c 2014. This manuscript version is made available under the CC-BY-NC-ND 4.0 license.

Accepted for Publication in Combustion and Flame, August 2014

Published article: http://dx.doi.org/10.1016/j.combustflame.2014.08.019

Chemical Kinetics Mechanism for Oxy-Fuel Combustion of Mixtures of

Hydrogen Sulfide and Methane
Dominik Bongartz, Ahmed F. Ghoniem∗
Department of Mechanical Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA 02139, USA

Abstract
Oxy-fuel combustion of sour gas, a mixture of natural gas (essentially methane (CH4 )), carbon dioxide (CO2 ), and
hydrogen sulfide (H2 S), could enable the utilization of large natural gas resources, especially when combined with
enhanced oil recovery. In this work, a detailed chemical reaction mechanism for oxy-fuel combustion of sour gas
is presented. To construct the mechanism, a CH4 sub-mechanism was chosen based on a comparative validation
study for oxy-fuel combustion. This mechanism was combined with a mechanism for H2 S oxidation, and the sulfur
sub-mechanism was then optimized to give better agreement with relevant experiments. The optimization targets
included predictions for the laminar burning velocity, ignition delay time, and pyrolysis of H2 S, and H2 S oxidation
in a flow reactor. The rate parameters of 15 sulfur reactions were varied in the optimization within their respective
uncertainties. The optimized combined mechanism was validated against a larger set of experimental data over a
wide range of conditions for oxidation of H2 S and interactions between carbon and sulfur species. Improved overall
agreement was achieved through the optimization and all important trends were captured in the modeling results. The
optimized mechanism can be used to make qualitative and some quantitative predictions on the combustion behavior
of sour gas. The remaining discrepancies highlight the current uncertainties in sulfur chemistry and underline the need
for more accurate direct determination of several important rate constants as well as more validation data.
Keywords: Sour gas, Sulfur, Hydrogen sulfide, Methane, Oxy-fuel, Kinetic modeling

1. Introduction

A large fraction of the world wide natural gas resources is currently not usable because it consists of ’sour’ natural
gas, or sour gas [1]. Unlike regular ’sweet’ natural gas, which is made up mainly of methane (CH4 ) and some
other light hydrocarbons, sour gas also contains large amounts of hydrogen sulfide (H2 S) and carbon dioxide (CO2 ).
Depending on the gas field, the volume fractions of these contaminants often vary between zero and 30% each [2].
Both substances are undesirable; H2 S is toxic and its combustion products (especially sulfur trioxide (SO3 )) can cause
severe corrosion problems, while CO2 lowers the heating value of the gas. If the concentrations of H2 S and CO2 get
too high, expensive purification processes are necessary if the natural gas is to be used in a conventional gas fired
power plant, which can make the use of the gas uneconomic.
Oxy-fuel combustion may provide a solution to this problem. In oxy-fuel combustion, the fuel is burned in pure
oxygen (O2 ) and some diluent (usually either CO2 , water (H2 O), or a mixture of these) instead of air. This yields
combustion products consisting mainly of CO2 and H2 O, which allows for easy separation of CO2 by condensation
of H2 O and makes oxy-fuel combustion an option for carbon capture and sequestration [3, 4]. Since the production
of O2 is energy intensive, oxy-fuel systems are likely to operate close to stoichiometry (equivalence ratio Φ ≈ 1) to

∗ Correspondingauthor. 77 Massachusetts Avenue, Room 3-344, Cambridge MA 02139, USA. Fax: +1-617-253-5981.
Email addresses: [email protected] (Dominik Bongartz), [email protected] (Ahmed F. Ghoniem)
avoid wasting either fuel or energy for air separation. Thus, the combustion temperature is controlled by adjusting the
amount of diluent that is added to the combustor.
In the case of sour gas, the problem of the low heating value can therefore be overcome when employing an
oxy-fired combustion strategy by recycling less diluent to the combustor. Furthermore, the fact that oxy-fuel systems
operate close to stoichiometry might alleviate corrosion issues associated with sulfur-containing combustion products.
If the carbon dioxide produced in the oxy-fuel process is used for enhanced oil recovery (EOR), an additional revenue
stream can be generated to counteract the higher cost and complexity of the process [5]. However, the use in EOR
also gives rise to different requirements for the combustion products, one important change being that the O2 content
has to be limited to ppm levels [6].
To investigate the potential of this approach, it is necessary to characterize the combustion behavior of sour gas
and in particular its chemical kinetics. We therefore need a detailed chemistry model that can capture most of the
phenomena that arise from the unusual fuel and combustion environment. Eventually, such a model can also serve
as a basis for the development of a reduced mechanism suitable for use in multi-dimensional computational fluid
dynamics (CFD) simulations. Specifically, a good model should fulfill the following requirements:
• Capture the effects of CO2 or H2 O diluted oxy-fuel conditions on the laminar burning velocity and the ignition
delay time of CH4 over a range of temperatures.
• Predict the formation and consumption of carbon monoxide (CO) and O2 under oxy-fuel conditions with rea-
sonable accuracy.
• Give accurate predictions for the laminar burning velocity and ignition delay time of H2 S over a range of
temperatures and for different diluents.
• Predict the formation of SO3 under the conditions of interest.
• Capture possible interactions between carbon and sulfur species and reproduce the correct relative oxidation
speed of CH4 and H2 S.
The purpose of this paper is to assemble and optimize a detailed chemical reaction mechanism for sour gas com-
bustion that fulfills these requirements. We first present a comparative study of mechanisms for oxy-fuel combustion
of CH4 to identify a suitable CH4 sub-mechanism. Next, we describe how this CH4 mechanism was combined with a
mechanism for the oxidation of H2 S and explain our optimization of the sulfur kinetics. Finally, we give a summary of
the performance of the combined mechanism against experimental data (including but not limited to the optimization
targets) on sulfur oxidation and interactions between CH4 and H2 S.

2. Mechanism for Oxy-Fuel Combustion of CH4

While there has been a considerable amount of work on modeling combustion of CH4 in air, so far only few
reaction mechanisms have been developed or at least tested specifically for the special conditions that are typical in
oxy-fuel combustion (that is combustion close to stoichiometry in the presence of large fractions of CO2 or H2 O).
This group of mechanisms includes the GRI-Mech 3.0 [7], the mechanism developed by Glarborg et al. [8–10], and
the mechanism developed by Dagaut et al. [11, 12]. However, to our knowledge there has been no comprehensive
comparative study yet to show how well these mechanisms capture the effects caused by the unusual combustion
environment.
In our case, the O-H subset of the mechanism is particularly important since it will also be used by the sulfur
part of the mechanism (see Section 3). Therefore, we also included the AramcoMech 1.3 mechanism by Metcalfe
et al. [13] in our study, since it is based on a recent O-H model [14] which (unlike the older mechanisms mentioned
above) inludes the latest advances in this area, such as, for example, the improved rate constant for the reaction
H + O2 = OH + O [15]. It should be noted that AramcoMech 1.3 does not contain a model for formation of nitrogen
oxides (NOx). However this is not expected to be a major problem when dealing with oxy-fuel combustion, since a
number of studies have found NOx formation to be negligible for realistic O2 purities [16–18].
Several studies have shown that adding CO2 to CH4 -air flames reduces the laminar burning velocity of CH4
through both thermal and chemical effects [19–21]. The extent of this reduction is captured correctly by all three
2
tested mechanisms (see Fig. 1). For the mechanism by Dagaut et al., we encountered numerical difficulties for all
burning velocity calculations.
There have been a number of studies on the impact of H2 O on the burning velocity of CH4 [22–25]. While all three
tested mechanisms can predict the decrease of the burning velocity through H2 O addition equally well at atmospheric
pressure, AramcoMech 1.3 performs better at elevated pressure (see Fig. 2).
Levy et al. [26] measured the ignition delay time of CH4 in different diluents and found that neither CO2 nor
H2 O have a significant effect. All four mechanisms considered here agree reasonably well with their measurements
in mixtures containing CO2 (see Fig. 3a) and both CO2 and H2 O (see Fig. 3b).
Recently, the influence of CO2 and H2 O on CO oxidation has been investigated by several researchers (e.g., [12,
27]). Our tests showed that all four mechanisms are able to reproduce the inhibiting effect of large CO2 concentrations
on CO oxidation observed by Abián et al. [27] both for lean and rich conditions (see Fig. 4; only the case with the
highest CO2 content is shown here for better readability). The experimental data is best reproduced by GRI-Mech
3.0. The Glarborg mechanism predicts the onset of CO oxidation to be at slightly lower temperatures than what is
observed experimentally.
The enhancing effect of moderate concentrations of H2 O on CO oxidation in the presence of CO2 in the same
experiment is captured qualitatively by all mechanisms (see Fig. 5; again, only the case with the highest H2 O content
is shown). The inhibiting effect of larger concentrations of H2 O on CO oxidation at very lean conditions (Φ ≈ 0.04)
observed by Glarborg et al. [28] is also captured by all models (see Fig. 6). In both cases, the Glarborg mechanism
again predicts lower temperatures for the onset of CO oxidation.
While previous cases studied the impact of CO2 and H2 O on CO oxidation, all mechanisms can at least qualita-
tively predict the effect of CO2 on CO concentrations in CH4 oxidation in the experiment of Glarborg and Bentzen [8]
(see Fig. 7). At stoichiometric and fuel-rich conditions, the temperature above which increased CO is observed is best
predicted by AramcoMech 1.3 and the Dagaut mechanism.
In summary, AramcoMech 1.3 provides the best overall agreement with the experimental data in the presented
relevant validation cases, especially for the burning velocity at elevated pressure and CO formation in CH4 oxidation.
Based on this observation and the fact that it contains the most recent O-H submodel of the mechanisms considered
here, we selected it as the basis for our mechanism for sour gas combustion.

3. Mechanism for H2 S Oxidation

To construct a mechanism for sour gas combustion, we chose to combine AramcoMech 1.3 (including its O-H
submodel) with a suitable mechanism for H2 S oxidation. However, testing this basic combined reaction mechanism
revealed discrepancies between experiment and modeling for important combustion parameters of H2 S such as laminar
burning velocity and ignition delay time. This motivated the tuning of the sulfur mechanism to improve the agreement
with experiments under relevant conditions. In this section, we first describe the selection of the H2 S mechanism and
the optimization procedure. Next, we present the performance of the mechanism for the optimization targets as well
as other experiments.

3.1. Mechanism Selection and Optimization

Sulfur oxidation chemistry is far less well established than hydrocarbon chemistry, but there has been a growing
interest in the topic in recent years and a number of detailed reaction mechanisms have been proposed, including
the work of Frenklach et al. [29]; Glarborg et al. [30]; Alzueta et al. [31]; the Leeds sulfur mechanism [32], and the
mechanism presented by Cerru et al. [33, 34]. Recently, Zhou et al. [35] presented a mechanism for H2 S oxidation
that puts more emphasis on the chemistry of disulfur species. Since these were found to be particularly important near
stoichiometric conditions, this mechanism was chosen as the basic sulfur mechanism.
Only the reactions containing sulfur-species were adopted from the mechanism of Zhou et al., since the O-H
reactions are already contained in AramcoMech 1.3. These reactions that determine the O-H radical pool are important
for both CH4 and H2 S oxidation and hence provide a means of interaction between the two fuels. The mechanism
does not include any species containing both carbon and sulfur such as carbonyl sulfide (COS) or carbon disulfide
(CS2 ), the influence of which Cerru et al. [33] found to be negligible. The resulting combined mechanism consists of
157 species and 1011 reactions.

3
 The thermodynamic data was taken from AramcoMech 1.3 and Zhou et al. [35], except for the enthalpy of forma-
tion of the mercapto radical (SH), which was set to ∆H0f,298 (SH) = 33.8 kcal/mol. This is in agreement with the recent
recommendations of Goos et al. [36] and Denis [37] while improving overall agreement with the experimental data
used for validation. Because of this change, we also had to adjust the rate parameters for the high-pressure limit of
the reaction SH + SH (+M) = HSSH (+M) since they were originally determined from the backward reaction using
the equilibrium constant, which is affected by the thermodynamic data [38]. The new rate constant is only about half
of the original one and has a higher activation energy.
As mentioned above, there where some important experiments for which significant discrepancies were observed
between the experimental data and the modeling results with this initial combined mechanism. To improve the overall
performance of the mechanism, the following cases were chosen as targets in a systematic optimization of the sulfur
kinetics:

• the laminar burning velocity of H2 S in air at atmospheric pressure [39–42]

• the ignition delay time of H2 S in air as measured by Frenklach et al. [29]
• the pyrolysis of H2 S in the experiment of Hawboldt et al. [43]

• the H2 S flow reactor experiment of Zhou et al. [35]

The exact points used, the corresponding target values, and the assigned weights are listed in Table 1. For the laminar
burning velocity (target 1), a target range was chosen to account for the scatter of the data points from the different
experiments. For the ignition delay time (targets 2–5), target ranges were chosen because the selection of a single
data point led to arbitrarily different results. For the flow reactor experiment of Zhou et al. [35], we only consid-
ered the cases with an equivalence ratio of Φ = 0.8 (targets 8–12) because they are closer to the conditions around
stoichiometry that occur in oxy-fuel combustion than the third case presented therein (Φ = 0.15).
The objective function f used in the optimization was defined as the sum of the quadratic deviations of the
modeling predictions (xi ) from the target values (Xi,min , Xi,max ) normalized by a characteristic value (Xi,char ), multiplied
by the respective weights (wi ):
h i
4 sgn(xi − Xi,min ) + sgn(xi − Xi,max ) · min (xi − Xi,min )2 , (xi − Xi,max )2
1  2
X
f = wi · 2
(1)
i
Xi,char

In most cases, Xi,char was set to be the target value (or the mean of the upper and lower limit of the target range).
However, for targets 10–12 a common value of 300 ppm was chosen in order to avoid a division by zero for target
12. In order to achieve good overall agreement, target 12 was therefore assigned an empirically determined increased
weight. Increased weight was also assigned to the laminar burning velocity (target 1) due to its importance in charac-
terizing the flame stabilization behavior of a fuel, also keeping in mind the intended future use of a reduced version of
the mechanism for CFD of turbulent flames. The sum in Eq. (1) did not necessarily include all targets listed in Table 1,
but different subsets were used to obtained different optimized versions of the mechanism by only considering certain
combinations of experiments. In the comprehensive model validation presented in the next section, three of these
versions were investigated. In the following, they are referred to as ’Optimization 1–12’, ’Optimization 1–7’, and
’Optimization 8–12’, where the numbers are the target numbers given in Table 1. These combinations correspond to
an optimization with respect to all experiments mentioned above, all experiments except the H2 S flow reactor of Zhou
et al. [35], and only the H2 S flow reactor, respectively.
The decision variables used in the optimization were the pre-exponential factors for the rate constants of 15
reactions in the sulfur mechanism to which the results for the target cases were particularly sensitive. To identify
these reactions, we conducted sensitivity analyses for targets 1–7. Sensitivity analyses for the flow reactor experiment
(targets 8–12) were already conducted by Zhou et al. [35]. The results of some of these analyses are summarized
in Fig. 8. The reactions highlighted in boldface (also listed in Table 2) were chosen to be varied in the systematic
optimization of the mechanism. Note that these also include the ten reactions Zhou et al. [35] found to be the most
important for their flow reactor experiment. We confirmed this in a similar analysis for the present combined model.
For the pressure dependent reactions SH + SH (+M) = HSSH (+M) and H + SO2 (+M) = HOSO (+M), the rate

4
constants for both high and low pressure limit were considered. For the reaction H2 S + S (+M) = HSSH (+M), only
the high pressure expression was included as a decision variable, since even at atmospheric pressure the reaction is
almost in the high-pressure limit [38]. Thus, a total of 17 decision variables were used in the optimization. The large
number of reactions that had to be varied to improve the agreement with the different experiments simultaneously is
a first indication of the significant uncertainties remaining in sulfur chemistry. An earlier attempt considering only 10
of the present 15 reactions was significantly less successful.
The constraints on the variations of the decision variables were determined from the uncertainties reported for the
original experiments or calculations wherever possible, to make sure that the values for the rate parameters determined
through the optimization are still in agreement with the more direct experimental or theoretical determinations. For the
reaction S + O2 = SO + O, the uncertainty of the pre-exponential factor was determined as +50%/-20% by comparison
with the experimental data used in the original fitting by Lu et al. [44]. In a similar fashion, we determined the
constraints on the variation of the pre-exponential factor of the reaction H2 S + H = SH + H2 as ±25% by comparison
with the experimental data used by Peng et al. [45]. For the reaction SO + O2 = SO2 + O, we reevaluated the fit to the
experimental data of Grillo et al. [46], Black et al. [47], and Garland [48]. The uncertainty of the pre-exponential factor
was determined to be about 10% by combining the higher and lower limits for the different rate constants considered in
the fitting. For the reaction H2 S + S = SH + SH, two duplicate reactions corresponding to different reaction channels
were defined in the original sulfur mechanism [35]. We determined the uncertainties in the pre-exponential factors
of each of these by varying them and comparing the resulting fit to the original data of Gao et al. [49] and Shiina et
al. [50]. The limits on the variation of the pre-exponential factor that were chosen are +20%/-40% for the first of the
two duplicate reactions and ±25% for the second reaction. The rate expression for the reaction H2 S + O = SH + OH
used in the mechanism was determined from theoretical calculations, but the comparison with experimental data
presented by Wang et al. [51] suggests an uncertainty of a factor of two. The rates of the other reactions were obtained
from theoretical calculations alone so that an uncertainty of a factor of three was assumed [38, 52].
We implemented a simple coordinate descent method (see e.g. [53]) to solve the optimization problem described
above, i.e. only one decision variable was varied at a time with the others being constant. This one-dimensional
line search problem was solved with the golden section method. The calculations needed for evaluating the objective
function were conducted using CHEMKIN-PRO [54]. Once an optimal value was found, the next variable was varied.
The order in which the decision variables were optimized was chosen randomly at the beginning of each loop over
the 17 decision variables. This procedure was repeated until the relative change in the objective function between
two consecutive loops over all variables was negligible (≤ 10−5 ). Typically, between 25 and 35 iterations over all
decision variables were necessary to achieve a converged solution, taking between 10 and 50 hours of CPU time total,
depending on the number of targets used.
Optimization with respect to the three different combinations of targets (’Optimization 1–12’, ’Optimization 1–7’,
and ’Optimization 8–12’) resulted in different values for the rate parameters that were varied (see Table 2). A complete
version of the combined mechanism which was found to give the best overall agreement, ’Optimization 1–12’ (see
Section 3.2), is available as supplementary material.
For a number of reactions, the rates were changed in the opposite direction and the optimal values differ by almost
a factor of ten depending on the combination of targets used. In particular, the reactions

H2 S + S = SH + SH
H + HSS = SH + SH
SH + SH(+M) = HSSH(+M)
S + SH = S2 + H
H2 S + O = SH + OH
S + O2 = SO + O

show conflicting changes in the different optimized versions while having a strong influence on the results, and thus
require a more accurate direct determination of their rate constants to resolve the current uncertainties.

5
3.2. Mechanism Performance
To demonstrate the performance of the combined mechanism, we first present the results for the basic combined
mechanism and the three optimized combined mechanisms (’Optimization 1–12’, ’Optimization 1–7’, and ’Optimiza-
tion 8–12’) in the cases that served as optimization targets. Next, we summarize additional validation cases for sulfur
oxidation and interactions between carbon and sulfur species. Where possible, emphasis was put on situations that are
relevant to oxy-fuel combustion. All modeling calculations were again conducted using CHEMKIN-PRO [54]. For
cases in which the different optimized versions led to different results, the calculations with all four mechanisms are
shown. In all other cases, only the computation using the recommended version ’Optimization 1–12’ is presented.

3.2.1. Optimization Targets

To our knowledge, the laminar burning velocity of H2 S in air (target 1) has been measured in only four studies [39–
42] and there is considerable scatter in the data. However, at soichiometric conditions (which are most important for
the intended application) most experiments suggest a value around 40 cm/s (see Fig. 9), meaning that the prediction
of the basic combined mechanism is too high. We found that this discrepancy could be removed by considering the
laminar burning velocity as a target in the optimization.
Frenklach et al. [29] measured the ignition delay time of H2 S in air at pressures of p = 30-45 bar under fuel-lean
and fuel-rich conditions (targets 2–5). For our modeling, the ignition delay time was taken to be the time needed for
a temperature rise of ∆T = 600 K similar to [34], but the results were found to be not sensitive to the exact definition.
Figure 10 shows that while the calculated order of magnitude is correct, the temperature dependence predicted by the
basic mechanism is somewhat too strong. Again, improved agreement was achieved in the optimization, especially if
targets 8–12 were not used.
Zhou et al. [35] also presented data on lean H2 S oxidation in an atmospheric flow reactor (targets 8–12). Figure 11
shows that while the trends of the prediction with the basic mechanism for the H2 S profiles are correct, the profiles are
shifted to lower temperatures, which is in agreement with the original findings of Zhou et al. This could be improved
by including the experiment as an optimization target. If the optimization was conducted only against the other 7
targets, the predicted conversion is much too low. The prediction quality for sulfur dioxide (SO2 ) is comparable to the
H2 S profiles (see Fig. 12). Hydrogen is overpredicted (see Fig. 13), in agreement with the results of Zhou et al.
The pyrolysis of H2 S has been the subject of a number of experimental (e.g., [43, 55]) and modeling (e.g., [50, 52])
studies. Compared to the experimental data of Hawboldt et al. [43] (targets 6–7), H2 S conversion predicted by the
basic mechanism tends to be too slow for most temperatures (see Fig. 14). The optimization was able to improve this
if the experiment was used as a target, but even optimization only against the flow reactor of Zhou et al. [35] led to
some improvements.

3.2.2. Validation: H2 S Oxidation

Very recently, Mathieu et al. [56] measured the ignition delay time of H2 /H2 S mixtures. The basic combined mech-
anism predicts a shorter ignition delay than measured at high temperature, while at p = 33 atm and low temperature
there is a change in the slope of the Arrhenius plot that is not observed experimentally (see Fig. 15). The agreement is
improved only for the version of the mechanism (Optimization 1–7) that was optimized against all targets except for
the flow reactor of Zhou et al. This is consistent with the observation in the experiment of Frenklach et al. [29] (cf.
Fig. 10) and highlights the difficulty of reconciling the flow reactor data and the ignition delay time. This was also
observed by Mathieu et al. [56], who were able to manually tune the same basic mechanism (the sulfur mechanism of
Zhou et al. [35] and the H2 mechanism of Kéromnès et al. [14]) to give even better agreement for their own data than
the present model, but at the expense of deteriorated predictions for the flow reactor [35] and also for H2 S pyrolysis
in the experiment of Hawboldt et al. [43]. Including the experiment of Mathieu et al. [56] as an optimization target
did not improve the overall agreement or clarify the conflicting trends in the reaction rates (cf. Table 2).
Merryman and Levy [57] measured the flame structures of premixed H2 S flames at low pressure. Figure 16 shows
that the prediction of the major species is good while the hydrogen (H2 ) concentration is somewhat overpredicted,
similar to the flow reactor results mentioned above. The predictions with the different optimized versions are virtually
identical.
While the primary combustion product of H2 S is SO2 , some SO3 can be formed during cooling of the flue gas
by further oxidation of SO2 in the presence of excess O2 . This is a major concern due to the role of SO3 in sulfuric

6
acid corrosion and low temperature hot corrosion [58–60]. The formation of SO3 has been predicted to be increased
in oxy-coal combustion compared to air combustion [61], mainly because SO2 is recycled with the flue gas, thus
leading to a higher sulfur concentration. However, this can be overcome by appropriate cycle design. Additional
improvements are expected for oxy-fuel combustion of gaseous fuels, since gas-fired plants can operate much closer
to stoichiometric conditions which reduces the excess O2 that is needed for SO3 formation.
Fleig et al. [62] recently investigated the formation of SO3 in a flow reactor over a wide range of conditions.
Figure 17 shows a comparison of the modeling results with some of their experimental data. It can be seen that the
mechanism can capture the qualitative changes with varying O2 and SO2 content, the influence of an environment with
high CO2 concentration, and the impact of H2 O. However, there remain some quantitative discrepancies, especially
at higher temperatures. This was also observed by Fleig et al. [62] using the kinetic model from [63]. We were not
able to remove this discrepancy by including these experiments as optimization targets.

3.2.3. Validation: Interactions Between CH4 and H2 S

One possible interaction between carbon and sulfur species is the inhibition of CO oxidation by SO2 [30, 31, 63,
64]. The combined mechanism can capture the inhibiting effect of SO2 on CO oxidation at stoichiometric conditions
observed by Giménez-López et al. [63] both in a nitrogen (N2 ) bath and in the presence of CO2 (see Fig. 18). Com-
parable agreement is achieved for the flow reactor experiment of Dagaut et al. [64]. The different optimized versions
lead to slight variations of the predicted profiles for the case with SO2 , but the overall quality of the prediction is not
altered significantly.
Chin et al. [65] measured the simultaneous oxidation of CH4 and H2 S in a flow reactor under fuel-rich conditions.
As can be seen from Fig. 19, the combined mechanism can predict the relative oxidation speed of the two fuels,
although there are some quantitative discrepancies especially for methane.
Fleig et al. [62] also investigated the influence of CO and CH4 on the oxidation of SO2 to SO3 . As shown in
Fig. 20, the model qualitatively predicts the increased SO3 formation at intermediate temperatures in the presence of
combustibles. The slightly varying extent of this effect when comparing different amounts of CO/CH4 is not captured
correctly.

4. Conclusion
A detailed chemical kinetic model has been developed for oxy-fuel combustion of sour gas by combining existing
mechanisms for CH4 and H2 S and optimizing the sulfur kinetics:
1. A comparative study of mechanisms for oxy-fuel combustion of CH4 showed that current mechanisms give
good agreement with the experiment for the impact of the typical oxy-fuel diluents on the laminar burning
velocity, ignition delay time, and CO concentrations.
2. AramcoMech 1.3 was chosen as the CH4 sub-mechanism for the combined model because of its good perfor-
mance at elevated pressure and its new O-H model.
3. The sulfur part of the mechanism was optimized by varying the rate parameters of 15 reactions within the
experimental or computational uncertainties. While it was possible to get good agreement for a combination of
the laminar burning velocity, ignition delay time, and pyrolysis data or for the H2 S flow reactor data [35], the
discrepancies were larger if a combination of those two groups was considered. For the recommended version
(’Optimization 1–12’), overall agreement for the oxidation of H2 S and formation of SO3 after optimization is
reasonable.
4. The mechanism can also predict phenomena related to the interaction between carbon and sulfur species. The
predicted relative oxidation speed of CH4 and H2 S is correct.
5. Some quantitative discrepancies remain that highlight the uncertainties that still exist in sulfur chemistry. In
particular, the rates of several important reactions were changed in opposite directions in the optimization,
depending on which targets were used. This underlines the need for more accurate direct determination of their
rate constant.
6. We also observe that the amount of experimental data on sulfur combustion chemistry that is available for vali-
dation is very limited. More reliable data on laminar burning velocities and flame structures for H2 S, especially
at higher pressure, as well as data on CH4 /H2 S flames under air and oxy-fuel conditions would be extremely
valuable in this respect.

7
Acknowledgments

The financial support by Siemens Energy, Inc. is gratefully acknowledged. We would like to thank Chenlai (Ryan)
Zhou, Olivier Mathieu, and Andreas Kronenburg for helpful discussions. We would also like to thank the reviewers
for their constructive comments that have led to significant improvements of this paper.

References
[1] W. F. J. Burgers, P. S. Northrop, H. S. Kheshgi, J. A. Valencia, Energy Procedia 4 (2011) 2178–2184.
[2] G. Hammer, T. Lübcke, R. Kettner, M. R. Pillarella, H. Recknagel, A. Commichau, H.-J. Neumann, B. Paczynska-Lahme, in: Ullmann’s
Encyclopedia of Industrial Chemistry, volume 23, Wiley-VCH, Weinheim, 2012, p. 740.
[3] M. B. Toftegaard, J. Brix, P. A. Jensen, P. Glarborg, A. D. Jensen, Prog. Energ. Combust. Sci. 36 (2010) 581–625.
[4] H. M. Kvamsdal, K. Jordal, O. Bolland, Energy 32 (2007) 10–24.
[5] J. Davison, Energy 32 (2007) 1163–1176.
[6] G. Pipitone, O. Bolland, Int. J. Greenh. Gas Con. 3 (2009) 528–534.
[7] G. P. Smith, D. M. Golden, M. Frenklach, N. W. Moriarty, B. Eiteneer, M. Goldenberg, C. T. Bowman, R. K. Hanson, S. Song, W. C. Gardiner,
Jr., V. V. Lissianski, Z. Qin, GRI-Mech 3.0, available at http://www.me.berkeley.edu/gri-mech/version30/text30.html.
[8] P. Glarborg, L. L. B. Bentzen, Energy Fuels 22 (2008) 291–296.
[9] T. Mendiara, P. Glarborg, Combust. Flame 156 (10) (2009) 1937–1949.
[10] Z. Tian, Y. Li, L.Zhang, P. Glarborg, F. Qi, Combust. Flame 156 (2009) 1413 – 1426.
[11] T. Le Cong, P. Dagaut, G. Dayma, J. Eng. Gas Turb. Power 130 (2008) 041502.
[12] T. Le Cong, P. Dagaut, Energy Fuels 23 (2009) 725–734.
[13] W. K. Metcalfe, S. M. Burke, S. S. Ahmed, H. J. Curran, Int. J. Chem. Kinet. 45 (2013) 638–675.
[14] A. Keromnes, W. K. Metcalfe, K. A. Heufer, N. Donohoe, A. K. Das, C.-J. Sung, J. Herzler, C. Naumann, P. Griebel, O. Mathieu, M. C.
Krejci, E. L. Petersen, W. J. Pitz, H. J. Curran, Combust. Flame 160 (2013) 995–1011.
[15] Z. Hong, D. F. Davidson, R. K. Hanson, Combust. Flame 158 (2011) 633 – 644.
[16] G. Richards, K. Casleton, B. Chorpening, Proc. Inst. Mech. Eng. A J. Power Energy 219 (2005) 121–126.
[17] T. Williams, C. Shaddix, R. Schefer, Combust. Sci. Technol. 180 (2008) 64–88.
[18] R. Lewis, K. Caselton, G. Richards, D. Straub, W. Rogers, in: 2001 Joint International Combustion Symposium, Kauai, HI.
[19] F. Liu, H. Guo, G. J. Smallwood, Combust. Flame 133 (4) (2003) 495–497.
[20] F. Halter, F. Foucher, L. Landry, C. Mounaı̈m-Rousselle, Combust. Sci. Technol. 181 (6) (2009) 813–827.
[21] V. R. Kishore, N. Duhan, M. R. Ravi, A. Ray, Exp. Therm. Fluid Sci. 33 (2008) 10–16.
[22] E. Albin, H. Nawroth, S. Göke, Y. D’Angelo, C. Paschereit, Fuel Process. Technol. 107 (2013) 27–35.
[23] A. N. Mazas, B. Fiorina, D. A. Lacoste, T. Schuller, Combust. Flame 158 (12) (2011) 2428–2440.
[24] V. Babkin, A. V’yun, Combust. Explos. Shock Waves 7 (1971) 339–341.
[25] T. Le Cong, P. Dagaut, in: Proceedings of ASME Turbo Expo 2008, pp. GT2008–50272.
[26] Y. Levy, E. Olchanski, V. Sherbaum, V. Erenburg, A. Burcat, J. Propul. Power 22 (3) (2006) 669–676.
[27] M. Abián, J. Giménez-López, R. Bilbao, M. U. Alzueta, Proc. Combust. Inst. 33 (2011) 317–323.
[28] P. Glarborg, D. Kubel, P. G. Kristensen, J. Hansen, K. Dam-Johansen, Combust. Sci. Technol. 110-111 (1) (1995) 461–485.
[29] M. Frenklach, J. H. Lee, J. N. White, W. C. Gardiner, Jr., Combust. Flame 41 (1981) 1–16.
[30] P. Glarborg, D. Kubel, K. Dam-Johansen, H.-M. Chiang, J. W. Bozzelli, Int. J. Chem. Kinet. 28 (1996) 773–790.
[31] M. U. Alzueta, R. Bilbao, P. Glarborg, Combust. Flame 127 (4) (2001) 2234–2251.
[32] Leeds University, The SOx mechanism extension to the Leeds Methane Oxidation Mechanism, available at
http://garfield.chem.elte.hu/Combustion/Combustion.html.
[33] F. G. Cerru, A. Kronenburg, R. P. Lindstedt, Proc. Combust. Inst. 30 (2005) 1227–1235.
[34] F. G. Cerru, A. Kronenburg, R. P. Lindstedt, Combust. Flame 146 (3) (2006) 437–455.
[35] C. Zhou, K. Sendt, B. S. Haynes, Proc. Combust. Inst. 34 (2013) 625–632.
[36] E. Goos, A. Burcat, B. Ruscic, Ideal gas thermochemical database with updates from active thermochemical tables, available at
http://garfield.chem.elte.hu/Burcat/burcat.html, 2013.
[37] P. A. Denis, J. Sulfur Chem. 29 (3-4) (2008) 327–352.
[38] C. Zhou, K. Sendt, B. S. Haynes, J. Phys. Chem. A 113 (2009) 8299–8306.
[39] D. S. Chamberlin, D. R. Clarke, Ind. Eng. Chem. 20 (10) (1928) 1016–1019.
[40] L. Cohen, Fuel 34 (1955) S119–S122.
[41] G. J. Gibbs, H. F. Calcote, J. Chem. Eng. Data 4 (3) (1959) 226–237.
[42] C. Flockenhaus, Gas Wärme Int. 18 (4) (1969) 153–156.
[43] K. A. Hawboldt, W. D. Monnery, W. Y. Svrcek, Chem. Eng. Sci. 55 (2000) 957–966.
[44] C.-W. Lu, Y.-J. Wu, Y.-P. Lee, R. S. Zhu, M. C. Lin, J. Chem. Phys. 121 (17) (2004) 8271–8278.
[45] J. Peng, X. Hu, P. Marshall, J. Phys. Chem. A 103 (1999) 5307–5311.
[46] A. Grillo, R. Reed, M. W. Slack, in: Proceedings of the 12th International Symposium on Shock Tubes and Waves, Jerusalem, July 12-19,
1979, Magnes Press, The Hebrew University, 1980, pp. 486–494.
[47] G. Black, R. L. Sharpless, T. G. Slanger, Chem. Phys. Lett. 93 (6) (1982) 598–602.
[48] N. L. Garland, Chem. Phys. Lett. 290 (1998) 385–390.
[49] Y. Gao, C. Zhou, K. Sendt, B. S. Haynes, P. Marshall, Proc. Combust. Inst. 33 (2011) 459–465.

8
[50] H. Shiina, M. Oya, K. Yamashita, A. Miyoshi, H. Matsui, J. Phys. Chem. 100 (6) (1996) 2136–2140.
[51] C. Wang, G. Zhang, Z. Wang, Q. S. Li, Y. Zhang, J. Mol. Struct.-THEOCHEM 731 (2005) 187–192.
[52] K. Sendt, M. Jazbec, B. S. Haynes, Proc. Combust. Inst. 29 (2002) 2439–2446.
[53] D. Bertsekas, Nonlinear Programming, Athena Scientific, 2nd edition, 1999.
[54] R. J. Kee, F. M. Rupley, J. A. Miller, et al., CHEMKIN-PRO 15112, Reaction Design, San Diego, CA, 2011.
[55] M. Binoist, B. Labégorre, F. Monnet, P. D. Clark, N. I. Dowling, M. Huang, D. Archambault, E. Plasari, P.-M. Marquaire, Ind. Eng. Chem.
Res. 42 (2003) 3943–3951.
[56] O. Mathieu, F. Deguillaume, E. L. Petersen, Combust. Flame 161 (1) (2014) 23–36.
[57] E. L. Merryman, A. Levy, J. Air Pollut. Control Assoc. 17 (12) (1967) 800–806.
[58] R. Stanger, T. Wall, Prog. Energ. Combust. Sci. 37 (2011) 69–88.
[59] G. Blythe, K. Dombrowski, SO3 Mitigation Guide Update, Technical Report 1004168, EPRI, Palo Alto, CA, 2004.
[60] N. Eliaz, G. Shemesh, R. M. Latanision, Eng. Fail. Anal. 9 (2002) 31–43.
[61] D. Fleig, F. Normann, K. Andersson, F. Johnsson, B. Leckner, Energy Procedia 1 (2009) 383–390.
[62] D. Fleig, M. U. Alzueta, F. Normann, M. Abián, K. Andersson, F. Johnsson, Combust. Flame 160 (6) (2013) 1142–1151.
[63] J. Giménez-López, M. Martı́nez, A. Millera, R. Bilbao, M. U. Alzueta, Combust. Flame 158 (1) (2011) 48–56.
[64] P. Dagaut, F. Lecomte, J. Mieritz, P. Glarborg, Int. J. Chem. Kinet. 35 (2003) 564–575.
[65] H. S. F. Chin, K. Karan, A. K. Mehrotra, L. A. Behie, Can. J. Chem. Eng. 79 (2001) 482–490.

9
Tables

Table 1: Targets used in the optimization of the combined mechanism.

No. Case Point Target Value Weight Reference
1 H2 S Laminar Burning Velocity Φ=1 42-46 cm/s 40 [39–42]
2 H2 S Ignition Delay Time Φ = 0.5, T = 950 K 1000-1600 µs 1 [29]
3 Φ = 0.5, T = 1200 K 60-90 µs 1 [29]
4 Φ = 1.8, T = 950 K 1600-2500 µs 1 [29]
5 Φ = 1.8, T = 1200 K 100-160 µs 1 [29]
6 H2 S Pyrolysis T = 1000 ◦ C, t = 0.64 s 18.6% H2 S Loss 1 [43]
7 T = 1150 ◦ C, t = 0.17 s 45.1% H2 S Loss 1 [43]
8 H2 S Flow Reactor, 520 ppm T = 1000 K 498 ppm H2 S 1 [35]
9 T = 1050 K 19 ppm H2 S 1 [35]
10 H2 S Flow Reactor, 325 ppm T = 1030 K 314 ppm H2 S 1 [35]
11 T = 1070 K 132 ppm H2 S 1 [35]
12 T = 1100 K 0 ppm H2 S 10 [35]

Table 2: Changes in the pre-exponential factors relative to the values in [35] for optimization against different target combinations.
Reaction Optimization 1–12 Optimization 1–7 Optimization 8–12
H2 S + H = SH + H2 0.96 3.00 0.33
H2 S + S = SH + SH (1) 1.20 0.83 1.00
H2 S + S = SH + SH (2) 0.78 1.25 0.75
S + SH =S2 + H 1.11 3.00 0.63
H + HSS = SH + SH (1) 2.82 3.00 0.67
H + HSS = SH + SH (2) 1.35 3.00 1.22
SH + SH (+M) = HSSH 0.17 0.17 1.55
Low pressure limit 3.00 1.11 1.50
SH + HSS = H2 S + S2 0.33 0.33 1.11
H2 S + S (+M) = HSSH (+M) 0.63 0.39 2.00
SH + HO2 = H2 S + O2 1.35 1.35 0.84
H2 S + O = SH + OH 0.50 0.77 2.00
S + O2 = SO + O 0.90 0.80 1.10
SO + O2 = SO2 + O 0.89 0.89 0.84
H + SO2 (+M) = HOSO (+M) 1.50 1.20 0.33
Low pressure limit 1.89 0.81 0.35
S2 + O = SO + S 1.20 1.35 1.68

10
Figures

1 AramcoMech 1.3

Burning Velocity Ratio [-]

GRI-Mech 3.0
Glarborg et al.

0.5

0
0 0.1 0.2
CO2 Mole Fraction [-]

Figure 1: The modeling results (lines) agree well with the experimental data (symbols) of Halter et al. [20] (◦) and Kishore et al. [21] (4) on the
reduction of the laminar burning velocity of stoichiometric CH4 -air flames through CO2 addition (T = 300 K, p = 1 atm).

p=1atm p=15atm
1 AramcoMech 1.3 1
Burning Velocity Ratio [-]

GRI-Mech 3.0
Glarborg et al.

0.5 0.5

0 0
0 0.1 0.2 0 0.1 0.2
H2O Mole Fraction [-] H2O Mole Fraction [-]

Figure 2: The modeling results (lines) agree well with the experimental data (symbols) of Albin et al. [22] (4,5), Mazas et al. [23] (), and Babkin
and V’yun [24] (◦) on the reduction of the laminar burning velocity of stoichiometric CH4 -air flames through H2 O addition at atmospheric pressure
(a). At high pressure, AramcoMech 1.3 preforms better than the other mechanisms (b).

11
 (a) (b)
3 3

Ignition Delay Time [µs]

10 10

2 2
10 10
AramcoMech 1.3
GRI-Mech 3.0
Glarborg et al.
1 Dagaut et al. 1
10 10
0.6 0.7 0.6 0.7
1000/Temperature [1/K] 1000/Temperature [1/K]

Figure 3: The modeling results (lines) agree well with the experimental data (symbols) of Levy et al. [26] on the ignition delay time of CH4 at
p = 7-9 bar, Φ = 0.3. (a) N2 dilution with 5.2% CO2 , (b) N2 dilution with 5.2% CO2 and 4.0% H2 O.

Φ=1.4 Φ=0.5 AramcoMech 1.3

1 GRI-Mech 3.0 1
Glarborg et al.
Dagaut et al.
CO/COinlet

0.5 0.5

0 0
800 1200 1600 800 1200 1600
Temperature [K] Temperature [K]

Figure 4: Comparison between modeling results (lines) and data (symbols) from the atmospheric CO/O2 /CO2 /H2 O/N2 flow reactor of Abián et
al. [27]. The modeling was done assuming plug flow with a flow rate of 1 lSTP /min and the temperature profiles given in [63]. Inlet conditions:
XCO ≈ 2000 ppm, XH2 O = 0.6%, XCO2 = 75%, balance N2 .

Φ=2 Φ=0.5 AramcoMech 1.3

1 GRI-Mech 3.0 1
Glarborg et al.
Dagaut et al.
CO/COinlet

0.5 0.5

0 0
1000 1200 1400 1000 1200 1400
Temperature [K] Temperature [K]

Figure 5: Comparison between modeling results (lines) and data (symbols) from the atmospheric CO/O2 /CO2 /H2 O/N2 flow reactor of Abián et
al. [27]. Inlet conditions: XCO ≈ 5100 ppm, XCO2 = 25%, XH2 O = 10%, balance N2 .

12
 1 AramcoMech 1.3

CO Mole Fraction [ppm]

GRI-Mech 3.0
Glarborg et al.
Dagaut et al.

0.5

0
1000 1100 1200
Temperature [K]

Figure 6: Comparison between modeling results (lines) and data (symbols) from the atmospheric flow reactor experiment of Glarborg et al. [28].
No full experimental temperature profiles were available so that a fixed residence time of τ = 64/T s was assumed, where T is the constant reactor
temperature in kelvin. Inlet conditions: XCO ≈ 1500/1630 ppm, XO2 = 2.0%, XO2 = 32.0%, balance N2 .
CO Mole Fraction [ppm]

Φ=0.5
2000

1000

0
CO Mole Fraction [ppm]

3000 Φ=1

1500

0
CO Mole Fraction [ppm]

4000 Φ=4

2000 AramcoMech 1.3

GRI-Mech 3.0
Glarborg et al.
Dagaut et al.
0
1200 1400 1600 1800
Temperature [K]

Figure 7: Comparison between modeling results (lines) and data (symbols) from the atmospheric CH4 /O2 /CO2 /N2 flow reactor of Glarborg and
Bentzen [8] on CO formation in CH4 oxidation in the presence of CO2 . The modeling was done assuming a residence time of τ ≈ 960/T s and
constant temperature. Inlet conditions: XCO ≈ 1000 ppm, XCO2 = 77-95 %, balance N2 .

13
 (a) S + O2 = SO + O (b) SH + HSS = H2S + S2
SO + O2 = SO2 + O H + HSS = H2S + S
H + O2 = O + OH S + SH = S2 + H
S2 + O = SO + S H2S + H = SH + H2
S + H2 = SH + H H2S + M = H2 + S + M
OH + H2 = H + H2O H + S2 + M = HSS + M
O + H2 = H + OH HSSH + H = H2S + SH
H2S + H = SH + H2 H2S + S (+M) = HSSH (+M)
S + SH = S2 + H SH + SH (+M) = HSSH (+M)
H + SO2 (+M) = HOSO (+M) H + HSS = SH + SH (2)
-0.5 -0.25 0 0.25 0.5 -0.1 -0.05 0 0.05 0.1
Flow Rate Sensitivity Maximum H2S Sensitivity

(c) H2S + S = SH + SH (1) (d) H2S + S = SH + SH (1)

H2S + S = SH + SH (2) lean S + O2 = SO + O lean
SH + HO2 = H2S + O2 rich SH + HO2 = H2S + O2 rich
S2 + O = SO + S S2 + O = SO + S
SH + O2 = H + SO2 SO + O2 = SO2 + O
SO + O2 = SO2 + O H2S + S = SH + SH (2)
S + O2 = SO + O S + SH = S2 + H
SH + HSS = H2S + S2 H2S + O = SH + OH
H2S + O = SH + OH H + HSS = SH + SH (2)
SH + SH (+M) = HSSH (+M) SH + SH (+M) = HSSH (+M)
-1000 -500 0 500 1000 -20 -10 0 10 20
Maximum Temperature Sensitivity Maximum Temperature Sensitivity

Figure 8: Sensitivity coefficients of the pre-exponential factor of different reactions. (a) Sensitivity with respect to the laminar burning velocity of
H2 S in air at p = 1 atm, Φ = 1; (b) Sensitivity with respect to the H2 S concentration in the pyrolysis experiment [43] at T = 1000 ◦ C; (c) Sensitivity
with respect to temperature in the ignition delay measurement [29] for T 0 =950 K; (d) Sensitivity with respect to temperature in the ignition delay
measurement [29] for T 0 =1200 K.
Laminar Burning Velocity [cm/s]

60

40

20

0
0.5 1 1.5 2
Equivalence Ratio

Figure 9: Laminar burning velocity of H2 S in air at atmospheric pressure. The symbols are the experimental data of Chamberlin and Clarke [39]
(×), Cohen [40] (◦), Gibbs and Calcote [41] (), and Flockenhaus [42] (4). The lines are results of the present modeling study: Basic mechanism
(solid), Optimization 1-12 (dashed), Optimization 1-7 (dash-dotted), Optimization 8-12 (dotted). Both Soret effect and multicomponent diffusion
were used for the modeling.

14
 Φ=0.5 Φ=1.8

Ignition Delay Time [µs]

3 3
10 10

2 2
10 10

0.85 0.95 1.05 0.85 0.95 1.05

1000/Temperature [1/K] 1000/Temperature [1/K]

Figure 10: Ignition delay time of H2 S in air as a function of temperature. The symbols are the experimental data of Frenklach et al. [29]. The lines
are results of the present modeling study: Basic mechanism (solid), Optimization 1-12 (dashed), Optimization 1-7 (dash-dotted), Optimization 8-12
(dotted). P = 30-45 atm. The pressure was varied along with the initial temperature to keep the partial density of H2 S constant.

400 600
(a) (b)
H2S Mole Fraction [ppm]

200 300

0 0
950 1050 950 1050 1150
Temperature [K] Temperature [K]

Figure 11: Comparison between the experimental data (symbols) for the atmospheric H2 S/O2 /N2 flow reactor of Zhou et al. [35] and the results
of the present modeling study: Basic mechanism (solid), Optimization 1-12 (dashed), Optimization 1-7 (dash-dotted), Optimization 8-12 (dotted).
The modeling was done using temperature profiles similar to the one shown in [35]. Inlet conditions: (a) XH2 S = 325 ppm, XO2 = 600 ppm; and
(b) XH2 S = 520 ppm, XO2 = 1000 ppm.

400 600
(a) (b)
SO2 Mole Fraction [ppm]

200 300

0 0
950 1050 950 1050 1150
Temperature [K] Temperature [K]

Figure 12: Comparison between the experimental data (symbols) for the atmospheric H2 S/O2 /N2 flow reactor of Zhou et al. [35] and the results
of the present modeling study: Basic mechanism (solid), Optimization 1-12 (dashed), Optimization 1-7 (dash-dotted), Optimization 8-12 (dotted).
Same conditions as in Fig. 11.

15
 400
(a) (b)
H2 Mole Fraction [ppm]

200

200
100

0 0
950 1050 950 1050 1150
Temperature [K] Temperature [K]

Figure 13: Comparison between the experimental data (symbols) for the atmospheric H2 S/O2 /N2 flow reactor of Zhou et al. [35] and the results
of the present modeling study: Basic mechanism (solid), Optimization 1-12 (dashed), Optimization 1-7 (dash-dotted), Optimization 8-12 (dotted).
Same conditions as in Fig. 11.

100
1150°C
1000°C
H2S Conversion [%]

50

0
0 0.5 1
Time [s]

Figure 14: Comparison between modeling results (lines) and experimental data (symbols) of Hawboldt et al. [43] for the pyrolysis of H2 S at
atmospheric pressure. Basic mechanism (solid), Optimization 1-12 (dashed), Optimization 1-7 (dash-dotted), Optimization 8-12 (dotted). Initial
conditions: XH2 S = 2.5%, balance N2 .

16
 p=13atm p=33atm

Ignition Delay Time [µs]

3 3
10 10

2 2
10 10

0.7 0.8 0.9 0.7 0.8 0.9

1000/Temperature [1/K] 1000/Temperature [1/K]

Figure 15: Ignition delay time of H2 /H2 S mixtures in Ar at different pressures. The symbols are the experimental data of Mathieu et al. [56]. The
lines are results of the present modeling study: Basic mechanism (solid), Optimization 1-12 (dashed), Optimization 1-7 (dash-dotted), Optimization
8-12 (dotted). Initial conditions: XH2 = 1%, XO2 = 1%, XH2 S = 0.16%, balance Ar.

0.2 0.08
H2S
O2
SO2
Mole Fraction [-]

0.1 0.04

H2 x2 H2O
SO x2
0 0
0 1 0 1 2
Distance [cm] Distance [cm]

Figure 16: Comparison between modeling results (lines) and experimental data (symbols) for the H2 S-air flame of Merryman and Levy [57].
P = 0.1 atm, Φ = 0.5. Plot was divided for better readability. The modeling used the experimental temperature profile and extrapolated upstream
to 300 K.

17
 3
40 (a) (b)
SO3 Mole Fraction [ppm] 20.3%

SO3/SOx Ratio [%]

O2 250ppm
5.0% 500ppm 2
3.0% 1000ppm
20 0.8%
0.1% 1

SO2
0 0
(c) (d)

SO3 Mole Fraction [ppm]

SO3 Mole Fraction [ppm]

20 8.7% 20
80%
H2O
CO2 1.1%
0%
0.1%

10 10

0 0
900 1100 1300 900 1100 1300 1500
Peak Temperature [K] Peak Temperature [K]

Figure 17: Comparison between modeling results (lines) and experimental data (symbols) for the atmospheric flow reactor of Fleig et al. [62]. The
modeling was done for a fixed flow rate of 1.011 lSTP /min using the experimental temperature profiles. Inlet conditions: (a) XSO2 = 1000 ppm,
XH2 O = 1.11%, varying O2 ; (b) XO2 = 3%, XH2 O = 1.11%, varying SO2 ; (c) XSO2 = 1000 ppm, XO2 = 3%, XH2 O = 1.11%, varying CO2 ;
(d) XSO2 = 1000 ppm, XO2 = 3%, varying H2 O. Balance N2 in all cases.

75% CO2 0% CO2

1 1
0.1%
0%
CO/COinlet

SO2
0.5 0.5

SO2

0 0
800 1200 1600 800 1200 1600
Temperature [K] Temperature [K]

Figure 18: Comparison between modeling results (lines) and experimental data (symbols) for the atmospheric flow reactor of Giménez-López et
al. [63]. The modeling was done at a constant flow rate of 16.7 cm3STP /s using the measured temperature profile. Inlet conditions: XCO = 2000 ppm,
XO2 = 1000 ppm, XH2 O = 0.6%, XCO2 = 75%/0%, varying SO2 , balance N2 .

18
 100

H2S

Conversion [%]
50

1.5% O2
CH4
1.0% O2
0
1000 1100 1200
Temperature [°C]

Figure 19: Comparison between modeling results (lines) and experimental data (symbols) for the flow reactor of Chin et al. [65]. P = 1.1 bar. Inlet
conditions: XH2 S = 2.45%, XCH4 = 2.45%, varying O2 , balance N2 ; τ = 0.6 s.

30
SO3 Mole Fraction [ppm]

Base
0.1% CO
0.1% CH4
20
0.3% CH4

10

0
900 1100 1300 1500
PeakTemperature [K]

Figure 20: Comparison between modeling results (lines) and experimental data (symbols) for the atmospheric flow reactor of Fleig et al. [62]. Inlet
conditions: XSO2 = 1000 ppm, XO2 = 3%, XH2 O = 1.11%, varying CO and CH4 , balance N2 .

19