Energy 135 (2017) 823e832

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Methane thermal decomposition in regenerative heat exchanger

reactor: Experimental and modeling study
Tiina Keipi a, *, Tian Li b, Terese Løvås b, Henrik Tolvanen a, Jukka Konttinen a
a
Laboratory of Chemistry and Bioengineering, Tampere University of Technology, P.O. Box 541, 33101 Tampere, Finland
b
Department of Energy and Process Engineering, Norwegian University of Science and Technology, Kolbjørn Hejes vei 1B, NO-7491 Trondheim, Norway

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

Article history: In this work, thermal decomposition of methane (TDM) was experimentally studied at nominal gas
Received 17 February 2017 temperatures of 1070 Ke1450 K in a non-catalytic laboratory test reactor. The purpose was to use a
Received in revised form simple kinetic mechanism to describe the TDM reaction, which could be applied in industrial reactor
8 May 2017
design. The experimental data was utilized to optimize global kinetic parameters describing the TDM
Accepted 30 June 2017
Available online 4 July 2017
reaction in the test reactor. For comparison, a 37-step reaction mechanism for TDM was adopted from the
literature. When analyzing experimental datasets from the literature, the optimized global kinetics
provided better agreement with the experimental data than the 37-step mechanism when the reactor
Keywords:
CCS
temperature profiles were defined in detail. Since the 37-step mechanism was not able to predict the
Hydrogen production solid carbon formation well enough, the mechanism was slightly adjusted according to a reaction flow
Kinetics and sensitivity analysis. Additionally, it was suggested that the 37-step mechanism can be improved by
Methane cracking optimizing the reaction mechanism by using a detailed experimental data of hydrocarbon formation in
Methane decomposition TDM achieved in an environment where the temperature profiles are fully defined.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction reaction (DHr0 ¼ þ76 kJ=mol) is the following [10]:

Nowadays, H2 is widely utilized in the chemical industry, and CH4 /CðsÞ þ 2H2 : (1)
furthermore, it has been proposed as a potential clean energy car-
rier in the future as it does not produce CO2 when combusted [1]. Instead of gaseous CO2, TDM produces solid carbon, which in
Transition towards hydrogen economy requires the development of some cases can be considered as a value-added product. According
hydrogen production, storage, distribution and utilization tech- to previous studies [11,12], TDM can be an economically feasible
nologies as well as regulatory changes [2e4]. Hydrogen production process only if the value of this product carbon is sufficiently high,
technologies that are both economically feasible and produce less i.e., 500e1000 EUR/tonne of carbon, which is comparable to the
CO2 emissions are needed. Potential larger-scale low-CO2 emission price of low-quality carbon black. When the CO2 emissions are
technologies for hydrogen production are water decomposition [5], considered, TDM has been found as a more preferable technology
water electrolysis by using renewable energy [6] and biomass for H2 production than the currently most applied steam methane
gasification [7]. Additionally, CO2 emissions of steam reforming of reforming (SMR), even if carbon capture and storage would be
natural gas, which is the currently most applied hydrogen pro- coupled with SMR [13e15].
duction technology [1], could be reduced when combined with CO2 A catalyst can be utilized to accelerate the TDM reaction, but the
capture and storage (CCS) [8]. benefit of non-catalytic TDM over catalytic reaction is that it avoids
Decarbonization of fossil fuels, and especially treating natural impurities in the product carbon originating from the catalyst.
gas by the thermal decomposition of methane (TDM), has been Regarding non-catalytic TDM, many studies ([16e22]) have focused
proposed as a potential technology in the transition towards H2 on concentrated solar power reactors, which could enable H2
economy [9]. The reaction equation for the endothermic TDM production with zero CO2 emissions. However, a possible reactor
concept for non-catalytic TDM is a regenerative heat exchanger
reactor, which was first presented by part of the authors [11].
* Corresponding author. Previously, the carbon particle removal to avoid reactor clogging
E-mail address: tiina.keipi@tut.fi (T. Keipi). has been identified as the main challenge restraining the industrial

http://dx.doi.org/10.1016/j.energy.2017.06.176
0360-5442/© 2017 Elsevier Ltd. All rights reserved.
824 T. Keipi et al. / Energy 135 (2017) 823e832

development of TDM [23], which this new concept could avoid. The simplification in the test reactor is that there is no bed material
bed material in the reactor enables both effective heat transfer and circulation. The substantially high heat capacity of the solid bed
conveys the product carbon out from the reactor. material stabilizes the reactor temperature and makes the reactor
Modeling the reaction kinetics involved is an important part of appropriate for experimental kinetic studies.
the TDM reactor design. Both the global kinetics [18,24] and more The primary part of the experimental setup was the vertically
detailed kinetic mechanisms [18,19] have been presented for non- mounted reactor tube made of Kanthal APM iron-chromium-
catalytic TDM modeling in the literature. However, the reaction aluminum (FeCrAl) alloy. The reactor tube had the following di-
temperatures in these studies have been from 1500 K to over 2000 mensions: the inner diameter of 73 mm, the outer diameter of
K. Therefore, those mechanisms cannot be directly applied to the 83 mm and the length of 2500 mm. In order to enable opening and
reaction conditions of a regenerative heat exchanger reactor, which closing of the reactor tube, there were flanges at both ends of the
operates at a temperature of 1000 Ke1500 K and an approximate tube. During the operation, the joints of the flanges were sealed
residence time of 1 se10 s. with copper seals.
In this study, the objective was to find, for reactor design pur- The middle part of the reactor tube was electrically heated with
poses, a relatively simple kinetic model to describe the non- three circular Fibrothal RAC 200/500 heating modules which were
catalytic TDM reaction below 1500 K. First, the simplest way to insulated with Fibrothal insulation modules. Each heating module
describe the reaction kinetics, a global mechanism, was studied in had a heating power of 19.7 kW. The modules were separately
order to find out its applicability on modeling TDM. The experi- controlled with an Eurotherm 7100 A thyristor unit combined with
mental data, which was achieved with a laboratory test reactor that an Eurotherm 3204 temperature controller.
mimics the operation of a regenerative heat exchanger reactor, was The reactor tube was filled with spherical ceramic beads with an
applied to optimizing the kinetic parameters for the global TDM approximate diameter of 10 mm. The ceramic beads served as a
reaction by using a specifically written MATLAB-based plug flow heat exchange material that stabilized the temperature and gas
reactor model. Second, the optimized global kinetics was compared flow profiles in the reactor tube. Moreover, a part of the product
with a rather simple 37-step reaction mechanism presented by carbon deposited on the surface of the beads. The void fraction, i.e.,
Ozalp et al. [19]. The 37-step reaction mechanism was further the proportion of empty space in the reactor, was separately
developed based on a reaction flow and sensitivity analysis to defined to be 0.4. The determination was done experimentally by
improve its ability to predict solid carbon formation at tempera- counting the number of beads that fit into a certain volume and
tures 1000 Ke2000 K. Thus, this article complements the previous estimating the void fraction mathematically.
knowledge in this field of study by providing (i) experimental data The temperature profile inside the reactor tube was measured
of non-catalytic TDM reaction at nominal gas temperatures of by using eight measuring points as shown in Fig. 1. Temperature
1070 Ke1450 K with specially defined temperature profiles, (ii) was measured continuously with K-type thermocouples and
kinetic parameters for the global TDM reaction that were success- recorded with a data logger in 2-s intervals. The experiments were
fully applied to predict methane conversion in non-catalytic TDM conducted with reactor maximum temperatures between 1070 K
and (iii) suggestions for improving the 37-step reaction mechanism and 1450 K. The gas mixture containing CH4 (purity 99.995%) and
for non-catalytic TDM presented in the literature. N2 (purity 99.95%) with different mixture ratios entered the reactor
at room temperature. For modeling purposes, the product gas
2. Experimental section temperature at the reactor end was assumed to be 30 K less than
that measured at the eighth measurement point, which was located
2.1. Experimental setup 10 cm below the reactor outlet. The assumption is based on the
extrapolation of the decreasing gas temperature at the reactor end.
The experimental work presented and discussed in this paper This assumption has a minor significance, since the gas tempera-
was conducted with an experimental setup demonstrated in Fig. 1. ture at the reactor end is well below the temperature where TDM
The laboratory test reactor mimics the regenerative heat exchanger reaction occurs.
reactor previously presented by part of the authors [11]. The main To control the gas flow rates, mass flow controllers (Bronkhorst)
in the range of 0e15 dm3/min were used. The product gas flowed
continuously out from the top end of the reactor tube. After exiting
the reactor, part of the outcoming gas was extracted for analyzing
purposes. The part of the product gas that was not analyzed was
bubbled through water to remove the remaining carbon particles
before discarding. The reactor was maintained inert during the
periods of heating and quenching with a constant N2 flow of 6 dm3/
min.
Prior to the gas analysis, the product gas flowed through an
inline filter with stainless steel housing that sifted 93% of the car-
bon particles with a diameter more than 0:01 mm out. A continuous
analysis of the product gas composition was conducted with a
Fourier transform infrared spectroscopy (FTIR) in 30-s intervals.
The examination procedure was as follows: the product gas was
pumped into the sampling unit during the first 10 s and the
composition of the sample gas was examined during the following
20 s. The FTIR analyzer has a limitation that it cannot detect bia-
tomic homonuclear molecules, such as N2 and H2 in this case.
Therefore, the methane conversion was calculated by comparing
the measured methane concentration values to the initial concen-
tration. The methane conversion a is defined in this article as
Fig. 1. A sketch of the laboratory scale test rector. follows:
 T. Keipi et al. / Energy 135 (2017) 823e832 825

(mol=cm3 ), kforward and kbackward are rate constants and n and m are
nCH4 ;inlet  nCH4 ;outlet
a¼ ; (2) the reaction orders () for forward and backward reactions,
nCH4 ;inlet respectively [26]. The rate constants for forward and backward
reactions were assumed to follow the Arrhenius law:
where nCH4 ;inlet and nCH4 ;outlet are respectively the amount of CH4
(mol) at the reactor inlet and outlet. k ¼ A,exp½  Ea =ðRu TÞ; (4)

2.2. Experimental arrangement where A is the frequency factor, Ea is the activation energy (kJ/mol),
Ru is the gas constant (8.314 J/molK) and T is temperature (K) [26].
A total of 48 separate cases were measured. The following pa- The rate constants and frequency factors use units of cm3, J, mol and
rameters were recorded in each case: the volumetric flow of CH4/N2 s.
mixture, the reactor temperature profile and the composition of the In this study, the experimental data was used to optimize the
product gas. Before each measurement case, a stable temperature Arrhenius parameters for the forward reaction and the reaction
profile was achieved. Similarly, the reported methane conversion orders for forward and backward reactions. Next, the kinetic pa-
was calculated based on an average of 10 stabilized gas composition rameters for the backward reaction were solved by using the
measurement points, which corresponds to a measurement period following relationship [26]:
of 2.5 min. Experiments were conducted with total gas volumetric
kforward
flows ranging from 5 dm3/min to 16 dm3/min measured at NTP kbackward ¼ : (5)
conditions, which resulted in residence times between 6 s and 21 s. Kc
The inlet gas composition, temperature profiles and corresponding The equilibrium constant Kc () values were taken from the
methane conversion values for each experimental case are pre- literature [27] and modified into the following exponential form:
sented in Appendix.
In order to ensure that the product carbon that deposited in the Kc ¼ 941; 272:1,exp ½  96; 425:0=ðRu TÞ: (6)
reactor did not affect the gas flow, the heat transfer, or the TDM
The gas flow in the laboratory test reactor was modeled as a plug
reaction, an oxygen flow was fed to the reactor between each
flow where the tracked element had a constant mass. The pressure
measurement to oxidize the product carbon that had formed dur-
drop dP (Pa) for a packed bed reactor was calculated as follows [28]:
ing the previous measurement. Furthermore, the effect of depos-
ited carbon during the experimental runs was minor since the mass
dP 150mVc 1:75rVc2
of the ceramic beads in the reactor was thousand times the theo- ¼ þ ; (7)
dx L 2c Lc
retical maximum carbon amount that could have been deposited
during an average 15-min experimental run. The focus in this study
where m is the dynamic viscosity (Ns/m2), r is density (kg/m3). The
was not on the product carbon quality, however, some qualitative
characteristic velocity Vc (m/s) and characteristic length L c (m)
analyses were conducted in order to verify (i) that carbon was
were defined as follows:
formed during the tests and (ii) based on elementary analysis the
carbon purity was high and it contained only a hint of impurities. m_
Vc ¼ (8)
rεv Ac
2.3. Measurement uncertainty evaluation
and
The uncertainty of the measurements originated mostly from  
two sources: measurement of the volumetric gas flow with the εv
L c ¼ dp ; (9)
mass flow controllers and the gas analysis with FTIR. The numeric 1  εv
values for measurement uncertainties were given by the manu-
facturer of each measurement device. In the case of measuring the where m_ is the gas mass flow (kg/s), Ac is the cross-sectional area of
volumetric gas flow the uncertainty was 0.5% whereas in the gas the bed (m2), dp is the diameter of the spherical bed particles (m)
composition analysis the measurement uncertainty had an abso- and εv is the void fraction (), i.e., the proportion of the empty
lute value of 0.2% points. The measurement uncertainties are space in the packed bed [28].
shown in the experimental results in Section 4. The error bars for
the measurement uncertainties were truncated when exceeding 3.2. Kinetic parameter optimization
the natural limits, i.e., conversion beyond values from 0 to 1, ac-
cording to the recommendation given by the Analytical Methods The kinetic parameters were optimized by using a particularly
Committee of the Royal Society of Chemistry [25]. written MATLAB-based function. Table 1 presents the optimized
kinetic parameters and the derived Arrhenius parameters for the
3. Modeling of the reaction kinetics backward reaction. In the MATLAB-based reactor model, a plug
flow element was tracked through the reactor length and the
3.1. Chemical kinetics methane conversion at the reactor end was compared with the
experimental values. The function minimized the square error
For reactor design purposes was the chemical kinetics of the
global methane decomposition reaction studied. The reaction rate r Table 1
(mol/s) was defined as: The kinetic parameters defined in Eqs. (3) and (4) achieved by optimization of the
global TDM reaction parameters to correspond the experimental data. The reaction
dnCH4 rate uses the units cm3, J, mol and s.
r¼ ¼ kforward ,cCH4 n  kbackward ,cH2 m ; (3)
dt A Ea (kJ/mol) n () m ()
12
following standard formulations where nCH4 is the amount of CH4 Forward 8.5708$10 337.12 1.123 e
Backward 1.1190$107 243.16 e 0.9296
(mol), t is time (s), cCH4 and cH2 are the concentrations of CH4 and H2
826 T. Keipi et al. / Energy 135 (2017) 823e832

between the calculated methane conversion and the measured separate numerical analysis was conducted in which the heat
values in each case. The optimization was conducted by using the transfer efficiency between the beads and gas was defined. In the
function ‘fminsearch’ in MATLAB. analysis, the test reactor was modeled as a packed bed reactor and
In the literature, activation energies of 208 kJ/mol [24] and the gas temperature was calculated based on heat transfer between
370 kJ/mol [18] have been reported when the kinetic parameters the gas and bed. As a simplification, the bed temperature was
for a global non-catalytic TDM reaction have been derived from defined equal with the measured reactor temperature profile. The
experiments conducted at temperatures above 1500 K. For a high- conclusion was that the gas temperature deviates from the bed
temperature (above 2000 K) rapid decomposition the activation temperature only at the beginning and at the end of the reactor
energies of 356 kJ/mole402 kJ/mol have been reported for a first where the reactor temperature changes steeply. However, the gas
order global reaction in the literature [24]. Thus, the activation temperature is practically the same as the bed temperature in the
energy defined in this study for the global reaction, 337 kJ/mol for part of the reactor where the temperature is above 1000 K. This is
lower temperature, is in line with the activation energies for non- also the most remarkable area considering the kinetic parameter
catalytic decomposition reaction reported by others. Significantly optimization. Thus, it is justified to neglect the heat transfer be-
lower values for the activation energy (147 kJ/mole162 kJ/mol) tween the gas and the reactor from the reaction kinetics analysis.
have been reported when carbon particles have been laden to a
vortex-flow at temperatures above 900 K [29], or 31 kJ/mole85 kJ/
mol when a carbon catalyst has been added to the solar powered 4. Results
reactor at temperatures above 723 K [30]. Furthermore, the reac-
tion order for the forward reaction defined in this study, 1.123, is 4.1. Experiments and global kinetics
close to value 1, which is generally presented for the global TDM
reaction [18,24,29]. The methane conversion values from the experiments and from
the modeling of the test reactor using the optimized global kinetics
3.3. Constant pressure reactor model are shown in Fig. 2. Each data point corresponds to one of the 48
experimental cases. The results achieved with the optimized global
The MATLAB-based model was suitable for studying the global kinetics did not depend on the model selection but both the plug
TDM reaction. However, in order to simulate the reactor with flow reactor model and the constant pressure reactor model pre-
detailed chemistry and enable the use of the pre-defined temper- dicted highly equal methane conversion values. The error bars
ature profiles from experiments, the homogeneous 0D constant indicate the uncertainty in the experiments. Since the measure-
pressure reactor model in LOGESoft was found appropriate [31]. In ment uncertainty in the gas composition analysis had an absolute
this model, the reactor was assumed to be closed system, and hence value that depended on the selected measurement range, the
the mass conservation equation was as follows: measurement uncertainty was the highest in cases where the
methane concentration in the input gas was the lowest. In order to
vm clarify the graphical format in Fig. 2, the cases are presented in the
¼ 0: (10)
vt order of increasing experimental methane conversion and the same
The mass fraction of the ith species in the gas mixture, Yi (¡), order is used throughout the paper. Since the global kinetic pa-
was rameters were optimized using the experimental data, the
modeling values follow well the experimental ones. In most of the
vYi ri Mi cases, the modeling values are within the limits of measurement
¼ ; (11)
vt r uncertainties in Fig. 2.
Four experimental cases are presented in Fig. 3 to exemplify the
where r is the density (kg/m3), ri is the reaction rate (mol/(m3s)) temperature profiles in the test reactor and to compare the
and Mi is the molecular mass (kg/mol) of the ith species in the gas experimental methane conversion values to the modeling values
mixture [32]. The thermodynamic properties of gases were calcu-
lated by utilizing the NASA polynomials [31].
Even though the constant pressure reactor model was applied, a
pressure drop does in fact occur in the test reactor. This pressure
drop was calculated during the kinetic parameter optimization and
according to the results the pressure drop along the total reactor
length was not more than 0.3%. Therefore, the constant pressure
reactor model was considered suitable for simulating the test
reactor.
Initially the temperature profiles in the test reactor were
measured as a function of position, but those were converted to
time-dependent profiles during the kinetic parameter optimiza-
tion. The gas temperature in the constant pressure model was set to
follow the time-dependent temperature profiles so that the
modeling results were directly comparable with the experimental
data. Furthermore, the gas temperature in the test reactor was
assumed to be uniform in the radial direction. This assumption was
considered valid since the gas velocity was rather low and the heat
capacity of the beads in the reactor was large compared to that of
the gas mixture. Additionally, the distances between the beads
were small. Fig. 2. Comparison of the experimental methane conversion with the values from the
Furthermore, it was assumed that the gas and bed temperatures modeling using the optimized global kinetics. Each case represents a single mea-
in the reactor were equal. In order to justify this assumption, a surement. Detailed data for each measurement is presented in Appendix.
 T. Keipi et al. / Energy 135 (2017) 823e832 827

Fig. 3. Comparison of the experimental methane conversion values to the modeling values with the optimized global kinetics in four example cases. The reactor temperature fits
were formed during the optimization process based on reactor temperature measurements.

with the optimized global kinetics. These cases were chosen as they kinetics, comparison was done to other mechanisms presented in
represent the high (Case 43 and Case 34) and low (Case 17 and Case the literature. Detailed mechanisms for the TDM reaction have been
16) nominal gas temperatures during the experiments. The devia- presented previously for instance by Rodat et al. [18] and Ozalp
tion between the experimental and modeling methane conversion et al. [19]. Among those, the mechanism by Ozalp et al. was chosen
values was smaller in cases where the reactor maximum temper- here since it was available with full kinetic parameter information.
ature was high, above 1300 K, and as a result, the methane con- This kinetic mechanism was originally developed by Olsvik and
version was high. This is mainly due to the global mechanism Billaud [33] who conducted non-catalytic methane decomposition
optimization procedure, where the function minimized the abso- experiments in a plug flow reactor at reactor maximum tempera-
lute difference between the model values and experiments, and ture of 1273 K with methane conversion values below 0.01 and
thus, weighted the high methane conversion values. modeled the TDM reaction with homogenous gas phase chemical
kinetics. The reaction kinetics model consisting of 36 reactions was
4.2. 37-Step reaction mechanism used to explain the decomposition of methane to other hydrocar-
bons and the modeling results correlated well with the
In order to evaluate the operation of the optimized global experiments.

Fig. 4. Comparison of methane conversion values between the own experiments and Fig. 5. Comparison of the methane conversion between Plevan's experimental results,
modeling with the 37-step reaction mechanism of Ozalp et al. [19]. Each case repre- modeling results with the optimized global kinetics and with the 37-step reaction
sents a single measurement. mechanism of Ozalp et al. [19].
828 T. Keipi et al. / Energy 135 (2017) 823e832

Later, Ozalp et al. [19] revised the reaction kinetics by adding an the experiments. Furthermore, the 37-step mechanism predicted
irreversible reaction that takes into account the solid carbon for- negligible solid carbon formation, even though the carbon forma-
mation. Ozalp et al. used the 37-step reaction mechanism to tion was detected experimentally. This coincides with the conclu-
simulate the non-catalytic methane decomposition at sion stated by Ozalp et al. [19] that this kinetic model cannot
1823 Ke2300 K with short residence times (<0.1 s). Even though predict solid carbon formation well at temperatures below 2000 K.
Ozalp et al. presented that solid carbon was formed during the In order to test the optimized global kinetics and the 37-step
experiments at 1823 K, their kinetic mechanism proposed negli- mechanism for different reactor conditions and different temper-
gible carbon formation at that temperature. Ozalp et al. [19] re- ature ranges, experiments from the literature [34] were chosen to
ported that the kinetic mechanism was able to predict solid carbon be modeled with both kinetics. These experiments were conducted
formation only at temperatures above 2100 K, which is well above by Plevan et al. [34], who studied TDM in a bubble column reactor
the temperature range used in this study and also in industrial and conducted additional tests with a blank tube at the nominal gas
reactors other than solar or plasma reactors. temperatures of 1023 K, 1123 K and 1173 K. Furthermore, Plevan
First, the 37-step mechanism was applied to model the TDM et al. conducted numerical studies of this setup and concluded that
reaction in the test reactor. The experimental and modeling results the global reaction mechanisms found in the literature were not
are shown in Fig. 4. According to the results, the 37-step mecha- sufficient to explain the TDM reaction, and proposed the utilization
nism constantly predicted lower methane conversion values than of a more detailed mechanism and reverse reactions.

Fig. 6. The reaction flow analysis (mol=m3 ) and sensitivity analysis for 100% CH4 at 1500 K and at reaction time of 10 s by using the 37-step reaction mechanism of Ozalp et al. [19].
The two most essential reactions in this study are marked with dashed circles. The sensitivity analysis of carbon production presents the five most effective reactions in the 37-step
reaction mechanism.
 T. Keipi et al. / Energy 135 (2017) 823e832 829

Here, the experimental conditions in the study of Plevan et al. Table 2

[34] are modeled with the optimized global kinetics proposed in The original and adjusted Arrhenius parameters (defined in Eq. (4)) for (Reaction 37)
in the reaction mechanism developed by Ozalp et al. [19].
this study and with the 37-step mechanism adopted from the
literature. The gas residence time in the reactor was calculated A Ea (kJ/mol) n () Ref.
based on the gas volumetric flows, reactor dimensions and reactor C2H / 2C þ H(37) 4.68$10 16 124 0.0 [19]
temperature profiles presented in the study of Plevan et al. The C2H / 2C þ H(37) 4.68$1015 62 0.0 This work
experimental and modeling results are presented in Fig. 5. The
optimized global kinetics shows especially good agreement with
the experimental data at temperatures 1023 K and 1123 K. In all
adjusted 37-step mechanism predicted the H2 production better
cases the 37-step mechanism predicts too low methane conversion
than the original 37-step mechanism and was also able to predict
similarly than is shown above in Fig. 4.
the solid carbon formation at lower temperatures. Moreover, the
The global kinetic parameters were optimized in this study
adjusted 37-step mechanism predicted acetylene decomposition,
based on experimental data that contains measured temperature
even though the rate of this reaction was slightly too high in
profiles, which due to the beads corresponded well with the gas
comparison with the experiments.
temperature profiles, and a reasonable correspondence was ach-
Second, the adjusted 37-step mechanism was applied to model
ieved between the model values and the experiments of Plevan
the experiments of Billaud and Gueret [37], who studied methane
et al. [34]. In contrast, the 37-step mechanism, which was originally
decomposition to other hydrocarbons with an electrically heated
developed by Olsvik and Billaud [33], is based on experiments
flow reactor of which temperature was reported as 1263 K. The
conducted at the reactor maximum temperature of 1273 K and
modeling values were mainly in a good agreement with the ex-
simulations conducted at the constant temperature of 1243 K,
periments as shown in Fig. 8. The modeling values deviated the
which is 30 K below the reactor maximum temperature in the
most from the experiments in the case of acetylene and benzene.
experiments. The inadequate temperature profile definition during
These deviations may be partly because the 37-step mechanism
the experiments and simulations might be the fundamental reason
consists of a limited amount of hydrocarbon species, and therefore,
for the deviation between the modeling and experimental values in
the mechanism was not able to predict the formation and decom-
Figs. 4 and 5. Studying this further would require revising the 37-
position of all the species measured by Billaud and Gueret.
step mechanism by using experimental data of hydrocarbon for-
Both in Figs. 7 and 8 the models predict faster methane
mation in TDM produced in an environment where the tempera-
decomposition and formation of other components that has been
ture profiles are defined in detail. However, the experimental data
detected during the experiments. The reason for this might be that
in this study was not sufficient for this purpose. Therefore, this
the temperature profiles in the experiments in Refs. [36,37] were
study focuses on another previously reported flaw of the 37-step
not constant even thought it was assumed in this study. Since only
mechanism, which is the inability to predict solid carbon forma-
one temperature value was given for each experiment in these
tion below 2000 K [19]. This mechanism was chosen for further
studies, modeling was conducted here in constant temperatures,
optimization since it is a rather simple mechanism and the full
which were equal to the given temperature levels. When analyzing
kinetic parameter information is publicly available. Therefore, this
Figs. 7 and 8, it appears that a temperature ramp might have
model could be utilized for prospective research and reactor design
occurred in the first part of the reactor during the experiments
purposes.
where the gas has warmed up from the inlet temperature to the
given temperature level. An increasing temperature ramp in the
4.3. Sensitivity analysis and mechanism improvement
measurements would cause a slower methane decomposition and
product gas formation in the first part of the reactor than a constant
In order to identify the most important reactions in the mech-
temperature profile.
anism concerning the formation of solid carbon, a reaction flow and
Finally, the adjusted 37-step mechanism was compared with the
sensitivity analyses were performed according to the work of
Soyhan et al. [35] (Fig. 6). The analyses were conducted at a refer-
ence temperature of 1500 K, considerably lower than the 2000 K
previously studied.
According to the reaction flow analysis, a high amount of ben-
zene was formed from acetylene whereas the route from acetylene
to C2 H and finally to solid carbon was minor. Furthermore, the
sensitivity analysis revealed how the benzene formation competed
with the carbon formation. Therefore, the reaction of solid carbon
formation from C2 H (Reaction 37) was adjusted in order to accel-
erate the solid carbon formation and, as a result, speed up the C2 H
formation from acetylene. As a consequence, this modification
reduced benzene formation from acetylene.
The activation energy of (Reaction 37) was decreased to speed
up the carbon formation and the frequency factor was adjusted
slightly by trial and error method. The resulting kinetic parameters
are shown in Table 2. Finally, the adjusted mechanism was applied
to model several TDM experiments presented in the literature.
First, the adjusted parameters were tested in an experimental
case (10% CH4 in Ar at 1823 K) presented by Rodat et al. [36], which
was modeled by using the original 37-step mechanism and com-
parison was conducted to the modeling results with the adjusted Fig. 7. Comparison of the experimental methane conversion data (10% CH4 in Ar at
37-step mechanism (Fig. 7). Adjusting the kinetics improved the 1823 K) from Rodat et al. [36] and the modeling results using the original and adjusted
agreement between the model values and experiments. The kinetic parameters for the 37-step reaction mechanism.
830 T. Keipi et al. / Energy 135 (2017) 823e832

Fig. 8. Comparison of the product gas composition from methane decomposition experiments at 1263 K and 1 atm (marker symbols) [37] and from modeling with the adjusted 37-
step reaction mechanism (solid lines).

simulation results of Rodat et al. [18], who studied methane results of Rodat et al. [18] in Fig. 9. A total mass fraction less than 1
decomposition with a highly detailed homogenous reaction in Fig. 9 indicates that hydrocarbons other than selected here have
mechanism (over 240 species and 14,000 reactions) and verified been formed.
this kinetic model with their experimental data achieved at Similarly as observed previously in Fig. 7, the original 37-step
1670 Ke1770 K. Both the original 37-step mechanism and the mechanism predicted low H2 and high acetylene formation as
adjusted 37-step mechanism are here studied in a case where the well as a slow decomposition of acetylene in comparison with the
constant temperature is 1700 K and residence time is 1  106 simulations of Rodat et al. [18]. The original 37-step mechanism
se100 s. The modeling results are compared with the simulation was not able to predict the solid carbon formation at the selected

Fig. 9. Results of methane decomposition modeling at 1700 K by using the original 37-step reaction mechanism (dash-dot lines) and the adjusted 37-step reaction mechanism
(solid lines) compared with the simulation results of Rodat et al. [18] (dashed lines).
 T. Keipi et al. / Energy 135 (2017) 823e832 831

temperature of 1700 K, but instead predicted formation of very high dx.doi.org/10.1016/j.energy.2017.06.176.

amounts of benzene, as was previously noticed in the reaction flow
and sensitivity analyses in Fig. 6. References
Adjusting the 37-step mechanism improved the agreement to
the simulation results of Rodat et al. [18]. The major improvement [1] International Energy Agency (IEA). Technology roadmap - hydrogen and fuel
cells. Tech. rep. 2015. https://www.iea.org/publications/freepublications/
of the adjusted kinetics was that it predicted carbon formation publication/TechnologyRoadmapHydrogenandFuelCells.pdf.
comparable to the Rodat's results and predicted a negligible [2] Neef H-J. International overview of hydrogen and fuel cell research. Energy
amount of benzene. Furthermore, the adjusted 37-step mechanism 2009;34:327e33. http://dx.doi.org/10.1016/j.energy.2008.08.014.
[3] Penner SS. Steps toward the hydrogen economy. Energy 2006;31:33e43.
predicted a more moderate formation and the faster decomposition http://dx.doi.org/10.1016/j.energy.2004.04.060.
of acetylene compared with the original 37-step mechanism. [4] da Silva Veras T, Mozer TS, da Costa Rubim Messeder dos Santos D, da Silva
In conclusion, it can be stated that the adjustment of the 37-step Cesar A. Hydrogen: trends, production and characterization of the main
process worldwide. Int J Hydrogen Energy 2017;42:2018e33. http://
mechanism provided better correspondence with the experimental
dx.doi.org/10.1016/j.ijhydene.2016.08.219.
and simulation results at temperatures below 2000 K. The solid [5] Orhan MF, Babu BS. Investigation of an integrated hydrogen production sys-
carbon formation in the 37-step mechanism followed a particularly tem based on nuclear and renewable energy sources: comparative evaluation
simple reaction path, but in general, the solid carbon formation of hydrogen production options with a regenerative fuel cell system. Energy
2015;88:801e20. http://dx.doi.org/10.1016/j.energy.2015.06.009.
from hydrocarbon pyrolysis is stated to proceed via formation of [6] Nastasi B, Basso GL. Hydrogen to link heat and electricity in the transition
PAHs, as presented in numerous experimental [38,39] and towards future smart energy systems. Energy 2016;110:5e22. http://
modeling studies [40e42]. Regarding the carbon formation, an dx.doi.org/10.1016/j.energy.2016.03.097.
[7] Doranehgard MH, Samadyar H, Mesbah M, Haratipour P, Samiezade S. High-
option to improve the 37-step mechanism could be to take into purity hydrogen production with in situ CO2 capture based on biomass gasi-
account these more detailed models for carbon formation in TDM. fication. Fuel 2017;202:29e35. http://dx.doi.org/10.1016/j.fuel.2017.04.014.
However, rather than adding complex PAH and soot formation [8] Blok K, Williams RH, Katofsky RE, Hendriks CA. Hydrogen production from
natural gas, sequestration of recovered CO2 in depleted gas wells and
mechanisms to the 37-step mechanism, for example, the kinetic enhanced natural gas recovery. Energy 1997;22:161e8. http://dx.doi.org/
model of Rodat et al. [18], where the PAH and soot formation are 10.1016/S0360-5442(96)00136-3.
included already, could be utilized. [9] Weger L, Aba nades A, Butler T. Methane cracking as a bridge technology to the
hydrogen economy. Int J Hydrogen Energy 2017;42:720e31. http://
dx.doi.org/10.1016/j.ijhydene.2016.11.029.
5. Conclusion [10] Muradov NZ. CO2-Free production of hydrogen by catalytic pyrolysis of hy-
drocarbon fuel. Energy & Fuels 1998;12:41e8. http://dx.doi.org/10.1021/
ef9701145.
In this paper, two simple kinetic mechanisms to describe the
[11] Keipi T, Hankalin V, Nummelin J, Raiko R. Techno-economic analysis of four
TDM reaction in a non-catalytic reactor were studied. The global concepts for thermal decomposition of methane: reduction of CO2 emissions
kinetic mechanism, where the kinetic parameters were optimized in natural gas combustion. Energy Convers Manag 2016;110:1e12. http://
based on experiments conducted in this study, was found to predict dx.doi.org/10.1016/j.enconman.2015.11.057.
[12] Triphob N, Wongsakulphasatch S, Kiatkittipong W, Charinpanitkul T,
methane conversion accurately in non-catalytic TDM experiments Praserthdam P, Assabumrungrat S. Integrated methane decomposition and
where the temperature profiles were defined. Therefore, this solid oxide fuel cell for efficient electrical power generation and carbon
mechanism will be chosen for future reactor design. The 37-step capture. Chem Eng Res Des 2012;90:2223e34. http://dx.doi.org/10.1016/
j.cherd.2012.05.014.
mechanism, which was adopted from the literature, has potential [13] Dufour J, Serrano DP, G alvez JL, Moreno J, García C. Life cycle assessment of
in a detailed modeling of hydrocarbon formation in TDM. However, processes for hydrogen production. Environmental feasibility and reduction of
two flaws were identified concerning this mechanism. First, the 37- greenhouse gases emissions. Int J Hydrogen Energy 2009;34:1370e6. http://
dx.doi.org/10.1016/j.ijhydene.2008.11.053.
step mechanism was unable to predict solid carbon formation, but [14] Dufour J, Galvez JL, Serrano DP, Moreno J, Martínez G. Life cycle assessment of
that was successfully improved in this study. Second, the temper- hydrogen production by methane decomposition using carbonaceous cata-
ature profile determination during the 37-step mechanism devel- lysts. Int J Hydrogen Energy 2010;35:1205e12. http://dx.doi.org/10.1016/
j.ijhydene.2009.11.093.
opment has been inadequate and that might be the reason for lvez JL, Gonzalez A, Soria E, Fierro JLG. Life cycle
[15] Dufour J, Serrano DP, Ga
inaccuracies between the modeling and experimental results assessment of alternatives for hydrogen production from renewable and fossil
detected in this study. Improvement would require producing sources. Int J Hydrogen Energy 2012;37:1173e83. http://dx.doi.org/10.1016/
j.ijhydene.2011.09.135.
experimental data of formation of numerous hydrocarbons in TDM
[16] Dahl JK, Buechler KJ, Weimer AW, Lewandowski A, Bingham C. Solar-thermal
in an environment where the reactor temperature profiles are dissociation of methane in a fluid-wall aerosol flow reactor. Int J Hydrogen
defined in detail, and finally, utilizing this data to optimize the ki- Energy 2004;29:725e36. http://dx.doi.org/10.1016/j.ijhydene.2003.08.009.
netic parameters of the 37-step mechanism again. [17] Dahl JK, Buechler KJ, Finley R, Stanislaus T, Weimer AW, A L, et al. Rapid solar-
thermal dissociation of natural gas in an aerosol flow reactor. Energy 2004;29:
715e25. http://dx.doi.org/10.1016/S0360-5442(03)00179-8.
Acknowledgments [18] Rodat S, Abanades S, Coulie  J, Flamant G. Kinetic modelling of methane
decomposition in a tubular solar reactor. Chem Eng J 2009;146:120e7. http://
dx.doi.org/10.1016/j.cej.2008.09.008.
The study presented in this paper was carried out in the Carbon [19] Ozalp N, Ibrik K, Al-Meer M. Kinetics and heat transfer analysis of carbon
Capture and Storage Program (CCSP) research program coordinated catalyzed solar cracking process. Energy 2013;55:74e81. http://dx.doi.org/
by CLIC Innovation Ltd. with funding from the Finnish Funding 10.1016/j.energy.2013.02.022.
[20] Abanades S, Flamant G. Hydrogen production from solar thermal dissociation
Agency for Technology and Innovation, Tekes. The financial support of methane in a high-temperature fluid-wall chemical reactor. In: Chemical
received from Fortum, Gasum, Helen, and Neste is gratefully engineering and processing: process intensification, vol. 47; 2008. p. 490e8.
acknowledged. Furthermore, the first author wishes to acknowl- http://dx.doi.org/10.1016/j.cep.2007.01.006. 10th French Congress on Chem-
ical Engineering.
edge the financial support of the Tampere University of Technology [21] Maag G, Zanganeh G, Steinfeld A. Solar thermal cracking of methane in a
(Finland) graduate school, Fortum Foundation and Walter Ahlstro €m particle-flow reactor for the co-production of hydrogen and carbon. Int J
Foundation. Last, the contribution of Juha Savolainen toward Hydrogen Energy 2009;34:7676e85. http://dx.doi.org/10.1016/
j.ijhydene.2009.07.037.
designing and installing the electrical installation and automation
[22] Hirsch D, Steinfeld A. Solar hydrogen production by thermal decomposition of
system of the experimental setup is acknowledged. natural gas using a vortex-flow reactor. Int J Hydrogen Energy 2004;29:
47e55. http://dx.doi.org/10.1016/S0360-3199(03)00048-X.
[23] Aba nades A, Rubbia C, Salmieri D. Technological challenges for industrial
Appendix A. Supplementary data
development of hydrogen production based on methane cracking. Energy
2012;46:359e63. http://dx.doi.org/10.1016/j.energy.2012.08.015.
Supplementary data related to this article can be found at http:// [24] Dahl JK, Barocas VH, Clough DE, Weimer AW. Intrinsic kinetics for rapid
832 T. Keipi et al. / Energy 135 (2017) 823e832

decomposition of methane in an aerosol flow reactor. Int J Hydrogen Energy http://dx.doi.org/10.1016/j.ijhydene.2015.04.062.

2002;27:377e86. http://dx.doi.org/10.1016/S0360-3199(01)00140-9. [35] H. Soyhan, P. Amne us, F. Mauss, C. Sorusbay, A skeletal kinetic mechanism for
[25] Measurement uncertainty and confidence intervals near natural limits, AMC the oxidation of iso-octane and N-Heptane validated under engine knock
Technical Briefs No. 26A (2008). URL www.rsc.org/amc. conditions, SAE Technical Paper1999-01-3484. http://dx.doi.org/10.4271/
[26] Turns SR. An introduction to combustion: concepts and applications. McGraw- 1999-01-3484.
Hill Companies, Inc; 2012. [36] Rodat S, Abanades S, Flamant G. Co-production of hydrogen and carbon black
[27] Villacampa JI, Royo C, Romeo E, Montoya JA, Angel PD, Monzo na A. Catalytic from solar thermal methane splitting in a tubular reactor prototype. Sol En-
decomposition of methane over Ni-Al2 O3 coprecipitated catalysts. React ergy 2011;85:645e52. http://dx.doi.org/10.1016/j.solener.2010.02.016.
Regen Stud Appl Catal A General 2003;252:363e83. http://dx.doi.org/ [37] Billaud F, Gueret C. Thermal decomposition of pure methane at 1263 K. Ex-
10.1016/S0926-860X(03)00492-7. periments and mechanistic modelling. Thermochim Acta 1992;211:303e22.
[28] Mills AF. Basics of heat and mass transfer. Prentice Hall; 1999. http://dx.doi.org/10.1016/0040-6031(92)87029-A.
[29] Trommer D, Hirsch D, Steinfield A. Kinetic investigation of the thermal [38] Bo€hm H, Jander H, Tanke D. PAH growth and soot formation in the pyrolysis of
decomposition of CH4 by direct irradiation of a vortex-flow laden with carbon acetylene and benzene at high temperatures and pressures: modeling and
particles. Int J Hydrogen Prod 2004;29:627e33. http://dx.doi.org/10.1016/ experiment 27. In: Twenty-seventh symposium (international) on combus-
j.ijhydene.2003.07.001. tion volume one; 1998. p. 1605e12. http://dx.doi.org/10.1016/S0082-
[30] Abanades S, Kimura H, Otsuka H. Kinetic investigation of carbon-catalyzed 0784(98)80570-5.
methane decomposition in a thermogravimetric solar reactor. Int J [39] Bo€hm H, Jander H. PAH formation in acetyleneebenzene pyrolysis. Phys Chem
Hydrogen Energy 2015;40:10744e55. http://dx.doi.org/10.1016/ Chem Phys 1999;1:3775e81. http://dx.doi.org/10.1039/A903306H.
j.ijhydene.2015.07.023. [40] Lucas P, Marchand A. Pyrolytic carbon deposition from methane: an analytical
[31] LOGE AB. Lund, LOGESOFT manual, book 1. 2016. approach to the chemical process. Carbon 1990;28:207e19. http://dx.doi.org/
[32] LOGE AB. Lund, LOGESOFT manual, book 2. 2016. 10.1016/0008-6223(90)90115-F.
[33] Olsvik O, Billaud F. Modelling of the decomposition of methane at 1273 K in a [41] Vlasov PA, Warnatz J. Detailed kinetic modeling of soot formation in hydro-
plug flow reactor at low conversion. J Anal Appl Pyrolysis 1993;25:395e405. carbon pyrolysis behind shock waves. Proc Combust Inst 2002;29:2335e41.
http://dx.doi.org/10.1016/0165-2370(93)80058-8. Proceedings of the 10th http://dx.doi.org/10.1016/S1540-7489(02)80284-X.
International Conference on Fundamental Aspects, Processes and Applications [42] Agafonov GL, Naydenova I, Vlasov PA, Warnatz J. Detailed kinetic modeling of
of Pyrolysis. soot formation in shock tube pyrolysis and oxidation of toluene and n-hep-
[34] Plevan M, Geißler T, Aba nades A, Mehravaran K, Rathnam RK, Rubbia C, et al. tane. Proc Combust Inst 2007;31:575e83. http://dx.doi.org/10.1016/
Thermal cracking of methane in a liquid metal bubble column reactor: ex- j.proci.2006.07.191.
periments and kinetic analysis. Int J Hydrogen Energy 2015;40:8020e33.