applications of metal oxides to a variety of industrial catalysts and their supports, structures, and electronic properties of their surfaces at the atomic and molecular levels, dynamic behaviors of adsorbates during adsorption and catalytic reactions at the surfaces and hence detailed reaction mechanisms still remain unclear. One of the most important catalytic properties of metal oxides is the acid-base property, which regulates the performance of the catalysts for the dehydrogenation and dehydration of alcohols, formic acid, ethanolamine, and so on.1 The acid-base property of heterogeneous oxide catalysts has been characterized by the decomposition reaction of formic acid as a probe reaction, because the products are regulated by the acid-base property of the oxide surface.2-8 Typically, the dehydration reaction (HCOOH f CO + H2O) occurs on acidic oxides such as Al2O3 and TiO2, while the dehydrogenation reaction (HCOOH f CO2 + H2) occurs on basic oxides such as MgO and ZnO. Various oxide single crystal surfaces have been used as model catalysts to identify active sites and reaction intermediates on the surfaces,9-45 and the HCOOH decomposition has also received much attention theoretically and experimentally to explore the acidic-basic reactivity and active sites of the single crystal surfaces.9-29 In contrast to the conventional acid-base catalysis concept based on oxide powder catalysts, the dehydration of HCOOH occurred on a basic MgO(100)9 and CeO2* Corresponding author. Telephone: +81-3-5841-4363. Fax: +81-35800-6892. E-mail: [email protected]. † The University of Tokyo. ‡ Osaka University.

(110) and (110) surfaces10,11 Furthermore, we found that the catalytic reaction of HCOOH on a TiO2(110) surface violated the uniqueness rule of the acidity-basicity of the catalysts.20-22 The HCOOH dehydration dominantly occurred under a lowpressure atmosphere of HCOOH and at high temperatures above 500 K, while the dehydrogenation of HCOOH occurred under a relatively high-pressure atmosphere and at low temperatures below 450 K. The kinetics of these reactions was completely different from each other; the dehydration was a unimolecular reaction with an activation energy of 120 kJ mol-1 (1.25 eV), whereas the dehydrogenation was a bimolecular reaction (first order to both the coverage and pressure of HCOOH) with an activation energy of 15 kJ mol-1 (0.16 eV). It remains to be uncovered why the acidic and basic reactions switch over on the same single crystal TiO2(110) surface. It may also be strange that an additional acidic HCOOH molecule promotes the dehydrogenation categorized as basic catalysis. We proposed the mechanism for the dehydration of HCOOH on a TiO2(110) surface by means of density functional theory (DFT) calculations and scanning tunneling microscopy (STM).23,42 Figure 1 shows the proposed mechanism for the HCOOH dehydration, which was the most stable pathway in the DFT calculations and strongly supported by STM observations. Briefly, HCOOH dissociatively adsorbs to form a bridging formate (HCOO-) with an O-C-O plane parallel to the [001] direction and a bridging OH group (A). Two OH groups merge to form H2O, and the produced H2O is desorbed, leaving an oxygen defect site (2ObH f H2O + Ob + O-defect site, B) on the surface in Figure 1. The bridging formate migrates on a one-dimensional Ti4+ row and occupies an oxygen defect site with an oxygen atom of the formate, where the O-C-O plane

10.1021/jp074524y CCC: $37.00 © 2007 American Chemical Society Published on Web 10/06/2007

16380 J. Phys. Chem. C, Vol. 111, No. 44, 2007

Figure 1. (A) Surface structure of TiO2(110) characterized by STM and NC- AFM. (B) Mechanism for the dehydration of HCOOH on the TiO2(110) surface.23 The decarboxylation of D to E is rate-determining.

rotates to become parallel to the [11h0] direction (C). The bridging formate at the defect site transforms into a monodentate formate, and the monodentate formate gives its H atom to the adjacent bridging O atom accompanied with decomposition into CO and bridging O (HCOOad(d)+ Ob f CO + Ob + HOb) (D f E), which is the rate-limiting step in the HCOOH dehydration. The STM images confirmed the existence of the three intermediate formates B-D and their migration and mutual transformations.42 The active site is the oxygen point defect, which was self-formed via water extraction and self-mediated via HCOOH dissociation during the catalytic dehydration cycle. In this article, we report the studied mechanism of the HCOOH dehydrogenation by DFT calculations and discuss the mechanism of the switchover of the reaction pathways between the dehydrogenation and dehydration on a single crystal TiO2(110) surface. 2. Computational Methods Density functional calculations with the GGA-PBE46 functional were performed with Materials Studio Dmol3 from Accelrys. The basis sets were DND and effective core potentials (DND is as accurate as 6-31G*).47 The real space cutoff radius was 0.45 nm. The convergence tolerance for SCF was 1.0 × 10-5 Ha. Those of geometry optimization for energy and maximum force were 2.0 × 10-5 Ha and 0.04 Ha/nm, respectively. A transition state (TS) between two immediate stable structures was first identified by linear synchronous transit48 and then cyclically refined by quadratic synchronous transit and conjugate gradient methods. Each TS was converged within 0.1 Ha nm-1. The lattice constants of the rutile TiO2 were optimized to be 0.655 × 0.296 nm. The slab method was employed to model a TiO2(110) surface, where the thickness

Uemura et al. of a vacuum layer was 1 nm. The slab thickness was decided to be three layers (one layer involves three atomic layers, i.e., O-Ti-O) according to the reason described later. The atomic arrangement of a bottom layer was fixed at the bulk arrangement. To evaluate adsorption energies of HCOOH on a TiO2(110) surface, a full covered surface by HCOOH (denoted as R1-R8) was used in the calculations; a p(1 × 1) periodicity for HCOOH adsorption was used for R1, R2, and R3, and a p(1 × 2) periodicity for HCOOH adsorption was for R4, R5, R6, R7, and R8. In the calculations of adsorption energies of HCOOH on a defective surface, an oxygen defect (O-defect) was formed by removing a bridging oxygen from a stoichiometric p(1 × 2) surface. For the transition state search, a p(1 × 4) or p(2 × 2) HCOOH surface was used, where the k-point meshes were 3 × 2 × 1 and 2 × 3 × 1, respectively. In the calculations for the transition state search on the defective surface, an oxygen defect was formed by removing a quarter of the bridging oxygens from a stoichiometric p(2 × 2) surface, and thermal smearing of moderate strength49 was imposed to improve the SCF convergence. To decide the optimum thickness of the slab in the present study, the energies of some adsorption states of HCOOH were compared on the p(1 × 1) HCOOH surface, with the thickness of three, five, and seven layers. In these calculations, the k-point mesh was 3 × 7 × 1. The chosen adsorption states were the most stable bridging formate and a quasi-stable monodentate formate produced by the dissociative adsorption of HCOOH, and a less stable molecularly form. Their adsorption energies were 1.872, 1.059, and 0.562 eV for the three-layer slab; 1.498, 0.899, and 0.475 eV for the five-layer slab; and 1.406, 0.840, and 0.477 eV for the seven-layer slab. Thus, the adsorption energies decreased with an increase in the slab thickness because of increasing surface relaxation, and they were almost convergent for the seven-layer slab. However, the difference between the values for the three- and seven-layer slabs was at most 0.466 eV. Furthermore, the order of the stability for the three states was unchanged with the slab thickness, and the difference in the relative stability between three- and seven-layer slabs was at most 0.381 eV. Therefore, we employed the most efficient three-layer slab throughout the calculations involving TS search. Independently, very recently, Perron et al. also reported that the surface energy on TiO2(110) was almost convergent at fourlayer slab.50 At the cost of the thin slab in deciding plausible reaction paths for HCOOH dehydrogenation, the difference of 0.5 eV in the activation barrier was regarded as a margin, and all the reaction paths with activation energies less than 0.5 eV larger than the smallest activation energy were considered as candidates. 3. Results and Discussion 3.1. Adsorption States of HCOOH on a TiO2(110) Surface. Before the calculations of the transition states, various adsorption configurations of HCOOH on a TiO2(110) surface were investigated to seek possible reactants and intermediates for the HCOOH dehydrogenation. Considering previous experimental and theoretical studies on adsorption structures on TiO2(110) surfaces,23,30-45 we calculated various types of adsorbing structures of HCOOH to examine the optimized adsorbate structures on the stoichiometric and defective TiO2(110) surfaces (note that the adsorbates on the defective surface have the signature of (d)). Table 1 summarizes their adsorption energies and the Mulliken charges of the hydrogen atoms of associative and dissociative adsorbates. R1 and R2 are molecularly adsorbed species, where R1 adsorbs on a 5-fold coordinated Ti4+ with a carbonyl group of

Mechanism for Decomposition of HCOOH on TiO2(110)

J. Phys. Chem. C, Vol. 111, No. 44, 2007 16381

TABLE 1: Adsorption Energies of HCOOH on the TiO2 (110) Surface and Mulliken Charges of Hydrogen Atomsa adsorbate conformation

adsorption energy/eV

charge of H1b

charge of H2c

R1 R2 R3 R4 R5(d)d R6(d)d R7(d)d R8(d)d gas phase

-0.559 -0.562 -1.059 -1.872 -1.584 -1.992 -0.292 -0.181

0.330 0.241 0.215 0.219 0.181 0.221 0.047 0.006 0.140

0.437 0.552 0.552 0.472 0.490 0.480 0.476 0.468 0.421

a R1 and R2: associative adsorption. R3-R8: dissociative adsorption. b H1 is the hydrogen atom of C-H. c H2 is the hydrogen atom of O-H groups. Note that there are two types of OH groups for molecularly adsorbed HCOOH and bridging OH at the surface. d Adsorption on oxygen defect sites.

formic acid, while R2 adsorbs with a hydroxyl group. The adsorption energies of the molecular adsorption states were similar and small, and they may be regarded as precursor states for dissociative adsorption of formic acid. For the dissociative adsorption on the stoichiometric surface, there were two configurations of R3 and R4. R3 is monodentate formate, which binds to a 5-fold coordinated Ti4+ with an O atom of the carbonyl group. R4 is a bridging formate, where two O atoms of a formate bind to the neighboring surface Ti4+ ions, and the bridging formate (R4) was most stable. The adsorption energies of R1-R4 were very similar to the values by previous DFT calculations, which confirms the quality of our calculation method.35 On the other hand, four dissociative configurations (R5(d)-R8(d)) were found to be specific to a defective surface. R5(d) offers an O atom to a defect site and another O atom directs to the surface normal, while R6(d) binds to a defect site with an O atom and another O atom binds to a 5-fold coordinated Ti4+ in front of the defect site. R7(d) adsorbs on a 5-fold coordinated Ti4+ with an O atom and on a defect site with a C-H hydrogen. R8(d) is a formate species, where the C-H bond of R7(d) is nearly broken (not shown). The most stable species on the defective surface was R6(d), while unstable R7(d) and R8(d) may be regarded as a transient configuration. This is due to an electron transfer from the neighboring Ti3+ ions at the defect site, which weakens the Lewis acidity of the 5-fold coordinated Ti4+ site in front of the defect site. Experimentally, only R4 and R6(d) were observed at room temperature on a TiO2(110) surface by means of Fourier transform reflection-absorption infrared spectroscopy43 and X-ray photoelectron diffraction.37,44 The majority was R4 species as imaged by STM41 at elevated temperatures. STM revealed increasing population of R6(d) and the existence of R5(d) as intermediate species for the unimolecular dehydration which was suggested to proceed by the steps R4 f R6(d) f R5(d).42 The unimolecular decomposition of R5(d) to CO + OH is ratedetermining, as suggested by STM41 and DFT calculations.23 In all the configurations, the C-H hydrogen atoms denoted as H1 are electron richer than the OH hydrogen atoms denoted as H2 (Figure 2 and Table 1), and particularly H1 of R7(d) and R8(d) is richest because of direct electron donation from the two Ti3+ ions. Except these species, the adsorption on the Ti4+ Lewis acid site decreased the electron negativity of H1, compared to that of gas-phase molecule. On the other hand, the most positively charged H2 was observed with R2 and R3, which hydrogen bonded with the neighboring surface oxygen atoms. The other H2 had a similar charge to that of a gas-phase molecule. The modification of the H charge through adsorption

Figure 2. Possible adsorption structures of HCOOH on a TiO2(110) surface. R1, R2, R3, and R4 are species on a stoichiometric surface, while R5(d), R6(d), and R7(d) are species on an oxygen defect surface. The red, dark gray, and white balls represent O, C, and H atoms, respectively.

was much larger for electron-negative H1 than that for H2. These charges on H1 and H2 were strongly related to the genesis of the bimolecular reaction path as discussed hereinafter. 3.2. Reaction Pathways for Bimolecular Dehydrogenation. The dehydrogenation of HCOOH on a TiO2(110) surface is given by the second-order rate equation which is proportional to the formate coverage and the gas-phase HCOOH pressure.21 We performed the transition state searches for the bimolecular dehydrogenation reaction between a HCOOH(g) or molecularly weakly adsorbed HCOOH and dissociatively adsorbed species on the basis of the adsorption states mentioned earlier. The energy diagrams of the obtained pathways on the stoichiometric and defective surfaces are shown in Figures 3 and 4, respectively. Two kinds of activation energies for each path are listed in Table 2; ∆Eel is the activation energy of the rate-limiting elementary step, which corresponds to the height of the transition state from the neighboring stable or quasi-stable state, and ∆Eap is the apparent activation energy, which is the height of the transition state from the zero-level reference energy, that is, the energy of a clean surface and two HCOOH(g). Note that the experimentally observed activation barrier 0.16 eV corresponds to ∆Eap. We summarize all the obtained pathways before concentrating on the discussion on plausible pathways. Path 1: Direct collision between a surface hydroxyl group and a gas-phase HCOOH molecule. Coadsorbed formate is a spectator in this path. The H1 atom of a gas-phase HCOOH molecule collides with the H2 atom of a bridging hydroxyl group to produce H2(g) and a hydrocarbonate CO3H-. Its ∆Eap was 0.585 eV. Path 2: Reaction between a bridging formate (R4) and a weakly adsorbed HCOOH (such as R1) at the adjacent Ti4+ ions. When the H2 atom of the weakly adsorbed HCOOH and the H1 atom of R4 react with each other, both the adsorbates take a configuration close to monodentate species at the transition state, finally forming H2(g), CO2(g), and an original adsorbate (R4). Its ∆Eap became very small, -0.353 eV (the negative value was partly due to the overestimation of the adsorption energies on the three-layer slab as mentioned above). We also systematically examined other combinations of reactants on the perfect surface, but could not find any transition states due to significant steric repulsion between reactants around

16382 J. Phys. Chem. C, Vol. 111, No. 44, 2007

Uemura et al.

Figure 3. Energy diagrams for the possible dehydrogenation pathways on the stoichiometric TiO2(110) surface. The zero-level energy (E0) as reference is defined as the sum of the energies of the clean surface and of two HCOOH(g).

the transition states. For example, when H2 of HCOOH(g) and H1 of R3 were chosen as a reactant pair, the steric repulsion between the O(-H) atom of HCOOH(g) and the O atoms of R3 became dominant to make the transition state search fail. Paths 3 and 4: Reactions between H1 of a monodentate formate at the defective surface (R5(d)) and H2 of a gas-phase molecule.

In Path 3, a gas-phase molecule approaches the surface with its hydroxyl group directing toward the vacuum side, while in Path 4, the hydroxyl group of an incoming HCOOH directs toward the surface. In these paths, the reaction products are H2(g), surface CO2, and monodentate species (R3). ∆Eap values were 1.508 and 1.257 eV for Paths 3 and 4, respectively.

Mechanism for Decomposition of HCOOH on TiO2(110)

J. Phys. Chem. C, Vol. 111, No. 44, 2007 16383

Figure 4. Energy diagrams for the possible dehydrogenation pathways on the oxygen defect TiO2(110) surface. The zero-level energy (E0) as reference is defined as the sum of the energies of the clean surface and of two HCOOH(g).

Path 5: Reaction between R7(d) and molecularly adsorbed HCOOH (such as R1).

The H2 atom of the gas-phase HCOOH molecule reacts with the H1 atom of R7(d) trapped at an oxygen defect. At the

16384 J. Phys. Chem. C, Vol. 111, No. 44, 2007

Uemura et al.

TABLE 2: Relationship between the Activation Energies and the Mulliken Charges of the H Atoms of Two Reactants at the Transition State path

H1

H2

∆Eel/eVa

∆Eap/eVb

1 2 3 3 5 6 7

0.14 0.219 0.221 0.221 0.047 -0.046 -0.046

0.552 0.437 0.437 0.552 0.437 0.552 0.552

2.708 2.662 2.424 2.173 0.952 1.782 0.907

0.585 -0.353 1.508 1.257 0.212 1.007 -0.041

a ∆Eel is the intrinsic activation energy between the TS and the adsorption state. b ∆Eap is the apparent activation energy that is the height of the TS from the initial state before adsorption.

transition state, the R1-like molecule begins to transform into a monodentate species (R3). Its ∆Eap was 0.212 eV. We also calculated the reaction between an R7(d) species and another molecularly adsorbed species such as R2 but could not get any transition state because of large steric repulsion between the reactants. Path 6: Reaction between R8(d) and a bridging hydroxyl group. Path 6 consists of two elementary steps: the first step is the C-H scission of R7(d) at the defect site, where CO2(g) is produced; the last step is the dehydrogenation between the isolated hydrogen atom at the defect site and a bridging hydroxyl group of another HCOOH. The rate-limiting step was the last one, whose ∆Eap was 1.007 eV. Path 7: Reaction between R8(d) and a H2 atom of a HCOOH molecule weakly adsorbed such as R2. The transformation process from R7(d) to R8(d) was the same as in Path 6. The rate-limiting step was the second step, where ∆Eap was as small as -0.041 eV. The transition state search between an R8(d) species and another molecularly adsorbed HCOOH such as R1 was not successful for reasons similar to those in the cases of other unsuccessful searches. 3.3. Important Factors To Decide the Activation Energy for the Dehydrogenation. As a result of the series of the transition state searches, we found two important factors to decide the height of the transition state. A factor is the charges of the reacting H atoms to produce H2(g). Since H1 and H2 are electron negative and positive, respectively, the dehydrogenation reaction tends to be a heterolytic reaction. This fact suggests that the polarization between the two H atoms was important. The relation between the activation energies and the charges of the reacting H atoms in Table 2 indicates that the H1 charge is critical for ∆Eel. Concretely, the defect-trapped H1 atom gave very small ∆Eel values in Paths 5 and 7 (Path 6 was an exceptional case, where the large distance between the reacting H atoms raises ∆Eel). The other factor is the amount of the adsorption energy, which affects the energy level of the transition states. ∆Eap of Path 2 was comparable to or even smaller than those of Paths 5 and 7, despite the much larger value of ∆Eel. This was because the stabilization of the transition state by the adsorption energies was most effective in Path 2. At the transition state in Path 2, the two reactant molecules have a monodentate-like configuration, whose adsorption energies should be ca. 1 eV, as listed in Table 1. As seen in the electron density distribution of the transition state in Path 2 (Figure 5), both the leaving and adsorbing reactants interact with surface 5-fold coordinated Ti4+ ions. On the other hand, at the rate-limiting transition states of Paths 5 and 7, the adsorption energy of only one monodentatelike reactant is available because of the instability of R7(d) and

Figure 5. Electron density at the transition state in the most plausible dehydrogenation pathway (Path 2). Note that both the reactant molecules at the transition state interact with the surface.

R8(d). Furthermore, the transition state in Path 2 seems not to have any notable steric repulsion or distortion because of the moderate distance between the two neighboring Ti4+ ions. On the contrary, the elongated distance between the two reacting H atoms in Paths 5 and 7 made the use of the adsorption energy states at the transition ineffective. 3.4. Dominant Pathway for the Dehydrogenation. The results of Table 2 conclude that Paths 2, 5, and 7 are plausible pathways with sufficiently low activation energies compared to the experimental value (0.16 eV). Path 2 on the stoichiometric surface is advantageous for the available adsorption energies at the transition state, whereas Paths 5 and 7 on the defective surface are stable because of the electron negativity of the defecttrapped H atom. In this part, we decide which pathway is the most plausible under the actual catalytic reaction conditions. While Path 2 proceeds on the three neighboring 5-fold coordinated Ti4+ ions at a perfect surface, Paths 5 and 7 require an oxygen defect site a priori. It is known that the concentration of oxygen defects on a TiO2(110) surface is typically only less than 5% under the dehydrogenation reaction condition.25 In addition, the defect formation by H2O removal from two OH groups observed at elevated temperature >500 K is not probable at the dehydrogenation temperature below 450 K.21,25,42 Considering that R6(d) species was the most stable one at the oxygen defects, most of the defects must be occupied by R6(d), particularly at the higher pressures of HCOOH under the dehydrogenation reaction conditions. Namely, Paths 5 and 7 cannot be reaction paths for the HCOOH dehydrogenation. The R6(d) species is a reaction intermediate for the HCOOH dehydration as reported previously.36 Thus, we rejected Paths 5 and 7 based on the lack of the reactants R7 and R8 under the reaction conditions. On the other hand, the concentrations of reactants, binding formate R4 and weakly adsorbed formic acid, in Path 2 were sufficient at the surface because no special sites are required for this path and one of the reactants, R4, is the most stable species. Furthermore, high HCOOH coverage by increasing HCOOH pressure increases the coadsorption state in such a manner that a weakly adsorbed HCOOH is located adjacently to an R4 species. Thus, we propose that Path 2 should be the most plausible dehydrogenation pathway under the catalytic reaction conditions.21,22 3.5. Mechanism of the Switchover of Reaction Paths. Finally, we discuss the mechanism of the switchover of the HCOOH decomposition pathways between the dehydrogenation

Mechanism for Decomposition of HCOOH on TiO2(110)

J. Phys. Chem. C, Vol. 111, No. 44, 2007 16385 4. Conclusions

Figure 6. Reaction pathways for the HCOOH decomposition on TiO2(110).

and the dehydration on a TiO2(110) single crystal surface (Figure 6). We suggested the bimolecular dehydrogenation mechanism, where the reaction proceeds on the three neighboring Ti4+ sites, with the aid of the large adsorption energy of the reactants. As this mechanism requires neither special sites nor unstable reactants, the concentration of the reactants as well as the apparent activation barrier are quite reasonable. The dehydrogenation pathways on the oxygen defect sites were discarded because the concentration of the active site was regarded to be too small at the dehydrogenation temperature below 450 K. On the other hand, the unimolecular dehydration pathway previously proposed by means of both DFT and STM23,42 is catalyzed by the oxygen defect sites that are interestingly produced in situ in the catalytic dehydration reaction above 500 K. There are two different active sites on the TiO2(110) surface that enable the switchover of the dehydration and dehydrogenation, depending on the reaction conditions (Figure 6): at a low temperature and a high HCOOH pressure, in situ formation of the oxygen defect sites does not occur, and hence the bimolecular dehydrogenation preferentially proceeds on the three neighboring surface Ti4+ sites; at a high temperature and a low HCOOH pressure, the bimolecular dehydrogenation is kinetically suppressed, and instead the unimolecular dehydration preferably proceeds on the defect sites. Unimolecular decomposition of HCOOH on the SnO2(110) surface with the same structure as the TiO2(110) surface was promoted by reduction of the SnO2(110),15 which indicates that the dehydration of HCOOH occurs at oxygen defects of the SnO2(110) surface in a manner similar to that on the TiO2(110) surface. Henderson26 and Bowker et al.27 stated that there was no evidence for the dehydrogenation of HCOOH on TiO2(110). This controversy to our previous results should be attributed to the difference in the employed reaction atmospheres. We performed the decomposition of HCOOH under the HCOOH atmospheres, whereas they examined it under ultrahigh vacuum. According to our results, the bimolecular dehydrogenation never proceeds in the absence of the gas-phase HCOOH. Kecske´s et al.24 reported that HCOOH was converted to HCHO on a defective TiO2(110) surface at 300 K and that the produced HCHO reacted with HCOOH(g), leading to CO and H2 at 473 K. We could not have any evidence of this process in the previous STM work.20-23 Wang et al. studied the decomposition of DCOOD at 500 K on TiO2(110) surfaces with different initial concentrations of the oxygen defects.25 They observed the production of CO2 + D2 on a perfect TiO2(110) surface, similarly to our result,20-22 while the product changed to CO + D2 as the concentration of the defect sites increased. These results on the defective TiO2(110) surface reproduce the dehydration reactivity of TiO2 power catalysts.8 We found a denude zone of 1.4 nm on a terrace from the edge of an atomicheight step on a TiO2(110) surface, which means that no dehydrogenation reaction proceeds unless there are flat areas larger than 2.8-nm dimension without steps on the surface of TiO2 powders.22 Thus, the reaction pathways of HCOOH decomposition depend on the defect concentration as well as the geometric arrangement of the TiO2 surface.

The mechanism of the bimolecular dehydrogenation reaction of HCOOH on a TiO2(110) single crystal surface was investigated by DFT calculations. As a result of a series of transition state searches, we found that the activation energy for the dehydrogenation was decided by the charges of the reacting H atom of adsorbates (HCOOH and HCOO-) and the amount of the adsorption energy available at the transition state. The most plausible reaction pathway was the one between a bridging formate adsorbed on two 5-fold coordinated Ti4+ ions and HCOOH molecule weakly adsorbed at the adjacent Ti4+ ion. The dehydrogenation occurs by acid-base interaction between the two adsorbates. This signifies that the dehydrogenation process does not require a special site such as an oxygen defect, which was the active site for unimolecular dehydration of HCOOH. The switchover of the reaction pathways between the dehydrogenation and the dehydration depending on reaction temperature and gas-phase pressure, which was observed on the perfect TiO2(110) surface, originates from the two different active sites on the surface: three adjacent surface Ti4+ ions for the dehydrogenation and an oxygen defect site for the dehydration. The difference in the reactivity between the TiO2(110) single crystal surface and TiO2 powder catalysts should be due to the concentration of oxygen defects and the dimension of flat terrace area without steps. The classical acid-base concept for oxide catalysis is not appropriate to explain the oxide catalysis and should be modified. References and Notes (1) Acid-Base Catalysis: Proceedings of the International Symposium on Acid-Base Catalysis; International Symposium on Acid-Base Catalysis, Sapporo, Japan, Nov 28-Dec 1, 1988; Tanabe, K., Hattori, H., Yamaguchi, T., Tanaka, T., Eds.; Kodansha: Tokyo, 1989. (2) Mars, P.; Scholten, J. F.; Zwietering, P. AdV. Catal. 1963, 14, 35. (3) Aramedia, M. A.; Borau, V.; Garcia, I. M.; Jimenez, C.; Marinas, A.; Marinas, J. M.; Porras, A.; Urbano, F. J. Appl. Catal., A 1999, 184, 115. (4) Patermarakis, G. Appl. Catal., A 2003, 252, 231. (5) Poulston, S.; Rowbotham, E.; Stone, P.; Parlett, P.; Bowker, M. Catal. Lett. 1998, 52, 63. (6) Halawy, S. A.; Al-Shihry, S. S.; Mohamed, M. A. Catal. Lett. 1997, 48, 247. (7) Haffad, D.; Chambellan, A.; Lavalley, J. C. J. Mol. Catal. A: Chem. 2001, 168, 153. (8) Munuera, G. J. Catal. 1970, 18, 19. (9) Peng, X. D.; Barteau, M. A. Catal. Lett. 1990, 7, 395. (10) Stubenrauch, J.; Brosha, E.; Vohs, J. M. Catal. Today 1996, 28, 431. (11) Lintuluoto, M.; Nakatsuji, H.; Hada, M.; Kanai, H. Surf. Sci. 1999, 429, 133. (12) Yamamoto, H.; Watanabe, N.; Wada, A.; Domen, K.; Hirose, C. J. Chem. Phys. 1997, 106, 4734. (13) Bandara, A.; Kubota, J.; Onda, K.; Wada, A.; Domen, K.; Hirose, C. Surf. Sci. 1999, 435, 83. (14) Xu, C.; Goodman, D. W. Catal. Today 1996, 28, 297. (15) Gercher, V. A.; Cox, D. F. Surf. Sci. 1994, 312, 106. (16) Persson, P.; Lunell, S.; Qjamae, L. Int. J. Quantum Chem. 2002, 89, 172. (17) Yoshimoto, M.; Takagi, S.; Umemura, Y.; Hada, M.; Nakatsuji, H. J. Catal. 1998, 173, 53. (18) Ludviksson, A.; Zhang, R.; Campbell, C. T.; Griffiths, K. Surf. Sci. 1994, 313, 64. (19) Dilara, P. A.; Vohs, J. M. J. Phys. Chem. 1993, 97, 12919. (20) Onishi, H.; Aruga, T.; Iwasawa, Y. J. Am. Chem. Soc. 1994, 115, 10460. (21) Onishi, H.; Aruga, T.; Iwasawa, Y. J. Catal. 1994, 146, 557. (22) Iwasawa, Y.; Onishi, H.; Fukui, K.; Suzuki, S.; Sasaki, T. Faraday Discuss. 1999, 114, 259. (23) Morikawa, Y.; Takahashi, I.; Aizawa, M.; Namai, Y.; Sasaki, T.; Iwasawa, Y. J. Phys. Chem. B 2004, 108, 14446. (24) Kecske´s, T.; Ne´meth, R.; Rasko´, J.; Kiss, J. Vacuum 2005, 80, 64. (25) Wang, Q.; Biener, J.; Guo, X. C.; Farfran-Arribas, E.; Madix, R. J. J. Phys. Chem. B 2003, 107, 11709. (26) Henderson, M. A. J. Phys. Chem. B 1997, 101, 221.

16386 J. Phys. Chem. C, Vol. 111, No. 44, 2007 (27) Bowker, M.; Stone, P.; Bennett, R.; Perkins, N. Surf. Sci. 2002, 511, 435. (28) Borowiak, M. A. J. Mol. Catal. A: Chem. 2000, 156, 21. (29) Borowaik, M. A. J. Mol. Catal. A: Chem. 1999, 139, 97. (30) Uetsuka, H.; Sasahara, A.; Yamakata, A.; Onishi, H. J. Phys. Chem. B 2002, 106, 11549. (31) Smith, R. D.; Bennett, R. A.; Bowker, M. Phys. ReV. B 2002, 66, 7. (32) Bates, S. P.; Kresse, G.; Gillan, M. J. Surf. Sci. 1998, 409, 336. (33) Gutierpagerez-Sosa, A.; Martinez-Escolano, P.; Raza, H.; Lindsay, R.; Wincott, P. L.; Thornton, G. Surf. Sci. 2001, 471, 163. (34) Ka¨ckell, P.; Terakura, K. Appl. Surf. Sci. 2000, 166, 370. (35) Ka¨ckell, P.; Terakura, K. Surf. Sci. 2000, 461, 191. (36) Bennett, R. A.; Stone, P.; Smith, R. D.; Bowker, M. Surf. Sci. 2000, 454, 390. (37) Thevuthasan, S.; Herman, G. S.; Kim, Y. J.; Chambers, S. A.; Peden, C. H. F.; Wang, Z.; Ynzunza, R. X.; Tober, E. D.; Morais, J.; Fadley, C. S. Surf. Sci. 1998, 401, 261. (38) Wang, L. Q.; Ferris, K. F.; Shultz, A. N.; Baer, D. R.; Engelhard, M. H. Surf. Sci. 1997, 380, 352. (39) Onishi, H.; Iwasawa, Y. Langmuir 1994, 10, 4414.

Uemura et al. (40) Onishi, H.; Fukui, K.; Iwasawa, Y. Bull. Chem. Soc. Jpn. 1995, 68, 2447. (41) Onishi, H.; Iwasawa, Y. Chem. Phys. Lett. 1994, 226, 111. (42) Aizawa, M.; Morikawa, Y.; Namai, Y.; Morikawa, H.; Iwasawa, Y. J. Phys. Chem. B 2005, 109, 18831. (43) Hayden, B. E.; King, A.; Newton, M. A. J. Phys. Chem. B 1999, 103, 203. (44) Chambers, S. A.; Thevuthasan, S.; Kim, Y. J.; Herman, G. S.; Wang, Z.; Tober, E.; Ynzunza, R.; Morais, J.; Peden, C. H. F.; Ferris, K.; Fadley, C. S. Chem. Phys. Lett. 1997, 267, 51. (45) Diebold, U. Surf. Sci. Rep. 2003, 48, 53. (46) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. ReV. Lett. 1996, 77, 3865. (47) Delley, B. J. Phys. Chem. 1990, 92, 508. (48) Halgern, T. A.; Lispscomb, W. N. Chem. Phys. Lett. 1977, 49, 225. (49) Delley, B. In Modern Density Functional Theory: A Tool for Chemistry; Theoretical and Computational Chemistry 2; Seminario, J. M., Politzer, P., Eds.; Elsevier: Amsterdam, 1995; pp 221-254. (50) Perron, H.; Domain, C.; Rogers, J.; Drot, R.; Simon, E.; Xatalette, H. Theor. Chem. Acc. 2007, 117, 565.