Chemisorption of HCl to the MgO(001) surface: A DFT studyw

Andreas Markmann,*aJacob L. Gavartin

band Alexander L. Shluger

b

Received 20th June 2006, Accepted 14th August 2006

First published as an Advance Article on the web 22nd August 2006

DOI: 10.1039/b608719a

We use plane wave and embedded cluster ab initio density functional calculations to study

adsorption, dissociation and diffusion of the HCl molecule on the MgO(001) surface. The two

methods yield comparable results for adsorption of an isolated HCl molecule and complement

each other when considering charged species and coverage effects. We find dissociative

chemisorption at a coverage smaller than 0.5 monolayer with a Cl� ion electrostatically coupled

to the OH� ion at the surface oxygen site. The adsorption energy of the Cl�� � �(OH)� complex is

1.5 eV and the activation energy of Cl� diffusion away from OH� is 0.6 eV. There is no

significant activation energy for rotation of Cl� around the adsorption site. At rising coverage, an

increase in dipole–dipole repulsion between HCl molecules leads to a lowering of the adsorption

energy per HCl and a change of binding towards hydrogen-bridge type as well as a lowering of

the activation energy for Cl� diffusion. OH� formed in the surface due to HCl adsorption has a

stretch frequency of 3083 cm�1 with Cl� associated and 3648 cm�1 with Cl� removed.

1. Introduction

Understanding the nature of adsorption of polar molecules at

oxide surfaces is important both fundamentally and for nu-

merous technological applications. Most of the experimental

and theoretical studies have been concerned with adsorption

of individual water molecules and water layers. Adsorption of

diatomic molecules, such as HCl and HBr, on MgO, alumina,

ice and other surfaces has also been studied due to various

issues pertaining to chemical processes at surfaces of atmo-

spheric and interstellar dust particles and geochemistry (see,

for example, ref. 1–3), mechanisms of photo-induced surface

processes,4,5 nanochemistry6 and catalysis.7–9 Adsorption of

HCl on the MgO (100) surface is perhaps one of the simplest

examples of such a system. However, even the basic features of

HCl adsorption, i.e. whether it is physisorption or chemisorp-

tion, whether the molecule remains intact or dissociates, how

this depends on coverage, still remain controversial.

Recently Wittig et al.10,11 recorded time-of-flight (TOF)

spectra of HCl scattered at the MgO(001) surface and deduced

an Arrhenius activation energy for desorption of Edes =

0.3 eV. On the other hand, Stark and Klabunde6 concluded

that HCl and HBr molecules chemisorb at MgO nanocluster

surfaces.

Interest in the adsorption behaviour of the MgO surface was

motivated also by its use as an efficient nanocrystalline dehy-

drohalogenation catalyst.7–9 Large surface area and enhanced

surface reactivity are responsible for a very high capacity of

MgO to destructively chemisorb volatile organic compounds.

It was observed that the adsorption rate of HCl varies widely

with temperature,7,8,12 so that chemisorption to special sites

has been suggested to be the dominant adsorption mechanism.

Adsorption of HCl to another polar surface, g-Al2O3, yields

appearance of a band attributed to hydroxyl (OH�) groups at

3500 cm�1 (104.9 THz).13,14 At the same time, no band

assignable to HCl was detected. This is indicative of dissocia-

tion of HCl due to adsorption. In later experimental work15

OH� groups formed at the surface were found to be mainly

acidic due to OH� � �Cl interaction.A study of HCl adsorption on a-Al2O3

16 yielded observa-

tion of a large range of desorption temperatures (300–500 K)

as the signature of adsorption at sites of different adsorption

energies (e.g. defect sites vs. terrace sites) without the possibility

for diffusion along the surface. A constant peak desorption

temperature from the Al2O3 surface at different levels of

coverage was interpreted to be the signature of associative

adsorption but, seemingly contradictively, the presence of

OH� signals was interpreted as a sign for dissociative adsorp-

tion.

HCl adsorption on MgO has been studied theoretically

by Chacon-Taylor and McCarthy.17 They used periodic

Hartree–Fock calculations with the Colle–Salvetti correlation

correction and predicted a relatively weak non-dissociative

physisorption of HCl at the oxygen site. They reported an

adsorption energy at one quarter monolayer coverage of about

0.48 eV. The predicted distance between H and O is 1.82 A.

Another physisorbed vertical configuration with chlorine

bonded to a surface magnesium ion was considered, with an

adsorption energy of 0.35 eV.

Both the experimental and theoretical data on HCl adsorp-

tion at the MgO(001) and alumina surfaces are not conclusive.

One can look for clues in other, similar systems. In particular,

water adsorption at MgO surfaces is perhaps best studied.

However, the character of water adsorption remains contro-

versial, too, as recently reviewed in ref. 18. It was also found to

a Theoretische Chemie, Technische Universitat Munchen,Lichtenbergstraße 4, 85 747 Garching, Germany.E-mail: andreas.markmann@ch.tum.de

bDepartment of Physics and Astronomy and London Centre forNanotechnology, University College London, Gower Street, London,UK WC1E 6BT

w The HTML version of this article has been enhanced with colourimages.

This journal is �c the Owner Societies 2006 Phys. Chem. Chem. Phys., 2006, 8, 4359–4367 | 4359

PAPER www.rsc.org/pccp | Physical Chemistry Chemical Physics

be defect related,19–23 although adsorption to a clean surface

terrace was observed at elevated vapour pressures.21,22 Theo-

retical calculations at low coverage predict weak non-disso-

ciative physisorption24–27 or even negative adsorption

energy.28,29 In contrast, both H+ and OH� ions adsorb readily

to the defect-free terraces.

Another polar molecule whose adsorption to the MgO(001)

surface has been studied theoretically is formic acid

(HCOOH). A second-order Møller–Plesset (MP2) calculation

of the interaction of formic acid with a quadratic cluster of 4 �4 MgO atoms performed by Nakatsuji et al.30 predicts a

minimum energy configuration in which the proton is disso-

ciated from the molecule and bound to a surface oxygen ion.

The adsorption energy predicted is about 2.2 eV. These cluster

calculations did not include the surface Madelung potential.

Recent, more accurate embedded cluster calculations31 predict

dissociative adsorption and formate ion formation at the

MgO(001) terrace. A plane wave density functional calcula-

tion of this system by Szymanski and Gillan32 demonstrated

that dissociative adsorption is more favorable than molecular

adsorption by about 0.3 eV. Experimentally, the dissociative

adsorption of formate on MgO(001) has been demonstrated

e.g. in ref. 33.

These results indicate that the character of adsorption can

depend on coverage and on the particular surface site of

adsorption, i.e. terrace, step, kink or a point defect, such as

a vacancy. In this paper, we study the mechanism of HCl

adsorption at the perfect MgO(001) surface terrace. Our

periodic plane wave and embedded cluster density functional

theory (DFT) simulations predict dissociative adsorption of

HCl with OH� formation due to the heterolytic splitting of

HCl at the surface.

The models used in this paper are described in section 2.

Results are presented in section 3 and discussed in section 4.

2. Models

To study the geometric and electronic structures of individual

HCl molecules and monolayers corresponding to different

surface coverage on the (001) MgO surface we employed two

methods: (i) periodic plane wave density functional theory

(PW-DFT) at the generalised gradient approximation (GGA)

level; and ii) the embedded cluster method (EC-DFT) with

atomic basis sets and the B3LYP hybrid density functional.

Both approaches have their advantages and limitations and

largely complement each other, as discussed in e.g. ref. 34.

PW-DFT calculations were performed in two rectangular

supercells containing four (001) MgO planes with surface

dimension of a0(4 � 4) and a0(6 � 6) (a0 = 2.11 A being the

nearest-neighbour Mg–O distance in the bulk crystal). With a

single HCl molecule in the supercell, this corresponds to a

coverage of 1/8 and 1/18 of a monolayer, respectively. The

vacuum gap between the slabs was set to twice the slab

thickness, which is sufficient to diminish interaction between

periodic images of the slab and leave sufficient space for a

desorbed molecule. The two layers of MgO not facing the

admolecule were fixed to the bulk structure, while the two

other layers were free to relax. We have used the VASP

package35,36 utilising PW-DFT with the PW’91 GGA density

functional by Perdew and Wang and ultrasoft pseudopoten-

tials of Vanderbilt type,37 as supplied by Kresse and Hafner.38

We used an energy cutoff of 396 eV and the G-point only in

the reduced Brillouin zone. The energy was converged to

an accuracy of 0.1 eV and the forces to an accuracy of

0.05 eV A�1.

In order to estimate the adsorption energy of a single HCl

molecule (zero coverage), the energies obtained in PW-DFT

calculations were corrected for a spurious dipole–dipole inter-

action between supercells. The correction energy was calcu-

lated using effective point charges and an algorithm due to

Kantorovitch39 with the effective charges –1.426 e at the

adsorption site oxygen, 0.426 e at hydrogen as parameterised

for a free OH� molecule by Saul and Catlow,40 where e

denotes the modulus of the electron charge. Formal ionic

charges were used for Cl, Mg, and all other O ions.

PW-DFT calculations as just described have two principal

limitations:

(1) The spurious electrostatic interaction arising from the

periodic boundary conditions is significant, especially in polar

systems such as those considered here. In extreme cases this

may lead to incorrect results even with the corrections just

discussed;34,41

(2) The self-interaction error inherent in all local and semi-

local functionals (including PW’91) may lead to qualitatively

incorrect predictions in systems where electronic interaction

leads to the formation of localised states.41–44

To quantify these shortcomings of the PW-DFT approach,

we also used the embedded cluster method (EC-DFT) de-

scribed in detail elsewhere.34,45–47 The setup is illustrated in

Fig. 1a. A quantum cluster (QC) is embedded in a finite array

of point charges located near the MgO lattice sites. The part of

the ions closest to the QC (region I) is treated within the shell

model. Ions of region I interact with each other and with the

atoms in the QC via shell model potentials of Stoneham and

Sangster.48 The positions of cores and shells of classical ions in

region I are optimised self-consistently with the ionic positions

and charge density in the QC. The ions outside region I

constitute region II and are represented by non-polarisable

point charges fixed at their lattice sites and serve to provide a

correct electrostatic potential on all ions in region I. The entire

simulated system measures 19 by 19 ions along the surface and

7 ions perpendicular to the surface (i.e. about 40 � 40 � 15

A3). In order to avoid artificial polarisation of the oxygen ions

in the QC,34,47 the boundary of the quantum cluster Mg29O13

(Fig. 1b and c) is formed by effective core potential (ECP)

Mg2+ cations.49 The standard basis sets 6–311G* for O, Cl

and H and 6–31G for Mg ions were employed. These basis sets

have been validated previously in studying the electronic

properties of defects at the MgO(001) surface. The electronic

structure and geometry of the QC was calculated using the

hybrid B3LYP density functional.50,51 The self-consistent

embedding procedure as described in this section has been

implemented in the computer program GUESS,46 which in-

corporates GAUSSIAN9852 for the electronic structure calcu-

lations.

A study of defects in an oxide surface comparing PW-DFT

and cluster results has been performed previously in ref. 53 but

there, pure cluster calculations (with no embedding) were used

4360 | Phys. Chem. Chem. Phys., 2006, 8, 4359–4367 This journal is �c the Owner Societies 2006

and the PW-DFT method was parameterised according to the

cluster results. In contrast, we use EC-DFT, which models

the surface more accurately than a pure cluster and compare

the results of PW-DFT and EC-DFT without fitting one to the

other.

3. Results

We start the discussion of our results from a brief description

of surface structures obtained using both methods. As evi-

denced by experiments, the surface does not reconstruct, so

anions and cations must retain the perfect crystal structure

parallel to the surface (x and y directions) but normal to the

surface (in the z direction), the symmetry is broken, so that

anions and cations can have altered z-coordinates near the

surface. The resulting surface relaxation and rumpling can be

expressed in percentages of the lattice constant. The surface

relaxation is the change dd12 of the spacing d12 between the

surface and subsurface layers. It is given as a percentage of the

perfect lattice spacing (half of the lattice constant) d

dd12 ¼d12 � d

d100%: ð1Þ

The surface rumpling D1 is defined as the difference between

the z-coordinates of anions and cations, again as percentage of

the ideal lattice spacing

D1 ¼zanion1 � zcation1

d100%: ð2Þ

The calculated surface relaxation and rumpling parameters for

MgO are +0.81% and+3.6% in PW-DFT, and+0.05% and

+2.9% in EC-DFT. Relaxation and rumpling in the EC-DFT

model are less reliable due to the constraint caused by the

unrelaxed ions of region II and the relatively small basis set on

the oxygen ions. Additionally, the surface relaxation para-

meter is very sensitive to the choice of d (eqn (1)), which is

difficult to pinpoint in EC-DFT as this method features no

periodicity. Given this limitation, the obtained rumpling para-

meters satisfactorily reproduce earlier LDA calculations and

LEED experiments on vacuum cleaved samples (see ref. 54 for

a review).

In order to identify the stable adsorption configurations of

the HCl molecule on the (001) MgO terrace, we examined a

number of possible conformations including HCl bonding to a

surface oxygen site, surface Mg site, or adsorbed over a plane

interstitial site. However, our calculations predict stable ad-

sorption only at a surface oxygen site with a formation of OH�

and Cl� ions in close proximity. Several configurations (ab-

breviated ‘‘conf.’’) are found to be stable or metastable and

will be discussed in the following: The symmetric, upright HCl

molecule over the adsorption site oxygen (conf. (1)) is meta-

stable and relaxes by symmetry breaking into conf. (2),

corresponding to an H–Cl axis tilted in the (100) direction

(towards a neighbouring magnesium ion) or into conf. (3)

where the H–Cl axis is tilted in the (110) direction (towards

plane interstitial position). Conf. (4) corresponds to HCl

heterolytically dissociated at the crystal surface with the pro-

ton sitting on top of an oxygen and the chlorine on top of a

next-nearest neighbour magnesium. Calculated configura-

tional parameters and corresponding adsorption energies of

these states are summarised in Fig. 2. PW-DFT energies are

shown for the HCl coverage of 1/8 and 1/18 monolayer,

respectively, as E4�4 and E6�6. EC-DFT energies are denoted

Eads(EC). We define the adsorption energy of a molecule at a

surface as the sum of the calculated energies of a free molecule

and a clean surface minus the calculated energy of the

adsorbed molecule. Thus, a bound system has positive adsorp-

tion energy.

Most PW-DFT calculations were performed with the

a0(4 � 4) unit cell, i.e. at 1/8 monolayer coverage due to com-

putational constraints. If not specified, this unit cell is implied

for all PW-DFT results presented below. The geometries

predicted by PW-DFT and EC-DFT agree very well, i.e. the

PW-DFT unit cells are large enough to approximate an

isolated molecule and the dipole–dipole correction yields a

good extrapolation to its energy. A significant improvement is

seen in the larger a0(6� 6) unit cell only for the subtle response

of the surface ions around the adsorption site to adsorption of

HCl. The distance between the proton and chlorine in the

associated adsorption geometries (1)–(3) is, at between 1.70

and 2.05 A, considerably larger than the 1.3 A in a free

molecule. The distance between the proton and the surface

oxygen at the adsorption site is 1.0 A, i.e. similar to the 0.96 A

in a free OH� molecule. The small distances of the proton to

the surface oxygen and the large magnitude of the adsorption

energies of above 1 eV suggest dissociative chemisorption

rather than physisorption as the dominating adsorption me-

chanism. The proton always points away from the surface in

Fig. 1 Setup of HCl on MgO(001) embedded cluster calculations. An Mg29O13 cluster was used. (a) Embedding of the quantum cluster into a

lattice of shell model ions, (b) cross-section through the quantum cluster, (c) three-dimensional view of the quantum cluster. The ECP magnesium

ions surround the cluster oxygen ions at the interface.

This journal is �c the Owner Societies 2006 Phys. Chem. Chem. Phys., 2006, 8, 4359–4367 | 4361

order to minimise the repulsion with the magnesium ions

surrounding the adsorption site oxygen.

The dissociative character of HCl adsorption is confirmed

by an analysis of the electronic structure of the ground state

geometry (conf. 2). Projected single-particle densities of states

(DOS) obtained in PW-DFT and EC-DFT calculations are

shown in Fig. 3a and b. The EC-DFT single-particle band gap

of 6.3 eV (B3LYP) may be juxtaposed with the MgO surface

exciton energy of 6.15–6.2 eV measured using electron-energy-

loss spectroscopy55,56 (compare with a bulk MgO band gap of

7.8 eV57). The band gap predicted by PW-DFT is much

smaller at 2.6 eV. This is a well known limitation of local

and semilocal (e.g. GGA) density functionals.

The adsorbed molecule does not induce states in the main

band gap. However, a few low lying gap-localised and valence

band resonant states are evident. Two states split from the

bottom of the second and valence band (at �20.2 eV and �7eV) are strongly localised on the oxygen adsorption site and

hydrogen. These are low lying states of the OH� molecule,

which are not much perturbed by the rest of the ions. There are

also two OH� states resonant with the valence band. The state

at �13 eV has a strong Cl 3s character and almost zero

contribution from the hydrogen ion. Three states related to

Cl 3p orbitals are seen as resonances near the top of the

valence band. This molecular orbital sequence confirms that,

despite the fact that the adsorbed H and Cl ions are close, HCl

adsorption is dissociative with the formation of an effectively

positive OH� ion and a negative Cl� electrostatically coupled

to it.

Table 1 summarises calculated ionic charges deduced from

natural population analysis (NPA) of the results obtained

from EC-DFT and from integration within ionic Wigner-Seitz

spheres for the results of PW-DFT. Again, charges on O at the

adsorption site, H and Cl consistent with dissociative adsorp-

tion are predicted by both methods.

Isosurface plots of the difference in PW-DFT charge density

between a calculation with an admolecule and one without for

conf. (2) are presented in Fig. 4a. In the free HCl molecule,

valence electron density is shared between hydrogen and

chlorine. Adsorption of HCl to the MgO(001) surface adds

no such density to the clean MgO surface, as can be seen from

the charge density difference plot. This means that the covalent

bond between H and Cl is effectively broken. Instead, a large

amount of charge density is donated from the surface oxygen

to the proton indicating formation of a covalently bonded

OH�molecule. Since OH� sits at the lattice position of an O2�

ion, it is positively charged (i.e. a hole) in the context of the

crystal lattice. Chlorine carries almost a full electron charge

and is bound to the OH� adsorption site due to electrostatic

attraction. The dipole due to this polarity between the OH�

hole and the Cl� charge is responsible for the lowering of the

adsorption energy per molecule at finite coverages.

Conf. (4), whose density difference plot is shown in Fig. 4b,

shows a pronounced change in polarisation of the underlying

oxygen crystal sublattice. As Cl� is well separated from the

OH� hole, its charge can now cause stronger polarisation of

the surrounding oxygen ions, which plays an important role in

the stabilisation of this configuration. The proton also has a

Fig. 2 Stable geometries of an adsorbed HCl molecule at the MgO terrace from a0(4 � 4) PW-DFT and EC-DFT calculations. Only atoms

surrounding the adsorption site are shown for clarity. Configuration energies EPW/EC-DFT, adsorption energies E4�4 and E6�6 for corresponding

PW-DFT supercells, dipole–dipole corrected adsorption energies Eads(corr), EC-DFT adsorption energies Eads(EC) (positive values indicate bound

geometries, all energies in eV). Geometry: angles and distances between atoms [PW-DFT results in braces]. Dipole–dipole interaction calculated

using an algorithm by Kantorovich39 with a point charge representation of the supercell and subtracted from energy of adsorbed system to arrive

at correction. No correction given for conf. 4, as there higher order terms in the multipole expansion dominate.

4362 | Phys. Chem. Chem. Phys., 2006, 8, 4359–4367 This journal is �c the Owner Societies 2006

more pronounced impact on the polarisation of its partner

oxygen. The latter is no longer polarised laterally by chlorine,

as the Cl� charge is effectively screened by the chlorine

neighbour polarisation.

Having established the structure of the stable states, we

proceed with a discussion of the activation barriers. The

barrier between configurations 2 and 3 gives an estimate for

the rotational barrier of the Cl� ion around the adsorption

site. It is very small, ato0.05 eV in both EC and PW-DFT, so

Cl� can be considered as freely rotating in a horizontal plane

around the OH bond. We also calculated the barriers for the

Cl� diffusion away from the OH� site. A direct transition

between conf. (3) and conf. (4) is not viable due to the strong

electrostatic repulsion between Cl� and the surface oxygen

O2�. The migration path via conf. (2) is found to be more

favourable energetically. A calculation with Cl� between

nearest and next-nearest neighbour Mg2+ ions (i.e. between

configurations 2 and 4) yields an energy less than 0.1 eV above

the energy of conf. (4). This corresponds to a barrier of 0.65 eV

(PW-DFT) and 0.78 eV (EC-DFT) for Cl� diffusion away

from the OH� site, and o0.1 eV for diffusion towards the

OH� site. The calculated dissociation energy of the HCl

molecule into H+ and Cl� at the MgO surface is 1.8 eV

(EC-DFT). The dissociated system comprises a proton, with

an adsorption energy of 12.3 eV and a Cl� ion with an

adsorption energy of 1.7 eV. Charged configurations were

not simulated with the PW-DFT method due to known

problems with compensation by charged backgrounds. In

contrast, the adsorption energies for H0 and Cl0 are 0.3 and

1.5 eV, respectively from both the EC and PW-DFT methods.

A diagram summarising all the processes discussed in the

previous paragraphs and their energies is shown in Fig. 5. It

can be checked that a process leading back to the starting

configuration in the diagram always results in a total transi-

tion energy of zero (reversing energies when proceeding

against the direction of an arrow). This diagram can be used

to assess processes not discussed so far. For example, removal

of chlorine as a charged ion from the surface with an asso-

ciated molecule costs 1.8 eV to dissociate OH� and Cl� at the

surface plus 1.7 eV to remove the isolated Cl� from the

surface, i.e. 3.5 eV in total. In comparison, removing Cl0 from

the adsorbed molecule as the neutral atom costs 3.5 eV alone

for dissociation at the surface plus 1.5 eV for desorption of

Cl0. This means that removal of Cl� as an anion is preferred

over the removal of Cl0 as a neutral atom and the potential

energy curves for removal of chlorine for the neutral and

anionic charge states do not cross. This is in stark contrast to

most molecular systems, for example the free HCl molecule,

where neutral separation (at 3.9 eV) is favoured over charged

separation (14.4 eV). However, both dissociation processes at

the surface are not very fast at room temperature due to their

high activation energy.

Next we discuss the effect of HCl surface coverage on the

adsorption energy. To this end, we have performed PW-DFT

calculations of up to 8 HCl molecules in the a0(4 � 4) surface

unit cell, and one molecule in the a0(6 � 6) unit cell. As the

molecule only binds to the oxygen sites, this corresponds to

coverages in the interval from 1/18th monolayer (5.6%) up to

a full monolayer of HCl molecules. Calculations predict

adsorption energies in configuration 2 (Fig. 3) to be 0.25,

0.83, 1.0 and 1.4 eV for 1/2, 1/4 , 1/8 and 1/18 monolayer,

respectively.

A coverage of 1/4 monolayer corresponds to that modelled

by Chacon-Taylor and McCarthy.17 Our value of 0.83 eV is

substantially larger than the 0.48 eV in ref. 17. However, for

1/4 monolayer coverage in conf. (2), the average OH� bond

length elongates by 0.1 A to 1.08 A and the average H–Cl

distance contracts by 0.3 A to 1.75 A. This indicates that HCl

adsorption is more like a hydrogen-bridge bond for high

coverages. Together with the tendency of Hartree–Fock to

underestimate binding energies, this explains that a different

geometry and physisorption were predicted in ref. 17.

Fig. 3 Electronic densities of state of the surface–molecule system in

conf. (2). (a) Projected densities of state (DOS) from PW-DFT. (b)

DOS from EC-DFT.

Table 1 Comparison of charges obtained from PW-DFT and EC-DFT for the tilted geometry, configuration (2). Charges are given in multiplesof the positive unit charge e. It can be seen that the spherical integration charges in the first row always deviate towards positive charge comparedto the NPA charges. This supports the NPA charges, as some (negatively charged) electron density is neglected for spherical integration

Method of calculation (population analysis) Cl H O in OH� Other O Mg

PW-DFT (spherical integration) �0.9 0.5 �1.4 �1.7 1.8EC-DFT (Mulliken analysis) �0.73 0.40 �1.10 �1.55 1.31EC-DFT (NPA) �0.85 0.47 �1.42 �1.79 1.71

This journal is �c the Owner Societies 2006 Phys. Chem. Chem. Phys., 2006, 8, 4359–4367 | 4363

This tendency can also be seen in the binding energy of the

OH� molecule, which is 2.9 eV with Hartree–Fock and 4.4 eV

with B3LYP. The B3LYP result is much closer to the value of

4.9 eV we calculated with Multi-Reference Configuration

Interaction (MRCI), the most accurate method we have used

for this molecule (all done with basis sets at the triple zeta

level).

Our results show that the coverage dependence of the

adsorption energy is dominated by the dipole–dipole repulsion

between the molecules which almost completely can be ac-

counted for by Ewald summation over the molecule effective

charges in the lateral directions. This is demonstrated in Fig. 6,

where we have plotted the difference between the adsorption

energy calculated with EC-DFT (1.54 eV, conf. 2) and

PW-DFT adsorption energies per molecule at different cov-

erages. This line is matched (taking into account error bars for

the crude point charge model) by the dipole–dipole repulsion

energy.

Coverage of every surface oxygen with an HCl molecule

(full monolayer) results in an unstable system, i.e. upon

relaxation the molecules migrate into the vacuum. However,

an additional HCl molecule can be adsorbed on top of a half

monolayer of HCl with an adsorption energy of 0.3 eV. The

additional HCl molecule is oriented with its positive (i.e.

hydrogen) end towards the underlying Cl� ion. As the lower-

most proton is bonded chemically to a surface oxygen and the

first and second Cl� layer are chemically equivalent, the

Fig. 4 Isosurface plots of the difference in charge density between

calculations of a surface with an admolecule and without. Surfaces

shown match a charge density difference value of rcutoff = 0.56 e A�3.

(a) Adsorption of an associated molecule, configuration (2). The

charge density change on the far oxygen ion is due to the PBC. (b)

Adsorption of a dissociated molecule, configuration (4) [Cl sits over

the magnesium ion (210) from the proton adsorption site]. The

apparent cutoff of the density on the leftmost O ion is due to a

limitation of the plot area. More oxygen ions are polarised in (b) which

partially accounts for the discrepancy between the differences in H–Cl

electrostatic energy and PW-DFT energy.

Fig. 5 Adsorption and dissociation energies of HCl. Negative energies stand for energy gained and positive ones for energy to spend. Left side:

dissociation into charged species. Right side: dissociation into neutral species.

Fig. 6 This comparison shows that the dipole–dipole interaction

energy is the dominant contribution in the coverage dependence of

the adsorption energy per admolecule. Conf. (2), dipole–dipole en-

ergies evaluated with a rigid ion model of the unit cell (hence the large

error bars). The energy difference is Eads(EC-DFT) � Eads(PW-DFT),

where Eads(EC-DFT) = 1.5 eV.

4364 | Phys. Chem. Chem. Phys., 2006, 8, 4359–4367 This journal is �c the Owner Societies 2006

proton in the middle mediates a hydrogen bond between the

first and second admolecule layer.

We have used the EC-DFT method to calculate vibrational

frequencies of some localised vibrational modes. Table 2

shows an overview of mode frequencies found for OH� in

conf. (2) and in the protonated surface, i.e. without the

associated chlorine ion. It can be seen that the non-transla-

tional modes have frequencies which are well-separated from

the top of the MgO surface slab vibrational band (22 THz in

shell model lattice dynamics58). This means they are unlikely

to couple strongly to the normal modes of the MgO substrate.

The electrostatic attraction of the chlorine ion softens the

translational and stretch modes of OH�, while it hardens the

librational modes with respect to the proton without asso-

ciated Cl�. The changes in the intramolecular OH� frequen-

cies are due mostly to the slight elongation of the OH� bond

which makes OH� reside in the sloping part of its (approxi-

mately Morse-type) potential.

In experimental adsorption of HCl to MgO nanocrystals6 at

high pressure, broad bands from 3000 to 3700 cm�1 were

measured. The broad bandwidth measured indicates a wide

range of OH� intramolecular distances present in the sample

due to dipole–dipole repulsion and chlorine diffusion.

Chlorine readily diffuses away from the adsorption site

(most likely to surface defects) at high coverage: a Cl� ion

can rotate freely around its associated OH� but it can also

rotate freely around an OH� ion at the neighbouring oxygen

lattice position as one is likely to be available at the high

coverages in these experiments, which can remove chlorine

from its adsorption site. Hence, the activation energy for Cl�

diffusion goes from 0.6 eV for an isolated molecule to below

0.05 eV for high coverage.

A band at 3750 cm�1 was present prior to HCl adsorption

due to surface contamination with OH�.6 Note that these

OH� species are not embedded in the surface like the one

formed due to HCl adsorption at the terrace. Bands at 1600

and 1700 cm�1 appeared on HCl adsorption which are in the

region of the water bending mode (1400–1600 cm�1, depend-

ing on isotopes).59 These are likely due to hydrochloric acid

protonating the contaminating OH� species already present at

the surface.

4. Discussion

The results presented above clearly demonstrate that at low

coverage the HCl molecule adsorbs dissociatively on the MgO

(001) terrace. Our DFT calculations predict adsorption of HCl

to the MgO(001) surface at low coverage to be caused pri-

marily by a chemical bond between a surface oxygen and the

molecular proton. The adsorption energy in the most stable

configurations, (2) and (3), is 1.5 eV. In unison with the

breaking of the covalent bond between H and Cl, their

distance increases to between 1.7 and 2.0 A, compared to 1.3

A in the free molecule. Chlorine is held at the adsorption site

by its electrostatic attraction to the OH� molecule, which

represents a hole in the context of the crystal. Hence HCl

at the surface is chemically dissociated but electrostatically

associated.

EC-DFT simulations of HCl adsorbed to the MgO(001)

surface confirmed and extended our PW-DFT results. The fact

that both methods using different basis sets and density

functionals predict the dissociative character of the adsorption

further supports our conclusions.

In comparison to molecular hydrogen, which adsorbs to

MgO dissociatively only at low-coordinated special sites or

vacancies60–63 and to water, which does so at steps, vacancies

or corners of nanocrystals,19–23 HCl adsorbs dissociatively at

the MgO(001) surface terrace. In this respect, HCl represents a

limiting case. Out of the HX compounds, where X is a halogen

ion, dissociative adsorption to the MgO(001) terrace may be

hindered only for HF, which has a dissociation energy of

almost 6 eV. For all other hydrohalogenic acids, dissociative

adsorption is to be expected.

The adsorption energy of HCl may be viewed as consisting

of the HCl vacuum dissociation energy, the protonation

energy and the adsorption energy of Cl� to the proton

adsorption site (Fig. 5). The accuracy of the surface protona-

tion component determines the accuracy of the total adsorp-

tion energy. It has been determined previously by embedded

cluster Hartree–Fock (HF) calculations64 at 10.4 eV and by

flame-ion mass spectrometry experiments65 at around 10 eV.

The difference between our EC-DFT result of 12.3 eV and HF

can be attributed to the tendency of Hartree–Fock to under-

estimate binding energies and of GGA DFT to overestimate

them slightly. The mixed B3LYP functional, however, is fairly

reliable, as demonstrated for the OH� molecule in section 3.

The same tendencies of the methods, together with lowering

of the adsorption energy due to coverage effects is responsible

for the previous prediction of physisorption of HCl on

MgO(001).17

The adsorption energy of 1.5 eV predicted by our results is

significantly higher than that estimated in molecular scattering

experiments. Wittig et al.10 derived an activation energy of

0.3 eV by applying the Arrhenius equation to the measured

desorption rate. As we have demonstrated that the adsorption

energy per HCl molecule decreases with coverage, it may be

supposed that the discrepancy to our predicted adsorption

energy was due to coverage. However, the experiment gives

Table 2 Overview of OH� mode frequencies from EC-DFT calculations. The electrostatic attraction to the associated chlorine ion softens theOH� translational and stretch modes and hardens the librational mode

OH� in HCl on MgO terrace OH� in protonated terrace

Mode Wavenumbers/cm�1 Mode Wavenumbers/cm�1

Translational modes 312–394 Translational modes 235–290Libration 1007 Libration 835Rotation 1021 Libration 835Stretch 3083 Stretch 3648

This journal is �c the Owner Societies 2006 Phys. Chem. Chem. Phys., 2006, 8, 4359–4367 | 4365

repeatable results for many molecular pulses,66 which indicates

that the surface coverage remains low throughout the experi-

ment. Due to the ratio of HCl in the molecular beam (1–2%)

the surface coverage is estimated to be of the order of 10�3

monolayers HCl.66 Coverage can therefore be excluded as a

reason for the discrepancy.

Instead we propose a dynamical mechanism: OH� at the

surface has vibrational and librational frequencies well sepa-

rated from the bulk phonon density of states of MgO. We

argue that such spectrally and spatially localised modes are

characterised by a low dissipation rate, and thus can retain

vibrational excitations for a considerably long time. The

smallest incident kinetic energy in the experiments of Wittig

et al. corresponds to a velocity of 763 m s�1. At an incidence

angle of 151, this corresponds to an approach velocity per-

pendicular to the surface of 737 m s�1. Assuming an interac-

tion region of 5 A above the surface (a generous estimate, as

the Madelung potential decays exponentially with distance),

the molecule has about 1.35 ps to reorient, dissipate its excess

energy and get trapped. If this succeeds, a chemical bond

between the proton and a surface oxygen can be formed. The

dissipated energy (adsorption energy + translational kinetic

energy of the molecule) is likely to be initially transformed into

a vibrational excitation of the OH� stretching mode. The

vibrational frequencies of OH� libration and vibration

(30–92 THz) are well above the maximum crystal frequency

(omax(MgO) = 22 THz). Therefore, excitation of these modes

may survive for many pico- or even nanoseconds due to the

spectral localisation of these modes. However, the activation

energy for desorption of a ‘hot’ molecule is the difference

between the vacuum level and the stored vibrational energy.

This can significantly reduce the measured desorption energy

in scattering experiments, assuming that mostly desorption of

‘hot’ molecules is observed. Accounting for the zero vibration

energy of the OH� stretch mode lowers the desorption barrier

by approximately 0.2 eV ( 1.4 eV � (0.38 eV/2)E 1.2 eV). The

desorption energy involving OH� in its first vibrational excited

state would lower this energy by a further 0.3 eV, and so on.

Thermalised rotational distributions at all incident energies

were previously interpreted as a sign of trapping–desorption.10

However, molecular dynamics67 predict that the collision of

randomly oriented diatomic molecules with a surface should

lead to such a rotational distribution naturally. The depen-

dence of the characteristic rotational temperature on the

temperature of the crystal can be ascribed to events where

molecules collide with oncoming surface atoms. Overall, this

demonstrates that the experimental observations in ref. 10, 11,

13–15 are not contradicting the theoretical predictions pre-

sented in this article. At special adsorption sites, such as steps

and kinks, the frequency of the stretch mode of OH� should

be increased due to a stronger chemical bond. This means that

the OH� stretch mode is spectrally separated further from the

MgO band, maintaining weak coupling. Hence, the qualitative

features of the dynamics of HCl adsorbed on a step or kink

oxygen should be similar to adsorption at the terrace.

Low-kinetic energy, low coverage trapping–desorption ex-

periments involving H35Cl and D37Cl may illuminate whether

HCl molecules dissociate and recombine (in mixed combina-

tions) at the surface prior to desorption. This would yield an

unambiguous answer to the question whether molecules dis-

sociate and whether chlorine can diffuse at the surface at long

time scales.

Acknowledgements

The authors are indebted to P. V. Sushko for his support

running GUESS and to L. Kantorovich for aiding the calcula-

tion of dipole–dipole interaction energies. Both colleagues

have provided useful further advice throughout the project.

W. Eisfeld has lent support for the MRCI calculation of the

OH� molecule. A. Markmann would like to thank the Physics

and Astronomy Department of University College London for

a studentship and the Delbrucksche Familienstiftung for

further financial support. J. L. Gavartin would like to thank

the Leverhulme trust for funding.

References

1 A. D. Pradhan, K. P. Kirby and A. Dalgarno, J. Chem. Phys.,1991, 95, 9009.

2 C. Toubin, S. Picaud and C. Girardet, Chem. Phys., 1999, 244, 227.3 J. B. Fegley, in Treatise on Geochemistry, ed. Andrew M. Davisand Heinrich D. Holland, Elsevier, Amsterdam, 2003, vol. 1,pp. 487–507.

4 T. Seideman and H. Guo, J. Chem. Phys., 1995, 103, 2745.5 M. Hintenender, F. Rebentrost, R. B. Gerber and R. Kosloff,J. Chem. Phys., 1995, 102, 578.

6 J. V. Stark and K. J. Klabunde, Chem. Mater., 1996, 8, 1913.7 U. Rappold and G. Luft, Chem. Eng. Technol., 1999, 22, 843.8 M. Li and H. Shaw, Ind. Eng. Chem. Res., 2000, 39, 1898.9 I. V. Mishakov, A. F. Bedilo, R. M. Richards, V. V. Chesnokov,A. M. Volodin, V. I. Zaikovskii, R. A. Buyanov and K. J.Klabunde, J. Catal., 2002, 206, 40.

10 M. Korolik, D. W. Arnold, M. J. Johnson, M. M. Suchan, H.Reisler and C. Wittig, Chem. Phys. Lett., 1998, 284, 164.

11 M. Korolik, M. M. Suchan, M. J. Johnson, D. W. Arnold, H.Reisler and C. Wittig, Chem. Phys. Lett., 2000, 326, 11.

12 Y. Kajiwara, N. Yoshida and S. Kishimoto, J. Colloid InterfaceSci., 1987, 118, 297.

13 J. B. Peri, J. Phys. Chem., 1965, 70, 1482.14 M. Tanaka and S. Ogasawara, J. Catal., 1970, 16, 157.15 A. Kytokivi and M. Lindblad, J. Chem. Soc., Faraday Trans.,

1995, 91, 941.16 J. W. Elam, C. E. Nelson, M. A. Tolbert and S. M. George, Surf.

Sci., 2000, 450, 64.17 M. R. Chacon-Taylor and M. I. McCarthy, J. Phys. Chem., 1996,

100, 7610.18 M. A. Henderson, Surf. Sci. Rep., 2002, 46, 1.19 W. Langel and M. Parrinello, Phys. Rev. Lett., 1994, 73, 504.20 J. L. Anchell and A. C. Hess, J. Phys. Chem., 1996, 100, 18317.21 P. Liu, T. Kendelewicz, J. Gordon, E. Brown and G. A. Parks,

Surf. Sci., 1998, 412/413, 287.22 P. Liu, T. Kendelewicz and J. Gordon E. Brown, Surf. Sci., 1998,

412/413, 315.23 F. Finocchi and J. Goniakowski, Phys. Rev. B, 2001, 64, 125426.24 D. Ferry, C. Girardet and J. Suzanne, Surf. Sci., 1998, 409, 101.25 S. Joyce and B. Kay, J. Chem. Phys., 1996, 105, 1295.26 M. A. Johnson, E. V. Stefanovich, T. N. Truong, G. Gunster and

D. W. Goodman, J. Phys. Chem. B, 1999, 103, 3391.27 M. I. McCarthy, G. K. Schenter, C. A. Scamehorn and J. B.

Nicholas, J. Phys. Chem., 1996, 100, 16989.28 A. L. Shluger, J. D. Gale and C. R. A. Catlow, J. Phys. Chem.,

1992, 96, 10389.29 J. Goniakowski, S. Bouette-Russo and C. Noguera, Surf. Sci.,

1993, 284, 315.30 H. Nakatsuji, M. Yoshimoto, M. Hada, K. Domen and C. Hirose,

Surf. Sci., 1995, 336, 232.31 M. L. Sushko, A. Y. Gal, M. Watkins and A. L. Shluger,

Nanotechnology, 2006, 17, 2062.

4366 | Phys. Chem. Chem. Phys., 2006, 8, 4359–4367 This journal is �c the Owner Societies 2006

32 M. Szymanski and M. Gillan, Surf. Sci., 1996, 367, 135.33 X. D. Peng and M. A. Barteau, Catal. Lett., 1991, 7, 395.34 A. Shluger, A. Foster, J. Gavartin and P. Sushko, inNano and Giga

Challenges in Microelectronics, ed. J. Greer, A. Korkin and J.Labanowski, Elsevier, Amsterdam, 2003.

35 G. Kresse and J. Hafner, Phys. Rev. B, 1993, 47, 558.36 G. Kresse and J. Furthmuller, Phys. Rev. B, 1996, 54, 11169.37 D. Vanderbilt, Phys. Rev. B, 1990, 41, 7892.38 G. Kresse and J. Hafner, J. Phys.: Condens. Matter, 1994, 6, 8245.39 L. N. Kantorovich, Phys. Rev. B, 1999, 60, 15476.40 P. Saul and C. R. A. Catlow, Philos. Mag. B, 1985, 51, 107.41 J. L. Gavartin, P. V. Sushko and A. L. Shluger, Phys. Rev. B, 2003,

67, 035108.42 G. Pacchioni, F. Frigoli, D. Ricci and J. A. Weil, Phys. Rev. B,

2001, 63, 054102.43 J. Lægsgaard and K. Stokbro, Phys. Rev. Lett., 2001, 86, 2834.44 J. L. Gavartin and A. L. Shluger, Phys. Rev. B, 2001, 64, 245111.45 P. V. Sushko and A. L. Shluger, Surf. Sci., 1999, 421(3), L157.46 P. V. Sushko, A. L. Shluger and C. R. A. Catlow, Surf. Sci., 2000,

450, 153.47 P. V. Sushko, J. L. Gavartin and A. L. Shluger, J. Phys. Chem. B,

2002, 106, 2269.48 A.M. Stoneham andM. J. L. Sangster, Philos. Mag. B, 1985, 52, 717.49 W. R. Wadt and P. J. Hay, J. Chem. Phys., 1985, 82, 284.50 A. D. Becke, J. Chem. Phys., 1993, 98, 5648.51 C. Lee, W. Yang and R. G. Parr, Phys. Rev. B, 1988, 37, 785.52 M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A.

Robb, J. R. Cheeseman, V. G. Zakrzewski, J. A. Montgomery, Jr,R. E. Stratmann, J. C. Burant, S. Dapprich, J. M. Millam, A. D.Daniels, K. N. Kudin, M. C. Strain, O. Farkas, J. Tomasi, V.Barone, M. Cossi, R. Cammi, B. Mennucci, C. Pomelli, C. Adamo,S. Clifford, J. Ochterski, G. A. Petersson, P. Y. Ayala, Q. Cui, K.

Morokuma, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B.Foresman, J. Cioslowski, J. V. Ortiz, A. G. Baboul, B. B. Stefanov,G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. Gomperts, R.L. Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A.Nanayakkara, C. Gonzalez, M. Challacombe, P. M. W. Gill, B.Johnson, W. Chen, M. W. Wong, J. L. Andres, C. Gonzalez, M.Head-Gordon, E. S. Replogle and J. A. Pople, GAUSSIAN 98(Revision a.7), Gaussian, Inc., Pittsburgh, PA, 1998.

53 R. Coquet and D. J. Willock, Phys. Chem. Chem. Phys., 2005, 7,3819.

54 A. R. Gerson and T. Bredow, Phys. Chem. Chem. Phys., 1999, 1,4889.

55 V. E. Henrich, G. Dresselhaus and H. J. Zeiger, Phys. Rev. B, 1980,22, 4764.

56 P. A. Cox and A. A. Williams, Surf. Sci., 1986, 175, L782.57 D. M. Roessler and W. C. Walker, Phys. Rev., 1967, 159, 733.58 A. Markmann, J. L. Gavartin and A. L. Shluger, unpublished

work.59 P. F. Bernath, Phys. Chem. Chem. Phys., 2002, 4, 1501.60 K. Sawabe, N. Koga, K. Morokuma and Y. Iwasawa, J. Chem.

Phys., 1994, 101, 4819.61 H. Kobayashi, D. Salahub and T. Ito, J. Phys. Chem., 1994, 98,

5487.62 H. Kobayashi and T. Ito, Nippon Kagaku Kaishi, 1996, 9, 765.63 E. N. Gribov, S. Bertarione, D. Scarano, C. Lamberti, G. Spoto

and A. Zecchina, J. Phys. Chem. B, 2004, 108, 16174.64 J. Oviedo, C. J. Calzado and J. F. Sanz, J. Chem. Phys., 1997, 108,

4219.65 Q. Chen and J. M. Goodings, Int. J. Mass Spectrom., 1998, 181,

181.66 M. Suchan, private communication.67 J. Blomer and A. E. Beylich, Surf. Sci., 1999, 423, 127.

This journal is �c the Owner Societies 2006 Phys. Chem. Chem. Phys., 2006, 8, 4359–4367 | 4367