Ketonic decarboxylation reaction mechanism: A combined experimental and DFT study

Article abstract of DOI:10.1002/cssc.201200419

The ketonic decarboxylation of carboxylic acids has been carried out experimentally and studied theoretically by DFT calculations. In the experiments, monoclinic zirconia was identified as a good catalyst, giving high activity and high selectivity when compared with other potential catalysts, such as silica, alumina, or ceria. It was also shown that it could be used for a wide range of substrates, namely, for carboxylic acids with two to eighteen carbon atoms. The reaction mechanism for the ketonic decarboxylation of acetic acid over monoclinic zirconia was investigated by using a periodic DFT slab model. A reaction pathway with the formation of a β-keto acid intermediate was considered, as well as a concerted mechanism, involving simultaneous carbon-carbon bond formation and carbon dioxide elimination. DFT results showed that the mechanism with the β-keto acid was the kinetically favored one and this was further supported by an experiment employing a mixture of isomeric (linear and branched) pentanoic acids. This way or that? Monoclinic zirconia has great potential as a catalyst for ketonic decarboxylation of carboxylic acids (see picture). A combined experimental and DFT study shows a route involving a β-keto acid intermediate as the kinetically preferred reaction pathway. Copyright

Full text of DOI:10.1002/cssc.201200419

CHEMSUSCHEM
FULL PAPERS
DOI: 10.1002/cssc.201200419
Ketonic Decarboxylation Reaction Mechanism: A
Combined Experimental and DFT Study
Angeles Pulido,* Borja Oliver-Tomas, Michael Renz,* Mercedes Boronat, and
Avelino Corma[a]
The ketonic decarboxylation of carboxylic acids has been car-
ried out experimentally and studied theoretically by DFT calcu-
lations. In the experiments, monoclinic zirconia was identified
as a good catalyst, giving high activity and high selectivity
when compared with other potential catalysts, such as silica,
alumina, or ceria. It was also shown that it could be used for
a wide range of substrates, namely, for carboxylic acids with
two to eighteen carbon atoms. The reaction mechanism for
the ketonic decarboxylation of acetic acid over monoclinic zir-
conia was investigated by using a periodic DFT slab model.
A reaction pathway with the formation of a b-keto acid inter-
mediate was considered, as well as a concerted mechanism, in-
volving simultaneous carbon–carbon bond formation and
carbon dioxide elimination. DFT results showed that the mech-
anism with the b-keto acid was the kinetically favored one and
this was further supported by an experiment employing a mix-
ture of isomeric (linear and branched) pentanoic acids.
Introduction
The conversion of biomass, instead of fossil fuels, into chemi-
cals and fuels is one of the most relevant research areas at the
present time and is important for both academia and indus-
try.[1–3] Biomass consists of highly oxy-functionalized molecules
and, for fuels and chemicals production, the carbon/oxygen
ratio needs to be increased. Thus, to upgrade biomass mole-
cules, they have to be initially dehydrated and hydrogenated,
whereas the molecular size has to be increased, by means of
CꢀC bond formation reaction, especially for the production of
liquid transportation fuels, and oxy functionalities have to be
removed. One of the routes that allows controled CꢀC cou-
pling and deoxygenation is decarboxylation of carboxylic
acids, also known as ketonic decarboxylation, in which two car-
boxylic acids, abundant in the biomass, are transformed into
a ketone with water and carbon dioxide as the only byprod-
ucts. The ketonic decarboxylation reaction can be written as
shown in Equation (1):
the reaction takes place between two different carboxylic acids
(R¼R’), then a mixture of three ketones is obtained: two sym-
metrical (RꢀCOꢀR and R’ꢀCOꢀR’) and one asymmetrical (Rꢀ
COꢀR’). Furthermore, note that fatty ketones, from decarboxy-
lation of fatty acids, are valuable items in the chemical industry
and find applications in ink manufacturing,[4] detergents,[5] or
in personal care products.[6] Many examples for ketonic decar-
boxylation starting from one carboxylic acid can be found in
industry. Dry distillation of calcium acetate to yield acetone
was reported in 1858 and has been used for the commercial
manufacture of acetone over 50 years.[7] However, acetone for-
mation is only one of the possible examples and production of
symmetrical and asymmetrical ketones (such as 2-pentanone,[8]
3-pentanone,[9] cyclopentanone[10,11] or propiophenone,[8,12]
among others[13–16]) reported. More recently, ketonic decarboxy-
lation has been proposed by Dumesic et al. as a key step for
the upgrading of valeric acid, which can be obtained, for in-
stance, from cellulose, to produce liquid transportation fuels.[17]
The primary product, namely 5-nonanone, can be hydrogenat-
ed to nonane and included in the kerosene fraction.
It has been observed that ketonic decarboxylation takes
place over a wide range of metal oxide catalysts (such as MgO,
BaO, CeO2, ZrO2, or TiO2, and mixtures thereof among
others),[13,15,18–36] although a preferred catalyst cannot be identi-
fied. Therefore, it is of paramount interest for the further devel-
opment of biomass transformations to optimize the catalytic
process for the conversion of acids into ketones and to select
the most suitable metal oxide catalyst. To control the catalytic
process, it is crucial to have a good understanding of the reac-
tion mechanism of ketonic decarboxylation and it is desirable
to have a deep insight into the most favorable reaction path-
ways of the mechanism, as well as possible competitive routes.
To date, controversy exists about the reaction mechanism that
If only one acid is used as a reactant in Equation (1) (R=R’),
then the product is a symmetric ketone (RꢀCOꢀR), whereas if
[a] Dr. A. Pulido, B. Oliver-Tomas, Dr. M. Renz, Dr. M. Boronat, Prof. A. Corma
Instituto de Tecnologꢀa Quꢀmica, Universitat Politꢁcnica de Valꢁncia
Consejo Superior de Investigaciones Cientꢀficas (UPV-CSIC)
Av. de los Naranjos s/n, 46022 Valencia (Spain)
Fax: (+34) 96 387 9444
E-mail: apulido@itq.upv.es
mrenz@itq.upv.es
Supporting Information for this article is available on the WWW under
http://dx.doi.org/10.1002/cssc.201200419.
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2013, 6, 141 – 151 141

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
takes place during ketonic decarboxylation. In 1939, Neun-
hoeffer and Paschke proposed a reaction mechanism for cyclo-
pentanone formation from adipic acid by means of a b-keto
acid intermediate, formed by abstraction of an a-hydrogen
atom, Ha (Scheme 1).[10] This mechanism has been further dis-
gated by Hites and Biemann, and they proposed a free-radical
mechanism in which alkyl and acyl radicals acted as initia-
tors.[38]
Valuable knowledge about the reaction mechanism of ke-
tonic decarboxylation can be extracted from the proposed
pathways taking place during adipic acid decarboxylation.[10,13]
However, this is a particular case and the transfer of this
knowledge to the ketonic decarboxylation of fatty acids in the
biomass is not so straightforward for the following reasons:
1) cyclopentanone can be obtained in excellent yield at much
lower reaction temperatures (250–3008C) than those required
for monocarboxylic acids (350 to 4008C); and 2) it is an intra-
molecular reaction, whereas we will be dealing mainly with in-
termolecular reactions during biomass transformations. There-
fore, we focused on the intermolecular ketonic decarboxylation
of monocarboxylic acids catalyzed by metal oxides. The use of
computational chemistry is highly desired to provide mecha-
nistic insight on the molecular scale, since no direct experi-
mental evidence about transition-state structures and other
mechanistic details can be obtained. Although some computa-
tional studies on ketonic decarboxylation over metal oxides
are available in the literature, they mainly focus on water and
acid adsorption,[39–41] and a detailed mechanistic investigation
is still missing. Herein, we report the excellent performance of
monoclinic zirconia oxide as a catalyst for ketonic decarboxyla-
tion with very high reaction yields and selectivity towards the
ketone, regardless of whether light or fatty acids, such as
acetic or stearic acid, were used as reactants. Moreover, the re-
action mechanism of acetone formation over monoclinic zirco-
nia (m-ZrO2) was investigated by means of a periodic DFT
model. The two previously proposed reaction mechanisms, de-
scribed above,[10,13] as well as competitive reaction routes[38]
were investigated, and a general overview of the ketonic de-
carboxylation reaction mechanism is given.
Scheme 1. Reaction mechanism for ketonic decarboxylation involving a) a b-
keto acid, proposed by Neunhoeffer and Paschke,[10] and b) the mechanism
proposed by Rand et al.[13]
cussed by other authors with different substrates and reaction
conditions, involving fixed-bed, continuous-flow reactors.[21,37]
Miller et al. reported ketonic decarboxylation of isobutyric acid,
lacking hydrogen atoms in the alpha position, over an aerogel
thoria catalyst for the formation of the asymmetric tert-butyl
isobutyl ketone with no evidence of di-tert-butyl ketone.[24]
They proposed that acids without alpha-hydrogen atoms may
be ketonically decarboxylated through a reaction occurring at
the beta carbon and resulting in an asymmetrical ketone.
Another argument in favor of this mechanism reported by
Koch and Leibnitz was the impossibility of obtaining a ketone
product from 2,2,5,5-tetramethyladipic acid or pivalic acid,[37]
molecules without any Ha atom. However, in 1962, the forma-
tion of 2,2,5,5-tetramethylcyclopentanone from 2,2,5,5-tetrame-
thyladipic acid was reported by Rand et al., and this cyclization
reaction could not be explained by the b-keto acid mechanism
previously proposed.[13] They suggested that the reaction sub-
strate was the mono-deprotonated diacid (Scheme 1).[13] Thus,
the deprotonated acid functionality is decarboxylated, forming
a carbanion that preferentially attacks the remaining carboxylic
group nucleophilically. However, an inconsistency was pointed
out for this proposed mechanism:[15] the carbanion formed
should be protonated to a large extent by the carboxylic acid
to form pentanoic acid (described as an alternative reaction
pathway in Scheme 1), but this product has never been report-
ed. A further development of Rand’s mechanism has been pro-
posed by one of us,[15] in which the decarboxylation and nucle-
ophilic attack occur in a concerted fashion, avoiding significant
drawbacks of Rand’s mechanism. It is also necessary to take
into account that, if the decarboxylation reaction involves tem-
peratures above 4508C, pyrolysis of the carboxylic acid will be
a competitive route leading to the formation of a series of un-
desired byproducts. Pyrolysis of calcium decanoate was investi-
Experimental Section
General
Acetic acid, pentanoic acid, decanoic acid, and stearic acid were
purchased from standard chemical suppliers, such as Acros or Al-
drich, and used as received. Monoclinic zirconium oxide and silicon
oxide were bought from ChemPur, Germany, as pellets with surface
areas of 100 and 240 m2 gꢀ1, respectively. CeO2 (nanopowder,
64 m2 gꢀ1) and Al2O3 (pellets, 330 m2 gꢀ1) were obtained from Al-
drich. ZrCeO4 (70 m2 gꢀ1) was synthesized by following a literature
procedure.[42]
Ketonic decarboxylation in a fixed-bed, continuous-flow re-
actor
The reaction apparatus is displayed in Figure 1. The catalyst (2.5 g,
pellets 0.4–0.8 mm) was diluted with silicon carbide and placed as
a fixed bed in a stainless-steel tube (0.93 cm external diameter).
The reactor was heated to reaction temperature T, and the feed
(5 mL) was passed through at a rate of 9.0 mLhꢀ1 together with
a nitrogen flow of 50 mLminꢀ1 at ambient pressure. The product
was condensed at room temperature and analyzed offline by gas
chromatography with dodecane as an external standard.
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2013, 6, 141 – 151 142

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
¯
Figure 2. PW91-optimized unit cell of the (111) facet surface of the m-ZrO2
model. Inequivalent zirconium (ZrA–D) and oxygen (Oa–g) atoms are indicated.
Distances of the surface vectors (u and v) are given in ꢁ. Zr and O atoms are
depicted in light and dark gray, respectively.
were below 0.015 eVꢁꢀ1 using only one irreducible k point and
a plane wave basis set with an energy cut off of 500 eV. Optimiza-
tion of the transition-state structure was performed with the dimer
method by using only first derivatives[51] and the nudged elastic
band method, as implemented by Henkelman and Jꢃnsson.[52] Elec-
tronic PW91 energies were further corrected by the addition of in-
termolecular dispersion energies, which were evaluated by using
the D3 method,[53,54] over the PW91-optimized geometries.
Figure 1. Schematic description of the fixed-bed, continuous-flow reactor
employed for gas-phase ketonic decarboxylation. FC=flow controller,
PC=manometer, TC=temperature controller.
Computational Methods
The stable phase of zirconium dioxide at low pressure between 0
and 11808C has monoclinic symmetry (space group P21/c) and its
experimental structure (lattice parameters a=5.150 ꢁ, b=5.212 ꢁ,
c=5.315 ꢁ, and b=99.238) was taken as the initial setup for mod-
eling bulk m-ZrO2. The DFT-optimized lattice vectors of m-ZrO2
were obtained from a Birch–Murnaghan fit[43,44] and were found to
be 1.2% larger than the experimental value (a=5.21 ꢁ, b=5.27 ꢁ,
and c=5.38 ꢁ). Periodic DFT calculations were carried out with the
VASP code[45] using the PW91 functional,[46,47] and core electrons
were described by using a projector-augmented-wave method
(PAW), as described by Blçchl[48] and adopted by Kresse and Jour-
bert.[49] For the Zr atoms, a total of 12 electrons (4s24p65s24d2)
were treated explicitly. DFT volume calculations were performed
by using a 4ꢂ4ꢂ4 k-point set, which corresponds to 16 irreducible
k-points, and a plane wave basis set with an energy cutoff of
800 eV.
Results and Discussion
With the aim of understanding the ketonic decarboxylation re-
action at the molecular scale, a combined theoretical and ex-
perimental study has been carried out. First, the performance
of several metal oxide catalysts, such as alumina, silica, ceria,
zirconia, and mixed zirconium cerium oxide, in the decarboxy-
lation of decanoic acid into 10-nonadecanone has been inves-
tigated experimentally when placed as fixed beds in a continu-
ous-flow reactor. From this series of experiments employing
different catalysts, and based on characteristics such as selec-
tivity towards ketone, reaction temperature, and conversion,
zirconia was selected as the best catalyst and submitted to fur-
ther investigations. Thus, the catalytic performance of zirconi-
um dioxide for ketonic decarboxylation of a series of carboxylic
acids, containing from two to eighteen carbon atoms, was
studied. Then, quantum chemistry was used to investigate, in
detail, the reaction pathways taking place during ketonic de-
carboxylation of acetic acid over zirconium dioxide by using
a periodic DFT-D model. Finally, the ketonic decarboxylation re-
action mechanism is discussed.
The relative stability of the seven inequivalent surfaces of m-ZrO2
has previously been investigated at the periodic DFT level by Chris-
tensen and Carter,[50] and the (111) surface has been found to be
¯
the most stable one. Therefore, the m-ZrO2 surface was modeled
¯
by using a periodic (111) facet slab model built up as follows: from
¯
the DFT-optimized bulk m-ZrO2, the (111) facet surface was built up
with surface vectors along the [110] and [101] directions with
lengths of 6.86 and 7.41 ꢁ, respectively (see Figure 2) and a dis-
tance of about 20 ꢁ between periodic images. This unit cell con-
tained three ZrO2 layers and had a chemical composition of
Zr12O24. The mechanistic investigation was carried out by using
Experimental study on ketonic decarboxylation over differ-
ent metal oxides
¯
¯
a periodic (111) facet slab model, denoted m-ZrO2(111), with a 2ꢂ
2ꢂ1 supercell and a composition of Zr48O96; see the Supporting In-
formation for additional details.
The ketonic decarboxylation of decanoic acid was carried out
at 350, 375, 400, and 4258C over zirconia, ceria, mixed zirconi-
um cerium oxide, silica, and alumina. The conversion and se-
lectivity for these reactions are displayed in Figure 3. Conver-
Geometry optimization was performed at the periodic PW91 level
by fixing the coordinates of the Zr and O atoms in the lowermost
ZrO2 layer, whereas all other atoms were fully relaxed until forces
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2013, 6, 141 – 151 143

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
Figure 4. Conversion and selectivity for the ketonic decarboxylation of differ-
ent carboxylic acids over ZrO2 (2.5 g) when employing a feed rate of
9 mLminꢀ1: a) acetic acid (C2H4O2; contact time W/F=0.26 ghgꢀ1
;
15.9 ghmolꢀ1), b) valeric (pentanoic) acid (C5H10O2; contact time W/
F=0.30 ghgꢀ1; 30.5 ghmolꢀ1), c) decanoic acid (C10H20O2; contact time W/
F=0.31 ghgꢀ1; 53.5 ghmolꢀ1), and d) stearic acid (C18H36O2; contact time W/
F=0.33 ghgꢀ1; 81.4 ghmolꢀ1).
rate (9.0 mLhꢀ1, see Figure 4). Hence, conversion was complete
in all cases at a reaction temperature of 4008C, for acetic acid
even at 3758C (see Figure 4a). The selectivity varied between
90 and 95%, with the exception of a reaction temperature of
4258C, where in many cases the selectivity dropped below
90%. Very similar results were obtained when the reaction was
compared at the same contact time for the weight of catalyst
and time per mol of substrate for the four carboxylic acids (for
details see the Supporting Information).
Figure 3. Conversion (C) and selectivity (S) for the ketonic decarboxylation of
decanoic acid, contact time weight of catalyst/feed rate (W/F)=0.31 ghgꢀ1
,
over different metal oxides: a) ZrO2, b) CeO2, c) CeZrO4, d) SiO2, and e) Al2O3.
sion increased when raising the temperature and full conver-
sion was observed at 4008C with zirconium oxide and zirconi-
um cerium mixed oxide, and at 4258C with cerium oxide (Fig-
ure 3a, c, and b, respectively). The mixed oxide CeZrO2 be-
haves exactly the same as a mixture of the pure oxides CeO2
and ZrO2 when the decline in conversion is considered below
4008C (cf. Figure 3c). For these three cases the selectivity
reached 90 to 95% in almost all cases. The other two oxides
employed, that is, SiO2 and Al2O3, were significantly less active.
With silica, conversion increased successively with temperature
up to 70% at 4258C, but the selectivity was never higher than
80% and varied between 65 and 80% (Figure 3d). Al2O3 gave
low conversions over almost the whole temperature range
(Figure 3e).
From these experimental results, it can be concluded that
monoclinic zirconium oxide is a suitable catalyst for the keton-
ic decarboxylation of carboxylic acids with a wide range of mo-
lecular weights, that is, from acetic acid to fatty acids.
Thus, to understand the ketonic decarboxylation reaction
mechanism at the molecular scale, the formation of acetone
(CH3ꢀCOꢀCH3), carbon dioxide (CO2), and water (H2O) from the
decarboxylation of two molecules of acetic acid (CH3ꢀCOOH)
was investigated by using DFT. Taking into account the experi-
mental results presented before, we have chosen m-ZrO2, mod-
¯
eled as a (111) facet surface by using a periodic slab model, m-
¯
The most favorable reaction conditions identified, with dec-
anoic acid as the substrate, were achieved with the use of
commercial pelletized m-ZrO2 at 4008C for which full conver-
sion was obtained together with 95% selectivity towards the
desired ketone. However, almost the same results were ach-
ieved with the zirconium cerium mixed oxide at the same tem-
perature or with cerium oxide at 4258C. Because the cerium
oxide and zirconium cerium mixed oxide had significantly
lower activity at temperatures below 4008C in comparison to
zirconium oxide, the latter was chosen for screening carboxylic
acids with different carbon chain lengths and a total carbon
atom number from two to eighteen. Similar trends were ob-
served for all carboxylic acids when passed at the same feed
ZrO2(111), as catalysts for the DFT study. Based on the previous-
ly proposed reaction mechanisms[10,13] described above for
adipic acid (see Scheme 1), decarboxylation of acetic acid was
investigated via formation of a b-keto acid intermediate and
through a concerted route, based on a modification of Rand’s
mechanism (see Scheme 2). Thus, intermediate and transition-
state structures along the reaction pathways following a b-
keto acid or a concerted route were identified, as well as other
competitive routes.
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2013, 6, 141 – 151 144

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
cation, orientation, and coordi-
nation of the molecule over the
surface, for which interaction
energies (DEPW91-D3
)
between
ꢀ67 and ꢀ172 kJmolꢀ1 were
obtained. The reactivity of 2-fc
(Oc) and 3-fc (Oa, Ob, Od, and Og)
surface oxygen atoms towards
acidic hydrogen abstraction was
also investigated. It was found
that H removal was a barrierless
process when the reaction took
place at the 2-fc coordinated
oxygen atoms, Oc, whereas
a
barrier of 8 kJmolꢀ1 was
Scheme 2. Reaction mechanism for acetic acid decarboxylation over m-ZrO2 via the b-keto acid (a) and concerted
found in the proximity of 3-fc
oxygen atoms, Oa and Og. It is
(b) routes.
also important to note that the intermediate adsorption spe-
cies formed, after acid deprotonation, in proximity to the 2-fc
oxygen atoms are about 30 kJmolꢀ1 more stable than those
DFT investigation of acetic acid decarboxylation over m-
ZrO2
In the case of gas-phase reactions on solid catalysts, adsorp- formed near the 3-f coordinated O atoms.
tion of reactant molecules over the catalyst surface is the first
step of the reaction mechanism. In the case of metal oxide cat-
b-Keto acid route
alysts, factors such as surface topology or surface metal and
oxygen atom coordination can play a crucial role in the reac- The reaction mechanism discussed first involves the formation
tion mechanism.[55] Thus, the reactivity of zirconium and of a b-keto acid intermediate, in line with the proposal by
oxygen atoms of the (111) facet surface of m-ZrO2 are dis- Neunhoeffer and Paschke,[10] and is referred to herein as the b-
¯
¯
cussed first. The unit cell of the periodic m-ZrO2 (111) surface keto acid route (Scheme 2). In Figure 5, the reaction pathway
slab model has four zirconium atoms (ZrA–D) and seven oxygen A!B!C!D!E!F!G!H!I shows schematically ketonic
atoms (Oa–g) per layer, which are not equivalent (see Figure 2). decarboxylation taking place along the b-keto acid route over
¯
Three out of the four zirconium atoms (ZrA, ZrB, and ZrC) are the m-ZrO2 (111) surface slab model. Intermediate and transi-
sixfold coordinated, denoted herein as 6-fc, whereas the last tion-state structures along the reaction pathway for acetic acid
¯
zirconium, ZrD, is 7-fc. Zirconium atoms in bulk m-ZrO2 are 7-fc, decarboxylation over the m-ZrO2 (111) surface are shown in
whereas oxygen atoms are 3- or 4-fc. Therefore, the ZrD atom Figure 6. Selected geometrical parameters are given in
is expected to be inert towards acetic acid molecule adsorp- Figure 7. A list of the intermediate and transition-state struc-
tion due to the lack of coordination vacancies, whereas the 6- tures involved in each step of the reaction pathway can be
fc zirconium atoms, ZrA–C, with one vacancy coordination, will found in the Supporting Information. The b-keto acid route
be active towards molecule adsorption. The seven inequivalent starts with the formation of the most stable acetic acid adsorp-
¯
surface oxygen atoms of the m-ZrO2 (111) surface are two-, tion complex (species 1 in Figures 6 and 7) with release of
three-, and fourfold coordinated. With regard to oxygen reac- 172 kJmolꢀ1. This intermediate 1 can be further stabilized by
tivity, the 4-fc oxygen atom, Oe, is expected to be almost inert interaction of the deprotonated oxygen atom with a second
due to a lack of vacancy coordination, as well as the 3-fc atom, zirconium atom to form a double-coordinated acetate inter-
Of, which is below the surface plane and molecule accessibility mediate 2. In the acetate intermediate 2, the ZrꢀO distances
is rather limited.
are about 2.25 ꢁ and the CꢀO distances are within 1.27–1.29 ꢁ.
Simple inspection of the geometry of intermediate 2 shown
Bearing in mind the previous description of zirconium and
oxygen atom reactivity, the adsorption of acetic acid over the in Figure 6 immediately reveals an acetate with a negative
¯
m-ZrO2 (111) surface was investigated. For all intermediate and charge. For the most relevant intermediate species, Bader’s
transition-state structures, the relative energy with respect to analysis of charge density is available in the Supporting Infor-
¯
reactants (m-ZrO2(111) surface and two gas phase molecules of mation and it has been used as a guide for naming the reac-
acetic acid) was evaluated at the PW91 (DEPW91) and at the tion intermediates. From acetate 2, abstraction of one out of
PW91-D3 (DEPW91-D3) levels, and are available in the Supporting the three a-hydrogen atoms, Ha, by the 2-fc surface oxygen
Information. Unless stated otherwise, PW91-D3 energies are atom leads to the formation of acetic acid dianion intermedi-
discussed below. The most stable acetic acid adsorption com- ate 4 via TS 3 (step B in Figure 5). This is an endothermic step,
plexes on the m-ZrO2 (111) surface were found when the or- DErxn =39 kJmolꢀ1, with an activation barrier of 75 kJmolꢀ1.
¯
ganic molecule interacted, through the carbonyl oxygen atom, Formation of the dianion for adsorbed acetic acid has been re-
¯
with one of the 6-fc surface zirconium atoms, ZrA, ZrB, or ZrC. cently investigated over the m-ZrO2 (111) surface at the period-
More than ten structures were investigated by varying the lo- ic DFT level where similar reaction energies and activation bar-
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2013, 6, 141 – 151 145

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
ecules and can be combined to
produce a molecule of water
(step E in Figure 5). From the
previously adsorbed H and OH
group, in the proximity of the
zirconium ZrC atom, water for-
mation was an endothermic
process
(DErxn ꢁ120 kJmolꢀ1
)
that needed an activation barri-
er of 136 kJmolꢀ1.
Theoretical investigations into
water adsorption on the m-ZrO2
¯
(111) surface has been previous-
ly carried out[39,41] and it has
been found that H2O can
adsorb to the zirconia surface
molecularly or dissociatively, de-
pending on the zirconium ad-
sorption site and water cover-
age. Water adsorption processes
¯
on the m-ZrO2 (111) surface
have been reported as exother-
mic ones, where energy is re-
leased in the range between 80
and 125 kJmolꢀ1 at the DFT
level, which is in good agree-
ment with our reported energy
barrier for water formation at
the zirconium ZrC site (PW91
values available in the Support-
ing Information). After water de-
sorption, an acetic acid dianion
and an acylium intermediate are
closely adsorbed on the zirconia
surface (structure 10). Then the
terminal CH2 carbon atom of
the dianion attacks the carbon
atom of the carbonyl group of
the acylium intermediate (step F
in Figure 5), leading to the for-
mation of the b-keto acid inter-
mediate (as a carboxylate, struc-
ture 12) via TS 11, where the
new CꢀC bond is already par-
Figure 5. Reaction pathways for ketonic decarboxylation of acetic acid (R=R’=CH3) over m-ZrO2 following the b-
keto acid (bottom) and concerted (top) routes. PW91-D3 reaction and activation (in italic and with superscript #)
*
*
)
energies of each step are shown. Zr and O atoms of the ZrO2 surface are represented by empty ( ) or filled (
tially formed with
a
bond
dots, respectively. For the sake of clarity, double bonds and charges are not displayed.
length of 1.825 A (see Figure 7).
This is an exothermic step
(ꢀ93 kJmolꢀ1) with an activa-
riers have been found.[40] The reaction pathway continues
(step C in Figure 5) with adsorption of a second molecule of
acetic acid (structure 5) in the proximity of the previously
formed dianion. The second molecule of acetic acid can be
easily dehydroxylated (Ea =37 kJmolꢀ1), leading to the forma-
tion of an adsorbed acylium intermediate 7 via TS 6 (step D in
Figure 5), in which the CꢀOH bond is already partially broken
with a length of 1.448 ꢁ. At this point, a hydrogen and hydrox-
yl group have already been removed from the acetic acid mol-
tion barrier of 57 kJmolꢀ1. From the b-keto carboxylate inter-
mediate 12, acetone enolate and CO2 can be formed (step G in
Figure 5) by breaking of the CꢀC bond (originally in the dia-
nion fragment) via TS 13 (see Figures 6 and 7). This is an endo-
thermic process (88 kJmolꢀ1) with an activation barrier of
108 kJmolꢀ1. Regardless of water formation at very unfavorable
locations, this is the rate-determining step of the b-keto acid
route. Acetone formation by transfer of an adsorbed H atom
to the acetone enolate intermediate has a barrier of only
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2013, 6, 141 – 151 146

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
the carbon–carbon distance is 1.555 ꢁ, but simulta-
neous dehydroxylation also takes place and the Cꢀ
OH distance is enlarged to 3.005 ꢁ (see Figures 8
and 9). This is an endothermic process,
+45 kJmolꢀ1, and it is the rate-determining step of
the concerted mechanism with an activation barrier
of 154 kJmolꢀ1; this is about 50 kJmolꢀ1 higher than
the rate-determining step of the b-keto acid route.
At a reaction temperature of 4008C (corresponding
to a thermal energy of ꢁ8 kJmolꢀ1), this could be
considered a significant energy difference. Closing of
the catalytic cycle by water, carbon dioxide, and ace-
tone desorption would proceed as previously de-
scribed for the b-keto acid route.
As an alternative reaction pathway for the con-
certed route, we investigated the addition of the
methyl carbanion of the adsorbed acetate species to
an adsorbed acylium intermediate (steps O!P in
Figure 5). This was
a highly endothermic step
(DErxn =121 kJmolꢀ1) with a large activation energy
barrier, 241 kJmolꢀ1. Although the adsorbed acylium
ion (Figure 9, structure 23) involves an activated nu-
cleophilic cationic carbon atom and a carbanion
attack (see TS 24 in Figures 8 and 9) could be ex-
pected to be favorable, the strong interaction be-
tween the cationic carbon atom and the oxygen
surface, C···O distance of 1.301 ꢁ, makes it a highly
stabilized species and quite inert towards nucleo-
philic attack.
Figure 6. PW91-optimized structures involved in the b-keto acid route of acetic acid de-
carboxylation over m-ZrO2. Zr and O surface atoms are depicted as light and dark gray
sticks, respectively, whereas H, C, and O atoms belonging to the acetic acid molecule are
depicted as white, light gray, and dark gray balls, respectively.
Competitive routes
17 kJmolꢀ1, in a process (step H in Figure 5) that is almost ther-
The radical route was proposed for high-tempera-
moneutral. The catalytic cycle is closed (step I in Figure 5) by
ture reactions (5008C), in which pyrolysis of carboxylic acids
starts to become a competitive route.[38] The high selectivity
observed for ketonic decarboxylation over ZrO2 indicates that
cracking reactions are not very important. However, with the
aim of completing the picture of the mechanism as much as
possible, radical fragmentation was investigated and the ques-
tion was considered as to whether surface-assisted reactions
could be competitive pathways of the b-keto acid or concerted
routes. For this purpose, alternative routes for the formation of
adsorbed methyl species and acetone were investigated.
On the one hand, homolytic breaking of the CꢀC bond in the
adsorbed acetate intermediate 26, generating an adsorbed
methyl group (structure 27; step Q in Figure 10), is a highly en-
dothermic process, 278 kJmolꢀ1, with an energy barrier greater
than 275 kJmolꢀ1. As an alternative pathway to homolytic
bond cleavage in the adsorbed acetate, surface-assisted activa-
tion of the C=C double bond of the acetic acid dianion inter-
mediate was considered (step R in Figure 10). This is also
a competitive route to the formation of b-keto acid intermedi-
ate 12 from the dianion intermediate. It was found that form-
ing a bond between the surface and the carbon atom of the
CH2 group of the dianion is an endothermic process
(120 kJmolꢀ1) with an activation energy of 272 kJmolꢀ1 (see
Figures 11 and 12). Note that these processes could occur
acetone desorption, DErxn =84 kJmolꢀ1.
Concerted route
A concerted reaction pathway, involving carbon–carbon bond
formation and carbon dioxide formation in the same step (J!
K!L!M!N in Figure 5), inspired by Rand’s mechanism,[13]
was also investigated for acetic acid decarboxylation over the
¯
(111) zirconia surface. Intermediate and transition-state struc-
tures along the reaction pathway were identified at the period-
ic DFT level and are shown in Figure 8, whereas some geomet-
rical parameters are given in Figure 9. The concerted route also
starts (step J in Figure 5) with the formation of a deprotonated
acetic acid molecule on the zirconia surface (structure 18) with
release of 160 kJmolꢀ1. Then another molecule of acetic acid
needs to be weakly adsorbed in close proximity (DErxn
=
ꢀ94 kJmolꢀ1, step K in Figure 5). Thus, from the surface-ad-
sorbed molecules (structure 19), it is possible that the forma-
tion of acetone (structure 21) occurs by transfer of the methyl
carbanion (labeled R in the upper part of Figure 5) to the
carbon atom of the carboxyl group of the other adsorbed mol-
ecule (step L in Figure 5). The formation of TS 20 in this step
shows that the new CꢀC bond is already partially formed and
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2013, 6, 141 – 151 147

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
lation of acetic acid for kinetic
reasons over the concerted
Rand’s route involving simulta-
neous carbon dioxide and
carbon–carbon bond formation.
To test DFT predictions, ketonic
decarboxylation of a mixture of
two C5 carboxylic acids, namely,
valeric acid and pivalic acid, was
also studied experimentally
over the m-ZrO2 catalyst. Valeric
acid, CH3ꢀ(CH2)3ꢀCOOH, has
two Ha, whereas pivalic acid,
(CH3)3CꢀCOOH, has no Ha
atoms (see Scheme 3). As men-
tioned before, when the reac-
tion takes place with a mix of
two acids, three ketones (Rꢀ
COꢀR, RꢀCOꢀR’ and R’ꢀCOꢀR’)
should be obtained in a statisti-
cal ratio of 25:50:25. Thus, from
our mixture of valeric and pival-
ic acid, we should obtain the
two symmetric ketones, CH3ꢀ
(CH2)3ꢀCOꢀ(CH2)3ꢀCH3 (5-nona-
none) and (CH3)3CꢀCOꢀC(CH3)3
(2,2,4,4-tetramethyl-3-penta-
Figure 7. Geometry of the PW91-optimized structures involved in the b-keto acid route of acetic acid (R=CH3) de-
carboxylation over m-ZrO2. Distances are given in ꢁ. Zr and O atoms of the ZrO2 surface are represented by
*
*
empty ( ) or filled ( ) dots, respectively. For the sake of clarity, double bonds and charges are not displayed.
none), and the asymmetric one,
CH3ꢀ(CH2)3ꢀCOꢀC(CH3)3 (2,2-di-
methyl-3-heptanone). However,
during cracking of the acetic acid molecule, but are not ex-
pected to be competitive for the formation of the b-keto acid
intermediate from the dianion intermediate. On the other
hand, addition of the methyl group from homolytic bond
cleavage to adsorbed acylium species 30 (see Figures 11 and
12), via TS 31, to form adsorbed acetone (structure 17) is an
exothermic process (step S in Figure 10), with the release of
119 kJ molꢀ1 and an activation barrier of 350 kJmolꢀ1 is
needed.
the product distribution experimentally observed is clearly dif-
ferent, as shown in Scheme 3, for valeric and pivalic acid. From
the results, it can be observed that 1) there is low conversion
of pivalic acid (20%) compared with almost full conversion of
valeric acid (99%); and 2) selectivity towards the symmetric
ketone from decarboxylation of valeric acid, that is, 5-nona-
none, is about 89% and towards the asymmetric ketone, 2,2-
dimethyl-3-heptanone, about 4%, whereas the symmetric
ketone of pivalic acid, that is, 2,2,4,4-tetramethyl-3-pentanone,
is not observed at all.
Reaction pathways for acetic acid decarboxylation through
the b-keto acid and concerted routes were also investigated at
the periodic PW91 level, in addition to the PW91-D3 data dis-
cussed above. The inclusion of intermolecular dispersion inter-
action energies preserves the energy profile obtained at the
PW91 level (see the Supporting Information). Thus, for a given
elementary step, differences between calculated PW91 and
PW91-D3 activation barriers are up to approximately
In principle, 5-nonanone would be accessible by ketonic de-
carboxylation of valeric acid through the b-keto acid or con-
certed Rand’s route, although based on our DFT results the b-
keto acid route is the most favorable reaction pathway. Thus,
the valeric acid dianion can be formed by removal of the
acidic H and one of the two a-hydrogen atoms. Then, the vale-
ric acid dianion could attack the valeric acid acylium cation to
form 5-nonanone as the product or the pivalic acid acylium
cation, forming 2,2-dimethyl-3-heptanone; in both cases via
the corresponding b-keto acid intermediate. Formation of both
adsorbed acylium ions is expected to be energetically accessi-
ble, due to the low activation energy found for acetic acid de-
hydroxylation. At this point, we cannot explain unambiguously
the higher selectivity observed for 5-nonanone (symmetric
ketone formation) with respect to 2,2-dimethyl-3-heptanone
(asymmetric) when compared with the statistical ratio. First,
a steric hindrance effect was considered due to the presence
15 kJmolꢀ1
.
Further experimental evidence for the b-keto acid route as
the ketone decarboxylation reaction mechanism
From the previously presented experimental data, we can see
that m-ZrO2 catalysts are promising candidates for biomass in-
dustrial transformations, but little information can be extracted
concerning the reaction mechanism itself. At this point, DFT in-
vestigations favor the b-keto acid route for ketonic decarboxy-
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2013, 6, 141 – 151 148

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
Figure 9. Geometry of the PW91-optimized structures involved in the con-
certed route of the acetic acid (R=R’=CH3) decarboxylation over m-ZrO2.
Distances are given in ꢁ. Zr and O atoms of the ZrO2 surface are represented
*
*
as empty ( ) or filled ( ) dots, respectively. For the sake of clarity, double
bonds and charges are not displayed.
Figure 8. PW91-optimized structures involved in concerted route of acetic
acid decarboxylation over m-ZrO2. Zr and O surface atoms are depicted as
light and dark gray sticks, respectively, whereas H, C, and O atoms belonging
to the acetic acid molecule are depicted as white, light gray, and dark gray
balls, respectively.
of the three methyl groups in the pivalic acid acylium cation,
which makes it less likely to have both reactant fragments (va-
leric acid dianion and pivalic acid acylium carbocation) in close
proximity. However, for a clearer understanding of this issue,
a deeper study is currently under development. The ketonic
decarboxylation of acetic and pivalic acid over m-ZrO2 is inves-
tigated at the periodic DFT level. Thus, formation of the sym-
metric, CH3ꢀCOꢀCH3, and asymmetric, CH3ꢀCOꢀCꢀ(CH3)3, ke-
tones is considered to take place by the b-keto acid and con-
certed route. Preliminary results show that activation barriers
for ketone formation involving acetic and pivalic acid decar-
boxylation are larger than those obtained for decarboxylation
of two molecules of acetic acid.
Figure 10. PW91-D3 reaction and activation (in italic and with superscript #)
energies of each step are shown (R=CH3). Zr and O atoms of the ZrO2 sur-
*
*
face are represented as empty ( ) or filled ( ) dots, respectively. For the
sake of clarity, double bonds and charges are not displayed.
certed Rand’s route, due to the lack of a-hydrogen atoms.
However, the product has not been observed and this strongly
supports DFT predictions that the b-keto acid route over the
Formation of 2,2,4,4-tetramethyl-3-pentanone from ketonic
decarboxylation of pivalic acid can only occur through the con-
¯
m-ZrO2 (111) surface catalyst is kinetically favored and pre-
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2013, 6, 141 – 151 149

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
Figure 12. Geometry of the PW91-optimized structures involved in the com-
petitive routes of acetic acid (R=CH3) decarboxylation over m-ZrO2. Distan-
ces are given in ꢁ. Zr and O atoms of the ZrO2 surface are represented by
*
*
empty ( ) or filled ( ) dots, respectively.
Finally, it has to be stated that, from DFT, it was found that
formation of the acetic acid dianion had an energy barrier of
75 kJmolꢀ1, whereas it was only 36 kJmolꢀ1 for the reverse re-
action—protonation of the dianion. Therefore, acetic acid di-
anion protonation competes with b-keto acid intermediate for-
mation that has a similar activation barrier (37 kJmolꢀ1). Thus,
although distribution of equilibrium species over the zirconia
oxide surface is expected to be influenced by reaction condi-
tions (such as gas pressure and temperature), it is concluded
that the ketonic decarboxylation of monocarboxylic acids into
ketones proceeds predominantly via b-keto acid intermediates.
Figure 11. PW91-optimized structures involved in the competitive routes of
acetic acid decarboxylation over m-ZrO2. Zr and O surface atoms are depict-
ed as light and dark gray sticks, respectively, whereas H, C, and O atoms be-
longing to the acetic acid molecule are depicted as white, light gray, and
dark gray balls, respectively.
ferred over the concerted Rand’s route. Pivalic acid decarboxy-
lation over aerogel thoria at 4908C has been reported by Miller
et al.,[24] leading mainly to the asymmetric tert-butyl isobutyl
ketone with no evidence of 2,2,4,4-tetramethyl-3-pentanone
formation. They explained that the reaction took place via a g-
keto acid. Alternatively, the formation of the product can be ra-
tionalized by pivalic acid isomerization with subsequent keton-
ic decarboxylation by the b-keto acid route, thus supporting
this mechanism further. In our case, this asymmetric ketone
has not been found, probably due to milder reaction
conditions.
Conclusions
Ketonic decarboxylation of carboxylic acids was carried out ex-
perimentally and studied theoretically by DFT calculations.
In the experiments m-ZrO2 was identified as a good catalyst,
involving high activity and high selectivity when compared
with other potential catalysts, such as silica, alumina, or ceria,
and it was shown that it could be used for a wide range of
substrates, namely, for carboxylic acids containing from two to
eighteen carbon atoms.
The reaction mechanism for
the ketonic decarboxylation of
acetic acid over m-ZrO2 was in-
vestigated by using a periodic
DFT slab model. A reaction path-
way with the formation of a b-
keto acid intermediate was con-
sidered, as well as a concerted
mechanism, involving simultane-
ous carbon–carbon bond forma-
tion and carbon dioxide elimina-
tion. DFT results show that the
Scheme 3. Product distribution for ketone decarboxylation of a mix of valeric and pivalic acid (1:1 ratio).
mechanism with the b-keto acid
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2013, 6, 141 – 151 150

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
´
[19] M. Glinski, J. Kijenski, A. Jakubowski, Appl. Catal. A 1995, 128, 209–217.
[20] A. Westfechtel, C. Breucker, B. Gutsche, L. Jeromin, H. Eierdanz, H. Bau-
mann, K. H. Schmid, W. Nonnenkamp, DE 4121117, 1993.
[21] R. Martinez, M. C. Huff, M. A. Barteau, J. Catal. 2004, 222, 404–409.
[22] K. Parida, H. K. Mishra, J. Mol. Catal. A 1999, 139, 73–80.
[23] A. Ignatchenko, E. Kozliak, ACS Catal. 2012, 2, 1555–1562.
[24] A. Miller, N. Cook, F. Whitmore, J. Am. Chem. Soc. 1950, 72, 2732–2735.
[25] S. Sugiyama, K. Sato, S. Yamasaki, K. Kawashiro, H. Hayashi, Catal. Lett.
1992, 14, 127–133.
was kinetically favored; this was further supported by an ex-
periment employing a mixture of isomeric pentanoic acids.
Therefore, the mechanism involving a b-keto acid was pro-
posed as the predominant mechanism for the selective ketonic
decarboxylation of monocarboxylic acids.
Acknowledgements
[26] R. Pestman, R. M. Koster, A. v. Duijne, J. A. Z. Pieterse, V. Ponec, J. Catal.
1997, 168, 265–272.
We thank MINECO (MAT2011-28009, Consolider Ingenio 2010-
MULTICAT, CSD2009-00050 and CTQ2011-27550) and the Spanish
National Research Council (CSIC, Es 2010RU0108) for funding.
Red EspaÇola de Supercomputaciꢂn (RES) and Centre de Cꢃlcul
de la Universitat de Valꢁncia are acknowledged for computation-
al facilities and technical assistance. A.P. and B.O.-T. thank
MINECO (Juan de la Cierva Programme) and CSIC (JAE Pro-
gramme), respectively, for their fellowships.
[27] D. M. Cowan, G. H. Jeffery, A. I. Vogel, J. Chem. Soc. 1940, 171–176.
[28] G. P. Hussmann, US 4754074, 1988.
[29] E. Mꢄller-Erlwein, B. Rosenberger, Chem. Ing. Tech. 1990, 62, 512–513.
[30] F. Wattimena, EP 85996, 1983.
[31] S. Rajadurai, Catal. Rev. Sci. Eng. 1994, 36, 385–403.
[32] J. Cressely, D. Farkhani, A. Deluzarche, A. Kiennemann, Mater. Chem.
Phys. 1984, 11, 413–431.
[33] J. C. Kuriacose, S. S. Jewur, J. Catal. 1977, 50, 330–341.
[34] V. Lorenzelli, G. Busca, N. Sheppard, J. Catal. 1980, 66, 28–35.
[35] E. Mꢄller-Erlwein, Chem. Ing. Tech. 1990, 62, 416–417.
[36] L. J. Gooßen, P. Mamone, C. Oppel, Adv. Synth. Catal. 2011, 353, 57–63.
[37] H. Koch, E. Leibnitz, Periodica Polytech. 1961, 5, 149–185.
[38] R. A. Hites, K. Biemann, J. Am. Chem. Soc. 1972, 94, 5772–5777.
[39] A. Ignatchenko, D. G. Nealon, R. Dushane, K. Humphries, J. Mol. Catal. A
2006, 256, 57–74.
Keywords: ab initio calculations · biomass · carboxylic acids ·
heterogeneous catalysis · ketones
[1] G. W. Huber, S. Iborra, A. Corma, Chem. Rev. 2006, 106, 4044–4098.
[2] A. Corma, M. Renz, C. Schaverien, ChemSusChem 2008, 1, 739–741.
[3] D. Martin Alonso, J. Q. Bond, J. A. Dumesic, Green Chem. 2010, 12,
1493–1513.
[4] S. L. Malhotra, R. W. Wong, M. P. Breton, Xerox Corporation, US 6461417,
2002.
[5] A. D. Tomlinson, Unilever PLC, WO 2001094522, 2001.
[6] W. Seipel, H. Hensen, N. Boyxen, Cognis Deutschland GmbH, DE
19958521, 2001.
[7] C. Friedel, Justus Liebigs Ann. Chem. 1858, 108, 122–125.
[8] H. Froehlich, M. Schneider, W. Himmele, M. Strohmeyer, G. Sandrock, K.
Baer, EP 2758113, 1979.
[40] A. V. Ignatchenko, J. Phys. Chem. C 2011, 115, 16012–16018.
[41] S. T. Korhonen, M. Calatayud, A. O. I. Krause, J. Phys. Chem. C 2008, 112,
6469–6476.
[42] J. C. Serrano-Ruiz, J. Luettich, A. Sepffllveda-Escribano, F. Rodríguez-Rein-
oso, J. Catal. 2006, 241, 45–55.
[43] F. D. Murnaghan, Proc. Natl. Acad. Sci. USA 1944, 30, 244–247.
[44] F. Birch, Phys. Rev. 1947, 71, 809–824.
[45] G. Kresse, J. Furthmꢄller, Phys. Rev. B 1996, 54, 11169–11186.
[46] J. P. Perdew, Y. Wang, Phys. Rev. B 1992, 45, 13244–13249.
[47] J. P. Perdew, J. A. Chevary, S. H. Vosko, K. A. Jackson, M. R. Pederson,
D. J. Singh, C. Fiolhais, Phys. Rev. B 1992, 46, 6671–6687.
[48] P. E. Blçchl, Phys. Rev. B 1994, 50, 17953–17979.
[9] H. Siegel, M. Eggersdorfer in Ullmann’s Encyclopedia of Industrial
Chemistry (Ed.: W. Gerhartz), Weinheim, VCH, 1990, 83–95.
[10] O. Neunhoeffer, P. Paschke, Ber. Dtsch. Chem. Ges. A/B 1939, 72, 919–
929.
[49] G. Kresse, D. Joubert, Phys. Rev. B 1999, 59, 1758–1775.
[50] A. Christensen, E. A. Carter, Phys. Rev. B 1998, 58, 8050–8064.
[51] G. Henkelman, H. Jꢃnsson, J. Chem. Phys. 1999, 111, 7010–7022.
[52] G. Henkelman, H. Jꢃnsson, J. Chem. Phys. 2000, 113, 9978–9985.
[53] S. Grimme, J. Antony, S. Ehrlich, H. Krieg, J. Chem. Phys. 2010, 132,
154104.
[11] Bayer & Co, DE 256622, 1911.
[12] C. A. Smith, L. F. Theiling Jr., US 4172097, 1979.
[13] L. Rand, W. Wagner, P. O. Warner, L. R. Kovac, J. Org. Chem. 1962, 27,
1034–1035.
[54] http://toc.uni-muenster.de/DFTD3/.
[55] M. A. Barteau, Chem. Rev. 1996, 96, 1413–1430.
[14] V. I. Yakerson, A. M. Rubinshtein, L. A. Gorskaya, GB 1208802, 1970.
[15] M. Renz, Eur. J. Org. Chem. 2005, 979–988.
[16] H. Lermer, W. Hoelderich, M. Schwarzmann, DE 3638005, 1986.
[17] J. C. Serrano-Ruiz, D. Wang, J. A. Dumesic, Green Chem. 2010, 12, 574–
577.
Received: June 18, 2012
Revised: September 14, 2012
Published online on November 30, 2012
[18] W. A. Beavers, US 20070100166, 2007.
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2013, 6, 141 – 151 151