Journal of CO₂ Utilization 21 (2017) 368–382

Contents lists available at ScienceDirect

Journal of CO2 Utilization

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

Thermal management of a Sabatier reactor for CO2 conversion into CH4: T

Simulation-based analysis
⁎
Duo Sun, David S.A. Simakov
Department of Chemical Engineering, University of Waterloo, Waterloo, ON, N2L 3G1, Canada

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

Keywords: Converting CO2-rich waste streams such as raw biogas, landfill gas and power plant flue gas into synthetic fuels
CO2 and chemicals will reduce greenhouse gas emissions, providing revenue at the same time. One option is to
Renewable natural gas convert CO2 into CH4 by hydrogenation via Sabatier reaction. This synthetic methane will be renewable if the H2
Sabatier reactor required for the reaction is generated via water electrolysis using solar and wind energy or hydroelectricity.
Molten salt
However, to realize the potential of this approach, a number of technological challenges related to the Sabatier
reactor design have to be resolved, including thermal management. The high exothermicity of the Sabatier
reaction can lead to reactor overheating. High temperatures are unfavorable to the exothermic and reversible
methanation process, resulting in low CO2 conversions. A simulation-based study of a Sabatier reactor was
performed in order to optimize the removal of heat, while maximizing CO2 conversion and CH4 production. The
heat-exchanger type packed bed reactor with internal cooling by a molten salt was simulated using a transient,
pseudo-homogeneous mathematical model. Reactor performance was evaluated in terms of CO2 conversion and
CH4 yield. The simulation results show that feed temperature, feed flow rate, and molten salt flow rate are the
crucial parameters affecting the reactor performance. For the optimized operating conditions, the model predicts
CO2 conversions and CH4 yields above 90% at high reactor throughputs, with space velocities up to 10,000 h−1.
A preliminary techno-economic evaluation is provided: opportunities and challenges are outlined.

1. Introduction electrochemical reduction of CO2 into a variety of products, including

formic acid and methanol, in an aqueous environment using water as a
The intensive use of fossil fuels leads to growing CO2 emissions that source of H2 for CO2 reduction [1,2]. This approach is apparently at-
accelerate Global Warming. It is absolutely necessary to capture an- tractive as it only uses water and CO2 as the starting materials and the
thropogenic CO2. Once captured, CO2 can be stored (e.g., in geological sun and (potentially) renewable electricity as a source of energy.
formations) but this approach requires high capital and operating costs, However, photochemical CO2 reduction has inherent limitations of
hindering its widespread commercialization. The alternative to CO2 solar energy utilization, while electro-chemical reduction of CO2 has
storage is CO2 conversion into synthetic fuels and chemicals that closes low efficiencies of electricity utilization. Both processes are limited by
the artificial carbon cycle. Being a harmful greenhouse gas, CO2 is also low CO2 solubility in water and have severe diffusion limitations [1,2].
an excellent source of carbon, non-flammable and non-corrosive. An alternative approach is thermocatalytic conversion that com-
Resources are abundant, including flue gases from coal- and natural bines the use of high temperatures with a heterogeneous catalyst,
gas-fired power plants, biogas and landfill gas (up to 40% CO2), and off- providing fast reaction rates and, therefore, allowing for large volume
gas streams in several industrial processes such as ammonia production production [3]. One of the thermocatalytic routes for CO2 conversion is
and fermentation. There are also large resources of CO2 accompanying methanation, also called the Sabatier reaction [3,4]. This reaction
natural gas and oil production wells; this CO2 is typically vented into converts CO2 into CH4 by hydrogenation, Eqs. (1)-(3) below. This
the atmosphere. synthetic methane will be renewable if the H2 required for the reaction
Conversion of captured CO2 into synthetic fuels and chemicals is an is produced via water electrolysis using renewable electricity (solar,
attractive avenue for reduction of CO2 emissions and, as such, this topic wind, or hydro), making the process sustainable. Despite the abundance
has attracted the interest of many research groups around the world of natural gas in certain countries (also thanks to fracking) and current
[1–3]. Many research efforts have focused on photochemical and low prices of fossil methane, CO2 conversion into renewable natural gas

⁎
Corresponding author.
E-mail address: dsimakov@uwaterloo.ca (D.S.A. Simakov).

http://dx.doi.org/10.1016/j.jcou.2017.07.015
Received 14 October 2016; Received in revised form 23 June 2017; Accepted 17 July 2017
Available online 18 August 2017
2212-9820/ © 2017 Elsevier Ltd. All rights reserved.
D. Sun, D.S.A. Simakov Journal of CO₂ Utilization 21 (2017) 368–382

Nomenclature GHSV Gas hourly space velocity, h−1

t Time, s
ac,HE Cooling tube surface-to-volume ratio, m−1 T Reactor temperature, K
ar,HE Cooling tube surface-to-packed volume ratio, m−1 Tc Coolant temperature, K
ar,HL Reactor surface-to-volume ratio, m−1 Te Environment temperature, K
Ac Total cross-sectional area of cooling tubes, m2 Uw Overall effective wall heat transfer coefficient, kJ/(m2 s K)
Aj Pre-exponential factor of the rate coefficient of reaction j, v Fluid velocity, m/s
units of kj V Compartment volume, m3
Bj Pre-exponential factor of the adsorption coefficient of yi Mole fraction of species i
species i, units of Kj z Reactor length coordinate, m
Ci Molar concentration of species i, mol/m3
Ct Total molar concentration, mol/m3 Greek letters
Cpc Coolant heat capacity, kJ/(kg K)
Cpg Gas heat capacity, kJ/(mol K) ε Packed bed porosity
d Wall thickness, m ϕ Thiele modulus
dp Catalytic pellet diameter, m ηj Effectiveness factor of reaction j
D Diameter, m ηST Steam turbine efficiency
Dae Effective axial dispersion coefficient, m2/s λ Thermal conductivity, kW/(m K)
Dm Gas molecular diffusivity, m2/s μ Viscosity, kg/(m s)
Ej Activation energy of reaction j, kJ/mol ρc Coolant density, kg/m3
Fi Molar flow rate of species i, mol/s ρg Gas molar density, mol/m3
Ftf Total feed molar flow rate, mol/s ρs Solid density, kg/m3
G Gravimetric (mass) flow rate, kg/s τb Packed bed tortuosity
hnc Natural convection heat transfer coefficient, kJ/(m2 s K)
hw Effective wall heat transfer coefficient, kJ/(m2 s K) Subscripts
ΔHi Adsorption enthalpy change of species i, kJ/mol
kae Effective axial heat dispersion coefficient, kJ/(m s K) c Coolant
kj Rate constant of reaction j eff Effective
Ki Adsorption constant of species i, bar−1 eq Equilibrium
Kj,eq Equilibrium constant of reaction j f Feed
L Reformer length, m g Gas
Nc Number of cooling tubes HE Heat exchange
Nu Nusselt number HL Heat loss
pi Partial pressure of gaseous species i, bar int Initial
P Pressure, bar MS Molten salt
Pr Prandtl number nc Natural convection
Re Reynolds number PB Packed bed
Rj Rate of reaction j, mol/(kg s) s Solid
Rg Gas constant, kJ/(mol K) r Reactor

(RNG) is still a useful strategy. CO2 + H2 ⇄ CO + H2 O ΔH∘298 = 41.2 kJ/mol (2)

First, this approach creates an artificial carbon cycle (similar to
photosynthesis if solar energy is used for H2 production) recycling the CO + 3H2 ⇄ CH 4 + H2 O ΔH∘298 = − 206.1 kJ/mol (3)
otherwise emitted CO2 in our energy distribution systems and, there-
The overall process is highly exothermic which requires efficient
fore, reducing the consumption of the fossil methane (natural gas).
heat removal, in order to facilitate CH4 production. In addition to this
Second, the produced RNG can be injected into the existing natural gas
thermodynamic constrain, reactor overheating can also lead to fast
infrastructure that provides virtually unlimited storage and transpor-
catalyst deactivation by sintering and coking. For nickel-based cata-
tation capacities not only for renewable energy but also for surplus and
lysts, operating temperatures should be kept below 550 °C to prevent
off-peak electricity. Third, converting CO2 into RNG can provide local
catalyst deactivation [5]. Historically, methanation systems were de-
energy independence, while RNG can be used as a transportation fuel as
signed for conversion of synthesis gas [6], but, recently, CO2 was also
well. Eventually, the renewable methane is combusted to produce
considered as a feedstock for methanation [5]. Two conventional con-
power and heat, and CO2. This CO2 can be potentially captured and
figurations of methanation reactors are fixed bed [7–11] and fluidized
converted back into RNG using renewable energy, thus using this
bed [12–14]. Microchannel reactors [15], monolith reactors [16–18],
carbon as a renewable energy carrier. Even though re-capturing of CO2
and three-phase slurry reactors [19,20] were also suggested as design
might be unsuitable in the near future for some uses, such as residential
solutions for carrying out methanation reactions. A comprehensive
heating, recycling the carbon captured from waste streams (in form of
overview of the abovementioned concepts is available in the literature
CO2) one more time as an energy carrier can reduce the consumption of
[21]. One of the most challenging problems in the design of methana-
fossil methane significantly. The price of the produced RNG is of course
tion reactors is thermal management.
a concern; this issue is addressed in Section 3.4.
For the adiabatic fixed bed reactor design, the primary approach is
The strongly exothermic Sabatier reaction, Eq. (1), is typically ac-
to use cascades of adiabatic reactors (typically 2–6) with intermediate
companied by the mildly endothermic reverse water gas shift, Eq. (2),
cooling and gas recirculation [21]. Though such systems can efficiently
and the exothermic CO methanation, Eq. (3).
control the process temperature, the use of multiple reactors, recycle
CO2 + 4H2 ⇄ CH 4 + 2H2 O ΔH∘298 = − 164.9 kJ/mol compressors, heat exchangers etc. results in system complexity and high
(1)
capital cost investment. An alternative approach is to use a cooled

369
D. Sun, D.S.A. Simakov Journal of CO₂ Utilization 21 (2017) 368–382

packed-bed reactor [9,10]. This method could potentially allow per- conventional heat carriers, such as steam and mineral oils, due to
forming methanation as a single-stage process but heat removal opti- higher operating temperatures, low working pressure, non-flamm-
mization to avoid the hot spot formation is a difficult task. Structured, ability, high density and heat capacity, and, importantly, good heat
microchannel and monolith reactors [15–18] have been developed to conductivity [22]. Mineral oils are unsuitable for removing heat from
overcome the problems of temperature control and overheating in the Sabatier reactor due to high operating temperature. Temperatures
conventional packed-bed reactors. However, while improved heat above 700 K are required to achieve high CO2 conversions in the Sa-
transfer characteristics and, ideally, nearly-isothermal operation sig- batier reaction (as it is shown by numerical simulations below). How-
nificantly simplify process control, the reactor itself becomes more ever, maximum temperatures for mineral oil-based heat transfer fluids
expensive because of the complicated deposition of the catalyst coating do not exceed 600 K. At higher temperatures, these heat transfer fluids
and the difficulty of replacing deactivated catalyst. will decompose. Steam can be theoretically used at high temperatures.
Fluidized-bed reactors [12–14] are widely suggested as ideal iso- However, it will require extremely high pressures to use saturated
thermal reactors due to the excellent transport characteristics, allowing steam which has good heat transfer characteristics. Regardless, hand-
for the use of a single reactor and simplifying process control. However, ling large amounts of steam, which can cause corrosion, fouling and
catalyst attrition remains a major obstacle toward implementation of scaling, is rather disadvantageous from the operating point of view.
fluidized beds. Another disadvantage is that fluidized-bed reactors can Molten salts, on the other hand, have their own disadvantages.
only be operated in a narrow range of operating conditions dictated by Molten salt can solidify if the temperature is below the melting point,
the fluidization velocity, restricting reactor throughputs. Three-phase clogging the tubes. Also, there is an upper limit of operation above
methanation reactors [19,20] provide excellent heat transfer char- which thermal decomposition will occur. To sum up, molten salts could
acteristics, allowing isothermal operation, but are restricted to a narrow be highly-efficient heat transfer fluids for heat removal from a Sabatier
operating window dictated by solvent/heat transfer fluid properties. reactor, due to their high heat capacity, excellent heat transfer char-
Other major drawbacks are the gas-to-liquid mass transfer resistance acteristics and suitable operating temperature range. For similar rea-
and the fluid decomposition and evaporation. sons, molten salts were suggested as a highly-efficient heat transfer
The actively-cooled packed bed reactor design can potentially pro- fluid for the low-temperature solar reforming which has similar oper-
vide a low-cost solution for the single-pass methanation process. ating temperature range [22–24].
However, heat removal optimization is required. Recently reported A conceptual representation of the entire methanation system
modeling studies did not account for coolant temperature variations, comprising the molten-salt cooled Sabatier reactor is shown in Fig. 1.
assuming constant temperature of the heat transfer fluid [9,10]. Such Hydrogen which is required for the Sabatier reaction can be generated
situation is only achievable when the heat transfer fluid flow rate is using renewable energy [25,26]. The reaction heat removed from the
very high [22]. On the other hand, process optimization will require reactor by the heat transfer fluid can be used to preheat the feed, and
minimizing the coolant flow rate, maximizing at the same time the can be also utilized for electricity generation and residential heating,
reactor performance. Fig. 1. In our study we focus first on the mathematical analysis of the
In the presented herein study, the heat-exchanger type Sabatier Sabatier reactor via numerical simulations. The reactor performance is
reactor cooled by molten salt is investigated via numerical simulations evaluated in terms of CO2 conversion and CH4 yield as a function of
in order to optimize the removal of heat, while maximizing CO2 con- operating parameters, such as feed temperature, feed flow rate, and
version and CH4 production rate. Molten salts are advantageous over molten salt flow rate. We performance a sensitivity analysis and a

Fig. 1. Conceptual process flow diagram of the renewable natural gas generation system. The Sabatier reactor converts a mixture of CO2 and H2 into CH4 and water. The reaction heat is
removed from the reactor by the molten salt recirculation system.

370
D. Sun, D.S.A. Simakov Journal of CO₂ Utilization 21 (2017) 368–382

preliminary optimization aimed at maximizing the reactor throughput single molten salt tube is described by Eq. (6). The corresponding
and CH4 yield. Finally, based on the output from the reactor analysis, boundary and initial conditions are listed in Eqs. (7) and (8). Ideal gas
we perform a preliminary techno-economic evaluation at the system behavior was assumed to calculate an average gas density. For the
level. The purpose of this assessment is to estimate the economic via- conditions used in our simulations the difference between the average
bility of the renewable natural gas (RNG) production using the analyzed molar density calculated from tabulated gas densities and the one cal-
reactor as an essential component of the integrated system. culated from the ideal gas law was negligible. Pressure drop was cal-
culated using Ergun equation, Eq. (9), with a fixed inlet pressure (ad-
2. Model formulation justed by a back pressure regulator in a practical situation). Variations
in the gas velocity due to change in number of moles are considered by
The suggested Sabatier reactor design is shown in Fig. 2. The reactor Eq. (10) [29]. Effective heat capacity in Eq. (5) is defined by Eq. (11);
is a heat-exchanger type packed bed internally cooled by molten salt Nc is the number of cooling tubes.
flowing in multiple cooling tubes. To minimize uncontrollable heat ∂Ci ∂ 2C ∂C
losses to the environment, the reactor is insulated by a layer of quartz ε = Dae 2i − εvg i + ρs (1 − ε ) ∑ ηj Rij
∂t ∂z ∂z j (4)
wool. Both the external and internal tubes are made of stainless steel.
The dimensions used in numerical simulations are listed in Table 1. ∂T ∂2T ∂T
(ρCp)eff = kae − ερg Cpg vg ∂z +
Since we used a one-dimensional mathematical model in our study, ∂t ∂z 2
we selected reactor dimensions in such a way that the radial reactor + ρs (1 − ε ) ∑ (−ΔHj ) ηj Rj − Nc Uw, HE ar , HE (T − Tc )
dimension is as small as possible. In order to minimize the heat transfer j
distance (to reduce the space between the cooling tubes), the minimum − Uw, HL ar , HL (T − Te ) (5)
number of cooling tubes was set to Nc = 12. For this number of tubes
and for the dimensions described in Table 1 with evenly distributed ∂Tc ∂ 2T ∂Tc
ρc Cpc = λ c 2c − ρc Cpc vc − Uw, HE ac, HE (Tc − T )
tubes, the distance between any two adjusted cooling tubes will be ∂t ∂z ∂z (6)
19 ± 1 mm for Nc = 12. Thermal conductivity of the packed bed
∂Ci ∂Ci
should prevent large radial gradients over that relatively small distance z=0 εvgf (Ci, f − Ci ) = −Dae z=L =0
∂z ∂z
[11]. The maximum of Nc = 22 was set to allow enough space for ∂T
∂T =0
catalytic pellets between the cooling tubes. Due to the high thermal ερg vgf Cpg (Tf − T) = −kae ∂z ∂z
conductivity of the molten salt, significant radial gradients are not ex- ∂Tc
ρc vcf Cc (Tc, f − Tc ) = −kc
∂Tc
∂z
=0
pected to develop within the coolant tubes (Dc = 0.02 m) [22]. ∂z (7)
Due to the high volumetric heat capacity of the molten salt, only a
relatively small fraction of the total reactor volume is required for the at t = 0: Ci (0, z ) = Ci,int
molten salt compartment (Table 1). The total reactor volume for the T (0, z ) = Tint
suggested dimensions is approximately 30 L, which is suitable for Tc (0, z ) = Tc,int (8)
medium scale applications. As it is later discussed in Section 3.4, the
single reactor is predicted to provide up to 100–500 kW power output, dP (1 − ε )2μg (1 − ε ) ρg 2
= −150 vg − 1.75 vg P (0) = Ptf
providing a potential for MW-scale applications using multiple reactors. dz dp2 ε 3 dp ε 3 (9)
A modular, multiple reactor parallel operation will be advantageous for
maintenance, since reactors can be conditioned or replaced without ∑ Ci (z )
shutting down the entire process. i
vg (z ) = vgf
A transient, one-dimensional, pseudo-homogeneous mathematical ∑ Ci, f
i (10)
model [27,28] was used to simulate the reactor. As opposed to as-
suming constant coolant temperature [9,10], the model accounts for (ρCp)eff = ερg Cpg + (1 − ε ) ρs Cps (11)
temperature variations in the heat transfer fluid. The model also in-
cludes the axial mass and heat dispersion, and accounts for the tem-
perature dependence of thermo-physical properties. Component mass 2.1. Kinetics
balance and energy balance for the packed bed compartment are given
by Eqs. (4) and (5), respectively; the temperature distribution in a Reaction rates are calculated using the commonly adopted kinetics

Fig. 2. Schematic of the molten salt-cooled, heat-

exchanger type packed bed Sabatier reactor.

371
D. Sun, D.S.A. Simakov Journal of CO₂ Utilization 21 (2017) 368–382

Table 1
Reactor dimensions.

Dr (m) Dc (m) dw (m) diw (m) dp (m) Nc L (m) VPB (m3) VMS (m3)

0.2 0.02 0.002 0.05 0.003 12–22 1 0.0255 0.0041

Dr and Dc denote the diameter of the packed bed (Dr) and coolant (Dc) compartments. L is the reactor length and Nc is the number of the molten salt coolant tubes. dw denotes the
thickness of the reactor wall and cooling tube, correspondingly; diw is the insulation layer thickness. V denotes the volume of the packed bed (VPB) and coolant (molten salt) (VMS)
compartments; dp is the catalytic pellet diameter.

−1
for methane steam reforming over the Ni/Al2O3 catalyst [30,31], Eqs. 1 d d 1 ⎞
Uw, HL = ⎛ ⎜ + w + iw + ⎟
(12a)–(12c). The kinetic expressions account for the reversibility of the
⎝ h wr kw kiw hnc ⎠ (16b)
reforming and water gas shift reactions and can be used therefore for
modeling the CO2 methanation reaction system, Eqs. (1)–(3). All The wall heat exchange coefficient between the catalytic bed and
parameters are tabulated in the literature, e.g., see Tables 3 and 4 in the coolant tube, Uw,HE, accounts for resistances of the fixed bed, the
Rodríguez et al. [32]. Intraparticle and interphase transport limitations coolant tube wall, and the molten salt, Eq. (16a). Similarly, the corre-
can be neglected for the size of catalytic pellets used in our simulations lation for the wall heat transfer coefficient, Uw,HL, accounts for re-
(dp = 3 mm, Table 1) [28]. sistances through the catalytic bed, the reactor wall, the quartz wool
insulation layer (Fig. 1) [37], accounting also for natural convection
k1 ⎛ pH3 2 pCO ⎞ 1 (hnc) [38]. Since the insulation layer natural convection resistances are
R1 = p p
2.5 ⎜ CH4 H2 O
−
pH2 K1, eq ⎟ den2 (12a) dominant in Eq. (16b), the wall heat transfer coefficient was nearly
⎝ ⎠
constant in all simulations: Uw,HL ≈ 0.01 W/(m2 K).
k2 ⎛ pH2 pCO2 ⎞ 1 The effective wall heat transfer coefficient for the reaction com-
R2 = ⎜pCO p H O − ⎟
partment (hwr) is estimated using the following correlation obtained in
pH2 ⎝ 2
K2, eq ⎠ den2 (12b)
the similar way as Eq. (15), using a complete set of the original cor-
k3 ⎛ pH2 pCO2 ⎞ 1 relations [27,28,35,36]:
R3 = ⎜p p2 − ⎟
pH3.52 ⎝ CH4 H2 O K3, eq ⎠ den2 (12c) hwr dp vg ρg dp
Nup = λg
= 24 + 0.34 Re0.77
p ; Rep= μg (17)
K H2 O p H2 O
den = 1 + K CO pCO + K H2 p H2 + K CH4 pCH4 + The effective wall heat transfer coefficient for the coolant tube (hwc)
p H2
is estimated using the following correlations [39–41]:

( )
kj = Aj exp
−Ej
Rg T
for Rec < 2030: Nuc = 3.66 +
0.065 Rec Prc (Dc / L)
1 + 0.04[Rec Prc (Dc / L)]2/3 (18a)

K = B exp ( ) −ΔHi
i i Rg T for 2030 < Rec < 4000: Nu wc = 0.012(Rec 0.87 − 280)Prc 0.4 [1

The internal effectiveness factor was calculated using the standard + (Dc / L)2/3] (18b)
expression for a spherical pellet [33], with the Thiele modulus defined
for Rec > 4000: Nuc = 0.027 Rec 0.8 Prc1/3 (18c)
for each reaction with the corresponding reaction rate constant (kj in
Eq. (12)): h wc Dc vc ρc Dc Cpc μc
Nu wc = ; Rec = ; Prc =
3 kˆj dp2 λc μc λc
ηj = ⎛ 1 −
1
⎞; ϕj =
ϕj tanh ϕj ϕj 4Dm
⎝ ⎠
k1 ρs (1 − ε ) k2 ρs (1 − ε ) Ptf k3 ρs (1 − ε )
kˆ1 = ; kˆ2 = ; kˆ3 = 2.3. Numerical procedure
Ptf ρg ε ρg ε Ptf ρg ε (13)
The model was solved using the MATLAB PDE solver with a second-
2.2. Transport parameters order accurate spatial discretization based on a fixed set of user-speci-
fied nodes and time integration done by the stiff ODE solver (ode 15s).
The effective axial mass dispersion coefficient is calculated using the Dependences of thermophysical properties (density, viscosity, heat ca-
following correlation [34]: pacity, diffusivity, and thermal conductivity) on temperature, pressure
and composition were accounted for using polynomial regressions fitted
Dae = ε ( Dm
τb )
+ 0.5dp vg ; τb =
1
ε 0.5 (14)
to the data on thermophysical properties from the literature
[39,42–44]. Molten salt properties were adopted from the data on
The expression for the effective axial heat dispersion coefficient commercially available molten salts (Dynalene, Inc. [45], Dynalene MS-
(kae), Eq. (15), was derived from the heat conductivity correlations 2).
developed for catalytic fixed beds [27,35,36]. Values of kae were cal-
culated using original correlations [35,36] in the relevant range of 3. Results and discussion
parameters, plotted versus particle Reynolds number (Rep), and fitted
using least squares analysis, resulting in the following correlation: Reactor performance was evaluated via numerical simulations as a
vg ρg dp function of operating parameters, in terms of CO2 conversion (XCO2),
kae = λ g (8 + 0.05 Re1.09
p ); Rep= μg selectivity to CH4 (SCH4), and CH4 yield (YCH4):
(15)
yCH4 + yCO
Wall heat transfer coefficients are determined by resistances in XCO2 =
series: yCH4 + yCO + yCO2 (19)
−1 yCH4
1 d 1 ⎞
Uw, HE = ⎛ ⎜+ w + ⎟ SCH4 =
⎝ h wr λ w h wc ⎠ (16a) yCH4 + yCO (20)

372
D. Sun, D.S.A. Simakov Journal of CO₂ Utilization 21 (2017) 368–382

yCH4 constant parameter. Variations of the coolant temperature within the

YCH4 =
yCH4 + yCO + yCO2 (21) reactor will lead to variations of the heat transfer flux, affecting
therefore the thermal distribution within the reactor. As heat is being
Outlet conversion, selectivity and yield are calculated based on the transferred from the packed bed to the coolant, the molten salt tem-
outlet mole fractions. In all simulations, the feed pressure was 500 kPa perature increases decreasing the inter-compartment temperature dif-
and the feed composition was set to the molar stoichiometric ratio of ference which is the driving force for the heat transfer. Mole fraction
H2/CO2 = 4. The molten salt feed temperature was set to its minimum profiles form a sharp front coinciding with temperature distribution,
operating temperature of 415 K [45]. All parameters are listed in Fig. 5c, d. CO2 and H2 are mainly consumed at the front, producing
Table 2. mainly CH4 and H2O, as well as a small fraction of CO. Downstream the
Variable operating parameters included initial temperature (Tint), reactor, there is additional consumption of H2, as CO is converted via
gas stream feed temperature (Tf), gas hourly space velocity (GHSV), and methanation, Eq. (3). Since the methanation reaction is highly exo-
the normalized cooling rate (Gc/Gc,0). Gas hourly space velocity (vo- thermic, downstream cooling will be crucial for reactor performance, as
lumetric feed flow rate divided by the packed bed volume) is defined as discussed in Section 3.3.4.
follows: Increasing the inlet temperature of the feed gas mixture (Tf) shifts
εvgf the location of the thermal front towards the reactor entrance, Fig. 6b.
GHSV =
L (22) Reactor performance change accordingly (lower panels in Fig. 6).
Slightly higher CH4 yield is obtained for higher feed temperature, but in
The reference molten salt gravimetric flow rate is calculated as-
overall very similar performance is predicted in both cases, Fig. 6.
suming that the heat generation rate (for complete CO2 conversion and
However, the case shown in Fig. 6a is expected to be less beneficial for
no CO formation) is equal to the rate of heat removal by the molten salt
reactor operation. First, almost a third of the reactor stays cold with no
(assuming that ΔTMS = 300 K; the operating range of the molten salt is
reactions occurring there. Second, a thermal front located at some lo-
415–758 K [45]):
cation within the reactor may be unstable and prone to downstream
yCO2 ΔHSR Ftf propagation which can result in reactor extinction, as discussed in
GMS,0 = ; Ftf = ρgf VPB GHSV
Cpc ΔTMS (23) Section 3.3.2.

3.3. Reactor performance

3.1. Model validation: approach to equilibrium
Reactor performance was evaluated as a function of variable para-
For long residence times, i.e. low space velocity, it is expected that
meters listed in Table 2: initial (start-up) temperature (Tint), inlet
the reactor performance would approach equilibrium. Comparison be-
temperature of the feed mixture (the composition is H2/CO2 = 4,
tween the simulation results obtained at GHSV = 100 h−1 and equili-
Table 2) (Tf), gas hourly space velocity (GHSV), and the molten salt
brium data [46] is shown in Fig. 3, in terms of CO2 conversion and
gravimetric flow rate (GMS) normalized to the reference rate (GMS/
selectivity to CH4 formation (parameters are listed in the figure cap-
GMS,0). Initial and feed temperatures are expected to affect the reactor
tion). Good agreement with the equilibrium data was obtained. CO2
performance significantly because of the high exothermicity of the Sa-
conversion is favored by high pressures and low temperatures; CH4
batier reaction system, Eqs. (1)–(3). Feed flow rates in the reaction and
selectivity drops rapidly for T > 850 K.
coolant compartments (GHSV and GMS) will be crucial parameters,
Because of a combined effect of the decrease of conversion and
defining the rates of heat generation and removal.
selectivity with increasing temperatures, CH4 yield drops rapidly for
T > 800 K, Fig. 4. Therefore, thermal management of the Sabatier
process is crucial. Reactor temperature should be kept sufficiently high 3.3.1. Start-up temperature
to accelerate catalytic reactions, but below 900 K, when considerably Though the Sabatier reaction is highly exothermic, preheating will
high CH4 yield is achievable at slightly elevated pressures, Fig. 4. be required in a practical situation in order to ignite the reactor (an
Moreover, for the reactor configuration studied here, molten salt op- example of the simulated reactor ignition is shown in Fig. 5a). The ef-
erating range becomes a crucial parameter. The coolant feed tempera- fect of the initial (start-up) temperature is shown in Fig. 7: outlet
ture should be obviously kept above the molten salt melting point but temperatures, conversion, selectivity, and yield are plotted versus in-
also well below the temperature of its thermal decomposition. The re- itial temperature. It is assumed that the entire reactor is heated to the
commended operating range for the type of molten salt used in the same temperature, which can be achievable by flowing the molten salt
simulations is 415–758 K, which is a typical operating range for molten through the cooling tubes before feeding the gas mixture. There is a
salts (some molten salts can be operated at slightly higher temperature sharp threshold temperature of 545 K. Above this temperature, reaction
but have higher freezing point, e.g., 500–838 K) [45]. The upper range ignition takes place, while further increase in the preheating tempera-
of this temperature regime is also favorable for CH4 formation, Fig. 4. ture has virtually no effect on the reactor performance.

3.2. Spatial profiles within the reactor 3.3.2. Effect of feed temperature
The gas feed temperature (Tf) is a crucial operating parameter af-
Typical spatial reactor profiles are presented in Fig. 5, showing the fecting the temperature distribution within the reactor, Fig. 6. It is
reactor ignition (Fig. 5a) and stationary spatial profiles of temperatures expected that below a certain threshold feed temperature reactions will
and mole fractions. For the set of parameters listed in Fig. 5, ignition
Table 2
occurs approximately at the first third of the reactor length, with sub-
Operating conditions.
sequent formation of a stationary thermal front after approximately
5 min of simulated time, Fig. 5a. The simulation predicts that steady Pf (kPa) H2/CO2 Tcf (K) Tint (K) Tf (K) GHSV (h−1) GMS/GMS,0
state temperatures are identical in the packed bed and molten salt
compartments, except for the small area at the temperature front, 500 4/1 415 500–600 450–650 100–50,000 0.1–1

Fig. 5b.
Pf denotes the feed pressure and H2/CO2 is the molar feed ratio of H2 to CO2. Tcf stands for
Importantly, the molten salt temperature varies significantly as a the coolant (molten salt) feed temperature. Tint and Tf are initial (start-up) and gas feed
function of the reactor length. This observation emphasizes the im- temperatures, respectively. GHSV is the gas hourly space velocity, Eq. (22). GMS/GMS,0
portance of accounting for the coolant temperature as a variable, not a stands for the normalized gravimetric flow rate of the molten salt, Eq. (23).

373
D. Sun, D.S.A. Simakov Journal of CO₂ Utilization 21 (2017) 368–382

Fig. 3. Approach to equilibrium: CO2 conversion (XCO2) and selectivity to CH4 (SCH4) obtained in numerical simulations with low space velocity (symbols) are compared to equilibrium
values (lines). Parameters in numerical simulations: Tint = 550 K, Tf = 600 K, GHSV = 100 h−1, GMS = GMS,0.

Ignition-extinction curves obtained by the gradual increase or de-

crease of the feed temperature are shown in Fig. 9, in terms of the
maximum and outlet packed bed temperature as a function of the feed
temperature. For low feed temperatures, which are below the ignition
threshold, both the maximum and outlet temperatures increase linearly,
until the ignition threshold of Tf > 500 K is reached. A small increase
in the feed temperature beyond this threshold leads to reactor ignition,
Fig. 9. Further increase in the feed temperature results in a moderate
decrease of the maximum and outlet temperature, which can be at-
tributed to the complex coupling of the reaction heat generation, inter-
compartment heat transfer, and heat removal by the molten salt. When
the feed temperature is gradually decreased following the top branch of
the ignition-extinction curve, extinction occurs at T < 480 K.
Since the extinction temperature is different from the ignition
threshold, there is a small region of bi-stability, where both the ignited
and extinguished states co-exists. For practical applications, it is fa-
vorable to operate the reactor well outside the region of bi-stability
where sudden reactor extinctions can occur as a result of fluctuations in
Fig. 4. Approach to equilibrium: CH4 yield (YCH4) obtained in numerical simulations with the feed temperature. Fluctuations in operating parameters are ex-
low space velocity (symbols) as compared to equilibrium values (lines). Parameters in pected to occur in real situations and other operating parameters can
numerical simulations: Tint = 550 K, Tf = 600 K, GHSV = 100 h−1, GMS = GMS,0. affect the extinction-ignition behavior as well, such as space velocity
[27,28].
not take place, similarly to the effect of preheating, Fig. 7. On the other
hand, exceedingly high feed temperatures can result in reactor over-
heating, negatively affecting product distribution of the highly exo- 3.3.3. Effect of space velocity
thermic Sabatier reaction system, Eqs (1–3). In a practical situation, it is Increasing space velocity is beneficial from the point of view of
desirable to keep the feed temperature as low as possible to save en- maximizing reactor throughput, even though it can negatively impact
ergy. Another important consideration is feed temperature fluctuations longevity. Nevertheless, it is important to understand the effect of in-
which are unavoidable in real systems. creasing throughput on the reactor performance. In the analyzed re-
The effect of a step-wise change of the feed temperature is de- actor configuration (Fig. 2), the coupling of the reaction heat genera-
monstrated in Fig. 8. For parameters listed in Fig. 8 and a feed tem- tion with the heat transfer between the compartments and heat removal
perature of Tf = 510 K, the simulation predicts formation of a sta- by the molten salt can result in nontrivial behavior.
tionary thermal front in the second half of the packed bed, left panels in The effect of increasing space velocity on reactor profiles is de-
Fig. 8 for t < 1 h. A step-wise increase of the feed temperature to monstrated in Fig. 10, where elevated cooling rate was used
Tf = 550 K immediately induces upstream front propagation with (GMS = 0.5GMS,0). At low space velocity, Fig. 10a, the inter-compart-
subsequent stabilization of the thermal front at the reactor entrance, ment heat transfer is very efficient, resulting in equal temperatures in
Fig. 8a. If the feed temperature is abruptly decreased to Tf = 450 K, the the packed bed and molten salt compartment, except for the small
front propagates downstream, eventually leading to the reactor ex- temperature difference at the reactor entrance, Fig. 10 (upper panel).
tinction, Fig. 8b. Such moving thermal fronts are known to form in For the parameters used in Fig. 10a, the simulation predicted highly
packed bed reactors with catalytic exothermic reactions; the propaga- efficient cooling. Both the gas stream and molten salt leave the reactor
tion is due to the combine effect of heat generation, axial heat disper- at the temperature which is only slightly higher than the molten salt
sion, and convection [27,28]. Understanding this phenomenon is im- feed temperature. Note, that heat losses to the environment was ac-
portant from the practical point of view, in order to avoid operating counted for in the current study, last tram in Eq. (5).
under conditions that can lead to the formation of downstream pro- Increasing the space velocity ten-folds leads to a very different
pagating thermal fronts that result in reactor extinction. distribution within the reactor, Fig. 10b. Though for the most of the
reactor, the temperatures are equal in both compartments, a very

374
D. Sun, D.S.A. Simakov Journal of CO₂ Utilization 21 (2017) 368–382

Fig. 5. The spatiotemporal profile of reactor ignition (a) and steady state spatial profiles of temperature (b), mole fractions (c), and mole fractions calculated on dry basis (d). TPB and TMS
in upper panels denote temperatures in the packed bed (PB) and molten salt (MS) compartments, respectively; yi denotes the mole fraction of a component i. Parameters in numerical
simulations: Nc = 22, Tint = 550 K, Tf = 450 K, GHSV = 1000 h−1, GMS = 0.1GMS,0.

significant temperature gradient exists in the first quarter of the reactor, Nc = 13, as reducing number of tubes will simplify reactor manu-
upper panel in Fig. 10b. The cooling is much less efficient, though same facturing and operation. Elevated rate of cooling was used,
ratio of GMS = 0.5GMS,0 is kept; the outlet temperature is ∼600 K. The GMS = 0.4GMS,0, as previous simulations showed insufficient heat re-
exceedingly high packed bed temperature at the reactor entrance af- moval for GMS =0.1GMS,0 (compare Fig. 6 and Fig. 10). The effect of the
fects the mole fraction distribution within the reactor. As a result of less space velocity on the outlet CO2 conversion, selectivity to CH4, CH4
efficient cooling, the outlet mole fraction of CH4 is lower, lower panel in yield is demonstrated in the left panel of Fig. 12; showing also the
Fig. 10b. For high space velocities, the heat removal could be limited by normalized pressure drop, ΔPN = (Pf − Pout)/Pf. Corresponding outlet
the heat transfer between the compartments, which is represented in temperatures are shown on the right panel of Fig. 12.
the model by the heat exchange terms, in Eqs. (5) and (6). For relatively low space velocities (GHSV < 5000 h−1), outlet CO2
Wall heat transfer coefficients (Uw,HL, Uw,HE) are calculated from conversions and CH4 yields higher than 90% are predicted, while full
correlations, Eqs. (16)–(18), and, therefore, cannot be arbitrary ma- selectivity to the formation of CH4 is predicted at the reactor outlet.
nipulated. The reactor geometry is fixed in this study except for the These results indicate that the cooling is sufficient to prevent the for-
number of cooling tubes (Nc) which can be varied within the restriction mation of CO. Indeed, as it can be seen from Fig. 12 (right panel), the
of the reactor design, Table 1. The effect of the number of coolant tubes outlet backed bed temperatures are below 700 K for GHSV < 5000
on the reactor performance at high space velocity h−1, for which the equilibrium predicts virtually no CO formation. Even
(GHSV = 10,000 h−1) is demonstrated in Fig. 11. For Nc = 22, both though some CO can still be formed in the hottest part of the reactor, it
streams (the reaction mixture and the coolant) exit the reactor at same is converted into CO2 downstream.
temperature. Decreasing the number of cooling tubes leads to slightly As space velocity is further increased, the conversion and yield drop
lower outlet temperatures, until a significant difference between the below 90% and the pressure drop starts to be significant, left panel of
outlet temperatures appear for Nc < 12. Though CO2 conversion and Fig. 12. The packed bed outlet temperature gradually increase, while
CH4 yield are not affected to a significant extent by the inefficient heat the outlet molten salt temperature decreases. For GHSV > 7000 h−1,
transfer, right panel in Fig. 11, the heat removal appears to be limited there is a substantial difference between outlet temperatures, indicating
by the insufficient heat exchange area for Nc < 12. Note that the si- insufficient heat removal due to the transport limitations, i.e., the rate
mulations predict CH4 yields over 90% at the elevated space velocity of of heat transfer through the walls of the cooling tubes is significantly
GHSV = 10,000 h−1, Fig. 11. lower than the rate of reaction heat generation. Importantly, the 80%
As it is highly desirable to maximize the reactor throughput to in- CH4 yield is still achieved at a high space velocity of
crease the rate of CH4 production and thus to reduce the capital cost GHSV = 15,000 h−1, while the outlet molten salt temperature is well
investment, reactor performance should be investigated over the range above the solidification point and below the thermal decomposition
of space velocities. In Fig. 12, the reactor performance is evaluated as a threshold; the operating range of the molten salt is 415–758 K [45].
function of space velocity ranging from 1000 1/h to 25,000 1/h using Above the threshold of GHSV > 15,000 h−1, the reactor performance

375
D. Sun, D.S.A. Simakov Journal of CO₂ Utilization 21 (2017) 368–382

Fig. 6. Steady state spatial profiles of temperature (upper panels) and calculated reactor performance (lower panels) for gas stream feed temperatures of Tf = 450 K (a) and Tf = 600 K
(b). TPB and TMS in upper panels) denote temperatures in the packed bed (PB) and molten salt (MS) compartments, respectively; XCO2, SCH4, and YCH4 denote conversion, selectivity, and
yield, Eqs. (19–21). Parameters in numerical simulations: Nc = 22, Tint = 550 K, GHSV = 1000 h−1, GMS = 0.1GMS,0.

drops sharply, which can be attributed to overcooling. Note that the gas reactor throughput, while keeping high CH4 yields. At the same time,
feed, which enters the reactor at a relatively low temperature of the molten salt flow rate should be minimized in order to reduce energy
Tf = 600 K, also contributes to the cooling, especially at high space investment in pumping and storage. However, decreasing the rate of
velocities. Pressure drop also increases considerably for GHSV > cooling will lead to the increase of the reactor temperature reducing
15,000 h−1, which can also contribute to the performance drop, as low CO2 conversion and, therefore, CH4 production. Therefore, optimiza-
pressures are thermodynamically unfavorable for the Sabatier reaction, tion of the molten salt flow rate, i.e., cooling rate, will be crucial for the
Figs. 3 and 4. Note that in all simulations the inlet pressure was fixed to reactor performance improvement and, therefore, for system optimi-
Pf = 500 kPa, Table 2. zation.
Effect of the cooling rate on the reactor performance is first ex-
amined for a fixed space velocity, Fig. 13. For low cooling rate of
3.3.4. Optimization of cooling rate GMS = 0.1GMS,0, Fig. 13a, the heat removal is insufficient and CO2 is
For process optimization, it is highly desirable to maximize the

Fig. 7. Effect of the initial reactor temperature (pre-heating, start-up temperature) on the steady state reactor performance in terms of the outlet temperature (left panel) and outlet
conversion, selectivity, and yield (right panel). Parameters: Nc = 13, Tf = 520 K, GHSV = 1000 h−1, GMS = 0.38GMS,0.

376
D. Sun, D.S.A. Simakov Journal of CO₂ Utilization 21 (2017) 368–382

Fig. 8. Moving thermal fronts propagating upstream (a), as a result of increasing feed temperature from Tf = 510 K to Tf = 550 K, and downstream (b), as a result of switching feed
temperature from Tf = 510 K to Tf = 450 K. Right panels show 2D representation of the spatio-temporal 3D patterns shown in left panels. Parameters: Nc = 13, Tint = 550 K,
GHSV = 1000 h−1, GMS = 0.38GMS,0.

Fig. 9. Ignition-extinction curves showing the packed bed maximum temperature (a) and the packed bed outlet temperature (b) as a function of the gas feed temperature. All results
shown were obtained after the steady state was achieved. Parameters: Nc = 13, Tint = 550 K, GHSV = 1000 h, GMS = .

only consumed at the reactor entrance. There is further consumption of consumption downstream the reactor, resulting in a dramatic im-
H2 downstream the reactor, but only due to the CO conversion, lower provement of CH4 production, lower panel in Fig. 13b. Once can con-
panel of Fig. 13a. Importantly, the molten salt temperature is above clude that a declining temperature profile is highly favorable for CH4
800 K for the most of the reactor, which is outside the operating range production via Sabatier reaction. From the thermodynamic point of
and thus will lead to thermal decomposition of the molten salt. In- view, relatively high temperatures in the first part of the packed bed
creasing the cooling rate ten-fold, to GMS = GMS,0, changes reactor favor CO2 conversion into CH4 but also to CO, Fig. 3. As temperature
profiles dramatically, Fig. 13b. The molten salt temperature is within its declines downstream the reactor, the CO formed is converted into more
operating range, upper panel in Fig. 13b and the packed bed tem- CH4, as it is predicted by the equilibrium, Fig. 3.
perature is cooled down along the reactor. As a result, in addition to the The reactor performance for low and high cooling rates is analyzed
CO2 consumption at the reactor entrance, there is additional CO2 in Fig. 14 over a wide range of space velocities, in terms of outlet

377
D. Sun, D.S.A. Simakov Journal of CO₂ Utilization 21 (2017) 368–382

Fig. 10. Stationary temperature profiles (upper panels) and mole fraction distributions (lower panels) for low space velocity, GHSV = 100 h−1 (a) and for elevated space velocity,
GHSV = 1000 h−1 (b). Parameters: Nc = 22, Tint = 550 K, Tf = 600 K, GMS = 0.5GMS,0.

conversion, selectivity, and yield (also showing pressure drop). From (Fig. 14, left panel). Since CH4 production is favored by high pressures
analyzing Fig. 14 in comparison with Fig. 12 (intermediate cooling (Figs. 3, 4), it is desirable to keep the entire reactor at high pressure.
rate), one can conclude that the sharp drop in the reactor performance Operation at low cooling rate is also disadvantageous from the point of
can be rather attributed to overcooling by the molten salt. As the view of conversion and selectivity. High conversion and selectivity are
cooling rate is increased from low (GMS = 0.08GMS,0, left panel in only attainable for low space velocity. As the space velocity is in-
Fig. 14) to intermediate (GMS = 0.38GMS,0, left panel in Fig. 12), and to creased, both CO2 conversion and selectivity to CH4 drop significantly,
high (GMS = 0.77GMS,0, right panel in Fig. 14), the performance drop resulting in CH4 yields below 60%. This can be attributed to insufficient
threshold decreases from GHSV = 36,000 h−1, to cooling that results in exceedingly high temperatures that suppress CH4
GHSV = 15,000 h , and GHSV = 8000 h−1.
−1
production. Overcooling, i.e. high cooling rate (Fig. 14, right panel) is
Pressure drop can also contribute to the drop in performance, par- also disadvantages. Though high CH4 yield is obtained, operation is
ticularly for the case of the low cooling rate: the pressure drop becomes limited to GHSV < 10,000 h−1.
very significant at the threshold space velocity of GHSV = 36,000 h−1 Since the reactor performance is strongly affected by the rate of heat

Fig. 11. Effect of the number of cooling tubes on the steady state reactor performance at high space velocity. Outlet temperatures of the packed bed (PB) and coolant, molten salt (MS)
compartments (a) and outlet conversion, selectivity, and yield (b) are plotted versus the number of cooling tubes (Nc) in the reactor. Parameters: Tint = 550 K, Tf = 650 K,
GHSV = 10,000 h−1, GMS = GMS,0.

378
D. Sun, D.S.A. Simakov Journal of CO₂ Utilization 21 (2017) 368–382

Fig. 12. Effect of space velocity on the steady state reactor performance. Outlet conversion, selectivity, yield, and normalized pressure drop (left panel) and outlet temperatures in the
packed bed and molten salt compartments (right panel) are plotted versus gas space velocity. Normalized pressure drop is defined as ΔPN = (Pf − Pout)/Pf. Parameters:
Nc = 13, Tint = 550 K, Tf = 600 K, GMS = 0.38GMS,0.

removal, cooling rate optimization is of crucial importance. The effect pronounced for the higher space velocity. Also, the sharp drop in the
of the cooling rate at high (fixed) space velocities is examined in reactor performance occurs at a lower GMS/GMS,0 ratio for
Fig. 15. Outlet conversion, selectivity, and yield are plotted versus the GHSV = 45,000 h−1.
molten salt flow rate for GHSV = 20,000 h−1 (Fig. 15a) and To summarize the findings discussed in two last sub-sections, it can
GHSV = 45,000 h−1 (Fig. 15b). The increase in the molten salt gravi- be concluded that both space velocity and cooling rate (molten slat flow
metric flow rate initially leads to a considerable improvement in CH4 rate) are crucial parameters affecting the reactor performance. The
production, which is followed by a sharp drop after a certain threshold, (desirable) increase in the space velocity unavoidably leads to the de-
similarly to the effect of space velocity, Figs. 12, 14. While for clining CH4 yield and, eventually, a threshold-like drop in performance
GHSV = 20,000 h−1 the selectivity to CH4 remains almost complete leading to conversions and yields which cannot be considered practical
over the entire range, GHSV = 45,000 h−1 increasing the molten salt (below 20%). Moderately elevated (GHSV < 10,000 h−1) space velo-
flow rate results in a very significant improvement, right panel in cities with intermediate cooling rates (GMS ≈ 0.5GMS,0) are re-
Fig. 15. In both cases, the optimum in the CH4 yield is observed, more commended for the reactor configuration analyzed here.

Fig. 13. Effect of the cooling rate on the reactor performance. Stationary temperature profiles (upper panels) and mole fraction distributions (lower panels) are shown for the relatively
low cooling rate, GMS = 0.1GMS,0 (a), and for the elevated cooling rate, GMS = GMS,0 (b). Parameters: Nc = 22, Tint = 550 K, Tf = 600 K, GHSV = 2000 h−1.

379
D. Sun, D.S.A. Simakov Journal of CO₂ Utilization 21 (2017) 368–382

Fig. 14. Effect of the heat removal rate on the steady state reactor performance. Outlet conversion, selectivity, yield, and normalized pressure drop are plotted versus space velocity for
relatively low cooling rate, GMS = 0.08GMS,0 (a) and elevated cooling rate, GMS = 0.77GMS,0 (b). Parameters: Nc = 13, Tint = 550 K, Tf = 600 K.

Fig. 15. Effect of the cooling rate on the steady state reactor performance at high space velocities. Outlet conversion, selectivity, and yield are plotted versus the molten salt flow rate for
GHSV = 20,000 h−1 (a) and GHSV = 45,000 h−1 (b). Ranges of the molten salt flow rate correspond to GMS = (0.2-2.3)GMS,0 (a) and GMS = (0.05-0.4)GMS,0 (b). Parameters:
Nc = 13, Tint = 550 K, Tf = 650 K.

3.4. Preliminary techno-economic evaluation amounts of CO2 into RNG which can be injected into the existing nat-
ural gas infrastructure. Suitable feedstocks include combustion flue
For practical applications, the Sabatier reactor analyzed in this gases, as well as biogas and landfill gas. Biogas and landfill gas already
study should be eventually incorporated into the entire renewable contain a significant fraction of non-fossil CH4 (up to 60%) and can be,
natural gas (RNG) production system represented in Fig. 1. Such a therefore, directly used for power generation via combustion. However,
system can be used for converting waste streams containing significant this approach has low efficiency due to the thermodynamic limitations

Fig. 16. Techno-economic evaluation in terms of the reactor power density, Eq. (24), and the renewable natural gas production cost, Eq. (25), as a function of the reactor throughput. The
production cost (right panel) is shown for a single reactor, three reactors, and 10 reactors, for the cumulative 10,000 h of operation. Parameters: Nc = 13, Tint = 550 K, Tf = 600 K,
GMS = 0.38GMS,0.

380
D. Sun, D.S.A. Simakov Journal of CO₂ Utilization 21 (2017) 368–382

of combustion engines. Alternatively, these feedstocks can be upgraded though the produced renewable natural gas (RNG) will be eventually
into RNG via pre-separation of CH4 (can be directly used as RNG) and burned to generate electricity and heat, the use of RNG reduces the
converting the remaining CO2 into more renewable methane. consumption of the fossil natural gas, reducing therefore CO2 emissions.
This approach has a potential to provide enormous storage and This is in particular true if the source of CO2 is not fossil, such as fer-
transportation capacities for the re-distribution of renewable energy. mentation flue gas or biogas.
Renewable sources such as solar and wind energy are abundant but As it can be seen from Fig. 16, increasing the number of reactors
their efficient utilization is challenging due to its transient nature re- reduces the production cost substantially. For the relatively small re-
sulting in the supply-demand mismatch. Fluctuations in power supply actor size analyzed herein, no additional molten salt pump or steam
can be compensated by storing off-peak power in batteries but this turbine is required up to ten reactors, and the cost of reactors is low. As
approach has limitations of energy density, cost, and scalability. An a result, when the number of reactors is increased, the capital cost re-
alternative route is to store the surplus electricity in the form of che- mains almost same, reducing the production cost. For a small number of
mical energy of the RNG. However, in order to assess the economic reactors, it is beneficial to operate at high space velocities. For ten re-
viability of this approach, the production cost of the resulted RNG actors, similar price is obtained over the entire range, though there is a
should be estimated. slight optimum for moderately high space velocity.
Herein, we provide only a preliminary techno-economic evaluation Importantly, the production cost drops down below 20 $/GJ,
by integrating the reactor analyzed in this study into the entire system, starting to approach the price of H2: 2 $/kg is equivalent to 16.6 $/GJ
Fig. 1. One of the important considerations is the reactor compactness, for H2, using the low heating value. From the thermodynamic point of
which can be evaluated by calculating the power density as a function view, the maximum of 83% of the H2 energy content can be recovered
of the reactor throughput (LHV stands for low heating value): via the Sabatier reaction (low heating values of CH4 and H2 are 802.3
KJ/mol and 241.8 kJ/mol, respectively); the rest is dissipated as heat,
FCH4, out LHVCH4
P= Eq. (1). In this case, the CH4 production cost will be ∼20 $/GJ. Because
Vr (24)
some heat is recovered from the molten salt stream for electricity
The calculated power densities shown in Fig. 16 (left panel) corre- generation, Fig. 1, Eq. (25), it is possible to reduce the production cost
spond to the results presented in Fig. 12. The power density increases even below 20 $/GJ even for incomplete CO2 conversion and CH4 yield.
nearly linearly for increasing reactor throughputs, as the rate of CH4 Note that the relatively low calculated RNG production cost is
production increases, for the fixed reactor size. Very high power density mainly a derivative of the assumed low H2 production cost of 2 $/kg.
is achieved at the optimum, before the reactor performance declines However, this cost coincides with the cost goal set by the US
sharply, resulting in a drop of CH4 production. Department of Energy (2–3 $/kg) [49] and might be provided in the
To evaluate CH4 production cost, we consider the most essential future by novel technologies such as high temperature water splitting
elements of the synthetic methane generation system shown in Fig. 1: a [50]. Another source of the estimated low RNG cost is the use of the
molten salt pump, a steam turbine, and the reactor itself. The H2 pro- compact, high power density reactor units placed in parallel. This
duction infrastructure cost is lumped in the price of H2. It is assumed configuration is different from the more conventional multiple reactor
that a pure stream of CO2 is available. The production cost was calcu- configuration with inter-stage cooling that would require much higher
lated using the following equation: capital investment and operating costs, contributing to significantly
higher production costs [51]. Some post-purification might be also re-
production cost = capital investement + pumping cost × t+ quired for the outlet stream depending on application, increasing the
+ [FH2, f × H2 price − ηST GMS CP, MS (TMS, out − TMS, f ) × electricity price] RNG production cost. Nevertheless, the lower limit of 20 $/GJ esti-
mated in our study coincides with some lower limit estimations re-
×t
ported in the public domain [52–54] and some recent studies on the
(25)
techno-economic evaluation and life-cycle analysis of CO2 methanation
The capital investment includes the cost of the reactor, molten salt systems (25–40 $/GJ) [55]. It should be mention that other studies
pump, and steam turbine. The reactor cost will not be a significant provide much higher estimates (30–120 $/GJ) [26].
contribution, since the price of Ni-based catalysts considered in our
study is relatively low and the reactor construction material is also low- 4. Concluding remarks
cost (stainless steel). A typical price of Ni-based catalysts is 50 $/kg
(e.g., Johnson Matthey). The packed bed volume is ∼ 25 L, which will The molten salt-cooled heat exchanger type packed bed reactor was
result in approximately $3000 cost of a single reactor (including $1500 suggested as a novel design configuration for synthetic methane pro-
catalyst cost and manufacturing). The molten salt pump and steam duction via Sabatier reaction. A mathematical model was defined and
turbine costs were assumed to be $40,000 and $58,000, respectively; analyzed using numerical simulations in order to determine the effects
these are typical prices for the small-to-medium scale equipment (e.g., of operating parameters and to find optimal operation regimes. The
[47,48]). The electricity consumption for pumping is considered in Eq. reactor performance was evaluated in terms of CO2 conversion and CH4
(25) as well, but this contribution was found to be rather insignificant yield, as a function of various parameters including feed temperature,
to another time-dependent cost which is H2 consumption. A (rather space velocity and cooling rate, which were found to be of crucial
optimistic) projection of 2 $/kg was used as a price of renewable H2. importance for the reactor performance. The simulations predicted that
The cost of H2 consumed can be partially recovered by electricity under certain optimized operating conditions it is possible to obtain
generation by a steam turbine, Fig. 1, using the reactor outlet molten CH4 yields of more than 90% at elevated, industrially relevant space
salt to generate steam. The overall efficiency of the electricity genera- velocities.
tion from the molten salt heat (ηST) was assumed to be 30% and the Optimizing the rate of cooling, i.e. the molten salt flow rate, is a
electricity celling price was fixed to 0.13 $/kWh. Labor, maintenance, challenging task. On one hand, high cooling rates produce a declining
and equipment replacement/depreciation costs were not included in temperature profile in the packed bed which is beneficial from the
our preliminary assessment. thermodynamic point of few: low temperatures facilitate methane for-
The production cost calculated using Eq. (25) (right panel in Fig. 16) mation in the exothermic Sabatier reaction. On the other hand, over-
clearly cannot compete with the current low prices of natural gas (∼2.5 cooling can lead to a sharp performance drop and even to reactor ex-
$/GJ), although historically the price of natural gas fluctuated well tinction. Low cooling rates should be avoided because they result in
above 10 $/GJ (in 2008). Also, the process analyzed in this study exceedingly high temperatures that negatively affect CH4 formation and
consumes CO2, which can be potentially used for carbon trading. Even can lead, in a practical situation, to the thermal decomposition of the

381
D. Sun, D.S.A. Simakov Journal of CO₂ Utilization 21 (2017) 368–382

molten salt. Increasing space velocity is beneficial from the point of and iIonic liquids for three-phase methanation and biogas purification processes,
Energy Fuel 27 (2013) 4705–4716.
view of increased reactor throughputs and, therefore, higher rate of CH4 [20] J. Lefebvre, M. Götz, S. Bajohr, R. Reimert, T. Kolb, Improvement of three-phase
production. However, high space velocities can lead to reactor extinc- methanation reactor performance for steady-state and transient operation, Fuel
Process. Technol. 132 (2015) 83–90.
tion in a threshold-like manner similar to the effect of the overcooling [21] S. Rönsch, J. Schneider, S. Matthischke, M. Schlüter, M. Götz, J. Lefebvre,
by molten salt. P. Prabhakaran, S. Bajohr, Review on methanation −From fundamentals to current
A preliminary techno-economic evaluation carried out at the system projects, Fuel 166 (2016) 276–296.
[22] S.A.M. Said, D.S.A. Simakov, M. Waseeuddin, Y. Roman-Leshkov, Solar molten salt
level predicted the production cost of the renewable natural gas (RNG) heated membrane reformer for natural gas upgrading and hydrogen generation: a
below 20 $/GJ, for multiple reactors and the optimized cooling rate and CFD model, Sol. Energy 124 (2016) 163–176.
[23] S.A.M. Said, M. Waseeuddin, D.S.A. Simakov, A review on solar reforming systems,
reactor throughput. The most significant contribution to the RNG pro-
Renewable Sustainable Energy Rev. 59 (2016) 149–159.
duction cost is expected to be the price of renewable H2. Therefore, [24] D.S.A. Simakov, M.M. Wright, S. Ahmed, E.M.A. Mokheimer, Y. Román-Leshkov,
reducing the production cost of renewable H2 can lead to a widespread Solar thermal catalytic reforming of natural gas: a review on chemistry, catalysis
and system design, Catal. Sci. Technol. 5 (2015) 1991–2016.
introduction of RNG into the markets. Importantly, most of the energy [25] G. Gahleitner, Hydrogen from renewable electricity: an international review of
content of H2 consumed in the reaction is recovered in the RNG pro- power-to-gas pilot plants for stationary applications, Int. J. Hydrogen Energy 38
duced and electricity co-generated (up to 80%). Since renewable H2 (2013) 2039–2061.
[26] M. Götz, J. Lefebvre, F. Mörs, A. McDaniel Koch, F. Graf, S. Bajohr, R. Reimert,
produced in a distributed manner is less suitable for storage and T. Kolb, Renewable power-to-gas: a technological and economic review, Renew.
transportation, converting this renewable H2 into RNG could be an Energy 85 (2016) 1371–1390.
[27] D.S.A. Simakov, M. Sheintuch, Design of a thermally balanced membrane reformer
attractive avenue for storage and re-distribution of the transient re- for hydrogen production, AIChE J. 54 (2008) 2735–2750.
newable sources such as solar and wind energy. [28] D.S.A. Simakov, M. Sheintuch, Model-based optimization of hydrogen generation
by methane steam reforming in autothermal packed-bed membrane reformer,
AIChE J. 57 (2011) 525–541.
Acknowledgement [29] R. Güttel, Study of unsteady-state operation of methanation by modeling and si-
mulation, Chem. Eng. Technol. 36 (2013) 1675–1682.
The authors acknowledge the support of the Waterloo Institute of [30] J. Xu, G.F. Froment, Methane steam reforming, methanation and water-gas shift: I.
Intrinsic kinetics, AIChE J. 35 (1989) 88–96.
Sustainable Energy (WISE) through the WISE – Cisco Systems Smart [31] S.S.E.H. Elnashaie, A.M. Adris, A.S. Al-Ubaid, M.A. Soliman, On the non-monotonic
Grid Research Fund. behaviour of methane steam reforming kinetics, Chem. Eng. Sci. 45 (1990)
491–501.
[32] M.L. Rodríguez, M.N. Pedernera, D.O. Borio, Two dimensional modeling of a
References membrane reactor for ATR of methane, Catal. Today 193 (2012) 137–144.
[33] G.F. Froment, K.B. Bischoff, Chemical Reactor Analysis and Design, Wiley, New
York, 1979.
[1] B. Kumar, M. Llorente, J. Froehlich, T. Dang, A. Sathrum, C.P. Kubiak, [34] R.J. Berger, J. Pérez-Ramı́rez, F. Kapteijn, A. Moulijn, Catalyst performance testing:
Photochemical and photoelectrochemical reduction of CO2, Annu. Rev. Phys. Chem. radial and axial dispersion related to dilution in fixed-bed laboratory reactors, Appl.
63 (2012) 541–569. Catal. A Gen. 227 (2002) 321–333.
[2] Y. Chen, N.S. Lewis, C. Xiang, Operational constraints and strategies for systems to [35] A.G. Dixon, D.L. Cresswell, Theoretical prediction of effective heat transfer para-
effect the sustainable, solar-driven reduction of atmospheric CO2, Energy Environ. meters in packed beds, AIChE J. 25 (1979) 663–676.
Sci. 8 (2015) 3663–3674. [36] O.R. Derkx, A.G. Dixon, Effect of the wall Nusselt number on the simulation of
[3] W. Wang, S. Wang, X. Ma, J. Gong, Recent advances in catalytic hydrogenation of catalytic fixed bed reactors, Catal. Today 35 (1997) 435–442.
carbon dioxide, Chem. Soc. Rev. 40 (2011) 3703–3727. [37] O.A. Sergeev, A.G. Shashkov, A.S. Umanskii, Thermophysical properties of quartz
[4] W. Wang, J. Gong, Methanation of carbon dioxide: an overview, Front. Chem. Sci. glass, J. Eng. Phys. Thermophys. 43 (1982) 1375–1383.
Eng. 5 (2011) 2–10. [38] S.W. Churchill, H.H. Chu, Correlating equations for laminar and turbulent free
[5] T. Schaaf, J. Grünig, M.R. Schuster, T. Rothenfluh, A. Orth, Methanation of CO2 – convection from a horizontal cylinder, Int. J. Heat Mass Transfer 18 (1975)
storage of renewable energy in a gas distribution system, Energy Sustain Soc. 4 1049–1053.
(2014). [39] A.F. Mills, Heat Transfer, 2nd ed., Prentice-Hall, New Jersey, 1999.
[6] J. Kopyscinski, T.J. Schildhauer, S.M.A. Biollaz, Production of synthetic natural gas [40] V. Gnielinski, New equations for heat and mass transfer in turbulent pipe and
(SNG) from coal and dry biomass −a technology review from 1950 to 2009, Fuel 89 channel flow, Int. Chem. Eng. 16 (1976) 359–368.
(2010) 1763–1783. [41] J.P. Holman, Heat Transfer, 9th ed., McGraw-Hill, Inc., New York, Boston, 2002.
[7] Y.-L. Kao, P.-H. Lee, Y.-T. Tseng, I.L. Chien, J.D. Ward, Design, control and com- [42] Y.S. Touloukian, R.K. Kirby, E.R. Taylor, T.Y.R. Lee, Thermophysical Properties of
parison of fixed-bed methanation reactor systems for the production of substitute Matter, The TPRC Data Series, Purdue University, 1977.
natural gas, J. Taiwan Inst. Chem. Eng. 45 (2014) 2346–2357. [43] R. Morrell, Handbook of Properties of Technical and Engineering Ceramics,
[8] I. Kiendl, M. Klemm, A. Clemens, A. Herrman, Dilute gas methanation of synthesis Stationery Office Books, London, 1987.
gas from biomass gasification, Fuel 123 (2014) 211–217. [44] G.J. Janz, Molten Salts Handbook, Elsevier, New York, 1967.
[9] D. Schlereth, O. Hinrichsen, A fixed-bed reactor modeling study on the methanation [45] Dynalene Molten Salts, Dynalene, Inc., 2016, https://www.dynalene.com/Molten-
of CO2, Chem. Eng. Res. Des. 92 (2014) 702–712. Salts-s/1831.htm.
[10] L. Kiewidt, J. Thöming, Predicting optimal temperature profiles in single-stage [46] J. Gao, Y. Wang, Y. Ping, D. Hu, G. Xu, F. Gu, F. Su, A thermodynamic analysis of
fixed-bed reactors for CO2-methanation, Chem. Eng. Sci. 132 (2015) 59–71. methanation reactions of carbon oxides for the production of synthetic natural gas,
[11] B. Hou, Y. Huang, X. Wang, X. Yang, H. Duan, T. Zhang, Optimization and simu- RSC Adv. 2 (2012) 2358.
lation of the Sabatier reaction process in a packed bed, AIChE J. 62 (2016) [47] Wabash Power Equipment Company. 510 KW Dresser-Rand Turbodyne, (2017)
2879–2892. https://www.wabashpower.com/inventory/turbine-generators/steam-turbines/
[12] M.C. Seemann, T.J. Schildhauer, S.M. Biollaz, Fluidized bed methanation of wood- 510-kw-dresser-rand-turbodyne.
derived producer gas for the production of synthetic natural gas, Ind. Eng. Chem. [48] Rheinhutte Pumps. Pumps for Molten Salts, (2017) http://www.rh-pumps.com/
Res. 49 (2010) 7034–7038. pump-application/molten-salts/.
[13] J. Li, L. Zhou, P. Li, Q. Zhu, J. Gao, F. Gu, F. Su, Enhanced fluidized bed metha- [49] J. Turner, G. Sverdrup, M.K. Mann, P.-C. Maness, B. Kroposki, M. Ghirardi,
nation over a Ni/Al2O3 catalyst for production of synthetic natural gas, Chem. Eng. R.J. Evans, D. Blake, Renewable hydrogen production, Int. J. Energy Res. 32 (2008)
J. 219 (2013) 183–189. 379–407.
[14] J. Kopyscinski, T.J. Schildhauer, S.M.A. Biollaz, Methanation in a fluidized bed [50] U.S. Department of Energy, Energy Department Announces $14 Million to Advance
reactor with high initial CO partial pressure: part I—experimental investigation of Hydrogen Fuel Technologies, (2016) https://energy.gov/eere/articles/energy-
hydrodynamics, mass transfer effects, and carbon deposition, Chem. Eng. Sci. 66 department-announces-14-million-advance-hydrogen-fuel-technologies.
(2011) 924–934. [51] M. De-Saint-Jean, P. Baurens, C. Bouallou, K. Couturier, Economic assessment of a
[15] K.P. Brooks, J. Hu, H. Zhu, R.J. Kee, Methanation of carbon dioxide by hydrogen power-to-substitute natural-gas process including high-temperature steam electro-
reduction using the Sabatier process in microchannel reactors, Chem. Eng. Sci. 62 lysis, Int. J. Hydrogen Energ. 40 (2015) 6487–6500.
(2007) 1161–1170. [52] Renewable natural gas (RNG) technology roadmap for Canada, http://www.cga.ca/
[16] M. Sudiro, A. Bertucco, G. Groppi, E. Tronconi, Simulation of a structured catalytic natural-gas-markets/renewable-natural-gas/2014.
reactor for exothermic methanation reactions producing synthetic natural gas, [53] Biomethane for Transportation Opportunities for Washington State, (2013) http://
Comput. Aided Chem. Eng. 28 (2010) 691–696. www.energy.wsu.edu/Documents/Biomethane_For_Transportation_
[17] C. Junaedi, K. Hawley, D. Walsh, S. Roychoudhury, CO2 reduction assembly pro- WWCleanCities.pdf.
totype using microlith-based Sabatier reactor for ground demonstration, 44th [54] Renewable Natural Gas for Transportation: An Overview of the Feedstock Capacity,
International Conference on Environmental Systems (2014). Economics, and GHG Emission Reduction Benefits of RNG as a Low-carbon Fuel,
[18] C. Junaedi, K. Hawley, S.A. Vilekar, S. Roychoudhury, Evaluation of CO2 adsorber, (2012) http://www.npc.org/FTF_Topic_papers/22-RNG.pdf.
Sabatier reactor, and solid oxide stack for consumable, propellant, and power [55] P. Collet, E. Flottes, A. Favre, L. Raynal, H. Pierre, S. Capela, C. Peregrina, Techno-
production – potential in ISRU architecture, 46th International Conference on economic and life cycle assessment of methane production via biogas upgrading
Environmental Systems (2016). and power to gas technology, Appl. Energy 192 (2017) 282–295.
[19] M. Götz, F. Ortloff, R. Reimert, O. Basha, B.I. Morsi, T. Kolb, Evaluation of organic

382