Article pubs.acs.org/JPCC
First-Principles Study of the Formamide Adsorption to the OxygenCovered (0001) Surface of Ruthenium Philip R. McGill† and Tilo Söhnel*,†,‡ †
School of Chemical Sciences, The University of Auckland, Private Bag 92019, Auckland, New Zealand Centre for Theoretical Chemistry and Physics, The New Zealand Institute for Advanced Study, Massey University, Auckland, New Zealand
‡
ABSTRACT: Density functional theory has been used to study the adsorption of formamide over the (0001) surface of ruthenium metal with a coadsorbed oxygen (1 × 2) overlayer. Molecular adsorption is found to be energetically preferred and in the case of a perfectly ordered oxygen overlayer proceeds mainly through hydrogen bonding between the amino hydrogen and surface oxygen as well as some weak coordinate bonding between the carbonyl oxygen and surface ruthenium. The most energetically favored dissociative mode is found to involve the transfer of a proton to metal HCP sites and coordinate bonding between both the formamide N and O atoms and surface Ru. The resulting adsorption energies are found to be too small to explain experimental desorption temperatures but may be improved somewhat by the inclusion of an empirical dispersion correction. The effect of vacancies and disorder in the oxygen overlayer on formamide adsorption has also been investigated. It is found that disorder in the oxygen overlayer yields higher adsorption energies and may explain a number of experimental observations.
1. INTRODUCTION Understanding the strength and nature of interaction between various functional groups of adsorbates and the surfaces to which they adhere is of fundamental importance to surface science. The amide group has major relevance to biochemistry: the peptide linkages in proteins, for example, are secondary amides. Formamide, being the most elementary amide, is an excellent model molecule to probe the nature of this interaction: its simplicity means it lacks both other functional groups and large carbon chains that may convolute a study. Formamide has also been shown to self-react to yield nucleobases under a variety of conditions,1 making its reactions relevant to origin of life theories. Whereas a number of experimental studies have examined formamide adsorption to metal,2 nonmetal,3 and metal oxide1b,4 surfaces, far less computational work has been conducted on formamide adsorption to surfaces. A theoretical study does exist for formamide adsorption over the Ag (111) surface.5 It was shown that pure GGA functionals greatly underestimate adsorption energies; however, the application of an empirical dispersion correction6 restores them to more expected levels − suggesting that much of the interaction between Ag and formamide may result from dispersion forces. Formamide was found to bind preferentially lying flat to the surface. However, whereas experimental studies exist for formamide adsorption to Ag nanoparticles,7 no such study exists over a well-defined single-crystal surface. Some of the most comprehensive experimental studies of formamide adsorption over surfaces were those on ruthenium. On the clean Ru (0 0 0 1) surface, two decomposition routes are reported.2e In the first route, formamide dissociatively © 2012 American Chemical Society
adsorbs at 80 K. This dissociation involves hydrogen bonded to the carbon atom. Further decomposition occurs when the temperature is raised to 250 K, leaving NH3, NH, CO, and H adatoms. The NH3 and CO desorb at temperatures of 315 and 480 K, respectively. The NH decomposes forming N and H adatoms, which undergo combinative desorption as H2 and N2 at 420 and 700 K. In the second decomposition route, formamide initially adsorbs in a molecular fashion through coordinate bonding between the O atom and surface Ru. Raising the temperature to 225 K results in a change to a different dissociative adsorption mode, suggested to be HCONH; this binds through the underbonded N atom to the surface. HCONH undergoes a second H dissocation at 300 K, leaving HCON. The HCON decomposes at 375 K, leaving CO and surface N and H atoms. The second decomposition route is reportedly only observed at higher coverages. The reaction pathway can be significantly altered, however, by the presence of coadsorbed oxygen. On the Ru (0 0 0 1) surface with a 2 × 1 oxygen overlayer, a different and singular reaction path is reported: formamide adsorbs molecularly at 84 K through coordinate bonding between the carbonyl O and surface.2d Raising the temperature to 225 K results in a change to a different molecular adsorption mode, suggested to involve coordinate bonds between the surface and both N and O atoms. Increasing the temperature to 250 K causes a transition to dissociative adsorption, whereby an amino hydrogen Received: March 21, 2012 Revised: May 24, 2012 Published: June 6, 2012 14368
dx.doi.org/10.1021/jp302729e | J. Phys. Chem. C 2012, 116, 14368−14381
The Journal of Physical Chemistry C
Article
dissociates to the surface metal atoms. These adsorption modes are shown schematically in Figure 1.
Figure 2. Schematic of the Ru (0 0 0 1) surface with 2 × 1 oxygen overlayer. The square box indicates the surface repeat unit. HCP and FCC surface sites are also indicated.
Figure 1. Formamide adsorption scheme to Ru (0001) O 2 × 1 suggested by Weinberg and Parmeter.
tional.9 Plane-wave basis sets were employed with a 30 Ryd kinetic cutoff energy and an augmented charge density cutoff of 300 Ryd. The all-electron potential was replaced with ultrasoft pseudopotentials.10 K-point sampling was conducted with a uniform 16 × 16 × 12 grid on the bulk unit cell and grids of the same density on the larger cells used for the surface. Occupancy was smeared across the Fermi energy by the method of Maizari and Vanderbilt11 with a smearing width of 0.005 eV. All slabs contained at least 15 Å vacuum space. Transition states were located with the climbing image nudged elastic band method.12 Vibrational normal modes were obtained through partial Hessian vibrational mode analysis. 13 Unless otherwise indicated, calculations were conducted using PWSCF.14 To examine the effect of dispersion interactions, we have applied the empirical dispersion correction by Grimme6 to selected calculations when noted.
These hydrogens combine and desorb as H2. Each change in adsorption mode is accompanied by competitive desorption at high coverages. At 420 K, the HCONH remaining on the surface undergoes decomposition through simultaneous CO and H2 evolution. The N atoms appear to remain on the surface until 570 K, at which point they combine and desorb as N2 gas. The adsorption of formamide to hydrogen precovered Ru(001) surface has also been previously studied using high-resolution electron energy loss spectroscopy and thermal desorption mass spectrometry.2f As with the oxygen precovered surface, the presence of hydrogen is found to modify the adsorption mode and surface chemistry. As with the oxygen precovered surface, formamide dosed at 80 K adsorbs through the carbonyl oxygen. Increasing the temperature to 200 K results in adoption of a different molecular adsorption mode through both the C and O atoms. Increasing the temperature again to 250 K results in a breaking of the formamide C−H bond and dissociation of the H to the surface, whereas temperatures above 300 K cause decomposition into CO, NH, and H. The NH further decomposes into isolated N and H, which desorb as H2 and N2 at higher temperatures. The appearance of two different (temperature controlled) molecular adsorption modes for formamide in the case of either O or H coadsorption has been explained in terms of steric effects of the coadsorbed atoms, which initially prevent the adoption of the most energetically favored configuration. It was suggested that once a high enough temperature is reached, the O or H on the surface becomes mobile can be displaced to allow for more energetically favorable formamide adsorption modes.2f Numerous other adsorbates have also been studied on the Ru (0 0 0 1) surface, by both experiment and theory. Despite this detailed experimental work available, no complementary computational study has been performed on formamide adsorption to ruthenium with or without coadsorbed oxygen. This article seeks to rectify this deficiency. Given that the reaction pathway actually appears simpler with the presence of coadsorbed oxygen, the oxygen precovered Ru (0 0 0 1) surface will be the focus of this study. The Ru (0 0 0 1) surface is formed from the close-packed termination of a hexagonally close-packed lattice. Oxygen atoms are energetically biased toward the “HCP sites” of the surface, at sufficient coverages forming an ordered oxygen overlayer.8 The Ru(0 0 0 1) surface with 2 × 1 oxygen overlayer is shown in Figure 2, with hexagonally close-packed (HCP) and face-centered cubic (FCC) surface sites indicated.
3. RESULTS AND DISCUSSION 3.1. Slab Thickness Tests. As an initial test of required slab thickness, the adsorption energy of oxygen at the 2 × 1 coverage has been calculated at various slab thicknesses, relative to an isolated oxygen atom. For a three layered slab, an adsorption energy of 522 kJ mol−1 was obtained. Increasing the slab thickness to four layers resulted in an increase in adsorption energy to 529 kJ mol−1; this increased by only 1 kJ mol−1 on further extending the slab to seven layers. The adsorption energy found here was slightly greater than that previously reported on a four-layered slab of value of 509 kJ mol−1, with differences probably due to either the use of differing exchange correlation functional or pseudopotentials or the relaxation methodology employed. The prior study used a four-layer slab with the bottom two layers of the slab fixed. Here all slab layers were allowed to relax. While not fully converged, the three-layer slab will be used for the remainder of the work on this surface because it allows for wider supercells to be used. This thickness has been employed by others for the study of graphene adsorbed to the Ru (0 0 0 1) surface.15 Because the adsorption of other molecules on this surface has been shown to be strongly influenced by the presence of disorder in the oxygen overlayer, a brief study of disorder in the overlayer will be conducted before moving onto formamide adsorption. Given the very large adsorption energy of oxygen on ruthenium, other forms of defects within the overlayer, such as vacancies, are unlikely on a well-dosed surface. 3.2. Oxygen Defects from the Disorder in the Oxygen Overlayer. The first step to probing the likelihood of oxygen displacement defect formation is an examination of the relative energies of various oxygen displacements. A 2 × 4 supercell of this surface was used, featuring a single oxygen displaced into
2. COMPUTATIONAL DETAILS All surface calculations were carried out within density functional theory using the PBE exchange correlation func14369
dx.doi.org/10.1021/jp302729e | J. Phys. Chem. C 2012, 116, 14368−14381
The Journal of Physical Chemistry C
Article
higher energies for oxygen accommodation and thus to a reasonable approximation may also be excluded. Under such assumptions, the statistical weight is given as the product of the possible combinations of vacancies in the main oxygen lattice and the combinations of displaced oxygen atoms in the unoccupied HCP site lattice. Using this to provide the entropic contribution to the free energy and seeking to minimize it with respect to defect number, a fractional expression for defect concentration can be found to be
empty HCP or FCC sites. Illustrations of the different displacements examined are shown in Figure 3, along with
⎛ E ⎞ n ≈ N exp⎜ − d ⎟ ⎝ 2kT ⎠
(1)
where Ed is the defect formation energy, N is the total number of HCP sites in the alternate rows, and n is the number of defects. This is the same elementary expression that is commonly used to estimate interstitials within the bulk of simple crystals, where the same number of relevant interstitial sites exist as main lattice points (for example, octahedral sites within a face-centered cubic lattice). The expression predicts ∼4% of all adsorbed oxygen will adopt this disordered configuration at 300 K and ∼1% at 225 K, where formamide begins to change modes of adsorption on this surface. These are acceptably small numbers to validate the assumptions used and potentially a large enough value to have some significant effect on the adsorption of species. However, only a very small fraction of defects would be present at 80 K when formamide is initially dosed. Recently, STM measurements have been carried out over a 1 × 2 oxygen covered ruthenium (0 0 0 1) surface at 7 K.16 It was found that domains of differently orientated 1 × 2 oxygen overlayers form. The oxygen defect concentration was somewhat higher than the values found above (perhaps 10%. However, most of the defects were present at the boundaries between domains. Within the domains, the defect concentration appeared around a few percent. While in good agreement with the displacement defect values above for 200−300 K, this is considerably higher than the close to zero defect concentration predicted at 7 K. This suggests that the surface oxygens may lose mobility at 200−300 K. A second important avenue of study is the kinetics of defect formation: specifically whether the oxygen atoms are likely to reach the equilibrium concentrations found above over a reasonably short time frame (relative to the experiment). Transition states for oxygen displacement have been calculated by way of the climbing image nudged elastic band method for oxygen displacement along the surface from ordered to FCCadjacent, FCC-adjacent to HCP-adjacent. Eight images were used to interpolate the reaction coordinate. The HCP-adjacent to FCC-distant and FCC-distant to HCP-distant barriers have not been evaluated as being due to the high energy of the FCCdistant state, which even without any additional activation barrier will make a transition from the HCP distant state to FCC distant state extremely rare. This suggests that rather than direct HCP-distant to HCP-distant diffusion steps, displaced oxygens move through a multistep process wherein either the vacancy migrates with the displaced oxygen or a new defect forms ahead of the originally displaced oxygen to allow the originally displaced oxygen to move without crossing the FCCdistant configuration. Table 1 shows the activation barriers for diffusion. For comparative purposes, a study of O diffusion on the clean Ru (0 0 0 1) surface using plane-wave basis sets and a
Figure 3. Top view of the Ru (0001) O-2 × 1 surface illustrating the examined forms of oxygen displacement defects in the oxygen overlayer: (a) oxygen displacement to adjacent FCC site, (b) oxygen displacement to far FCC site, (c) oxygen displacement to near HCP site, and (d) oxygen displacement to far HCP site.
relative energies of the displacement configurations trialed. Displacement into empty row HCP sites is the most energetically favored form of defect formation, with an energy cost of only 16 kJ mol−1. The favoritism of the empty row HCP sites is not unexpected: oxygen has been established previously to preferentially bind to the HCP sites, and placing an oxygen in an FCC site brings it particularly close to neighboring oxygen atoms of the ordered layer, which presumably also results in strong nearest-neighbor repulsion. The latter point is underscored by vacancy adjacent FCC sites being only 2 kJ mol−1 higher in energy than the HCP sites. The main difference between adjacent FCC sites and distant FCC sites is the presence of nearest-neighbor oxygens in the distant case. The most elementary means of qualitatively estimating the extent of such disorder is to assume that the energy of defect creation is independent with respect to both the number of defects and where the interstitial surface oxygen finds itself located. Such assumptions should be reasonable in the low defect concentration limit. As the HCP sites have been shown to be favored for oxygen adsorption, it is further assumed that the surface oxygen species only adsorb to these sites. The last approximation is justified given the relative energies of the HCP and FCC sites: empty row HCP sites are the most energetically favored sites to receive a displaced oxygen atom. Whereas FCC sites immediately adjacent to the oxygen vacancy appear to have almost the same energy, the number of such sites will only be proportionate to the number of vacancies and as such have far lower statistical weight than the available HCP sites, legitimizing their exclusion. Similarly those FCC sites not immediately adjacent to point vacancies appear to have far 14370
dx.doi.org/10.1021/jp302729e | J. Phys. Chem. C 2012, 116, 14368−14381
The Journal of Physical Chemistry C
Article
Table 1. Activation Barriers for Oxygen Atom Diffusion on Ru (0 0 0 1) O 2 × 1 Surface diffusion path
activation barrier (kJ mol−1)
ordered → FCC adjacent ordered ← FCC adjacent FCC adjacent → HCP adjacent FCC adjacent ← HCP adjacent
66 47 48 47
Table 2. Adsorption Energies for Formamide on the Ru (0 0 0 1) O-2 × 1 Surface adsorption energy (kJ mol−1)
PW91 exchange correlation functional obtained an activation energy of 52 kJ mol−1,17 of similar order to the values obtained here on the O 2 × 1 covered surface. Whereas no experimental studies capable of atomic resolution have examined oxygen diffusion over the O 2 × 1-covered surface, Scanning tunneling microscopy has been previously used to examine isolated oxygen atom diffusion over an otherwise clean Ru (0 0 0 1) surface.18 It was found that at room temperature oxygen diffuses too quickly to be resolved. Given that STM typically takes a period of order seconds to complete a single examination, this implies a time-average diffusion rate of many steps per second. Invoking classical transition-state theory19 with the assumption that the vibrational partition functions for O in its ordered state and ordered → FCC adjacent transition state completely cancel and using the activation barrier found above, one obtains a time-average diffusion rate of 30 s−1 at 300 K, consistent with the rapid diffusion observed in the experiment. The diffusion rate drops to M7 > M6 > M2. When including the dispersion correction, the order becomes M7 > M3 > M4 > M6 > M5 > M2. The “flat” adsorption modes (M7, M6) show the largest stabilization from the correction because they place all nonhydrogen atoms in formamide (O, C, N) the closest to the surface ruthenium. The strong stabilization of these configurations as a result of applying the dispersion correction is particularly important because the M6 and M7 adsorption modes closely resemble the high-temperature molecular adsorption mode suggested by Parmeter and Weinberg. 3.4. Dissociative Formamide Adsorption. The dissociative adsorption of formamide has also been examined. Dissociative adsorption configurations are shown in Figure 5, whereas bond lengths and angles of the dissociative configurations are given in Table 5. Two configurations were found in the presence of a perfectly ordered oxygen overlayer. In both configurations, formamide was assumed to dissociate a hydrogen from the amino group to the surface. The remainder of formamide was assumed to bind through the O and N atoms of formamide to the exposed ruthenium atoms between rows of coadsorbed oxygens. The configurations differed in the location of the dissociated hydrogen. In the D1 configuration, the dissociated hydrogen was placed on the hcp sites of ruthenium metal. In the D2 configuration, the dissociated hydrogen was placed on the coadsorbed oxygen. A third configuration where hydrogen adsorbed to the fcc sites on ruthenium was also trialed, however the dissociated hydrogen relocated to the hcp sites (adopting the D1 configuration) during geometry optimization. Because adsorption in this manner necessitated that formamide adopt an adsorption configuration between oxygen rows (as opposed to bridging between ruthenium atoms for the carbonyl and surface oxygen atoms for the amino hydrogen), the 2 × 2 cell used in molecular adsorption configurations was found to be too short in the second dimension and surplus to requirements in the first dimension. As such, a 1 × 4 cell was used instead. The D1 configuration was energetically favored over the D2, in agreement with the experimental observation that no OH stretch appears on the vibrational spectra and no water desorbs from the surface. However both trialed configurations were found to have higher energies than the gaseous formamide, suggesting that formamide should preferentially desorb rather than dissociate. This is in clear opposition to experiment, where formamide converts from molecular to dissociative adsorption modes as the temperature is increased above 250 K.2d The introduction of disorder in the overlayer was found to lead to
Figure 5. Formamide dissociative adsorption modes.
disordered areas and that significant adsorption to the perfectly ordered surface simply does not happen. However, it may also be that interactions not captured by the GGA, such as London dispersion interactions, contribute significantly to the adsorption energy. Prior evidence exists to support such a point of view in the case of formamide adsorption. In a study of formamide adsorption to silver, it was found that DFT using a GGA produced lower than expected adsorption energies and that the application of an empirical London dispersion correction was needed to make the computed results conform to experimental expectations.5 To examine the effect of dispersion forces, we have applied the same empirical correction6 to the smallest examined cell size for each molecular adsorption configuration. These adsorption energies are also shown in Table 2. All adsorption configurations show a significant increase in adsorption energy. The correction also changes the order of stability of the various adsorption configurations: without the dispersion correction, the order of molecular adsorbed modes on a 2 × 2 supercell is M3 > M4 >
Table 5. Bond Distances of Formamide Adsorbed to the Ru(0 0 0 1) O2 × 1 Surface in Configurations D1−D4a D1 PBE CO C−N C−H(3) N−H(1) N−H(2) Ru−O Ru−N O(surf)−H(1) O−C−N O−C−H a
D2 PBE-D
PBE
1.278 1.312 1.106
1.278 1.311 1.105
1.278 1.312 1.106
1.021 2.185 2.136
1.02 2.177 2.126
1.021 2.187 2.145 0.986 126.3 115.7
126.2 115.7
126.1 115.8
D3 PBE
D4 PBE-D
PBE
PBE-D
1.26 1.347 1.104
1.261 1.347 1.103
1.284 1.312 1.105
1.285 1.312 1.104
1.024 2.157 2.238, 2.275
1.023 2.148 2.232, 2.262
1.023 2.240, 2.845 2.101
1.022 2.241, 2.808 2.086
123.6 117.8
123.4 118
124.8 116.2
124.7 116.3
PBE indicates the use of standard DFT with no dispersion correction. PBE-D indicates an empirical dispersion correction has been applied. 14374
dx.doi.org/10.1021/jp302729e | J. Phys. Chem. C 2012, 116, 14368−14381
The Journal of Physical Chemistry C
Article
Figure 6. Charge density difference plot showing an isosurface of the change in charge density on formamide adsorption in the M3 configuration. Red areas denote a gain in electron density on adsorption, while blue areas denote a loss. The isosurface was produced at a level of ±0.002 electrons/ Bohr.3
significantly more stable modes of adsorption. Two adsorption configurations were trialed in which a single surface oxygen was removed. Both modes assumed hydrogen dissociated to surface Ru rather than O because ruthenium was shown to be the preferred adsorption site for hydrogen on the perfect surface. The modes were differentiated by whether the N atom or O atom of formamide was directly above the ruthenium atom exposed by the surface oxygen removal. In the case of the D3 configuration, the formamide oxygen was placed above the exposed Ru. The D4 configuration had the formamide nitrogen above the exposed ruthenium. The D3 configuration was marginally more energetically stable. As with the molecular adsorption modes, the dissociative adsorption modes were substantially stabilized by the use of an empirical London dispersion correction. The D1 configuration was no longer less energetically favored than desorption, whereas the D3 desorption configuration had approximately the same adsorption energy as the M3 and M4 molecular adsorption modes. 3.5. Changes to the Electronic Structure on Formamide Adsorption. The M3 adsorption configuration represents the strongest binding configuration without dispersion corrections and remains one of the strongest adsorption configurations after dispersion corrections are added. As such, it will be used to study the nature of formamide adsorption with the surface. A charge density difference plot for formamide adsorbed in the M3 configuration is shown in Figure 6, whereas the electronic density of states for the M3 configuration is shown in Figure 7. Additionally, a Löwdin population analysis23 has been conducted. The total Löwdin population on formamide displays a change of −0.3 electrons when adsorbed as compared with the gas phase. The overall chemical interaction between formamide and the surface is therefore one of charge donation to the surface. The net change in the population of all surface oxygens is +0.17 electrons, hence more than half of the net change transferred to the surface has been accommodated at the surface oxygen atoms. Of this, there is a +0.09 electron gain for the oxygen atom hydrogen bonding to the formamide molecule. The charge density difference plot clearly illustrates this: the largest gain in electric charge is present over the surface oxygen atom. This runs counter to conventional expectations, which would suggest the major charge acceptor is the atom coordinate-
Figure 7. Electronic density of states for formamide adsorbed in the M3 configuration. Solid red line: total DOS; dashed line: formamide partial DOS.
bonded to. Instead, the ruthenium atom coordinate-bonded to has a smaller net gain of +0.05 electrons. To examine the bonding mechanism between ruthenium and the adsorbate, the change in Löwdin populations on the atomic orbitals of the coordinate bonded ruthenium atom will now be examined. The largest change is a loss of population from the dz2 orbital (−0.19 electrons), with the second greatest change being a gain in population on the pz orbital (0.11 electrons). The remainder of the d orbitals show small population increases (0.03 to 0.04 electrons), whereas the other p orbitals show
