Surface Science 602 (2008) 2639–2642

Contents lists available at ScienceDirect

Surface Science

journal homepage: www.elsevier .com/locate /susc

The chemistry of chlorine on Ag(111) over the sub-monolayer range: Adensity functional theory investigation

Alberto Roldán a,b, Daniel Torres a, Josep M. Ricart b, Francesc Illas a,*

a Departament de Química Física and Institut de Química Teòrica i Computacional (IQTCUB), Universitat de Barcelona, C/Martí Franquès 1, E-08028 Barcelona, Spainb Departament de Química Física i Inorgànica, Universitat Rovira i Virgili, C/Marcel.lí Domingo s/n, 43007 Tarragona, Spain

a r t i c l e i n f o

Article history:Received 30 May 2008Accepted for publication 22 June 2008Available online 27 June 2008

Keywords:Chlorine promotionPartial oxidationEpoxidationDFTDensity functionalHalogens on metals

0039-6028/$ - see front matter � 2008 Elsevier B.V. Adoi:10.1016/j.susc.2008.06.014

* Corresponding author.E-mail address: francesc.illas@ub.edu (F. Illas).

a b s t r a c t

Chlorine adsorption on Ag(111) as a function of coverage has been studied by means of periodic densityfunctional theory. Incorporation of Cl into the substrate leading to a surface AgCl film has also been con-sidered. It is concluded that at low coverage (hCl < 0.2 ML) on-surface adsorption is favoured over Cl pen-etration while at higher coverage on-surface and subsurface adsorption become both thermodynamicallyand kinetically favoured. Implications for the Cl promoted silver catalyzed ethylene partial oxidation arediscussed.

� 2008 Elsevier B.V. All rights reserved.

1. Introduction

Heterogeneously catalyzed reactions carried out for industrialproduction of a variety of bulk chemicals usually involve multipleadditives to induce an enhancement in activity and/or selectivitytowards the desired product [1–3]. The effect of these additivesas promoters or poisons is well recognized in the chemical industryalthough the molecular mechanisms responsible for the chemicalchanges induced by surface additives are still poorly understood.Ethylene epoxidation by partial oxidation is a prototypical exam-ple. Industrially, the process is carried out exclusively with silverparticles dispersed on alumina with typical low selectivity whichis significantly increased upon addition of both alkali promotersand Cl moderators [4]. In fact, adding chlorine to the catalyst hasbeen shown to increase the selectivity to the desired product [5],whereas alkali metals have also been shown to increase selectivityalthough at the expense of the loss of some activity [5]. Among thealkali-metal salts incorporated during the catalyst preparation,caesium has been reported to be especially effective and severalexplanations for the selectivity enhancement have been suggested[6,7].

The molecular mechanism of the reaction catalyzed by Ag isnowadays well understood both from theory [8–11] and recentexperimental advances [12–15] even if some aspects of the reac-tion mechanism are not completely clear [16]. However, although

ll rights reserved.

the interaction of Cl with well defined Ag surfaces has attracted aconsiderable attention from theory [17–20], much less is knownabout everything that goes beyond the low-coverage regime andon-surface adsorption [21,22]. The need for a detailed study ofAg surfaces chlorination is readily recognized when one realizesthat Cl is added in the form of chlorinated hydrocarbons and it isthe decomposition of these compounds which deposit chlorineatoms on the catalyst surface. More important, the well-knownneed to maintain the initial optimized level of chlorine coverageby continuous replenishment from the gas phase indicates thatCl is continuously being lost from the catalyst, due possibly to dis-solution into the subsurface or to a removal by the hydrocarbondiluents used in the reactor [15]. For these reasons, it is possibleto over chlorinate the silver surface to levels larger than one mono-layer. This over chlorination is perceived as leading to an inactiveAgCl surface layer, poisoning the catalytic reaction [23].

Therefore, understanding the silver chlorination seems to beessential to better understand the role of chlorine as a promoterof silver-catalyzed ethylene epoxidation. The purpose of this con-tribution is precisely to investigate the chemistry of chlorine atomson Ag(111) surface with a particular focus on the initial incorpora-tion of Cl atoms to the basal plane of Ag. We investigate on-surfaceas well as subsurface chlorine structures, comparing adsorptionand relaxation properties. To get further insight into the formationof a subsurface chlorinated-phase, we employed thermodynamicas well as kinetic arguments. This permits us to estimate the Clconditions under which Cl migration to the subsurface will occur.Finally, the first stage on the formation of an inert AgCl phase is

2640 A. Roldán et al. / Surface Science 602 (2008) 2639–2642

considered and the Cl conditions for the phase transition to occurare provided.

2. Surface models and computational details

Periodic density functional calculations have been carried out tostudy chlorine adsorption on Ag(111) for a wide range of coveragesituations including chlorine penetration to subsurface. TheAg(111) surface has been modelled by four-layer slabs periodicallyrepeated in the direction perpendicular to the surface with fiveequivalent layers of vacuum between successive slabs. Two differ-ent supercells, p(2 � 2) and p(4 � 4) were considered. Adsorptionwas allowed only on one of the two surfaces of the slab model ex-posed and the resulting dipole model removed by a proper electro-static potential adjustment [24]. The top two layers of the slabwere fully allowed to relax. Previous studies, for instance recentwork concerning oxygen adsorption on the surface of coinage met-als [25], show that results are converged with respect to slab thick-ness and vacuum width.

The total energy density functional (DF) calculations were per-formed using the VASP code [26,27]. The exchange-correlation en-ergy and potential were described by means of the generalizedgradient approximation (GGA) with the PW91 implementation[28]. The electron density has been represented through the usualKohn-Sham (KS) approach, the KS valence states were expanded ina plane-waves basis set with kinetic energy below 315 eV; the ef-fect of the ionic cores on the valence electron density was takeninto account by means of the PAW method [29]. The reciprocalspace was described with a Monkhorst-Pack grid [30] of2 � 2 � 1 and 4 � 4 � 1 k-points, for large p(4 � 4) and smallp(2 � 2) cells, respectively. These settings have proven to lead tosufficiently accurate results [25].

The average adsorption energy per chlorine atom was calcu-lated, with respect to the Cl2 molecule, according to

Eb ¼1

NClEcell�Cl � Ecell � NCl

12

ECl2

� �; ð1Þ

where NCl is the number of Cl atoms per unit cell, ECl2 is the total en-ergy of the Cl2 molecule and Ecell�Cl is the total energy of the system,while Ecell is the energy of the clean slab.

3. Result and discussion

3.1. Surface versus subsurface Cl on Ag(111)

We model chlorine incorporation into the Ag(111) surface byfirst addressing the on-surface and subsurface adsorption overthe sub-monolayer range. For this purpose, Cl coverage wasprogressively increased from 0.06 to 1 ML. Results for averagebinding energies for the different Cl coverage considered are givenin Table 1.

Regarding the on-surface adsorption, we find Cl adsorption tobe the most favoured on hollow sites present on the basal surface,both fcc and hcp sites being energetically equivalent, within the

Table 1Average binding energies as a function of Cl coverage (Eb, in eV) for on-surface (fcc)and subsurface (Oh, Th I, Th II) adsorption; see text for site description

h (ML)

0.06 0.25 0.50 0.75 1

fcc �1.60 �1.53 �0.95 �0.45 �0.18Oh 1.29 1.30 0.01 �0.14 �0.04Th I 1.53 1.46 0.35 0.37 0.27Th II 1.57 1.50 0.52 0.33 0.23

accuracy of the present DF techniques. Increasing coverage givesrise to a pronounced reduction in adsorption energy due to repul-sive lateral interaction between the negatively charged adsorbates[20]. Nevertheless, adsorption energies remain exothermic up to�1 ML (see Fig. 1). For subsurface Cl occupation three possible sitesare considered; these are an octahedral site (Oh), coordinated withsix Ag atoms, and two different tetrahedral sites (Th I and Th II). Thefirst one (Th I) is directly located between the first and second Aglayers just below the hcp on-surface sites, while the second one(Th II) is located just below Ag atoms of the first substrate layer.For the whole coverage range considered in the present work, sur-face penetration to incorporate Cl to the Oh hollow site is energet-ically more favoured than on Th by �0.2 eV; whereas Clincorporation to the Th sites is energetically equivalent. Results inTable 1 show that compared to subsurface adsorption, on-surfaceadsorption is notably more favoured for low coverage, while athigh coverage on-surface and subsurface adsorption become prac-tically equal; this is a new and rather unexpected result although itis in agreement with earlier experimental work by Bowker andWaugh [31] concerning Cl adsorption on Ag(111) which indicatethat Cl transport from surface to bulk is facile even at low temper-ature. The low stability of subsurface sites is likely to be related tothe structural distortion of the silver substrate induced by subsur-face Cl occupation. Table 2 displays the substrate relaxation causedby on-surface and subsurface Cl for the whole range of coverageconsidered, measured as the percentage change of the first inter-layer distance. The relaxation induced by on-surface adsorptionis moderate as already known from previous studies [20]. How-ever, the lattice expansion induced by incorporation of Cl to sub-surface varies from �60%, at low coverage, to nearly �100% athigh coverage. Note that the lattice expansion for subsurface Cloccupation is larger that the corresponding value for other atomic

Fig. 1. (a) Average binding energy of chlorine (Eb) as a function of the total chlorinecoverage for the fcc on-surface site and Th I, Th II as well as Oh subsurface sites. (b)Cell relaxation for the adsorption of chlorine on the different sites as a function oftotal chlorine coverage. The lines are drawn to guide the eye.

Table 2Substrate relaxation for the on-surface adsorption (fcc) and subsurface adsorption onthe three interstitial sites (Oh, Th I, Th II) as a function of chlorine coverage

d12(%)

h (ML) 0.06 0.25 0.50 0.75 1

fcc �1.94 �2.30 �2.26 �2.48 �2.80Oh 1.66 56.96 86.16 93.60 96.50Th I 9.40 62.76 80.97 74.00 98.57Th II 6.96 59.14 67.47 73.90 97.82

The relaxation is measured as the percent change in the distance between the twotopmost atomic layers (d12).

A. Roldán et al. / Surface Science 602 (2008) 2639–2642 2641

species such as oxygen, which rises up to a maximum of �30%, athigh coverage [32].

The results above for the relative stability of subsurface Cl are inagreement with experimental evidence for chlorine migration tothe subsurface [23,31]. In order to get further insight on the rela-tive energy of the different subsurface phases, we focus now onthe initial incorporation of Cl to the subsurface and analyze the sta-bility of a set of structures with increasing Cl content correspond-ing to a coverage ranging from 0.06 to 1 ML. Following the similarprocedure employed by Todorova et al. [33] to study the incorpo-ration of oxygen on Pd, we have considered a set of structures ob-tained by placing the chlorine atoms in any combinatorial way onsubsurface and on-surface sites, between the first and the secondsubstrate layers, and obtained the minimum average binding en-ergy per Cl atom by appropriate energy minimization of each pos-sible structure. We employed a p(2 � 2) unit cell with increasingnumber of Cl atoms per unit cell from 1 to 4. Chlorine adatomswere distributed between the on-surface and the subsurface lead-ing to a large number of site combinations, which obviously de-creases upon increasing Cl coverage. In each case, the structure

Fig. 2. (a) Calculated lowest values for the average binding energy (Ebmin) as a

function of the total chlorine coverage, for structures with Cl located only on thesurface and structures with mixed on/subsurface chlorine. The lines are drawn toguide the eye. (b) Calculated value for the adsorption enthalpy (DHAds) with respectto Cl coverage. The horizontal line represents the experimental value for theformation enthalpy of AgCl (DHf

AgCl).

with minimum energy has been chosen and results are summa-rized in Fig. 2. Occupation of subsurface sites will start at a givencoverage corresponding to the structure with all Cl atoms locatedon the surface which will have similar stability to that with Clatoms occupying subsurface sites at the same coverage. Fig. 2(a)shows that the pure on-surface adsorption is always most favour-able in the whole sub-ML coverage range considered, but also indi-cates that migration of chlorine from surface to subsurface is likelyto occur after completion of a full surface chlorine layer. To furtherinvestigate the conditions which lead to the formation of subsur-face Cl, we studied the kinetic barrier for the penetration of Clthrough the surface, from on-surface fcc site to subsurface hcpoctahedral site at two different coverage. The transition state forsubsurface Cl migration corresponds to the geometric structurewith the Cl atom in the plane of the silver surface layer. Analysisof the results for different initial values of Cl coverage shows thatthe energy barrier for surface penetration is strongly dependenton the coverage decreasing from �3 eV for low coverage(0.06 ML) to �1.5 eV at high coverage (1 ML). The large decreasein the energy barrier upon increasing coverage is also accompaniedby a large decrease in the reaction energy for the migration pro-cess, changing form �3 to �0.5 eV, when going from 0.06 to 1 ML.

3.2. Subsurface Cl on Ag(111) versus AgCl phases

The results presented in the previous section show that uponincreasing coverage a redistribution of chlorine atoms betweenthe surface and subsurface will occur. These results have been ob-tained for surface models of the Ag(111) surface. Nevertheless, it isclear that they have important implications for the Cl-doped Agcatalyst used in ethylene epoxidation. In particular, it seems logicalto expect that the relative occupation of surface and subsurfacesites by Cl atoms will affect the useful lifetime of the catalyst. Infact, when chlorine is added in large amounts, AgCl phase mayeventually appear. These arguments call for a deeper investigationof the thermodynamics of the formation of the AgCl phase that issupposed to lead to the deactivation of the catalyst.

We investigate the thermodynamic aspects of the formation ofthe AgCl phase following the thermodynamic model proposed byCarlisle et al. [34] and applied by Scheffler et al. to the study ofthe stability of oxide phases on metals [35]. The transition betweena chemisorbed chlorine phase and the appearance of an AgCl film isthermodynamically determined by a critical value of the Cl cover-age, (hc). This critical coverage for the transition corresponds to thevalue for which the differential heat of adsorption of the process(DHAds) equals the value of the formation enthalpy of the AgClphase (DHf

AgCl; the experimental value for AgCl is �1.32 eV [36])as in Eq. (2)

DHAgClf ¼ DHAdsðhcÞ: ð2Þ

Following Carlisle et al. [34], it is possible to make a rough estimatethe DHAds (hc) value for through a linear Taylor series expansion asin Eq. (3).

DHAdsðhÞ ¼ EbðhÞ þoEb

oh

� �h: ð3Þ

From the calculated values for the bonding energy as a function ofcoverage shown in Fig. 2(b) one can roughly evaluate the criticalcoverage fulfilling Eq. (2) as �0.20 ML. Therefore, it appears thatthe chemistry of Cl on Ag(111), and hence its role as a selectivitypromoter, is strongly dependent on the chlorine coverage. In caseCl coverage is maintained low enough, Cl atoms will be presenton the surface and its role as a selectivity promoter may probablybe due to the on-surface species. However, if Cl coverage is not care-fully controlled, the interaction with Ag(111) will readily lead to

2642 A. Roldán et al. / Surface Science 602 (2008) 2639–2642

the formation of an AgCl phase, poisoning the silver catalyst for theformation of partial oxidized compounds.

4. Concluding remarks

In the present work the energetic and thermodynamic aspectsof initial incorporation of Cl to the basal plane of Ag(111) surfacehave been studied by means of periodic density functional calcula-tions. We have studied the Cl adsorption as a function of coverageand investigated the incorporation of Cl into the substrate leadingto a surface AgCl film. To this end, we have analyzed the adsorptionproperties over the sub-monolayer range. The present results allowus to conclude that at low coverage (hCl < 0.2 ML) on-surfaceadsorption is favoured over Cl penetration. At high Cl coverage,on-surface and subsurface adsorption become almost isoenergeticand the migration of chlorine from the surface to the basal plane ofAg(111) appears to be thermodynamically favoured, in agreementwith experimental evidence [31]. In addition, increasing the Cl cov-erage also leads to a significant decrease of the energy barrier for Clpenetration. Hence, high Cl coverage favours surface penetrationboth thermodynamically and kinetically. Addition of Cl in largeamounts will rapidly overcome the critical 0.2 ML coverage andformation of an inert AgCl phase may eventually occur. Clearly, thishas strong implications in the catalytic partial oxidation of ethyl-ene; excessive chlorination will lead to the deactivation of thecatalyst.

Acknowledgments

A.R. and D.T. thank the Universitat Rovira i Virgili and Universitatde Barcelona, respectively for pre-doctoral fellowships. Financialsupport has been provided by the Spanish Ministry of Educationand Science (Grants CTQ2005-08459-CO2-01 and -02, UNBA05-33-001) and by Generalitat de Catalunya (2005SGR-00697,2005SGR-00104 and Distinció per a la Promoció de la Recerca Univer-sitària de la Generalitat de Catalunya granted to F.I). Part of the com-puter time was provided by the Centre de Supercomputació deCatalunya (CESCA) and Barcelona Supercomputing Centre (BSC).

References

[1] R.B. Grant, C.A.J. Harbach, R.M. Lambert, S.A. Tan, J. Chem. Soc. Faraday Trans. I83 (1987) 2035R.

[2] S. Hawker, C. Mukoid, J.P.S. Badyal, R.M. Lambert, Surf. Sci. 219 (1989) L615.[3] L.S. Byskov, J.K. Nørskov, B.S. Clausen, H. Topsøe, J. Catal. 187 (1999) 109.[4] J.G. Serafin, A.C. Liu, S.R. Seyedmonir, J. Mol. Catal. A 131 (1998) 57.[5] R.B. Grant, R.M. Lambert, Langmuir 1 (1985) 29.[6] S. Linic, M.A. Barteau, J. Am. Chem. Soc. 126 (2004) 8086.[7] W.S. Epling, G.B. Hoflund, D.M. Minahan, J. Catal. 171 (1997) 490.[8] S. Linic, M.A. Barteau, J. Am. Chem. Soc. 125 (2003) 4034.[9] M.-L. Bocquet, D. Loffreda, J. Am. Chem. Soc. 127 (2005) 17207.

[10] M.-L. Bocquet, A. Michaelides, D. Loffreda, P. Sautet, A. Alavi, D.A. King, J. Am.Chem. Soc. 125 (2003) 5620.

[11] D. Torres, N. Lopez, F. Illas, R.M. Lambert, J. Am. Chem. Soc. 127 (2005) 10774.[12] A. Klust, R.J. Madix, Surf. Sci. 600 (2006) 5025.[13] A. Klust, R.J. Madix, J. Am. Chem. Soc. 128 (2006) 1034.[14] F.J. Williams, D.P.C. Bird, A. Palermo, A.K. Santra, R.M. Lambert, J. Am. Chem.

Soc. 126 (2004) 8509.[15] J.R. Monnier, J.L. Stavinoha Jr., G.W. Hartley, J. Catal. 226 (2004) 321.[16] S. Linic, H. Piao, K. Adib, M.A. Barteau, Angew. Chem. Int. Ed. 43 (2004) 2918.[17] A. Ignaczak, J. Electroanal. Chem. 495 (2001) 160.[18] N.H. de Leeuw, C.J. Nelson, C.R.A. Catlow, P. Sautet, W. Dong, Phys. Rev. B 69

(2004) 045419.[19] H. Fu, L.L. Jia, W.N. Wang, K. Fan, Surf. Sci. 584 (2005) 187.[20] A. Migani, F. Illas, J. Phys. Chem. B 110 (2006) 11894.[21] H. Fu, L. Jia, W. Wang, K. Fan, Surf. Sci. 584 (2005) 187.[22] L.L. Jia, Y. Wang, K.N. Fan, J. Phys. Chem. B 107 (2003) 3813.[23] H. Piao, K. Adib, M.A. Barteau, Surf. Sci. 557 (2004) 13.[24] G. Kresse, J. Furthmuller, Comput. Mat. Sci. 6 (1996) 15.[25] D. Torres, K.M. Neyman, F. Illas, Chem. Phys. Lett. 429 (2006) 86.[26] G. Kresse, J. Hafner, Phys. Rev. B 47 (1993) 558.[27] G. Kresse, J. Furthmuller, Phys. Rev. B 54 (1996) 11169.[28] J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson, M.R. Pederson, D.J. Singh, C.

Fiolhais, Phys. Rev. B 46 (1992) 6671.[29] P. Blöchl, Phys. Rev. B. 50 (1994) 17953.[30] H. Monkhost, J. Pack, Phys. Rev. B 47 (1993) 558.[31] M. Bowker, K.C. Waugh, Surf. Sci. 134 (1983) 639.[32] W.X. Li, C. Stampfl, M. Scheffler, Phys. Rev. B 67 (2003) 045408.[33] M. Todorova, K. Reuter, M. Scheffler, Phys. Rev. B 71 (2005) 195403.[34] C.I. Carlisle, T. Fujimoto, W.S. Sim, D.A. King, Surf. Sci. 15 (2000) 470.[35] M. Todorova, W.X. Li, M.V. Ganduglia-Pirovano, C. Stampfl, K. Reuter, M.

Scheffler, Phys. Rev. Lett. 89 (2002) 096103.[36] D.R. Lide (Ed.), CRC Handbook of Chemistry and Physics, Internet Version 2007,

(87th ed.), Taylor and Francis, Boca Raton, FL, 2007 (http://www.hbcpnetbase.com).