Glycerolysis of methyl oleate on MgO: Experimental and theoretical study of the reaction selectivity

Article abstract of DOI:10.1016/j.jcat.2015.01.001

The liquid-phase MgO-promoted glycerolysis of methyl oleate, a fatty acid methyl ester (FAME), to give acylglycerol products was studied both, experimentally and by density functional theory (DFT). Catalytic results showed that strongly basic low coordination O^2- surface sites participate in kinetically relevant steps of the glycerolysis reaction. Changes in the selectivity toward the different mono- and diglyceride isomers were investigated by varying the reaction conditions. The main product was always α-glyceryl monooleate (α-MG), a monoglyceride with the ester fragment at one of the terminal positions of the glycerol molecule; the β-MG isomer, with the ester substituted at position 2 was obtained in much lower amounts. The molecular modeling of glycerol (Gly) and FAME adsorptions as well as of the glycerolysis reaction was carried out using periodic DFT calculations and a model of stepped MgO surface. Results indicated that FAME was more weakly adsorbed than Gly; the latter adsorbs on a coordinatively unsaturated surface O^2- site with O-H bond breaking at position 2 of the Gly molecule, giving therefore a surface β-glyceroxide species. Calculations explained the apparent contradiction between the preferential formation of the α-MG isomer and the energetically favored dissociation of the secondary OH group of Gly that leads to the β-glyceroxide species. They predict that the β-glyceroxide species participates in the pathways conducting to both, α- and β-MG isomers. Synthesis of α-MG occurs by C-O coupling of β-glyceroxide with FAME at one of the two primary OH groups of the β-glyceroxide species. Two transition states (TS) and a tetrahedral intermediate (TI) are involved in both, α-MG and β-MG isomer formation. However, the pathway toward β-MG is limited by the large sterical effects associated to the TI formation. Contrarily, the TI leading to α-MG is relatively easy to form.

Full text of DOI:10.1016/j.jcat.2015.01.001

Journal of Catalysis 323 (2015) 132–144
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Glycerolysis of methyl oleate on MgO: Experimental and theoretical
study of the reaction selectivity
b
b
c
b,
P.G. Belelli a, , C.A. Ferretti , C.R. Apesteguía , R.M. Ferullo , J.I. Di Cosimo
⇑
⇑
a Group of Materials and Catalytic Systems (GMSC) – IFISUR (UNS-CONICET), Av. Alem 1253, 8000 Bahía Blanca, Argentina
b Catalysis Science and Engineering Research Group (GICIC) – INCAPE (UNL-CONICET), Santiago del Estero 2654, 3000 Santa Fe, Argentina
c INQUISUR (UNS-CONICET), Av. Alem 1253, 8000 Bahía Blanca, Argentina
a r t i c l e i n f o
a b s t r a c t
Article history:
The liquid-phase MgO-promoted glycerolysis of methyl oleate, a fatty acid methyl ester (FAME), to give
acylglycerol products was studied both, experimentally and by density functional theory (DFT). Catalytic
results showed that strongly basic low coordination O2ꢀ surface sites participate in kinetically relevant
steps of the glycerolysis reaction. Changes in the selectivity toward the different mono- and diglyceride
Received 6 November 2014
Revised 2 January 2015
Accepted 3 January 2015
isomers were investigated by varying the reaction conditions. The main product was always
a-glyceryl
monooleate ( -MG), a monoglyceride with the ester fragment at one of the terminal positions of the glyc-
erol molecule; the b-MG isomer, with the ester substituted at position 2 was obtained in much lower
amounts.
The molecular modeling of glycerol (Gly) and FAME adsorptions as well as of the glycerolysis reaction
was carried out using periodic DFT calculations and a model of stepped MgO surface. Results indicated
that FAME was more weakly adsorbed than Gly; the latter adsorbs on a coordinatively unsaturated sur-
face O2ꢀ site with OAH bond breaking at position 2 of the Gly molecule, giving therefore a surface b-glyc-
eroxide species. Calculations explained the apparent contradiction between the preferential formation of
a
Keywords:
Glycerolysis
Monoglyceride
Isomer selectivity
DFT
Base catalysis
Molecular modeling
Periodic calculation
the
to the b-glyceroxide species. They predict that the b-glyceroxide species participates in the pathways
conducting to both, - and b-MG isomers. Synthesis of -MG occurs by CAO coupling of b-glyceroxide
with FAME at one of the two primary OH groups of the b-glyceroxide species. Two transition states
(TS) and a tetrahedral intermediate (TI) are involved in both, -MG and b-MG isomer formation. How-
ever, the pathway toward b-MG is limited by the large sterical effects associated to the TI formation. Con-
trarily, the TI leading to -MG is relatively easy to form.
a-MG isomer and the energetically favored dissociation of the secondary OH group of Gly that leads
a
a
a
a
Ó 2015 Elsevier Inc. All rights reserved.
1. Introduction
Gly is traditionally included in the formulation of food, cosmetic
and pharmaceutical products, liquid detergents, and antifreeze [1].
Other recent applications comprise the production of hydrogen [2],
liquid fuels [3], fuel additives [4], and chemicals [1,5,6]. Further-
more, environmental concerns are the driving force to replace a
petroleum-based feedstock such as mineral Gly with its renewable
biomass-derived substitute in the synthesis of new or conventional
products.
Acylglycerols, the glycerol estersof fatty acids, canbe synthesized
by glycerolysis of fatty acid methyl esters (FAME) with Gly, as shown
in Scheme 1. In particular, the monoesters or monoglycerides (MG)
are an attractive option to transform bio-Gly into valuable chemi-
cals; they are widely used as surfactants due to their emulsifying
properties that help hydrophilic and lipophilic substances mix
together. Therefore, they find application in many food, detergent,
plasticizer, cosmetic, and pharmaceutical formulations [5,7].
During biodiesel synthesis by oil or fat transesterification,
glycerol (Gly) is obtained as a coproduct representing 10% of the
biodiesel production.
Lately, the increasing participation of biofuels in the energy
matrix of many countries has generated a Gly surplus with the
consequent drop of the unrefined Gly price. Thus, many efforts
have been devoted recently to develop new applications intended
to convert glycerol into value-added chemicals with the purpose to
improve the economics of biodiesel production.
⇑
Corresponding authors. Fax: +54 291 4595142 (P.G. Belelli), +54 342 4511170
(J.I. Di Cosimo).
E-mail addresses: patricia.belelli@uns.edu.ar (P.G. Belelli), dicosimo@fiq.unl.edu.
ar (J.I. Di Cosimo).
http://dx.doi.org/10.1016/j.jcat.2015.01.001
0021-9517/Ó 2015 Elsevier Inc. All rights reserved.

P.G. Belelli et al. / Journal of Catalysis 323 (2015) 132–144
133
Glycerolysis of FAME is a base-catalyzed transesterification
reaction. In a previous work [8], the glycerolysis reaction was stud-
ied using methyl oleate (C18:1) as FAME and MgO as catalyst. We
investigated the chemical nature of the base sites responsible for
the catalytic activity of MgO, both experimentally and by density
functional theory (DFT). The catalytic results showed the preferen-
a N2 flow for 18 h at 773 K to obtain high-surface area MgO. The
sample was finally ground and sieved, and the particles with aver-
age particle size of 177–550
experiments.
lm were used in the catalytic
The BET surface area (SA) was determined by N2 physisorption
at 77 K in a NOVA-1000 Quantachrome sorptometer. The structural
properties of MgO were determined by X-ray diffraction (XRD)
tial formation of
a-MG, Scheme 1, indicating that the FAME-Gly
coupling with formation of a new CAO bond is produced mostly
at the primary hydroxyl of the glycerol molecule after OAH bond
breaking. Contrarily, the DFT calculations using the cluster model
approximation [8] predicted that upon Gly adsorption on MgO,
the OAH bond of the secondary hydroxyl is dissociated, forming
a surface b-glyceroxide, Scheme 2, that might give rise to a reaction
pathway leading b-MG.
To get insight into this selectivity issue, in this work, we con-
tinue our studies of the methyl oleate glycerolysis on MgO. We
report our investigations about the effect of different operational
variables on the glyceride isomer selectivity. Furthermore, molec-
ular modeling using periodic calculations supports the preferential
using a Shimadzu XD-D1 diffractometer equipped with CuKa radi-
ation source (k = 0.1542 nm).
MgO base properties were measured by temperature pro-
grammed desorption (TPD) and infrared spectroscopy (FTIR) of
CO2. For the TPD experiment, the sample was pretreated in situ
in a N2 flow at 773 K to remove water and carbonates formed dur-
ing storage, cooled down to room temperature, and then exposed
to a flowing mixture of 3% of CO2 in N2, until surface saturation
was achieved (5 min). Weakly adsorbed CO2 was removed by
flushing with N2. Finally, the temperature was increased to 773 K
at a ramp rate of 10 K/min. Desorbed CO2 was converted in CH4
on a methanation catalyst (Ni/Kieselguhr) and then analyzed using
formation of
a-MG and allows us to postulate reaction mecha-
a flame ionization detector (FID). Total base site number (Nb,
lmol/
nisms for -MG and b-MG formation. The reasons for the former
a
g) was measured as the evolved CO2 obtained by integration of TPD
curves. The total base site density in l
nb = Nb/SA.
contradictory results are therefore elucidated since calculations
show that the b-glyceroxide surface species participates in the for-
mation of both MG isomers.
mol/m2 was calculated as
The chemical nature of adsorbed surface CO2 species was deter-
mined by FTIR after CO2 adsorption at room temperature and
sequential evacuation at increasing temperatures. The sample
was pressed in a wafer and degassed in vacuum at 773 K for 1 h
and then cooled down to room temperature. The spectrum of the
pretreated catalyst was then taken. After admission of 5 kPa of
CO2 and evacuation at 298, 373, 473, and 573 K, the CO2 adsorption
spectra were recorded at room temperature. Spectra of the
adsorbed species were obtained by subtracting the catalyst spec-
trum. Data were collected in a Shimadzu FTIR Prestige-21 spec-
2. Experimental and computational methods
2.1. Catalyst preparation and characterization
Magnesium oxide catalyst was prepared by hydration with dis-
tilled water of commercial MgO (Carlo Erba 99%; 27 m2/g). Details
are given elsewhere [9]. The resulting Mg(OH)2 was decomposed in
Scheme 1. Monoglyceride (MG) synthesis by glycerolysis of FAME and consecutive reactions to diglycerides (DG) and triglyceride (TG).

134
P.G. Belelli et al. / Journal of Catalysis 323 (2015) 132–144
Scheme 2. Formation of surface glyceroxide species from glycerol and FAME surface activation on MgO.
trometer using an inverted T-shaped cell fitted with CaF2 windows
that contains the sample pellet. The absorbance scale was normal-
ized to 20 mg.
following equations (nj, mol of product j; MG, monoglycerides
(both isomers); DG, diglycerides (both isomers); TG, triglyceride):
nMG þ 2nDG þ 3nTG
XFAME ð%Þ ¼
SMG ð%Þ ¼
ꢂ 100
nMG þ 2nDG þ 3nTG þ nFAME
2.2. Catalytic tests
nMG
ꢂ 100
nMG þ 2nDG þ 3nTG
The glycerolysis of a fatty acid methyl ester (methyl oleate,
Fluka, >60.0%, with 86% total C18 + C16 esters as determined by
gas chromatography) with glycerol (Aldrich, 99.0%) was carried
out at 483–503 K in a seven-necked cylindrical glass reactor with
mechanical stirring equipped with a condenser to remove the
methanol generated during reaction. The 4-phase reactor was
operated in a batch regime for the liquid and solid phases at atmo-
spheric pressure. A continuous flow of N2 (35 ml/min) was used to
remove the methanol produced out of the reactor. More details are
given elsewhere [10]. Gly/FAME molar ratios of 2–6 and a catalyst/
FAME ratio (Wcat=n0FAME) of 8 g/mol were used. Catalyst was pre-
treated ex situ at 773 K for 6 h to remove adsorbed water and car-
bon dioxide and kept overnight at 373 K in flowing N2 until use and
then quickly transferred to the reactor without exposing it to air to
start the reaction. The reactor was assumed to be perfectly mixed.
Eleven samples of the reaction mixture were extracted and ana-
lyzed during the 8-h catalytic run.
2nDG
nMG þ 2nDG þ 3nTG
SDG ð%Þ ¼
STG ð%Þ ¼
ꢂ 100
ꢂ 100
3nTG
nMG þ 2nDG þ 3nTG
Yj ð%Þ ¼ XFAME ꢂ Sj ꢂ 100
2.3. Computational details and models
The Vienna Ab-initio Simulation Package (VASP) [12–14] was
used to perform all the periodic calculations reported in this work.
The Kohn–Sham equation was solved using a plane wave basis set.
A good convergence was achieved with a cutoff energy of 350 eV
for the kinetic energy. The projector augmented wave (PAW)
method developed by Blöch [15] was used to describe the core
electron effect in the electronic density of the valence electrons
[16]. PAW is a frozen core method which considers the exact shape
of the valence wave functions instead of pseudo-wave functions.
The exchange and correlation effects were described by the gener-
alized gradient approximation (GGA) using the functional
expressed by Perdew, Burke and Ernzerhof (PBE) [17]. The two-
dimensional Brillouin integrations in the reciprocal space were
performed on a grid of 3 ꢁ 2 ꢁ 1 Monkhorst–Pack special k-points
[18]. The Methfessel–Paxton method with a smearing width of
Reaction products were a- and b-glyceryl monooleates (mono-
glycerides, MG), 1,2- and 1,3-glyceryl dioleates (diglycerides, DG),
and glyceryl trioleate (triglyceride, TG). Samples were silylated to
improve compound detectability. Details on the analytical proce-
dures are given elsewhere [11]. Silylated samples were analyzed
by GC in a SRI 8610C gas chromatograph equipped with a flame
ionization detector (FID), on-column injector port, and a HP-1 Agi-
lent Technologies 15 m ꢁ 0.32 mm ꢁ 0.1
lm capillary column.
Quantification was carried out using authentic standards of glyc-
eryl trioleate (Sigma, 65.0%) and a commercial mixture of glyceryl
mono- and dioleate of known composition (Fluka, 71.5% MG and
25.3% DG).
r
= 0.2 eV was applied, and the reported total energies were then
extrapolated to ? 0 eV [19].
r
Conversion (XFAME, referred to the total content of esters in the
reactant), selectivity (S), and yield (Y) were calculated through the
A slab representing a monoatomic step on the MgO (100) sur-
face was constructed following the MgO (510) plane. The resulting

P.G. Belelli et al. / Journal of Catalysis 323 (2015) 132–144
135
supercell contains 90 atoms distributed in three layers. We have
observed that the addition of a forth layer produces a negligible
variation of the superficial energy. The cell is sufficiently large to
avoid lateral interaction between adsorbed molecules. The vacuum
region between neighboring slabs in the perpendicular direction to
the surface was set to ꢃ15 Å, corresponding to seven ideal bulk
layers. The gas-phase geometries of isolated Gly and FAME were
completely optimized using a 20 ꢁ 20 ꢁ 20 supercell. Periodic cal-
culations indicate that the (100) surfaces of alkaline-earth oxides
present very slight relaxation [20,21]. For this reason, only the
atoms corresponding to the first layer were allowed to relax while
all the adsorbed molecules were completely optimized. The relax-
ation of atoms placed at the edge of stepped MgO produces the
angle increase from 90° (unrelaxed surface) to 101.6° and 104.0°
for O4CMg6CO5C and Mg4CO6CMg5C angles, respectively, in good
agreement with the results of Chizallet et al. [22].
forms on metal–oxygen acid–base sites such as Mg5cAO5c pairs,
which are predominant on the terrace sites of MgO. This species
shows a symmetric OACAO stretching at 1320–1340 cmꢀ1 and
an asymmetric OACAO stretching at 1610–650 cmꢀ1. Bicarbonate
(Bic) species formation involves surface hydroxyl groups and
shows a CAOH bending mode at 1220 cmꢀ1 as well as symmetric
and asymmetric OACAO stretching bands at 1480 cmꢀ1 and
1650 cmꢀ1, respectively [25,26].
The FTIR analysis of CO2 adsorbed at room temperature and
evacuated at 298, 373, 473, and 573 K was carried out on MgO.
Fig. 2 shows the spectra obtained in the 1900–1100 cmꢀ1 carbon-
ate region. The overlapping broad infrared bands confirmed the
presence of different carbonate species, i.e., base sites of different
nature formed on the non-uniform surface of MgO, as explained
above (Fig. 1). Bic species disappeared after evacuation at 373 K
revealing the weak basic features of the surface OH groups of
MgO. In contrast, U.C. and B.C. species remained on the surface
even after evacuation at 573 K; U.C. species were more resistant
to evacuation at high temperatures. Thus, results in Fig. 2 suggest
the formation of weak, medium, and strong surface oxygen species
on MgO, with the following base strength order: low coordination
O2ꢀ anions > oxygen in Mg2+AO2ꢀ pairs > OH groups.
The adsorption energy (Eads
evaluated according to the following total energy difference:
where ‘‘molecule’’ is
)
of Gly or FAME was
Eads = E(molecule–MgO) ꢀ E(MgO) ꢀ E(molecule(g))
,
either Gly or FAME. We also define the reaction energy (Ereac) as
the relative energy with respect to the Gly + FAME + MgO as iso-
lated species. In all the cases, negative values indicate exothermic
processes.
Quantification of the contribution of these oxygen species was
carried out by deconvolution of the TPD profile (not shown here)
in three desorption bands, reaching maximum desorption rates
at about 400, 450, and 550 K. The total base site density (nb) was
Complementary calculations including solvation effects were
performed; the results are presented at the end of Section 3.4.3.
Three FAME molecules were added to mimic the solvent. Two of
these molecules were located above the adsorbed Gly and FAME
species and the third one interacting with the b-GlyOx–FAME
contact.
3.47
the density of weak (nOH), medium (nMgAO), and strong (nO) base
sites, was 0.55
mol/m2, 1.26 mol/m2, and 1.66 mol/m2, respec-
l
mol/m2, and the contribution of each peak, identified as
l
l
l
tively. Thus, the MgO surface contains mainly medium and strong
base sites. In a recent paper [8], we demonstrated that the density
of weak, medium, and strong base sites dramatically depends on
the MgO activation procedure. An increase of the calcination tem-
perature generates smother MgO surfaces that resemble the steps
of the surface depicted in Fig. 1. Thus, surface defects such as low
coordination oxygen atoms located in corners or edges disap-
peared upon treatment at high temperatures causing a decrease
of nO. As a consequence, the total base site density and the overall
basicity of a MgO sample calcined at high temperatures (873 K) are
lower than those of a sample treated at 673 K.
3. Results and discussion
3.1. Characterization of MgO
The hydration of commercial MgO and subsequent decomposi-
tion of the resulting Mg(OH)2 followed by stabilization at 773 K
gave rise to a MgO catalyst with an increased BET surface area
(SA) of 189 m2/g. This is attributed to the development of a porous
structure caused by the gaseous evolution during Mg(OH)2 decom-
position [23]. Also, the final MgO catalyst presents a pore volume
of 0.38 ml/g and a mean pore diameter of about 80 Å. The X-ray
diffractogram of the calcined material (not shown here) confirmed
the presence of only a MgO periclase phase (ASTM 4-0829).
The surface base properties of MgO were investigated by com-
bination of TPD and IR measurements the CO2 preadsorbed at room
3.2. The glycerolysis reaction. MgO active site for monoglyceride
synthesis
MgO was tested in the glycerolysis of FAME in a reactor consist-
ing of four phases: the solid catalyst; the bottom liquid layer
formed by the Gly phase; the top liquid fatty phase containing
FAME and glyceride products; and the gas phase containing meth-
anol that is continuously removed from the reactor. FAME is not
soluble in Gly and Gly is slightly soluble in the FAME phase, and
therefore, the reaction occurs in the upper phase where both reac-
tants coexist. More details of the glycerolysis reaction are given
elsewhere [10].
temperature. The total base site number, Nb
by integration of TPD profile (not shown) was 655
(l
mol/g), calculated
mol/g.
l
On a perfect MgO (100) surface, Mg2+ and O2ꢀ are five-coordi-
nated ions (Mg5c and O5c), but on the surface of the high-surface
area MgO catalyst used here, both ions are also present with coor-
dination numbers (L) lower than 5 depending on the location in
corners or edges, so that L is 5, 4 or 3 for ions in terrace, edge or
corner sites, respectively, as sketched in Fig. 1.
Fig. 3 shows a catalytic experiment carried out with MgO at typ-
ical reaction conditions (493 K). Results presented in Fig. 3 are the
FAME conversion (XFAME) and yield of the different glycerides (Yj,
In a previous work [24], we used FTIR of CO2 preadsorbed at
room temperature to study the chemical nature of surface oxygen
species on similar high-surface area MgO catalyst. By this tech-
nique, several CO2 adsorbed species formed on different surface
oxygen-containing species or on oxygen anions with different
coordination numbers could be detected. We identified at least
three different CO2 adsorption species: unidentate and bidentate
carbonates, and bicarbonate. Unidentate carbonate (U.C.) forma-
tion requires coordinatively unsaturated oxygen anions, such as
those present in corners or edges (O3c or O4c) and exhibits a sym-
metric OACAO stretching at 1360–1400 cmꢀ1 and an asymmetric
OACAO stretching at 1510–1560 cmꢀ1. Bidentate carbonate (B.C.)
j =
rates, reaching ꢃ95% conversion at the end of the 8 h-run. The pre-
dominant formation of monoglycerides can be observed, with
a-MG, b-MG, 1,2-DG, and 1,3-DG). MgO converts FAME at high
a-
MG being the main isomer (Y -MG = 53%) over b-MG. Diglycerides
a
appeared mainly as 1,3-DG and to a lesser extent as 1,2-DG. No
TG formation was detected at any reaction condition on MgO.
The non-zero initial slope (slope at t = 0) of the Yb-MG vs. time
curve (open triangles in Fig. 3) suggests that b-MG is a primary
product of the glycerolysis reaction formed catalytically and
directly from FAME and not consecutively from a-MG. On the con-

136
P.G. Belelli et al. / Journal of Catalysis 323 (2015) 132–144
Fig. 1. Scheme of a stepped MgO (100) surface with OLc and MgLc ions in different positions (L: coordination number). Terrace sites: O5c, Mg5c; edge sites: O4c, Mg4c; corner
sites: O3c, Mg3c
.
100
80
60
40
20
0
80
60
40
20
0
B.C.
U.C.
XFAME
0.2
Bic
α-MG
1,3-DG
298K
1,2-DG
β-MG
373K
0
1
2
3
4
5
6
7
8
Time (h)
473K
573K
Fig. 3. FAME conversion and glyceride yields [MgO calcined at 773 K, Gly/
FAME = 6.0, T = 493 K].
1800
1600
1400
1200
MG formation rate and nO presented in Fig. 4 suggests that the
reaction rate-determining step involves the participation of strong
base sites. Thus, monoglyceride synthesis by glycerolysis of methyl
oleate is mainly promoted by coordinatively unsaturated oxygen
anions present in corners or edges of the non-uniform surface of
MgO.
Wave number (cm-1)
Fig. 2. FTIR spectra of CO2 adsorbed at room temperature on MgO and desorbed at
increasing evacuation temperatures [MgO calcined at 773 K; B.C.: bidentate
carbonate; U.C.: unidentate carbonate; Bic: bicarbonate].
trary, the zero initial slope of the Y1,2-DG and Y1,3-DG curves indi-
3.3. Effect of the experimental conditions on the selectivity toward
glycerides
cates consecutive formation from
a- and b-MG as postulated in
Scheme 1. However, isomer interconversion at higher reaction
times cannot be ruled out, neither the existence of disproportion-
ation reactions [10].
In previous papers [8,23], we investigated the chemical nature
of the MgO base sites involved in the glycerolysis reaction. We
found a decline of the catalytic activity measured on MgO samples
calcined at increasing temperatures, following a trend similar to
that observed for the strong base site density (nO), as explained
in Section 3.1. In fact, the linear correlation between the initial
We have studied in the past that diffusional limitations can be
neglected at the experimental conditions used in the present work
(small particle size and high stirring rate). Besides, as most transe-
sterification reactions, the glycerolysis of Gly with methyl oleate
toward glyceryl monooleate and methanol, Scheme 1, is thermo-
neutral [10]. Thus, in a kinetically controlled reaction system, a
change in the reaction temperature would affect mainly the overall
kinetic rate constant, as predicted by the Arrhenius equation, but

P.G. Belelli et al. / Journal of Catalysis 323 (2015) 132–144
137
26 and 19 kcal/mol for MG and DG formation, respectively, on
MgO catalysts with similar basic properties as the one used here
[10]. The effect of temperature is also noticeable on the distribu-
tion of DG, showing a higher contribution of the 1,3-DG isomer
at lower temperatures.
0.150
0.125
0.100
0.075
0.050
According to Scheme 1, the stoichiometry of the MG synthesis
requires a Gly/FAME molar ratio of 1:1. However, a glycerol excess
is often used to drive the reaction forward and to reach high FAME
conversions. In fact, we have shown that using a Gly/FAME = 1 at
493 K, the MGs formed in the early stages of the reaction are read-
ily converted to DGs [31]. This is explained by total Gly consump-
tion in the reaction zone (the FAME phase) which favored the
consecutive reaction of MG with FAME over the glycerolysis of
FAME with Gly, as sketched in Scheme 1. Furthermore, the Gly sol-
ubility in the FAME phase measured at 493 K increased from
0.86 wt.% at Gly/FAME = 1 to 1.15 wt.% at Gly/FAME = 4.5. Thus,
Gly/FAME ratios >1 are expected to favor the selective formation
of MG. However, the effect of the glycerol excess on the catalytic
performance can be hardly predicted because glycerol is more
hydrophilic than FAME and might adsorb strongly on the MgO sur-
face, thereby blocking the active sites.
Here, we are reporting the effect of varying the Gly/FAME ratio
at 493 K. The 8 h-tests were carried out at a constant Wcat=nF0AME
ratio and changing the molar amount of Gly in order to obtain
Gly/FAME molar ratios in the range of 2–6. The XFAME at the end
of the run increased with Gly/FAME so that values of 78%, 85%,
90%, and 93% were measured for Gly/FAME ratios of 2, 3, 4.5, and
6, respectively. These results confirm that high Gly/FAME ratios
increase the availability of Gly to react with FAME in the reaction
zone.
0.5
1.0
1.5
2.0
2.5
3.0
nO, density of strong base sites (μmol/m2)
Fig. 4. Initial MG formation rate as a function of the density of strong base sites
[MgO calcined at 673, 773, and 873 K, Gly/FAME = 4.5, T = 493 K].
also the Gly solubility in the fatty phase. Therefore, the reaction
will proceed faster as Gly solubility in the fatty phase increases,
which occurs at increasing temperatures [27,28]. Measurements
of the Gly solubility in the FAME phase showed indeed a Gly con-
centration of just 1.15 wt.% at 493 K which slightly increases to
1.65 wt.% at 523 K. Therefore, we anticipate a positive effect of
the reaction temperature increase on the reaction rate in the FAME
phase. However, the effect on the selectivity is less evident and for-
mation of other competing products is likely to occur at high tem-
peratures, such as the undesirable formation of polyglycerols
above 533 K [29].
At the typical reaction conditions of Fig. 3,
a-MG is the main
The selectivity to glycerides as a function of the Gly/FAME ratio
at 35% and 50% FAME conversion is shown in Fig. 6a and b, respec-
tively. Results suggest a negligible effect of the reactant ratio on
product but in order to investigate the effect of the reaction tem-
perature on the selectivity, catalytic tests were performed on
MgO at 483, 493, and 503 K using a Gly/FAME ratio of 4.5. As
expected, FAME conversion increased with the reaction tempera-
ture reaching 80% at 483 K, 90% at 493 K, and 96% at 503 K after
8 h of reaction. Fig. 5 shows the product distribution in terms of
selectivity as a function of the reaction temperature at two XFAME
the glyceride distribution, with a-MG being always the main prod-
uct. Similarly, a SMG/SDG ratio of ꢃ13 was rather independent of the
Gly/FAME ratio at 35% conversion, Fig. 6a. However, at 50% conver-
sion, the SMG/SDG ratio dropped to a constant value of ꢃ6 as a con-
sequence of the increasing selectivity to DG at higher conversions.
In conclusion, results of Figs. 5 and 6 indicate that the reaction
temperature has a more marked influence on the selectivity than
the reactant ratio. Furthermore, a-MG is always the main reaction
product in the wide range of experimental conditions, reaching
selectivities of up to 85%.
levels. MG and in particular the
483–503 K. The selectivity to the
a
a
-MG isomer predominate at
-MG isomer is enhanced by
increasing the reaction temperature at the expense of diglycerides,
whereas selectivity to b-MG remains unchanged. Thus, the
increase of the reaction temperature from 483 to 503 K remarkably
affects the SMG/SDG ratio. For instance, at XFAME = 35%, Fig. 5a, the
ratio increases from 4.5 to 17.9. Similar results are observed at a
conversion level of 50%, Fig. 5b. The enhancement of the SMG/SDG
ratio upon a temperature increase is consistent with a higher acti-
vation energy of the MG synthesis step compared to that of DG,
Scheme 1 [30]. In fact, we have previously calculated values of
3.4. Theoretical studies of the kinetics of FAME glycerolysis on MgO
As explained above, the linear correlation between the catalytic
activity and nO of Fig. 4 clearly suggests that on MgO, the glycerol-
90
90
80
70
60
25
20
15
10
5
25
(a)
(b)
α-MG
α-MG
80
20
70
60
15
10
5
20
10
0
20
10
0
1,3-DG
1,2-DG
β-MG
β-MG
1,3-DG
1,2-DG
485
0
0
480
490
495
500
505
480
485
490
495
500
505
Temperature (K)
Temperature (K)
Fig. 5. Effect of the reaction temperature on glyceride selectivity and on SMG/SDG selectivity ratio at FAME isoconversion. (a) XFAME = 35%; (b) XFAME = 50% [Gly/FAME = 4.5].

138
P.G. Belelli et al. / Journal of Catalysis 323 (2015) 132–144
20
15
10
5
90
80
70
90
80
70
20
15
10
5
(a)
(b)
α-MG
α-MG
1,3-DG
β-MG
10
0
10
0
β-MG
1,2-DG
4
1,3-DG
4
1,2-DG
0
0
2
3
5
6
2
3
5
6
Gly/FAME ratio
Gly/FAME ratio
Fig. 6. Effect of the Gly/FAME ratio on glyceride selectivity and on SMG/SDG selectivity ratio at different FAME isoconversion. (a) XFAME = 35%; (b) XFAME = 50% [T = 493 K].
ysis reaction mostly occurs on strong base sites such as low coor-
dination O2ꢀ anions. The role of these surface species is expected to
be the OAH bond dissociation of the hydroxyl groups present in
the Gly molecule with formation of a surface glyceroxide and a
sented as containing four different adsorption sites (Fig. 1): the ter-
race site with Mg5cAO5c pairs (L = 5) that corresponds to medium
strength base sites; the edge and the O-apical corner sites model-
ing the strongly basic O4c (L = 4) and O3c (L = 3) sites, respectively,
and a Mg-apical corner (Mg3c; L = 3) that models a Lewis acid site.
In agreement with the catalytic results, the molecular modeling of
Gly adsorption on those four different cluster sites predicted that
dissociative chemisorption of the OAH bond occurs only on strong
base sites (edge sites) whereas non-dissociative adsorption takes
place on medium strength base sites such as those of terrace sites.
The calculations also showed that none of the primary OH bonds
dissociates in a barrierless fashion on either the MgO terrace or
edge sites and that the strong Gly-edge site interaction caused
the dissociative adsorption at the secondary OH group. Thus, due
to OAH bond breaking, two new surface species were formed: a
hydroxyl group between the abstracted H and a surface oxygen,
proton, as depicted in Scheme 2, for primary (a-glyceroxide) and
secondary (b-glyceroxide) hydroxyls. Gly contains three hydroxyl
groups, and therefore, the Gly dissociative adsorption can take
place in theory, with the OAH bond breaking at one, two, or the
three OH groups.
On the other hand, as in many base-catalyzed reactions, surface
Lewis acid sites, i.e., the Mg2+ cations, participate in the stabiliza-
tion of negatively charged reaction intermediates (the glyceroxide
anion). Another possible role of the Mg2+ cations is the activation of
the FAME molecule and polarization of its C@O bond, as sketched
in Scheme 2, so that to facilitate the attack of the glyceroxide anion
to the positively charged carbonyl carbon of FAME.
The synthesis of MG and DG requires the formation of a new
CAO bond between FAME and Gly after cleavage of the OAH bonds
of Gly. Figs. 3, 5 and 6 show that the main product of the glycerol-
Os, and a b-glyceroxide stabilized on a surface cation, Mgs
(Scheme 2). Consequently, these previous calculations do not pro-
vide an answer for the experimental results showing the preferen-
ysis of methyl oleate was always the
a
-glyceryl monooleate (
a
-
tial formation of a-MG.
MG), followed by the 1,3 isomer of glyceryl dioleate (1,3-DG) in a
wide range of experimental conditions. Clearly, the primary OH
groups of Gly were involved in the formation of those products
by the reaction sequence of Scheme 1. Time ago, we studied the
reduction of ketones by hydrogen transfer reactions on MgO using
alcohols as a hydrogen source and found that aliphatic secondary
alcohols were better hydrogen donors than the primary ones
[32]. Similar findings were reported by other authors [33]. An
explanation for this outcome is that secondary alcohols have better
reducing capacities than primary alcohols because the correspond-
ing dialkyl ketones present higher reduction potentials than alde-
hydes [34,35]. Based on that, now, we would have expected the
secondary OAH bond of Gly to be preferentially activated giving
b-MG, and probably 1,2-DG, as the main products.
Although those previous calculations with cluster models
allowed us to evaluate comparatively the Gly and FAME adsorp-
tions, they present limitations mainly due to the large number of
atoms involved. Besides, these models show the so-called ‘‘edge
effects’’ because the surfaces are modeled by a finite number of
atoms. Extended surfaces are better represented by periodic calcu-
lations. That is why in the present work, we have widen our studies
using the VASP package to predict the energetic profile of the glyc-
erolysis reaction and to calculate kinetic parameters in order to
elucidate the predominant formation of a-MG.
3.4.1. Adsorption of Gly and FAME on MgO
Fig. 7 shows different views of the slab used to model the
stepped MgO surface.
The preferential formation of a-MG and 1,3-DG compared to b-
MG and 1,2-DG is possibly a consequence of the higher accessibil-
ity and number of primary OH groups in the Gly molecule or may
be due to steric hindrance for CAO coupling at the secondary OH
group. Consequently, the search for a suitable and supported
explanation for this behavior is the motivation for the theoretical
calculations of the following sections.
To elucidate the selectivity issue, we carried out calculations
within the density functional theory (DFT) formalism using the
Gaussian-03 package [8] and the cluster model approximation.
The molecular modeling of the Gly and FAME adsorptions on a
MgO (100) surface was performed using four different clusters.
Thus, the surface of the high-surface area MgO catalyst was repre-
Firstly, Gly and FAME adsorption processes were modeled inde-
pendently. For the Gly adsorption modeling, different geometrical
modes were tested and five geometries with diverse interactions
between Gly and the MgO surface were obtained. Optimized geo-
metrical structures for Gly adsorption on the stepped MgO surface
are shown in Fig. 8, and the results are summarized in Table 1. The
different configurations are labeled as nOH(m), where n and m rep-
resent the amount and position of the OH groups interacting with
the surface, respectively. Thus, m is 1 or 3 for primary hydroxyls
and 2 for the secondary one. In order to compare with the distances
reported in Table 1, the geometry for the isolated Gly molecule was
also calculated; the obtained intramolecular distances resulted to

P.G. Belelli et al. / Journal of Catalysis 323 (2015) 132–144
139
relaxed
frozen
Fig. 7. Top view (top) and side view (bottom) of the slab used as model for the
stepped MgO surface. Red and green balls correspond to and Mg atoms,
O
respectively. (For the interpretation of the references to color in this figure legend,
the reader is referred to the Web version of this article.)
be the following: d(CAC) ꢄ 1.52 Å, d(CAO) ꢄ 1.44 Å, d(CAH) ꢄ 0.98 Å,
and d(OAH) ꢄ 1.09 Å.
In all the cases, Gly interacts preferentially with Mgs and Os ions
at the edge of the stepped MgO surface. Dissociative adsorption
modes present a stronger interaction with the surface than non-
dissociative adsorption modes. In agreement with previous results
obtained using the cluster model approximation [8], the present
calculations predict no spontaneous dissociative adsorption modes
involving the primary OAH groups. The weakest interaction with
the surface corresponds to the non-dissociative adsorption indi-
cated as 3OHa (Fig. 8d). In this configuration, both primary hydro-
2ꢀ
xyl groups form hydrogen bonds with the strongly basic O sites,
4c
2ꢀ
and the secondary one interacts with a terrace O ion with a
5c
longer hydrogen bond. In the non-dissociative adsorption
2OH(2,3) (Fig. 8c), the secondary OH group has a short hydrogen
bond (1.43 Å) with the OAH distance longer than in free glycerol
(1.09 Å). In this case, its Eads indicates a stronger interaction than
3OHa.
In the other three geometries, Gly interacts dissociatively by
breaking the OAH bond of the secondary hydroxyl group. As a
result, a surface OH at the edge is formed together with the b-glyc-
2+
eroxide species (b-GlyOx) which strongly interacts with Mg
.
4c
Besides, the adsorption becomes more stable as the number of pri-
mary hydroxyl groups interacting with the surface increases
through the formation of hydrogen bonds. Indeed, the 1OH(2)
geometry does not have hydrogen bonds (Fig. 8a), 2OH(1,2) has
only one (Fig. 8b), and in the most stable geometry, 3OHb, both pri-
mary OH groups are linked with the surface (Fig. 8e). Besides, in all
the cases, an additional hydrogen bond is formed between the O
and H atoms formerly belonging to the secondary hydroxyl. This
interaction contributes to the high stability of the glyceroxide spe-
cies. As it was previously reported [8], the OAH bond dissociation
is a required condition in the reaction mechanism and it is more
Fig. 8. Optimized geometrical structures of Gly adsorbed on stepped MgO. (a)
dissociative adsorption, mode 1OH(2); (b) dissociative adsorption, mode 2OH(1,2);
(c) non-dissociative adsorption, mode 2OH(2,3); (d) non-dissociative adsorption,
mode 3OHa; (e) dissociative adsorption, mode 3OHb. Red, green, cyan, and yellow
balls are O, Mg, H, and
C atoms, respectively. (For the interpretation of the
references to color in this figure legend, the reader is referred to the Web version of
this article.)
probably to take place on low coordination O2ꢀ sites, in particular
nitude of Eads increases, Table 1. This trend indicates that Gly can
easily dissociate on steps of MgO to form the b-GlyOx species. In
the past, the interaction of Gly and related molecules was theoret-
ically studied on monoatomic steps of CaO. Contrary to our results
2ꢀ
at edge O sites.
4c
From these results, we can observe that as the d(OAH) increases
(and at the same time the d(HAOs) and d(OAMgs) decrease), the mag-

140
P.G. Belelli et al. / Journal of Catalysis 323 (2015) 132–144
Table 1
Adsorption energy (Eads) and bond distances (d) found to the adsorption of Gly on the stepped MgO surface.
Entry
Mode, nOH(m)
n
m
Eads (eV)
d(HAOs) (Å)
d(OAMgs) (Å)
d(OAH) (Å)
1 (Fig. 8a)
2 (Fig. 8b)
3 (Fig. 8c)
4 (Fig. 8d)
5 (Fig. 8e)
1OH(2)
2OH(1,2)
2OH(2,3)
3OHa
1
2
2
3
3
2
1, 2
2, 3
1, 2, 3
1, 2, 3
ꢀ1.57
ꢀ1.59
ꢀ1.33
ꢀ0.53
ꢀ1.82
1.02
1.03
1.43
1.62
1.00
2.01
1.91
2.14
3.09
1.92
1.54
1.56
1.09
0.99
1.84
3OHb
on stepped MgO, these authors observed that in the most preferred
situations, the complete deprotonation of Gly takes place [36,37].
Similar to the calculations with the Gly molecule, the molecular
modeling of isolated and adsorbed FAME was carried out. The
FAME molecule used in the catalytic experiments was methyl ole-
ate, the methyl ester derived from oleic acid in which the long tail
contains eighteen carbon atoms and one unsaturation (C18:1).
However, for modeling purposes, a shorter saturated methyl ester
containing three carbon atoms in the acyl group was used (C3:0).
The calculated intramolecular distances of FAME as an isolated
molecule are: d
¼ 1:22 Å, d
¼ 1:43 Å, d
¼ 1:43 Å),
ðC@OÞ
ðOACH3
ꢄ 1:10 Å.
Þ
ðCAOCH3Þ
d
ðCACÞ
ꢄ 1:52 Å, and d
ðCAHÞ
Optimized structures of FAME adsorbed on a stepped MgO sur-
face are shown in Fig. 9, and the results are summarized in Table 2.
From different initial geometries, four adsorption modes were
obtained: two with the FAME molecule adsorbed perpendicular
to the step direction, and other two situations adsorbed along
the edge. In all the situations, FAME adsorbs non-dissociatively.
In cases with FAME adsorbed perpendicular to the edge, the inter-
actions through the C@O group (labeled as FAMEC@O, Fig. 9a) and
through the O of the methoxy fragment (labeled as FAMEO,
Fig. 9b) were found. On the other hand, when FAME was adsorbed
along the edge, in one of the situations, the FAME molecule is ori-
ented ‘‘standing’’ over the step (with the OAC@O plane perpendic-
ular to the (100) surface, indicated as FAMEstand, Fig. 9c); in the
other configuration, the molecule is tilted onto the terrace
(FAMEtilt, Fig. 9d). Among all the situations, the most stable adsorp-
tion mode resulted to be FAMEC@O, in which the MgsAO distance
presents the lowest value (d
in the FAME internal geometry with respect to the free molecule.
¼ 2:08 Å), with slight changes
ðOC@OAMgsÞ
By examining the net atomic charges, we have observed that when
2+
FAME in this configuration interacts with the step Mg site, polar-
4c
ization of the C@O group (as depicted in Scheme 2) does not pro-
ceed to a high extent in comparison with isolated FAME.
By comparing the Eads values of Tables 1 and 2, we conclude
that, as in our previous work [8], Gly is more strongly adsorbed
than FAME.
3.4.2. FAME adsorption on MgO after b-glyceroxide formation
Once the Gly and FAME adsorptions were independently stud-
ied, we modeled both species adsorbed side by side on the step
in order to study the glycerolysis reaction. For this reason, we
now consider the consecutive adsorption of FAME on the stepped
MgO surface, with Gly previously adsorbed as the highly stable
b-GlyOx species (3OHb geometry, Fig. 8e and Table 1, entry 5).
Two different types of arrays were considered depending on the
desired product:
a-MG or b-MG. We found two stable situations
in which either the primary or the secondary OH groups of b-GlyOx
interact with the carbonyl carbon of FAME. The optimized struc-
tures are schematized in Figs. 10a and 11a, respectively. For
a-
MG formation, the FAME molecule was oriented perpendicular to
the stepped surface, such as the FAMEC@O geometry, with the car-
bonyl group close to one primary OH group of b-GlyOx, Fig. 10a;
while for the b-MG formation, FAME was oriented lying and paral-
lel to the MgO step, such as the FAMEtilt geometry, with the C@O
group near to the secondary oxygen of b-GlyOx, Fig. 11a. In the first
Fig. 9. Optimized structures of FAME adsorbed on stepped MgO. (a) Mode
FAMEC@O; (b) mode FAMEO; (c) mode FAMEstand; and (d) mode FAMEtilt
.

P.G. Belelli et al. / Journal of Catalysis 323 (2015) 132–144
141
Table 2
ꢀ0.59 and ꢀ0.28 eV, respectively. It is interesting to note that
the FAME adsorption over MgO containing the b-GlyOx species is
in the first case 0.10 eV more stable than onto pure stepped MgO
(adsorbed as an isolated molecule at the FAMEC@O mode, Table 2,
entry 1). This behavior indicates that the pair b-GlyOx/FAMEC@O
Adsorption energy (Eads) and bond distances (d) found to the adsorption of FAME on
the stepped MgO surface.
Entry
Mode
Eads (eV)
d
(Å)
d
(Å)
ðC@OÞ
ðOC@OAMgsÞ
1 (Fig. 9a)
2 (Fig. 9b)
3 (Fig. 9c)
4 (Fig. 9d)
FAMEC@O
FAMEO
FAMEstand
FAMEtilt
ꢀ0.49
ꢀ0.28
ꢀ0.29
ꢀ0.41
2.08
4.52
2.11
2.15
1.23
1.22
1.23
1.24
formation which ultimately yields to
favored.
a-MG is energetically
3.4.3. Reaction mechanisms of FAME glycerolysis
The above-mentioned A and Ab geometries were used as start-
ing points to study the reaction paths to form the a-MG and b-MG
products, respectively (Figs. 10 and 11). The reaction profiles for
both reactions are shown in Fig. 12. In this scheme, the above-
defined reaction energy (Ereac) is expressed relative to the energy
of the Gly, FAME, and MgO as isolated species.
a
situation, an optimized distance of 2.98 Å was obtained between
the carbon atom of the C@O group of FAME and the oxygen atom
of primary OH of b-GlyOx species. In the second array, owing
mainly to steric effects, the distance between the carbon atom of
the C@O group of FAME and the oxygen atom of the secondary
OH of b-GlyOx is much longer (4.15 Å). The optimized paired b-
GlyOx/FAMEC@O and b-GlyOx/FAMEtilt structures were labeled as
We first studied the complete reaction sequence for the
a-MG
formation. Optimized geometries of all the involved species are
presented in Fig. 10. The reaction occurs through two successive
A
and Ab geometries, respectively.
a
The A geometry (Fig. 10a) which finally leads to
a
-MG is calcu-
a
steps: the formation of a tetrahedral intermediate (TI ), and the
a
lated to be 0.31 eV more stable than the one that leads to b-MG (Ab
geometry, Fig. 11a). Taking into account that the main difference
between both geometries is the FAME orientation, we calculated
the energy of the FAME adsorption coming from the gas phase,
on the stepped MgO surface containing preadsorbed Gly (3OHb
final desorption of the products,
a-MG and CH3OH. A tetrahedral
intermediate species was also observed by de Lima et al. [38],
when the homogeneous transesterification reaction between Gly
and methoxide and ethoxide anions was theoretically evaluated.
The formation of TI takes place by the dissociation of a primary
a
geometry). For FAME in the A and Ab geometries, the values are
a
OH group of the adsorbed b-GlyOx and the creation of a new
(b)
(a)
Aα
TS1α
β-GlyOx
2.98
FAME
2.13
1.34
1.26
1.69
TS2α
TIα
(c)
(d)
1.46
1.52
1.67
1.83
2.23
1.70
1.55
1.61
1.95
Bα
(e)
(f)
Dα
α-MG
1.35
1.98
α-MGOx
CH3O-
3.53
CH3OH
Fig. 10. Optimized structures of the species involved in the FAME glycerolysis reaction to form
a-MG. (a) b-GlyOx–FAME coadsorption; (b) transition state 1 (TS1 ); (c)
a
tetrahedral intermediate (TI ); (d) transition state 2 (TS2 ); (e) a-MG and CH3OH. Bond distances in Å.
a
-MGOx and CH3Oꢀ; and (f) gas phase
a
a

142
P.G. Belelli et al. / Journal of Catalysis 323 (2015) 132–144
(a)
(c)
(e)
(b)
FAME
TS1β
Aβ
1.36
1.23
4.15
1.28
1.80
2.09
2.00
2.20
β-GlyOx
1.03
1.54
1.83
(d)
TIβ
TS2β
1.41
1.32
2.53
1.51
2.02
2.04
1.34
1.15
2.15
1.13
1.04
1.54
2.16
1.33
1.81
1.74
(f)
β-MG
Bβ
Dβ
CH3O-
β-MGOx
1.24
1.34
4.52
1.44
1.06
2.22
1.93
2.05
CH3OH
Fig. 11. Optimized structures of the species involved in the FAME glycerolysis reaction to form b-MG. (a) b-GlyOx–FAME coadsorption; (b) transition state 1 (TS1b); (c)
tetrahedral intermediate (TIb); (d) transition state 2 (TS2b); (e) b-MGOx and CH3Oꢀ; and (f) gas phase b-MG and CH3OH. Bond distances in Å.
CAO bond (Fig. 10c). TI is strongly connected with the surface by
a
means of two OAMgs bonds, one by the O atom formerly belonging
to b-GlyOx and other by the C@O group of the FAME fragment. Our
0.5
0.0
Dβ
Dα
results indicate that TI is formed from the initial configuration by
0.15
a
Gly +FAME
(g)
(g)
β-MG +CH OH
(g)
(g)
3
crossing an energetic barrier of 0.43 eV, corresponding to the first
transition state (TS1 , Fig. 10b). The geometry of the activated
-0.33
a
α-MG +CH OH
(g)
(g)
3
complex involves the stretching of the OAH bond belonging to
the primary hydroxyl group, the approach of the carbonylic C atom,
and the reorientation of the hydrocarbon chain. The relative strong
hydrogen bonds between the O of the formerly primary hydroxyl
-0.5
-1.0
-1.5
-2.0
-2.5
-3.0
TS2β
-1.08
0.30
TS2α
TS1β
-1.15
TIβ
-1.38
Cβ
-1.48
of Gly fragment with the resulting surface OH at both TS1 and
a
Cα
-1.56
-1.62
0.96
TI species help stabilize these structures.
a
The second reaction step, indicated as B geometry in Fig. 10e,
TS1
α -1.99
a
0.59
Aβ
-2.11
0.31
comprises the
a-monoglyceroxide species (a-MGOx) formation
TIα-2.21
0.43
-2.42
Bα-2.32
Bβ-2.55
and the releasing of CH3Oꢀ; the latter ends linked to the edge
through a OAMgs bond of 1.99 Å. The finally formed CAO bond
which connects the acyl group with the rest of the ester molecule
results in a distance of 1.35 Å. The transition state associated with
Aα
Reaction coordinate
this step (TS2 , Fig. 10d) requires higher energy (0.59 eV) than the
a
first one (0.43 eV). This reaction step is slightly exothermic
(ꢀ0.11 eV). The activated complex geometry involves the separa-
Fig. 12. Reaction energy profile for FAME glycerolysis toward
formation.
a-MG and b-MG

P.G. Belelli et al. / Journal of Catalysis 323 (2015) 132–144
143
tion of the methoxide group and its simultaneous approach to an
from A to TI is calculated to be 0.41 eV with explicit solvation,
a a
and 0.43 eV without it. Regarding the geometries, the most
affected distance is the CAO bond between b-GlyOx and FAME
2+
edge Mg cation. To complete the reaction,
a
-MG and methanol
molecules must be formed as adsorbed species from -MGOx
and CH3Oꢀ, respectively. This stage of the reaction involves the for-
mation of two new OAH bonds: one at the -MG molecule (sec-
4c
a
fragments (A geometry, Fig. 10a) which is slightly shortened by
a
a
0.05 Å with solvation. The OAH distance of the primary hydroxyl
of b-GlyOx does not change along the reaction step. The other dis-
tances are altered by 0.03 Å at most. Therefore, the solvent seems
to stabilize all the species in a similar manner in such a way that
the energy differences are affected only slightly.
ondary OH group of the Gly fragment) and the other belonging to
methanol. The required energy on this step is 0.76 eV (Fig. 12, evo-
lution between B and C geometries). The subsequent desorption
a
a
of both molecules needs again of additional energy (Fig. 12, evolu-
tion between C and D geometries). In Fig. 12, the energy of the
a
a
D
geometry (pictured in Fig. 10f) corresponds to the total energy
a
4. Conclusions
of the separated fragments, i.e., a-MG, methanol, and bare MgO.
Let us now consider the formation of b-MG. The corresponding
optimized geometries of all the species are shown in Fig. 11. As it
was previously mentioned, the FAME molecule is far away from
The catalytic results showed that the MgO active site for the
glycerolysis reaction is a strongly basic oxygen anion, i.e., a coord-
inatively unsaturated oxygen located on corners or edges of the
MgO surface. In the catalytic tests, the main monoglyceride isomer
the secondary OH group of b-GlyOx. In this structure (Ab geome-
2+
try), both molecules are sharing the same Mg cation at the
4c
was always the a-glyceryl monooleate (a-MG) whereas the diglyc-
stepped MgO surface (Fig. 11a).
The first step of this reaction consists also in the formation of
the tetrahedral intermediate, TIb, through the activated complex
eride isomer was the 1,3-glyceryl dioleate (1,3-DG).
The glyceride products form by CAO coupling of FAME with
glycerol at one of the glycerol OH groups. The preferential forma-
tion of glycerides containing the ester fragment at positions 1
and/or 3 of the glycerol molecule could not be extensively modified
by changing the reaction temperature or reactant ratio in a wide
range.
TS1b (Fig. 11b). In TS1b, the new formed CAO bond is at 1.80 Å.
2+
Besides, the Mg at the edge is significantly puckered up. Two
4c
hydrogen bonds between both primary OH groups and the oxide
surface are present in this activated complex (with oxygen anions
at edge and terrace); despite of that, the required energy to form
In agreement with the catalytic results, the calculations pre-
TIb (0.96 eV) is more than twice the one needed to form TI
a
sented here predict that
a-MG formation is favored over that of
(0.43 eV). The tetrahedral intermediate TIb is sterically hindered
(Fig. 11c), because the new CAO bond connects two bulky chains
at very short distance (1.51 Å). As is clear from Fig. 11c, this config-
b-MG, in spite of the fact that the glycerol secondary OAH bond
cleavage is energetically favored. These results are explained by
the theoretical modeling by considering that:
uration is highly unstable in comparison with TI (Fig. 10c).
a
The second reaction step, that comprises the b-monoglycerox-
ide (b-MGOx) formation and the release of CH3Oꢀ, geometry Bb
(Fig. 11e) is highly exothermic (ꢀ1.17 eV). In TS2b geometry, the
outgoing methoxide group is stabilized by a strong hydrogen bond
with the surface OH formed during the initial chemisorption
(Fig. 11d). The new CAO bond at TIb is reduced to 1.41 Å, and the
O atom belonging to C@O maintains the link with the surface but
more enlarged (2.16 Å). Due to steric effects, TS2b is highly unsta-
ble and the TIb transforms into b-MGOx and CH3Oꢀ surpassing a
low barrier (0.30 eV). Such a structure was called Bb geometry
(Fig. 11e), in which b-MGOx is bonded to the surface mainly
(i) Gly adsorption occurs preferentially through the dissocia-
tion of its secondary OH group on the unsaturated basic sites
at the edge of a stepped MgO surface. This initial step is bar-
rierless, and a b-glyceroxide species is formed.
(ii) The b-glyceroxide species participates in formation of both,
a
-MG and b-MG.
(iii) The interaction of b-glyceroxide with FAME occurs more
favorably through one of the two remaining primary OH
groups of the b-glyceroxide species because the secondary
O atom is more sterically hindered and hence, less available
to react. For this reason, a primary OH group of glycerol
finally participates in the CAO bond formation giving prefer-
through one primary OH group of the Gly fragment, whereas the
2+
methoxide species results doubly bonded to Mg and OH. Notice
4c
entially
a-MG instead of b-MG.
2+
that the C@O group practically loses its interaction with the Mg
4c
(iv) The tetrahedral intermediate TI leading to
a-MG is rela-
a
located at the edge of the surface. For the final formation of b-
MG and methanol molecules, the energetic requirement is
1.07 eV (evolution between Bb and Cb geometries, Fig. 12), a value
tively easy to form due to its relaxed geometry, which allows
a strong interaction with the surface.
(v) The limiting step of the pathway to b-MG is clearly the for-
mation of the tetrahedral intermediate TIb, owing to large
sterical effects. In fact, once this intermediate disappears,
the energy abruptly falls down.
higher than that for the a-MG/methanol adsorbed pair (0.76 eV). In
addition, the last step, Db geometry (Fig. 11f), that comprises the
release of b-MG and methanol species to the gas phase, demands
the highest energy cost of the whole process.
Finally, we have performed complementary calculations to test
possible FAME solvent effects. Since the reaction takes place in a
dilute solution of Gly in FAME, we have measured the influence
of other FAME molecules in one particular path, namely, the first
Acknowledgments
Authors thank Universidad Nacional del Sur (UNS), Universidad
Nacional del Litoral (UNL), Consejo Nacional de Investigaciones
Científicas y Técnicas (CONICET) and Agencia Nacional de Promo-
ción Científica y Tecnológica (ANPCyT) of Argentina, for the finan-
cial support of this work.
step of reaction to produce
a-MG (Fig. 10a–c). For that, we have
explicitly added three FAME molecules surrounding the b-GlyOx/
FAME pair (A geometry). Taking into account the large size of
a
FAME molecules, three seems to be a sufficient number to evaluate
short-range effects due to the solvent. The results showed a stabil-
ization value of about 0.4 eV in the reaction energies (measured
with respect to the isolated fragments) of all the species involved
References
[1] A. Corma, S. Iborra, A. Velty, Chem. Rev. 107 (2007) 2411–2502.
[2] J.W. Shabaker, G.W. Huber, J.A. Dumesic, J. Catal. 222 (2004) 180–191.
[3] A. Corma, G.W. Huber, L. Sauvanaud, P. O’Connor, J. Catal. 247 (2007) 307–327.
[4] R.S. Karinen, A.O.I. Krause, Appl. Catal. A: Gen. 306 (2006) 128–133.
[5] Y. Zheng, X. Chen, Y. Shen, Chem. Rev. 108 (2008) 5253–5277.
in this step: A , TS1, and TI . However, the difference among the
a
a
species is essentially the same: the reaction energy difference
between A and TI is 0.20 eV with additional FAME molecules
a
a
and 0.21 eV without them (Fig. 12). Besides, the activation energy

144
P.G. Belelli et al. / Journal of Catalysis 323 (2015) 132–144
[6] F. Jerome, Y. Pouilloux, J. Barrault, ChemSusChem 1 (2008) 586–616.
[25] R. Philipp, K. Fujimoto, J. Phys. Chem. 96 (1992) 9035–9038.
[26] C. Morterra, G. Ghiotti, F. Boccuzzi, S. Coluccia, J. Catal. 51 (1978) 299–313.
[27] T. Kimmel, Kinetic Investigation of the Base-catalyzed Glycerolysis of Fatty
Acid Methyl Esters, PhD Thesis, Institut fur Chemie, TU, Berlín, Germany, 2004.
[28] A.E. Andreatta, L.M. Casás, P. Hegel, S.B. Bottini, E.A. Brignole, Ind. Eng. Chem.
Res. 47 (2008) 5157–5164.
[7] A. Corma, S.B.A. Hamid, S. Iborra, A. Velty, J. Catal. 234 (2005) 340–347.
[8] C.A. Ferretti, S. Fuente, R. Ferullo, N. Castellani, C.R. Apesteguía, J.I. Di Cosimo,
Appl. Catal. A: Gen. 413–414 (2012) 322–331.
[9] J.I. Di Cosimo, V.K. Díez, C.R. Apesteguía, Appl. Catal. A: Gen. 137 (1996) 149–
166.
[10] C.A. Ferretti, R.N. Olcese, C.R. Apesteguía, J.I. Di Cosimo, Ing. Eng. Chem. Res. 48
(2009) 10387–10394.
[29] J. Barrault, Y. Pouillox, J. Clacens, C. Vanhove, S. Bancquart, Catal. Today 75
(2002) 177–181.
[11] C.A. Ferretti, C.R. Apesteguía, J.I. Di Cosimo, J. Argentine Chem. Soc. 98 (2011)
16–28.
[30] G.F. Froment, K.B. Bischoff, Chemical Reactor Analysis and Design, second ed.,
John Wiley and Sons, New York, 1990. p. 176.
[12] G. Kresse, J. Hafner, Phys. Rev. B 47 (1993) 558–561.
[13] G. Kresse, J. Hafner, Phys. Rev. B 48 (1993) 13115–13118.
[14] G. Kresse, J. Hafner, Phys. Rev. B 49 (1994) 14251–14268.
[15] P. Blochl, Phys. Rev. B 50 (1994) 17953–17979.
[16] G. Kresse, D. Joubert, Phys. Rev. B 59 (1999) 1758–1775.
[17] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865–3868.
[18] H.J. Monkhorst, J.D. Pack, Phys. Rev. B 13 (1976) 5188–5192.
[19] M. Methfessel, A.T. Paxton, Phys. Rev. B 40 (1989) 3616–3621.
[20] P. Broqvist, H. Grönbeck, I. Panas, Surf. Sci. 554 (2004) 262–271.
[21] N.V. Skorodumova, K. Hermansson, B. Johansson, Phys. Rev. B 72 (2005)
125414.
[31] C.A. Ferretti, A. Soldano, C.R. Apesteguía, J.I. Di Cosimo, Chem. Eng. J. 161
(2010) 346–354.
[32] J.I. Di Cosimo, A. Acosta, C.R. Apesteguía, J. Mol. Catal. A: Chem. 234 (2005)
111–120.
[33] J.K. Bartley, C. Xu, R. Lloyd, D.I. Enache, D.W. Knight, G.J. Hutchings, Appl. Catal.
B: Environ. 128 (2012) 31–38.
[34] H. Adkins, F.W. Cox, J. Am. Chem Soc. 60 (1938) 1151–1159.
[35] C.F. de Graauw, J.A. Peters, H. van Bekkum, J. Huskens, Synthesis (1994) 1007–
1017.
[36] M. Calatayud, A.M. Ruppert, B.M. Weckhuysenfor, Chem. Eur. J. 15 (2009)
10864–10870.
[22] C. Chizallet, G. Costentin, M. Che, F. Delbecq, P. Sautet, J. Phys. Chem. B 110
(2006) 15878–15886.
[23] J.I. Di Cosimo, V.K. Díez, C. Ferretti, C.R. Apesteguía, Catalysis 26 (2014) 1–28.
[24] J.I. Di Cosimo, V.K. Díez, M. Xu, E. Iglesia, C.R. Apesteguía, J. Catal. 178 (1998)
499–510.
[37] M. Calatayud, Catal. Today 152 (2010) 88–92.
[38] E.F. de Lima, J.W.deM. Carneiro, C. Fenollar-Ferrer, S. Miertus, S. Zinoviev, N.C.
Om Tapanes, D.A.G. Aranda, Fuel 89 (2010) 685–690.