You are on page 1of 19

Chemical Engineering Journal 429 (2022) 132278

Contents lists available at ScienceDirect

Chemical Engineering Journal


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

Kinetic study of glycerol steam reforming catalyzed by a Ni-promoted


metallurgical residue
Alex Desgagnés, Maria C. Iliuta *
Department of Chemical Engineering, Université Laval, Québec, QC G1V0A6, Canada

A R T I C L E I N F O A B S T R A C T

Keywords: The kinetics of the glycerol steam reforming reaction catalyzed by a nickel-promoted metallurgical residue was
Glycerol steam reforming studied in a fixed-bed reactor at atmospheric pressure. The temperature was varied between 480 ◦ C and 580 ◦ C.
Ni-based catalyst The experimental reaction rates were correlated via different heterogeneous kinetic models based on Langmuir-
Metallurgical waste valorization
Hinshelwood-Hougen-Watson mechanisms that consider single- or dual-site adsorption of the reactants. The
Kinetic modeling
Langmuir-Hinshelwood-Hougen-Watson
mechanism that provided the most mathematically accurate and thermodynamically consistent results is initi­
reaction mechanism ated by dissociative adsorption of glycerol, which cause the breaking of two C–C bonds in the glycerol molecule
upon adsorption. This model claims that the dehydrogenation of an adsorbed glycerol intermediate is the rate-
determining step of the global reaction. A macroscopic analysis of the glycerol to gas consumption rates also
validated the activation energies identified by the successful heterogeneous model by fitting the experimental
data to a straightforward power law model. The activation energy was found to be 66.1 kJ⋅mol− 1 with a partial
order with respect to glycerol of 0.63.

C3 H8 O3(g) + 3H2 O(g) ↔ 7H2(g) + 3CO2 (g) r


ΔH298K = 128kJ∙mol− 1
(R1)

Many authors agree that this overall reaction can be broken down
1. Introduction into an endothermic glycerol pyrolysis step (R2) followed by the
water–gas shift reaction (R3) which converts carbon monoxide into
In the current environmental crisis in which the world finds itself, the additional hydrogen due to the excess steam present in this process
concern to reduce our energy dependence on fossil fuels by exploiting [5–9].
more ecological and sustainable alternatives is becoming increasingly C3 H8 O3(g) ↔ 3CO(g) + 4H2(g) r
ΔH298K = 251kJ∙mol− 1
(R2)
important. The hydrogen fuel (H2) will most certainly play a pivotal role
in the global energy transition towards green and less polluting re­ CO(g) + H2 O(g) ↔ CO2 (g) + H2 (g) r
ΔH298K = − 41kJ∙mol− 1
(R3)
sources due to some outstanding assets as an energy carrier [1].
Nevertheless, the global hydrogen production is currently being almost Inevitably, undesirable products such as methane, which reduces the
entirely provided by the reforming of fossil methane [2]. Consequently, hydrogen yield, and solid elemental carbon, that can accelerate catalyst
the benefits of hydrogen as a low- (or no–) carbon energy carrier become deactivation, are generated through secondary reactions in this process.
environmentally justified only if it is generated from sustainable re­ A variety of condensable organic compounds including formaldehyde,
sources [3] such as glycerol, the main by-product in the production of acrolein, allyl alcohol, and hydroxyacetone can also be derived from the
biodiesel from the transesterification process of animal or vegetable fats. decomposition of glycerol in the GSR process [5].
In fact, the growing worldwide production of biodiesel is generating Due to the wide range of possible reaction paths in this process, the
staggering quantities of glycerol that far exceed market demand. The selection of a suitable catalyst capable of converting a maximum amount
production of hydrogen by glycerol steam reforming (GSR) is therefore of glycerol while maintaining a high yield of hydrogen over a long
an interesting avenue to recover this residual organic material while period of time is paramount. Nickel-based catalysts are widely reported
reducing the net production cost of biodiesel [4]. in the literature since this metal demonstrates high activity in the GSR
The stoichiometry of the overall GSR reaction claims that 7 mol of reaction [5,6,8,10–12], which is mainly attributed to its good ability to
hydrogen can be theoretically produced per mole of glycerol (R1). cleave the C–H, O–H and C–C bonds present in glycerol [13]. Nickel

* Corresponding author.
E-mail address: maria-cornelia.iliuta@gch.ulaval.ca (M.C. Iliuta).

https://doi.org/10.1016/j.cej.2021.132278
Received 4 May 2021; Received in revised form 18 August 2021; Accepted 1 September 2021
Available online 8 September 2021
1385-8947/© 2021 Elsevier B.V. All rights reserved.
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

Nomenclature R ideal gas constant, 8.314 J⋅mol− 1⋅K− 1


R2 regression coefficient, dimensionless
A pre-exponential factor of the Arrhenius equation (eq. (5)), T temperature, K
variable units Ts temperature at the surface of the catalyst, K
dp catalyst particle diameter, m Tb temperature in the bulk gas phase, K
DAB gas diffusivity, m2⋅s− 1 U gas superficial velocity, m⋅s− 1
De effective diffusivity, m2⋅s− 1 W catalyst weight, g
CC3 H8 O3 ,s glycerol concentration at the surface of the catalyst, X active site, dimensionless
mol⋅m− 3 XG glycerol conversion to gaseous products, dimensionless
CC3 H8 O3 ,b glycerol concentration in the bulk gas phase, mol⋅m− 3 xC3 H8 O3 glycerol feed molar fraction, dimensionless
Cpmix isobaric molar heat capacity of gas mixture, J⋅mol− 1⋅K− 1 YH2 hydrogen yield, dimensionless
Ea activation energy, J⋅mol− 1 Yl yield of liquid product l , dimensionless
F molar flow rate, mol⋅min− 1 yi,k molar fraction associated with compound i, and
F1− ϕ critical value of the Fisher distribution experimental condition k, dimensionless
G gas phase mass velocity, kg⋅m− 2⋅s− 1
Subscripts
GHSV gas hourly space velocity, std L⋅gcat-1⋅h− 1
in at the reactor inlet
h heat transfer coefficient, J⋅m− 2⋅s− 1⋅K− 1
out at the reactor outlet
k rate constant of the Arrhenius equation (eq. (5)), variable
PL associated to power law model
units
exp experimental
k’ pseudo rate constant given in the power law model (eq.
cal calculated
(10) and eq. (11)), mol⋅min− 1⋅gcat-1⋅atm-α
kc mass transfer coefficient, m⋅s− 1 Greek letters
krxj rate constant associated to heterogeneous kinetic model j, α partial order of reaction with respect to glycerol,
variable units dimensionless
KAi.j apparent adsorption constant associated with reactant i ϕ confidence interval, dimensionless
and heterogeneous kinetic model j (M1-M12), variable units ΔHr298K standard enthalpy of reaction, J⋅mol− 1
KDi.j apparent desorption constant associated with product i and ΔHAi.j enthalpy of adsorption associated with reactant i and
heterogeneous kinetic model j (M1-M12), variable units heterogeneous kinetic model j (M1-M12), J⋅mol− 1
K’Ai.j actual adsorption constant associated with reactant i ΔHDi.j enthalpy of desorption associated with product i and
derived in the development of the mechanism of heterogeneous kinetic model j (M1-M12), J⋅mol− 1
heterogeneous kinetic model j (M1-M12), variable units ΔSAi.j entropy of adsorption associated with reactant i and
K’Di.j actual desorption constant associated with product i heterogeneous kinetic model j (M1-M12), J⋅mol− 1⋅K− 1
derived in the development of the mechanism of ΔSDi.j entropy of desorption associated with product i and
heterogeneous kinetic model j (M1-M12), variable units heterogeneous kinetic model j (M1-M12), J⋅mol− 1⋅K− 1
L0 liquid flow rate of the reactant feed mixture, std mL⋅min− 1 λe effective thermal conductivity of catalyst, J⋅K− 1⋅m− 1⋅s− 1

n order of reaction, dimensionless μ dynamic viscosity, Pa⋅s


OF objective function calculated as (eq. (6)), dimensionless ρb catalyst bulk density, kgcat⋅m− 3 of reactor
p average partial pressure, atm ρc catalyst density, kgcat⋅m− 3 of catalyst
P total pressure in the reactor, atm ρg gas density, kg⋅m− 3
Q0 total gas flow rate introduced in the reactor, std mL⋅min− 1 υ degree of freedom, dimensionless
r consumption or formation rate, mol⋅min− 1⋅gcat-1
σ2 variance, variable units
rj glycerol to gas consumption rate calculated with
ωi,k weight factor associated with compound i, and
heterogeneous kinetic model j, mol⋅min− 1⋅gcat-1
experimental condition k, dimensionless
ri,k consumption of formation rates associated with compound
i, and experimental condition k, mol⋅min− 1⋅gcat-1

competes with noble metals such as Rh, Ru, Pt or Pd, which are also (hereinafter labeled as UGSO, abbreviation for Upgraded Slag Oxide),
highly active for GSR reaction and selective for hydrogen. However, Ni- which is generated in powder form at a rate of tens of thousands of tons
based catalysts have a head start over these precious metal-based cata­ per year by the Rio Tinto Fer & Titane metallurgical complex in Sorel-
lysts for use in industrial applications, as they are much more accessible Tracy (Quebec, Canada), is currently stored in large mine waste sites
and less expensive. The choice of the support is also crucial since it without marketable use. Although slightly variable, the composition of
dictates the dispersion of the active phase, which is directly related to UGSO includes mostly iron (Fe), aluminum (Al) and magnesium (Mg)
the catalytic activity. The surface properties of catalytic supports or crystallized as spinel mixed oxides (MgFe2O4 and FeAl2O4) or free per­
promoters can mitigate the defects observed in some active metals. In iclase (MgO) [5]. By promoting it with Ni, UGSO can act as an effective
the case of nickel, the presence of basic promoters (e.g., MgO, CaO) catalyst in the GSR process, offering a promising way to valorize the
[10,14] can decrease the concentration of acidic sites conducive to coke valuable elements it contains in a sustainable development perspective.
formation, while its vulnerability to sintering [14] can be tempered by When compared to an alumina-supported Ni-based commercial catalyst,
using lanthanoids oxides-based supports such as CeO2 [15] or La2O3 it has been determined that the presence of active Mg and Fe oxide
[16]. species in Ni-UGSO catalyst could favor the water gas shift reaction as
In our previous studies [5,10,17,18], interesting conclusions have well as coke gasification [10], thus increasing the hydrogen yield and
been reported regarding the use of a Fe-Mg bearing metallurgical res­ long-term stability. In addition, a study on the effect of active metal in
idue as a catalytic support and promoter in the GSR process. This residue UGSO-based catalysts revealed that the very slightly lower performance

2
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

Table 1
UGSO elemental composition (wt. %) and textural properties of fresh UGSO, calcined UGSO and 5 % Ni-UGSO calcined catalyst (Data retrieved from [5,10]).
UGSO (wt. %) Fe Mg Al Ca Mn V Ti

31.26 17.49 5.35 1.05 0.98 0.92 0.61


Catalyst Specific surface area Average pore diameter Average particle size
2
SBET (m /g) Dp (nm) (nm)
Fresh UGSO 29.8 17.2 202
Calcined UGSO 12.0 40.5 499
5 % Ni-UGSO 9.9 52.8 604

of a 5 % Ni-UGSO catalyst compared to a 1 % Rh-UGSO catalyst (both in ICP-MS analysis) as well as its textural properties, alongside those of the
mass percentages) could not economically justify the use of a rhodium- calcined 5 % Ni-UGSO catalyst, are shown in Table 1. It can be observed
based catalyst given the low price of nickel compared to this noble metal that the heat treatment carried out during calcination is the main cause
[18]. Without neglecting the effect of the parameters studied so far in of the decrease in the specific surface area of UGSO compared to its
the GSR process using this innovative and eco-friendly catalytic mate­ initial state. The Ni incorporation has a lesser impact, but still contrib­
rial, the reaction mechanism and kinetics component have not yet been utes to slightly decrease the specific surface area, and to increase the size
developed. of the catalyst particles, which is probably due to the blocking of the
The study of reaction and kinetic mechanics is even more important UGSO pores by nickel particles [5].
since it represents a further step towards the design and modeling of
reactors that could be industrially implemented. In recent years, only a 2.2. Experimental setup & procedure
few studies have been published in the literature concerning the kinetics
of the GSR process. Although some authors have come to similar con­ The experimental laboratory set-up used to perform the GSR cata­
clusions regarding the reaction mechanisms that best reflect reality, lytic tests is schematized and described in detail elsewhere [10].
others have reached quite different outcomes [19]. These discrepancies Concisely, 0.5 g of catalyst (weight used for each experiment) was put on
are undoubtedly the result of differences in the properties of the studied a grid covered with quartz wool located at the mid-length of a 40 cm
catalysts that lead to observable behavioural changes in surface reactive long stainless-steel tubular reactor (internal diameter = 11 mm). Prior to
species, and thus to different findings in the kinetic studies. Neverthe­ the GSR reaction test, the catalyst was reduced under a 22 % vol. H2/Ar
less, the usefulness and significance of a kinetic study is even more atmosphere at 650 ◦ C for 1.5 h. Argon (carrier gas) and hydrogen
justified when a new catalytic material is investigated, as is the case in (reduction gas) flow rates were adjusted by mass flow controllers, while
the present work. The validity of the power law empirical model has temperature of the catalytic bed was controlled by an electric tubular
been revealed by many authors [6–8,20–25] for estimating the kinetic furnace combined with a temperature controller and a thermocouple.
constants and activation energies of the overall GSR process reaction. Afterwards, an argon purge was carried out for 30 min to flush the
However, little attention has been paid to the mechanistic consider­ hydrogen from the lining and allow the reactor to cool down to the
ations of the reaction. desired reaction temperature (480–580 ◦ C). To begin the GSR reaction,
In this work, an in-depth experimental and modelling study of the the liquid phase reactant solution (99.5 % pure glycerol and water) was
kinetics of the GSR process over a Ni-based catalyst supported on the fed to the reactor via a high-performance liquid chromatography pump.
UGSO metallurgical residue in a fixed-bed reactor was carried out. A water-to-glycerol feed molar ratio of 9 (steam-to-carbon feed molar
Different reaction pathways were proposed to develop new or literature- ratio of 3) was maintained for all GSR experiments. The upper part of the
inspired expressions of reaction rates based on Langmuir-Hinshelwood- reactor (the first 18 cm above the catalyst bed) was considered long
Hougen-Watson heterogeneous kinetic mechanisms. A macroscopic enough to evaporate all reactants before they reached the catalyst bed.
analysis was also performed by applying a power law model to correlate Unless stated otherwise, 52 std mL⋅min− 1 (at 298.15 K and 1 atm, i.e., at
the experimental rates of glycerol consumption and generation of the std conditions) of Ar was mixed in the reactor with the vaporized
main gaseous products. The reaction mechanisms developed in this reactant feed mixture (liquid flow rate L0 varied from 0.05 to 0.11 std
study were adjusted to the experimental data in order to obtain the ki­ mL⋅min− 1) during catalytic tests. At the exit of the reactor, the gas flow
netic parameters associated with each model, as well as to elucidate passed through a cold trap used to capture the condensable products of
which of them best reflects reality. The findings from this research will the reaction. A mass flow meter measured the total gas flow rate exiting
benefit the design and operation of GSR reactors that would be operated the reactor and the concentration of its components was recorded on-
on a larger scale. line by a microGC (Agilent Technologies, 3000A) that analyzed sam­
ples taken automatically every 15 min. Liquid product analysis was
2. Materials and methods conducted using a high-performance liquid chromatograph (HPLC;
Shimadzu SPD-M20A HPLC) equipped with a refractive index detector
2.1. Catalyst preparation (RID).
Glycerol conversion to gaseous products (XG ) and hydrogen yield
The UGSO residue was promoted with nickel following a solid-state (YH2 ) were determined to quantify and compare the different catalytic
impregnation method described in detail elsewhere [5,10]. Briefly, activity performances observed during the study. In the present work,
UGSO powder (previously ground and sieved to particle size less than the values of these parameters were obtained after 2.5 h of reaction,
75 µm) was mixed and further ground by hand using an agate mortar when steady-state conditions were well established. Glycerol conversion
with the Ni precursor (Ni(NO3)2⋅6H2O, Fisher, 99.9 %) at ambient to gaseous products (eq. (1)) was calculated based on the carbon balance
temperature for 10 to 15 min. A few drops of acetone were also added to of the gaseous products detected by the microGC:
the mixture to ensure good homogeneity. The resulting solid was then
calcined at 600 ◦ C under air atmosphere for 2 h. The 5 % Ni-UGSO XG =
FCO2 + FCO + FCH4
(1)
(weight %) catalyst was used throughout the whole kinetic study, as it 3∙FC3 H8 O3 in
has proven to be the most efficient (higher glycerol conversion and where F is the molar flow rate of the different compounds.
hydrogen yield) among other catalysts with different nickel contents The hydrogen yield (eq. (2)) was calculated based on the theoretical
[5]. The elemental chemical composition of UGSO residue (obtained by and maximum amount of H2 that can be obtained by steam reforming of

3
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

Table 2 Table 3
Assessment of the absence of internal mass and heat transfer effects using the Assessment of the absence of external heat and mass transfer effects using Mears
Weisz-Prater criterion. criterion.
Transfer Massa, b Heatb, c Transfer Massa Heatb
phenomenon phenomenon
⃒( ⃒ ⃒ ⃒
Criterion (rC3 H8 O3 exp )ρc dp 2 ⃒ − ΔHr )d 2 ρ (r ⃒ Criterion (− rC3 H8 O3 exp )dp ρb n ⃒( − ΔHrxn )(− rC3 H8 O3 exp )dp ρb Ea ⃒
⃒ 298K p b C3 H8 O3 exp )Ea ⃒ < 0.15 ⃒⃒ ⃒ < 0.15
<1 ⃒ ⃒ < 0.1 2kc CC3 H8 O3 ,b 2
4De CC3 H8 O3 ,s ⃒ λe RTs 2 ⃒ 2hRTb ⃒
Calculated value 0.0033 Calculated value 0.0004 0.0002
< 1∙10− 4
a
a
De = 0.1 × DAB where DAB is calculated using equation developed by Fuller, The mass transfer coefficient kc is calculated from kc = JD ∙U∙Sc2/3 , where
( )
Shettler, and Giddings [29]. 1 0.765 0.365 μ ρ ∙U∙dp
JD = + , Sc = , Re = g μ .
b
The highest (− rC3 H8 O3 exp ) value was used to estimate both criterions in the ε Re0.82 Re0.386 ρg ∙DAB
b
worst conditions. The heat transfer coefficient h is calculated based on the J-factor Chilton and
c
λe = 10 × λg h Cpmix ∙μ
Colburn analogy: JH = JD = ∙Pr2/3 wherePr =
Cpmix ∙G λg
1 mol of glycerol according to the stoichiometry of the overall reaction
(R1): 3. Experimental results
FH 2
Y H2 = (2) 3.1. Mass and heat transfer effects
7∙FC3 H8 O3 in

The kinetic regime in which the catalytic tests were performed Heterogeneously catalyzed chemical reactions such as GSR involving
favored the conversion of glycerol to condensable products at different gas phase reactants and a solid catalyst occur through a three-phase
extents depending on the selected conditions. The yield of each of the mass transfer mechanism: (i) external diffusion of the reactants in the
detected liquid products (eq. (3)) was calculated based on the carbon gas film at the surface of the catalyst, (ii) internal diffusion of the re­
balance of the liquid products detected by HPLC: actants through the pores of the catalyst followed by the chemical re­
action on the active sites, and (iii) diffusion of the products from the
moles of C in liquid product ↕
Y↕ = (3) catalyst surface to the bulk gas phase. In a kinetic study, the influence of
3∙FC3 H8 O3 in
these mass transfer processes should be minimized to identify the
where l can be glycerol, acrolein, allyl alcohol, hydroxyacetone or intrinsic kinetic parameters of the reaction. In catalytic reactions, the
formaldehyde. particle size of the catalyst is among the most significant contributors to
The reaction rates were estimated based on the conversion to gaseous the internal mass transfer resistance. For particles smaller than 1 mm, it
products by differential analysis of the plug-flow reactor equation [26]. has been reported that the effects of internal diffusion on catalyzed gas-
Since the weight of catalyst used is very small (as is the height of the phase reactions are not significant [26,28]. In the present study, since
bed), the differential reactor model assumes that the reactant concen­ the catalyst particles were all smaller than 75 µm, the resistance due to
trations vary linearly in the bed. Therefore, a reaction rate that is internal diffusion could be expected to be negligible compared to the
spatially uniform in the bed and based on average concentrations is surface reaction. The Weisz-Prater criteria calculated from the experi­
considered. For a given conversion and reactant flow rate, the steady- mental data validated this assumption that no significant concentration
state mole balance on glycerol is similar to that for continuous-stirred gradient exists within the catalyst pellets [27]. For the same reason, a
tank reactors, and is given as follows [27]: zero-temperature gradient was expected within the particles due to their
small size, as evidenced by compliance with the calculated criterion [8].
[Flow ratein ] − [Flow rateout ] − [Rate of reaction] The criterion employed to evaluate the internal effects of mass and heat
= [Rate of accumulation] (4.1) transfer is presented in Table 2.
To ensure the absence of external diffusion effects, a preliminary test
[ ] [ ]
FC3 H8 O3 in − FC3 H8 O3 out was carried out to confirm that the total gas velocity did not influence
[(
Rate of reaction
) ] the glycerol conversion (and thus the reaction rates) achieved in steady

Weight of catalyst
∙(Weight of catalyst) state [30]. For the same W/FC3 H8 O3 in ratio, where W and FC3 H8 O3 in refer to
the catalyst weight and the glycerol molar feed rate, respectively, the
=0 (4.2)
conversion difference was within the experimental error range (less than
[ ] [ ]
FC3 H8 O3 in − FC3 H8 O3 out − rC3 H8 O3 ∙dW = 0, where W is the catalyst 0.5 %) between two tests performed with different argon flow rates (52
weight. and 190 std mL⋅min− 1), and therefore, different total gas flow rates Q0
By substituting FC3 H8 O3 out = FC3 H8 O3 in − FC3 H8 O3 in ∙dXG , we find: (166 and 304 std mL⋅min− 1). Since these tests were performed using the
highest temperature and the lowest W/FC3 H8 O3 in ratio values covered in
r C 3 H8 O3 =
FC3 H8 O3 in ∙dX G
(4.3) the kinetic study (580 ◦ C and 17.866 g⋅h⋅mol− 1), it was agreed that the
dW effect of external diffusion was ultimately negligible for all other tests in
With the following assumption: dX G
≈ ΔXG
when ΔW→0, we find the this work. Based on empirical data, the Mears criterion is another way to
dW
investigate whether the effect of external mass transfer from the bulk gas
ΔW
experimental glycerol to gas consumption rate (rC3 H8 O3 exp ) defined as
phase to the catalyst surface can be neglected. Similarly, Mears also
follows (eq. (4.4)):
proposes that the bulk fluid temperature will be essentially the same as
XG at the catalyst surface if the criterion associated with heat transfer is
rC3 H8 O3 exp = ( ) (4.4)
W met. These homologous versions of Mears’ criterion are presented in
FC3 H8 O3 in
Table 3.

4
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

Fig. 1. Experimental glycerol conversion to gaseous products (XG ) and hydrogen yield (YH2 ) as a function of reaction temperature and liquid flow rate of feed
reactants (L0 ). The total conversion of glycerol considers the combined effect of thermal decomposition (XG,thermal ) and catalytic conversion (XG,cat ). For each L0 is
associated a standard feed gas flow (Q0 ), a space–time (W/FC3 H8 O3 in ), a gas hourly space velocity (GHSV) and a glycerol feed molar fraction (xC3 H8 O3 in ).

3.2. GSR kinetic data each detected liquid by-products for the investigated temperatures,
while the reaction scheme shown in Fig. 2B illustrates the possible
The experimental data of the kinetics study are presented in Fig. 1. routes taken to achieve the formation of these compounds. As expected,
Reaction temperatures ranging from 480 ◦ C to 580 ◦ C were tested for the yield of liquid products is higher when the reaction temperature is
different space time values defined as the W/FC3 H8 O3 in ratio (g⋅h⋅mol− 1). lower since the energy supplied to the system is not sufficient to convert
These were adjusted by varying the liquid feed rate of the glycerol and glycerol completely to the desired gaseous products (i.e., H2 and CO2). It
water solution. As expected, at a given temperature, glycerol conversion can also be observed that most of the glycerol that is not converted into
to gas and hydrogen yield increased significantly alongside the W/ CO2, CO or CH4 is not transformed into other liquid products either,
FC3 H8 O3 in ratio. Similarly, for a constant value of L0 , the increase in re­ since it is the primary component of the liquid condensate.
action temperature resulted in increases in XG and hydrogen yield, As shown in Fig. 3, the experimental rate of reaction is influenced by
confirming that the endothermic (R1) reaction is indeed prevailing in the reaction temperature and the partial pressure of glycerol. For a
the reaction system. Given the temperatures of the reforming process similar variation in pC3 H8 O3 , the increase in reaction rate is more pro­
and the high thermal instability of glycerol, some decomposition to nounced at higher reaction temperatures. At first sight, the experimental
gaseous products is likely to occur before reaching the catalytic bed consumption rates do not appear to align linearly with glycerol partial
[31,32]. Therefore, the total XG includes the combined effect of thermal pressure, as evidenced by the y-intercept of the trend curves, which
decomposition and catalytic conversion. From Fig. 1, it can be noticed deviates progressively from zero as the temperature increases. This
that the formation of gaseous products via the thermal route is naturally observation allows us to exclude the possibility that the reaction system
amplified as the temperature increases. However, the latter remains investigated here is of order 1 with respect to glycerol, as it was the case
modest compared to the catalytic contribution. Glycerol conversion to for other authors who studied the GSR reaction [6,7,23]. Moreover, we
gaseous products ranged from 23.51 % to 98.99 %, while hydrogen can assume that this partial order is greater than zero since the reaction
yields varied from 17.17 % to 77.07 % across the range of flow rates and rate increases with the partial pressure of glycerol. The methods used for
temperatures tested in the study. For the test performed at 580 ◦ C with a the identification of the order of the reaction as well as the other kinetic
liquid flow rate of 0.05 std mL⋅min− 1, the hydrogen yield obtained parameters are addressed in detail in the following section.
contradicted the general trend since it was slightly lower than that ob­
tained at 555 ◦ C. The analysis of the reaction products led to the hy­ 4. Kinetic modeling
pothesis that the reverse water–gas shift reaction could be favored above
555 ◦ C, which could results in a direct decrease of the hydrogen yield, 4.1. Estimation of kinetic parameters methodology
and an increase in CO concentration [33]. However, this phenomenon
was not observed in the tests performed with lower space times. The establishment of an appropriate kinetic model requires the
Tests carried out at temperatures below 555 ◦ C led to the formation identification of the parameters that most accurately fit the experi­
of liquid products including glycerol (unconverted), acrolein, allyl mental data. Most often, in heterogeneous kinetic modeling, these pa­
alcohol, hydroxyacetone, and formaldehyde. Other undetectable prod­ rameters include rate or adsorption constants that are valid only for
ucts (such as glyceraldehyde, 3-hydroxypropionaldehyde or acetalde­ specific reaction conditions. In most laboratory and industrial reactions,
hyde) are possibly present in trace amounts. Fig. 2A reports the yields of the rate constants are assumed to be a function of temperature only [27]

5
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

Fig. 2. Liquid by-products analysis: A) Yield of liquid products for the investigated temperatures (GHSV = 12.444 std L⋅gcat-1⋅h− 1) and B) Possible reaction pathways
in glycerol steam reforming.

6
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

Fig. 3. Experimental glycerol to gas consumption rate as a function of glycerol partial pressure at different temperatures. Trend lines have been added to highlight
the lack of proportionality of the relationship near the origin.

and their dependence can be defined by the Arrhenius equation (eq. (5)): desorption constants), the program minimized an OF function value
separately for each heterogeneous model. The more the value of the
(5)
− Ea
k = Ae function OF is minimized, the higher the regression coefficient R2 should
RT

where k is the rate constant, A is the pre-exponential factor, Ea is the be on the parity plot of the calculated and experimental reaction rates.
activation energy of the reaction, and R is the ideal gas constant. Other As a primary criterion, models that had a R2 ≤ 0.98 were rejected. The
constants, such as a reaction order, can be identified through modelling validity of the mathematical solutions respecting this first criterion was
as is the case in the power law model. The glycerol reforming rate as well then verified by a residual analysis. Solutions whose relative residuals
as the formation rates of the main gas products were first fitted to a (calculated according to equation (eq. (8))) showed a trend or were not
power law model. The estimation of the model parameters (i.e., reaction normally distributed around a mean of zero were deemed invalid and
orders and rate constants) was performed using a non-linear regression were discarded.
program to minimize an objective function (OF) inspired by the ( )
rexp − rcal
Levenberg-Marquardt algorithm, which is based on the sum of the Relative residuals = (8)
rexp
squares of the relative residues (eq. (6)):
( )2 For the models that were able to meet the two criteria listed above,
∑n ∑ q rexpi,k − rcali,k an analysis of variance (ANOVA) was performed using Fisher’s law test
OF = ωi,k ∙ (6) on the residuals of the models. The ANOVA was performed to compare
i=1 rcali,k
two models, model A having a higher sum of squares of residuals than
k=1

where rexp and rcal are, respectively, the experimental and calculated model B (SSRB < SSRA ). The Fisher’s law test states that the improve­
consumption or formation rates associated with n different compounds i ment brought by model B is significant with a percentage of confidence
(C3H8O3, H2, CO2, CO, and CH4, excluding water, which was not of 100⋅(1 − ϕ) if the following condition is met (eq. (9)):
quantified at the reactor’s outlet) and q experimental conditions k. The
SSRA /zA σ2 A
weight factors (ωi,k ) are included to give equal importance to the rates of FA− B = = > F1− ϕ (υA , υB ) (9)
each product. These vary in inverse proportion to the molar fraction of SSRB /zB σ2 B
each compound at the given conditions (yi,k ), and are calculated as fol­ ∑
where SSR = qk=1 (rexp,k − rcal,k )2 ,υ is the degree of freedom, z is the
lows [34,35] (eq. (7)): number of parameters to determine in the corresponding model, σ2 is the
1 variance, ϕ = 0.1 and the F-value was calculated using tabulated data
ωi,k = (7) for Fisher distribution [36].
yi,k
Subsequent analysis of the remaining solutions was then performed
For modeling heterogeneous kinetic expressions, the objective to ensure that they were thermodynamically realistic. Indeed, the so­
function to be minimized is essentially the same as for the power law lutions provided by the resolution algorithm were acquired on a math­
model but does not include the product formation rates since the ematical basis. It is therefore essential to ensure that the rate and
developed expressions describe the overall reaction rate, i.e., the rate at adsorption constants of the retained models have positive values and
which glycerol is reformed to gaseous products based on the rate- present logical trends in relation to the temperature variation. For
determining steps. By varying the values of the kinetic parameters instance, the rate constants should always increase with temperature
present in the heterogeneous rate expressions (rate, adsorption, and according to Arrhenius’ law in contrast to the adsorption equilibrium

7
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

constants which should decrease as stated by Van’t Hoff’s relation due to nickel-based catalyst, a dual-site molecular adsorption of steam and
the exothermic nature of adsorption. The selection of models that best glycerol with a rate-determining surface reaction was the most adequate
reflect reality is detailed and justified in the Discussion section. representation of the reaction mechanism. In comparison, dissociative
adsorption of steam was considered more adequate in the work of Cheng
et al.[8] who studied the reaction catalyzed by nickel on an alumina
4.2. Proposed kinetic models support. The results supported that dissociative adsorption of water
occurred preferentially on acidic sites, while molecular adsorption of
The parameters and expressions in this kinetic study were estab­ glycerol occurred on basic sites, also present on the catalyst surface.
lished from experimental data obtained or calculated in the well- Sundari et al. [7] and Vaidya et al. [37] who used Ru-based catalysts,
established steady state only. Therefore, under the assumption that hypothesized a single-site molecular adsorption of glycerol followed by
the partial pressures of the compounds vary linearly over the length of its reaction with water to form an intermediate complex. The decom­
the catalyst bed, the average partial pressures in the catalytic bed were position of this complex was considered the rate-determining step in the
considered to adjust the rate law expressions for all the proposed kinetic reaction.
models. In this study, mechanisms involving the reversible adsorption of
water and glycerol, surface reactions as well as the desorption of prod­
4.2.1. Power law model ucts were developed and derived into Langmuir-Hinshelwood-Hougen-
The experimental glycerol to gas consumption rates were first Watson-type kinetic expressions to correlate the experimental data.
adjusted to a power law model. Due to the large stoichiometric excess of Molecular or dissociative adsorption of the reactants on identical single
water over glycerol in the feed solution for all tests carried out in this or dual sites were considered to establish different reaction mechanisms.
study (steam/glycerol molar ratio of 9), a zero-order reaction with A partial first order with respect to each respective reactant was
respect to water was assumed. The power law model can therefore be considered for all the elementary steps involved in each mechanism. In
defined as follows (eq. (10)): more than three quarters of heterogeneous reactions that are not limited
by mass transfer effects, the rate-determining step is a surface reaction
rPL− = k’ ∙pαC3 H8 O3 (10)
C 3 H 8 O3
[27]. In light of this statement, in this work, eight different rate law
where rPL− C3 H8 O3 is the calculated glycerol to gas consumption rate expressions were developed assuming that surface bimolecular reactions
(mol⋅min− 1⋅gcat-1), k’ is the pseudo rate constant (mol⋅min− 1⋅gcat-1⋅atm- involving the reactants (or intermediates of their decomposition) were
α
) of the reaction, pC3 H8 O3 is glycerol partial pressure (atm) and α is the the limiting steps (M1-M8). In the mechanism developments detailed
reaction order. At first sight, a α-value between 0 and 1 was expected below, the active sites are represented by X , while the main and sec­
after observing the results in Fig. 3. An analogous power law expression ondary gaseous products are labeled in red and orange, respectively. In
was also used to model the experimental formation rates of the main models where dual-site adsorption is considered, X1 are the active sites
gaseous products i (H2, CO2, CO, and CH4) (eq. (11)). for glycerol while X2 represent the active sites for water. The presence of
CH4 and CO in non-negligible quantities in the gaseous products was
rPL− i = k’ ∙pαC3 H8 O3 (11) considered by the presence of secondary reaction pathways in the
development of all mechanisms. We also highlight that it is possible to
4.2.2. Heterogeneous kinetic models retrieve the global reaction of GSR (R1) by doing the stoichiometric
To date, the glycerol steam reforming reaction has been studied balance for each elementary step.
using several different catalytic systems, which is probably the reason Model (M1): Single-site molecular adsorption of glycerol and
why no consensus has yet been reached on which reaction mechanism dissociative adsorption of steam
actually prevails. For Adesina et al. [22], who used a bimetallic cobalt/

8
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

Model (M2): Single-site dissociative adsorption of glycerol and


steam

Model (M3): Single-site dissociative adsorption of glycerol and mo­


lecular adsorption of steam

9
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

Model (M4): Single-site molecular adsorption of glycerol and steam

Model (M5): Dual-site molecular adsorption of glycerol and disso­


ciative adsorption of steam

10
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

Model (M6): Dual-site dissociative adsorption of glycerol and steam

Model (M7): Dual-site dissociative adsorption of glycerol and mo­


lecular adsorption of steam

11
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

Model (M8): Dual-site molecular adsorption of glycerol and steam (M9) suggests an elementary first step of dehydrogenation of glycerol to

form adsorbed CHOH-COH-CH2OH, followed by its decomposition


initiated by two simultaneous C–C bond cleavage that is considered as
The reaction mechanisms presented so far consider the rate- the rate-determining step. Model (M10) proposes that cleavage of the
determining step of the reaction to be the bimolecular reaction be­ first C–C bonds occurs upon glycerol adsorption. The rate determining
tween the adsorbed species derived from water and glycerol. However, step in model (M10) is considered to be the dehydrogenation of adsorbed
several studies performed on reforming processes have shown that the CH2OH, which releases the first H2 molecules. Model (M11) suggests a
decomposition of hydrocarbons or oxygenated hydrocarbons is often the molecular adsorption of glycerol on a single active site. The breaking of
rate-determining step in the reaction [7,38–41]. In this respect, models a first C–H bond within this adsorbed glycerol molecule is considered as
(M9) to (M12) were developed considering the decomposition of glycerol the rate-determining step. Finally, model (M12) is similar to model (M9)
in various forms as the rate-determining step of the reaction. In each of but proposes a dissociative adsorption of glycerol on two active sites,
these mechanisms, it is possible to separate the reaction into two sub­ which detach a single H from the initial glycerol molecule (compared to
sequent stages, i.e., i) the decomposition and dehydrogenation of glyc­ two for model (M9)). The subsequent breaking of the first C–C bond is
erol on the catalyst surface that ultimately leads to the formation of considered to be the rate-determining step of the reaction in model
adsorbed CO species, followed by ii) the water–gas shift (WGS) reaction (M12). In models (M9) to (M12), it is assumed that the adsorbed water
between adsorbed water and CO species to form the reaction products reacts almost immediately with adsorbed CO species via the water–gas
(CO2 and H2). The aforementioned two-phase reaction sequence was shift reaction. This causes the rate expressions developed for these
first assumed and validated by Byrd et al. [41] using a Ru-based catalyst, models to be unaffected by the partial pressure of water. The de­
and then was later verified by Pant et al. [40] using a nickel-based velopments of these reaction mechanisms are detailed below.
catalyst. The difference between models (M9) to (M12) lies in the Model (M9): Single-site dissociative adsorption of glycerol (no C–C
behavior of glycerol upon its dissociative adsorption to the surface of the bond cleavage) and molecular adsorption of steam, first and second C–C
active site, and in the considered rate-determining step. First, model bond breaking as rate-determining step

12
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

Model (M10): Single-site dissociative adsorption of glycerol (with


C–C bond cleavage) and molecular adsorption of steam, adsorbed Model (M12): Single-site dissociative adsorption of glycerol and
CH2OH dehydrogenation as rate-determining step molecular adsorption of steam, first C–C bond breaking as rate-

determining step

Model (M11): Single-site molecular adsorption of glycerol and mo­


lecular adsorption of steam, first C–H bond breaking as rate-determining
step

13
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

to gas consumption rates (rj ) as a function of the rate constants krxj , the
The rate law expressions developed from the 12 different reaction apparent adsorption and desorption constants (KAi.j and KDi.j , respec­
scenarios are presented in Table 4. These equations provide the glycerol tively) and the partial pressures (p) of the individual compounds. The

Table 4
Rate expressions developed for proposed heterogeneous kinetic models.
Model Rate expression Hypothesesa

(M1) 1
krx1 pC3 H8 O3 pH2 O 2
r1 =
1
( )2
1 + KA1.1 pC3 H8 O3 + KA2.1 pH2 O 2 + KD3.1 pCO2
(M2) 1 1
krx2 pC3 H8 o3 2 pH2 O 2
r2 =
1 1
( )2
1 + KA1.2 pC3 H8 o3 2 + KA2.2 pH2 O 2 + KD3.2 pCO2
(M3) 1
krx3 pC3 H8 o3 2 pH2 O
r3 =
( 1 1 )2
pCO2 pH2
1 + KA1.3 pC3 H8 o3 2 + KA2.3 pH2 O + KD3.3 + KD5.3 pH2 2
pH2 O
(M4) krx4 pC3 H8 O3 pH2 O
r4 =
( 1 )2
pCO2 pH2
1 + KA1.4 pC3 H8 O3 + KA2.4 pH2 O + KD3.4 + KD5.4 pH2 2
pH2 O
(M5) 1
krx5 pC3 H8 O3 pH2 O 2
r5 =
1
( )( )
1 + KA1.5 pC3 H8 O3 + KD3.5 pCO2 1 + KA2.5 pH2 O 2 + KD4.5 pCO2
(M6) 1 1
krx6 pC3 H8 o3 2 pH2 O 2
r6 =
1 1
( )( )
1 + KA1.6 pC3 H8 o3 2 + KD3.6 pCO2 1 + KA2.6 pH2 O 2 + KD4.6 pCO2
(M7) 1
krx7 pC3 H8 o3 2 pH2 O
r7 =
( 1 1 )( 1)
pCO2 pH2
1 + KA1.7 pC3 H8 o3 2 + KD3.7 + KD5.7 pH2 2 1 + KA2.7 pH2 O + KD6.7 pH2 2
pH2 O
(M8) krx8 pC3 H8 O3 pH2 O
r8 =
( 1 )( 1)
pCO2 pH2
1 + KA1.8 pC3 H8 O3 + KD3.8 + KD5.8 pH2 2 1 + KA2.8 pH2 O + KD6.8 pH2 2
pH2 O
(M9) 1
krx9 pC3 H8 o3 3
r9 =
1
( )3
1 + KA1.9 pC3 H8 o3 3 + KD3.9 pCO2
(M10) 2
krx10 pC3 H8 o3 3
r10 =
1
( )2
1 + KA1.10 pC3 H8 o3 3 + KD3.10 pCO2
(M11) krx11 pC3 H8 O3
r11 = ( )2
1 + KA1.11 pC3 H8 o3 + KD3.11 pCO2
(M12) 1
krx12 pC3 H8 o3 2
r12 =
1
( )2
1 + KA1.12 pC3 H8 o3 2 + KD3.12 pCO2
Colors: , ; S: Single-site adsorption; D: Dual-site adsorption
a
: ●: Molecular adsorption; : Dissociative adsorption; : Bimolecular surface reaction is determining the reaction rate, : A step included in the decomposition of
glycerol is determining the reaction rate

14
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

Fig. 4. Confirmation of the validity of the power law model: calculated versus experimental reaction rate of glycerol to gas consumption, Arrhenius plot and
identified kinetic parameters.

subscript j indicates the associated model number while the subscript i reaction of 0.63 were identified for the temperature range (480–580 ◦ C)
refers to the compound involved in the adsorption or desorption (glyc­ tested in this study. Keeping in mind that a catalyst provides an alter­
erol: i = 1, water: i = 2, carbon dioxide: i = 3 or 4, hydrogen: i = 5 or 6). native pathway for the reaction with a lower activation energy, it is
The presence of two possible subscripts associated with CO2 and H2 in logical to encounter distinct activation energy values from one publi­
the dual-site models means that desorption of these products can occur cation to another in the literature since several parameters such as the
on either type of site. catalyst used, the method of preparation, or the tested temperature
range can change the rate-determining step of a reaction. Despite this
5. Discussion statement, similarities can be seen between the Ea values obtained in this
study and those from the work of Cheng et al. [8] (59.8 kJ⋅mol− 1 over a
The kinetic parameters identified for the power law model for 15 % Ni/Al2O3 catalyst, between 450 and 550 ◦ C), Foo et al. [21] (67.2
glycerol reaction rates are shown in Fig. 4, which clearly demonstrates kJ⋅mol− 1 over a 15 % Co/Al2O3 catalyst, between 450 and 550 ◦ C) and
the validity of this method due to the high R2 coefficient between the Adesina et al (63.3 kJ⋅mol− 1 over a 5%Co-10 %Ni/Al2O3 catalyst, be­
experimental and calculated rate values as well as the compliance with tween 500 and 550 ◦ C) . However, the overall reaction orders identified
Arrhenius’ law. An activation energy of 66.1 kJ⋅mol− 1 and an order of being different (0.82 for Cheng et al. [8] and 0.47 for Foo et al. [21])

Fig. 5. Confirmation of the validity of the power law model: calculated versus experimental reaction rate of product formation, and identified kinetic parameters.

15
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

Table 5
ANOVA results for discrimination of models (M10), (M11) and (M12).
Relation(A → B) (M10)→(M12) (M11)→(M12) (M9)→(M12) (M10)→(M11) (M10)→(M9) (M11)→(M9)
a
F1− ϕ (υA , υB ) 3.45 3.45 3.45 3.45 3.45 3.45
FA− B 8.59 6.94 3.97 1.24 2.16 1.75
Model A significantly improves the goodness of fit compared to B Yes Yes Yes No No No
a
: ϕ = 0.1 was chosen for this study

may suggest that the mechanisms involved or the rate-determining step where K’Ai.j : actual adsorption constant with respect to reactant i and
may also be different [19]. model j,
The kinetic parameters describing the formation rate of the gaseous K’A1.9 = 4∙KA1.9 3 (model M9)
products were also determined through power law modeling using (eq.
K’A1.10 = KA1.10 3 (model M10)
(11)) and are presented in Fig. 5. It can be noticed that the Ea values for
K’A1.11 = KA1.11 (model M11)
H2, CO2 and CO are close to the value of 66.1 kJ⋅mol− 1 calculated for
( ) − ΔH
glycerol, which proves that the formation of these products comes ln K’Di.j =
Di.j
+
ΔSDi.j
(12b)
mainly from the consumption of glycerol via reactions (R1) and (R2). As RT R
for CH4, a higher activation energy of 90.7 kJ⋅mol− 1 and significantly where K’Di.j : actual desorption constant with respect to product i and
slower formation rates were observed. This observation is not surprising
model j,
since the 5 % Ni-UGSO catalyst has shown very low selectivity towards
K’D3.9 = KD3.9 − 3 (model M9)
methane at temperatures below 630 ◦ C in our previous studies [5].
K’D3.10 = KD3.10 − 3 (model M10)
Positive values of reaction orders for all compounds indicate that their
rate of formation is directly favored by the partial pressure of glycerol K’D3.11 = KD3.11 − 3 (model M11)
( )
during the reaction. While H2 and CO2 exhibit α approaching that of By plotting ln K’A1.j as a function of 1/T, it is possible to assess the
glycerol consumption, higher values for CO and CH4 suggest that their glycerol adsorption enthalpy (ΔHAi.j ) and entropy (ΔSAi.j ) variation of
formation rates are more strongly influenced by the presence of glycerol. the system. Yet, since adsorption is an exothermic process, the slope of
According to the heterogeneous kinetic model analysis, models (M1), this graph should be positive (ΔHAi.j < 0). We also know that the value
(M2), (M4), and (M8) were rejected since they could not significantly of the y-intercept should be negative since the adhesion of gas molecules
minimize the objective function OF compared to the other models (R2 ≤ to the surface decreases the randomness, and thus the entropy of the
0.98 on the parity plot). The residual analysis allowed the rejection of system. To ascertain the validity of the desorption constants associated
model (M3), (M5), (M6) and (M7) since the residuals were either biased with CO2, a similar procedure, but with an opposite reasoning, was
by a trend or not averaged near zero. Model (M7), for instance, exhibits performed. Positive enthalpies and entropies were then expected.
both flaws: the residuals associated with the tests performed at 505 ◦ C Through this exercise, a negative entropy value was determined from
are biased by an observable trend and those calculated at 480 ◦ C are not the Van’t Hoff plot of the constant associated with carbon dioxide
averaged near zero (Fig. S1). desorption in the (M9) model (ΔSD3.9 < 0). As this observation was
As a result, models (M9), (M10), (M11) and (M12) were the only four found to be unrealistic, the (M9) model had to be rejected. For all the
that met the two criteria stated above. It can be noted that the activation reasons mentioned above, models (M10) and (M11) were considered the
energies calculated for models (M9) and (M11) (78.6 kJ⋅mol− 1 for M9 and most realistic.
50.4 kJ⋅mol− 1 for M11) are farther than the Ea acquired from the The accuracy of these last two models (M10 and M11) in estimating
macroscopic power law analysis (66.1 kJ⋅mol− 1), while Ea values for the experimental data is clearly observable in the parity plot shown in
(M10) and (M12) are closer (62.8 kJ⋅mol− 1 for M10 and 58.4 kJ⋅mol− 1 for Fig. 6. The Arrhenius and Van’t Hoff plots showing the dependencies of
M12). Further statistical analysis based on the variance of the estimated the rate and adsorption constants on temperature do not reveal any
rates of models (M9), (M10), (M11) and (M12) was performed using the obvious anomaly at first sight. The kinetic and thermodynamic param­
condition of (eq. (9)) to assess whether their results were significantly eters calculated from the trend curves of the latter plots are also pre­
different. Analysis by Fisher’s law test showed that models (M9), (M10) sented in Fig. 6. Although the sign of the calculated values respects the
and (M11) significantly improved the goodness of fit compared to model thermodynamic logic, it can be noticed that the enthalpies as well as the
(M12). The latter was therefore rejected. However, model (M10), which entropies of adsorption and desorption identified are quite different for
showed the lowest SSR, was not able to significantly improves the model (M10) and (M11). According to model (M11), it would mean that
goodness of fit according to the calculated criteria when compared to an energy of 197.4 kJ would be released from the system per mole of
model (M9) and model (M11). The ANOVA results are presented in glycerol adsorbed. This value, which seems rather extreme, is in fact
Table 5. Note that F1− ϕ (υA − υB ) is the same for all relations since models much higher than the one identified in the kinetic study of Cheng et al.
(M9), (M10), (M11) and (M12) each had 5 degrees of freedom (υ9 = υ10 = using a Ni-Co based catalyst (-39.93 kJ⋅mol− 1) [22]. As for CO2
υ11 = υ12 = 5). desorption, the high enthalpy value of 297.1 kJ⋅mol− 1 also seems un­
Further investigation was therefore needed to identify which likely when compared to other values reported in the literature for
mechanism was more realistic. An analysis of the adsorption and materials known to have better affinities with CO2 than our catalyst,
desorption constants was performed using the Van’t Hoff relation (eq. such as CD-MOF-2 (113.5 kJ⋅mol− 1) [42] or 5A zeolite (44.0 kJ⋅mol− 1)
(12a) for adsorption and eq. (12b) for desorption) to validate the [43]. For these reasons, although models (M10) and (M11) were able to
calculated values for models (M9) to (M11). Here, the analysis was per­ predict the experimental data almost equally well, it may be concluded
formed on the actual adsorption and desorption constants. These, which that model (M10) is more plausible since the associated kinetic and
are directly obtained from the development of the mechanisms in thermodynamic parameters are more realistic, and therefore, quite
elementary steps, differ from the apparent constants which appear in the possibly more reliable. Table 6 provides a summary of the reasons for
rate expressions of Table 4. The relations between actual and apparent rejecting each kinetic model.
constants are showed below: In terms of the mechanics of the GSR process, (M10) model suggests
( ) − ΔH ΔSAi.j that the reaction process occurs in two distinct stages including the
(12a)
Ai.j
ln K’Ai.j = + decomposition/dehydrogenation of glycerol followed by the WGS
RT R

16
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

Fig. 6. Parity plot of the calculated versus the experimental rates of glycerol to gas consumption for model (M10) and (M11), Arrhenius plot, Van’t Hoff plot (filled
symbols refer to desorption constants, empty symbols refer to adsorption constants) as well as the identified kinetic and thermodynamic parameters.

Table 6
Justification to accept/reject each of the heterogeneous kinetic models.
Model (M1) (M2) (M3) (M4) (M5) (M6) (M7) (M8) (M9) (M10) (M11) (M12)

Reason to reject 1 1 2 1 2 2 2 1 4 X 5 3

Legend: 1: The model failed to fit the experimental data with R2 > 0.98
2: The residuals exhibited a noticeable trend or are not averaged near zero
3: ANOVA revealed a significant difference in goodness of fit compared to other models
4: The model presented thermodynamic anomalies (ΔSdes < 0)
5: The model presented kinetics parameters whose values are unlikely
X: Selected model.

reaction between adsorbed water and CO species. In relation to the model (62.8 kJ⋅mol− 1). This suggest that the mechanisms involved or
conclusions drawn in our previous work [5], the first step could be the rate-determining steps of the reaction could have similarities,
attributed to the active nickel phase owing to its good ability to break although divergent conclusions have been reported in each case.
the C–C and C–H bonds present in the glycerol molecule. The second
phase, on the other hand, could be favored by the presence of Fe in the 6. Conclusions
UGSO support which enhances the hydrogen yield of the reaction by
promoting the conversion of adsorbed CO species via the WGS pathway. In this study, the kinetics of the glycerol steam reforming reaction
It is also assumed that the rate-determining step of the reaction is the catalyzed by a nickel-promoted metallurgical residue (5 % Ni-UGSO)
dehydrogenation of the adsorbed CH2OH intermediate, which initially were experimentally investigated in a fixed-bed reactor at atmospheric
originates from the primary decomposition of the glycerol molecule pressure wherein the reaction temperature was varied between 480 and
occurring upon its adsorption. As mentioned earlier, other studies 580 ◦ C and the reactant feed solution between 0.05 and 0.11 std
published in the literature have already reported that the rate- mL⋅min− 1. Langmuir-Hinshelwood-Hougen-Watson reaction mecha­
determining step in the reforming reaction was found to include the nisms were developed to establish detailed heterogeneous kinetic
decomposition of an oxygenated hydrocarbon intermediate [7,38–41]. models that incorporate single- or dual-site adsorption of reactants as
Moreover, the activation energies of the glycerol steam reforming re­ well as desorption of products on the catalyst surface. The modeling
action calculated in the work of Cheng et al. [8] (59.8 kJ⋅mol− 1), Ade­ results suggested that the reaction mechanism of the GSR process de­
sina et al. [22] (63.3 kJ⋅mol− 1), and Foo et al. [21] (67.2 kJ⋅mol− 1) are velops in two stages: i) glycerol decomposition and dehydrogenation,
found to be very close to those identified in this study, either by the which could be favored by the active Ni phase, followed by ii) WGS
power law model (66.1 kJ⋅mol− 1) or by the best heterogeneous kinetic reaction with water in the gas phase, which could be enhanced by the

17
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

presence of Fe in the catalytic support. According to the selected model, [7] R. Sundari, P.D. Vaidya, Reaction Kinetics of Glycerol Steam Reforming Using a
Ru/Al2O3 Catalyst, Energy Fuels 26 (7) (2012) 4195–4204, https://doi.org/
the reaction is initiated by a dissociative adsorption of glycerol which
10.1021/ef300658n.
leads to the cleavage of a two C–C bond on the surface of the 5 % Ni- [8] C.K. Cheng, S.Y. Foo, A.A. Adesina, Steam reforming of glycerol over Ni/Al2O3
UGSO catalyst. The rate-determining step, considered as the dehydro­ catalyst, Catal. Today 178 (1) (2011) 25–33, https://doi.org/10.1016/j.
genation of an adsorbed glycerol intermediate, will ultimately lead to cattod.2011.07.011.
[9] J.M. Silva, L.S. Ribeiro, J.J.M. Órfão, S. Tosti, M.A. Soria, L.M. Madeira, From
the formation of adsorbed CO species, which will react with adsorbed sorption-enhanced reactor to sorption-enhanced membrane reactor: A step towards
steam to produce the main reaction products, H2 and CO2. Glycerol and H2 production optimization through glycerol steam reforming, Chem. Eng. J. 368
steam adsorption are assumed to occur on identical single sites. Our (2019) 795–811, https://doi.org/10.1016/j.cej.2019.02.178.
[10] Sahraei O.A.Z., Larachi F., Abatzoglou N., Iliuta M.C., Hydrogen production by
model was able to estimate the reaction rates with high accuracy while glycerol steam reforming catalyzed by Ni-promoted Fe/Mg-bearing metallurgical
presenting realistic kinetic parameters that exhibited logical trends with wastes, Appl. Catal. B 219 (2017) 183-193, doi.org/10.1016/j.
the tested temperatures according to the laws of thermodynamics. An apcatb.2017.07.039.
[11] N.D. Charisiou, G. Siakavelas, L. Tzounis, B. Dou, V. Sebastian, S.J. Hinder, M.
activation energy of 62.8 kJ⋅mol− 1 was calculated as well as a pre- A. Baker, K. Polychronopoulou, M.A. Goula, Ni/Y2O3–ZrO2 catalyst for hydrogen
exponential factor of 77.4 mol⋅min− 1⋅gcat-1⋅atm− 2/3 in relation with production through the glycerol steam reforming reaction, Int. J. Hydrogen Energy
the Arrhenius relation. The dissociative adsorption of glycerol is quan­ 45 (17) (2020) 10442–10460, https://doi.org/10.1016/j.ijhydene.2019.04.237.
[12] N.D. Charisiou, K.N. Papageridis, L. Tzounis, V. Sebastian, S.J. Hinder, M.A. Baker,
tified by an enthalpy of − 34.1 kJ⋅mol− 1 and an entropy of − 39.3 J⋅mol- M. AlKetbi, K. Polychronopoulou, M.A. Goula, Ni supported on CaO-MgO-Al2O3 as
1⋅K− 1, while the desorption of CO2 features an enthalpy of 41.1 kJ⋅mol− 1 a highly selective and stable catalyst for H2 production via the glycerol steam
and an entropy of 12.1 J⋅mol-1⋅K− 1. A macroscopic analysis was also reforming reaction, Int. J. Hydrogen Energy 44 (1) (2019) 256–273, https://doi.
org/10.1016/j.ijhydene.2018.02.165.
performed by applying a power law model to correlate the experimental
[13] P.V. Mathure, S. Ganguly, A.V. Patwardhan, R.K. Saha, Steam Reforming of
rates of glycerol to gas consumption and generation of the main gaseous Ethanol Using a Commercial Nickel-Based Catalyst, Ind. Eng. Chem. Res. 46 (25)
products (H2, CO2, CO, CH4). A pre-exponential factor of 31.4 (2007) 8471–8479, https://doi.org/10.1021/ie070321k.
mol⋅min− 1⋅gcat-1⋅atm-α, a reaction order of 0.56 with respect to glycerol [14] Q. Chen, J. Zhang, B. Pan, W. Kong, Y. Chen, W. Zhang, Y. Sun, Temperature-
dependent anti-coking behaviors of highly stable Ni-CaO-ZrO2 nanocomposite
as well as an activation energy of 64.7 kJ⋅mol− 1 were determined as the catalysts for CO2 reforming of methane, Chem. Eng. J. 320 (2017) 63–73, https://
kinetic parameters associated with the consumption of glycerol rate. doi.org/10.1016/j.cej.2017.03.029.
This Ea value is very close to that identified for the agreed heterogeneous [15] C. Zhang, S. Li, G. Wu, J. Gong, Synthesis of stable Ni-CeO2 catalysts via ball-
milling for ethanol steam reforming, Catal. Today 223 (2014) 53–60, https://doi.
model, which confirms the validity of the methods used. org/10.1016/j.cattod.2013.08.013.
[16] B. Valle, A. Remiro, A.T. Aguayo, J. Bilbao, A.G. Gayubo, Catalysts of Ni/α-Al2O3
and Ni/La2O3-αAl2O3 for hydrogen production by steam reforming of bio-oil
Declaration of Competing Interest aqueous fraction with pyrolytic lignin retention, Int. J. Hydrogen Energy 38 (3)
(2013) 1307–1318, https://doi.org/10.1016/j.ijhydene.2012.11.014.
[17] M. Aissaoui, O.A.Z. Sahraei, M.S. Yancheshmeh, M.C. Iliuta, Development of a Fe/
The authors declare that they have no known competing financial
Mg-bearing metallurgical waste stabilized-CaO/NiO hybrid sorbent-catalyst for
interests or personal relationships that could have appeared to influence high purity H2 production through sorption-enhanced glycerol steam reforming,
the work reported in this paper. Int. J. Hydrogen Energy 45 (36) (2020) 18452–18465, https://doi.org/10.1016/j.
ijhydene.2019.08.216.
[18] O.A.Z. Sahraei, A. Desgagnés, F. Larachi, M.C. Iliuta, A comparative study on the
Acknowledgments performance of M (Rh, Ru, Ni)-promoted metallurgical waste driven catalysts for
H2 production by glycerol steam reforming, Int. J. Hydrogen Energy 46 (63)
(2021) 32017–32035, https://doi.org/10.1016/j.ijhydene.2021.06.192.
Financial support provided by the Natural Sciences and Engineering [19] J.M. Silva, M.A. Soria, L.M. Madeira, Challenges and strategies for optimization of
Research Council of Canada (NSERC), the Alexander Graham Bell Can­ glycerol steam reforming process, Renew. Sust. Energ. Rev. 42 (2015) 1187–1213,
ada Graduate Scholarships Master’s program (NSERC CGS M), and the https://doi.org/10.1016/j.rser.2014.10.084.
[20] S. Adhikari, S.D. Fernando, A. Haryanto, Kinetics and Reactor Modeling of
Centre for Innovation and Research on Carbon Utilisation in Industrial
Hydrogen Production from Glycerol via Steam Reforming Process over Ni/
Technologies (CIRCUIT) is gratefully acknowledged. CeO2Catalysts, Chem. Eng. Technol. 32 (4) (2009) 541–547, https://doi.org/
10.1002/ceat.v32:410.1002/ceat.200800462.
[21] C.K. Cheng, S.Y. Foo, A.A. Adesina, H2-rich synthesis gas production over Co/
Appendix A. Supplementary data Al2O3 catalyst via glycerol steam reforming, Catal. Commun. 12 (4) (2010)
292–298, https://doi.org/10.1016/j.catcom.2010.09.018.
Supplementary data to this article can be found online at https://doi. [22] C.K. Cheng, S.Y. Foo, A.A. Adesina, Glycerol Steam Reforming over Bimetallic Co-
Ni/Al2O3, Ind. Eng. Chem. Res. 49 (21) (2010) 10804–10817, https://doi.org/
org/10.1016/j.cej.2021.132278.
10.1021/ie100462t.
[23] P.N. Sutar, P.D. Vaidya, A.E. Rodrigues, Glycerol-Reforming Kinetics Using a Pt/C
References Catalyst, Chem. Eng. Technol. 33 (10) (2010) 1645–1649, https://doi.org/
10.1002/ceat.v33:1010.1002/ceat.201000055.
[24] C.D. Dave, K.K. Pant, Renewable hydrogen generation by steam reforming of
[1] R. Yukesh Kannah, S. Kavitha, Preethi, O. Parthiba Karthikeyan, G. Kumar, N.
glycerol over zirconia promoted ceria supported catalyst, Renewable Energy 36
V. Dai-Viet, J. Rajesh Banu, Rajesh Banu, Techno-economic assessment of various
(11) (2011) 3195–3202, https://doi.org/10.1016/j.renene.2011.03.013.
hydrogen production methods - A review, Bioresour. Technol. 319 (2021) 124175,
[25] G.S. Go, H.J. Lee, D.J. Moon, Y.C. Kim, Glycerol steam reforming over Ni–Fe–Ce/
https://doi.org/10.1016/j.biortech.2020.124175.
Al2O3 catalyst for hydrogen production, Res. Chem. Intermed. 42 (1) (2015)
[2] D. Zeng, F. Kang, Y. Qiu, D. Cui, M. Li, L. Ma, S. Zhang, R. Xiao, Iron oxides with
289–304, https://doi.org/10.1007/s11164-015-2324-7.
gadolinium-doped cerium oxides as active supports for chemical looping hydrogen
[26] M.C. Iliuta, I. Iliuta, B.P.A. Grandjean, F. Larachi, Kinetics of Methane
production, Chem. Eng. J. 396 (2020), 125153, https://doi.org/10.1016/j.
Nonoxidative Aromatization over Ru-Mo/HZSM-5 Catalyst, Ind. Eng. Chem. Res.
cej.2020.125153.
42 (14) (2003) 3203–3209.
[3] O.A.Z. Sahraei, K. Gao, M.C. Iliuta, Application of industrial solid wastes in catalyst
[27] H.S. Fogler, Elements of Chemical Reaction Engineering -, Fifth Edition, Pearson
and chemical sorbent development for H2/syngas production by conventional and
Education Inc., 2016.
intensified steam reforming, New dimensions in production and utilization of
[28] O. Levenspiel, Chemical Reaction Engineering -, Third Edition, John Wiley & Sons,
hydrogen, Elsevier, Cambridge, 2020.
1999.
[4] F. Yang, M.A. Hanna, R. Sun, Value-added uses for crude glycerol–a byproduct of
[29] E.N. Fuller, P.D. Schettler, J.C. Giddings, New Method for Prediction of Binary Gas-
biodiesel production, Biotechnol. Biofuels 5 (13) (2012), https://doi.org/10.1186/
Phase Diffusion Coefficients, Ind. Eng. Chem. Res. 58 (5) (1966) 18–27.
1754-6834-5-13.
[30] S. Tadepalli, R. Halder, A. Lawal, Catalytic hydrogenation of o-nitroanisole in a
[5] O.A.Z. Sahraei, A. Desgagnés, F. Larachi, M.C. Iliuta, Ni-Fe catalyst derived from
microreactor: Reactor performance and kinetic studies, Chem. Eng. Sci. 62 (10)
mixed oxides Fe/Mg-bearing metallurgical waste for hydrogen production by
(2007) 2663–2678, https://doi.org/10.1016/j.ces.2006.12.058.
steam reforming of biodiesel by-product: Investigation of catalyst synthesis
[31] V. Chiodo, S. Freni, A. Galvagno, N. Mondello, F. Frusteri, Catalytic features of Rh
parameters and temperature dependency of the reaction network, Appl. Catal. B
and Ni supported catalysts in the steam reforming of glycerol to produce hydrogen,
279 (2020) 119330, https://doi.org/10.1016/j.apcatb.2020.119330.
Appl. Catal. A 381 (1–2) (2010) 1–7, https://doi.org/10.1016/j.
[6] C. Wang, B. Dou, H. Chen, Y. Song, Y. Xu, X. Du, T. Luo, C. Tan, Hydrogen
apcata.2010.03.039.
production from steam reforming of glycerol by Ni–Mg–Al based catalysts in a
fixed-bed reactor, Chem. Eng. J. 220 (2013) 133–142, https://doi.org/10.1016/j.
cej.2013.01.050.

18
A. Desgagnés and M.C. Iliuta Chemical Engineering Journal 429 (2022) 132278

[32] Y.S. Stein, M.J. Antal Jr., M. Jones Jr., A Study of the Gas-phase Pyrolysis of metal-dominated reaction pathways, React. Chem. Eng. 3 (6) (2018) 883–897,
Glycerol, J. Anal. Appl. Pyrol. 4 (4) (1983) 283–296, https://doi.org/10.1016/ https://doi.org/10.1039/c8re00145f.
0165-2370(83)80003-5. [39] R.O. Idem, N.N. Bakhshi, Kinetic modeling of the production of hydrogen from the
[33] Sadanandam G., Sreelatha N., Phanikrishna Sharma M.V., Kishta Reddy S., Srinivas methanol-steam reforming process over Mn-promoted co-precipitated Cu-Al
B., Venkateswarlu K., Krishnudu T., Subrahmantam M., Durga Kumari V., Steam catalyst, Chem. Eng. Sci. 51 (1996) 3697–3708.
Reforming of Glycerol for Hydrogen Production over Ni/SiO2 Catalyst, ISRN [40] K.K. Pant, R. Jain, S. Jain, Renewable hydrogen production by steam reforming of
Chem. Eng. 2012 (2012) 1-10, doi.org/10.5402/2012/591587. glycerol over Ni/CeO2 catalyst prepared by precipitation deposition method,
[34] L. Oar-Arteta, A.T. Aguayo, A. Remiro, A. Arandia, J. Bilbao, A.G. Gayubo, Kinetics Korean J. Chem. Eng. 28 (9) (2011) 1859–1866, https://doi.org/10.1007/s11814-
of the steam reforming of dimethyl ether over CuFe2O4/γ-Al2O3, Chem. Eng. J. 306 011-0059-8.
(2016) 401–412, https://doi.org/10.1016/j.cej.2016.07.075. [41] A. Byrd, K. Pant, R. Gupta, Hydrogen production from glycerol by reforming in
[35] L.H. Hosten, G.F. Froment, Parameter Estimation in Multiresponse Models, supercritical water over Ru/Al2O3 catalyst, Fuel 87 (13–14) (2008) 2956–2960,
Periodica Polytechnia Chemical Engineering 19 (1–2) (1975) 123–136. https://doi.org/10.1016/j.fuel.2008.04.024.
[36] W.W. Hines, D.C. Montgomery, D.M. Goldsman, C.M. Borror, Probability and [42] D. Wu, J.J. Gassensmith, D. Gouvêa, S. Ushakov, J.F. Stoddart, A. Navrotsky, Direct
statistics in engineering -, 4th ed.,, John Wiley & Sons, Inc., 2003. Calorimetric Measurement of Enthalpy of Adsorption of Carbon Dioxide on CD-
[37] P.D. Vaidya, A.E. Rodrigues, Kinetics of Steam Reforming of Ethanol over a Ru/ MOF-2, a Green Metal-Organic Framework, J. Am. Chem. Soc. 135 (18) (2013)
Al2O3 Catalyst, Ind. Eng. Chem. Res. 45 (19) (2006) 6614–6618, https://doi.org/ 6790–6793, https://doi.org/10.1021/ja402315d.
10.1021/ie051342m. [43] H. Deng, H. Yi, X. Tang, Q. Yu, P. Ning, L. Yang, Adsorption equilibrium for sulfur
[38] M.D. Zhurka, A.A. Lemonidou, J.A. Anderson, P.N. Kechagiopoulos, Kinetic dioxide, nitric oxide, carbon dioxide, nitrogen on 13X and 5A zeolites, Chem. Eng.
analysis of the steam reforming of ethanol over Ni/SiO2 for the elucidation of J. 188 (2012) 77–85, https://doi.org/10.1016/j.cej.2012.02.026.

19

You might also like