Azobenzene versus 3,3',5,5'-tetra-tert-butyl-azobenzene (TBA) at Au(111): Characterizing the role of spacer groups
Erik R. McNellis, Christopher Bronner, Jorg Meyer, Martin Weinelt, Petra Tegeder, Karsten Reuter
((submitted to PCCP)
Azobenzene versus 3,3’,5,5’-tetra- tert -butyl-azobenzene (TBA) at Au(111):Characterizing the role of spacer groups
Erik R. McNellis, Christopher Bronner, J¨org Meyer, Martin Weinelt,
2, 3
Petra Tegeder, and Karsten Reuter
1, 4 Fritz-Haber-Institut der Max-Planck-Gesellschaft, Faradayweg 4-6, D-14195 Berlin (Germany) Fachbereich Physik, Freie Universit¨at Berlin, Arnimallee 14, D-14195 Berlin, (Germany) Max-Born-Institut, Max-Born-Str. 2A, D-12489 Berlin (Germany) Department Chemie, Technische Universit¨at M¨unchen, Lichtenbergstr. 4, D-85747 Garching (Germany)
We present large-scale density-functional theory (DFT) calculations and temperature programmed desorptionmeasurements to characterize the structural, energetic and vibrational properties of the functionalized molec-ular switch 3 , (cid:48) , , (cid:48) -tetra- tert -butyl-azobenzene (TBA) adsorbed at Au(111). Particular emphasis is placedon exploring the accuracy of the semi-empirical dispersion correction approach to semi-local DFT (DFT-D) inaccounting for the substantial van der Waals component in the surface chemical bond. In line with previous find-ings for benzene and pure azobenzene at coinage metal surfaces, DFT-D significantly overbinds the molecule,but seems to yield an accurate adsorption geometry as far as can be judged from the experimental data. Com-paring the trans adsorption geometry of TBA and azobenzene at Au(111) reveals a remarkable insensitivity ofthe structural and vibrational properties of the − N = N − moiety. This questions the established view of the roleof the bulky tert-butyl-spacer groups for the switching of TBA in terms of a mere geometric decoupling of thephotochemically active diazo-bridge from the gold substrate. I. INTRODUCTION
The ultimate goal of nanotechnology is control overmolecular-scale mechanical and electronic components. Anoft-proposed route to this goal is the construction of com-ponents from controllable single molecules. An importantclass of such molecules with obvious applications is formedby those that have properties that are reversibly and bi-stablymodifiable by external stimuli – so-called molecular switches .One example is the azobenzene molecule (H C -N = N-C H ),which can be bi-stably photo-isomerized between its planar, C h symmetric trans and torsioned-twisted, C cis conform-ers in both solution and gas-phase. The high yield and sta-bility of this reaction have rendered azobenzene an archetypeof molecular switch research with proposed technical appli-cations including e.g. light-driven actuators and informationstorage media .For many such applications, switching of molecules at solidinterfaces – adsorbed at metal surfaces, for example – isof particular interest. Unfortunately however, the switchingproperties of azobenzene have proven highly sensitive to theadsorbate-substrate interaction: Even at nearly chemically in-ert close-packed noble metal surfaces, switching of surface-adsorbed azobenzene by light has never been achieved, andswitching by excitation with a scanning tunneling microscope(STM) tip has been successful only at Au(111) . A naturalroute to restoring the adsorbate switching ability is to fur-ther decouple the frontier π , n and π ∗ orbitals responsible for the gas- or liquid phase photo-isomerization from thesubstrate electronic structure. With these frontier orbitalslargely located at the central diazo ( − N=N − ) bridge an intu-itive idea to achieve such a decoupling is to functionalize themolecule with bulky spacer groups that prevent a closer en-counter of the photochemically active unit with the substrate.This is precisely the notion behind the arguably to date moststudied such adsorbate, 3 , (cid:48) , , (cid:48) -tetra- tert -butyl-azobenzene(TBA) . TBA consists of azobenzene functionalized with FIG. 1: Perspective views of trans and cis azobenzene and TBA,together with an illustration of the diazo-bridge bond length d NN andthe dihedral CNNC angle ω . The latter is defined as the smallest an-gle between two planes spanned by the − N = N − bridge and − C − N − and − N − C − bonds to the first and second phenyl-ring, respectively.The C atoms of the phenyl-rings have been darkened to allow easierdistinction of azobenzene backbone and functional butyl groups inTBA. four tert -butyl (-C-(CH ) ) groups in the phenyl-ring meta po-sitions as illustrated in Fig. 1. These ’table legs’ were indeedfound to enhance the switching efficiency of the adsorbedspecies, as e.g. indicated by the successful TBA switchingby light at Au(111) . However, attempts at STM-tip in-duced switching at ostensibly comparable substrates such asAg(111) and Au(100) have been unsuccessful, indicatingthat TBA is not significantly more robust to specifics of thesubstrate interaction than pure azobenzene.These circumstances beg the question, how and to what de-gree the TBA butyl groups really ’decouple’ the photochromicmoiety from the substrate. For a corresponding atomic-scaleunderstanding the detailed characterization of the adsorbategeometry and binding constitutes a prerequisite and is the a r X i v : . [ c ond - m a t . m t r l - s c i ] F e b main objective of the present contribution. In contempo-rary surface science, corresponding analyses are increasinglyperformed by quantitative first-principles electronic structurecalculations. Particularly density-functional theory (DFT)with present-day local or semi-local exchange-correlation (xc)functionals has developed into an unparalleled workhorse forthis task, with often surprising accuracy particularly with re-spect to structural properties of the surface adsorption system.For TBA at Au(111) corresponding calculations are alreadychallenged by the extension of the functionalized moleculeand the simultaneous necessity to describe the metal bandstructure within a periodic supercell approach. On a more fun-damental level, the real limitation comes nevertheless fromsizable dispersive van der Waals (vdW) contributions to thesurface chemical bond as characteristic for organic moleculescontaining highly polarizable aromatic ring systems . Withlocal and semi-local xc functionals inherently unable to ac-count for such contributions skewed, if not qualitativelywrong results must therefore be suspected. We have recentlyquantified this for azobenzene at the close-packed coinagemetal surfaces . Comparison to detailed structural andenergetic reference data from normal-incidence x-ray stand-ing wave (NI-XSW) and temperature programmed desorp-tion (TPD) measurements for azobenzene at Ag(111) demon-strates that the prevalent semi-local DFT xc treatment leadsindeed to a significant underbinding with key structural pa-rameters deviating by more than 0.5 ˚A .For system sizes as those implied by the adsorptionof azobenzene or TBA an appealing and computationallytractable possibility to improve on this situation is a semi-empirical account of dispersive interactions within the frame-work of so-called DFT-D schemes . In this approach thevdW interactions not described by present-day xc function-als are approximately considered by adding a pairwise inter-atomic C R − term to the DFT energy. At distances belowa cut-off, motivated by the vdW radii of the atom pair, thislong-range dispersion contribution is heuristically reduced tozero by multiplication with a short-range damping function.While the applicability of this approach to adsorption at metalsurfaces is uncertain ( vide infra ), our recent benchmark studyfor azobenzene at Ag(111) revealed that in particular themost recent DFT-D scheme due to Tkatchenko and Scheffler(TS) yields excellent structural properties, albeit at a notableoverbinding .In this contribution we further explore the generality ofthis finding by analyzing the adsorption geometry, vibra-tions and energetics of TBA at Au(111). Comparisonagainst our recent near edge x-ray absorption fine structure(NEXAFS) and high-resolution electron energy loss spec-troscopy (HREELS) measurements, as well as a completeanalysis of new TPD data confirms the accurate structural andvibrational predictions reached by the DFT-D TS scheme. Theagain obtained significant overbinding furthermore supportsthe interpretation that the neglect of metallic screening isthe main limitation in the application of this scheme to the ad-sorption of organic molecules at metal surfaces. Comparingthe TBA data to those for pure azobenzene at Au(111) we finda qualitatively different adsorption geometry for the functional backbone in the case of the cis isomer. For the trans isomer,on the other hand, we determine an intriguing structural andvibrational insensitivity of the photochemically active centraldiazo-bridge to the presence of the bulky spacer groups. Therole of the latter for the switching efficiency is therefore moresubtle than simple geometric decoupling and will be the topicof continuing work in our groups. II. METHODA. Theory
The DFT-D methodology followed in this work corre-sponds exactly to that employed in our preceding work for theadsorption of azobenzene. We therefore restrict ourselves hereto a brief account and refer to our previous publications for an in-depth description of the underlying concepts andtechnical details.The DFT calculations were performed using a plane-wavebasis set and library ultrasoft pseudopotentials as imple-mented in and distributed with the CASTEP code. The gen-eralized gradient approximation (GGA) to the xc-functionalwith the parametrization suggested by Perdew, Burke andErnzerhof (PBE) was used throughout. In the spirit of theDFT-D approach, the lack of vdW interactions in this semi-local functional is approximately corrected with an additionalanalytical, two-body inter-atomic potential. At long range,this potential equals the leading C R − -term of the Londonseries, where R is the inter-atomic distance, and C a so-calleddispersion coefficient. At short range, this long range poten-tial is matched to the DFT inter-atomic potential by a damp-ing function f ( R , R ) , typically modulated by the vdW radii R of the atom pair. In this work we use the material-specific C and R parameters, as well as the damping function formsuggested by Tkatchenko and Scheffler (henceforth denom-inating corresponding results as PBE+TS). This scheme ac-counts to some degree for the bonding environment througha Hirshfeld-analysis based adjustment of the C parameters,which in our previous work on azobenzene at coinage metalsurface gave a superior performance compared to other DFT-D schemes .The analytical form of the dispersion correction poten-tial brings the advantage that dispersion-corrected total ener-gies of geometries that are fully relaxed with respect to thedispersion-corrected forces can be obtained (employing an in-house extension to the CASTEP code) at essentially the samecomputational cost as a regular GGA calculation. On the otherhand the assumption of a simple two-body form for the dis-persion potential inherently neglects the effect of electronicscreening of the vdW interactions , which particularly for thehere studied adsorption at metal surfaces is expected to leadto an overestimation of the binding energy . A second limi-tation of the semi-empirical DFT-D approach might arise foradsorbate molecules which also interact covalently with thesubstrate. This typically leads to molecule-substrate bond dis-tances that are so short that the uncertainties in the heuristicdamping function of the dispersion term may mingle in an un-controlled way with deficiencies of the employed semi-localDFT xc functional.The surface calculations were performed within supercells,using (111)-oriented metal slabs with ( × ) surface unit-cellsand at least 18 ˚A vacuum. We verified that lateral interactionsbetween adsorbed TBA and its periodic images are negligbleat the GGA-PBE level. In the dispersion correction poten-tial corresponding interactions were also switched off, so thatour calculations should give a fair account of TBA adsorptionin the low-coverage limit. As in our preceding work we ne-glected the subtle effects of the long-range Au(111) surfacereconstruction. Full geometry optimizations (to within resid-ual forces below 30 meV/ ˚A) of all molecular degrees of free-dom were correspondingly performed with TBA adsorbed onone side of a four layer bulk-truncated slab. Test calculationswith appropriately saturated tert -butyl groups indicated onlya weakly expressed site specificity of these TBA functionalgroups at Au(111). This suggests that the photochemicallyactive diazo-bridge plays a prominent role in anchoring themolecule at the metal substrate. The geometry optimizationswere correspondingly initiated with the TBA diazo-bridge lat-erally placed as in the previously determined optimal adsorp-tion geometry of azobenzene, corresponding to a 1:1 N-metalatom coordination . Harmonic vibrational spectra of the ad-sorbed species in these relaxed geometries (and in the gas-phase) were subsequently obtained from numerical Hessianscalculated by finite differences and neglecting any degrees offreedom of substrate atoms. To efficiently parallelize over the432 displacements required for each spectrum, we have inter-faced with the Atomistic Simulation Environment includingthe ’phonopy’ extension to it . Not aiming to reproduce theHREELS intensities , simple Lorentzian broadening with awidth of 1 meV was applied to the spectra for better visual-ization.For the energetic and electronic structure calculations thethus determined relaxed adsorbate geometries were invertedin the bottom layer, forming inversion-symmetric seven layerslabs with adsorbates at both sides. This zeroes the internaldipole moment of the slab and results in a substantially im-proved substrate electronic structure. The two central ener-getic quantities obtained with the resulting structures are theadsorption energy E ads = (cid:2) E azo@ ( ) − E ( ) (cid:3) − E azo ( gas ) , (1)and the relative stability of adsorbed cis (C) and trans (T) con-former ∆ E C − T = (cid:2) E azo@ ( ) ( C ) − E azo@ ( ) ( T ) (cid:3) . (2)Here E azo@ ( ) is the total energy of the relaxed, double-sided azobenzene-surface system, E ( ) the total energy ofthe clean slab, and E azo ( gas ) the total energy of the corre-spondingly relaxed gas-phase isomer (computed within Γ -point sampled (
35 ˚A ×
45 ˚A ×
35 ˚A ) supercells). Where ap-plicable, TS DFT-D corrections calculated in optimized fourlayer slab and gas-phase geometries are added to E azo@ ( ) and E azo ( gas ) , respectively. Both quantities were also con-sistently zero-point energy corrected with the previously ob-tained vibrational frequencies. In the convention of Eq. (1)the adsorption energy of either cis or trans isomer at the sur-face is thus measured relative to its stability in the gas-phaseat both pure PBE and dispersion corrected PBE+TS levels oftheory, and a negative sign indicates that adsorption is exother-mic. Consistently, a negative sign of ∆ E C − T indicates a higherstability of the cis isomer. Convergence tests show that atthe employed plane wave cutoff of 450 eV and ( × × ) Monkhorst-Pack (MP) grid both energetic quantities are nu-merically converged to within ±
30 meV.
B. Experiment
The TPD measurements were carried out under ultrahighvacuum conditions. The Au(111) crystal was mounted ona liquid nitrogen cooled cryostat, which in conjunction withresistive heating enables temperature control from 90 K to750 K. The crystal was cleaned by a standard procedure ofAr + sputtering and annealing. The TBA was dosed by meansof a home-built effusion cell held at 380 K at a crystal temper-ature of 260 K. In the TPD experiments, the substrate was re-sistively heated with a linear heating rate of 1 K/s and desorb-ing TBA was detected with a quadrupole mass spectrometer atthe TBA-fragment mass of 190 amu (3,5-di- tert -butyl-phenylion). This procedure was repeated for different dosing timescorresponding to different initial TBA-coverages.As further discussed below, three desorption features ( α – α ) are observed in the TPD. They are assigned to desorp-tion from the multilayer ( α ) and the first monolayer (ML)( α + α ), where α represents the desorption of ≈
10% ofthe monolayer coverage (for details see Fig. 5 and Ref. 10).The NEXAFS and HREELS measurements of Refs. 24,25discussed in Section III were performed at a coverage of 1.0and 0.9 ML, respectively, which were prepared by heating themultilayer-covered surface to 340 K or to 420 K.
III. RESULTS AND DISCUSSIONA. Adsorption Geometry
Our previous work on azobenzene at coinage metal surfacespoints to an understanding of the surface chemical bond interms of a balance of four major contributions: A covalentbond between the diazo-bridge and the metal, the vdW attrac-tion between the metal and the phenyl-rings, the Pauli repul-sion between the phenyl-rings and the metal, and the ener-getic penalty due to the distortion of the gas-phase moleculargeometry . This understanding should largely carry overto TBA at Au(111), with the bulky tert-butyl groups partic-ularly adding to the vdW attraction and the molecular defor-mation upon adsorption. Key structural parameters to char-acterize the adsorption geometry for both trans and cis iso-mer are therefore the location and orientation of the centraldiazo-bridge, as well as the orientation of the planes of the two
FIG. 2: Side view of adsorbed cis
TBA at Au(111), illustrating keystructural parameters defining the adsorption geometry (see text):The vertical height z i of the two diazo-bridge N atoms, as well asthe out-of-horizontal phenyl plane bend angles ˜ ω i . phenyl-rings with respect to the surface. As further illustratedin Figs. 1 and 2 we concentrate on the − N=N − bond length d NN and the vertical N atom − surface plane distances z i forthe prior. Here, the indices i = i = z = z reflects an diazo-bridge that is oriented parallel to the surface,as we obtain for the more symmetric trans isomer through-out. To specify the position of the phenyl-rings we focus onthe CNNC dihedral angle ω and the out-of-horizontal bendangles ˜ ω i of the two ohenyl planes, with the same conventionfor the index i = , ω = ◦ indicates a planar TBA molecule,while ˜ ω i = ◦ denotes a phenyl-ring that lies parallel to thesurface plane.Table I summarizes the detailed geometric parameters ob-tained from our calculations with the exception of the diazo-bridge bond length. For the latter we consistently computeonly insignificant changes away from the TBA gas-phasevalue of 1.29 ˚A ( trans ) and 1.28 ˚A ( cis ) at both PBE andPBE+TS level of theory. While this indicates an overall mi-nor activation of the − N=N − bond upon adsorption, it alsodemonstrates an insensivity to the degree of vdW interactionsaccounted for in the calculations. Starting the more detailedpresentation with the trans isomer, Fig. 3 displays a perspec-tive view of the overall TBA adsorption geometry. In linewith the interpretation of STM , surface vibrational andNEXAFS data, the central azobenzene moiety essentiallymaintains its gas-phase planarity, with its long axis aligned toabout 5 ◦ along the direction of close-packed atom rows on the Trans @ Au(111) z = z ( ˚A) ω ( ◦ ) ˜ ω = ˜ ω ( ◦ )TBA (PBE) 3.97 172 9TBA (PBE+TS) 3.22 169 12TBA (Exp.) − − ± Cis @ Au(111) z ( ˚A) z ( ˚A) ω ( ◦ ) ˜ ω ( ◦ ) ˜ ω ( ◦ )TBA (PBE) 3.25 3.74 10 21 79TBA (PBE+TS) 2.34 2.85 8 24 81TBA (Exp.) − − − ± ± . Additionally shown are the out-of-horizontal phenyl plane bend angles ˜ ω i as determined recently byNEXAFS measurements .FIG. 3: Perspective view of the trans TBA adsorption geometry atAu(111). (111) surface. Also consistent with the preferential orienta-tion deduced from NEXAFS the tert-butyl groups contact thesurface with one C-H bond pointing towards the substrate. Atthe pure PBE level of theory, the central diazo-bridge corre-spondingly floats at a rather high height of about 4 ˚A parallelto the surface. This is further away than for pure azobenzene,where this height was 3.5 ˚A . With the PBE functional notproviding attractive vdW components to the molecule-surfaceinteraction, the main effect of the bulky tert-butyl groups con-forms therefore with the intuitive expectation to simply lift thefunctional azobenzene backbone further away from the sur-face.This picture is obviously prone to change, once some ac- FIG. 4: Perspective view of the cis
TBA adsorption geometry atAu(111). count of vdW attraction is added to the theoretical description.At the PBE+TS level of theory, the height of the diazo-bridgeis indeed significantly reduced to 3.22 ˚A, which is neverthe-less still large on the scale of a typical N-transition metal bondlength. In contrast, the differential interaction with the surfaceis not significantly altered by the additional bonding compo-nent. The slight bending of the azobenzene moiety away fromthe ideal gas-phase planarity is basically the same in boththe PBE and PBE+TS adsorption geometry. Here, the par-allel orientation of the diazo-bridge to the surface plane andthe corresponding slight upward bending of both phenyl-ringsby ∼ ◦ are in very good agreement to the recent NEXAFSexperiments . Indirectly, this corroborates some of the as-sumptions made in the NEXAFS data analysis, without whichtwice as large tilt angles would have resulted . On the otherhand, due the insensitivity of both structural parameters dueto the degree of dispersive interaction in the calculations thisagreement with experiment does unfortunately not allow anyconclusion as to the accuracy of the description provided bythe semi-empirical TS scheme. This is particularly unfortu-nate, as the latter predicts an intriguing similarity of the diazo-bridge height for TBA (3.22 ˚A) and pure azobenzene (3.28 ˚A),cf. Table I. If correct, this obviously largely contradicts theafore mentioned intuitive view of the bulky spacer groups interms of ’table legs’ that decouple the photochemically activeunit in a geometric sense.The similarity between functionalized and pure moleculedoes not extend to the adsorption geometry of the cis isomerat Au(111). As shown in Fig. 4 this geometry is skewed forTBA with both the position and orientation of the two phenyl-rings distinctly different. This feature is consistently obtainedat both PBE and PBE+TS level of theory, while for pureazobenzene both theoretical descriptions agreed on an adsorp-tion structure with only a small sideward tilt away from a C rotational symmetry around the central diazo-bridge. In theasymmetric adsorption mode of TBA the lower phenyl-ringis tilted out of the surface plane, yet without significant tor- sion that would place its two butyl groups at different heightsabove the surface. This is much different for the upper phenyl-ring, which does not only stand essentially upright, but is sotorsioned that one of its butyl groups points largely towardsthe diazo-bridge and the other one away from it. Overall, theinternal structure of the azobenzene backbone in this adsorbedstate is therefore very similar to that in gas-phase TBA witheven the CNNC dihedral angle ω not much affected by the sur-face potential. Another noteworthy feature of the skewed ad-sorption mode is the prominent tilt of the central diazo-bridge,i.e. in contrast to adsorbed trans and cis azobenzene and ad-sorbed trans TBA the two N atoms are at a notably differentvertical height, cf. Table I. Qualitatively, this overall struc-ture is perfectly consistent with the understanding reached inthe recent STM and NEXAFS measurements. As tothe NEXAFS, the agreement is even quantitative in all re-spects. The determined out-of-surface-plane bend angles ofboth phenyl-rings agree very well with the assignments made,cf. Table I. This also extends to the inclination of the cen-tral CNNC plane with respect to the surface plane . For this,NEXAFS determines about 60 ◦ , close to the 61 ◦ and 68 ◦ de-termined at the PBE and PBE+TS level, respectively.To some extent unfortunate and similar to the situation forthe trans conformer, none of these qualitative features, as wellas tilt and inclination angles are very sensitive to the descrip-tion of vdW interactions in the calculations. Both PBE andPBE+TS give essentially identical results. The major fea-ture introduced again by the account of vdW attraction is anessentially rigid downward shift of the entire molecule byabout 0.9 ˚A. This is slightly larger than is the case for trans TBA ( ∼ . cis -azobenzene(2.23 ˚A). With such prominent differences between PBE andPBE+TS, measurements of the vertical heights as was doneby NI-XSW for pure azobenzene would therefore again pro-vide a critical benchmark to judge on the accuracy of the ac-count of vdW interactions introduced by the semi-empiricalTS scheme. In turn, however, precisely the lacking sensitiv-ity of the qualitative features, as well as tilt and inclinationangles of both trans and cis adsorption geometries suggeststhat they are not much affected by the shortcoming of the em-ployed semi-local xc functional with respect to dispersive in-teractions. This increases the confidence that a quite reliableunderstanding of the adsorption geometry of both conformershas been reached by the present calculations. B. Energetics
In view of the preceding study on azobenzene it is clearthat the binding energetics provided by the semi-empiricalDFT-D approach deserves particular scrutiny. We performedTPD measurements to obtain an experimental reference valuefor the binding energy of the thermodynamically favored trans TBA on Au(111). Figure 5 shows the TPD spectra as a func-tion of TBA surface coverage recorded at the fragment mass of190 amu ((C H ) -C H + ), with a linear heating rate of 1 K/s. Temperature [K] Q M S i n t en s i t y o f m a ss [ a r b . un i t s ] (cid:2) = 0.1 MLE = 1.65 ± 0.08 eV des Slope = -E / R des
T [K ] x 10 -1 -1 -3 l n (r) [ a r b . un i t s ] de s (cid:3) (cid:3) (cid:3) (cid:4) (cid:3) (cid:4) (cid:3) (cid:5) (cid:3) (cid:5) TBA/Au(111)
FIG. 5: Thermal desorption spectra (raw data) of TBA adsorbed onAu(111) at different coverages as recorded with a linear heating rateof 1 K/s at the fragment-mass of 190 amu ((C H ) -C H + ). Theinset shows the desorption rate (ln ( dr des / dt ) plotted against the re-ciprocal temperature at a coverage θ of 0.1 monolayer (ML). Theslope of the line, - E des / R with R the gas constant, then determinesthe activation energy for desorption E des at this coverage, in this case E des (0.1 ML) = 1.65 ± In the low coverage regime a broad desorption peak ( α ) is ob-served around 542 K which extends to lower temperature withincreasing coverage, as has also been observed for TBA onAg(111) and other azobenzene (derivatives) on noble metalsurfaces . This behavior is attributed to repulsive inter-actions between the adsorbed molecules (for example due todipole-dipole interactions).After saturation of the desorption peak α a second feature α develops at 394 K. Further increase in coverage leads tosaturation of this peak and to the appearance of a new featurearound 308 K labeled as α . The α peak increases in heightand width with increasing coverage, showing a typical zero-order desorption behavior (data not shown). We therefore as-sign the α peak to desorption from the multilayer, while α and α are associated with desorption from the monolayer.The α component represents the desorption of about 10 % ofthe monolayer coverage and is attributed to desorption out ofa densely packed TBA structure .In order to derive the activation energy for desorption, E des ,of TBA on Au(111) in the low-coverage regime from this TPDdata we utilize the so-called complete analysis , which hasthe advantage that knowledge about the absolute coverage isnot required. For this analysis a family of TPD curves aremeasured as a function of initial coverage ( θ i ) as shown in Fig.5. These curves are subsequently used to construct a familyof θ ( t ) curves via A p ∝ (cid:82) ∞ r des ( t ) dt = θ with A p the peak areaand r des the desorption rate. Due to the known linear heat- Trans CisE ads ∆ E C − T E ads Gas-phase PBE − − PBE+TS − − @ Au(111) PBE − .
16 0.52 − . − .
00 0.99 − . − . ± . − − TABLE II: Energetics of TBA in gas-phase, and adsorbed atAu(111). All numbers are in eV. ing rate (in our case 1 K/s), this corresponds to a knowledgeof θ ( T s ) as a function of θ i , where T s is the surface tempera-ture. Following this, an arbitrary coverage value θ is chosenthat is contained in each of the desorption curves. The des-orption rate at this coverage, r des ( θ ) , and the temperature atwhich this rate was obtained, T , are then read off from eachof these desorption curves. Plotting ln [ r des ( θ )] versus 1 / T (Arrhenius-plot) finally allows to determine E des ( θ ) as fol-lows directly from the Polanyi-Wigner equation . This is ex-emplarily shown for a coverage of 0.1 ML in the inset of Fig.5, which yields an activation energy of 1.65 ± ≤ E des = . ± E ads for trans TBA may bereferenced. As apparent from Table II the values obtained atthe PBE and PBE+TS level are very consistent with the find-ings of the previous studies of benzene and pure azobenzeneat coinage metal surfaces : The lack of vdW interactionsin the semi-local PBE functional leads to an overall negligi-ble binding of both TBA isomers, with the actual values for E ads in fact essentially identical to those computed for pureazobenzene at Au(111) before . Adding dispersive attractionwithin the DFT-D approach these binding energies are dramat-ically increased, such that in the end the intended dispersion’correction’ amounts to more than 90 % of the total bindingenergy. Compared to the experimental reference the PBE+TSapproach overbinds the trans isomer by about as much asPBE underbinds. In absolute numbers this is a disconcert-ing deviation of more than 1 eV. In our previous work we hadassigned the corresponding overbinding determined for pureazobenzene to the neglect of metallic screening in the DFT-Dapproach . This argument was largely based on the intrigu-ing accuracy of the PBE+TS adsorption geometry comparedto the bond distances derived from NI-XSW measurements.As discussed above, such a conclusion on the determined TBAadsorption geometry is presently not possible, as all hithertomeasured structural parameters are not very sensitive to theadditional vdW attraction provided by the PBE+TS scheme.On the other hand, the additional butyl groups lead to a signif-icantly increased overall binding energy of TBA compared toazobenzene, and in both cases PBE+TS overbinds by roughly40 % (azo: 1.71 eV vs. 1.08 eV ; TBA: 3.00 eV vs. 1.70 eV).Such a rather geometry-unspecific deviation between theoret-ical and experimental data quite well fits the anticipated effectof an overestimated uniform background potential. The latteris attributed to the neglect of screening of vdW contributionsfrom more distant substrate atoms.For gas-phase TBA PBE+TS yields a somewhat smaller cis - trans energy difference ∆ E C − T than for pure azobenzene,0.29 eV compared to 0.49 eV , respectively. This arises pre-dominantly from the additional vdW attractions due to thebutyl groups in the bend cis conformer, cf. Fig. 1, and isnot found at the PBE level of theory, where ∆ E C − T = .
58 eVfor both TBA and azobenzene. Compared to the gas-phasereference, the stability difference of the two TBA conformersis with 0.99 eV substantially increased at Au(111) at PBE+TSlevel of theory. This is primarily due to the larger vdW at-traction possible in the planar trans adsorption mode and wasequivalently obtained for pure azobenzene adsorption . Thisinterpretation in terms of vdW is corroborated by the essen-tially unchanged ∆ E C − T of gas-phase and adsorbed TBA atPBE level of theory, where these bonding contributions areabsent, cf. Table II. Not withstanding, in light of the sus-pected overestimation of the vdW attraction within PBE+TSwe also expect that this increase of ∆ E C − T upon adsorptionis overestimated. Extrapolating the roughly 40 % overshootseen in the trans TBA binding energy, we would thereforecautiously conclude on only a moderate ∼ . − . cis - trans energy difference upon adsorption ofTBA at Au(111). While the therewith implied higher stabil-ity of adsorbed trans TBA is consistent with the existing ex-perimental data, there are to date unfortunately no dedicatedmeasurements of ∆ E C − T against which this estimate could becompared. C. Vibrations
The picture arising from the geometric and energetic char-acterization of TBA at Au(111) points to a surface chemicalbond that is predominantly due to unspecific dispersive inter-actions. Particularly for the trans conformer, the structuralsimilarity of the azobenzene and TBA adsorption complexwith respect to the photochemically active diazo-bridge moi-ety (together with the in both cases negligible − N=N − bondactivation) questions the oft-quoted function of the tert-butyl’spacers’ in terms of a geometric decoupling from the surface.Complementary information confirming this view can comefrom an analysis of the surface vibrational modes. Figure 6compares corresponding results computed for trans TBA atAu(111) at PBE+TS level of theory with experimental datafrom HREELS . At first glance, we find overall agreementwith all major features nicely reproduced by the calculations.As had already been noticed in the experimental work, the vi-brational spectrum exhibits only minor changes between ad-sorbed cis and trans TBA and even TBA in the condensedphase . This is similarly obtained in the calculations, asillustrated by the spectrum for cis
TBA at Au(111) also shownin Fig. 6.The detailed inspection of the C-H vibrational modes in
0 50 100 150 200 250 300 350 400 I n t e n s it y ( a r b . un it s ) Energy (meV)Trans, PBECis, PBE+TSTrans, PBE+TSCis, PBECis, Exp.Trans,Exp.
350 365 380 395
FIG. 6: Comparison of calculated surface vibrational modes of trans (top, red lines) and cis (center, blue lines) TBA at Au(111) at PBE(dashed lines) and PBE+TS (solid lines) level of theory to the corre-sponding HREELS spectra from ref. 25 (bottom, black solid lines.Upper: trans , Lower: cis ). Center inset: The highest energy peaksof the spectrum, shown at larger scale. Note the C-H stretch peak onthe left. The spectra have been vertically displaced for clarity. Notaiming to reproduce the HREELS intensities, the computed modesare convoluted with a Lorentzian broadening of 1 (meV). the energy range of 350 to 400 meV reveals nevertheless thesubtle influence of the Au(111) substrate on the vibrationalspectra of the adsorbed species. Based on the calculations thethree frequency bands centered at 368, 378, and 390 meV areassigned to the symmetric and asymmetric C-H vibrations ofthe tert-butyl groups, as well as the C-H bending modes atthe phenyl-rings, respectively. There is a clear change of theformer two vibrational bands when going from the PBE tothe PBE+TS description, cf. the inset of Fig. 6. The overallreduced distance of the TBA molecule to the gold substrateupon inclusion of the attractive vdW forces leads to a clearsplitting of the C-H vibrations of the tert -butyl moieties. Inthe adsorbed species the frequencies of the C-H bonds point-ing towards the substrate are red shifted by 5-10 meV andthe corresponding vibrations develop low-energy sidebands.These subtle but important changes are confirmed by the ex-perimental HREELS spectra. Due to the mainly in-plane char-acter of the C-H phenyl-ring bending modes we observe onlya small shoulder at 390 meV. In contrast the symmetric andasymmetric C-H stretch vibrations of the tert -butyl legs at 368and 380 meV result in pronounced dipole-active bands. Upon trans to cis isomerization the number of C-H bonds point-ing towards and noticeably contacting the substrate is roughlyhalved, cf. Fig. 2. This leads in both experiment and theoryto a reduction of the low-energy side bands best seen for thesymmetric C-H stretch vibrations which constitute the lowestband at 368 meV. As the splitting of the C-H stretch vibrationshappens only at a reduced distance to the substrate (PBE vs.PBE+TS adsorption structure) it confirms the importance ofvdW interaction in the bonding of TBA. Unfortunately thisis not a quantitative benchmark for the accuracy of our semi-empirical PBE+TS approach.More insight into the specific bonding of the diazo-bridgemoiety can specifically come from a detailed analysis of the − N=N − stretch mode. However, identically obtained withPBE and PBE+TS, the computed value of 184 meV in ad-sorbed trans and cis TBA again reflects a negligible activationof the NN bond at the surface. Compared to the respectivegas-phase values, the mode red-shifts only by about 3 meVfor both isomers. Correspondingly, the obtained shift betweenthis stretch mode in trans and cis
TBA at Au(111) is very sim-ilar to the one computed for free TBA, 6 vs. 5 meV, respec-tively. This in turn compares very well to the 8 meV mea-sured for TBA in the condensed phase with IR and Ramanspectroscopy . Such a negligible change of the cis - trans stretch frequency difference upon adsorption is not found forpure azobenzene at Au(111). Here, the closer encounter ofthe diazo-bridge in the cis adsorption geometry softens thestretch mode by more than 6 meV. With the mode unaffectedin adsorbed trans azobenzene, this leads to a concomitant in-crease of the cis - trans stretch frequency shift compared to thegas-phase. In cis TBA at Au(111) a corresponding softeningdoes not occur as the diazo-bridge does not come as close tothe surface in the skewed adsorption mode. In this respect,one can really attest some geometric decoupling effect to thebutyl groups, yet only for the cis isomer. For the trans iso-mer though, an equivalent analysis of the diazo-bridge againstsurface stretch or other torsional azo modes shows always thesame similarity between TBA and azobenzene at Au(111) asfor the just discussed − N=N − stretch mode. Also the charac-terization of the surface vibrational properties supports there-fore the understanding that the effect of the bulky tert -butylgroups for the switching properties of trans TBA at Au(111)must be more subtle than a mere geometric decoupling of thecentral photochemically active molecular moiety.
IV. SUMMARY AND CONCLUSIONS
A methodological motivation of this work was to furtherexplore the capabilities of the semi-empirical dispersion cor-rection approach to semi-local DFT in describing the adsorp-tion of complex organic molecules at metal surfaces. For thiswe have presented a detailed characterization of the geomet-ric, energetic and vibrational properties of the trans and cis isomers of the azobenzene derivate TBA at Au(111). Thefindings are in all respects in line with the experience frompreceding studies on benzene and pure azobenzene at coinagemetal surfaces : The additional account of attractive vdWinteraction introduced by the PBE+TS scheme leads to a sig-nificant modification of the adsorption geometry, primarily interms of bringing the molecule closer to the surface. The con-comitantly dramatically increased binding energy is notablyoverestimated compared to the reference value from TPDmeasurements. For azobenzene at Ag(111) a correspondingoverbinding – in fact in relative terms very much comparableto the one found here for TBA at Au(111) – was attributedto the neglect of metallic screening of dispersive interactionsin the semi-empirical DFT-D approach . This argument waslargely based on the very accurate PBE+TS adsorption ge-ometry as compared to detailed structural data from NI-XSWmeasurements. Such a conclusion is, unfortunately, not yetpossible in full for the here studied TBA at Au(111). Thequalitative features of the determined trans and cis adsorptiongeometries and even detailed tilt and inclination angles agreeall very well with existing data from STM , HREELS and NEXAFS experiments. The energetic lowering of theC-H stretch vibrations of the tert -butyl groups pointing to-wards the substrate is only observed in the PBE+TS approach.This points towards the importance of the vdW componentintroduced by the DFT-D approach. However, no experi-mental reference data exist to date for the vertical height ofthe adsorbed molecule, that would allow to directly confirmthe present working hypothesis, namely that the PBE+TS ap-proach is a very useful and computationally tractable tool toprovide accurate structural information for adsorbed complexorganic molecules.If this working hypothesis proves correct and the deter-mined PBE+TS adsorption geometries are indeed accurate,the prevailing preconception of the role of the bulky func-tional groups for the experimentally observed improved iso-merization ability of adsorbed TBA needs revision. Moti-vated by the gas-phase structure, this view discusses the butylgroups as spacers that help to decouple the molecular switchfrom the metal substrate. For the TBA cis isomer we indeedcompute a skewed adsorption mode, in which one phenyl-moiety is much more tilted, up to the point of standing es-sentially perpendicular to the surface plane. This is quite dif-ferent to the more symmetric adsorption mode of cis azoben-zene at Au(111), such that here one can attest some geomet-ric decoupling from the surface due to the butyl ’table legs’.Quite distinctly, the almost planar adsorption mode of trans TBA is very much comparable to the adsorption geometry ofpure azobenzene. Particularly for the photochemically activediazo-bridge moiety this similarity goes even down to essen-tially unchanged structural and vibrational properties. In thisrespect, a core message arising from the present study is thatespecially for the photo-switching of trans
TBA the effect ofthe bulky spacer groups must be more subtle than the antici-pated mere geometric decoupling of the functional backbonefrom the metal surface.
V. ACKNOWLEDGEMENTS
Funding by the Deutsche Forschungsgemeinschaft throughSfb 658 - Elementary Processes in Molecular Switches at Sur- faces - is gratefully acknowledged. Ample computing timehas been provided at the HLRB-II architecture of the LeibnizSupercomputing Centre in Garching. Y. Yu, M. Nakano and T. Ikeda,
Nature , 2003, , 145 Z. F. Liu, K. Hashimoto and A. Fujishima,
Nature , 1990, ,658–660 T. Ikeda and O. Tsutsumi,
Science , 1995, , 1873–1875 B.-Y. Choi, S.-J. Kahng, S. Kim, H. Kim, H. W. Kim, Y. J. Song,J. Ihm and Y. Kuk,
Phys. Rev. Lett. , 2006, , 156106–4 T. Ishikawa, T. Noro and T. Shoda,
J. Chem. Phys. , 2001, ,7503–7512 A. Cembran, F. Bernardi, M. Garavelli and L. Gagliardi,
J. Am.Chem. Soc. , 2004, , 3234–3243 T.A. Jung, R.R. Schlittler, and J.K. Gimzewski,
Nature , 1997, , 696–698 F. Moresco, G. Meyer, K.-H. Rieder, H. Tang, A. Gourdon, andC. Joachim,
Phys. Rev. Lett. , 2001, , 672–675 M. Alemani, M.V. Peters, S. Hecht, K.-H. Rieder, F. Moresco,and L. Grill, it J. Am. Chem. Soc. , 2006, , 14446–14447 S. Hagen, F. Leyssner, D. Nandi, M. Wolf and P. Tegeder,
Chem.Phys. Lett. , 2007, , 85–90 M. J. Comstock, N. Levy, A. Kirakosian, J. Cho, F. Lauterwasser,J. H. Harvey, D. A. Strubbe, J. M. J. Fr´echet, D. Trauner, S. G.Louie and M. F. Crommie,
Phys. Rev. Lett. , 2007, , 038301–4 P. Tegeder, S. Hagen, F. Leyssner, M. V. Peters, S. Hecht,T. Klamroth, P. Saalfrank and M. Wolf,
Appl. Phys. A , 2007, ,465–472 M. Alemani, S. Selvanathan, F. Ample, M. V. Peters, K.-H.Rieder, F. Moresco, C. Joachim, S. Hecht and L. Grill,
J. Phys.Chem. C , 2008, , 10509–10514 S. J. Jenkins,
Proc. R. Soc. A , 2009, , 2949–2976 S. Kristy´an and P. Pulay,
Chem. Phys. Lett. , 1994, , 175–180 E. McNellis, J. Meyer, A. Dehghan Baghi and K. Reuter,
Phys.Rev. B , 2009, , 035414–4 E. R. McNellis, J. Meyer and K. Reuter,
Phys. Rev. B , 2009, ,205414–10 G. Mercurio, E. R. McNellis, I. Martin, S. Hagen, F. Leyssner,S. Soubatch, J. Meyer, M. Wolf, P. Tegeder, F. S. Tautz andK. Reuter,
Phys. Rev. Lett. , 2010, , 036102–4 Q. Wu and W. Yang,
J. Chem. Phys. , 2002, , 515–524 S. Grimme,
J. Comput. Chem. , 2004, ibid. , 1787–1799 P. Jureˇcka, J. ˇCern´y, P. Hobza and D. Salahub,
J. Comput. Chem. , 2007, , 555–569 N. Atodiresei, V. Caciuc, J.-H. Franke and S. Bl¨ugel,
Phys. Rev.B , 2008, , 045411–9; N. Atodiresei, V. Caciuc, P. Lazi´c andS. Bl¨ugel, Phys. Rev. Lett. , 2009, , 136809–4 A. Tkatchenko and M. Scheffler,
Phys. Rev. Lett. , 2009, ,073005–4 R. Schmidt, S. Hagen, D. Brete, R. Carley, C. Gahl, J. Doki´c,P. Saalfrank, S. Hecht, P. Tegeder and M. Weinelt,
Phys. Chem.Chem. Phys. , 2010, ( submitted ) L. ´Ov´ari, M. Wolf and P. Tegeder,
J. Phys. Chem. C , 2007, ,15370–15374 D. Vanderbilt,
Phys. Rev. B , 1990, , 7892–7895 S. J. Clark, M. D. Segall, C. J. Pickard, P. J. Hasnip, M. I. J.Probert, K. Refson and M. C. Payne,
Z. Kristallogr. , 2005, ,567–570 J. P. Perdew, K. Burke and M. Ernzerhof,
Phys. Rev. Lett. , 1996, , 3865–3868 J. J. Rehr, E. Zaremba and W. Kohn,
Phys. Rev. B , 1975, ,2062–2066 S. R. Bahn and K. W. Jacobsen,
Comput. Sci. Eng. , 2002, ,56–66; also see A. Togo, Phonopy (2009). http://phonopy.sourceforge.net/ H. J. Monkhorst and J. D. Pack,
Phys. Rev. B , 1976, , 5188–5192 The central CNNC atoms of the azobenzene backbone are onlyprecisely located within one plane for a dihedral angle ω =
0. Forthe small, but finite angles ω determined in the calculations, wetherefore use a simple arithmetic average to determine the inclina-tion of the CNNC plane in the comparison to the NEXAFS data. L. ´Ov´ari, J. Schwarz, M.V. Peters, S. Hecht, M. Wolf, and P.Tegeder,
Int. J. Mass. Spectrom. , 2008, , 223–228 K. Christmann,
Introduction to Surface Physical Chemistry ; Top-ics in Physical Chemistry, H Baumg¨artel, E.U. Franck, W.Gr¨unbein (Eds.), Steinkopff, Springer, Darmstadt, New York,1991, pp 152. R. K¨ubler, W. L¨uttke, and S. Weckherlin,
Z. Elektrochem. , 1960, , 650–658 B. Kellerer, H.H. Hacker, and J. Brandm¨uller,
Indian. J. PureAppl. Phys. , 1971,9