Carbon−Carbon Bond Formation on Model Titanium Oxide Surfaces

9828

J. Phys. Chem. C 2008, 112, 9828–9834

Carbon-Carbon Bond Formation on Model Titanium Oxide Surfaces: Identification of Surface Reaction Intermediates by High-Resolution Electron Energy Loss Spectroscopy Hengshan Qiu, Hicham Idriss,† Yuemin Wang,* and Christof Wo¨ll Physical Chemistry I, Ruhr-UniVersity Bochum, 44780 Bochum, Germany ReceiVed: February 14, 2008; ReVised Manuscript ReceiVed: April 2, 2008

The interaction of CH2O with perfect and defective TiO2(110) surfaces (produced by overannealing and Ar ion sputtering methods) was studied by thermal desorption spectroscopy, high-resolution electron energy loss spectroscopy (HREELS), and density functional theory (DFT) calculations. Exposing the perfect TiO2(110) surface to CH2O at 100 K leads to the formation of physisorbed CH2O and to polymerization of CH2O, yielding paraformaldehyde. The latter is bound to the 5-fold coordinated surface Ti atoms and is found to decompose and release CH2O at about 270 K. On the defective TiO2(110) surface, CH2O adsorbs more strongly on oxygen vacancy sites, ultimately forming a diolate (-OCH2CH2O-) species, as demonstrated by HREELS. The assignment of the vibrational frequencies was aided by theoretical calculations on the DFT-B3LYP level. Upon heating to higher temperatures, this species undergoes deoxygenation, resulting in ethylene formation. 1. Introduction Chemical reactions occurring on surfaces of titanium oxide have been studied in great detail, and many specific information regarding atomic arrangement, electronic states, surface relaxation, and point defects such as “oxygen vacancies” are now available.1–3 The most studied surface of rutile TiO2 is the (110) surface that contains alternating rows of 5-fold coordinated Ti atoms (Ti5c) and 2-fold coordinated bridging O atoms (O2c). This surface is generally regarded as a prototype model for rutile TiO2. This surface however, no matter how carefully prepared, contains oxygen vacancies in ultrahigh vacuum (UHV), with their density depending on the surface annealing temperature and oxygen partial pressure.4–6 A large fraction of these oxygen vacancies can be healed by water dissociation yielding two surface hydroxyls per one surface oxygen vacancy.3,7,8 Many catalytic reactions rely on the dynamic equilibrium between the addition and removal of oxygen atoms on the surface. For example, CO oxidation in a Mars Van Krevelen type on RuO2(110) surfaces9–11 as well as oxidative dehydrogenation of alcohols12 and hydrocarbons13 all rely on the dynamics between surface oxygen vacancies and gas phase oxygen pressure. In surface science, for most of these oxidation/ reduction processes the decomposition (dissociation) pathways are studied where the number of moles consumed is equal or less to the number of moles produced. This preference is in part due to what it generally referred to as the pressure gap; simply stated as high-vacuum favoring dissociation and highpressure favoring association. However, some counter examples are persistent in surface reactions where surface-adsorbate interactions lead to building up of large molecules via carbon-carbon, carbon-oxygen, and carbon-nitrogen bond formation. These coupling reactions are among the most essential chemical reactions in nature. In UHV, carbon-carbon bond formation was reported on O-defected TiO214–19 and UO2 * Corresponding author phone: +49(0)2343224217; e-mail: wang@ pc.rub.de. † On leave from the Department of Chemistry, The University of Auckland, Private Bag 92019, Auckland, New Zealand.

single crystals.20–23 Oxidative coupling of acetylene to furan was also detected on a Pd(111) single crystal surface.24 Organometallic and organic chemists recognized early on that Ti compounds are active for coupling reactions of carbonyl compounds via carbon-carbon bond formation. During this catalytic process Ti atoms are oxidized, ultimately to Ti4+, a reaction termed the McMurry reaction.25,26 Orbital symmetry is given as the reason for the high activity of the surface Ti species, although detailed studies are not reported. It is also worth mentioning that this reaction is not observed on many other transition metals, and of all the transition metals investigated, Ti is the most active. The reductive coupling of carbonyls on reduced TiO2(001) single crystals was studied by temperature-programmed desorption (TPD) and X-ray photoelectron spectroscopy (XPS).15–20 The results indicated that a diolate type species (-OCH2CH2O-) is most likely formed as an intermediate between the carbonyl (R(R′)CdO) and the symmetric hydrocarbon (R(R′)CdC(R′)R). This conclusion was based on the appearance of an XPS C1s signal at about 286.5 eV, a value which is found between the C1s of hydrocarbons (ca. 285 eV) and that of carbonyls (ca. 288 eV). However, photoelectron spectroscopy provides indirect evidence of the surface species formed during the reaction. No unambiguous evidence for the formation of a diolate species on TiO2 single crystal surfaces has been reported. High resolution electron energy loss spectroscopy (HREELS) has been extensively used to characterize adsorbed species on metal single crystal surfaces (see, e.g., refs 27–32). In contrast to metals, the application of this technique to oxide surfaces is rather scarce. This lack of information is to a large extent due to the intense Fuchs-Kliewer phonon losses33 that make the adsorbate-related losses very difficult to detect. Two methods to overcome this problem have been proposed in the literature: Fourier deconvolution of combination losses34,35 and collection of spectra under conditions where impact scattering is enhanced.36 More recently, using a combination of both methods, high-quality HREELS data have been reported for different adsorbates on ZnO and TiO2 surfaces.37–43 The successful application of HREELS to oxides and the advanced understanding of the surface structure of TiO2(110)

10.1021/jp801327b CCC: $40.75  2008 American Chemical Society Published on Web 06/10/2008

Carbon-Carbon Bond Formation on Titanium Oxide rutile single crystal allow for a detailed investigation of one of the most fundamental and essential reactions in chemistry: the carbon-carbon bond formation reaction. Association of molecules to higher ones is not only central to industrial catalysis and chemical manufacture in general but also at the origin of any process in nature where complex compounds are made from simpler monomers or precursors. In this work we focus on the reaction of formaldehyde, the simplest of the carbonyls, on the surfaces of stoichiometric and reduced rutile TiO2(110), using HREELS and thermal desorption spectroscopy (TDS). It is shown that formaldehyde can form ethylene by reductive coupling (2CH2O + 2VO f CH2dCH2 + 2Os), where VO denotes an O vacancy and s for surface. This reaction, although the simplest of all reductive carbonyl coupling reactions, is complex and is not understood on an elementary level. A key intermediate between the adsorbed formaldehyde and the desorbing ethylene is a diolate species (-OCH2CH2O-) bonded to Ti atoms. This stable intermediate species has not been unambiguously identified on any metal oxide single crystal surface prior to this work. 2. Experimental Section All experiments were performed in a UHV system, which consists of two chambers. The upper chamber is equipped with a low energy electron diffraction (LEED) optic, an Ar ion sputter gun, and a quadrupole mass spectrometer (Pfeiffer, Prisma) for TDS experiments. The base vacuum in the upper chamber is 5 × 10-11 mbar. The mass spectrometer is placed in the main chamber and differentially pumped with an ion pump. When carrying out TDS experiments, the sample surface is moved to approach the orifice of a shroud of the mass spectrometer within a distance of about 1 mm. The lower chamber holds the HREEL spectrometer (Delta 0.5, SPECS, Germany) with straight-through resolution close to 1 meV (8.065 cm-1).44–47 The base vacuum in the lower chamber is 2 × 10-11 mbar. The TiO2 single crystal (2 mm × 5 mm × 7 mm) was mounted on a Ta plate (thickness, 0.5 mm) by specifically designed thin Ta foils, with the thermocouple contacting the sample on the lateral side. The sample temperature can be changed from 90 K (cooled by liquid nitrogen) to 900 K (heated by W filament from the backside of the Ta plate). The TiO2(110) surface was cleaned by many sputtering (1 keV, 2 × 10-5 mbar) and annealing (800 K) cycles until a sharp (1 × 1) LEED pattern was obtained. Oxidation at 800 K, in an ambient of 1 × 10-6 mbar of molecular oxygen, was also used between two sputtering and annealing cycles. For the clean surface, contamination was below the detection limits of HREELS. To create a defective surface, overannealing and argon ion sputtering methods were used. The overannealing temperature was fixed at 900 K. For the Ar ion sputtering method, the energy was fixed at 600 eV. In the course of this work, different ion fluxes were used to modify the resulting concentration of O vacancies on the TiO2(110) surface. CH2O was produced by heating N-(hydroxymethyl)-benzamide (also known as benzamidomethanol, C6H5CONHCH2OH), in a glass tube connected to the UHV apparatus, to 90-95 °C for 3 min. Prior to the experiment, the compound was thoroughly degassed by overnight pumping while heated to 60-70 °C.48 Upon heating to 90 °C, CH2O molecules are released in the gas phase. It was opted to use this precursor rather than paraformaldehyde because it has been found to be easier to avoid a coexposure to water. Mass spectrometry was routinely used to check the compositions of the gas resulting from the benzamide decomposition, no evidence of products above m/e 30 was

J. Phys. Chem. C, Vol. 112, No. 26, 2008 9829

Figure 1. Stick-and-ball model of TiO2(110) in perspective view together with surface oxygen vacancies and reaction intermediate, diolate species formed after formaldehyde activation on two adjacent bridge O-vacancies. (Ti: bright gray ball, O: red ball).

found. In mass spectra, CH2O yields three intense peaks at m/e 28 (CdO), 29 (HCdO), and 30 (H2CdO). In the experiment described below, just mass 29, which is the most intense peak, was monitored to detect desorption of CH2O. Dosing of CH2O was executed by backfilling the preparation chamber through a leak valve. Exposures are given in units of Langmuir (L) (1 L ) 1.33 × 10-6 mbar s). The Fourier deconvolution process used to deconvolute the EELS data was similar to that reported in ref 34. In the present work, the elastic peak was always used as the response function, since this method was found to be superior to any simulation (Gaussian or Lorentzian) of the response function. Quantum chemical vibrational frequency calculations were conducted for (OCH2CH2O)-TiCl2 (cluster 1) and TiCl2-OCH2-CH2O-TiCl2 (cluster 2). A previous work has specifically studied the coupling of formaldehyde to diolate over these clusters at the DFT/B3LYP level with a 6-31G* basis set,26 but in that work no frequencies were calculated. Cluster 1 has the two oxygen atoms attached to the same Ti atom, whereas cluster 2 has the two oxygen atoms connected to separate Ti atoms. Geometry optimizations were conducted up to a max gradient of less than 8 × 10-6 using DFT/B3LYP with the following Gaussian basis sets; 6-31G**, 6-31+G*, 6-311+G**, and 6-311+G**. As is the case for many organic compounds, the 6-31G** was found to be adequate. All electrons of Ti atoms were fully treated without the use of a pseudopotential. Calculations were conducted using the Spartan 04 and Spartan 06 code. Gas phase ethylene glycol and glycolate dianions were also computed for comparison using the same method. 3. Results and Discussion Figure 1 shows a stick-and-ball model of the rutile TiO2(110) surface structure in a perspective view; the surface consists of 2-fold coordinated bridging O atoms (O2c), 3-fold coordinated lattice O atoms, and 5-fold coordinated Ti atoms (Ti5c). When the surface is heated to high temperatures, a number of bridging O atoms desorb (depending on the annealing temperature)4,6 from the surface, thus creating O vacancies. Each O vacancy is stoichiometrically associated with the reduction of two Ti4+ cations to Ti3+ cations. In this work, additionally, an alternative method to reduce the surfaces, namely argon ion sputtering, was employed. 3.1. CH2O Adsorption on a Perfect TiO2(110) Surface. Figure 2 presents TD spectra recorded after exposing the defectfree TiO2(110) surface to various amounts of CH2O at 100 K. With increasing exposure to CH2O, three peaks located at 290,

9830 J. Phys. Chem. C, Vol. 112, No. 26, 2008

Qiu et al. TABLE 1: Vibrational Energies (meV) and Mode Assignments of CH2O Adsorbed on Fully Oxidized TiO2(110) Surfaces at 100 K assignment paraformaldehyde physisorbed formaldehyde

a

Figure 2. TD spectra of CH2O adsorbed on the perfect TiO2(110) surface at 100 K with various exposures. The heating rate was 1.5 K/s. 1 L ) 1.33 × 10-6 mbar s. The inset shows the relative coverage of CH2O as function of exposure.

Figure 3. HREEL spectra of (a, b) clean TiO2(110) surface, (c) 5 L CH2O adsorbed on the perfect TiO2(110) surface at 100 K. After curve c, the sample was subsequently annealed to (d) 200 K, (e) 260 K, and (f) 400 K. Curve a is the raw spectrum, and the Fourier deconvoluted spectra are shown in curves b-f. In curves c and d, the surface phonon at 189 meV is not completely removed. All the spectra were taken at 100 K in specular direction, with an incidence angle of 55° and with a primary energy of 10 eV.

268, and 128 K are observed for m/e 29 (the parent ion of CH2O). The relative coverage of CH2O was determined from the ratio of the integrated peak areas (see inset of Figure 2). The following masses were also investigated but no signals were detected: m/e 2, 18, 26, 27, 44. A small amount of m/e 31 (CH3OH) was observed, but its intensity was less than 0.6% of that of m/e 29 and, as a result, was not further considered. At low coverage only one formaldehyde desorption peak at 290 K is seen; increasing surface exposure results in an increase of the total desorption at the same temperature. At higher exposures (2 L and above) a peak at 268 K develops and becomes dominant at 5 L. The peak at 128 K observed at high coverage is assigned to physisorbed CH2O. It is also worth noting that for the 5 L exposure the main desorption peak becomes very broad between 150 and 350 K. This hints to a nonuniform distribution of CH2O species on the TiO2(110) surface, which will be discussed later. In Figure 3, HREEL spectra are presented that were recorded for the stoichiometric TiO2(110) surface before and after adsorption of formaldehyde at 100 K. The raw spectrum of the

modes

T ) 100 K

T ) 200 K

ν(C-O) δ(CH2) ν(CH2) ω(CH2) F(CH2) δ(CH2) ν(CdO) ν(CH2)

137 177 363 143 156 185a 210 363

137 174 363

Not resolved from the surface phonon.

clean TiO2(110) surface (curve a) shows intense primary surface phonon bands at 45, 54, and 94 meV, as well as their combination and multiple excitations. The latter could be completely removed by Fourier deconvolution (curve b), and the corresponding spectrum demonstrates the presence of a clean, contamination-free TiO2(110) surface (Note that in curves c, d, and f the surface phonon at 189 meV is still detected as a weak feature). After exposure of 5 L CH2O at 100 K, new loss peaks are detected at 137, 143 (shoulder), 156, 174, 185, 210, and 363 meV (Figure 3c). By comparison with IR results of free CH2O molecules,49 the loss features at 143, 156, 210, and 363 meV are assigned to wagging ω(CH2), rocking F(CH2), stretching ν(CdO), and asymmetric stretching νa(CH2) modes of physisorbed CH2O, respectively. This assignment is confirmed by the sequential annealing experiment shown in Figure 3d, where all the physisorbed CH2O-related bands are found to disappear upon heating the sample to 200 K. This observation is consistent with the assignment of the low temperature TDS peak at 128 K to physisorbed species (Figure 2). Accordingly, the remaining loss features at 137, 174, and 363 meV in Figure 3d must be related to chemisorbed CH2O. As was shown previously by many groups,50,51 CH2O adsorbed on metal or metal oxide surfaces at low temperature often polymerizes to yield paraformaldehyde. In our HREELS data, the loss peak at 137 meV is a typical feature of the O-C-O units in paraformaldehyde and corresponds to the C-O stretching vibration. A chemisorbed CH2O monomer is generally bonded to the oxide surface in an oxygen end-on η1(O) configuration,50,52 giving a typical ν(CdO) mode at around 205 meV due to the slightly reduced CdO bond. On the basis of our vibrational data, the presence of isolated chemisorbed CH2O molecules on TiO2(110) can be excluded. After further annealing to 260 K (Figure 3e) no new peaks show up, and the losses at 137, 174 and 363 meV decrease in intensity. The two desorption states at 268 and 290 K in TDS are therefore assigned to the decomposition of the same surface species, paraformaldehyde. All HREELS data and the corresponding assignments are summarized in Table 1. To further assert the adsorption site of paraformaldehyde, coadsorption experiments of CH2O and CO were conducted (Figure 4). It is known that CO is adsorbed on 5-fold coordinated Ti5c cations.42,53 Thus, if CH2O also adsorbs on the metal ions, then competitive adsorption will take place, which is expected to follow thermodynamic laws; the species with the stronger adsorption energy will prevail. Figure 4 presents the TDS data obtained after exposing the TiO2(110) surface first to 2 L of CH2O and then to 2 L CO at 100 K. No CO desorption is found in the TD spectrum, confirming that the Ti5c sites are occupied by CH2O molecules, thus preventing the adsorption of CO. The inset of Figure 4 shows a TD spectrum of 2 L CO adsorbed on

Carbon-Carbon Bond Formation on Titanium Oxide

Figure 4. TD spectra recorded after coadsorption of CO and CH2O on the perfect TiO2(110) surface: 3 L CH2O on TiO2(110) at 100 K followed by 2 L CO at 100 K. The inset shows TD spectrum of 2 L CO on defect-free TiO2(110) at 100 K.

Figure 5. TD spectra of CH3OH on a perfect TiO2(110) surface after (a) only CH2O adsorption and (b) coadsorption of CH2O and 1000 L atomic hydrogen at 120 K. The solid line shows the guide for eyes.

the clean TiO2(110) surface at 100 K. A desorption peak at 130 K is detected and corresponds to a CO binding energy of ∼33 kJ/mol (obtained by applying a standard Redhead-analysis with a frequency factor of 1 × 1013 s-1), in good agreement with previous results on fully oxidized TiO2(110).42,54 As expected, similar results are observed if the order of adsorption is reversed, indicating that all O atoms in the paraformaldehyde oligomer chain bind to Ti5c sites. It should be noted that there is a mismatch between the large Ti-Ti separation in TiO2 (2.96 Å along the [001] direction) and the CH2O unit length in (CH2O)n of 2.35 Å.55 As a result, many of the interior oxygen atoms in the polymer chain must be bound to the surface via electrostatic interactions. The appearance of two desorption phases (268 and 290 K) and the broadening of the TDS peak are likely to originate from the paraformaldehyde oligomer chains with different lengths and/or different modes of interaction. A coadsorption of CH2O and atomic hydrogen was also carried out on the perfect TiO2(110) surface at 100 K. After 2 L CH2O adsorption on a TiO2(110) surface at 100 K, the modified surface was further exposed to 1000 L atomic hydrogen at 100 K, performed by dissociating H2 on a hot tungsten filament in line-of-sight from the substrate as described in previous works.42,56 The resulting TDS data (desorbed species, peak shapes, and desorption temperatures) were found to be similar to those recorded after exposure to 2 L CH2O at 100 K. The only significant difference is the observation of a small amount of CH3OH (see Figure 5), which reveals the formation of methanol through the hydrogenation of formaldehyde. 3.2. CH2O Adsorption on Defective TiO2(110) Surfaces. 3.2.1. CH2O Adsorption on OWer-Annealed TiO2(110) Surface. As was mentioned above, bridge O vacancies can be created by heating the sample to high temperatures.4–6 The increase in

J. Phys. Chem. C, Vol. 112, No. 26, 2008 9831

Figure 6. TDS data of 5 L CH2O adsorbed on overannealed TiO2(110) surfaces at 100 K. (a) CH2O; (b, c) C2H4. The sample was overannealed (a, b) at 900 K for 5 min and (c) at 900 K for 10 min.

vacancy density is not associated with a considerable rearrangement of the surface, as indicated by results from both STM and LEED.1 Dosing of 5 L CH2O to a TiO2(110) surface that has been annealed at 900 K for 5 min yields significantly different TDS data. The results presented in Figure 6b show a new feature in the high temperature region. A peak of mass 26 is observed at 620 K and corresponds to the desorption of C2H4. TDS results recorded after longer annealing time (900 K, 10 min) only reveal a small increase in the intensity of this peak, accompanied by a slight temperature shift (see Figure 6c). The same desorption species (C2H4) was also observed in previous work,19 and the yield of C2H4 as a reduction product was found to be proportional to the concentration of surface O vacancies. In Figure 6c, the ratio of ethylene to formaldehyde is found to be equal to 3.7%. Correction factors for mass spectrometer sensitivity have been considered. Assuming full saturation of all Ti5c sites with formaldehyde and that the formation of one ethylene molecule leads to two oxygen atoms filling O vacancies, we calculated an initial O-defect concentration of 7.4% (or ≈3.9 × 1013 O vacancy/cm2). 3.2.2. CH2O Adsorption on a Sputtered TiO2(110) Surface. Figure 7 displays TD spectra recorded after exposing a series of Ar-ion sputtered TiO2(110) surfaces to CH2O at room temperature. Besides a very small amount of molecular desorption of CH2O (curve a), desorption of ethylene at about 550 K is clearly observed (curve b), even at the weakest sputtering condition used (4.0 µA min). As shown in the inset of Figure 7B, increasing the sputtering intensity results in considerable increase of the ethylene desorption (formation), reaching a maximum value of 19% at about 40 µA min. Molecular desorption of CH2O from the same surface, in contrast, is found to clearly decrease with increasing sputtering intensity. The corresponding maximum concentration of oxygen vacancies created by Ar+ sputtering is computed to be equal to 38%. Compared with the results on the fully oxidized (perfect) and overannealed TiO2(110) surfaces (Figure 8), the yield of ethylene dramatically increases for the sputtered surface (note that for the perfect TiO2(110) surface no ethylene desorption is detected), whereas the molecular desorption of CH2O is largely reduced. These observations strongly indicate that the deoxygenation reaction of CH2O occurs on surface oxygen vacancies. This finding is further supported by the coadsorption experiments with H2O. Water adsorption on the TiO2(110) surface has been intensively studied, and it was shown that the adsorption on O vacancy sites at room temperature causes a dissociation of H2O molecules, forming two OH species (and the filling of the O vacancies).57,58 We have conducted TDS experiments of formaldehyde (2 L) on an Ar-ion sputtered TiO2(110) surface that

9832 J. Phys. Chem. C, Vol. 112, No. 26, 2008

Qiu et al.

Figure 9. HREELS data recorded after exposing differently modified TiO2(110) surfaces to 5 L CH2O at 300 K, followed by annealing to 400 K. (a) The fully oxidized (perfect) surface; (b) Ar-ion sputtered TiO2(110) surface. Both HREELS experiments were recorded at room temperature. Curve (b) was given as a raw spectrum. All the spectra were taken at 300 K in specular direction with an incidence angle of 55° and with a primary energy of 10 eV. Figure 7. TDS data recorded after exposing a sputtered TiO2(11) surface to 5 L CH2O at 300 K with various sputtering parameters (indicated in the lower part): (A) CH2O; (B) C2H4. The energy of the argon ions was fixed at 600 eV, and the sputtering parameters were given in the product between sputtering ion current and sputtering time (µA min). The inset shows the integrated areas of CH2O and C2H4 as function of sputtering intensity.

TABLE 2: Assignments of HREELS Data for the Diolate Molecule Formed on Defective TiO2(110) Surfaces and Comparison with Other Works from Ethylene Glycol and the Computed (at the DFT/B3LYP-6-31G**) Frequencies from a Ti2Cl4C2H3O2 Clustera

modes νas(C-O) νs(C-O) ν(C-C) ν(C-C)+ω(CH2) δ(CH2) νs(CH2) νas(CH2) a

Figure 8. TDS results obtained after exposing differently modified TiO2(110) surfaces to 5 L CH2O at 100 K: (a) fully oxidized (perfect) TiO2(110); (b) prior-overannealed to 900 K; (c) prior-sputtered TiO2(110); and (d) the sputtered surface was exposed first to 2 L H2O at 300 K and then to 2 L CH2O at 300 K. Panel A: CH2O; panel B: C2H4.

has been pre-exposed to 2 L of water at room temperature. No desorption of ethylene was observed (see Figure 8d), as expected from the above reaction mechanism requiring the presence of O vacancies. To elucidate the mechanism of formaldehyde activation at O vacancies, we have carried out HREELS experiments. Figure 9 presents a comparison of vibrational data recorded after CH2O adsorption on perfect and defective surfaces. After exposing the perfect TiO2(110) surface to 5 L CH2O at 300 K, followed by annealing to 400 K, the corresponding HREEL spectrum only shows the features of a clean surface; clearly, CH2O has completely desorbed at this temperature, in good agreement with

ethylene glycol O-CH2-CH2-O HREELS computed (liquid)59 on Mo(110)60 data results 129 135 107 181 356 364

128 111 179 352 361

138 143 b

174 180 353 363

137 145 114 174 180-186 378 385

All numbers are given in meV. b Not resolved.

the TDS results. However, the HREEL spectrum (Figure 9b) obtained on the defective TiO2(110) surface prepared by Ar+ sputtering exhibits a number of distinct loss features, revealing the presence of chemisorbed species formed via CH2O adsorption on O vacancy sites. The fact that no ν(CdO) mode (at around 210 meV) was observed hints toward a dissociative adsorption of CH2O on O vacancies. On the basis of the similarity of the frequencies in the HREELS data with those of ethylene glycol, both in liquid59 and adsorbed on Mo(110)60 (see Table 2), we propose the presence of a diolate species (-OCH2CH2O-). This assignment is clearly supported by DFT calculations. As shown in Table 2, the computed frequencies of diolate in a TiCl2-OCH2CH2O-TiCl2 cluster are in good agreement with the experimental results. The losses observed at 138, 143, 180, 353, and 363 meV are assigned to νas(-O-CH2-H2C-O-), νs(-O-CH2-H2C-O-), δ(CH2), νs(CH2), and νas(CH2) modes, respectively. The peak at 174 meV is attributed to a vibration of mixed character involving C-C stretching and CH2 wagging. The diolate species can be formed through activation of two formaldehyde molecules adsorbed at adjacent O vacancy sites (see Figure 1). Actually, this type of double O vacancy has been observed on a TiO2(011)-2 × 1 surface prepared by electron irradiation in a recent STM study.61 In our work, the oxygen vacancies on TiO2(110) are created mainly by Ar+ sputtering, and a large vacancy concentration of about 38% of a monolayer is achieved. More recently, the diffusion barrier (Eb) of bridge

Carbon-Carbon Bond Formation on Titanium Oxide SCHEME 1

O vacancies on TiO2(110) was determined by both STM and DFT calculations62 to be equal to 1.15 eV, revealing that oxygen vacancies exhibit rather limited mobility at room temperature. The corresponding hopping rate (h) can be calculated according to h ) V exp(-Eb/kbT), where V is the preexponential factor (1.58 × 1012 s-1),62 and kb is the Boltzmann constant. On the basis of the given data, a hopping rate of bridge oxygen vacancies is estimated to be as low as ∼7 × 10-8 s-1 at 300 K. This leads to a nonuniform distribution of O vacancies created by Ar+ sputtering at 300 K, revealing the presence of double or multiple O vacancies on the sputtered TiO2(110) surface. We propose that, upon annealing, the diolate species undergoes deoxygenation, leading to the formation of C2H4 as the final product. Two oxygen atoms of the diolate species are left on the surface to fill up two O vacancies. In addition, it should be noted that the reductive reaction of CH2O to C2H4 on the defective TiO2(110) surface could be complicated and other reaction mechanisms can also be involved. In particular, the TiO2(110) surface prepared by overannealing should be dominated by the single O vacancies due to the repulsive interaction between vacancies.62 In that regard it is worth discussing the mechanism invoked for the reductive coupling of carbonyl reactions in general. Organometallic chemists have studied this reaction in details. This reaction is also known as the McMurry reaction,25 although several variants of it have been reported, in particular, by the Lippard group.63 On the basis of the crystal structure and spectroscopic and kinetic methods, two following mechanisms have been identified (Scheme 1), where [M] is a zerovalent metal. The net reaction involves the loss of four electrons from the surface (or the compound) reducing the carbonyl compound to the symmetric olefin. The radical species has been observed in numerous works.25,64 It has to be stressed that these two mechanisms exist depending on the nature of the metal and the reaction conditions. In the case of M ) Ti, mechanism 1 predominates, whereas when M ) W or U, mechanism 2 is observed. It is clear that both mechanisms require heavily reduced metals and strong π-d orbital interactions. Our present work indicates that the reductive coupling on the heavily reduced TiO2(110) surface is efficient. On the basis of our spectroscopic results, this reaction most likely involves diolate type species as the intermediate (mechanism 1). Now the question arises why the overannealed surface still shows some reductive coupling activity (albeit quite small). Since the concentration of two nearest neighbor oxygen point defects would be far too small to account for the reaction yield of about 3.7%, one has to invoke a diffusion of a surface intermediate species from one defect center to the other. This process is presented in Scheme 2, with details as follows: A. Oxygen point defects are further created by overannealing TiO2(110) to 900 K (represented by an open square). These defects have been reported to involve two Ti atoms, each with a +3 oxidation state. These two Ti atoms are, however, not equivalent; one is along the [001] direction (formerly 6-fold

J. Phys. Chem. C, Vol. 112, No. 26, 2008 9833 SCHEME 2: Formation of Diolate Species on the Over-annealed TiO2(110) Surface

coordinated to oxygen atoms) and the other is a Ti 5-fold coordinated along the [110] direction.65 B. Formaldehyde preferentially adsorbs on the oxygen defects along the [001] direction. The adsorption of formaldehyde leads to the possible formation of a radical species as previously identified by EPR over Ti complexes for the same reaction.64 C. The reaction of the initially adsorbed formaldehyde with an adjacent bridging oxygen atoms leads to the formation of a dioxy-methylene species (-OCH2O-). These species have been observed from formaldehyde over several metal oxides by FTIR66 and are postulated as a reaction intermediate for the coupling of formaldehyde to ethylene over the reduced TiO2(110) surface.54 D. The diffusion of these species is facilitated at higher temperatures and would involve the breaking and making of one carbon-oxygen bond. E. The reaction of two adjacent species would result in the formation of a diolate species. F. Upon farther heating, breaking of two carbon-oxygen bonds, facilitated by the energy gained to oxidize the surface, would lead to the formation of ethylene that instantaneously desorbs from the surface. Finally, an additional weak desorption peak of ethylene is observed in the TDS data at around 160 K (see Figure 8), indicating that the reductive reaction of CH2O to C2H4 also occurs in the low temperature region. This finding may indicate the existence of more active defect sites on TiO2(110) such as lattice O vacancies created by Ar+ sputtering. 4. Conclusions In summary, formaldehyde adsorption on perfect and defective TiO2(110) surfaces has been studied by TDS and HREELS.

9834 J. Phys. Chem. C, Vol. 112, No. 26, 2008 On the perfect TiO2(110) surface, it was found that adsorption of CH2O at 100 K leads to physisorbed CH2O and to polymerization of CH2O, yielding paraformaldehyde. The latter is bound to the surface via O-Ti bonds at the 5-fold coordinated surface Ti5c ions. Upon heating, the paraformaldehyde oligomer chain decomposes and releases CH2O. On defective surfaces, created by both overannealing or Ar+ sputtering, formaldehyde is more strongly adsorbed at O vacancy sites. For the sputtered surface a reaction intermediate species, diolate (-OCH2CH2O-), is clearly identified by HREELS. The observed vibrational frequencies are in good agreement with those obtained by DFT calculations. Upon heating, a reductive reaction takes place to form ethylene, which desorbs from the surface. Further studies for other organic molecules on TiO2(110) are needed in order to elucidate whether similar carbon-carbon coupling reactions forming olefins take place at the reduced sites, surface oxide vacancies. In addition, our results demonstrated that HREELS is not only a powerful method to study adsorption and reaction of molecules at perfect oxide surfaces but also can be used to characterize the interaction of adsorbates with defect sites. Acknowledgment. This work was funded by the German Research Foundation (DFG) within SFB 558 “Metal-Substrate Interactions in Heterogeneous Catalysis”. H.Q. thanks the IMPRS of SurMat for a research grant. References and Notes (1) Diebold, U. Surf.Sci. Rep. 2003, 48, 53. (2) Idriss, H.; Barteau, M. A. AdV. Catal. 2000, 45, 261. (3) Bikondoa, O.; Pang, C. L.; Ithnin, R.; Muryn, C. A.; Onishi, H.; Thornton, G. Nat. Mater. 2006, 5, 189. (4) Epling, W. S.; Peden, C. H. F.; Henderson, M. A.; Diebold, U. Surf. Sci. 1998, 412/413, 333. (5) Henderson, M. A.; Epling, W. S.; Perkins, C. L.; Peden, C. H. F.; Diebold, U. J. Phys. Chem. B 1999, 103, 5328. (6) Henderson, M. A.; Epling, W. S.; Peden, C. H. F.; Perkins, C. L. J. Phys. Chem. B 2003, 107, 534. (7) Brookes, I. M.; Muryn, C. A.; Thornton, G. Phys. ReV. Lett. 2001, 87, 266103. (8) Maksymovych, P.; Mezhenny, S.; Yates, J. T., Jr. Chem. Phys. Lett. 2003, 382, 270. (9) Wendt, S.; Seitsonen, A. P.; Kim, Y. D.; Knapp, M.; Idriss, H.; Over, H. Surf. Sci. 2002, 505, 137. (10) Fan, C. Y.; Wang, J.; Jacobi, K.; Ertl, G. J. Chem. Phys. 2001, 114, 10058. (11) Wang, J.; Fan, C. Y.; Jacobi, K.; Ertl, G. J. Phys. Chem. B 2002, 106, 3422. (12) Idriss, H.; Seebauer, E. G. Catal. Lett. 2000, 66, 139. (13) Feyel, S.; Schroder, D.; Schwarz, H. J. Phys. Chem. A 2006, 110, 2647. (14) Idriss, H.; Pierce, K.; Barteau, M. A. J. Am. Chem. Soc. 1991, 113, 715. (15) Idriss, H.; Libby, M.; Barteau, M. A. Catal. Lett. 1992, 15, 13. (16) Idriss, H.; Kim, K. S.; Barteau, M. A. J. Catal. 1993, 139, 119. (17) Idriss, H.; Barteau, M. A. Stud. Surf. Sci. Catal. 1993, 78, 463. (18) Idriss, H.; Pierce, K.; Barteau, M. A. J. Am. Chem. Soc. 1994, 116, 3063. (19) Lu, G.; Linsebigler, A.; Yates, J. T. J. Phys. Chem. 1994, 98, 11733. (20) Chong, S. V.; Idriss, H. J. Vac. Sci. Technol. A 2000, 18, 1900. (21) Chong, S. V.; Idriss, H. J. Vac. Sci. Technol. A 2001, 19, 1933. (22) Chong, S. V.; Idriss, H. Surf. Sci. 2002, 504, 145. (23) Senanayake, S. D.; Chong, S. V.; Idriss, H. Catal. Today 2003, 85, 311.

Qiu et al. (24) Ormerod, R. M.; Lambert, R. M. Catal. Lett. 1990, 6, 121. (25) McMurry, J. E. Chem. ReV. 1989, 89, 1513. (26) Stahl, M.; Pidun, U.; Frenking, G. Angew. Chem., Int. Ed. 1997, 36, 2234. (27) Jacobi, K.; Bedu¨rftig, K.; Wang, Y.; Ertl, G. Surf. Sci. 2001, 472, 9. (28) Wang, Y.; Lafosse, A.; Jacobi, K. 2002, 507, 773. (29) Badescu, S. C.; Salo, P.; Ala-Nissila, T.; Ying, S. C.; Jacobi, K.; Wang, Y.; Bedu¨rftig, K.; Ertl, G. Phys. ReV. Lett. 2002, 88, 136101. (30) Wang, Y.; Jacobi, K. Surf. Sci. 2002, 513, 83. (31) Badescu, S. C.; Jacobi, K.; Wang, Y.; Bedu¨rftig, K.; Ertl, G.; Salo, P.; Ala-Nissila, T.; Ying, S. C. Phys. ReV. B 2003, 68, 205401. (32) Wang, Y.; Jacobi, K. J. Phys. Chem. B 2004, 108, 14726. (33) Fuchs, R.; Kliewer, K. L. Phys. ReV. 1965, 140, A2076. (34) Cox, P. A.; Williams, A. A. Surf. Sci. 1985, 152/153, 791. (35) Crook, S.; Dhariwal, H.; Thornton, G. Surf. Sci. 1997, 382, 19. (36) Wu, M. C.; Estrada, C. A.; Goodman, D. W. Phys. ReV. Lett. 1991, 67, 2910. (37) Wang, Y.; Meyer, B.; Yin, X.; Kunat, M.; Langenberg, D.; Traeger, F.; Birkner, A.; Wo¨ll, Ch. Phys. ReV. Lett. 2005, 95, 266104. (38) Wang, Y.; Muhler, M.; Wo¨ll, Ch. Phys. Chem. Chem. Phys. 2006, 8, 1521. (39) Wang, Y.; Kova´cˇik, R.; Meyer, B.; Kotsos, K.; Stodt, D.; Staemmler, V.; Qiu, H.; Traeger, F.; Langenberg, D.; Muhler, M.; Wo¨ll, Ch. Angew. Chem, Int. Ed. 2007, 46, 5624. (40) Wang, Y.; Xia, X.; Urban, A.; Qiu, H.; Strunk, J.; Meyer, B.; Muhler, M.; Wo¨ll, Ch. Angew. Chem., Int. Ed. 2007, 46, 7315. (41) Wo¨ll, Ch. Prog. Surf. Sci. 2007, 82, 55. (42) Yin, X.-L.; Calatayud, M.; Qiu, H.; Wang, Y.; Birkner, A.; Minot, C.; Wo¨ll, Ch. ChemPhysChem 2008, 9, 253. (43) Wang, Y. Z. Phys. Chem. 2008, 222 (5–6). (44) Wang, Y.; Jacobi, K.; Ertl, G. J. Phys. Chem. B 2003, 107, 13918. (45) Paulus, U. A.; Wang, Y.; Bonzel, H. P.; Jacobi, K.; Ertl, G. J. Phys. Chem. B 2005, 109, 2139. (46) Wang, Y.; Jacobi, K.; Scho¨ne, W.-D.; Ertl, G. J. Phys. Chem. B 2005, 109, 7883. (47) Jacobi, K.; Wang, Y.; Ertl, G. J. Phys. Chem. B 2006, 110, 6115. (48) Van Veen, A. C.; Zanthoff, H. W.; Hinrichsen, O.; Muhler, M. J. Vac. Sci. Technol. A 2001, 19 (2), 651. (49) Nakanaga, T.; Kondo, S.; Saeki, S. J. Chem. Phys. 1982, 76, 3860. (50) Truong, C. M.; Wu, M.-C.; Goodman, D. W. J. Am. Chem. Soc. 1993, 115, 3647. (51) Bryden, T. R.; Garrett, S. J. J. Phys. Chem. B 1999, 103, 10481. (52) Idriss, H.; Kim, K. S.; Barteau, M. A. Surf. Sci. 1992, 262, 113. (53) Sorescu, D. C.; Yates, J. T. J. Phys. Chem. B 1998, 102, 4556. (54) Linsebigler, A.; Lu, G. Q.; Yates, J. T. J. Phys. Chem. 1996, 100, 6631. (55) Abe, A.; Mark, J. E. J. Am. Chem. Soc. 1976, 98, 6468. (56) Kunat, M.; Burghaus, U.; Wo¨ll, Ch. Phys. Chem. Chem. Phys. 2004, 6, 4203. (57) Schaub, R.; Thostrup, P.; Lopez, N.; Lægsgaard, E.; Stensgaard, I.; Nørskov, J. K.; Besenbacher, F. Phys. ReV. Lett. 2001, 87, 266104. (58) Brookes, I. M.; Muryn, C. A.; Thornton, G. Phys. ReV. Lett. 2001, 87, 266103. (59) Brown, N. F.; Barteau, M. A. J. Phys. Chem. 1994, 98, 12737. (60) Queeney, K. T.; Arumainayagam, C. R.; Weldon, M. K.; Friend, C. M.; Blumberg, M. Q. J. Am. Chem. Soc. 1996, 118, 3896. (61) Dulub, O.; Batzill, M.; Solovev, S.; Loginova, E.; Alchagirov, A.; Madey, T. E.; Diebold, U. Science 2007, 317, 1052. (62) Zhang, Z.; Ge, Q.; Li, S.-C.; Kay, B. D.; White, J. M.; Dohna´lek, Z. Phys. ReV. Lett. 2007, 99, 126105. (63) Carnahan, E. M.; Protasiewicz, J. D.; Lippard, S. J. Acc. Chem. Res. 1993, 26, 90. (64) Villiers, C.; Ephritikine, M. Angew. Chem., Int. Ed. 1997, 36, 2380. (65) Di Valentin, C.; Pacchioni, G.; Selloni, A. Phys. ReV. Lett. 2006, 97, 166803. (66) Idriss, H.; Hindermann, J. P.; Kieffer, R.; Kiennemann, A.; Vallet, A.; Chauvin, C.; Lavalley, J. C.; Chaumette, P. J. J. Mol. Catal. 1987, 42, 205.

JP801327B