Académique Documents
Professionnel Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: This paper investigates the concept of using piston engines as chemical reactors to convert CO2 to more
Received 17 January 2019 useful substances. Conditions are presented under which significant conversions of CO2 and CH4 to syn-
Revised 6 March 2019
thesis gas (CO and H2 ) can be achieved in internal combustion engines (ICE). Numerical optimization of
Accepted 20 May 2019
the process based on detailed chemistry is employed to identify initial values of temperature, pressure
and gas composition which theoretically yield maximal CO2 conversions. The optimization results are
Keywords: used to guide experiments in a rapid compression machine (RCM) setup that is used in place of a real
CO2 reforming ICE. The optimization predicts that the addition of molecular oxygen enables the endothermal dry re-
Internal combustion engine forming reaction (CO2 + CH4 2 H2 + 2 CO) to proceed by delivering the required energy from burning
Rapid compression machine
a part of the fuel. Despite combustion taking place, experiments show that over 50% of the CO2 con-
Numerical optimization
tained in mixtures of Ar, O2 , CH4 , DME (dimethylether) and CO2 can be converted to syngas with H2 O as
Detailed chemical kinetics
a by-product.
© 2019 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
1. Introduction reforming, these systems have other disadvantages [3–5]. For ex-
ample, nickel catalysts are susceptible to coking, while cobalt cat-
The environmental effects of CO2 emissions are a major prob- alysts are reported to suffer from deactivation by oxidation [6,7].
lem of industrialized societies. Beside its appearance as a waste- An exploration of alternative methods for CO2 reforming there-
product of hydrocarbon combustion [1], CO2 also occurs as one fore appears worthwhile. The physico-chemical properties of the
of the main byproducts in chemical industry. CO2 reforming, or overall reaction (Eq. 1) impose important constraints for such a
dry reforming, has gained increased attention in recent years as it method:
opens a route to utilize the CO2 as a carbon source for more use-
ful products [2], and therefore represents a possibility to reduce
1. Hypothetically, if all kinetic constraints are removed, the system
CO2 emissions. Dry reforming is strongly endothermic, and the net
will instantaneously approach chemical equilibrium. The chem-
reaction can be formulated as
ical composition at this equilibrium, and therefore also the ex-
CO2 + CH4 2 H2 + 2 CO H = 247 kJ mol−1 . (1) tent of conversion, depends on the thermodynamic conditions.
For the endothermal reaction (Eq. (1)), chemical equilibrium
Besides offering the benefit of removing undesired CO2 , this re-
moves further to the right (i.e. to the side of CO and H2 )
action is also attractive for applications that require synthesis gas
with increasing temperature (at constant pressure). At constant
with low H2 /CO ratios.
temperature, equilibrium is shifted to the right when pressure
In the chemical industry, reforming processes are generally car-
is decreased. Therefore, based on equilibrium considerations,
ried out using catalysts that often contain noble metals such as
high-temperature and low-pressure conditions favor high CO2
rhodium and palladium. However, as in the case of dry reform-
conversions. Starting from unreacted “cold” low-pressure con-
ing, the use of precious metal catalysts is often cost prohibitive at
ditions, input of energy is hence required to shift equilibrium
commercial-scale. Although current research suggests that less ex-
to the desired state.
pensive catalysts based on nickel or cobalt could be used in dry
2. Some of the elementary reaction steps underlying Eq. (1) have
very high activation energies. Due to the Arrhenius-like de-
∗
Corresponding author. pendence of elementary reaction rate coefficients, high tem-
E-mail address: deutschmann@kit.edu (O. Deutschmann). peratures are required to allow those reactions to proceed at
URL: http://www.itcp.kit.edu/deutschmann (O. Deutschmann) sufficient speeds and approach equilibrium.
https://doi.org/10.1016/j.combustflame.2019.05.031
0010-2180/© 2019 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
H. Gossler, S. Drost and S. Porras et al. / Combustion and Flame 207 (2019) 186–195 187
Compression engines have the potential to fulfill the above- outline a concept in which engines form the key component of
mentioned requirements, and also feature some more properties compact gas-to-liquids plants. The concept allows the plant to be
which make them appropriate candidates as reactors: operated at a significantly smaller scale in a profitable way, due
to the lower costs associated with engines compared to conven-
• The required input of energy (item (1) above) can be provided
tional chemical reactors. Furthermore, a unique feature of their ex-
in the form of mechanical compression work. This work can
perimental work was the ability to preheat the inlet gas to over
be supplied within relatively short times (down to the scale
480° C.
of milliseconds). Within these time scales, heat loss processes
There are also numerous publications that studied the in-
can interfere strongly with chemical kinetics. The ability for fast
cylinder production of syngas via partial oxidation numeri-
repetition also means that high mass-flows of CO2 can be con-
cally, for example [16–18]. In their modeling study, Hegner and
verted.
Atakan [18] focused on thermodynamic aspects of a polygenera-
• The high temperatures required to overcome kinetic barriers
tion process to produce hydrogen. The authors claim that exergetic
(point (2) above) can be provided in engines via compression.
efficiencies of up to 80% are achievable compared to the separated
• In contrast to conventional large-scale chemical reactors, com-
production of power, heat and hydrogen. Their study was later ex-
pression engines can be operated in a very flexible manner, tol-
tended by accompanying experiments in a rapid compression ma-
erating quite large variations of inputs (mass flow rates, input
chine [14]. Here, dimethyl ether (DME) was used as a fuel addi-
pressure and temperature).
tive to allow methane conversions at moderate, technically acces-
• Engines are also able to react promptly to changes in input con-
sible compression ratios and inlet temperatures, as was done in
ditions, and their operation can be adapted to optimal opera-
the present work. Apart from syngas, small amounts (2 mol %) of
tion.
ethylene were produced in the experiments.
Also, from an economic point of view, engines have the advan- In another modeling study, numerical optimization was applied
tage of being cheap as the technology is well developed and ro- to maximize the hydrogen yield at the end of the expansion stroke
bust. of an HCCI engine [17]. In the present study we expand on the ap-
Note that, the higher the conversion, i.e., the more the reaction proach in [17]. In contrast to previous studies that investigated the
(Eq. (1)) is moved to its right side, the more net input of energy use of engines as reformers to produce syngas, this work focuses
is required for the process. In energy conversion technology, situa- on converting an undesirable species (CO2 ). The production of syn-
tions where a temporal and/or local surplus of an otherwise “unus- gas, which goes along with the CO2 conversion, is regarded as a
able” form of energy (e.g., heat or work) is available for exactly this side effect (albeit a positive one) only.
purpose, occur quite frequently. The availability of energy from “re- The objective of this paper is to demonstrate the concept of us-
newable” energy sources such as solar and wind power are subject ing a piston engine process to convert CO2 to syngas by dry re-
to high fluctuations. This leads to gaps between energy demand forming. We employ a coupled simulation/experimental approach
and availability. When there is an abundance of (electric) energy, in which simulations, together with numerical optimization, are
the energy should ideally be stored in order to compensate times used to identify operation conditions that theoretically yield opti-
where the renewable energy sources are not able to meet the de- mal conversions. A rapid compression machine (RCM) is used as an
mand of energy. engine simulator to assess whether the predicted conversion levels
While the literature contains enormous amounts of studies con- can be realized in a process of semi-technical scale. Rapid com-
cerning ICEs operated under conventional (i.e. total-oxidation) con- pression machines are devices which are frequently employed to
ditions, substantially less research has been dedicated to their use investigate chemical reactions under engine-like conditions [19].
as chemical reactors. Recently, a review article has been published In the remainder of this paper we will first describe the mod-
that gives an overview of the past research on fuel reforming in elling approach, which involves simulation of an engine process
conjunction with internal combustion engines [8]. Most of the pre- with detailed chemistry. The procedure for using the simulation
vious work on in-cylinder fuel reforming has focused on burning to identify optimal engine operation points is then outlined. Next,
rich fuel-air mixtures to produce synthesis gas via partial oxida- the RCM experiment setup is described. The results presented af-
tion following the reaction terwards show that CO2 conversion levels of over 50 percent are
1 possible, both in theory and in the experiments, with suitably cho-
CH4 + O2 2 H2 + CO H = −35.7 kJ mol−1 . (2) sen initial composition of the mixture and initial temperatures and
2
pressures.
The in-cylinder production of syngas has been shown to work
in engine experiments by numerous groups, e.g. [9–15]. In these
studies, various ways to achieve stable operating conditions (i.e. 2. Methodology
low cycle-to-cycle variations) were investigated. Spark ignition (SI),
compression ignition (CI) and homogeneous charge compression In an engine, there are many operating parameters that can be
ignition (HCCI) were successfully applied to convert methane- varied independently. Examples include the engine speed (rpm),
based fuels to syngas. the inlet gas temperature (T0 ), the inlet pressure (p0 ) and the over-
Karim and Wierzba [11] operated an engine on rich mix- all gas composition (mole fractions, x). Finding operating points
tures of methane and oxygenated air, where the mixture was ig- that will lead to high CO2 conversions is difficult due to the
nited by injecting diesel fuel (CI mode). Yang et al. [12] were vast number of combinations of possible operating parameters
able to demonstrate syngas production in an HCCI process with and the typically strongly nonlinear dependence of the chemical
added CO. In a combined modeling and experimental study, Wie- conversion on these parameters. For this reason, conducting em-
mann et al. [15] investigated a single-cylinder engine operated pirical experiments on a trial-and-error basis is not a practical
on methane/air using spark ignition and homogeneous charge approach.
compression-ignition. In their experiments, n-heptane was added Instead, the approach taken in this paper is to employ numer-
to reduce auto-ignition temperatures and to stabilize engine oper- ical optimization in conjunction with detailed chemical kinetics
ation. simulations to predict promising operating points. The optimal op-
In a study by a group of researchers at MIT [13], the use of erating points proposed by the model are then verified in a rapid
an engine as a reformer is motivated by economical factors. They compression machine setup.
188 H. Gossler, S. Drost and S. Porras et al. / Combustion and Flame 207 (2019) 186–195
accounted for using the global heat transfer model according to q˙ (t ) = h(t ) · Apiston + Ahead + Aliner (t ) T (t ) , (8)
Woschni [20].
The modeling approach gives rise to the following system of where h is the heat transfer coefficient that is obtained by the
coupled differential equations that must be solved simultane- Woschni relation (in W/m2 /K, see below). Furthermore, A is the
ously: area of the particular engine part as denoted by the index. The
area of the liner is the cylindrical part of the combustion cham-
dni
= V · ω˙ i , (3) ber, and is therefore dependent on time (cf. Fig. 1). The tempera-
dt ture difference in the above equation applies to the gas and wall
temperatures, i.e.
dT n˙ i H i (T ) − n˙ i RT − pV˙ − q˙
=
i
. (4) T (t ) = Tgas (t ) − Twall . (9)
dt cV
The temperature Twall was taken to be equal to the inlet tempera-
The volume V enters the simulation as a temporal constraint; it
ture of the gas (462 K) for all simulations; in practice, typical wall
is computed from tabulated time/volume pairs that are obtained
temperature variations in engine-like devices [22] are small rela-
from the experimental setup:
tive to the gas temperature Tgas , so that they do not impede the
V = Vcyl (t ) . (5) realism of the model.
H. Gossler, S. Drost and S. Porras et al. / Combustion and Flame 207 (2019) 186–195 189
compression time plus a certain hold time. We formulate the opti- Table 2
Parameters of the RCM setup.
mization problem as
Bore B 82 mm
nCO2 (tf ) Stroke L 74 mm
max XCO2 = 1 − , (13)
x0 nCO2 (t0 ) Compression ratio rc 8 to 12
Displaced volume Vd 0.4 L
subject to the linear constraints Clearance volume Vc 6.6 cm3
Initial temperature T0 300 K to 462 K
x0,i = 1 , (14) Initial pressure p0 0.5 to 1.5 bar
i
The equilibria represent the states that would result in the RCM
9x0,DME − 1x0,CH4 = 0 , (15)
if the system would instantaneously react to the maximum possi-
ble extent, always adapting to transient thermodynamical condi-
tions (U(t) and V(t)), which constantly change during the process
x0,i ≥ 0 , (16) due to compression and heat losses.
The equilibrium for given values of Ufix and Vfix is given by the
state of maximum entropy at those conditions. While a direct so-
x0,Ar ≥ 0.02 . (17) lution of U, V equilibrium via entropy maximization is possible, it
is sometimes plagued by numerical problems, especially for rich
The linear constraints formulated above are to ensure that the
mixtures. On the other hand, for the related problem of finding the
sum of all mole fractions equals unity (Eq. (14)), that the fuel
equilibrium for given values of T, p (by minimizing Gibbs enthalpy),
is composed of 90% CH4 and 10% DME by volume (Eq. 15), and
a highly robust and efficient algorithm is known [39]. Therefore,
that each component can only take on positive values (Eq. (16)).
we used this as a numerical tool to obtain U, V-equilibria by an
Furthermore, Eq. (17) constrains the amount of argon to above
iterative procedure. The first step involves expressing the equilib-
2%. This last constraint is in place to reflect requirements of the
rium moles (ni ) for isothermal-isobaric conditions as a function of
accompanying experiments, where a certain amount of helium
T and p:
is required as a reference for the gas analysis instrument (see
Section 3 for more details). For simplicity, the optimization does ni = n˜ i,eq (T , p) (i = 1, . . . , NS ) (19)
not additionally take helium into account. Instead, a part of the ar-
To obtain the equilibrium for given U and V rather than for given T
gon is substituted by helium in the experiments.
and p, the nonlinear system of two simultaneous equations in the
The optimization problem is solved using the derivative-based
two unknowns T, p
SQP algorithm from the MATLAB Optimization Toolbox (see Chap-
ter 18 in [35] for a description of the algorithm). The actual simula-
NS
tion code is part of the DETCHEM package [36] and is written in C, U i (T ) · n˜ i,eq (T , p) − Ufix = 0 , (20)
C++ and Fortran. The code computes the objective function value i=1
0.45 45 40
CO2 conversion / %
0.4 40 35
0.35 35
30
0.3 30
Pressure / bar
25
Volume / L
0.25 25
20 RCM
0.2 RCM 20
RCEM
0.15 15 15
RCEM 0.6 0.8 1 1.2 1.4 1.6
0.1 10 Initial pressure / bar
RCEM
0.05 RCM 5 Fig. 3. Comparison of CO2 conversions for RCM and RCEM operation modes at
T0 = 462 K. With increasing initial pressures, ignition occurs and the CO2 conversion
0 0
0 50 100 150 200 tends towards the same value (32%) for both modes of operation. Mole fractions of
Time / ms the gas before compression: 8% CO2 , 14.8% CH4 , 1.6% DME, 15.1% O2 , 55.5% Ar, 5%
He.
Fig. 2. Exemplary volume and pressure profiles of the experimental setup using an
ignitable CH4 /O2 /Ar mixture, highlighting the differences between RCM and RCEM
modes of operation. At t = 50 ms top-dead center is reached, which corresponds to sion. Therefore, no pressure decrease after TDC (which would be
an engine running at a speed of 600 rpm. common in a typical RCM experiment when ignition delay times
are measured) is visible.
After compression (from t = 50 ms), the difference between
measures the accumulated partial pressures. The volume of the RCM and RCEM modes becomes evident. In RCM mode, isochoric
mixing vessel and its total interior pressure are both one order of conditions prevail, with the volume remaining at the constant min-
magnitude larger than in the combustion chamber, resulting in the imum value of 0.05 L. The pressure decay in this region is solely
gas mass inside the vessel being two orders of magnitude larger. due to heat loss through the cylinder walls. In RCEM mode, the
This allows performing multiple RCM shots with the same initial expansion phase immediately follows the compression phase (hold
mixture. time tH = 0), like in a piston engine. The volume increases again af-
The machine is equipped with a creviced piston that swallows ter reaching its minimum value at TDC, resulting in a much faster
the roll-up vortex, helping to grant homogeneous thermal condi- pressure decay than in the RCM case. However, the piston only par-
tions in the compressed gas [42–44]. Furthermore, the combustion tially retracts due to geometrical constraints of the setup. Conse-
chamber is surrounded by a thermostatically controlled oil bath to quently, the volume does not reach its initial value of 0.45 L. Fur-
keep the temperature at the desired level. thermore, it can be seen that the piston slightly oscillates during
A knee lever system pushes the piston towards the right into the expansion phase.
the cylinder, thus compressing the gas in the combustion cham- After an experiment, the piston is retracted and it is waited un-
ber. Maximum compression is reached when the knee lever is fully til the temperature has fallen to near its pre-compression value.
stretched and aligned with the piston rod and the piston itself (top Then, samples from the in-cylinder gas mixture are extracted and
dead center, TDC). A bumper prevents the knee lever from moving analyzed in a gas chromatographic (GC) analyzer (Agilent 490 Mi-
past the stretched position further downwards in order to keep the cro GC). Using helium as an inert reference and argon as a carrier
piston from retracting back. In this way, isochoric conditions are gas, this GC allows to measure multiple species, including CO2 and
met after the compression phase. We refer to this mode of opera- lower hydrocarbons.
tion as “RCM mode”. Experiments were carried out using the gas mixtures of CH4 ,
Alternatively, the bumper can be removed from the setup after CO2 , O2 , DME and Ar predicted by the optimization. For simplic-
some adjustable hold time (tH = 0, . . . , ∞), and the knee lever can ity, helium was not included in the optimization. Therefore, 2 to
then move downwards past the stretched position immediately af- 5% of the argon was substituted by helium in the experiments to
ter the compression phase, which in effect retracts the piston after be able to quantify the mixture’s exact composition from the chro-
the hold time. We refer to this operation mode as “RCEM mode”, matograms. The CO2 conversion was then calculated according to
short for rapid compression expansion machine mode. This mode, Eq. (12).
with compression and subsequent expansion, is similar to a recip-
rocating internal combustion engine; various realizations of such 4. Results
rapid compression-expansion machines have been presented, and
have been used as a simplified, well-controlled replacement of pis- 4.1. RCM and RCEM operation modes
ton engines [45].
Figure 2 shows typical volume and pressure profiles of the If operated with short hold times, RCEM mode more closely
setup using an ignitable CH4 /O2 /Ar mixture, highlighting the dif- resembles conditions in a real engine than RCM mode. However,
ferences between RCM and RCEM operation modes. The solid lines running experiments in RCEM mode over prolonged time periods
denote the volume (left axis) and the dashed lines denote the pres- causes excessive wear to the setup. Therefore, RCM mode is pre-
sure (right axis). The blue lines are the profiles recorded in RCM ferred in order to avoid damage to the setup. This leads to the
mode (i.e. with the bumper in place), while the red lines were question whether results obtained in RCM mode can be meaning-
recorded in RCEM mode. The traces start at t = 0, where the cylin- fully extrapolated to real engine conditions.
der is filled with the unreacted mixture. In the compression phase Figure 3 shows a comparison between both operation modes. In
(until t = 50 ms), the volume and pressure curves of both modes these measurements, the gas mixture (specified in the figure cap-
coincide. As the volume decreases from its initial value of 0.45 L to tion) was compressed at constant initial temperature (T0 = 462 K)
0.05 L, the pressure rises from 1.0 bar to about 20 bar as per near- at varying initial pressures. Afterwards, the gas was analyzed and
adiabatic compression. In these examples, ignition occurs in both the CO2 conversions were obtained. The measurements were car-
operation modes near TDC (at 50 ms), detectable by a sharp pres- ried out in RCM mode (red crosses) and in RCEM mode (blue
sure rise up to 40 bar. Here, the ignition directly follows compres- circles).
192 H. Gossler, S. Drost and S. Porras et al. / Combustion and Flame 207 (2019) 186–195
10 15 20 25 30 35 40 45 50 CO2 conversion of 51.0% is shown in the top graph (blue line). Only
55
Experiment CH4 , DME and O2 enter the formula to calculate φ ; The oxygen
50
CO2 conversion / %
Optimal
required here to attain similarly high CO2 conversions, we would
20 2.4 not have been able to quantify the results.
15 2.3
4.3. Evolution of states during the RCM process
10 2.2
0 50 100 150 200 250 300 librium. In contrast, the detailed kinetics simulations predict that
45
during cool-down CO2 stays near its low value of the “hot” phase,
40 CH (kin.)
4
where both equilibrium and detailed simulations coincide and pre-
CO (equil.) H (kin.) CO (kin.) dict xCO2 ≈ 5%.
35 2
The observed high CO2 conversion levels exceeding 50% in both
H (equil.)
Mole fractions / %
30 2
the detailed chemistry simulations and the experiment therefore
25 CO (equil.)
reflect the tendency to preserve the chemical composition of equi-
H O (kin.)
libria (that are reached in the “hot” phase of the process) through-
20 CO2 (kin.) 2
H O (equil.)
2
out the expansion. The lower temperatures that result during
CO (equil.)
15 2 expansion quench CO2 producing reactions, so that the low CO2
levels from earlier “hot” equilibria are preserved.
10
Interestingly, heat loss alone leads to a sufficiently fast tem-
CO2 (kin.)
5 CO2 (equil.)
perature decay, allowing this effect to be observed in the RCM,
CH (equil.)
4
even in absence of the expansion stroke. The expansion stroke lets
the mixture attain low temperatures much faster than heat loss
through the cylinder walls alone and therefore may present an op-
Temperature / K
2000 T (equil.)
tion to further improve CO2 conversions. However, when the RCM
1500 is operated in this way (RCEM mode), ignition timing becomes a
key issue (see also Fig. 3).
T (equil.)
1000 Figure 6 shows that CO2 conversion occurs on a timescale of
T (kin.) a few milliseconds. Within this timescale, a mixture with initially
T (kin.)
500 uniform temperature distribution will develop a non-uniform dis-
0 50 100 150 200 250 300 tribution due to heat transfer from the gas to the cylinder walls.
Time / ms Despite the fact that the simulation model does not take into ac-
Fig. 5. Temporal evolution of selected species (top) and temperature (bottom) dur- count effects of thermal stratification, the experiment and simula-
ing a RCM process, showing detailed kinetics (solid lines, “kin.”) and equilibrium tion show a good agreement in achieved CO2 conversion.
(dashed lines, “equil.”) simulations. Initial conditions correspond to the optimal val- At a first glance, this agreement would be surprising. There is
ues for CO2 initial mole fraction of 20%: T0 = 462 K, p0 = 1.0 bar, 20% CO2 , 39.15% a plausible explanation, however: We propose that for the studied
CH4 , 4.35% DME, 33.0% O2 , 3.6% Ar.
conditions, the detailed dynamics of combustion and heat release
during the in-cylinder process are not crucial for the final amount
0 20 40 60 80 80.5 81 81.5 82 100 150 200 of CO2 conversion. The important requirements for the conversion
45
are that
40 CH 4 (kin.)
CO (equil.) • the gas reaches a sufficiently high maximum temperature to be
35 H
2 able to overcome kinetic barriers and approach chemical equi-
Mole fractions / %
30 librium
CO (equil.) • the cool-down phase is fast enough to let the chemical state
25
freeze at a point of some high-temperature equilibrium
20 CO (kin.)
2
H O
2
These requirements can be achieved in different regions of the
15 CO2 (equil.) in-cylinder mixture (e.g., central or closer to walls), even if they
10
are affected by heat transfer to a different extent, and with differ-
CO2 (kin.) ent timing during the cycle. For instance, even near-wall regions,
5 CO (equil.)
2
CH (equil.)
which during an engine cycle first would remain relatively cold
4
by heat transfer, can later attain the high temperatures required
for efficient CO2 conversion by thermal contact to burned regions.
Temperature / K
2000 T (equil.)
They can auto-ignite by mixing with hot burned gas, and then fur-
ther increase their temperature by chemical reaction.
1500
5. Conclusions
1000
T (kin.) Our work shows that CO2 conversion by dry reforming yielding
500 CO and H2 can be carried out in piston engines. As a by-product,
0 20 40 60 80 80.5 81 81.5 82 100 150 200 water is produced. From this viewpoint, the engine can be seen
Time / ms
as a multi-purpose chemical converter, which is both capable of
Fig. 6. Temporal evolution of selected species (top) and temperature (bottom) dur- producing power (in conventional combustion mode), but also able
ing a RCM process, showing detailed kinetics (solid lines, “kin.”) and equilibrium to deliver useful products (syngas) from CO2 .
(dashed lines, “equil.”) simulations. Same data as in Fig. 5, but with differently A combination of numerical simulation and optimization was
scaled time axis, helping to better discern the different phases of the RCM process.
used to identify operating conditions that yield high conversions,
Center section: “hot” phase. Right section: “cool-down” phase.
exceeding 50%; experiments in a rapid compression machine show
that these predictions are realistic, both with respect to the opti-
of zero up to xCO2 ,eq = 7%. This is in accordance with the princi- mal conditions and with respect to the achieved conversions. The
ple of Le Chatelier and Braun for the endothermal CO2 reforming optimization approach was not only helpful to reduce the number
reaction CO2 + CH4 2 CO + 2 H2 . Also, the equilibrium mole of experiments, but also to determine whether CO2 conversion in
fraction of CO2 rises during the cool-down phase to approx. 16%, an engine is possible in the first place.
which is close to its pre-compression value. Hence, no notable One result of the optimization is that addition of oxygen to the
CO2 conversion would result if the system always followed equi- initial mixture of CH4 and CO2 helps improve the conversion by
194 H. Gossler, S. Drost and S. Porras et al. / Combustion and Flame 207 (2019) 186–195
[24] H. Wang, X. You, A.V. Joshi, S.G. Davis, A. Laskin, F. Egolfopoulos, C.K. Law, USC [35] J. Nocedal, S. Wright, Numerical optimization, Springer Series in Operations
Mech Version II. High-temperature combustion reaction model of H2 /CO/C1-C4 Research and Financial Engineering, 2, Springer-Verlag, New York, 2006.
compounds, 2007, http://ignis.usc.edu/USC_Mech_II.htm. [36] O. Deutschmann, S. Tischer, C. Correa, D. Chatterjee, S. Kleditzsch, V.M. Janard-
[25] H.J. Curran, S.L. Fischer, F.L. Dryer, The reaction kinetics of dimethyl ether. II: hanan, N. Mladenov, H.D. Minh, H. Karadeniz, M. Hettel, H. Gossler, DETCHEM
Low-temperature oxidation in flow reactors, Int. J. Chem. Kinet. 32 (12) (20 0 0) Software Package v2.7 (www.detchem.com), 2018,
741–759, doi:10.1002/1097- 4601(2000)32:12<741::AID- KIN2>3.0.CO;2- 9. [37] R.J. Hogan, Fast reverse-mode automatic differentiation using expression tem-
[26] C.K. Westbrook, M. Mehl, W.J. Pitz, G. Kukkadapu, S. Wagnon, K. Zhang, Multi- plates in C++, ACM Trans. Math. Softw. 40 (4) (2014) 26:1–26:16, doi:10.1145/
fuel surrogate chemical kinetic mechanisms for real world applications, Phys. 2560359.
Chem. Chem. Phys. 20 (16) (2018) 10588–10606, doi:10.1039/C7CP07901J. [38] A.C. Hindmarsh, P.N. Brown, K.E. Grant, S.L. Lee, R. Serban, D.E. Shumaker,
[27] B. Chan, J.M. Simmie, Barriometry – an enhanced database of accurate bar- C.S. Woodward, SUNDIALS: Suite of nonlinear and differential/algebraic equa-
rier heights for gas-phase reactions, Phys. Chem. Chem. Phys. 20 (16) (2018) tion solvers, ACM Trans. Math. Softw. 31 (3) (2005) 363–396, doi:10.1145/
10732–10740, doi:10.1039/C7CP08045J. 1089014.1089020.
[28] D.L. Baulch, C.T. Bowman, C.J. Cobos, R.A. Cox, T. Just, J.A. Kerr, M.J. Pilling, [39] S.B. Pope, Gibbs function continuation for the stable computation of chemical
D. Stocker, J. Troe, W. Tsang, R.W. Walker, J. Warnatz, Evaluated kinetic data equilibrium, Combust. Flame 139 (3) (2004) 222–226.
for combustion modeling: Supplement II, J. Phys. Chem. Ref. Data 34 (3) (2005) [40] P. Deuflhard, Newton methods for nonlinear problems: Affine invariance and
757–1397, doi:10.1063/1.1748524. adaptive algorithms, Springer, 2011, doi:10.1007/978- 3- 642- 23899- 4.
[29] B.S. Haynes, H.G. Wagner, Soot formation, Prog. Energy Combust. Sci. 7 (4) [41] M. Werler, L. Cancino, R. Schießl, U. Maas, C. Schulz, M. Fikri, Ignition delay
(1981) 229–273, doi:10.1016/0360-1285(81)90 0 01-0. times of diethyl ether measured in a high-pressure shock tube and a rapid
[30] Y. Hidaka, K. Sato, Y. Henmi, H. Tanaka, K. Inami, Shock-tube and modeling compression machine, Proc. Combust. Inst. 35 (1) (2015) 259–266, doi:10.1016/
study of methane pyrolysis and oxidation, Combust. Flame 118 (3) (1999) 340– J.PROCI.2014.06.143.
358, doi:10.1016/S0010-2180(99)00010-3. [42] C.-j. Sung, H.J. Curran, Using rapid compression machines for chemical kinetics
[31] C. Heghes, C1 -C4 hydrocarbon oxidation mechanism, University of Heidelberg, studies, Prog. Energy Combust. Sci. 44 (2014) 1–18, doi:10.1016/j.pecs.2014.04.
Germany, 2006 (Ph.D. thesis). 001.
[32] Z. Zhao, M. Chaos, A. Kazakov, F.L. Dryer, Thermal decomposition reaction and [43] D. Lee, S. Hochgreb, Rapid compression machines: Heat transfer and suppres-
a comprehensive kinetic model of dimethyl ether, Int. J. Chem. Kinet. 40 (1) sion of corner vortex, Combust. Flame 114 (3–4) (1998) 531–545, doi:10.1016/
(2008) 1–18, doi:10.1002/kin.20285. S0 010-2180(97)0 0327-1.
[33] S. Porras, R. Schießl, U. Maas, A chemical kinetic modeling study of ignition in [44] G. Mittal, M.P. Raju, C.-J. Sung, Vortex formation in a rapid compression ma-
fuel-rich methane/dimethyl ether mixtures for polygeneration processes, 8th chine: Influence of physical and operating parameters, Fuel 94 (2012) 409–417,
European Combustion Meeting 2017 (2017). Dubrovnik, Croatia doi:10.1016/J.FUEL.2011.08.034.
[34] D. Kaczmarek, S. Porras, F. Sen, T. Kasper, R. Schießl, U. Maas, B. Atakan, [45] S.S. Goldsborough, C.J. Potokar, The Influence of Crevice Flows and Blow-By
Kinetische Untersuchung der partiellen Oxidation von Methan/DME- und on the Charge Motion and Temperature Profiles Within a Rapid Compression
Methan/Ethanol-Gemischen, 28. Deutscher Flammentag (2017). Darmstadt, Expansion Machine Used for Chemical Kinetic (HCCI) Studies, SAE Technical
Germany Paper Series (2007-01-0169) (2007) 776–790, doi:10.4271/2007- 01- 0169.