Article pubs.acs.org/JPCC
Deprotonated Glycine on Cu(111): Quantitative Structure Determination by Energy-Scanned Photoelectron Diffraction D. A. Duncan,† M. K. Bradley,† W. Unterberger,‡ D. Kreikemeyer-Lorenzo,‡ T. J. Lerotholi,§ J. Robinson,† and D. P. Woodruff*,† †
Physics Department, University of Warwick, Coventry, CV4 7AL, United Kingdom Fritz-Haber Institut der MPG, Faradayweg 4-6, D14195 Berlin, Germany § School of Chemistry, University of Witwatersrand, PO Wits, Johannesburg, 2050, South Africa ‡
ABSTRACT: The local adsorption site of the deprotonated simple amino acid glycine (glycinate) on Cu(111) has been investigated quantitatively by O 1s and N 1s energy-scanned photoelectron diffraction (PhD). The nitrogen atom is found to adsorb in a near-atop site with a Cu−N bond length of 2.02 ± 0.02 Å. However, based on the PhD data alone there is some ambiguity in the adsorption sites occupied by the oxygen atoms, although at least one of these atoms occupies a near-atop site with a Cu−O bond length of 2.00−2.02 ± 0.02−0.07 Å. Density functional theory (DFT) calculations have also been conducted on simple models (a low-coverage (3 × 3) phase of noninteracting molecules and a higher-coverage ordered (4 × 4) structure). The structural conclusions of the DFT calculations proved to be very sensitive to the use of different functionals and failed to resolve the structural ambiguity of the PhD analysis fully, but a single tridentate-bonding structural model appears to be most consistent with the PhD, DFT, and supporting spectroscopic information from previous studies.
1. INTRODUCTION The adsorption behavior of simple amino acids on metals is of significant interest in understanding how biological molecules interact with inorganic materials. Most biological molecules are far too complicated for study by quantitative surface structural techniques, but investigations of these simpler molecules may provide some insight into more general interaction trends. Amino acids have the general formula of H2N−CH(−R)− COOH, where R is a species of varying complexity and functionality. In the case of glycine, R is a single hydrogen atom, making glycine the simplest of the amino acids. Glycine is also the only amino acid that does not contain a chiral center. On Cu(110) and Cu(100) glycine is deprotonated to form a glycinate species H2NCH2COO, which bonds to the surface through the amino N atom and both carboxylate O atoms. This tridentate bonding geometry creates a chiral ’footprint’1,2 and leads to two different enantiomers of the adsorbed molecule, related to one another by a mirror reflection, as in conventional chiral molecules. However, because these two metal surfaces are themselves achiral, each of the two surface enantiomers is equally probable and the resulting adsorbate layer is racemic and must fail to show any of the optical or biologically distinct interactions that characterize enantiopure molecules.3 The local adsorption sites of glycinate on Cu(100) and Cu(110) have previously been determined1,2 by energy scanned photoelectron diffraction (PhD), the nitrogen and two oxygen atoms all occupying singly coordinated sites. On the squaremesh (100) surface there is a particularly close match of the underlying Cu atom mesh to the relative locations of the N and O atoms within the glycinate species, and all these atoms © 2012 American Chemical Society
occupy sites that are very close to atop surface Cu atoms. On the rectangular-mesh (110) surface the increased Cu−Cu spacing in the [001] azimuth is much larger than the N−O distance in glycinate, and the molecule adsorbs with the N atom very close to an atop site but with the two O atoms significantly displaced (by ∼0.8−1.0 Å) from atop sites, albeit still singly coordinated to the underlying Cu atoms. On the triangular-mesh (111) surface it is unclear what local adsorption geometry may be expected, and this is the issue the present study seeks to resolve. An early LEED (low-energy electron diffraction) study of the Cu(111)/glycine systems showed that, after deposition at room temperature, an (8 × 8) diffraction pattern is observed.4 Later STM (scanning tunnelling microscopy) experiments showed that this (8 × 8) pattern may be attributed to the coexistence of three equivalent domains of an (8 × 4) structure.5 The STM images also indicate that the (8 × 4) unit mesh contains two quite similar (4 × 4) components. Each (4 × 4) area was assumed to contain several glycinate species. The STM study also revealed formation of a (2√13 × 2√13) phase on annealing the surface to 400 K for 10 min. Significantly higherresolution STM images have very recently been published6 of this phase which show it to be comprised of molecular trimers that can also be seen at lower coverages, and simulated images are consistent with each molecule ‘lying down’ on the surface. RAIRS (reflection−absorption infrared spectroscopy) measureReceived: January 12, 2012 Revised: April 19, 2012 Published: April 20, 2012 9985
dx.doi.org/10.1021/jp300377x | J. Phys. Chem. C 2012, 116, 9985−9995
The Journal of Physical Chemistry C
Article
ments7 provided strong evidence that, as on Cu(100) and (110), the glycine is deprotonated upon adsorption onto Cu(111) in both of these structural phases. Structural models have been proposed on the basis of the STM images, but these are largely speculative, and the images obtained contain no information on the adsorbate−substrate registry. We note, however, that the structural models suggested for the (8 × 4) phase involve only glycinate species ‘lying down’ on the surface with tridentate bonding to the surface through both O atoms and the N atom, although symmetry considerations require that at least two different local geometries are involved. In the (2√13 × 2√13) phase it was suggested5 originally that some of the glycinate species adopt an O−O-bidentate configuration, bonding to the surface only through the two O atoms, but the more recent STM study6 provides much stronger evidence for this phase also involving only lying-down species. Here we present the results of a quantitative experimental local adsorbate structure investigation of the Cu(111)(8 × 4)glycinate phase using N 1s and O 1s PhD8,9 and the results of some related DFT calculations of this system The PhD technique7,8 exploits the coherent interference of the directly emitted component of a photoelectron wavefield from an adsorbate atom with other components of the same wavefield elastically scattered by the surrounding atoms. Scanning the photon energy leads to changes in the photoelectron energy and thus the photoelectron wavelength, causing different scattering paths to switch in and out of phase with the directly emitted wavefield. The resulting modulations in the measured intensity can be interpreted in terms of the different scattering paths associated with a particular adsorption geometry through the use of multiple scattering simulations for different trial structures.
powder (Sigma-Aldrich, >99% purity) in a simple glass tube evaporator to 400 K, while the sample was held at room temperature, followed by brief annealing to 400 K. An ordered (8 × 8) pattern was observed by LEED in initial characterization experiments, but this quickly faded with continued exposure to the electron beam. In view of this evidence of electron-beam damage, subsequent preparations for the PhD measurements used the same dosing method but the surface order was not checked by LEED. The surface coverage as judged by SXPS was essentially independent of whether or not the initial deposition produced a multilayer film of glycine, the 400 K anneal consistently producing a saturation submonolayer coverage. PhD modulation spectra were obtained by measuring photoelectron energy distribution curves (EDCs) of the O 1s and N 1s peaks, at 4 eV steps in photon energy, over the photoelectron kinetic energy range of 50−350 eV for a number of different polar emission angles in the [21̅ 1] and [110̅ ] azimuths. These data were processed following our general PhD methodology (e.g., refs 7 and 8) in which the individual EDCs are fitted by one or more Gaussian peaks, a Gauss error function (step), and a template background. The integrated areas of each of the individual peaks were then plotted as a function of photoelectron kinetic energy and used to define a smooth spline which represents the nondiffractive intensity and instrumental factors. The spline was then subtracted from and used to normalize the integrated areas to provide the final PhD modulation spectrum. DFT calculations were performed using the planewave pseudopotential code CASTEP.11 Three different functionals were applied, namely, the local density approximation (LDA) and two implementations of the generalized gradient approximation (GGA), using the the revised Perdew−Burke− Ernzerhof exchange-correlation functional (RPBE)12 and the Perdew and Wang exchange and correlation functional (PW91).13 Ultrasoft pseudopotentials from the CASTEP pseudopotential library were used throughout to minimize computation time. Additional calculations were conducted using a different set of pseudopotentials generated by the on the fly pseudopotential generator supplied with the CASTEP distribution. While calculations based on the different pseudopotentials led to changes in the absolute energy differences of different structural models, the qualitative results were unchanged, so only the results from the library pseudopotentials are shown here. Calculations based on the experimentally observed (8 × 4) unit mesh proved too demanding for the computational resources available in this study, so instead calculations were conducted on two simpler models. Specifically, calculations using a (3 × 3) mesh containing a single glycinate species (coverage 0.11 ML) were undertaken to identify the preferred adsorption geometry for ‘isolated’ species, while calculations using a (4 × 4) unit mesh containing three glycinate species (coverage 0.187 ML) provided a means of assessing the possible influence of intermolecular interactions. This is closely similar to the previously proposed models for the (8 × 4)glycinate structure which contained three glycinate species in each (slightly different) (4 × 4) subunit. These trimers are also similar to those believed to account for the more recent STM images,6 even at low average coverages. A kinetic energy cutoff of 400 eV and a 6 × 6 × 1 {4 × 4 × 1} Monkhorst−Pack kpoint sampling mesh were used to relax structures consisting of 5 layer single-sided (3 × 3) {(4 × 4)} slabs containing a total of
2. EXPERIMENTAL AND COMPUTATIONAL DETAILS The experiments were conducted in an ultra-high-vacuum surface science end station equipped with typical facilities for sample cleaning, heating, and cooling. This instrument was installed on the UE56/2-PGM-2 beamline of the BESSY-II synchrotron radiation source, which comprised a 56 mm period undulator followed by a plane grating monochromator.10 The sample was mounted approximately 1 m beyond the focus of the beamline optics to ensure the beam was defocused, greatly reducing the flux density and associated radiation damage. Sample characterization in situ was achieved by LEED and SXPS (soft X-ray photoemission spectroscopy) using the incident synchrotron radiation, SXPS also being used to check that no significant radiation damage occurred in the adsorbed molecular layers over the duration of the experiments. The SXPS and PhD measurements were obtained using an Omicron EA-125HR 125 mm mean-radius hemispherical electrostatic analyzer. The analyzer was equipped with sevenchanneltron parallel detection and mounted at a fixed angle of 60° to the incident radiation in the same horizontal plane as that of the polarization vector of the incident radiation. The sample could be rotated about its surface normal (to change the azimuthal angle) and about its vertical axis (to change the polar angle), allowing (simultaneous) variation of incidence and electron collection directions. A clean, well-ordered Cu(111) surface was prepared from an oriented and a polished crystal slice by the usual combination of Ar+ ion bombardment and brief annealing to 950 K to give a sharp (1 × 1) LEED pattern and a SXP spectrum devoid of impurities. Sample dosing was achieved by heating glycine 9986
dx.doi.org/10.1021/jp300377x | J. Phys. Chem. C 2012, 116, 9985−9995
The Journal of Physical Chemistry C
Article
45 {80} Cu atoms and 1 {3} glycinate species in the unit supercell. The bottom two layers were constrained to the calculated bulk fcc Cu structure (lattice parameters 3.655 (GGA-RPBE), 3.634 (GGA-PW91), and 3.519 Å (LDA)), while a vacuum gap of at least 16 Å in all models ensured negligible interaction between periodic images of the slab. Structures were considered to be converged when the forces fell below a tolerance level of 0.02 eV Å−1. Tests using double-sided slabs, with adsorbed molecules on both sides, revealed no significant differences in the relative energies of different structures, demonstrating that no dipole corrections were required. As the glycinate species is not stable in the gas phase (and indeed a DFT calculation undertaken on this gas-phase species found it to dissociate), absolute adsorption energies are not very meaningful, so we focus here only on the energetic dif ferences, ΔEmodel, between the adsorption energy per molecule for each model structure, Ea(model), and the adsorption energy per molecule, Ea(O−O‑bidentate), in the (3 × 3) lowcoverage phase of the ‘upright’ O−O-bidentate adsorption geometry in which only the two oxygen atoms are bonded to the surface and the nitrogen atom is a significant (nonbonding) distance from the metal surface. ΔEmodel = Ea(model) − Ea(O−O‐bidentate)
(1)
Ea(model) = (Eadsorbate + slab − Eslab)/Nglycinate
(2)
Figure 1. SXP spectra for the prepared Cu(111)/glycinate surface. Nominal calibration of the absolute binding energies was performed as described in the main text. O 1s, N 1s, and C 1s spectra were measured at 650, 500, and 400 eV, respectively, in the normal emission direction.
attributable to photoelectron diffraction, but a very similar effect is to be seen in published spectra from the interaction of serine with Cu(110),16 alanine with Cu(110),17,18 and glycine with Pd(111).19 It is possible that some of the intensity for the carboxylate C emission is in unresolved shakeup features at lower kinetic energy. Both the N 1s and the O 1s SXP spectra show only a single peak. A single O 1s peak clearly indicates that the molecule is deprotonated to produce the glycinate species but also indicates that the two O atoms in each molecule are in chemically similar sites on the surface. The single N 1s peak is consistent with the fact that there is only one N atom in each molecule, but we may also infer that if the adsorbed glycinate does occupy more than one local geometry; the associated N and O sites are chemically very similar. To gain quantitative information from PhD it is necessary to perform calculations using multiple scattering theory, comparing the results for different structural models with the experimental data. However, some information can be obtained by inspection of the PhD spectra. Specifically, when the angle between an emitter, a nearest-neighbor backscatterer, and the measurement direction is close to 180°, strong long-period modulations can be expected associated with a single dominant backscattering pathway. Figure 2 shows the seven N 1s and five O1s experimental PhD spectra that show the strongest modulations. The strongest modulations from both the nitrogen and the oxygen emitter atoms arise at or near normal emission and have similar long-range periodicity. This may be taken to imply that both the N and the O atoms occupy atop or near atop sites, relative to an underlying Cu backscattering atom. However, it is also notable that the modulations in the O 1s PhD spectra are significantly weaker than those measured from N 1s. This leads us to suggest that at least one O atom is likely to be significantly more laterally displaced from an exact atop site than the single N emitter atom or indeed that one O atom occupies a distinctly different (probably low symmetry) site. Interestingly, a similar qualitative effect in the normal emission PhD modulation amplitudes for N 1s and O 1s was seen in the measurements of glycine on Cu(100), with the O 1s modulations about a factor of 2 weaker,2 but in the present case of glycine on Cu(111) the effect is much larger with a difference of a factor of ∼5. Less surprisingly, perhaps, the O 1s PhD modulations from glycine on Cu(111) are a factor of ∼3 weaker than those seen for formate, HCOO, on the same surface,20 as this simpler species bonds to the surface only through the two carboxylate O atoms. Nevertheless, these comparisons highlight not only the effect of the constraints
Nglycinate is the number of glycine species per unit cell, and Eslab and Eadsorbate+slab are the total energies of the complete supercell slab without and with the adsorbate molecule. However, in order to provide a comparison with the results of Rankin and Sholl14 for glycine on Cu(110) we calculate the absolute adsorption/reaction energy quoted by them defined as Ereaction = {(Eadsorbate + slab − Eslab)/Nglycinate} − Eglycine + 0.5E H2
Of course, this quantity fails to take account of the fact that the detached acid H atoms must, at least initially, be adsorbed on the Cu surface before finding a second H atom to desorb as the molecular species, so it is not a true adsorption energy. For the calculations of Ereaction on Cu(111) the total energies of the free glycine and hydrogen molecules, Eglycine and EH2, were calculated in a 15 × 15 × 15 Å3 unit cell.
3. RESULTS 3.1. Characterization by SXPS and Qualitative analysis of PhD Data. The SXP spectra from the prepared Cu(111)/ glycinate surface are shown in Figure 1. Due to some calibration problems in the electron spectrometer controller we were unable to determine reliable absolute binding energies experimentally, and all of the nominal binding energies recorded (the difference between the monochromator energy and the measured kinetic energy) have been displaced by a constant amount such as to yield the same O 1s and C 1s binding energies as those reported by Hasselström et al.15 for glycine on Cu(110). The C 1s spectra contain two peaks, with the higher binding energy peak corresponding to the carboxylate carbon and the lower binding energy peak arising from the carbon bonded to the amine species. The origin of the very significant difference in the intensity of the two components is unclear, as the effect seems too large to be 9987
dx.doi.org/10.1021/jp300377x | J. Phys. Chem. C 2012, 116, 9985−9995
The Journal of Physical Chemistry C
Article
Figure 2. Experimental N 1s and O 1s PhD spectra from the Cu(111)/glycinate surface. Note that the modulation amplitudes of the O 1s spectra have been scaled up by a factor of 5.
imposed on the carboxylate−Cu bonding by the presence of an additional N−Cu bonding in glycinate but also the effect of the difference in these constraints on the (100) and (111) surfaces. 3.2. DFT Study of an Isolated Glycinate Species. While the PhD technique has been shown to be extremely effective in determining the local adsorption structure of simple systems with no prior assumptions regarding the likely solution, complex systems such as those involving multiple adsorption sites are more challenging and more likely to lead to ambiguities or increased imprecision in the solutions. In these situations, the results of DFT calculations concerning the predicted minimum energy structure and particularly the many subtle subsurface relaxations that may accompany adsorption can prove extremely valuable. For example, in a PhD study of the structure of deprotonated analine on Cu(110)21 comparison of the experimental results with those of DFT calculations proved invaluable in identifying both the agreed underlying structure (involving two distinct coexistent local geometries) and the fundamental discrepancy in the chemisorption bond lengths. Bearing in mind previous suggestions that the Cu(111)/glycine system also involves at least two distinct local adsorption sites it was felt that the results of DFT calculations for this system may prove valuable. However, as remarked in the Introduction, DFT calculations for a (8 × 4) unit mesh are beyond the computational facilities available to this study (a 5-layer supercell would contain a total of 214 atoms), so initial calculations were instead performed on an isolated glycine molecule within a (3 × 3) unit mesh. In view of the generally held view that intermolecular hydrogen bonding plays a major role in the ordering of surface phases of adsorbed amino acids, this may seem to be a serious oversimplification. However, we note that PhD is intrinsically a local structural probe, and identifying the preferred adsorbate− substrate geometry in the absence of intermolecular interactions may also provide a useful set of starting structures for the PhD simulations. Five different stable adsorption geometries were identified by these calculations, and schematic diagrams of these are shown in Figure 3, while their relative adsorption energies are shown
Figure 3. Schematic diagrams of the minimum energy Cu(111)/ glycinate models in an isolated (3 × 3) overlayer as calculated in GGARPBE. Largest (copper colored) atoms are the substrate copper atoms, with the other atoms scaled in size with their atomic number. Specifically, the O atoms are red, N atoms are blue, C atoms are black, and H atoms are white.
in Table 1. These values (in meV) are relative to the O−Obidentate configuration, which was found to be the most favorable geometry in the GGA-RPBE calculations; positive values indicate a less favorable structure. This structure, in which the amino N atom is far from the surface and not Table 1. Comparison of the Relative Adsorption Energies, ΔEmodel (in meV), of the Minimum Energy Cu(111)/ Glycinate Structures Found in DFT Calculations for the Low-Coverage (3 × 3) and Higher-Coverage (4 × 4) Phases Using Various Functionals (see text)a functional model (cell size) O−Obidentate hollow-O− Nbidentate atop-O− Nbidentate atoptridentate bridgetridentate
RPBE (3 × 3)
RPBE (4 × 4)
PW91 (3 × 3)
PW91 (4 × 4)
LDA (3 × 3)
LDA (4 × 4)
DFT-D PW91 (4 × 4)
0
15
0
29
0
37
−204
219
121
57
−46
−215
−386
−429
264
234
177
95
−238
−397
272
219
83
7
−339
−412
242
330
39
85
−287
−319
−246
a Energies given are defined by eq 1 and relative to the adsorption energy for the low-coverage O−O bidentate structure for each functional. Positive values indicate less stable structures. Schematics of the models can be found in Figures 3 and 4. Note that in the LDA (3 × 3) calculations no stable atop-O−N-bidentate or atop-tridentate structures were found: calculations started with these geometries converged on the bridge-tridentate structure.
9988
dx.doi.org/10.1021/jp300377x | J. Phys. Chem. C 2012, 116, 9985−9995
The Journal of Physical Chemistry C
Article
Table 2. Comparison of the Structural Parameters of the Lowest-Energy Low-Coverage Cu(111)(3 × 3)-Glycine Structures Obtained in DFT Calculations Using Different Functionalsa O−O-bidentate R-PBE
PW 91
hollow-O−N-bidentate LDA
dCu−N (Å) zCu−N (Å) xyCu−N (Å) dCu−O1 (Å)
2.04
2.01
1.93
zCu−O1 (Å)
2.03
2.00
1.93
xyCu−O1(Å)
0.19
0.20
0.15
dCu−O2 (Å)
2.02
1.99
1.92
zCu−O2 (Å)
2.02
1.99
1.92
xyCu−O2 (Å)
0.16
0.15
0.09
φN−C−C (deg) θNCC‑surf (deg) θCOO‑surf (deg) ψO−O‑surf (deg)
113 89.4 84.4 0.5
113 86.5 85.8 0.2
112 85.2 86.3 0.3
atop-O−Nbidentate
R-PBE
PW 91
LDA
P-RBE
PW 91
2.21 2.19 0.33 2.24 2.21 2.19 1.57 1.52 1.62 1.59 1.61 1.47
2.15 2.12 0.34 2.15 2.16 2.16 1.48 1.44 1.55 1.57 1.61 1.51
2.04 2.03 0.29 2.06 2.05 2.12 1.41 1.46 1.54 1.50 1.46 1.45
2.15 2.15 0.14 1.98
2.10 2.10 0.13 1.96
1.95
0.30
3.56
116 52.4 66.0 58.1
3.50
116 52.2 67.5 59.4
3.29
116 50.4 58.4 51.9
3.63
115 48.5 47.9 46.3
atop-tridentate
bridge-tridentate
PW 91
R-PBE
PW 91
LDA
2.29 2.27 0.36 2.10
2.18 2.15 0.36 2.07
2.20 2.20 0.09 2.12
2.16 2.16 0.10 2.10
2.05 2.05 0.07 2.03
1.93
2.09
2.06
2.11
2.09
2.02
0.31
0.20
0.17
0.15
0.16
0.17
2.13
2.10
2.09
2.04
0.42
0.48
2.33 2.67 2.03 2.23 1.14 1.46 108 52.4 33.6 3.0
2.20 2.54 1.88 2.08 1.15 1.45 108 44.0 35.3 2.8
2.08 2.47 1.81 2.00 1.02 1.45 107 46.4 32.7 2.9
3.55
115 48.9 43.9 43.4
R-PBE
105 71.1 37.6 0.3
105 63.8 35.9 0.1
a Note that in the LDA (3 × 3) calculations no stable atop-O−N-bidentate or atop-tridentate structures were found. d values are interatomic distances, z values are interlayer spacings, and xy values are lateral offset values from atop a surface Cu atom. In structures in which an oxygen atom is close to a bridge or hollow site the two or three Cu−O nearest-neighbour parameter values are given. O1 is the oxygen atom that is closer to the N atom, φ values are bond angles, θ is the angle between two planes, and ψ is the angle between two atoms and a plane.
involved in the adsorbate−substrate bonding, is clearly not consistent with the N 1s PhD data; our qualitative evaluation of these data indicate that the N atoms must lie close to atop a surface Cu atom at a bonding distance (∼2 Å), and this conclusion is fully supported by the results of multiple scattering simulations reported below. However, while the GGA-PW91 calculations also find this O−O-bidentate structure to be the lowest-energy configuration, the results of the LDA calculations are quite different with a tridentate geometry having a much lower total energy. Insofar as the quantity Ereaction provides a measure of the total adsorption energy the values obtained for GGA-RPBE (−0.014 eV), GGA-PW91 (−0.274 eV), and LDA (−0.499 eV) seem to indicate much weaker chemisorption than on Cu(110); the equivalent value we find for the lowest-energy (tridentate) adsorption structure on Cu(110) using GGA-PW91 is −0.995 eV. This latter value is smaller than that (−1.496 eV) reported by Rankin and Sholl, although we note that the value proved to be sensitive to full optimization of the minimum energy structure of the free glycine molecule; we found a value for Ereaction of −1.461 eV for a structure that proved to be a local, rather than global, energy minimum. The large difference in the predictions of the GGA and LDA calculations is clearly a major issue for the present study, although, as we show below, the problem is less severe in the calculations for the higher-coverage phase that more nearly describes the experimental situation. Differences in the absolute adsorption energies of these two approaches are well known, with LDA generally overbinding and GGA-RPBE, in particular, underbinding, an effect that also leads to small differences in
predicted bond lengths. This difference leads to shorter predicted bond lengths in LDA, as seen in the present case in Table 2, which shows a comparison of the detailed structural parameters found for the stable structures in these (3 × 3) DFT calculations. However, these different DFT methods generally lead to the same predicted lowest-energy model structures. Indeed, in the best-known example in which DFT calculations predict a simple adsorption structure at variance with that found experimentally, namely, that of CO on Pt(111), for which there is ample experimental evidence that atop sites are occupied at low coverage, a range of GGA and LDA calculations all result in the hollow site having the lowest energy.22 However, for the case of step site adsorption at high coverage the GGA and LDA calculations do lead to different preferred local sites (atop and bridge), while at higher coverages on the (111) surface GGA-RPBE and GGA-PW91 calculations also lead to different predicted site occupations. However, very significant differences in predicted molecular orientation between LDA and GGA have been observed previously for CN on Cu(111) and Ni(111);23 GGA calculations found the favored bonding geometry to be monodentate with the C−N axis perpendicular to the surface, but LDA calculations favored a bidentate geometry with the C−N axis near parallel to the surface, consistent with the experimentally determined geometry. Notice that this result has close parallels with the present results in that LDA favors a geometry in which there is a higher coordination of the adsorbed molecular ‘footprint’. While our results are from the first application of DFT calculations to glycine on Cu(111), there have been previous 9989
dx.doi.org/10.1021/jp300377x | J. Phys. Chem. C 2012, 116, 9985−9995
The Journal of Physical Chemistry C
Article
Figure 4. (4 × 4) structural models used in DFT calculations. Coloring is as in Figure 3. Bold lines show a single (4 × 4) unit mesh and dashed lines a second such mesh. Two sets of lines together define a (8 × 4) unit mesh (although the periodicity of the structures shown is (4 × 4)).
Table 3. Comparison of the Structural Parameters of the Minimum Energy Cu(111)(4 × 4)-Glycine Structures Found in DFT Calculations Using Different Functionalsa O−O-bidentate RPBE
a
PW 91
LDA
dCu−N(Å) zCu−N(Å) xyCu−N(Å) dCu−O1(Å)
2.04
2.01
1.94
zCu−O1(Å)
2.03
2.00
1.93
xyCu−O1(Å)
0.20
0.2
0.20
dCu−O2(Å)
2.02
1.994
1.92
zCu−O2(Å)
2.02
1.99
1.92
xyCu−O2(Å)
0.15
0.17
0.13
φN−C−C(deg) θNCC‑surf(deg) θCOO‑surf(deg) ψO−O‑surf(deg)
112 87.6 84.7 0.5
113 89.3 87.9 0.7
112 88.9 87.1 0.6
hollow-O−N-bidentate R-PBE 2.18 2.18 0.18 2.38 2.12 2.26 1.65 1.62 1.65 1.71 1.36 1.54
3.92
113 45.1 76.5 70.3
PW 91 2.13 2.12 0.16 2.33 2.09 2.19 1.57 1.56 1.51 1.72 1.39 1.52
3.78
112 49.8 74.5 68.2
LDA
atop-O−N-bidentate R-PBE
PW 91
2.15 2.14 0.25 1.9
2.04 2.03 0.19 2.13 2.00 2.14 1.49 1.51 1.52 1.53 1.32 1.52
3.59
LDA
PW 91
2.10 2.09 0.21 1.95
2.02 1.99 0.37 1.94
2.28 2.21 0.57 2.23
1.93
1.92
1.79
0.40
0.37
0.73
3.35
112 49.0 74.9 68.5
atop-tridentate RPBE
117 60.8 42.9 39.6
3.42
116 60.2 40.2 36.8
3.03
117 46.8 36.0 32.5
bridge-tridentate
LDA
RPBE
PW 91
LDA
2.19 2.11 0.58 2.19
2.10 1.98 0.69 2.11
2.19 2.19 0.16 2.10
2.13 2.12 0.17 2.10
2.05 2.02 0.32 2.02
2.16
2.12
2.05
2.09
2.08
2.01
0.53
0.54
0.52
0.25
0.24
0.21
2.14
2.10
2.05
2.12
2.07
2.00
0.33
0.37
0.42
104 72.1 34.2 1.8
104 73.6 37.2 1.7
104 70.0 36.2 1.4
2.50 2.65 2.14 2.28 1.29 1.35 108 52.1 30.3 6.5
2.33 2.57 1.98 2.14 1.22 1.42 108 48.3 23.3 3.2
2.12 2.79 1.97 2.19 0.79 1.72 108 40.8 25.1 0.8
Parameters are as defined in the caption to Table 2.
that only model structures with the C−C axis approximately parallel to the surface were explored.13,24,25 By contrast, a DFT cluster calculation (using a Cu16 cluster to represent the Cu(110) surface) did find that the tridentate geometry was favored over the O−O-bidentate geometry by 0.73 eV.26 In order to provide a more direct comparison with our results for the Cu(111) surface we therefore also ran a GGA-RPBE calculation on a single glycin(at)e species in a (3 × 2) unit mesh on the Cu(110) surface; the results yielded an energy
DFT studies of adsorbed glycine (and alanine) on Cu(110) and Cu(100), including both GGA and LDA calculations. On these two surfaces glycine is known from experiment to adopt the tridentate bonding configuration, and this structure has also been identified in the DFT studies. However, in these earlier DFT slab calculations the focus has been on the adsorbate− substrate registry and whether structures that involve two molecules per unit mesh form homochiral or heterochiral domains, and in at least some of these calculations it appears 9990
dx.doi.org/10.1021/jp300377x | J. Phys. Chem. C 2012, 116, 9985−9995
The Journal of Physical Chemistry C
Article
low-coverage (3 × 3) phase the two GGA calculations are in qualitative agreement and differ from the predicted structure obtained in LDA, but in the higher-coverage (4 × 4) phase it is the GGA-PW91 and LDA results that are in qualitative agreement and at odds with the predicted structure obtained in GGA-RPBE. In view of the much weaker adsorption that these results indicate for the glycinate species on Cu(111), relative to Cu(110), it is possible that dispersion forces, not accounted for in standard DFT calculations, may play a role in both the molecule−substrate and the intermolecular interactions. To explore this possibility we therefore also repeated some of the calculations using the DFT-D option for the PW91 functional available in CASTEP;28 the associated semiempirical corrections are not regarded as rigorous and universally applicable methods but have been found to provide a description of these additional forces that fall ‘in the right ballpark’.28 This addition of dispersion forces was found to significantly increase the adsorption energy for all geometries, as reflected in the larger (negative) value of Ereaction of −0.502 eV for this DFT-D calculation for the O−O-bidentate geometry in the (lowcoverage) (3 × 3) cell compared with the use of the same PW91 functional without dispersive forces (−0.274 eV). Table 1 shows the (negative) relative adsorption energies increase for all structures in the (4 × 4) models. The results of this GGAPW91 DFT-D calculation also lead to significant increases in the adsorption energies in the (4 × 4) structure relative to the isolated molecule O−O-bidentate value (Table 1), reflecting increased intermolecular interactions. This effect leads to this O−O-bidentate geometry being significantly more energetically disfavored, and indeed, the atop tridentate geometry is the second most favorable structure according to these calculations, being only 17 meV/molecule less favored than the hollow O− N-bidentate structure. Note that while the DFT-D calculations lead to significant changes in some of the energies, the changes in optimized atomic positions are minimal, in all cases being less than 0.02 Å. In summary, it appears that for the (4 × 4) calculations that take some account of intermolecular interactions it is only the GGA-RPBE calculations that favor the one structure, namely, O−O-bidentate, which is clearly inconsistent with the PhD experimental data. Both the GGA-PW91 (with and without dispersion forces included) and the LDA calculations for the (4 × 4) phase favor other structures that may prove to be more consistent with the experimental data. This is explored below. 3.4. Quantitative Analysis of PhD Data. Proper quantitative structure determination using the PhD technique requires the use of multiple scattering simulations for different trial structures, and these were performed using the computer codes developed by Fritzsche.29−32 These are based on expansion of the final state wave function into a sum over all scattering pathways that the electron can take from the emitter atom to the detector outside the sample. The agreement between theory and experiment was quantified using an objective reliability factor (R factor),7,8 similar to that defined by Pendry for quantitative LEED studies.33 This R factor is defined such that a value of 0 corresponds to perfect agreement and a value of 1 corresponds to uncorrelated data. Typically, the best determined structures have R factors in the range 0.2− 0.4, although the best fit that can be achieved is heavily dependent on the complexity of the structure and the amplitude of the modulations. Simple systems with emitter atoms in single high-symmetry adsorption sites typically show
preference from the tridentate geometry over the O−Obidentate geometry of 0.11 eV. The quantitative difference in this value from the results of the cluster calculation are probably not very significant; more important is the fact that this same GGA-RPBE approach leads to different preferred bonding geometries on the (111) and (110) surfaces. 3.3. DFT Study of a Higher-Coverage (4 × 4) Overlayer. While calculations for a (8 × 4) unit mesh structure were beyond the computational resources available in this study, we remarked above that the STM images of this phase5 showed the structure to be composed of two rather similar (4 × 4) units. The only difference in these two (4 × 4) units, in the specific structural model suggested in the STM paper, was that the glycinate species were located over either hcp or fcc hollow sites (i.e., over second- or third-layer Cu atoms). Indeed, this proposed model is actually homochiral, i.e., all the glycinate species have the same ‘footprint’ chirality. In fact, as neither the (111) surface nor the free glycine molecule is chiral, there must be an equal number of glycinate species of each footprint chirality, and one way to achieve this would be to have an overall heterochiral ordered phase. This could be achieved if the two (4 × 4) subunits of the (8 × 4) unit mesh had opposite chirality. Either of these models leads us to believe that calculations based on only a (4 × 4) subunit may be expected to provide much of the required insight into the additional role of intermolecular interactions on the surface. A set of calculations was therefore performed on a (4 × 4) supercell containing 3 glycinate species in locally equivalent sites related by the 3-fold rotation axis of the substrate (Figure 4). Five starting structures were used, corresponding to the five local structures identified in the low-coverage (3 × 3) calculations (Figure 3). The resulting relative adsorption energies for each of the higher-coverage (4 × 4) models are also included in Table 1. The local structural parameters of the glycinate species in the low-coverage (3 × 3) and highercoverage (4 × 4) structures are shown in Tables 2 and 3, respectively. These show that the effect of the intermolecular interactions on the local structure is generally small, consistent with our initial rationale for performing the low-coverage calculations. However, Table 1 shows that the influence of the intermolecular interactions on the relative adsorption energies is far more significant, consistent with previous reports of strong intermolecular hydrogen bonding in the Cu(110)/ glycinate system.27 As might be expected, for most structures these interactions lead to stronger bonding to the surface, indicative of attractive intermolecular interactions, but for the O−O-bidentate bonding geometry the total surface bonding is weakened in all calculations (although this geometry remains the most favorable energetically in the GGA-RPBE calculations). However, while in the GGA-RPBE calculations the energetic advantage of this favored geometry over the other structures remains quite large (in most cases >0.2 eV), the GGA-PW91 calculations reveal much smaller energy differences between the different structures. Indeed, using this functional the hollow-O−N-bidentate geometry is most favorable; relative to this structure the atop-tridentate geometry is only disfavored by 53 meV. These two geometries are also the lowest-energy structures in the LDA calculations. The effect of the intermolecular interactions in determining the energetically most favorable local geometry is therefore very significant, but there remain important differences in the predicted lowestenergy geometries based on the different functionals. For the 9991
dx.doi.org/10.1021/jp300377x | J. Phys. Chem. C 2012, 116, 9985−9995
The Journal of Physical Chemistry C
Article
strong PhD modulations and yield the lowest R factors; complex systems involving multiple and/or low-symmetry adsorption sites show weak modulations, and in these cases, the best fits commonly correspond to much higher values of the R factor. PhD modulations are generally dominated by the influence of substrate scattering, with intramolecular scattering being much weaker. In view of the reasonably strong modulations seen in the N 1s PhD spectra, we therefore first ran calculations to try to determine the approximate site of the N atom (within the molecule) on the surface. Bearing in mind the qualitative assessment in section 3.1, suggesting that the N species is in a near atop site, this initial determination of the N adsorption site was pursued by a grid search of parameter space for an isolated N atom placed on the surface close to an atop site. The position of the N atom was defined in spherical coordinates with the nearest-neighbor Cu atom as the origin, the three defining parameters being the bond length between the N and the Cu atom (dCu−N), the tilt of the Cu−O bond away from the surface normal, and the azimuthal orientation of the bond around the surface normal. The results showed no significant sensitivity of the calculated PhD spectra to the azimuthal rotation, but there was a strong dependence on both the dCu−N length and the polar angle. The best fit, with an R factor of 0.17, was found with values for dCu−N and the polar angle of 2.02 ± 0.02 Å and 12 ± 5°, respectively. This combination of parameters means that the N atom is laterally displaced by 0.42 ± 0.18 Å from an exact atop site. Including the nearest-neighbor C atom within the molecule as an additional scatterer in the calculation did not significantly affect the fit. Figure 5 shows a comparison of the seven experimental N 1s PhD spectra of Figure 2 with the results of the multiple scattering simulations for this best-fit structure. Notice that this experimental Cu−N bond length (2.02 ± 0.02 Å) is significantly shorter than that predicted by
the GGA-RPBE calculations for the structures that involve direct N−Cu bonding (2.15−2.29 Å), consistent with the expectation of underbinding with this functional. The measured bond length is actually much closer to the comparable predictions of the LDA calculations, while the GGA-PW91 calculations show intermediate values. The much weaker PhD modulations in the O 1s PhD spectra are clearly indicative of multiple sites and/or low-symmetry sites on the surface, although the similarity of the periodicity of the O 1s and N 1s spectra at normal emission also indicate that at least one O atom is in the vicinity of an atop site. Initial searches of possible tridentate structures prior to performing the DFT calculations proved rather inconclusive. Further calculations based on the stable structures found in the DFT calculations were therefore performed. Note that in these calculations it was assumed that the glycinate layer on the Cu(111) surface was racemic; calculations were therefore averaged over the local structures shown in both Figures 3 and 4 and their mirror images. Because PhD is intrinsically a local structural probe it cannot distinguish between an intrinsically heterochiral structure (such as that that would be created if the (8 × 4) ordering was due to adjacent (4 × 4) units of opposite chirality) and the alternative situation of homochiral domains of the two enantiomers. Of course, we may anticipate that the exact structural parameters found in the DFT calculations may not be correct, so the parameter space around these computed minimum-energy structures was explored to optimize the agreement between the PhD experimental data and the resulting multiple scattering simulations. To achieve this our implementation of the automated particle swarm optimization (PSO) algorithm34 was used, constraining the Cartesian coordinates of the two O atoms to lie within 0.2 Å in the x, y, and z directions of the theoretically predicted structures derived from the DFT calculations for the isolated (3 × 3) overlayer. Initially the calculations were performed with only the two O atoms on the surface (acting as both emitters and scatterers), but final refinement was performed including scattering from both C atoms of the glycinate. Only the nearest-neighbor (carboxylate) C atom influenced the simulated spectra, so the more distant N atom was not included. Although the presence of the carboxylate C atom did have a significant effect on the simulated spectra, they were very insensitive to its exact location. No acceptable structure starting from the hollow-O−N-bidentate model was found in these structural optimizations, but three acceptable fits were found for structures originating from the bridge-tridentate, atop-O− N-bidentate, and atop-tridentate models, with R values of 0.24, 0.24, and 0.23, respectively. Figure 6 shows a comparison of the five experimental O 1s PhD spectra of Figure 2 compared with the results of the multiple scattering simulations for these three best-fit structures. Note that all three models are also consistent with the adsorption site determined for the N atom. The associated experimentally determined structural parameter values for these three models are shown in Table 4 and compared with values obtained from the GGA-PW91 DFT (4 × 4) calculations (which are generally intermediate between the values obtained from the GGA-RPBE and LDA calculations). On the basis of the PhD data analysis alone, all of these structures are equally acceptable but additional information may provide a basis for further discrimination. One such piece of information is the results of the previous (RAIRS) vibrational spectroscopic study of this system.6 An
Figure 5. Comparison of the experimental N 1s PhD modulation spectra of Figure 2 with the results of the theoretical simulations for the best-fit local site of the N emitter atom, including its nearestneighbor C atom scatterer. 9992
dx.doi.org/10.1021/jp300377x | J. Phys. Chem. C 2012, 116, 9985−9995
The Journal of Physical Chemistry C
Article
the two O atoms are clearly in quite distinct local bonding environments. In this configuration one would expect the asymmetric C−O stretching mode to appear very strong in the RAIR spectra as this mode becomes strongly dipole active or, indeed, the two C−O stretches could become uncoupled. The RAIRS spectra, however, show an asymmetric C−O stretching mode with a frequency and relative intensity that is very similar to that seen on Cu(100) on which glycinate adopts a tridentate bonding. It therefore seems unlikely that this O−N-bidentate bonding is present on the Cu(111) surface. We may also note that it is likely that an O−N-bidentate bonding geometry would lead to sufficient differences in the environment of the two inequivalent O atoms to cause a splitting of the O 1s photoemission peak that is not observed. The remaining question is therefore whether there is any basis on which to distinguish the two tridentate bonding models in which the key difference is the location of the O atoms on the surface. In this regard, the data of Table 4 are relevant. The structures found in the PhD analysis were based on a reoptimization of the detailed structural parameters of models found to correspond to stable and reasonably lowenergy structures in the DFT calculations. It is therefore instructive to compare the resulting parameter values of the PhD and DFT solutions; while we may certainly anticipate bond length differences of as much as 0.1−0.2 Å, much larger differences in several parameters might indicate a more fundamental inconsistency. In this regard, two aspects of the parameters for the bridge-tridentate model are questionable. One is the distance of the bridging O atom from the nearestneighbor Cu atom which in the PhD analysis is 2.5 ± 0.1 Å; this distance is clearly too long to correspond to a chemisorption bond with the surface and is actually significantly longer than the value found in DFT analysis. This contrasts with the fact that the DFT calculations appear to consistently yield chemisorption bond lengths that are too long. Indeed, the near-atop O atom in this structure is found in the PhD analysis to have an associated Cu−O bond length of 2.02 ± 0.04, consistent with chemisorption and quite close to the value found in the PW91 calculations. In effect, therefore, the nominal bridge-tridentate structure found to be compatible with the PhD data is not tridentate in its bonding and is a fundamentally different structure from that identified in the DFT calculations. One further significant difference between the values of a parameter in the PhD and DFT analysis is the lateral offset of the N atoms from an atop site, xyCu−N. An excellent theoretical fit to the N 1s PhD data was obtained, with a value for this parameter of 0.42 ± 0.18 Å. This value is in good agreement only with that found in the DFT calculations for the atop-tridentate model, providing a further basis for regarding this solution as the most probable one.
Figure 6. Comparison of the experimental O 1s PhD modulation spectra of Figure 2 with the results of the theoretical simulations for the three alternative models giving the lowest R factors. Note that all three structures are compatible with the local site of the N atom, giving rise to the best fit to the N 1s PhD data (Figure 5).
Table 4. Structural Parameter Values Obtained from the Three Models Giving the Lowest R Factors in Simulations of O 1s and N 1s PhD Spectra, Compared with the Values Obtained in the GGA-PW91 DFT Calculations for the (4 × 4) Phasea atop-O−Nbidentate
atop-tridentate
bridge-tridentate
PhD expt.
DFT PW91
PhD expt.
DFT PW91
PhD expt.
DFT PW91
dCu−N (Å) zCu−N (Å) xyCu−N (Å) dCu−O1 (Å) zCu−O1 (Å) xyCu−O1 (Å) dCu−O2 (Å)
2.02(3) 1.96(2) 0.42(18)
2.10 2.09 0.21
2.02(3) 1.96(2) 0.42(18)
2.19 2.11 0.58
2.02(3) 1.96(2) 0.42(18)
2.13 2.12 0.17
1.99(3)
1.95
2.00(7)
2.19
2.02(4)
2.10
1.99(3)
1.92
1.98(7)
2.12
2.02(4)
2.08
0.1(2)
0.37
0.3(4)
0.54
0.1(4)
0.24
2.01(7)
2.10
2.5(1)
2.33
zCu−O2 (Å)
3.47(5)
2.00(4)
2.07
3.0(1) 2.31(7)
2.57 1.98
0.37
2.55(7) 1.0(3)
2.14 1.22
1.6(3)
1.42
xyCu−O2 (Å)
3.42
0.2(3)
4. GENERAL DISCUSSION AND CONCLUSIONS Using energy-scanned photoelectron diffraction we clearly identified the N adsorption site of the glycinate species on the Cu(111) to be close to atop; while on the basis of this technique alone there is some ambiguity over the exact bonding sites of the O atoms, it is clear that glycinate bonds to the surface through both the N and the O atoms. Through the use of DFT calculations to identify several stable adsorption geometries of glycinate on this surface we have shown that three possible models are potentially consistent with both these calculations and the PhD data, although the relative total energies of the different structures are found to be strongly
a
Parameters are as defined in the caption to Table 2. Values in parentheses are the error estimates in the last digit(s) of the parameter value.
important difference between the three models is that in the O−N-bidentate bonding model the O−C−O plane of the carboxylate species is tilted far out of the surface plane, while 9993
dx.doi.org/10.1021/jp300377x | J. Phys. Chem. C 2012, 116, 9985−9995
The Journal of Physical Chemistry C
Article
different specific sites or due to local disorder. On the other hand, DFT calculations indicate that the corrugation of the surface potential across the surface experienced by tridentatebonded glycinate is rather weak, so those molecules that are not locked in the ordered phase may have quite a high surface mobility. Of course, in contrast to the apparent success in identifying a single local geometry glycinate that may be dominant on Cu(111), the large variations in the predictions of the DFT calculations using different functionals is clearly a cause for concern. At least one previous example of this effect has been reported for CN adsorption on Ni and Cu,21 and of course, there may be other unreported examples. In the present case it seems that the GGA-PW91 calculations (probably including dispersion forces) may be the most reliable, favoring structures generally consistent with the experiment, although it is the structure calculated to have the second lowest energy that is most consistent with experiment. Of course, LDA calculations favor the same structures, but this approach is generally thought to be less suitable for molecular adsorption on metal surfaces. The well-known problems associated with DFT calculations of CO adsorption on some transition metals (e.g., ref 20) appear to have a somewhat different source, but both are a timely reminder of the potential hazards of over-reliance on this approach in isolation, despite its many successes. Evidently, it is the combination of such theory when combined with experimental results that is most likely to provide reliable structure determination.
dependent on the functionals used. In particular, GGA-RPBE calculations find the energetically favored model to be an ‘upright’ O−O-bidentate geometry, a result that is inconsistent with the PhD data that clearly shows the N atoms to adopt a bonding site on the Cu(111) surface. By contrast, LDA and GGA-PW91 calculations (both with and without inclusion of dispersion forces) find the two most favored geometries to be the hollow-O−N-bidentate and atop-tridentate models. Only the second of these is consistent with the PhD results, while the former also seems to be inconsistent with the results of previously published RAIRS experiments and our own O 1s SXPS. We therefore conclude that the most probable local adsorption geometry is the atop-tridentate model. The experimentally determined chemisorption bond lengths found for this structure are essentially identical to those found in similar studies of glycinate on Cu(100) and both glycinate and alaninate on Cu(110). Specifically, the Cu−N bond lengths found in these systems (2.04 ± 0.02, 2.04 ± 0.02, and 2.02 ± 0.03 Å1,2,19) are to be compared with the value in the present study of 2.02 ± 0.02 Å. These are also similar to values found for pyridine, C5H5N (2.00 ± 0.02 Å),35 and 2-methyl-pyridine, C5H4(CH3)N (2.04 ± 0.02 Å),36 on Cu(110). As might be expected, the Cu−N bond length found for ammonia, NH3, on Cu(111) is longer (2.09 ± 0.03 Å),37 although this is not the case for ammonia on Cu(110), which shows a shorter bond length of 2.00 ± 0.04 Å.38 The Cu−O chemisorption bond lengths found in this study (2.00−2.02 ± 0.02−0.07 Å) are also very similar to those found for glycinate on Cu(100) (2.05 ± 0.02 Å) and Cu(110) (2.02 ± 0.04 and 2.00 ± 0.04 Å)1,2 and alaninate on Cu(110) (2.02 ± 0.03 Å).19 Indeed, these Cu−O bond lengths are also consistent with those of the simplest carboxylate, formate HCOO, on the Cu(111) surface (1.99 ± 0.04 Å).18 These comparisons further reinforce our conclusion that glycinate adopts a tridentate bonding to Cu(111) involving both carboxylate O atoms as well as the amino N atom. It is interesting, though, that these bond lengths for the simple amino acids are the same on all three low-index faces of Cu. This contrasts with the shorter Cu−O bond lengths found for carboxylates on Cu(110), namely, formate (1.90 ± 0.03 Å),18 acetate, CH3COO (1.91 ± 0.04 Å),39 and benzoate, C6H5COO (1.91 ± 0.02 Å).40 The significance of these shorter bond lengths is discussed elsewhere;18 the difference from the amino acids may be a consequence of the ‘lying-down’ geometry associated with the extra substrate bonding through the N atom. While the results of this work appear to identify one specific preferred adsorption geometry, we should recall that the observed (8 × 4) periodicity of the ordered phase of glycinate on Cu(111) probably implies that at least two different geometries are occupied. One possibility is that the two geometries simply correspond to the two different enantiomers of a tridentate-bonded glycinate species; this possibility is already included in our PhD simulations. However, it is also possible that two or more distinct sites are occupied. In this regard we should note that the best-fit PhD simulations for the three structures of Table 4 were all performed with root-meansquared vibrational amplitudes for the O atoms of 0.13 ± 0.02 Å, significantly larger than those of bulk Cu atoms at room temperature (0.08 Å). The associated Debye−Waller factor does not, of course, distinguish between dynamic and static displacements of the atoms, so this apparent vibrational enhancement may be a result of static displacements either of
■
AUTHOR INFORMATION
Corresponding Author
*E-mail: d.p.woodruff@warwick.ac.uk. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS The authors acknowledge the partial support of the Engineering and Physical Sciences Research Council (U.K.) for this work. The computing facilities were provided by the Centre for Scientific Computing of the University of Warwick with support from the Science Research Investment Fund.
■
REFERENCES
(1) Booth, N. A.; Woodruff, D. P.; Schaff, O.; Gießel, T.; Lindsay, R.; Baumgärtel, P.; Bradshaw, A. M. Surf. Sci. 1998, 397, 258−269. (2) Kang, J.-H.; Toomes, R. L.; Polcik, M.; Kittel, M.; Hoeft, J. T.; Efstathiou, V.; Woodruff, D. P.; Bradshaw, A. M. J. Chem. Phys. 2003, 118, 6059−6071. (3) Kean, W. F.; Lock, C. J. L.; Howard-Lock, H. E. Lancet 1991, 338, 1565−1568. (4) Atanasoska, L. L.; Buchholz, J. C.; Somorjai, G. A. Surf. Sci. 1978, 72, 189−207. (5) Zhao, X.; Yan, H.; Zhao, R. G.; Yang, W. S. Langmuir 2003, 19, 809−813. (6) Kanazawa, K.; Taninaka, A.; Huang, H.; Nishimura, M.; Yoshida, S.; Takeuchi, O.; Shigekawa, H. Chem. Commun. 2011, 47, 11312− 11314. (7) Efstathiou, V.; Woodruff, D. P. Surf. Sci. 2003, 531, 304−318. (8) Woodruff, D. P.; Bradshaw, A. M. Rep. Prog. Phys. 1994, 57, 1029−1080. (9) Woodruff, D. P. Surf. Sci. Rep. 2007, 62, 1−38. (10) Weiss, M. R.; Follath, R.; Sawhney, K. J. S.; Senf, F.; Bahrdt, J.; Frentrup, W.; Gaupp, A.; Sasaki, S.; Scheer, M.; Mertins, H.-C.; 9994
dx.doi.org/10.1021/jp300377x | J. Phys. Chem. C 2012, 116, 9985−9995
The Journal of Physical Chemistry C
Article
Abramsohn, D.; Schäfers, F.; Kuch, W.; Mahler, W. Nucl. Instrum. Methods A 2001, 467−468, 449−452. (11) Clark, S. J.; Segall, M. D.; Pickard, C. J.; Hasnip, P. J.; Probert, M. J.; Refson, K.; Payne, M. C. Z. Kristallogr. 2005, 220, 567−570. (12) Zhang, Y.; Yang, W. Phys. Rev. Lett. 1998, 80, 890−890. (13) Perdew, J. P.; Burke, K.; Wang, Y. Phys. Rev. B 1996, 54, 16533− 16539. (14) Rankin, R. R.; Sholl, D. S. Surf. Sci. 2004, 548, 301−308. (15) Hasselström, J.; Karis, O; Weinelt, M.; Wassdahl, N.; Nilsson, A.; Nyberg, M.; Pettersson, L. G. M.; Samant, M. G.; Stöhr, J. Surf. Sci. 1998, 407, 221−236. (16) Eralp, T.; Shavorskiy, A.; Zheleva, Z. V.; Held, G.; Kalashnyk, N.; Ning, Y.; Linderoth, T. R. Langmuir 2010, 26, 18841−18851. (17) Allegretti, F.; Polcik, M.; Sayago, D. I.; Demirors, F.; O’Brien, S.; Nisbet, G.; Lamont, C. L. A.; Woodruff, D. P. New J. Phys. 2005, 7, 109−1−19. (18) Jones, G.; Jones, L. B.; Thibault-Starzyk, F.; Seddon, E. A.; Raval, R.; Jenkins, S,J; Held, G. Surf. Sci. 2006, 600, 1924−1935-L8. (19) Gao, F.; Li, Z.; Wang, Y.; Burkholder, L.; Tysoe, W. T. J. Phys. Chem. C 2007, 111, 9981−9991. (20) Kreikemeyer-Lorenzo, D.; Unterberger, W.; Duncan, D. A.; Bradley, M. K.; Lerotholi, T. J.; Robinson, J.; Woodruff, D. P. Phys. Rev. Lett. 2011, 107, 046102−1−4. (21) Sayago, D. I.; Polcik, M.; Nisbet, G.; Lamont, C. L. A.; Woodruff, D. P. Surf. Sci. 2005, 590, 76−87. (22) Feibelman, P. J.; Hammer, B.; Norskov, J. K.; Wagner, F.; Scheffler, M.; Stumpf, R.; Watwe, R.; Dumesic, J. J. Phys. Chem. B 2001, 105, 4018−4025. (23) Harrison, M. J.; Woodruff, D. P.; Robinson, J. Surf. Sci. 2006, 600, 340−347. (24) Rankin, R. R.; Sholl, D. S. Surf. Sci. 2005, 574, L1−L8. (25) Rankin, R. R.; Sholl, D. S. J. Phys. Chem. B 2005, 109, 16764− 16773. (26) Nyberg, M.; Hasselström, J.; Karis, O.; Wassdahl, N.; Weinelt, M.; Nilsson, A.; Pettersson, L. G. M. J. Chem. Phys. 2000, 112, 5420− 5427. (27) Nyberg, M.; Odelius, M.; Nilsson, A.; Pettersson, L. G. M. J. Chem. Phys. 2003, 119, 12577−12585. (28) McNellis, E. R.; Meyer, J.; Reuter, K. Phys. Rev. B. 2009, 80, 205414−1−10. (29) Fritzsche, V. Surf. Sci. 1992, 265, 187−195. (30) Fritzsche, V. J. Phys.: Condens. Matter 1990, 2, 1413−1424. (31) Fritzsche, V. Surf. Sci. 1989, 213, 648−656. (32) Fritzsche, V.; Pendry, J. B. Phys. Rev. B 1993, 48, 9054−9057. (33) Pendry, J. B. J. Phys. C: Solid State Phys. 1980, 13, 937−944. (34) Duncan, D. A.; Choi, J. I. J.; Woodruff, D. P. Surf. Sci. 2012, 606, 278−284. (35) Giessel, T.; Schaff, O.; Lindsay, R.; Baumgärtel, P.; Polcik, M.; Bradshaw, A. M.; Koebbel, A.; McCabe, T.; Bridge, M.; Lloyd, D. R.; Woodruff, D. P. J. Chem. Phys. 1999, 110, 9666−9672. (36) Terborg, R.; Polcik, M.; Hoeft, J. T.; Kittel, M.; Pascal, M.; Kang, J. H.; Lamont, C. L. A.; Bradshaw, A. M.; Woodruff, D. P. Surf. Sci. 2000, 457, 1−10. (37) Baumgärtel, P.; Lindsay, R.; Gießel, T.; Schaff, O.; Bradshaw, A. M. J. Phys. Chem. B 2000, 104, 3044−3049. (38) Booth, N. A.; Davis, R.; Toomes, R.; Woodruff, D. P.; Hirschmugl, C.; Schindler, K.-M.; Schaff, O.; Fernandez, V.; Theobald, A.; Hofmann, P.; Lindsay, R.; Giebel, T.; Baumgärtel, P.; Bradshaw, A. M. Surf. Sci. 1997, 387, 152−159. (39) Weiss, K.-U.; Dippel, R.; Schindler, K.-M.; Gardner, P.; Fritzsche, V.; Bradshaw, A. M.; Kilcoyne, A. L. D.; Woodruff, D. P. Phys. Rev. Lett. 1992, 69, 3196−3199. (40) Pascal, M.; Lamont, C. L. A.; Kittel, M.; Hoeft, J. T.; Terborg, R.; Polcik, M.; Kang, J. H.; Toomes, R.; Woodruff, D. P. Surf. Sci. 2001, 492, 285−293.
9995
dx.doi.org/10.1021/jp300377x | J. Phys. Chem. C 2012, 116, 9985−9995