摘要:

Mg clusters on MgO surfaces study of the nucleation mechanism with MIES and ab initio calculations L. N. Kantorovich,a A. L. Shluger,a P. V. Sushko,ab J. Gué nster,c P. Stracke,d D. W. Goodmanc and V. Kempterd a Department of Physics and Astronomy University College L ondon Gower Street L ondon UK W C1E 6BT b T he Royal Institution of Great Britain 21 Albemarle Street L ondon UK W 1X 4BS c Department of Chemistry T exas A&M University College Station T X77842-3012 USA d Physikalisches Institut der T U Clausthal L eibnizstraêe 4 D-38678 Clausthal-Zellerfeld Germany 173 Faraday Discuss. 1999 114 173»194 Received 23rd April 1999 We combined experimental studies using ultraviolet photoelectron spectroscopy (UPS) metastable impact electron spectroscopy (MIES) and temperature programmed desorption (TPD) with ab initio calculations of metal adsorption on the perfect MgO surface and at defect sites in order to elucidate the role of surface defects in the initial stages of nucleation and growth of metal clusters at oxide surfaces.MgO –lms (2 nm thick) grown on Mo and W substrates were used as a prototype system. The MIES and UPS (HeI) spectra were collected in situ and the growth of Mg clusters was observed by monitoring the dynamics of additional MIES peaks during Mg deposition. TPD experiments were made in order to monitor the surface coverage by Mg clusters and to determine the Mg desorption energies. Interpretation of the results was made on the basis of theoretical modelling using density functional theory (DFT) calculations in both periodic and embedded cluster models.The geometric and electronic structures of the surface terrace F-centre positively charged anion vacancy and step edge at the MgO(001) surface were calculated and their role in adsorption and clustering of Mg atoms on this surface was studied. The absolute position of the top of the surface valence band of MgO with respect to the vacuum was calculated and compared with the MIES results. The MIES spectra were modelled on the basis of surface density of states (SDOS). The calculated SDOS predicted the location of additional peaks in the band gap and their shift as a function of Mg concentration on the surface in agreement with the MIES data. The desorption energies of Mg atoms from small Mg clusters formed at step edges are found to be about 1.3 eV atom~1.Comparison between the theoretical results and the experimental data suggests preferential initial adsorption of Mg atoms at steps and kinks rather than at charged and neutral vacancies. At larger exposures these Mg atoms serve as the nucleation sites. This journal is( The Royal Society of Chemistry 2000 1 Introduction Understanding of the mechanisms of growth and parameters of the geometric and electronic structures of metal clusters and layers on metal oxide surfaces is important for a number of technological applications. In particular metal addition to oxides leads to an enhanced reactivity via electron transfer to a variety of adsorbed molecules leading to the formation of radical anion species.1,2 The interaction between metal clusters and metal oxide supports plays a key role in catalysis3,4 and microelectronics.5 The interaction between metal atoms and oxide surfaces is important for understanding the mechanisms of their segregation,6,7 diÜusion of metal atoms on insulators,8 formation of metal-induced point defects on oxides,2,9 and the formation of nanoparticles inside semiconductors and insulators.10 Surprisingly little is known about the structure of metal/oxide interfaces in particular the initial stages of the metal adsorption types of the metal adsorption sites the nature of bonding to oxides and between the adsorbed metal atoms.5,11 Although it is clear that the defect sites such as surface vacancies step edges and kinks play a signi–cant role at least at the early stages of metal growth on oxides the number distribution and structure of these defects is very difficult to control experimentally.On the other hand most of the existing theoretical calculations are concerned with metal adsorption on ideal oxide surfaces (see for example refs. 12 and 13) and only very few treat metal adsorption on defective oxide surfaces (for a review see ref. 14). In particular Ferrari and Pacchioni15 performed cluster Hartree»Fock calculations of MgO surfaces with point defects (neutral and charged anion and cation vacancies) and studied the charge transfer between the Rb atoms and these defective surfaces. While the neutral point defects are not very reactive the charged anion vacancies can ionize metal atoms provided the electron affinity of the defect is larger than the ionisation potential of the metal atom at the surface.Thus the interaction of metal atoms with surface point defects can greatly alter their surface diÜusion behaviour and consequently can be responsible for cluster nucleation in the neighbourhood of defect sites. In this study we combined several experimental techniques with theoretical modelling in order to elucidate the early stages of Mg cluster formation on the (001) surface of MgO thin –lms. Metastable impact electron spectroscopy (MIES) using He* (1s2s) projectiles is particularly useful for these purposes because it probes the very top surface layer and is sensitive to quite small concentrations of adsorbed species. It is naturally combined with ultraviolet photoelectron spectroscopy (UPS) using HeI as the light source.Unlike UPS (HeI) MIES is extremely sensitive to features resulting from the charge density of s-electrons speci–c for adsorption of alkali and alkaline earth atoms. Correlation of the spectroscopic data with the results of the temperature programmed desorption (TPD) experiments is illuminating for the interpretation of the spectral features and understanding of the initial stages of the metal cluster growth. To understand better the experimental data and to construct a model of metal adsorption we performed ab initio electronic structure calculations using the density functional theory (DFT) and the method of pseudopotentials. The embedded cluster model16,17 was employed for DFT calculations of the surface ionisation energies which are compared with the position of the top of the valence band with respect to the vacuum level of the MgO –lm determined from the MIES data.Using the periodic DFT calculations we treated the adsorption of up to –ve Mg atoms on the perfect (001) surface near an anion vacancy and at a neutral surface F-center. In order to elucidate the role of extended surface defects in the Mg cluster growth we modelled the adsorption of up to four Mg atoms at a step edge. To facilitate comparison with the experiment in all these cases we analysed the energetics and the geometric and electronic structures of the metal clusters adsorbed on the surface and near defects and calculated the surface density of states (SDOS).What have we learned regarding the adsorption of metal on MgO thin –lms from this complex study? Both the experimental MIES spectra and the results of our calculations give similar energies for the position of the top of the valence band (about 6.5 eV) of the MgO –lm with respect to the vacuum level. This is an important parameter for adsorption photochemistry and interface studies on MgO –lms. In the context of the present paper one can compare this value with the 3s electron ionisation energy of the Mg atom (7.6 eV) and use a simple argument in order to predict the nature of the chemical bonding of Mg atoms on the MgO surface. This qualitative prediction Faraday Discuss. 1999 114 173»194 174 was then con–rmed using the results of our DFT calculations which demonstrate the formation of bonding and anti-bonding states due to the interaction of adsorbed Mg atoms with the surface oxygen ions.The electronic states due to Mg adsorption manifest themselves in the MIES spectra which can be approximately interpreted using the calculated SDOS. The attachment energies of Mg atoms to Mg clusters formed at diÜerent surface sites are compared with the TPD data. The best agreement is achieved with the energies at the step edge. These are the largest adsorption energies we found. Combined with the MIES data this suggests that the initial nucleation of Mg clusters happens at step edges. The paper is organised as follows. In Section 2 we give a brief account of the experimental techniques used in this study and present the experimental results.The theoretical methods and the results of calculations are described in Section 3. The discussion of the experimental and theoretical results and conclusions are presented in Section 4. A preliminary account of some of our results is presented in ref. 18. 2 Experimental results 2.1 Experimental techniques The apparatus used in these studies has been described previously.19,20 Brie—y it is equipped with a cold-cathode gas discharge source which also serves (i) for the production of metastable He (3S/1S) (E*\19.8/20.6 eV excitation energy) with thermal kinetic energy required for MIES and (ii) as a source for ultraviolet photoelectron spectroscopy UPS(HeI) E*\21.2 eV. The intensity ratio 3S/1S is found to be 7 1.Additionally the apparatus is equipped with X-ray photoelectron spectroscopy (XPS) Auger electron spectroscopy (AES) and low energy electron diÜraction (LEED). Metastable and photon contributions within the beam were separated by means of a time-of-—ight method using a mechanical chopper. The MIES and UPS spectra were acquired with incident photon/metastable beams 45° with respect to the surface normal. The kinetic energy of the electrons emitted in the direction normal to the surface is measured by employing a hemispherical analyzer (Leybold EA10/100) with an energy resolution of 250 meV for MIES/UPS. Collection of each MIES/UPS spectrum requires approximately 140 s. A second apparatus described elsewhere,21 is equipped with a MIES/UPS source of the same type as described above and a setup for TPD was used to calibrate the Mg coverages.The TPD spectra were collected with a diÜerentially pumped quadrupole mass –lter in line-of-sight to the sample while ramping the sample temperature linearly by 3 K s~1. In addition this second apparatus is equipped with XPS AES and LEED. The qualitative interpretation of the results of the MIES experiments is based essentially on a model of refs. 22 and 23 which is shown schematically in Fig. 1. Excited He* atoms approach the surface with thermal velocities. At the distances of about 2.5»4.0 ” when there is a considerable t overlap between the surface (n is a band k the wavevector) and the He t wavefunctions nk 1s surface electrons tunnel into the 1s He hole states. This may happen from all surface energy levels that are higher than the He 1s level.In the case of MgO all electronic states of the O 2p VB participate (see Fig. 1). The energy gained in this so-called Auger de-excitation (AD) process is transferred to the electron occupying the He 2s level which is emitted in the same process. The kinetic energies Ekin of the emitted electrons are measured in MIES experiments and their distributions constitute MIES spectra. Conventionally though the electron spectra are presented vs. the binding energy scale which refers to the Fermi energy EF of metallic substrate as shown in Fig. 1. Experimentally the Fermi energy EF is a –xed point on the energy scale and corresponds to the maximum kinetic energy at which electrons can be measured with MIES and UPS from a metallic substrate.Since substrate and analyser are in electrical contact E appears at the same F kinetic energy irrespective from the substrate work function i.e. for all Mg exposures. Thus presenting the spectra with a binding energy scale with E as origin allows the change of the F work function (due to for example adsorption or charging) to be determined from the shift of the high-energy cutoÜ of the spectra. In addition the absolute value of the work function can be determined from the energetic distance between this cutoÜ and the point on the energy scale that equals the excitation energy (19.82 eV) of the probe atom. The maximum binding energy with respect to the vacuum level Evac probed by He* equals its excitation energy minus the binding energy of the He* 2s electron.175 Faraday Discuss. 1999 114 173»194 Fig. 1 Energy diagram for a He* probe atom in front of a surface of insulator. Left side energy levels of the isolated He and Mg atoms and surface density of states in the valence band (VB). Also shown is the position of the Fermi level EF in the insulator band gap; U is the work function of the surface. Middle binding energies Ebin\EF[E of electrons involved in the Auger de-excitation process are usually presented with respect to this axis which has its origin at EF . Right side schematic of the experimental spectrum of kinetic energies of the electrons emitted in the AD process (Ekin). Zero kinetic energy corresponds to a binding energy of 19.82 eV with respect to the vacuum level (or (19.82[U) eV with respect to EF).2.2 Electron spectroscopy MgO layers with an approximate thickness of 2 nm were prepared by evaporation of Mg on Mo(100) and W(110) substrates at room temperature followed by a subsequent annealing at 800 K in oxygen ambient. The MIES and UPS spectra measured on as-prepared MgO –lms are very similar to those obtained on MgO single crystals.24 Fig. 2 shows the MIES and UPS spectra collected during the exposure of the MgO –lm grown on the Mo(001) surface to Mg atoms at 100 K; no signi–cant changes are found at 300 K. As discussed previously,24 the MIES spectrum acquired from the MgO –lm prior to the Mg exposure (bottom spectrum) re—ects the MgO SDOS as seen via an AD process. Thus the spectral feature with binding energies between 4 and 10 eV with respect to E is due to the ionisation of the MgO valence band states with O 2p F character.The large peak in the MIES spectra located between 10 and 17 eV is to some extent aÜected by secondary and scattered electrons and will not be considered in the following discussion. One important characteristic of the electronic structure of our MgO –lm is the position of the top of the valence band with respect to the vacuum level. It depends on the surface preparation and is difficult to determine using conventional methods (see for example the discussion in ref. 25). It can be determined using the MIES spectra and the following considerations. The distance (as seen in Fig. 2) between E and the top of the valence band is about 4 eV.The distance between F E and the vacuum level i.e. the work function U is determined from the high energy cutoÜ of the F spectra. The work function measured for the MgO –lms grown on the Mo(100) and W(110) substrates is equal to 2.7 eV. Therefore for the MgO –lms used in our study we –nd the top of the valence band at 6.7^0.4 eV with respect to the vacuum level. As a consequence of the Mg dosing an additional peak located at ^2 eV above the top of the valence band (2 eV binding energy) develops within the band gap. Since the work function of the MgO –lm (2.7 eV) is considerably smaller than the He 2s binding energy no unoccupied states are available at the surface into which resonant transfer of the 2s electron can take place even when taking into account an eventual shift of the 2s level during the He* interaction with the surface.Consequently the band gap feature arises from the AD process which involves electrons from the Faraday Discuss. 1999 114 173»194 176 Fig. 2 MIES and UPS spectra acquired from the MgO surface as a function of the Mg exposure. In MIES the bottom spectrum shows the clean MgO surface ; the topmost the fully covered one; UPS vice versa. Inset (top panel) work function change vs. exposure time. The dashed spectra were acquired near the work function minimum. occupied states below the Fermi level due to adsorbed Mg atoms. At larger Mg exposures the MgO valence band emission between 4 and 10 eV weakens considerably. The disappearance of the O 2p structure in the topmost MIES spectrum in Fig.2 indicates that the entire surface is covered by Mg; the shape of this spectrum is very similar to that for Mg –lms. Also shown in Fig. 2 (see inset) is the work function change during the Mg exposure. When Mg atoms are dosed to the oxide the work function decreases by ^0.5 eV while the valence band intensity decreases by ^10% (dotted MIES spectra). Simultaneously the top of the valence band shifts towards larger binding energies by approximately the same amount. This coincidental shift of the valence band structures and the high binding energy cutoÜ indicates a band bending eÜect rather than a real change of the work function. Such a band bending can be attributed to the creation of additional states on the surface.5 At larger exposures the work function plateaus at a 177 Faraday Discuss.1999 114 173»194 level typical for metallic Mg –lms (3.6 eV) which is consistent with a model in which Mg islands grow with lateral bonding similar to bulk Mg. In order to gain more detailed information about the changes in the low binding energy region during Mg dosage to the MgO surface MIES spectra were also acquired using a lower Mg evaporation and a higher energy resolution. Fig. 3 summarizes the data obtained at the low evaporation rate. As shown in Fig. 3 both the energy position and the peak width depend (weakly) on the exposure time. An observed feature has a Gaussian shape with 1.8 eV FWHM over the entire studied exposure range. Since this feature appears well separated both from the valence band maximum and EF the species responsible for this structure exhibits nonmetallic behavior.The band gap feature can be detected until its intensity falls below a level of 10~3 of that of the valence band O 2p emission of the clean MgO surface. Additional measurements show that the band gap feature is stable up to 500 K and has virtually disappeared upon heating to 600 K. In the valence band region the UPS measurements (see Fig. 2) provide similar information to MIES i.e. an attenuation of the MgO substrate intensity at increasing Mg coverages and a shift of the O 2p structure which follows essentially the work function change of the substrate. On the other hand due to the fact that UPS probes the average character of several top layers a signi–- cant contribution from the MgO substrate is still noticeable at maximum Mg coverages.In addition the valence band of the clean MgO surface reveals a two peak structure between 4 and 10 eV which is discussed in detail ref. 24. However the most obvious diÜerence is the absence of the Mg-induced band gap feature in UPS; only a small intensity increase in the binding energy range up to 4 eV is observed at high Mg coverages. This can be attributed to the fact that unlike MIES UPS probes not just the surface layer but the rather deeper layers and in addition is very insensitive to metallic s states.26 A similar band gap feature was observed in the MIES spectra after dosing with Na atoms; however it is less stable thermally and disappears between 350 and 400 K upon annealing. Li and Cs additives on the MgO surface also demonstrate similar band gap features.21 2.3 Temperature programmed desorption TPD is an excellent technique for determining surface coverage and studying the interaction between adsorbates themselves and with the surface.Fig. 4 displays TPD and MIES spectra acquired from the Mg-covered MgO surface at the same Mg exposure. Starting at the uppermost MIES spectrum acquired for the clean MgO surface the Mg exposure increases monotonically towards the bottom spectrum. In the MIES spectrum for the highest Mg coverage (bottom spectrum) the O 2p band has essentially disappeared indicating complete coverage of the MgO surface by Mg. The relative peak area of the corresponding TPD Mg feature (m/z\24) is 264 times larger than the Mg feature in the uppermost TPD spectrum which corresponds to the Fig.3 Intensity energetic position width (FWHM) of the band gap feature vs. time. Faraday Discuss. 1999 114 173»194 178 Fig. 4 Comparison between TPD and MIES data. The TPD spectra were taken from the surface characterized with MIES. The topmost MIES spectrum shows the clean MgO surface ; the Mg coverage increases monotonically towards the bottom spectrum. detection threshold of our mass –lter. However even at this very low coverage the Mg-induced band gap feature appears fully developed in MIES. Analysis of the Mg TPD peaks reveals an exponential increase in intensity towards higher temperature with increasing coverage (leading edge behaviour) indicating that the desorption follows zeroth-order kinetics.Since this behaviour is typically observed for desorption with scission of adsorbate»adsorbate bonds this result suggests that even at the lowest Mg concentrations accessible to TPD the formation of 3D islands occurs. Layer-by-layer growth would lead to a Fig. 5 Arrhenius plot of the desorption rates obtained in the complete analysis28 of the TPD data in Fig. 4. Inset calculated desorption activation energy vs. relative Mg coverage. 179 Faraday Discuss. 1999 114 173»194 two-peak structure in TPD (as is actually observed for Na/MgO at 100 to 150 K27). The formation of 3D islands is also supported by a non-linear Mg uptake vs. exposure time. The TPD data have been analyzed using the so called complete analysis28 in which an Arrhenius plot of the natural logarithm of the desorption rates yields a straight line for a particular Mg coverage (see Fig.5). The slope of each line corresponds to the activation desorption energy Edes for that particular coverage. The inset in Fig. 5 shows that the desorption energy increases initially from about 1.0 eV at very low coverages up to 1.4 eV. This latter value is close to 1.45 eV the heat of sublimation for bulk Mg. Even at the lowest exposure accessible for TPD the band gap feature in MIES is fully developed and the work function has traversed through its minimum. The band gap feature as a function of exposure smoothly transforms into the spectrum characteristics for the full metallic coverage. Our TPD results suggest that this feature is due to Mg clusters which for sufficiently large exposures acquire metallic properties.Because the same kind of feature is already present in MIES spectra at much lower exposures it is reasonable to assume that at these low exposures MIES also detects the presence of clusters at the surface. Thus comparison of Figs. 4 and 5 also supports the formation of Mg clusters with Mg»Mg bond strength similar to that in the Mg bulk metal. 3 Theoretical results 3.1 Theoretical models Before we go into a detailed description of the theoretical methods and the results of calculations let us brie—y summarise the experimental data. Essentially we are presented with two types of data. The TPD results tell us that the adsorbed Mg clusters have quite large desorption energies of individual Mg atoms which increase with the coverage.The MIES spectra demonstrate a feature that evolves with the Mg concentration transforming at large Mg exposures into the spectrum characteristic of full metallic coverage. To construct an atomistic model of adsorption of Mg atoms and further growth of metallic clusters on the MgO –lm surface we can also use the STM images29 of MgO –lms grown on Mo(001). They demonstrate that these surfaces are very rough and contain 3D MgO islands and a lot of steps within the islands. Such surfaces would normally also have a number of anion and cation vacancies. The anion vacancies can be –lled by electrons from the metal substrate forming charged or neutral F-centres. The latter can be ionised directly during the AD process or by trapping electron holes ; in addition holes can localise near cation vacancies (forming the well known stable V-centres).In this study we assumed that the Mg atom adsorption takes place on terraces near charged anion vacancies neutral F-centres and at step edges. Comparing the calculated Mg atom adsorption energies with the TPD data we can approximately deduce which sites are most favoured. The spectroscopic MIES and UPS data though require a much more complex analysis. The MIES spectra of the MgO –lms before Mg dosing contain information about their electronic structure. In particular the largest kinetic energy of emitted electrons corresponds to the AD process which involves electrons from the highest occupied states localised at the top surface layer which we call for simplicity ììthe top of the valence bandœœ.The position of the top of the valence band with respect to the vacuum determined from the MIES data is 6.7^0.4 eV. We can calculate the lowest ionisation energy of the surface terrace and compare it with these data. Another prediction that one can deduce from the experimental results is that the top of the valence band and the 3s states of Mg atom have close energies. This suggests formation of partially covalent bonding of Mg atoms to the surface which can also be checked theoretically. More detailed analysis of MIES spectra is less straightforward.30 In the AD process which is the only one that we took into account in the present paper a surface electron is transferred into the 1s hole state of the He* 1s2s atom and the excited He 2s electron is ejected.Since only one surface electron is involved in the process it is often assumed that the AD MIES spectra to a good extent re—ect the SDOS.30,31 More accurate static30,32 and dynamic33h35 theoretical models suggest that like in the TersoÜ and Hamann model of STM,36 the density of states projected on the 1s function of the He* atom and integrated over the incoming trajectory of that atom would better represent the probability of the electron tunneling from the surface to the He*. Our Faraday Discuss. 1999 114 173»194 180 calculations31,37 have demonstrated that the shape of the experimental MIES spectrum of MgO is diÜerent from the bulk and the surface DOS and is indeed similar to the SDOS projected on the 1s He* function.However the relative energies of diÜerent features in the three DOS and in the experimental spectra remain very similar. Therefore we believe that the MIES features due to localised band gap states of adsorbed atoms and their relative position with respect to the top of the valence band are likely to be well reproduced by SDOS. 3.2 Theoretical methods Density functional theory38 is widely used in surface studies (see for example ref. 39). Two models within the DFT method have been used in our work. Due to the importance of the electron correlation for the hole states we employed the DFT to calculate the surface ionisation energies within an embedded cluster model.16,17 To model the Mg adsorption and MIES spectra a periodic model and the realisation of the DFT in the VASP code40h43 are more appropriate.In the embedded cluster calculations a cluster of up to 50 atoms (quantum cluster) was treated quantum mechanically using DFT. It was embedded into a –nite cluster (region I) of 12]12]6 ions treated in a polarisable ion model. Pair potentials44 were used to calculate the interactions between these ions and the shell model45 to treat the polarisable oxygen ions. The quantum cluster and region I were embedded into an outer region of frozen ions which makes the total number of ions in the system 20]20]8. The eÜective charges on all classical ions were ^2e (e is the electron charge). This setup provides the correct values for the Madelung potential and its gradients on the ions in region I and in the quantum cluster.As an example in Fig. 6 we present the largest quantum cluster Mg37O13 . The matrix elements of the electrostatic potential of the rest of the system including the dipole contributions from the polarized oxygen ions in region I are included in the Kohn»Sham equations implemented in the modi–ed Gaussian94 code.46 The B3LYP functional47 was employed to calculate the electronic structure of quantum clusters. All electrons of oxygen ions and those of the magnesium ions (shown as black in Fig. 6) were described using the 6-31G standard Gaussian basis set.48 To facilitate calculations of large quantum clusters the magnesium ions (shown as gray in Fig. 6) were treated using the pseudopotentials of Wadt and Hay49 and the 1s function described by two contracted Gaussians.In smaller clusters Mg atoms that had less than two nearest quantum oxygens were treated in the similar way. Periodic DFT calculations were performed in the slab geometry where an in–nite stack of slabs separated by vacuum gap (to suppress the mutual interaction between parallel slabs) is considered in the z-direction. Within each slab the unit cell is periodically repeated in two dimensions. The ionisation energy. Open circles represent oxygen ions black and gray circles are magnesium ions with diÜerent Fig. 6 The top view on the Mg37O13 quantum cluster used in the embedded cluster calculations of surface basis sets (see text). 181 Faraday Discuss. 1999 114 173»194 Kohn»Sham electronic orbitals t are expanded into plane waves and the plane wave coefficients nk are varied to obtain the minimum of the energy for the given ionic geometry.A number of unoccupied states are also included in the variational procedure as it speeds up the calculations and at the same time allows us to consider possible metallisation in the system. The whole system is brought to a mechanical equilibrium by minimising the forces acting on every atom in the unit cell until they become less than 0.1 eV ”~1. Only valence electrons are treated explicitly which is achieved by using non-local ìì soft œœ Vanderbilt pseudopotentials.50,51 The advantage of using these pseudopotentials instead of the norm-conserving ones38 is that it is possible to have a relatively E small cutoÜ cut\400 eV (especially for oxygen) which speeds up the calculations by at least a factor of four.The generalised gradient approximation (GGA)52 was used in all calculations which is especially important for surfaces.39 The cell sizes vacuum widths and system geometries for all systems studied in this paper are discussed in detail below. In all calculations the interionic distance of d0\2.122 ” was used to specify the surface unit cells. This was found in ref. 53 to be the equilibrium distance for the bulk MgO using the same GGA functional as in the present study. The calculations were performed in the following way. First the singlet ground state of the reference system (perfect surface step etc.) without adsorbed metal atoms was considered. After it was relaxed to mechanical equilibrium and Mg atoms were introduced into the calculation.The whole system was relaxed again except for the atoms in the bottom layer of the slab which were –xed in the positions corresponding to the perfect system to simulate the crystal bulk. From two to four k-points in the plane (2D) Brillouin zone have been normally used in all such ground-state calculations. This has been shown54,55 to be sufficient for the cell sizes considered here. The ground state calculations give adsorption energies relaxed geometries and the electronic densities. The latter were analysed using the general visualisation tool LEV00,56 which facilitates construction and analysis of arbitrary 3D objects speci–ed on a grid such as electronic densities and wavefunctions. In addition to the usual contour maps of for example electron density we have also found it very informative to integrate the charge density into spheres of diÜerent radii around various positions within the simulating cell and to compare these results with those obtained for other positions of the same or other similar system (cf.ref. 54). The densities of states were calculated using a method of tetrahedra outlined in ref. 54. Brie—y using the point-group symmetry of the cell a necessary mesh of k-points was generated for the plane Brillouin zone and the wavefunctions and energies of the surface electrons were recalculated again for all non-equivalent k-points using the VASP code. In most cases our systems have no symmetry at all and a mesh of 13 k-points corresponding to 750 tetrahedra in the plane Brillouin zone was used.Then the SDOS was calculated using LEV00. It is smeared by a Gaussian to simulate the eÜect of phonon broadening57 at room temperature using a smearing parameter equal to 0.3 eV. All periodic calculations were performed on the T3E parallel supercomputers in the Edinburgh Parallel Computer Center and at the University of Manchester under the Computer Services for Academic Research (CSAR) initiative. 3.3 Calculation of the surface ionisation energy To calculate the surface ionisation energy we used several quantum clusters of increasing size as listed in Table 1. Four of them had quantum oxygens only in the surface plane and in the case of Mg29O13 four additional oxygen ions were added to the Mg25O9 cluster in the second plane to check whether this will have any signi–cant eÜect.After calculation of the perfect lattice each cluster was ionised and the diÜerence in the total energy with the ionized state was calculated in two approximations with IP(I) and without IP(0) a self-consistent account of the electronic part of the polarisation in region I which corresponds to the ìì vertical œœ ionisation potential. Only the oxygen polarisation was included and treated classically in the shell model. As the hole delocalisation increases one would expect the polarisation energy *IP to decrease. Except for the smallest cluster which contains only one oxygen ion in all the ionised clusters the hole was delocalised by all quantum oxygen ions. Increasing the cluster size we should approach the limit of a completely delocalised band hole state.For the smaller clusters the hole was distributed almost evenly by all the oxygen ions. However in the case of the Mg37O13 this distribution was more complex which re—ects the fact that as the hole state becomes more delo- Faraday Discuss. 1999 114 173»194 182 Table 1 The ionisation energies calculated using diÜerent quantum clusters in eV Cluster IP(I)b IP(0)a W d *IPc 2.11 1.03 0.77 0.71 0.51 9.75 7.89 7.29 7.18 7.03 0.68 2.02 2.26 2.82 3.13 7.64 6.86 6.52 6.47 6.52 Mg5O1 Mg17O5 Mg25O9 Mg29O13 Mg57O25 a In the calculation of IP(0) the lattice polarisation outside the quantum cluster was not included. b IP(I) are calculated taking into account the classical lattice polarisation.c *IP is the diÜerence between the two which re—ects the hole localisation and consequently the lattice polarisation energy. d W is the valence band width. calised its eigenvalue approaches the top of the valence band. As a result several quasi-degenerate hole distributions become possible which also hampers the convergency of calculations. As one can see in Table 1 this however does not signi–cantly aÜect the calculated energies. The ionisation energies in both approximations –rst decrease sharply as the cluster size increases and then change slowly. The diÜerence in ionisation energies between the completely localised hole state in the smallest quantum cluster and the most delocalised state in the largest cluster should approximately correspond to half of the valence band width W which is also presented in Table 1.This approximately holds in our calculations. 3.4 Adsorption of Mg atoms on the perfect MgO (001) surface and near a surface F-centre A supercell consisting of three layers of oxygen and Mg atoms (eight surface unit cells in every layer) with the vacuum width between slabs equivalent to three additional layers was used to simulate the perfect MgO(100) surface. Up to four Mg atoms were added to this supercell to model the Mg adsorption. A similar setup was used to model the surface F-centre. The latter was created by removal of one oxygen atom from the topmost surface layer in the slab. The whole system was relaxed and its energy was used as the reference energy in the calculations of the corresponding adsorption energies.A detailed account of the surface F-centre calculations with the same method is given in ref. 54. In this paper we focus on the results related to Mg adsorption. The adsorption energies of one Mg atom on the perfect surface and near the surface F-centre are shown in Table 2. On the regular surface the most stable position of the Mg atom is above the surface oxygen. The on top F-centre position is much less stable and although there is a shallow energy minimum corresponding to the adsorption energy shown in Table 2 most of the Mg atoms would probably prefer the nearest oxygen sites to F centres. On top of the F-centre the Mg atom is pulling up part of the electron density from the vacancy.Above the oxygen site the adsorbed Mg atom is about 2.3” ” above the surface plane with the oxygen ion displaced by about 0.25 towards it. No signi–cant relaxation of other surface ions was found. The calculated barrier for Table 2 Total adsorption energies for all systems studied in eV 5 Mg 4 Mg 3 Mg 2 Mg 1 Mg Surface 4.05 2.84 2.51 4.98 3.64 1.69 2.00 3.62 2.35 0.96 0.71 3.05 1.25 0.51 0.21 1.26 0.51 Perfect surface With F-center With step With anion vacancy 183 Faraday Discuss. 1999 114 173»194 diÜusion of an Mg atom along the perfect surface is only 0.26 eV with the barrier point at about 3.2 ” above the centre of the surface unit cell. The equilibrium distance between Mg atoms in a free Mg molecule obtained in our calcu- 2 lations is 3.75 ” which is much larger than the distance between two nearest oxygen ions and smaller than that between next-nearest ions.Therefore when more than one Mg atom is added to the system their lateral interaction does not allow them to occupy the most energetically favourable positions above the oxygen ions. As a result we –nd that the potential energy surface in the lateral direction above the surface is very —at with many local minima. Typical geometries for adsorption of four Mg atoms on the perfect surface and near the F-centre are shown in Fig. 7(a) (b). As one can see the geometries obtained are similar in both cases. Every Mg atom occupies a surface area containing one surface oxygen.It is positioned above the surface plane in the range 2.2»3.2 ” depending on the particular arrangement of the Mg atoms and on whether the nearest surface oxygens are displaced signi–cantly towards them from the surface plane. The adsorption energies found for the two systems are summarized in Table 2. When two Mg atoms are added to the terrace the adsorption energy per adsorbed Mg atom does not increase more than twice but is actually a little smaller than the double adsorption energy for one Mg due to the repulsion of Mg atoms. Then it grows slowly reaching 0.7 eV Fig. 7 Typical geometries for the adsorption of four Mg atoms on the terrace (a) and near the F-centre on the —at surface (b) ; for one (c) and two (d) Mg atoms at the anion vacancy. Oxygens are shown as open balls surface Mg atoms as shaded smaller balls and adsorbed Mg atoms are shown as black circles.Faraday Discuss. 1999 114 173»194 184 atom~1 when four atoms are added (in this comparison we are using an averaged parameter that corresponds to the dissociation of adsorbed cluster into free atoms). A similar tendency is also observed for the surface with the F-centre. This is due to the creation of mutual bonding between the adsorbed Mg atoms and between them and the surface oxygens. In order to demonstrate the character of this bonding we show in Fig. 8 the contour plot of the valence electronic density of the system with one Mg atom adsorbed on the terrace above a surface oxygen. The strong contribution of the electron density from the surrounding oxygen ions into that of the adsorbed Mg atom is clearly visible.To analyse the bonding further one can integrate the charge density around the adatoms. For one adsorbed Mg atom on the perfect surface this does not indicate any signi–cant charge transfer to the surface. However a detailed analysis of the occupied orbitals in this system reveals that the adatom participates in the states at the top of the O 2p valence band (VB) whereas there is a signi–cant contribution of the surface oxygens nearest to the adatom in the charge density localised on the Mg atom. This picture can be further clari–ed by examining the calculated DOS for this system shown in Fig. 9 (the lowest curve). The last occupied state n manifests itself in the DOS as a feature about 1.2 eV from the o valence band maximum.The corresponding partial density n(r)\&k ownk o2 contains contributions from both the adsorbed Mg atom and the surface oxygens underneath. While integrating on(r) in spheres of diÜerent radii we –nd that it does not account for the total charge around the adatom so that part of the density comes from the valence band states. On another hand there is a considerable localisation of this density on the nearest surface O atoms. When more than one Mg atom is adsorbed the electron density is easily shared between them. In all cases we found a rather diÜuse density around the adatoms with strong highly localised peaks on the nearest surface oxygens. In the DOS these mixed states manifest themselves as a set of peaks in the gap which (after smearing) show up as a broad peak around 2 eV above the VB maximum.One can also notice a considerable distortion of the O 2p valence band due to the Mg adsorption this is seen as a bump at about [1.5 eV. on(r) associated with the features With three Mg atoms per simulation cell the partial densities in the gap are still well localised. Adsorption of four Mg atoms in our setup corresponds to half of the monolayer. In this case the last occupied state in the system is very diÜuse and is spread over most of the simulation cell. This state is mainly due to all four adsorbed Mg atoms and the surface oxygens nearest to them. In the DOS we –nd that this state has a considerable width (of over 2 eV) and overlaps both with other adatom-related local states at lower energies and with the Fig.8 The contour plot of the valence electronic density of a Mg atom adsorbed above a surface oxygen atom on the terrace (in units of 10~2 electron ”~3). To guide the eye the surface atoms are connected by a dashed line. The cut has been made along [010] axes perpendicular to the surface plane. To avoid high peaks on oxygens the density has been chopped at 0.2 electron ”~3. Distances are in ”. 185 Faraday Discuss. 1999 114 173»194 Fig. 9 DOS (arbitrary units) for the perfect surface with up to four adsorbed Mg atoms. The DOS is aligned so that zero energy corresponds to the unsmeared top of the VB. unoccupied states. Although strictly speaking the unoccupied states in the DFT do not have clear physical meaning we believe that this behaviour is an indication of the beginning of the system metallisation.This is because with four atoms it is already possible to construct a con–guration of adsorbed metal atoms in which the distances between the neighboring Mg atoms in the central and adjacent cells are of the same order of magnitude. Then some of the occupied states become very diÜuse in the direction of the short distance between the metal atoms and the system as a whole becomes conductive. This eÜect is similar to the one in the percolation theory of conductance in disordered systems. Although some features of the DOS for the system with the F-centre are diÜerent it nevertheless retains the same character. In particular in the DOS for one adsorbed Mg atom (see Fig. 10 the lowest curve) there are two peaks in the gap at 0.9 and 2.6 eV above the VB maximum.They correspond to the bonding and antibonding states between the Mg atom and the F-center electrons. These states also contain a signi–cant portion of the density localised on the nearest surface oxygens. Every new atom added to the system generates an additional peak in the gap of the DOS. After smearing (see Fig. 10) all these peaks form a broad feature in the gap approximately 2 eV above the VB maximum. The states that make up the defect band are quite diÜuse. They spread over all Mg atoms (some states have bonding some an antibonding character with respect to the adatoms) and have signi–cant localisation in the anion vacancy and especially on the nearest surface oxygens.While integrating the partial density associated with the states that form the broad feature in the gap we have not been able to account for all the density which means that the electrons from the VB themselves have signi–cant localisation on the adsorbed atoms. The VB states therefore are strongly perturbed. This manifests itself in the distortion of the O 2p band seen in Figs. 9 and 10. Similar to the perfect surface we found that metallisation starts to form when four or –ve Mg atoms are added to the supercell with the surface F-centre the defect states in the gap interact more strongly and are spread over larger energy intervals so that they overlap with unoccupied states. Thus we conclude that upon adsorption of Mg atoms chemical bonding is formed between the Mg atoms and that at a coverage of roughly half a monolayer the adsorbed layer may become conductive.Faraday Discuss. 1999 114 173»194 186 Fig. 10 DOS (arbitrary units) for the perfect surface containing one neutral F-center per simulation cell with up to –ve adsorbed Mg atoms. The DOS is aligned so that zero energy corresponds to the unsmeared top of the VB. 3.5 Adsorption of Mg atoms near the anion vacancy Adsorption of Mg atoms near the F-center is similar to that on the perfect surface partly because the F-center bears roughly the same charge (two electrons) as the lattice oxygen O2~ ion. A doubly positively charged anion vacancy Va may interact diÜerently with adsorbed metal atoms. However modelling of charged systems in the periodic model is less straightforward.Formal procedures (see for example refs. 58 and 59) which one can use to study charged systems in periodic boundary conditions are not strictly applicable to surfaces and to slab geometries. In the bulk one can remove the Coulomb interaction energy between the extra charge across the simulation cells by dividing it by the relative permittivity e0 . It is not that clear however how to model the electronic polarisation in the surface case. Therefore in this paper we adopted a diÜerent procedure. It is well known that in real systems charged defects tend to be compensated by other defects having the opposite charge. To compensate the anion vacancy at the slab surface we formed a cation vacancy Vc at the slab centre. This makes the whole system neutral at the expense of introducing a dipole moment in the cell.The dipole moment is going to be large both along the surface and normal to the surface as we want to separate the two vacancies from each other as much as possible. To check the eÜect of the dipole moment on the energetics geometry and the DOS we have run extensive tests. They have demonstrated that the eÜect on the energetics and geometries of adsorption is insigni–cant and leads only to a shift as a whole of the calculated DOS. Therefore we used this setup and the 80-atom supercell modelling a –ve layer slab in further calculations of the Mg adsorption. The cation vacancy was created in the middle layer of the slab and the oxygen ion was removed in the top layer from the lattice site which is most separated from the cation vacancy two layers underneath.The distance between the two vacancies is 3d0\6.4 ” in all calculations. After the geometry relaxation this system was treated as the reference for further calculations of the Mg atomœs adsorption. No signi–cant electron density is localised in the anion vacancy. One Mg atom is adsorbed between the vacancy and the nearest surface oxygen as shown in Fig. 7(c). The adatom is located 2.4 ” above the surface plane and there is a considerable upward displacement of the nearest oxygen atom. The analysis of the electronic density revealed that two 187 Faraday Discuss. 1999 114 173»194 electrons of the Mg atom are strongly pulled towards the vacancy and form a diÜuse electronic cloud localised in the area containing both the Mg atom and the anion vacancy.Because of this direct charge transfer from the adatom to the surface the adsorption energy is more than 0.5 eV (see Table 2) which is much greater than in the case of adsorption on the F-centre. In the DOS for this system shown in Fig. 11 (the lowest curve) one can notice a feature just above the VB top which is due to this single diÜuse state. Because the wavefunction associated with this state penetrates more into the surface than in the other two cases studied above the extra peak in the gap was found only 0.35 eV above the VB maximum and after smearing appears as a shoulder in Fig. 11. When one more Mg atom is added to the system at least two con–gurations are possible. If the Mg atom –nds a surface oxygen within the proximity of the anion vacancy it shares its electrons with the anion vacancy and the Mg atom already adsorbed nearby.The total adsorption energy (Table 2) is more than doubled. This con–guration is depicted in Fig. 7(d). Another possibility is that the second Mg atom adsorbs further away from the vacancy. In this case the adsorption will happen on a terrace above one or two surface oxygens and the adsorption energy increases only up to 0.81 eV which is less than one would expect from the results for the perfect surface. The corresponding DOS for these two con–gurations shown in Fig. 11 look very similar after smearing and one can notice the development of a feature about 1 eV above the VB maximum. Addition of more Mg atoms leads to the formation of mutual electronic states between them and with the surface and to a substantial gain in adsorption energy as seen in Table 2.In the DOS shown in Fig. 11 (two upper curves) one can clearly see the development of a defect band around 1 eV above the VB maximum. It is closer to the VB edge because the defect is positively charged. 3.6 Adsorption of Mg atoms near a monolayer step Finally let us turn to the Mg adsorption at a monolayer step. The simulation cell used in these calculations contained 44 lattice sites (for 22 oxygens and 22 Mg atoms) arranged in three layers as in ref. 54. Each layer as can be seen in Fig. 12(a) goes like a d -high staircase containing a 0 3d -long and in–nitely wide terraces. One Mg atom is adsorbed just in front of the step facing two 0 Fig.11 DOS (arbitrary units) for the surface containing one anion vacancy (compensated by a cation vacancy in the middle of the slab) per simulation cell with up to four adsorbed Mg atoms. The DOS is aligned so that zero energy corresponds to the unsmeared top of the VB. Faraday Discuss. 1999 114 173»194 188 Fig. 12 Three-layer slab system used in the study of the Mg adsorption at the monolayer step (a) the geometry for one (b) and four (c) Mg atoms adsorbed at the step. Notations as in Fig. 7. surface oxygens as shown in Fig. 12(b) with the substantial energy gain of 1.26 eV. This is the biggest adsorption energy we obtained for a single Mg atom on the MgO surface. The two oxygens nearest to the adatom slightly displace towards it (by about 0.03 d0).However the displacements of the nearest surface Mg ions (up to 0.05 d0) shown by arrows in Fig. 12(b) are more Fig. 13 DOS (arbitrary units) for the monolayer step system with up to four adsorbed Mg atoms. The DOS is aligned so that zero energy corresponds to the unsmeared top of the VB. 189 Faraday Discuss. 1999 114 173»194 o Fig. 14 The contour plot of the partial charge density n(r) associated with the defect state in the gap for a single Mg atom adsorbed at the step. The cut has been made perpendicular to the direction of the step through the adsorbed Mg atom and the two nearest surface oxygens. To guide the eye broken lines indicate the surface structure. Other notations are as in Fig. 8. substantial. This con–guration is very stable the system total energy is by about 1 eV higher if the Mg atom is adsorbed on the terrace or just above the center of the unit cell on the step edge.The DOS shown in Fig. 13 (the lowest curve) demonstrates a single peak at about 2 eV above the top of the VB. It is made of the orbitals of the adatom and the two surface oxygens nearest to it. The partial density on(r) associated with this state is rather diÜuse on the adatom but forms sharp peaks on the surface oxygens as shown in Fig. 14. The step simulation cell contains two equivalent positions at the step edge (see Fig. 12(b)). It is therefore not surprising that the second Mg atom prefers to stick at this position as well. The adsorption energy presented in Table 2 increases over 1.5 eV atom~1.The band at 3 eV above the VB in the DOS (see the second curve from the bottom in Fig. 13) becomes broader. However as the third and the fourth Mg atoms are adsorbed no such positions are available in the simulation cell and the additional Mg atoms are forced to occupy the terrace sites. The typical geometry for four adsorbed Mg atoms is shown in Fig. 12(c). Nevertheless the adsorption energy increases substantially (see Table 2). One can also notice that a second defect band in the DOS around 2 eV above the VB maximum is developed due to Mg adsorption on the terrace sites. This band consists of several states equal to the number of adsorbed Mg atoms. Note that the states responsible for the defect features in the gap of the DOS are localised in the direction perpendicular to the step.However due to a relatively small size of simulation cell along the direction of the step the states in question are delocalised in this direction. Nevertheless we have not found any signs of the metallisation at this coverage. 4 Discussion Let us start with a brief discussion of theoretical results. First we note a qualitative agreement between our results for the Mg adsorption on the MgO surface terrace and those obtained by Musolino et al.12 for adsorption of Cu (n\1 . . . 4) on MgO using a DFT based method. n Interestingly some of the calculated geometries and adsorption energies for clusters and also the diÜusion parameters for one Cu atom on the surface are close even quantitatively to our results for the Mg adsorption despite the diÜerence in the Cu and Mg electronic structures.The results,12 though demonstrate a richer variety of adsorption con–gurations which we did not fully explore in this study. A qualitative agreement also exists with the embedded cluster DFT calculations of M clusters (M\Cu Ag Ni Pd) on MgO by Matveev et al.13 Both the results of Matveev et al. 4 Faraday Discuss. 1999 114 173»194 190 and our calculations predict polarisation of adsorbed metal atoms (see Figs. 8 and 14). However our analysis suggests a more covalent character for chemical bonding of Mg clusters with the 4 MgO surface than suggested for other metals in ref. 13. We start our comparison with the experimental data from the perfect surface. Although the results of calculations for the surface ionisation energies (see Table 1) are in good agreement with the position of the top of the valence band with respect to the vacuum experimentally determined from the MIES data (6.7^0.4 eV) this agreement is not conclusive.Our results do not show a fast convergency with the quantum cluster size. Although its further increase is not feasible we believe that the values given in Table 1 are already representative. Thermal —uctuations and the surface roughness broaden signi–cantly the band edge which is re—ected in the UPS spectra (see Fig. 2) and in the experimental error in the determination of the position of the top of the valence band from the MIES spectra. This broadening masks the ìì electronic œœ band edge which can be obtained as a limit in our calculations (see for example ref.60). Similar to refs. 18 30 31 we made the assumption that the MIES spectra due to the AD process re—ect to a good approximation the SDOS. This in fact does not hold for the states in the lower part of the valence band and those below the valence band. As one can see in Figs. 9»11 the calculated surface DOS has a pronounced maximum at these energies which is almost completely absent in Fig. 2(a). This is because the wavefunctions of lower energy states decay faster into the vacuum than those of the higher energy states as discussed in detail elsewhere.37 Nevertheless we believe that the position of the defect states in the band gap with respect to the top of the valence band can be reproduced more reliably. The comparison of the numerical results and the experimental MIES spectra thus suggests that the band gap feature is due to adsorbed Mg atoms and small Mg clusters.The SDOS calculated for the Mg adsorption on the terrace near the F-center and at the charged anion vacancy predict a shoulder at about [1.5 eV (see Figs. 9»11). However despite the fact that the UPS spectrum for the perfect surface is well reproduced by the SDOS such a shoulder is not seen in the UPS spectra after the Mg exposure shown in Fig. 2(b). They also do not demonstrate any visible band gap features. We attribute this to the fact that UPS probes mostly deeper surface layers which are not aÜected by the Mg adsorption. Another reason is that UPS is very insensitive to metallic s states.26 Our theoretical results suggest that individual Mg atoms adsorbed on terraces are fairly mobile at room temperature (the calculated adiabatic barrier for diÜusion is only 0.26 eV).The largest adsorption energies were obtained for individual Mg atom adsorption at the step edge. These results indicate that at very low coverages one can expect more Mg atoms to be adsorbed at step edges than at terraces. However as the Mg exposure increases the energy gained due to attachment of each atom to existing clusters on the terrace also increases (see Table 2). So for instance to desorb one atom from the four atom cluster on the terrace requires 1.15 eV and from the terrace close to the step edge 1.36 eV (see Fig. 12(c)). As one can see in Fig. 5 the desorption activation energies as derived from the TPD measurements increase with the Mg exposure from about 1.0 eV at relatively low coverages to 1.4 eV for almost metallised surface.As discussed in Section 2.3 the TPD data suggest that these energies correspond to desorption of individual Mg atoms from Mg clusters and metallic layer. Although it is tempting to directly compare the theoretical results with the TPD data this is impossible due to the large variety of both adsorption sites and cluster geometries. Since TPD measures the smallest desorption energies comparison with these data without proper simulation of desorption kinetics can be only qualitative. It suggests that the calculated desorption energies of individual Mg atoms are certainly in the range of desorption energies determined from TPD.Based on these results it seems plausible to assume that –rst the metal atoms occupy all available sites immediately in front of the step edges (decoration of MgO islands or clusters). They form chemical bonds with the nearest surface oxygens accompanied by the considerable lateral interaction between the adsorbed Mg atoms. This assumption is supported by comparison of the dependence of the calculated SDOS on the character of Mg adsorption with the MIES spectra. According to the calculations when Mg atoms are adsorbed at step edges this should result in the development of the defect band in the band gap about 2.5»3 eV above the VB maximum (see Fig. 13). As all such sites are occupied by the adsorbed atoms the nearby terrace sites also become gradually –lled.A band about 2 eV above the top of the VB should develop due to newly adsorbed Mg atoms (see Fig. 9). Since the number of terrace sites is much bigger than that of the 191 Faraday Discuss. 1999 114 173»194 edge sites the peak at 3 eV above the VB should soon become less visible. This implies that the Mg related peak in the MIES spectra should shift to lower energies as the Mg concentration increases. Careful analysis of the experimental spectra demonstrates that this indeed is the case. It is interesting to note a new structure formed by Mg atoms adsorbed at the step edge (see Fig. 12(c)) the oxygen vacancy created by the three adsorbed atoms A»C»B and the step Mg ion. The electron density plot shown in Fig. 15 clearly shows a considerable localisation of the electron density in the ìpocketsœ created by the Mg atoms and in the ìvacancyœ.The band gap feature corresponding to these states is seen in Fig. 13. In principle a band gap feature of similar shape and energetic position could also result from Auger de-excitation of point defects F-centers in particular. However no indication of this feature is seen at the clean surface (which certainly is not free of point defects). We also note that because of the proximity of the metal substrate and also because the position of its Fermi level is several eV above the top of the MgO VB one can expect that in our experiments all electronic traps such as anion vacancies and hole centres will be quickly –lled by electrons tunnelling from the metal. The speed of this should depend though on the –lm thickness.This will not be the case for other types of experiments e.g. on single crystals. Our results for Mg/MgO allow for the following more general qualitative considerations. One of the key parameters responsible for the type and strength of bonding between metal species and oxide surface is the ionisation energy of a metal atom with respect to the valence and conduction bands of the oxide surface. Provided the valence level of the metal atom is in resonance with the occupied valence band states the bonding will feature covalent and polarisation contributions. This holds in the present case of Mg/MgO and for other metals on MgO with an ionization potential of about 7 eV and larger. On the other hand if the metal level is in resonance with unoccupied states of the conduction band as is the case for alkali atoms adsorbed on TiO2 a charge transfer from the alkali metal atoms to the oxide can be expected.For the alkali/TiO case this leads to the reduction of the Ti 2 cation (and the appearance of band gap states due to Ti3` 3d formation). Thus there is ionic chemisorption in this and similar cases. MIES studies on such systems are currently in progress at TU Clausthal. In cases like Mg/TiO2 both charge transfer from the metal atom to the cation and hybridisation of the Mg 3s and O 2p states might occur simultaneously resulting in mixed covalent and ionic bonding. At present we are attempting to verify this prediction in a joint MIES and UPS study on Mg/TiO2 .Fig. 15 The contour plot of the valence charge density for the step system with four adsorbed Mg atoms (labelled) as shown in Fig. 12. The cut is made parallel to the terrace in such a way that all four adsorbed atoms are crossed (note that they are not exactly at the same height). Broken lines indicate positions of the atoms at the upper terrace. The adsorbed atoms are in front of this terrace. Other notations as in Fig. 8. Faraday Discuss. 1999 114 173»194 192 Finally we note that the presence of extended surface defects steps in particular will most likely not change these qualitative considerations. However as our calculations for Mg/MgO indicate they may aÜect the adsorption strength quite appreciably and therefore decide the adsorption kinetics. Acknowledgements LNK is supported by EPSRC.PVS would like to acknowledge the support by Kodak. This work has been performed in the framework of a joint collaborative project between University College London and the TU Clausthal Germany funded by the British Council (grant ARC 887) and DAAD. We are grateful to A. M. Stoneham and I. V. Abarenkov for useful discussions and to I. I. Tupitsyn and G. Tsikarishvili for help in calculations and to A. S. Foster for useful comments on the manuscript. We gratefully acknowledge the allocation of computer time on the Cray T3E provided by the High Performance Computer Initiative through the Materials Chemistry consortium. References 1 E. Giamello A. Ferrero S. Collucia and A. Tecchina J. Phys. Chem. 1991 95 9385. 2 D.Murphy and E. Giamello J. Phys. Chem. 1995 99 15172. 3 D. W. Goodman Chem. Rev. 1995 95 523. 4 C. Xu and D. W. Goodman Handbook of Heterogeneous Catalysis eds. G. Ertl H. Knoé zinger and J. Weitkamp VCH Weinheim 1997 vol. 2 p. 826. 5 C. T. Campbell Surf. Sci. 1997 387 136. 6 R. Souda Y. Hwang T. Aizawa W. Hayami K. Oyoshi and S. Hishita Surf. Sci. 1997 387 136. 7 T. Suzuki S. Hishita K. Oyoshi and R. Souda Surf. Sci. 1997 391 L1243. 8 T. Kizuka and N. Tanaka Surf. Sci. 1997 386 249. 9 E. Giamello and D. Murphy Mol. Eng. 1994 4 147. 10 Y. Kanzawa T. Kageyama S. Takeoka M. Fujii S. Hayashi and K. Yamamoto Solid State. Commun. 1997 102 533. 11 G. K. Wertheim Z. Phys. D At. Mol. Clusters 1989 12 319. 12 V. Musolino A. Selloni and R. Car J. Chem. Phys.1998 108 5044. 13 A. V. Matveev K. M. Neyman G. Pacchioni and N. Roé sch Chem. Phys. L ett. 1999 299 603. 14 A. M. Stoneham and J. H. Harding Proc. 9th W orld Ceramics Congress (CIMT EC) Florence June 1998 in press. 15 A. M. Ferrari and G. Pacchioni J. Phys. Chem. 1996 100 9032. 16 A. L. Shluger P. V. Sushko and L. N. Kantorovich Phys. Rev. B Condens. Matter 1999 59 2417. 17 A. L. Shluger L. N. Kantorovich A. I. Livsits and M. J. Gillan Phys. Rev. B Condens. Matter 1997 56 15332. 18 J. Gué nster J. Stultz S. Krischok D. W. Goodman P. Stracke and V. Kempter J. V ac. Sci. T echnol. A 1999 A17 1657. 19 W. Maus-Friedrichs M. Wehrhahn S. DieckhoÜ and V. Kempter Surf. Sci. 1990 237 257. 20 W. Maus-Friedrichs S. DieckhoÜ and V. Kempter Surf. Sci. 1991 249 149.21 D. Ochs M. Brause W. Maus-Friedrichs and V. Kempter J. Electron Spectrosc. Relat. Phenom. 1998 88ñ91 757. 22 P. A. Zeijlmans van Emmichoven P. A. A. F. Wouters and A. Niehaus Surf. Sci. 1988 195 115. 23 P. Eeken J. M. Fluit A. Niehaus and I. Urazgilœdin Surf. Sci. 1992 273 160. 24 D. Ochs W. Maus-Friedrichs M. Brause J. Gué nster V. Kempter V. Puchin A. Shluger and L. Kantorovich Surf. Sci. 1996 365 557. 25 P. A. Cox and A. A. Williams Surf. Sci. L ett. 1986 175 L782. 26 Electron Specroscopy. T heory T echniques and Applications ed. C. R. Brundle and A. D. Baker Academic Press London 1977»79 vol. 1»4. 27 S. Krischok and W. Goodman to be published. 28 D. A. King Surf. Sci. 1975 47 384. 29 M. C. Gallaher M. S. Fy–eld J. P. Cowin and S. A.Joyce Surf. Sci. 1995 339 L909. 30 Y. Harada S. Masuda and H. Ozaki Chem. Rev. 1997 97 1897. 31 D. Ochs W. Maus-Friedrichs M. Brause J. Gué nster V. Kempter V. Puchin A. Shluger and L. Kantorovich Surf. Sci. 1996 365 557. 32 H. D. Hagstrum Phys. Rev. 1954 96 336. 33 Z. L. Mis� kovicç and R. K. Janev Surf. Sci. 1986 166 480. 34 A. T. Amos K. W. Sulston and S. G. Davison Adv. Chem. Phys. 1989 76 335. 35 B. L. Burrows A. T. Amos Z. L. Mis� kovicç and S. G. Davison Phys. Rev. B Condens. Matter 1995 51 1409. 193 Faraday Discuss. 1999 114 173»194 36 J. TersoÜ and D. R. Hamann Phys. Rev. B Condens. Matter 1985 31 805. 37 L. N. Kantorovich A. L. Shluger P. V. Sushko and A. M. Stoneham Surf. Sci. submitted. 38 M. C. Payne M. P. Teter D. C. Allan T. A. Arias and J.D. Joannopoulos Rev. Mod. Phys. 1992 64 1045. 39 M. J. Gillan L. N. Kantorovich and P. J. D. Lindan Curr. Opin. Solid State Mater. Sci. 1996 1 820. 40 G. Kresse and J. Furthmué ller Comput. Mater. Sci. 1996 6 15. 41 G. Kresse and J. Furthmué ller Phys. Rev. B Condens. Matter 1996 54 11169. 42 G. Kresse Thesis Technische Universitaé t Wien 1993. 43 G. Kresse and J. Hafner Phys. Rev. B Condens. Matter 1993 47 RC558. 44 A. L. Shluger A. L. Rohl D. H. Gay and R. T. Williams J. Phys. Condens. Matter 1994 6 1825. 45 B. G. Dick and A. W. Overhauser Phys. Rev. 1958 112 603. 46 Gaussian94 (Revision E.1) G. W. Trucks H. B. Schlegel P. M. W. Gill B. G. Johnson M. A. Robb J. R. Cheeseman T. A. Keith G. A. Petersson J. A. Montgomery K. Raghavachari M. A. Al-Laham V. G. Zakrzewski J. V. Ortiz J. B. Foresman J. Cioslowski B. B. Stefanov A. Nanayakkara M. Challacombe C. Y. Peng P. Y. Ayala W. Chen M. W. Wong J. L. Andres E. S. Replonge R. Gomperts R. L. Martin D. J. Fox J. S. Binkley D. J. Defrees J. Baker J. P. Stewart M. Head-Gordon C. Gonzalez and J. A. Pople Gaussian Inc. Pittsburgh PA 1995. 47 A. D. Becke J. Chem. Phys. 1993 98 5648. 48 W. J. Hehre R. Ditch–eld and J. A. Pople J. Chem. Phys. 1972 56 2257. 49 W. R. Wadt and P. J. Hay J. Chem. Phys. 1985 82 284. 50 D. Vanderbilt Phys. Rev. B Condens. Matter 1990 41 7892. 51 G. Kresse and J. Hafner J. Phys. Condens. Matter 1994 6 8245. 52 J. P. Perdew J. A. Chevary S. H. Vosko K. A. Jackson M. R. Pederson D. J. Singh and C. Fiolhais Phys. Rev. B Condens. Matter 1992 46 6671. 53 L. N. Kantorovich M. J. Gillan and J. A. White J. Chem. Soc. Faraday T rans. 1996 92 2075. 54 L. N. Kantorovich J. Holender and M. J. Gillan Surf. Sci. 1995 343 221. 55 L. N. Kantorovich and M. J. Gillan Surf. Sci. 1997 374 373. 56 L. N. Kantorovich unpublished data. 57 G. D. Mahan Phys. Rev. B Condens. Matter 1980 21 4791. 58 M. Leslie and M. J. Gillan J. Phys. C Solid State Phys. 1985 18 973. 59 G. Makov and M. C. Payne Phys. Rev. B Condens. Matter 1995 51 4014. 60 G. K. Wertheim J. E. Rowe D. N. E. Buchanan and P. H. Citrin Phys. Rev. B Condens. Matter 1995 51 13675. Paper 9/03241J Faraday Discuss. 1999 114 173»194

ISSN:1359-6640    

DOI:10.1039/a903241j

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

A microcalorimetric study of the heat of adsorption of copper on well-de�ned oxide thin �lm surfaces MgO(100) p(2�1) oxide 195 on Mo(100) and disordered W oxide JeÜrey T. Ranney David E. Starr Jana E. Musgrove Dan J. Bald and Charles T. Campbell* Department of Chemistry University of W ashington Box 351700 Seattle WA 98195»1700 USA Introduction Calorimetric measurements of the heats of adsorption of reactant molecules or metal atoms on surfaces provide important thermodynamic data. Such data greatly enhance both our understanding of the thermodynamic driving forces which in—uence molecular surface transformations (i.e. catalytic reactions) and of the properties of adsorbed metal –lms (i.e. oxide-supported metal catalysts). Until fairly recently detailed studies of the heat of adsorption at low coverages on highly de–ned surfaces were not possible.However a new technique of single crystal adsorption calorimetry has been introduced as a powerful surface science tool for measuring the heats of adsorption on well-de–ned surfaces. This technique was –rst developed by King and coworkers1h5 and its applications have been reviewed.6,7 In this technique the heat generated by the adsorption of a pulse of gas containing as little as 1»2% of a monolayer (ML) can be measured by monitoring Faraday Discuss. 1999 114 195»208 Received 1st April 1999 The heats of adsorption as a function of coverage have been determined for copper adsorption onto several well-de–ned oxide thin –lm surfaces at room temperature by microcalorimetric measurements.The heats of adsorption are accurately determined as a function of coverage with resolution of 2% of a monolayer. For all three oxide surfaces investigated MgO(100) a p(2]1) molybdenum oxide –lm on Mo(100) and disordered W oxide the initial heat of copper adsorption is much lower than the heat of sublimation for Cu (337.4 kJ mol~1). On MgO(100) the initial Cu heat of adsorption in the –rst 2»4% of a monolayer is 240 kJ mol~1 and increases rapidly to the heat of Cu sublimation. Auger spectroscopy shows that Cu grows on MgO(100) as two-dimensional (2-D) islands until B0.3 monolayers where it switches to the growth of 3-D islands at which point the heat of adsorption of Cu reaches B92% of its heat of sublimation. The room temperature sticking probability of Cu on MgO was also investigated as a function of coverage and determined to be [0.99.On the ordered p(2]1) oxide of molybdenum on Mo(100) the initial Cu heat of adsorption is 287 kJ mol~1. The heat of adsorption then decreases slightly to 278 kJ mol~1 in the –rst 15% of a monolayer after which it rapidly increases to the heat of sublimation. Similarly on the disorder W oxide surface the initial heat of Cu adsorption was 280 kJ mol~1 at 300 K. These results are compared to Pb adsorption on the same oxide thin –lms and are discussed in the context of important factors in—uencing metal island growth. This journal is( The Royal Society of Chemistry 2000 the transient temperature rise caused by the adsorption on an ultrathin single-crystalline sample.Several methods can be used to measure the heat input into the sample as a result of adsorption such as infrared (IR) detection as originally used by King and coworkers,1h6 or by using pyroelectric LiTaO single crystals as heat sensitive substrates.8 This microcalorimetric technique is partic- 3 ularly useful for non-reversible adsorption processes where temperature programmed desorption (TPD) and equilibrium adsorption isotherm experiments fail to provide adsorption energies. Another difficulty with TPD studies is that the surface structure of the adsorbed species present at the desorption temperature may and often does diÜer from the structure of the adsorbate of interest present at lower temperatures. This is particularly true in cases for molecules that decompose on the surface during heating and for metal adsorption where thermodynamic equilibrium achieved at high temperatures often provides for vastly diÜerent structures than are observed by kinetically limited adsorption and –lm growth at lower temperatures.In addition for many systems the underlying surface is modi–ed during the heating process either by incorporation of the adsorbate into the surface such as alloying or mixed oxide formation or by the decomposition of the surface which can be observed during TPD of metals on oxide –lms or of metals on polymer surfaces. King and co-workers1h6 have used this technique primarily to investigate the adsorption of gaseous molecules (reactants) on single crystal metal (model catalyst) surfaces.We describe here calorimetric measurements of metal atom adsorption energies on well-de–ned oxide surfaces using a diÜerent heat detector than that developed by King and co-workers.1h6 Characterization of interfacial systems and the understanding of the interactions between metal –lms and a variety of substrates used as supports is important for a host of technologies including catalysis semiconductor processing and biotechnology. From a catalytic standpoint understanding the interactions of metals with underlying support oxides is necessary in order to characterize and predict how the oxide/metal interactions in—uence the catalyst morphology and catalytic properties. Several reviews of this topic have been published detailing a variety of interesting systems and important eÜects.9h14 In the case of metal adsorption on well-de–ned thin oxide –lms very little is known concerning the metal heats of adsorption and most of the existing experimental data is derived from TPD experiments with the aforementioned difficulties.Scanning microelectronic calorimeters have been developed to study metal –lms and small metal particles.15h17 These studies provide important information on the heat of fusion for the adsorbed metal islands and investigate melting point changes as a function of particle size for small metal particles. However these studies are not able to provide the heat of adsorption that is obtained from single crystal adsorption calorimetry. In previous communications we have presented a modi–cation that extends the microcalorimetric adsorption technique for the study of metal adsorption18,19 and reported the microcalorimetric heats of adsorption for a variety of metal/metal and metal/ oxide systems.18h24 Systems previously investigated include Pb adsorption on Mo(100) MgO(100) and the p(2]1) oxide of Mo(100) all studied at room temperature.18h23 The observed diÜerences between Pb adsorption on the metal and adsorption on the oxide surfaces follows the expected trend where metals interact strongly with other metals but only weakly with oxide surfaces.In this communication those heats of Pb adsorption on oxide surfaces are compared to the adsorption energies of Cu on similar surfaces in order to address the factors that in—uence the growth of metal –lms on oxide thin –lm surfaces.We report here a calorimetric measurement of the coverage-dependent heat of adsorption of copper onto two diÜerent metal oxide single crystal surfaces a thin –lm of MgO(100) grown on a Mo(100) crystal and an ordered thin –lm of Mo oxide produced by mild thermal oxidation of Mo(100). To gain a more complete understanding of the factors that determine the heat of adsorption and play an important role in determining the –lm growth mode it is important to compare data from diÜerent systems. The results reported here are compared to our previous results for Cu adsorption on an oxidized but disordered W(100) surface,21 highlighting a trend in adsorption energies for Cu adsorption on oxide thin –lm surfaces. Lead adsorption has also been studied on these oxide surfaces and shows similar results to the Cu/oxide systems indicating a trend for metal adsorption on oxides that may once more systems are tabulated allow better understanding of cic properties from a metal/support interaction standpoint.Magnesium oxide is one of a large group of oxide materials that is often used as relatively inert supports for active metals in catalytic applications. The Cu/MgO(100) system is a well-studied Faraday Discuss. 1999 114 195»208 196 2O3 representative model catalytic system. Numerous studies both experimental and theoretical of Cu thin –lms on oxide materials and on MgO(100) in particular have been previously published.25h34 The conclusions in the literature for Cu –lm growth on MgO are inconsistent although they often show generally similar results.Data for the growth of Cu on MgO(100) have been interpreted to be epitaxial,30 Volmer»Weber (3-D particle formation VW)25 and Stranki-Krastanov (layer-bylayer in the –rst monolayer with 3-D particles growing on top of the surface wetting layer SK).27,28,31 Surfaces of MgO(100) have been studied using both thin –lms of MgO(100) grown on a suitable refractory metal surface26 and on cleaved MgO crystals.25,27h31 LEED XPS TPD EELS AES and a variety of other spectroscopic techniques have been employed to study Cu on MgO(100) over a range of substrate temperatures and Cu coverages. At high coverage Cu islands or –lms grow which have a Cu(100) structure.30 Below B180 K evidence of isolated Cu adatoms on the MgO surface has been presented,34 while at higher temperatures Cu islands form with the Cu atoms becoming mobile near 200 K and with a diÜusion barrier height of around 0.5 eV.The more recent studies indicate that at room temperature Cu islands grow in either VW type 3-D growth mode25,26 or with SK growth.27,28 The formation of a Cu2O species at low Cu coverages has been suggested,27,28 although this assignment which is typically based on AES peak shifts is questionable when considering the importance of –nal-state eÜects on these small particles as discussed below and elsewhere.9 In theoretical studies several groups32,33 estimated the properties of small Cu particles adsorbed on MgO(100). These theoretical studies suggest that on defect-free surfaces the Cu will preferentially bind to surface oxygen atoms and that the growth of 3-D islands is energetically preferred even for small particles (i.e.VW growth). Additionally these studies suggest that at low coverages a Cu(100) like overlayer is not expected.32 Copper adsorption has also been studied on a variety of other oxides including Al ZnO TiO2 and SiO2 ,26,29,35h39 and 3-D Cu particle growth in the sub-monolayer regime is typically reported. Surfaces of MgO(100) have also been studied with respect to several other metals such as Pd which are reviewed in detail elsewhere.10 Molybdenum –nds use in a variety of catalytic applications and molybdenum oxide as a catalyst is active for partial oxidation reactions. The p(2]1) structure formed by light oxidation of Mo(100) represents a model molybdenum oxide thin –lm surface for studying Cu adsorption at room temperature.Oxide surfaces of Mo(100) have been studied previously to understand their structure and chemical/catalytic activity.40h46 This thin p(2]1) oxide is an B2 atomic layer thick oxide of molybdenum,40h46 with B50% oxygen in each of the two top layers of the oxide.40 Studies using STM support a missing-row reconstruction model of the Mo(100) surface upon oxygen adsorption and annealing.41,42 Experimental The ultra-high-vacuum (UHV) chamber and single crystal microcalorimeter apparatus used to collect the heat of adsorption data and the experimental protocol have been described in detail elsewhere18h20 and are only brie—y described here. The single crystal adsorption microcalorimeter is a modi–cation of the microcalorimetric system –rst reported by King and coworkers.1h4 One key diÜerence is the novel highly sensitive pyroelectric detection method which utilizes a thin pyroelectric polymer ribbon for sensing the transient temperature change.47 This detection method does not preclude high temperature treatment of our sample and should also be operable at cryogenic temperatures where IR detection becomes difficult.Detailed analysis of the pulse shape of this detectorœs response is presented elsewhere.24 In Fig. 1 a schematic of the calorimetric system is shown illustrating the calorimetric detector and a simpli–ed schematic of the chopped metal atom beam source. The pulsed metal calorimeter train consists of a chopped collimated beam of metal atoms from an eÜusive source which is incident onto a thin oxide –lm grown on a 1 micron thick Mo(100) sample.The thin Mo sample is mounted to a manipulator that allows us to move the sample into the calorimetric position and the various pre-treatment/preparation and electron spectroscopy positions. The pyroelectric detector ribbon used to measure heat input to the sample is gently pressed into contact with the back of the sample and allowed to come to thermal equilibrium. The sensitivity of the ribbon to heat input into the sample is then calibrated by introducing a laser pulse of known energy onto the sample with the same geometric and temporal pro–le as the metal pulse. The calorimetric measurements are then taken by impinging a 197 Faraday Discuss.1999 114 195»208 Fig. 1 A schematic of the pulsed metal atom source and the associated apparatus used to obtain the heat of adsorption and sticking probability. The apparatus consists of a high —ux Knudsen cell (TsourceB1700 K) and a series of apertures that provide a collimated source of metal atoms that is chopped in 100 ms pulses by a chopper wheel. The microcalorimeter detector is brought into contact with the sample to measure the heat input from the metal atom pulses. The metal —ux is measured by a quartz crystal microbalance. The BaF window allows separation of the heat input from radiation due to the hot metal source. The laser/prism2 assembly allows for absolute calibration of the detector signal against a known heat input to the sample.The QMS is used to measure the probability that metal atoms do not stick when they strike the sample. pulsed metal atom beam onto the surface and measuring the heat input for each metal atom pulse. By –rst blocking the metal atom —ux with the BaF window shown in Fig. 1 the contribution to 2 the measured heat attributable to the oven radiation is determined. The absolute metal —ux is determined with a quartz crystal microbalance (QCM). We are then able to determine the enthalpy of adsorption at 300 K by subtracting out the oven radiation correcting for the kinetic energy of the incident atoms to 300 K and making the appropriate addition to account for changes in pressure times volume. These corrections have been explained in detail elsewhere.18 The copper charge to the Knudsen-cell eÜusion metal source for these experiments was copper shot of 99.999% purity (metal basis).In order to accurately determine the heat of adsorption we need to accurately know the amount of metal in each pulse (metal —ux) and the sticking probability of the incident metal atoms as a function of metal coverage in addition to having an absolute calibration of the detector in contact with the sample to a known heat input. The metal —ux is determined by direct measurement of the —ux with the QCM and the sticking probability is determined by our line-of-sight modi–cation of the King and Wells method.48 The chamber is equipped with a UTI 100C quadruple mass spectrometer (QMS) which provides a suitable method to monitor any re—ected (non-sticking) metal for calibration of sticking probability in experiments similar to our microcalorimetric experiments.The QMS also allows TPD of the metal –lms to be performed. The QMS is in a line of sight position at the magic angle (to minimize the eÜect of any angular distribution49) as shown in Fig. 1. In this position we can directly monitor the angle-integrated —ux of any re—ected metal atoms which do not stick to the sample. It is important for this measurement to have nearly normal incidence of the atom beam so that non-sticking atoms in both trapping/desorption and the quasi-elastic/quasi-specular channels have Bcosn h angular distributions where h is the polar angle and n is an exponent between one and nine. By comparing the time-integrated signal (per pulse) with the integrated TPD signal of a metal –lm of known coverage (as determined by the QCM —ux]dose time on a surface with unit sticking) we can determine the sticking probability.A temperature correction is also applied to the signal (P1/T 1@2) to adjust the QMS signal for the diÜerences in velocities between the re—ected signal and the TPD signal. In all cases we assume that the metal atoms leave the surface at the surface temperature.10 The chamber is also equipped with LEED and AES which are used to characterize the order and cleanliness of the samples as well as to investigate the Cu metal –lm structure. The MgO(100) –lms were grown for these experiments by exposing the Mo(100) sample at 300 K to a —ux of magnesium metal (B4]1014 atoms cm~2 min~1) from a thermal evaporation Faraday Discuss.1999 114 195»208 198 source under a background pressure of 3]10~7 Torr of oxygen similar to the method described elsewhere.26,50 Subsequently the MgO –lms were thermally annealed to B750 K by electronbeam heating under UHV to order the –lm. This procedure yielded MgO LEED patterns similar to those reported by Wu and coworkers50 for MgO(100) –lms grown on Mo(100). We estimate from AES that the MgO –lms grown using this procedure are B40 Aé thick. The Mo(100) and p(2]1)O»Mo(100) thin –lms also showed LEED patterns characteristic of well-ordered surfaces. The oxygen concentration on the oxidized molybdenum surface was con–rmed by AES. Results and discussion Adsorption of Cu on MgO(100) In Fig.2 the observed sticking probabilities of Cu on MgO(100) at room temperature as a function of coverage are presented. One monolayer (ML) of Cu is de–ned here as the packing density of a Cu(111) plane in bulk Cu 1.77]1015 atoms cm~2. The initial sticking coefficient is determined to be B0.996 and the sticking coefficient slowly increases with coverage towards unit sticking. Although a signi–cant amount of noise appears in the sticking coefficient as a function of coverage only sticking probabilities above 0.99 are observed so this scatter is due to the very tiny re—ected signals. The sticking probabilities that we have observed for copper on the MgO(100) thin –lms are higher than has been previously reported for this and similar systems.26,29 For room temperature adsorption of Cu on a MgO(100) thin –lm very similar to ours Wu et al.26 report a value of 0.82 for the initial sticking probability while Zhou et al.29 report a value of 0.5 for S on a cleaved o MgO(100) surface which certainly has a much lower defect density than our thin –lm.There are several factors that may contribute to the higher sticking probabilities we observe. First if our MgO –lm has a larger defect density than that of Zhou et al.,29 we would expect a greater number of islands to nucleate due to the larger defect density. Since greater island density would eÜectively lower the average diÜusion length between islands for Cu atoms in a mobile precursor state (isolated Cu atoms on MgO) this would result in a higher sticking probability assuming that the sticking probability is governed by competition between desorption and addition to copper islands.This argument for higher sticking with defect density is supported although rather loosely by Zhou et al.,29 who introduced ììdefectsœœ to the MgO(100) surface by way of small coverages of molybdenum and observed a dramatic increase to unit sticking. Second and possibly more important is that the instantaneous metal —ux we employ is much higher than the —ux used Fig. 2 The measured sticking coefficients of copper as a function of copper coverage on the MgO(100) thin –lm at room temperature. Each point represents approximately 2.7% of a monolayer of Cu atoms. A best –t curve through the sticking probability is shown as a guide. 199 Faraday Discuss. 1999 114 195»208 in these other experiments.For instance we routinely operate at a —ux during the metal pulse equivalent to one monolayer every 5 s. Wu et al.26 report a —ux equivalent to 1 ML in 160 s 32 times lower than our —ux. Common models for metal island growth9,51 suggest that the island density is proportional to (—ux/diÜusion)1@2h1@3. Given that the diÜusion constants for Cu should be the same on these similar MgO –lms these models suggest an island density 3.1»5.7 times larger in our experiment neglecting any diÜerences in defect density. One can easily see how this large increase in island density could have a dramatic eÜect on the sticking probability assuming some eÜective diÜusion length in which a metal atom searches for an island or desorbs. One additional concern is the apparent saturation of the sticking coefficient at 0.82 which has been observed previously upon cooling.26 This might indicate a small (18%) error in the absolute calibration since one would expect the sticking to increase all the way to unity at lower temperatures.On a similar system SchaÜner et al. estimated a room temperature sticking probability of 0.6 for Ag on MgO(100) thin –lms.52 Contrasting our results with those of SchaÜner et al. for silver on MgO we see a similar trend. In the Ag/MgO case the Ag —ux was 50 times less than our Cu —ux. Also Ag is expected to have a lower desorption energy and therefore a lower sticking probability on MgO than Cu due to Agœs lower sublimation energy. (This trend was observed for Pb adsorption on MgO»see below.) Regardless of the factors leading to the high sticking probability which we observe this sticking probability allows us to calculate the adsorption energies assuming a unit sticking for copper on the oxide surfaces.Once we know the sticking probability we can then determine the copper heat of adsorption as a function of coverage. In Fig. 3 we present the heat of adsorption of a sequence of Cu pulses onto a room temperature MgO(100) thin –lm. The inset to Fig. 3 shows an enlargement of the experimental heat of adsorption in the low coverage range. The heat of adsorption in the –rst 2»4% of a monolayer of copper is 240 kJ mol~1 almost 100 kJ mol~1 lower than the Cu heat of sublimation of 337.4 kJ mol~1. As the copper coverage increases the heat of adsorption rapidly increases reaching 304 kJ mol~1 (B90% of the *H of sublimation) by 0.25 ML and 321 kJ mol~1 (B95% of the *H of sublimation) by 0.7 ML.The heat of adsorption levels oÜ at B98% of the Cu heat of sublimation by 2»3 ML and slowly increases thereafter reaching the heat of sublimation by 4.5 Fig. 3 The calorimetric heat of adsorption (standard enthalpy of adsorption) of Cu on a thin MgO(100) overlayer (grown on the Mo(100) surface) as a function of copper coverage at room temperature. The inset shows a blow up of the low coverage heat of adsorption highlighting the rapid increase in adsorption energy. Faraday Discuss. 1999 114 195»208 200 ML. Fig. 3 clearly shows that the heat of adsorption of the Cu metal atoms is a strong function of coverage at low coverages and the heat rapidly approaches the expected *H of sublimation by a coverage of several ML.To better understand the growth of Cu on the MgO(100) thin –lms we examined the metal –lm growth as a function of coverage with AES. In Fig. 4 we show representative Cu(MVV) AES spectra of –lms of diÜerent coverages of Cu grown at room temperature on the MgO(100) thin –lm. The spectra show an increase in the Cu peak to peak intensity with increasing Cu coverage. Upon careful examination of the AES data one also observes a shift in the peak position to higher energy with Cu coverage. A trend line is shown in Fig. 4 to highlight this shift of B2.5 eV with increasing coverage. Above B10 ML of Cu the Cu(MVV) transition energy remains constant at 61 eV and is attributed to bulk-like Cu particles.The AES intensities have been analyzed by comparison to a layer-by-layer model for the growth of Cu –lms. Fig. 5 compares the experimental Cu AES intensities (data points) to the expected intensity for a layer-by-layer growth model (trend line) both as a function of Cu coverage. In this model 5.0 Aé was used as the escape depth for electrons from Cu in this energy range.53 Up to B0.3 ML the AES data –t the layer-by-layer model reasonably well. Above this coverage a break is observed in the AES data indicating that the growth has switched from 2-D to 3-D in this coverage range. Since the Cu AES signal falls to about half the value expected for the layer-bylayer growth model by B3 ML of Cu these data also show that Cu grows in 3-D islands.The AES signal indicates also that at 3 ML of Cu up to B30»40% of the MgO surface remains free of Cu and B25% is free at 10 ML. This type of transition from 2-D islands to 3-D islands during the growth of metals is often seen on oxides.9,38 The 2-D islands are kinetically allowed but are not thermodynamically preferred,9 as also shown by the lower bonding strength of Cu to the oxide relative to the Cu»Cu bonding. Previously published data for the growth of Cu on MgO(100) surfaces have been interpreted to indicate various growth modes including layer-bylayer VW and SK growth.25h34 Based on our results and much of this literature we favor a growth for Cu on MgO(100) at 300 K which is characterized as kinetically controlled 2-D island growth up to a coverage of B0.3 ML.Above 0.3 ML the islands thicken into 3-D particles and impinging Cu atoms add on top of these islands faster than they cover the remaining bare oxide surface. This is consistent with both the calorimetric heats which reach nearly the heat of subli- Fig. 4 Representative Cu(MVV) AES spectra of –lms of diÜerent coverages of Cu grown at room temperature on the MgO(100) thin –lm. A 2.5 eV shift to increasing energy is observed with increasing Cu thickness up to Cu coverages of between 10 ML (not shown) and 14 ML above which the peak position remains constant. For thicker –lms the kinetic energy of 61.0 eV is attributed to bulk Cu. The trend line highlights the shift. 201 Faraday Discuss. 1999 114 195»208 Fig. 5 The Cu AES peak to peak intensities for Cu on MgO(100) as in Fig.4 plotted vs. Cu coverage and compared to the layer-by-layer growth model. The Cu signal shows a deviation from the layer-by-layer model at a Cu coverage of approximately 0.3 ML. Above 0.3 ML the AES data suggest that the growth mode has switched from 2-D island to a 3-D growth mode. The Cu AES intensities are normalized to the intensity from a clean bulk Cu –lm. mation by 0.3 ML as expected for Cu addition onto large 3-D islands. This type of growth could be interpreted as either VW or as SK up to some coverage less than a complete monolayer although a non-equilibrium growth model is most appropriate at least at low copper coverages. The relatively low initial heat of adsorption observed in Fig.3 for Cu on the MgO(100) surface indicates that the interactions between copper atoms/islands and the MgO substrate are much weaker than the Cu»Cu interactions. These lower metal»oxide interaction energies clearly indicate that large 3-D Cu particles should be thermodynamically favorable for this system and any epitaxial or 2-D –lm growth could only result from a kinetically limited –lm growth process. This highlights the tremendous advantage in determining heats of adsorption using direct calorimetric measurements since a TPD approach would be likely to lead to ripening of the Cu islands/ particles prior to desorption resulting in a measurement not truly representative of the nonequilibrium structures formed during metal deposition at room temperature. 2O C2O formation at low coverages we feel that the formation of auspecies At this point we reconsider the shift observed in the Cu AES peak position.In the literature some reports27,28 have suggested that at initially low coverages of Cu on MgO(100) Cu is adsorbed as a Cu2O like species. This interpretation ultimately comes from shifts in the AES peak positions or a shift in the associated Auger parameter. Although we also observe a shift that might be interpreted as Cu is somewhat unlikely. The cause of the shift in AES peak position is most likely a result of –nal state relaxation eÜects during the Auger process which are expected and indeed cause these shifts for small neutral 2-D and 3-D metal particles.9,10 Comparing this to a previous study of Cu thin –lms on ZnO(0001)»O we observed that at the critical coverage of 0.5 ML of Cu (where 2-D island growth switches to a 3-D dominated growth) a shift of 1.8 eV is observed in the position of the Cu AES peaks relative to bulk Cu.38 This shift is not caused by the formation of Cu2O on the ZnO substrate but is the result of –nal state relaxation eÜects.In the Cu on MgO(100) experiments reported here at a coverage of 0.3 ML of Cu we also see a break from the layer-by-layer or 2-D growth mode based on the AES and we observe a shift of B2.0 eV from the bulk Cu peak position. This is in excellent agreement with the magnitude of the shift at the 0.5 ML critical Faraday Discuss. 1999 114 195»208 202 coverage of Cu on ZnO where 3-D growth begins and strongly suggests that the shift in peak position is due not to Cu2O formation but instead is due to –nal state relaxation eÜects for 2-D charge-neutral Cu islands on the MgO(100) surface as well.This interpretation is in agreement with Zhou and Gustafsson,25 both in terms of the Cu –lm growth mode determination and the unlikelihood of Cu formation. In the work of Zhou and Gustafsson25 on a cleaved MgO(100) 2O surface with a much lower defect density than on our epitaxially growth thin MgO –lms the authors found that the Cu islands grew in a VW mode even at coverages as low as 0.14 ML and that at a Cu coverage of 0.99 ML the islands were on average B5 layers thick. The higher defect density on our surface may lead to 2-D growth up to the B0.3 ML which we observe. Theoretical studies –nd very limited charge transfer even for Cu clusters of just four atoms.32,33 n *adsH) From the integral calorimetric heat of adsorption up through multilayer coverage (& the adhesion energy for the metal on the oxide surfaces (Eadh) can be estimated within a simple thermodynamic formula derived elsewhere9,20 Eadh\[(n*subH[&n *adsH)/A](1]f )p where p is the surface energy of the bulk metal (176 lJ cm~2),54 A is the area covered by the metal –lm which contains n atoms and f is the surface roughness factor of this multilayer –lm.For the 7 ML Cu –lm the integral heat of adsorption is 0.0009378 J or an average of 330.7 kJ mol~1 which is spread over a geometric beam area of 0.138 cm2. To obtain an accurate value for the adhesion energy between the metal –lm and the oxide surface it is necessary to have a good estimation of both the absolute area covered by the multilayer metal –lm and the roughness factor ( f ) of the metal –lm.20 From the AES data we estimate that 25% of the MgO surface is not covered by the Cu –lm at a coverage of 7 ML therefore we scale the geometric surface area by 0.75.Although we do not measure the surface roughness directly we can make some reasonable approximations and show that it is near 1.0 so that we can estimate the adhesion energy. The multilayer –lm from which our adhesion energy is calculated covers B75% of the surface and is B7 ML thick on average. The density of islands in such systems is typically9,10 B1012 cm~2 which means that the island centers are separated by 10~6 cm from the nearest neighboring island centers.This is a length equivalent to the thickness of B50 ML. The island diameters are then about the same length since they must be overlapping if only 25% of the surface is uncovered. As spherical caps islands with an average height of B7/M(50)](0.75)N\18.5% of their diameter have a surface area that is B1.14 times their projected area which provides a crude estimate of their surface roughness factor. A roughness factor even closer to unity would result if we assumed that our Cu islands obtain similar aspect ratios to 3-D Cu mounds grown upon deposition of Cu multilayers on Cu(100).55 Using 1.14 for the roughness factor the adhesion energy of Cu on the MgO(100) surface is estimated as B192 lJ cm~2. This estimation is highly sensitive to the exact surface coverage and the roughness of the –lm.If for instance the island density is actually four times larger it would give a roughness factor of B1.5 and an adhesion energy of B250 lJ cm~2. Let us consider whether the Cu heats on MgO(100) can be modeled within a nearest-neighbor bond additivity model. For simplicity we will model the Cu particles as growing as fcc(111) platelets where Cu atoms in the –rst layer have one Cu»MgO bond downward. When a Cu atom adds to either a large 2-D island or a large Cu island on an existing Cu(111) platelet it will be assumed that it adds at a kink site such that it forms three Cu»Cu bonds parallel to the surface. When adding on a large Cu island on an existing Cu(111) platelet it also forms three Cu»Cu bonds downward to the Cu layer below for a total adsorption energy equal to the strength of 3]3\6 Cu»Cu bonds and also equal to the bulk sublimation energy of Cu B337.4 kJ mol~1.This gives B56.2 kJ mol~1 per Cu»Cu bond. If we assume that at the highest coverage where Cu growth is still following a layer-by-layer model (0.3 ML) that the Cu is adding predominantly to large 2-D Cu islands then the measured heat of adsorption at this point (310 kJ mol~1) must correspond to the formation of one Cu»MgO downward bond plus three Cu»Cu bonds parallel to the surface. Since the latter contribute 56.2 kJ mol~1 each to this energy within this bond-additivity model the Cu»MgO bond energy must then equal B310»3(56.2)\141.4 kJ mol~1. Within this model the increase in adsorption energy from a value of B240»310 kJ mol~1 in the –rst 0.3 ML (where only 2-D Cu exists according to AES) simply corresponds to the increase in the number of Cu nearest neighbors as the 2-D Cu island size increases.The initial heat of 240 kJ mol~1 corresponds then to (240»141.4)/56.2\1.75 nearest neighbors bonds on average. This value suggests 203 Faraday Discuss. 1999 114 195»208 that the initial heat corresponds to the formation of 2-D Cu clusters ranging in size from B7»10 atoms. (Forming Cu dimers gives an average of only 1/2 Cu»Cu bond per Cu atom.) Density functional calculations estimate an average adsorption energy of 206 kJ mol~1 for the Cu atoms in a 9-atom 2-D island on Mg(100),32 and 156 kJ mol~1 in a 4-atom cluster.33 Within this model the adhesion energy between Cu and MgO(100) would simply correspond to 141.4 kJ mol~1 of Cu»MgO bonds at the interface.Given that the Cu(111) packing density is 1.77]1015 Cu atoms cm~2 then this converts to 415 lJ cm~2. This is B215% larger than the actual observed adhesion energy of 192 lJ cm~2. We attribute this diÜerence to the failure of the bond-additivity concept here. Probably the heat of adsorption of 310 kJ mol~1 measured at 0.3 ML coverage corresponds to much stronger Cu»Cu bonds than the average bulk value of 56.2 kJ mol~1 so that the downward Cu»MgO bond energy is actually somewhat weaker than the value of 141.4 kJ mol~1 (estimated assuming pair-wise bond additivity) by B50% or B71 kJ mol~1. Density functional calculations32 indicate that the Cu atoms preferentially reside on top of the oxygen atoms on the MgO(100) surface with a bond energy of B84 kJ mol~1 which is in reasonable agreement with this value.Lead adsorption on this same MgO(100) thin –lm surface has also been studied and is reported elsewhere.22 The initial heat of adsorption for lead on the room temperature MgO(100) surface is B100 kJ mol~1 and rapidly increases with Pb coverage to the *H of sublimation of bulk Pb of 195.2 kJ mol~1. Similar to the Cu –lm growth on MgO(100) the initial heat of Pb adsorption is B100 kJ mol~1 lower than the heat of sublimation. In this respect lead should also exhibit a similar equilibrium growth morphology on the MgO(100) surface. Indeed the AES data for Pb on MgO(100) support a VW like growth mode.22,23 Additionally the growth of 3-D lead particles at low coverages is supported by the observed sticking probabilities for Pb on the MgO(100) surface.At coverages above 1.0 ML the sticking probability remains well below unity even though lead sticks to pure lead with unit probability. This suggests large patches of bare MgO still exist.22,23 Qualitatively the heats of adsorption and growth morphology for Cu seen here are very similar to the results for Pb on the MgO(100) surface. Adsorption of Cu on the p(2�1) oxide of Mo(100) As a comparison to the heat of adsorption of Cu on MgO(100) we have also investigated the heat of adsorption of Cu on the p(2]1)O overlayer on the Mo(100) surface. This overlayer is produced by a saturation dose (B3 L) of O at room temperature followed by annealing to 1075 K and is 2 believed to be essentially a bilayer of molybdenum oxide on the underlying Mo(100) structure.40h46 In Fig. 6 we show the heat of adsorption as a function of coverage for Cu on the p(2]1)O»Mo(100) surface at room temperature. The initial heat of adsorption is 286.9 kJ mol~1 (average of the –rst 3 pulses). The heat of adsorption slowly decreases by 3% to 277.8 kJ mol~1 up to a coverage of 0.15 ML. Above 0.15 ML of copper the adsorption energy increases reaching 92% of the heat of sublimation of bulk Cu by 1.0 ML and saturating at a value of 332.5 kJ mol~1 within 2% of Cuœs heat of sublimation. The nearly constant Cu heat of adsorption for the –rst 15% of a ML followed by a rapid increase in the heat of adsorption toward the sublimation energy has also been observed for lead adsorption on this same p(2]1)O overlayer on Mo(100).22,23 For lead adsorption an initial low value of 145 kJ mol~1 is observed again and stays relatively constant until B0.15 ML above which the heat of adsorption rapidly approaches the heat of lead sublimation of 195.2 kJ mol~1.These similar results suggest that some fundamental aspect of metal adsorption on the p(2]1) molybdenum oxide surface must lead to this interesting feature. The similarity between Cu and Pb in this respect is surprising given the diÜerences in covalent bonding usually observed associated with their diÜerent electronic con–gurations (positions in the Periodic Table). The similarity suggests that direct covalent bonding does not dominate the metal attraction to the oxide (for the MgO(100) the p(2]1)O»Mo(100) and the oxidized W) but instead that the attraction arises mainly from the polarizability of the metal responding the Madelung potential of the oxide surface.Observed trends in measured adhesion and interfacial energies between metals and oxide surfaces (as determined by contact angle methods) have been explained based on a similar physical mechanism for interfacial bonding.56 For Pb adsorption on the (2]1) molybdenum oxide thin –lm AES showed that the growth initially followed a layer-by-layer curve vs. coverage. But at a coverage of B0.3 ML the Faraday Discuss. 1999 114 195»208 204 Fig. 6 The calorimetric heat of adsorption of Cu as a function of coverage at room temperature on the Mo(100) single crystal surface pre-covered with an ordered p(2]1) overlayer of oxygen.The inset to the –gure shows a enlargement of the Cu adsption in the low coverage regime up to 1.0 ML. Pb/Mo AES ratio dropped below the layer-by-layer curve showing that above B0.3 ML 3-D islands of Pb predominantly grow.22 This morphology change was used to explain the abrupt * of Pb observed at about this same coverage:22 below B0.15 ML the Pb increase in the adsH adatoms bond directly onto the oxide surface (probably forming 2-D islands) and above B0.3 ML they add on top of existing Pb islands thus having larger heats of adsorption similar to the *subH. We attribute the increase in the *adsH of Cu seen here between 0.2 and 1.0 ML to a very similar phenomenon the transition from the growth of mainly 2-D to mainly 3-D Cu islands.The growth of Cu on both the Zn- and O-terminated faces of ZnO(0001)57,58 has been shown to undergo a more abrupt 2-D to 3-D transition in growth at coverages of B0.3 and B0.55 ML respectively. A kinetic model to explain such a transition has been suggested9,59 which relies on an energetic diÜerence qualitatively similar to that measured here for Cu on this Mo oxide thin –lm surface and for Pb on the same oxide. Comparing the results for Cu on the two oxide surfaces and to results to be published elsewhere for Pb on these oxide surfaces22,23 allows us to develop a better understanding of the important factors in metal growth on oxides. In general for all four of the aforementioned systems as well as for metals on hydroxylated MgO(100) thin –lms,60 the metal adsorbate interacts much more weakly with the oxide surface than with other metal atoms.This fact is observed through a lower heat of adsorption on the clean surfaces that rapidly approaches the metal heat of sublimation as the coverage increases to a point where the adsorbing metal can interact preferably with large 3-D metal islands which allow for high coordination of the adsorbing atoms with the underlying substrate. Very small metal particles often exhibit diÜerent catalytic properties than larger metal particles which behave more like bulk metals in catalytic reactions.9h14 An example of this interesting phenomena is oxide supported gold particles for both propylene and CO oxidation.61,62 Certainly numerous eÜects contribute to the diÜerent catalytic properties of smaller metal particles.It has long been thought that smaller particles will be more aggressive in the chemisorption of reactants since the average coordination number of the metal atoms in smaller particles is lower than in larger particles. Stronger adsorption on smaller particles is indeed often observed.9,11 However another equally important factor is the role of the underlying oxide support in determining the 205 Faraday Discuss. 1999 114 195»208 morphology and chemical properties of the metal particles. Here we have demonstrated that the initial adsorption energy of copper on oxides is much lower than the adsorption energy of the metal onto itself (i.e.the sublimation energy). This much weaker bonding to the surface for the special case of metal atoms in 2-D islands compared to metal atoms in the top plane of a 3-D metal island of comparable area eÜectively places them at a lower ììeÜective coordination numberœœ as discussed elsewhere.11 Due to this ììeÜective lower coordinationœœ of metal atoms in 2-D islands one can expect the metal atoms to be more aggressive in their chemisorption properties (i.e. they bond molecules fragments and atoms more strongly). Thus it is easy to understand how the catalytic properties might diÜer. From this perspective the metal adsorption energies provide signi–cant insight into how the oxide-metal interactions contribute to the chemisorption and catalytic properties of oxide-supported ultrathin metal particles.This eÜect of metal»oxide interactions is to be clearly distinguished from an eÜect of particle size alone. Thus we predict based on these energetics that a 2-D Cu island of the same size on these oxide surfaces will be more aggressive (i.e. less noble) in its chemisorption properties than a 2-D Cu island of the same size and shape supported on another transition metal or on Cu itself as a 3-D island. Indeed comparing the TPD characteristics of small probe molecules from 2-D Cu islands on ZnO single crystals59 with those from 2-D Cu islands on Ru(0001)63 or bulk-like Cu surfaces indicates this to be the case. Conclusion The calorimetric heats of Cu adsorption have been determined as a function of coverage for a series of well-de–ned oxide surfaces at 300 K.In all cases the Cu bonding to the oxide surfaces is found to be a lower energy interaction than Cu»Cu bonding. On MgO(100) thin –lms the initial room temperature heat of adsorption for Cu is 240 kJ mol~1 and rapidly increases with coverage eventually approaching the Cu heat of sublimation of 337.4 kJ mol~1. Cu appears to form 2-D islands on the MgO substrate up to a coverage of B0.3 ML above which 3-D island formation is observed. The coverage dependence of the heat of adsorption suggests that a pair-wise additive nearest neighbor bonding model is insufficient to model the adsorption energies. For Cu adsorption on the ordered p(2]1) oxide of Mo(100) the initial room temperature heat of adsorption is 285 kJ mol~1 and stays relatively constant up to a coverage of 0.15 ML where it rapidly begins to increase to the sublimation energy.Similar to these results the heat of Cu adsorption on a disorder oxide thin –lm on W(100) shows an initial low value of 280 kJ mol~1 which increases with coverage as reported elsewhere.20 These Cu adsorption data have been compared with calorimetric results for lead adsorption on these same oxide –lms in order to investigate the eÜect of the metal on the trends observed for the Cu/oxide systems. In both the MgO(100) and p(2]1) oxide of Mo(100) cases qualitatively similar results are observed for Cu and Pb adsorption. For both Cu and Pb on the MgO(100) thin –lm the initial metal adsorption energy is approximately 100 kJ mol~1 less than the *H of metal sublimation.On the p(2]1) oxide both metals appear to either grow as 2-D islands up to B0.15 ML or else the metal atoms titrate defect sites on the oxide surface. For both metals the initial heat of adsorption is approximately 55 kJ mol~1 less than the *H of metal sublimation. This low initial value stays relatively constant (or slightly decreasing) up to metal coverages of 15% of a monolayer. More detailed trends relating growth mode and chemical properties to heats of adsorption both for diÜerent metals and diÜerent substrates will be developed as the database for these microcalorimetric experiments expands. The weaker interactions between the underlying oxide –lm and the metal overlayer are signi–cant in controlling the growth of the metal –lms.These interactions may also lead to modi–ed catalytic properties of thin –lms and small metal particles when compared to bulk metal like catalysts and contribute to the more aggressive chemisorption properties of thin metal –lms on oxide supports. Acknowledgements The National Science Foundation is acknowledged for –nancial support of this research project. We would like to thank Dr. Hans Coufal and Professor David King for useful discussions on the calorimetric experimentation and Jacques Chevallier for supplying the thin Mo(100) metal single crystals used in these experiments. Faraday Discuss. 1999 114 195»208 206 References 1 C. E. Borroni-Bird N. Al-Sarraf S. Andersson and D. A. King Chem. Phys. L ett. 1991 183 516. 2 C. E.Borroni-Bird and D. A. King Rev. Sci. Instrum. 1991 62 2177. 3 A. Stuck C. E. Wartnaby Y. Y. Yeo and D. A. King Phys. Rev. L ett. 1995 74 578. 4 A. Stuck C. E. Wartnaby Y. Y. Yeo J. T. Stuckless N. Al-Sarraf and D. A. King Surf. Sci. 1996 349 229. 5 N. Al-Sarraf J. T. Stuckless C. E. Wartnaby and D. A. King Surf. Sci. 1993 283 427. 6 W. A. Brown R. Kose and D. A. King Chem. Rev. 1995 95 797. 7 S. Cernyœ Surf. Sci. Rep. 1996 26 1. 8 S. J. Dixon-Warren M. Kovar C. E. Wartanaby and D. A. King Surf. Sci. 1994 307ñ9 16. 9 C. T. Campbell Surf. Sci. Rep. 1997 227 1. 10 C. R. Henry Surf. Sci. Rep. 1998 31 231. 11 C. T. Campbell Curr. Opin. Solid State Mater. Sci. 1998 3 439. 12 P. L. J. Gunter J. W. Niemantsverdriet F. H. Ribero and G. A. Somorjai Catal. Rev.-Sci.Eng. 1997 39 77. 13 H.-J. Freund Angew. Chem. Int. Ed. Engl. 1997 36 452. 14 D. R. Rainer and D. W. Goodman J. Mol. Catal. 1998 131 259 15 S. L. Lai J. Y. Guo V. Petrova G. Ramanath and L. H. Allen Phys. Rev. L ett. 1996 77 99. 16 S. L. Lai G. Ramanath L. H. Allen P. Infante and Z. Ma Appl. Phys. L ett. 1995 67 1229. 17 S. L. Lai J. R. A. Carlsson and L. H. Allen Appl. Phys. L ett. 1998 72 1098. 18 J. T. Stuckless N. A. Frei and C. T. Campbell Rev. Sci. Instrum. 1998 69 2427. 19 J. T. Stuckless D. E. Starr D. Bald and C. T. Campbell Mater. Res. Soc. Symp. Proc. 1997 440 103. 20 J. T. Stuckless D. E. Starr D. Bald and C. T. Campbell J. Chem. Phys. 1997 107 5547. 21 J. T. Stuckless D. E. Starr D. Bald and C. T. Campbell Phys. Rev. B Condens. Matter 1997 56 13497.22 D. E. Starr J. T. Ranney J. E. Musgrove D. J. Bald and C. T. Campbell in preparation. 23 D. E. Starr J. E. Musgrove D. J. Bald J. T. Ranney and C. T. Campbell in preparation. 24 J. T. Stuckles N. Frei and C. T. Campbell Sens. Actuators B in press. 25 J. B. Zhou and T. Gustafsson Surf. Sci. 1997 375 221. 26 M.-C. Wu W. S. Oh and D. W. Goodman Surf. Sci. 1995 330 61. 27 T. Conard J. Ghijsen J. M. Vohs P. A. Thiry R. Caudano and R. L. Johnson Surf. Sci. 1992 265 31. 28 T. Conard J. M. Vohs P. A. Thiry and R. Caudano Interface Anal. 1990 16 446. 29 J. B. Zhou H. C. Lu T. Gustafsson and E. Garfunkel Surf. Sci. 1993 293 L887. 30 J.-W. He and P. J. Moller Surf. Sci. 1986 178 934. 31 I. Alstrup and P. J. Moller Appl. Surf. Sci. 1998 33/34 143.32 (a) V. Musolino A. Selloni and R. Car Surf. Sci. 1998 402»404 413; (b) V. Musolino A. Selloni and R. Car J. Chem. Phys. 1998 108 5044. 33 (a) A. V. Matveev K. M. Neyman and G. Pacchioni Chem. Phys. L ett. 1999 299 603; (b) G. Pacchioni and N. Roé sch J. Chem. Phys. 1996 104 7329. 34 M.-H. SchaÜner F. Patthey W.-D Schneider and L. G. M. Pettersson Surf. Sci. 1998 402»404 450. 35 J. B. Zhou H. C. Lu T. Gustafsson and E. Garfunkel Surf. Sci. 1997 382 21. 36 Y. Wu E. Garfunkel and T. Madey J. V ac. Sci. T echnol. A. 1996 14 1662. 37 J.-W. He and P. J. Moller Surf. Sci. 1987 180 411. 38 K. H. Ernst A. Ludviksson R. Zhang J. Yoshihara and C. T. Campbell Phys. Rev. B Condens. Matter 1993 47 13782. 39 U. Diebold J.-M. Pan and T. E. Madey Phys. Rev. B Condens.Matter 1993 47 3868. 40 E. Bauer and H. Poppa Surf. Sci. 1975 48 31. 41 H. Xu and K. Y. S. Hg Surf. Sci. 1996 356 19. 42 H. Xu and K. Y. S. Hg Surf. Sci. 1996 355 L305. 43 C. Zhang M. A. Van Hove and G. A. Somorjai Surf. Sci. 1985 149 326. 44 S. L. Miles S. L. Bernasek and J. L. Gland J. Phys. Chem. 1983 87 1626. 45 R. M. Henry B. W. Walker and P. C. Stair Surf. Sci. 1985 155 732. 46 B. W. Walker and P. C. Stair Surf. Sci. 1981 103 315. 47 H. J. Coufal R. K. Grygier D. E. Horne and J. E. Fromm J. V ac. Sci. T echnol. A 1987 5 2875. 48 D. A. King and M. G. Wells Surf. Sci. 1972 29 454. 49 S. W. Pauls and C. T. Campbell Surf. Sci. 1990 226 250. 50 M.-C. Wu J. S. Corneille C. A. Estrada J.-W. He and D. W. Goodman Chem. Phys. L ett. 1991 182 472. 51 J. A. Venables Philos. Mag. 1973 27 697. 52 M.-H. SchaÜner F. Pattey and W. D. Schneider Surf. Sci. 1998 417 159. 53 S. Tanuma C. J. Powell and D. R. Penn Surf. Interface Anal. 1991 17 911. 54 W. R. Tyson and W. A. Miller Surf. Sci. 1977 62 267. 55 J.-K. Zuo and J. F. Wendelken Phys. Rev. L ett. 1997 78 2791. 56 F. Didier and J. Jupille Surf. Sci. 1994 314 378. 57 J. Yoshihara J. M. Campbell and C. T. Campbell Surf. Sci. 1998 406 235. 58 C. T. Campbell and A. Ludviksson J. V ac. Sci. T echnol. A 1994 12 1825. 207 Faraday Discuss. 1999 114 195»208 59 S. C. Parker A. W. Grant V. A. Bonzie and C. T. Campbell Surf. Sci. in press. 60 J. E. Musgrove D. E. Starr J. T. Ranney D. J. Bald and C. T. Campbell in preparation. 61 T. Hayashi K. Tanaka and M. Haruta J. Catal. 1998 178 566. 62 M. Haruta Catal. T oday 1997 36 153. 63 D. W. Goodman and C. H. F. Peden J. Chem. Soc. Faraday T rans. 1 1987 83 1967. Paper 9/02649E Faraday Discuss. 1999 114 195»208 208

ISSN:1359-6640    

DOI:10.1039/a902649e

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

2 Cu atoms and clusters on regular and defect sites of the SiO surface. Electronic structure and properties from �rst principle 2 calculations Gianfranco Pacchioni,a Nuria Lopezb and Francesc Illasb a Dipartimento di Scienza dei Materiali Universita` di Milano-Bicocca Istituto Nazionale di Fisica della Materia via Cozzi 53-20125 Milano Italy. E-mail gianfranco.pacchioni=mater.unimib.it b Departament de Quimica Fisica Universitat de Barcelona Marti i Franques 1 08028 Barcelona Spain Received 24th March 1999 The interaction of isolated Cu atoms and small Cu clusters from Cu to Cu with the dehydroxylated surface of silica has been investigated by means of cluster models density functional calculations. The regular non defective surface shows very low reactivity towards Cu atoms; the binding is largely due to polarization mechanisms.This implies that impinging Cu atoms will easily diÜuse on the surface and re-evaporate unless trapped at a defect site. In fact strong bonds are formed between Cu atoms and clusters and some typical point defects at the SiO surface. We have analyzed two of these defects the non-bridging oxygen site (NBO) 3Si»O~ and the E@ center corresponding to a Si singly occupied sp3 dangling bond 3Si~. The Cu clusters interacting with these paramagnetic centers are signi–cantly perturbed by the bonding at the interface as shown by the diÜerent geometrical structures of supported compared to gas-phase clusters. Some observable consequences of the cluster deposition in particular the appearance of states in the gap of the material are discussed.1 Introduction The metal/oxide interface represents a very important aspect of surface science.1h3 It is related to a variety of technological applications from catalysis and gas sensors to microelectronics from coating and protection of metals to colloidal chemistry and non-linear optics phenomena. The experimental study of oxides using surface science techniques is complicated because oxides are usually insulators and are brittle. This causes experimental difficulties related to sample charging and heating. An approach that has been used in recent years to study oxides and metal/support interactions under controlled conditions is to deposit oxide thin-–lms on a metal surface.1h3 Important information about the nature of oxide surfaces comes also from the study of polycrystalline materials.4 In this case the high surface area of powders allows the use of techniques like electron paramagnetic resonance (EPR) which would be too difficult to use on single crystals because of the insufficient small number of paramagnetic centers present on the surface.5,6 Theoretical studies in particular if performed with –rst-principle approaches provide a useful complement to experiment for the understanding of the microscopic aspects of the metal/oxide interaction.7h13 The quantum-mechanical description of the interaction of metal atoms or clusters 5 Faraday Discuss.1999 114 209»222 209 2 This journal is( The Royal Society of Chemistry 2000 with the surface of oxide materials is still in its infancy and only a few –rst-principle studies have been performed on this subject so far.This is also because of the lack of accurate structural information. In the last few years however several experimental investigations on single crystal or thin –lm oxide surfaces have been reported opening the way to a direct comparison of observations with theoretical predictions.1h3 Another important aspect of the quantum-mechanical simulation of these systems (which severely limits the applicability of parametrized force-–elds and semi-empirical approaches) is the great variety of supporting oxides from very ionic e.g. MgO to largely covalent e.g. SiO2 or intermediate e.g. TiO or Al2O3 . Finally even when the geometri- 2 cal and electronic structure of the ideal oxide surface is well characterized this is not sufficient to fully describe the early stages of the metal deposition since these processes often involve defects low-coordinated sites and surface irregularities.Since many catalysts consist of small particles on high surface-area powders of SiO or Al2O3 an understanding at the microscopic level of the 2 metal/oxide interaction in particular at the defect sites is an essential prerequisite for an understanding of the catalytic activity. Recently we started a systematic quantum-mechanical study of the interaction of Cu atoms and clusters with regular and defect sites of the dehydroxylated SiO surface using ab initio Hartree» 2 Fock (HF) and gradient-corrected density functional theory (DFT) cluster calculations.14h16 The choice of studying Cu on SiO is due to the existence of experimental data on these systems17h20 2 and also to our recent studies on the electronic structure and spectral properties of the point defects in bulk silica.21h23 We have investigated the interaction of isolated metal atoms with regular and defect sites of SiO2 ; then we considered the deposition of small metal clusters.In this paper we brie—y review the results of our previous investigations,14h16 and we present new results on the interaction of Cu clusters on surface defects. The adsorption sites considered are regular bridging oxygens at the dehydroxylated surface 3Si»O»Si3 the non-bridging oxygen site (NBO) 3Si»O~ and the E@ defect corresponding to a Si singly occupied sp3 dangling bond 3Si~.Both the E@ and NBO defects are paramagnetic and therefore detectable by EPR spectroscopy. They have attracted great interest over the past 20 years because of their role in the degradation of Si/SiO2 interfaces in microelectronics devices or in the absorption of light in optical –bers.24 In fact very similar defects are present in the bulk of amorphous and crystalline silica,25 on the surface of mechanically activated SiO2 ,26 of SiO2 thin –lms,17 or of a-quartz single crystals surface after Ar` bombardment.27 Proof of the existence of these centers comes from typical EPR spectra and optical absorption bands at 2 eV (NBO) and 5.8 eV (E@). It has been suggested that these defect centers are the primary cause of the interface bond formation.2 Computational approach To describe the regular and defect centers of the SiO surface we used cluster models derived from 2 a ” -quartz. The cluster broken bonds were saturated by H atoms placed 0.98 from the O atoms along the O»Si bond directions of the perfect crystal. The H atoms were kept –xed during the geometrical optimization to provide a simple representation of the mechanical embedding of the solid matrix. The optimization was done without any symmetry constraint. The details of the cluster used for each site are given below. Below we also present a detailed analysis of the dependence of the results on cluster size. In general the adsorption of the metal cluster has little eÜect on the other geometrical parameters of the SiO substrate model; in particular the average 2 Si»O»Si angles remain around 145° i.e.very close to that of a-quartz.28 The cluster electronic structure was computed at the spin polarized DFT level using Beckeœs three parameter hybrid non-local exchange functional29 combined with the Lee»Yang»Parr30 gradient-corrected correlation functional (B3LYP). Hartree»Fock calculations were performed for comparative purposes. The basis sets used were all electron 6-31G* on all Si and O atoms31 and 3-21G32 on the terminal H atoms. For the Cu clusters we used an eÜective core potential ECP,33 which explicitly includes in the valence the 3s2 3p6 3d10 4s1 electrons of Cu. The Cu basis set is [8s5p4d/3s3p2d].33 For the atomic Cu adsorption a [14s11p6d/8s6p4d] all electron basis set was used.34 The adhesion energies of the cluster to the silica defects as well as the cohesive energies of the metal clusters were not corrected by the basis set superposition error BSSE.35 Only for the isolated Cu atom have we estimated the BSSE which is of the order of 0.5 .Since the bonding of the Cu atoms or clusters to the defect centers is de–nitely larger than this error the general Faraday Discuss. 1999 114 209»222 210 conclusions should not be aÜected. All the calculations were performed with the Gaussian94 program package.36 E@ (3Si~) 1-T model 2.365 2.238 1.00 2.32 3 Isolated Cu atoms on regular and defect sites of SiO2 3.1 Regular sites The interaction of isolated Cu atoms with the regular surface sites of silica has been modeled by the (HO)3Si»O»Si(OH)3 cluster Fig.1. Cu atoms were adsorbed ìon-topœ of the two-coordinated bridging oxygen. At the HF level the potential energy curve is purely repulsive ; at the DFTB3LYP level the interaction energies are of the order of B0.6 eV but after inclusion of the BSSE the binding energies are extremely small B0.1 eV Table 1; the equilibrium Cu»O distance is rather long. A full geometry optimization of the Cu atom position without imposing any constraint leads to an adsorption geometry where the adsorbate interacts with more than one surface oxygen.15 This is probably what happens in reality Cu atoms interact mainly through polarization mechanisms with two or more surface oxygens. Thus the regular sites of silica are very unreactive towards Cu atoms and in general towards metal atoms.It is worth noting that our calculations indicate for the adsorption of Cu on the O atoms of silica a bond strength even lower than that of the rather unreactive MgO (001) surface.37 This means that Cu atoms deposited on silica from the gas phase will not be trapped at the regular sites but rather will diÜuse on the surface remain trapped at a defect or re-evaporate. Fig. 1 Cluster model of (a) a bridging oxygen on the silica surface and (b) a Cu atom interacting with the bridging oxygen. Table 1 Adsorption properties of Cu on regular sites (bridging oxygen) NBO (3Si»O~) and E@ (3Si~) centers on the SiO surfacea 2 Bridging oxygen NBO (3Si»O~) 2-T model 2-T model Method r(SiO2»Cu)/” Unbound 2.460 Unbound 0.08 HF B3LYP HF B3LYP 1.944 1.855 2.56 3.79 D (BSSE)/eV D (BSSE)/eV e\dissociation energy corc e a r\Distance between Cu and bridging oxygen D rected by the BSSE (see text).Faraday Discuss. 1999 114 209»222 211 3.2 Non-bridging oxygens Non-bridging oxygens at the surface of silica 3Si»O~ probably represent the most important defects for the reactivity of mechanically activated SiO2 . By using metastable impact electron spectroscopy experiments (MIES) these broken bonds have been proposed as the centers where impinging metal atoms are trapped.38 The 3Si»O~ group has been represented by an (HO)3Si»O»Si(OH)2»O~ cluster see Fig. 4 in ref. 15. Cu forms strong covalent bonds with the 3Si»O~ groups having a value of 3.8 eV obtained from B3LYP calculations using a model with two tetrahedra and being BSSE corrected Table 1 and with an important polarization towards oxygen of the bonding electrons (partial charge transfer).A net residual positive charge forms on adsorbed Cu atoms as shown by the data of Mulliken population.15 The fact that the Cu atom donates charge to the substrate favors the appearance of other interactions with the neighboring surface O atoms. The Cu adsorbate binds mostly electrostatically with the exposed two-coordinated O atoms of the surface which then become eÜectively three-coordinated. This corresponds to the closure of a ring where the Cu atom binds to two surface oxygens; in a sense the Cu replaces a missing Si atom in the lattice. 3.3 Surface Eº centers Si sp3 dangling bonds 3Si~ or E@ centers are present on the (0001) and (1010) surfaces of a-quartz 27 on mechanically activated silica26 as well as on UHV grown thin SiO –lms.17 They can be 2 detected by means of EPR and show a characteristic hyper–ne coupling constant of B470 G with the 29Si nuclide and an intense electronic transition around 6 eV.26 The 3Si~ surface radical is very reactive towards molecular species ; a similar reactivity is also expected towards metal atoms.In fact on the [(HO)3Si»O]3Si~ model of a surface E@ center Fig. 2(b) Cu forms a rather strong bond of B1 eV in HF and B2.3 eV at the B3LYP level (BSSE corrected values Table 1). The formation of such a strong bond is re—ected in the rather short Si»Cu distance B2.2 ” Table 1.The bonding arises from the coupling of the Si sp3 singly occupied orbital and the metal 4s open shell orbital with formation of a bonding level. This level gives rise to a resonance in the band gap of the material well above the O 2p band.15 The character of the Cu»Si bond is largely covalent but the polarization of the bonding electrons is towards the metal at variance with the NBO center at least according to the Mulliken charges Table 1. Unlike the NBO case the Cu atom does not bind to other O atoms of the SiO surface because the Si»Cu bond does not have 2 the required —exibility for ring closure. Cu passivates the Si dangling bonds without leading to new ring structures. 3.4 Cluster size dependence and QM/MM models The results described above were obtained with relatively small clusters.It remains to be established to what extent the results obtained on small models can be extended to the real surface. To answer this question we studied the dependence of the results on the cluster size. This was done using clusters of various size but also by employing a hybrid quantum-mechanical/molecular mechanics approach QM/MM.14 The method used is the integrated molecular orbital molecular mechanics IMOMM method.39 The analysis was performed for various defects. Here we review the main features of the E@ center. Some of the clusters employed to describe the E@ defect are derived from the structure of a-quartz in particular (HO) Si~ and [(HO)3Si»O]3Si~ Fig. 2; these 3 models contain 1 and 4 tetrahedral Si atoms respectively and are thereafter denoted for brevity as 1-T and 4-T.We also considered a molecular model of silica (H7Si8O12)~ octahydrosilasesquioxane OSQ which possess a ìcubicœ Si8O12 unit Fig. 2(c). OSQ and derived compounds by substitution of terminal H atoms with alkyl or organometallic groups can be considered as molecular models of SiO surfaces.40 In the clusters derived from quartz the posi- 2 tion of the H atoms was –xed. No constraint except for symmetry was imposed in the optimization of OSQ. The calculations were performed at the restricted (Open) HF and (spin polarized) DFT-B3LYP levels. Since the emphasis here is on the cluster size dependence the results were not corrected by the BSSE. We described the (»O)3Si unit nearest to Cu as QM the rest as MM see Fig.2. The Faraday Discuss. 1999 114 209»222 212 Fig. 2 Models of the E@ center 3Si~ (a Si dangling bond) on the silica surface. (a) 1-tetrahedron model; (b) 4-tetrahedra model; (c) octahydrosilasesquioxane model. When a hybrid QM/MM approach has been used the QM part has been enclosed within a grey line. broken bonds of the QM and MM regions were always –lled with H atoms. The IMOMM calculations were performed with a program built from modi–ed versions of the Gaussian92/ DFT41 (QM part) and MM3(92)42 (MM part) programs. MM calculations used the MM3(92) force –eld.43h45 For further details see ref. 14. In the section 3.3 we have shown that Cu binds to an E@ center through direct coupling of the singly occupied Si sp3 dangling bond with the Cu 4s orbitals.Even at the HF level and using the smallest (HO) Si~ QM model the bonding is relatively strong 1.58 eV (but the overestimate due to 3 the BSSE is about 0.6 eV) with a Si»Cu distance of 2.37 ” Table 2. The Si»O distances B1.67 ” are only slightly elongated with respect to the free cluster 1.65 ”. Modest changes are found in the Si»O»Si angles. At the B3LYP level there is an increase of D and a shortening of the Si»Cu bond e length while the other geometrical parameters remain essentially stable. This shows that the bonding is described in a qualitatively correct manner at the HF level. For this reason we have restricted the analysis of the cluster size dependence to HF results. Going from the minimum 1-T cluster to the larger 4-T one Fig. 2 does not result in signi–cant changes in the adsorption properties Table 2.In particular the adsorption energy computed at the HF level goes from 1.58 eV to 1.64 eV by changing the cluster size Table 2. Also the OSQ model gives bonding properties nearly identical to those of the smaller model; the binding energy in fact is 1.63 eV and the Si»Cu distance 2.347 ” nearly coincident with the HF values obtained with the other clusters see Table 2. This provides a clear sign of the local nature of the Si»Cu bonding and of the minor eÜect that this bond has on the surface structure. 213 Faraday Discuss. 1999 114 209»222 Table 2 Adsorption properties of Cu atoms on a E@ center 3Si~ on the silica surface as a function of cluster sizea a(Cu»Si»O)/ b(Si»O»Si)/ degrees Method Model r(Si»Cu)/” r(Si»O)/” degrees D /eV e 2.37 2.24 2.34 (HO)3Si~ 1-T QM HF (HO)3Si~ 1-T QM B3LYP [(HO)3Si»O]3Si~ 4-T QM HF QM HF 1.58 3.08 1.64 1.63 1.98 1.61 128 124 145 154 125 143 111 110 113 114 113 109 1.67 1.66 1.66 1.66 1.66 1.65 (H7Si8O12)~ OSQ 2.35 [(HO)3Si»O]3Si~ 4-T QM/MM HF 2.36 QM/MM HF 2.34 (H7Si8O12)~ OSQ a QM\quantum-mechanical treatment ; MM\molecular mechanics treatment.All the results described so far have been obtained with a full QM treatment. To test the validity of the hybrid QM/MM approach we considered the 4-T surface model and the OSQ cluster. In these two clusters the QM part coincides with the minimum 1-T cluster ; the rest is treated at the MM level.The results obtained with the QM/MM models are close to the fully QM ones. However some small changes are found in D and in the Si»O»Si angle for the 4-T e QM/MM model Table 2. In fact D is about 0.3 eV higher than in all the other models con- e sidered and the Si»O»Si angle is smaller by about 20°. A closer inspection shows that this energy diÜerence is due to the MM part of the cluster.14 Despite the small inaccuracy in one of the models adopted the QM/MM approach provides qualitatively similar results at a much lower computational cost. To summarize the results show that the bond of Cu with the E@ center of the SiO surface is 2 local ; the results do not change signi–cantly as the size and shape of the cluster is varied. The hybrid QM/MM approach works very well if the de–nition of the QM part is such that all the important quantum-mechanical interactions between surface and adsorbate are included.When this is done the results are extremely close to those of a full QM treatment. The success of the mixed QM/MM approach is clearly due to the local nature of the Cu/SiO bond specially at 2 defective sites. 4 Cu clusters interacting with SiO Eº centers 2 In this section we consider the interaction mode of small Cu clusters from Cu to Cu5 with the 2 E@ center represented by the (HO)3Si»O»(HO)2Si»O»(OH)2Si~ cluster Fig. 3(a). The results for the E@ are compared with those on the adsorption of Cu clusters on the NBO sites reported previously.16 We have seen above that the bonding of a single Cu atom with these two centers is rather strong.It is interesting therefore to examine the modi–cations occurring in a small cluster as a consequence of bonding with the surface. The Cu clusters were geometrically optimized at the DFT-B3LYP level in the gas-phase and on the oxide support. 4.1 Cluster geometries 2) in Table 3 is somewhat larger than that of the Cu atom At the Cu /SiO interface the clusters are directly bound to the E@ and NBO centers with rather n 2 diÜerent distances r(Si»Cu) B2.34^0.06 in E@ and r(O»Cu) B1.95^0.07 ” in NBO Table 3. However all the clusters are anchored to the surface through more than one Cu atom see Figs. 3»5 and Fig. 2»6 in ref. 16. This is a general characteristic ; in fact in addition to the direct O»Cu or Si»Cu covalent bonds weaker interactions occur between the metal cluster and the bridging oxygens of the surface.In NBO the partial charge transfer from Cu to SiO leads to a depletion of 2 electronic charge from the metal cluster. The distance between one of the Cu atoms of the cluster and a bridging oxygen see r(Cu»O directly interacting with the NBO see r(Cu»O1) in Table 3. When the Cu clusters interact with a Si dangling bond however the distances of the cluster from the bridging oxygens are de–nitely larger than for the NBO center.16 In the case of E@ the interaction with the bridging oxygens is probably weaker than for NBO and mainly due to the polarization within the metal group induced by the negatively charged O atoms. Faraday Discuss. 1999 114 209»222 214 Fig. 3 (Top) (HO)3Si»O»(OH)2Si»O»(OH)2Si~ model of a Si dangling bond (E@ center) on the silica surface.Small spheres represent the terminal hydrogen atoms dark spheres oxygen atoms large spheres silicon atoms. (Bottom) Model of a supported Cu atom (larger sphere) on an E@ center 3Si»Cu. We consider now the geometrical changes within the metal unit induced by the interaction with the substrate. The addition of a second atom to the 3Si»Cu complex results in the formation of a supported Cu dimer Fig. 4(a) ; the ground state of 3Si»Cu is doublet while gas-phase Cu has a 1 ” &g ` ground state. The Cu»Cu distance in the supported molecule 2.34 ” is about 0.1 longer 2 than in gas-phase 2.26 ”. Notice that the same molecule adsorbed on an NBO center shows an elongation of almost 0.2 ” consistent with a stronger bond with this surface defect and a more Table 3 Selected bond distances of NBO the Cu/SiO interface 2 r(Si»O1)/” r(O1»Cu)/”a r(O2»Cu)/”b r(Si»Cu)/”a r(O2»Cu)/”b a Shortest distance(s) of non-bridging oxygen O with the Cu atom(s) of the cluster.1 b Shortest distance of a bridging oxygen O2 from a Cu atom of the cluster. 2 (3Si»O~) and E@ (3Si»~) centers interacting with Cu clusters at n 3Si»O»Cu 3Si»O~ 3Si»O»Cu4 3Si»O»Cu3 3Si»O»Cu2 1.606 1.865 1.614 1.869 1.673 » 1.622 2.017 2.076 2.116 1.676 2.001 2.038 3.033 2.256 2.090 » 3Si~ 3Si»Cu 3Si»Cu3 3Si»Cu4 3Si»Cu2 2.279 2.315 2.097 2.329 2.193 2.337 2.205 Faraday Discuss. 1999 114 209»222 3Si»O»Cu5 1.633 1.994 2.004 2.595 3Si»Cu5 2.376 2.217 215 3 cluster (larger spheres) on a E@ center 3Si»Cu2 .The optimal Cu»Cu ”. C (Bottom) Model of a supporteducluster (larger spheres) on an E@ center 3Si»Cu3 . Fig. 4 (Top) Model of a supported Cu2 distance is given in The optimal Cu»Cu distances are given in ”. The existence of a second bonding interaction between the cluster and a bridging oxygen of the silica surface has also been represented by a solid line. pronounced perturbation of the metal dimer. The second Cu atom of the dimer interacts weakly with a bridging oxygen Fig. 3; the Cu»O distance 2.20 ” is very close to that found for the NBO case Table 3 and the spin is largely localized on the Cu unit Table 4. Gas-phase Cu is bent 3 2 C2v with an internal angle of 75.7° and Cu»Cu distances of 2.326 ” 3 ` is a closed shell equilateral triangle with Cu»Cu distances of 2.394 ”.The addition of 2 3Si»Cu surface complex (doublet) results in the closed shell 3Si»Cu3 3 while Cu a Cu atom (doublet) to the system Fig. 4(b). The distances within supported Cu are considerably elongated with respect to the gas-phase unit but the cluster retains the C2v structure with two long and one short Cu»Cu distances Fig. 4(b). On an NBO center on the contrary we found a diÜerent adsorption mode with two Cu atoms of Cu interacting with the NBO (see Fig. 4 in ref. 16) ; on NBO Cu assumes 3 3 an almost perfect equilateral triangular geometry with internal angles of 60^1° and Cu»Cu distances of about 2.4 ”.Thus a quite diÜerent structure is found for the same cluster interacting with the two defect centers. Free Cu has a singlet ground state and a planar rhombic structure. The Cu»Cu distances are of 2.455 and the short Cu»Cu diagonal of the rhombus is 2.308 When deposited on an E@ . center of SiO remains nearly planar but one of the Cu»Cu distances is markedly elongated to 2.72 ” Fig. 5(a). This corresponds to having two Cu atoms of the cluster interacting directly with the surface one forming a bond with the Si atom and the second with a bridging oxygen atom Fig. 5(a). Also in this case the structure is diÜerent from that of an NBO center where Cu ” ” 4 2 Cu4 4 Faraday Discuss. 1999 114 209»222 216 Table 4 Charge q and spin distribution in free and SiO supported Cu clusters 2 3Si»O»Cu 3Si»O~ 3Si»O»Cu5 3Si»O»Cu4 3Si»O»Cu2 n 3Si»O»Cu3 [0.71 [0.05 » [0.71 ]0.08 » [0.71 ]0.13 » [0.37 » 0.93 (O) q(O) q(Cu)a Spin density [0.71 [0.01 0.03 (O) 0.96 (Cu4) [0.70 ]0.11 0.08 (O) 0.89 (Cu2) 3Si~ 3Si»Cu 3Si»Cu5 3Si»Cu4 3Si»Cu3 3Si»Cu2 ]0.84 ]1.36 [0.11 ]1.17 [0.18 ]1.06 [0.24 0.78 (Si) q(Si) q(Cu)a Spin density ]1.17 [0.11 0.11 (Si) 0.86 (Cu4) ]1.18 [0.21 0.18 (Si) 0.73 (Cu2) a Average values.assumes a pseudo-tetrahedral shape with Cu»Cu distances scattered in the wide range 2.38»2.61 ”; the structure on the NBO can be better described as that of a bent rhombus (butter—y) (see Fig. 5 in ref. 16).3Si»Cu has a doublet ground state with the unpaired electron almost entirely delo- 4 calized over the four Cu atoms with little spin density on the Si atom of the surface Table 4. cluster (larger spheres) on an E@ center 3Si»Cu4 . The optimal Cu»Cu ”. C (Bottom) Model of a supporteducluster (larger spheres) on an E@ center 3Si»Cu5 . Fig. 5 (Top) Model of a supported Cu4 distances are given in The optimal Cu»Cu distances are given in ”. The existence of a second bonding interaction with a bridging 5 oxygen of the silica surface has been indicated by a solid line. 217 Faraday Discuss. 1999 114 209»222 The last cluster considered is Cu5. C5 u has a planar trapezoidal structure obtained by Free the internal angles are close to 60° 5 2.45 ” are practically coincident with ” and very similar to those of the 3Si»O»Cu surface complex adding a Cu atom in the plane containing the rhombic Cu4; and the Cu»Cu distances go from 2.404 to 2.475 ” C Fig.5(b). The structure of supporteduwas 5 obtained starting from that of 3Si»O»Cu by replacing the NBO oxygen by the –fth Cu. The 4 geometry optimization leads to an almost —at pentamer which resembles that of free Cu5 Fig. 5(b). This structure is not too diÜerent from that found on an NBO center. Cu is bound with two 5 Cu atoms to the surface with the Si atom of the E@ center and with a bridging oxygen see Fig. 5(b). On average the Cu»Cu distances of supported Cu those of the free cluster 2.44 4 (NBO). This suggests that the geometrical distortions within the metal cluster due to bonding with diÜerent substrate defects disappear quite rapidly as the cluster size increases.4.2 Adhesion atomization and nucleation energies Ead\[[E(Cun/SiO2)[E(Cun)[E(SiO2)] (positive values of Ead correspond from cluster to cluster are due to the open ad 3 Cu5 and which favors the direct coupling of the unpaired electron on 4 have to Small Cu clusters more polarizable than a single atom are expected to interact more strongly with the non-defective SiO surface. Still the role of defects for the diÜusion adhesion and nucle- 2 ation processes is crucial. All Cu clusters considered from Cu to Cu interact with the E@ center 2 5 with adhesion energies which go from 1.7 to 3.3 eV Table 5. These values are always smaller than for the same clusters interacting with NBO Table 5.The adhesion energies (Ead) were computed for the fully optimized supported clusters with respect to the ground state of the equilibrium gas-phase clusters to bound states). The relatively large oscillations in E shell character of Cu Cu the metal with that of the paramagnetic surface center ; closed shell clusters Cu and Cu 2 ìopenœ their con–guration in order to form a direct covalent bond. It is also possible that polarization interactions with the two-coordinated bridging O atoms diÜer from cluster to cluster. The energy required to atomize the cluster is de–ned as De e /atom\[M[E(Cu /SiO n 2)[nE(Cu) [E(SiO )]/nN. The D /atom in metal clusters increases with cluster size and converges to the 2 cohesive energy of the bulk metal for very large metallic aggregates.For a supported cluster the atomization energy provides a measure of the additional stability of the cluster due to the bond at the interface. The values of D /atom reported in Table 5 show an additional stabilization of the e supported compared with the free cluster which is more or less constant for all clusters. Even for a supported pentamer there is a non-negligible contribution from the bond at the interface to the overall stability of the cluster towards atomization. The stronger bonds of the clusters with the NBO centers compared to the E@ explain the larger value of D /atom although the eÜect is not e very pronounced Table 5. An important quantity determining the mechanism of cluster growth is the nucleation energy Enuc de–ned as the energy gain due to the addition of an isolated Cu atom to a supported Cun cluster (Enuc\[[E(Cun/SiO2)[E(Cu)[E(Cunv1/SiO2)]).Recently accurate microcalorimetric E D atomization Table 5 Adhesion E ad /atom and nucleation nuc ener- e gies of gas-phase and supported Cu clusters on NBO and E@ centers at the SiO surface 2 Cu Cu3 Cu5 Cu4 Cu2 E /eVa ad D /atom/eVb e NBO 3Si»O~ E@ 3Si~ Free NBO 3Si»O~ 4.71 3.04 1.41 2.35 2.02 1.83 3.55 2.58 1.31 2.19 1.95 2.21 3.46 3.32 1.00 2.16 2.11 1.00 3.16 1.65 1.01 2.59 1.83 2.02 3.96 2.52 » 3.96 2.52 » Enuc/eVc E@ 3Si~ Free NBO 3Si»O~ E@ 3Si~ 3.00 2.68 2.30 2.02 1.30 2.67 1.21 1.14 » » /atom\[M[E(Cu /SiO2)[nE(Cu)[E(SiO2)]/nN.e n c E a b E Dad\[[E(Cun/SiO2)[E(Cun)[E(SiO2)]. nuc\[[E(Cun/SiO2)[E(Cu)[E(Cun~1/SiO2)]. Faraday Discuss. 1999 114 209»222 218 3Si»Cu5 . Fig. 6 Valence density of states of a model of the E@ center on the silica surface of a free Cu cluster and of a The zero of the scale has been aligned with the top of the O 2p band. 5 supported Cu cluster 5 measurements of the heat of adsorption of a metal atom to a metal cluster supported on an oxide surface have been reported.46 Therefore the nucleation energy is a quantity that is becoming available also through experimental studies even for small aggregates. So far heats of adsorption of Cu atoms on Cu/MgO have been measured.46 On oxide surfaces nucleation is believed to occur through diÜusion of isolated atoms or eventually dimers; therefore it is useful to compare the nucleation energy for free and supported clusters.The addition of an extra Cu atom leads to a stabilization that is larger for the supported than for the free Cu clusters Table 5 with the exception of the dimer (the energy gain for the process Cu]Cu]Cu is obviously larger than for the 2 3Si»Cu]Cu]3Si»Cu one because of the closed shell nature of the Si»Cu bond). This is an 2 important conclusion which shows the role of the substrate in the growth process of a supported particle. In fact since isolated Cu atoms are weakly bound to the regular SiO surface they will 2 diÜuse with low activation barriers. The diÜusion process will stop only at defects or at sites where nucleation has already started.It should be noted that the nucleation energy seems to increase with the cluster size for the supported clusters going from about 1 eV for the formation of the dimer to about 2.7»3.0 eV for the pentamer. In other words the energy gain of the process SiO2»Cun]Cu]SiO2»Cun`1 seems to increase for larger n. This is partly because as the cluster becomes larger the cohesive energy increases due to the increase of the coordination number of the atoms in the cluster ; another eÜect however is that larger clusters are more polarizable and the interactions with the bridging oxygens also increase. Of course by further increasing the cluster size the nucleation energy will tend –rst to that of the corresponding isolated Cu particles and then for larger crystallites to the cohesive energy of the bulk metal.In other words the perturbation induced by the strong bond with the surface defect is rapidly screened by the conduction band electrons of the metal cluster and the electronic modi–cations induced by the bond with the surface defect disappear for aggregates of a few tens of atoms. 4.3 Valence band and gap states Recent MIES experiments on metal atoms and clusters deposited on oxide surfaces have shown the appearance of new states in the gap of the oxide material.38 These states can be attributed to the population of new defects at the surface of the oxide e.g. F centers in MgO to the presence of new metal»oxide bonds at the interface or to features typical of a metal particle (occupied d states etc.).It is therefore of interest to analyze the presence of states in the gap after metal deposition. In principle cluster calculations are not adequate to determine the gap energies because of the lack of periodic boundary conditions. Nevertheless it is possible to estimate the size of the gap in a cluster calculation from the HOMO»LUMO separation derived from one-electron orbital energies. It is well known that one-electron energies do not provide a good approximation of an excited state problem like the determination of the optical gap. In particular Hartree»Fock calculations largely overestimate the gap (by a factor 2 easily) while DFT approaches underestimate it. With a cluster model of the non-defective SiO surface we compute a ììHOMO»LUMOœœ gap of 2 219 Faraday Discuss.1999 114 209»222 5 . Fig. 7 Valence density of states of a model of the NBO center on the silica surface of a free Cu cluster and The zero of the scale has been aligned with the top of the O 2p band. 5 3Si»O»Cu of a supported Cu cluster 5 8.1 eV underestimated by only 10% with respect to the experimental one B9 eV.47 The computed gap is smaller than in HF but larger than in a pure DFT calculation because of the use of a hybrid DFT approach (B3LYP) where the HF exchange is partially mixed in with the DFT exchange. Kohn»Sham orbital energies can also be used to determine the valence density of states DOS by convolution with Gaussian functions of the one-electron energy spectrum. In Figs. 6 and 7 we have reported the DOS of the SiO support of free Cu 3SiO2»Cu5 surface 5 and of the 2 complex for both the E@ and NBO centers.The Cu DOS has been determined for the optimal 5 structure of the gas-phase cluster ; the DOS of Cu with the structure adopted on the support is 5 almost identical at least with the kind of Gaussian broadening used (0.75 eV). The DOS of the E@ center shows a small feature in the gap about 2.5 eV above the top of the O 2p valence band Fig. 6. This state corresponds to the singly occupied Si sp3 level. On the contrary no gap states are present in the DOS spectrum of NBO Fig. 7 since the O dangling bond gives rise to a resonance within the O 2p valence band. The qualitative diÜerences in the O 2p valence DOS of the SiO2 substrate clusters Fig. 6 and 7 re—ect the slightly diÜerent shape of the substrate models.For both interface bonds 3Si»Cu (E@) and 3Si»O»Cu (NBO) a series of new levels of dominantly Cu 3d 5 5 character appears above the O 2p valence band and extends into the gap. These states are located about 1 eV above the top of the O 2p valence band. The shape and the position of these states is virtually identical in the two cases and similar to those of the unsupported Cu cluster. This shows 5 unambiguously that the gap features are due to the supported metal particle and not to the interface bond. Looking at similar DOS curves for smaller clusters (not shown) we observe a progressive broadening of the peak above the O 2p levels due to metal states. Therefore it is expected that by growing even larger clusters these features will then appear as broad bands rather than as well resolved peaks.5 Conclusions We have performed density functional calculations on the interaction of small Cu clusters with regular and defect sites of the surface of dehydroxylated silica. The non-defective sites the bridging oxygen atoms 3Si»O»Si3 are rather unreactive towards adsorbed metal atoms. This is fully consistent with measurements of the sticking coefficient of Cu on SiO Zhou et al. found that at 300 2 . K only one third of the initially incident Cu atoms stick to the surface ;19 Xu and Goodman observed that the sticking depends markedly on the temperature varying from 0.6 at 90 K to 0.1 at 400 K.17 Both studies agree with the fact that the bonding of Cu with the clean surface is weak and that sticking occurs only at the defect sites.Indeed strong bonds form between Cu atoms and surface defects. We have considered here two of the dominant point defects at the silica surface the Si dangling bond (the E@ center) and the non-bridging oxygen (an O dangling bond). Both centers are paramagnetic and are possible sites where the nucleation of the cluster begins. We have reported new data on the interaction of small Cu clusters with the E@ center and we have Faraday Discuss. 1999 114 209»222 220 compared the characteristic features with those of the same clusters interacting with the NBO centers. Cu clusters containing from 2 to 5 metal atoms form strong bonds with both E@ centers and non-bridging oxygens although the adhesion energy with these latter centers is higher.The interaction arises in part from the formation of a covalent polar bond between the metal cluster and the paramagnetic center either E@ or NBO and in part from the polarization interaction of the cluster electron density which leads to direct although weaker bonds with a two-coordinated oxygen of silica. As a result due to the interface bond the fragmentation energy of the supported cluster increases compared to the free gas-phase counterparts. The eÜect of the interface bond on the metal»metal distance of the cluster is not large and tends to disappear as the size of the cluster increases. Hence while the Cu»Cu distances in supported Cu and Cu are larger than in the free 2 3 clusters in Cu and in particular in Cu the average Cu»Cu distances are close to those of the 4 5 unsupported clusters.On the other hand the shape of the supported clusters can diÜer substantially from that of the gas-phase units. This is true in particular for the Cu clusters interacting with the NBO centers where the interaction is stronger. We expect that by growing larger Cu particles these will assume nearly spherical three-dimensional structures since the metal»metal bonds will dominate over the weak electrostatic Cu»SiO interactions. In this respect point 2 defects on the SiO surface act as strong anchoring sites for the entire cluster limiting the diÜusion 2 process and favoring nucleation. These results are consistent with a Volmer»Weber growth mode of Cu overlayers on silica with formation of 3D particles.48 The bonding of Cu clusters with surface defects has some consequences that in principle could be observed experimentally.The most important one is that new states appear in the wide gap of SiO These states are located 1»2 eV above the top of the O 2p valence band and extend into the 2 . gap. Depending on the cluster size however the feature corresponding to the occupied metal states can give rise either to sharp or to broad bands. As the cluster size increases we expect a considerable broadening of the features. Thus a strong coverage dependence of the width of the states in the gap is expected. Acknowledgements N.L. is grateful to the ììDireccioç General de Recerca Generalitat de Catalunyaœœ for supporting her visit to the University of Milano.G.P. thanks the University of Barcelona for an invited professor position. Financial support from the Spanish ììMinisterio de Educacioç n y Cienciaœœ project CICyT PB95-0847-C02-01 ììAccioç n Integrada Hispano-Italiana HI1998-0042œœ Italian INFM (Project PAIS) Italian MURST (Co–n Area 03) ììGeneralitat de Catalunyaœœ projects 1997SGR00167 is fully acknowledged. Part of the computer time was provided by the ììCentre de Supercomputacioç de Catalunyaœœ CESCA and ììCentre Europeu de Paralel-lisme de Barcelonaœœ CEPBA through a research grant from the University of Barcelona. References 1 H. J. Freund Angew. Chem. 1997 109 444. 2 R. Lambert and G. Pacchioni (ed.) Chemisorption and Reactivity on Supported Clusters and T hin Films NAT O ASI Ser. Ser. E 1997 331.3 C. T. Campbell Surf. Sci. Rep. 1997 27 1. 4 L. Marchese S. Coluccia G. Martra E. Giamello and A. Zecchina Mater. Chem. Phys. 1991 29 437. 5 E. Giamello M. C. Paganini D. M. Murphy A. M. Ferrari and G. Pacchioni J. Phys. Chem. 1997 101 971. 6 J. H. Lunsford Catal. T oday 1990 6 235. 7 G. Pacchioni and N. Roé sch Surf. Sci. 1994 306 169. 8 G. Pacchioni and N. Roé sch J. Chem. Phys. 1996 104 7329. 9 N. Lopez and F. Illas J. Phys. Chem. B 1998 102 1430. 10 V. Musolino A. Selloni and R. Car J. Chem. Phys. 1998 108 5044. 11 C. Li R. Wu A. J. Freeman and C. L. Fu Phys. Rev. B Condens. Matter 1993 48 8317. 12 Y. Li D. C. Landgreth and M. R. Pederson Phys. Rev. B Condens. Matter 1995 52 6067. 13 I. Yudanov G. Pacchioni K. Neyman and N. Roé sch J. Phys.Chem. B 1997 101 2786. 14 N. Lopez G. Pacchioni F. Maseras and F. Illas Chem. Phys. L ett. 1998 294 611. 15 N. Lopez F. Illas and G. Pacchioni J. Am. Chem. Soc. 1999 121 813. 16 N. Lopez F. Illas and G. Pacchioni J. Phys. Chem. 1999 in press. 221 Faraday Discuss. 1999 114 209»222 17 X. Xu and D. W. Goodman Appl. Phys. L ett. 1992 61 1799. 18 X. Xu J. W. He and D. W. Goodman Surf. Sci. 1993 284 103. 19 J. B. Zhou H. C. Lu T. Gustafsson and E. Garfunkel Surf. Sci. L ett. 1993 293 L887. 20 J. B. Zhou T. Gustafsson and E. Garfunkel Surf. Sci. 1997 372 21. 21 G. Pacchioni and G. Ierano` Phys. Rev. L ett. 1997 79 753. 22 G. Pacchioni and G. Ierano` Phys. Rev. B 1998 57 818. 23 G. Pacchioni G. Ierano` and A. Marquez Phys. Rev. L ett. 1998 81 377. 24 T he Physics and T echnology of Amorphous SiO2 ed.J. Arndt R. Devine and A. Revesz Plenum New York 1988. 25 L. Skuja J. Non-Cryst. Solids 1998 239 16. 26 V. A. Radsig Chem. Phys. Rep. 1995 14 1206. 27 F. Bart M. Gautier F. Jollet and J. P. Durand Surf. Sci. 1994 306 342. 28 Y. Le Page L. D. Calvert and E. J. Gabe J. Phys. Chem. Solids 1980 41 721. 29 A. D. Becke J. Chem. Phys. 1993 98 5648. 30 C. Lee W. Yang and R. G. Parr Phys. Rev. B Condens. Matter 1988 37 785. 31 R. Ditch–eld W. J. Here and J. A. Pople J. Chem. Phys. 1971 54 724. 32 M. S. Gordon J. S. Binkley J. A. Pople W. J. Pietro and W. J. Hehre J. Am. Chem. Soc. 1982 104 2797. 33 P. J. Hay and W. R. Wadt J. Chem. Phys. 1985 82 299. 34 A. J. H. Wachters J. Chem. Phys. 1970 52 1033. 35 S.F. Boys and F. Bernardi Mol. Phys. 1970 19 553. 36 M. J. Frisch G. W. Trucks H. B. Schlegel P. M. W. Gill B. G. Johnson M. A. Robb J. R. Cheesman T. A. Keith G. A. Petersson J. A. Montgomery K. Raghavachari M. A. Al-Laham V. G. Zakrzewski J. V. Ortiz J. B. Foresman J. Cioslowski B. B. Stefanov A. Nanayakkara M. Challacombe C. Y. Peng P. Y. Ayala W. Chen M. W. Wong J. L. Andres E. S. Reploge R. Comperts R. L. Martin D. J. Fox J. S. Binkley D. J. Defrees J. Baker J. P. Stewart M. Head-Gordon C. Gonzalez and J. A. Pople GAUSSIAN 94 Gaussian Inc. Pittsburgh PA 1997. 37 N. Lopez F. Illas N. Roé sch and G. Pacchioni J. Chem. Phys. 1999 110 4873. 38 M. Brause D. Ochs J. Gué nster T. Mayer B. Braun V. Puchin W. Maus-Friedrichs and V. Kempter Surf. Sci. 1997 383 216. 39 F. Maseras and K. Morokuma J. Comput. Chem. 1995 16 1170. 40 C. Marcolli and G. Calzaferri J. Phys. Chem. B 1997 101 4925. 41 M. J. Frisch G. W. Trucks H. B. Schlegel P. M. W. Gill B. G. Johnson M. W. Wong J. B. Foresman M. A. Robb M. Head-Gordon E. Repogle R. Gomperts J. L. Andres K. Raghavachari J. S. Binkley C. Gonzalez R. L. Martin D. J. Fox D. J. Defrees J. Baker J. P. Stewart and J. A. Pople GAUSSIAN 92/DFT Gaussian Inc. Pittsburgh PA 1993. 42 N. L. Allinger MM3(92) QCPE Bloomington IN 1992. 43 N. L. Allinger Y. H. Yuh and J. H. Lii J. Am. Chem. Soc. 1989 111 8551. 44 J. H. Lii and N. L. Allinger J. Am. Chem. Soc. 1989 111 8566. 45 J. H. Lii and N. L. Allinger J. Am. Chem. Soc. 1989 111 8576. 46 J. T. Ranney D. E. Starr J. E. Musgrove D. J. Bold and C. T. Campbell Faraday Discuss. 1999 114 195. 47 N. F. Mott J. Non-Cryst. Solids 1980 40 1. 48 X. Xu S. Vesecky and D. W. Goodman Science 1992 258 788. Paper 9/02374G Faraday Discuss. 1999 114 209»222 222

ISSN:1359-6640    

DOI:10.1039/a902374g

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

General Discussion Prof. Jennison opened the discussion of Dr Mackrodtœs paper When we were working on high-T c materials we found that explicit inclusion of correlation such as through a many band extended Hubbard model which could be solved on a small cluster was essential to obtain the correct value for the superexchange as determined for example by neutron scattering.1 One important energy in this model is the screened two-hole interaction on the oxygen ion which we determined from the O(KVV) Auger lineshape to be UB6 eV. Since this quantity is likely to be similar for these materials I might expect antiferromagnetic properties such as the Neç el temperature to be poorly given without this added layer of theory. Could you comment on this please ? 1 E. B. Stechel and D.R. Jennison Phys. Rev. B 1988 38 4632. Dr Mackrodt responded This is an important point and I would like to mention three things in this connection. The –rst is that the inclusion of electron correlation beyond UHF certainly increases the superexchange coupling as Illas and co-workers have shown from cluster calculations. 1 Whether this is the case in solids to the same extent remains to be established. The second point is that as with other aspects of magnetism notably the local moment the superexchange coupling energy cannot be measured directly but has to be extracted from experiment on the basis of a model so that direct comparisons of calculated and ìmeasuredœ values need to be viewed with some caution. Finally there have now been UHF calculations for quite an extensive range of systems including MnO NiO a-Fe2O3 a-Cr2O3 KCuF3 CaCuO2 Sr2 CuO3 SrCuO2 LiCuO2 LiMnO2 LaMnO3 and all the known polymorphs of MnS and without exception these calculations have predicted the observed low temperature magnetic ordering.In addition UHF calculations for MnO and NiO2 also yield good agreement with the measured rhombohedral distortion below the Neç el temperature due to spin»lattice interaction. Where there are ìmeasuredœ values of the moment in every case other than for Cu (S\1 systems for which it 2) is known that zero-point quantum —uctuations renormalise the moment appreciably (in Sr CuO 2 3 for example Kojima et al.,3 have deduced a moment of B0.06 k compared with the formal value of 1 kB) the computed spin moments are in good agreement with the ìmeasuredœ values.B 1 I. D. Moreira and F. Illas Phys. Rev. B 1997 55 4129; C. de Graaf F. Illas R. Broer and W. C. Nieupoort J. Chem. Phys. 1997 106 3287. 2 M. D. Towler N. L. Allan N. M. Harrison V. R. Saunders W. C. Mackrodt and E. Apra` Phys. Rev. B 1994 50 5041. 3 K. M. Kojima Y. Fudamoto M. Larkin G. M. Luke J. Merrin B. Nachumi Y. J. Uemura N. Motoyama H. Eisaki S. Uchida K. Yamada Y. Endoh S. Hosoya B. J. Sternlieb and G. Shirane Phys. Rev. L ett. 1997 78 1787. Dr Shluger asked (1) Unrestricted Hartree»Fock (UHF) is known to overemphasise hole localisation in many cases and electron correlation is needed to address the question of hole localisation. Could you please comment on this ? (2) Charge transfer transition energies can be aÜected by lattice repolarisation due to the charge transfer.What eÜect could lattice polarisation have on your results ? Dr Mackrodt answered (1) UHF calculations do suggest a substantial degree of localisation which might be reduced by the further inclusion of electron correlation. However such localisation and the implications of this for the activation energy of hole hopping for example are certainly compatible with experimental data including the recent 7Li NMR measurements for LixNi1~xO.1 223 Faraday Discuss. 1999 114 223»243 This journal is( The Royal Society of Chemistry 2000 (2) There will certainly be some eÜect due to lattice repolarisation but separating this eÜect from the direct interaction of charge transfer excitons which result from the transition could be quite difficult.1 M. Corti S. Marini A. Rigamonti and F. Tedoldi Phys. Rev. B 1997 56 11056. Prof. Pacchioni asked How sensitive are direct and superexchange interactions to changes in the lattice parameter due to the epitaxial growth of the thin NiO –lm? Dr Mackrodt replied Superexchange interactions are extremely sensitive to changes in the lattice parameter so that one might expect substantial diÜerences in the stability of the magnetic order of NiO –lms grown epitaxially on diÜerent substrates such as MgO and CaO. Prof. Harrison said The stability of the localised state in Li Ni7O8 is due to a competition between the energy gain due to lattice distortion (0.3 eV) and the additional correlation energy of the totally symmetric (delocalised) state.In this case it seems very likely that the UHF solution is correct and a self-trapped hole forms. Dr Egdell said Experimentally Li-doped NiO is not a metallic conductor. Thus the holes are localised rather than itinerant. Dr Noguera said As regards the localisation of a hole in NiO Hartree»Fock (HF) and DFT calculations give contradictory results. While it is true that HF usually overestimates localisation eÜects as noted by Dr Shluger DFT overestimates delocalisation eÜects because the self interaction of the electrons is not correctly subtracted. It thus seems that standard ab initio methods at present are not fully satisfactorily in this respect. Prof. Thornton asked For low dimensional structures rumpling could be signi–cant.Do you have a feel for how they might aÜect your results ? Dr Mackrodt replied There is no experimental evidence of appreciable rumpling of NiO(100) but if there were circumstances in which this did occur I would expect the magnetism to be aÜected since the superexchange interaction is very sensitive to deviations from linearity of Ni(C)» O»Ni(B) and to a lesser extent the d]d excitation energies which is sensitive to the local coordination. However I would not expect any major changes in the electronic structure including the d electron con–guration single particle energy levels or high spin insulating behaviour nor in the nature and stability of hole states. Prof. Freund asked From an experimental point of view it is desirable to have predictions on the change in magnetic surface structures as a function of temperature.Are such predictions feasible ? Dr Mackrodt answered From the diÜerences in energy between diÜerent magnetic orderings it should be possible to give crude estimates of the changes in magnetic surface structures with temperature including that of the disordered state though the precision would not be high. Dr Egdell said The new technique of spin polarised metastable He atom diÜraction is now yielding information about magnetic ordering of surfaces. For example the surface ionic layer for NiO(100) shows a (6]n) magnetic superstructure not found in the bulk.1 With the related technique of inelastic scattering of metastable spin polarised helium atoms it might be possible to measure surface spin wave dispersions and superexchange parameters.2 1 A.Swan M. Maryowski W. Franzen M. Elbatanouny and K. M. Harbui Phys. Rev. L ett. 1993 71 1250. 2 M. Elbatanouny C. Murthy C. R. Willis S. Kais and V. Staemmler Phys. Rev. B 1998 58 7391. Dr Mackrodt responded These very exciting new developments will certainly have an important impact on our understanding of ultrathin –lms and provide fresh impetus for further more extensive calculations of magnetism at the surface and in thin –lms. Faraday Discuss. 1999 114 223»243 224 Prof. Campbell asked Your Fig. 9 shows a surface energy of B1.2 J m~2 for MgO(100). There are experimental values of 0.6»0.9 J m~2 for alumina and 0.3»0.6 J m~2 for silica.1 Based on similar wetting behaviour I expect MgO(100) to be close to the value of silica or B0.4 J m~2.Can you comment on this diÜerence? 1 C. T. Campbell Surf. Sci. Rep. 1997 227 1. Dr Mackrodt answered Iœm not clear as to why you expect the surface energy of MgO(100) to be close to the value of silica but our value is certainly within the range of experimental values (1.04»1.2 J m~2) reported by Tosi1 and as mentioned in our paper the temperature dependence is consistent with data for rocksalt (100) surfaces (ref. 58 of our paper). 1 M. P. Tosi Solid State Phys. 1964 16 1. Prof. Madey said In Fig. 9 you plot surface energy as a function of slab thickness and –nd convergence to a limiting value at B10 layers. Have you computed other properties of the NiO –lms (bandgap density of states magnetic properties) as a function of slab thickness ? At which thicknesses do you –nd convergence? Dr Mackrodt replied It is straightforward and relatively inexpensive to carry out atomistic simulations as a function of slab thickness and this we have done for quantities such as the surface phonon densities of states and impurity segregation free energies for which 10 layers takes us beyond the convergence limit.For quantities computed from –rst principles electronic structure calculations testing the convergence to the limit is very much more expensive. However such studies that we have been able to make suggest that for a three layer slab the central layer is a good approximation to the bulk (see Table 2 of our paper for example) and I would certainly expect the local electronic properties of the central layer of a –ve layer slab to be virtually indistinguishable from those of the bulk.Prof. Diebold opened the discussion of Prof. Neddermeyerœs paper (1) What were the thickest –lms you could achieve in both the CoO and NiO case and can you comment on whether the –lms are n-type or p-type semiconducting? (2) Can you comment on the nature of the defects and the mechanism of motion? Do you think that some of the motion is tip-induced ? Prof. Neddermeyer responded (1) For the measured samples the maximum coverage was nominally 10 ML for both NiO and CoO. Samples of such thickness could not be measured easily however since instabilities of the tunnelling current frequently occurred possibly due to insufficient conductivity.The actual thickness of the –lms at the measuring position could not be determined and therefore 10 ML should be considered as an upper limit. The doping of the –lms will depend on the preparation conditions. In the case of NiO islands the I/V curves indicate n-type conductivity (see Fig. 3(b) of our paper) for CoO –lms the Fermi level was observed in the centre of the band gap (as determined from the dependency of the contrast as a function of the sample bias). (2) For CoO we observe two kinds of defects (see Fig. 7 of our paper) atomic defect structures (as emphasised by the circles) and more extended irregularities (as displayed on that part of the surface extending to the right upper corner of the images). While the former ones are probably related to vacancies in the surface layer the latter ones might result from small changes of the registry of the CoO –lms with regard to the Ag(100) substrate.It has to be emphasised that both kinds of feature are preferentially observed on surfaces which have not yet been fully annealed. In our opinion the changes seen in the images are thermally induced and might be caused by the sample»tip interaction only to a minor extent. Dr Castell said As referenced in your paper we have observed characteristic defects on Lidoped NiO(001) cleavage surfaces. We observe second nearest neighbour brightening around point defects (which we believe are Li dopants) and on top of S001T step edges. Do you observe similar defects on the NiO(001) –lms you have grown? 225 Faraday Discuss.1999 114 223»243 Prof. Neddermeyer responded In the case of the NiO –lms we observe defects which probably are related to vacancies in the surface layer (see Fig. 2 of our paper). A second nearest neighbour brightening is not seen for these defects. This would be consistent with the special nature of Li dopants in your previous studies which should not be present in our –lms. Dr Egdell asked Could you comment on the fact that the tunnelling spectrum in Fig. 3 of your paper is characteristic of an n-type semiconductor. Bulk NiO is usually considered to be a p-type material. Prof. Neddermeyer replied Although the tunnelling spectrum shown in Fig. 3(b) (full line) would be consistent with n-type conductivity in the islands the data might not be simply related to the electronic states of the NiO islands.While the diÜerences on the Ag(100) substrate and NiO islands may qualitatively be understood by the metallic and oxidic nature of the measuring position respectively for a detailed analysis of the spectra the nature of the tip has to be known more precisely. We believe that the characteristics shown in Fig. 3 are obtained with a tip which was not metallic on the apex. In this case the rise of the tunnelling current (at both polarities) will be shifted to higher values of the sample bias due to the additional band gap of the tip apex and a lack of electrons at the Fermi level. 2O3 Prof. Jennison asked In your Fig. 4 you show compact islands of oxidized Co on Ag(100). At lower temperatures one might expect dendritic growth to occur as Jué rgen Behm saw for Al islands on Ru(001).1 Have you looked at island shapes at lower temperatures? 1 R.J. Behm unpublished work. Prof. Neddermeyer replied We did not perform deposition experiments below room temperature. However our experiments showed that already deposition at room temperature does not lead to well ordered oxidic –lms. For deposition at even lower temperature other methods (e.g. pulsed laser deposition) probably have to be applied. Prof. Thornton asked When you observe the substrate by tunnelling through the CoO layer do you have a measure of the oxide –lm thickness ? Prof. Neddermeyer answered In our experience monatomic step heights as measured on threedimensional oxide islands agree with the expected value (around 0.22 nm).However measurements of Ag/oxide step heights are in—uenced by diÜerences of the density of states of Ag and the oxides and consequently show drastic changes with the sample bias. An accurate determination of the thickness of oxide –lms by the STM measurements alone is therefore difficult. Dr Shluger said I would like to comment that in the system which you are studying STM of oxide –lm on metal substrates image forces should be very strong. This can lead to ion instabilities at oxide surface and their jumps on the tip. Your remarks on tip changes during imaging supports this conclusion. I have two questions. How did the tip»surface interactions in your experiments aÜect the defect diÜusion shown in your video? Can you desorb ions from the surface at large applied voltages ? Prof.Neddermeyer replied At low measurement speed most of the defects moved to a neighbouring position while scanning the surface. Upon increasing the measuring speed the number of such diÜusion jumps per image decreased. This means that the defect movement can at most to a small extent be induced by the sample»tip interaction (whose in—uence cannot be excluded however). At higher sample»tip voltages U (e.g. U[3 V and U\[3 V) the oxidic –lms were often destroyed during the measurement. A systematic study of this eÜect has not yet been performed. Dr Weiss asked Fig. 4(a) displays an STM image taken after Co deposition at room temperature where four diÜerent grey levels can be seen. On which basis did you assign these levels to Ag and the O/Co precursor ? Faraday Discuss.1999 114 223»243 226 Prof. Neddermeyer answered The most direct evidence for Ag or oxide species as seen in the STM image can be deduced from step heights and contrast changes as a function of sample bias and tip conditions. In addition in some cases the oxidic layers showed a rougher surface structure as compared with the smooth appearance of the metallic substrate. Note that the Ag/Ag step height is always found in agreement with the expected value (0.2 nm). Dr Venables said With this preparation procedure can you exclude the possibility of Ni and Co metal buried in the Ag substrate ? For example the microstructure of Fig. 4(a) looks remarkably similar to that produced by depositing pure Co on noble metals (Cu or Ag) in the absence of oxygen.Prof. Neddermeyer responded Ni and Co atoms may indeed be incorporated in the uppermost atomic layer of the Ag(100) substrate. We have studied the growth behaviour of the clean transition metal on Ag(100) (without the presence of O) in each case. Surfaces as shown in Fig. 4(a) could only be obtained with sufficiently high partial pressure of O. To our knowledge and experience at room temperature clean transition metal/Ag(100) (or Cu(100)) systems show distinct diÜerences. For example the tendency of three-dimensional growth of the transition metal deposit may be recognised in contrast to experiment with the presence of O described above where a more two-dimensional growth mode of the deposit is found.Prof. Hayden opened the discussion of Prof. WoodruÜœs paper You have suggested that the similarity in the PhD for the two nitrogen transitions is evidence that they derive from the same species. Have you measured the PhD for the nitrogen transition which exists alone (which you ascribe to nitrogen atoms) and compared it with the NO derived levels ? Dr Lindsay responded We have recorded PhD data from the single N 1s feature at normal emission. We found that there was little or no oscillation within experimental error. This is in direct contrast to the two normal emission modulation functions extracted from the N 1s doublet feature which exhibit signi–cant oscillations. Prof. Thornton said You could presumably improve the precision of your measurements by using the O 1s signal for PhD.Do you resolve two or three O 1s peaks which would allow you to do this ? Prof. WoodruÜ responded We would certainly expect to be able to substantially improve the precision of the location of the O atom within the NO if we had measured the photoelectron diÜraction for the O 1s component (or components) associated with the molecule. Unfortunately in these experiments our spectral resolution was inadequate to make this separation so no such measurements were recorded. In the future we certainly expect that the availability of undulator radiation at a third-generation synchrotron radiation source such as BESSY II will provide the necessary combination of high —ux and spectral resolution to make these measurements possible.Prof. Freund said Angle resolved XPS studies in particular under grazing excidence have revealed indications for the presence of the O 1s ionisation of oxygen of adsorbed NO/NiO. However due to the substrate oxygen the signal is hard to discern from the background. Prof. Pettersson asked Could you amplify your comment that the presumed atomic nitrogen species does not give rise to oscillations in the spectra ? In light of the discrepancies between experiment and theory for this system it would be of importance to have as much information as possible also on diÜerent adsorbates. It might be that if you can get atomic nitrogen on the NiO as a decomposition product this would be easier to treat theoretically and could give insight into the speci–c problems associated with adsorption on NiO as compared to MgO.Would it be possible for you to make a determination of the structure of the proposed N/NiO product? Prof. WoodruÜ responded It might be possible to obtain structural information on the (presumed) atomic N species using photoelectron diÜraction but we did not make the necessary measurements. In general when we are conducting such a study we measure the PhD spectra in 227 Faraday Discuss. 1999 114 223»243 many (typically 10»20 or more) diÜerent emission directions. The modulations are typically strongest in directions correcting to 180° back-scattering from a near-neighbour in the substrate while in directions far from such geometries the modulations can be quite weak. The actual structure analysis is then typically conducted by –tting 5»10 of these spectra including some of those with the strongest modulations over a range of emission directions.In the present case our main concern was to establish that this N species was a distinct entity from the adsorbed NO. We recorded the associated PhD spectrum in normal emission and found weak modulations whereas both the N 1s peaks we attribute to the NO show strong modulations in this direction. This therefore implies that the emitting N atom is in quite a diÜerent local site. Indeed we might infer that it is not atop a surface Ni atom (so normal emission is not a favoured back-scattering direction). Without more data we really cannot say more about this species. 2~ at a much higher temperature.Prof. Joyner said The explanation involving photodissociation of NO and its subsequent reactivity that WoodruÜ proposes seems very reasonable and in line with the known surface chemistry of NO. It could be relatively simply con–rmed by temperature programmed desorption. If his explanation is solid he should observe desorption of N2O (probability at a relatively low temperature) and N Prof. WoodruÜ responded I agree that this would be interesting (although it is not clear that a temperature would exist at which N2O would be formed as a stable surface species) but no TPD measurements were attempted in our experiments. Prof. Friend asked (1) Is it possible that in ambient NO there are multiple species ? It is possible that other techniques e.g. IR would reveal multiple species on the surface ; therefore the two peaks in the absence of NO ambient may not be same as in a vacuum.Is it possible to perform photoelectron diÜraction in NO ambient? (2) To what extent will ì—oppyœ vibrational modes aÜect your precision ? Prof. WoodruÜ answered (1) From the present experiments we can only say that the surface species produced in the stable condition achieved by a low NO ambient pressure appears to be the same as on a freshly prepared surface at UHV so we believe all the previous characterisation work is valid for our surface. More generally the ability to conduct the photoelectron diÜraction in ambient pressures of reactant gas is limited by three factors the mean-free-path for the electrons passing from the sample to and through the electron energy analyser ; the maximum operating pressure for the electron (channeltron) detector ; the contamination of the beamline and ultimately the electron storage ring pressure.In practice I think one could devise ways of working up to about 10~6 mbar. (2) Large amplitude vibrational modes are certainly important in PhD and can limit precision. Vibrations introduce dephasing in the scattering interferences which are treated through a Debye» Waller factor. Because the scattering is basically a local process the Debye»Waller factors are dependent on the relative movement of emitter and scatterer and are thus directionally dependent. In fact the present case of a species adsorbed atop provides a good example of a situation in which we believe large amplitude vibrational modes do commonly have a signi–cant in—uence.Notice that the PhD spectra in the present case show strong modulations at normal emission but these amplitudes fall oÜ rapidly as the oÜ-normal emission angle is increased. We have seen this behaviour in many atop adsorbates on metal and semiconductor surfaces and an important factor appears to be the large amplitude frustrated translational mode parallel to the surface which introduces a large Debye»Waller factor for scattering events involving signi–cant components of the scattering path parallel to the surface thus damping out PhD modulations at oÜ-normal emission angles. Of course this also leads to lower precision in determining the mean position of the emitter along the directions of these large vibrations.Indeed we have found a few cases where it is not possible to distinguish between adsorption sites displaced by up to about 0.2 ” oÜ a high-symmetry position and large amplitude vibrational motions in the same direction. Prof. Freund said The system NO/NiO has been looked at with a variety of experimental methods (ref. 85 of the Introductory Lecture). From refs. 111 and 113 of the Introductory Lecture Faraday Discuss. 1999 114 223»243 228 it is clear that the system undergoes photo-desorption and both peaks in the N 1s ionisation spectrum belong to the same species. Prof. Friend asked Have there been other types of experiments to con–rm that there is a single NO species ? The existence of a single type of NO is very unusual because there are many possible types of NO binding.Prof. WoodruÜ responded There has been a great deal of characterisation of this adsorption system performed previously by a variety of methods (notably by Freund and co-workers as cited in the Introductory Lecture). I believe these show rather clearly that there is a single NO adsorbed species under the conditions of our study. As a general statement I should say that it is only sensible to address surface structural problems with PhD after this level of pre-characterisation and this has been our general policy. In the case of oxide surfaces this certainly will be a limiting factor for us in the near future especially as we also really need a reasonable starting model for the structure of the clean surface which is not always available.The PhD technique is specialised and while this gives it real strengths in the area of structure determination it also means it is blind to other complexities. In addition if one has multiple species or no clear idea of what the surface species are this would be a major added complication to the range of models to be tested in a PhD analysis. Prof. Jennison said I note that the error bars on the angular position of the NO are very large (could you comment on this ?) but I want to point out another source of information on the tilt. In electronically stimulated desorption the partition of energy into translational vs. rotational modes is quite sensitive to the tilt and considerable theoretical work has recently been done on this system.1 Concerning the height disagreement with theory this molecule is relatively weakly bound and perhaps dispersive forces not well described by DFT could play a role in causing DFT to be in error and while these forces could be included accurately in quantum chemical calculations here accurate geometric relaxation of the surface cannot be done because the cluster must be small.Perhaps a hybrid approach is indicated. 1 T. Kluner H.-J. Freund V. Staemmler and R. KosloÜ Phys. Rev. L ett. 1998 80 5208. Prof. WoodruÜ said As remarked in response to Thornton earlier our data for the present system comprise only N 1s PhD which means that the –rst order information is the N site on the surface. In general we would obtain the position of the O atom in the NO in a largely independent fashion from O 1s PhD but in the present experiment this was not possible because we could not separate this from the O 1s signal from the oxide substrate.The molecular orientation and N»O bond length in the present case can thus only be obtained from the weak intramolecular (mainly multiple) scattering eÜects which are really second-order eÜects. More generally of course we do measure PhD from each of the elements within the molecular adsorbate and obtain much better precision in such parameters. Prof. Pettersson commented I should point out that in the theoretical work by Staemmler et al. in ref. 1 the calculations were indeed carried out at the con–guration interaction (CI) level so that dynamical correlation was included.I have performed large-scale multireference CI calculations (unpublished work) using diÜerent models including embedded clusters with one or two Ni2` cations considering rumpling and possible dimer formation but the results always indicate physisorption rather than a bond formation. Furthermore in the comparison with CO chemisorption the theoretical results do not indicate any diÜerence in bonding; both show positive vibrational shifts in contradiction with experiment for NO and both show only weak bonding where experiment clearly shows both adsorbates rather strongly bound and with NO substantially more strongly bound than CO. In light of the recent TDS data from Wichtendahl et al.2 and the present contribution by Prof. WoodruÜ it seems very clear that we now have some very –rm experimental calibration points and that indeed something is missing in the theoretical description.Since the presently determined Ni ” »N distance is rather shorter than what has been obtained theoretically earlier (0.2 compared 229 Faraday Discuss. 1999 114 223»243 to CI calculations by Staemmler et al.1 I have made a set of calculations (also including correlation) at the proposed geometry; the model was an embedded NiO cluster and the NO was 5 assumed to tilt at 45°. The results are negative however for the doublet state the energy was found to increase in going to the shorter distance ; introducing rumpling with the Ni relaxing into the lattice had insigni–cant eÜects ; moving the NO to tilt between two oxygens similarly negligible ; allowing an expansion (10%) of the lateral distance between the surface oxygens beneath the oxygen of the NO gives a very small energy lowering; expanding (10%) the oxygens near the nitrogen gives somewhat more but still only in the meV range; –nally investigating the quartet state either transforming as A@ or AA ( C symmetry) gave higher energies.It should be pointed out s that although NO has a rich chemistry with e.g. dimer formation with a high electron affinity on many oxide surfaces or dinitrosyl formation this possibility can be ruled out by the available experimental data. Thus it would seem that some new idea or understanding of the properties of these surfaces is sorely needed for the theoretical description of the experimental data. 1 H.Kuhlenbeck G. Odoé rfer R. Jaeger G. Illing M. Menges Th. Mull H.-J. Freund M. Poé hlchen V. Staemmler S. Witzel C. Scharfschwerdt K. Wennemann T. Liedtke and M. Neumann Phys. Rev. B 1991 43 1969. 2 R. Wichtendahl M. Rodriguez-Rodrigo U. Haé rtel H. Kuhlenbeck and H.-J. Freund Surf. Sci. 1999 423 90; R. Wichtendahl M. Rodriguez-Rodrigo U. Haé rtel H. Kuhlenbeck and H.-J. Freund Phys. Status Solidi A 1999 173 93. Prof. WoodruÜ responded This discrepancy is clearly potentially important. The one parameter to which our measurements should be most sensitive and which should therefore yield high precision is the Ni ” »N distance. A diÜerence of 0.2 is clearly well outside our precision. As a very general comment (directed at theoreticians generally) it is clear that in comparisons of theoretically computed and experimentally determined structures a proper understanding of the errors is of paramount importance.Experimental surface crystallographers expend a lot of energy in attempting to quantify their errors and it would be a big help if one could do the same for theoretical values. Of course experimentalists usually quote only their random errors and theoreticians may argue that all of their errors are systematic and therefore not quanti–able. I can well imagine that this is the issue in the present case. I wonder more generally however whether the gradient of the total energy with change in a structural parameter in a theoretical calculation might provide some indication of how reliable diÜerent theoretically obtained parameter values might be expected to be.Certainly this should help if one can estimate the precision of the calculated (relative) total energies of the structures. Prof. Bowker said I notice that the limits on the tilt angle include the limit of a 90% tilt (normal molecule). This derives from the very —at dependence of the R factor on tilt angle between 40 and 90°. Is this due to error or due to a soft vibrational mode with large amplitude? Would high resolution angle scan photoelectron diÜraction at –xed photon energy give a better idea of the tilt angle ? Prof. WoodruÜ responded As no vibrational mode of this type is included in the calculations this cannot account for the observed insensitivity. The reason is probably more to do with the dependence of the scattering cross-section of the O atom with scattering angle.I should stress again however that our insensitivity to the NO orientation is not surprising in the present case because we have only measured the N 1s PhD; more surprising is that we have any signi–cant sensitivity at all ! I am sure we could obtain a much more reliable orientation with separate O 1s PhD data. However it is also true that one could try to use angle-scan X-ray photoelectron diÜraction (XPD) from the N 1s signal using a conventional laboratory X-ray source. Under these conditions the photoelectron diÜraction is dominated by intramolecular forward scattering and has been used quite successfully to determine some molecular orientations. In the present case however the tilt angle away from the surface normal is quite large so if the tilt is azimuthally random as we expect the forward scattering peaks from diÜerent molecules would be smeared out over an azimuthal ìridgeœ.This could lead to a very weak modulation perhaps not observable experimentally. Dr Carley communicated I have a comment about the assignment of the two N 1s features as arising from a single adsorbed species since work at CardiÜ and elsewhere studying NO adsorp- Faraday Discuss. 1999 114 223»243 230 tion on diÜerent metals as a function of exposure and temperature suggests that two diÜerent molecular species may coexist. These species have been tentaively identi–ed as NO adsorbed in ì linear œ and ìbentœ con–gurations (with some theoretical justi–cation) and I wonder how sensitive the experimental technique described in this paper would be in discriminating between such species especially given the large error quoted for the derived tilt angle for the N»O bond.As the authors comment the presence of a weakly bound (ì linear œ) and strongly bound (ìbentœ) species would explain the observed photon beam induced changes in a simple way and avoid the need for the complex reaction scheme which they have to invoke. Prof. WoodruÜ communicated in response Because the N 1s PhD data are primarily sensitive to the location of the N atom to the near-neighbour substrate backscattering atoms it is certainly true that a model which attributed both the N 1s features to NO in atop sites (at the same Ni»N distance) could be compatible with our data as shown in particular in Fig.2. One would need to perform further calculations on the level of agreement between the calculations for the diÜerent orientations and the PhD spectra from the individual N 1s peaks. On the other hand the low R factor for the comparison of the experimental PhD spectra from the two N 1s peaks the low R factor for the experiment»theory comparison for our optimum structure (0.09) and the large value for the R factor when these experimental data (averaged over the two peaks) is compared with the model of a perpendicular N»O bond (0.30»0.40 ; see Fig. 5 of the paper) suggests that this model will not prove satisfactory. I also believe that previous work on this surface (as opposed to on other metal surfaces) by other methods strongly suggests that only a tilted species is present.Moreover I should stress that our data are certainly not consistent with simple photodesorption of one of these species as implied in the second part of this question ; we remark in our paper only that this explanation in terms of two species appears ìat –rst sight œ to present an explanation. However the PhD measurements show clearly that the species which remains after extended irradiation under UHV conditions is clearly not an atop NO species as discussed in the paper and in the oral discussion. This alternative picture of the basic adsorption states therefore actually makes the situation more rather than less complicated. Prof. Finnis opened the discussion of Dr Renaudœs paper Can you say anything about the mis–t dislocations in the Ag/MgO interface ? I am recalling the TEM observations suggesting a semicoherent interface.Dr Renaud responded Yes indeed we have studied the interface of fairly thick (B1500 ”) Ag –lms on the MgO(001) surface by grazing incidence X-ray scattering.1 We con–rm that the interface is semicoherent with a well ordered array of interfacial mis–t dislocations which yields very nice rods of diÜraction from this interfacial superlattice. However the HRTEM study concluded an erroneous orientation and Burgerœs vector of the dislocations along S100T directions which was interpreted as arising from an alternate epitaxial site for Ag on MgO on top of O and on top on Mg. We have unambiguously demonstrated that the dislocations are instead oriented along S110T directions with 12 S110T Burgerœs vectors which is what is expected from the O-lattice theory and corresponds to only one adsorption site above O ions of the last MgO(001) plane.1 G. Renaud P. Gueç nard and A. Barbier Phys. Rev. B 1998 58 7310. Prof. Freund said You have convincingly shown how GIXS can be used to extract information on the static structure. Do you see routes towards the extraction of dynamical information from the data? Dr Renaud replied We have recently performed grazing incidence small angle X-ray scattering (GISAXS) measurements in real time in situ during the growth of several metals on an MgO(001) surface and dynamically. This was achieved thanks to a 2D CCD detector and to fairly low deposition rates of the order of 0.1 ” min~1.These were the very –rst dynamical and in situ GISAXS experiments which should provide very nice information on the morphology of the metallic islands during growth but not however on the structure. We intend to try GIXS measurements at wide angles using a 2D detector in the near future. However a lot has to be done in 231 Faraday Discuss. 1999 114 223»243 order to deduce quantitative structural data from such measurements and this is still an open –eld. Prof. Goodman said This method is obviously the most elegant and preferred approach for addressing metal structure on oxides. However many of us oftentimes are limited to more available techniques for such structural diagnostics such as TEM. Here however one can imagine serious problems with beam damage.With respect to Pd/MgO Henry has examined this system in great detail with TEM and the results seem to be entirely consistent with your results for Pd/MgO. Is this not the case and if so it would seem to imply that TEM very well may be generally useful for metals on oxides ? Dr Renaud responded I would say that it is an elegant method but not the most elegant one since every method has its drawbacks and limitations and GIXS can not provide all the answers by itself. In the case of Pd/MgO all our results are indeed perfectly consistent with those of Henry obtained by TEM. This is certainly because Pd is rather strongly bound to the MgO and because all specimen preparations and TEM studies of Henry were very careful.However when the metal is less bonded to the substrate such as in the Ag/MgO(001) case TEM studies can clearly lead to erroneous results either because of modi–cations of the interface during sample preparation or because the preferred state of the interface in a ” B50 thick cross-section sample defer from that of the sample before it was thinned (this is the case of the Ag/MgO(001) interface) or because of radiation induced mobility. Another clear advantage of the grazing incidence X-ray scattering technique is that it can be performed in situ during the growth and that at all stages of the growth the structure and the morphology can be characterised in great detail in a nondestructive way as opposed to TEM experiments. Prof. Jennison said You remarked that theorists often assume commensurate metal overlayers.I just want to point out that information obtained from such calculations is still useful. For example while the large lattice mismatch of Cu on TiO causes an incommensurate overlayer the 2 relatively small mismatch for Ru on Al2O3 would likely lead to large areas of commensurability well described by the calculations with strain then relieved by periodic mis–t dislocations. Dr Renaud responded I fully agree with your remark. However in the Ag/MgO(001) case that I mentioned the metal»oxide bond was weak as compared to the metal»metal one and to the surface tension eÜect. In that particular case it might be important to allow the Ag islands to relax in theoretical calculations as is experimentally observed at all stages of the growth despite the relatively small lattice parameter mismatch of 3%.Prof. Thornton asked In your data analysis you assume that the MgO(100) substrate is unchanged on metal adsorption. How realistic is this assumption given that the metal»oxygen interaction will be signi–cant ? Dr Renaud answered The MgO(001) substrate was assumed to be unchanged on metal adsorption mainly because the possible eÜect of a relaxation or a rumpling of the last layer would in any case be totally negligible as compared to the eÜect of metal adsorption. We have studied this relaxation and rumpling eÜect.1 Clearly even if the rumpling or relaxation would reach 10% the eÜect on the CTRs would still be negligible with respect to metal adsorption because they are only visible on the ìweak CTRsœ whose intensity is proportional to the square of the diÜerence between the atomic scattering factors of O and Mg i.e.4]4\16 electrons as compared to the square of the scattering power of Ag for instance of 47]47\2200! In addition all calculations and our experiments show that the MgO(001) surface is extremely stable with only a very small rumpling (B1%) and relaxation (B [0.6%). This is not expected to be strongly modi–ed upon adsorption especially Ag which is very weakly bound. Moreover Spiess2 has calculated the rumpling and relaxation induced upon adsorption of a single Ag atom on top of O. The relaxation was found to be zero and the rumpling 1.5% which tends to con–rm Faraday Discuss. 1999 114 223»243 232 this hypothesis of very small modi–cations of the substrate upon adsorption in the case of MgO(001).1 O. Robach G. Renaud and A. Barbier Surf. Sci. 1998 401 227. 2 L. Spiess Surf. Rev. L ett. 1996 3 1365. Prof. Campbell asked Can you please say what percentage of the Pd is in the –rst ML at the total coverage of 0.5 ML deposited and at 1.0 ML deposited (i.e. are 2D islands dominating at 0.5 ML Pd). Is it a uniform Pd ML at 1.0 ML of Pd or is some of the Pd in the second and third layer then? Dr Renaud answered The –ts of the CTRs (crystal truncation rods) allow determination of this percentage of Pd in each layer only for that portion of the Pd –lm that is lattice-matched parallel to the substrate. At 0.5 ML deposited 90% is in the –rst ML 10% in the second.At 1 ML deposited 50% is in the –rst ML B40% in the second and B10% in the third. Hence 2D islands indeed dominate at 0.5 ML Pd but for larger deposited amounts the second and third layers start to build. Dr Venables said You state in Section IV that these metals (Ag Pd Ni) are expected to grow on MgO(001) in the island growth mode yet you observe the –rst monolayer of Pd to be pseudomorphic. It is of considerable interest to know whether this 2D layer is a result of kinetic rather than thermodynamic mechanisms. Have you con–rmed that there is a kinetic barrier by annealing the (incomplete) –rst monolayer and observing that it transforms into 3D islands ? This can also be the case I believe for Ag as –rst discussed by Jupille and co-workers.1 In that case the 2D layer can be suppressed when impurities nucleate 3D islands earlier than otherwise.1 F. Didier and J. Jupille Surf. Sci. 1994 587 307; A. M. Flank R. Delauney P. Lagarde M. Pompa and J. Jupille Phys. Rev. B 1996 53 R1737. Dr Renaud replied In the case of Ag we have clearly observed that even a very small increase of the temperature above room temperature induces a fast diÜusion with an associated increase of island size and increase of the height-to-width ratio. We have indeed annealed a 2 ML thick –lm at 50 °C which readily resulted in large 3D islands. We have worked under extremely clean conditions and always found 3D growth even from the very beginning of deposition. However again only the –rst plane is occupied for a coverage less than 0.2 ML; then the second plane starts to build until B0.5 ML deposited and above 0.5 ML the third and further planes are also occupied.When we did not anneal under O so that residual C contamination was present the growth was 2 more 3D but the most important fact is that a large part of the Ag grew in the (111) instead of the (001) orientation. For a substrate with residual Ca segregation almost all the Ag is (111) oriented while on a very clean and —at substrate all the Ag grows in the (001) orientation. In the case of Pd only the –rst 0.5 ML is —at and pseudomorphic and this has clearly been shown by Henry to be a kinetic rather than a thermodynamic eÜect since growth at higher temperature readily results in 3D islands. Prof. Diebold asked How did you prepare the substrate ? Do you believe that you donœt have any hydroxylated point defects ? Dr Renaud replied The preparation of the substrates is described in ref.1. MgO(001) 15]15]0.5 mm3 supplied by Earth Chemical (Japan) are –rst annealed in air at 1500 °C for several hours which results in single crystals of very high crystalline quality as shown by highresolution X-ray measurements as well as X-ray topography. The Bragg peaks are all resolutionlimited to a FWHM of B0.001° even near the surface under grazing incidence conditions and in most cases there is only one grain through the whole sample. This also results in extremely —at surfaces but which are however contaminated because of surface segregation of many bulk contaminants (mainly Ca but also P Si K .. .). In order to remove these contaminants while keeping the high crystal and surface quality we next proceed to in situ ion bombardment in the UHV chamber at the temperature of 1500 °C which is required to allow a high surface mobility and 233 Faraday Discuss. 1999 114 223»243 thus keep a —at and crystalline surface during the process. The –nal procedure before measurements or deposition is a 20 min long annealing at B850 °C under a partial pressure of oxygen B1]10~4 Torr. The samples were characterised by X-ray re—ectivity CTRs measurements and STM all in UHV without any exposure to air and found to have an rms roughness of 2.2 ” which corresponds to a single atomic high step. The average terrace size between steps is 6000 ” although UHV STM revealed some single atomic-plane deep holes within the terraces a few tens of a- ngstroms wide.In the case of the Ag/MgO(001) and MgO(001) clean surface measurements great care was taken to ensure a very low pressure B2]10~11 Torr obtained by several weeks of bake-out followed by very careful degassing with the result of negligible partial pressure of water and hydrogen below 10~13 Torr according to our quadrupolar measurements. We also took care that the samples were always maintained in UHV after the ion bombardment preparation and hence I am fairly con–dent that there were no hydroxylated point defects. 1 O. Robach G. Renaud and A. Barbier Surf. Sci. 1998 401 227. Prof. Hayden said You observe a discontinuity in the interfacial distance and metal interplane distance during Pd growth at 4»5 ML.You ascribe this to relaxation of mis–t at the edge of islands. Since the measurement is sensitive to the buried interface could this relaxation not extend signi–cantly into the structure from the edge of the islands ? Is not sensitivity to the buried interface a signi–cant advantage that your technique brings to understanding such growth? Dr Renaud responded Of course the relaxation of the mis–t extends into the structure of the whole island even far from the edges. Additional metal planes are introduced near the edges of the islands in order to relax the mis–t but this induces a relaxation even at the buried interface in between two dislocations. What is very nice with the grazing incidence X-ray scattering technique is that it is sensitive both to the structure and to the morphology over the whole –lm thickness and at all stages of the growth.Hence in good cases it can assess the correlation between structural relaxation and change of morphology or change in the interfacial parameters such as interfacial distance and interplane distance. 2O3(0001) Dr Renaud commented This comment concerns the discussion on the accuracy of the relaxations determined by X-ray crystal truncation rods analysis on the Al surface published previously.1 The –rst point is that the measurements have been performed three times on two diÜerent samples. The –rst two measurements were done at LURE on a sample obtained by a 3 h long annealing in air at 1500 °C in two states before and after annealing up to 900 °C for 20 min in 10~5 Torr of O in the UHV chamber.The third measurement was done on the ID3 beamline at 2 the ESRF on another sample also annealed in air for 3 h at 1500 °C and next annealed at 900 °C for 20 min in 10~5 Torr of O in the UHV chamber. The ESRF experimental conditions were 2 de–ned such that the CTR measurements were very accurate with a large redundancy of measured structure factors. A total of 9 CTRs were measured at ESRF as opposed to only 3 at LURE with a lesser accuracy. Given the uncertainty on the parameters deduced from –ts of the LURE measurements the same interplanar relaxations were deduced from the three diÜerent measurements. It has to be mentioned that the CTRs are extremely sensitive to these relaxations.This sensitivity will be discussed in a forthcoming paper.2 However in all three cases the base pressure was B10~10 Torr with no particular care as regards the residual partial pressure of water or hydrogen. Since it was recently shown that hydrogen always seems to be present on the surface3 and that it could induce a ìde-relaxationœ it would certainly be interesting to perform new measurements making sure that no water nor hydrogen is adsorbed on the surface. Another point was the absolute accuracy on the experimental relaxation. It has to be mentioned that the –ts were not perfect which given the quality and the reproducibility of the measurements would indicate that the model is not sophisticated enough (we recall that it has only 5 main parameters 4 out-of-plane relaxations and one in-plane).Faraday Discuss. 1999 114 223»243 234 As regards comparison with the calculated relaxations I think the best way is to recalculate the CTRs by using the calculated relaxations and compare to the experimental data which I intend to do in the near future. 1 P. Gueç nard G. Renaud A. Barbier and M. Gautier-Soyer Mater. Res. Soc. Symp. Proc. 1996 437 15; P. Gueç nard G. Renaud A. Barbier and M. Gautier-Soyer Surf. Rev. L ett. 1997 5 321. 2 G. Renaud in preparation. 3 J. Ahn and J. W. Rabalais Surf. Sci. 1997 388 121. Prof. Flavell opened the discussion of Dr Kantorovichœs paper I would be grateful if you could comment further on the agreement between the experimental MIES spectra and the calculated SDOS. Given that the tunnelling probability of the electron from the surface to the He* is involved it seems surprising that the agreement between the two is so good.Dr Kantorovich responded The full answer to this question is presented in a paper1 that has recently been submitted for publication. Therefore we shall give only a very brief answer here. In the case of the He* projectile and MgO surface the main process which is responsible for the electron emission observed as the MIES spectrum is the Auger de-excitation (AD) process. In this process only one surface electron is involved which tunnels from the surface to –ll in an empty He 1s state. Therefore using very simple intuitive arguments one can show that the transition rate for this process R(E R) for a given position of the projectile R with respect to the surface and a certain kinetic energy E of the emitted electron to a good approximation is proportional to the surface DOS (SDOS) projected on this orbital (1) R(E R)PD1s(E[*E R) is the excitation (1s2]1s2s) energy of the He atom which is 19.82 eV.The *E\E where HeR[EHe projected DOS is de–ned as (2) D occ 1s(e R)\; o Stk ot1sT o2 d(e[ek) k where t e and are one-particle orbitals and energies of the surface electrons and the summation k k is carried out over all occupied states i.e. in the case of MgO all electrons from the upper valence band (VB) contribute. To calculate MIES one has to account for the fact that transition can actually happen with a certain probability at any time along the trajectory of the projectile.For a given trajectory R(t) of the He* atom the probability to undergo the transition at the time t is given by the product of the probability to survive before this instant (the so-called escape probability) on the incoming part of the trajectory (3) Pesc(R(t))\expA[P= dt@PdER(E R(t@))B R(t) and the probability to undergo the transition during the time interval between t and t]dt which is R(E R(t)) dt. Then integration with respect to time gives the total probability for the AD process for the given trajectory. Finally one has to account for all possible trajectories so that the MIES spectrum is given as (4) P(E)\TP0 R(E R(t))Pesc(R(t)) dtU ~= where S… … …T corresponds to averaging with respect to all possible trajectories and we integrate over the whole incoming part of the trajectory.The two functions in eqn. (4) (above) have a rather diÜerent behaviour the escape probability changes smoothly from unity at large distances from the surface to zero at close approach while the transition rate decays exponentially away from the surface. The integrand in eqn. (4) is therefore peaked at some distance from the surface Rmp called the most probable target distance which corresponds to the highest probability for the electron emission with energy E in the AD 235 Faraday Discuss. 1999 114 223»243 process. Our calculations demonstrate a weak dependence of Rmp on energy E. Therefore the time integral in eqn. (4) appears to be approximately proportional to the SDOS R(E Rmp) projected on the He 1s orbital centred at the target distance from the surface.Note that target distances depend on the particular trajectory and are in the range of 2 ” »4 above the surface. The exponential decay into the vacuum of the electronic wavefunctions t in eqn. (2) is deter- k mined by their binding energies [ek . The states with large binding energies (near the bottom of the VB) decay faster and are eÜectively cut oÜ in the spectrum. Therefore the largest contribution to the MIES spectra is made by the states in the gap and near the VB top. Since the SDOS and the projected DOS R(E Rmp) have similar structures the relative position of defect related peaks in the gap with respect to the top of the VB in the MIES spectra can be predicted with good accuracy by SDOS.This conclusion has been checked in ref. 1 where we directly compared the calculated SDOS and MIES for a peroxide defect O22~ formed at the MgO(001) surface upon adsorption of an O atom which binds to a surface oxygen (see also ref. 55 in our paper). It was found that a defect related feature above the top of the VB present in the SDOS also appears in the MIES at the same position with respect to the top of VB. We think that the agreement between the SDOS and the experimental MIES demonstrated in our paper in the case of Mg adsorption on the MgO(001) surface also supports this conclusion. 1 L. N. Kantorovich A. L. Shluger P. V. Sushko and A. M. Stoneham Surf. Sci. 1999 in press. Prof. Hermann asked Did you allow the atoms in the vicinity of the F-centres to relax and how large was the geometric eÜect.What was the eÜect on the electronic structure of the F-centre ? How does an adsorbing Mg stabilise near the F-centre ? Dr Kantorovich answered Yes we did. For all systems studied the complete geometry relaxation was performed as described in detail in Section 3.2. In the case of the F-centre at the MgO(001) surface the relaxation appears to be insigni–cant (about 1% of the Mg»O distance) due to well localised electronic density in the anion vacancy which is equivalent to two electrons. Note that both valence and the F-centre electrons contribute to the charge in the vacancy as explained in ref. 54 of our paper. When a Mg atom is added above the F-centre it pulls out some electronic density from the vacancy and forms bonding and antibonding r-type states which can be well recognised in the calculated system DOS.The stabilisation energy (see Table 2 of our paper) is only 0.21 eV which is half of that for the perfect surface. We did not calculate however the barrier between the two con–gurations (above the terrace oxygen and the F-centre) which we expect to be of the order of 0.1»0.2 eV. Prof. Pacchioni said The adsorption energy of a Mg atom on an anion vacancy is only 0.5 eV according to your calculations (Table 2). An anion vacancy in MgO F2` is very electron de–cient and based on simple electrostatic arguments I would expect the following process to occur. F2`]Mg]F`]Mg` The Mg` ion can then become stabilized at a nearby oxygen. Since both Mg` and F` are characterized by the presence of unpaired electrons this can be modeled only considering a triplet state.Dr Kantorovich replied Indeed simple electrostatics suggest that one may expect signi–cant charge transfer to the anion vacancy from the adatom due to a more favourable Madelung potential in the vacancy. According to our calculations this does not happen. When a Mg atom is added to the MgO(001) surface with an anion vacancy and the singlet state of the whole system is considered it indeed loses some electronic charge which is pulled towards the vacancy. However by integrating the electron density in the vacancy we were able to account for only about 0.5 electron there. Note that the Mg atom is not adsorbed directly above the vacancy but is displaced along the [110] direction towards the nearest oxygen as shown in Fig.7(c) of our paper. Faraday Discuss. 1999 114 223»243 236 We also considered this system in the triplet state and found that it lies 4.3 eV higher in energy. In this state the Mg atom is atop the nearest to the vacancy oxygen. We did not study the distribution of the electronic density in the triplet state so that I cannot comment on this. The fact that the charge transfer from Mg to the vacancy is not very signi–cant as the simple electrostatic argument would suggest is explained by at least two factors. First of all the large charge transfer would lead to creation of both positively charged F` centre at the surface and the Mg` adatom nearby and this would contribute with a positive Coulomb interaction energy.Secondly the electron affinity of the surface vacancy is much smaller as compared to the ionisation potential of the Mg atom which is 7.6 eV. We have calculated the electron affinity of the surface anion vacancy using the embedded cluster method (described in refs. 16 and 17 of our paper) and found that it is only 3.4 eV. Prof. Joyner asked What are the relative densities of steps and kinks in the experimental study compared to surface vacancies ? I would expect there to be many more steps than vacancies therefore adatoms have little choice other than to link at steps almost irrespective of the energetics. Dr Kantorovich answered We think that Prof. Joyner is absolutely right and the number of sites available for adsorption at steps and terraces is much larger than the possible number of vacancies at the surface.This is also supported by the STM images of similarly prepared surfaces of MgO –lms presented in ref. 29 of our paper. They clearly show that these surfaces are very rough and contain 3D MgO islands and a lot of steps within the islands. Our results demonstrate that steps and terrace sites are the most important for the growth of Mg clusters on these surfaces. Prof. Kempter added The MgO –lm preparation procedure used in our experiments leads to a small number of point defects in general including vacancies. Normally we start with a Mg-rich MgO –lm. The –lm is exposed to molecular oxygen and then to a subsequent annealing procedure. As a consequence the resulting surface is inert to the exposure of additional oxygen as well as to temperature.This preparation procedure produces valence band spectra both in MIES and UPS that are identical to those for MgO(001). In particular both EELS1 and MIES (ref. 21 and 24 of our paper) do not display any features in the band gap above the VB maximum of the MgO –lm that could be attributed to point defects as in particular to F` and F centres. Therefore we expect that the number of extended defects steps in particular is indeed much larger than that of point defects. Note that in principle the adatoms could link to terrace ions as well. The energetics shows however that this is not favoured at the initial stages of the Mg cluster formation. 1 M. C. Wu C. M. Truong and D. W. Goodman Phys. Rev.B Condens. Matter 1992 46 12688. Dr Noguera said How do you explain the dramatic change in the Mg»Mg interaction that you obtain between the gas phase state and the adsorbed state ? Could your calculations reproduce the weak bonding of the small unsupported clusters (which should be of the van der Waals type below a critical size) ? Dr Kantorovich replied We did not perform a comprehensive study of free Mg clusters. However we did consider a free Mg molecule and indeed found a very small binding energy of 2 0.15 eV at the interatomic distance of 3.75 ”. There is every reason to believe that the binding of other free Mg clusters will be also weak. I believe the situation is diÜerent at the MgO surface due to covalent bonding between the metal atoms and the surface oxygens.Prof. Pettersson said It seems very unlikely that the gas phase Mg dimer bonding should have any in—uence on the geometry at the surface. Since in the gas phase the bonding is only through van der Waals interactions it is very weak 429.6 cm~1.1 This energy is so much smaller than the computed interaction with the substrate that the geometry should be determined only by the bonding to the substrate. The situation could change if the dimer is ionized but the neutral system represents interaction between two closed-shell systems and is very weak. 237 Faraday Discuss. 1999 114 223»243 1 K. C. Li and W. C. Stwalley J. Chem. Phys. 1973 59 4423. because when more than one Mg atom is adsorbed there is a distribution of distances d E Dr Kantorovich responded I believe this is not entirely true.The Mg»Mg bond is indeed very weak. However one has to consider here also the whole potential energy curve of this system EMghMg(d) between them. We –nd in our calculations that the energy MghMg(d) changes weakly for distances d[d0 where d0\3.75 ” and is the equilibrium distance between the two metal atoms in a free molecule. However the repulsion of Mg atoms at closer distances shows much steeper character. The typical distances between adsorbed Mg atoms at the surface (which are indeed primarily determined by the interaction with the surface oxygens) are in the range of 3.1»3.3 ” i.e. they are closer than the gas-phase equilibrium distance d0 . At this distance range the repulsion between two Mg atoms is about 0.1 eV.Therefore it is not bonding but rather repulsion between Mg atoms which is the second factor aÜecting the geometry at the surface. Prof. Asscher opened the discussion of Prof. Campbellœs paper Would it be possible to measure single Cu atom heat of adsorption by avoiding surface diÜusion by cooling the sample? Is it not expected that a single atom would be partially ionized and therefore its binding energy should be larger than that of a neutral atom? Maybe even larger than a neutral dimer? Prof. Campbell responded There is pretty good evidence from related metal-on-oxide systems (ref. 9 of our paper) that the metal adatoms are cationic when present at such tiny coverages (\2% of a monolayer) and that they remain isolated. However they adsorb as nearly neutral species as the coverage increases (although still at extremely low coverages) where they are also observed to cluster together (ref.9 of our paper). This suggests that isolated adatoms are more stable when cationic than when neutral as you propose. Our –rst data point for measuring their heats of adsorption averages the heats of adsorption for the –rst B2% of a monolayer contained in that –rst pulse of the metal atom beam. By the end of that pulse we generally already have clustering of the metal atoms but the –rst bit is probably cationic. The low heat for this –rst pulse suggests that if these cationic adsorbed metal adatoms have a higher heat of adsorption than adatoms in the clusters it is not dramatically higher. Prof. Freund said You have made a very important contribution to our understanding of metal oxide interaction and I expect major insight from the method you presented.Is it conceivable to combine a calorimetric measurement with mass selected exposure of metal clusters to the surface ? Prof. Campbell replied Thank you. Yes possibly. I think some groups are now preparing low-energy beams of mass-selected clusters of sufficient intensity to make such measurements feasible. I would guess that nice signals could be obtained if the —ux in such a beam were sufficient to deposit B3]1011 atoms of metal per pulse within a pulse width of no more than 0.4 s and at a repetition period of no shorter than 2 s. With a good contact our heat detection limit (B3 standard deviations) is about 2]10~8 J so in principle even much smaller —uxes could be measured but the signal-to-noise ratio in the results would not be so beautiful.Lately we have seen even much much lower detection limits than that with an aluminium foil as the sample substrate. Detectivity seems to be higher for softer samples which makes sense since it depends directly on the heat transfer coefficient at the contact between the polymer detector and the sample. Prof. Kempter said Our results for Mg/MgO indicate a strong dependence of the heat of adsorption on the presence of extended defects as in particular steps. I wonder to what extent your results obtained for Cu/MgO (–lm) are representative for Cu on a MgO(100) single crystal. In fact your values obtained for the heat of adsorption compare very well with ours for Mg/MgO where MgO is indeed a rough surface.Prof. Campbell responded I agree. We probably have a highly defective MgO(100) surface. Since to my knowledge there is no way to get a single crystal of MgO(100) that is 1 cm in diameter but thin enough (\6 lm) to use for adsorption calorimetry we were forced to use a 3.0 Faraday Discuss. 1999 114 223»243 238 nm thick MgO(100) thin –lm grown on an ultrathin Mo(100) single crystal. Such –lms have highly defective surfaces and give broad LEED spots indicating average terrace widths of \10 nm. Prof. Jennison said You mentioned that the metal Auger spectrum can be shifted several eV in energy if the island sizes are small on an oxide substrate due to limitations in –nal state screening.I would like to show some data from the group of JeÜ Kelber which illustrate this eÜect. Note in the lower portion of Fig. 1 for Cu deposition on SiO2 just the eÜect you mentioned the peak in the Auger parameter plot moves continuously with deposition time (as the metal island sizes increase) until it meets the asymptotic position of metallic copper. Now note the contrast with Cu deposited on the heavily hydroxylated sapphire surface in the top portion of the –gure no shift occurs with deposition time but a second peak of metallic Cu grows starting when greater than 1/3 ML is deposited. This indicates to us that the –rst peak is oxidized Cu(I) and is not due to an island size eÜect. 1 J. A. Kelber et al. Surf. Sci. in press. Fig. 1 Cu(LMM) evolution during Cu deposition on (a) sapphire(0001) and (b) SiO with deposition rate at 2 0.03 ML Cu min~1.Deposition temperature\300 K. Due to diÜerential charging on sapphire surface the Auger parameter for Cu(0) on sapphire is diÜerent from that on SiO2 . (Reproduced from ref. 1 with permission.) 239 Faraday Discuss. 1999 114 223»243 Prof. Hayden replied An alternative to a single shifting peak with coverage would be that there is a growth of a second peak at high coverages. Note that the high coverage result spectrum shows a distinct shoulder. Prof. Campbell also replied to Prof. Jennison Yes that is one possible explanation. The result you show from Kelberœs group is however unlike the result obtained by Madeyœs group also for Cu on alumina (ref. 36 of our paper) where a continuous shift in the Auger parameter with increasing coverage from 1848.4 to 1851.1 eV was observed and attributed entirely to –nal state eÜects.My group also observed a large increase in the Cu Auger parameter up to 1851.6 eV with coverage of Cu on Zn(0001)»O which was proven by work function and band bending measurements to be due entirely to diÜerences in –nal state screening as neutral Cu clusters increased in size (ref. 38 of our paper) and not due to the change from Cu(I) to Cu(O). The persistence of an Auger peak due to Cu(I) at high Cu coverages on such a stable oxide is rather unusual but it could be understood if there were a small amount of oxygen impurity in the background gases which might oxidise the adsorbed Cu. Prof. Madey addressed Prof.Campbell I am curious about your high sticking probability ([0.9 for Cu on MgO(001) which is higher than the values reported by others on cleaved MgO(001) and on MgO –lms similar to yours. You attribute this to your higher incident —ux. Can you rule out the possibility that you have a higher defect density than the other investigators ? Do you have adequate signal-to-noise that you could check this by measuring the sticking probability at lower —uxes? Prof. Campbell responded Yes I think that the reason might be that we had higher defect density than in ref. 29 of our paper which used a very nice bulk crystal surface prepared by cleavage in UHV and reported a sticking probability of only 0.5. Our models for the coverageand temperature-dependences of sticking probabilities of Pb on MgO(100) (ref.22 of our paper) which are similar to models in ref. 29 would predict that higher island density and therefore defect density should increase the sticking probability. Since we prepared our MgO(100) following the exact same recipe as in ref. 26 it is less likely that the defect density could explain the diÜerences between the value of 0.82^0.05 reported there and our value of [0.995. It is interesting that this value stays constant at B0.82 in ref. 26 between B400 and 100 K and does not increase upon cooling as expected (and as observed above 400 K until it saturates at B0.82 at 400 K). The saturation at 0.82 suggests a possible calibration problem in ref. 26 wherein the Cu TPD intensity corresponding to a unit sticking probability was assigned to a value of 0.82 instead.I assume that this could happen due to diÜerences in mass spectrometer tuning or angular distributions between the calibrant and the real experimental test case. Prof. Goodman said It is noteworthy that your calorimetry data at the low coverage limit suggests that the MgO –lms synthesized in your experiments have a relatively low density of defects i.e. 1%. This is encouraging with respect to using –lms rather than cleaved or polished oxide surfaces for chemical and physical surface science measurements. Prof. Campbell replied Yes I agree. I think these are good surfaces for most tests. However our kinetic models mentioned above predict that 1% defect sites can rather dramatically aÜect the sticking probabilities.Prof. Finnis asked Would there be scope with your kind of calorimetry to investigate the heats of any phase transitions in thin –lms perhaps induced by the adsorption? Prof. Campbell answered Yes I think our sensitivity and temperature control has become so good that we could also follow temperature-induced phase transitions and those would be very exciting applications of the calorimeter. Dr Henderson said In your data for Pb adsorbed in the clean vs. hydroxylated/hydrated Mg(100) surface the heat of adsorption on the latter is less than that of the former. Is it possible that some of the heat generated by metal adsorption could have gone to desorption of the Faraday Discuss. 1999 114 223»243 240 layer thus rendering a lower apparent heat of Pb adsorption? Have you looked for H2O OH/H2O desorption or done TPD after the fact to test for this ? Prof Campbell responded No we havenœt done that but itœs a good idea.As with all these adsorption energy measurements we do it would be far better to have more information about the structure than we do. This is a –rst-generation instrument and we are hoping to have more structural characterisation tools on later versions. Prof. Kirschner asked If you had a system showing layer-by-layer growth could you imagine observing oscillations of the heat of adsorption with one monolayer periodicity ? Prof. Campbell replied Yes We have actually seen something like that for the case of Pb on Mo(100) which grows layer-by-layer for 2 monolayers.There is a rather sharp change in the heat of adsorption near the completion of both the –rst and second monolayers (ref. 21 of our paper). Dr Venables said You have discussed the possibility that you might measure the adsorption energy for Cu adatoms on MgO at low coverage. May I say probably in agreement with you that you will be able to do this in a room temperature experiment since neutral Cu adatoms are very mobile with an activation energy expected to be in the range 0.1»0.2 eV atom~1 (B10»20 kJ mol~1). With these low values island density of a few ]1012 cm~2 are formed at a coverage below 10~3 ML. In the comparable case of Pd/MgO(001) recent AFM experiments coupled with a rate equation analysis have put an upper limit of 0.2 eV atom~1 from the observation of the nucleation density at TB200 K.1 Could you comment on what one might expect if Cu` adions are present in addition to neutral Cu adatoms and could you observe any such eÜects ? 1 J.A. Venables G. Haas H. Brune and J. H. Harding Mater. Res. Soc. Symp. 1999 570 51. Prof. Campbell responded I agree the activation energies for diÜusion of an isolated metal adatom are very tiny and one needs a very low temperature to prevent them from clustering at 0.02 ML. Like you I am scared that 100 K may not be cold enough. We can work at smaller pulses now (down to 0.002 ML) albeit with reduced signal-to-noise so that may help. In principle we can do these experiments at 20 K or below although we can not deliver the liquid He needed for that yet. Prof. Catlow opened the discussion of Prof.Pacchioniœs paper I note that in your simulations you held the positions of H atoms –xed which I am sure is the correct procedure. Do you have any feeling for how the results are in—uenced by diÜerent choices of geometry of your cluster or in other words diÜerent choices of model for the surface structure ? Prof. Pacchioni responded The general features of the interaction should not change by changing the model of the substrate since the Cu atoms and clusters form relatively strong local bonds with the paramagnetic point defects at the silica surface. However we observed that as a consequence of a partial charge transfer from the Cu atoms to the non-bridging oxygen atoms at the surface the Cu atoms and clusters interact also with some neighbouring bridging oxygens with formation of ìringœ structures.The stability of these ring structures will depend on the topology of the surface and on the strain present in the rings. Other models of the surface where more relaxation is possible could result in larger adsorption energies. Prof. Pettersson said I like the fact that the coupling to the rest of the crystal matrix has been included and investigated in your models. The restrictions induced by the lattice connectivity on the displacements of atoms in response to the formation of defects or bond dissociation processes are important and should be included in the description. We have recently investigated the dissolution of the siloxane bridge in quartz and –nd that the barrier is strongly aÜected by the connectivity of the participating silica units.For instance in going from a singly connected to doubly and triply connected units the barrier increases by 5 and 10 kcal mol~1 respectively.1 This is due to the reduced freedom to relax the structure in the transition state. Similar considerations should apply in your study and I am glad that this point is brought up and illuminated in your paper. 241 Faraday Discuss. 1999 114 223»243 1 A. Pelmenschikov H. Strandh L. G. M. Pettersson and J. Leszczynski to be published. Dr Shluger said For modelling adsorption from the gas phase for example Cu cluster surface centre the correct presentation of the relative energies of the surface defect and adsorbing atom states is important. However your cluster model which does not take into account the crystalline potential does not treat the surface defect states correctly.How could this aÜect the adsorption energy and character of adsorption? Prof. Pacchioni responded Our assumption is that the charge separation which is present within a cluster model of the silica surface includes the most important contributions to the local potential around the defect centre and that the contribution of the long-range potential is not large in particular if one wants to simulate the surface of amorphous silica. The assumption that the Madelung –eld is not of major importance in this material is based on the fact that several properties of point defects in silica (optical excitations hyper–ne interactions defect formation energies vibrational spectra etc.) have been nicely reproduced using cluster models saturated by H atoms,1h7 without including the crystalline potential.In this respect we consider SiO a 2 material characterised by covalent polar bonds certainly not an ionic solid. Of course since the bond of a Cu atom at the defect sites has a partial charge transfer character in particular when we consider the non-bridging oxygen site this can be aÜected by the Madelung potential of the crystalline phase. In this respect some change in the absolute value of the adsorption energy can be expected when this term is properly taken into account. 1 G. Pacchioni and G. Ierano` Phys. Rev. L ett. 1997 79 753. 2 G. Pacchioni and G. Ierano` Phys. Rev. B 1997 56 7304. 3 G. Pacchioni and G.Ierano` Phys. Rev. B 1998 57 818. 4 G. Pacchioni G. Ierano` and A. M. Marquez Phys. Rev. L ett. 1988 81 377. 5 G. Pacchioni and R. Ferrario Phys. Rev. B 1998 58 6090. 6 G. Pacchioni and M. Vitiello Phys. Rev. B 1998 58 7745. 7 G. Pacchioni and M. Vitiello J. Non-Cryst. Solids 1999 245 175. Prof. Kempter said You pointed out some observable consequences of the Cu-adsorption to the defects taken into account in particular the presence of band gap states. In order to observe these states their intensity relative to that of the valence band emission would be needed. Could you comment on this point ? Prof. Pacchioni responded Our observation of the appearance of states in the gap when metal atoms are adsorbed on the silica surface is based on the qualitative analysis of the Kohn»Sham orbitals and does not allow us to estimate the relative intensity of these bands.Prof. Jennison said I agree with your statement that SiO is best described as having polarised 2 covalent bonding in contrast with the highly ionic oxides. Aside from theoretical analyses which I believe all agree in this sense another piece of evidence comes from Auger O(KVV) spectral data where the peak position and shape is quite diÜerent in SiO from those of the ionic oxides such as 2 MgO. Concerning your Table 4 I am interested in the ionicity of the single Cu atom attached to the Si»O~ system. The oxygen gains considerable charge (from [0.37 to [0.71) but the Cu only loses 0.13 which doesnœt add up. Could you please clarify the meaning of the reported charges and comment on the ionicity ? Prof.Pacchioni responded The reason why the charge of the Cu atom and of the non-bridging oxygen do not add up is that we did not report the values of all the atoms present in the model. The Cu atom donates charge to the non-bridging oxygen (0.3 electrons if we consider the change in population of the NBO center) but as a consequence of this charge transfer it interacts also mostly electrostatically with a bridging oxygen of the surface see Fig. 4b in ref. 1. The bridging oxygen donates some charge to the Cu atom which results to be less positively charged than it would be without this additional interaction. However Mulliken charges are not very reliable and should be considered with some care. More than the net charges computed with the Mulliken Faraday Discuss.1999 114 223»243 242 scheme in my opinion what shows the existence of a charge transfer is the shift of the core levels of the oxygen atoms. The occurrence of a partial charge transfer is thus quite evident from the data although a quantitative estimate of the amount of charge transfer is not easy. 1 N. Lopez F. Illas and G. Pacchioni J. Am. Chem. Soc. 1999 121 813. Prof. Thornton made a general comment There is a growing body of evidence that 2D clusters of size 7»8 atoms are particularly stable. Prof. Pacchioni added The suggested existence of Cu and Pd clusters on TiO raises a ques- 8 8 2 tion which has been around for a while in the clusters community i.e. the existence of ìmagic numbersœ corresponding to particularly stable aggregates.Do we have ìmagic numbersœ also for clusters on oxide surfaces ? Another question is related to the electronic structure of Cu clusters. Since the Cu atom has a 2S ground state clusters with an odd number of atoms have an open shell ground state at variance with clusters with an even number of atoms which can be diamagnetic. This leads to some oscillations in the properties as a function of cluster size (nucleation energy ionisation potential etc.) In principle also the heat of adsorption of Cu on MgO should exhibit these oscillations in the low coverage regime. Prof. Campbell responded I donœt know if there will be magic cluster sizes but I agree that there will probably be oscillations in stability with cluster size.Perhaps we could measure their stabilities and see such oscillations if we could get cold enough to control cluster sizes down to such small sizes. This may be possible already with our calorimeter with the proper combination of metal and oxide. Prof. Freund remarked With respect to the question whether it is possible to prepare deposited clusters with a small number of metal atoms I would like to mention that deposition of Rh or Ir onto alumina –lms at 50 K leads to clusters containing 1»7 atoms. After exposure to CO the IRAS are recorded and the analysis of vibrational spectra from isotopically labelled species is in line with the idea of clusters containing 1 2 and 3 (up to seven metal atoms) (refs. 157 and 158 of the Introductory Lecture). Prof. Goodman said With respect to the use of carbonyl precursors in the synthesis of metal and mixed-metal clusters I should mention our recent work on the synthesis of Ru clusters on TiO using a Ru carbonyl precursor. Although this study is in progress preliminary results show 2 3 clearly this carbonyl precursor route to be particularly suitable for the synthesis of highly dispersed metal clusters. In fact preliminary STM studies show several examples that suggest formation of a signi–cant population of Ru cluster species on the TiO surface. Although of course it 3 2 is impossible to comment on the carbon content of the imaged species the following STM certainly suggests that ring-opening of the Ru carbonyl could occur to the apparent epitaxial growth of 3 a linear Ru cluster. 3 Prof. Campbell commented I just wanted to make a general comment that theory seems to be getting closer to experiments these days on oxide surfaces. For example the energy of small Cu clusters on MgO(100) that we measured (see our paper presented here) are pretty well reproduced in DFT calculations by Musolino and Car and by values discussed here by Pacchioni.1 1 Musolino and Car Surf. Sci. 1998 402ñ4 413. 243 Faraday Discuss. 1999 114 223»243

ISSN:1359-6640    

DOI:10.1039/a908285i

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Oxygen-induced restructuring of rutile TiO (110) formation mechanism atomic models and in¢©uence on surface chemistry 1 Introduction The rutile TiO (110) surface has evolved as one of the most important model systems for metal 2 oxide surfaces. Titanium dioxide is used in gas sensing catalysis and photocatalysis where surface phenomena play an important role. In 1994 when the surface science of metal oxides was reviewed by Henrich and Cox TiO (110) was already an intensely studied system.1 Since this time 2 its popularity has increased steadily partly because bulk single crystals can be reduced easily 2 Min Li,a Wilhelm Hebenstreit,a Ulrike Diebold,*a Michael A. Hendersonb and Dwight R. Jennisonc a Department of Physics T ulane University New Orleans L A 70118 USA.E-mail diebold=mailhost.tes.tulane.edu b Paci¡©c Northwest National L aboratory Richland W A 99352 USA c Sandia National L aboratories Albuquerque NM 87185-1413 USA Faraday Discuss. 1999 114 245¡í258 245 Received 5th May 1999 The rutile TiO (110) (1]1) surface is considered the prototypical iwell-de¡©ned©« system in 2 the surface science of metal oxides. Its popularity results partly from two experimental advantages (i) bulk-reduced single crystals do not exhibit charging and (ii) stoichiometric surfaces as judged by electron spectroscopies can be prepared reproducibly by sputtering and annealing in oxygen. We present results that show that this commonly applied preparation procedure may result in a surface structure that is by far more complex than generally anticipated.Flat (1]1)-terminated surfaces are obtained by sputtering and annealing in ultrahigh vacuum. When re-annealed in oxygen at moderate temperatures (470¡í660 K) irregular networks of partially connected pseudohexagonal rosettes (6.5]6 Ae wide) one-unit cell wide strands and small (Btens of Ae ) (1]1) islands appear. This new surface phase is formed through reaction of oxygen gas with interstitial Ti from the reduced bulk. Because it consists of an incomplete kinetically limited (1]1) layer this phenomenon has been termed irestructuring©«. We report a combined experimental and theoretical study that systematically explores this restructuring process. The in¡ªuence of several parameters (annealing time temperature pressure sample history gas) on the surface morphology is investigated using STM.The surface coverage of the added phase as well as the kinetics of the restructuring process are quanti¡©ed by LEIS and SSIMS measurements in combination with annealing in 18O-enriched gas. Atomic models of the essential structural elements are presented and are shown to be stable with ¡©rst-principles density functional calculations. The eUect of oxygen-induced restructuring on surface chemistry and its importance for TiO and other bulk-reduced oxide materials is brie¡ªy 2 discussed. This journal is( The Royal Society of Chemistry 2000 (which conveniently prevents charging) and partly because of a desire to perform experiments on a ìwell-characterized systemœ. The preparation of a clean atomically —at TiO (110) (1]1) surface with a controlled defect 2 density is very important for surface chemistry experiments.Normally sputtering and annealing in ultrahigh vacuum (UHV) or oxygen at high temperatures are used. Many authors have published preparation recipes as an example we cite the one given by Pan et al. :2 ììThe stoichiometric (or nearly perfect) surface was obtained by sputtering with 500 eV Ar` then annealing to 1000 K for 3 min in 2]10~6 Torr of O2 and –nally cooling down to room temperature in the same oxygen atmosphere. XPS showed sharp Ti 2p peaks with no indication of reduced Ti states. œœ Most of the STM studies on TiO (110) performed so far3h12 have focused on UHV annealed 2 surfaces and the nature of the (1]2) reconstruction that evolves at high temperatures.Our STM measurements showed that oxygen-annealed surfaces prepared using Pan et al.œs recipe2 were considerably rougher than vacuum-annealed samples and that the appearance of the surface varied greatly when seemingly the same procedure was applied. In order to understand this phenomenon we re-annealed —at surfaces (prepared by UHV annealing at high temperature) at moderated temperatures in oxygen gas and found a pronounced morphology change which we have termed ìrestructuringœ. A brief report of STM results and conclusions was given previously,13 and a more complete account (including data not shown here) will be published elsewhere.14 In this paper we apply a combination of STM LEIS and SSIMS to explore systematically the in—uence of preparation parameters (temperature annealing time oxygen pressure and reduction state of the crystal) on surface restructuring.Ab initio total-energy density functional calculations are used to test the geometric model for restructured surfaces and to explore its electronic structure. As the main conclusion we –nd that both the surface structure and morphology of a TiO (110) surface depend sensitively on the oxidation conditions as well as the history of the 2 crystal. Speci–cally we –nd a new structure (termed ìrosette networkœ) that consists of an incomplete TiO (110) layer where all atoms are in approximate bulk-like positions but some are 2 missing in a regular fashion. Compared to a (1]1) structure the rosette networks exhibit quite diÜerent bonding geometry coordination number undercoordinated sites and degree of covalency.Hence TiO (110) surfaces prepared by annealing in oxygen may not resemble the —at 2 (1]1) terminated surfaces that are often assumed in interpretation of surface chemistry experiments. With this work we would like to provoke thoughts and invite comments on how surface chemistry may be aÜected by the presence of such rosette networks and to what extent our –ndings may be transferable to other bulk oxide systems with facile transport of metal interstitials. 2 Experimental and calculation methods Both pure The experiments were performed in two UHV systems described elsewhere.15,16 Polished TiO2 single crystals from three diÜerent vendors have been used which exhibited a blue color after an initial high-temperature anneal (950 K); details on sample mounting can be found elsewhere.13v15 Before each experiment the sample was prepared with sputtering (1000 eV Ar` IsampleB8.3 lA cm~2 20 min) and annealing to 880 K in UHV for 30 min which yielded a —at TiO (110) 2 (1]1) surface with atomically —at large terraces (up to 500 Aé wide) as shown in Fig.1. The alternating white and dark rows along the [001] direction are located at the positions of 5-fold coordinated Ti atoms and 2-fold bridging O atoms of the (1]1) structure respectively.11 Some bright rows typically several tens of nanometers long are scattered across terraces or connected to step edges. These appear upon UHV annealing of relatively dark crystals.17 In reference to the features observed upon annealing to higher temperature,4h7 we call these rows (1]2) strands.Smooth step edges are oriented along the [116 1] and the [001] direction.12 Only Ti and O signals were detected by XPS and LEIS indicative of a clean sample surface after such a treatment and LEED showed a sharp (1]1) pattern. 16O 18O gas and isotopically enriched (18O2 16O2\93% 7%) were gas 2 2 employed in oxygen exposure experiments. Gas dosing was performed by back–lling the chamber. To quantify the 18O surface content we took LEIS spectra14 which show clearly separated 18O and 16O peaks. These LEIS measurements do not aÜect the uptake of 18O with carefully controlled parameters (total ion —uence of D1.6]10~15 cm~2 per measurement and beam energy of Faraday Discuss. 1999 114 245»258 246 Fig.1 of STM image (500 Aé ]500 Aé ) TiO (110) sputtered and annealed in UHV for 30 min at 880 K. The 2 surface shows a regular (1]1) termination. The few bright lines are referred to as (1]2) strands. 1000 eV). Static SIMS experiments were performed in a separate chamber16 with a diÜerentially pumped ion gun utilizing a 500 V Ar` beam with an ion —ux in the nA cm~2 regime. No ion current was measured at the sample without Ar —owing through the gun even with a 10~6 Torr chamber pressure of O2. O2 This indicates that virtually no entering the gun from the chamber was exposed to the crystal as ions. During a typical experiment the total Ar` ion exposure was maintained below about 5% of a monolayer in order to minimize any potential eÜects due to sputter damage.Secondary ions generated by sputtering were monitored with a quadrupole-based Extrel C50 spectrometer. The electronic structure calculations used the massively parallel Gaussian-based code QUEST (quantum electronic structure)18 and density functional theory in the local density approximation (LDA) as described in refs. 14 and 19. Successful tests of computational accuracy included comparisons with the results of Ramamoorthy et al.,20 who used a plane wave code and diÜerent pseudopotentials. Local densities of states (LDOS) are found by the standard projection on the local Gaussian bases. Integrating these to the Fermi level then results in local electron populations having diagonal (same atom) and oÜ-diagonal (interatomic) parts which provide information on the degrees of ionicity and covalency in local interactions.3 Results 3.1 Oxygen-induced change of surface structure In the following we present LEIS and STM results after re-annealing at various temperatures for a —at UHV-annealed surface similar to the one displayed in Fig. 1. The following procedure was employed each time sputtering UHV annealing lowering the sample temperature to the speci–ed value exposing to 18O gas (1]10~6 mbar) for a speci–ed time and cooling in UHV to room 2 temperature. 18O exposure at 500 K for 10 min produces bright features evenly distributed on the (1]1) 2 substrate as shown in Fig. 2(a). Most features are assembled into short aggregates (B40 Aé ) roughly oriented along the [116 0] direction. In addition a few scattered (1]1) islands (ca.40 Aé ]30 Aé marked by arrows in Fig. 2(a)) can be seen on top and in between the large (1]1) terraces. The step edges of the (1]1) terrace become more irregular as compared to a UHVannealed surface (Fig. 1). After annealing in 18O at 520 K (Fig. 2(b)) distinctly diÜerent morpho- 2 logical features appear on the large (1]1) terraces. Patches of rosette-like networks13 (labeled R typically 30 Aé wide and elongated along the [116 0] direction) dominate see Fig. 4 below. Located in between are small (1]1) islands (of typical size 60 Aé ]40 Aé ). In addition some white clusters are found on top of the (1]1) islands. Fig. 2(c) (annealing at 550 K) shows an even rougher 247 Faraday Discuss. 1999 114 245»258 2 Fig. 2 STM images (500 Aé ]500 Aé ) of a TiO (110) surface.All surfaces were pretreated by sputtering and 2 annealing in UHV at 880 K for 30 min. 18O (1]10~6 mbar) was dosed at (a) 500 K (b) 520 K (c) 550 K (d) 660 K for 10 min (e) 710 K for 15 min and (f ) 830 K for 20 min. 18 surface consisting of many layers of somewhat larger (1]1) islands (B80 Aé ]60 Aé ) partially connected to each other and roughly oriented along the [116 0] direction. Between and on top of these islands are network patches (R) also elongated along the [116 0] direction. The —at (1]1) substrate (still discernible in Fig. 2(b)) can no longer be identi–ed in this image. After annealing in O at 660 K (Fig. 2(d)) the (1]1) phase dominates the surface. The (1]1) islands are connected 2 to each other to form large (1]1) terraces with larger network patches (ca.100 Aé ]80 Aé ) appearing on top of terraces. STM results14 from a sample annealed in p O2\1]10~6 mbar at 660 K for 5 10 and 20 min gave images very similar to the one displayed in Fig. 2(d) (10 min). (This justi–es the comparison between annealing for 10 min (Fig. 2 (a)»(d)) and the somewhat longer annealing times at higher temperatures in Fig. 2(e) and (f )). Oxygen exposure at 710 K for 15 min leads to a dramatic morphological change (Fig. 2(e)) with much larger (1]1) islands and straight step edges (some of which are reconstructed as is observed on UHV annealed surfaces A 12). A number of [001]-oriented bright strands (typically 70 é long) are distributed uniformly on top of (1]1) terrace and extend out from the step edges across the lower terrace.An even higher annealing temperature (Fig. 2(f )) yields large —at (1]1) terraces with a few white clusters and bright strands on top of the bright [001]-oriented rows (the Ti sites) of the substrate. Such strands are also visible in the small scale image (Fig. 4 below). Annealing in 18O leads to incorporation of 18O into the sample surface. LEIS 18O peak areas 2 (after normalization of the total LEIS O signal to 100%) that correspond to the images displayed in Fig. 2 are shown in Fig. 3. The 18O content increases with temperature up to a maximum value of 75% for Fig. 2(d). Concurrent with the transition to —at larger (1]1) islands and the absence of rosette networks the 18O content decreases again. All the surfaces displayed in Fig. 2 exhibit a (1]1) LEED pattern.XPS results of an oxygenannealed surface revealed no diÜerence in the Ti 2p peak position or shape as compared to a surface annealed for 30 min in UHV at 880 K. This may be because of the insufficient sensitivity of our XPS-setup. Previous measurements2,21 showed a shoulder indicative of Ti3` species (attributed to oxygen vacancies) on the UHV-annealed surfaces. Faraday Discuss. 1999 114 245»258 248 Fig. 3 The surface concentration of 18O (measured with LEIS) on TiO (110) surfaces prepared as in Figs. 2 2(a)»(f ). A small scale STM image of both network patches and (1]1) islands is shown in Fig. 4. The small isolated (1]1) island (ca. 60 Aé ]30 Aé ) at the center is partially connected to network patches. The network is atomically resolved as arrays of inter-connected pseudo-hexagonal units (named rosette R).Usually one rosette (marked in Fig. 4) is composed of six bright spots with a width equal to the substrate unit cell in the [116 0] direction (6.5 Aé ) and twice as long as the substrate unit cell in the [001] direction (2]3 Aé ). Some bigger rosettes composed of more than six bright spots appear occasionally. Incomplete rosettes are incorporated into the edge of the (1]1) islands but rosettes never appear within an island. The rosettes have the same height as the (1]1) terraces. The dark centers of the rosettes appear on top of the bright rows (on top of the 5-fold coordinated Ti atoms) of the underlying TiO (1]1) layer. Structural models of these 2 rosettes are presented below (Figs.8 and 9). Some short bright strands ( A ca. 10 é long) are either connected to or located between network patches and (1]1) islands. Usually one of the six bright spots is missing at the connection between a rosette and a strand. In a forthcoming paper,22 we argue that the strands exhibit the same structure as the double ridges of the TiO (110) (1]2) reconstruction.7 2 2 Fig. 4 A small scale STM image (200 Aé ]200 Aé ) of network patches (R) strands and (1]1) islands after annealing in 1]10~6 mbar 18O at 570 K for 25 min. 249 Faraday Discuss. 1999 114 245»258 Fig. 5 The 18O surface concentration during annealing in 6.7]10~7 mbar 18O at various temperatures 2 monitored with SSIMS. 3.2 Kinetics of the restructuring process A study of the initial 18O incorporation with annealing time was performed using static secondary ion mass spectrometry (SSIMS) in a separate chamber.The sample was prepared by sputtering and annealing in 16O (6.7]10~7 mbar) at the temperature range 477»815 K. The 16O gas was 2 2 18O pumped out the chamber was then back–lled with gas (6.7]10~7 mbar) which was 2 pumped out after 260 s. During the whole procedure the 18O surface content was monitored (Fig. 5). The 18O uptake occurs more rapidly with increasing annealing temperature and decreases again above 669 K. This is consistent with the LEIS results displayed in Fig. 3 where a maximum Fig. 6 SSIMS of 18O surface concentration during annealing in 18O for rutile crystals with diÜerent colors. 2 The color of a crystal is a measure of its bulk defect concentration.Faraday Discuss. 1999 114 245»258 250 2 (1]10~6 mbar) at 570 K for 10 18O2 Fig. 7 STM images (500 Aé ]500 Aé ) of (a) a light blue and (b) a dark TiO (110) sample prepared by sputtering and UHV annealing at 970 K for 20 min followed by annealing in min. of 18O surface content was found after annealing for 10 min at 710 K. From Fig. 5 the 18O uptake rate was determined using a (1[H) dependence. The rate increases quickly below 566 K slows down and decreases above 669 K. From these rates the 18O activation energy is estimated to be 19 kcal mol~1. The uptake rate is strongly dependent on crystal ìageœ (i.e. reduction state)17 ìFreshœ aspurchased TiO crystals are transparent. With increasing numbers of sputtering/annealing cycles 2 they change in color from light to dark blue to metallic grayish.This color change is caused by the creation of color centers in the bulk and can be used as a quantitative measure for the degree of bulk reduction. Fig. 6 shows that darker more reduced samples incorporate 18O at a much faster rate than lighter more stoichiometric samples. Displayed in Fig. 7 are two TiO (110) samples with 2 a diÜerent degree of bulk reduction. (These samples were mounted next to each other on one sample platen and sputtered and annealed simultaneously to ensure exactly the same treatment. A more detailed and quantitative investigation of the relationship between sample color type of bulk defects and surface properties is being published in a forthcoming paper.17) A drastically diÜerent appearance is visible by STM; annealing in oxygen a light blue sample (top in Fig.7) shows basically a (1]1) surface termination whereas the much darker sample (Fig. 7 bottom) is quite covered with rosettes. Both samples incorporate 18O however (Fig. 6). 4 Discussion 4.1 Geometric model for rosette-like network structure A model for the rosette network has already been presented in a previous paper.13 The main features are shown in Fig. 8. It consists of an incomplete TiO layer where the O and Ti atoms 2 251 Faraday Discuss. 1999 114 245»258 Fig. 8 Atomic model (top and side view) for a restructured surface. A bulk-terminated (1]1) island is shown on the right side. The network patch (R) on the left side consists of an incomplete TiO (110) (1]1) layer and contains only atoms at bulk positions.Small white balls are Ti atoms. Shadowed large balls represent oxygen 2 atoms and darker shading indicates higher z-positions. The rectangle outlines the unit cell of the (1]1) structure. The hexagons connect Ti atoms in similar positions on both islands. Atoms missing in the network are marked with large crosses on the (1]1) island. are missing in a regular fashion and all the remaining atoms are in bulk-like positions. In Fig. 8 two islands are placed onto a (1]1) surface the left one representing a rosette network with missing atoms and the right one the regular (1]1) structure. First consider the (1]1) island. Titanium atoms are drawn as small white balls and oxygen atoms as dark balls. We chose this shading for easier comparison with STM images where Ti sites are generally imaged bright.11 As is visible in the side view oxygen atoms at higher locations are shaded darker.For example the bridging oxygen atoms covering every other Ti row in the regular (1]1) structure are shown as black balls. The (1]1) structure has a rectangular unit cell as outlined on the far right of Fig. 8. The four 6-fold coordinated Ti atoms on the corners of the unit cell are covered by bridging oxygen atoms and only the center Ti (5-fold coordinated) is visible by STM. Outlined (with full lines) on the (1]1) island in Fig. 8 is a hexagon connecting four 5-foldcoordinated Ti and two six-fold-coordinated Ti atoms (underneath the bridging oxygens). Suppose the two bridging oxygen atoms marked with crosses as well as the six-fold coordinated Ti atom in the center are missing.Then one would end up with six Ti atoms arranged in a quasi-hexagon (which is 6.5 A Aé (one unit cell) wide and 6 é (two unit cells) high as observed in the experiment). Exactly these atoms are missing in the ìRœ network structure shown in Fig. 4 which consists only of O and Ti atoms in bulk-like positions. So conversely adding one TiO unit into the hexagon 2 (full-lines) of the network structure of Fig. 8 generates the regular (1]1) structure. A second kind of hexagon with Ti atoms at the corners is drawn with dashed lines on both islands of Fig. 8. Here one TiO unit (one 6-fold-coordinated Ti atom at the center and two ìconnectingœ bridging oxygen atoms counted as one oxygen atom at the sides) is missing in the network structure.Hence the rosette-structure is simply an incomplete TiO (110) layer with some atoms missing. 2 4.2 Electronic structure calculations for a rosette on top of TiO2(1�1) (110) surface To test the stability of the proposed rosette structure we performed LDA calculations using the 258-atom supercell shown in Fig. 9. A single rosette with six Ti atoms and twelve O atoms is Faraday Discuss. 1999 114 245»258 252 Fig. 9 Supercell used for electronic structure calculations (a) top-view (b) side-view. 2 arranged on top of a TiO2(1]1) (110) substrate. The bottom three TiO2 layers (3rd»5th layer in Fig. 9(b)) were frozen at the bulk positions and the top two layers of TiO and the rosette were allowed to relax geometrically by minimizing the force of each atom.Relaxations of selected atoms are illustrated in Figs. 9(a) and 9(b). The rosette shows substantial horizontal relaxations (between 0.1 Aé and 0.6 Aé Fig. 9(a)) with a general tendency to collapse towards the center. From the side view (Fig. 9(b)) all Ti atoms (1»6) in the rosette relax downward by 0.3 é Aé . A All inner O atoms (9,12,16,17) sink by 0.3 while 15 and 253 Faraday Discuss. 1999 114 245»258 18 sink slightly (0.1 Aé ). The other outer O atoms (7,8,10,11,13,14) rise by 0.4 Aé . Thus the rosette is shrinking and buckling to reach its equilibrium position. This results in a shortening of Ti»O bond lengths by 0.1 to 0.2 Aé as compared to the bulk. With the rosette on top of the substrate the relaxation of Ti and O atoms in the –rst layer is similar to the –rst-principle calculations of the clean (1]1) (110) surface.23 The existence of rosettes causes small relaxations in the second and third layers.The LDOS of bulk atoms (O131 and Ti130 not labeled in Fig. 10(b)) –rst layer atoms (O35 and Ti19) and the DOS of rosette atoms (averaged O Ti1 and Ti3) are shown in Fig. 10. In each case the Ti states dominate the unoccupied conduction band and the O states the valence band similar to other theoretical works on the TiO2(1]1) (110) surface.23h28 When interpreting our STM images (Fig. 4) we assumed that the Ti atoms in the rosette are also imaged bright. The results displayed in Fig. 10 justify this assumption. In the rosette the empty Ti 3d states dominate the conduction band and from the theory of TersoÜ and Hamann,29 the tunneling current is found to be proportional to the surface LDOS at the position of the tip.The valence band width of rosette O atoms is narrower as compared to O atoms on the (1]1) surface and in the bulk and its shape is also diÜerent. Such changes should clearly be visible in UPS valence band measurements especially if performed under conditions where the photon energy is varied to increase sensitivity to T3d derived states.30 The gap width for both O 2p and Fig. 10 LDOS of selected atoms within the slab in Fig. 9. Faraday Discuss. 1999 114 245»258 254 Ti 3d states in the rosette is wider than that in the bulk as shown in Fig. 10. When analyzing the oÜ-diagonal LDOS of selected Ti»O pairs14 (not shown here) we –nd a signi–cant increase in the strength of covalent interaction between Ti and O atoms in the rosette as compared to the bulk.We believe that it is this interaction which broadens the gap moving the local conduction band minimum higher as seen in Fig. 10. 4.3 Mechanism Only a few previous accounts of oxygen-induced morphological and structural changes of TiO (110) have been given. A brief Letter has been published by this group.13 Engel and co- 2 workers have observed cross-linked row structures along the [116 0] direction after annealing the TiO (110) (1]2) phase in oxygen (1]10~7 Torr) at 1000 K followed by heating in UHV.3,4 2 Onishi and Iwasawa9 observed hill-like structures when exposing the TiO (110) (1]1) surface to 2 O gas (B1]10~7 mbar) at 800 K.With time these features were transformed into added rows 2 and new (1]1) terraces. This is consistent with our STM results above 710 K (Fig. 2(e)) where only added strands and (1]1) terraces are visible. Onishi and Iwasawa proposed a re-oxidation scheme where Tin` (nO3) interstitial atoms from the reduced bulk diÜuse to the surface where they react with O gas to form hills added rows and new terraces. Consumption of Ti interstitials 2 by reaction with surface oxygen produces a concentration gradient that results in a net diÜusion current of these Ti interstitials towards the surface. The same mechanism i.e. segregation of Tin` combined with reaction with gaseous oxygen is responsible for the formation of rosettes strands and (1]1) islands in our experiments.The rate of surface restructuring is aÜected by the surface concentration of both reaction partners Ti interstitials and O molecules. The Ti segregation rate 2 depends on temperature (in—uencing diÜusion to the surface) number of Ti interstitials (reduction state of the crystal) and the chemical potential of oxygen (the oxygen pressure). The restructuring process can be regarded as the manifestation of reoxidation of a reduced crystal at the atomic scale. It results in the growth of additional TiO layers at the surface with Ti 2 coming from the reduced bulk and oxygen from the gas phase. The kinetic processes and energetics that govern nucleation growth and morphology of deposited –lms are well studied.31 This case is special ; because one constituent of the newly added –lm comes from the bulk the kinetics of bulk diÜusion must be taken into account as well.(Extensive TiO bulk studies32h39 revealed 2 titanium interstitial ions as well as oxygen vacancies in a reduced TiO crystal. Bulk diÜusion 2 studies40h48 and a recent SSIMS investigation49 show that Ti interstitials are the major diÜusive species in TiO rutile and not O vacancies.) For heavily reduced crystals the added features 2 nucleate mainly on terraces as is visible from their random distribution in low-coverage images (Fig. 2(a)). This suggests that Ti interstitials are driven out in vertical direction from the bulk to the surface where they react with oxygen. Some step-—ow growth occurs also as evidenced by the relatively rough step edges that develop already at low temperatures and/or gas exposures.Step- —ow dominates the growth on less reduced light crystals (Fig. 7(a)). The rosette networks are the precursors to the added (1]1) islands especially at lower temperatures (\660 K). The transition from rosettes to the (1]1) structure is straightforward as discussed above (Section 4.1) one simply needs to add additional atoms to rosettes. This should happen easily upon arrival of new Ti interstitials on the surface even without additional incorporation of O from the gas phase. At low enough temperatures the overall island shape of rosette 2 network patches appear to be preserved during the transition to a (1]1) island hence the preferred orientation of the (1]1) islands. When the temperature rises above D700 K (Fig.2(e)) the (1]1) islands assume a more square shape with step edge orientations typical for hightemperature UHV-annealed surfaces.12 There are three distinct regions for the rate of 18O uptake as reported in the context of Fig. 5. Initially the rate increases with increasing temperature. At higher temperatures a second process kicks in that –rst slows and then decreases the reaction rate with annealing temperature. This is also apparent in Fig. 3 where the total 18O uptake for a given exposure time decreases above 700 K. Annealing an 18O-rich surface in UHV decreases the 18O surface content with a clear break point in the depletion rate around 740 K.22 The 18O can leave the surface via two routes exchange with 16O from the bulk and desorption into the gas phase.The observed changes in surface morphology suggest that the latter process is dominant at high temperatures. Note that 255 Faraday Discuss. 1999 114 245»258 the slowdown in 18O uptake under oxygen-rich conditions occurs concurrently with the disappearance of rosettes ; upon annealing in 18O at 710 K (Fig. 2(e)) only strands are formed on the 2 surface. Similarly rosettes transform into (reduced Ti strands when a 18O-restructured 2O3) surface is annealed in UHV at 690 K.22 A mere scrambling between surface and bulk oxygen atoms cannot lead to a surface reduction. The assumption that desorption of oxygen from the surface occurs at temperatures above 700 K is also supported by other studies. It has been reported that surface point defects are created when TiO (110) (1]1) surfaces are annealed in UHV at 2 temperatures above 700 K.2 Xu et al.10 observed a (1]2) reconstructed surface between 700 and 800 K which reversibly converts to the (1]1) surface.They attributed this conversion to O desorption into the gas phase and/or Ti diÜusion into the bulk. This is also consistent with our results ; at temperatures above 830 K only bulk-terminated (1]1) terraces exist (Fig. 2(f )) indicating that the row structure is also a metastable phase that converts into the most stable (1]1) terrace with temperature. 5 Conclusion and open questions 2 The results presented in this paper clearly indicate that both the oxidation conditions and the history of the TiO (110) sample have a signi–cant bearing on the morphology of the surface.It is 2 somewhat frustrating that even this ìbest characterizedœ of all metal oxide systems is not yet completely understood and that characterization with spectroscopic and diÜraction techniques is not sufficient to reveal the great diÜerences in atomic structure that can form through annealing in oxygen. The STM images displayed in Fig. 7 are a good example. Under exactly the same oxidation conditions a very light blue crystal exhibits only a (1]1) structure while a dark crystal is covered with rosettes. Hence it is possible to deliberately prepare either a stoichiometric TiO (110) (1]1) surface or one covered with the metastable structure. The observed variations in the surface structure with O pressure crystal temperature and bulk 2 defect density are so vast that we suspect chemistry of the TiO (110) surface should be signi–- 2 cantly variant for samples oxidized under diÜerent conditions.For example the issue of whether water is molecularly or dissociatively adsorbed on TiO (110)2,16,50h52 may be signi–cantly 2 clouded in the literature because of studies in which the morphology of the surface was unknowingly disordered by the presence of the rosettes and/or strands observed in this study by STM. Another example where restructured surfaces may exhibit quite diÜerent chemistry from a (1]1) surface is the adsorption of pyridine. Pyridine molecules bind more strongly at 4-fold coordinated Ti atoms at step edges as compared to the 5-fold coordinated Ti on the —at (1]1) surface as shown recently by Suzuki et al.53 Possibly pyridine interacts strongly with rosette networks where 4-fold coordinated Ti atoms are prevalent (Fig.9). Metal overlayer –lm growth on TiO (110) may be aÜected as well,54 e.g. the roughness induced by oxygen annealing of dark 2 crystals may in—uence nucleation and growth of overlayers. The rosette networks may also provide special adsorption sites for metal atoms e.g. it is not inconceivable that one could place a single metal atom in the center of the rosette displayed in Fig. 9. On a larger scale the results in this study suggest that subsurface (interstitial) Ti is fairly labile in TiO rutile especially as the bulk concentration of these species increases. The bulk of small 2 rutile particles could therefore act as sinks for excess Ti under reductive conditions with this Ti returning to the surface under oxidative conditions.Such cycling of Ti between the bulk and surface should signi–cantly in—uence surface properties of small crystalline particles as suggested by results in this work but should also aÜect the bulk electrical and photoabsorptive properties. Rutile TiO (110) may not be the only system where such oxygen-induced morphology changes 2 occur. A study of the reoxidation mechanisms of other bulk-reduced materials may provide an attractive playing –eld for surface scientists where rich and interesting metastable structures may be expected. 6 Acknowledgement This work was supported in part by NSF-CAREER and DoE-EPSCoR. The SSIMS work was supported by the US Department of Energy Office of Basic Energy Sciences Division of Materials Sciences and was conducted at the William R.Wiley Environmental Molecular Sci- Faraday Discuss. 1999 114 245»258 256 ences Laboratory a Department of Energy user facility funded by the Office of Biological and Environmental Research. Paci–c Northwest National Laboratory is a multiprogram national laboratory operated for the US Department of Energy by Batelle Memorial Institute under Contract DE-AC06-76-RLO 1830. Sandia is a multiprogram laboratory operated by Sandia Corporation a Lockheed Martin Company for the United States Department of Energy under Contract DE-AC04-94AL85000. The electronic structure calculation was partially supported by a Laboratory Directed Research and Development project.References 1 V. E. Henrich and P. A. Cox T he Surface Science of Metal Oxides Cambridge University Press Cambridge 1994. 2 J.-M. Pan B. L. MaschhoÜ U. Diebold and T. E. Madey J. V ac. Sci. T echnol. A 1992 10 2470. 3 A. Szabo and T. Engel Surf. Sci. 1995 329 241. 4 M. Sander and T. Engel Surf. Sci. L ett. 1994 302 263. 5 D. Novak E. Garfunkel and T. Gustafsson Phys. Rev. B Condens. Matter 1994 50 5000. 6 P. W. Murry N. G. Condon and G. Thornton Phys. Rev. B Condens. Matter 1995 51 10989. 7 H. Onishi and Y. Iwasawa Surf. Sci. L ett. 1994 313 783. 8 H. Onishi K. Fukui and Y. Iwasawa Bull. Chem. Soc. Jpn. 1995 68 2447. 9 H. Onishi and Y. Iwasawa Phys. Rev. L ett. 1996 76 791. 10 C. Xu X. Lai G. W. Zajac and D. W. Goodman Phys. Rev. B Condens. Matter 1997 56 13464.11 U. Diebold J. F. Anderson K.-O. Ng and D. Vanderbilt Phys. Rev. L ett. 1996 77 1322. 12 U. Diebold J. Lehman T. Mahmoud M. Kuhn G. Leonardelli W. Hebenstreit M. Schmid and P. Varga Surf. Sci. 1998 411 137. 13 M. Li W. Hebenstreit and U. Diebold Surf. Sci. 1998 L414 L951. 14 M. Li L. Groê W. Hebenstreit U. Diebold M. A. Henderson D. R. Jennison P. A. Schultz and M. P. Sears Surf. Sci. 1999 437 173. 15 L. Zhang M. Kuhn and U. Diebold Surf. Sci. 1997 371 223. 16 M. A. Henderson Surf. Sci. 1996 355 151. 17 M. Li W. Hebenstreit U. Diebold and M. A. Henderson Surf. Sci. submitted. 18 M. P. Sears and P. A. Schultz at Sandia National Laboratories Albuquerque NM 87185-1111. E-mail mpsears=sandia.gov 19 C. Verdozzi D. R. Jennison P. A.Schultz M. P. Sears J. C. Barbour and B. G. Potter Phys. Rev. L ett. 1998 80 5615; ibid. 1999 82 799. 20 M. Ramamoorthy D. Vanderbilt and R. D. King-Smith Phys. Rev. B Condens. Matter 1994 49 16721. 21 L. Wang D. R. Baer and M. H. Engelhard Surf. Sci. 1994 320 295. 22 M. Li W. Hebenstreit and U. Diebold Phys. Rev. B Condens. Matter submitted. 23 K.-O. Ng and D. Vanderbilt Phys. Rev. B Condens. Matter 1997 56 10544. 24 O. Gué lseren R. James and D. W. Bullett Surf. Sci. 1997 377ñ379 150. 25 S. Munnix and M. Scheits Phys. Rev. B Condens. Matter 1984 30 2202. 26 N. Yu and J. W. Halley Phys. Rev. B Condens. Matter 1995 51 4768. 27 D. Vogtenhuber R. Podloucky A. Neckel S. G. Steinemann and A. K. Freeman Phys. Rev. B Condens. Matter 1994 49 2099. 28 D. Vogtenhuber R.Podloucky and J. Redinger Surf. Sci. 1998 402ñ404 798. 29 J. TersoÜ and D. R. Hamann Phys. Rev. B Condens. Matter 1985 31 805. 30 Z. Zhang and V. E. Henrich Phys. Rev. B Condens. Matter 1991 43 12004. 31 J. A. Venables Surf. Sci. 1994 299/300 798. 32 M. Aono and R. R. Hasiguti Phys. Rev. B Condens. Matter 1993 48 12406. 33 F. Millot M. G. Blanchin R. Tetot J. F. Marucco B. Poumellec C. Picard and B. Touzelin Prog. Solid State Chem. 1987 17 263. 34 P. Kofstad J. L ess-Common Metals 1967 13 635. 35 L. A. Bursill M. G. Blanchin and D. J. Smith Proc. R. Soc. L ondon Ser. A 1984 391 373. 36 L. A. Bursill and D. J. Smith Nature (L ondon) 1984 309 319. 37 J. Sasaki N. L. Peterson and K. Hoshino J. Phys. Chem. Solids 1985 46 1267. 38 H. B. Huntington and G.A. Sullivan Phys. Rev. L ett. 1965 14 177. 39 D. J. Smith L. A. Bursill and M. G. Blanchin Philis. Mag. A 1984 50 473. 40 D. J. Derry D. G. Lees and J. M. Calvert J. Phys. Chem. Solids 1981 42 57. 41 D. J. Neild P. J. Wise and D. G. Barnes J. Phys. D Appl. Phys. 1972 5 2292. 42 H. Kolem and O. Kanert Z. Metallkd. 1989 80 227. 43 T. S. Lundy and W. A. Coghlan J. Phys. Colloq. 1973 299. 44 K. Hoshino N. L. Peterson and C. L. Wiley J. Phys. Chem. Solids 1978 39 457. 45 D. A. Venkatu and L. E. Poteat Mater. Sci. Eng. 1970 5 258. 46 J. R. Akse and H. B. Whitehurst J. Phys. Chem. Solids 1978 39 457. 257 Faraday Discuss. 1999 114 245»258 47 M. Arita M. Hosoya M. Kobayashi and M. Someno J. Am. Ceram. Soc. 1979 62 443. 48 M. Someno and M. Kobayashi Springer Ser. Chem. Phys. 1979 9 222. 49 M. A. Henderson Surf. Sci. 1999 419 174. 50 L. Wang D. R. Baer M. H. Engelhard and A. N. Shultz Surf. Sci. 1995 344 237. 51 M. A. Henderson L angmuir 1996 12 5093. 52 R. L. Kurtz R. Stockbauer and T. E. Madey Surf. Sci. 1989 218 178. 53 S. Suzuki Y. Yamaguchi H. Onishi K. Fukui T. Sasaki and Y. Iwasawa Catal. L ett. 1998 50 117. 54 C. T. Campbell Surf. Sci. Rep. 1997 27 1. Paper 9/03598B Faraday Discuss. 1999 114 245»258 258

ISSN:1359-6640    

DOI:10.1039/a903598b

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

The selective adsorption and kinetic behaviour of molecules on TiO (110) observed by STM and NC-AFM 2 Yasuhiro Iwasawa,* Hiroshi Onishi,§ Ken-ichi Fukui Shushi Suzuki and Takehiko Sasaki Department of Chemistry Graduate School of Science T he University of T okyo Hongo Bunkyo-ku T okyo 113-0033 Japan. Fax ]81-3-5800-6892; E-mail iwasawa=chem.s.u-tokyo.ac.jp Received 19th April 1999 2 In the present paper we report a kinetic aspect of molecules on terraces and steps of a TiO (110)-(1]1) surface as observed by scanning probe microscopy which may be relevant to oxide catalysis. When the TiO surface heated at 400»450 K was exposed to a formic acid ambient of 1»2]10~6 Pa small particles were formed on terraces. Post-reaction STM observation revealed that the particle formation was strongly suppressed in the vicinity of single-atom height steps.The suppressive eÜect of the steps ranged 2.4 nm into the terrace. Formate ions (a possible reaction intermediate) were imaged in-situ during formic acid exposure at the reaction temperature. The local density of the formates and hence the product distribution which were very low near the steps were simulated by a model. The particles produced were suggested to be carbonates CO32~ on the bridging oxygen atoms of the TiO2(110) surface. Introduction 2 There are a variety of applications of oxides to industrial materials and processes e.g. optical applications electronic applications catalytic materials (such as photocatalysts environmental catalysts and so on) electrodes adsorbing materials magnetic applications ceramic applications coatings environmental applications pigments cosmetics containers for nuclear materials.1,2 TiO is a typical oxide and has been widely used in many types of technologies.However there is little atomic- and molecular-level information on the structure of metal oxide surfaces which should be uncovered scienti–cally. The great potential of scanning probe microscopy (SPM) to directly observe surface structures and reactions at a high resolution has been demonstrated on metal oxide surfaces.1,2 Individual atoms that constitute a sputter-annealed TiO (110)-(1]1) surface were resolved by scanning tunneling microscopy (STM)3 and atomic force microscopy (AFM) operated in a frequency modulated non-contact mode (NC-AFM),4 as shown in Fig.1. Each Ti atom with –ve-fold coordination shown in the bright rows of 0.65 nm separation to each other and parallel to the [001] direction was clearly observed at a regular interval of 0.30 nm by STM. The atomically resolved NC-AFM image reproduced the (1]1) unit 0.65 nm]0.30 nm with the alignment of the bridging oxygen atoms. Several dark sites on the bright lines in the NC-AFM image could be vacancies of the bridging oxygen atoms. Further the molecules and 2 § Present address Kanagawa Academy of Science and Technology KSP East-404 Sakado Takatsu-ku Kawasaki-shi 213-0012 Japan. Faraday Discuss. 1999 114 259»266 259 2 This journal is( The Royal Society of Chemistry 2000 Fig. 1 The atomically resolved images of the (110)-(1]1) surface of the rutile TiO2 .(a) A constant current topography by STM.3 Scan area 12]12 nm2 sample bias voltage ]1.0 V tunneling current 0.3 nA. The diagonal scratches are artifacts. (b) NC-AFM.4 Scan area\9.4]9.4 nm2 V (c) A stoichiometric truncation of the (110) surface. s\0 V frequency shift\80 Hz. 2 reaction intermediates adsorbed on a TiO (110) surface can be imaged at temperatures P300 K 2 because their diÜusion is fairly restricted over metal oxide surfaces.5 For example formate isolatedly adsorbed at room temperature (NC-AFM),6 formate ions migrating at room temperature (STM)7,8 and acetate intermediates decomposing at 540 K (STM)9 have been visualized on TiO (110) surfaces. The high-temperature dynamic behavior of oxide surfaces has also been suc- 2 cessfully imaged with a TiO (110) surface at 800 K.10 The most promising ability of SPM expected in relation to catalysis research is to discriminate and visualize the reaction events occurring at diÜerent sites on a surface.The enhancement of catalytic reactions due to coordinatively unsaturated and atom-rearranged metal centers (like steps) has long been postulated since the ì active centerœ concept of Tailor.11 In fact the sitespeci –c adsorption of pyridine promoted on four-fold coordinated Ti centers at the steps with in particular azimuth orientations on TiO (110) has been visualized.12 The steps on which the pyri- 2 dine molecules adsorbed run parallel to the [112] and [112] directions. Similar activity has been observed on the steps parallel to the [113] [114] and [115] directions.In contrast the steps of the [111] [110] and [001] directions were not active for the adsorption. In the present paper we report a long-range suppressive eÜect of single-atom height steps of a TiO (110) surface for a formic acid decomposition reaction that takes place over the terraces far 2 distant from the steps. Experimental The experiments were performed in an UHV-compatible scanning tunneling microscope (JEOLJSTM4500VT) with an electrochemically etched tungsten tip. Constant-current topography was Faraday Discuss. 1999 114 259»266 260 continuously imaged at a rate of 33 s per frame and recorded in video. A polished TiO (110) wafer 2 of 6.5]1]0.25 mm3 (Earth Chemicals Co. Japan) gave a sharp (1]1) LEED pattern after repeated cycles of Ar` sputtering (3 keV 0.3 lA) and vacuum annealing at 900 K.The temperature of the wafer was monitored with an IR radiation thermometer. A small de–ciency in oxygen concentration 0.001% was estimated from the bulk resistivity of the blue crystal 2 ) m.13 Research grade HCOOD gas was puri–ed through trap-thaw cycles. row for illustration. Results and discussion When the TiO (110) surface heated at 450 K was exposed to a formic acid ambient of 1.3]10~6 2 Pa small particles were formed on the terraces. Fig. 2 shows the constant-current topography observed on the surface reacted with the atmosphere for 10 min. The exposed surface was cooled to room temperature (RT) in the presence of formic acid vapor and the atmosphere was evacuated.The topography was determined at RT. The diagonal zig-zag line over the frame was a single-atom height (0.32 nm) step between the (left) upper and the (right) lower terraces. The surface was saturated with adsorbed formate ions. The smallest protrusions presented in a gray scale and arranged in a (2]1) order are the formate ions.14 Many blips larger and brighter than the formate were observed on the terraces. The brighter particles can be assigned to the product of a reaction that took place on the exposed surface since they were not observed before the exposure at 450 K. Lines of slight contrast parallel to the [001] direction are the added Ti2O3 doublestrand rows inevitably formed on sputter-annealed (1]1) surfaces.3 They were also observed at the surface before the reaction.The spatial distribution of the product particles is quite interesting. The local number density depended on the distance from the step. This was obvious on the upper terrace. A few particles appeared in the proximity of the step whereas they were produced at random on the terrace a long way from the step. The near-step area on the terrace looks a depletion belt and is completely covered with formate ions as a result. This non-uniform distribution of the product particles was a reproducible phenomenon. Depletion belts of similar width were always observed near the steps on the surface (see Fig. 2) and also on similarly reacted TiO surfaces. 2 To elucidate the non-uniform distribution quantitatively the local number density of 204 particles observed in a zoomed-out image was plotted as a function of the distance from the step in a histogram (Fig.3). The densities observed in the segments nearest and second-nearest to the step Fig. 2 A post-reaction STM image of the TiO (110) surface heated at 450 K and exposed to 1.3]10~6 Pa HCOOD gas for 10 min. The constant current topography was determined on the surface cooled to room 2 temperature. Scan area 25]25 nm2 sample bias voltage ]2.0 V tunneling current 0.10 nA. One product particle is marked by the white circle. One (2]1) unit of the formate monolayer is shown with the solid rectangle. A solid line is superimposed on a Ti2O3 261 Faraday Discuss. 1999 114 259»266 Fig. 3 Local number density of the product (brighter) particles as a function of the distance from the step.in Fig. 3 were reduced from the value averaged over the third-nearest or farther segments. The degree of the reduction was outside of the random —uctuation predicted from statistics. The particle was immobile at the reaction temperature (450 K). Fig. 4(a) shows an STM image of a reacted surface with product particles recorded at 650 K in vacuum. At this temperature the formate ion decomposed unimolecularly,2,15 leaving no product particles on the surface. Sequential imaging revealed that the migration of the particles was prohibited even at this elevated temperature. When the surface was heated at 700 K the particles disappeared probably decomposing as shown in Fig. 4(b). A product particle stayed at the place where it was formed.These results indicate that the particle formation reaction that occurred on the terrace far from the step was suppressed in nm-proximity of single-atom height steps on TiO (110). This may be a 2 novel phenomenon. The inverse eÜect step-induced enhancement of a reaction has long been demonstrated.16 A typical example of step-induced enhancement has recently been visualized on NO/Ru(001).17 The nitric oxide molecules were dissociated at the coordinatively unsaturated Ru atoms exposed to steps. The resultant nitrogen atoms migrate from the dissociation sites (steps) into the terraces. The localized distribution of the N atoms was imaged in a post-reaction STM observation.17 The long range of the suppressive eÜect of the single-atom height steps on TiO (110) is another 2 interesting feature.Fig. 3 shows that the suppressive eÜect ranges over eight lattice units i.e. 2.4 nm. It is quite important to know how an atom-size surface singularity like the step aÜects the reactivity over nanometer proximity. In situ STM observation was tried during the particle formation reaction in order to obtain information on the location and transport of formate ions as a possible intermediate of the reaction. A sputter-annealed surface was heated at 420 K and formic acid vapor of 1.6]10~6 Pa was dosed in the microscope chamber at t\0. Fig. 5 shows an image of the exposed surface observed at t\230 s. The number of formate ions imaged as small bright dots increased with exposure time. The number density of formates was often reduced in the proximity of the steps.The diÜusion of the formate ions is activated at this temperature. We can thus assume that the non-uniform distribution of the product particles results from the non-uniform distribution of the formate intermediates. The next question is how the step regulates the formate density. The most simple interpretation is the electrostatic eÜect. The absence of Ti and O ions that would have existed there if the upper Faraday Discuss. 1999 114 259»266 262 Fig. 4 STM images of the reacted TiO surface observed in vacuum at (a) 650 K and (b) 700 K. Scan area 2 100]100 nm2 sample bias voltage ]2.6 V tunneling current 0.05 nA. terrace continued beyond the step causes a –nite modulation of the electrostatic –eld near the step. Another method of interpretation is a diÜusion-derived picture.The formate ion may be consumed very fast at the step site by recombinative desorption or a decomposition reaction as is postulated in the step-induced enhancement of reaction. The local number density of the formate ion has to reduce near the step since adsorption-desorption equilibrium and surface diÜusion are coupled so as to determine the local population of adsorbed species. We have simulated the phenomenon by eqn. (1) under the boundary condition of h\0 at x\0 (h particle coverage; x\distance from the step end) assuming a fast desorption from the step. The coverage change of formate ions against the reaction time is given by eqn. (1) including diÜusion adsorption and desorption terms. k D is the diÜusion constant is the rate constant for adsorption a (1) dh dt \D d d x 2h 2 ]kaP(1[h)[kdh and k is the rate constant for desorption.Coverage dependency on the distance was calculated by d using the equation. The simulated curve for h against x is shown in Fig. 3. The simulated curve reproduced the histogram well. In the steady-state eqn. (2) is obtained and the depletion-zone 263 Faraday Discuss. 1999 114 259»266 rows. Fig. 5 An in situ STM image of the TiO (110) surface reacted at 420 K in 1.6]10~6 Pa HCOOD gas. The 2 reactant atmosphere was introduced at t\0 s and the topography was observed at the reaction temperature at t\230 s. Scan area 20]20 nm2 sample bias voltage ]1.7 V tunneling current 0.2 nA. One formate ion is marked by the white circle.The seven thick white lines parallel to the [001] direction are Ti2O3 length j is given approximately by eqn. (3) where h is ka P/(kaP]kd). (2) Adh dxB at x\0 \ J = D(k kaP aP]kd) (3) j\Adh h= dxB at x\0 \ Jk J aP D ]kd (4) j d max\Sk D \S2 k k h L c d When P becomes close to zero j becomes maximum (jmax) and jmax is given by eqn. (4). D is kh L c2/2 where kh is a rate constant for hopping migration of formate ions and L c is the lattice constant of TiO (110). When the ratio kh/kd is 100 the calculated j reproduces the observed 2 max depletion-zone length of 2.4 nm. Finally it should be considered what the product particle is. Fig. 6 shows a zoomed-in image of a reacted surface. The atom-scale location of the product can be identi–ed by using the coexisting (2]1)-formate structure as a scale.A formate ion in the (2]1) monolayer is adsorbed on two –ve-fold coordinated Ti atoms in a bridge form with an O-C-O bond angle of 126°.14,18v20 The product particles were adsorbed at the on-top position of the bridge oxygen atoms. Cross-section analysis revealed that the topographic height of the product was higher by 0.15 nm than the top of the (2]1)-formate monolayer. The formate ion itself exhibits protrusions of 0.14 nm height from the surface in similar tunneling conditions.14 The accordance in topographic size suggests that the product particle and the formate ion are similar in physical size. In the steady-state reaction of formic acid on TiO (110) the main products were H and CO in the temperature range 420-550 2 2 2 K (the activation energy\15 kJ mol~1).15 These considerations allowed us to assume that a CO (CO32~) group molecule adsorbed at the on-top site of the bridge oxygen atom forming carbonate 2 makes the product particle as illustrated in Fig.6(c). The STM image of the particles in Fig. 6 seems elongated not spheric. A similar elongated feature was observed for carbonate (CO32~) Faraday Discuss. 1999 114 259»266 264 Fig. 6 A zoomed-in STM topography of the reacted TiO surface shown in Fig. 2. (a) Raw image scan area 2 7]8 nm2 sample bias voltage ]2.0 V tunneling current 0.1 nA. (b) The (1]1) lattices of the TiO surface 2 which are scaled on the (2]1) order of the formate monolayer are imposed with solid lines.(c) A model of carbonate as the product particle. Top and side views are shown. CO2 .21 Carbonate formation22 and produced on a Na-deposited TiO (110) surface exposed to 2 release15 have been observed on formate-covered TiO surfaces during heating. 2 CO2 Acknowledgement This work was supported by Core Research for Evolutional Science and Technology (CREST) of the Japan Science and Technology Corporation (JST). References 1 Y. Iwasawa Stud. Surf. Sci. Catal. 1996 101 21. 2 Y. Iwasawa Catal. Surveys Jpn. 1997 1 3. 3 H. Onishi K. Fukui and Y. Iwasawa Bull. Chem. Soc. Jpn. 1995 68 2447. 4 K. Fukui H. Onishi and Y. Iwasawa Phys. Rev. L ett. 1997 79 4202. Faraday Discuss. 1999 114 259»266 265 5 V. E. Henrich and P. A. Cox T he Surface Science of Metal Oxides Cambridge University Press Cambridge 1994.6 K. Fukui H. Onishi and Y. Iwasawa Chem. Phys. L ett. 1997 280 296. 7 H. Onishi and Y. Iwasawa L angmuir 1994 10 4414. 8 H. Onishi K. Fukui and Y. Iwasawa Colloids Surf. A 1996 109 335. 9 H. Onishi Y. Yamaguchi K. Fukui and Y. Iwasawa J. Phys. Chem. 1996 100 9582. 10 H. Onishi and Y. Iwasawa Phys. Rev. L ett. 1996 76 791. 11 H. S. Taylor Proc. R. Soc. L ondon Ser. A 1925 108 105; J. Am. Chem. Soc. 1931 53 578. 12 S. Suzuki Y. Yamaguchi H. Onishi K. Fukui T. Sasaki and Y. Iwasawa Catal. L ett. 1998 50 117. 13 H. Onishi and Y. Iwasawa Jpn. J. Appl. Phys. 1994 33 L1338. 14 H. Onishi and Y. Iwasawa Chem. Phys. L ett. 1994 226 111. 15 H. Onishi T. Aruga and Y. Iwasawa J. Catal. 1994 146 557. 16 G. A. Somorjai Introduction to Surface Science and Catalysis Wiley New York 1994. 17 T. Zambelli J. Wintterlin J. Trost and G. Ertl Nature (L ondon) 1996 273 1688. 18 Q. Guo I. Cocks and E. M. Williams J. Chem. Phys. 1997 106 2924. 19 S. Thevuthasan G. S. Herman Y. J. Kim S. A. Chambers C. H. F. Peden Z. Wang R. X. Ynzunza E. D. Tober J. Morais and C. S. Fadley Surf. Sci. 1998 401 261. 20 B. E. Hayden A. King and M. A. Newton J. Phys. Chem. B 1999 103 203. 21 H. Onishi T. Aruga C. Egawa and Y. Iwasawa J. Chem. Soc. Faraday T rans. 1 1989 85 2597. 22 K. S. Kim and M. A. Barteau J. Catal. 1990 125 353. Paper 9/03106E Faraday Discuss. 1999 114 259»266 266

ISSN:1359-6640    

DOI:10.1039/a903106e

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Scanning tunnelling microscopy studies of the reactivity of the TiO (110) surface Re-oxidation and the thermal treatment of 2 2 2 metal nanoparticles R. A. Bennett P. Stone and M. Bowker* Department of Chemistry University of Reading Reading UK RG6 6AD. E-mail M.Bowker=Reading.ac.uk 2 Received 10th May 1999 2 We have employed variable temperature scanning tunnelling microscopy (STM) to probe the surface structure of the TiO (110) surface with clean adsorbate and metal covered terminations. The aim of the work is to understand the nature of catalysis on supported metal oxide catalysts for which a good model is an admetal on a single crystal oxide surface. For Pd overlayers annealing in vacuum shows the formation of metal particles with nanometer sized dimensions which are comparable to those seen in real catalysts.The clean TiO (110) surface has two commonly observed terminations the (1]1) bulk truncation the (1]2) reduced and reconstructed surface. Less commonly for very reduced crystals the formation of ordered defects occurs leading to crystallographic shear planes. We have explored all of these surfaces by low energy electron diÜraction (LEED) and STM to provide structural information while we have employed dynamic imaging of the surface in reactive conditions at elevated temperature to assess the chemistry. We –nd that oxygen rich atmospheres promote a re-growth of the surface that has important consequences for the surface chemistry and morphology. The oxidation and reduction of the support in this system has been shown to modify the reactive properties of the supported metal and we relate our observations to the strong metal support interaction (SMSI).2 Introduction The surface structure of TiO (110) has recently been the subject of much interest due to the wide use of titania industrially as a white pigment in paint and cosmetics as a support for photocatalysts and as a biocompatible interface for medical implants. The behaviour of these disparate functions are critically determined by the surface properties of the oxide which as we shall show are dependent upon bulk defect concentrations. The ability to prepare conducting TiO samples in UHV has led to a wealth of studies on its surface structure and chemistry.1h16 However there has been relatively little work exploiting the eÜect of the variable bulk stoichiometry of reducible oxides on surface structure and reactivity.9 We have therefore investigated the reactivity of reduced samples at elevated temperatures in UHV conditions.We have identi–ed a novel reoxidation scheme with strong temperature dependences. Furthermore we are particularly interested in the use of TiO as a support for (photo)catalysts. To this end we have prepared model catalysts of Pd nanoparticles on TiO (110) surfaces and run the same reactions to investigate the in—uence of the supported particles. We –nd temperature-dependent enhancements in the rate of 267 Faraday Discuss. 1999 114 267»277 2 This journal is( The Royal Society of Chemistry 2000 re-oxidation around the Pd particles which we interpret as spillover of oxygen from metal to support.In addition the nature of the re-oxidation scheme de-activates the Pd particles. Experimental STM experiments were performed using an Oxford Instruments variable temperature STM contained within an UHV chamber equipped with additional facilities for Ar` ion sputtering low energy electron diÜraction (LEED) and Auger electron spectroscopy (AES) described in detail elsewhere.17 The chamber was ion pumped to produce a typical base pressure of 1]10~10 mbar. TiO (110) crystals (PIKEM UK) were repeatedly sputtered (600 eV RT) and annealed (1200 K) 2 to initially produce near stoichiometric light blue crystals which repeatedly produced (1]1) LEED patterns after annealing (10 s 1270 K).In time continued sputter»anneal18 cycles further reduced the sample which then displayed a dark-blue colour and (1]2) reconstructions after annealing. Crystals displaying Magneli phases resulting from crystallographic shear plane formation could be manufactured by prolonged annealing at 1300 K at which point they appeared dark blue/black.5 The sample was heated radiatively using a tungsten –lament situated close to the rear of the sample to attain temperatures \1000 K while scanning. E-beam heating by biasing the –lament was used to reach the extreme temperatures for annealing. Pd was deposited from a home-built source composed of high purity Pd wire tightly wrapped around a resistively heated W wire. Pd was deposited onto the TiO surface at 673 K and subsequently annealed at 773 K for 15 min.2 Results and discussion Surface structures Early STM studies of TiO reported a variety of interpretations of the images but more recently 2 the dominant contributions to the tunnelling have been established.8,9 In normal state images the empty states are of 3d character and so Ti4` ions are imaged whereas the bridging O atoms which are geometrically closer to the tip appear dark. Fig. 1 shows the (1]1) termination the image contrast is dominated by tunnelling into the 5-fold coordinated Ti sites which appear as rows in the S001T direction. In the following images the tunnelling parameters are such that the detail imaged within the surface structures is dominated by tunnelling into the Ti 3d states i.e. positive sample bias.Fig. 1 ” Small area scan (47 square) showing atomic resolution on the (1]1) terminated TiO (110) surface 2 (1 nA 1000 mV). Faraday Discuss. 1999 114 267»277 268 S116 0T direction the cross-linked (1]2) reconstruction forms antiphase domain 2O3 On reduction of the crystal the surface reconstructs to form a (1]2) termination that we have recently shown can exist as two distinct reconstructions which occur for diÜering levels of reduction. For low levels of reduction added rows of stoichiometry Ti2O3 are formed extending from step edges Fig. 2A which generate a true (1]2) surface reconstruction.2,3 The rows have a low overall corrugation (B1»1.5 ”) are centred on the bright rows of the (1]1) (i.e. the 5-fold co-ordinated Ti) and are relatively unreactive to oxygen exposure.For higher levels of reduction a cross-linked (1]2) reconstruction is formed in which the rows running in the S001T direction are periodically linked roughly B12 lattice units in the S001T direction Fig. 2B.6,7,11,19 These links most commonly occur as cross-shaped features in the trough between rows Fig. 2C. The reconstruction is an added row of TiO2 again centred on the 5-fold co-ordinated Ti of the underlying (1]1) surface and displays a larger corrugation in STM which is close to that expected for a normal step height (B3.1 ”).19 These surfaces display excellent LEED patterns Fig. 2D with extra spots due to the ordered array of cross-links. It is interesting to note that the two (1]2) reconstructions have markedly diÜerent accommodations of antiphase domain boundaries running in the walls composed of paired cross-links Fig.3 whereas the added Ti type (1]2) reconstruction terminates domains with bright points.20 Both types of reconstruction have similar S001T directed Fig. 2 The two (1]2) reconstructions on the reduced TiO surface. A Added rows of Ti2O3 growing out from step edges which are formed for small departures from stoichiometry (1 nA 1000 mV). B Cross-linked 2 BTiO added rows of TiO formed on the surface for departures from stoichiometry xB10~4. The cross- 2 2~x links form a well ordered array separated by 12 unit cells in the S001T direction (0.1 nA 1000 mV). C Close up of the cross-linking structure showing double rows of Ti within each added row which are periodically cross-linked in the S116 0T direction by bright features in the trough between rows (1 nA 1000 mV).D LEED pattern with distinctive 12th order spots and streaking in the S001T direction. 269 Faraday Discuss. 1999 114 267»277 Fig. 3 Antiphase domain boundaries on the cross-linked (1]2) surface. Domain walls appear as paired cross-links running in the S116 0T direction. Domain walls running in the S001T direction have an extra unit cell width separation between added rows (0.1 nA 1000 mV). boundaries with a single lattice spacing gap appearing in the S116 0T direction between the added rows. The cross-linked (1]2) surface also has occasional links that bridge these antiphase domain boundaries. This wide bridging is more prevalent when the surface is re-oxidising.For slightly higher levels of reduction the cross-linked (1]2) reconstruction remains but is accompanied by the formation of defects running diagonally across the images Fig. 4A B. These are the beginnings of the formation of crystallographic shear (CS) planes which arise from the ordered clustering of defects within a d0 metal oxide (TiO2 V2O5 MoO3 and WO3) crystal.1,21,22 The formation of CS planes is also seen in real catalyst TiO2 supports after reduction23 and so we believe that using the TiO at this stoichiometry represents a good support 2 material for model catalysts. The formation of CS planes occurs in reduced crystals with departures from stoichiometry of around TiO2~x x[10~4. We shall concentrate on the reactivity of surfaces with about this level of reduction in later sections.For more extreme levels of reduction the CS planes have recently been shown to terminate at the surface in a well ordered array of Fig. 4 Crystallographic shear plane pairs running diagonally across the image formed on the cross linked surface by slow cooling from 1200 K. Narrow strips of (1]1) terminated surface appear between the CS plane pairs. A Large area image (1000 ” square) showing variety of directions that CS planes may take (0.2 nA 1000 mV). B Close up (150 ” square) of (1]1) strip between planes (1.5 nA 1000 mV). Faraday Discuss. 1999 114 267»277 270 half-height steps5,24h26 on TiO (110). Such ordering of the CS planes produces Magneli phases of 2 variable stoichiometry depending upon separation of the planes and produce complex LEED patterns.5 Reduction Vacuum annealing is well known to reduce TiO surfaces and is commonly applied to produce 2 suitably conductive crystals for STM imaging.In Fig. 5 we show three images of the cross-linked (1]2) surface maintained at 1000 K in UHV taken at time t\0 1980 s and 2580 s. Towards the top right of all three images are a small group of pairs of CS planes that are used as markers in the images as they remain –xed in position (they are surface terminations of extended bulk defects). The cross-links of the (1]2) are not apparent as at this temperature they are mobile on the timescale of scanning. Analysis of the images shows that the step edges move and change shape. Material is predominantly lost from step edges parallel to the S116 0T with movement of S001T edges arising from removal of the (1]2) rows from the S116 0T ends.Some TiO is redis- 2 tributed and extends from the step edges in the vicinity of the shear planes which may be due to an enhanced stability of steps that incorporate a shear plane displacement such that the step Fig. 5 ” Sequential images (1000 square) taken at 1000 K of the cross-linked (1]2) surface reducing by loss of oxygen to the vacuum. A Time t\0 showing CS plane group and several step edges (0.1 nA 1500 mV). B Time t\1980 s showing movement of step edges predominantly in the S001T direction (0.1 nA 1500 mV). C Time t\2580 s showing continued movement of step edges and the 1-dimensional shortening of the added row exposed on the S001T directed step edge (0.1 nA 1500 mV).271 Faraday Discuss. 1999 114 267»277 Fig. 6 Sequence of images showing re-oxidation of the cross-linked surface maintained at 673 K in the presence of B4]10~7 mbar O2 . A»D Cross-links increase in number and aggregate to form small islands of (1]1) within the (1]2) terraces. E»F (1]1) islands coalesce to form extended (1]1) terraces upon which bright points form. These points grow to form rows on the surface which cluster to form a cross-linked (1]2) surface. Tunnelling conditions A B and F (0.1 nA 1000 mV) C and D (0.1 nA 1500 mV) and E (0.1 nA 1400 mV). edges have 1 the normal height. The average rate of retraction of the S116 0T step edges calculated 2 from these images is ” B0.021^0.006 s~1. The net loss of material from the surface implies that oxygen vacancies are not the only defects introduced by vacuum annealing.The production of oxygen vacancies which diÜuse into the bulk Faraday Discuss. 1999 114 267»277 272 (and consequently the diÜusion of oxygen to the surface) would maintain the surface structure and therefore no loss of material would be seen. However we have measured a loss of material which implies removal of both O and Ti ions from the surface. Oxygen loss may be accommodated by vacancy formation but as the surface reconstruction is not changed the Ti must dissolve into the bulk and form interstitial Ti ions. Further evidence for the presence of interstitial Ti ions is presented below. Re-oxidation Fig. 6 shows a sequence of images of the same area taken at 673 K (the images presented are only a selection taken from the full sequence of over 100 frames) showing the reaction of a cross-linked (1]2) surface with a low pressure of oxygen admitted to the chamber (the oxygen pressure varies slightly throughout the sequence but is approximately 4]10~7 mbar).Fig. 6A shows the surface prior to reaction with ordered cross-links antiphase domain boundaries and a small depression of (1]1) within the terrace which will act as a marker. The reaction begins with an increase in number of the cross-links between the rows which also begin to bridge the antiphase domain boundaries directed in the S001T direction Fig. 6B. Additionally small areas of (1]1) have grown out from the step edges into the lower (1]2) terrace. As the cross-links increase in density islands of (1]1) nucleate and grow within the (1]2) Fig.6C. Many (1]1) islands form on the terraces and grow preferentially in the S001T direction along the cross-links. These islands coalesce to form larger islands Fig. 6D which spread to cover the terraces. At this temperature the (1]1) nearly completes a layer before the next stage of the re-oxidation is observed. In this second phase bright points appear on top of and towards the centre of the (1]1) islands Fig. 6E. The bright points are stable and act as nucleation centres for the formation of bright rows that grow in the S001T direction Fig. 6F. The regions in which these rows aggregate form a (1]2) surface on top of the (1]1) surface. Neighbouring (1]2) rows cross-link to form regions with the initial surface termination which spreads across the terrace.The reaction now proceeds in a cyclic fashion with small (1]1) islands nucleating and growing within the (1]2) terraces. It is possible to grow many layers in this way. The most striking feature of this re-oxidation reaction is that the surface clearly grows. Step edges move out across the lower terraces troughs in the (1]2) structure –ll in to form the (1]1) termination new growth is seen on top of the (1]1) terraces in the form of rows which aggregate to form a new (1]2) surface. If oxygen vacancies were the dominant defect structure re-oxidation would simply replace the vacancy with oxygen without growth of the surface. Growth necessitates the incorporation of Ti into the new layer.Thus Ti interstitials which were distributed throughout the crystal during reduction are captured by the ambient oxygen as they diÜuse to the surface region where they are incorporated into the growing surface. Nanoparticle reaction Re-oxidation and reduction cycles are fundamental processes in catalysis the most familiar being the so called Mars van Krevelen mechanism whereby oxygen is inserted from an oxide catalyst into a reactant to form the product. The resulting oxygen vacancy is healed by adsorption of oxygen from the gas or adsorbed phases. In the latter case oxygen atoms may be supplied from a second component in the catalyst system and this is termed spillover. Spillover of reactants from one component to another can result in enhanced reactivity due to synergy of action.27 We have used nanoscale Pd particles supported on reduced TiO (110) as a model catalyst 2 surface and to investigate spillover by the re-oxidation of the support.Fig. 7 shows images of a Pd nanoparticle covered (1]2) reconstructed TiO (110) surface maintained and imaged at 573 673 2 and 773 K (each taken from a sequence of over 40 images). The –rst panel at each temperature shows the particles just before admission of oxygen at B2]10~7 mbar. The second panel in the group shows each surface at a later time during the re-oxidation reaction with the –nal state of the surface being displayed in the last panel. Starting with the experiment at 573 K panel A shows three particles on the surface two at step edges and one in the terrace.After admission of the gas the terrace slowly reacts to form small (1]1) islands within the (1]2) terraces. These small (1]1) islands then promptly develop 273 Faraday Discuss. 1999 114 267»277 Fig. 7 Three sequences of images taken at 573 K (A»C) 673 K (D»F) and 773 K (G»I) showing the in—uence of Pd nanoparticles on the re-oxidation of the surface.§ At 573 K (A»C) TiO preferentially regrows close to the particles and completely encapsulates them leaving raised features above the terrace which itself only 2 grows slowly. At 673 K (D»F) particles again become buried but the region of Pd enhanced re-growth spreads out a substantial distance from the particle resulting in raised terraces with (1]1) terminations. The surface far from the particles grows slower.At 773 K (G»I) the terrace grows in a layer-by-layer fashion with little preferential growth around the particles. Tunnelling conditions A (0.1 nA 1000 mV) B and C (0.1 nA 2000 mV) D (0.1 nA 1500 mV) E (0.1 nA 2000 mV) F (0.1 nA 1000 mV) and G»I (0.1 nA 1500 mV). § An STM movie corresponding to panels D»F is available as electronic supplementary information. See http ://www.rsc.org/suppdata/fd/1999/267 Faraday Discuss. 1999 114 267»277 274 Fig. 7 Continued. bright points on top of them as another layer begins to grow. This behaviour is identical to that seen for the reduced surface re-oxidising with no Pd present. By panel B the step edges are obscured by the disordered re-growth. However the regions closely surrounding the Pd particles appear to have grown faster with the appearance of steps leading up to the Pd particles.The small particle to the lower left is almost buried. In the last panel the same particle has been buried by the growth of the titania while the particle in the centre of the terrace has clearly experienced enhanced growth in the area ” B50 around the particle. There are B5 layers of growth up the 275 Faraday Discuss. 1999 114 267»277 side on this particle from the terrace which itself has grown 3 layers since the beginning of the reaction giving a total of 8 layers growth at the particle. Panels D»F show images from a similar sequence taken at 673 K. Far from the particles the surface behaves as described above with the sequential growth of (1]1) terraces and the formation of bright points on the newly created terraces which aggregate to form a new (1]2) layer.However panel E shows that the reaction leads to a more pronounced peripheral growth around each particle. This extends ” B100 from each particle until the peripheries merge to produce a (1]1) terrace. The new terraces which form and grow outwards from the particles always display the (1]1) termination. By panel F the particles have preferentially grown TiO around them- 2 selves until they have become buried. In total there has been B7 layers of growth around the particle and this occurs approximately 16 times as fast as the growth on the non-Pd-covered surface at this temperature. Panel In the –nal three panels G»H the surface is reacting at 773 K with 2]10~7 mbar O2 . G shows many particles supported on a surface which also displays several CS planes.Again the reaction is started and the reaction sequence followed. In this case little growth is observed preferentially around the particles. Panel H shows the particle-covered surface after the complete growth of the –rst (1]1) layer and the beginnings of the growth of a new (1]2) layer. The (1]2) rows nucleate and grow in a similar manner to the particle free surface although there seems to be a slight preference for nucleating at the particle. Interestingly the nucleation of the growth at the particle diÜers at this temperature to the experiments at lower temperature in that (1]1) terraces do not form preferentially at the particle. At 773 K the growth is always of the (1]2) rows which suggests that the particles are nucleation centres for growth but do not contribute an excess of spillover oxygen which can be incorporated to directly form (1]1) terraces.In panel I the smaller particles have been covered over by TiO growth (B3 layers). The regions 2 surrounding the particles show little preferential growth but do show the slight preference for nucleating the new layers growth as (1]2) rows. Continuing this reaction at higher oxygen pressures buries the nanoparticles on the surface but the growth rate at the particles periphery is the same as on the clean surface. The observation of three regimes reactivity as a function of temperature for this system allows some discussion of the mechanism. At low temperatures (573 K) the Pd nanoparticles rapidly become covered by TiO to form an encapsulated particle that protrudes from the terrace.At 673 2 K TiO preferentially grows outwards from the nanoparticle in a layer-by-layer fashion. Even- 2 tually the particles are covered by the re-growth. The rate of growth due to the Pd particles presence is B16 times as fast as that observed for the clean surface and so the surface shows a roughened terrace structure over each particle. At 773 K there is little or no preferential growth of (1]1) at the particles. However the nanoparticles do appear to nucleate the formation of the next (1]2) layer which then grows in the same fashion as on the clean surface. The particles again become covered but due to the layer-by-layer growth which is not preferential around the particles —at terraces result that display the normal TiO structure.These observations and the 2 departures from the clean surface reaction can be rationalised within a scheme which predominantly considers the action of oxygen adsorption and desorption on the Pd particles. At 773 K oxygen desorbs rapidly from single crystal Pd(111) (peak B800 K28) and so with the small oxygen gas phase pressure we expect a low steady state surface coverage of oxygen on the Pd nanoparticles (Pd particles preferentially display low index facets predominantly (111)29). The low steady state coverage on the particle implies that the rate of spillover of oxygen atoms from the particles to the support will also be low. Thus no Pd enhanced growth is observed as the entire —ux of oxygen to the surface is directly from the gas phase and the reaction scheme of the clean surface is followed.In contrast at 673 K oxygen cannot desorb rapidly from the Pd nanoparticle and a large steady state coverage is produced. However the oxygen atoms on the Pd can make excursions oÜ the particle and diÜuse over the neighbouring TiO surface until they capture a Ti 2 interstitial and become incorporated into the growing layer. This produces an enhancement in growth rate over that of the clean surface as the sticking probability of oxygen on Pd is over an order of magnitude greater than on the reduced TiO surface. Thus the Pd particles supply 2 oxygen atoms to the TiO which allows the growth of the fully oxygen terminated (1]1) struc- 2 ture. At 573 K the Pd particles remain a reservoir of oxygen however the lower temperature reduces the extent of diÜusion oÜ the particle and onto the support.Furthermore the lower Faraday Discuss. 1999 114 267»277 276 temperature results in a less extended diÜusion of oxygen on the TiO and so the growth is 2 restricted to the area in the immediate vicinity of the nanoparticle. Conclusions We have shown that non-stoichiometry in TiO2~x samples is accommodated by the dissolution of Ti interstitials into the bulk of the rutile crystal structure. During re-oxidation these ions are extracted and grow on the surface in their normal crystallographic positions. This has wide ranging implications for catalysis and gas sensing since in both areas models of reactivity and sensing ability are based on descriptions of the formation and annihilation of oxygen vacancies.We have shown that titanium interstitials accommodate non-stoichiometry and are key to accurate descriptions of oxidation and reduction at elevated temperature on this surface. The regrowth behaviour has been used to probe spillover of reactive oxygen species from supported metal particles. In this case we show that the reactivity increase due to spillover oxygen is strongly temperature dependent and controlled by the adsorption/desorption from the Pd particle. We note that nanoscale particles supported on this oxide can be buried by re-growth of titania and are thus removed from the surface»a kind of strong metal support interaction (SMSI). Acknowledgements PS would like to thank Oxford Instruments PLC for the provision of a CASE award and we thank the EPSRC for support.References 1 G. S. Rohrer V. E. Henrich and D. A. Bonnell Science 1990 250 1239. 2 H. Onishi and Y. Iwasawa Phys. Rev. L ett. 1996 76 791. 3 H. Onishi and Y. Iwasawa Surf. Sci. 1994 313 L783. 4 D. Novak E. Garfunkel and T. Gustafsson Phys. Rev. B Condens. Matter 1994 50 5000. 5 R. A. Bennett S. Poulston P. Stone and M. Bowker Phys. Rev. B Condens. Matter 1999 59 10341. 6 M. Sander and T. Engel Surf. Sci. 1994 302 L263. 7 A. Szabo and T. Engel Surf. Sci. 1995 329 241. 8 U. Diebold J. F. Anderson K. O. Ng and D. Vanderbilt Phys. Rev. L ett. 1996 77 1322. 9 K. O. Ng and D. Vanderbilt Phys. Rev. B Condens. Matter 1997 56 10544. 10 P. W. Murray N. G. Condon and G.Thornton Phys. Rev. B Condens. Matter 1995 51 10989. 11 C. L. Pang S. A. Haycock H. Raza P. W. Murray G. Thornton O. Gué lseren R. James and D. W. Bullet Phys. Rev. B Condens. Matter 1998 58 1586. 12 R. E. Tanner M. R. Castell and G. A. D. Briggs Surf. Sci. 1998 412/413 672. 13 M. A. Henderson Surf. Sci. 1999 419 174. 14 M. A. Henderson Surf. Sci. 1995 343 L1156. 15 W. S. Epling C. H. F. Peden M. A. Henderson and U. Diebold Surf. Sci. 1998 412/413 333. 16 M. Li W. Hebenstreit and U. Diebold Surf. Sci. 1998 414 L951. 17 M. Bowker S. Poulston R. A. Bennett P. Stone A. H. Jones S. Haq and P. Hollins J. Mol. Catal. A Chem. 1998 131 185. 18 Sputtering while hot ([800 K) greatly increases the rate of reduction of the crystal due to preferential sputtering of surface O atoms and replenishment by diÜusion of Ti to the bulk. 19 R. A. Bennett P. Stone and M. Bowker Phys. Rev. L ett. 1999 82 3831. 20 R. E. Tanner M. R. Castell and G. A. D. Briggs Surf. Sci. 1998 412/413 672. 21 L. A. Bursill and B. G. Hyde in Progress in Solid State Chemistry ed. H. Reiss and J. O. McCaldin Pergamon New York 1972 vol. 7 p. 177. 22 M. Reece and R. Morrell J. Mater Sci. 1991 26 5566. 23 S. Bernal F. J. Botana J. J. Calvino C. Loç pez J. A. Peç rez-Omil and J. M. Rodrïç guez-Izquierdo J. Chem. Soc. Faraday T rans. 1996 92 2799. 24 G. S. Rohrer V. E. Henrich and D. A. Bonnell Surf. Sci. 1992 278 146. 25 H. Noé renberg R. E. Tanner K. D. Schierbaum S. Fischer and G. A. D. Briggs Surf. Sci. 1998 396 52. 26 H. Noé renberg and G. A. D. Briggs Surf. Sci. 1998 402»404 738. 27 B. Delmon Surf. Rev. L etts. 1995 2 25. 28 X. Guo A. HoÜman J. T. Yates Jr. J. Chem. Phys. 1989 90 5787. 29 K. Wolter O. Seiferth H. Kuhlenbeck M. Baé umer H.-J. Freund Surf. Sci. 1998 399 190. Paper 9/03731D 277 Faraday Discuss. 1999 114 267»277

ISSN:1359-6640    

DOI:10.1039/a903731d

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Oxygen-induced morphological changes of Ag nanoclusters supported on TiO (110) 2 Xiaofeng Lai Todd P. St.Clair and D. Wayne Goodman* Department of Chemistry T exas A&M University P.O. Box 30012 College Station TX 77842-3012 USA 2 2 2 Received 8th April 1999 2 2 2 The eÜect of in situO exposure on TiO (110)-supported Ag nanoclusters was investigated using X-ray photoelectron spectroscopy (XPS) and scanning tunneling microscopy (STM). An oxygen-induced cluster ripening was observed by STM after Ag/TiO (110) was exposed to 10.00 Torr O for 2 h in an elevated-pressure reactor. The Ag clusters exhibit a clear bimodal size distribution after O exposure due to Ostwald ripening some clusters increase in size while other clusters decrease in size.The cluster density also increased 5»15% after O exposure indicating redispersion simultaneously occurs with ripening. It is shown that intercluster transport is likely accomplished through the formation of Ag2O. 1. Introduction A primary goal of heterogeneous catalysis is to elucidate the relationship between catalyst structure and activity. However the achievement of this objective is often complicated by the morphological and structural changes that surfaces sometimes undergo at reaction pressures. These adsorbate-induced surface restructurings result from the reorganization of surface atoms near the adsorption site of a chemisorbed atom or molecule. Depending on the strength of the surface» adsorbate bond the surface may experience diÜerent degrees of restructuring from weak local relaxation to massive transportation of surface atoms.In the case of ìì real worldœœ catalysts which often consist of small metal particles supported on oxide substrates one is not only concerned with changes associated with the substrate surface but the supported metal clusters as well. Metal nanoclusters contain a relatively small number of atoms thus the surface atoms can be easily in—uenced because of their low coordination number making supported nanoclusters vulnerable to adsorbate-induced morphological changes. If this restructuring of metal clusters results in cluster growth then active metal surface area has been lost and the catalyst may become deactivated. The physical and chemical properties of supported metal clusters and the interactions between these clusters and diÜerent substrates have been widely investigated.1h9 The scanning tunneling microscope (STM) is particularly suited for studying such interactions because it can directly probe cluster morphology and structure.For example STM results indicate that exposure of CO pressures between 10~3 and 10~1 mbar to Rh/TiO (110)[(1]2) led to a signi–cant agglomeration of Rh clusters.10 This phenomenon was attributed to the formation of Rh»CO bonds (185 kJ mol~1) that promote disruption of the weaker Rh»Rh bonds (44.5 kJ mol~1). Similar eÜects have also been observed for Ir/TiO (110)[(1]2).11 2 2 In a recent study of CO oxidation over model Au catalysts the well-known structure sensitivity and the stability of Au nanoclusters to CO and O were investigated.12 STM in conjunction with 279 2 Faraday Discuss.1999 114 279»284 This journal is( The Royal Society of Chemistry 2000 reaction kinetics measurements revealed that the structure sensitivity was related to a quantum size eÜect with respect to the thickness of the Au islands maximum reactivity was observed for Au clusters with diameters of B3.5 nm and heights less than 3 atomic layers. It was observed in the course of these CO oxidation experiments that O exposure resulted in sintering of the Au nano- 2 clusters. Fig. 1 shows two STM images of 0.25 monolayer (ML) Au/TiO (110) before (top image) 2 and after (bottom image) exposure to 10.00 Torr O for 2 h. The eÜect of the O exposure was a 2 2 decrease in cluster density (number of clusters per unit area) and an increase in cluster size from 2.6 nm]0.7 nm (diameter]height) to 3.6 nm]1.4 nm.These results demonstrate the utility of STM for studying supported metal nanoclusters by establishing a correlation between the structural electronic and reactivity properties of model Au catalysts. Furthermore given the sintering eÜect of O on Au nanoclusters it is expected that other metal nanocluster systems could also be 2 susceptible to O -induced morphological changes. 2 Silver is a metal whose interaction with O environments is of interest primarily because of its 2 use as an industrial catalyst for two important oxidation reactions ethylene oxidation to ethylene Fig. 1 STM images of Au clusters on TiO2(110)[(1]1) (2.0 V 2.0 nA). The dosing —ux is 0.083 ML min~1 and the Au coverage is 0.25 ML.Top Fresh Au/TiO (110) deposited at room temperature and then annealed to 600 K. Bottom 0.25 ML Au/TiO (110) exposed to 10.00 Torr O for 2 h at 300 K. TheO exposure results 2 2 2 in decreased cluster density and increased cluster size. 2 Faraday Discuss. 1999 114 279»284 280 epoxide and methanol oxidation to formaldehyde.13h15 To gain additional insight into the role that O plays in cluster growth Ag clusters deposited on a TiO (110) surface were characterized 2 2 using XPS and STM before and after O exposures. 2 2. Experimental Experimental details have been published elsewhere.6,7,9,12 Brie—y the experiments were performed in a combined elevated pressure reactor-UHV system with a base pressure of 5]10~11 Torr.The system is equipped with a double pass cylindrical mirror analyzer (CMA) for Auger electron spectroscopy (AES) and X-ray photoelectron spectroscopy (XPS) reverse view low energy electron diÜraction (LEED) optics and an Omicron UHV-STM. A TiO (110) single crystal 2 (Commercial Crystal Laboratories Inc.) typically prepared by cycles of Ar ion bombardment and vacuum annealing to 1100 K was found to be sufficiently clean —at and conductive for electron spectroscopies and STM studies. Ag clusters were evaporated onto the TiO (110) surface from a 2 source containing high purity Ag wire wrapped around a Ta –lament that could be heated resistively. The Ag doser was extensively outgassed prior to use. The experimental Ag —ux of 0.125 ML min~1 was calibrated with a Re(0001) substrate using AES and STM.One ML Ag coverage corresponds to 1.39]1015 atoms cm~2. A combined elevated pressure reactor-UHV system was used for studies of high-pressure gas exposures and reactions. Following preparation and characterization in the primary UHV chamber the sample was transferred in situ into the elevated-pressure reactor through doublestage diÜerentially pumped Te—on sliding seals. This pumping arrangement provides a convenient way of performing elevated-pressure adsorption and reaction kinetics studies in the range of 1]10~10 to 1]103 Torr while maintaining UHV pressures in the main chamber. 3. Results and discussion The top image in Fig. 2 shows an STM constant current topograph (CCT) image of 2.0 ML Ag deposited on TiO (110) in UHV conditions at ambient temperature.Three-dimensional (3D) 2 hemispherical Ag clusters with relatively homogeneous sizes were observed both on —at terraces and step edges. The Ag clusters have an average size of B4.8 nm]2.6 nm (diameter]height) corresponding to B1900 atoms per cluster. In addition to the 3D Ag clusters bare patches of substrate were also visible indicating an island growth mode (Volmer»Weber growth mode) for Ag on TiO (110) consistent with our previous reports for other transition metals on 2 (110).6,7,9,12 XPS results (not shown) indicated a Ag 3d TiO (Eb) of 368.1 eV binding energy 2 5@2 consistent with metallic silver. After deposition of 2.0 ML Ag on the TiO (110) substrate and subsequent investigation by 2 STM the sample was transferred to the elevated-pressure reactor and exposed to 10.00 Torr O at 2 ambient temperature for 2 h.The sample was then transferred back to the UHV chamber and examined by XPS and STM. XPS results (not shown) indicated no Ag 3d5@2 Eb shift after O2 exposure consistent with other XPS studies of spent Ag catalysts.14 The bottom image in Fig. 2 presents an STM topograph of O -exposed Ag/TiO (110) and clearly reveals that the exposure 2 2 dramatically aÜected the cluster sizes. A bimodal size distribution of cluster sizes is apparent with some clusters increasing in size while others decrease. Furthermore a small increase (5»15%) in cluster density was observed indicating that redispersion is occurring as well. Fig. 3 shows the size distributions calculated from the STM images of Ag clusters before and after high-pressure O exposure.Initially Ag clusters exhibited a unimodal size distribution from 2 2.0 to 6.5 nm with a maximum diameter B5.0 nm. However after O exposure a bimodal size 2 distribution was observed with one size domain from 1.0 to 5.0 nm and another from 5.0 to 11 nm. The smaller clusters in the range 1.0»5.0 nm have a higher density and a narrower size distribution with an average Ag cluster size of B3.0 nm] B1.1 nm (B260 atoms). The larger clusters however have a lower cluster density and a broader size distribution with an average size of B6.7 nm] B3.1 nm (B4200 Ag atoms per cluster). It is noteworthy that the total cluster volume before and after O exposure calculated from the STM images agrees to within 10% error.2 Since Ag oxidation catalysts are typically used in elevated O pressure conditions an analysis of 2 these induced morphological changes is of general interest to the catalysis community. 281 Faraday Discuss. 1999 114 279»284 for 2 h at 300 K. The O exposure results in a bimodal size distribution certain clusters increase in size at 2 Fig. 2 STM images of Ag clusters on TiO2(110)[(1]1) (2.0 V 1.0 nA). The dosing —ux is 0.125 ML min~1 and the Ag coverage is 2.0 ML. Top Fresh Ag/TiO (110) prepared at room temperature. Homogeneous Ag clusters are observed with a uniform size distribution. Bottom 2.0 ML Ag/TiO (110) exposed to 10.00 Torr 2 O2 the expense of other clusters. 2 In general cluster growth of supported metal catalysts can proceed by two processes.First clusters can migrate along the surface until they collide with other clusters resulting in coalescence. Second cluster growth can occur by intercluster transport or Ostwald ripening which is capillarity driven. In this case the reduction of the total surface free energy by intercluster transport occurs such that certain clusters grow larger at the expense of other clusters.4 Thus in light of the bimodal size distribution observed following O exposure Ostwald ripening is likely 2 the cause of the Ag cluster growth. Regardless of the cause cluster growth results in catalysts with decreased active surface areas leading to a decline in catalytic activity. Intercluster transport of atomic (or molecular) species can occur by either surface diÜusion along the substrate or vapor phase transport.Under vacuum or reducing conditions the transport between Ag clusters can only occur in the form of free metallic Ag atoms and the driving force should be related to the Ag vapor pressure. However the Ag vapor pressure depends exponentially on the energy required to break Ag»Ag metal bonds and transfer a Ag atom to the vapor Faraday Discuss. 1999 114 279»284 282 Fig. 3 The size distributions of 2.0 ML Ag/TiO (110) before and after 10.00 Torr O exposure. Top Fresh 2 Ag clusters with a maximum volume B5.0 nm cluster diameter. Bottom Ag clusters after O2 exposure 2 exhibiting a bimodal distribution with average cluster diameters of B3.0»3.5 nm and 6.5»7.0 nm respectively. * phase i.e. the sublimation energy which is B285 kJ mol~1.Obviously such a high sublH(Ag) energy barrier suggests that intercluster transport by free Ag atoms will be very slow at room temperature. This result is in accordance with our STM observations that Ag clusters are generally stable in UHV conditions. In an oxidizing environment the situation is quite diÜerent. For example Wynblatt16 showed that growth of Pt particles in O environments occurred through the formation of volatile PtO 2 2 . Platinum oxide has a lower sublimation energy than platinum metal and therefore serves as the mechanism by which intercluster transport occurs. Unfortunately to the best of our knowledge no vapor pressure or sublimation energy data is available for silver oxide rendering it difficult to directly compare such values with Ag metal.However it will be shown that the formation of Ag2O O2 from Ag particles in 10.00 Torr is at least expected thermodynamically and thus may account for the intercluster transport discussed above. by reaction with oxygen at room tem- Thermodynamically Ag can form silver oxide (Ag2O) perature. This can be illustrated by considering the following simple reaction (1) 2Ag(s)]12O2(g)]Ag2O(s) which has a negative standard free energy of formation of Ag2O (*G298\[11.2 kJ mol~1) at room temperature. The equilibrium constant K for the above reaction can be expressed as (2) P KP\aP(aAg)~2(pO)~1@2 are the activities of Ag2O and Ag respectively (both values are unity) and p is O K into the standard *G equation p where a and a P Ag the equilibrium partial pressure of oxygen.Substitution of yields (3) DGT\12RT ln pO2 value in eqn. (4) for T\298 K. By substituting2 DG where T is the absolute temperature and R is the universal gas constant. Ag oxidation can only occur under the present conditions if the partial pressure of O in the cluster environment is pO2 higher than the 298\[11.2 kJ mol~1 into eqn. (4) the equilibrium oxygen partial pressure pO2 was calculated to be 1.23]10~4 atm (0.094 Torr). Since the partial pressure in the cluster environment (10.00 Torr) is higher than the equilibrium oxygen partial pressure at 300 K then the oxidation of bulk Ag is thermodynamically allowed. 283 Faraday Discuss. 1999 114 279»284 An additional eÜect that that must be considered is the in—uence of the Ag cluster curvature on the free energy.The decrement Dg in free energy due to cluster curvature is given by (4) Dg\2pM/(or) where p is the surface energy M is the atomic weight o is the density and r is the cluster curvature radius.17 At room temperature taking p\1400 erg cm~2 18 and o\10.5 g cm~3 *G298(r) the standard free energy of formation of Ag for Ag clusters with curvature radius r (in nm) is given 2O by (5) DG298(r)\DG298[Dg\[11.2[28.8/r(kJ mol~1) For an average cluster diameter of 5.0 nm (r\2.5 nm) the value of DG298(r) is doubled (DG298(r)\ [22.7 kJ mol~1). As the cluster size decreases further to 3.0 nm (r\1.5 nm) the o DG298(r) o value increases by a factor of 1.7. Therefore the driving force for oxidation of absolute Ag nanoclusters is increased signi–cantly at room temperature when accounting for cluster curvature eÜects.While this analysis shows that Ag2O O2 formation from Ag and is possible at room temperature information about the rate of Ag formation is not available. 2O 4. Conclusions In summary Ag nanoclusters on TiO (110) are likely to form Ag2O when exposed to 10.00 Torr 2 O at room temperature. Ag atoms in the form of volatile oxide molecules can thus be transported 2 between clusters at higher rates than metallic Ag atoms under vacuum or reducing conditions. In principle this somewhat volatile oxide may diÜuse more easily along the surface of the support or through the vapor phase from higher-energy sites to lower-energy sites. Consequently Ostwald ripening is observed when the Ag nanoclusters are exposed to oxygen.Evidence for redispersion was also found as the cluster density increased slightly following O exposure. The STM results 2 provide convincing evidence that the TiO (110)-supported Ag clusters are exceptionally reactive to 2 O and reinforce the notion that nanoclusters are particularly susceptible to adsorbate-induced 2 restructurings. 5. Acknowledgements We acknowledge with pleasure the support of this work by the Department of Energy Office of Basic Energy Sciences Division of Chemical Sciences and the Robert A. Welch Foundation. Paper 9/02795E References 1 D. R. Rainer C. Xu and D. W. Goodman J. Mol. Catal. A Chem. 1997 119 307. 2 C. T. Campbell Surf. Sci. Rep. 1997 27 1. 3 P. Wynblatt and N. A. Gjostein in Progress in Solid State Chemistry ed.J. O. McCaldin and G. Somorjai Pergamon Press New York 1975 9 p. 21. 4 P. Wynblatt R. A. Dalla Betta and N. A. Gjostein in T he Physical Basis for Heterogeneous Catalysis ed. E. Drauglis and R. I. JaÜee Plenum Press New York 1975 p. 501. 5 R. Persaud and T. E. Madey Chem. Phys. Solid Surf. 1997 8 407. 6 C. Xu X. Lai G. W. Zajac and D. W. Goodman Phys. Rev. B Condens. Matter 1997 56 13464. 7 X. Lai C. Xu and D. W. Goodman J. V ac. Sci. T echnol. A 1998 16 2562. 8 M. Valden and D. W. Goodman Isr. J. Chem. 1998 38 285. 9 X. Lai T. P. St.Clair M. Valden and D. W. Goodman Prog. Surf. Sci. 1998 59 25. 10 A. Berkoç G. Menesi and F. Solymosi J. Phys. Chem. 1996 100 17732. 11 A. Berkoç and F. Solymosi Surf. Sci. 1998 411 L900. 12 M. Valden X. Lai and D. W. Goodman Science 1998 281 1647. 13 S. Cheng and A. Clear–eld J. Catal. 1985 94 455. 14 V. I. Bukhtiyarov I. P. Prosvirin R. I. Kvon S. N. Goncharova and B. S. Balœzhinimaev J. Chem. Soc. Faraday T rans. 1997 93 2323. 15 V. I. Bukhitiyarov A. I. Boronin I. P. Prosvirin and V. I. Savchenko J. Catal. 1994 150 268. 16 P. Wynblatt Acta Metall. 1976 24 1175. 17 J. D. Verhoeven in Fundamentals of Physical Metallurgy Wiley New York 1975 p. 169. 18 J. D. Verhoeven in Fundamentals of Physical Metallurgy Wiley New York 1975 p. 202. Faraday Discuss. 1999 114 279»284 284

ISSN:1359-6640    

DOI:10.1039/a902795e

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

First principles simulations of titanium oxide clusters and surfaces Tristan Albaret Fabio Finocchi and Claudine Noguera L aboratoire de Physique des Solides (UMR CNRS 8502) Ba� t. 510 Universiteç Paris-Sud 91405 Orsay France 1 Introduction TiO is an important material for reasons of both fundamental interest and potential technologi- 2 cal applications. For example TiO is widely used in catalysis photocatalysis protective surface 2 coating etc.1 The stoichiometric oxide is a non-magnetic insulator with a gap of about 3 eV. The valence and conduction bands have mixed Ti and O character revealing the mixed iono-covalent character of this oxide.2 Titania has a wide range of possible oxygen stoichiometries. Eventually ordered phases such as Ti2O3 or TiO are produced which are no longer insulating.While TiO is metallic,3 Ti2O3 behaves either as a semi-conductor or a metal as a function of the temperature. This is often assigned to a Mott»Hubbard-type transition although its exact nature is still controversial. 4,5 The TiO (110) face is the most stable surface of rutile. It has emerged as one of the most 2 important oxide surfaces which makes it a model system to explore the surface physics and chemistry of transition metal oxides.1,6 There is evidence that a simple charge-neutral truncation is the correct structure for the stoichiometric (1]1) surface although the amount of surface relaxation remains controversial.7 Non-stoichiometry on the other hand leads to reconstructed surface structures for which several models have been recently debated.8 This surface has served as the support for metal deposition and systematic trends as a function of the metal electronegativity have been derived.9 molecule,12 positively charged TinO2n~d ` (n\1»7 d\0»4) clusters have been Unsupported titanium oxide clusters have been considered both theoretically and experimentally.Aside from the many works devoted to the TiO molecule,10,11 and from the experimental study of the TiO2 investigated with mass spectrometry and collision induced dissociation.13 Recently anionic TiOy~ (y\1»3) molecules and (TiO2)n~ (n\1»4) clusters have been produced and studied using anion photoelectron spectroscopy.14 This has permitted the determination of the electron affinities and 285 Faraday Discuss. 1999 114 285»304 Received 19th April 1999 Electronic and structural properties of TiO species of various sizes charges and 2 stoichiometries ranging from TinOm x clusters (n\1»3 m[n\0 1 x\[1 0 ]1) to bulk rutile and its (110) surface have been obtained by total energy calculation based on the density functional theory (DFT) in the local density and local spin density approximations (LSDA) and complemented by a Bader-type analysis of the total electronic density.Attention has been focused on the electron distribution to better understand how the ionocovalent character of the Ti»O bonding and the screening properties vary as a function of the size of the system the atomic coordination and the surface orientation. This journal is( The Royal Society of Chemistry 2000 and Ti excitation gaps of these clusters as a function of their size.From a theoretical point of view TiOy molecules (y\1»2) have been studied using various methods.15 Ab initio Hartree»Fock calculations have been performed on neutral and cationic Ti clusters (n\1»3 d\0»1) in order nO2n~d to determine the equilibrium geometries ionization potentials and vibrational spectra.16 By using an ab initio DFT-LSDA approach we have obtained preliminary results on the geometry and electronic properties of the lowest energy stoichiometric Ti isomers.17 2O4 3O6 An open question related to both systems (free clusters and surfaces) is the extent of charge redistribution when they are either ionised or non-stoichiometric. To de—ect free clusters into a mass spectrometer for example a preliminary ionization of the clusters has to be performed so that it is actually the abundance of positively charged species that is directly probed.Similarly anion spectroscopy is performed on negatively charged species. Whether the excess charge in free clusters is delocalised on all atoms or is mainly trapped on speci–c sites has been addressed for charged and/or non-stoichiometric NaCl,18h20 NaF21,22 and Li2O23 clusters which are very ionic compounds with large gaps. However the general answer to this question is related to the screening properties of the systems under consideration. On surfaces usually no direct charge injection is performed but in the presence of oxygen vacancies or metal deposits charge redistributions take place the amount and spatial extent of which are directly related to the electronic structure of the oxide.For example TiO has recently 2 been examined in relation to alkali deposition. The adsorption characteristics and the charge transfer from K to TiO2(100)24 and from Na K and Cs atoms to TiO (110)25h29 have been studied. From a theoretical point of view both K30 and Na31 adsorption on2 TiO (110) have been 2 considered. The aim of the present paper is to analyse the charge redistribution eÜects in free clusters and surfaces of titanium dioxide to better understand how the ionocovalent character of Ti»O bonding and screening properties of the oxide vary as a function of the size of the system the atomic coordination number and the surface orientation. Section 2 gives the details of the computational method that we use which is based on the DFT.Some emphasis will be given to the choice of the pseudopotentials that account for the interaction between valence and core electrons. We will also describe the charge analysis which is performed subsequently to investigate the electron or hole redistributions. Sections 3 presents our results for the neutral cationic and anionic (TiO lowest energy isomers (n\1»3) and the addition of one oxygen atom on the 2)n neutral clusters. We also give results on the geometry and charge analysis of bulk TiO and the 2 clean (1]1) TiO (110) surface. Sections 4 5 and 6 are devoted to a discussion of our results. They 2 focus on the modi–cations of the degree of covalency of the Ti»O bonds as a function of the diÜerent environments on the localization of the excess charges (holes or electrons) and on screening eÜects respectively.2 Computational method The electronic structure calculations have been performed within the density functional theory by using both local density approximation (LDA) and local spin density approximation (LSDA) for the exchange and correlation energy. In particular LSDA is used for clusters with an odd number of electrons and for some even»numbered clusters that could present a non spin»paired ground state. We have thus checked the stability of the calculated electronic structure with respect to spontaneous spin polarization. The Kohn»Sham orbitals are expanded in a plane»wave basis set and soft norm»conserving pseudopotentials are used to describe the interaction between the ionic cores (consisting of 1s oxygen atomic states 1s 2s and 2p titanium states) and the valence electrons.The interaction between core and valence electrons is accounted for by norm-conserving pseudopotentials in the Kleinman»Bylander form32 including s p and d components for oxygen and titanium. The reference state is 1s22s22p6 for O and Ar 3d4 for Ti. A local reference potential is chosen (d component for oxygen s for titanium). We followed the prescriptions given by Troullier and Martins33 in the generation of pseudopotentials. The core radii chosen for O (Ti) are 1.38 a (a0\Bohr (1.30) a and 1.38 (1.80) a for the s p and d components respectively 0 1.60 (1.40) 0 0 radius). The O pseudopotential was generated for the neutral atom and extensively tested in previous studies.23,34 Faraday Discuss.1999 114 285»304 286 The inclusion of the titanium 3s and 3p states in the valence was found necessary to obtaia good agreement with the experiments and advanced calculations performed on TiO.15 This appeared to be especially important to ensure the transferability of the pseudopotential in systems having sites of various coordination numbers. For bulk rutile titanium oxide it is well known that many pseudopotential types give results in good accordance with respect to the experiments and usually the diÜerences with the latter are not discussed in terms of the pseudopotential characteristics but rather in terms of the approximation used for the exchange-correlation energy.The LDA for example is known to underestimate interatomic distances by 1% or 2% and the GGA (generalized gradient approximation) to overestimate them. To perform a more stringent test than bulk rutile we chose the TiO molecule which is an open shell system of very small size for which experimental data are available.11 The ground state electronic con–guration is 3* and the equilibrium distance and vibrational frequency are equal to 3.06 a and 1000 cm~1 respectively. We 0 have built a –rst pseudopotential (I) including Ti 3s 3p 4s and 3d states in the valence and a second one (II) only including Ti 4s and 3d states in the valence. Both pseudopotentials yield the right symmetry ground state and the discrepancies with experimental values are [3% (]6%) for the interatomic distance and ]9% (]0.1%) for the vibrational frequency for the I(II) pseudopotentials.II is seen to give abnormally high equilibrium distances. These diÜerences between the two pseudopotentials show the active contribution of Ti 3s and 3p electrons in the Ti»O bonding. More precisely we have checked that using pseudopotential I the Ti 3p orbital which points in the bond direction slightly hybridizes with an O 2p orbital resulting in a ppr bonding orbital. Evidence for this hybridization is also supported by the energy splitting between the Ti 3p r and p states which decreases from 2.90 eV to 0.45 eV when increasing the interatomic distance from 2.60 a to 3.20 a0 . As a consequence of the interaction between 0 Ti 3p and O 2p orbitals and the enhanced electron density along the bond the ordering and relative splitting of the electronic states depends sensitively on the inclusion of the Ti 3p states.Ti 3s electrons on the other hand do not form hybrids but are polarized toward oxygen and it was not possible to take account of this polarization without treating these states explicitly. In addition their removal may cause technical problems like the appearance of ghosts states.35 Our results using pseudopotential I compare well with previous CI (con–guration interaction) calculations, 15 regarding the symmetry and the shape of the electronic wavefunction. As a consequence in the following we adopt the pseudopotential I. The logarithmic derivatives show an optimal transferability over a wide energy range well above the atomic vacuum level.Although the number of electron states and the energy cut-oÜ needed to get convergence with respect to the plane-wave basis set are high as a result of the inclusion of the 3s and the 3p Ti electrons in the valence band they participate in the bonding and can modify the computed structural parameters especially for low-coordinated atoms as it is the case for clusters and surfaces. and with A supercell geometry with a face-centered cubic cell of lattice parameter equal to 35 a0 was used for the cluster calculation and integration in the Brillouin zone was performed using the ! point. Convergence on the total energy as a function of the cut-oÜ energy was checked a precision E of 0.1 eV on total energy is reached for cut\60 Ry and of 0.1 eV on total energy diÜerence between isomers at Ecut\40 Ry.The calculations were performed with Ecut\60 Ry for TiO2 and (TiO Ecut\40 Ry for larger clusters. 2)2 The charged clusters are simulated by introducing a compensating uniform background which takes into account the –nite contribution of the background»background interaction to the long range part of the electrostatic energy.36 This method allows a faster convergence of the total energy as a function of the size of the unit cell as con–rmed by extensive tests on charged molecules. We checked that the size of the unit cell is sufficient to ensure that spurious interactions between neighboring periodic images do not bias the computed energy diÜerences between the various isomers and the calculated interatomic distances.In order to –nd out the ground state con–gurations we started from structures suggested in ref. 13 by simple pair-potential simulations. The geometries were then relaxed until the atomic forces did not exceed 0.01 eV Aé ~1 and the calculated electronic structures thus refer to the stable con- –gurations optimized in a fully self»consistent way. 2O2)»O The (110) unreconstructed face of rutile is simulated by a slab containing three O»(Ti units consisting of 18 atoms in total (see Fig. 4 below) in the two-dimensional unit cell plus a vacuum whose width (B5.5 ”) is enough to prevent spurious interactions between periodic images. 287 Faraday Discuss. 1999 114 285»304 Two k ñ points in the irreducible two»dimensional Brillouin zone are used for the charge integration.A full geometry optimization of all TiO layers is performed in order to determine the 2 equilibrium geometry of each surface con–guration. The search for the energy minimum is stopped when atomic forces are less than 0.01 eV Aé ~1. Our charge analysis follows the scheme proposed by Bader,37 who introduced the concept of atoms-in-molecules or solids. According to Bader an atom is seen as an open system that can be described by a Schroé dinger equation. Consequently its volume must be enclosed by a surface through which no electron —ux passes. The mathematical condition which de–nes the partitioning of space into atomic basins is thus (2.1) +ä o(rñ ) … nñ \0 in which o r n (rñ ) is the total electron density and ñ the normal to the surface at ñ .Usually and more speci–cally in ionic systems each atomic basin includes one nucleus. In the course of our study we got no evidence that this condition breaks. Bader analysis includes a topological discussion of the electronic density in terms of critical points that we will not use in the present work. We will only focus on the determination of the atomic basins and the calculation of the integral of the total electronic density within each basin. i(j) on which o(Rä i(j)) decreases monotonically. The space is therefore partitioned in From a numerical point of view in order to obtain the total electronic density o(rñ ) we sum the valence density issued from our DFT calculation and the core level density obtained from the pseudopotential generation code.o(rñ ) is written on a regular grid and is integrated in each atomic basin as follows. Starting from a grid point close to a given nucleus j the point Rä 0(j) of the mesh corresponding to the density maximum is determined. Then a simple algorithm determines the mesh points Rä domains D(j) consisting of all paths on which the sign of +ä o(Rä i l (j)) … lñ i is constant where ñ i is the elementary displacement along the paths. The electron density is then summed up within each domain and the contributions from the border regions are shared between the neighbouring basins. By using this method on a cubic grid with a parameter of 0.15 a0 we obtain the charge integral with a precision better than 0.1 electron per basin. (3.1) 3 Charge distribution in titanium oxide results 3.1 Neutral (TiO2)n clusters (n= 1ñ3) C and S symmetry.We have previously described17 the geometry and energetics of the low energy isomers issued from the geometry optimization for n\2 and 3. They are represented in Fig. 1. For (TiO2)2 the C C two con–gurations have C and symmetry respectively. The isomer is non-planar as a 2v 3v 2v result of the bent con–guration of the TiO trimer. For (TiO2)3 we have found two low energy 2 isomers of s 4 Integration of the electron density in the regions D(j) associated with each atom j yields the Bader charges borne by j in the clusters. We –nd titanium charges in the range []1.5 ]2] and the oxygen charge in the range [[0.7 [1.1]. These values are far from the formal charges ]4 and [2 assigned in the fully ionic limit to Ti and O in titanium oxides of TiO stoichiometry.2 This is an indication that the Ti»O bonds have a large part of covalent character in agreement with resonant photoemission results.2 To quantify the relative part of ionic and covalent characters we do not directly compare the charge values which are the complex result of O»Ti electron transfer through orbital hybridization and local environment factors such as coordination numbers. In order to discriminate the diÜerent eÜects we analyze the charges in the following way. Starting from the purely ionic limit in which Ti and O have ]4 and [2 charges we take into account the eÜect of Ti»O orbital hybridization between –rst neighbour atoms by introducing parameters D which describe electron transfers per bond.For example in the TiO molecule by symmetry a single D is required 2 and the atom-in-molecule charges are equal to Q QTi\]4[2D O\[2]D According to this picture each O2~ atom donates D electrons back to the Ti4`. In this simple case the orbital hybridization is included in D and the coefficient in front of D in eqn. (3.1) is the Faraday Discuss. 1999 114 285»304 288 Fig. 1 Low energy isomers of (TiO clusters (n\1»3). Ti atoms are drawn in pale grey and O atoms in dark grey in all ball and stick representations. The inequivalent electron transfers per sites are labelled accord- 2)n ing to their values in Table 1. In charged and non-stoichiometric clusters are indicated the values of the excess electron density (positive sign) or excess hole density (negative sign) on the atoms on which the electron or hole(s) are localized (see text).For the sake of clearness we have indicated the excess electron density on the three bridging oxygens in Ti2O4 `-C3v ` and Ti2O5-C only once although these oxygens are equivalent. For example in Ti2O4 -C3v there exist three inequivalent electron transfers D that we label D atom coordination number. We have shown38,39 that in a tight-binding approach D is a monotb/( e onic function of the ratio C[eA) between the resonance integral b associated to the orbital hybridization and the diÜerence in energy between the cation and anion atomic orbitals under consideration. D vanishes in the ionic limit (either b\0 or eC[eA ]O). Its magnitude thus characterizes the strength of covalent bonding.When inequivalent bonds are present in a system distinct parameters D are required and the charge of an atom-in-molecule can be obtained from j the values of the electron transfers from or to its Z –rst neighbours. 1 and D (see Fig. 1). The atoms bear charges equal to 3 D2 QTi\4[3D3 QTi\4[D1[3D2 (3.2) QO\[2]D2]D3 QO\[2]D1 289 Faraday Discuss. 1999 114 285»304 Fig. 1.»Continued for 4-fold and 3-fold coordinated Ti respectively (–rst line) and for 1-fold and 2-fold coordinated oxygens respectively (second line). More generally we have used the relations (3.3) QO\[2]; Dj j QTi\4[; Dj j with the sum index running over bonds in which the atom is involved. This has allowed us to derive the values of the D parameters in the –ve neutral (TiO clusters (n\1»3) that are col- 2)n lected in Table 1 together with the computed corresponding bond lengths.The D parameters and consequently the degree of covalency of the Ti»O bonds vary in large i proportions ranging from 0.2 to 1.26 electrons. The same is true for the interatomic distances which vary from 3.0 to 3.9 a0 . Inequivalent local environments of the atoms thus induce strong changes in the mixed ionocovalent character of the Ti»O bonds. This is akin to rather covalent systems such as TiO2 and would not have been found in very ionic clusters. clusters (n= 1ñ3) 3.2 Anionic (TiO2)n ó and cationic (TiO2)n ë The addition or removal of an electron does not modify the cluster conformations deeply.Bond lengths and bond angles vary slightly while the electronic structure changes dramatically. Faraday Discuss. 1999 114 285»304 290 Table 1 Electron transfers per bonds D and interatomic distances (in atomic units) in neutral anionic cationic (TiO2)n and in non-stoichiometric Ti clusters (n\1»3) in bulk TiO rutile and at the (110) nO2n`1 2 surface. The charges borne by the atoms may be derived from the Dj using eqns. (3.3) for neutral stoichiometric clusters eqn. (3.4) for anionic clusters and eqn. (3.5) for cationic or non-stoichiometric clusters. 3-Cs n\1 2 ` 2~ TiO TiO 3-Cs TiO TiO Peroxo-TiO D1(d1) 1.24 (3.076) .94 (3.134) 1.16 (3.083) 0.78 (3.25) 0.54 (3.380) 2 D2(d2) n\2 Ti 2O6-Cs 1.23 (3.106) Ti2O4~ 2O4-C2v 2O4 ` Ti Ti2O5-Cs Peroxo-Ti DD D1(d1) 1.24 (3.045) 1.03 (3.101) 0.98 (3.07) 0.66 (3.30) 0.48 (3.385) 2(d2) 0.52 (3.426) 0.46 (3.461) 0.61 (3.413) 0.46 (3.51) 0.55 (3.404) 3(d3) 0.68 (3.35) 0.52 (3.416) D4(d4) 1.0 (3.16) 1.22 (3.098) Ti2O4~ 2O4-C3v Ti2O4 ` Ti2O5-C n\2 Ti D DD 1(d1) 1.07 (3.075) 1.02 (3.133) 1.32 (3.00) 0.85 (3.247) 2(d2) 0.29 (3.692) 0.37 (3.666) 0.22 (3.795) 0.44 (3.505) 3(d3) 0.90 (3.307) 0.65 (3.398) 0.87 (3.306) 3O7-C Peroxo-Ti3O7-Cs 1.26 (3.092) 0.21 (3.712) 0.34 (3.610) 0.84 (3.27) 0.67 (3.313) 0.15 (4.041) 0.17 (3.844) Ti 3O7-Cs n\3 T 3O6 ` i3 O6-Cs Ti3O6~ Ti Ti3O7-C3v Peroxo-Ti D1(d1) 1.26 (3.11) 1.18 (3.145) 1.0 (3.175) 0.76 (3.251) 0.54 (3.38) D2(d2) 0.44 (3.434) 0.40 (3.456) 0.54 (3.416) 0.57 (3.424) D3(d3) 0.31 (3.639) 0.51 (3.541) 0.40 (3.552) D4(d4) 0.86 (3.264) 0.53 (3.330) 0.84 (3.273) D5(d5) 0.63 (3.324) 0.42 (3.401) 0.54 (3.373) 0.33 (3.615) D6(d6) 0.2 (3.932) 0.22 (3.843) 0.25 (3.808) D7(d7) n\3 Ti 3O6 ` 3O6-S4 Ti3O6~ Ti3O7-Cs Peroxo-Ti 0.53 (3.416) 0.57 (3.386) 0.55 (3.402) 0.50 (3.419) 1.24 (3.095) DD D1(d1) 1.24 (3.107) 1.06 (3.144) 0.92 (3.145) 0.62 (3.31) 0.48 (3.382) 2(d2) 0.50 (3.42) 0.42 (3.45) 0.65 (3.37) 0.46 (3.493) 0.53 (3.41) 3(d3) 0.53 (3.405) 0.55 (3.405) 0.52 (3.40) 0.60 (3.39) 0.53 (3.421) DD 4(d4) 5(d5) D6(d6) 1.1 (3.115) n\O (110) Surface 0.17 (3.826) 0.42 (3.616) 0.28 (3.888) 0.47 (3.405) 0.43 (3.585) 0.29 (3.669) Bulk D DD 1(d1) 0.35 (3.668) 0.54 (3.399) 2(d2) D DD 3(d3) 21(d21) 31(d31) 4(d4) D41(d41) DD 5(d5) 0.29 (3.737) 51(d51) 0.46 (3.577) clusters the –lled electronic states are mainly oxygen-derived In neutral stoichiometric (TiO2)n with a non-negligible contribution from titanium orbitals.For the sake of conciseness we refer to them as valence band states which is formally correct only for extended systems. Symmetrically in the neutral clusters the lowest empty states (conduction band states in the following) have mainly Ti character. In anionic clusters this state (HOMO) becomes half-–lled. Direct visualization of the HOMO shows the localization of the additional electron. A typical picture relevant for Ti3O6~-S4 is shown in Fig. 2 where we can see that the probability of –nding the electron is equally shared on the two titaniums equivalent by symmetry.More generally we have found that the HOMO wavefunction is highly localized on one or a few titaniums with a d-orbital-like shape consistently with a Ti non-bonding character for the HOMO. When inequivalent titaniums are present the electron is generally localized on the titanium(s) with the lowest coordination number ( and i.e. the three-fold coordinated Ti in Ti2O4 -C3v Ti3 O6 -Cs and the two three-fold coordinated Ti in Ti3O6-S4). The distribution of the added electron Nie can be unambiguously determined for all the anionic clusters and is given in Fig. 1. Due to the perturbation brought in by the negative charge all electron transfers along the Ti»O bonds are modi–ed and we have calculated them from the following expression (3.4) Q Q O\[2]; Dj j Ti\4[Ne[; Di i 291 Faraday Discuss.1999 114 285»304 Fig. 2 Highest occupied orbital in Ti3O6~-S4~. Ti and O atoms are represented as small light grey atoms and in black respectively. with Ne the excess electron number on the Ti site under consideration. The resulting values of the electron transfers per bonds are written in Table 1. As a general trend D decreases in the vicinity of the perturbed sites and the bonds are elongated. In cationic clusters the removal of an electron is associated with the creation of a hole in the valence band. Due to electron»electron interactions the valence band states are strongly rearranged and the hole wave function cannot be simply expanded in one or few wavefunctions of the neutral cluster.However the hole is fully spin polarized so that the polarization density f(rñ )\oè(rñ )[oé(rñ ) gives a good hint of the hole localization. Interestingly the sign of f(rñ ) agrees almost everywhere with that of the net spin showing that the hole creation aÜects the total electronic distribution nearly independently of the spin. As a general trend (see Fig. 1) we –nd that the hole is shared between the two terminal oxygens (one-fold coordinated oxygens) in all clusters 2O4-C3v. We de–ne positive parameters except Ti Nih equal to the hole densities on atom i. We have written their values in Fig. 1 with a ì[œ sign to emphasize the decrease in electron numbers. Due to the perturbation that results from the positive charge of the cluster all electron transfers along the Ti»O bonds are modi–ed.We calculate them through (3.5) Q QO\[2]Nh]; Dj j Ti\4[; Di i where Nh is the hole concentration on the O site under consideration. The charge transfers D are collected in Table 1. As a general trend D decreases in the vicinity of perturbed oxygen sites and the corresponding bond lengths increase. clusters 3.3 Non-stoichiometric TinO2në1 Starting from the –ve stoichiometric clusters described above we have considered the addition of one oxygen atom. In each case several distinct isomers are formed. We have represented in Fig. 1 those of lowest energy with their space groups. In most cases the additional oxygen binds to a Fig. 3 Surfaces of equal spin polarisation in Ti2O5-C.Ti and O atoms are represented as small light grey atoms and in black respectively. Faraday Discuss. 1999 114 285»304 292 single titanium atom the only exception being the peroxide cluster Ti3O7 -Cs . The conformation of the clusters is not deeply modi–ed by the oxygen addition but as we will see below important modi–cations of bond lengths take place. The electronic structure of the non-stoichiometric clusters is characterized by a valence band with two holes. As in the case of cationic stoichiometric clusters we have determined the localization of these holes by studying the spin density distributions. The case of Ti2O5-C is represented in Fig. 3. In all cases the hole localization is very similar to what it is in cationic stoichiometric clusters i.e.essentially on the terminal oxygens. Ti2O5-C is slightly more complex due to a sharing of the hole density on the –ve oxygens of the clusters. We have indicated in Fig. 1 the hole distribution on the diÜerent atoms. In most cases and especially in complex cases such as Ti2O5- C the estimation of the N results from a procedure in two steps with a –rst rough determination ih followed by a re–nement as commented in Section 4. As for the cationic clusters we have derived the modi–ed values of the charge transfer per bonds using eqns. (3.5) and we have collected them in Table 1. Actually two families of non-stoichiometric clusters may be discriminated according to whether an O»O bond forms or not. When the latter case occurs an energy stabilization takes place and the isomers with such O»O bond have the lowest energy.These O»O bonds are known in the condensed phase or in oxidized metallic clusters.40,41 When the O group bears a [1 2 charge it is named superoxide anion and when its charge is [2 peroxide anion. The O»O bond lengths are diÜerent in the two cases A A d being of the order of 1.34 é and 1.49 é for superoxides and peroxides respectively. For the –ve isomers which present such an O group we –nd that 1.483\ 2 d/”\1.498 which unambiguously points towards the existence of an O22~ peroxide group. The hole distribution reported in Fig. 1 supports this conclusion since the two holes are entirely localized on the O group. In Section 5 the charge distribution in these clusters will be further 2 discussed. 3.4 TiO bulk and the (110) surface 2 Bulk TiO rutile is tetragonal and may be characterized by lattice constants a and c and by an 2 internal parameter u which governs the location of the oxygen atoms.We have computed the structural properties of rutile by allowing a full atomic relaxation. The calculations are performed at a 60 Ry cut-oÜ and are fully converged with respect to the Brillouin zone sampling. The results are compared to the experimental data and to previous calculations in Table 2. One can note that (i) our results well agree with the experimental data,42 from which they diÜer by a slight underestimate of less than 1% of the lattice parameters a and c. This is usual when using the LDA. (ii) Our results agree almost perfectly with those obtained by Ramamoorthy and coworkers,43 who also treated the semicore Ti 3s and 3p electrons self-consistently.On the other hand the lattice parameters obtained by Glassford and Chelikowsky,44 who kept the semi-core Ti electrons frozen are overestimated. We attribute this disagreement mainly to the diÜerent treatment of the 3s and the 3p Ti electrons. As regards the (110) surface there is a longstanding controversy about its surface conformation. On one hand grazing X»ray diÜraction experiments7 give a large inward relaxation for the bridging O which was then seen to be consistent with shadowing eÜects in the diÜraction of O` ions.45 On the other hand most of the theoretical calculations agree with the experimental positions for the Ti ions while those computed for the oxygens show quantitative discrepancies.For these Table 2 Equilibrium lattice parameters of bulk rutile Theory Ref. 43 This work Ref. 44 Experiment Ref. 42 a/” c u /” 4.556 2.930 0.306 4.567 2.932 0.307 4.653 2.965 0.305 4.5936 2.9587 0.3048 293 Faraday Discuss. 1999 114 285»304 Table 3 Ionic displacements (in ”) of the surface atoms at the (110) surface of rutile with respect to the bulk-truncated surface This work Ref. 43 Ref. 7 [110] Theoretical [110] [110] [110] [110] Experimental [110] Experimental lattice parameter [110] Theoretical lattice parameter [110] 6-fold Ti 5-fold Ti Bridging O In plane O 0 0 0] 0.04 0 0 0] 0.06 0 0 0 0.16^0.05 ]0.13 [0.17 [0.06 ]0.13 ]0.11 [0.15 [0.12 ]0.12 ]0.12 [0.15 [0.09 ]0.14 ]0.12^0.05 [0.16^0.05 [0.27^0.08 ]0.05^0.05 0 0 0] 0.05 reasons we carried out a careful relaxation of our slab by using both theoretical and experimental lattice constants a and c in order to assess whether the theoretical results could be biased by the various computational ingredients.The atomic positions were relaxed until the atomic forces do not exceed B5]10~3 eV ”~1. Our –nal results are in line with previous LDA calculations. In Table 3 our computed ionic displacements are collected and compared to those obtained by Ramamoorthy and coworkers,43 and to those inferred by X»ray surface diÜraction.7 One can note that a substantial agreement exists between the calculations.Our calculation performed at the experimental lattice parameters is slightly closer to the experimental result though the gap between the latter and our numerical values is still signi–cant. For both geometries bulk and surface we have applied the same charge analysis as in clusters to determine the electron numbers for each atom-in-molecule. The electron transfers per bond Di for each class i of equivalent bonds are obtained using the same formalism as in the neutral stoichiometric clusters (eqn. (3.3)). In the bulk the Ti atoms have six neighbouring oxygens placed at two inequivalent distances so that only the average of D can be determined. As regards the (110) surface Fig. 4 indicates the labels used to specify the inequivalent bonds.We have reported the values deduced for the D in the lower part of Table 1. An enhancement of covalency is i evidenced on some bonds and more speci–cally on the bond which links the bridging oxygens to the underlying titaniums. To summarize in the various systems that we have considered the charge —uctuations are rather weak and do not exceed 0.3 electrons. The only exception is Ti2O4 -C3v in which the charge diÜerence between the two inequivalent titaniums is 0.75 electrons. On average the Ti charge is close to ]1.7 and the oxygen charge is close to [0.85. The rough constancy of these values results from a partial cancellation between the large variations of the electron transfers per bonds Fig. 4 Electron transfers per bonds on the (110) surface of TiO2 .Ti and O atoms are represented in light grey and dark grey respectively. Faraday Discuss. 1999 114 285»304 294 and the large variations in the coordination numbers of the atoms. 4 Iono covalent character of the TiñO bonds The electron transfer per bond D varies over a wide range as a function of the type of the system (clusters (110) surface bulk) of the net charge of the stoichiometry and of the speci–c conformation. This shows that the Ti»O bond in some cases has a large part of covalent character (for example in TiO2) while in some other cases it is much more ionic (for example one of the central bonds in Ti3O6-Cs). In this section we point out the factors that drive these strong variations of D. Morever we show that there is a relation between the electron transfers per bond and the bond lengths valid for all the systems under consideration which connects a theoretical concept the mixed ionocovalent character of a bond quanti–ed by the non-measurable quantities D with the i experimentally accessible structural parameters di .4.1 Driving factors underlying electron transfers As shown in refs. 38 and 39 in a tight binding picture the electron transfer per bond D is a b/(e monotonic function of the ratio C[eA) between the resonance integral b associated with the orbital hybridization and the energy diÜerence between the cation and anion atomic orbitals under consideration. b is strongly dependent upon the interatomic distances. The atom eÜective levels e and e are renormalized with respect to their values in the neutral atoms by Coulomb C A and exchange interactions due to the surrounding electrons.eC[eA . Compact environments are expected to induce large MadeeC[ eA values and thus small electron transfers per bond D. We have dV Mad (i) for the actual geometry by q for all anions related by qTi]2qO\0.46 Coulomb eÜects show up in the Madelung potential exerted by the neighbouring charges on a given atom. Since in most compounds anions are surrounded by cations and vice-versa the Madelung potential is positive on an anion and negative on a cation which shifts their eÜective levels towards lower and higher energies respectively. For a given pair of Ti and O atoms connected by the ith bond the diÜerence dV Mad (i) between the Madelung potential acting on Ti and O renormalizes the diÜerence lung potentials i.e.large checked this idea in the neutral stoichiometric clusters and shown the correlation between D and Mad (i) in Fig. 5. Actually the Madelung potential results from a self-consistent relation between dV i the charges and the total potentials. In order to point out the role of the morphology on the renormalization of the atomic levels we have computed the using a constant charge q for all cations and The Ti O current value for q has no importance in the following discussion apart from an overall scaling in Ti Fig. 5. We have also added the values calculated in bulk rutile which result from Madelung potential data given in ref. 47 correctly renormalized by the qTi charge value and the values that we have calculated for the (110) surface.Considering the large range of morphologies of the systems under consideration from the TiO molecule in which Ti is 2-fold coordinated to the bulk in which the 2 local symmetry is nearly octahedral Fig. 5 clearly demonstrates that our computed dV Mad (i) and thus the cluster morphologies are the relevant parameters that drive the values D of the electron i transfers per bond. 4.2 Correlation between electron transfers and bond lengths In addition the Madelung potential also depends upon the bond lengths d and so do the resonance integrals b which enter the expression for D in a simple tight binding approach. Again the relationship between the D and the d values is self-consistent. Strong hybridizations (associated with large D) are expected to yield high bond energies and thus small interatomic distances.To check this correlation we report in Fig. 6 all the D and d values collected in Table 1 for the neutral or charged stoichiometric clusters as well as for the non-stoichiometric ones the bulk and the surface. b/(eC[eA) which The trend is a rapid decrease of D as a function of d. If we consider the ratio drives the value of D both its numerator and its denominator decrease when the bond lengths are Faraday Discuss. 1999 114 285»304 295 2 on the Ti and O in the neutral stoichiometric clusters (–lled circles) in bulk rutile TiO (plus) and on the Fig. 5 Electron transfer per bond D as a function of the diÜerences dVMad in the Madelung potentials acting TiO (110) surface (diamonds).2 elongated. We can conclude that the variations of the resonance integrals b are stronger since as a whole D follows the trend given by b. In addition the dispersion of the points in Fig. 6 is very small considering the large diÜerences among the systems that were investigated. We consider that this excellent correlation between D Fig. 6 Electron transfer per bond D as a function of the Ti»O bond length d in neutral (–lled circles) anionic (down triangles) and cationic (up triangles) stoichiometric clusters in non-stoichiometric clusters (stars) in bulk rutile TiO (plus) and at the (110) surface (diamonds). 2 Faraday Discuss. 1999 114 285»304 296 and d whatever the morphology and the size of the system is a strong indication that the D parameters although they are not measurable quantities are very relevant quantities to discuss the nature of the anion»cation bonding especially in these systems which at the same time are highly inhomogeneous and present a mixed ionocovalent character.As a matter of fact we have used this correlation in several instances. In Ti3O6 -Cs clusters in Nih T in complex cases. This happened fori which an undetermination exists in the linear system of equations which gives the D values by i using Fig. 6 we have assumed a reasonable value of D on one bond (usually the longest one) and deduced all other electron transfers per bonds from the equations. We have also used Fig. 6 to re–ne the values for the excess hole distribution 2O5-C where the holes are shared between two types of oxygens (Fig.3) in Ti2O5 -Cs where the same occurs although to a lesser extent and to estimate the diÜerence of hole concentrations in nonstoichiometric clusters with O22~ groups. However among the 92 values reported in Table 1 and Fig. 6 85 are unambiguously determined. To summarize the charge analysis in the titanium oxide –nite and in–nite systems has allowed us to study the bonding characteristics as a function of the size the local environment and the atom coordination number. We have found that the degree of covalency of the Ti»O bonds quanti–ed by the values of the electron transfers per bond * strongly varies as the local environ- i ment of the atoms changes and is driven by the value of the Madelung potentials acting on the atoms.Electron transfers per bond and bond lengths are monotonic functions of each other through a ìì universal œœ relationship valid from the molecule to the bulk whatever the charge state and the stoichiometry. This conclusion supports the use of this universal curve as a predictive tool for other isomers or geometries. 5 Excess charge localization in TinO2n and TinO2në1 clusters The localization of excess charges in free clusters has been considered in alkali-halides NaCl18h20 and NaF21,22 and Li2O23 systems. In most cases these excess charges arose from an excess of metal atoms. The authors found that the excess electrons are very delocalized in the clusters and that excess metal atoms induce a kind of metallization of the clusters.However for some special geometries and stoichiometries namely cuboid Nan Fn~122 with a single anion vacancy the electrons remain trapped at the vacancy site. This also happens when a neutral oxygen vacancy is created in the bulk or at the surface of MgO.34,48 The excess electron in anionic clusters has a quasi-atomic d-type wave function (see Fig. 2) which is highly localized on titaniums and presents a non-bonding Ti»O character as exempli–ed in Fig. 2. Such Ti»O non-bonding states of Ti character are usually found to form the bottom of the conduction band in defective systems. When inequivalent titaniums are present the excess electron rests on the titanium(s) with the lowest coordination number which have the weakest Madelung potential. Symmetrically in cationic or non-stoichiometric clusters the hole wave function has a quasiatomic p-type shape and is localized on oxygens.It is a Ti»O non-bonding wave function of O character as expected at the top of the valence band in systems where the oxygen atoms have lost some of their cation neighbours. In most clusters when inequivalent oxygens are present the hole rests on the oxygen(s) with the lowest coordination number or more precisely on the oxygen(s) which experience the smallest Madelung potential. Ti2O4 `-C3v and Ti2O5-C seem not to obey this rule. As proved by a direct visualization the HOMO consists of O»O hybridized 2p orbitals with a non-bonding Ti»O character. This is not due to the existence of especially short O»O bonds but rather to the presence of three atoms which take part in the hybridization.The O»O antibonding state is thus higher in energy than the non-bonding 2p orbital of the terminal oxygen. The hole is therefore shared between the bridging oxygens with an additional weight on the terminal oxygens in the case of Ti2O5-C. This eÜect is visible in the clusters with the C geometry and not in the other clusters because their two Ti are 3v bridged by three oxygens. 2 In the non-stoichiometric clusters with an O group it is found that the two holes are completely shared between the two oxygens of the2O group with no noticeable weight on other oxygen atoms. This supports the prediction of the existence of a peroxide O22~ group. We will see in the next section that the actual charge on the two oxygens is reduced from the [2 value by Faraday Discuss.1999 114 285»304 297 2O4-C2v . We assign this eÜect to the enhanced electroscreening eÜects. The short-hand notation O22~ has thus to be understood in this framework as indicative of the hole localization. The comparison between non-stoichiometric clusters of similar conformation which either present or do not present a peroxide group shows that the proximity of the two oxygens in the O group enhances the hole localization on the group. This is especially 2 clear on the Ti2O5 clusters issued from Ti static interaction between the two oxygens of the O group that raises their eÜective levels and to 2 the strong hybridization eÜects which raise the O»O antibonding states. (6.1) Ni\Ni0]Niext]Niscr i (6.2) dNiscr\^; (Dj[Dj0) j 6 Screening eÜects When atoms molecules or solids are submitted to an electrostatic perturbation the ground state electron distribution changes to screen the perturbation lowering the energy of the system.The fruitful framework to understand and describe these eÜects is the theory of the relative permittivity. In the homogeneous electron gas the basic model for simple metals the perturbing potential is totally screened at large distance. The characteristic screening length is proportional to the inverse of the Thomas»Fermi wave vector. In semiconductors or insulators the existence of a gap in the spectrum of electronic excitations and the strong inhomogeneities of the electron density make the description of screening more involved.It is known that the degree of screening is incomplete and related to the value of the optical relative permittivity e=. In addition local –eld eÜects are important which are a manifestation of the discrete atomic structure of the material.49 For example screening of a test charge located close to an anion is diÜerent from what takes place close to a cation. Also in small –nite systems such as molecules or clusters there exists an eÜect called anti-screening.50 In the case of a positive test charge for example the electron cloud moves towards the site of the perturbation so that there are regions especially those far from the perturbation where the perturbing potential can be reinforced. 6.1 Strength and characteristics of screening eÜects The charge analysis performed in the titanium oxide clusters that we have considered allows us to get some insight into the strength and characteristics of screening eÜects.In analogy with the theory of electrostatic screening for charged stoichiometric clusters we write the electron number on a given atom i as the sum of the electron number i0 modi–cation which includes the charge perturbation N in the neutral cluster (the ìì unperturbed charge œœ) plus a Next and the contribution from screening Nscr. The charge perturbation is related to the excess number of electron N or holes N that we ie ih * in the neutral ones) as follows i have determined previously. By using eqns. (3.4) (3.5) and (6.1) one can relate the screening charge to the variations Dj[Dj0 of the electron transfers induced by the perturbation (Dj in the charged clusters and j0 The sum runs over the bonds that link atom i to its neighbours and the ^ sign refers to cations (]) and anions (-).The screening charge thus appears as due to the modi–cation of the electron transfers per bonds. Moreover the sign of Dj[Dj0 indicates the direction of displacements of electrons to screen the perturbation. A positive (negative) value is associated to a displacement from (to) the oxygens toward (from) the titaniums. The largest variations Dj[Dj0 are of the order of 0.3 electrons. These large values are consistent with the breakdown of Koopmanœs theorem as discussed in ref. 17. The general trend (see Fig. 7) is a displacement of electrons away from a perturbed site which bears an excess of electrons and toward a site with an appreciable hole density with a magnitude which decreases as a function of the distance from the perturbation.In non-stoichiometric clusters the description of screening eÜects is slightly more involved. The oxygen addition is considered as the perturbation together with the formation of new Ti»O bonds. One thus has to compare the –nal non-stoichiometric cluster to an initial state which Faraday Discuss. 1999 114 285»304 298 Fig. 7 Screening eÜects in charged (TiO clusters (n\1»3). Top panel addition of an electron to form anionic clusters ; lower panel subtraction of an electron to form cationic clusters. The arrows indicate the 2)n B electron displacements to screen the perturbation.One two or three arrows are drawn as a rough indication of the magnitude of the electron transfers. consists of a neutral stoichiometric cluster plus an oxygen atom far away. At variance with charged stoichiometric clusters the perturbing charge Next is thus ](2[Nh) and along the newly i * formed bond(s) j0\0. Screening eÜects in non-stoichiometric clusters present the same characteristics as in charged stoichiometric clusters. 6.2 Mechanism of screening eC[eA of the neighbouring anion and cation. Similarly in charged or nonj[ Dj0 of the electron transfers eC[eA induced by the perturbation. When eC[eA increases the The screening mechanism6 can be understood on the basis of the model already used in Section 4 to rationalize the magnitude of the electron transfers in terms of the energy diÜerences between the eÜective levels stoichiometric clusters we can understand the modi–cations D through the modi–cations of bond is more ionic and electrons are backdonated by Ti to O.When eC[eA decreases the bond gets more covalent and electrons are transferred from O to Ti. j1 An excess (depletion) of electrons raises (lowers) all the energy levels with a magnitude which is a decreasing function of the distance from the perturbed site i0 . As depicted schematically in Fig. 8 whatever the sign of the perturbing charge the bonds j involving i become more ionic (D 1 0 decreases). Further away the energy level diÜerences (DE in Fig. 8) vary with alternating signs j which results in oscillating variations Dj[Dj0 of the electron transfers associated with oscillating variations of bond lengths (Fig.6). 299 Faraday Discuss. 1999 114 285»304 Fig. 8 Modi–cation of the positions of the eÜective atomic levels (dashed lines positions in the neutral cluster plain line positions in the charged clusters) as a function of the distance from the site where the perturbation is localised. Top panel addition of an electron on a Ti ; lower panel subtraction of an electron to an O. The arrows indicate the direction of the electron displacements to screen the perturbation. These oscillating variations of Dj[Dj0 allow a cooperative displacement of the electrons to screen the perturbation. This mechanism becomes less and less eÜective as the diÜerence eC[eA (eC[eA ]O) the charges increases i.e.as the ionicity grows. In the limit of perfect ionic bonding coincide with their formal values and no screening takes place. In TiO species that we have Table 4 Ratios X\oNscr/Next o between the screening charge and the charge perturbation on the i site on which it is localized in i Ti n\1 2~ TiO 0.60 n\2 Ti 0.68 2O4~-C2v n\2 Ti 0.75 2O4~-C3v n\3 Ti 0.88 3O6~-Cs n\3 Ti 0.71 3O6~-S4 a The perturbed sites are Ti when x\[1 and O when x\1 or m[n\1. In non-stoichiometric clusters screening on oxygens belonging to the stoichiometric clusters is written on the –rst line (in the order from left to right in Fig. 1) and screening on the added oxygen is written on the second line. Faraday Discuss. 1999 114 285»304 300 nOm x clustersa 3-Cs 2 ` TiO 0.17 TiO 0.70 0.59 0.53 Ti2O4 `-C2v Ti2O5-Cs 0.70 0.70 0.56 0.27 Ti2O4 `-C3v Ti2O5-C 0.79 0.52 0.54 Ti3O7-C3v 0.76 0.52 Ti3O6 `-Cs 0.64 Ti3O6 `-S4 0.57 Ti3O7-Cs 0.67 0.79 0.58 2 Peroxo-TiO3-Cs 0.70 2O5-Cs 0.54 Peroxo-Ti 0.76 0.48 3O7-Cs 3O7-C Peroxo-Ti 0.49 0.40 Peroxo-Ti 0.64 0.48 Peroxo-Ti3O7-Cs 0.76 0.48 considered the ionic charges are far from their formal values as can be expected from a compound with a narrow gap and screening eÜects are important.6.3 Local –eld eÜects In Table 4 we report the values of the ratios X\oNscr/Next o between the screening charge and i i the charge perturbation on the site i on which it is localized.This ratio is an indication of the strength of screening eÜects on inequivalent perturbed sites for the various clusters under study. It is independent of the magnitude of the perturbation in a linear response theory framework. The results given in Table 4 raise several issues. 1. X is in the range 0.17»0.88 and shows large variations as a function of the nature of the perturbed site (type coordination number etc). This is the indication of a strongly inhomogeneous screening consistently with the cluster morphologies. It reveals the importance of local –eld eÜects. 2 ` Ti2O4 `-C2v and Ti3O6 `- 2. For cationic stoichiometric clusters of increasing sizes (i.e. TiO S4 in the order) the hole is shared between the two terminal oxygens. The bigger the distance between them the larger X.As shown in the Appendix this result is due to a kind of destructive interference eÜect between the hole perturbing potentials on the atomic eÜective levels. 3. As far as holes are concerned screening is much more efficient in neutral non-stoichiometric clusters than in cationic stoichiometric ones. This may be rationalized by Madelung –eld arguments which are explained in the Appendix. In most non-stoichiometric clusters X is close to 0.75 on the regular oxygen sites which is indicative of strong screening. 7 Conclusions Electronic and structural properties of TiO species of various sizes charges and stoichiometries TinOm x clusters (n\1»3 m[n\0,1 x\[1 0 ]1) bulk rutile and its (110) surface including 2 have been obtained by total energy calculation based on the density functional theory in the local density and local spin density approximations.Using a Bader-type analysis of the total electronic density we have focused our attention on the electron distribution to better understand how the ionocovalent character of the Ti»O bonding and screening properties vary as a function of the size of the system the atomic coordination numbers and the surface orientation. We have found that the mixed ionocovalent character of the Ti»O bonds varies dramatically according to the local environment of the atoms and can be quanti–ed by a parameter D which represents an electron transfer per bond due to orbital hybridization. We found a unique correlation between the values of D as issued from the charge analysis and the bond lengths for all the systems considered here.Moreover we have rationalized the variations in D from bond to bond through arguments based on the values of the Madelung potentials on the diÜerent atoms. On charged clusters we have found that the excess charge is localized in a quasi-atomic state (d-type orbital on Ti and p-type on O) on the atom(s) with the lowest Madelung potential. Nonstoichiometric neutral clusters with an excess O have two holes in the valence band. As regards the hole localization a striking analogy is found with cationic stoichiometric clusters. It is worth emphasizing the existence of low energy isomers with a short O»O bond. The bond lengths which are close to 1.49 Aé whatever the cluster size and our charge analysis point towards the presence of peroxide O22~ groups.We described screening eÜects around localized electrostatic perturbations in charged or nonstoichiometric clusters and proposed a mechanism for electron redistribution based on the behaviour of the Madelung potential. It predicts an increase of ionicity around the perturbed site while further away the bonds become alternately more covalent and more ionic. This allows an electron displacement to take place away from a negatively charged perturbed site and towards a positively charged site. This mechanism for screening in materials with a gap in the excitation spectrum is markedly diÜerent from screening in simple metals which takes place naturally through the very delocalized character of the electron wave functions in metals.We have found that in TiO2 species screening processes are very efficient as can be expected from a compound with a relatively narrow gap and are very dependent upon the chemical nature and the location of the perturbed site in the cluster thus exemplifying the importance of local –eld eÜects. 301 Faraday Discuss. 1999 114 285»304 Finally it should be stressed that these small TiO clusters represent ideal model systems to 2 study the ionocovalent character of the anion»cation bonding and the screening properties in insulating compound systems. They provide a wide range of conformations associated to very diÜerent local environments of the atoms. In addition the non-negligible part of covalent character in the Ti»O bonding in bulk rutile is responsible for the strong responses to electrostatic perturbations and for the large electron redistributions.We have seen in this work that the understanding on covalency gained in the clusters extends to bulk and clean surfaces. Current work is under progress to show that the same is true as regards surface modi–cations by adsorption or non-stoichiometry. (9.1) X\KD[D0 Next K 8 Acknowledgements Calculations were performed on the Cray C98 at the IDRIS computational centre in Orsay (projects 984089 and 994089). We acknowledge fruitful discussions with C. Lecomte and M. Souhassou. eC[eA A being the oxygen and C the neighbouring Ti. It has been said that D 0 b/(eC[eA) i.e. a monotonic decreasing function F (9.2) D[D0BF@](eCA[eCA 0 ) (9.3) XBKF@](eCA[eCA 0 ) Next K (9.4) XPUO[ 1 d (9.5) XPUO[ 1 d [dV 9 Appendix In this Appendix we give a theoretical background to local –eld eÜects in anionic and nonstoichiometric clusters.Local –eld eÜects are characterized by the ratio X\oNscr/Next o between i i the screening charge and the charge perturbation on the site on which it is localized. We –rst consider cationic clusters and restrict ourselves to the analysis of local screening around a one-fold coordinated oxygen. The variation D[D of the electron transfer results from the eÜect of the perturbation on the energy diÜerence is a monotonic increasing function of the ratio e of CA\eC[eA . In a linear response theory framework the variations of D are obtained by diÜerentiation of this function and We consider the electrostatic contribution of Next to (eCA[eCA 0 ).It is equal to the diÜerence between the Hartree intra-atomic term UONext on the O and an interatomic one Next/d on the Ti if no other holes are present in the vicinity If another atom bearing a non-zero hole density is present in the neighbourhood such as in TiO2 ` Ti2O4 `-C2v and Ti3O6 `-S4 its contribution has to be taken into account via a potential V1Next on the oxygen and V2Next on the titanium. In the geometries under consideration dV \V2[V1[0 and its value grows as the two holes get closer. Smaller values of X result. This explains the evolution of X (X\0.17 0.53 and 0.64) in TiO2 ` Ti2O4 `-C2v and Ti3O6 `-S4 . It can be said that the two holes interact destruc- the series tively on screening.Faraday Discuss. 1999 114 285»304 302 We then consider the diÜerence in screening of holes in positively charged clusters and nonstoichiometric ones. Actually the values of X are very diÜerent in the two cases despite the fact that in many instances the same oxygen atoms are involved. Quite systematically screening of the non-stoichiometry is more efficient than screening of a positive charge. We assign this diÜerence to the fact that no global positive charge exists in the non-stoichiometric clusters which remain neutral as shown in the following way. We again restrict ourselves to the analysis of local screening around a one-fold coordinated oxygen and use eqn. (9.3) to estimate X.The electrostatic contribution of Next to (eCA[eCA 0 ) is of U the order of ONext as regards the oxygen levels. The titanium levels on the other hand experience a diÜerent Madelung contribution according to whether the cluster is cationic or nonstoichiometric. In the former case the Madelung contribution is of the order of Next/d as shown above. In the second case it amounts to NextW with W \1/d due to the in—uence of a surrounding negative charge which balances Next and assures the neutrality of the cluster. As a result in non-stoichiometric clusters (9.6) XPUO[W is larger than in cationic clusters. References 1 V. E. Henrich and P. A. Cox T he Surface Science of Metal Oxides Cambridge University Press Cambridge 1994. 2 Z. Zhang S. P.Jeng and V. E. Henrich Phys. Rev. B Condens. Matter 1991 43 12004; J. Nerlov G. M‘ller Surf. Sci. 1996 348 28. Qingfeng and P. J. 3 S. R. Barman and D. D. Sarna Phys. Rev. B Condens. Matter 1994 49 16141. 4 L. F. Mattheiss J. Phys Condens. Matter 1996 8 5987 and references therein. 5 M. Catti G. Sandrone and R. Dovesi Phys. Rev. B Condens. Matter 1997 55 16122. 6 C. Noguera Physics and Chemistry at Oxide Surfaces Cambridge University Press Cambridge 1996; Physique et Chimie des Surfaces dœOxydes Eyrolles 1995 collection Alea Paris. 7 G. Charlton P. B. Howes C. L. Nicklin P. Steadman J. S. G. Taylor C. A. Muryn S. P. Harte J. Mercer R. MacGrath D. Norman T. S. Turner and G. Thornton Phys. Rev. L ett. 1997 78 495 and references therein. 8 M. Li W. Henbenstreit and U.Diebold Surf. Sci. 1998 414 L951; C. L. Pang S. A. Haycock H. Raza P. W. Murray G. Thornton O. Gulseren R. James and D. W. Bullet Phys. Rev. B Condens. Matter 1998 58 1586; H. Onishi and Y. Iwasawa Surf. Sci. 1994 313 L783. 9 U. Diebold J. M. Pan and T. E. Madey Surf. Sci. 1995 331ñ333 845. 10 M. Barnes A. J. Mercer and G. F. Metha J. Mol. Spectrosc. 1997 181 180; L. A. Kaledin J. E. McCord and M. C. Heaven J. Mol. Spectrosc. 1995 173 499. 11 A. J. Merer Annu. Rev. Phys. Chem. 1989 40 407. 12 N. S. McIntyre K. R. Thompson and W. Weltner Jr. J. Phys. Chem. 1971 75 3243. 13 W. Yu and R. B. Freas J. Am. Chem. Soc. 1990 112 7126. 14 H. Wu and L. S. Wang J. Chem. Phys. 1997 107 8221. 15 C. W. Bauschlicher P. S. Bagus and C. J. Nelin Chem. Phys.L ett. 1983 101 229; R. Bergstroé m S. Lunell and L. A. Eriksson Int. J. Quant. Chem. 1996 59 427; M. V. Ramana and D. H. Phillips J. Chem. Phys. 1988 88 2637. 16 A. Hagfeldt R. Bergstroé m H. O. G. Siegbahn and S. Lunell J. Phys. Chem. 1993 97 12725. 17 T. Albaret F. Finocchi and C. Noguera Appl. Surf. Sci. 1999 144»145 672. 18 U. Landman D. Scharf and J. Jortner Phys. Rev. L ett. 1985 54 1860. 19 S. Pollack C. R. C. Wang and M. M. Kapper Z. Phys. D 1989 12 241. 20 R. N. Barnett H. P. Cheng H. Hakkinen and U. Landman J. Phys. Chem. 1995 99 7731. 21 G. Rajagopal R. N. Barnett and U. Landman. Phys. Rev. L ett. 1991 67 727. 22 E. C. Honea M. L. Homer P. Labastie and R. L. Whetten Phys. Rev. L ett. 1989 63 394. 23 F. Finocchi and C. Noguera Phys. Rev. B Condens.Matter 1998 57 14646; F. Finocchi and C. Noguera Eur. Phys. J. D 1999 9 in press. 24 K. Prabhakaran D. Purdie R. Casanova C. A. Muryn P. J. Hardman P. L. Wincott and G. Thornton Phys. Rev. B Condens. Matter 1992 45 6969. 25 R. Souda W. Hayami T. Aizawa and Y. Ishizawa Surf. Sci. 1993 285 265. 26 R. Heise and R. Courths Surf. Sci. 1995 333 1460. 27 J. Nerlov Q. F. Ge and P. J. M‘ller Surf. Sci. 1996 348 28. 28 J. Nerlov S. V. Christensen S. Weichel E. H. Pedersen and P. J. M‘ller Surf. Sci. 1997 371 321. 29 H. Onishi and Y. Iwasawa Catal. L ett. 1996 38 89; Surf. Sci. 1997 371 321. 30 P. J. D. Lindan J. Muscat S. Bates N. M. Harrison and M. Gillan Faraday Discuss. 1997 106 135. 303 Faraday Discuss. 1999 114 285»304 31 M. A. San Miguel C. J. Calzado and J.F. Sanz Int. J. Quant. Chem. 1998 70 351. 32 L. Kleinman and D. M. Bylander Phys. Rev. L ett. 1982 48 1425. 33 N. Troullier and J. L. Martins Phys. Rev. B Condens. Matter 1991 43 1993. 34 F. Finocchi J. Goniakowski and C. Noguera Phys. Rev. B Condens. Matter 1999 59 5178. 35 X. Gonze R. Stumpf and M. Scheffler Phys. Rev. B Condens. Matter 1991 44 8503. 36 M. Leslie and M. Gillan J. Phys. C. 1985 18 973. 37 R. F. W. Bader Chem. Rev. 1991 91 983; R. F. W. Bader in Atoms in Molecules-A Quantum T heory Oxford University Press Oxford 1990. 38 C. Noguera A. Pojani F. Finocchi and J. Goniakowski in Chemisorption and reactivity on supported clusters and thin –lms T owards an understanding of microscopic processes in catalysis ed. R. M. Lambert and G. Pacchioni Kluwer ASI series E Applied Sciences 1997 331 455. 39 A. Pojani F. Finocchi and C. Noguera to be published. 40 F. A. Cotton and G. Wilkinson in Advanced Inorganic Chemistry Wiley New York 5th edn. 1980. 41 S. D. Elliott and R. Ahlrichs J. Chem. Phys. 1998 109 4267. 42 S. C. Abrahams and J. L. Bernstein J. Chem. Phys. 1971 55 3206. 43 M. Ramamoorthy R. D. King-Smith and D. Vanderbilt Phys. Rev. B Condens. Matter 1994 49 7709. 44 K. M. Glassford and J. R. Chelikowsky Phys. Rev. B Condens. Matter 1992 46 1284. 45 B. Hird and R. A. Armstrong Surf. Sci. 1997 385 L1023. 46 Given the small —uctuations of the actual charges around a mean value our approach may be thought of as a perturbation calculation on the eÜective atomic levels. 47 J. Q. Broughton and P. S. Bagus J. Electron Spectrosc. Relat. Phenom. 1980 20 261. 48 G. Pacchioni A. M. Ferrari and G. Ierano Faraday Discuss. 1997 106 155. 49 M. S. Hybertsen and S. G. Louie Phys. Rev. B Condens. Matter 1987 35 5585. 50 G. Onida L. Reining R. W. Godby R. Del Sole and W. Andreoni Phys. Rev. L ett. 1995 75 818. Paper 9/03066B Faraday Discuss. 1999 114 285»304 304

ISSN:1359-6640    

DOI:10.1039/a903066b

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

The in½uence of soft vibrational modes on our understanding of oxide surface structure N. M. Harrison,*§a X.-G. Wang,b J. Muscatîa and M. Schefflerb a CCL RC Daresbury L aboratory Daresbury W arrington UK W A4 4AD b Fritz-Haber-Institut der Max-Planck Gesellschaft Faradayweg 4-6 D-14195 Berlin-Dahlem Germany 1 Introduction Titanium dioxide (TiO2) is an interesting and industrially important material which has been studied extensively in recent years. This interest is in part due to the existing applications of TiO2 as a white pigment and as a catalyst support1,2 but also due to the many new applications currently under investigation. Recent examples include self cleaning paint coatings,3 catalytically active paving stones,4 solar cells5 and water disinfection.6 TiO is also of interest as a model 2 transition metal oxide.It is readily reduced in the bulk and at the surfaces resulting in the occupation of Ti d orbitals which had a profound eÜect on the physical and electronic structure. 7h9 However the bulk structure of TiO is simple relative to many oxides and thus a variety 2 of empirical and –rst principles theories can readily be used to compute its physical and chemical properties. It is therefore an excellent model system displaying many of the properties of more complex oxides which can be studied relatively easily using a variety of experimental and theoretical techniques. Many of the important and useful properties of TiO depend on the physical and electronic 2 structure of its surfaces. The structure of the most stable (110) surface has attracted enormous § Also at Department of Chemistry Imperial College London London UK SW72AY.î Present address CSIRO Minerals Box 312 Clayton South Victoria 3169 Australia. 305 Faraday Discuss. 1999 114 305»312 Received 5th August 1999 We examine the reasons for the poor quantitative agreement between the structures predicted from the minimum energy con–guration of –rst principles calculations and those deduced from surface X-ray diÜraction experiments for the structure properties of the TiO (110) surface. In order to con–ne all numerical approximations very large scale 2 all-electron –rst principles calculations are used. We –nd a very soft anisotropic and anharmonic surface rigid-unit vibrational mode which involves displacements of the surface ions of approximately 0.15 ” for thermal vibrations corresponding to room temperature.It is concluded that in order to perform an accurate comparison between theory and experiment for this and perhaps other oxide surfaces it will be necessary to take account of such anisotropic vibrations in models used to interpret experimental data. In addition the contribution of the vibrational entropy to the surface free energy is likely to be signi–cant and must be taken into account when computing surface energies and structures. This journal is( The Royal Society of Chemistry 2000 interest in recent years. Experimental studies have included low energy electron diÜraction (LEED),10 scanning tunneling microscopy (STM),11h13,7 surface X-ray diÜraction (SXRD)14 and ion scattering.15 There have also been a large number of –rst principles theoretical studies.16h20 Early contributions included a periodic Hartree»Fock study16 within a linear contribution of atomic orbitals (LCAO) formalism and a density functional theory study (DET) which used a full potential linear augmented plane wave (FP-LAPW) method.17 In each case the model systems studied were rather small and the surfaces only partially relaxed.More recently plane-wave (PW) pseudopotential calculations have been used based on both the local density (LDA) and generalised gradient (GGA) approximations to DFT.18h20 These calculations included extensive relaxations of the surface structure and were based on larger structural models. Despite these extensive eÜorts the agreement between computed and measured structures is semi-quantitative at best.The extensive experience of calculations on bulk oxides which has been built up in recent years leads one to expect that DFT and HF calculations will reproduce experimental bond lengths to somewhat better than 0.1 ”. At the (110) surface the inward relaxation of the bridging oxygen ion determined by SXRD is” ” [0.27 while most calculations –nd a relaxation of less than [0.1 . The current article is concerned with a detailed examination of this discrepancy. Transition metal oxides represent a signi–cant challenge to –rst principles calculations. The localised nature of the oxygen 2p and in particular the titanium 3d states makes the PW pseudopotential method particularly demanding.The expansion of localised orbitals in plane waves requires large kinetic energy cut-oÜs to converge the total energy. In addition the separation of the sp and d electron eigenvalues in the periodic system is dependent on the choice of the atomic reference state from which the pseudopotential is constructed. Great care must be taken to understand the eÜect of these approximations on computed material properties. Recently Hamann21 had discussed in detail calculations for bulk TiO and concluded that computed geometries and 2 energies varied signi–cantly with the choice of the local component in the pseudopotential. With these difficulties in mind we have chosen to use two complementary all electron techniques to study the TiO (110) surface.Firstly the FP-LAPW method22 employs a basis set con- 2 sisting of plane waves which (inside atom-centered non-overlapping spheres) are matched continuously in value and slope to an expansion in terms of spherical harmonics (here up to lmax wf \10) and numerical solutions of the radial Schroé dinger equation. This basis set has maximum —exibility and ensures the high accuracy of the calculations. Secondly the LCAO method which employs a basis set of atom centred Gaussian functions for which a hierarchy of basis sets approaching complete convergence has recently been developed and tested for TiO2 surfaces.24,25 By con–ning all numerical approximations we provide de–nitive DFT-GGA results for this surface. Using the energy surface obtained we are able to examine the comparison of theory and experiment and thus resolve this long standing problem.The next section contains details of the computational methods used the results are then presented and discussed and our conclusions are summarised in the –nal section. 2 Methodology In this section we give details of the structural model used to describe the (110) surface and of the FP-LAPW and LCAO methods used to perform our calculations. 2.1 Structural model The (110) surface is modelled as a slab periodic in [001] and [16 10] directions but –nite in the [110] direction (as shown in Fig. 1). As essential requirement for quantitative studies is that computed properties are fully converged with respect to the thickness of the slab. A systematic series of tests in which the structures of slabs of varying thickness were fully relaxed revealed that for a slab containing 21 atomic layers (i.e.7 O»Ti2O2»O layers) the surface energy was converged to better than 0.1 J m~2 (6 eV ”~2) and geometric displacements to better than 0.02 ”.23h25 2.2 FP-LAPW In the FP-LAPW calculations a supercell periodic in 3 dimensions was used in which the slab geometry described above was repeated in the [110] direction with slabs separated by a large vacuum region of 9 ” to ensure that there were no signi–cant interactions between the slabs. The Faraday Discuss. 1999 114 305»312 306 Fig. 1 A section through a (110) surface viewed in the [001] direction. The surface is based on a 21 layer slab with a mirror plane through the centre of the slab. All symmetry inequivalent ions in the top half of the slab are labelled.size of this structural model is signi–cantly larger than that used in previous studies of surface structures within the FP-LAPW method. These calculations have been made feasible by recent developments and improvements of the method.26 (RTi MT\0.90 ” ROMT\0.80 ”) and there- Emax wf was mandatory to ensure good numerical accuracy. The electron In order to minimise the number of k-points required to converge the surface energy care was taken to ensure a systematic cancellation of errors between the calculations on the slab and bulk crystal. This is achieved by describing the bulk crystal with a unit cell corresponding as closely as possible to that used to describe the surface and using identical computational parameters in both sets of calculations.A bulk unit cell with six times the volume of the primitive cell was used. With this arrangement we found that a uniform k-point mesh with three points in the irreducible part of E the Brillouin zone was adequate. A kinetic energy cut-oÜ for the plane-wave basis of max wf \22 Ry was used. This is a rather high value for such huge systems. However because of the large surface relaxations we had to use rather small muffin-tin spheres fore a large value for density and potential are expanded in lattice harmonics up to lmax pot \6 inside the spheres and the wavefunctions are expanded in angular momenta up to lmax wf \10. The electron density and potential in the interstitial region are expanded in plane waves up to 144 Ry.The core states are treated fully relativistically. The Ti 3s 3p and O 2s which are represented by local orbitals as well as the valence states (Ti 3d 4s O 2p) are treated scalar-relativistically. The relaxations of all atoms in the slab were considered and the surface structure was determined by relaxing the entire system to equilibrium. All the atoms were relaxed according to the force directions and total energy minimization until all atom forces for a geometry fall below a certain limit. The process of the structure optimization has been described in ref. 27. 2.3 LCAO The LCAO calculations were performed with the CRYSTAL program.28 In contrast to the FP-LAPW calculations the slab geometry was modelled as periodic in two dimensions and –nite in the third removing the need to de–ne a vacuum gap.307 Faraday Discuss. 1999 114 305»312 The main approximation in the LCAO formalism is the choice of the local basis set used to expand the Bloch orbitals of the crystal. The basis set is made up of atom centered Gaussian functions with s p or d symmetry. A systematic hierachy of basis sets was developed in a recent study of the TiO (100) surface.24 In this study it was shown that sets employing two basis func- 2 tions to describe the valence electrons (so called double valence»DV) can predict surface ionic relaxations to an accuracy of 0.02 ” compared to the basis set limit. Tests for the (110) surface con–rm these conclusions and so in the current study a DV basis set has been used the details of which are given elsewhere.24,25,29 The total energy of the bulk crystal and surface were explicitly converged with respect to k-point sampling.A Pack»Monkhorst mesh28,30 of order 4 which yields 10 k-points in the irreducible Brillouin zone of a (110) slab and 36 in that of the bulk crystal were used. This procedure of converging the bulk and slab energies explicitly with respect to k-space sampling removes the reliance on a systematic cancellation of errors when computing surface properties. CRYSTAL computes matrix elements of the Coulomb exchange and correlation matrix elements by direct summation over the in–nite periodic lattice. Very efficient computational schemes for truncating the lattice summations have been developed.31 The accuracy of the summation is based on overlap criteria for the atomic orbitals.Details of the control of these criteria have been described elsewhere.32h34,24 In the current study the criteria were chosen to achieve an accuracy in the relative energies of the surface and bulk structures of the order of 1 meV per cell.35 The surface relaxations were performed using an adapted conjugate gradient minimisation algorithm36 to a tolerance of 0.01 ” in atomic positions and 10~5 eV in the total energy. 2.4 The Exchange-correlation functional The main results of this article have been computed using the GGA functionals recently introduced by Perdew Burke and Ernzerhof (PBE).37 In addition a number of alternative treatments of the electron exchange and correlation interactions have been used in order to establish the sensitivity of key results.Within the LCAO formalism Hartree»Fock (i.e. non-local exchange with no treatment of correlation) and LDA calculations were also performed.38,39,25 Within the FP-LAPW formalism the GGA functional proposed by Perdew and Wang (PWGGA)40 was also used. We –nd that the structural properties of the bulk crystal and the surface are very insensitive to the choice of functional. The PBE and PWGGA approaches agree to well within the numerical tolerances and diÜerences between HF LDA and GGA are con–ned to less than 0.02 ” in any surface or bulk displacement.25 It is likely that diÜerences between previous calculations which have been assigned to diÜering treatments of exchange and correlation are in fact due to incomplete convergence of the calculations.16h20 3 Results and discussion The atomistic structure of the TiO (110) surface is depicted in Fig.1 in which labels are assigned 2 to the atoms in the surface region. The relaxations of the top few layers computed here and in a number of recent studies are compared to those deduced from surface X-ray diÜraction experiments in Table 1. At –rst sight the most notable feature of this data is that the agreement between theory and experiment is poor. This is particularly true for the position of the bridging oxygen ion (O(3)) for which the computed relaxation is never more than” ” [0.16 while the experiment –nds [0.27 . On closer examination it is the discepancy between the various theoretical approaches which gives most cause for concern.This is especially true for the current study in which as stated above great care has been taken to control the eÜects of all numerical tolerances on two diÜerent allelectron approaches. Nevertheless the relaxation of the bridging oxygen ion is computed to be [0.02 or [0.16 ” and the relaxation of the six-fold coordinated Ti-ion directly ììbeneathœœ it (Ti to be 0.23 or 0.08 ” in the LCAO and FP-LAPW methods respectively. These variations (1)) are signi–cantly larger than the numerical errors that one would expect. It is however evident that the Ti»O separation in the [110] direction is more consistent (this is reported in the –nal row of Table 1). In the current study this is 1.04” ” and 1.03 in the FP-LAPW and LCAO calculations Faraday Discuss.1999 114 305»312 308 from bulk terminated positions) of the ions at the (110) (1]1) surface computed within the FP-LAPW and LCAO formalisms ” Table 1 The displacements (in compared to previous plane wave calculations and those deduced from surface X-ray diÜraction dataa,b SXRD14 PW-GGA20 PW-LDA18 LCAO-PBE FP-LAPW-PBE [1 6 10] [110] [1 6 10] [110] [1 6 10] [110] [1 6 10] [110] [1 6 10] [110] Label ^0.05 ^0.05 ^0.08 Ti(1) Ti(2) O(3) O(4) O(5) O(6) Ti(7) Ti(8) O(9) O(10) O(11) O(12) Ti(13) Ti(14) O(15) O(16) O(17) O(18) O(19) O-Tia 0.12 [0.16 [0.27 0.05 0.05 0.05 0.07 0.23 [0.11 [0.02 0.18 0.18 0.03 0.12 0.23 [0.17 [0.02 0.13 0.13 0.02 0.14 » » » [0.04 0.04 0.13 [0.17 [0.06 0.13 0.13 0.08 [0.23 [0.16 0.09 0.09 » » » ^0.08 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 [0.16 0.16 0.00 0.00 0.00 0.00 0.07 [0.05 0.05 0.00 0.00 0.00 0.00 0.03 [0.06 0.06 0.00 0.00 0.00 0.00 0.03 ^0.05 ^0.05 ^0.08 [0.05 0.05 Faraday Discuss.1999 114 305»312 ^0.08 » » [0.07 0.06 [0.08 0.02 [0.09 0.07 [0.13 » » » » 0.02 » » » » 0.05 ^0.04 ^0.04 ^0.08 [0.06 0.03 ^0.06 0.00 ^0.06 [0.05 [0.02 [0.07 ^0.06 0.00 [0.09 ^0.08 » 0.03 » 0.00 » » » » » » » » » [0.09 0.00 0.02 0.02 [0.10 0.00 0.03 0.03 » » ^0.06 [0.05 [0.04 [0.04 [0.03 [0.03 [0.01 [0.03 0.00 0.00 0.00 0.00 0.02 [0.03 0.00 0.00 0.00 0.00 [0.01 0.05 [0.06 0.01 0.01 0.01 0.01 [0.04 0.02 [0.08 » » » » » » » » ^0.07 » » » » [0.12 0.00 » » » » » » » » » » » » » » » » » [0.07 [0.03 [0.03 [0.02 0.02 0.00 0.00 [0.02 [0.02 [0.02 [0.02 0.02 » » » » » » » » » » » » » » » » [0.01 0.89^0.13 1.03 1.09 1.03 1.04 309 a The last row gives the O(3)-Ti(1) separation in the [110] direction.b The labels and directions refer to Fig. 1. respectively while 0.89 ” ^0.13 is deduced from the SXRD experiment and ion scattering measurements. 15 This observation leads one to the hypothesis that the energy surface with respect to vertical displacement of the bridging oxygen is rather —at and that the displacement involves a cooperative motion of the surface ions which at the least involves the six-fold coordinated surface Ti-ion (Ti(1)).In order to explore this possibility a number of relaxations have been performed in which the position of the bridging oxygen ion has been constrained but the surrounding surface ions fully relaxed. For reasons of efficiency these calculations were performed on a relatively small system containing 9 atomic layers. The minimum energy structure of this system is somewhat displaced from that of larger systems but it displays all of the features necessary to examine the qualitative structure of the energy surface.The resultant energy surfaces computed within both the FP-LAPW and LCAO formalisms are displayed in Fig. 2. The LCAO calculations on this smaller system were performed with a very large basis set (the TVAEd basis set reported in ref. 24 and 25). In order to give some feeling for the energy scale a line representing a typical room temperature (k thermal energy per degree of freedom BT \0.025 eV) has been drawn on the –gure. The energy surface is sufficiently —at for thermal vibrations to leave the minimum undetermined to about 0.15 ”. From this it is clear that the discrepancy between diÜerent theoretical approaches is due to the difficulty in –nding an absolute energy minimum in this very —at energy surface. We may associate the —at energy surface with a highly anisotropic and anharmonic surface vibrational mode.The nature of the mode is easily seen from an animation of the atomic positions. The displacements explored at thermal energies approximately corresponding to room temperature are displayed in Fig. 3. During this vibration the O(3)»Ti(1) separation along [110] remains very close to 1.03 ” and the separation of Ti and O is also nearly constant. The (1) (6) displacements of Ti(2) O(4) and O are very small. Thus to the –rst approximation we may (5) understand the vibration as a ìì rigid unit modeœœ of the square planar TiO unit containing Ti 4 (1) and its four nearest neighbours (O(3) O(6)) and their periodic images (this is depicted in the lower left panel of Fig. 3). An immediate consequence of the rigid unit mode is that the structure of the TiO (110) surface 2 apparent in the experimental probes applied at –nite temperature does not correspond to the minimum energy con–guration computed within a total energy calculation.In order to make such a comparison further treatment of the eÜect of surface vibrational modes on our interpretation of experimental data must be explored. In the case of SXRD and LEED experiments this necessitates the modelling of an anharmonic thermal vibration which is also highly anisotropic. The current practice is to –t the diÜraction rods within an harmonic and often isotropic Debye»Waller model which is inadequate for the current case. A quantitative interpretation of STM images will require a treatment of the tip»surface interaction as this is likely to result in signi–cant distortions of the surface structure.Fig. 2 The relative energy of the (110) surface computed within the LCAO and FP-LAPW formalisms for various –xed positions of the bridging oxygen ion relative to the unrelaxed bulk terminated position. Faraday Discuss. 1999 114 305»312 310 Fig. 3 The approximate room temperature thermal motions of the atoms within the soft rigid unit mode of the (110) surface.° The atom labels correspond to those in Fig. 1. The oxygen atoms comprising the rigid unit are labelled in the lower left panel. In addition the free energy associated with soft surface modes cannot be neglected. This has been demonstrated in recent –rst principles free energy calculations on the Ag(111) surface.In this system the minimum energy structure corresponds to a contraction of the outer layer spacing of [1.0% while the free energy minimum at 1150 K yields an expansion of 6.3%»a shift in the interlayer spacing of 0.16 ”.41 We expect a signi–cantly larger eÜect at the TiO (110) surface due 2 to the presence of a soft anharmonic surface vibrational mode. 4 Conclusion There is poor quantitative agreement between the structures predicted from the minimum energy con–guration of –rst principles calculations and those deduced from X-ray diÜraction experiments for the Ti0 (110) surface. We –nd that a very soft and anharmonic surface rigid-unit vibrational 2 mode involves displacements of the surface ions of approximately 0.15 ” for thermal vibrations corresponding to room temperature.In order to perform an accurate comparison between theory and experiment for this and perhaps other surfaces it will be necessary to take account of such anisotropic vibrations in models used to interpret experimental data. In addition the contribution of the vibrational entropy to the surface free energy is likely to be signi–cant and must be taken into account when computing surfaces energies and structures. References 1 V. E. Henrich and A. F. Cox T he Surface Science of Metal Oxides Cambridge University Press Cambridge 1993. 2 A. Fujishima and K. Honda Nature (L ondon) 1972 238 37. 3 New Scientist 1998 8. 4 New Scientist 1997 15. 5 New Scientist 1998 11. 6 I. M. Butter–eld P. A. Christensen A. Hamnett K. E. Shaw G. M.Walker S. A. Walker and C. R. Howarth J. App. Electrochem. 1997 27 385. 7 R. A. Bennett P. Stone N. J. Price and M. Bowker Phys. Rev. L ett. 1999 82 3831. 8 J. Muscat N. M. Harrison and G. Thornton Phys. Rev. B 1999 59 15457. 9 A. T. Paxton and L. Thien-Nga Phys. Rev. B 1998 57 1579. 10 V. E. Henrich and R. L. Kurtz Phys. Rev. B 1981 23 6280. 11 H. Onishi and Y. Iwasawa Surf. Sci. 1994 313 L783. 12 P. W. Murray N. G. Condon and G. Thornton Phys. Rev. B 1995 51 10989. 13 U. Diebold J. F. Anderson K. O. Ng and D. Vanderbilt Phys. Rev. L ett. 1996 77 1322. 14 G. Charlton P. B. Howes C. L. Nicklin P. Steadman J. S. G. Taylor C. A. Muryn S. P. Harte J. Mercer R. McGrath D. Norman T. S. Turner and G. Thornton Phys. Rev. L ett. 1997 78 495. 15 B. Hird and R.A. Armstrong Surf. Sci. L ett. 1997 385 L1023. ° Two movies of these motions are available as electronic supplementary information. See http :// www.rsc.org/suppdata/fd/1999/305 311 Faraday Discuss. 1999 114 305»312 16 P. Reinhardt and B. A. Hess Phys. Rev. B 1994 50 12015. 17 D. Vogtenhuber R. Podloucky A. Neckel S. G. Steinemann and A. J. Freeman Phys. Rev. B 1994 49 2099. 18 M. Ramamoorthy D. Vanderbilt and R. D. King-Smith Phys. Rev. B 1994 49 16721. 19 P. J. D. Lindan N. M. Harrison M. J. Gillan and J. A. White Phys. Rev. B 1997 55 15919. 20 S. P. Bates G. Kresse and M. J. Gillan Surf. Sci. 1997 385 386. 21 D. R. Hamann Phys. Rev. B 1997 56 14979. 22 P. Blaha K. Schwarz P. Sorantin and S. B. Trickey Comput. Phys. Commun. 1990 59 399. 23 J.Muscat and N. M. Harrison in preparation. 24 J. Muscat N. M. Harrison and G. Thornton Phys. Rev. B 1999 59 2320. 25 J. Muscat PhD Thesis University of Manchester 1999. 26 M. Petersen and M. Scheffler 1999 to be published. 27 B. Kohler S. Wilke M. Scheffler R. Kouba and C. Ambrosch-Draxl Comput. Phys. Commun. 1996 94 31. 28 R. Dovesi V. R. Saunders C. Roetti M. Causa` N. M. Harrison R. Orlando and E. Apra` CRY ST AL 95 Userœs Manual University of Turin Turin 1996. 29 http://www.dl.ac.uk/TCS/Software/CRYSTAL/. The CRYSTAL Basis set library 1998. 30 J. D. Pack and H. J. Monkhorst Phys. Rev. B 1977 16 1748. 31 C. Pisani R. Dovesi and C. Roetti Hartree»Fock Ab Initio T reatment of Crystalline Systems Springer- Verlag Berlin 1988 vol. 48. 32 R. Orlando R. Dovesi C. Roetti and V. R. Saunders. J. Phys. Condens. Matter 1990 2 7769. 33 R. Dovesi M. Causa R. Orlando C. Roetti and V. R. Saunders J. Chem. Phys. 1990 92 7402. 34 Quantum Mechanical Ab Initio Calculation of the Properties of Crystalline Materials ed. C. Pisani Springer-Verlag Berlin 1996 vol. 67. 35 Explicitly the ITOL parameters were set to 10~6 10~6 10~6 10~6 and 10~12. 36 C. Zhu R. H. Byrd P. Lu and J. Nocedal L -BFGS-B»Fortran Subroutines for L arge Scale Bound Constrained Optimisation Dept of Elec. Eng. and Comp. Sci Northwestern University Illinois 1994. 37 J. P. Perdew K. Burke and M. Ernzerhof ACS Symposium-Series 1996 629 453. 38 P. A. M. Dirac Proc. Cambridge Philos. Soc. 1930 26 376. 39 J. P. Perdew and A. Zunger Phys. Rev. B 1981 23 5048. 40 J. P. Perdew J. A. Chevary S. H. Vosko K. A. Jackson M. R. Pederson D. J. Singh and C. Fiolhais Phys. Rev. B 1992 46 6671. 41 J. Xie S. Gironcoli S. Baroni and M. Scheffler Phys. Rev. B 1999 59 970. Paper 9/06386B Faraday Discuss. 1999 114 305»312 312

ISSN:1359-6640    

DOI:10.1039/a906386b

出版商:RSC    

年代:1999

数据来源: RSC