Molecular Catalysis 458 (2018) 189–197

Contents lists available at ScienceDirect

Molecular Catalysis
journal homepage: www.elsevier.com/locate/mcat

Kinetic and mechanistic study of triose sugar conversion on Lewis and

Brønsted acid solids
E.A. Pighin, J.I. Di Cosimo, V.K. Díez ∗
Catalysis Science and Engineering Research Group (GICIC), INCAPE, UNL-CONICET, CCT CONICET Santa Fe, Colectora Ruta Nac. 168, km 0, Paraje “El Pozo”,
(3000) Santa Fe, Argentina

a r t i c l e i n f o a b s t r a c t

Article history: The effect of catalyst acid site nature on the reaction kinetics of the liquid-phase conversion of dihy-
Received 10 August 2017 droxyacetone with ethanol was investigated using Brønsted and Lewis acid solids. The reaction proceeds
Received in revised form 11 October 2017 through a complex reaction network involving a sequence of consecutive and parallel reaction steps.
Accepted 21 November 2017
Main final products were ethyl lactate and pyruvic aldehyde diethyl acetal. Different catalysts such as
Available online 1 December 2017
the cesium salt of tungstophosphoric acid (Cs-HPA), Amberlyst resin and alumina-supported tin and zinc
oxides were used. Catalysts were characterized by ICP, X-ray diffraction, N2 physisorption, TPD of NH3
Keywords:
and FTIR of adsorbed pyridine. Cs-HPA contains only Brønsted acid sites and alumina-supported tin and
Dihydroxyacetone
Ethyl lactate
zinc samples show exclusively Lewis acidity. During dihydroxyacetone conversion, the selectivities to
Lewis and Brønsted solid acid catalysts the final products strongly depend on the nature of the surface acid sites present on the solids. Brønsted
Kinetic modeling acid solids favor the synthesis of pyruvic aldehyde diethyl acetal with final yields of up to 93%, whereas
the ethyl lactate is the main product on Lewis acid solids, reaching final yields of about 51%. A kinetic
model based on a pseudohomogeneous mechanism was proposed to interpret the experimental catalytic
data. The comparison of the kinetic rate constants obtained on both kinds of acid solids confirms that on
Brønsted catalysts the route toward pyruvic aldehyde diethyl acetal is clearly favored. On Lewis solids,
the synthesis of ethyl lactate occurs selectively via isomerization of the pyruvic aldehyde hemiacetal
intermediate.
© 2017 Elsevier B.V. All rights reserved.

1. Introduction Alkyl lactates are biodegradable nontoxic compounds with high

boiling points; these properties make them excellent green sol-
There is an increasing worldwide interest in deriving chemicals vents. They are used as plasticizer [4] and in surfactant and in
from biomass. Materials containing cellulose and starch are attrac- herbicide formulations. Also, they are employed in the cosmetic
tive candidates for its transformation in more valuable compounds. industry as skin whitener and in acne treatments [5]. In particu-
These resources can be converted in sugars in biorefineries; sugars lar, EL is used as green solvent, plasticizer, and in pharmaceutical
in turn are molecules that can give rise to a number of chemicals and and cosmetic formulations. On the other hand, PADA is employed as
liquid fuels [1,2]. In fact, sugars resulting from cellulose enzymatic intermediate in the synthesis of folic acid [6] and of an osteoporosis
hydrolysis [1,3], can be converted by fermentation in many plat- inhibitor [7].
form compounds for the production of valuable C3-C6 derivatives. In the last years, many works describe the reactions between
Particularly, glucose can be transformed by retroaldol reactions triose sugars and short chain alcohols using solid acid catalysts.
in short carbon chain polyfunctional compounds containing C O The literature reports the use of zeolites [8–11], mixed oxides [12]
and OH groups. Thus, dihydroxyacetone (DHA), a monosaccharide and substituted mesoporous materials such as Al and Ga/MCM-41
that belongs to the group of trioses, can be obtained from glucose [13] as catalysts. More recently, it has been postulated the use of
and subsequently transformed in more valuable oxygenates such Sn-containing catalysts such as mesoporous materials (Sn-MCM-41
as ethyl lactate (EL) and pyruvic aldehyde diethyl acetal (PADA), [13,14], Sn-SBA-15 [15], Sn-Si mixed oxides [16]), clays like mont-
Scheme 1. morillonite [17] and zeolites (Sn-MFI [18], Sn-BEA [19–23]) which
showed to be highly active and selective toward alkyl lactates.
However, the use of mesoporous and zeolitic materials during the
conversion of DHA presents several draw backs such as the costly
∗ Corresponding author.
and time demanding synthesis procedures. Given this, in a previ-
E-mail address: verodiez@fiq.unl.edu.ar (V.K. Díez).

https://doi.org/10.1016/j.mcat.2017.11.026
2468-8231/© 2017 Elsevier B.V. All rights reserved.
190 E.A. Pighin et al. / Molecular Catalysis 458 (2018) 189–197

Scheme 1. Reaction pathways for the conversion of C3 monosaccharides via dehydration, ethanol nucleophilic addition and esterification reactions on solid catalysts with
Lewis and Brønsted acid properties [25].

ous work [24] we investigated the conversion of DHA and ethanol 2. Experimental
on inexpensive alumina-supported Sn catalysts obtained by incip-
ient wetness impregnation. We showed that on these Lewis acid 2.1. Catalyst synthesis and activation
materials high ethyl lactate yields (≈70%) are obtained at mild con-
ditions. We focused there on the role of the surface Sn species in Alumina-supported tin and zinc oxides were prepared by incipi-
the generation of more and stronger Lewis acid sites than those ent wetness impregnation of commercial ␥-Al2 O3 Cyanamid Ketjen
present on the alumina support. Also, we demonstrated that the CK 300. Aqueous solutions of SnCl4 ·5H2 O (Aldrich, 98%) and
ethyl lactate yield is enhanced as the Sn loading increases, confirm- Zn(NO3 )2 (Aldrich, 98%) were used; samples were coded as SnAl
ing the participation of these sites in rate-limiting steps. Although and ZnAl, respectively. The SnAl and ZnAl precursors were ther-
the DHA conversion reaction has been extensively studied in our mally decomposed overnight (18 h) in flowing air at 573 K. The
and other previous works, very few efforts have been made so far to chemical content of Sn and Zn in both alumina-supported metal
determine the kinetics and the mechanistic features of the different oxides was around 7.0 wt.%.
reaction steps involved in the reaction pathway of Scheme 1. ZnO was obtained by decomposition of commercial
Given this, we have focused on the kinetic study of the DHA (ZnCO3 )2 (Zn(OH)2 )3 (Anedra) in air at 623 K. On the other hand,
conversion reactions on Lewis acid solids (Sn/Al2 O3 catalysts) [25]; SnO2 was prepared by precipitation of a solution of SnCl4 ·5H2 O
we analyzed the effect of Sn loading and reaction temperature on (Aldrich, 98%) with aqueous solution of urea (Anedra, 99%). The
the ethyl lactate yield and on the reaction kinetics and postulated precipitate was then decomposed in air at 573 K overnight in order
a pseudohomogeneous mechanism that gave a suitable descrip- to obtain the corresponding pure tin oxide. More details are given
tion of the complex reaction system (Scheme 1). We found that the elsewhere [26].
kinetic parameter associated to ethyl lactate formation increases The cesium salt of the tungstophosphoric acid
with the number of Lewis acid sites confirming that surface Sn (Cs2.5 H0.5 PW12 O40 ), coded as Cs-HPA, was prepared by slowly
species are responsible for the catalytic activity of these materials. adding an aqueous solution of Cs2 CO3 (Sigma-Aldrich, PA 99.9%)
However, there was still a lack in the literature of kinetic stud- to an aqueous solution of H3 PW12 O40, HPA (Merck, PA). The
ies of DHA conversion investigating comparatively solid catalysts precipitate was heated overnight at 353 K and then calcined in N2
with Brønsted and Lewis acidity. This is why, in this work, we flow at 573 K.
continue our investigations on the reaction kinetics involved in A commercial Amberlyst 35W resin (Rohm and Haas) was used;
the DHA conversion but comparing the catalytic performance of pellets were crushed and sieved to retain particles between 180
solids with Brønsted (cesium salt of tungstophosphoric acid and and 480 ␮m. The resulting powder was treated in a N2 flow at 373 K
Amberlyst 35W resin) and Lewis (Sn/Al2 O3 , and Zn/Al2 O3 ) acid before the catalytic test.
properties. Our goal was to elucidate the effect of the catalyst acid
nature on determining the selectivity toward the final product, EL
or PADA. Furthermore, we performed a kinetic study of the reaction
system by modeling experimental data using a pseudohomoge- 2.2. Catalyst characterization
neous model. We focused on the competitive pathways leading to
PADA and EL presented in Scheme 1 and identified the steps of the BET surface areas (SA) were measured by N2 physisorption at
reaction network promoted by Lewis or Brønsted sites. 77 K using an Autosorb Quantachrome 1-C sorptometer. The cata-
lyst Sn or Zn chemical content was analyzed by inductively coupled
plasma (ICP-OES). The sample structural properties were analyzed
by X-Ray Diffraction (XRD) technique using a Shimadzu XD-D1
 E.A. Pighin et al. / Molecular Catalysis 458 (2018) 189–197 191

instrument with nickel filtered Cu K␣ radiations between 10 and ties were calculated as Sj = Cj /Cj , where Cj is the concentration of
80◦ . product j.
Total acid site numbers (na ) were measured by TPD of NH3 . Sam- The initial DHA conversion rate (rDHA0 , mmol/h gcat ) was calcu-
ples were thermally treated in He at the corresponding calcination lated from the initial slope of the XDHA vs. tW/n0 DHA (corrected time)
temperature and then exposed to a 1.01% NH3 /He flow at 373 K dur- curve, where W is the catalyst weight.
ing 30 min in order to reach surface saturation. Weakly adsorbed
NH3 was removed by flushing with He. The temperature was then 2.4. Kinetic modeling and statistical analysis
increased from 373 K to 973 K for Cs-HPA and Al2 O3 samples, using
a ramp rate of 10 K/min. In the case of SnAl and ZnAl, the tem- A kinetic study was carried out by modeling the catalytic data
perature was increased up to 593 K and then kept constant at this using a first-order pseudohomogeneous model. More details will
temperature for 1 h. NH3 concentration in the reactor effluent was be given in Section 3.3.
monitored using a mass spectrometer (MS) detector. The system of differential equations was solved numerically
Identification of the chemical nature of the acid sites present using the Microsoft Excel Solver software and the Euler method.
on the catalyst surface was carried out by Infrared Spectroscopy This calculation tool uses the Generalized Reduced Gradient (GRG)
(IR) of pyridine adsorbed at room temperature and evacuated at algorithm for optimizing non-linear problems.
increasing temperatures. Experiments were performed with a Shi- The experimental relative molar concentrations of all the
madzu FTIR Prestige-21 spectrophotometer. An inverted T-shaped ∗ ) were compared with the values
species during the reaction (Cjobs
cell containing the sample wafer and fitted with CaF2 windows was ∗
predicted by the model (Cjcalc ). The kinetic parameters (ki ) of each
used. In a typical experiment, the sample wafer was evacuated at
reaction step of Scheme 1 were estimated by minimizing the sum of
573 K and then cooled down to room temperature to take the cat-
the squared errors (SSE) [29] between the experimental and model
alyst spectrum. After that, the sample was exposed to 0.12 kPa of
data as:
pyridine at room temperature and then evacuated consecutively at   2
∗ ∗
298, 373, 423, 473 and 573 K. The resulting spectrum at each evac- SSE = Cjcalc − Cjobs (1)
uation temperature was recorded at room temperature. Spectra
alldata
of the adsorbed species were obtained by subtracting the catalyst
spectrum. In all the spectra, the absorbance scales were normalized samples
to 20-mg wafers. 0
where CDHA is the initial DHA concentration in the reactor expressed
 
in mmol DHA/cm3 . The relative concentration of compound j Cj∗
2.3. Catalytic testing 0 . The coefficient of determination (R2 ) gives
is defined as Cj /CDHA
the fitting quality [29] and was calculated using Eq. (2):
Liquid-phase conversion of dihydroxyacetone, DHA (Aldrich,
n ∗ ∗ ) 2
97%) and ethanol (Merck, 99.8%) was carried out at 353 K and at
i=1
(Cjcalc − C̄jobs
2
an autogenous pressure of 250 KPa in a batch PARR reactor. A solu- R = n 2
(2)
∗
(Cjobs ∗ )
− C̄jobs
tion of DHA in ethanol with DHA/ethanol = 0.023 (molar ratio) was i=1
loaded in the reactor. The catalyst/DHA weight ratio was 43%. Cat- ∗
where C̄jobs is the mean of measured values.
alysts were thermally treated ex-situ in air at the corresponding
calcination temperature to remove adsorbed water. After introduc- The discrimination between models was carried out using the
ing the reactant mixture, the reactor was sealed and flushed with model selection criterion (MSC) [30], according to Eq. (3):
N2 and then the mixture was heated up to the reaction tempera- ⎡  2 ⎤
∗
˙ CJobs ∗
− C̄Jobs
ture under stirring (400 rpm). Then, the catalyst as ground powder ⎢ ⎥ 2p
was added to the reaction mixture to start the reaction. Inter- and MSC = ln ⎣  2 ⎦ − m (3)
∗
˙ CJobs ∗
− CJcalc
intra-particle diffusional limitations were verified to be negligible.
For this purpose, the Mears number [27] was calculated to eval-
uate the extent of the external mass transfer resistance. On the where p is the number of parameters; m is the number of exper-
other hand, the Weisz-Prater criterion [28] for a first order reaction imental observations. The MSC parameter is used to compare
and spherical catalyst particles was used to verify the absence of statistically different models so that the larger the MSC value, the
intra-particle mass transfer limitations. More details are given else- better the fit and the more appropriate the model for interpreting
where [25]. During the experiments, 13 samples of ≈0.5 mL were the data.
extracted from the reactor. Product identification was achieved
using a Thermo Scientific Trace 1300 GC with a Thermo Scien- 3. Results and discussion
tific TR-5MS capillary column coupled to a Thermo Scientific ISQ
QD MS unit. Product analysis and quantification were done using a 3.1. Catalyst characterization
7890A Agilent Technologies GC equipped with a FID detector and
a Carbowax Amine 30 M capillary column. The concentration, Cj , The solid acid catalysts prepared by different methods and
of the main reaction products (pyruvic aldehyde (PA), ethyl lac- commercial Amberlyst resin were characterized; their chemical,
tate (EL), pyruvic aldehyde hemiacetal (PAHA), pyruvic aldehyde structural, textural and acid properties were investigated. Results
diethyl acetal (PADA) and glyceraldehyde diethyl acetal (GLADA)) are summarized in Table 1, where properties of the ␥-Al2 O3 sup-
was calculated by this technique. For quantification purposes, n- port are included as reference. Catalyst metal contents (Zn and Sn)
octanol was used as external standard. were around 7.0 wt.%. Through the different procedures used for
Product yields (Yj , mole of product j/mole of DHA fed) were preparing the catalysts, solids withsurface area (SA) values higher
calculated as Yj = Sj XDHA where XDHA is the dihydroxyacetone con- than 140m2 /g were obtained. The SnAl and ZnAl samples presented
version (XDHA = (n0 DHA − nDHA )/n0 DHA , where n0 DHA are the moles of lower surface areas than the ␥-Al2 O3 support, probably due to a
dihydroxyacetone initially loaded and nDHA are the moles of dihy- partial pore blockage by metal species during the impregnation
droxyacetone at the reaction time t) and Sj is the selectivity of procedure. On the other hand, the substitution of the protons of
product j (Sj , mole of product j/mole of DHA reacted). Selectivi- the bulk HPA structure by Cs+ cations promotes a 16-fold increase
192 E.A. Pighin et al. / Molecular Catalysis 458 (2018) 189–197

Table 1
Chemical, textural and acidic properties of solid catalysts.

Catalyst Surface area, SA (m2 /g) Metal loadinga (wt.%) Surface acidic properties

Acid site number, na b (␮mol/g) Acid site nature d

L/(L + B) (%) Lewis band

position (cm−1 )

␯ 19b ␯ 8a

SnAl 184 6.9 165 100.0 1453 1617

ZnAl 183 6.6 59 99.2 1450 1613
Cs-HPA 143 – 47 0.0 – –
Amberlyst 35W 52 – 5200c nd nd nd
␥-Al2 O3 230 – 24 99.1 1450 1616

B: Brønsted sites, L: Lewis sites; nd: not determined.

a
by ICP.
b
by TPD of NH3 .
c
data provided by the supplier.
d
by FTIR of pyridine desorbed at 423 K.

of SA value (SAHPA = 9 m2 /g), which is in agreement with a previous

report [31]. In addition to the significant higher SA, bulk Cs-HPA
was employed instead HPA because the former is insoluble in the
alcoholic medium of the reaction, in contrast to the known solu-
bility of bulk tungstophosphoric acid in oxygen-containing polar
solvents [32,33].
The structural properties of ZnAl and SnAl catalysts and pure
Al2 O3 , SnO2 and ZnO oxides were analyzed by XRD. The XRD pat-
terns of tin and zinc alumina-supported oxides (see Fig. S1 in
Supplementary Material) exhibit broad signals corresponding to
quasi-amorphous ␥-Al2 O3 . No crystalline tin phases were observed
on SnAl sample thereby indicating that Sn species are well dis-
persed on the alumina surface forming small SnO2 domains not
detectable by XRD. Furthermore, the XRD pattern of the ZnAl sam-
ple shows a barely perceptible ZnO phase indicating a probably
lower dispersion of the ZnO species on alumina. On the other hand,
the structure of Cs-HPA was confirmed by XRD analysis (Fig. S1)
and also the XRD pattern obtained for this sample was consistent
with those reported previously in the literature [34].
The surface acid properties of SnAl, ZnAl and Cs-HPA samples
were investigated by combining TPD of NH3 preadsorbed at 373 K
and FTIR of pyridine adsorbed at room temperature. Characteriza-
tion of Amberlyst 35W resin by these techniques was not possible
due to the low thermal stability of this sample. The NH3 TPD profile
of Cs-HPA (see Fig. S2 in Supplementary Material) showed sev-
eral desorption peaks at temperatures in the range of 450–850 K,
reflecting the presence on this sample of surface acid species
that bind NH3 with different strength. Furthermore, Cs-HPA is the
sample with the strongest acid sites, with NH3 still desorbing at
Fig. 1. FTIR of pyridine adsorbed at room temperature and evacuated at 423 K on
temperatures higher than 815 K. On the other hand, the NH3 TPD Lewis and Brønsted acid catalysts.
profiles of SnAl and ZnAl (Fig. S2) exhibit a broad asymmetric des-
orption peak around 580–600 K.
For the analyzed catalysts, the total amount of NH3 desorbed
from the surface was measured by integration of the TPD curves after evacuation at 423 K are shown in Fig. 1. Cs-HPA showed the
and it was taken as an indication of the total acid site number, typical bands of pyridinium ion formed by protonation of pyridine
na (␮mol/g). The na values are reported in Table 1. Sample SnAl on Brønsted acid sites. These bands are the ring vibrations modes at
exhibits the highest na value. In fact, the addition of tin species to 1636, 1608, 1537 and 1485 cm−1 [35]. On the other hand, only the
the Al2 O3 support produces almost a 7-fold increase of na whereas bands at 1622, 1615 and 1455 cm−1 , assigned to pyridine coordi-
the incorporation of zinc on Al2 O3 generates a much lower increase nated on the oxide cations, were detected in the spectra of the SnAl
of na (2.5 times, Table 1). This result is in line with the lower dis- and ZnAl samples. Thus, SnAl and ZnAl contain Lewis acid centers
persion of the ZnO species compared to those of SnO2 observed by [36]. This was confirmed by calculating the relative contribution of
XRD. Later, the differences in acidic properties between SnAl and the Brønsted (B) and Lewis (L) acid sites, L/(L + B), by integration of
ZnAl catalysts will be further discussed. the bands at ∼1540 and ∼1450 cm−1 , respectively, Table 1. Results
The identification of the chemical nature of the surface acid sites show that Lewis acid centers represent about 99–100% of the total
of SnAl, ZnAl, Al2 O3 and Cs-HPA samples was performed by analyz- acid site number of samples SnAl and ZnAl, whereas Cs-HPA does
ing the FTIR spectra of adsorbed pyridine after admission at 298 K not exhibit Lewis acidity. On the other hand, commercial Amberlyst
and sequential evacuation at increasing temperatures; the spectra 35W, which could not be characterized, is a solid with well-known
Brønsted acidic properties confirmed in a previous work [37].
 E.A. Pighin et al. / Molecular Catalysis 458 (2018) 189–197 193

Moreover, an analysis of the strength of the Lewis acid sites a final YEL value of 70%. Here, the SnAl catalyst with 6.9 wt.% Sn
was performed for the SnAl, ZnAl and Al2 O3 samples based on reached complete DHA conversion at 7 h of reaction at 353 K, giv-
the position of the adsorbed pyridine bands at around 1450 cm−1 ing a moderate final YEL of 51% and minor amounts of the other
and 1620 cm−1 , that correspond to ␯19b and ␯8a vibration modes, products.
respectively, Fig. 1. On SnAl these bands shift to higher frequencies Our catalytic results obtained on Lewis and Brønsted acid solids
compared with ZnAl (Table 1), thereby suggesting stronger acid (Table 2) are in line with previous works reported in the litera-
sites on the former [38]. ture. Pescarmona et al. [8] and Clippel et al. [14] found that zeolites
with high concentration of strong Brønsted sites catalyze the for-
3.2. Effect of the acid site nature on catalytic performance during mation of the diacetal product, in agreement with the findings of
DHA conversion Table 2 for Cs-HPA and Amberlyst 35W. On the other hand, zeo-
lites combining mild Brønsted acidity with Lewis acidity arising
The effect of the chemical nature (Brønsted or Lewis) of the from extra-framework aluminum species, promote the formation
catalyst surface acid sites during the conversion of dihydroxyace- of alkyl lactates with yields between 54 and 59% [8].
tone (DHA) was investigated. The catalytic performance of different Regarding Lewis acid properties, the effect of changing the
Brønsted and Lewis acid solids was evaluated under identical reac- Lewis metal cation of the alumina-supported catalysts was also
tion conditions. These conditions were selected based on previous investigated by preparing and characterizing Sn- and Zn-promoted
experiments and literature where mixed oxides, zeolites and meso- materials. Zeolites and mesoporous materials containing structural
porous materials were used in this reaction [12–14,23,24,25]. Pure Sn species have been widely reported in the literature for convert-
Al2 O3 was also tested for comparative purposes. ing trioses into alkyl lactates [13,14,19]. That was the input to select
In previous work [25], we postulated the reaction pathway for Sn as a Lewis acid cation. However, Sn in the SnAl catalyst reported
triose conversion depicted in Scheme 1. Dihydroxyacetone and here has a different chemical environment compared to zeolites
glyceraldehyde (GLA), (step 1) are easily interconvertible and lead or MCM-41 materials. We also selected Zn as a promoter of Lewis
to similar product distribution [8]. Therefore, they are equally suit- acidity based on the fact that several authors claimed that ZnBEA
able as substrates for the synthesis of ethyl lactate (EL) and pyruvic [39] and zincosilicates are strong Lewis acids [40,41] and therefore
aldehyde diethyl acetal (PADA). Thus, global conversion of trioses they are attractive catalytic materials for the conversion of sugars,
during catalytic experiments, XDHA , involves, in fact, the conver- in particular, for the Lewis acid- or Brønsted/Lewis acid-catalyzed
sion of DHA and GLA. The following step is the triose dehydration isomerization of glucose to fructose [41].
to pyruvic aldehyde, PA (step 2 in Scheme 1). From PA the reac- The two Lewis acid solids investigated here (SnAl and ZnAl)
tion pathway might proceed toward EL by re-arrangement with exhibit values of EL/PADA ratio (Table 2) between 2.2 and 3.0, con-
incorporation of an alcohol molecule (step 3), or toward pyruvic firming that on these samples EL is the main reaction product at
aldehyde diethyl acetal, PAHA (step 4) and PADA (step 5) by addi- low conversions. However, the ZnAl sample was considerably less
tion of an alcohol molecule in each of these reaction steps. Step 6 active to convert DHA than SnAl, showing an initial DHA conver-
is the isomerization of PAHA to EL. sion rate value ≈10 times lower. Similarly, on ZnAl the final XDHA
Table 2 summarizes the catalytic results obtained with all the and YEL values were clearly lower than those of SnAl.
acid solids investigated. The alumina support resulted poorly active The characterization results of the Lewis acid samples (Table 1)
under our reaction conditions in spite of being a Lewis acid solid show an important effect of the Sn promotion on the enhanced
(Fig. 1). Thus, it seems that the Al3+ cations are not acidic enough acidic properties (number and strength of the acid sites) of SnAl
to transform DHA. compared to Al2 O3 . That can be explained considering that the cat-
0
Different activities (rDHA values in Table 1) were obtained alyst Lewis acidity is related to the oxide electronegativity (␹oxide ).
depending on the Brønsted or Lewis acid properties of the different According to Sanderson, the electronegativity of an oxide with the
solid materials. However, as will be discussed later, the nature of formula My Ox , is: ␹oxide = [(␹M )y (␹O )x ]1/(y+x) [42]. Time ago, we cal-
the acidity is not the only factor determining activity. culated ␹oxide for a series of My Ox single oxides and showed an
Main products obtained from triose conversion were the inter- acceptable correlation between ␹oxide and na , in spite of the former
mediate products PA and PAHA, and the terminal products EL and being a bulk property [43]. The ␹oxide calculated for Al2 O3 (2.53)
PADA (Scheme 1). As can be seen in Table 2, on the catalysts with was similar to that of ZnO (2.38) and lower than that of SnO2 (2.85).
0
high rDHA values (SnAl, Cs-HPA and Amberlyst 35W), the final yields Thus, more acidic SnO2 forming highly dispersed surface Sn species
(Yj ) toward EL and PADA, are strongly dependent on the acid nature is responsible for the noticeable increase of the surface acidic prop-
of the active sites. erties of the SnAl sample compared to the support (Table 1). On the
Brønsted solids favored the synthesis of PADA over that of EL; on contrary, surface promotion of Al2 O3 with the less acidic zinc oxide
Cs-HPA and Amberlyst 35W, both with strong Brønsted acid prop- produces only a slight increase of the acid site number (Table 1),
erties, PADA was the major product. In particular, yields of PADA of thereby explaining the poor catalytic performance of this catalyst.
93% were reached at the end of the 7 h run on Amberlyst resin. How- In addition to the lower acidity of ZnO compared to SnO2 , the
ever, at higher conversion levels Cs-HPA resulted less active and marked difference between the catalytic performance of SnAl and
selective than Amberlyst, giving other undesirable products. The ZnAl samples could be attributed to the different way in which Sn
selective formation of PADA on these solids is not a matter of con- and Zn species interact with the support during impregnation and
version level since the EL/PADA ratio calculated at low conversions subsequent calcination procedures. Bulkier ZnO oxide species were
(Table 2) for all the catalysts showed similar trend. generated on the ZnAl sample, as indicated by the XRD data, sug-
The performance of Lewis acid solids during triose conversion gesting a lower affinity of the Zn2+ cations for the alumina surface.
was investigated in detail in previous works [24,25]. We prepared Furthermore, Hayashi and Sasaki [44] reported the
and characterized similar SnAl catalysts and studied the effect of homogeneously-catalyzed conversion of trioses with alcohols.
varying the tin loading (between 1.4–7.6 wt.%), tin precursor (SnCl2 Working with different alcoholic solutions of metal chlorides
or SnCl4 ) and reaction temperature (between 353 and 373 K) on the (MCln ) they found that most of the chlorides, such as ZnCl2 , were
activity and product distribution. We demonstrated that the SnAl inactive for the reaction and only tin chlorides (SnCln ) were able to
catalysts produce selectively EL and that the catalytic performance produce high yields of alkyl lactates. They attributed the distinct
of these materials can be improved by increasing Sn content to performance of the SnCln salts to their ability to strongly coordi-
7.6 wt.% [24] and the reaction temperature to 373 K [25], to reach
194 E.A. Pighin et al. / Molecular Catalysis 458 (2018) 189–197

Table 2
Results obtained during DHA conversion on different acid solid catalysts.

Catalyst Initial DHA conversion EL/PADA selectivity ratioa Catalytic results at t = 7 h

rate, r0 DHA (mmol/hg)
DHA conversion, XDHA (%) Product yields, Yj (%)

PA PAHA EL PADA GLADA

SnAl 30.4 3.0 98.6 15.7 0.2 50.9 11.1 16.0

ZnAl 3.6 2.2 19.8 2.2 2.5 1.6 0.7 2.8
Cs-HPA 26.4 0.2 70.9 3.3 4.6 0.3 41.9 0.2
Amberlyst 35W 54.3 0.1 98.8 0.3 0.4 3.5 93.1 0.2
Al2 O3 2.5 – 11.0 1.3 6.7 3.1 0.0 0.0

Reaction conditions: T = 353 K, DHA/ethanol (molar) = 0.023, Wcat /WDHA = 43 wt.%.

a
at XDHA ≈10–20%.

Fig. 2. DHA conversion (XDHA ) and product yields (Yj ) vs reaction time on Amberlyst 35W (A) and SnAl (B). [T = 353 K, DHA/ethanol (molar) = 0.023, Wcat /WDHA = 43 wt.%].
Points, experimental results; solid lines, model predictions.

nate to PA to form a specific intermediate in which a 1,2-hydride The kinetic modeling of the experimental data obtained dur-
shift occurs to predominantly yield the alkyl lactate. ing DHA conversion on Brønsted and Lewis acid solids was carried
It seems then that whereas Zn-based materials promote the iso- out considering different assumptions for the pseudohomogeneous
merization of sugars, they are not able to promote the complex model. The only products of triose (DHA and GLA) conversion were
reaction pathway involved in the ethyl lactate synthesis. PA, EL, PAHA, PADA, GLADA; unidentified products present in low
concentration in the reaction mixture were grouped as OTHERS.
First order kinetics respect to the DHA and products were con-
3.3. Reaction kinetic modeling and surface mechanisms sidered; taking into account that the concentration of ethanol is
in excess respect to the stoichiometric concentration, the initial
ethanol concentration (CEtOH0 ) remains constant during the typical
Fig. 2 shows the catalytic tests on Amberlyst 35W and SnAl, rep-
resenting typical Brønsted and Lewis acid solids, respectively. In 7-h run and therefore, the reaction order for ethanol was taken as
agreement with the reaction pathways postulated in Scheme 1, the zero.
shape of product yield vs time curves of Fig. 2, shows that PA and Initially, we followed here a strategy consisting of considering all
PAHA are intermediate compounds. On the other hand, the zero ini- the steps of Scheme 1 as reversible reactions, but small values of the
tial slopes of the EL and PADA yield curves and the fact that these kinetic constants for the reverse reactions of steps 2, 3, 5 and 6 were
curves do not exhibit maxima, suggests they are secondary and obtained. Then, we repeated the calculations assuming that these
terminal products of the reaction. According to these results and were irreversible steps whereas steps 4, 7 and 8, were considered
as mentioned in Section 3.2, the conversion of DHA with ethanol as reversible transformations. Finally, in an attempt to elucidate
seems to proceed by an initial dehydration step and further rear- which of the two possible pathways toward EL, either the direct
rangement into PA. PA is then converted into the EL or PAHA via synthesis from PA (step 3) or the isomerization of PAHA (step 6),
ethanol addition; the latter can undergo consecutive esterification was kinetically favored under our reaction conditions, we carried
to PADA. Similar reaction pathways were proposed previously in out two more calculations of the kinetic parameters assuming in
the literature [8,9,14]. one case that k6 = 0 and that k3 = 0 in the other. We compared the
In an attempt to analyze comparatively the performance of our kinetic and statistical parameters in both cases and we found that
Lewis and Brønsted catalysts and to gain insight into the reaction the best results were obtained considering that on both catalysts
mechanism, a kinetic study was carried out by modeling the cat- the synthesis of EL occurs exclusively from PAHA via isomerization
alytic data of Fig. 2 using a pseudohomogeneous model. One of the reaction. Thus, the highest value of the coefficient of determination
purposes of the kinetic modeling was to identify which of the reac- (R2 ) and the lowest value of the sum of the squared errors (SSE)
tion steps of the reaction network (Scheme 1) is promoted by Lewis were obtained when the kinetic constant of the direct synthesis of
or Brønsted acid sites. EL from PA was considered equal to zero (k3 = 0).
 E.A. Pighin et al. / Molecular Catalysis 458 (2018) 189–197 195

The mass balances for the different components of reaction mix- Table 3
Kinetic and statistical parameters from modeling of typical Lewis and Brønsted acid
ture, are given by the differential Eqs. (4)–(10):
catalyzed conversion of DHA with ethanol.
∗
n0DHA dCDHA
= −r2 − r7 − r8 (4) Parameter (cm3 g−1 h−1 ) Catalyst
W dt
SnAl Amberlyst 35W
∗
n0DHA dCPA
= r2 − r4 (5) k2 43.20 ± 3.19 87.50 ± 5.30
W dt k4 ∗ 1160.00 ± 568.00 7220.00 ± 5590.00
∗ k5 ∗ 45.40 ± 27.10 275.00 ± 45.90
n0DHA dCEL
= r6 (6) k6 200.00 ± 109.00 24.50 ± 6.33
W dt k7 ∗ 17.90 ± 3.05 0.00 ± 0.00
∗ k8 16.40 ± 4.33 31.90 ± 10.80
n0DHA dCPAHA k- 4 8650.00 ± 441.00 0.00 ± 0.00
= r4 − r5 − r6 (7)
W dt k-7 6.79 ± 4.26 0.00 ± 0.00
∗ k-8 25.60 ± 11.60 90.30 ± 42.80
n0DHA dCPADA
= r5 (8) R2 0.98410 0.99332
W dt
SSE 0.07425 0.05569
∗
n0DHA dCGLADA MSC 3.10374 5.11438
= r7 (9)
W dt Reaction conditions: T = 353 K; P = 250 kPa; DHA/ethanol (molar) = 0.023,
∗ Wcat /WDHA = 43 wt.%.
n0DHA dCOTHERS
= r8 (10)
W dt
where Cj∗ is the relative concentration of compound j (Cj /CDHA 0 ); r ,
2 of PA. Then, the alcohol molecule attacks the terminal carbonyl of
r4 , r5 , r6 , r7 and r8 are the reaction rates of step 2, 4, 5, 6, 7 and 8, the adsorbed PA species leading to formation of PAHA. The follow-
respectively, in Scheme 1, expressed in mmol/hg. ing step involves the activation of PAHA on a proton site, forming
On the other hand, the mathematics equations for the proposed a carbocation intermediate that after releasing a H2 O molecule
model are given in Eqs. (11)–(16) before re-parametrization of Cj to consecutively reacts with another ethanol molecule to give the
dimensionless relativeconcentrations Cj∗ ; ki and k−i represent the terminal product, PADA.
kinetic constant of the forward and the reverse reaction, respec- The pathway toward EL via PAHA isomerization (step 6 in
tively, and k4∗ , k5∗ and k7∗ are parameters involving the ethanol Scheme 1) is clearly more favored on catalysts with Lewis acid
concentration: properties. In effect, k6 was almost 10 times higher on SnAl than
on the Amberlyst resin. These results are in line with the catalytic
r2 = k2 CDHA (11) experiments on SnAl (Fig. 2B and Table 2), since EL is obtained with
r4 = k4∗ CPA − k−4 CPAHA (12) high yields on Lewis acid sites. The EL synthesis occurs mainly by
activation and intramolecular rearrangement of PAHA (step 6 in
r5 = k5∗ CPAHA (13) Scheme 1) via a Meerwein-Ponndorf-Verley-Oppenauer (MPVO)
r6 = k6 CPAHA (14) mechanism [19]. To a lesser extent, EL could be obtained by direct
nucleophilic addition of ethanol to PA, followed by an intramolec-
r7 = k7∗ CDHA − k−7 CGLADA (15) ular Cannizzaro reaction (step 3 in Scheme 1), as was previously
r8 = k8 CDHA − k−8 COTHERS (16) reported using liquid [44] and solid catalysts with Lewis acid prop-
erties [8]. However, as explained above, the best modeling results
k4∗ = CEtOH
0
k4 (17) were obtained considering k3 = 0 and therefore, the contribution of
this mechanism was ruled out.
k5∗ = 0
CEtOH k5 (18)
Based on our results a surface mechanism to interpret the con-
k7∗ = 0
CEtOH k7 (19) version of PAHA toward EL on the SnAl catalyst is presented in
Scheme 3. The participation of an anionic intermediate interact-
The system of differential equations was solved numerically ing with surface tin species and that unlikely on other Lewis acids
using the methods described in Section 2.4. The nine kinetic param- such as Zn-containing catalysts, is postulated. This intermediate,
eters (ki and k−i ) calculated as described above are presented in formed by bidentate coordination of PAHA on a Sn atom, under-
Table 3. Results of Table 3 show that the value of k2 is similar for goes an intramolecular rearrangement with a 1,2-hydride shift to
both catalysts, suggesting that the initial dehydration of DHA is give rise to EL.
favored on either the Lewis acid sites of SnAl or on the Brønsted The kinetics results of Table 3 were summarized in Scheme 4
acid sites present on the Amberlyst resin. Similar findings were where the most favored reaction pathways during DHA conversion
reported by Dapsens et al. [11] and Clippel et al. [14]. Formation of on Lewis (Scheme 4A) and Brønsted (Scheme 4B) acid catalysts
GLADA (step 7) on the other hand, does not proceed on Brønsted are depicted. In Scheme 4, the main terminal products obtained
sites and occurs with a low rate constant on Lewis acid sites. are highlighted and the thickness of the arrow lines indicates the
The pathway toward PADA from the intermediate compound relative importance of the reaction steps on each catalyst.
PA occurs faster on Brønsted acids than on Lewis catalysts, as indi- Using these kinetic parameters, the DHA conversion and product
cated by the much higher values of k4∗ (step 4) and k5∗ (step 5) yield values predicted by the model were calculated as a function
calculated for the resin. Furthermore, the modeling results show of time. They are shown in Fig. 2 as solid lines. A good visual agree-
that step 4 is essentially irreversible on the resin. These results ment between experimental (points) and calculated (lines) data
can be mechanistically interpreted by considering that the tradi-

was obtained. The parity plot of Fig.3 involving all the experi-
tional homogeneously catalyzed esterification mechanism on H+ ∗ ∗
mental Cjobs and calculated Cjcalc relative concentrations of
sites applies on solids with Brønsted acidity [45]. Thus, in Scheme 2
a surface mechanism that interprets the conversion of PA to PAHA reactant and products for SnAl and Amberlyst 35W gave a statisti-
and the subsequent esterification of PAHA with ethanol is postu- cal confirmation of the model goodness of fit (R2 = 0.9723). Thus, the
lated, previous adaptation from the literature. The conversion of proposed model gives a suitable description of the DHA conversion
PA to PAHA requires H+ sites that activate the terminal C O group reactions on both, Lewis and Brønsted acid solids.
196 E.A. Pighin et al. / Molecular Catalysis 458 (2018) 189–197

Scheme 2. Postulated surface pathway for conversion of pyruvic aldehyde into pyruvic aldehyde diethyl acetal on Brønsted acid solids.

Scheme 3. Postulated surface pathway for synthesis of ethyl lactate on SnAl from pyruvic aldehyde hemiacetal via MPVO mechanism. Adapted from ref. [25].

A Lewis catalysts B Brønsted catalysts

EL EL

6 6
4 5 4 5
PA PAHA PADA PA PAHA PADA
2 2
DHA DHA
7
1 GLADA 1
8 8
GLA GLA
OTHERS OTHERS

Scheme 4. Simplified reaction pathways for triose conversion from kinetic modeling results on SnAl catalyst (A) and Amberlyst resin (B).

4. Conclusions mer more active. Brønsted acid solids favor the synthesis of pyruvic
aldehyde diethyl acetal, whereas ethyl lactate is the main product
The liquid-phase conversion of triose sugars is efficiently pro- on Lewis acid solids.
moted both on Brønsted and Lewis solid acid catalysts. The The kinetics of dihydroxyacetone conversion on Lewis and
reaction proceeds through a complex reaction network compris- Brønsted acid solids can be well described by a pseudohomoge-
ing a sequence of consecutive and parallel reaction steps with neous mechanism and first order rate expressions. Brønsted and
formation of ethyl lactate and pyruvic aldehyde diethyl acetal as Lewis sites promote the initial triose dehydration toward the pyru-
final products. The cesium salt of tungstophosphoric acid and a vic aldehyde intermediate with similar rates. The values of the
commercial Amberlyst resin were investigated as Brønsted acid kinetic rate constants for the following reaction steps confirm that
solids whereas alumina-supported tin and zinc oxides were tested on Brønsted solids the route toward pyruvic aldehyde diethyl acetal
as Lewis acid catalysts. The catalytic activity of these materials occurs at high rates. On Lewis solids, the synthesis of ethyl lactate
depends on the chemical nature (Brønsted or Lewis) of the acid via isomerization of the pyruvic aldehyde hemiacetal intermediate
sites, but this is not the only factor affecting it. Although Sn and predominates. Different reaction mechanisms occur on Brønsted
Zn oxides are both Lewis acids, the acid properties, as well as the and Lewis catalysts. On Brønsted solids, the traditional esterifica-
distinct coordination with reaction intermediates, make the for- tion mechanism on surface H+ sites takes place, transforming triose
 E.A. Pighin et al. / Molecular Catalysis 458 (2018) 189–197 197

[7] M.S. Jensen, M. Palucki, N.R. Rivera, K. Wells, Y. Xiao, Y. Wang, Ch. Yang, N.
Yasuda, WO 2202191 (2000), to Merck & Co., Inc.
[8] P.P. Pescarmona, K.P.F. Janssen, C. Delaet, C. Stroobants, K. Houthoofd, A.
Philippaerts, C. De Jonghe, J.S. Paul, P.A. Jacobs, B.F. Sels, Green Chem. 12
(2010) 1083–1089.
[9] R.M. West, M.S. Holm, S. Saravanamurugan, J. Xiong, Z. Beversdorf, E.
Taarning, C.H. Christensen, J. Catal. 269 (2010) 122–130.
[10] K.P.F. Janssen, J.S. Paul, B.F. Sels, P.A. Jacobs, Stud. Surf. Sci. Catal. 170 (2)
(2007) 1222–1227.
[11] P.Y. Dapsens, B.T. Kusema, C. Mondelli, J. Pérez-Ramírez, J. Molec. Catal. A:
Chem. 388–389 (2014) 141–147.
[12] A.M. Mylin, S.I. Levytska, M.E. Sharanda, V.V. Brei, Catal. Comm. 47 (2014)
36–39.
[13] L. Li, C. Stroobants, K. Lin, P.A. Jacobs, B.F. Sels, P.P. Pescarmona, Green Chem.
13 (2011) 1175–1181.
[14] F. de Clippel, M. Dusselier, R. Van Rompaey, P. Vanelderen, J. Dijkmans, E.
Makshina, L. Giebeler, S. Oswald, G.V. Baron, J.F.M. Denayer, P.P. Pescarmona,
P.A. Jacobs, B.F. Sels, J. Am. Chem. Soc. 134 (2012) 10089–10101.
[15] C.M. Osmundsen, M.S. Holm, S. Dahl, E. Taarning, Proc. R. Soc. A (2012) 1–17.
[16] N. Godard, A. Vivian, L. Fusaro, L. Cannavicci, C. Aprile, D.P. Debecker, Chem
Cat Chem (2017), http://dx.doi.org/10.1002/cctc.201601637, in press.
[17] J. Wang, Y. Masui, M. Onaka, Appl. Catal. B: Environ. 107 (2011) 135–139.
[18] H.J. Cho, P. Dornath, W. Fan, ACS Catal. 4 (2014) 2029–2037.
[19] E. Taarning, S. Saravanamurugan, M. Spangberg Holm, J. Xiong, R.M. West,
C.H. Christensen, Chem. Sus. Chem. 2 (2009) 625–627.
[20] J. Dijkmans, M. Dusselier, D. Gabriels, K. Houthoofd, P.C.M.M. Magusin, S.
Huang, Y. Pontikes, M. Trekels, A. Vantomme, L. Giebeler, S. Oswald, B.F. Sels,
ACS Catal. 5 (2015) 928–940.
[21] C. Hammond, S. Conrad, I. Hermans, Angew. Chem. Int. Ed. 51 (2012)
11736–11739.
Fig. 3. Parity plot for experimental data and model predictions (closed symbols: [22] S. Tolborg, A. Katerinopoulou, D.D. Falcone, I. Sádaba, C.M. Osmundsen, R.J.
Amberlyst 35W resin; open symbols: SnAl catalyst). Davis, E. Taarning, P. Fristrup, M.S. Holm, J. Mater. Chem. A 2 (2014)
20252–20262.
[23] W.N.P. van der Graaff, G. Li, B. Mezari, E.A. Pidko, E.J.M. Hensen,
sugars into pyruvic aldehyde hemiacetal and diethyl acetal with ChemCatChem 7 (7) (2015) 1152–1160.
the release of H2 O molecules. On the other hand, on catalysts with [24] E. Pighin, V.K. Díez, J.I. Di Cosimo, Appl. Catal. A: General 517 (2016) 151–160.
Lewis acid properties, the pyruvic aldehyde hemiacetal interme- [25] E. Pighin, V.K. Díez, J.I. Di Cosimo, Catal. Today 289 (2017) 29–37.
[26] O. Acarbas, E. Suvacı, A. Dogan, Ceram. Int. 33 (2007) 537–542.
diate coordinates with metal cations (particularly Sn4+ ) forming a [27] D.E. Mears, Ind. Eng. Chem. Proc. Des. Dev. 10 (4) (1971) 438–447.
cyclic intermediate that undergoes an intramolecular rearrange- [28] P.B. Weisz, C.D. Prater, Adv. Catal. 6 (1954) 143–196.
ment with a 1,2-hydride shift, yielding ethyl lactate. [29] R.E. Walpole, R.H. Myers, S.L. Myers, K.E. Ye, Probability & Statistics for
Engineers & Scientists, 8th Ed., Pearson Prentice Hall, 2007.
[30] S. Armenise, E. García-Bordeje, J.L. Valverde, E. Romeo, A. Monzón, Phys.
Acknowledgements Chem. Chem. Phys. 15 (2013) 12104–12117.
[31] I. Kozhevnikov, Chem. Rev. 98 (1998) 171.
[32] S. Tatematsu, T. Hibi, T. Okuhara, M. Misono, Chem. Lett. (1984) 865–868.
Authors thank the Agencia Nacional de Promoción Científica
[33] Y. Izumi, K. Hisano, T. Hida, Appl. Catal. A: Gral. 181 (1999) 277–282.
y Tecnológica (ANPCyT), Argentina (Grant PICT 1857/15), CON- [34] S. Soled, S. Miseo, G. McVicker, W.E. Gates, A. Gutierrez, J. Paes, Chem. Eng. J.
ICET, Argentina (grant PIP 11220090100203/10) and Universidad 64 (1996) 247.
[35] B. Bachiller-Baeza, J.A. Anderson, J. Catal. 228 (2004) 225–233.
Nacional del Litoral, Santa Fe, Argentina (grant CAID PI 007-040/05)
[36] S. Toppi, C. Thomas, C. Sayag, D. Brodzki, K. Fajerwerg, F. Le Peltier, C. Travers,
for financial support of this work. G. Djéga-Mariadassou, J. Catal. 230 (2005) 255–268.
[37] V.K. Díez, C.R. Apesteguía, J.I. Di Cosimo, Catal. Lett. 123 (2008) 213–219.
[38] K. Tanabe, M. Misono, Y. Ono, H. Hattori, New solid acids and bases their
Appendix A. Supplementary data
catalytic properties, in: K. Tanabe, M. Misono, Y. Ono, H. Hattori (Eds.), Stud.
Surf. Sci. and Catal., vol. 51, 1989 (p. 82 Elsevier ISBN: 978-0-444-98800-3).
Supplementary data associated with this article can be found, [39] J. Penzien, A. Abraham, J.A. van Bokhoven, A. Jentys, T.E. Muller, C. Sievers, J.A.
in the online version, at http://dx.doi.org/10.1016/j.mcat.2017.11. Lercher, J. Phys. Chem. B 108 (2004) 4116–4126.
[40] T. Takewaki, L.W. Beck, M.E. Davis, J. Phys. Chem. B 103 (1999) 2674–2679.
026. [41] M. Orazov, M. Deimund, M.E. Davis 24th., North American Catalysis Society
Meeting Proceeding, 14–19th, June 2015, Pittsburgh (USA), O-W-A-15, 2015.
References [42] R.T. Sanderson, Chemical Bonds and Bond Energy, 2nd Ed., Academic Press,
New York, 1976.
[43] V.K. Díez, P.A. Torresi, P.J. Luggren, C.A. Ferretti, J.I. Di Cosimo, Catal. Today
[1] A. Corma, S. Iborra, A. Velty, Chem. Rev. 107 (2007) 2411–2502. 213 (2013) 18–24.
[2] E.L. Kunkes, D.A. Simonetti, R.M. West, J.C. Serrano-Ruiz, C.A. Gärtner, J.A. [44] Y. Hayashi, Y. Sasaki, Chem. Commun. (2005) 2716–2718.
Dumesic, Science 322 (2008) 417. [45] R.T. Morrison, R.N. Boyd, Organic Chemistry, 6th Ed., Prentice-Hall, New York,
[3] G.W. Huber, S. Iborra, A. Corma, Chem. Rev. 106 (2006) 4044–4098. 1959, pp. 680.
[4] K.T. Narayana, US Patent 6342626, 2002.
[5] T.N.B. Kaimal, P. Vijayalakshmi, B. Ramalinga, A. A. Laxmi, US Patent N◦
6342626, 2002.
[6] D.I. Weisblad and B.J. Magerlein, US 2631149 (1953), to Kalamazoo
Corporation.