surface science

ELSEVIER Surface Science 343 (1995) 261-272

A theoretical study of the adsorption of oxalic acid on TiO 2

Adil Fahmi a,1, Christian Minot a,,, Patrick Fourr6 b9 Patrice Nortier b
a Laboratoire de Chimie Organique Thdorique, Universit~ P. et M. Curie, Bolte 53, URA 506 CNRS, 4 Place Jussieu, F-75252 Paris Cedex
05, France
b Service de Synth~se Mindrale, RhSne-Poulenc Recherches, 52 rue de la Hale Coq, F-93308 Aubervilliers, France

Received 7 February 1995; accepted for publication 31 July 1995

Abstract

This paper presents results of periodic Hartree-Fock calculations on the adsorption of oxalic acid on rutile and anatase
TiO 2 structures. One dimensional polymers are used as models for TiO 2 bare surfaces. Oxalic acid undergoes dissociative
adsorption leading to the oxalate bonded to two adjacent Ti atoms, which seems to be the most stable form of oxalic acid on
TiO 2 surfaces. The adsorption is stronger on the anatase polymer than on the ruti!e polymer and thus the crystal growth of
TiO 2, in the presence of oxalic acid, leads to the rutile structure.

Keywords: Ab initio quantum chemical methods and calculations; Carboxylic acid; Low index single crystal surfaces; Titanium oxide

1. Introduction growth is therefore oriented towards the faces that

The effect of foreign substances on the growth of
crystals is a well-known phenomenon [1-3]. There
are several possible mechanisms by which an addi-
tive can inhibit the crystal growth reaction. Two
main mechanisms are often used to describe the
inhibitory effect [4]: in the first one, the additive
forms stable complexes with one of the precipitating
ions and decreases solution supersaturation; in the
second one, the additive is adsorbed on the active
growth sites. During this later process, the additive
poisons the surface where the adsorption takes place.
The adsorption inhibits the growth on that face pre-
venting the deposition of the crystal molecules. The
* Corresponding author.
Present address: Eindhoven University of Technology, De-
partment of Inorganic Chemistry and Catalysis, P.O. Box 513,
5600 MB Eindhoven, The Netherlands.
interact weakly with the additive. In the case of the
competition between the formation of two polymor-
phic structures (for instance rutile and anatase struc-
tures for TiO2), the growth is oriented towards the
least active structure. Here rises the molecular recog-
nition phenomenon at inorganic surfaces [5] that has
the potential to provide control over crystal growth
processes. A detailed understanding of recognition
processes leads to the choice of the additive which
can direct the growth towards the desired crystalline
structure.
This work was stimulated by the neefi for more
effective chemicals to control the titanium dioxide
crystallization. Rutile and anatase are the main crys-
tal structures of TiO 2, rutile is the most stable one
[6]. During TiO 2 crystallization, some carboxylic
acids [7] are preferentially adsorbed on futile and the
growth is oriented toward~ a~atase. However, the
rutile formation is still observed in presence of ox-

0039-6028/95/$09.50 © 1995 Elsevier Science B.V. All rights reserved

SSDI 0039-6028(95 )00813-6
262 A. Fahmi et al. / Surface Science 343 (1995) 261-272
alic acid. The same trend was found in the TiO 2
sol-gel process [8]. It is therefore suggested that
oxalic acid is preferentially adsorbed on anatase,
blocking the anatase growth.
In this paper we present a theoretical study of the
adsorption of the oxalic acid on bare rutile and
anatase surfaces. The structure of the oxalic acid on
TiO., is analyzed. The crystal growth orientation will
in principle be found by the comparison of the
adsorption energies of the oxalic acid on the two
structures. We first present the oxalic acid molecule
(HOOC-COOH), the hydrogen oxalate (HOOC-
COO-) and the oxalate (C.,O42-) ions (Sections
3-5). We discuss the gas phase acid-base properties
of the oxalic acid (Section 6) and we present the
models for the bare TiO,. surfaces (Section 7). Next,
we report the results of the molecular adsorption
(Section 8). The various dissociative adsorptions are
analyzed in Sections 9 and 10. The last section is
devoted to oxalic acid decomposition in the gas
phase and on the TiO2 surface.
2. Method of calculation
The calculations are carried out at the SCF level
with a restricted Hartree-Fock hamiltonian using the
core pseudopotential from Durand-Barthelat [9]. For
the gas phase properties of oxalic acid, we use the
MONSTERGAUSS [10] program to optimize the
geometry of the various free anions which can be
generated by the oxalic acid. Next we use the CRY~,-
TAL [11] program to compute the electronic prope~'-
ties o: the optimized molecules and also for the
adsorption studies. The basis sets are the same as in
Table 1
Optimized geometrical parameters of the oxalic acid (in ,~ and deg), the hydrogen oxalate and the oxalate ions
Bond length Oxalic acid
or angle ("A"conformation)
C=C 1,504 (1,536)
C=O 1,206 (1,202)
C-O 1,33 (1,291)
O=C-C 124,7 (121,7)
O-C-C ! 11,5 (112)
The experimental values [21] arc indicated in parentheses. The three values for the CO bond in the hydrogen-oxalate correspond to single
bonds (the first one is that of the hydroxyl group, the last one is cis with respect to that group). The same convention is used for the OCC
angles,
our previous studies [6,12,13], a PS-31G [14] basis
set for the oxygen and the carbon atoms and the
standard 4-31G basis set for the hydrogen atom. For
the titanium atom the basis set consists of five d
primitives contracted to a (4/1) basis set and a
single 4sp shell with an exponent of 0.484; the 4sp
shell is almost vacant for the titanium ions. These
basis sets were optimized for the TiO 2 bulk structure
[6]. One might think that they are of medium quality.
However, this represents for a periodic calculation a
better level than it would be for a molecular calcula-
tion. Indeed, very diffuse orbitals are not required as
for a molecular calculation, since they are responsi-
ble for the basis set linear dependence [15].
3. The oxalic acid molecule
The oxalic acid has a planar geometry. According
to hydrogen atoms orientation, two conformations
are very close in energy (see Scheme 1). For the first
one, the hydrogen bonding takes place within each
carboxyl group, while for the second one, hydrogen
atoms connect different carboxyl groups. The first
one (conformation A) is more stable by 1.7 kca|/mol
with the PS-31G basis sets.
The conformation B, with hydrogen bonding (O-
O 2.7 and H . . . O ~ 2.29 ,g,) is believed to be the
most stable according to electron diffraction [16] and
spectroscopic analysis [16,17]. Despite of numerous
theoretical studies, the most stable conformation is
still controversial [18-20]. The use of ab initio all
electron calculations shows a very small energy dif-
ference between the two conformations. At the
MP2/4-31G level [19] the conformation A is more
Hydrogen oxalate Oxalate Oxalate
(planar) (orthogonai) (planar)
1,556 1,530 1.596
1,212 - -
1.361-1,239-1,268 1,274 1.269
I i !.8 117.5 117.7
127,7-11 l,l-118 117.5 117.7
 A. Fahmi et al./ Surface Science 343 (1995) 261-272
0 O--H
.-d -
\H
conformation A conformation B
Scheme 1.
stable by 0.8 kcal/mol and at the M P 2 / 6 - 3 1 G ,
level [20] the conformation B is more stable by 2.5
kcal/mol.
The optimized parameters for the conformation A
with the PS-31G basis sets are shown in Table 1.
The overall agreement with experimental data [21] is
satisfactory except the position of the hydrogen atom
that gives errors of 0.1 A for the O - H bond length
and 10° for the C - O - H angle. Errors of the same
magnitude were found by other authors [19].
4. The hydrogen oxalate ion, HOOC-COO-
The hydrogen oxalate ion has a planar geometry.
The hydrogen atom is oriented in the way that allows
ydrogen bonding (O-O = 2.52 and H . . . O - - 2 . 0 4
) between the two carboxyl groups (see Scheme 2).
This bonding is strong because of the negative charge
on the oxygen atom.
The energy difference between the planar and the
orthogonal conformations is 9.9 kcal/mol. This later
value can be considered as an estimation of the
hydrogen bonding energy. The O . . . H overlap popu-
/--\
io oxxo___,,,o
%. o attractor substituent, the oxalic acid proton affinity is
planar conformation orthogonal conformation
Scheme 2.
263
lation is 0.048. The optimized parameters for the
planar conformation are shown in Table 1.
5. The oxalate ion, C20 z-
Using the 6-31G, and 6-31 + G basis sets, De-
war et al. [22] found that the orthogonal D2d geome-
try of the oxalate is more stable than the planar D2h
geometry. By a metal-cation complex the geometry
can be modified towards the planar oxalate. The
energy difference between the two conformations are
4.9 (6-31G,) and 6.3 (6-31 + G) kcal/mol. With
the PS-31G basis sets, we found the same trend, a
stabilization of the D20 geometry by 2.9 kcal/mol.
The optimized parameters are shown in Table 1. The
C - C bond length is longer for the planar geometry
because of the oxygen-oxygen repulsion.
6. Gas phase properties of oxalic acid
6.1. Gas phase proton affinities
1here are two basic sites in the oxalic acid: a
carboxylic and a hydroxylic oxygen atom. When a
single proton interacts with the oxalic acid molecule,
the first proton affinity is involved:
HOOC-COOH + H + ..o HOOC-C 0 (OH) 2,
A Hc,,i,: = - 180.5 kcal/mol
(the proton is attached to a carbonyl group),
HOOC-COOH + H + --* HOOC-CO(OH 2 ) +'
AHc.~lc = - 168.2 kcal/mol
(the proton is attached to a hydroxyl group).
The proton is located on a carbonyl group, the
most basic site. Since the second COOH group is an
small with respect to that of the formic acid (A H -
- 184.1 kcal/mol; experimental value: - 182.8
kcal/moi [23]) or that of water ( A H = - 1 8 2 . 5
kcal/mol; experimental value: - 1 7 3 kcal/mol
[23,24]).
264 A. Fahmi et ai./ Surface Science 343 (1995) 261-272
When two protons interact with the oxalic acid,
the best protonation mode also involves the carbonyl
groups:
HOOC-COOH + 2H + --* (HO)2C+-C + (OH)2,
AHcatc -- - 244.1 kcal/mol
(protonation of two carbonyl groups),
HOOC-COOH + 2H + --* + ( H 2 0 ) O C - C + (OH)e,
A Helle = - 231.4 kcai/mol
(protonation of one hydroxyl and one carbonyl
groups),
HOOC-COOH + 2H + --* + (H20)OC-CO(OH~,) +,
AHcalc = - 218.3 kcal/mol
(protonation of two hydroxyl groups).
From these results we expect several modes for
the molecular adsorption of oxalic acid on TiO2. The
mode that should be the best involves two carbonyl
groups.
6.2. Acidic cleavages of oxalic acid
Since the COOH group is an attractor, the hydro-
gen oxalate ion, HCOO-COO =, is more stable than
the formate, HCOO =, or hydroxide, HO °, ions,
Therefore, the first acidic cleavage of oxalic acid is
J 11 11
0 C O 0
ruble polymer
less endothermic than that of formic acid or that of
water:
H20 -* H O - + H +, AHcalc- 423.8 kcal/mol
(experimental value: 390.8 kcal/mol [25,26]),
HCOOH -* H C O O - + H +,
AHcalc - 356.7 kcal/mol
(experimental value: 345.2 kcal/mol [26]),
HCOO-OOCH -* H O O C - C O O - + H +,
AHc~alc- 328.8 kcal/mol.
The second acidic cleavage of oxalic acid is very
endothermic:
H C O O - O C C - ---, - O O C - C O O - + H +,
AHcalc - 459.5 kcal/mol.
7. Models for the TiO z surfaces
We model the TiO 2 surfaces by one dimensional
polymers [6,12]: a linear polymer for rutile and a
zigzag polymer for anatase (see Scheme 3). These
models have been checked by a study on the clean
surface [27]; it is concluded that atoms from the
TiO 2 polymers do not differ by much from the bulk
and are still bound objects relative to the free atoms.
We have already used this model for the studies of
adsorption on the TiO 2 surfaces [12,13,28].
11
G 0
j o--T --o
~ ~ o
anatasc polymer
_ . presentedas undistortedsys~m
~ormc optimizedsuucture, Ti-O-Tiis 152°.
Scheme 3.
 A. Fahmi et al./ SurfaceScience 343 (1995) 261-272
The rutile polymer is made of rectangular pieces
with a long edge (3.0065 A, the c parameter of the
bulk) in the direction of the chain and a short one
(2.4875 A) in theoperpendicular direction. The Ti-O
distances, 1.951 A, are the equatorial distances from
the bulk structure. For the anatase polymer, the
Ti-Ti translation vector is the a parameter of the
bulk, 3.7629 A. This latter system involves both
equatorial Ti-O bonds, 1.939 A, along the axis of
the polymer and apical ones, 1.995 A, in the perpen-
dicular direction. These parameters are taken from
the optimized bulk structures [6]. The coordination of
the Ti atoms is four and is the same for the two
polymers. For the oxygen atoms, the coordination is
two in the linear polymer, while, in the zigzag
polymer, half of them is singly-coordinated and the
other half is triply-coordinated. On real surfaces, the
coordination of the Ti atoms is usually five (it is four
on the (001) surface of ruffle) and that of the oxygen
atoms is most often two or three. Previous studies
[12,13] have shown that the reduction of the coordi-
nation of the atoms in the polymer models contribute
to overestimate the adsorption energies. The energy
shift is comparable for various adsorption modes and
thus does not affect the relative stabilities.
These polymers are obviously far from being a
complete representation of crystallographic faces al-
though they are significantly better than a cluster
model. However, they contain titanium and oxygen
atoms with the oxidation numbers + IV and -11,
respectively. The linear polymer can be recognized
as the main feature of the ruffle (110) and (100)
faces; it contains the linearity of the TiO 2 sequence
that is specific of rutile. The zigzag polymer can be
found on the anatase (100) and (010) faces. We
consider that the study of the polymer models should
give information about the activities of the TiO2
surfaces.
The ruffle polymer is more stable than the anatase
polymer; the calculated total bonding energies are 83
and 68 kcal/mol, respectively [12]. The water ad-
sorption study [12] shows that the metal cations from
the ruffle polymer are the best Lewis acidic sites and
the singly-coordinated anions from the anatase poly-
mer are the best basic sites.
A basic adsorbing species would approach the
titanium atom perpendicularly to the polymer plane
and an acidic adsorbing species would app~::oach the
265
oxygen atom in the polymer plane. We will consider
the adsorption of one molecule of oxalic acid on a
cell of two TiO 2 units of L~e polymer. Adsorption
energies have been calculated according to the ex-
pression:
Ead s ffi ETiO2 4" EH2C20, -- E(H2C202/TiO2),
where E(H~C204/TiOe) is the total energy of the ad-
sorbate/substrate system, ETi02 is the total energy
of the substrate and Eu2c204 is the total energy of
the free oxalic acid molecule in its equilibrium ge-
ometry. A positive Eads value corresponds to a stable
adsorbate/substrate system.
For the geometry of adsorbed oxalic acid, the
following methodology has been used. We first per-
formed a molecular calculation of the adsorbate
species (without the substrate) using the MONSTER-
GAUSS program. The adsorbate species is defined
as the oxalic acid molecule plus one or two protons
simulating the Ti metal cations. Positions of oxygen
atoms by which the oxalic acid is attached to the
surface and those of protons are fixed; the positions
of all the other atoms are optimized. Finally, this
geometry is transferred to the CRYSTAL calcula-
tions. The Ti-O distances between the oxalic acid
and the surface are taken from water adsorption [12]
( T i - O - 2.12 ~. for the molecular adsorption and
Ti-O = 1.81 A for the dissociative adsorption). The
study of the formic acid adsorption shows that these
parameters, optimized for H20, are transferable to a
carboxylic acid [29].
8. Molecular adsorption on futile
Titanium dioxide is an amphoteric compound,
endowed with acidic Ti(+IV) and basic O ( - I I )
sites. Surface chemistry may thus involve Lewis and
Brcnsted acid-base reactions. Bare TiO 2 surfaces
dissociatively adsorb water to form surface hydroxyl
groups [12]. These surfaces correspond to the real
catalytic species more often than bare surfaces. The
first step in the interaction of a carboxylic acid
(RCOOH) with TiO 2 surfaces is the formation of the
carboxylate (RCOOads) [30-34]. The proton is trans-
ferred to an oxygen atom of the surface. The second
step is the decomposition of the carboxylate. The
structure of the carboxylate and the thermodynamic
266 A. Fahmi et al./ Surface Science 343 (1995) 261-272

H~~ .........',,,,~:)...,,,I,
H Lewis acid-base reaction, the Lewis acid is the
metal cation and the Lewis base is an oxygen atom
from the oxalic acid. There is a correlation between
the molecular adsorption and the gas phase proton
affinity of the carboxylic acid. The oxalic acid can
interact with the surface by one or two carboxyl
groups. These two adsorption modes are investi-
gated.

(a) 8.1. Adsorption by one fimction

This function can be either a carbonyl group (Fig.

.... "m,llll~..,,,lill H
H CO(OH:) + species (a model for the adsorption
(b)
Fig, 1, Molecular adsorption of the oxalic acid through a single
function: (a) the ¢arbonyl group, and (b) the hydroxyl group,
of the reaction leading to its formation are not well
determined.
in this section we d,~al with the molecular adsorp-
tion: the oxalic acid is not dissociated on the surface.
The molecular adsorption can be described as a
la) or a hydroxyl group (Fig. lb). The geometry of
the adsorbed oxalic acid corresponds to that of
H O O C - C(OH) + species (a model for the adsorp-
tion through the carbonyl) or to that of H O O C -
through the hydroxyl), where the proton simulates a
metal cation of the polymer. The best conformations
for the adsorbed oxalic acid are the orthogonal ones
where the non-adsorbed carboxyi group is parallel to
the polymer plane. In these conformations, the oxy-
gen-oxygen repulsion is the weakest.
As expected from the gas phase proton affinities
of oxalic acid, the adsorption by the carbonyl is more
favourable than that by the hydroxyl. The adsorption
energies are respectively 23.0 and 7.0 kcai/mol
(exothermic adsorptions). The molecular adsorption
of the oxalic acid is weak in comparison with that of
the formic acid (31.2 and 26.2 kcal/m01 for the
adsorption by the carbonyl and by the hydroxyl
respectively). The deconjugation in the orthogonal

H H

(a) (b)
Fig, 2, Molecular adsorptionof the oxalic acid through two functions:(a) two carbonyl groups, and (b) the carbonyl group and the hydroxyl
group,
 A. Fahmi et al./ Surface Science 343 t1995) 261-272 267

conformation of the oxalic acid increases the differ-
ence expected from the proton affinities.
8.2. Adsorption by two Jimctions
From the gas phase proton affinities (Section 6.1),
we have seen that the protonation of two hydroxyl
groups is less favourable than that involving two
carbonyl groups. Thus, the corresponding adsorption
should be less favourable than that involving car-
bonyl groups. Because of the proton-proton repul-
sion between two nearest oxalic acid molecules, the
calculation of the adsorption by two hydroxyl groups
requires the use of a low coverage (one oxalic acid
per 4 TiO 2 units). The geometry of the adsorbed
oxalic acid corresponds to that optimized for
(HO)2C+-C+(OH)2 species with the constraint O -
O = 3.0065 ,g, (the translation parameter of the poly-
mer). The adsorption energy is weak, 25 kcal/mol.
At higher coverage (one oxalic acid per 2 TiO 2
units), the two adsorbed functions are either two
carbonyl groups (Fig. 2a) or one carbonyl and one
hydroxyl groups (Fig. 2b). Adsorption energies are
similar, 48.1 k c a l / m o l for the first adsorption mode
and 48.0 k c a l / m o l for the second one.
The adsorption through two functions for both
modes is more favourable than the adsorption through
one function.
9. Dissociative adsorption on futile
The dissociative adsorption can be described as a
Br0nsted acid-base reaction where the proton is

~
H
'.,,*,i**lllll~ .... Illll H

(a) (b)

(c)
Fig. 3. Various adsorption modes of the hydrogen oxalate: (a) The hydrogen oxalate is adsorbed through a single oxygen atom (CO), the
conformation is orthogonal. (b) The hydrogen oxalate is adsorbed through the two oxygen atoms of the carboxylategroup, the conformation
is planar. (c) The hydrogen oxalate is adsorbed through two oxygen atoms, one from the carboxylate group and the other one from qle
carboxyl group, the conformation is planar.
 268 A. Fahmi et aL /Surface Science 343 (1995) 261-272
transferred tO an oxygen atom of the oxide (a
Br¢nsted base) and the carboxylate is adsorbed on a
metal cation. On the contrary to molecular adsorp-
tion, dissociative adsorption involves simultaneously
the acidic and basic properties of both the oxalic acid
and the T i O 2 surface.
9.1. Adsorption of the hydrogen oxalate ion,
HOOC-CO0 -
The adsorption of the hydrogen oxalate ion can be
seen as the result of a dissociative adsorption of the
oxalic acid molecule on the TiO_, surface (HOOC-
C O O H a d s ~ H O O C - C O 0 ~ + H ads)"
+ The hydrogen
oxalate is adsorbed on a Ti ion and the proton goes
to a bridged oxygen atom of the polymer. When the
hydrogen oxalate is adsorbed through a single oxy=
gen atom from the carboxylate group (see Fig. 3a,
the carboxyl group is parallel to the polymer plane),
the adsorption energy is very small, 18.7 kcal/mol.
When the other oxygen atom of the carboxylate is
also adsorbed (see Fig. 3b), the adsorption energy is
83 kcal/mol. This adsorption mode is general for
monocarboxylic acids such as the formic acid [29,35].
The heat of adsorption is slightly inferior to that of
the formic acid, 85 kcal/mol, because of the attrac-
tive effect of the COOH group. Since the oxalic acid
is a diacid, the adsorption may involve the carbonyl
of the carboxyl group instead of that of the carboxyl-
ate (see Fig. 3c). Then, the adsorption energy be-
comes slightly larger, 85.7 kcal/mol.
in this later adsorption mode, the geometry of the
adsorbate is planar with a strong internal hydrogen
bonding. The oxygen-hydrogen distance is 2.1 A
with a Mulliken overlap population, OP--0.023.
From this OP, we expect a high value for the hydro-
gen bonding. For instance, in the water dimmer
(H,O), [12] the hydrogen bond has an OP of 0.017
and a bonding energy of 7.3 kcal/mol.
Clearly, the best adsorption mode for the hydro-
gen oxalate involves two oxygen atoms. These oxy-
gen atoms can be either from the carboxylate group
or from the carboxylate and the carboxyl groups.
9.2. Adsorption of the oxalaw ion, C20 ~ -
The formation of the oxalate on the TiO 2 surface
can be seen as the result of two successive acidic
H
Fig. 4. Adsorption of tile oxalate in a twisted conformation, the
twist angle is 15° .
cleavages. The oxalate is complexed in a twisted
geometry by two metal cations; the twisted angle is
15°. The protons are transferred to oxygen atoms of
the polymer (Fig. 4). The complexation of the ox-
alate ion by two metal cations was found on hematite
[36,37], no data are available for TiO 2.
There are different modes for the protonation of
the polymer as already found for the dissociative
adsorption of water [12]. In the best proton distribu-
tion, protons alternate on each side of the polymer.
The adsorption energy of the oxalate is 101.7
kcal/mol; it is larger than that of the hydrogen
oxalate.
We have also calculated the adsorption of the
oxalate ion through the two oxygen atoms of the
same CO, group. This adsorption is not favourable,
the adsorption energy is only 44.0 kcal/mol. This
system should be unstable and decompose into two
CO, molecules). The CC Mulliken overlap popula-
tion is very negative, O P - -0.15; this value sug-
gests a repulsion between the two CO, groups. The
adsorption of the oxalate by one CO, group,
(TiO)2C-CO 2, shows similarities with the (HO).,C-
CO., system obtained by a proton transfer in the
oxalic acid molecule; such system is unstable and
undergoes decarboxylation as found from previous
calculations [ 19].
From these calculations, we conclude that the
oxalate is more stable than the hydrogen oxalate on
TiO, surfaces.
9.3. Adsorption formally resulting from basic
cleavages, HOOC- +CO and OC +- +CO
The heterolytic dissociation of an acid can be
either acidic (O-H cleavage) or basic (C-OH cleav-
 A. Fahmi et a l . / Surface Science 343 (1995) 261-272
P gas phase, such cleavage is less endothermic than the
© The adsorption energy is 44.4 kcal/mol. When
I ....0
H'2~44A
a)
/H
,o--I- I-
O ~ Ti 0]/// Ti
f HOOC-CO + ~ +O C - C O + + HO -,
b)
Fig. 5. (a) Dissociative adsorption of the oxalic acid associated
with a basic cleavage(HOOC-COOH ~ HOOC-CO~., + HO~).
HOOC=+ CO is adsorbed in a bridging position. (b) Dissociative
adsorption of ~heo×alicacid associatedwith two cleavages(acidic
and basic: HOOC=COOH~ H,,d,, + + - 02C-CO,,,*t,,+ HO,~,).
Since the C=C bond is elongated, this adsorption mode leads to
the decompositionof the oxalic acid on the surface.
age). The basic cleavage of the oxalic acid on a TiO 2
surface leads to a doubly-coordinated carboxylate
and a singly-coordinated hydroxyl group (Fig. 5a).
There is an internal hydrogen bonding in the carbox-
ylate; the oxygen-hydrogen distance is 2.44 A and
the overlap population is 0.011.
Avoiding to speculate on the validity of a mecha-
nism involving a basic cleavage for an acidic species,
we can conceive this adsorption mode simply as an
alternative distribution of fragments that may be
generated from an acidic cleavage followed by an
exchange between the OR and OH groups on the
surface. Let us note however in favour of a basic
cleavage mechanism, that, for the oxalic acid in the
269
acidic cleavage by 35 kcal/mol:
HOOC-COOH --->HOOC-CO ÷+ H O - ,
A H = 294.13 kcal/mol.
the adsorption is by one function, the basic cleavage
(Fig. 5a) is more favourable than the acidic cleavage
(Fig. 3a) (see Section 9.1). However, while the
singly-coordinated carboxylate that results from an
acidic cleavage leads to the adsorption of a second
function (Figs. 3b and 3c) and to an increase of the
adsorption energy by a factor of five, the absence of
similar possibility for the basic cleavage makes this
process unfavourable.
The adsorption of the dication +OC-CO + that
results from two successive basic cleavages is un-
favourable (Ead s = 34.8 kcal/mol). The correspond-
ing gas phase reaction is highly endothermic:
AH = 510.9 kcai/mol.
Finally, we have considered the adsorption of the
- O 2 C - C O + species that results from the dehydra-
tion of the oxalic acid (acidic and basic cleavages).
This would be a possibility to have the adsorption of
two functions, one of them resulting from a basic
cleavage. The corresponding geometry IS shown in
Fig. 5b. The corresponding gas phase reaction is less
endothermic than two acidic or basic cleavages.
HOOC-COOH ~ -O2C-CO+ + HO- + H +,
A H = 418.2 kcal/mol.
The calculated adsorption energy is 71.6
kcal/mol; this value is inferior to that obtained for
the oxalate. As the - O 2 C - C O + species is a charged
representation of the CO2-CO complex, this adsorp-
tion mode raises the question of the decomposition
of the oxalic acid on the surface. We shall discuss
this matter in Section 11.
10. Adsorption of the oxalate on anatase
Since the oxalate is more stable than the oxalic
acid or the hydrogen oxalate on the futile polymer,
270 A. Fahmi et al. / Surface Science 343 (1995) 261-272
C~o H~°
!/
H
Fig. 6. Dissociative adsorption of the oxalic acid on the anatase
polymer. The geometry of the oxalate is planar and the protons are
in the polymer phme.
we limit our discussion in this section to the oxalate
adsorption. Dissociative adsorption of the oxalic acid
on the zigzag chain of anatase involves the protona-
tion of singly-coordinated oxygen atoms (which are
more basic than doubly-coordinated oxygen atoms of
the rutile polymer [12]). The protons are in the
polymer plane (Fig. 6). The geometry of the oxalate
is planar. The oxygen-oxygen distance when the
oxygen atoms lie above the titanium atoms is long
(3.10 ~) as compared with that of the rutile case
(3.01/~, the c parameter from the bulk) and closer to
that of the oxalate in the gas phase (2.14 J.).
The adsorption energy, 128.2 kcal/mol, is larger
than that on the rutile polymer by 26.5 kcal/mol.
The anatase structure appears to be more reactive
because of the basicity of the singly-coordinated
oxygen atom [ 12]. The distortion of the oxalic acid is
not the driving force. Indeed, the adequation of the
oxalic acid to the surface is not better on the anatase
surface than on the rutile one. On the contrary, on
the futile polymer the oxalate is twisted and the O-O
distance becomes close to that of the oxalate in the
gas phase. Such an adaptation is not possible on the
anatase polymer since the molecule remains planar.
The large O - O distance in the adsorbate is however
not that destabilizing, since the calculated adsorption
energy is large.
The difference in reactivity between the two
phases allows us to explain the crystallization pro-
cess of TiO, in presence of oxalic acid. Since the
adsorption is stronger on anatase, the oxalic acid is
adsorbed on the first crystallites of anatase and by
that the anatase formation is inhibited. The adsorp-
tion on rutile is weaker than on anatase, whence the
crystallization leads to rutile structure.
11. Decomposition of oxalic acid
11.1. Gas phase decomposition
In the gas pi~as~, the thermal decomposition of
oxalic acid yields CO_, + HCOOH [38]. The study of
UV photolysis [39] suggests that two processes oc-
cur, yielding CO, + HCOOH and CO 2 + CO + H 2O.
The calculated enthalpies for these reactions, in
agreement with previous calculations [19], show
exothermic reactions:
HOOC-COOH --* CO 2 + HCOOH,
AHcalc = - 12.1 kcal/mol
(experimental value: - 9 . 6 kcal/moi [40]; AHc.qc =
- 1 6 . 6 kcal/mol when intermolecular interactions
between products are allowed),
HOOC-COOH ~ CO., + CO + H 2O,
A Hc,,ic - - 7.2 kcal / mol
(experimental value: - 3.3 kcal/mol [40]; AH~,,i,, --
16.0 kcal/mol when intermolecular interactions
between products are allowed).
1i.2. Decomposition on the Ti02 surfaces
In order to estimate the enthalpies of these two
reactions on TiO., surfaces, we first calculate adsorp-
tion energies of all the products, H 20, HCOOH, CO,
and CO., independently. For such calculations, we
used the rutile polymer and consider the best adsorp-
tion mode for each adsorbate: dissociative adsorption
for water ( H , O ~ H,ds + OHad~) and formic acid
(HCOOH ~ H,id,~ + HCOO, d,~) at saturation ( 0 - 1).
For water, the hydroxyl group is adsorbed above a Ti
atom and the proton is transferred to an oxygen atom
of the polymer [ 12]. For formic acid, the carboxylate
is adsorbed between two Ti atoms and the proton is
transferred to an oxygen atom of the polymer [29].
For CO [13] and CO 2 [28] the best adsorption modes
correspond to perpendicularly adsorbed molecules
above Ti atoms. We have considered both the 0 - 1
 A. Fahmi et aL / Surface Science 343 (1995) 261-272 271

Table 2 12. Conclusion

Adsorption energies in kcai/mol for various molecules on the
rutile polymer at 0 = 1 and 0 = 1 / 2 coverages
Adsorbed system 0= 1 0 = 1/2 In this paper, we have presented a periodic
H 2Odiss 63 a _
Hartree-Fock ab initio calculation of the adsorption
HCOOHdiss 85 b _ of the oxalic acid on TiO 2 surfaces. We have used
CO 15.2 c 19.3 c polymers as models for rutile and anatase surfaces.
CO 2 10.4 d 19.0 d Various adsorption modes for oxalic acid molecule,
CO., + HCOOH diss 95.4 104.0 hydrogen oxalate and oxalate ions were investigated.
CO 2 + CO + H20 diss 88.6 101.3 The best adsorption modes always involve two oxy-
gen atoms of the adsorbate. The dissociative adsorp-
H 2C204 diss 101.7 -
tion is more favourable than the molecular adsorp-
The subscript "diss" refers to dissociatively adsorbed species. tion; our results clearly show that the oxalate com-
a Ref. [12]. plexed by two metal cations is the most stable form
b Ref. [29].
c Ref. [13].
of the oxalic acid on TiO 2 surfaces. This is a study
d Ref. [281. of the thermodynamically stable adsorption. We rec-
ognize that kinetics may limit the accessibility of the
stable configurations. The decomposition reactions
of the oxalic acid that are exothermic in the gas
and 0 = 1 / 2 coverages. The various heat of adsorp- phase are not favoured by the adsorption on the
tions are given in Table 2. The sum of these adsorp- oxide surface. The comparison of the adsorption
tion energies give numbers that are very close to the energy of the oxalate on rutile and anatase polymers
value for the adsorption energy of the oxalate, 101.7 explains crystallization experiments of TiO 2 in pres-
kcal/mol. From these values and the gas pha~se ence of oxalic acid. The adsorption energy is larger
values, we derive the following values for the disso- on anatase. Therefore, the oxalic acid is adsorbed on
ciation on the surface: the growing anatase preventing the formation of the
bulk structure. The adsorption on ruffle is weaker
HOOC-COOHdiss - , CO 2 + HCOOHoiss, and the crystal growth leads to the ruffle structure.
AH~a~ = - 5 . 8 kcal/mol ( - 14.4 at 0 = 1/2)
(to be compared with - 16.6 kcal/mol),
HOOC-COOHdi~ -~ CO 2 + CO + H 2Odi.~,,, References
AHcalc = +5.9 kcal/mol ( +6.8 at 0 = 1 / 2 )
(to be compared with - 16 kcal/mol). [1] P. Hartman, in: Adsorption et Croissance Cristallin¢. Colloq.
Intern., CNRS No. 152 (CNRS, Paris, 1965) p. 477.
The surface decomposition does not seem to be
[2] A. Julg and B. Deprick, J. Cryst. Growth 62 (1983) 587.
favoured by the adsorption. According to these val- [3] B. Deprick-Cote, J. Langlet, J. Caillet, J. Brcges, E. Kassab
ues, only the decomposition leading to the formate and R. Constanliel, Theor. Chim. Acta 82 (1992) 435.
and the carbon dioxide would remain possible. [4] Z. Amjad, Langmuir 7 (1991) 600.
Finally, we performed a calculation on the coad- [5] RJ. Davey, S.N. Black, L.A. Bromely, D. Collier, B. Dobbs
and J.E. Rout, Nature 353 (1991) 549.
sorption of the three molecules CO2ads + COads + (H [6] A. Fahmi, C. Minor, B. Silvi and M. Causa, Phys. Rev. B 47
+ HO),,ds on the polymer; this requires a unit cell (1993) 11717.
containing three titanium atoms and thus, the adsorp- [7] P. Fourr6, Rapport Interne Rh6ne-Poulenc CRA 1896 (1988).
tion energy has to be compared with an adsorption [8] T. Lopez, E. Sanchez, P. Bosch, Y. Meas and R. Gomez,
energy for the oxalate larger than 101.7 kcal/mol Mater. Chem. Phys. 32 (1992) 141.
[9] P. Durand and J.C. Barthelat, Theor. Chim. Acla 38 (1975)
since the adsorption energy increases when the cov-
283.
erage decreases. The calculated value is only 31.8 [10] R.A. Poirier and M. Peterson, Monslergauss Department of
kcal/mol showing that coadsorption effects are not Chemistry Memorial University of New foundland, St. John's,
favourable at this coverage (saturation). Newfoundland, Canada.
272 A. Fahmi et aL / Surface Science 343 (1995) 261-272
[Itl (a) R. Dovesi, C. Pisani, C. Roetti, M. Causa and V.R.
Saunders, CRYSTAL 88, QCPE program No 577, Blooming-
ton, Indiana (1989); (b) R. Dovesi, V.R. Saunders and C.
Roetfi, CRYSTAL 92 User Documentation, Torino and
Daresbury (1992); (c) C. Pisani, R. Dovesi and C. Roetti, in:
Hartree-Fe~ k Ab initio Treatment of Crystalline Systems,
Lecture Netes in Chemistry, Vol. 48 (Springer, Heidelberg,
1988)o
[12] A. Fahmi and C. Minot, Surf. Sci. 304 (1994) 343.
[131 A. Fahmi and C. Minot, J. Organometai. Chem. 478 (1994)
67.
[14] Y. Bouteiller, C. Mijoule, M. Nizam, J.C. Barthelat, J.P.
Daudey, M. Pelissier and B. Silvi, Mol. Phys. 65 (1988) 295.
[151 G. Aissing and H.J. Monkhost, int. J. Quant. Chem. 43
(1992) 733.
[161 Z. Nahlovska, B. Nahlovsky and T.G. Strand, Acta Chem.
Scand. 24 (1970) 2617.
[171 R.L. Redington and T.E. Redington, J. Mol. Struct. 48
(1978) 165.
[181 D. Ajo, G. Condorclli, !. Fragala and G. Granozzi, J. Mol.
Struct. 37 (1977) 160.
[19] T. Kakumoto, K. Saito and A. Imamura, J. Phys. Chem. 91
(1987) 2366.
[20] J. Nieminen, M. Rasanen and J. Murto, J. Phys. Chem. 96
(1992) 5303.
[211 L. Leiserowitz and F. Nader, Angew. Chem. Int. Ed. I1
(1972) 514.
[221 M.J.S. Dewar and Y.-J. Zheng, J. Mol. Struct. 209 (1990)
157.
[23l D.H. Aue and M.T. Bowers, in: Gas Phase Ion Chemistry,
Vol. 2, Bd. M.T. Bowers (Academic Press, New York, 1979)
p.l.
[241 J.L. Ozment and A.M. Schmiedekamp, Int. J. Quant. Chem.
43 (1992) 783.
[251 J. Phys. Chem. Ref. Data, Vol. 6, Sup. No 1 (1977).
[26] J.E. Bartmess and R.T. Mclver, Jr., in: Gas Phase Ion
Chemistry, Vol. 2, Ed. M.T. Bowers (Academic Press, New
York, 1979)p. 87.
[27] P. Reinhardt and B.A. Hess, Phys. Rev. B 50 (1994) 12015.
[281 C. Minor, A. Fahmi and J. Ahdjoudj, The Synergy between
Dynamics and Reactivity at Clusters and Surfaces, Ed. L.
Farrugia, NATO-AS! C465 (1995) p. 257.
[29] A. Fahmi and C. Minot, in preparation.
[3o1 H. Onishi, T. Aruga, C. Egawa and Y. lwasawa, Surf. Sci.
193 (1988) 33.
[31] M. Primet, P. Pichat and M.V. Mathieu, J. Phys. Chem. 75
(1971) 1221.
[32] K.S. Kim and M.A. Barteau, Langmuir 4 (1988) 945.
[33] K.S. Kim and M.A. Barteau, Langmuir 6 (1990) 1485.
[34] M.A. Enriquez and J.P. Fraissard, J. Catal. 79 (1983) 462.
[35] H.Onishi, T. Aruga and Y. lwasawa, J. Catal. 146 (1994)
557.
[36] N. Kally and B. Matijevic, Langmuir 1 (1985) 195.
[37l N. Kalk, and E. Matijevic, Langmuir 1 (1985) 201.
[38] G. Lapidus, D. Barton and P.E. Yankwich, J. Phys. Chem. 89
(1964) 1863,
[39] S. Yamamoto and R.A. Back, J. Phys. Chem. 89 (1985) 622.
[401 S.W. Benson, in: Thermochemical Kinetics,2nd ed. (Wiley,
New York, 1976).