Chemical Engineering Journal 329 (2017) 165–177

Contents lists available at ScienceDirect

Chemical Engineering Journal

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

Heat removal and catalyst deactivation in a Sabatier reactor for chemical

fixation of CO2: Simulation-based analysis
Duo Sun, Faisal Mohamed Khan, David S.A. Simakov ⇑
Department of Chemical Engineering, University of Waterloo, Waterloo, ON N2L 3G1, Canada

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

Sabatier reactor for CO2 conversion is

modelled considering catalyst
deactivation.

The reactor is a molten salt-cooled
heat-exchanger type packed bed.

Biogas fed reactor is analyzed
numerically over the range of
operating parameters.

Feed and cooling rates are crucial
parameters affecting the reactor
performance.

The model predicts CH4 yields of 80–
90% over the 10,000 time-on-stream.

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

Article history: Thermo-catalytic hydrogenation of CO2 into synthetic CH4 via Sabatier reaction is an attractive route to
Available online 28 June 2017 reduce fossil fuels consumption and to limit greenhouse gas emissions, while potentially providing rev-
enue. Hydrogen required for the reaction can be generated by water electrolysis using renewable or low
Keywords: carbon footprint electricity. With respect to CO2 methanation, a number of technological challenges have
Sabatier reactor to be resolved in order to make this technology economically viable. The highly exothermic nature of the
CO2 conversion Sabatier reaction makes heat removal a challenging task. Another major problem is catalyst deactivation
Packed-bed reactor
by coking caused by CH4 cracking at elevated temperatures. Maximizing CH4 production over extended
Catalyst deactivation
periods of operation will require highly efficient heat removal to facilitate CH4 production, while mini-
mizing catalyst deactivation. In this study, a heat-exchanger type, molten salt-cooled packed bed reactor
is analyzed using a transient mathematical model. The model considers inter-compartment heat
exchange and catalyst deactivation by coking, assuming the use of a Ni/Al2O3 catalyst. The reactor per-
formance was investigated in terms of CO2 conversion and CH4 yield for the case of the feed containing
CO2 and H2, and for the case of the reactor fed with biogas (40% CO2 and 60% CH4) and H2. The model
predicts that, though the catalyst deactivation leads to a substantial decline in the reactor performance,
CH4 yields and CO2 conversions over 80% are achievable after 10,000 h of operation.
Ó 2017 Elsevier B.V. All rights reserved.

1. Introduction fuels and chemicals [1]. The current usage of CO2 as a feedstock
in industries is limited to processes such as synthesis of urea, sal-
The increasing levels of global CO2 emissions have prompted icylic acid and polycarbonates [1]. Conversion of CO2 into synthetic
research in utilizing CO2 as a feedstock for generating synthetic CH4 via methanation, also called the Sabatier reaction, has recently
gained increasing interest as a process being more advantageous
thermodynamically as compared to other reactions which form
⇑ Corresponding author.
higher hydrocarbons or alcohols [2].
E-mail address: dsimakov@uwaterloo.ca (D.S.A. Simakov).

http://dx.doi.org/10.1016/j.cej.2017.06.160
1385-8947/Ó 2017 Elsevier B.V. All rights reserved.
166 D. Sun et al. / Chemical Engineering Journal 329 (2017) 165–177

Nomenclature

a activity factor Rg gas constant, kJ/(mol K)

ac;HE cooling tube surface-to-volume ratio, m1 SV gas hourly space velocity, h1
ar;HE cooling tube surface-to-packed volume ratio, m1 t time, s
ar;HL reactor surface-to-volume ratio, m1 T reactor temperature, K
Ac total cross-sectional area of cooling tubes, m2 Tc coolant temperature, K
Aj pre-exponential factor of the rate coefficient of reaction Te environment temperature, K
j, units of kj Uw effective wall heat transfer coefficient, kJ/(m2 s K)
Bj pre-exponential factor of the adsorption coefficient of v fluid velocity, m/s
species i, units of K j V compartment volume, m3
Ci molar concentration of species i, mol/m3 z reactor length coordinate, m
Ct total molar concentration, mol/m3
C pc coolant heat capacity, kJ/(kg K) Greek letters
C pg gas heat capacity, kJ/(mol K) e catalyst bed porosity
d wall thickness, m / Thiele modulus
dp catalytic pellet diameter, m gj effectiveness factor of reaction j
D diameter, m k thermal conductivity, kW/(m K)
Dae effective axial diffusion coefficient, m2/s l viscosity, kg/(m s)
Dm gas molecular diffusivity, m2/s qc coolant density, kg/m3
Ej activation energy of reaction j, kJ/mol qg gas molar density, mol/m3
G gravimetric (mass) flow rate, kg/s qs solid density, kg/m3
hnc natural heat convection coefficient, kJ/(m2 s K) sb catalyst bed tortuosity
hw effective wall heat transfer coefficient, kJ/(m2 s K)
DHi adsorption enthalpy change of species i, kJ/mol Subscripts
kae effective axial thermal conductivity, kJ/(m s K) c coolant
kj rate constant of reaction j eff effective
Ki adsorption constant of species i, bar1
eq equilibrium
K j;eq equilibrium constant of reaction j f feed
L reactor length, m g gas
Nu Nusselt number
HE heat exchange
pi partial pressure of gaseous species i, bar HL heat loss
P reactor pressure, bar int initial
Pr Prandtl number nc natural convection
Ptf total feed gas pressure, bar
p packed bed
Re Reynolds number s solid
Rj rate of reaction j, mol/(kg s) r reactor

The exothermic Sabatier reaction, Eq. (3), is accompanied by the in the feed gas stream. [9,10]. Sintering also plays an important
mildly endothermic reverse water gas shift, Eq. (2), and the role in the deactivation of Ni-based catalysts. Studies have indi-
strongly exothermic CO methanation, Eq. (1). cated that Ni sintering proceeds by chemical sintering, via the for-
mation of Ni(CO)4 [11].
CO þ 3H2 $ CH4 þ H2 O
While operating at relatively low temperatures typically
DH298 ¼ 206:1 kJ=mol ð1Þ employed in CO2 methanation (700–800 K) and in the absence of
sulphur (can be removed from feed gas by adsorbents), catalytic
CO2 þ H2 $ CO þ H2 O activity of Ni is mainly reduced by coking. Carbon deposition deac-
DH298 ¼ þ41:2 kJ=mol ð2Þ tivates the catalyst by fouling the catalyst surface, blocking catalyst
pores and disintegrating the catalyst support [9]. The three major
CO2 þ 4H2 $ CH4 þ 2H2 O forms of carbon deposited on Ni catalysts are encapsulating hydro-
carbons films formed by polymerization at temperatures below
DH298 ¼ 164:9 kJ=mol ð3Þ
750 K, whiskerlike carbon formed at temperatures higher than
The overall process is highly exothermic which requires effi- 725 K, and pyrolytic carbon formed by cracking of hydrocarbons
cient heat removal, in order to facilitate CH4 production. In addi- above 875 K [9]. The most thermodynamically probable reactions
tion to this thermodynamic constrain, reactor overheating can involved in carbon deposition in the methanation process are
also lead to fast catalyst deactivation by sintering and coking. CH4 cracking, Eq. (4), Boudouard reaction, Eq. (5), and CO reduction
Methanation of CO2 over catalysts based on various group VIII (reverse gasification), Eq. (6) [12]:
metals (e.g., Rh, Ru, Co) supported on TiO2, SiO2, Al2O3, CeO2, and CH4 $ C þ 2H2
ZrO2 has been studied as a possible method to produce synthetic DH298 ¼ þ74:8 kJ=mol ð4Þ
natural gas [3–7]. However, Ni-based catalysts have typically been
used in commercial processes due to the low costs, high activity 2CO $ C þ CO2
and good selectivity to CH4 formation [1,8]. Despite the abovemen-
DH298 ¼ 173:3 kJ=mol ð5Þ
tioned advantages, the implementation of Ni-based catalysts for
CO2 methanation is limited because of high tendency to deactiva-
CO þ H2 $ C þ H2 O
tion by poisoning, coking and sintering [9]. Poisoning of Ni cata-
lysts typically occurs due to the presence of sulphur compounds DH298 ¼ 131:3 kJ=mol ð6Þ
 D. Sun et al. / Chemical Engineering Journal 329 (2017) 165–177 167

Deactivation of Ni-based catalyst by coking has been studied by microchannel reactors, and fixed beds [21]. The Sabatier reactor
various researchers and our understanding of the carbon formation design suggested in the current study is shown in Fig. 1. The reac-
mechanisms is relatively complete. Claridge et al. investigated the tor is a heat-exchanger type packed bed internally cooled by mol-
effect of temperature on the rate of carbon deposition by the Bou- ten salt flowing in multiple cooling tubes. To minimize
douard reaction and CH4 cracking reaction. Their results indicated uncontrollable heat losses to the environment, the reactor is insu-
that carbon deposition by CH4 cracking dominates at relatively lated by a layer of quartz wool. Both the external and internal
high temperatures, while carbon formation via Boudouard reaction tubes are made of stainless steel.
mainly occurs at lower temperatures, with both process contribut- The dimensions used in numerical simulations are listed in
ing equally to carbon formation at 890 K [13]. Table 1. Since we used a 1D mathematical model in our study,
Other studies investigated the effect of various feed conditions on we selected reactor dimensions in such a way that the radial heat
the Boudouard reaction showing that low CO partial pressures, high transfer dimension is minimized. Due to the small diameter of the
H2/CO ratios, and high partial pressures of CO2 reduce the carbon for- cooling tubes (Dc = 20 mm) and high thermal conductivity of the
mation rate via the CO disproportionation. These findings indicate molten salt, significant radial gradients are not expected to develop
that the major cause of coking in the CO2 methanation process is within the molten salt [22]. In order to minimize the heat transfer
CH4 cracking [14,15]. Analysis of the reaction mechanisms involved distance in the packed bed, the number of cooling tubes was set to
in carbon deposition via CH4 cracking indicated that the rate- N = 13. For this number of tubes and for the dimensions described
limiting step is the dissociative adsorption of CH4 on the catalyst in Table 1 with evenly distributed tubes, the distance between any
with H2 adsorption acting in competition. The rate of the reaction two adjusted cooling tubes will be 19 ± 1 mm. Thermal conductiv-
was found to be a function of temperature and partial pressure of ity of the packed bed should prevent large radial gradients over
H2 and CH4. It is worth mentioning that CH4 cracking on Ni surfaces that relatively small distance [23]. In general, 1D models of packed
can be used to generate carbon nanotubes and H2 [16], with the cok- beds are believed to be capable of describing at least the qualitative
ing threshold directly related to the size of Ni nanocrystallites [17]. trends [24].
The goal of the present study was to analyze the effect of the Ni- A transient, one-dimensional, pseudo-homogeneous mathe-
based catalyst deactivation via CH4 cracking on reactor perfor- matical model [18,19] was used to simulate the reactor. The model
mance under various operating conditions including space veloc- accounts for axial mass and heat dispersion and for the tempera-
ity, pressure, and cooling rate. A heat-exchanger type, multiple- ture dependence of thermo-physical properties. Component mass
tube Sabatier reactor internally cooled by the molten salt was sug- balance and energy balance for the packed bed compartment are
gested as an effective design solution for heat removal. A transient given by Eq. (7) and Eq. (8), respectively. The model accounts for
pseudo-homogeneous mathematical model [18,19] was used to temperature variations of the heat transfer fluid, as opposed to
simulate the reactor performance. To simulate catalyst deactiva- assuming constant coolant temperature [24,25]. The temperature
tion via CH4 cracking, first order deactivation rate was assumed distribution in a single molten salt tube is described by Eq. (9).
[20]. The reactor performance was analyzed over extended operat- Note that the energy balance for the packed bed includes the heat
ing periods, up to 10,000 h time on stream, in terms of CO2 conver- loss term (Te = 298 K is the temperature of the environment) in
sion, selectivity to CH4 formation, and CH4 yield. Our study provide addition to the heat exchange term.
new insights into the operation of the molten salt-cooled Sabatier
@C i @2Ci @C i X
reactor subject to catalyst deactivation by coking. Importantly, our
e ¼ Dae 2  ev g þ að1  eÞqs gj Rij ð7Þ
mathematical model accounts for temperature variations of the @t @z @z j
heat transfer fluid, as well as for the catalyst deactivation. The
results of numerical simulations predict that, with the optimized
ðqC p Þeff ¼ kae @@z2T  eqg C pg v g @T
2
@T
heat removal, it is possible to operate the reactor over extended @t @z
þ
periods of time and with reasonably high CH4 yields, even using X
þ að1  eÞqs ðDHj Þgj Rj ð8Þ
the Ni-based catalyst that is susceptible to deactivation by coking. j

 U w;HE ar;HE ðT  T c Þ  U w;HL ar;HL ðT  T e Þ

2. Model formulation
@T c @2T c @T c
Various reactor configurations were suggested for CO2 metha- qc C pc ¼ kc 2  qc C pc v c  U w;HE ac;HE ðT c  TÞ ð9Þ
@t @z @z
nation, including fluidized beds, slurry reactors, honeycombs,

Fig. 1. Schematic of the molten salt-cooled, heat-exchanger type packed bed Sabatier reactor.
168 D. Sun et al. / Chemical Engineering Journal 329 (2017) 165–177

Table 1
Reactor dimensions.

Dr (m) Dc (m) dw (m) diw (m) dp (m) N (m) L (m) Vr (m3) Vc (m3)
0.2 0.02 0.002 0.05 0.003 13 1 0.0255 0.0041

D and L denote diameter and length of the packed bed (Dr) and coolant (Dc) compartments; dw denote thickness of the reactor wall and cooling tube; diw is insulation layer
thickness; V denote volume of the packed bed (Vr) and coolant compartment (Vc); N is the number of the molten salt coolant tubes; dp is the (spherical) catalytic pellet size.

rffiffiffiffiffiffiffi
z¼0 ev g ðC i;f  C i Þ ¼ Dae @C@zi z¼L @C i
¼0   ^ d2
@z
gj ¼ /3j 1
 /1j /j ¼
k j p

eqg v g C pg ðT f  TÞ ¼ kae @T @z
@T
@z
¼ 0 ð10Þ tanh /j 4Dm
ð16Þ
qc v c C c ðT c;f  T c Þ ¼ kc @T@zc @T c
@z
¼ 0 ^1 ¼ kp
k s ð1eÞ
1 qffiffiffiffi ^2 ¼ k2 qs ð1eÞPtf
k q e
^3 ¼ kp
k s ð1eÞ
3 qffiffiffiffi
P q e
tf g g P q e
tf g

t ¼ 0 C i ð0; zÞ ¼ C i;int
Tð0; zÞ ¼ T int ð11Þ
2.2. Catalyst deactivation kinetics
T c ð0; zÞ ¼ T c;int
A general form of the catalyst activity factor multiplying reac-
Pressure drop was accounted for using Ergun equation, Eq. (12).
tion terms in Eqs. (7), (8) is written as follows:
Velocity correction was calculated using Eq. (13). Effective heat
capacity in Eq. (8) is defined by Eq. (14).
da
dP ð1  eÞ lg ð1  eÞqg 2
2  ¼ rd ad ð17Þ
¼ 150 v g  1:75 vg ð12Þ dt
dz dp e
2 3 dp e3
Assuming 1st order deactivation (d ¼ 1), the following expres-
Ct
v g ¼ v gf ð13Þ sion for the catalyst activity factor is derived by integration:
C t;f
a ¼ expðrd tÞ ð18Þ
ðqC p Þeff ¼ eqg C pg þ ð1  eÞqs C ps ð14Þ As it was mentioned in the Introduction, for temperatures
above 700 K low CO partial pressures, the deactivation of Ni-
2.1. Catalytic kinetics based catalysts is predominantly caused by the accumulation of
the filamentous carbon formed via CH4 cracking, Eq. (4). The Bou-
Reaction rates are calculated using the commonly adopted douard reaction and the CO reduction (reverse gasification), Eqs.
kinetics for methane steam reforming over the Ni/Al2O3 catalyst (5), (6), require CO as a source of carbon. However, in all our sim-
[26,27], Eqs. (15a)(15c). These kinetic expressions account for the ulations the amount of CO produced was less than 1 mol% (mole
reversibility of the reforming and water gas shift reactions and fraction less than 0.01). On the other hand, CH4 concentrations
can be used without modification for modeling of the CO2 metha- obtained in our simulations were relatively high, as well as H2 con-
nation reaction system, Eqs. (1)–(3). All parameters are tabulated centrations (see Figs. 2, 4, and 6), thus the carbon gasification by H2
in the literature, e.g., see Tables 3 and 4 in Rodríguez et al. is expected to be significant.
(2012) [28] that include parameters for the rate constants, adsorp- The reactions of carbon gasification by H2O and CO2, reverse
tion constants, as well as the temperature dependence of the equi- reactions in Eqs. (5), (6), can potentially occur during methanation
librium constants. because CO2 is a reactant and H2O is a product. However, these
! reactions are highly endothermic, as compared to the mildly
k1 p3H2 pCO 1 exothermic carbon gasification by H2. Both steam and CO2 gasifica-
R1 ¼ 2:5 pCH4 pH2 O  ð15aÞ
pH2 K 1;eq den2 tion are thermodynamically favorable at temperatures above
  950 K, when the corresponding equilibrium constant is signifi-
k2 pH pCO 1
R2 ¼ pCO pH2 O  2 2 2
ð15bÞ cantly higher than 1 [29]. On the other hand, hydrogasification
pH2 K 2;eq den (reverse reaction in Eq. (4)) is thermodynamically favorable below
  that temperature [29]. Since in our simulations maximum temper-
k3 pH pCO 1
R3 ¼ 3:5 pCH4 p2H2 O  2 2 2
ð15cÞ atures did not exceed 750 K, the H2-induced gasification is
pH2 K 3;eq den
expected to be the predominant process for carbon removal.
Consequently, the catalyst deactivation in the present study
K H2 O pH2 O
den ¼ 1 þ K CO pCO þ K H2 pH2 þ K CH4 pCH4 þ was assumed to be solely induced by CH4 cracking, while consider-
pH2 ing the reverse reaction of gasification by H2. The following power
    law rate expression was adopted from the literature [30]:
Ej DHi  
kj ¼ Aj exp K i ¼ Bi exp pCH4 Ed
Rg T Rg T rd ¼ kd kd ¼ kd0 exp ð19Þ
pH2 Rg T
Estimation the intraparticle and interphase transport limita-
tions [18] has shown that intraparticle temperature gradients The activation energy (Ed = 178 kJ/mol), and pre-exponential
and interphase transport limitations can be neglected for the size factor (kd0 = 2.35  108 min1) were extracted from the experi-
of catalytic pellets used in our simulations (dp = 3 mm, Table 1), mental data available in the literature (Fig. 5 in Borghei et al.
in the entire range of parameters studied. However, it has been (2010) [20]). Note that these parameters are not expected to reflect
found that the internal (intraparticle) diffusion resistance cannot exactly the rate of carbon formation under the conditions used in
be neglected. This limitation was accounted for using the effective- our simulations, as these parameters were estimated for methane
ness factor defined below (the dimensionless Thiele Modulus, /j cracking (in the presence of H2), not for CO2 methanation. How-
for the reaction j, is defined with the modified reaction rate con- ever, as kinetic data on catalyst deactivation during the CO2 metha-
stants; the original rate constants (kj) are given by Eq. (15)): nation reaction is scarce, we selected these parameters for our
 D. Sun et al. / Chemical Engineering Journal 329 (2017) 165–177 169

Fig. 2. Spatial profiles of mole fractions (upper panels) and temperatures (lower panel) for TOS = 10 h (a) and TOS = 10,000 h (b). TPB and TMS denote temperatures in the
packed bed (PB) and molten salt (MS) compartments. Parameters: Tf = 650 K, GHSV = 1000 h1, Pf = 5 bar, Gc = 0.4Gc,0.

model. From the catalyst deactivation viewpoint, our study is The wall heat exchange coefficient between the catalytic bed
intended for the investigation of main trends rather than exact pre- and the coolant tube, Uw,HE, accounts for resistances of the fixed
diction of the reactor behaviour. Still, since these are experimen- bed, the coolant tube wall, and the molten salt, Eq. (22a). Similarly,
tally estimated kinetic parameters, the model is expected to have the correlation for the wall heat loss coefficient, Uw,HL, accounts for
certain predictive ability. resistances through the catalytic bed, the reactor wall, and the
quartz wool insulation layer [34], accounting also for heat losses
2.3. Transport parameters from the reactor external surface via natural convection (hnc)
[35]. Since the insulation layer resistance and natural convection
The effective axial mass dispersion coefficient is calculated are dominant in Eq. (22b), the wall heat loss coefficient was nearly
using the following correlation [31]: constant in all simulations: Uw,HL  0.01 W/(m2 K).
  The effective wall heat transfer coefficient for the reaction com-
Dm 1
Dae ¼ e þ 0:5dp v g sb ¼ ð20Þ partment (hwr) is estimated using the following correlation
sb e0:5 obtained in the similar way as Eq. (21), using a complete set of
The expression for the effective axial heat dispersion coefficient the original correlations [19,32,33]:
(kae), Eq. (21), was derived from the heat conductivity correlations
hwr dp v g qg dp
developed for catalytic fixed beds [19,32,33]. Values of kae were Nup ¼ ¼ 24 þ 0:34 Re0:77 Rep ¼ ð23Þ
calculated using original correlations [32,33] in the relevant range kg p
lg
of parameters, plotted versus particle Reynolds number (Rep), and
The effective wall heat transfer coefficient for the coolant tube
fitted using least squares analysis, resulting in the following
(hwc) is estimated using the following, adopted from the literature,
correlation:
correlations [36–38]:
v g qg dp
kae ¼ kg ð8 þ 0:05Rep1:09 Þ Rep ¼ ð21Þ 0:065Rec Prc ðDc =LÞ
lg Rec < 2030 Nuc ¼ 3:66 þ ð24aÞ
1 þ 0:04½Rec Prc ðDc =LÞ2=3
Wall heat transfer coefficients are determined by resistances in
series, accounting for contributions of the fixed bed, molten salt, 2030 < Rec < 4000 Nuwc
tube walls, and insulation layer, Fig. 1: 2=3
 1 ¼ 0:012ðRe0:87
c  280ÞPr0:4
c ½1 þ ðDc =LÞ  ð24bÞ
1 dw 1
U w;HE ¼ þ þ ð22aÞ
hwr kw hwc Rec > 4000 Nuc ¼ 0:027Re0:8 1=3
c Prc

 1 hwc Dc v c qc Dc C pc lc ð24cÞ

Nuwc ¼ Rec ¼ Prc ¼
U w;HL ¼
1
þ
dw diw
þ þ
1
ð22bÞ kc lc kc
hwr kw kiw hnc
170 D. Sun et al. / Chemical Engineering Journal 329 (2017) 165–177

Fig. 3. Temporal evolution of the reactor performance (a) and spatiotemporal profiles of the packed bed temperature (b), Sabatier reaction rate (c), and catalyst activity (d).
XCO2 , SCH4 , and YCH4 denote conversion, selectivity, and yield, Eqs. (25)–(27); R3, Eq. (15c) is shown as normalized reaction rate. Parameters: Tf = 650 K, GHSV = 25,000 h1,
Pf = 5 bar, Gc = 0.4Gc,0.

2.4. Numerical procedure via methanation reactions via endothermic steam reforming reac-
tions, Eq. (1), (3), making heat removal more efficient and decreas-
The model was solved using the MATLAB PDE solver with a ing the packed bed temperature. Lower reactor temperatures will
second-order accurate spatial discretization (finite elements) result in lower rate of the catalyst deactivation, Eq. (19). Dry CH4
based on a fixed set of user-specified nodes (100 nodes were used) reforming is not considered as it is not expected to occur to a sig-
and time integration done by the stiff ODE solver (ode 15s). Depen- nificant extent at relatively low temperatures considered in this
dences of thermophysical properties (density, viscosity, heat study, below 800 K [43].
capacity, diffusivity, and thermal conductivity) on temperature, For the case of the reactor fed with the mixture of CO2 and H2,
pressure and composition were accounted for using polynomial the reactor performance is evaluated in terms of CO2 conversion
regressions fitted to the data on thermophysical properties from (XCO2 ), selectivity to CH4 (SCH4 ), and CH4 yield (YCH4 ) which can be
the literature [39–41]. Molten salt properties were adopted from calculated based on molar fractions:
the data on commercially available molten salts (Dynalene, Inc.
yCH4 þ yCO
[42], Dynalene MS-2). X CO2 ¼ ð25Þ
yCH4 þ yCO þ yCO2
3. Results and discussion
yCH4
SCH4 ¼ ð26Þ
The effect of catalyst deactivation and heat removal on the reac- yCH4 þ yCO
tor performance is first analyzed for the case of the reactor fed with
pure CO2 and H2. In practical applications this approach will yCH4
Y CH4 ¼ ð27Þ
require CO2 separation from the carbon-rich stream such as landfill yCH4 þ yCO þ yCO2
gas, biogas or flue gas. Another approach is to feed a mixed feed-
stock such as raw biogas, only pre-treated to remove H2S to pre- For the case of the reactor fed with biogas, since the inlet mix-
vent catalyst poisoning by sulphur. In the analysis of the reactor ture already contains CH4 it is more convenient to evaluate the
fed with raw biogas (after sulphur removal) it is assumed that bio- reactor performance solely in terms of CH4 produced from CO2.
gas contains 60% CH4 and 40% CO2. This case, which is more com- The following equation (based on molar rates) was used to calcu-
plicated than the case of pure CO2 feed, is studied in more detail, late the yield of CH4 for the case of the biogas-fed reactor:
specifically investigating the effects of space velocity, pressure, F CH4 ;out  F CH4 ;in
and heat removal rate on the reactor performance in terms of Y CH4 ¼ ð27aÞ
F CO2 ;in
CH4 yield.
The presence of CH4 in the feed is expected to induce catalyst In all simulations, the initial temperatures (Eq. (11), Tint, Tc,int)
deactivation by coking, Eq. (19). On the other hand, the reaction were set to 550 K. The molten salt feed temperature was set to
system will be less exothermic as CH4 can react with H2O produced its minimum operating temperature of 415 K [42]. The feed molar
 D. Sun et al. / Chemical Engineering Journal 329 (2017) 165–177 171

Fig. 4. Spatial profiles of mole fractions (upper panels) and temperatures (lower panel) for TOS = 10 h (a) and TOS = 10,000 h (b), for the biogas-fed reactor. TPB and TMS are
temperatures of the packed bed (PB) and molten salt (MS). Parameters: Tf = 650 K, GHSV = 1000 h1, Pf = 5 bar, Gc = 0.4Gc,0.

Fig. 5. Spatiotemporal profiles of the packed bed temperature (a) and catalyst activity (b) for the biogas-fed reactor. Parameters: Tf = 650 K, GHSV = 5000 h1, Pf = 2 bar,
Gc = 0.4Gc,0.

stoichiometric ratio of H2/CO2 = 4 was kept in all simulations, sion and no CO formation) is equal to the rate of heat removal by
including biogas feed for which the feed composition in terms of the molten salt (assuming that DTMS = 300 K; the operating range
gas molar fractions was 8/13 H2, 2/13 CO2, and 3/13 CH4. All oper- of the molten salt is 415–758 K [42]):
ating parameters are listed in Table 2.
Variable operating parameters include the gas hourly space yCO2 DHSR F tf
Gc;0 ¼ F tf ¼ qgf V PB GHSV ð29Þ
velocity (GHSV), feed pressure (Pf), and normalized cooling rate C pc DT MS
(Gc/Gc,0). Gas hourly space velocity (volumetric feed flow rate
divided by the packed bed volume) is defined as follows:
3.1. Catalyst deactivation effect: Reactor fed with pure CO2 and H2
ev gf
GHSV ¼ ð28Þ
L Spatial profiles of mole fractions and temperatures in the
The reference molten salt gravimetric flow rate is calculated packed bed and cooling (molten salt) compartments for TOS = 10 h
assuming that the heat generation rate (for complete CO2 conver- and TOS = 10,000 h are shown in Fig. 2; TOS stands for time on
172 D. Sun et al. / Chemical Engineering Journal 329 (2017) 165–177

Fig. 6. Temporal evolution of the reactor outlet composition as dry basis mole fractions (a) and CH4 yield, Eq. (27a) (b) for the biogas-fed reactor. Parameters: Tf = 650 K,
GHSV = 5000 h1, Pf = 2 bar, Gc = 0.4Gc,0.

Table 2
Operating conditions.

H2/CO2 Tf [K] Tcf [K] Gc/Gc,0 Pf [bar] GHSV [h1] TOS [h]
4/1 650–700 415 0.08–0.77 1–20 1000–25,000 10–10,000

H2/CO2 denotes the ratio of feed mole fraction of H2 to CO2; Tf and Tcf stand for the feed temperature of reactants and coolant. Variable parameters are the normalized molten
salt flow rate (Gc/Gc,0), feed pressure (Pf), space velocity (GHSV), and time-on-stream (TOS).

stream. At TOS = 10 h, Fig.2a, the simulation predicts that reactions GHSV = 1000 h1 (Fig. 2), CO content in the outlet stream is negli-
mainly take place in the vicinity of the reactor entrance as it is evi- gible, as it can be seen from the nearly complete selectivity to CH4,
dent from sharp consumption of H2 and CO2 accompanied by a Fig. 3a. The absence of significant amounts of CO can be attributed
spike in the temperature profile. to the relatively low reactor temperature, Fig. 2b, which is achieved
The hot spot is formed at the reactor entrance and the differ- due to the active heat removal via the multiple molten salt tubes,
ence between the temperatures in the packed bed and molten salt Fig. 1. There is still a hot spot formation, but its temperature
compartments is significant, indicating that the rate of heat gener- decreases over the course of time, Fig. 2b.
ation is significantly higher than the rate of inter-compartment The decrease of temperature is due to catalyst deactivation that
heat transfer. For z/L > 0.4, nearly same temperatures are obtained results in significantly lower reaction rates, Fig. 3c. The reaction
in the packed bed and molten salt compartments. A small differ- mainly takes place at the location of the hot spot, which is
ence in the outlet temperatures (slightly lower for the packed expected due to the mutual dependence of reaction and heat gen-
bed) is because of heat losses to the environment, Eq. (8). The out- eration rates: lower reaction rates results in less heat generation
let stream contains mainly H2O and CH4, alongside with a rela- and otherwise. The catalyst remains fully active at the reactor
tively small fractions of H2 and CO2, upper panel in Fig. 2a. A entrance due to the high concentration of H2 in the reactor feed,
small amount of CO, which is formed at the hot spot, is consumed Eqs. (18), (19). However, as the reaction takes place, Fig. 3c, the cat-
downstream the reactor via either water gas shift or CO methana- alyst activity drops sharply forming a steep front, Fig. 3d. Down-
tion, Eqs. (1), (2). stream the reactor, there is certain recovery in the catalyst
After TOS = 10,000 h, very different spatial profiles are formed, activity due to the lower temperature; recall the exponential
Fig.2b, which can be attributed to catalyst deactivation. This obser- dependence of the deactivation rate on temperature Eqs. (18),
vation emphasizes the importance of performing dynamic simula- (19). The initial sharp decay in activity in almost the entire reactor,
tions that account for the temporal decay of catalyst activity, Eqs. Fig. 3d, is associated with the highly nonlinear (exponential)
(7), (8), as opposed to steady state simulations. By analyzing spatial dependence of the catalyst activity on time, Eq. (18). To interpret
profiles at TOS = 10,000 h one can conclude that the catalyst deac- the mathematical formulation, less active sites remain available
tivation affects the reactor performance very significantly: no CO for the reaction as a result of the catalyst deactivation, slowing
formed and the reactions take place over the entire catalytic bed, down the reaction rate. The decline in the reaction rate slows down
gradually consuming H2 and CO2. As a result, even though much the rate of deactivation resulting in stabilization, Fig. 3d.
less CO2 is converted in the vicinity of the reactor entrance, CH4 In summary, the effect of catalyst deactivation is clearly observ-
production only drops by less than 30%, as it can be deduced from able in our model. Note that since we assumed catalyst coking as
analyzing outlet CH4 fractions (compare upper panels in the only mechanism for catalyst deactivation, Eqs. (17)–(19), the
Fig.2a and b). The effect of catalyst deactivation is also evident catalyst activity profile in Fig. 3d reversely coincide with the nor-
from comparing the temperature profiles in the packed bed at malized carbon deposition profile (carbon deposited normalized
TOS = 10 and 10,000 h, Fig. 2. Lower maximum temperature at by the maximum possible deposited carbon amount). If the cata-
TOS = 10,000 h can be attributed to the lower catalyst activity lyst is fully active, there will be no carbon deposited, if the activity
resulting in less heat generation. is very low, the catalyst surface will be almost completely covered
Temporal evolution of the reactor performance at elevated by carbon. The deactivation time scale is relatively long for the
space velocity (GHSV = 25,000 h1) is analyzed in Fig. 3. The effect operating parameters used in the simulations (Table 2) and for
of catalyst deactivation is clearly observable in Fig. 2a: CO2 conver- the adopted deactivation rate expression, Eq. (19). Importantly,
sion and CH4 yield decline significantly over time (shown in log though the catalyst is almost fully deactivated in the vicinity of
scale). Note that XCO2 and YCH4 shown in this plot are calculated the hot spot, Fig. 3d, there is a small region at the reactor entrance
at the reactor outlet (z/L = 1). Similar to the case of where the activity remains almost complete due to the high
 D. Sun et al. / Chemical Engineering Journal 329 (2017) 165–177 173

concentration of H2 in the reactor feed. As a result, the reaction still nation reactor and the product stream is expected to contain CO2,
takes place, though at a significantly lower rate, Fig. 3c. For long CH4, CO and H2. Representative profiles (on dry basis) are shown in
operation times, the activity and reaction profiles are stabilizing Fig. 6 (same operating conditions as in Fig. 5). In the beginning of
and the reactor still operates at 55% CH4 yield after 10,000 h the reactor operation the simulation predicts the product stream
TOS. Still, the reactor performance drops by 30% due to the cata- composed of 65% CH4, 30% H2, 5% CO2, and negligible fraction
lyst deactivation, Fig. 3a. of CO. Depending on downstream application, H2 might need to be
removed from the outlet stream, but the CO2 concentration is low.
3.2. Reactor fed with biogas Because of the catalyst deactivation the outlet composition gradu-
ally changes, resulting in 45% CH4 and H2, and slightly above 10%
In this section, reactor fed with biogas (60% CH4, 40% CO2) is of CO2. The CH4 yield, Eq. (27a) is above 80% in the beginning, but
investigated, assuming that catalyst-poisoning impurities and declines to below 60% after 10,000 h on stream, Fig. 6b. To sum up,
water were removed from the feed stream. See Table 2 and related the model predicts that reasonably high CO2 conversions and CH4
description for details on parameters used in the simulations. All yields are obtainable in the molten salt-cooled packed bed reactor
results presented in Section 3.2 were obtained with the simulated fed with raw biogas (after Sulphur removal).
biogas feed; the CH4 yield is calculated using Eq. (27a). Typical pro-
files are shown in Fig. 4, for short-term operation (TOS = 10 h) and 3.2.1. Effect of space velocity
for the extended time on stream (TOS = 10,000 h). The obtained It is highly desirable to operate at high space velocities in order
spatial distributions are quite different from those obtained with to maximize the reactor throughput. However, high space veloci-
the pure CO2 feed, Fig. 2. One of the most noticeable distinctions ties typically result in lower conversions (because of low residence
is the significantly lower temperature of the hot spot (lower panels times) and, in some case, can result in reactor overcooling, causing
in Fig. 4). extinction. Higher rates of CH4 production will also accelerate cat-
In principle, lower temperatures could result from the alyst deactivation. Therefore, it is of crucial importance to deter-
endothermic steam reforming reactions, right-to-left in Eqs. (1), mine reactor performance over the wide range of space
(3), i.e. CH4 reacting with H2O produced in the methanation reac- velocities. In Fig. 7, the reactor performance is evaluated as a func-
tion. However, from the analysis of the spatial profiles (upper pan- tion of space velocity for different operation times.
els in Fig. 4) it can be concluded that the lower temperature is At low space velocities (GHSV < 5000 h1), CH4 yields up to 95%
rather the result of lower CO2 concentration (dilution) imposed are predicted in the beginning of operation, declining to 60–80%
by the use of the biogas feed and 1/4 CO2/H2 stoichiometry. These yield after 10,000 h on stream, Fig. 7a. In fact, operating at
constrains result in the CO2 feed concentration of 2/13, as opposed GHSV = 1000 h1 (residence time less than 5 s) allows obtaining
to 1/5 for the case of pure CO2 feed. CH4 yield in the 80–95% range over the 10,000 h operation period.
For TOS = 10 h, the simulation predicts that the outlet stream is As the reactor throughput is increased beyond GHSV = 5000 h1,
composed of mainly CH4 and water, also containing less than 10% the reactor performance continues to decline, still allowing work-
of H2 and a small fraction of unreacted CO2. After 10,000 h on ing in the 50–80% CH4 yield range. Above GHSV = 20,000 h1, CH4
stream, CO2 conversion drops significantly, as it is evident from production rate drops sharply.
higher CO2 and H2 fractions in the outlet stream, Fig.4b. Similar The effect of the catalyst deactivation is apparent from the
to Fig. 2, this change is attributed to the catalyst deactivation, as observation of temporal evolution, Fig. 7b. At low space velocity
it can be also seen from the disappearance of the hot spot near (1000 h1), when CH4 production rate is relatively low, catalyst
the reactor entrance, lower panel in Fig.4b. While for the short deactivation is slow, losing only 15% of CH4 yield over the course
operation time most of H2 and CO2 are consumed in the first third of 10,000 h. For GHSV = 10,000 h1, the CH4 yield is lower from the
of the reactor (upper panel in Fig.4a), reactant consumption is beginning which can be attributed to kinetic limitations (shorter
more distributed after 10,000 h on stream, (upper panel in Fig.4b). residence times, less than 1 s). The slope of the decay is similar
Spatiotemporal profiles of the reaction compartment tempera- to that obtained with GHSV = 1000 h1. For GHSV = 20,000 h1,
ture and catalyst activity for higher space velocity (GHSV = the starting CH4 yield is still above 60%, but the decay slope is
5000 h1) are shown in Fig. 5. The temperature distribution is qual- much sharper. For GHSV = 25,000 h1, the decay is slow, but it is
itatively similar to that obtained with the pure CO2 feed, Fig. 3b, attributed to very low CO2 conversion; under these conditions
though significantly lower temperatures are obtained. The activity the reactor is clearly limited kinetically or/and overcooled. Since
profile, on the other hand, is qualitatively different; compare in all simulations the feed temperature was 650 K or higher, the
Fig. 5b with Fig. 3d. First of all, for the case of the biogas feed there overcooling should be attributed with the heat removal by the
is a decline of the catalyst activity at the reactor entrance, which molten salt (discussed in more detail in Section 3.2.3).
can be attributed to the presence of a significant fraction of CH4
in the feed. The sharp front at the reactor entrance is still formed. 3.2.2. Effect of pressure
Second, while the activity declines at the inlet, it steadily increases Using raw biogas as a feed could be an attractive alternative to
in the entire catalytic bed over the course of time. the use of pure CO2 as a feedstock, since it does not require costly
This can be explained by the mutual feedbacks between the CO2 separation; gas clean-up to remove H2S will be still required.
reaction rate, the rate of heat generation, and catalyst deactivation However, raw biogas contains a high fraction of CH4 (60%) which
rate. As the catalyst is being deactivated, the reaction rate can cause catalyst deactivation immediately at the reactor
decreases, resulting in lower temperatures. Lower temperatures entrance due to the high partial pressure of CH4 accompanied by
result in a slower reaction rate, less CH4 production and, therefore, the high feed temperature, Eq. (19). High operating pressures will
slower deactivation. This interplay eventually slows down the pro- result in high CH4 partial pressures over the entire catalytic bed
cess of catalyst deactivation, leading to the recovery and stabiliza- affecting the catalyst deactivation rate. On the other hand, in many
tion of the catalyst activity profile. These finding emphasize the cases it is beneficial to run a process at elevated pressures to max-
importance of the use of dynamic models for the design of reactors imize the process throughput. The effect of the operating pressure
that involve catalytic systems susceptible to deactivation. on the reactor performance is investigated in this section.
For practical application, it will be important to identify the Fig. 8 demonstrates the effect of feed pressure on the CH4 yield
composition of the outlet stream over the course of the reactor at GHSV = 10,000 h1, Fig. 8a, showing also corresponding catalyst
operation. Water can be easily condensed downstream the metha- activity profiles obtained at TOS = 10,000 h, Fig.8b. It can be seen
174 D. Sun et al. / Chemical Engineering Journal 329 (2017) 165–177

Fig. 7. Effect of space velocity on the reactor performance for the biogas-fed reactor: CH4 yield, Eq. (27a), versus GHSV for TOS = 100, 1000, and 10,000 h (a) and temporal
evolution for GHSV = 1000, 10,000, 20,000, 25,000 h1 (b). Parameters: Tf = 650 K, Pf = 2 bar, Gc = 0.4Gc,0.

that increasing the feed pressure initially results in a relatively activity profiles at TOS = 10,000 h, Fig. 8b, shows an interesting
small decrease in the CH4 yield, Fig. 8a. This is apparently in con- trend; the activity actually increases with pressure, except for
tradiction with the Sabatier reaction equilibrium that predicts the small region in the vicinity of the reactor inlet.
higher CO2 conversions for increasing pressures since there is a These results can be interpreted by examining the spatiotempo-
decrease in number of moles, Eq. (3); pressure drop, Eq. (12), ral profiles of the catalytic bed temperature and catalyst activity,
was negligible. Fig. 9 (for Pf = 10 bar). Strong catalyst deactivation at the reactor
However, high pressures are expected to accelerate catalyst entrance is evident, as it is expected for the reactor fed with biogas.
deactivation. Indeed, it can be seen from Fig. 8a that not only the Since the rate of this deactivation is pressure dependent, Eq. (19),
initial CH4 yield but also the rate of its decline are higher for higher the decay of the catalyst activity and, therefore, of the CH4 yield
pressures, indicating that catalyst deactivation plays role in poor is expected to be faster for higher pressures, as it is observed in
reactor performance at high pressures. Examining the catalyst Fig. 8a. The catalyst deactivation also results in the temperature

Fig. 8. Effect of feed pressure on methane yield (a), and spatial profiles of catalyst activity for TOS = 10,000 h (b), for the biogas-fed reactor. Parameters: Tf = 700 K,
GHSV = 10,000 h1, Gc = 0.4Gc,0.

Fig. 9. Spatiotemporal profiles of catalyst activity (a) and packed bed temperature (b) for the biogas-fed reactor. Parameters: Tf = 700 K, Pf = 10 bar, GHSV = 10,000 h1,
Gc = 0.4Gc,0.
 D. Sun et al. / Chemical Engineering Journal 329 (2017) 165–177 175

decrease, Fig. 9b. Eventually, the temperature near the reactor increasing operating pressure is not necessary a good strategy for
entrance is almost equal to the feed temperature, meaning that the reactor fed with biogas, because of high CH4 content in the feed
no ignition occurs. The bed temperature gradually decreases and accelerated rate of catalyst deactivation induced by high par-
downstream the reactor due to the heat removal by the molten salt tial pressures of CH4.
cooling tubes, eventually resulting in temperatures below 700 K.
These temperatures are too low to obtain high conversion. The rea- 3.2.3. Effect of cooling rate
son for the high catalyst activity in the second half of the reactor, Efficient heat removal is of vital importance for the operation of
Fig. 9a, is that the temperature is relatively low there resulting in the highly exothermic Sabatier reactor, but it can also affect the
a low rate of the CH4 cracking and coking, Eq. (19). To sum up, catalyst activity indirectly. Heat removal decreases the reactor

Fig. 10. Reactor performance versus heat removal rate: CH4 yield, Eq. (27a), as a function of the normalized molten salt flow rate, Eq. (29), for TOS = 100, 1000, and 10,000 h
(a) and temporal evolution for Gc/Gc,0 = 0.08, 0.40, 0.54, and 0.88 (b); for the biogas-fed reactor. Parameters: Tf = 650 K, Pf = 2 bar, GHSV = 10,000 h1.

Fig. 11. Reactor performance for the biogas-fed reactor in terms of CH4 yield, Eq. (27a) in the 2D operation domain of reactor throughput (GHSV) and heat removal rate
(Gc/Gc,0) for TOS = 100 h (a), 1000 h (b), and 10,000 h (c). Fixed parameters: Tf = 650 K, Pf = 2 bar.
176 D. Sun et al. / Chemical Engineering Journal 329 (2017) 165–177

temperature favoring CH4 production according to the thermody- sidered inter-compartment heat exchange and catalyst deactiva-
namic equilibrium, since the Sabatier reaction is highly exother- tion by coking, assuming the use of the Ni/Al2O3 catalyst. The
mic, Eq. (3). However, increased CH4 production can potentially reactor performance was investigated for the case of the feed con-
lead to faster catalyst deactivation due to CH4 cracking and coking taining CO2 and H2, and for the case of the reactor fed with biogas
of the catalyst surface. In all simulations results shown until this (40% CO2 and 60% CH4) and H2, in terms of CO2 conversion and CH4
point, the cooling rate was set to an intermediate value of Gc = 0.4- yield. The effects of space velocity, operating pressure, and molten
Gc,0; Gc,0 is defined by Eq. (29). In this section, the effect of the cool- salt flow rate on the reactor performance were investigated in
ing rate on the reactor performance is investigated. detail for the case of the reactor fed with biogas. The CH4 yield
In Fig. 10, the reactor performance is evaluated in terms of CH4 was mapped in the dynamic two-dimensional domain of the two
yield, Eq. (27a), as a function of the normalized cooling rate, Gc/Gc,0, crucial operating parameters: space velocity and cooling rate.
Eq. (29). It can be seen that increasing the cooling rate promotes The model predicts that the reactor can be operated with CH4
CH4 production due to the more efficient heat removal, Fig. 10a. yields ranging from 80 to 90% at intermediate space velocities
The improvement is very significant, ranging from 60 to 90% over ranging from 1000 to 10,000 h1 for extended periods of operation,
the course of the 1000 h operation period. However, for extended up to 10,000 h time on stream. Further increase in reactor through-
operation (TOS = 10,000 h) the increase of the cooling rate above put results in a sharp decline of the reactor performance mainly
GC/GC,0 = 0.54 actually reduces the CH4 production, resulting in a due to the kinetic limitations and overcooling. High pressures,
distinct optimum, Fig. 10a. This finding can be attributed to the which are favorable for CH4 production according to the equilib-
effect of catalyst deactivation that is accelerated at higher CH4 pro- rium and beneficial for practical applications, actually causes faster
duction rates. It is also evident from the examination of the tempo- catalyst deactivation due to the higher partial pressure of CH4,
ral evolution, Fig. 10b. For GC/GC,0 = 0.88, the decay rate is initially resulting in a poor reactor performance.
similar to those obtained with slower cooling, but after approxi- The molten salt flow rate, i.e. cooling rate, is a crucial parameter
mately 5000 h TOS, the slope changes and the deactivation affecting the reactor performance. The model predicts that the CH4
becomes rapid. This finding indicates again the importance of the yield can be improved by 60–90% over short operating periods. For
use of dynamic modeling for catalytic systems subject to extended, >1000 h time on stream operation, the effect of heat
deactivation. removal is more complicated and distinct maxima in CH4 produc-
tion are observed. Scanning the reactor performance over extended
ranges of space velocities (reactor throughput) and molten salt
3.2.4. Optimization of heat removal and reactor throughput
flow rates (heat removal rate) shows a non-trivial, dynamically
The existence of the catalyst deactivation makes the problem of
evolving map. An algorithm for the reaction performance evalua-
the reactor performance optimization rather complicated, since
tion is currently under development.
optimal conditions could change during the course of reactor oper-
ation. The two important parameters, which and relatively easily
adjustable in a practical situation, are the space velocity (reactor Acknowledgements
throughput) and the molten salt flow rate (cooling rate). In this
section, the reactor performance in terms of CH4 yield is optimized The authors acknowledge the support of the Waterloo Institute
using a dynamic scan over the 2D operational window of the cool- of Sustainable Energy (WISE), Ontario, Canada, through the WISE –
ing rate vs. space velocity, Fig. 11. Cisco Systems Smart Grid Research Fund.
The analysis of the obtained maps shows that obtaining 80–90%
CH4 yields over extended periods of continuous operation should
be possible. For that, the reactor should be operated below References
GHSV = 5000 h1 and above Gc/Gc,0 = 0.5. For short-term operation,
[1] W. Wang, S. Wang, X. Ma, J. Gong, Recent advances in catalytic hydrogenation
it is possible to operate at elevated space velocities, with cooling of carbon dioxide, Chem. Soc. Rev. 40 (2011) 3703–3727.
rates ranging from Gc/Gc,0 = 0.3–0.6. The short-term operation [2] T. Inui, Highly effective conversion of carbon dioxide to valuable compounds
strategy is in principle possible if there is a possibility of periodic on composite catalysts, Catal. Today 29 (1996) 329–337.
[3] G.D. Weatherbee, C.H. Bartholomew, Hydrogenation of CO2 on group VIII
regeneration of the catalytic bed. For intermediate space velocities metals: IV. Specific activities and selectivities of silica-supported Co, Fe, and
ranging from GHSV = 10,000–20,000 h1, there are distinct max- Ru, J. Catal. 87 (1984) 352–362.
ima in the CH4 yield versus cooling rate. [4] Y. Zhang, G. Jacobs, D.E. Sparks, M.E. Dry, B.H. Davis, CO and CO2 hydrogenation
study on supported cobalt Fischer-Tropsch synthesis catalysts, Catal. Today 71
Generally speaking, analysis of such maps can determine the (2002) 411–418.
operation window where the maximal CH4 yield can be obtained [5] P.J. Lunde, F.L. Kester, Carbon dioxide methanation on a ruthenium catalyst,
over a certain period of operation. Depending on demand, various Ind. Eng. Chem. Proc. Des. Dev. 13 (1974) 27–33.
[6] K.R. Thampi, J. Kiwi, M. Graetzel, Methanation and photo-methanation of
strategies can be implemented. For example, a certain domain carbon dioxide at room temperature and atmospheric pressure, Nature 327
where initially the CH4 yield is very high but decays significantly (1987) 506–508.
at 10,000 h can be selected for short-term operation. For extended [7] F. Solymosi, A. Erdöhelyi, T. Bansagi, Methanation of CO2 on supported
rhodium catalyst, J. Catal. 68 (1981) 371–382.
operation times, it is beneficial to select a domain where the CH4
[8] G.A. Mills, F.W. Steffgen, Catalytic methanation, Catal. Rev. 8 (1974) 159–210.
yield is maintained relatively high over the entire period of opera- [9] C. Bartholomew, Catalyst deactivation and regeneration, Kirk-Othmer
tion. The selection can be quantitatively constrained by maximiz- Encyclopedia of Chemical Technology, 2003.
[10] R. Dalla Betta, A. Piken, M. Shelef, Heterogeneous methanation: steady-state
ing the overall CH4 production, integrated over the entire period
rate of CO hydrogenation on supported ruthenium, nickel and rhenium, J.
of time. Further investigation is currently underway in order to Catal. 40 (1975) 173–183.
develop a thorough optimization algorithm. [11] M. Agnelli, M. Kolb, C. Mirodatos, CO hydrogenation on a nickel catalyst.: 1.
Kinetics and modeling of a low-temperature sintering process, J. Catal. 148
(1994) 9–21.
[12] M.N. Pedernera, J. Piña, D.O. Borio, Kinetic evaluation of carbon formation in a
4. Concluding remarks
membrane reactor for methane reforming, Chem. Eng. J. 134 (2007) 138–144.
[13] J.B. Claridge, M.L. Green, S.C. Tsang, A.P. York, A.T. Ashcroft, P.D. Battle, A study
A mathematical model of the Sabatier reactor for conversion of of carbon deposition on catalysts during the partial oxidation of methane to
CO2 into synthetic CH4 was defined and analyzed by numerical synthesis gas, Catal. Lett. 22 (1993) 299–305.
[14] K. Nikooyeh, R. Clemmer, V. Alzate-Restrepo, J.M. Hill, Effect of hydrogen on
simulations. The reactor was a heat-exchanger type packed bed carbon formation on Ni/YSZ composites exposed to methane, Appl. Catal. A:
cooled by multiple tubes with flowing molten salt. The model con- Gen. 347 (2008) 106–111.
 D. Sun et al. / Chemical Engineering Journal 329 (2017) 165–177 177

[15] J.-W. Snoeck, G. Froment, M. Fowles, Steam/CO2 reforming of methane. Carbon [28] M.L. Rodríguez, M.N. Pedernera, D.O. Borio, Two dimensional modeling of a
filament formation by the Boudouard reaction and gasification by CO2, by H2, membrane reactor for ATR of methane, Catal. Today 193 (2012) 137–144.
and by steam: kinetic study, Ind. Eng. Chem. Res. 41 (2002) 4252–4265. [29] S. Lee, J.G. Speight, S.K. Loyalka, Gasification of coal, Handbook of Alternative
[16] M. Ermakova, D.Y. Ermakov, G. Kuvshinov, L. Plyasova, New nickel catalysts for Fuel Technologies, second ed., CRC Press, Boca Raton, FL, 2014.
the formation of filamentous carbon in the reaction of methane [30] M.C. Demicheli, E.N. Ponzi, O.A. Ferretti, A.A. Yeramian, Kinetics of carbon
decomposition, J. Catal. 187 (1999) 77–84. formation from CH4-H2 mixtures on nickel-alumina catalyst, Chem. Eng. Res.
[17] J.R. Rostrup-Nielsen, Steam Reforming Catalysts: An Investigation of Catalysts 46 (1991) 129–136.
for Tubular Steam Reforming of Hydrocarbons: A Contribution From the [31] R.J. Berger, J. Pérez-Ramı´rez, F. Kapteijn, A. Moulijn, Catalyst performance
Research Laboratory of Haldor Topsøe A/S, Teknisk Forlag, 1975. testing: radial and axial dispersion related to dilution in fixed-bed laboratory
[18] D.S.A. Simakov, M. Sheintuch, Model-based optimization of hydrogen reactors, Appl. Catal. A Gen. 227 (2002) 321–333.
generation by methane steam reforming in autothermal packed-bed [32] A.G. Dixon, D.L. Cresswell, Theoretical prediction of effective heat transfer
membrane reformer, AIChE J. 57 (2011) 525–541. parameters in packed beds, AIChE J. 25 (1979) 663–676.
[19] D.S.A. Simakov, M. Sheintuch, Design of a thermally balanced membrane [33] O.R. Derkx, A.G. Dixon, Effect of the wall Nusselt number on the simulation of
reformer for hydrogen production, AIChE J. 54 (2008) 2735–2750. catalytic fixed bed reactors, Catal. Today 35 (1997) 435–442.
[20] M. Borghei, R. Karimzadeh, A. Rashidi, N. Izadi, Kinetics of methane [34] O.A. Sergeev, A.G. Shashkov, A.S. Umanskii, Thermophysical properties of
decomposition to COx-free hydrogen and carbon nanofiber over Ni–Cu/MgO quartz glass, J. Eng. Phys. Thermophys. 43 (1982) 1375–1383.
catalyst, Int. J. Hydrogen Energy 35 (2010) 9479–9488. [35] S.W. Churchill, H.H. Chu, Correlating equations for laminar and turbulent free
[21] S. Rönsch, J. Schneider, S. Matthischke, M. Schlüter, M. Götz, J. Lefebvre, P. convection from a horizontal cylinder, Int. J. Heat Mass Transfer 18 (1975)
Prabhakaran, S. Bajohr, Review on methanation – from fundamentals to 1049–1053.
current projects, Fuel 166 (2016) 276–296. [36] V. Gnielinski, New equations for heat and mass transfer in turbulent pipe and
[22] S.A.M. Said, D.S.A. Simakov, M. Waseeuddin, Y. Roman-Leshkov, Solar molten channel flow, Int. Chem. Eng. 16 (1976) 359–368.
salt heated membrane reformer for natural gas upgrading and hydrogen [37] A.F. Mills, Heat Transfer, second ed., Prentice-Hall, New Jersey, 1999.
generation: a CFD model, Solar Energy 124 (2016) 163–176. [38] J.P. Holman, Heat Transfer, ninth ed., McGraw-Hill Inc, New York, Boston,
[23] B. Hou, Y. Huang, X. Wang, X. Yang, H. Duan, T. Zhang, Optimization and 2002.
simulation of the Sabatier reaction process in a packed bed, AIChE J. 62 (2016) [39] G.J. Janz, Molten Salts Handbook, Elsevier, New York, 1967.
2879–2892. [40] Y.S. Touloukian, R.K. Kirby, E.R. Taylor, T.Y.R. Lee, Thermophysical Properties of
[24] D. Schlereth, O. Hinrichsen, A fixed-bed reactor modeling study on the Matter, Purdue University, The TPRC Data Series, 1977.
methanation of CO2, Chem. Eng. Res. Des. 92 (2014) 702–712. [41] R. Morrell, Handbook of Properties of Technical and Engineering Ceramics,
[25] L. Kiewidt, J. Thöming, Predicting optimal temperature profiles in single-stage Stationery Office Books, London, 1987.
fixed-bed reactors for CO2-methanation, Chem. Eng. Sci. 132 (2015) 59–71. [42] Dynalene Molten Salts, Dynalene, Inc., https://www.dynalene.com/Molten-
[26] J. Xu, G.F. Froment, Methane steam reforming, methanation and water-gas Salts-s/1831.htm, 2016.
shift: I. intrinsic kinetics, AIChE J. 35 (1989) 88–96. [43] D.S.A. Simakov, M.M. Wright, S. Ahmed, E.M.A. Mokheimer, Y. Román-Leshkov,
[27] S.S.E.H. Elnashaie, A.M. Adris, A.S. Al-Ubaid, M.A. Soliman, On the non- Solar thermal catalytic reforming of natural gas: a review on chemistry,
monotonic behaviour of methane—steam reforming kinetics, Chem. Eng. Sci. catalysis and system design, Catal. Sci. Technol. 5 (2015) 1991–2016.
45 (1990) 491–501.