Combined Theoretical and Experimental Study on the Adsorption of

ARTICLE pubs.acs.org/JPCA

Combined Theoretical and Experimental Study on the Adsorption of Methanol on the ZnO(1010) Surface J. Kiss,† D. Langenberg,‡ D. Silber,‡ F. Traeger,‡ L. Jin,‡ H. Qiu,‡ Y. Wang,‡,§ B. Meyer,*,†,|| and Ch. W€oll*,‡,^ †

Lehrstuhl f€ur Theoretische Chemie, Ruhr-Universit€at Bochum, 44780 Bochum, Germany Lehrstuhl f€ur Physikalische Chemie I, Ruhr-Universit€at Bochum, 44780 Bochum, Germany § Lehrstuhl f€ur Technische Chemie, Ruhr-Universit€at Bochum, 44780 Bochum, Germany Interdisziplin€ares Zentrum f€ur Molekulare Materialien (ICMM) and Computer-Chemie-Centrum (CCC), Friedrich-Alexander-Universit€at Erlangen-N€urnberg, N€agelsbachstr. 25, 91052 Erlangen, Germany ^ Institut f€ur Funktionelle Grenzfl€achen (IFG), Karlsruher Institut f€ur Technologie (KIT), Kaiserstr. 12, 76131 Karlsruhe, Germany

)

‡

ABSTRACT: The structure, dynamics, and energetics of methanol adlayers on the nonpolar ZnO(1010) surface have been studied by He-atom diffraction (HAS), high-resolution electron energy loss spectroscopy (HREELS), thermal desorption spectroscopy (TDS), and density functional calculations. The experimental and theoretical data consistently show that at temperatures below 357 K methanol forms an ordered adlayer with a (2  1) periodicity and a coverage of one monolayer in which half of the methanol molecules are dissociated. The ordering of the methanol molecules is governed by repulsive interactions between the methyl groups of the methanol molecules. This repulsive interaction is also responsible for the formation of a second, low-density phase at higher temperatures with half monolayer coverage of undissociated methanol which is stable up to 440 K.

’ INTRODUCTION Over the past years, substantial progress has been achieved in understanding the adsorption of small molecules on oxide surfaces on the basis of detailed atomistic calculations and a subsequent validation of the theoretical results by comparison with experimental data obtained for model systems.1 Among the most studied examples are the surfaces of ZnO and TiO2 where for a number of small molecules (CO, CO2, H2O) a rather satisfying description could be obtained.1
10 OH-containing molecules are a particularly challenging case with regard to the structures and geometries formed after adsorption, because for these species the threshold for hydrogen abstraction has been shown to be fairly low on oxide surfaces. Accordingly, for most oxides it is rather the rule than the exception that water dissociates upon contact with the surface.11 Recently, for the case of H2O adsorbed on the nonpolar ZnO(1010) surface, it has been demonstrated that on this substrate a subtle interplay between hydrogen bonding, the formation of stable oxygen bonds to substrate Zn ions, and the formation of intermolecular hydrogen bonds is present. These different interactions lead to the formation of a rather complicated adsorbate structure consisting of an equal fraction of dissociated and intact molecules.3,12
14 Encouraged by the excellent agreement between ab initio calculations based on density functional theory (DFT) and experimental data obtained from He-atom scattering (HAS) and high-resolution electron energy loss spectroscopy (HREELS) for r 2011 American Chemical Society

the case of water adsorption on ZnO(1010), we have studied another, more complex adsorbate on this important surface of zinc oxide, namely, methanol. Methanol is expected to yield a much more complex situation than water since replacing one of the water hydrogen atoms by a methyl group will lead to substantially more demanding sterical requirements of the adsorbate. Also the vibrational dynamics of a methanol adsorbate is more complicated due to the larger number of atoms and the higher total mass. These considerations already suggest that a detailed investigation of this complex system should start with an analysis of the low-temperature phase where entropic contributions are minimal and where reactions going beyond OH bond scission can be largely excluded. In addition to providing a more challenging test case for the applicability and validity of theoretical calculations, methanol itself is a very important molecule. Methanol is a desired product in heterogeneous catalysis. The presently most important process is based on syngas (CO/CO2/H2), which is converted to methanol using ZnO powders as active catalyst15 or as support for catalytically active Cu particles. This technological interest Special Issue: J. Peter Toennies Festschrift Received: January 6, 2011 Revised: April 2, 2011 Published: April 19, 2011 7180

dx.doi.org/10.1021/jp200146v | J. Phys. Chem. A 2011, 115, 7180–7188

The Journal of Physical Chemistry A has triggered several studies in the past on the interaction of methanol with ZnO single-crystal surfaces16
21 and ZnO powders.22,23 The focus of these investigations, however, has been on the inverse process of methanol synthesis, methanol decomposition, and the reactivity of ZnO toward methanol, which requires surface defects and temperatures above 450 K. Only one recent study has focused on the low-temperature adsorption of methanol on the ideal parts of ZnO surfaces.24 Using scanning tunneling microscopy (STM) it was observed that at room temperature methanol dominantly forms a twodimensional island structure on the terraces of well-oxidized ZnO(1010) surfaces with the molecules residing on the surface Zn atom rows.24 The structure of the methanol islands could not be resolved, but occasionally a weak contrast with a (1  1) periodicity was obtained. In addition, a second ordered phase of adsorbed methanol molecules was found which points toward a more complicated adsorption behavior of methanol on ZnO(1010) compared to water. The second structure is a minority phase which covers only about 3% of the surface. It consists of linear chains of methanol molecules located between the Zn atom rows, and it is accompanied by line defects of missing ZnO dimers along the Zn rows which form after methanol desorption.24 Clearly, the STM results call for a thorough theoretical analysis of this complex adsorbate system, similar to the case of H2O adsorbed on ZnO(1010).3,12
14 In the present paper, we demonstrate by using DFT calculations in combination with a rather complete set of experimental data, HAS, HREELS, and thermal desorption spectroscopy (TDS), that we can provide a reliable description of the lowtemperature (below 450 K) methanol adsorbate phases with a very satisfying agreement between experiment and theory. The analysis of the theoretical results confirms the concern raised above, namely, that the system is substantially more complicated than the water adlayer and that the additional steric requirements brought about by the methyl group lead to rather different intermolecular interactions and, as a result, to a rather different patching of the molecular species within the adlayer. In this paper we focus exclusively on the low-temperature behavior. The complex phenomena occurring at temperatures above 450 K, where methanol decomposes at defect sites,16
23 will be the subject of a subsequent paper.

’ EXPERIMENT The He-atom scattering technique and apparatus have been described in detail in refs 25
28. Briefly, a differentially pumped supersonic molecular beam source is attached to a scattering chamber and to a drift tube. At the end of the drift tube a magnetic mass spectrometer for the detection of the scattered He atoms is mounted. Whereas the total scattering angle (sum of incident and exit angles) is fixed, the angle of incidence, Θi, can be changed by rotating the sample around an axis normal to the scattering plane. The scattering chamber contains a second quadrupole mass spectrometer, LEED, XPS, and a differentially pumped sputter gun. The HREELS and TDS experiments were performed in a different UHV apparatus consisting of two chambers separated by a valve.29 The base pressure was 2  10
11 mbar in both chambers. The upper chamber is equipped with an argon ion sputtering gun, LEED optics, and a quadrupole mass spectrometer for TDS experiments. The lower chamber houses a HREELS spectrometer of the latest Ibach design (Delta 0.5, SPECS, Germany) with a straight-through energy resolution of

ARTICLE

1 meV (see ref 30). HREEL spectra were recorded at an angle of incidence of 55 with respect to the surface normal in specular geometry. The energy of the primary electrons was set to 10 eV. The ZnO substrates used in this study (obtained from MaTeck crystals) have been cleaned by a procedure consisting of cycles of Arþ sputtering (600 V, ≈ 1 μA, Ts = 700 K, 60 min) followed by annealing in oxygen (1  10
6 mbar, Ts ≈ 850 K, 10 min) and in UHV (Ts = 850 K, 2
15 min). The crystals were never heated above 900 K to avoid an excessive loss of oxygen. The crystals exhibited a light yellow color, revealing the presence of a small amount of oxygen vacancies in the bulk. Methanol was purified by repeated freeze
pump
thaw cycles. Exposure of the sample to molecular methanol was carried out by backfilling the scattering chamber (typically at 1
5  10
10 mbar) through a leak valve and a capillary which was placed in line-of-sight (distance 5 cm) from the substrate surface. The use of a capillary reduced the contamination of the measuring chamber and led to an effectively 10 times higher pressure in front of the crystal surface. This was verified by comparing measurements with and without a capillary.

’ THEORY The DFT calculations were carried out with the Car
Parrinello Molecular Dynamics code CPMD.31,32 The exchange-correlation contributions to the total energy were treated in the generalizedgradient approximation (GGA) using the PBE functional of Perdew, Burke, and Ernzerhof.33 As has been shown previously, the PBE functional is very well suited for describing accurately the equilibrium structure and binding energy of water on ZnO surfaces3,12,13 and of hydrogen bonded systems, for example, water dimers3 and ice.34 The wave functions for the valence electrons were represented by plane waves with kinetic energy of up to 25 Ry, and Vanderbilt-type ultrasoft pseudopotentials35 were used to describe the core
valence interactions. For the calculations of vibrational frequencies, the cutoff energy was increased to 30 Ry. The k-point mesh for the Brillouin-zone integrations had a density of (4  2  1) with respect to one surface unit cell. For more details on the computational setup, see refs 13, 36, and 37. The mixed-terminated ZnO(1010) surface is nonpolar and consists of characteristic rows of ZnO dimers along the crystallographic [1210]-direction (see Figures 1 and 2). All ZnO surface structures were represented by periodically repeated slabs with a thickness of six ZnO layers separated by a 15 Å vacuum region. The theoretical bulk lattice parameters of a = 3.282 Å, c = 5.291 Å, and u = 0.3792 were used for the lateral extension of the slabs. The atoms in the upper half of the slabs together with the adsorbed molecules were fully relaxed by minimizing the atomic forces, whereas the atoms in the lower half were kept fixed at their bulk positions. Convergence was assumed when the largest residual force component was less than 0.005 eV/Å. Careful tests were done to check the convergence of this calculational setup. Increasing the plane wave cutoff energy to 30 Ry or the density of the k-point mesh to (6  4  1) changed the methanol adsorption energies by less than 0.002 eV. Using a slab thickness of eight instead of six layers lowered the adsorption energies by about 0.01 eV, which is mainly due to the additional relaxation energy since now four instead of three surface layers are relaxed. ’ RESULTS Experiment. Helium diffraction data recorded for the adsorbate-free ZnO(1010) surface and after adsorption of a methanol 7181

dx.doi.org/10.1021/jp200146v |J. Phys. Chem. A 2011, 115, 7180–7188

The Journal of Physical Chemistry A

Figure 1. Side and top views of the most stable atomic configurations of a single methanol molecule adsorbed on the ZnO(1010) surface. The methanol binding energy is given in the upper right corner. The substrate Zn and O ions and the methanol C, O, and H atoms are shown as small gray, large red, green, blue, and white spheres, respectively. In the top views, the atoms in the second surface layer are depicted in lighter colors.

monolayer are displayed in Figure 3. The angular distributions along the [1210]-azimuth (left panel) have been measured at a surface temperature of Ts = 200 K using an incident wave vector of 7.7 Å
1. The angular distribution for the clean surface along the [0001]-direction (right panel) has been recorded at a surface temperature of 250 K and an incident He beam wave vector of 7.8 Å
1 and those for the methanol monolayer at 200 K and 7.9 Å
1. Along the [1210]-direction, distinct half order diffraction peaks are seen. Half order diffraction peaks have also been reported after water adsorption on ZnO(1010) (see ref 3). The absolute intensities of the elastic diffraction peaks are lower by about a factor of 10 for adsorbed methanol compared to a water monolayer, which reflects that methoxy groups are vibrationally softer than hydroxyl groups. Also, the relative He diffraction intensities are substantially different for the two systems. This is expected since the rather different shape of the two molecular layers will lead to different corrugation profiles. In Figure 4, a He-atom reflectivity curve recorded during desorption of the methanol adlayer is plotted (He-TDS). The first stepwise increase of the reflectivity is observed at a temperature of 357 K which corresponds to an activation energy for desorption of 1.0 eV, assuming a pre-exponential factor of 1013 s
1 and that the Redhead formula for first-order desorption can be applied in this case. Closer inspection of the He-TDS data recorded for methanol desorption reveals the presence of a second step for which the increase in the reflectivity is substantially smaller. The second step is located at 440 K, which

ARTICLE

Figure 2. Side and top views of the most stable atomic configurations of a methanol monolayer adsorbed on the ZnO(1010) surface. The binding energy per methanol molecule is given in the upper right corner. The same color coding as in Figure 1 is used.

Figure 3. Left panels: Angular distributions of the clean surface and a methanol monolayer on the mixed-terminated ZnO(1010) surface at a surface temperature of Ts = 200 K, measured with an incident He beam wave vector of ki = 7.7 Å
1. Right panels: Angular distributions of clean and methanol covered ZnO(1010) along the [0001]-direction, Ts = 250 K, ki = 7.8 Å
1 (clean) and 200 K, 7.9 Å
1 (methanol monolayer).

corresponds to an activation energy for desorption of 1.2 eV, again assuming the Redhead expression. In addition to the He-TDS curve in Figure 3, Figure 5 shows angular distributions along [1210] after adsorption of methanol at 180 K and subsequent heating to 300, 350, and 450 K. The different form factors of adsorbate layers with various coverages lead to changes in the relative intensities of the peaks. Note that the temperatures given in the figures cannot be compared 7182

dx.doi.org/10.1021/jp200146v |J. Phys. Chem. A 2011, 115, 7180–7188

The Journal of Physical Chemistry A

ARTICLE

Figure 4. He-TDS of the desorption of a methanol monolayer. The heating rate was 1 K/s.

Figure 6. Angular distribution recorded before the adsorption and after the desorption of a methanol monolayer. The initial measurement was carried out at 400 K. Then the surface was cooled to 180 K in the presence of methanol and heated again to 450 K. The background intensity has been subtracted, and the specular peak has been normalized to one. Lower panel: Difference of the normalized angular distributions after and before adsorption (ki = 7.7 Å
1).

Figure 5. He angular distributions along [1210] after adsorption of methanol at 180 K and subsequent heating to 300, 350, and 450 K (ki = 7.7 Å
1).

directly to those of the TDS experiment which is driven by a rather quick temperature ramp. Specifically, the angular distribution at 350 K can be assigned to the coverage after the first desorption step, and at 450 K the specular intensity is already nearly at maximum. At a surface temperature of 350 K, no half order diffraction peaks are visible anymore. However, the form factor is clearly different from the form factor of the ZnO surface after desorption. In Figure 6, He angular distributions are compared for two cases: before adsorption of methanol and after desorption of a methanol monolayer. To be able to compare the diffraction intensity of first- and second-order peaks in relation to the specular peak, the specular peak has been normalized to one. In the lower panel, the difference between the two curves is shown. Both representations show first- and second-order diffraction peaks which have grown compared to the specular peaks, which means that the form factor is not the same. Figure 7 shows the conventional TDS data of methanol (mass 31) on the perfect ZnO(1010) surface obtained after

Figure 7. TD spectrum of methanol (mass 31) obtained after exposing the clean ZnO(1010) surface to 10 L of methanol at 95 K. The heating rate was 1.5 K/s.

exposure to 10 L of CH3OH at 95 K. The spectrum exhibits an intense desorption peak at 145 K which is not saturated and shifts slightly to higher temperatures with increasing methanol dose. This finding demonstrates the presence of multilayer methanol species weakly bound to the ZnO surface exhibiting zero-order desorption kinetics. In the higher-temperature region, two additional desorption states of methanol are detected: one dominant desorption peak at 370 K and one broad feature centered at about 440 K. Assuming a preexponential factor of 1013 s
1 and first-order kinetics, the activation energies for the desorption of methanol from the two states are estimated to amount to 1.01 eV (98 kJ/mol) and 1.21 eV (117 kJ/mol), respectively, in excellent agreement with the He-TDS results (note that the heating rates in the conventional TDS and the He-TDS experiment were 1.5 and 1.0 K/s, respectively). 7183

dx.doi.org/10.1021/jp200146v |J. Phys. Chem. A 2011, 115, 7180–7188

The Journal of Physical Chemistry A

Figure 8. HREEL spectra recorded after exposing the clean ZnO(1010) surface (curves a and b) to 10 L of methanol at 95 K and subsequent annealing to the indicated temperatures (curves c, d, and e). Curve a is the raw spectrum for the clean surface; the Fourier deconvoluted spectra are shown in curves b
e. The spectra were recorded at 95 K in specular direction with an incidence angle of 55 and with a primary electron energy of 10 eV.

To gain more insight into the interaction of methanol with ZnO(1010), we have carried out additional experiments using vibrational spectroscopy. Figure 8 displays the deconvoluted HREELS data obtained after CH3OH adsorption on the clean ZnO(1010) surface at 95 K and then heating to the indicated temperatures. The raw spectrum of ZnO(1010) is dominated by the intense primary Fuchs
Kliewer phonon mode38 at 556 cm
1 and its multiple excitations at 1112 and 1668 cm
1 (see Figure 8a). The latter are completely removed by Fourierdeconvolution (see Figure 8b), which makes it possible to clearly identify the molecular species adsorbed on the ZnO surface. The Fourier-deconvolution method used in our study was similar to that reported in the literature.39 Exposing ZnO(1010) to 10 L of methanol at 95 K leads to the appearance of a number of new bands at 1047, 1160, 1465, 2842, 2941, and 3280 cm
1 (see Figure 8c). They are characteristic for the formation of multilayer CH3OH species and are assigned to the ν(C
O), F(CH3), δs(CH3), νs(C
H), νas(C
H), and ν(O
H) modes, respectively.40 The O
H stretching mode is observed at 3280 cm
1 which is shifted significantly to lower frequency with respect to the gas phase value (3682 cm
1).41 The large red-shift of the ν(O
H) mode together with its broadening reveals strong intermolecular hydrogen bonding within the methanol multilayer. Upon heating to 245 K, the multilayer methanol is completely desorbed, as demonstrated by the disappearance of the characteristic vibrational modes at 1047 and 3280 cm
1 (see Figure 8d), in line with the TDS results in Figure 7. In the C
O stretch vibration region a new loss appears at 1068 cm
1 with a shoulder on the low-frequency side at about 1020 cm
1, while the νs(C
H) and νas(C
H) modes are slightly red-shifted to 2815 and 2915 cm
1, respectively (see Figure 8d). Interestingly, a weak feature is detected at 3635 cm
1 which increases in intensity after further heating to room temperature (see Figure 8e). This frequency is very similar to those reported previously for OH species on the same surface formed by adsorption of atomic hydrogen (3645 cm
1).42,43 Therefore, we assign the 3635 cm
1 band to surface hydroxyl species formed

ARTICLE

on ZnO(1010). This finding together with the observation of two ν(C
O) modes indicates a partial dissociation of methanol leading to the coexistence of methanol and methoxy species, as demonstrated by the following DFT calculations. Theory. We begin our study of the adsorption properties of methanol on the nonpolar ZnO(1010) surface by comparing the stability of different adsorption modes of isolated molecules. A single methanol molecule was placed in a (4  2) surface unit cell. The periodic images of the molecules are separated by 13.1 and 10.6 Å in the [1210]- and [0001]-direction, respectively, so that the direct interactions between the molecules are negligible. Similar as in our study of water adsorption on ZnO(1010) (see ref 13), we started atomic structure optimizations from many different initial configurations of dissociated and undissociated molecules. The results for the most stable relaxed configurations together with the calculated binding energies are shown in Figure 1. The coordination of the ions in the ZnO surface dimers is reduced to three from its bulk value of four. The surface Zn and O ions therefore act as Lewis acid and Lewis base sites, respectively.5 From this point of view, one would expect that the most favorable adsorption mode of methanol is via a heterolytic dissociation of the O
H bond with a subsequent protonation of a surface O2
and an adsorption of the methoxy anion at a Zn2þ Lewis acid site. However, the methanol binding energy for this configuration is only 0.56 eV (see Figure 1a). The configuration is even metastable. The methanol binding energy increases to 0.70 eV if the methoxy anion moves into a bridging position between two Zn cations (see Figure 1c). Another possibility to saturate the broken surface bonds of the ZnO surface dimers is via a heterolytic split of the methanol C
O bond to give an adsorbed OH
at Zn2þ sites (here only the bridging position is stable), and the methyl group forms a 3-fold coordinated methoxy anion with a surface O2
ion (see Figure 1b). Also for this case the methanol binding energy of 0.58 eV is rather low. Clearly the most favorable adsorption mode for isolated methanol molecules is a molecular adsorption as shown in Figure 1d with a binding energy of 0.97 eV. The methanol molecules sit in a “key
lock”-type configuration above the trenches of the ZnO(1010) surface. The O atom of the methanol molecule coordinates via its lone electron pair to a surface Zn2þ cation, and the H atom forms a hydrogen bond across the trench to a surface O2
anion. A dissociation of the methanol molecule in which the H atom is transferred across the trench to the surface O2
is not stable. In geometry optimizations starting from such a dissociated configuration, the methanol molecules recombine without encountering an energy barrier. The binding energies and the sequence of stability of the different molecular and dissociative adsorption modes for single methanol molecules are in very good agreement with the results of a recent DFT study of Pala and Metiu.44 Overall, in the limit of isolated molecules, methanol and water show a very similar adsorption behavior on ZnO(1010): in both cases, molecular adsorption is most stable with almost identical binding energies, whereas dissociation of the molecules is unfavorable.13 In the next step, we investigated the coverage dependence of methanol adsorption on ZnO(1010). Geometry optimizations were performed for different periodic arrangements of methanol molecules which were placed across the trenches of the ZnO(1010) surface as in the most stable configuration for isolated molecules (see Figure 1d). The methanol binding energy increases slightly if rows of methanol molecules in the [0001]-direction are formed which saturate simultaneously the 7184

dx.doi.org/10.1021/jp200146v |J. Phys. Chem. A 2011, 115, 7180–7188

The Journal of Physical Chemistry A Zn and O ions of a row of ZnO dimers. For (4  1), (3  1), and (2  1) arrangements of undissociated methanol molecules, a binding energy of 0.99 eV was obtained. The same attractive interaction between adsorbed molecules along [0001] was also found for water.13 For these structures up to a coverage of 1/2 monolayer, the methanol adsorption remains molecular. Dissociated structures with protons of the methanol molecules transferred across the trench to a surface O2
ion recombine in the geometry optimization. When the coverage of methanol is increased to a full monolayer, the binding energy for a (1  1) arrangement of undissociated molecules drops to 0.81 eV (see Figure 2a). The reason for the lower binding energy is the steric repulsion between the methyl groups of methanol molecules in neighboring [0001]-rows. This is now very different from the behavior of water: in contrast to methanol, water can form hydrogen bonds between neighboring molecules in the [1210]-direction which gives rise to an attractive interaction.13 In the case of water, the attractive interaction between neighboring molecules leads to the formation of water pairs in which one of the water molecules is dissociated. At full monolayer coverage, water films therefore show a (2  1) periodicity.3,13 To probe if for methanol also a superstructure can form, now due to a repulsive interaction, short molecular dynamics runs were performed to search for alternative configurations. A slab with (4  2) surface unit cells containing eight methanol molecules was used for the ab initio Car
Parrinello molecular dynamics simulations.31,32 All CPMD runs were started from the relaxed (1  1) arrangement of undissociated water molecules as shown in Figure 2a. The calculations were run for several picoseconds at a temperature of 300 K. In all simulations it was observed that some of the methanol molecules spontaneously dissociate and transfer a proton across the trenches to a surface O2
, whereas recombination rarely occurred. Several typical structural motives containing dissociated methanol molecules were taken from the CPMD trajectory and a subsequent geometry optimization was performed. Altogether we found three different new adsorbate structures with a lower energy than the molecular (1  1) monolayer (see Figure 2). All three structures consist of pairs of dissociated/undissociated methanol molecules which gives rise to a (2  1) periodicity. The first structure shown in Figure 2b is the structural analogue to the most stable configuration of water monolayers on ZnO(1010): The undissociated methanol molecule remains in its key
lock position across the trenches of the ZnO surface. The second methanol molecule donates its proton to a surface O2
across the trench, and the methoxy anion moves closer to the Zn cation, thereby reducing the steric repulsion of the methyl group with the neighboring undissociated methanol molecules. Compared to the molecular (1  1) monolayer the binding energy per molecule increases by 0.03 eV. In the second structure (see Figure 2c) the methoxy anion of the dissociated methanol molecules moves above the ZnO surface dimer. The O atoms of the ZnO surface dimers are protonated by the dissociation of the neighboring methanol molecules along the [0001]-lines. These OH groups change their orientation from [0001] to [0001] and form hydrogen bonds to the O atom of the methoxy anion sitting above the same ZnO dimer. The bond of the underlying ZnO dimer breaks, and instead new ZnO bonds across the trenches are formed. Despite this strong distortion of the ZnO surface structure, the binding energy per methanol molecule is higher by 0.03 eV compared to the water-like partially dissociated structure of Figure 2b.

ARTICLE

Table 1. Vibrational Frequencies (in cm
1) for Undissociated Methanol Molecules and Methoxy Groups in the Most Stable Configuration of a Methanol Monolayer Adsorbed on ZnO(1010) as Shown in Figure 2d vibration mode

methanol

methoxy

νas(C
H)

2991/3016

2928/2957

νs(C
H)

2919

2872

δas(CH3)

1434/1443

1434/1452

δs(CH3)

1403

1419

F(CH3)

1124/1172

1133/1140

ν(C
O)

1055

1016

Clearly the most stable configuration with a binding energy of 0.94 eV per methanol molecule is a derivative of the first partially dissociated structure. As shown in Figure 2d, the undissociated methanol molecule has rotated out of its key
lock configuration and forms now a hydrogen bond with the O atom of a neighboring methoxy group instead of with the surface O2
ion across the trench. By this reorientation, the steric repulsion of the methyl groups can be reduced to a minimum. Interestingly, the binding energy per methanol molecule in the full monolayer of 0.94 eV is slightly lower than the value of 0.99 eV at half monolayer coverage. This is in contrast to water where the highest binding energy per molecule is found for the monolayer. The reason is that water can form two H bonds per molecule at monolayer coverage, whereas for methanol only one H bond can exist and the methyl groups have to arrange in such a way that the steric repulsion is minimized. The consequence of this difference is that while for water the monolayer coverage is the thermodynamically most stable configuration up to the desorption temperature, methanol shows a sequence of thermodynamic stable structures with increasing temperature: first, a monolayer of methanol with half of the molecules dissociated at low temperatures, then a low-density phase of [0001] rows of undissociated methanol molecules with a maximum of half monolayer coverage at higher temperatures and finally full desorption of methanol.13 Finally, for the most stable structure of a methanol monolayer adsorbed on ZnO(1010), Figure 2d, the vibrational frequencies of the dissociated and undissociated molecules have been calculated in harmonic approximation using a finite difference scheme. The adsorbate atoms together with the Zn and O ions in the two topmost surface layers were displaced by 0.01 Å into the three Cartesian directions, and the dynamical matrix was determined from the resulting atomic forces. The calculated frequencies are summarized in Table 1. The C
H vibrations are in reasonablly good agreement with the HREELS results. Interestingly, for the C
O stretching mode ν(C
O), different values of 1055 and 1016 cm
1 are found for the undissociated and dissociated molecules, respectively, which agree very closely with the two new losses at 1068 and 1020 cm
1 that appear in the HREELS experiments after desorption of the methanol multilayers.

’ DISCUSSION The He-atom scattering data clearly reveal that the adsorption of methanol leads to the formation of (at least) two different adlayer phases on defect-free terraces of the ZnO(1010) surface. The first, low-temperature (LT) phase is stable up to temperatures well above room temperature and exhibits a (2  1) 7185

dx.doi.org/10.1021/jp200146v |J. Phys. Chem. A 2011, 115, 7180–7188

The Journal of Physical Chemistry A periodicity. On the basis of the theoretical results, we conclude that this (2  1) LT structure does not correspond to a half monolayer coverage but to a full monolayer of methanol, i.e., one molecule per surface unit cell. Our calculations show that the repulsion energy between methanol molecules on neighboring adsorption sites is much lower than the methanol binding energy to the surface. Thus, at low temperatures, the energy gain per surface area upon methanol adsorption is much larger for a full methanol monolayer than at half monolayer coverage. Instead of only having adsorbed methanol molecules on every second surface site, the calculations suggest that the (2  1) periodicity in the LT phase is brought about by the presence of an equal number of protonated and deprotonated molecules (i.e., methanol and methoxy species) within a full methanol monolayer, which are arranged in the fashion shown in Figure 2d. The barrier for the dissociation of every second methanol molecule within the monolayer is rather low since this process occurred spontaneously in our short molecular dynamics runs. Therefore, it is justified to analyze the TDS measurements on the basis of a firstorder kinetics, and the calculated binding energies for the methanol molecules can be compared directly to the experimental activation energies for desorption. Above 357 K, the HeTDS data indicate that a restructuring of the adlayer occurs which is accompanied by the desorption of methanol molecules as seen by conventional TDS. The activation energy for methanol desorption of 1.0 eV, derived from both the first steep rise in the He reflectivity at 357 K (He-TDS) and the first desorption maximum at 370 K in conventional TDS experiments, agrees very well with the calculated binding energy per methanol molecule of 0.94 eV for the structure depicted in Figure 2d. The HREELS results provide spectroscopic evidence for the coexistence of intact and dissociated methanol molecules within the (2  1) LT phase on ZnO(1010). HREELS data shown in Figure 8d and e indicate two C
O stretching bands located at 1068 and 1020 cm
1, respectively, which suggest the presence of intact methanol as well as methoxy species. A very satisfying agreement exists between these experimental values and the theoretical results. The theoretical frequencies amount to 1055 and 1016 cm
1 for the undissociated and dissociated molecules, a difference of 13 and 4 cm
1 relative to the experimental values, respectively. The deprotonation of methanol is further confirmed by the emergence of an OH band located at 3635 cm
1, which is attributed to hydroxyl species formed on the ZnO(1010) surface.14,42,43 It should be noted that the ν(O
H) mode for molecular methanol is not detected in the HREELS measurements. This mode is expected to exhibit a strong broadening and a significant red-shift in frequency due to the hydrogen bonding to neighboring O atoms as confirmed by the present DFT calculations. We believe that this (2  1) LT phase is identical with the island structure observed at room temperature in a recent STM study of methanol adlayers on the same substrate reported by Iwasawa and co-workers.24 The registry between the dissociated/ undissociated methanol molecules and the Zn ions of the substrate in our structural model (see Figure 2d) is in full agreement with the STM measurements. Presumably, the (2  1) periodicity was not seen in the STM experiments because the resolution was not sufficient to resolve the details of the molecular arrangements. In the STM measurements only sometimes a vague (1  1) contrast could be obtained,24 which is probably the result of a time average of different structures due to the dynamic behavior of the adsorbate layer, as was observed previously for water layers adsorbed on ZnO(1010).12 Unfortunately, the

ARTICLE

interpretation of the STM data reported in ref 24 was not supported by a comparison to theoretical simulations of the STM images (see, e.g., refs 3 and 12). The nature of the (2  1) LT structure containing an equal fraction of intact methanol and methoxy species is quite similar to the adlayer structure resulting from the exposure of the same substrate to water. Also in this case, every second molecule was found to be dissociated.3,12
14 A more detailed comparison of the structural models obtained from the theoretical calculations for water and for methanol, however, reveals two substantial differences between the apparently similar systems, originating mainly from the more demanding sterical requirements of the methyl group in methanol. First, while for water the formation of pairs of dissociated and undissociated molecules in the (2  1) structure is driven by an attractive interaction between the water molecules due to the formation of H bonds, for methanol the (2  1) periodicity is the result of a rather substantial repulsion between the methanol methyl groups. Consequently, one would expect that the two adlayers should show a different behavior upon heating: while for water the full monolayer is the thermodynamically most stable phase up to the desorption temperature, for methanol the repulsive interactions should give rise to an additional low-coverage phase for a small temperature range before full desorption occurs. Second, while the attractive interaction between the water molecules leads to only one stable (2  1) structure for the water adlayer, the steric repulsion between methanol molecules on neighboring surface sites can be reduced in several ways which gives rise to at least three different stable (2  1) overlayers that are close in energy (see Figure 2). Therefore, at room temperature the lowest-energy structure of the methanol monolayer, Figure 2d, will exhibit a rather high density of structural defects, in particular, methanol molecules with reoriented OH groups (see Figure 2b) and methoxy groups that have moved above the ZnO surface dimers (see Figure 2c). Altogether, the methanol adlayer will show a considerably higher degree of structural disorder than a water monolayer. This higher degree of disorder should contribute to a reduction of diffraction peak intensities, in addition to the substantially higher Debye
Waller factor for methanol, which is due to the presence of vibrations at lower frequencies (resulting from the larger mass of methanol) than in the case of water. This is in full accord with the experimental data, where the absolute intensities of the elastic diffraction peaks in HAS are found to be reduced by a factor of 10 compared to a water monolayer (see Figure 3). At 440 K the TDS experiments show a second methanol desorption peak. Below 500 K only desorption of methanol is observed in conventional TDS, but no indication of other species which might stem from a decomposition of methanol or a reaction between adsorbed species was found. Since the structural change of the surface accompanied by the desorption of methanol at 440 K is also detected in the HAS experiments via an increase in the He-atom reflectivity and a change of the surface form factor, the desorbing methanol molecules must originate from the terraces of the ZnO(1010) surface and not from defect sites (e.g., vacancies or steps). Thus, we can conclude that above 370 K, the temperature of the first methanol desorption peak in the conventional TDS experiments, a second, low-density, hightemperature (HT) methanol adlayer phase is formed on the ZnO(1010) terraces which is stable up to 440 K. Indeed, the formation of such a second, low-density adlayer phase at higher temperatures would be expected from the presence of a repulsive interaction between methanol molecules on neighboring 7186

dx.doi.org/10.1021/jp200146v |J. Phys. Chem. A 2011, 115, 7180–7188

The Journal of Physical Chemistry A adsorption sites. The theoretical calculations predict for this lowdensity HT phase a structure in which every second [0001]-row of surface ZnO dimers is saturated by undissociated methanol molecules which gives rise to a methanol coverage of 0.5 monolayers and a (2  1) periodicity as for the LT phase. Unfortunately, above 300 K the intensity of the He-atom diffraction peaks is so strongly reduced as a result of Debye
Waller effects that the HAS experiments are not able to confirm whether the low-density HT phase still exhibits a (2  1) periodicity or not. Furthermore, the calculated binding energy per methanol molecule increases only from 0.94 eV for the most stable structure of a methanol monolayer to 0.99 eV in this half monolayer HT structure, whereas a Redhead analysis of the desorption peaks in the TDS experiments yields values of 1.0 and 1.2 eV for the activation energy for methanol desorption for the two phases. Although direct experimental evidence for the presence of a low-density HT phase—in contrast to the case of the (2  1) LT structure—is lacking, all existing experimental results are fully consistent with a HT phase of half a monolayer of methanol as proposed on the basis of the theoretical calculations. In particular, the results obtained by conventional TDS clearly demonstrate that the temperature-induced formation of the HT phase is accompanied by the desorption of methanol as expected from a reduction of a coverage of 1.0 for the LT phase to 0.5 for the HT phase. It is also interesting to note that after full desorption of methanol the clean ZnO(1010) surface shows a somewhat different form factor than before methanol adsorption (see Figure 6). This indicates that the methanol adsorption and desorption have induced some structural modification in the ZnO(1010) surface layer, even though the temperature was never raised above 450 K. Iwasawa and co-workers came to a similar conclusion in their STM study:24 after the desorption of the methanol molecules from the linear chain minority phase structure, they observed the appearance of defect lines of missing ZnO dimers which were not present before methanol adsorption.

’ CONCLUSIONS The adsorption of methanol on the mixed-terminated ZnO(1010) surface has been investigated with He-atom scattering, HREELS, TDS, as well as with density functional theory calculations. At low temperatures, a monolayer of methanol molecules with a (2  1) periodicity is formed which consists of an equal fraction of undissociated and dissociated (methoxy) species. This phase is thus similar to the case of water adsorbed on the same surface,3,12
14 although the driving force for the formation of the (2  1) periodicity is not the hydrogen-bonding mediated attractive interaction in the case of adsorbed water molecules but the steric repulsion of the methanol methoxy species. Upon heating, a sequence of two major desorption peaks is observed, the first one at 357/370 K and a second one at 440 K, with corresponding activation energies for methanol desorption of 1.0 and 1.2 eV, respectively. On the basis of the theoretical calculations, the high-temperature phase below 440 K is identified as a low-density structure with a methanol coverage of half a monolayer. ’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected]; christof.woell@ kit.edu.

ARTICLE

’ ACKNOWLEDGMENT This work was supported by the German Research Foundation (DFG) via the Collaborative Research Center SFB 558 “Metal
Substrate Interactions in Heterogeneous Catalysis”. Computational resources were provided by Bovilab@RUB (Bochum). ’ REFERENCES (1) W€oll, Ch. Prog. Surf. Sci. 2007, 82, 55. (2) Staemmler, V.; Fink, K.; Meyer, B.; Marx, D.; Kunat, M.; Gil Girol, S.; Burghaus, U.; W€oll, Ch. Phys. Rev. Lett. 2003, 90, 106102. (3) Meyer, B.; Marx, D.; Dulub, O.; Diebold, U.; Kunat, M.; Langenberg, D.; W€oll, Ch. Angew. Chem. 2004, 116, 6810; Angew. Chem., Int. Ed. 2004, 43, 6642. (4) Kunat, M.; Meyer, B.; Traeger, F.; W€oll, Ch. Phys. Chem. Chem. Phys. 2006, 8, 1499. (5) Wang, Y.; Kovacik, R.; Meyer, B.; Kotsis, K.; Stodt, D.; Staemmler, V.; Qui, H.; Traeger, F.; Langenberg, D.; Muhler, M.; W€ oll, Ch. Angew. Chem. 2007, 119, 5722; Angew. Chem., Int. Ed. 2007, 46, 5624. (6) Wang, Y.; Xia, X.; Urban, A.; Qiu, H.; Strunk, J.; Meyer, B.; W€oll, Ch.; Muhler, M. Angew. Chem. 2007, 119, 7456; Angew. Chem., Int. Ed. 2007, 46, 7315. (7) Diebold, U. Surf. Sci. Rep. 2003, 48, 53. (8) Kunat, M.; Traeger, F.; Silber, D.; Qiu, H.; Wang, Y.; van Veen, A. C.; W€oll, Ch.; Kowalski, P. M.; Meyer, B.; H€attig, C.; Marx, D. J. Chem. Phys. 2009, 130, 144703. (9) Kowalski, P. M.; Meyer, B.; Marx, D. Phys. Rev. B 2009, 79, 115410. (10) Sun, C.; Liu, L.-M.; Selloni, A.; Lu, G. Q.; Smith, S. C. J. Mater. Chem. 2010, 20, 10319. (11) Henderson, M. A. Surf. Sci. Rep. 2002, 46, 5. (12) Dulub, O.; Meyer, B.; Diebold, U. Phys. Rev. Lett. 2005, 95, 136101. (13) Meyer, B.; Rabaa, H.; Marx, D. Phys. Chem. Chem. Phys. 2006, 8, 1513. (14) Wang, Y.; Muhler, M.; W€oll, Ch. Phys. Chem. Chem. Phys. 2006, 8, 1521. (15) Kurtz, M.; Strunk, J.; Hinrichsen, O.; Muhler, M.; Fink, K.; Meyer, B.; W€oll, Ch. Angew. Chem. 2005, 117, 2850; Angew. Chem., Int. Ed. 2005, 44, 2790. (16) Zwicker, G.; Jacobi, K.; Cunningham, J. Int. J. Mass Spectrom. Ion Processes 1984, 60, 213. (17) Akhter, S.; Cheng, W. H.; Lui, K.; Kung, H. H. J. Catal. 1984, 85, 437. (18) Akhter, S.; Lui, K.; Kung, H. H. J. Phys. Chem. 1985, 89, 1958. (19) Vest, M. A.; Lui, K. C.; Kung, H. H. J. Catal. 1989, 120, 231. (20) Vohs, J. M.; Barteau, M. A. Surf. Sci. 1986, 176, 91. (21) Vohs, J. M.; Barteau, M. A. Surf. Sci. 1988, 197, 109. (22) Ueno, A.; Onishi, T.; Tamaru, K. Trans. Faraday Soc. 1971, 67, 3585. (23) K€ahler, K.; Holz, M. C.; Rohe, M.; Strunk, J.; Muhler, M. Chem. Eur. J. Chem. Phys. 2010, 11, 2521. (24) Shao, X.; Fukui, K.; Kondoh, H.; Shionoya, M.; Iwasawa, Y. J. Phys. Chem. C 2009, 113, 14356. (25) W€oll, Ch. J. Phys.: Condens. Matter 2004, 16, S2981. (26) Traeger, F. Chem. Phys. Chem. 2006, 7, 1006. (27) Toennies, J. P. Surface Phonons, Springer Series of Surface Science; W., K., Wette, F. W. D., Eds.; Springer: Berlin, 1991. (28) Hinch, B. J.; Lock, A.; Madden, H. H.; Toennies, J. P.; Witte, G. Phys. Rev. B 1990, 42, 1547. (29) Wang, Y.; Jacobi, K.; Ertl, G. J. Phys. Chem. B 2003, 107, 13918. (30) Wang, Y.; Jacobi, K.; Sch€ one, W.-D.; Ertl, G. J. Phys. Chem. B 2005, 109, 7883. (31) Marx, D.; Hutter, J. Ab Initio Molecular Dynamics; Cambridge University Press: New York, 2009. (32) Hutter, J. CPMD program package, see www.cpmd.org. 7187

dx.doi.org/10.1021/jp200146v |J. Phys. Chem. A 2011, 115, 7180–7188

The Journal of Physical Chemistry A

ARTICLE

(33) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865; Erratum: Phys. Rev. Lett. 1997, 78, 1396. (34) Hamann, D. R. Phys. Rev. B 1997, 55, R10157. (35) Vanderbilt, D. Phys. Rev. B 1990, 41, 7892. (36) Meyer, B.; Marx, D. Phys. Rev. B 2003, 67, 035403. (37) Kiss, J.; Witt, A.; Meyer, B.; Marx, D. J. Chem. Phys. 2009, 130, 184706. (38) Wang, Y. Z. Phys. Chem. 2008, 222, 927. (39) Cox, P. A.; Williams, A. A. Surf. Sci. 1985, 152/153, 791. (40) Falk, M.; Whalley, E. J. Chem. Phys. 1961, 34, 1554. (41) Herzberg, G. In Infrared and Raman Spectra; Van Nostrand: New York, 1956. (42) Wang, Y.; Meyer, B.; Yin, X.; Kunat, M.; Langenberg, D.; Traeger, F.; Birkner, A.; W€oll, Ch. Phys. Rev. Lett. 2005, 95, 266104. (43) Noei, H.; Qiu, H.; Wang, Y.; L€offler, E.; W€oll, Ch.; Muhler, M. Phys. Chem. Chem. Phys. 2008, 10, 7092. (44) Pala, R. G. S.; Metiu, H. J. Catal. 2008, 254, 325.

7188

dx.doi.org/10.1021/jp200146v |J. Phys. Chem. A 2011, 115, 7180–7188