Atomic scale insights on chlorinated γ-alumina surfaces

Article abstract of DOI:10.1021/ja8019593

The thermochemistry of chlorinated γ-alumina surfaces is explored by means of density functional calculations as a function of relevant reaction conditions used in experiments and in high-octane fuel production in the refining industry such as hydrocarbon isomerization and reforming. The role of chlorine as a dope of the Bronsted acidity of γ-alumina surfaces is investigated at an atomic scale. Combining infrared spectroscopy and density functional theory calculations, the most favorable location of chlorine atoms on the (110), (100) and (111) surfaces of γ-alumina is found to result either from direct adsorption or from the exchange of basic hydroxyl groups. Moreover, the modification of the hydrogen bond network upon chlorine adsorption is put forward as a key parameter for changing the Bronsted acidity. In a second step, we use a thermodynamic approach based on DFT total energy calculations corrected by the chemical potentials of HCl and H₂O to determine the adsorption isotherms of chlorine and the relative surface concentration of hydroxyl groups and chlorine species on the γ-alumina surfaces. The determination of chlorine content as a function of temperature and partial pressures of H₂O and HCl offers new quantitative data required for optimizing the state of the support surface in industrial conditions. The mechanisms of chlorination are also discussed as a function of reaction conditions.

Full text of DOI:10.1021/ja8019593

Published on Web 07/23/2008
Atomic Scale Insights on Chlorinated γ-Alumina Surfaces
Mathieu Digne,† Pascal Raybaud,*,‡ Philippe Sautet,§ Denis Guillaume,‡ and
Herve´ Toulhoat
Direction Physique et Analyse, Direction Catalyse et Se´paration, IFP, Rond-point de
l’e´changeur de Solaize, BP 3, 69390 Solaize, France, UniVersite´ de Lyon, Institut de Chimie,
Laboratoire de Chimie, Ecole normale supe´rieure de Lyon, CNRS 46, alle´e d’Italie,
69364 Lyon cedex 07, France, and Direction Scientifique, IFP, 1-4 aVenue de Bois-Pre´au,
92852 Rueil-Malmaison cedex, France
Received March 17, 2008; E-mail: pascal.raybaud@ifp.fr
Abstract: The thermochemistry of chlorinated γ-alumina surfaces is explored by means of density functional
calculations as a function of relevant reaction conditions used in experiments and in high-octane fuel
production in the refining industry such as hydrocarbon isomerization and reforming. The role of chlorine
as a dope of the Brønsted acidity of γ-alumina surfaces is investigated at an atomic scale. Combining
infrared spectroscopy and density functional theory calculations, the most favorable location of chlorine
atoms on the (110), (100) and (111) surfaces of γ-alumina is found to result either from direct adsorption
or from the exchange of basic hydroxyl groups. Moreover, the modification of the hydrogen bond network
upon chlorine adsorption is put forward as a key parameter for changing the Brønsted acidity. In a second
step, we use a thermodynamic approach based on DFT total energy calculations corrected by the chemical
potentials of HCl and H2O to determine the adsorption isotherms of chlorine and the relative surface
concentration of hydroxyl groups and chlorine species on the γ-alumina surfaces. The determination of
chlorine content as a function of temperature and partial pressures of H2O and HCl offers new quantitative
data required for optimizing the state of the support surface in industrial conditions. The mechanisms of
chlorination are also discussed as a function of reaction conditions.
Introduction
ment of the γ-Al2O3 properties is at the core of many future
challenges for the refining, petrochemical, and energy industry.
The metastable γ-polymorph of alumina (γ-Al2O3) is an
important oxide material used as the support of numerous
catalytic active phases such as metals or transition metal sulfides.
From the mesoscale to the nanoscale, γ-Al2O3 exhibits very
versatile physical chemical properties1 required for refining
catalytic processes, such as hydrotreatment to reduce the sulfur
content of diesel,2 or naphtha reforming or isomerization to
increase the gasoline octane number.3,4 Moreover, the naphtha
reforming process involving dehydrogenation reactions of
alcanes and naphtenes is also the main source of hydrogen in
refineries. The hydrogen balance will remain an increasing
concern of refiners within the context of biofuels production or
coal liquids refining, which need higher consumption of H2 for
reducing the oxygen content. As a consequence, the improve-
At the mesoscale, the high specific area of the γ-Al2O3
support (usually around 200 m2 per gram) favors the high
dispersion degree of the active-phase crystallites exhibiting
nanometer sizes or even less, as for reforming catalysts where
the metallic active phase is made of small clusters for which
the Pt-Pt coordination number measured by X-ray absorption
spectroscopy (XAFS) is less than 7.5,6 At the atomic scale, the
acidic-basic properties of the oxide support can directly be
involved during the catalytic reactions. Again, a relevant
example is highlighted by bifunctional reforming catalysts such
as γ-alumina-supported bimetallic PtX bimetallic (X ) Re, Ir,
or Sn). The (de)hydrogenation and hydrogenolysis reactions are
taking place on the metallic active phase, whereas the acidic
properties of the support enhance the cracking and isomerization
reactions.4,5 For these reasons, understanding how to control
the acidic-basic properties of the support’s surfaces at the
atomic scale is a challenging fundamental scientific issue with
practical industrial applications. In the case of industrial
reforming and isomerization catalysts, the surface acidity of
γ-Al2O3 is modified by adsorbing halogens (such as Cl or F)
on the surface. A large number of modifying agents have been
tested for different catalytic applications, even if chlorine
remains one of the most widely used dopes to modify the surface
† Direction Physique et Analyse, IFP.
‡ Direction Catalyse et Se´paration, IFP.
§ Universite´ de Lyon, Institut de Chimie.
Direction Scientifique, IFP.
(1) Euzen, P.; Raybaud, P.; Krokidis, X.; Toulhoat, H.; Loarer, J.-L. L.;
Jolivet, J.-P.; Froidefond, C. In Handbook of Porous Solids; Schu¨th,
F., Sing, K. S. W., Weitkamp, J., Eds.; Wiley-VCH Verlag GmbH:
Weinheim, 2002; Vol. 3.
(2) Prins, R. In Handbook of Heterogeneous Catalysis; Ertl, G., Kno¨zinger,
H., Weitkamp, J., Eds.; Wiley-VCH Verlag GmbH: Weinheim, 1997;
Vol. 4.
(3) Ertl, G., Kno¨zinger, H., Weitkamp, J., Eds. The Handbook of
Heterogeneous Catalysis; Wiley-VCH: Weinheim, 1997.
(4) Marcilly, C. Acid-Base Catalysis: Application to Refining and Petro-
chemistry; Edition Technip: Paris, 2005; Vol. 2, Chapter 10.
(5) Sinfelt, J. H. In Handbook of Heterogeneous Catalysis; Ertl, G.,
Kno¨zinger, H., Weitkamp, J., Eds.; Wiley-VCH: Weinheim, 1997; Vol.
3, Chapter 3.9.
(6) Lynch, J. Oil Gas Sci. Technol. 2002, 57, 281.
9
11030 J. AM. CHEM. SOC. 2008, 130, 11030–11039
10.1021/ja8019593 CCC: $40.75
2008 American Chemical Society

Atomic-Scale Insights on Chlorinated γ-Alumina Surfaces
A R T I C L E S
acidity. For instance, the chlorination of alumina is used to
improve alkylation7 or cracking8 reactions. The chlorination of
the γ-alumina surfaces is accomplished by the use of various
chlorinating agents (HCl, CCl4, or AlCl3). The quantity of
chlorine fixed and the resulting acidity depend on the chlorinat-
ing agents and of the reaction conditions (temperature, partial
pressure of chlorinating agents, and water) used for the
chlorination. For reforming, the quantity of chlorine added is
about 1-2 wt %, whereas for isomerization, the chlorine content
may reach 6-7 wt %. The chlorination occurs always by
chemical exchange of the surface hydroxyl (-OH) groups.4,9
Since this exchange is reversible, the chemical potential of water
in the reaction environment must be controlled because an
excess of water could be detrimental to the stability of chlorine
at the surface. In addition, it is also known that the chlorine
addition may prevent the sintering of metallic clusters by
ensuring high dispersion state and low Pt-Pt coordination
number.6 In industrial conditions, to maintain this high disper-
sion state, the catalyst is continuously regenerated under a
chlorinated atmosphere. Moreover, depending on the chlorine
content of the catalyst’s support, this coordination number may
reach values below 4.6,10 Moreover, XAFS studies seem to
indicate that the nature of the support may modify the electronic
features of small Pt clusters11 and, as a result, the hydrogen
adsorption property of those clusters.12 Recent density functional
theory (DFT) studies have underlined the role of the hydration
state of the γ-alumina surfaces on the adsorption and diffusion
properties of Pdn (1 e n e 5) clusters.13,14 The hydration state
influences the distribution of hydroxyl (-OH) groups and Lewis
acid aluminum sites. In particular, it has been shown that the
adsorption mode of ethylene on γ-alumina-supported Pd4
clusters depends on the hydroxylation state of the γ-alumina
surfaces.15 As a consequence, further improvements of metallic
active phases highly dispersed on doped γ-alumina surfaces are
expected to be directly linked with deeper understanding of the
nano- and atomic-scale properties of the doped surfaces. A
challenging question concerns the role of the hydroxylation state
and crystallographic orientation of the γ-alumina surfaces when
doped by anionic and cationic species. However, the use of small
Pt4/(X2O)3 clusters (X ) F, H, Na) as proposed in refs 12 and
16 cannot be fully satisfactory to describe the surface state of
the support. This approach may raise questions about the
overestimate of the effect of the X2O ligand and the simulta-
neous underestimate of surface rearrangements which modify
the nature of the chemical interactions between Pt clusters and
the doped γ-alumina surfaces. In contrast, recent periodic DFT
calculations have brought many relevant insights about the
bulk17–19 and surface properties 20–22 of γ-alumina under realistic
environment. In particular, the effect of various reaction
conditions such as temperature, partial pressures of reactants
20,21,23–25
(such as H2O, H2S, H2)
were included to determine
the precise surface state of γ-alumina. More recently, we have
also proposed a rational model based on DFT calculations and
infrared (IR) characterization to explain the poisoning of the
γ-alumina surface acidity by the sodium cation.26 In particular,
Na cations were found to be located in an inner sphere complex
of surface hydroxyls and oxygen atoms after exchanging protons
from surface µ3-OH groups. This model’s chemistry significantly
differs from that of the cluster model used in reference 12.
However, even if the adsorption of HCl molecules on ideal
dehydrated R-alumina (0001) surface has been investigated by
periodic DFT simulation,27 the challenging study of chlorinated
γ-alumina surface remained to be undertaken.
According to the crucial role of chlorine as surface modifier
of γ-alumina properties, the determination of reliable surface
models of chlorinated γ-alumina is an important question. In
the context of reforming, the balance between H2O and HCl is
known to be critical for optimizing the physical chemical
properties of the metallic active phase. In this paper, we thus
propose to use density functional theory (DFT) and infrared
(IR) analysis of chlorinated γ-alumina nanocrystalline samples
synthesized in well-defined conditions in order to build realistic
theoretical models of chlorinated γ-Al2O3 surfaces for various
chlorine coverages. First the stable location of the chlorine
elements on the surface is determined, and the impact of the
surface OH stretching frequencies is discussed with respect to
the IR experimental results. The influence of chlorine on surface
acidity is also studied by means of pyridine adsorption calcula-
tions. Then, a thermodynamic model is proposed in combination
with DFT results to determine the domains of stability of
hydroxyl and chlorine species at the surface. Finally, the
calculation of chlorine contents of the γ-alumina surface as a
function of temperature and the partial pressures of H2O and
HCl is presented. It is shown how this model is expected to
find practical applications in the reforming and isomerization
processes, where these parameters are fixed in order to control
the catalytic surface state.
Theoretical Methods
Total Energy Calculation. The computational method used
for total energy calculation is coherent with the one used in
our previous published works where the hydroxylated surfaces
(17) Krokidis, X.; Raybaud, P.; Gobichon, A.-E.; Rebours, B.; Euzen, P.;
Toulhoat, H. J. Phys. Chem. B 2001, 105, 5121.
(18) Wolverton, C.; Haas, K. C. Phys. ReV. B 2001, 63, 024102.
(19) Paglia, G.; Buckley, C. E.; Rohl, A. L.; Hunter, B. A.; Hart, R. D.;
Hanna, J. V.; Byrne, L. T. Phys. ReV. B 2003, 68, 144110.
(20) Digne, M.; Sautet, P.; Raybaud, P.; Euzen, P.; Toulhoat, H. J. Catal.
2002, 211, 1.
(7) Clet, G.; Goupil, J. M.; Szabo, G.; Cornet, D. J. Mol. Catal. A: Chem.
1999, 148, 253.
(8) Guillaume, D.; Gautier, S.; Alario, F.; Deve`s, J. M. Oil Gas Sci.
Technol. 1997, 54 (4), 537.
(9) Gates, B. C.; Katzer, J. R.; Schuit, G. C. A. Chemistry of Catalytic
Processes; McGraw-Hill: New York, 1979.
(21) Digne, M.; Sautet, P.; Raybaud, P.; Euzen, P.; Toulhoat, H. J. Catal.
2004, 226, 54.
(10) Berdala, J.; Freund, E.; Lynch, J. P. J. Phys. (Paris) 1986, 47, C8–
269.
(22) Valero, M. C.; Digne, M.; Sautet, P.; Raybaud, P. Oil Gas Sci. Technol.
2006, 61, 535.
(11) Ramaker, D. E.; Oudenhuijzen, M. K.; Koningsberger, D. C. J. Phys.
Chem. B 2005, 109, 5608.
(23) Arrouvel, C.; Digne, M.; Breysse, M.; Toulhoat, H.; Raybaud, P. J.
Catal. 2004, 222, 152.
(12) Oudenhuijzen, M. K.; Bokhoven, J. A. v.; Miller, J. T.; Ramaker, D. E.;
Koningsberger, D. C. J. Am. Chem. Soc. 2005, 127, 1530.
(13) Valero, M. C.; Raybaud, P.; Sautet, P. J. Phys. Chem. B 2006, 110,
1759.
(24) Arrouvel, C.; Breysse, M.; Toulhoat, H.; Raybaud, P. J. Catal. 2004,
226, 260.
(25) Joubert, J.; Fleurat-Lessard, P.; Delbecq, F.; Sautet, P. J. Phys. Chem.
B 2006, 110, 7392.
(14) Valero, M. C.; Raybaud, P.; Sautet, P. Phys. ReV. B 2007, 045427.
(15) Valero, M. C.; Raybaud, P.; Sautet, P. J. Catal. 2007, 247, 339.
(16) Thomson, J.; Webb, G.; Webster, B. C.; Winfield, J. M. J. Chem.
Soc., Faraday Trans. 1995, 91, 155.
(26) Digne, M.; Raybaud, P.; Sautet, P.; Guillaume, D.; Toulhoat, H. Phys.
Chem. Chem. Phys. 2007, 9, 2577.
(27) Alavi, S.; Sorescu, D. C.; Thompson, D. L. J. Phys. Chem. B 2003,
107, 186.
9
J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008 11031

A R T I C L E S
Digne et al.
Figure 1. Local structures of the hydroxylated γ-alumina surfaces: (a) γ-Al2O3(100) with θOH ) 8.8 OH ·nm-2 and (b) γ-Al2O3(110) with θOH ) 11.8
OH·nm-2; (c) γ-Al2O3(111) with θOH ) 14.7 OH ·nm-2 in absence of chlorine. The most relevant surface OH groups are labeled: AlX stands for aluminum
atoms surrounded by X oxygen atoms and HO-µn for OH groups linked to n aluminum atoms.
hydroxyl coverages as high as 8.8 OH·nm-2 on the (100)
Table 1. Substitution of the Surface Hydroxyl Groups by Chlorine:
surface, 11.8 OH ·nm-2 on the (110) surface and 14.7 OH ·nm-2
Effect of the Local Environment
site
surface
ωcalc (cm-1
)
∆rE (kJ·mol-1
)
on the (111) surface as shown in Figure 1.
Thermochemistry of Chlorinated Surfaces. For low chlorine
content, a first approach to study the chlorine effect is to assume
HO-µ1-AIIV
(110)
(100)
(110)
(110)
(100)
(110)
(110)
(100)
(111)
(111)
(111)
3842
3777
3736
3707
3589
3206
3717
3616
3752
3732
3712
+19
+2
HO-µ1-AlVI
HO-µ1-AlV
-4
that the total surface coverage (θH O + θHCl) remains constant
2
HO-µ2-AlVI
+29
+95
+42
+31
+12
+119
+58
+26
during chlorination (this assumption will be discussed in detail
in the Results section). The surface coverage depends on the
experimental conditions: temperature and partial pressures of
water and hydrogen chloride. According to the slab supercells
used and presented in Figure 1, the substitution of a single OH
HO-µ3-AlVI
HO-µ1-AlIV (H-bonded)
H2O-µ1-AlV
H2O-µ1-AlVI
HO-µ3-AlVI
group increases the chlorine coverage, θHCl, by 1 to 2 Cl·nm-2
.
HO-µ2-AlVI
HO-µ1-AlVI
On the basis of this assumption, the substitution of the
different surfaces hydroxyl groups has been considered:
of non doped γ-alumina surface were determined.20,21 The
calculations are based on density functional theory (DFT), as
implemented in the Vienna Ab initio Simulation Package
(VASP),28,29 with the Perdew-Wang 91 (PW91) generalized
gradient-corrected exchange-correlation functional.30,31 Ultrasoft
pseudopotentials are used for each atom. Periodic boundary
conditions are set, leading to an infinite periodically repeated
system. The wave functions of the Kohn-Sham Hamiltonian
are expanded on a plane wave basis set with an energy cutoff
of 300 eV. For each atomic configuration, atomic forces are
calculated via the Hellmann-Feynman theorem, and the
geometry optimization is performed with a conjugate-gradient
algorithm.
Surface Models. The (hkl) surfaces, simulated using a slab
model, are generated with a least eight atomic layers in the (hkl)
direction. A vacuum of 12 Å is set between two periodically
repeated slabs. Each slab is kept symmetric to avoid unphysical
dipolar interaction between two consecutive slabs. In this study,
we focus on the three preferentially exposed surfaces of
γ-alumina in working conditions: the (100), (110), and (111)
surfaces which were identified by neutron diffraction analysis
and electron microscopy.32,33 Our previous DFT works have
shown that the hydroxyl concentration at the surfaces depends
on temperature and the partial pressure of water.21,23 According
to the temperatures exposed during chlorine exchange and
infrared experiments (Vide infra), the surfaces may exhibit
-Aln(OH)- + HClgas f -Aln′(Cl)- + H2Ogas
(1)
where n (respectively, n′) indicates the coordination numbers
of the surface OH (respectively, Cl) species.
The total energy balance of reaction 1 can be written as
follows:
∆Eexch ) E(-Aln′(Cl)-) + E(H2O) - E(-Aln′(OH)-) -
E(HCl) (2)
Considering the fact that the entropy contributions of the H2O
and HCl molecules are very close over a wide range of
temperatures,34 the total energy variation as calculated by eq 2
is a first reasonable approximation to evaluate the thermody-
namic affinity of chlorine for exchanging different surface OH
groups.
For higher chlorine contents depending on the relative partial
pressures of H2O and HCl, a more complex thermodynamic
model must be proposed. The stabilization energy is first
calculated with respect to the reference energy, E(surf), of the
fully dehydrated and dechlorinated surface used as reference:
Estab ) E(surf + nH OH2O + nHClHCl) - E(surf) -
2
n
H OE(H2O) - nHClE(HCl) (3)
2
where E(surf + nH OH2O + nHClHCl) is the total energy of the
2
surface with nH O (respectively, nHCl) molecules of H2O
2
(respectively, HCl) adsorbed.
The adsorption and coadsorption energies of H2O and HCl
for different hydroxyl and chlorine coverage are thus systemati-
cally calculated by eq 3. Then, to investigate the thermochem-
istry of chlorinated γ-alumina surfaces, the effects of entropies,
(28) Kresse, G.; Furthmu¨ller, J. Phys. ReV. B 1996, 54, 11169.
(29) Kresse, G.; Furthmu¨ller, J. Comput. Mater. Sci. 1996, 6, 15.
(30) Perdew, J. P.; Wang, Y. Phys. ReV. B 1992, 45, 13244.
(31) Perdew, J. P.; Chevary, J. A.; Vosko, S. H.; Jackson, K. A.; Pederson,
M. R.; Singh, D. J.; Fiolhais, C. Phys. ReV. B 1992, 46, 6671.
(32) Beaufils, J. P.; Barbaux, Y. J. Chim. Phys. 1981, 78, 347.
(33) Nortier, P.; Fourre, P.; Saad, A. B. M.; Saur, O.; Lavalley, J.-C. Appl.
Catal. 1990, 61, 141.
(34) Lide, D. R., Ed. Handbook of Chemistry and Physics; CRC Press:
New York, 1995.
9
11032 J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008

Atomic-Scale Insights on Chlorinated γ-Alumina Surfaces
A R T I C L E S
Figure 2. Local structures of chlorinated γ-alumina surfaces: (a) γ-Al2O3(100) for 2.15 Cl ·nm-2 and 6.45 OH ·nm-2 with Cl in µ1 position (a.1) and in µ3
position (a.2), (b) γ-Al2O3(110) for different chlorine and hydroxyl contents: (b.1) 1.5 Cl ·nm-2 and 4.5 OH ·nm-2, (b.2) 3.0 Cl ·nm-2 and 8.9 OH ·nm-2
(b.3) 5.9 Cl ·nm-2 and 5.9 OH·nm-2, (c.1) γ-Al2O3(111) for 1.2 Cl ·nm-2 and 13.5 OH ·nm-2 (Al-Cl distances are quoted in Å).
,
193 m2 ·g-1. Prior to additivation, the sample was calcined at 803
K under flowing dry air for 2 h. Chlorination was performed by
submitting the sample to a H2O-HCl-air stream at 773 K for 2 h.
HCl is obtained by the in situ decomposition of 1,2-dichloropropane
in air, leading to CO2, H2O, and HCl. After a calcination at 773 K
for 2 h, chlorine content, determined by X-ray fluorescence, was
1.43 wt %. It was thus possible to control the chlorine content at
temperature, and partial pressures found in the working condi-
tions must be added. For that purpose, we use the following
thermodynamic model including the chemical potential of H2O
and HCl molecules in gas phase.
Gstab(nH O,nHCl) ) Estab(nH O,nHCl) - nH OµH O - nHClµHCl
2
2
2
2
(4)
the calcination stage by an atmosphere containing water and
36,37
chlorinating agents as also observed by others
following the
where
chemical process represented by the direct equation (1). A similar
operation (called oxychloration) is used to regenerate the industrial
catalyst.
Next, a partial dechlorination of the support was performed by
calcining the sample at 803 K under air with about 20.000 wt ppm
of H2O. By varying the time of calcinations, γ-Al2O3 with 1.14
and 0.85 wt % of chlorine, were obtained according to the reverse
equation (2).
Infrared (IR) Spectroscopy. Fourier transform infrared spec-
troscopy has been performed using Digilab TTS80 spectrometer.
Samples were used as self-supporting wafers of 20 mg, activated
in situ 2 h under vacuum of 10-6 mbar at 773 K. Spectra were
recorded at room temperature with a resolution of 4 cm-1, and the
number of scans was fixed to 300.
µ
H O(T, pH O) ) h0H O - TsH O(T) + RT ln(pH O ⁄ p0)
2 2 2 2 2
(5)
An analogous expression for the chemical potential of HCl is
used.
In the Results section, the number of adsorbed molecules will
be normalized by surface area, resulting in coverage values.
Theoretical Characterization of Hydroxyl Group Acidity. To
evaluate the intrinsic Brønsted acidity of OH groups on the
chlorinated surface, the stretching frequencies of OH groups
are calculated, using an harmonic approach. The Hessian matrix
is calculated by a numerical finite-difference method: each ion
is displaced of 0.005 Å around its equilibrium position, in the
three directions of space. Due to the high anharmonicity of OH
bonds, these harmonic values are corrected by an anharmonicity
correction term of 80 cm-1 as calculated for similar OH groups
of boehmite γ-AlOOH.35
Results and Discussion
Low Chlorine Coverages. The enthalpy variation, associated
to eq 1, is reported in Table 1. The optimized structures for the
most relevant configurations after chlorine exchange are reported
in Figure 2. In most cases, the substitution of OH by Cl is an
endothermic process. The general trend can be resumed as
follows, whatever the crystallographic planes considered. (i) The
substitution of the µ3-OH and µ2-OH groups is a endothermic
process (from +29 to +119 kJ ·mol-1), (ii) the substitution of
the µ1-OH groups is an athermic or slightly endothermic process
(from -4 to +36 kJ·mol-1), (iii) the existence of hydrogen
bonds tends to destabilize the substitution by chlorine atoms,
and (iv) the substitution of chemisorbed water molecules is
slightly endothermic and leads to the dissociation of the HCl
molecule.
Furthermore the adsorption energy of the pyridine molecule
was calculated to further characterize the Brønsted acidity of
the chlorinated surface by means of the following relationships:
∆Eads ) E(surf + pyridine) - E(surf) - E(pyridine) (6)
where E(surf) is the total energies of the surface, E(pyridine)
of the isolated gas-phase pyridine, and E(surf + pyridine) of
the adsorbed molecule on the surface. A negative value,
corresponding to an exothermic process, indicates an energeti-
cally favored adsorption process.
Experimental Section
Preparation of Chlorinated γ-Alumina. The starting material
was a γ-Al2O3 sample from Axens with a specific surface area of
(36) Castro, A. A.; Scelza, O. A.; Benvenuto, E. R.; Baronetti, G. T.; Parera,
J. M. J. Catal. 1981, 69, 222.
(37) Castro, A. A.; Scelza, O. A.; Baronetti, G. T.; Fritzler, M.; Parera,
J. M. Appl. Catal. 1983, 6, 347.
(35) Raybaud, P.; Digne, M.; Iftmie, R.; Wellens, W.; Euzen, P.; Toulhoat,
H. J. Catal. 2001, 201, 236.
9
J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008 11033

A R T I C L E S
Digne et al.
Figure 3. Effect of chlorine content on the high-frequency region of the IR spectrum. A pure γ-alumina sample is compared to samples with increasing
chlorine content.
Results (i) and (ii) can be explained by the differences
between the optimal distances for Al-OH bonds (1.8-2.0 Å)
and for Al-Cl bonds (2.2-2.4 Å). For a µ1 species, a single
distance has to be optimized: the Cl atom relaxes on the surface
in order to match the optimal Al-Cl distance (for instance, 2.39
Å for the AlVI-µ1 group of the (100) surface, Figure 2 a.1). For
the µ2 and µ3 groups, two or three Al-Cl have to be optimized,
the aluminum atoms remaining mainly fixed due the alumina
network. The geometrical optimization leads to strong local
deformations, energetically unfavorable (Figure 2 a.2). The
impact of the hydrogen bond network on the substitution
enthalpy is due to the fact that the O-H· · ·Cl hydrogen bond
is weaker than the O-H· · ·OH bond. As a consequence, the
substitution of a hydrogen bond involving OH groups (donor
or acceptor) is more difficult than the substitution of a similar
OH group without hydrogen bonds. This effect is well illustrated
by the (111) surface, which exhibits a high hydroxyl coverage
(14.7 OH ·nm-2) and a high density of hydrogen bonds: the
substitution is always more endothermic, compared to the less
hydroxylated (100) and (111) (+119 kJ ·mol-1 for the (111)
HO-µ3-AlVI compared to +95 for the (100) HO-µ3-AlVI, +58
kJ·mol-1 for the (111) HO-µ2-AlVI compared to +29 for the
(110) HO-µ2-AlVI,...).
This part of our study provides an accurate localization of
the chlorine on the γ-alumina surfaces. Chlorine atoms form
surface Cl-µ1-Al groups in substitution of HO-µ1-Al groups.
Moreover, the presence of strong hydrogen-bond networks (as
found on the (111) surface) makes the Cl substitution energeti-
cally less favorable.
Vibrational Spectroscopy Analysis. The infrared spectra of
pure γ-alumina exhibits five high-frequency bands 3794, 3775,
3730, 3686, and 3670 cm-1 corresponding to the OH stretching
frequencies (Figure 3 a). According to our previous ab initio
simulations carried out,21 these bands have been assigned to
(110) HO-µ1-AlIV, (100) HO-µ1-AlVI, (110) HO-µ1-AlV, HO-
µ2, and HO-µ3. Numerous examples illustrating the parallel
between IR experiments and DFT simulations can be found in
a recent review.38 In what follows, we focus on the impact of
chlorine on this high-frequency region. The introduction of 0.7
Cl·nm-2 leads to the total disappearance of the bands located
at 3794 and 3775 cm-1. Increasing the Cl coverage, the intensity
of the band at 3730 cm-1 is decreasing, relative to the intensity
of HO-µ2 and HO-µ3 bands. Using the previous calculations of
chlorine exchange, the interpretation of these experimental
observations is straightforward. At very low chlorine content
(around 1 Cl·nm-2), the first OH species involved in the
exchange are preferentially the (110) HO-µ1-AlIV, the (110) HO-
µ1-AlV, and the (100) HO-µ1-AlVI groups. From a pure
spectroscopic analysis of OH stretching bond, these groups are
considered as the most basic ones, being also intrinsically the
most labile. The corresponding frequencies of these groups thus
disappear from the spectra. At the same time, the bands assigned
to the HO-µ2 and HO-µ3 groups are only slightly modified by
chlorine addition. A broadening of these bands is noticed for
the higher chlorine content. We propose the following explana-
tion of this broadening by the fact that on hydroxylated surfaces,
some OH groups are hydrogen bond donors toward HO-µ1
groups. This hydrogen bond induces a significant OH lengthen-
ing of the hydrogen bond donor group and a lowering of the
stretching frequency: for instance 1.02 Å and 3206 cm-1 for a
(110) HO-µ1-AlIV group. During chlorination, the HO-µ1 group
is exchanged into a Cl-µ1 group with a simultaneous weakening
of the hydrogen bond. For the same (110) HO-µ1-AlIV group,
the OH length decreases from 1.02 to 1.00 Å. Its frequency
increases from 3206 cm-1 to about 3566 cm-1. As a conse-
quence, the weakening of the hydrogen bonds due to the
presence of chlorine increases the stretching frequency of some
OH groups and explains the broadening of the bands in the
3700-3600 cm-1 region.
Pyridine Adsorption. A fruitful comparison of DFT calcula-
tions21 for pyridine adsorption with IR characterization39 was
recently provided for non-chlorinated γ-alumina surfaces. For
the chlorine-free surface, it was observed that pyridine molecules
adsorb on both Lewis and Brønsted acidic sites; however, the
formation of pyridinium ions was not observed. FTIR of
pyridine adsorbed on chlorinated γ-alumina revealed that the
acidity is increased40 and that the IR spectrum can exhibit a
(39) Morterra, C.; Magnacca, G. Catal. Today 1996, 27, 497.
(40) Berteau, P.; Kellens, M. A.; Delmon, B. J. Chem. Soc., Faraday Trans.
1991, 87, 1425.
(38) Meyer, R. J. Vib. Spectrosc. 2007, 43, 26.
9
11034 J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008

Atomic-Scale Insights on Chlorinated γ-Alumina Surfaces
A R T I C L E S
Figure 4. Adsorption of pyridine on the OH-µ2-AlVI group of the (100)
γ-alumina surface: (a) without exchanged chlorine atom on OH-µ1-AlVI
,
(b) with exchanged chlorine atom on the µ1-AlVI group and hydrogen bonded
complex, and (c) with exchanged chlorine atom on the µ1-AlVI group and
pyridinium ion formation. The Al-Cl, O-H and N-H distances are quoted
in Å.
band at 1540 cm-1, due to the formation of the pyridinium ion.41
In what follows, our main focus is to explore by DFT simulation
how the surface chlorination would influence adsorption of basic
probe molecules such as pyridine.
Figure 5. Surface stabilization energy as a function of the chlorine
coverage. The dark lines symbolize the iso-total coverage curves (θH O
+
2
θHCl constant) and dashed lines symbolize the iso-water coverage curves
(θH O constant). (Insert) Surface mechanisms associated when one moves
2
from one point of the diagram to an adjacent point.
The Brønsted acidity has been studied on the (100) surface
for 2.2 OH·nm-2 and 2.2 Cl·nm-2. This surface corresponds
to the dissociative adsorption of a single HCl molecule. Before
chlorination, the surface exhibits one HO-µ1-AlVI group and one
HO-µ2-AlVI group. The HO-µ2 is a hydrogen bond donor toward
the oxygen atom of the HO-µ1 groups. For both groups, pyridine
adsorbs on hydroxyl groups and forms the O-H· · ·N hydrogen-
bonded complex. The exothermic adsorption energy is -15
kJ·mol-1 for the HO-µ2 group (see Figure 4 a). After chlorina-
tion, the HO-µ1-AlVI group is converted into a Cl-µ1-AlVI group,
and the HO-µ2-AlVI group remains as a Brønsted site. The
adsorption energy of pyridine via hydrogen bond on this site
(figure 4 b) is strongly enhanced (-50 kJ ·mol-1), compared to
the non-chlorinated surface (-15 kJ ·mol-1). Even if this group
remains a hydrogen bond donor, the hydrogen bond O-H· · ·Cl
formed with Cl is weaker, compared to the O-H· · ·O before
chlorination. This implies that the proton of the O-H group
becomes more labile after chlorination. Hence, in presence of
pyridine, the stretching of the O-H bond length is increased
by +0.05 Å after chlorination. Moreover, by transferring the
proton to the nitrogen atom of pyridine, inducing the pyridinium
ion formation (Figure 4 c), the adsorption is even more stabilized
at -57 kJ·mol-1 against -50 kJ·mol-1 for hydrogen-bonded
graphs, it contains the most exchangeable hydroxyl groups in a
reasonably high concentration. Hence, to build the DFT
thermodynamic model quantifying the surface concentration of
chlorine in realistic p(H2O), p(HCl), and temperature (T)
conditions, we will consider only the (110) surface. Chlorine
coverages are explored up to 5.9 Cl ·nm-2. At this coverage,
each aluminum site is bonded to a single chlorine atom.
These values of the stabilization energies, Estab and Gstab, as
defined in the Theoretical Methods section, are reported in
Figure 5 for the same partial pressure of H2O and HCl and the
two relevant temperatures used either in isomerization (T ) 400
K) or in reforming (T ) 800 K).
The following general trends can be pointed out. First, when
the total number of adsorbed molecules, i.e. the total coverage,
increases, the average stabilization energy increases, meaning
that the surface tends to saturate the coordination sphere of the
Al and O surface atoms. Then, the substitution of the first OH
group by one chloride anion can be considered as an athermal
process, whatever the total coverage. This result is consistent
with the conclusions obtained for the low chloride content
surfaces and gives an extrapolation for the larger Cl coverages.
When the chlorine amount increases (beyond the chlorine
coverage of 1.4 Cl ·nm-2), the substitution of OH groups
becomes an endothermic process. For a given total coverage,
the surface is more and more destabilized when the amount of
chloride anions increases due to their increasing lateral repulsive
interaction. The fcc sublattice of oxygen atoms of γ-alumina
exhibits O-O distances between 2.6 and 2.9 Å. In contrast,
the reference AlCl3 lamellar material42 exhibits minimum Cl-Cl
distances of 3.1 and 3.3 Å. As a consequence, for large chlorine
amounts, the Cl-Cl distances on the surface layer become
frustrated, and repulsive anionic interactions destabilize the
system. Furthermore, in the case of the hydroxylated surfaces,
the lateral interactions of OH groups were partially stabilized
by hydrogen bonds. As already shown in the previous para-
graphs, the chlorine substitution destabilizes the hydrogen-bond
network which also contributes to endothermic energy. Finally,
for highly chlorinated alumina, the formation of the AlCl3 phase
cannot be excluded. Even if this is beyond the scope of the
current work, we suggest that the growth of such a phase would
complex formation. In line with the IR experiments reported in
40,41
the literature
and the OH stretching frequency analysis
described in the previous paragraph, the weakening of the
hydrogen-bond network by chlorine doping (making the HO-
µ2-AlVI group more available) may be at the origin of the acidity
increase as evaluated by pyridine adsorption.
Energy Models of Higher Chlorination Level of γ-Alumina. In the
previous paragraphs, we focused on surfaces with low or
moderate chlorine content (in the range 1-2 wt %). However,
it is well-known that the alumina surface acidity can be enhanced
by increasing the chlorine content even more. For instance, the
isomerization catalyst requires a strong acid function: the
chlorine content can usually reach 8 wt %, whereas for
reforming catalysts the chlorine content is around 2 wt %.4
It is also known experimentally that the (110) surface of
γ-alumina nanocrystallites is predominant (about 70-80%
according to32,33). Furthermore, as shown in previous para-
(41) Clet, G.; Goupil, J. M.; Cornet, D. Bull. Soc. Chim. Belg. 1997, 134,
223.
(42) Troyanov, S. I. Russ. J. Inorg. Chem. 1992, 37, 121.
9
J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008 11035

A R T I C L E S
Digne et al.
Hence, this expression corresponds to a Tempkin-Frumkin
adsorption isotherm type,43,44 which assumes that the heat of
adsorption decreases linearly with coverage due to adsorbate/
adsorbate interactions. RT is a fitting parameter, and ∆H0 is the
initial heat of adsorption which decreases when the initial
hydroxyl coverage increases: from 350 kJ ·mol-1 for a hydroxyl-
free surface to 124 kJ ·mol-1 for a surface with 8.9 OH ·nm-2
.
Our results show that a linear Tempkin-type of adsorption
isotherm would only be approximate and would not take into
the attractive hydrogen bonds (OH · · ·Cl) compensating the
repulsive adsorbate-adsorbate interaction.
Hydroxyl and Chlorine Surface Concentrations in Reaction
Conditions. This analytical expression, combined with the equa-
tion of stabilization energy (eq 7) is used to determine the stable
surface termination and the species surface concentration as a
function of temperature and hydrogen chloride and water partial
pressures. Two partial water pressures relevant for the reforming
Figure 6. Heat of adsorption of HCl as a function of the Cl surface coverage
and for different initial hydroxyl coverage (expressed in OH·nm-2). The
lines correspond to the analytical expressions derived from eq 7 and the
black squares to the ab initio calculated heats of adsorption. (Inset) Ab
initio stabilization energy versus the fitted stabilization energy (see the text
for parameters).
and isomerization processes have been considered: PH O ) 1
2
and PH O ) 0.001 bar (Figures 7 and 8). Concerning surface
2
hydroxyls (Figure 7), low temperatures favor high surface
concentration. The effect of HCl partial pressure on the OH
concentration depends on the water pressure and temperature.
In wet atmospheres (PH O ) 1 bar and T below 600 K), the OH
concentration remains roughly constant whatever the HCl partial
pressures (except very high HCl partial pressure), whereas for
2
require a process involving aluminum ion extraction (resulting
from strong surface reconstruction) followed by crystallization
resulting from combination of Al and Cl into AlCl3.
low water pressure (PH O ) 0.001 bar), the concentration
2
increases with increasing HCl partial pressure as a result of HCl
dissociative adsorption on Al and O free sites. Nevertheless,
the order of magnitude is not drastically different for the two
cases. This means that the number of potential Brønsted acid
sites is relatively stable, whatever the water content. As shown
in the previous section, the acid strength of theses sites strongly
depends on the chlorine content. Concerning the chlorine
concentration, the comparison between the two water partial
pressures demonstrates the strong impact of residual water on
In order to obtain a continuous description of surface
chlorination, instead of discrete values for specific surface
coverages, the stabilization energies, reported in the inset of
Figure 6, have been fitted using the following analytical
expression:
Estab(θH O,θHCl) ) -R ln(1 + AθH O) - ꢀ ln(1 + AθHCl) +
2
2
γ ln(1 + AθH O) ln(1 + AθHCl
)
(7)
2
chlorine coverage (Figure 8). For PH O ) 1 bar, it is impossible
2
where Estab is the stabilization energy, A the geometric area of
the (110) surface model, and R, ꢀ, γ are the fitted parameters.
Using standard fitting procedures, the following values are
obtained: R ) 404, ꢀ ) 350, and γ ) 163 kJ·mol-1. As shown
in the inset, the correlation is excellent (R2 ) 0.9907).
At this stage, it is required to add some comments about the
physical interpretation of the proposed analytic expression 7. It
is usual in experimental studies on adsorption isotherms to report
the heat of absorption of gaseous molecules as a function of
the molecular coverage. In our simulation, the heat of adsorption
corresponds actually to the derivative of the stabilization energy.
Assuming that the surface initially exhibits a known hydroxyl
to obtain high chlorine coverage, even for low temperature and
high HCl partial pressure. For PH O ) 0.001 bar, high chlorine
2
coverage (>3 Cl·nm-2) can be achieved even at 800 K.
Impact for the Reforming and Isomerization Processes. The
catalysts for reforming are always derived from platinum
deposited on chlorinated alumina. Chlorine is added to promote
the acid function of alumina. The chlorine content is usually
close to 1 wt %, and the operating temperatures are between
750 and 820 K. The chlorine contents (expressed in wt %) of
the alumina support and assuming a specific area of 200 m2 ·g-1
,
have been calculated as a function of HCl partial pressure and
for an average temperature of 800 K (Figure 9). At this
temperature, for water partial pressures between 0.001 and 1
bar, the differences of chlorine contents remain moderate and
roughly constant when the HCl partial pressures increase
(approximately 1 Cl wt %). If the contacting gas is correctly
dried, the 1 Cl wt % target is reached, even for low HCl content.
Platinum deposited on chlorinated γ-Al2O3 is also industrially
used as catalyst for light paraffin isomerization. In this case,
the acidic function of the catalyst must be enhanced by
increasing the chlorine content of the industrial catalyst which
comprises between 5 and 12 wt %. The isomerization reactions
are performed at low temperature, which comprises between
390 and 450 K.4 The chlorine content has been calculated for
a temperature of 400 K (Figure 9).
coverage (θH O), the heat of adsorption of hydrogen chlorine
2
on the (100) γ-alumina surface as a function of the Cl surface
coverage is given by:
∂Estab(θH O, θHCl
)
ꢀ - γ ln(1 + AθH O
)
2
2
∆H(θHCl) ) -
)
(
)
∂θHCl
1 + AθHCl
θH2
O
(8)
The heat of adsorption for different values of θH O (expressed
2
in OH ·nm-2) is plotted in Figure 6, evidencing the strong
impact of the initial hydroxyl coverage on the HCl adsorption
isotherm. For the low HCl coverage limit, the heat of adsorption
can be expressed as:
∆H(θHCl) ≈ ∆H0(θH O)(1 - RTθHCl
)
(9)
(43) Slygin, A.; Frumkin, P. Acta Physiochim. URSS 1935, 3, 791.
(44) Tempkin, M. I.; Pyshev, V. Acta Physiochim. URSS 1940, 69, 217.
2
9
11036 J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008

Atomic-Scale Insights on Chlorinated γ-Alumina Surfaces
A R T I C L E S
Figure 7. Hydroxyl coverage, in OH ·nm-2, as a function of temperature and hydrogen chloride partial pressure for PH O ) 1 bar and PH O ) 0.001 bar.
2
2
Figure 8. Chlorine coverage, in Cl·nm-2, as a function of temperature and hydrogen chloride partial pressure for PH O ) 1 bar and PH O ) 0.001 bar.
2
2
Figure 9. Chlorine content of γ-alumina (expressed in wt %, assuming a specific area of 200 m2 ·g-1) as a function of hydrogen chlorine partial pressure
for different conditions: for low residual water content (PH O ) 0.001 bar, full line) and for high residual water level (PH O ) 1 bar, dotted line), for T )
2
2
400 K (circles) and T ) 800 K (squares).
Compared to the reforming case, the residual water content
has a drastic impact on the chlorine content: for low temperature,
the difference between a dry and a wet atmosphere can be as
large as 6 Cl wt %. For moderate HCl partial pressure, the
chlorine content required for isomerization can only be reached
ature and partial pressures is crucial for an optimal control of
the surface state of γ-alumina in industrial conditions.
Mechanisms of Surface Chlorination. In what follows we
address the open question of surface chlorination mechanisms,
often discussed in literature. Indeed, gaseous HCl molecules
can react with the surface through the exchange of one Al-OH
group by one Al-Cl group with the release of one water
molecule (eq 1). A second chemical scenario is the dissociative
adsorption of the HCl molecule on free Al sites adjacent to O
sites, leading to the formation of new Al-Cl and Al-OH
for dry atmospheres (PH O ) 0.001 bar).
2
The thermodynamic model constructed from DFT calculations
of HCl and H2O coadsorption isotherms renders coherent and
quantitative trends in surface species distribution for two
industrial supports. The interpretation of the effect of temper-
9
J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008 11037

A R T I C L E S
Digne et al.
Figure 10. Mechanisms of surface evolution of the (100) γ-alumina surface as a function of the final temperature and HCl pressure (at fixed PH O equal to
2
0.001 bar). The diagram assumes that the initial surface corresponds to the stable termination for T ) 800 K, and PHCl ) 0.016 bar (see text for explanations).
The vertical dashed line corresponds to P(HCl) ) 0.25 bar.
surface groups. In the latter case, the appearance of new OH
groups, as observed by IR spectroscopic studies by some
authors,45,46 supports the dissociative adsorption mechanism.
In the former case, other authors have shown that, in some
experimental conditions, the equilibrium chlorine content as a
function the H2O/HCl molar ratio can be interpreted, assuming
that only exchanges between OH and Cl surface groups are
involved.37 Moreover, by combining FTIR and 1H MAS NMR
it has been found that the exchange reaction occurs whatever
the experimental conditions and that the adsorption mechanism
becomes significant when the reaction temperature decreases.47
More recently, by applying IR spectroscopy, TPD, and inelastic
neutron scattering spectroscopy for η-alumina,48 the existence
of both mechanisms has been confirmed and correlated to the
nature of the Lewis acid site. Using the previously established
stability diagrams of OH and Cl species on the (110) surface,
it is possible to also obtain insights about the mechanisms of
surface chlorination. The nature of the involved mechanisms is
strongly dependent on the initial and final conditions (temper-
ature, water and hydrogen chlorine pressures) that are consid-
ered. As these mechanisms are crucial for the reforming process,
we consider the stable termination of the (110) γ-alumina surface
to 1.7 (H2O+HCl)·nm-2. This iso-coverage curve is represented
by the red curve in Figure 10. Along this curve, the substitution
is the only possible mechanism: OHfCl for PHCl > 0.016 bar
and ClfOH for PHCl < 0.016 bar. As qualitatively expected,
when the HCl partial pressure increases (respectively decreases),
the total pressure increases (respectively decreases); thus, the
temperature must be increased (respectively decreased), if one
wishes to maintain the total coverage constant. For temperatures
higher than the iso-coverage temperature, the total coverage
tends to decrease. For instance, if the final HCl pressure is equal
to 0.25 bar (log (PHCl/P°) ) -0.6), the iso-coverage temperature
is about 845 K. For final temperatures between 845 and 1000
K, water molecules will desorb and some OH surface groups
will be substituted by Cl surface groups. As a consequence,
the surface is chlorine-enriched, and the total coverage decreases,
leading to the formation of surface unsaturated groups. For final
temperatures higher than 1000 K, both HCl and H2O molecules
will desorb. For temperatures lower than the iso-coverage
temperature, the total coverage tends to increase. Keeping the
same example (log (PHCl/P°) ) -0.6), for temperatures lower
than 845 K, the surface will be chlorine-enriched. For temper-
atures between 725 and 845 K, OHfCl substitution and HCl
dissociative absorption will occur simultaneously: for temper-
ature higher than 800 K, the substitution is the predominant
mechanism. For temperature lower than 725 K, the absorption
of both HCl and H2O molecules will occur, leading to the
progressive saturation of coordinately unsaturated surface sites.
Similar results can be obtained for final HCl pressures lower
than the initial HCl pressure (0.016 bar): the chlorine surface
content will tend to decrease. As a conclusion, the nature of
the mechanism involved during surface chlorination is complex
and is strongly affected by the initial and final thermodynamic
close to 800 K, with PH O ) 0.001 bar andPHCl ) 0.016 bar
2
(θH O ) 1.0 H2O·nm-2 and θHCl ) 0.7 HCl·nm-2) as the starting
2
conditions used in typical reforming units. The different domains
of surface evolution as a function of the final temperature and
HCl pressure (the water pressure remaining constant) are
reported in Figure 10. For a given temperature, a single HCl
pressure allows to keep the total coverage (θH O + θHCl) equal
2
(45) Peri, J. B. J. Phys. Chem. 1966, 70, 1483.
(46) Tanaka, M.; Ogasawara, S. J. Catal. 1970, 16, 157.
(47) Kyto¨kivi, A.; Lindblad, M.; Root, A. J. Chem. Soc., Faraday Trans.
1995, 91, 941.
(48) McInroy, A. R.; Lundie, D. T.; Winfield, J. M.; Dudman, C. C.; Jones,
P.; Parker, S. F.; Lennon, D. Catal. Today 2006, 114, 403.
9
11038 J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008

Atomic-Scale Insights on Chlorinated γ-Alumina Surfaces
A R T I C L E S
conditions. The pure substitution of Cl and OH surface groups
can only be obtained for a limited set of experimental conditions.
for the temperature used in reforming, the chlorine content is
limited to upper values of 2% for realistic HCl partial pressures.
These quantitative insights are crucial for a better control of
the chemical balance between water and chlorine surface
concentration on the catalytic support. Furthermore, depending
on the reaction conditions and on the initial thermodynamical
state of the surface, our study has determined the different
domains of the chlorination mechanisms (substitution process
versus dissociative addition process).
Our model approach provides new insights in the competition
between H2O and HCl on a γ-alumina surface and on the
influence of Cl groups on the acidity of the remaining hydroxyls.
The combination between a thermodynamic model and DFT
total energy calculations is hence a powerful tool to map the
various regimes of adsorption and substitution for these two
molecules, and therefore to obtain a predictive model of the
surface chemical properties in the temperature and pressure
conditions used in the experiment. We hope that these new
structural models of the chlorinated alumina surfaces can deserve
future theoretical and experimental investigations of the reactiv-
ity on Cl-γ-alumina-supported metal particles such as those used
in reforming catalysis.
Conclusions
This DFT study of chlorinated γ-alumina surfaces has
revealed new concepts for the chemical process involved in
chlorination of γ-alumina surfaces in various conditions of
temperature and partial pressures of water and HCl as found at
laboratory- and industrial-scale processes (catalytic isomerization
and reforming). At the laboratory scale and low chlorine
contents, the combination of DFT calculations of vibrational
frequencies of hydroxyl groups and IR spectroscopy has revealed
that the most basic µ1-OH groups present on the (100) and (110)
surfaces are preferentially exchanged. The weakening of the
hydrogen-bond network after chlorination has been found to
play a key role in the change of surface Brønsted acidity. As
simulated by pyridine adsorption, the formation of the pyri-
dinium ion on µ2-OH sites was favored after chlorine exchange,
due to the loss of the hydrogen bond of the µ2-OH species with
the neighboring basic µ1-OH groups once exchanged by Cl.
For higher chlorination contents as industrially used in
isomerization or reforming conditions, we have determined
numerous configurations of adsorbed and exchanged chlorine
on γ-alumina surfaces with various hydration degrees. The
competitive effects of water and HCl partial pressures were thus
put forward, thanks to a thermodynamic model including DFT
values of total energies and the chemical potential of the two
reactants. When considering a morphology model for γ-alumina
nanoparticles, this approach has led to the quantitative evaluation
of chlorine content as a function of partial pressures of HCl
and temperature. For a temperature close to isomerization, the
value of chlorine content can reach up to 6% chlorine, whereas
Acknowledgment. Part of this work was undertaken within the
Groupement de Recherche Europe´en “Dynamique Mole´culaire
Quantique Applique´e a` la Catalyse”, a joint project of IFP-CNRS-
TOTAL-Universita¨t Wien. We thank the Institut du De´veloppement
et des Ressources en Informatique Scientifique (IDRIS) for provid-
ing computational resources. M.D. and P.R. thank P. Euzen and
P.-Y. Le Goff from Axens for fruitful discussions on the industrial
reforming process.
JA8019593
9
J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008 11039