You are on page 1of 10

Chemical Engineering Science 116 (2014) 118–127

Contents lists available at ScienceDirect

Chemical Engineering Science


journal homepage: www.elsevier.com/locate/ces

Reaction mechanism for glycerol dehydration in the gas phase over


a solid acid catalyst determined with on-line gas chromatography
Isabelle Martinuzzi a,n, Yassine Azizi a,1, Jean-François Devaux b, Serge Tretjak c,
Orfan Zahraa a, Jean-Pierre Leclerc a
a
Laboratoire Réactions et Génie des Procédés, UMR 7274 CNRS – Université de Lorraine, 1 rue Grandville, B.P. 20451, 54 001 Nancy Cedex, France
b
Arkema, CRRA Pierre-Benite, B.P. 63, 69493 Pierre-Bénite Cedex, France
c
Arkema, CRDE, B.P. 61005, 57501 Saint-Avold Cedex, France

H I G H L I G H T S G R A P H I C A L A B S T R A C T

 We discover by-products of the


glycerol dehydration reaction which
are not yet cited in the literature.
 We used specific on-line gas chro-
matography to quantify most of the
compounds.
 We pass over the catalyst many
products of the glycerol dehydration
in order to identify pathways of the
reaction mechanism.
 We determine a reaction mechanism
of glycerol dehydration with
detailed pathways.
 We determine compounds responsi-
ble of carbon deposit on the catalyst.

art ic l e i nf o a b s t r a c t

Article history: Gas phase glycerol dehydration to form acrolein over a solid acid catalyst was studied to understand the
Received 5 November 2013 formation of by-products and to establish a detailed reaction mechanism. The experiments were
Received in revised form conducted in an isothermal fixed bed reactor operated under various conditions (temperature, space-
7 January 2014
time velocity, concentration). From a methodological point of view, many secondary products of the
Accepted 25 April 2014
reaction were passed separately over the catalyst to understand the different pathways of glycerol
Available online 4 May 2014
dehydration. Our scientific contribution to the analytical method is that all of the products were
Keywords: analyzed simultaneously using a multivalve on-line gas chromatograph equipped with a flame
Acrolein ionization detector and a thermal conductivity detector. Two products were quantified using high
Catalysis
performance liquid chromatography, and the unknown products (2-methyl 2-cyclopenten-1-one,
Dehydration
3-methyl 2-cyclopenten-1-one and 2,3-butanedione) were identified by gas chromatography–mass
Gas phase
Glycerol spectrometry. Several new compounds were found. Based on the determined products, a detailed
Reaction mechanism reaction mechanism was proposed.
& 2014 Elsevier Ltd. All rights reserved.

n
Corresponding author. Tel.: þ 33 3 83 17 50 42.
E-mail addresses: isabelle.martinuzzi@free.fr (I. Martinuzzi), yassine.azizi@ircelyon.univ-lyon1.fr (Y. Azizi), jean-francois.devaux@arkema.com (J.-F. Devaux),
serge.tretjak@arkema.com (S. Tretjak), orfan.zahraa@univ-lorraine.fr (O. Zahraa), jean-pierre.leclerc@univ-lorraine.fr (J.-P. Leclerc).
1
Present address: Yassine AZIZI, IRCELYON, 2 Avenue Albert Einstein, 69626 Villeurbanne Cedex, France.

http://dx.doi.org/10.1016/j.ces.2014.04.030
0009-2509/& 2014 Elsevier Ltd. All rights reserved.
I. Martinuzzi et al. / Chemical Engineering Science 116 (2014) 118–127 119

1. Introduction on-line GC with three detectors: two thermal conductivity detec-


tors (TCDs) for analyzing H2 and CO, CO2, N2 and one FID for
Glycerol is a by-product of oleochemistry and biodiesel analyzing C1 to C6 hydrocarbons. Liquid samples were also
production which increased drastically over the last decade. analyzed using GC equipped with a FID detector and GC–MS. In
Because 1 T of 1st generation biodiesel coproduces 0.1 T of addition, an IR cell was used to quantify CO2 during the catalyst
glycerol, glycerol production from biodiesel reached 860 kT in regeneration step.
Europe in 2011 according to the European Biodiesel Board This paper proposes a reaction network for glycerol dehydra-
Website (2013); approximately 2100 kT/year were produced tion over an industrial solid acid catalyst. This mechanism is based
worldwide in 2010–2012, and production is expected to grow on our experiments that measure the impact of various operating
to 3600 kT/year in 2022 according to OECD-FAO Agricultural parameters on the products selectivities: temperature, glycerol
outlook Website (2013). Glycerol is already used in the food, concentration, gas hourly space velocity (GHSV) and oxygen
cosmetics and pharmaceutical industries, but the world produc- concentration. We also passed many secondary products over
tion of glycerol far overtakes its consumption in existing applica- the catalyst separately to understand the different pathways for
tions. Consequently, excess glycerol is either stored or burnt and glycerol dehydration. The novel aspect of this paper is the
researchers are motivated to valorize this product. The synthesis analytical method employed: the whole products were analyzed
of acrylic acid from glycerol is a promising application. Indeed, with a multivalve on-line GC equipped with one FID and one TCD.
the global demand for acrylic acid is estimated at approximately Two products were quantified using high performance liquid
4200 kT/year according to Arkema investor day (2012); so that an chromatography (HPLC). The unknown products have been iden-
important part of the acrylic acid production could be synthe- tified by GC–MS.
tized from glycerol.
The route from glycerol to acrylic acid comprises glycerol
dehydration to form acrolein, followed by oxidation to form acrylic 2. Materials and methods
acid. The first reaction has been conducted in the gas phase over
solid acid catalysts (zeolites, heteropolyacids, and mixed oxides); 2.1. Catalyst characterization
the selectivity for acrolein is often near 70%, but 98% selectivity is
possible, according to the review by Katryniok et al. (2010). The solid acid catalyst used during the experiments was an
Understanding the product formation would help to improve the industrial catalyst known as ZR24. The Hammett acidity of the
selectivity for acrolein and to exert better control over the catalyst was measured between  8.2 and  5.6. In this study, the
secondary products. Several reaction mechanisms have been solid catalyst was crushed to form particles from 300 to 350 mm.
already proposed (Chai et al., 2007; Corma et al., 2008; To identify the presence of acidic and basic sites on the catalyst,
Deleplanque et al., 2010; Suprun et al., 2009; Tsukuda et al., NH3-temperature programmed desorption (NH3-TPD) and CO2-TPD
2007; Wang et al., 2009). The experiments have largely been were utilized, respectively.
conducted in a fixed bed reactor, except for the work published by The temperature programmed desorptions were performed in
Corma et al. (2008); this group used also a fluid catalytic cracking a U-shaped quartz catalytic reactor. The measurements were
reactor. Each author has used a different catalyst, but the reaction carried out using a mass spectrometer (VG Gasslab 300 quadru-
mechanisms can still be compared. Wang et al. (2009) used a pole mass spectrometer). Before the measurements, approxi-
vanadium phosphate oxide, Deleplanque et al. (2010) and Suprun mately 200 mg of the sample was heated to 100 1C for 30 min
et al. (2009) used other mixed oxides, such as titanium phosphate, under flowing helium (30 ml/min). Afterward, NH3 (respectively
aluminum phosphate and iron phosphate, Chai et al. (2007) and CO2) adsorption was carried out over 30 min with 0.8% NH3/He
Tsukuda et al. (2007) used silica supported heteropolyacids and (respectively 2% CO2/He). The physically adsorbed ammonia
Corma et al. (2008) used a ZSM-5 zeolite. Despite the wide range (or CO2) was removed by purging with helium at 100 1C for
of catalysts, the authors agreed on two different dehydration 30 min. The NH3-TPD (or CO2-TPD) of the samples was carried
pathways, leading to either acetol or 3-hydroxypropanal (3-HPA), out by linearly increasing the temperature from 100 to 850 1C at
which is further dehydrated to form acrolein. The retro-aldol 20 1C/min under 30 mL/min of helium while measuring the
condensation of 3-HPA into formaldehyde and acetaldehyde was evolved NH3 (or CO2) using mass spectrometry. The results in
also considered as a pertinent reaction (Geng et al., 2012; Laino Fig. 1 show that ZR24 is an acidic catalyst with both weak and
et al., 2011; Nimlos et al., 2006; Paine et al., 2007). Furthermore, strong acid sites: we observed a strong NH3 desorption between
almost all of the authors agreed that acetol was hydrogenated to 200 1C and 800 1C. However, the catalyst contains more strong
form 1,2-propanediol, followed by further dehydration to form
acetone. However, the detected secondary products were variable,
depending on the reaction conditions and the analytical systems.
Moreover, the interpretation of these formation pathways varied
between authors.
Due to the difficulty of analyzing various products, including
permanent gases (O2, N2, CO, CO2) and condensable products (up
to C6 compounds), we will summarize selected methods cited in
the literature. Wang et al. (2009) and Deleplanque et al. (2010)
used cold traps to measure condensable products off-line with
a gas chromatography mass spectrometer (GC–MS) coupled to
a flame ionization detector (FID). Three different on-line gas
chromatographs (GC) were needed to measure volatile com-
pounds, such as nitrogen, oxygen, CO, CO2, acrolein, acetaldehyde
and acetic acid. Suprun et al. (2009), Chai et al. (2007) and Tsukuda
et al. (2007) analyzed the products condensed in a cold trap with
an off-line GC equipped with a FID detector; GC–MS was used to
identify the unknown products. Corma et al. (2008) used an Fig. 1. NH3-TPD and CO2-TPD profiles of ZR24 samples.
120 I. Martinuzzi et al. / Chemical Engineering Science 116 (2014) 118–127

Fig. 2. Experimental pilot scheme.

acid sites than weak acid sites because more NH3 desorption was
observed after 350 1C, indicating strong acid sites. A small quantity
of strong basic sites was also observed; the CO2-TPD shows a
maximum at 600 1C.

2.2. Experimental set-up

The experimental set-up is original and represented Fig. 2. The


experiments were performed in a stainless steel isothermal fixed
bed reactor with a 1 cm inner diameter and a 12 cm length.
A 95 mm stainless steel porous frit was used to retain the catalyst
particles. Three thermocouples were placed in the reactor to
measure the temperature precisely. One was placed in the catalytic
bed and the others were placed slightly above it. A preheated and
a static mixing zone 10 cm long and 1 cm in diameter was placed
before the entrance of the reactor.
The mass flow controllers (Bronkhorst) were located before the
preheated zone to maintain the nitrogen, oxygen and methane
flow-rates at the desired values. Methane was used as an internal
standard during GC. Methane remains inert toward our experi-
Fig. 3. Gas chromatography scheme.
mental conditions. A high performance liquid chromatography
(HPLC) pump (Jasco – PU 2080) was used to carry the liquid
reactants to the entrance of the preheated zone. analysis and a FID for condensable gases analysis, as well as an
At the reactor outlet, a warmed pipe was used to avoid the HP-innowax (30 m  0.53 mm  1 mm) to separate the condensa-
condensation of products before the analysis in the on-line GC. After ble gases from one another and from non-condensable gases, a GS-
GC, the products were trapped in cold water (1–2 1C) within two cold Q (30 m  0.53 mm) to separate CO2 from the other permanent
collectors in series or stored in a bin. This bin contained water and gases and an HP-molesieve (30 m  0.53 mm  0.5 mm) to separate
potassium hydroxide at pH 14 to neutralize the acrolein. The reactor the other permanent gases (Fig. 3). This complex system required
pressure was set with a regulator at the end of the system. three six port valves. The first functioned as an injector with
a 250 ml sampling loop to transfer the product samples to the GC.
2.3. Analytical devices The second was used to direct the sample toward the FID or the
columns connected to the TCD. The third was used to separate the
The gases were analyzed on-line and continuously with permanent gases: the GS-Q retained CO2, but not N2, CO, O2, CH4;
a tailor-made gas chromatograph (Agilent GC 7890 A). This device those were directed to the HP-molesieve for separation. CO2
was equipped with two detectors and three separation columns exposure on the HP-molesieve must be avoided to prevent
(Agilent): a TCD for permanent gases (N2, CO, O2, CH4, CO2) blockages. To generate the calibration curves, compounds with
I. Martinuzzi et al. / Chemical Engineering Science 116 (2014) 118–127 121

water, N2 and CH4 at the desired concentrations were passed line GC and was present in the mixture (1 vol%). The pressure was
through the experimental set-up. The compounds were quantified constant during the experiments: 2.7 bar, absolute. In addition,
by comparing their peak area to the CH4 peak area. 0.966 g catalyst was added to the reactor. All of the experiments
A high performance liquid chromatograph (Shimadzu HPLC, were performed twice to verify the repeatability, except experi-
with PDA detector and LiChrosphers 60 RP-select B (25 m  ment 2 which was performed three times.
4 mm), Merck column) was used off-line to quantify the methyl- Analyzing the glycerol dehydration products using GC, HPLC
glyoxal and formaldehyde. A gas chromatograph coupled with and GC–MS allowed us to identify the products cited in the
mass spectrometer (Shimadzu GC–MS – QP2010) was used off-line literature: acrolein, acetaldehyde, propanal, acetone, methanol,
to identify the unknown reaction products. A HP-innowax formaldehyde, methylglyoxal, allylic alcohol, hydroxypropanal,
(30 m  0.32 mm  0.25 mm) column was used to compare the 1,3-dioxan-5-ol, acetic acid, phenol, ethanol, methacrolein, CO,
GC peaks with the GC–MS peaks. After the products were CO2, dioxanes, furan derivatives and 2-cyclopenten-1-one. We
identified on the GC–MS, they were injected into the GC and the have also identified several products in lower quantities that
GC–MS to verify their retention time and assert their presence. have not been cited elsewhere: 2-methyl 2-cyclopenten-1-one,
3-methyl 2-cyclopenten-1-one and 2,3-butanedione.
2.4. Calculations Table 2 shows the influence of the different parameters on the
carbon product selectivity. Because the catalyst was deactivated
The compound conversion (1), the carbon product selectivity a few hours after the beginning of the experiments, the selectivity
(2) and the carbon mass balance (3) were calculated according to was calculated 75 min after the glycerol injection, corresponding
the following expressions: to the first analysis of the cold traps.

ni input  ni output
Xi ¼ 100 ð1Þ 3.1.1. Effect of glycerol concentration
ni input
The influence of the glycerol concentration on the selectivity
nj output cj was studied with experiments 1–3 at 270 1C, 2.7 bar (absolute)
SCj ¼ 100 ð2Þ
ni input  ni output ci and a GHSV of 18,000 h  1. The feed was 6 vol% of oxygen, 40 vol%
of water, 2, 3 and 4 vol% of glycerol, respectively, and the
CB ¼ ∑ SCj ð3Þ remainder was nitrogen. The acetaldehyde, acetone, 2,3-butane-
j
dione, formaldehyde, CO and CO2 selectivities decrease when the
where: ni input and ni output are the molar flow of the reactant glycerol concentration increases. The propanal remains constant.
(e.g.: glycerol) at the input and output of the reactor; nj is molar No trend is apparent for the acrolein, methylglyoxal, phenol or
flow of the product; cj and ci represent the number of carbon 3-methyl 2-cyclopenten-1-one, selectivities. However, the hydro-
atoms in a molecule of product j and reactant i. xypropanal, allylic alcohol and “others” selectivities increase
The gas hourly velocity is calculated according to (4); Qv total is slightly with the glycerol concentration. Experiment 3 shows that
the total flow of the gases expressed in a normal liter per hour, and the glycerol conversion is 95%. Knowing that we had full conver-
V is the apparent volume of the crushed catalyst in liters. sion during these experiments after 34 min, we concluded that the
catalyst deactivates more rapidly when the glycerol concentration
Q V total
GHSV ¼ ð4Þ increases.
V catalyst

3.1.2. Effect of GHSV


Experiments 2, 4 and 5 show the influence of the GHSV on the
3. Catalytic results and discussion product selectivities. These experiments were realized at 270 1C,
2.7 bar (absolute) and GHSV of 18,000; 27,000; 36,000 h  1,
3.1. Glycerol dehydration respectively. The feed was 6 vol% of oxygen, 40 vol% of water and
3 vol% of glycerol with nitrogen making up the remainder. Table 2
The experiments were conducted in a catalytic fixed bed shows that acetol, hydroxypropanal, allylic alcohol, phenol and
reactor, as previously described. The experiments were run “others” increase slightly with the GHSV. Acrolein, methylglyoxal,
between 270 1C and 308 1C in the catalytic bed with glycerol acetaldehyde, propanal, acetone, 2,3-butanedione; formaldehyde,
concentrations between 2 and 4 vol%, oxygen concentrations 2 cyclopenten-1-one,3 methyl, CO2 and CO selectivities decrease
between 0 and 6 vol%, and a gas hourly space velocity (GHSV) when the GHSV increases. The conversion of glycerol decreases
between 18,000 h  1, and 36,000 h  1 (Table 1). Therefore, the total slightly when the GHSV increases.
volumetric flow varied between 20 NL h  1 and 40 NL h  1, while
the catalyst volume was constant at 1.1 mL. The water concentra-
tion was constant at 40 vol%, and nitrogen was used to complete 3.1.3. Effect of reaction temperature
the mixture. Methane was used as an internal standard for the on- Experiments 2, 6 and 7 show the influence of the reaction
temperature on the product selectivities. These experiments
occurred at 270 1C, 291 1C and 308 1C, respectively, under 2.7 bar
Table 1
Experimental conditions.
(absolute) and a GHSV of 18,000 h  1. The feed was 6 vol% of
oxygen, 40 vol% of water, 3 vol% of glycerol and the rest was
Units nitrogen. The acrolein, acetol, hydroxypropanal, 2 cyclopenten-1-
one, 3 methyl and “others” decrease slightly when the tempera-
Temperature 1C 270–308 71
ture increases. Acetol and hydroxypropanal are very reactive
Pressure bar 2.7 7 0.05
(absolute) products and acetol is known to generate many secondary pro-
Particles diameter mm 300–350 ducts. Therefore, when the temperature increases, these com-
Catalyst mass g 0.966 7 0.0005 pounds must be completely consumed to provide the secondary
Qv NL h  1 20–40 7 0.5 products. The acetaldehyde, formaldehyde and phenol selectivities
GHSV h1 18,000–36,000 7 500
Glycerol/H2O/N2/O2/CH4 vol% 2–4%/40%/50–56%/0–6%/1% 7 0.1%
increase with the temperature. The CO and CO2 selectivities
increase with the temperature, indicating that C–C bond cleavage
122 I. Martinuzzi et al. / Chemical Engineering Science 116 (2014) 118–127

Table 2
Carbon product selectivities for glycerol conversion: the influences of the glycerol concentration, GHSV, temperature and oxygen concentration.

Experiments 1 2 3 4 5 6 7 8 9

vol% Glycerol 2 3 4 3 3 3 3 3 3
GHSV (h  1) 18 000 18 000 18 000 27 000 36 000 18 000 18 000 18 000 18 000
Qv (NL/h) 20 20 20 30 40 20 20 20 20
Temperature (1C) 270 270 270 270 270 291 308 270 270
vol% Water 40 40 40 40 40 40 40 40 40
vol% Oxygen 6 6 6 6 6 6 6 3 0
Glycerol conversion (%) 100% 100% 95% 90% 80% 100% 100% 94% 92%

Molar carbon selectivity (%)


Acrolein 69 76 73 67 63 74 60 82 77
Acetol 0.022 2.4 2.3 3.0 3.9 0 0 2.3 7.5
Hydroxypropanal 1.0 1.0 1.4 0.94 1.7 0 0 0.45 0.37
Methylglyoxal 4.8 7.8 6.2 7.0 6.0 6.5 9.3 5.4 1.0
Acetaldehyde 1.1 0.73 0.56 0.61 0.54 2.1 2.5 0.77 0.53
Propanal 0.32 0.30 0.29 0.30 0.25 0.45 0.30 0.38 0.66
Acetone 0.020 0.015 0.013 0.015 0.012 0.027 0.018 0.021 0.19
2,3-butanedione 0.061 0.026 0.015 0.024 0.019 0.067 0.053 0.035 0.027
Allylic alcohol 0.063 0.11 0.11 0.14 0.15 0.039 0.085 0.16 0.69
Formaldehyde 0.48 0.41 0.33 0.31 0.27 1.5 1.3 0.35 0.19
Phenol 0.063 0.037 0.050 0.079 0.12 0.060 0.061 0.084 0.13
Acetic acid 0 0 0 0 0 0 0.48 0 0
2 cyclopenten-1-one, 3 methyl 0.016 0.13 0.084 0.093 0.080 0.0084 0 0.085 0.059
CO2 1.0 0.77 0.55 0.53 0.48 1.6 2.1 0.58 0
CO 0.98 0.58 0.36 0.38 0.37 2.1 3.3 0.38 0
Others 2.5 4.2 5.4 5.9 5.9 1.1 0.22 5.9 7.7
Carbon balance (%) 81 94 88 86 83 90 80 99 96

is easier at high temperature. No trend is apparent for the experiments; for a given experiment, the temperature was con-
methylglyoxal, propanal, acetone, 2,3-butanedione, allylic alcohol stant throughout the run. The experiments were run (a) with and
selectivities. Acetic acid is only present at high temperatures, (b) without oxygen. The molar ratio of water, O2, N2, and CH4 was
indicating that high temperatures are favorable for oxidizing 35/5/57–59/1 for case (a) and 35/0/62–64/1 for case (b). For each
acetaldehyde to form acetic acid. compound, the feed concentration is presented Table 3.
The carbon mass balances were calculated as the sum of the
measured product selectivities, explaining why their values some-
3.1.4. Effect of oxygen concentration
times appeared low when the conversion was low. Indeed, the error
Experiments 2, 8 and 9 show the influence of the oxygen
is important when the conversion is low due to the precision of the
concentration on the product selectivities. These experiments
measurements taken during gas chromatography. Few authors have
occurred at 270 1C, 2.7 bar (absolute) and a GHSV of 18,000 h  1.
estimated the carbon mass balance, where they assumed that the
The feed included 6, 3 and 0 vol% of oxygen, respectively, 40 vol%
sum of the compound selectivities equals 100% (Chai et al., 2007;
of water, 3 vol% of glycerol and the rest was nitrogen. The
Corma et al., 2008; Suprun et al., 2009; Tsukuda et al., 2007): in this
hydroxypropanal, methylglyoxal, acetaldehyde, formaldehyde,
case, the selectivity of the “unknown compounds” has been
and 2 cyclopenten-1-one,3 methyl contents increase when the
deduced after determining the selectivities of the known com-
oxygen content increases. Oxidation products such as CO and CO2
pounds. Only Deleplanque et al. (2010) and Wang et al. (2009)
increase as well. Acetol, propanal, acetone, allylic alcohol, phenol
calculated the real carbon deposit for the dehydration of glycerol;
and “others” decrease when the oxygen concentration increases.
the values ranged from 66 to 97%. In the present work, the carbon
These results are confirmed by those of Deleplanque et al. (2010):
mass balances were satisfactory considering that only the measured
acetaldehyde, CO and CO2 are favored by oxygen, in contrast to
concentrations were taken into account.
acetol and propanal, which are disfavored by oxygen.

3.2. Reaction pathway 3.2.1. Reactions in the presence of oxygen


The conversion of reactants, the product selectivities and the
To understand the formation of the by-products and to deter- carbon mass balances of the reactions with 5% oxygen are
mine the reaction pathways, the conversion of several compounds presented in Table 3 (light grey lines). The reactions of most
was assessed after their passage over the catalyst. The compounds compounds, except for acetic acid, allylic alcohol and methanol,
were acetol, acetaldehyde, acetic acid, acetone, acrolein, propanal, have high selectivities for CO (15–35%) and CO2 (20–85%) due to
formaldehyde, methylglyoxal, allylic alcohol, 2 cyclopenten-1-one, the presence of copious oxygen. The following observations show
3 methyl, phenol, 1,2-propanediol, methanol and hydroxypropa- the extent of reaction for each compound and the obtained final
nal. Even if 1,2-propanediol was not observed in the reaction products.
products, it is interesting to see what type of product it forms over Acetic acid: X¼0%. Acetic acid did not react on the catalyst with
the catalyst because many authors (Chai et al., 2007; Corma et al., or without oxygen. This compound should be a final product, as
2008; Deleplanque et al., 2010; Tsukuda et al., 2007; Wang et al., confirmed by Wang et al. (2009).
2009) have suggested that it is an intermediate in the reaction Acetaldehyde: X¼ 55%. Acetic acid (13%) was most likely formed
converting acetol into acetone. The experimental conditions by oxidation of acetaldehyde, as proposed by Deleplanque et al.
included a GHSV of 18,000 h  1, 276–290 1C and 2.7 bar (absolute). (2010).
The wide temperature range in the catalyst bed is due to the Acetone: X¼60%. Acetone primarily underwent total oxidation
difficulty of maintain exactly the same temperature for all the to form CO2. The selectivity in acetaldehyde (17%) presumes that
I. Martinuzzi et al. / Chemical Engineering Science 116 (2014) 118–127 123

Table 3
Catalytic results using different products in the absence and presence of 5% oxygen.

The light grey lines represent reactions in the presence of oxygen (5%) and the white lines represent reactions in the absence of oxygen. The grey cells denote selectivities
above 10%.
124 I. Martinuzzi et al. / Chemical Engineering Science 116 (2014) 118–127

acetaldehyde was most likely formed through a C–C bond cleavage 3-methyl 2-cyclopenten-1-one: X ¼100%. 3-methyl 2-cyclo-
in acetone. We did not observe acetic acid formation, in contrast to penten-1-one is a cyclic compound that was easily transformed
Corma et al. (2008) and Wang et al. (2009); these authors worked into CO (20%) and CO2 (45%) in the presence of oxygen. Because
at a higher temperature (350 1C and 300 1C, respectively), favoring acetaldehyde (5%) and acetone (3%) were created, acetone and
acetic acid. Even if the low introduced quantity and the moderate acetaldehyde might be involved in its formation. This supposition
conversion of acetone prevent us from detecting acetic acid, our is debatable but the formation of 3-methyl 2-cyclopenten-1-one is
mass carbon balance indicates that acetic acid formed with a low not described in the literature, this hypothesis is a starting point.
selectivity ( o1%) if at all. This pathway is therefore negligible. Furthermore, 2-cyclopenten-1-one derivatives can be created
Methylglyoxal: X¼ 100%. Despite its high reactivity, its selectivity through an aldol condensation between ketones and/or aldehydes
during the dehydration of glycerol could reach 9%. Acetaldehyde (Cai and Xie, 2011).
(15%) was most likely formed through a C–C bond cleavage. Phenol: X¼99%. According to Suprun et al. (2009)and Lauriol-
Similarly, acetic acid was most likely formed by methylglyoxal Garbey et al. (2011), phenol may be formed from a Diels–Alder
because its high selectivity (20%) reveals that it could not be created reaction between acetone and acrolein. However, a Diels–Alder
by only acetaldehyde. reaction is reversible, and the diene and alkene may be reformed.
Acetol: X¼ 100%. The full conversion of acetol agrees with the Because acrolein is a stable product, we should have detected it
lack of acetol formed during the glycerol dehydration reaction at but did not. However, we found acetone (5%) and acetaldehyde
higher temperatures (trials 6 and 7 of Table 2). The C–C bond (3%). Therefore, acetone and probably acetaldehyde must be
cleavage of acetol generates acetaldehyde (15%); the acetaldehyde involved in the formation of phenol.
is then oxidized to form acetic acid (30%). Because acetic acid was 1,2-propanediol: X¼100%. The high selectivities for acetalde-
more common than in the acetaldehyde conversion, it must be hyde (30%) and acetone (1.5%) confirm the suppositions of
formed by other compounds, such as methylglyoxal. Deleplanque et al. (2010): 1,2-propanediol may be an unstable
Formaldehyde: X ¼30%. Formaldehyde was only somewhat intermediate from the transformation of acetol to form acetalde-
reactive and was mainly oxidized to CO/CO2 (65%) and hydro- hyde and acetone.
genated to form methanol (35%), as proposed by Chai et al. (2007). Methanol: X¼ 25%. The moderate selectivity of formaldehyde
Hydroxypropanal þacrolein: Hydroxypropanal disappeared (30%) implies that methanol was dehydrogenated to form formal-
completely in the presence or absence of oxygen. A mixture of dehyde. The carbon mass balance was 30% because we only
these two compounds was prepared in the laboratory because observed formaldehyde (30%).
hydroxypropanal is not commercially available. This solution was
prepared at room temperature by reacting water with acrolein
over a few days. The solution could not be titrated; it contained 3.2.2. Reactions in the absence of oxygen
water, acrolein and hydroxypropanal. The main products were In Table 3 (white lines), the conversion of reactants, the
acrolein and acetaldehyde in both cases. Formaldehyde was also products selectivities and the carbon mass balances of the reac-
present. Therefore, hydroxypropanal was dehydrated to form tions in the absence of oxygen are shown. Interesting observations
acrolein; a retro-aldol condensation could form acetaldehyde were found with merits that will be discussed later. The selectiv-
and formaldehyde (Geng et al., 2012; Laino et al., 2011; Nimlos ities in CO and CO2 drastically decreased compared to the case
et al., 2006; Paine et al., 2007). Hydroxypropanal was an unstable with oxygen, although some CO/CO2 was formed. In general,
intermediate, but was still properly detected using our on-line GC a rapid deactivation of the catalyst occurred in the absence of
under our experimental conditions. oxygen; in addition, for some compounds, an important carbon
Acrolein: X ¼20%. This compound seems less reactive than the deposit on the catalyst and an unsatisfactory carbon mass balance
others. Propanal (3.2%) was one reaction product formed by the were observed. To determine whether our results are acceptable,
hydrogenation of acrolein, as cited by some authors (Deleplanque we measured the CO and CO2 during catalyst regeneration to
et al., 2010; Tsukuda et al., 2007; Wang et al., 2009); phenol (1.4%) estimate the average amount of carbon deposited over time during
could also be obtained through a Diels–Alder reaction between the reaction. The carbon mass balance including the deposited
acrolein and acetone (2.9%) (Suprun et al. (2009)). Due to the high carbon was determined and calculated after the regeneration step
selectivity for acetaldehyde (21%), it was most likely formed at 350 1C, 2.7 bar (absolute), 0.6 NL h  1 of O2 and 19.4 NL h  1 of
directly from acrolein. However, the most commonly proposed N2 that occurred until the CO and CO2 completely disappeared.
pathway is a retro-aldol condensation of hydroxypropanal to form Acetaldehyde: X¼ 38%. The carbon mass balance was 3% for the
acetaldehyde and formaldehyde (Geng et al., 2012; Laino et al., two experiments. The carbon mass balance including the depos-
2011; Nimlos et al., 2006; Paine et al., 2007). Because 8.7% of ited carbon was 79%, suggesting that, this compound produced
formaldehyde was detected in our case, acrolein might have been a significant amount of coke on the catalyst surface.
hydrated to generate hydroxypropanal before the latter under- Acetone: X¼30%. Only acetaldehyde (10%) was observed on the
went a retro-aldol condensation. In that case, the formaldehyde analytical devices. The carbon mass balance was 10% before the
would have undergone consecutive reactions, in contrast to what regeneration and 20% afterward. The conversion of acetone was
was observed with formaldehyde over the catalyst (only 30% relatively low, implying that a small error in the measurements
conversion). Therefore, acetaldehyde might have been formed created a large error in the selectivities and the mass carbon
only partially by a hydration – retro aldol pathway. balance. However, the determination of the probable pathways
Allylic alcohol: X¼100%. Allylic alcohol was always observed in using the formed products was more important here, even if their
low quantities during the glycerol dehydration; this component quantification is not accurate.
seemed very reactive. Under our conditions, allylic alcohol was Methylglyoxal: X¼100%. This compound gave 5% 2,3-butane-
clearly transformed into propanal (70%) through a rearrangement. dione as the major product and except for the conversion of
Allylic alcohol could dehydrogenate to form acrolein (4%), in glycerol, this reaction was the only process in which this product
contrast to the literature proposing the reverse reaction (Chai was observed, implying it was most likely derived from
et al., 2007; Corma et al., 2008; Deleplanque et al., 2010; Tsukuda methylglyoxal.
et al., 2007). Indeed, allylic alcohol was not detected as a product Acetol: X¼ 100%. Some furan derivatives were observed at low
of acrolein conversion in our experiments or those of Wang et al. concentrations and were not represented in the table; they are
(2009). included with the “others”. Acetone (25%) was present in high
I. Martinuzzi et al. / Chemical Engineering Science 116 (2014) 118–127 125

quantities only in the absence of O2; no acetic acid was detected. were responsible of the 3 methyl 2-cyclopenten-1-one formation
Finally, the presence of propionic acid (3%) implied that acetol in the reaction of glycerol.
underwent a rearrangement to generate propionic acid. Phenol: X ¼45%. The carbon mass balance was 2% before
Formaldehyde: X¼30%. The major products were the same as in regeneration because only acetone (2%) was detected. Only acet-
presence of oxygen; the only difference was the appearance of one might be involved in the formation of phenol. Indeed,
small amounts of acrolein and methacrolein. acetaldehyde was present only with oxygen, indicating that
Acrolein: X¼ 30%. The main products were acetaldehyde (2.6%), acetone formed acetaldehyde in this case. The carbon mass
propanal (2%) and formaldehyde (0.9%), similar to the case with balance after regeneration was 75%. Therefore phenol is a com-
oxygen. The carbon mass balance before regeneration was 9% and pound that leads to carbon deposits.
20% afterward. 1,2-propanediol: X¼100%. The propanal selectivity (21%) was
Propanal: X¼70%. The main products were acetaldehyde (2%), much higher than in the reaction with oxygen, while the acet-
acetone (1.5%), formaldehyde (0.7%) and approximately 4% aldehyde selectivity (10%) appeared much lower, confirming the
unknown products. The carbon mass balance before regeneration possible C–C bond cleavage of propanal to form acetaldehyde in
was 9% and 70% afterward. Coke formed, similar to the case for the presence of oxygen. This result suggests that 1,2-propanediol
acetaldehyde. dehydrated to form propanal directly.
Allylic alcohol: X ¼100%. Propanal (80%) and acrolein (6%) were Methanol: X¼15%. Only formaldehyde was observed. Neither
the major products. Moreover, acetaldehyde (0.1%) and formalde- CO nor CO2 were detected during regeneration, indicating that
hyde (0.1%) were also present as minor products, similar to the carbon deposits were not formed during the reaction. Due to the
reaction with oxygen. However, acetone (0.1%) has been found low methanol conversion, it was not possible to quantify the
only in the reaction without oxygen. Acetone could be transformed products precisely.
into acetaldehyde in the presence of oxygen.
3-methyl 2-cyclopenten-1-one: X ¼100%. The carbon mass bal-
ance was 22% before regeneration and 81% afterward. This product 3.2.3. Reaction mechanism proposition
could be responsible for carbon deposition on the catalyst surface. Some mechanisms for the dehydration of glycerol were pro-
The products of this reaction were comparable to those of the posed in the literature (Chai et al., 2007; Corma et al., 2008;
reaction with oxygen with an additional product (2 cyclopenten-1- Deleplanque et al., 2010; Suprun et al., 2009; Tsukuda et al., 2007;
one) present in a significant quantity (3%). Similar to the reaction Wang et al., 2009). Based on these previous proposals and the
in the presence of oxygen, acetaldehyde (2%) and acetone (2%) above results in the presence and absence of oxygen, we proposed

Fig. 4. Proposed reaction mechanism.


126 I. Martinuzzi et al. / Chemical Engineering Science 116 (2014) 118–127

a reaction mechanism starting from glycerol (Fig. 4). To clarify this acrolein and acetone to explain phenol formation, we propose
scheme, we decided not to represent reactions leading to CO and that phenol must be created using only acetone, based on our
CO2 because almost all of the products lead to these compounds. experiments. Finally, for the new products identified during our
The two products in parenthesis (2,3 hydroxypropanal and 1,2 analysis, we proposed plausible pathways: ethanol must be
propanediol) should be reaction intermediates but were not formed via the hydrogenation of acetaldehyde, cyclopentenes
detected during the experiments. Methylglyoxal was present in must be formed via aldol condensations between acetone and
large amounts during the catalytic conversion of glycerol with acetaldehyde and, even though further analysis is required to
oxygen. Because acetol formed only traces of methylglyoxal, this understand the formation of methacrolein and 2,3-butanedione,
pathway cannot be the only one to create the large amount of we suggest that methacrolein was formed from acrolein and 2,3-
methylglyoxal. Therefore, we proposed a reaction intermediate butanedione was formed from methylglyoxal.
that could be 2,3-hydroxypropanal (also proposed by Corma et al. Moreover, the conversion of these compounds over the catalyst
(2008)), forming most of methylglyoxal from glycerol. highlighted that certain compounds are responsible for depositing
We did not observe allylic alcohol in these experiments; carbon. Understanding of this phenomenon is essential for future
indicating that it was generated directly from glycerol. We industrial applications. Further research is required to understand
observed the presence of 1,3 dioxan-5-ol sometimes during the the deactivation processes.
glycerol reaction, indicating that it was created by cyclizing
glycerol and formaldehyde (Deleplanque et al., 2010).
The proposed mechanism resembles aspects of those of the Nomenclature
literature, but we also found new and significant pathways for the
well-known products: the dehydrogenation of allylic alcohol to
CB [%] Carbon balance
form acrolein, the isomerization of allylic alcohol to form propanal,
ci Represents the number of
the dehydration of 1,2-propanediol to form propanal and the C–C
carbon atoms in molecule of
bond cleavage of propanal to form acetaldehyde. We proposed that
reactant, i
phenol was created by only acetone, based on the products
cj Represents the number of
obtained in the phenol conversion.
carbon atoms in molecule of
Moreover, for the new compounds found in this study, we
product, j
proposed their formation: acetaldehyde hydrogenation led to
GHSV [NL h  1 L  1] (Nm3 s  1 m  3) Gas hourly space velocity
ethanol, as formol led to methanol; methacrolein was formed by
I Reactant
acrolein; 2,3-butanedione was formed by methylglyoxal and the
J Reaction product, j
cyclopentenes were created by an aldol condensation of acetone
ni input [mol s  1] Molar stream of the reactant
and acetaldehyde.
i at the input of the reactor
ni output [mol s  1] Molar stream of the reactant
i at the output of the reactor
4. Conclusion
nj [mol s  1] Molar stream of product j
Qv[NL h  1] (Nm3 s  1) Volumic flow
Glycerol dehydration to form acrolein in gas phase over a solid
SCj [mol%] Carbon selectivity of
catalyst was studied to determine the reaction mechanism. The
product j
products of the reaction and the reactants were analyzed on-line
Vcatalyst [L] (m3) Volume of the catalyst
continuously using a tailor-made gas chromatograph (Agilent GC
Xi [mol%] Conversion of product, i
7890A) equipped with two detectors: a TCD for permanent gases
∑j Sum of products, j
(N2, CO, O2, CH4, CO2) and a FID for condensable gases. The
primary advantage of this analytical set-up is that most of the
compounds are simultaneously analyzed and quantified while
using only one device. The influences of the glycerol concentration
(2, 3, 4 vol%), GHSV (18,000 h  1, 27,000 h  1, 36,000 h  1), the Acknowledgments
temperature (270 1C, 291 1C, 308 1C) and the oxygen concentration
(0, 3, 6 vol%) on the product selectivities were studied. The carbon The financial support from the region of Lorraine in France
mass balances were satisfactory (between 81% and 99%). Thanks to (Agence de Mobilisation Économique) is gratefully acknowledged.
these experiments, several compounds that are not still included We are grateful for the scientific support from Arkema. We also
in the published mechanism, such as ethanol, 2,3-butanedione, thank the members of the mechanical workshop, the electronics
2-methyl 2-cyclopentene-1-one, 3 methyl 2-cyclopentene-1-one workshop and the analytical team in the laboratory for their
have been identified via GC–MS. To determine the reaction professionalism during the pilot tests. We also acknowledge the
mechanism for glycerol dehydration, several products from the KinCom and Greener teams in the laboratory for their scientific
main reaction were passed over the catalyst individually with and and technical support, as well as Mr. Richard Lainé for his help on
without oxygen. These products included acetol, acetaldehyde, the pilot, Mr. Christophe Peloux and Mr. Michel Mercy for their
acetic acid, acetone, acrolein, propanal, formaldehyde, methyl- help with the GC instrument.
glyoxal, allylic alcohol, 3-methyl 2-cyclopenten-1-one, phenol, 1,2-
propanediol, methanol and hydroxypropanal. The conversion of
these compounds and the selectivities of the products revealed References
some non-evident pathways, allowing us to propose a reaction
Arkema investor day, 2012 〈http://www.arkema.com/export/sites/global/.content/med
mechanism for glycerol dehydration. For the well-known pro-
ias/downloads/investorrelations/en/finance/arkema-investor-day-2012-coating-so
ducts, some new pathways are proposed: allylic alcohol dehydro- lutions-va.pdf〉.
genation forms acrolein, allylic alcohol isomerization forms Cai, X.H., Xie, B., 2011. Synthesis of 2-cyclopentene-1-one derivatives by aldol
propanal, the dehydration of 1,2-propanediol forms propanal and condensation. Res. J. Chem. Sci. 1, 120–122.
Chai, S.H., Wang, H.P., Liang, Y., Xu, B.-Q., 2007. Sustainable production of acrolein:
C–C bond cleavage in propanal forms acetaldehyde. Contrary to investigation of solid acid-base catalysts for gas-phase dehydration of glycerol.
the literature reports citing a Diels–Alder reaction between Green Chem. 9, 1130–1136, http://dx.doi.org/10.1039/b702200j.
I. Martinuzzi et al. / Chemical Engineering Science 116 (2014) 118–127 127

Corma, A., Huber, G.W., Sauvanaud, L., O'Connor, P., 2008. Biomass to chemicals: Nimlos, M.R., Blanksby, S.J., Quian, X., Himmel, M.E., Johnson, D.K., 2006. Mechan-
catalytic conversion of glycerol/water mixtures into acrolein, reaction network. isms of glycerol dehydration. J. Phys. Chem. A 110, 6145–6156, http://dx.doi.org/
J. Catal. 257 (163–171), http://dx.doi.org/10.1016/j.jcat.2008.04.016. 10.1021/jp060597q.
Deleplanque, J., Dubois, J.-L., Devaux, J.-F., Ueda, W., 2010. Production of acrolein OECD-FAO Agricultural outlook Website, 2013. 〈http://www.oecd.org/site/oecd-
and acrylic acid through dehydration and oxydehydration of glycerol with faoagriculturaloutlook〉.
mixed oxide catalysts. Catal. Today. 157 (351-358), http://dx.doi.org/10.1016/j. Paine III, J.B., Pithawalla, Y.B., Naworal, J.D., Thomas Jr., C.E., 2007. Carbohydrate
cattod.2010.01.012. pyrolysis mechanisms from isotopic labeling Part 1: the pyrolysis of glycerine:
European Biodiesel Board Website, 2013. 〈http://www.ebb-eu.org/stats.php〉. discovery of competing fragmentation mechanisms affording acetaldehyde and
Geng, Z., Zhang, M., Yu, Y., 2012. Theoretical investigation on pyrolysis mechanism formaldehyde and the implications for carbohydrate pyrolysis. J. Anal. Appl.
of glycerol. Fuel 93, 92–98, http://dx.doi.org/10.1016/j.fuel.2011.08.021.
Pyrolysis 80, 297–311, http://dx.doi.org/10.1016/j.jaap.2007.03.007.
Katryniok, B., Paul, S., Bellière-Baca, V., Rey, P., Dumeignil, F., 2010. Glycerol
Suprun, W., Lutecki, M., Haber, T., Papp, H., 2009. Acidic catalysts for the
dehydration to acrolein in the context of new uses of glycerol. Green Chem.
dehydration of glycerol: activity and deactivation. J. Mol. Catal. A: Chem. 309
12, 2079–2098, http://dx.doi.org/10.1039/c0gc00307g.
(71–78), http://dx.doi.org/10.1016/j.molcata.2009.04.017.
Laino, T., Tuma, C., Curioni, A., Jochnowitz, E., Stolz, S.A., 2011. Revisited picture of
Tsukuda, E., Sato, S., Takahashi, R., Sodesawa, T., 2007. Production of acrolein from
the mechanism of glycerol dehydration. J. Phys. Chem. A 115, 3592–3595, http:
//dx.doi.org/10.1021/jp201078e. glycerol over silica-supported heteropoly acids. Catal. Commun. 8, 1349–1353,
Lauriol-Garbey, P., Postole, G., Loridant, S., Auroux, A., Belliere-Baca, V., Rey, P., Millet, J. http://dx.doi.org/10.1016/j.catcom.2006.12.006.
M.M., 2011. Acid-base properties of niobium-zirconium mixed oxide catalysts for Wang, F., Dubois, J.-L., Ueda, W., 2009. Catalytic dehydration of glycerol over
glycerol dehydration by calorimetric and catalytic investigation. Appl. Catal. B – vanadium phosphate oxides in the presence of molecular oxygen. J. Catal. 268,
Environ. 106, 94–102, http://dx.doi.org/10.1016/j.apcatb.2011.05.011. 260–267, http://dx.doi.org/10.1016/j.jcat.2009.09.024.

You might also like