DFT Studies of Adsorption of Benzoic Acid on the Rutile (110) Surface: Modes and Patterns
DDFT Studies of Adsorption of Benzoic Acid onthe Rutile (110)
Surface: Modes and Patterns
Xiang Zhao (Ivan) ∗ , † , ‡ and David R. Bowler ∗ , † , ‡ , ¶ London Centre for Nanotechnology, 17-19 Gordon Street, London, WC1H 0AH,Department of Physics & Astronomy, University College London, Gower Street, London,WC1E 6BT, and UCL Satellite, International Centre for Materials Nanoarchitectonics(MANA), National Institute for Materials Science (NIMS), 1-1 Namiki, Tsukuba, Ibaraki305-0044, Japan
E-mail: [email protected]; [email protected] ∗ To whom correspondence should be addressed † LCN ‡ UCL ¶ MANA Satellite, UCL a r X i v : . [ c ond - m a t . m t r l - s c i ] J u l bstract Adsorption of benzoic acid on the (110) surface of rutile, both unreconstructed and (1 × sites has been studied using DFT simulations, as implemented through the Vienna Abinitio Simulations Package (VASP). In order to study the effects of hydrogen bonding andVan der Waals forces in influencing the relative stabilities of different adsorbate overlayersuperstructures, these studies were performed through Local Density Approximation (LDA),Generalized Gradient Approximation(GGA) and DFT-D2. Through the calculations, it wasfound out that although the optimized structures corresponded with the proposed modelsfor the experimental results, the relative energetic stabilities of different overlayer structureshave shown some differences with the experimental results. In the GGA calculations, theoverlayer structures involving the benzene rings aligned along the (001)-direction were shownto be more stable than those aligned perpendicular to it, regardless of whether the benzoateswere arranged in (1 ×
1) or (1 ×
2) symmetries on the (1 × × × Introduction
Titanium dioxide (TiO ) has a history of applications since the early 19th century. Al-though its initial usage was confined to use as a white pigment, its applications diversifiedover the course of the last few decades, including but not limited to biological implants, pho-tocatalysis and dye-sensitized solar cells (DSSCs), based on the refractive, biocompatible,photocatalytic and photoelectric properties of the material. Recent emphasis on the lat-ter applications spurred research work in the surface properties and reactions of crystallineTiO , understanding of which is crucial in their realization.2n the case of DSSCs, this often involves the interactions of these surfaces with sensi-tizing dyes, such as triscarboxy-ruthenium terpyridine [Ru(4,4’,4”-(COOH) -terpy)(NCS) ],the ”black dye”. Recent developments have also been made in bio-sensitized solar cells(BSSCs), which are similar to DSSCs in operating principles, but with the synthetic dyeoften used in the latter replaced by one that is biologically derived. One class of suchbiomolecules are the cyclic tetrapyrroles, which encompass porphyrins and chlorins. Manyof such compounds contain a ring structure, with one or more carboxylic functional groupswhich play importance roles in anchoring the dye to the TiO surfaces, benzoic acid’s adsorp-tion on TiO can thus offer an exemplar case study for how this entire class of compoundscan interact with TiO surfaces.Out of all the TiO surfaces, the rutile (110) is the most energetically stable and naturallyoccurring, and thus forms the substrate surface of choice in such adsorption studies. Therutile (110) surface, though not prone to reconstruction, can undergo (1 ×
2) reconstructionupon annealing under UHV conditions, with the models proposed including added Ti O and Ti O rows. As such, the focus of our studies shall be on the adsorption of benzoic acidon these two types of rutile (110) surfaces, in terms of the binding mode and the overlayerpatterns.Experimental studies have been performed on such adsorptions, using LEED, ESDIADand STM by Guo et al, and later by Grinter et al using STM. In Guo’s studies, theoptimal adsorption mode observed was in line with those of other small carboxylates i.e.dissociative bidentate bridging binding (BB), involving the two carboxyl O atoms, afterdeprotonation on one of them, binding to two neighbouring five-coordinated Ti sites (Ti )along the [001] rows of such sites.In terms of adsorption patterns, through LEED, it was also observed that a p(2 × ×
2) pattern was observed instead. Such discrepancy could be accounted for by thefact that LEED is a scattering technique dependent upon the charge distribution of the3urface, while STM surveys the apex of the surface and the adsorbates, and the fact that thephenyl rings of the benzoates were rotated to align with the [0¯11] direction to form dimerstructures. The rotation of the phenyl rings and the resultant dimer structures are due tothe interactions of the lower phenyl H atoms and the π orbital of the neighbouring benzoate,facilitated by the bridging O anions.Building upon these findings, a later STM study of these adsorptions was performed, with the aims of studying orientations of the benzene rings of the adsorbate overlayer, aswell as the adsorption of benzoates on the (110)-(1 ×
2) reconstructed surface. The bindinggeometry of the acid was once again confirmed, but the elongation of the phenyl ringsalong the [1¯10] direction was only significant in the case of the reconstructed surface, dueto improved hydrogen bonding of the phenyl H atoms to the surface O atoms as a resultof closer proximity. There was also no formation of dimer structures at high coverages asreported in the much earlier study - which could be due to the higher dosing temperatureused this time, nor was there any observation of adsorption of dissociated H + ions. On the(110)-(1 ×
2) reconstructed surface, it was also observed that benzoic acid adsorbs and bindsthe same way as on the unreconstructed surface.Although DFT studies of the adsorption of benzoic acid on the rutile (110) surface hadbeen performed by Troisi et al, these were performed with the aim of comparing the effectsof using different DFT implementations on the computed results on electronic structures ofsuch benzoate-TiO complexes, rather than the physical structure of the adsorbates on thesurface which forms the main theme of experimental research work. In our work, we willexpand upon this topic by investigating the proposed models for surface adsorbate struc-tures and patterns described in the experimental work, using DFT as implemented throughVASP. After presenting our methodology, we calculate and compare the energetics of theseconfigurations obtained using LDA, GGA and vdW, as well as producing high resolutionsimulated STM images to provide comparisons with these results, before concluding.4 ethodology Our DFT investigations were carried out using the Vienna Ab initio Simulations Package(VASP) . Out of all the elemental pseudopotential files, oxygen has the highest E cut valueat 395.700 eV, hence this value was chosen as the cut-off energy E cut in all our simulations.Iterative electronic relaxations are done such that energy differences ( E diff ) between twosuccessive steps should not exceed 10 − eV, while relaxations are said to be achieved whenthe RMS force on each atom falls below 0 .
03 eV/˚A.In our calculations, LDA, GGA(PBE) and GGA+DFT-D2-based methods were used togive three different sets of results for comparison of the effects of hydrogen bonding and Vander Waals forces, as GGA and DFT-D2 methods can account for these forces much better, asthe interactions as a result of these forces were implied in the experimental STM studies. In order to simulate the rutile TiO substrate, an 8 Ti layer slab terminated by two(110) surfaces on each end was chosen, as studies of the rutile (110) surface through DFThave revealed that surface energies converge at around this thickness. This was done afterperforming bulk relaxation of the rutile crystal, where K-sampling values of 12 × ×
16 wereused, in order to reflect the approximate inverse ratio of the dimensions of the real space unitcell. The optimized lattice constants obtained through LDA calculations were found to be a = 4 .
546 ˚A and c = 2 .
925 ˚A for rutile, reflecting a contractions of 0 .
03 ˚A to 0 .
05 ˚A from theexperimental values, which are expected for DFT using LDA as reported in other researchliterature, as well as being within ±
2% error margin. As for GGA based calculations, theoptimized lattice constants were a = 4 .
615 ˚A and c = 2 .
966 ˚A, representing an expansion of0.04 ˚A, which is expected due to GGA’s underestimation of cohesive energies of insulators.The new lattice parameters were then used to set up the slabs for surface relaxations. 8 Tilayer slabs were set up using four such cells with the terminated ends autocompensated. Forthe (1 × O added-row structure as assumed in. We then set the benzoates up such that they correspond to the following six configurations,which involve the phenyl rings being aligned either with the [001] direction(denoted (cid:107) ) or5he [1¯10] direction (denoted ⊥ ), as well as being aligned such that they form (1 × i.p. ), or with the neighbouring [001]-rows of benzoates aligned one Ti step out with respect to each other (denoted o.p. ).The different monolayer patterns are denoted as follows, with the relaxed GGA structuresshown in the relevant figures (LDA and GGA relaxed structures show little different): • BB (cid:107) , i.p. : Phenyl rings of the benzoates are aligned with the [001] direction, with thebenzoates of neighbouring [001] rows in step with each other, as seen in Figure 1 • BB (cid:107) , o.p. : Phenyl rings of the benzoates are aligned with the [001] direction, with thebenzoates of neighbouring [001] one Ti site out of step with each other, as seen inFigure 2 • BB ⊥ , i.p. : Phenyl rings of the benzoates are perpendicular to the [001] direction, withthe benzoates of neighbouring [001] rows in step with each other, as seen in Figure 3 • BB ⊥ , o.p. : Phenyl rings of the benzoates are perpendicular to the [001] direction, withthe benzoates of neighbouring [001] one Ti site out of step with each other, as seenin Figure 4 • BB (cid:107) , × : Phenyl rings of the benzoates are aligned with the [001] direction, using the(1 × • BB ⊥ , × : Phenyl rings of the benzoates are perpendicular to the [001] direction, usingthe (1 × E ads ), theenergies of three simulation cells of the above defined dimensions, consisting of the slab andthe adsorbate ( E slab+mol ), just the slab ( E slab ) and just the adsorbate ( E mol ) in neutral formrespectively, were first calculated, obtaining E ads by:6 ads = E mol + slab − ( E mol + E slab ) (1)As the latest STM studies did not report H adsorption on the protruding bridgingO s, these simulations were also run with the dissociated H removed, with stabilities ofeach mode compared in terms of the total energy of the simulation cell E cell . In addition,the ionic adsorption energies of E ion, ads the different modes will be compared against oneanother, in a similar fashion to (1), with E mol being replaced by that corresponding to theenergy of a single benzoate anion in the simulation cell. This gives us completeness in thecomparisons for adsorption energies.Upon relaxations of these different modes of adsorptions under the above described setups, simulated STM images were then generated based on the electronic structures of therelaxed adsorbate-surface complex, at a bias voltage of +1.5V. Results
LDA Calculations
For the adsorption structures, the optimized structures for the six BB modes correspondedlargely with the models proposed for the STM images, in that the carboxylate groupremained anchored via two Ti -O carboxy bonds, while the benzene rings remained alignedmore or less along the [001] and the [1¯10] directions. Slight deviations were observed in theBB (cid:107) , i.p. and the BB (cid:107) , × modes, in that the benzene rings were not exactly aligned along the[001] directions and instead being slightly rotated about the [110] axis. In addition to theslight rotations, in all of the BB (cid:107) modes, the benzene rings underwent slight planar rotationsabout the [1¯10] axis. These are likely due to the sideways repulsions between the H atomson the benzene rings. 7n terms of adsorption energetics, the adsorption energies ( E ads ) were obtained (with thevalues obtained through GGA calculations) as displayed in Table 1. When the dissociatedHs were removed, ionic adsorption energies ( E ion, ads ) were obtained and presented in Table2. Table 1: Adsorption energies for dissociative adsorptions of benzoic acid on therutile (110) surface, in eV/˚A, as calculated through LDA, with and without theco-adsorbed H.AdsorptionMode E ads ( dis ) E ads ( no H )BB (cid:107) , i.p. (Figure 1) -2.30 -3.93BB ⊥ , i.p. (Figure 2) -3.49 -3.99BB (cid:107) , o.p. (Figure 3) -0.31 -3.01BB ⊥ , o.p. (Figure 4) -1.25 -4.23BB (cid:107) , × (Figure 5) -2.20 -1.94BB ⊥ , × (Figure 6) -2.00 -0.91From the figures for adsorption energetics alone, the LDA results did not completelycorroborate with the STM studies. In the former, it was observed that the BB (cid:107) , i.p. mode should be more stable than the BB ⊥ , i.p. configuration, our current results of DFTcalculations however contradict these. This could possibly be explained by the fact in suchan arrangement, the benzene rings are close to each other, resulting in repulsion betweenthe neighbouring rings, thereby destablizing the configuration. By rotating, this source ofdestablization was removed.In the BB o.p. superstructures, the results showed agreement with the observations thatthe benzene rings would be rotated along the [1¯10] direction, with E ads (BB ⊥ , o.p. ) being0.06 eV more stable than the value for E ads (BB (cid:107) , o.p. ). The difference however, was muchsmaller than expected taking into consideration of the predomination of the BB ⊥ , o.p. patternin the STM images.For adsorption on the reconstructed surface, stable dissociative bidentate bridging ad-sorption of benzoic acid on the surface was observed at the Ti sites between the recon-structed ridges along the [001] directions. The adsorption energy values of the BB (cid:107) , × and8he BB ⊥ , × patterns however, still contradicted the STM observations, with the non-rotatedmodes being almost 0.5 eV more stable than the rotated ones.When the dissociated hydrogens were removed, ionic adsorption energies ( E ion, ads ) wereobtained using the method described in (1) and presented in Table 2. Table 2: Adsorption energies for adsorptions of benzoates on the rutile (110) surface, in eV/˚A, as calculated through GGA, with and without the co-adsorbedH. AdsorptionMode E ads ( dis ) E ads ( no H )BB (cid:107) , i.p. (Figure 1) -2.78 -3.16BB ⊥ , i.p. (Figure 2) -1.29 -1.55BB (cid:107) , o.p. (Figure 3) -1.56 -1.10BB ⊥ , o.p. (Figure 4) -1.31 -0.65BB (cid:107) , × (Figure 5) -1.15 -1.80BB ⊥ , × (Figure 6) -0.70 -1.25When the same simulations were rerun with the dissociated hydrogens removed, how-ever, slightly different pictures emerged as the BB ⊥ , × mode became more energeticallystable than the BB (cid:107) , × mode, as was observed in the STM studies. The BB i.p. modes alsobecame energetically comparable, while the BB (cid:107) , o.p. mode became slightly more energeti-cally favourable than the BB ⊥ , o.p. . This suggests that the presence and absence of H atomshas influences on the overall stabilities of the adsorption structures,even without hydrogenbonding taken into account.The discrepancies between the calculated results and the experimental findings can beattributed to LDA as a method for DFT calculations that does not include hydrogen bonding,as well as π - π interactions, which in this case are those between neighbouring benzene rings.These are interactions that were proposed as significant factors in stablizing the adsorptionstructures, thougheven without being taken into account, rotation of the benzene rings wasseen to have major effects on the energetic stabilities of the adsorptions.9igure 1: The views of the GGA-optimized BB (cid:107) , i.p. mode of adsorption of benzoic acid onthe rutile (110) surface, through the [001], [1¯10] and the 110 directions.Figure 2: The views of the GGA-optimized BB ⊥ , i.p. mode of adsorption of benzoic acid onthe rutile (110) surface, through the [001], [1¯10] and the 110 directions.Figure 3: The views of the GGA-optimized BB (cid:107) , o.p. mode of adsorption of benzoic acid onthe rutile (110) surface, through the [001], [1¯10] and the 110 directions.10igure 4: The views of the GGA-optimized BB ⊥ , o.p. mode of adsorption of benzoic acid onthe rutile (110) surface, through the [001], [1¯10] and the 110 directions. GGA Calculations
In terms of physical structures, all the four modes of adsorption, save the BB (cid:107) , o.p. mode,exhibit adsorption structures close to those following the STM studies. Notable differencesexist between the GGA-optimized adsorption configurations for the BB (cid:107) modes, and theidealized structures proposed to account for the STM observations. In both of the BB (cid:107) modes, buckling of the phenyl rings on the [001] plane was observed. In the BB (cid:107) , o.p. mode inparticular, a rotation of the phenyl ring around the [001] axis was observed. This could havecome about as a result of the closer distances between the neighbouring phenyl rings of theadjacent benzoates along the [001] rows, resulting in steric repulsions between the phenylH atoms. In the BB ⊥ modes, such deviations from the idealized structures as proposedto model the STM studies, with the phenyl rings almost exactly aligned along the [1¯10]direction. This can be due to the fact that the phenyl rings of the neighbouring benzoatesalong the [1¯10] direction are further apart.One particular interesting observation was the tilting of the H atoms in the cases ofBB (cid:107) , o.p. , BB ⊥ , i.p. and the BB ⊥ , o.p. modes, towards the carboxyl O atoms of the respectivemodes, which were not observed in the LDA calculations. This is possibly due to the hy-drogen bonding between the deprotonated H, and the carboxyl O. Such significant tilting,however, was not observed for the BB (cid:107) , i.p. mode. The tilting observed in the former three11odes, could be attributed towards the hydrogen bonding interactions between the depro-tonated H, and the carboxyl O atoms. This however raises the question of why the samehydrogen bonding could not produce the same tilt on the H atom. This is possibly due tothe fact that in the BB (cid:107) , i.p. mode, the almost perfectly [001]-aligned phenyl rings, stericallyhinders such an interaction through interactions between the rings’ own H atoms and thecarboxyl oxygens.When the dissociated hydrogens are stripped from the surfaces, little structural changeswere observed in all the modes mentioned. This was also true when Van der Waals’ forceswere taken into account when DFT-D2 was turned on, as the phenyl rings were observedto retain their orientations once the calculations attained relaxation. These thereby suggestthat neither the presence or absence of H on the O sites had had significant effects on thephysical orientations of the phenyl rings.In terms of the adsorption energetics, the GGA-based calculations have shown that theadsorption energies have decreased from those of over 2 eV/benzoate, to values much closerto 1.5 eV, ,
19 20 and. In terms of the comparative energetic stabilities of the differentadsorption modes, the GGA calculations have revealed that the BB (cid:107) modes were more ener-getically stable, than the BB ⊥ modes, in all the three pairs of modes. This is in clear contrastto the LDA results, where the BB ⊥ modes were the more energetically stable configurationsout of all the three pairs. Part of the reason why in GGA the BB (cid:107) modes were more energet-ically stable, lies in the fact that now the inter-benzene ring distance between neighbouringrings along the [110] direction was increased, resulting in less repulsion between them.For the case of BB i.p. modes, the results agree with the conclusions drawn from STM,that the BB (cid:107) , i.p. mode is the more commonly observed mode in the (1 × . ⊥ , i.p. mode was calculated to be much moreenergetically stable (by 1.2 eV). The comparative greater energetic stability in the BB (cid:107) , i.p. mode can now be described in terms of greater spacing between the neightbouring benzene12ings, being increased by nearly 0 . (cid:107) and BB ⊥ modes, although the adsorption energy values werecloser to the above range, they did not agree with the models proposed in the experimentalstudies, in which both the BB o.p. and the BB × modes had the BB ⊥ modes were describedto be the predominant form observed in STM. This is however explainable by the presenceof co-adsorbed H on the O on the [001] ridges, which serve to repel H atoms on the rotatedbenzene rings.Figure 5: The views of the GGA-optimized BB (cid:107) , i.p. mode of adsorption of benzoic acid onthe rutile (110) − (1 ×
2) reconstructed surface, through the [001], [1¯10] and the 110 directions.When the acid was dissociatively adsorbed onto the rutile (110) in the manners of BB o. p. modes, similar to the results produced in LDA, the difference in terms of the energeticstabilities of this pair of BB (cid:107) and BB ⊥ modes was the narrowest of the three at 0 .
25 eV,though with the BB (cid:107) , o. p. mode being more energetically stable this time. This gap is also13igure 6: The views of the GGA-optimized BB ⊥ , i.p. mode of adsorption of benzoic acid onthe rutile (110) − (1 ×
2) reconstructed surface, through the [001], [1¯10] and the 110 directions.14arger than that reported for the corresponding pair in the LDA calculations, where theBB ⊥ , o. p. mode was more energetically stable by just 0 .
05 eV/molecule. The GGA resultssuggest that rotation of the benzene rings by 90 ◦ in order to avoid the steric repulsionbetween the rings along the [001] directions, as well as the weak H interactions between thebenzyl H and the O of the [001] rows, do not energetically stabilize the structure, and thatrotation of benzene rings themselves can serve as an energetically destabilizing factor.For the adsorption on the reconstructed surface, while the BB (cid:107) , × mode was calculated tobe the more energetically stable, as compared to the BB i.p. modes, the energetic favourabilitywas much lower at 0 .
25 eV/molecule. Furthermore, as compared to the corresponding pairof BB × modes relaxed in the LDA calculations, which showed that the BB ⊥ , × mode wasmore energetically stable by 0 . O reconstructedridges. This shows that the presence of H atoms on the O ridges does not affect hydrogenbonding interactions between most parts of the surface and any part of the benzoate. Interms of ionic adsorption energetics themselves, the comparative energetic stability of theBB (cid:107) over the corresponding BB ⊥ mode was confirmed, while in all of the cases, adsorptionof benzoate ions without the co-adsorption of the dissociated H had a greater energeticallystabilizing effect than dissociative co-adsorption of the both ions on the same surface withthe same geometry.One interesting thing to note is that while for the BB i. p. and BB o. p. modes, the E ads ( noH ) values are more , while that for the BB × modes the E ads ( no H ) values appeared tobe far larger than the corresponding E ads values, over twice values of the E ads obtainedfor dissociative co-adsorption. As discussed earlier, a partial restoration of symmetry ofthe Ti O reconstructed ridges was observed following the removal of the dissociated H,15nd this perhaps provided the additional energetic stabilization. However, due to the largeincrease in the values of E ads involved, and the values for E ads ( no H ) differed little regardlessof simulation cell used and the same E slab values were used for both the dissociative co-adsorptive and the ionic adsorptive cases, further investigations shall be needed to determinethe exact cause of this unusually large increase in the adsorption energy when the dissociatedhydrogens are not taken into account. GGA-DFTD2 Calculations
Although Van der Waals’ forces are weak in comparison to hydrogen bonding when comesto inter-ionic interactions between adsorbates, they can come into play when large non-polarcomponents come close to each other, and especially so in the case of phenyl rings withextensive π -orbitals being close with each other, as seen in Figures 1-6. Indeed, takinginto account of Van der Waals’ forces through GGA+DFTD2 calculations has improved theenergetic stabilities for all six different modes, each being at least 0.7 eV/molecule morestable after implementation of GGA-DFTD2 instead of just GGA (see Table 3). In spiteof these changes, the BB (cid:107) modes are still more energetically stable compared to their BB ⊥ counterparts, when the co-adsorbed hydrogens are taken into account.When just considering the case of ionic adsorption, the changes in E ads values becomeeven more pronounced, and especially so in the case of benzoates’ adsorptions on the (1 × ads values abnormally large, but also that theBB ⊥ , × mode has now become significantly more energetically stable as compared to theBB (cid:107) , × mode (-7.87 vs -5.14 eV/benzoate). These anomalies shall be discussed in furtherdetail in the ´’Discussions´’ section. Simulated STM Results
In the experimental STM studies, the STM images produced presented the sites of benzoateadsorption as bright spots, elongated along the directions of the alignments of the phenyl16 able 3: Adsorption energies for ionic adsorptions of benzoates on the rutile (110) surface, in eV/˚A, calculated using GGA+DFTD2, with ”*” representinganomalies in values.AdsorptionMode E ads ( dis ) E ads ( no H )BB (cid:107) , i.p. (Figure 1) -3.55 -3.95BB ⊥ , i.p. (Figure 2) -2.12 -2.52BB (cid:107) , o.p. (Figure 3) -2.48 -2.78BB ⊥ , o.p. (Figure 4) -2.23 -2.55BB (cid:107) , × (Figure 5) -2.04 -5.14*BB ⊥ , × (Figure 6) -1.72 -7.87*rings, with no direct visual evidence of co-adsorbed hydrogens, except for low coverageson the (1 × rows, with circular-shaped dimspots representing the 1s orbitals of H and dumbell-shaped ones representing the 2p orbitals17f the O atoms.In the BB i. p. modes, whose simulated STM images are as shown in Figure 7, revealedorientations of rectangular rows of bright spots, with their widths aligned along the planesof the benzene rings determined through DFT calculations. The patterns of the bright spotsrepresenting the benzene rings however differed between the BB (cid:107) , i. p. and the BB ⊥ , i. p. modes,with the former as parallel trios of bright spots along the [001]-direction, and the latter astwo central bright spots sandwiched by two larger merged dimmer spots aligned along the[1¯10]-direction.In the BB o. p. modes, the simulated STM images generated are as shown in Figure 8. Thebright spots representing the (2 ×
2) overlayer symmetry can be observed in both cases, inagreement with STM results obtained from experimental studies, where such features wereobserved in some regions alongside the bright dots representing the (1 ×
1) symmetry on theunrecontructed surface. The orientations of the bright spots again reflected the orientationsof the benzene rings as shown in Figures 3 and 4. However, whereas the benzene ring wasimaged as duo rows of bright spots in the BB (cid:107) , i. p. mode, in the BB ⊥ , i. p. mode the darkerspots around the central bright spots merged, forming kidney-shaped spots.For the case of BB × modes, in both cases the reconstructed Ti O ridge appeared aslarge bright bands along the [001] directions, with features showing the ridges being largelyindistinguishable. Again, although the double rows of three bright spots were observed whenthe ring was aligned along the [001] direction, the dimmer pairs of spots on the sides of thecentral bright spots became merged in the case of the benzene ring being aligned with the[1¯10] direction. Discussions
For adsorption energetics, GGA and GGA+DFTD2 simulations have shown the BB (cid:107) , i. p. to be the most energetically stable out of all the BB modes of adsorptions on the unrecon-18igure 7: Simulated STM images for the BB (cid:107) , i. p. (left) and BB ⊥ , i. p. (right) modes, at high(top) and low (bottom) charge valued isosurfaces, i.e. high and low tunneling current,revealing the positions of the benzoates (top) and the orientations of the benzene rings(bottom) respectively. 19igure 8: Simulated STM images for the BB (cid:107) , o. p. (left) and BB ⊥ , o. p. (right) modes, at high(top) and low (bottom) charge valued isosurfaces, revealing the positions of the benzoates(top) and the orientations of the benzene rings (bottom) respectively.20igure 9: Simulated STM images for the BB (cid:107) , × (left) and BB ⊥ , × (right) modes, in thelimit of low charge isosurfaces, revealing the orientations of the benzene rings.structed surface, when co-adsorbed with the dissociated hydrogens. Being the most observedpattern in the experimental STM studies of the unreconstructed surface, this is an expectedresult. In terms of the simulations themselves, this shows that when hydrogen bondingand/or Van der Waals’ forces are taken into account, the BB (cid:107) , i. p. mode is the most stable.Examining the three pairs of BB (cid:107) and BB ⊥ modes of adsorption, in GGA and GGA+DFTD2-based calculations, it is clear that on the non-reconstructed surface, simulations have shownthat for both (1 × × (cid:107) modes were shown to bemore energetically stable in both cases. These observations, in both the experimental andcomputational DFT studies, can be accounted for by the long distances between the O atoms and the nearest phenyl H atoms when the benzene ring is rotated, making hydrogenbonding between the two too weak to be a significant stablizing factor.In terms of the (2 × (cid:107) , o. p. modeswere shown to be more energetically stable through GGA and GGA+DFTD2 calculations.This again is in line with the experimental observations, which revealed alignment of thebenzene rings with the [001] direction. The proposed reasons for such observations wereagain that the nearest distances between the O atoms and the phenyl H atoms being toofar apart to make hydrogen bonding a significant stablizing force. In comparison the BB (cid:107) , i. p. (cid:107) , o. p. mode appeared less energetically stable (see Figures 1 and2). This again agrees with the findings of experimental STM studies, where such patternswere comparatively rare relative to the (2 × × (cid:107) , × configuration that was moreenergetically stable in all the three pairs of simulations, instead of the [1¯10]-aligned benzenerings as observed in the experimental STM images. The likely explanation for this werethe asymmetric distortions in the Ti O reconstruction ridges produced in the optimizedstructures, with the predicted energetic stablization effects from hydrogen bonding betweenthe O atoms on the reconstructed ridges and the nearest phenyl H being less than predicted.Perhaps the single greatest anomaly in the results lies in the abnormally large ionicadsorption energy values for benzoates on the 1 × i. p. , BB o. p. and BB × configurations.For both the BB i. p. and the BB o. p. modes, in the limit of sufficiently low valued chargeisosurfaces such that the benzene rings start to be imaged distinctively, the co-adsorbed Hatoms lose their distinguishability from their surroundings, whereas in the limit of isosurfacesof high charge values, the vacant O sites could be distinguished from the ones bound bydissociated H by the 2p orbitals imaged. The simulated STM images in the low charge valuelimits reveal features which are consistent with those observed in the experimental STM22mages produced of benzoates on the unreconstructed surface, most notably the elongationsof the bright regions along the directions of alignments of the benzene rings, as well as theabsence of features reflecting the co-adsorption of dissociated H. This suggests that perhapseven if STM images do not directly image the co-adsorbed H ions, it does not mean thatsuch co-adsorption did not take place.On the reconstructed surfaces, the simulated STM images revealed central bright bandsrepresenting the Ti O reconstructed rows similar to those observed in the experimentalSTM images, and that no distinguishable features representing co-adsorption of dissociatedH ions. This is however still in agreement with experimental results as co-adsorbed H ionswere not imaged at high coverages. Again, the lack of direct visual evidence for co-adsorptionof dissociated protons does not necessarily imply complete absence of it. Conclusions
We have studied the different patterns for the bidentate bridging dissociative adsorptionof benzoic acid on both the unreconstructed and (1 × ×
1) and (2 ×
2) symmetries. In all the three DFT methods studied, both withand without the dissociated H, the BB (cid:107) , i. p. pattern was invariably found to be the mostenergetically stable out all the six BB patterns of adsorption. These results support STMstudies which revealed that on the unreconstructed (110) surface, where the prevalentpattern was that of BB i. p. , with the orientation of the benzene rings aligned along the [001]direction.In the case of (1 × (cid:107) , × mode was found to bemore stable, in contrast to the experimental STM studies’ results where the benzene ringswere observed to have rotated by 90 ◦ , with hydrogen bonding between the phenyl hydrogensand the O atoms along the reconstructed ridges being the reason for the stablization of23his orientation.In the case of simulated STM images, the same overlayer patterns of bright dots wasreproduced for the proposed models, elongations of the bright spots representing the align-ments of benzene rings were observed, which were not obvious in the case of experimentalSTM studies. These alignments were only deduced by line profiles taken along the [001] andthe [110] directions in the experimental STM studies. In the different limits of high andlow-valued charge isosurfaces, the simulated STM images have also revealed the carboxylateand the benzene ring groups respectively, while in experimental STM images, these struc-tures were not clearly distinguished in the bright spots representing the sites of benzoateadsorption. Acknowledgement
We thank Dr Geoff Thorton from the Chemistry Department of University College Lon-don for providing the model for the rutile (110)-(1 × References (1) Diebold, U.
Surf. Sci. , , , 53-229(2) Wang, Z-S, Yamaguchi, T.; Sugihara, H.; Arakawa, H. Langmuir , , , 4272-4276(3) Thavasi, V.; Lazarova, T.; Filipek, S.; Kolinski, M.; Querol, E.; Kumar, A.; Ramakr-ishna, S.; Padr´os, E.; Renugopalakrishnan, V. J. Nanosci. Nanotechnol. , , ,1-9(4) Fujishima, A.; Honda, K. Nature 1972, , 37-38245) Fujishima, A.; Zhang, X.; Tryk, D. Surf. Sci. Rep. (6) Md. K. Nazeeruddin.; P. Pechy.; M. Gr¨atzel.
Chem. Commun. , 1997, 1705-1706(7) Lundqvist, M. J.; Nilsing, M.; Lunell, S.
J. Phys. Chem. B , , , 20513-20525(8) Perron, H.; Domain, C.; Roques, J.; Drot, R.; Simoni, E.; Catalette, H. Theor. Chem.Acc. , , , 565-574(9) Onishi, H.; Iwasawa, Y. Surf. Sci. , , , L783-L789(10) Fukui, K-I.; Onishi, H.; Iwasawa, Y. Surf. Sci. , , , L783-L789(11) Fukui, K-I.; Onishi, H.; Iwasawa, Y. Chem. Phy. Lett. , , , 296-301(12) Wendt, S.; Schaub, R.; Matthiesen, J.; Vestergaard, E. K.; Wahlstr¨om, M. D.; Ras-mussen, P.; Thostrup, P.; Molina, L. M.; Lægsgaard, E.; Stensgaard, I.; Hammer, B.;Besenbacher. Surf. Sci. , , , 226-245(13) Guo, Q.; Cocks, I.; Williams, E. M. Surf. Sci. , J. Phys. Chem. C , , 10201026(15) Martsinovich, N.; Jones, D. R.; Troisi, A. J. Phys. Chem. C , , 22659-22670(16) Phys. Rev. B , , , 11169(17) J. Comput. Chem. , , , 20442078(18) K¨ackell, P.; Terakura, K. Surf. Sci. , , , 191-198(19) K¨ackell, P.; Terakura, K. App. Surf. Sci. , , , 370-375(20) Fukui, K-I.; Onishi, H.; Iwasawa, Y. Appl. Suf. Sci. , , 259-264(21) Sasahara, A.; Uetsuka, H.; Onishi, H. Appl. Phys. A , ,
72 [Suppl.]72 [Suppl.]