Binding Site Transitions Across Strained Oxygenated and Hydroxylated Pt(111)

Abstract The effects of strain σ on the binding position preference of oxygen atoms and hydroxyl groups adsorbed on Pt(111) have been investigated using density functional theory. A transition between the bridge and FCC binding occurs under compressive strain of the O/Pt(111) surface. A significant reconstruction occurs under compressive strain of the OH/Pt(111) surface, and the surface OH groups preferentially occupy on‐top (bridge) positions at highly compressive (less compressive/tensile) strains. Changes to magnetisation of the O‐ and OH‐populated surfaces are discussed and for O/Pt(111) oxygenation reduces the surface magnetism via a delocalised mechanism. The origins of the surface magnetisation for both O‐ and OH‐bearing systems are discussed in terms of the state‐resolved electronic populations and of the surface charge density.


Introduction
Pt-based materialsa re effective catalysts fort he oxygen reduction reaction (ORR). [1,2] Early experimental reports demonstrated the highly active character of Pt 3 Ni(111) [3] and across as equenceo fr elated Pt 3 M( M = Ni, Co, Fe, Ti,a nd V) surfaces. [4] The ORR mechanism concerns the hydrogenation of O 2 and mainly occursa long either af our-electron reduction pathway producing H 2 Oo ra long at wo-electron pathway producing H 2 O 2 .The mechanisms of these pathways, however,are not understood and this lack of understanding is, in part, due to the number of concurrent processes that occur during the reaction. Recent studies of the hydrogenation of Oa nd OH on Pt(111) [5] have shown that Grotthus [6] diffusion of H + through aqueous adlayers on the Pt(111)a re instrumental in the ORR, but have failed to identify the rate limiting step(s) in the reaction. STM studies [7] demonstrated the propagationo f' reaction fronts' across the Pt(111)s urface. However, empiricalm odelling [7] of these fronts show significant qualitative disagreements with the experimental data and identify that numerous length-scale phenomena need to be included in the model to accurately describet he experimental observations. These phenomenainclude, at larger length scales, OH and H 2 Oisland formation and, at shorter length scales, the effects of attractive adsorbate-adsorbate interactions caused by hydrogen bonds between adsorbed molecules.
The mechanism is evidently complex even on ac lean, un-strainedP t(111)s urface.T he experimentalo bservations of reactivity on real (i.e. core-shell nanoparticle) are then further complicatedb ecause of the phase segregation in the nanoparticle. This means that the surface of the nanoparticle is typically a pure metal, whereas the bulk of the nanoparticle is an alloy. This causes al attice mismatch between the surfacea nd the bulk, which will strain the surface. The paradigmo fe xplaining ORR on nanoparticle surface is, therefore, complex. However, to reduce this problem, recent studies have investigated the behaviour of strained bulk alloys, extended phase-segregated surfaces and strained surface systems. The purpose of the work is to identify the behaviour of Oa nd OH on strained Pt(111)t oestablish aframework around which the ORR mechanism on strained nanoparticle surfacesc an be discussed. The currents urveyw ill overview some of the current work in these topics before leading into am ore detailed discussion of the O/ Pt(111)a nd OH/Pt(111), which are the subject of the computational sections of this work.

Characterisation of the Pt-Alloy Bulk and Surface
The studies outlined earlier in this Introduction have focussed on experimental characterisation of the catalysts. Am ore systematict heoretical approacht ou nderstanding why these catalysts work so welli sd aunting. The principle reasons for these difficulties are the number of degrees of freedom that need to be considered-sucha sb ulk and surface stoichiometry,s hape, and topology,a nd the magnetic character of the systemwhen investigating ap articulargroup of catalysts.
Recent theoretical studies have startedtoaddress these difficulties by surveying the bulk alloy.T hese investigations have varied the lattice parameter both above and below its equilibrium value and have consequently put the crystal into as tate of either tensile or compressives train, respectively.C hanging the lattice parameter in this way will change the interatomic The effects of strain s on the binding positionp reference of oxygen atoms and hydroxyl groups adsorbed on Pt(111)h ave been investigated using density functional theory.Atransition between the bridge and FCC binding occursu nder compressive strain of the O/Pt(111)s urface.Asignificant reconstruction occurs under compressive strain of the OH/Pt(111)s urface, and the surfaceO Hg roups preferentially occupy on-top (bridge) positions at highly compressive (less compressive/tensile) strains.C hanges to magnetisation of the O-andO H-populated surfaces are discussed and for O/Pt(111)o xygenation reduces the surfacem agnetism via ad elocalised mechanism. The origins of the surface magnetisation for both O-and OH-bearing systemsa re discussed in terms of the state-resolved electronic populations and of the surface charge density. distances within the crystal and the amount of overlap between the electronic orbitals of neighbouring atoms, which consequently will affect the electronic character of the entire system.I nvestigations into the ordered phases of Ni x Pt 1Àx (x = 0.25, 0.5, and 0.75) [8] andb oth Pt x Fe 1Àx and Pt x Co 1Àx [9] have been recently performed. The investigations focussed on the magnetisation of the unit cell and on the occupancy of the fully state-resolved atomic orbitals. For each case, the magnetisation was shown to be carried by the Pt andN i, Fe or Co d states, with all the other states remaining non-magnetic. The magnetism increased as the strain becamei ncreasingly tensile and was accompanied by charget ransfer between the magnetic quantum number-resolved dorbitals. Investigations of strain-induced changes in the magnetisation of alloys have also been performed on rare-earth alloys.S tudies of the CeNi 5 system [10] have shown that the magnetism of the alloy is carried by the Ce fa nd Ni ds tates. However,u nder strain, the spin moment associated with the Ni ds tates changes far more significantly than those associate with the Ce fs tates. These observations underlie the importance of the dorbital. The directionality of either the do rfstates is evidently insufficientt o ensure sensitivity of the spin momento fe ither to the strain state of the crystal. The radial extenta nd symmetry of each state must also be factorsa nd may serve as ad esign criterion for proposed the catalysts.
The technique of 'strain-engineering', or investigating the changes of the character of the system when it is subjected to compressive or tensile strains, mimics the surfaces of coreshell nanoparticles. In the current context, the technique is a simplification to render the systems more accessible to computationals tudy.I ncluding ordered selvedge and bulk alloy layers increases the degrees of structural freedom significantly.F urther,P t x M 1Àx (M = metal)s ystemsa re often experimentally disordered and with ar ange of stoichiometries. [8,9] The treatment of alloys with combinations short-and long-range order (i.e. disordered and ordered alloys, respectively) is aw ell-known problem;t heoretically,s tatistical treatments [11] have been applied successfully.D ensityf unctional theory (DFT) treatments, however,r equire ag reater level of structurald efinition and consequently structurala pproximations are often used. Contemporary DFT studies of Pt 3 M(111)( M= Ag, Au, Co, Cr,C u, Fe, Ir,M n, Mo, Ni,Pd, Re, Rh, Ru, Ti,V ) [12] have addressed these problemsb ys imulating five-layer slabs with as ingle Pt surface layer and an ordered PtM selvedger egion, and the calculated segregation energies were shown to be in agreement with experimental results. Subsequent studies [13] demonstrated that the oxidisation of the surface disrupts the surfacel ayer though this disruption was characterised morei nterms of the segregation energiesa nd layer spacing. This disruption, however,i sr eflected in the current work particularly in the surface reconstruction seen for OH/Pt(111).
The works discussed in the previousp aragraph reduced the nanoparticle surface to ap articular facet-notably,t he (111) [12] -and then investigated the behaviour of this facet with ap ure metal surfaces patterning alloy selvedge and bulk regions.Akey designq uestion in this approachi sh ow thick does the pure metal surface covering need to be?T his param-eter can be controlled both experimentally and in computational treatments. The pure metal surface covering will exhibit strain effects because the selvedge and bulk lattice parameters may be differentt ot he bulk lattice parameter of pure metal, and ligand (i.e. electronic structure) effects due to both the strainingo ft he pure metal and to 'seeping' of the alloy wavefunctionf rom regionsb elow the surface. Computational studies of A 3 B( whereA= Pt and Pd, and B = Cu, Ag and Au) [14] and of alloy-core@Pt nanoparticles [15] have shown that strain effects may persist for surface trilayers though ligand effects are more quickly damped with overlayer thickness and become markedlyl ess significanta fter single or bilayer surfacec overings. Studies of Pt(111)s urface layers on Pt 25 Ni 75 (111) [16] have shown ligand effects are significant for a2monolayer (ML) surface Pt thickness, but strain effects dominate for 3-4 ML thicknesses.T hese observations suggest the range of validity of the approachu sed in the current work is for Pt overlayers which have thicknesses ! 3ML.
DFT studies have recently been used to investigate the behaviour of surfaceHatoms acrossasequence of strained Pt [17] and other pure transition-metal [18] slabs. These studies were pertinent not only to the ORR reaction, which has been highlighted in the current work, but also for surface hydrocarbon chemistry [19] anda saprecursor to hydrogen storage. [20,21] The DFT studies showed that the preferred binding position of H atoms in the H/Pt(111)s ystem can be tuned to be either ontop or FCC by applying either compressive or tensile strain to the Pt(111)s urface. [17] Similarc hanges in the preferred Hb inding site wereo bserved on Pd and Ir fors trains of up to 2% and on Fe, Rh, Ag and Os for strains of up to 5%. [18] Consequentially,t he reactivity of the Ha tom may to change under the application strain. This is because the valency between the Ha tom and the surface will change with binding position, which will redistribute both the charge surrounding the H atom, whichi sa ssociated with the H-surface bond, and that which is not.T he latter component of charge will be more directly involvedi nr eactions between the Ha tom and either other adsorbates via the Langmuir-Hinshelwood mechanism or with gas-phase particles via the Eley-Rideal mechanism.
The sensitivity of Hb inding position to strain is also af undamental interest for bulk systems. Studieso ft he behaviour of the hydrides of alkali and alkaline earth metals under strain [22] have highlighted the superconducting behaviour of certain alloys under pressure. Evolutionary algorithmsh aveb een used to identify as eries of novel phases for MH n with n > 1a nd M = Li, Na, K, Rb, Cs, and for MH n with n > 2a nd M = Mg, Ca, Sr,B a. In addition, ah exagonal high-pressurep hase of rhodiumh ydride has recently been predicted by using DFT. [23] Within the alkaline earth metal hydrides,M g-based materials have significant applicationsi ne nergy storage. The materials have been characterised experimentally and have been seen to exist in rutile, [24] a-PbO 2, [25] and cubic [26] phases as well as in two forms of orthorhombic [27] and in non-stoichiometric forms. [28] There clearly is considerable diversity in the structure of hydrides both in bulk form and when patterned on surfacesand the application of this diversity is potentially of great interest. The phenomenono fHbindings ite/pressure dependence is, therefore, of current and significant interest, in both applied and fundamental fields. The accompanying fields of the O/ Pt(111)a nd OH/Pt(111)s ystems under pressure will be addressedi nt his work. Consequently,S ection 1.2. will review these systems.

Oxygenated and HydroxylatedPt(111)
The interaction of oxygen with Pt(111)h as been extensively investigated and ac ogent pictureo ft he gas/surface interaction has startedt oe merge. Early studies [29] presented the temperature-dependent behaviouro ft he interaction, whereby weakly adsorbed molecular oxygen forms on Pt(111)a tt emperatures below 120 Ka nd the adsorption of molecular oxygen occurs at temperatures in the range 150-500K and as ubsurface oxide forms at temperatures between 1000 and 1200K.T he hightemperature behaviour was later investigated using surfaces that had been prepared through differing methods of oxidation. [30] The onset decomposition of the oxygen-bearing surfaces was observed at 400 K, thoughn os urface oxygen was observed at 1070 K.
The dissociated phase was shown by low-energy electron diffraction (LEED)t of orm an ordered p(2 2)-Ol ayer [29,31] for coverages up to 0.25 ML. XPS studies [32] have shown that higher coverage states of oxygen have the same chemical state as those in coverages up to 0.25 ML. It was postulated [33] that the high coverage state forms throughd irect dissociation, whereas the lower coverage state forms via am olecularly adsorbed precursor. The similarities in the electronic state of the high and low coverageo xygen states were further evidenced by work function change (Df)s tudies, [34] whichs howedt hat Df varies linearly with oxygen coverage between 0a nd 1ML, which is consistentw ith the adsorption of as ingle surface species.
Later studies [35] demonstrated that an ovel high oxygen concentration state could be formed by exposing the Pt(111)s urface to thermally cracked oxygen at room temperatures. These finding reflecte arlier observations that the maximumo xygen concentration could be increased by exposing the surfacet o an electron beam [29,31] or by increasing the surfacet emperature/oxygen dosing pressure. [32] The necessity of investigating these dependencies is symptomatic of the need to couple fundamental surface science investigationsw ith those that are performed at more catalytically relevant temperatures and pressure. This can be seen in recent ReactorSTM studies, [36] which have identified the formationo ft wo high-coverages urface oxidesa th igh oxygen pressures (up to 5bar) and surface temperatures of 300-538 K.
Consequently,t he current study will focus on oxygen coverages of 0.25 ML. In this phase, the oxygen bindingp ositionh as been determined experimentally [37] to be at the FCC site. This binding position has also been predicted by using DFT. [38][39][40] The current study will also focus on the bonding mechanism and the mechanism of charger elocation that accompany the formationo fO< C-< Pt andO H ÀPt bonds. This is in deference to the more common theoretical studies that currently exist in the literature, which have soughtt oe lucidate the mechanisms of phase formation. [38] The binding mechanism between oxygen and transition-metal surfaces can proceed through more than one mechanism in deference to the binding mechanism of other molecules such as H, [41] whichb inds covalently. This multiplicity has been demonstrated explicitlyf or Ob inding to Ni(111)a nd Ni(100) clusters. [42] Studies of O/Pt(111) [40] system have suggestedt hat oxygen binding is accompanied by the exchange of charge between the Op and Pt do rbitals and that there is some suggestion of multiplicity in the bonding mechanism.T he current study will investigate these suggestedb inding mechanisms and also their behaviour as the Pt(111)s urfaceiss trained.
Hydroxyl formation on Pt(111)c an be achieved by first dosingo xygen onto the clean Pt(111)s urface and then reacting the resulting surface with water to form OH/Pt(111). [43][44][45] Experimentally,h owever,apure OH layer cannot be formed, [43] as hydroxyl is av ery good proton acceptor andf orms strong OHÀH 2 Ob onds by Hd onationf rom the adsorbed water towards the adsorbed OH group.T heoretical investigations [46] of low coverages (1/9 ML) of OH have shown that the bridge and on-top binding sites of the Pt(111)s urface are approximately degenerate. The binding energy at these sites is approximately 2.25 eV.H igher coverage theoreticals tudies [46] (1/2 to 1ML) showedt hat Hb ondingb etween adjacent OH groups causes an enhancement of the OH chemisorption energy and ap reference for binding at the on-top site.
Experimental studies [47] using scanning tunnelling microscopy (STM) and high-resolution electron energy loss spectroscopy (HREELS) have shown that the on-top adsorption site is most likely for coverages of 2/3 ML. Consequently, the current work will focus on OH coverages of 0.25 ML, where the formation of ah ydrogen-bond network between adjacent OH groups is anticipated to be minimal. Analysis of the OH bending mode in the HREELS investigations [47] showed that the OH molecule is tilted from the surfacenormal. Tilting has been discussedt heoretically with earlyD FT studies, [48,49] indicating that OH tendedt ob ind in an upright position in hollow sites whereas tilting is seen to lower the binding energy when the OH is bound to the on-top or bridge sites. [50] In the current work the degree of tilt of the OH group and the orientation of the OH bond with respect to the Pt(111)s urface will be allowed to fully relax.
In the remainder of this work, the computational detailsa re outlined and then investigations of the strained extended surface O/Pt(111) andO H/Pt(111)s ystems are presented in turn and discussed. The work will then concludew ith ap resentation of the key findings of these computational investigations.

Computational Details
The DFT simulations presented in this work were performed using the plane-wave Quantum Espresso package. [51] The kinetic energy cut-offs for wave-functions and for charge density and potential were set to 75 and 300 Ry,r espectively.B rillouin zone integration was performed on a( 6 6 1) grid using af irst-order Methfessel-Paxton [52] smearing of 0.02 Ry.T he PBE exchange correlation func- tional within the generalised gradient approximation (GGA) was used throughout this work and the core electrons were described with norm-conserving pseudopotentials. [53] Using this approach, the equilibrium bulk lattice constant of Pt was determined to be L 0 = 3.980 ,w hich is comparable to the experimental value of the lattice constant of 3.92 .A ll results presented in the current work were obtained by using fully spin-polarised simulations. The van der Waals correction has not been included in the current study as the focus of the work is on the OÀPt bond at low Oa nd OH coverages. The van der Waals interaction might have ag reater significance for higher coverages, particularly in regimes where the H atoms are more closely bound to other surface species. The assumption of low surface coverage is to simplify the system and to ensure that the focus of the current study is on the OÀPt bond and its behaviour under strain. The current study consequently excludes the complicating effects of lateral (adsorbate-adsorbate) interactions. Effects such as the presence of water molecules in the vacuum region are similarly excluded for the same reason;i nclusion of greater particle densities would be more appropriate to molecular dynamics for which the focus is not on the behaviour of the adsorbate-substrate bond, or the binding position of the adsorbate.
The surface was modelled using slabs containing seven layers of Pt atoms. Subsequent slabs were separated by av acuum with a width of approximately ten lattice constants. During relaxation, the central layer of Pt atoms were fully constrained. The remaining atoms were allowed to relax freely.C ompressive and tensile strains were applied by changing the lattice constant L by using Equation (1): where s is the strain and s 2 À0:05; þ0:05 ½ .B yu sing this convention, the applied strain was termed tensile (compressive) for s > 0 (s < 0). Equation (1) consequently defines the distance between Pt atoms in the central layer of each slab at the beginning of each simulation and this distance was not subsequently changed during the simulation.  2 2) Surface supercells were used for each simulation, and each of these supercells contained only as ingle Oa tom or OH group. This low concentration (0.25 ML) reduces the effect of lateral interactions between the adsorbates, and at this concentration no evidence was found of Ht ransfer between adjacent OH groups.
The Oa nd OH binding energies E B were defined by Equation (2): where E Ads=Pt is the total energy of the Oo rO Hb earing (2 2)-Pt(111)s lab, E Ads is the total energy of an isolated Oa tom or OH group, and E Pt is the total energy of ac lean, fully relaxed (2 2)-Pt(111)s lab. The factors of 2a nd 1=2a ccount for the binding of O atoms/OH groups on either side of the slab.
To characterise electronic changes to the surfaces under strain, the projected density of states (PDOS) curves for the pc omponents of the surface Oa toms and the dcomponents of their nearest Pt atoms were analysed by calculating the magnetic quantum number resolved occupancy N l,m using Equation (3): where gl ; m; m s ; s; E ðÞ is the spin-resolved PDOS, l,m and ms are the angular,m agnetic and spin quantum numbers, respectively, and E is energy.T or educe the analysis the angular quantum number resolved occupancy was calculated by using Equation (4): The analysis procedure outlined in Equations (3) and (4) will characterise the electronic state of the system in energy space. To obtain ar eal-space characterisation the difference electron density 1 diff was calculated by using Equation (5): where 1 is the electron density of the strained (s ¼ 6 0) or unstrained (s = 0) system. x s is the normalised position coordinate, and is defined by Equation (6): x is the absolute coordinate used during the calculation of 1.
The work function f of the oxygen-or hydroxyl-covered slab was calculated by using Equation (7): E F and V(1) are the Fermi energy and the potential at ah eight of five lattice constants above the slab, respectively.T he work func- T he curvess how that at equilibrium (s = 0) the preferred bindingp osition of the oxygen atom is at the FCC position. This is in agreement with previouse xperimental [38] and theoretical DFT [38][39][40] investigations. Thecurves also show that, as the strain becomes increasingly compressive (s < 0), the preferred binding position changes fromt he FCC to the bridge position and that this transition occurs at s = À0.03. The OÀPt bond length for FCCand bridge-bound Ow as 2.07 and 2.02 ,r espectively,f or the unstrained (s = 0) surface, andv aried by < 0.01 from these values as s was varied acrosst he interval[À0.05, + 0.05].
The changes in the surface geometry discussed in the previous paragraph are accompanied by changes in the surface electronic structure. Figure 3s hows the angular and magnetic quantum number resolved oxygen and surfaceP ts tate populations, N l and N l,m ,r espectively,a safunction of strain s for the (2 2)-O/Pt(111)s ystem.T he N l shown in Figures 3a-e show that the strain-dependentm agnetisation of the surface layers developsi nt he Op and Pt ds tates. Consequently,d iscussion of the N l,m will focus on the states. The occupation of the spin and down components are linked to the magnetic momento n each atom from the relationship that the magnetic moment equals difference between the total spin up andt he total spin down charge.
Bridge-bound oxygen is energetically preferred for s À0.03. The N l in Figure 3a show that in the s > À0.03 interval the Op states become spin split and within the s À0.03 interval the Op states are non-magnetic. This behaviour is reflected in the behaviour of the N l of the Pt ds tates shown in Figure 3b,w hich show at ransition between zero and non-zero spin splitting at s = À0.03.
FCC-bound oxygen is energetically preferred for s > 0.03. However,t he N l in Figure 3c show that across the entire range of s the Op states undergo considerably less spin splitting than their bridge-boundc ounterparts. Equivalently,F igure 3d showst hat the N l of the Pt ds tates spin split above s = À0.03, but less than their counterparts in the bridge bound case showni nF igure 3b.
These observations indicate that the local magnetic moment of the surface oxygen atoms and their nearest-neighbour surface Pt atoms are dependento nt he amount of strain s applied to the surface. To elucidate the origin of this change in the local magnetic moment, Figure 3e shows the N l of the clean surfaceP ta toms. In this case, there is very clears pin splitting across the entire range of s. The amount of spin splitting is also very clearly greater than the amount of Pt ds pin splitting seen at equal s for each of the oxygenated surfaces. The development of al ocal Pt ds tate spin moment under s is, therefore, ac onsequenceo ft he application of strain to either the clean or the oxygenated surfaces. The Pt dm agnetisation is reduced by the presenceo fs urfaceo xygen. The Op state magnetisation has been shown to minimise under s as the energetically preferred binding positions tend to carry low oxygen magnetic moments.
To quantify the amount of intra-orbital charge transfer that accompany these changes in N l ,F igure 3s hows the magnetic quantum number resolved Oa nd their nearest-neighbour surface Pt state populations N l,m for Oa toms boundi nF igure 3a and 3b bridge,a nd Figure 3c and 3d on-top sites. For comparison, the N l,m for surfaceP ta toms on clean Pt(111)i ss hown in Figure3e. Figure 3a showst hat the magnetic momentl ocalised on the Oa toms when bound in the bridge site increases as the population of each of the spin up component of the Op z ,p x and p y states increases and, most significantly,t he spin down component of the Op x state decreases. For Ob ound in the FCC site, the magnetic momentl ocalised on the Oa tom is relativelysmall when compared with the bridge-bound Om agnetic moment shown in Figure 3a and the Pt magnetic moments shown in Figure 3b-3 e. However,amoment does develop as the strain s becomes increasingly tensile above s = À0.03 and is predominantly attributed to increases (decreases) in the spin up components Op z (p x and p y )s tate(s), and increasesi nt he spin down component of the Op z ,p x and p y states.
Comparison of N l in Figures 3a-3 es hows that, fort he lowest energy structure at each s,t he surfacem agnetisation is predominantly attributed to the Pt atoms. Furthermore, comparisono ft he Pt N l in Figures 3b,3da nd 3e shows that Oa dsorptiont ends to reduce the Pt moments. Figure 3e shows that for the clean Pt(111)s urface, the Pt d xy andd x 2 Ày 2 states and the Pt d zy and d zx states are degenerate. Figures 3b and 3d show that this degeneracy is lifted by the presence of surface oxygen. The degree of splitting (i.e.t he energetic difference between the Pt d xy and d x 2 Ày 2 states, or the energetic difference between the Pt d xy and d x 2 Ày 2 states) is approximately monotonic between the bridge-bound O ( Figure 3b)a nd FCC-bound   ( Figure 3d). This is highly suggestive that the mechanism for splitting is delocalised, as al ocalised or directional mechanism would be expected to change as the binding positionc hanged from two-to three-fold. Figure 3d also shows quantitatively that, within the energetically preferredr egion where the magnetic moment is significant (i.e. when the Oi sF CC bound and s !À0.03), the largestc hanges in spin up populations are to the Pt d z 2 state. Lesser,t houghc omparable, changes affect the spin up Pt d xy and d x 2 Ày 2 states, whereas the largest changes in the spin down population are to the Pt d zy and d zx states. The significance of these latter qualitative differences is not yet fully realised,b ut it may be conjected that spin-polarised injections to the surfaces may be an effective mechanism of inducing either mechanical or electronic strain.
The discussion so far has focusedo nt he s-dependent changes to thes urface structureo ft he O/Pt(111)s ystem and the accompanying changes in the surface magnetisation. For the bridge-boundO /Pt(111)s ystem, Figure 4a shows the bonding( antibonding) Op states in ab and spanning EÀE F % À6eVt oÀ4.5eV( EÀE F % < M-> 1eVt o+ 2eV). The occupation of the spin up Oa ntibonding state increases with s. This can be inferred qualitatively from Figure 4a,b ut is shown quantitatively in the Os pin up N l,m panel of Figure 3a.T he occupation of the spin down Op antibonding state decreases with s and this change can be seen qualitatively and quantitatively in the same way.T here is an et increasei nt he antibonding population, and this increasew eakens the OÀPt bond and contributes to the decreasesi nt he magnitude of the oxygen binding energy E B for the bridge-boundc urve shown in Figure 2. These decreasesc ontribute to the shift in preferred binding positionf rom the bridge to the FCC site as s increases.
The bonding and antibonding Op states occupy similar energy bands for the FCC-bound O/Pt(111)s ystem. This is shown in Figure 4b.S ignificantly,t he antibondings tate for the FCC-bound O/Pt(111)s ystem does not span the Fermi level (EÀE F = 0eV) in the way seen for the bridge-bound O/Pt(111) system.C onsequently,i ts occupation is not sensitivet os-dependentc hanges in the Fermi level in the way that the state was sensitivef or the bridge bound case. The population of both the spin up and down Op bondings tates increase as s increases. This can be seen qualitativelyb yt he heighto ft hese states in Figure 4b and quantitatively by the N l,m in Figure 3c. This strengthens the OÀPt bond and causes the increasesi n the magnitude of the oxygen binding energy E B for the FCC curve shown in Figure 2.
In both the bridge-and FCC-bound cases, the nearest-neighbour surface Pt atom developsdstate density in the energy intervals spanned by the bonding and antibonding Op states. This can be seen by comparing either of the Pt panelso fF ig-  Figure 4c shows that the clean surfaceP td states become narrower as the strain s becomes increasing tensile.T his narrowing is not seen for the second layer Pt atoms, which are also shown in Figure 4c ands hows that the surfacel ayer becomes less metallica ss increases. The formation of bonding OÀPt interactions across EÀE F À 4.5 eV reverses this trend around the nearest-neighbours urface Pt site for the range of tensile s investigated in the current work. However,a th igherv alues of tensile s,t he formation of bonding OÀPt interactions may become problematic because of this tendency of the clean surfaceP td states to narrow.B ulkbound oxygen may be less affected. Figure 5s hows the difference electron density 1 diff for the (2 2)-O/Pt(111)s ystem with the Oa toms bound in the twofold bridge and the FCC sites. For Figure 5a,t he Oa toms are bound in the two-fold bridge site and 1 diff shows that charge accumulates in lobes whose axes are perpendicular to the Pt surfacef or tensile strains of s =+0.01 and + 0.03, anda re depleted from the surface-paralleln odes. These changes are reflected in changes to the N l,m for the bridge-bound O/Pt(111) system shown in Figure 3a,w here an increase( decrease) in the total population of the Op z (p x )s tates is observed as s be-comes increasingly tensile.Amore nominal change is noticed in the population of the Op y state.
For Figure5b, the Oa toms are bound in the FCC site and 1 diff shows that charge transfer occurs predominantly between lobes whose axes are parallel to the Pt surfacef or all strains. Charge accumulationi so bserved for s = À0.03 in node closest to the O-nearest-neighbour surfacePtbond. This node, however,b ecomesd epleted as s increases and charge accumulation occurs on the opposite surface parallel node. These charge transfers are accompanied by increases in the N l,m for the Op z states shown in Figure 3c.C onsequently,i th as been seen that the effect of increasing s on 1 diff is to increase the charged ensity in lobes centred on the Oa tom and whose axes are perpendicular (parallel) to the Pt(111)s urface for the bridge-a nd FCC-bound systems, respectively.O ccupation of the Op z state increases as the strain becomes increasingly tensilea nd charge accumulation favoursd irections that are alignedw ith the Onearest-neighbour Pt directionf or increasingly compressive strains. Figure 6s hows the work function f and the work-function difference f-f 111 ,w here f 111 is the work of the clean Pt(111) surface, of the (2 2)-O/Pt(111)s ystem under strain. Changes in f owing to s will have two components.T he first component is attributedt oS moluchowski smoothing. [54] This smoothing reduces the amplitude of the Pt surfacew avefunction as the surfacel attice constant increases, that is, as s becomes more  tensile.T his effect can be seen clearly in Figure 6a,a st he clean Pt(111)curve shows adecrease in f as s increases. Mechanistically,t his reduction occurs because, as the surface wavefunctionb ecomes smoother,t he accompanying surface dipole and, equivalently,t he potentialb arriert or emoval of an electron from the bulk of the crystal to the vacuum,r educes. The second contribution to the changes in the f is attributedt o the charge redistribution shown in terms of 1 diff in Figure 5. To evaluatet he relative importance of these two contributions the work-functiond ifference fÀf 111 was calculated and is shown in Figure 6b.T he changes caused by the difference fÀf 111 are predominantly attributed to changes in 1 diff and change by % + 0.1 and + 0.2 eV for the bridge-a nd FCCbound systems, respectively,a ss changes from À0.05 to + 0.05. Across the same strain interval, f changes by approximately À0.25 and À0.2 eV for the bridge-andF CC-bound systems, respectively,a nd clearly shows that the effects are comparable. Figure 7s hows the hydroxyl (OH) binding energy E B as af unction of strain s for the (2 2)-OH/Pt(111)s ystem. The curvesi n Figure 7f or binding of OH in the unreconstructed HCP,F CC, bridge and on-top positions were obtained by initially relaxing the OH group above the binding position with constraints applied to the Pt and Oa toms that preventedm otion parallel to the (111)p lane.O nce al ocal minimumh ad been approximately found by using these constrained minimisations, the local minimum was more accurately determinedb yr emoving the (111)-directed constraintst ot he Pt and Oa toms and restarting the minimisation process.

OH/Pt(111)
However,m easurements during the determination of the unreconstructed curveso utlined in the previous paragraph showedt hat ag lobal minimum could be obtained by allowing the Pt atoms to reconstruct parallel to the (111)p lane. To in-vestigate this reconstruction, the OH group and surface Pt atoms werei nitially displaced along the orthogonal [À110] and [1 1 À1] directions before relaxation, and then weren ot constrained during the subsequent relaxation. These crystalline directions and the displacementv ectors d [À110] and d [11 À1] are shown in Figure 8. Af ull sampling of the possible reconstructions was determined by increasing the initial displacements d [À110] and d [11 À1] incrementally four times across the unit cell, resultingi n( 4 + 1) (4 + 1) = 25 relaxations forO Hb ound in each of the high symmetrys ites.
The resulting reconstruction is shown in Figure 8. The Pt atoms reconstruct predominantly along the [À110] direction for large compressive strains (s = À0.05) and along both the [À110] and [À110] directions for smaller compressives trains. The magnitude of the reconstruction along the [À110] for large strains is comparable to the Pt-Pt distance in that direction (2.814 ). At larger compressive strains (s < À0.05), as ignificant out-of-plane motion of the surfaceP ta toms developed. Quantification of this reconstructionw as not straightforward as, because of the magnitude of the out-of-plane motion of the Pt atoms, the reconstruction began to extend across surfaced istances greater than those described by a( 2 2) unit cell. The OÀPt interaction is consequently more significant the Pt-Pt interaction between the surface and second layers for compressive strainso fu pt os = À0.05, and becomesi ncreasingly more significant that the interaction betweena djacent surfaceP ta toms for more compressives trains. On the reconstructed surface, the OÀPt bond lengths were 2.07-2.05 for    Figure 9s hows the angular and magnetic quantum number resolved oxygen and surfaceP ts tate populations, N l and N l,m , respectively,a safunction of strain s for the (2 2)-OH/Pt(111) system.F igures 9a and 9b show these populations for the reconstructed OH/Pt(111)s ystem and Figures 9c and 9d show the populations for the on-top bound OH/Pt(111)s ystem, which are included to highlight differencesb etween populations of the on-top position on both reconstructeda nd non-reconstructed Pt surfaces. In the reconstructed system, the OH group is bound at the on-top position for strains s < 0.01 and is bound at the bridge position for strains s ! 0.01. Consequently,there should be some comparison between the orbital populations of the reconstructed and non-reconstructed surfaces, which may distinguish between general trends and those that are more closely relatedt he preferred site occupancy. Figure 9s hows as plitting in the N l of the Op and Pt d states, whichi ss imilar to that seen for the O/Pt(111)s ystem and shown in Figure 3. This shows that qualitatively the presence of Hd oes not prevent the development of magnetism. This is ac ommon phenomenon and is attributed to the delocalised Hsstates. However,t he Pauli effect betweent he Oa nd their nearest-neighbour Pt atoms is evidently stronger than this Hd amping effect for these OH/Pt(111)s ystems. Figures 9a  and 9b show acleartransition in N l,m ,asthe surfaceundergoes the structural transition between on-top bound OH (s < 0.01) and bridge bound OH (s ! 0.01). Figure 9a shows that this transition is accompanied by as ignificant decrease (increase) in the population of the oxygen p y (p z )s tate for both spin components as strain s becomes increasingly tensile. Figure 9c shows as imilar trend in the Op z state for on-top bound OH. However,i nt his latter case the changes in the Op z ,p x and p y states are comparatively smaller.
Consequently,t he Op z state is generallye lectrophilic as s becomes increasingly tensile, and changes in the valenceo f the OH group as it movesf rom on-top to the bridge site increase this tendency.S moluchowski [54] smoothing arguments would suggest that as s and the surfacel attice constant increasest he Pt component of the surfacew avefunction will smootha nd less chargef rom this wavefunction would encroach on the Oa tom. This suggests that, electrostatically,t he population of the Op z state is limited by the presenceo fP t charge from the surface wavefunction. Sterically,t he Oa tom increases its valence with the surfaceP ta toms in the bridge positionc ompared to its valence in the on-top position.T his increasei nv alence is accompanied by charge flow from the Op y state into the Op z state. By comparing Figure 9a and 9c, it is clear that the charge transfers into the Op z state from the Op y state, which is enabled by the steric changes on the surface that occur as the OH migrates to between the bridge and the on-top positioniscentral to the strain-dependantcharacter of the OH/Pt(111)s ystem. Figure 9b shows ac lear transition in the N l,m for both spin components of the nearest-neighbour Pt d z 2 and d zx states as the surface undergoes the structuralt ransition between ontop bound OH (s < 0.01) and bridge-bound OH (s ! 0.01). N l,m for the clean Pt surface have been presented in Figure 3e and show degeneracy between the Pt d xy and d x 2 Ày 2 states as well as the Pt d zx and d zy states for all s. These degeneracies are lifted on the reconstructed OH/Pt(111)s urface (Figure 9b), thougho nly the Pt d zx andd zy degeneracy is lifted for the unreconstructed OH/Pt(111)surface, as shown in Figure 9d.
To distinguish between effects that are causedb yt he Pt reconstruction and those that are from the OH group, the s < 0.01(compressively strained) portions of Figures 9b and 9d are compared. The reason for this comparison is becuase the OH group is binding in the on-top position for both the reconstructed and the non-reconstructed surfaces in this strain interval. Figure9bs hows that the reconstruction lifts the degeneracy between the Pt d xy and d x 2 Ày 2 states as well as the Pt d zx and d zy states when compared to Figure 9d,w hich shows al esser lifting between the Pt d zx and d zy states and no lifting of the degeneracy between the Pt d xy and d x 2 Ày 2 states.
Generally,t he changes in registry betweent he surfacea nd second-layer Pt atoms accompanying the reconstruction may be expectedt oc hange the symmetry of the surface wavefunction and effect the degeneracy changes. The magnitudeo ft he energy differences between the Pt d zx and d zy states and the Pt d xy and d x 2 Ày 2 states would not be directly estimated by using ap urely symmetric argument, but are now quantified in Figure 9b.H owever,t he amount of degeneracyl ifting in the less compressive( s ! 0.01) region is affected by the binding position of the OH group, particularly for the Pt d zx and d zy states whose separation in the s ! 0.01 region is significantly more than for s < 0.01.
Overall, the effect of the change in binding positiono ft he OH group from the on-top to the bridge site is to increaset he occupancy of the Pt d z 2 andr educe the population of the Pt d zx states for both spin polarisations. The change in the bindingp osition also decreases the level of the degeneracy between the Pt d zx andd zy states and, to al esser extent, the the Pt d xy and d x 2 Ày 2 states. Changes in the registry between the surfacea nd second-layer Pt atoms during the reconstruction also contributes to this loss of degeneracy,b ut to al esser extent. Figure 10 shows the projected density of states (PDOS) g for the reconstructed andn on-reconstructed on-top bound OH/ Pt(111)s ystems. The valence orbitals of the OH groups can be seen in the compressed s = À0.05 OPDOS curvesofFigure 10 a in the interval E À E F ðÞ 2 À8:0 ½ eV.Asignificant delocalisation exists within these orbitals evidence the wide feature spanning E À E F ðÞ 2 À5:0 ½ eV.T his delocalisation is removed for s > À0.03 and replaced by am ore localised orbital at EÀE F % À4.5 eV.I nspection of the non-reconstructed PDOS curvesi n  0.05-0.03 longert han bindingi nt he on-top position at the same s on the non-reconstructed surface anda re, in part, a consequence of the lower degeneracy of the delocalised orbitals seen on the reconstructed surfacew hen compared to the more localised orbitals seen on the non-reconstructed surface. Bond lengthening will also be ac onsequence of the increase in electron density accompanying hydrogenation. This is also evidenceb yc omparing the OÀPt bond lengths of the OH/ Pt(111)a nd those for on-top bound O/Pt(111). In the latter case, the OÀPt bond length was 1.86 on the compressed (s = À0.05) surface, decreasing to 1.85 when s =+0.05. These comparisons show that the OÀPt bond length is largely governed by the binding position, degree of hydrogenationo f the Oa nd the reconstruction of the surface, and are only nominally determined by strain s. Figure 11 shows the difference electron density 1 diff for the (2 2)-OH/Pt(111)s ystemw ith the oxygen atoms are bound across the a) reconstructed Pt(111)s urface, and b) the on-top sites of the unreconstructed surface. The character of the delocalised OH orbitals identified and discussed in the earlier discussion of the PDOS curves shown in Figure 10 are immediately apparent by comparing the s = À0.03 panel of Figure 11 a with the more tensile (s = À0.01, + 0.01 and + 0.03) panelsi n the same figure. In these latter panels, 1 diff shows al ocalised accumulation around the Oa nd Ha toms, which increases with s. This compares to the s = À0.03 panel of Figure 11 a, where charge depletion is evident around the Oa nd Ha toms, and charge accumulation between the Oa toms and the Pt surface layer.T he increasei nc hargea round the extended surface is consistentw ith the appearance of delocalised features in Figure 10. This mechanism is unique to binding in the on-top positiono ft he reconstructed surface compared to on-top binding on the non-reconstructed surface. This is evident by inspection of the s = À0.03 panel of Figure 11 b, which does not show charge accumulationb etween the Oa tom and the surfaceP tl ayer,a nd whichs hows al ower amount of charge depletion around the Oa nd Ha toms. The low symmetry,o r 'cusping', of the feature around the Oa tom and between the Oa nd Ha toms in Figure 11 ai sa ttributedt ot he differences in the angle of elevation of the OH bond (H atom) with respect to the surface Pt(111)p lane between bridge-bound OH (s = 0) and the on-top bound OH (s = À0.03). Removing this low symmetry from the O-H features introduces low symmetry features elsewhere in the panel showing 1 diff .Q ualitatively, the  characteristics of the feature-thati ti dentifies ar egion of significant charge depletion-remains unchanged between the low and highsymmetry presentations.
Charge transfer will also contribute to changes in the work function f of the surface. Figure 12 shows the work function f and the work functiond ifference fÀf 111 ,w here f 111 is the work of the clean Pt(111)s urface, of the reconstructed and ontop bound (2 2)-OH/Pt(111)s ystems. Changes in f under strain have two components, one owing to Smoluchowski smoothing [54] and the second due to the charge redistribution shown in terms of 1 diff in Figure 11.I nF igure 12 a, the f for each of the reconstructed and on-top bound OH/Pt(111)a nd the clean Pt(111)s ystems shows ag eneral reduction as s becomes increasingly tensile. This reduction is attributed to Smoluchowski smoothing [54] and it's mechanism was discussed earlier when considering the work functionb ehaviouro ft he O/ Pt(111)s ystem in Figure6.Asharp transition is seen for the f and fÀf 111 curvesf or the reconstructed O/Pt(111)s ystem and shown in Figures 12 aa nd 12 b, respectively.T hist ransition is attributed to the change in the binding position of the OH group and in the developmento ft he Pt surfacer econstruction. Changes in valency between the OH group and the Pt surfacea st he OH group movesb etween the on-top and the bridge site may be anticipated significantly by the surface dipole layer and consequently the work functiono ft he sample. These changes in the surface dipole layer were shown in terms of 1 diff in Figure 11 a, particularly as the charge accumulationt hat developsb etween the Oa toms and the Pt surface layer in the s = À0.03 panel.T hisc hargea ccumulation was discussed earlier;h owever, the work function behaviour presentedi nF igure 12 quantifies the effect of this accumulation.