A detailed chemical reaction mechanism for the oxidation of hydrocarbons and its application to the analysis of benzene formation in fuel-rich premixed laminar acetylene and propene flames

On the basis of existing detailed kinetic schemes a general and consistent mechanism of the oxidation of hydrocarbons and the formation of higher hydrocarbons was compiled for computational studies covering the characteristic properties of a wide range of combustion processes.

Computed ignition delay times of hydrocarbon–oxygen mixtures (CH4-, C2H6-, C3H8-, n-C4H10-, CH4 + C2H6-, C2H4, C3H6-O2) match the experimental values.

The calculated absolute flame velocities of laminar premixed flames (CH4-, C2H6-, C3H8-, n-C4H10-, C2H4-, C3H6-, and C2H2-air) and the dependence on mixture strength agree with the latest experimental investigations reported in the literature.

With the same model concentration profiles for major and intermediate species in fuel-rich, non-sooting, premixed C2H2-, C3H6- air flames and a mixed C2H2/C3H6 (1:1)-air flame at 50 mbar are predicted in good agreement with experimental data.

An analysis of reaction pathways shows for all three flames that benzene formation can be described by propargyl combination.


With growing computing power modeling of various physical and technical systems becomes approachable or more realistic.

In the field of combustion research modeling provides a strong tool for the analysis of chemical and physical phenomena and predictions of properties of combustion systems.

Considerable progress has been made in the development of detailed reaction mechanisms being used in a manifold of scientific and industrial applications.

One aspect is modeling of hydrocarbon fuel oxidation including soot formation.

Soot emissions of combustion engines are suspected to cause cancer and research on soot formation is in the focus of both science and industry.

Warnatz and coworkers1–4 have developed a widely-used mechanism for the oxidation of C1-C4-hydrocarbon fuels which is validated for flame speeds, ignition delay times and species concentration profiles.

The GRI-mechanism5 is an extensively tested mechanism geared to methane and natural gas oxidation including NOx-chemistry and used as reference mechanism in many computational studies.

In 1987 Frenklach and Warnatz presented a kinetic model predicting PAH (poly-aromatic hydrocarbon) concentration profiles in sooting acetylene flames.6

Ten years later Wang and Frenklach published a GRI based detailed reaction mechanism describing fuel oxidation, benzene formation and PAH mass growth7 being the most extensive study on soot formation in C2-flames.

This mechanism was moderately updated in .20008

Benzene is commonly accepted to be a key species in PAH formation while the pathways to benzene are still discussed.

Most studies9–11 favour the combination of propargyl radicals C3H3 + C3H3 → A1 (reaction (R383), see mechanism in Table 1) to be the main source of benzene (A1, first aromat) while Wang and Frenklach7 demonstrated that in acetylene flames under certain conditions benzene formation can be described via alternative routes such as C4H5 + C2H2 → A1 + H, reaction (R449).

An experimental investigation of Atakan et al12. of a fuel rich propene flame indicates that additional reaction sequences via C6Hx,(x>6) are involved in benzene formation.

In a modeling study13 of benzene formation for C2 and C3 low pressure hydrocarbon flames (acetylene, ethylene, and propene) Pope and Miller found that besides propagyl combination the reaction C3H3 + C3H5 (allyl) → A1 + H + H reaction (R*) contributes significantly to benzene formation.

The pioneering work of the named authors and many others has provided modeling tools for various aspects of hydrocarbon oxidation yet the application of every kinetic model is limted to its range of test cases.

Thus it is desirable to provide a detailed reaction mechanism which is capable to predict flame structures, e.g. species concentration profiles, and global kinetic properties like flame speeds and ignition delay times simultaneously for a broad range of fuels.

Analysis of reaction pathways becomes more reliable with a larger set of test cases.

Therefore the aim of this study is to demonstrate that existing models can be improved when expanded to a broader range of targets.

This approach is confirmed by a study of Qin et al14. for C3 fuel-oxidation.

To obtain an optimal C3 data set, simultaneous optimisation of the kinetics of small and intermediate sized fuels was necessary.

It will be shown that on the basis of the mechanisms of Warnatz and coworkers and Wang and Frenklach a kinetic scheme can be compiled that fulfils the demands named above.

The new mechanism predicts (a) the flame speeds of laminar premixed hydrocarbon–air flames for the fuels methane, ethane, ethylene, acetylene, propane, propene and butane in good agreement with the latest experimental data, (b) ignition delay times and species concentration profiles for various test cases of Warnatz’s mechanism in good agreement with experiments and (c) the concentration profiles of major and intermediate species in premixed, fuel-rich non-sooting acetylene– and propene–air flames and a mixed acetylene/propene–air flame including the propargyl radical and benzene.

The reaction pathways leading to benzene are analysed by application of sensitivity and reaction flow analysis.

The results are compared with those of other experimental and computational studies with respect to the fuel, the temperature range and the vital role of the propargyl radical as a key intermediate in benzene formation.

The species concentration profiles or maximum concentration/position data are taken from flame chemistry studies by Atakan et al. obtained with molecular beam sampling mass spectroscopy The calculations were performed with the flame and ignition codes used in the Chemical Kinetics group in Lund40.

Chemical model

The detailed reaction mechanism consists of 569 reversible reactions and 96 chemical species.

The base for the C1-/C2- and C4-chemistry is the latest version of Warnatz’s mechanism.4

The kinetic data relevant for benzene formation and destruction including additional species and C2–C4-reactions is based on the mechanism7 of Wang and Frenklach.

The C3-chemistry is a new compilation.

The species nomenclature of C1–C4-species corresponds to ref. 2 and 3, thermodynamic data were taken from the Sandia National Laboratories database.15

For species introduced with the chemistry of Wang and Frenklach’s mechanism the nomenclature corresponds to ref. 16 and thermodynamic properties are taken from .ref. 7

The motivation to replace the GRI-C1-C2-chemistry in the original Wang–Frenklach-scheme is its limited reliability for fuels different from methane/natural gas as highlighted by the authors.5

In order to predict simultaneously flame velocities, ignition delay times and species concentration profiles for various fuels and flame stoichiometries an unspecific, broadly validated C1–C4-chemistry is appropriate.

Due to the extended application area the new compilation of the chemical scheme demands some modifications of the kinetic data.

The rate coefficients were taken either from literature or estimated based on analogous reactions.

However, the most of the key reactions accounting for benzene and benzene precursor formation are identical with the model of Wang and Frenklach as described in .ref. 7

Below those changes are discussed which have the greatest impact on the flame speeds of hydrocarbon flames and the species concentration profiles in the acetylene, propene, and the acetylene/propene flames.

These reactions are listed for convenience in Table 1, where they are cited with their original source.

In order to keep the main body of the paper short the full mechanism and the thermodynamic data are given as an electronic supplement.

Results and discussion

Flame speed

Current values of flame speeds for hydrocarbons in air at atmospheric pressure differ significantly from older data.

In the present study we adopted largely the experimental burning velocity data collected in the Princeton counterflow flames.17,18

The C1–C4 mechanism of Warnatz and coworkers, not validated against current flame speed data, overpredicts flame speeds for all fuel–air mixtures under study.

The key change of the Warnatz C1–C2-chemistry is the replacement of the kinetic data for the reaction 3CH2 + O2 by the values reported by Dembrowsky and Wagner.19

The former rate coefficients20 and reaction pathways are recommended for temperatures up to 1000 K which is far below flame temperatures.

The values of ref. 19 are results of shock tube experiments between 1000–1800 K and thus more appropriate for modeling high temperature hydrocarbon oxidation.

They show in contrast to the old values a negative activation energy and the total rate is slower by one order of magnitude at 1000 K. The new experimental data for CH3 + O features the branching path producing CO which effects strongly the mass flow in methane flames.

This reaction has been extensively studied in the last decade21,22 and the latest results of Hoyermann and coworkers23 are included in the mechanism:CH3 + O → HCHO + HCH3 + O → CO + H2 + HThe reactionC2H5 + H → CH3 + CH3in Warnatz’s mechanism has already been identified by Egolofopoulos et al17. to raise ethane flame speeds significantly, when comparing predictions of laminar burning velocities on the basis of different detailed mechanisms and the experimental data.

Using the rate coefficient for the backward reaction20CH3 + CH3 → C2H5 + Has recommended by Egolofopoulos reduces the flame speed by 2–4 cm s−1.

The C2H5 + H rate is calculated via the equilibrium constant.

Fig. 1 reflects the very good agreement between predicted and experimental flame speeds for methane and ethane.

These above-named changes to the C1–C2-chemistry lead also to a good match between calculated and measured flame speeds for acetylene and ethylene flames.

Especially the agreement in the dependence of the flame speed and mixture strength for ethylene under fuel rich conditions, displayed in Fig. 2, shows that flame speed and flame structure predictions, as discussed below, did not add up to a conflict with flame structure calculations.

The overprediction of the acetylene flame speeds for fuel-rich mixtures is due to the added HCCO decomposition being also discussed in the flame structure part.

The named changes to the C1–C2 chemistry and the new C3-chemistry compilation implemented to predict acetylene-/propene-flame structures lead also to good agreement between calculated and measured flame speeds of propane, propene and butane flames displayed in Fig. 3.

Ignition delay times

In order to prove the quality of the presented mechanism and its general applicability this scheme was chosen without further modifications for the calculation of ignition delay times of a variety of fuels.

As demonstrated in Fig. 4 the absolute values and the temperature dependence (i.e. apparent activation energy) of the experimentally determined ignition delay times are well reproduced by the calculations.

It is noteworthy to state that the original Warnatz scheme gave similar results with minor deviations for propene and butane air mixtures.

Flame structure

The structure of laminar premixed fuel-rich acetylene-, propene- and acetylene/propene-air flames34,12,35 was analyzed.

In the following a comparison of predicted and measured species concentrations is given with emphasis on major and intermediate species relevant for benzene formation and PAH growth.

Where experimental data is available measured and predicted concentration profiles are displayed in figures.

For many species only positions and values of the maximum concentration are provided.

These data are summarized for all species in Table 2.

The experimental position data is slightly shifted to account for probe effects.

A summary of the conditions of the three well-studied burner stabilized laminar premixed flames simulated in the present study is given in Table 3.

The associated temperature profiles are shown in Figs. 5–7.

For the present investigation the energy conservation equation was not solved explicitly and the temperature profiles were extracted from the experimental data.

The profiles actually used in the calculations (dashed lines in Figs. 5–7) are raised by 100–200 K which is well within the uncertainity of the applied temperature diagnostics, T(OH-LIF), T(CO,H2-Raman).12

This leads to better results for the partial equilibrium determined C2H2 concentrations in the post-flame region.

The reaction pathways leading to benzene are traced back by application of reaction sensitivity and mass flow analysis.

An analysis of the main fuel consuming pathways is included.

The results are illustrated in flow diagrams.

Flame 1: Acetylene

Good agreement between predicted and the experimental concentrations is found for the major species H2, H2O, CO. The calculated concentration of CO2,(x=12 mm) is overpredicted by approximately 60%.

The addition of the HCCO decomposition reactionHCCO → CH + COreduced the CO2 concentration maximum by 30%.

This effect is due to a decreased CO2 formation via the competitive reactionHCCO + O2 → H + CO + CO2which is a major source of CO2 under fuel rich conditions.

Its high sensitivity on CO2 formation is due to a high HCCO channel fraction of the reaction C2H2 + O (reaction (R111)) being the main fuel consuming path.

It should be noted that the influence of the HCCO decomposition reaction depends strongly on the thermodynamic properties of HCCO and the fuel/air ratio.

Moreover, it is not limited to acetylene flames: Our calculations for methane/air flames and a comparison of experimental results and GRI predictions36 show that CH concentrations are reduced for lean mixtures and raised for fuel rich mixtures.

Additional kinetic and thermodynamic studies for this high sensitive reaction are desirable.

Furthermore the early formation of CO2 in acetylene flames reported by Hoyerman et al. in 197237 can be explained via the sequence of reactions (R111) and (R107).

In a recent experimental study using time-resolved FTIR it was demonstrated by Osborn38 that CO2 is the main product for reaction (R107).

Fig. 8 shows that the concentrations of benzene, A1 and the benzene precursors C3H3, C4H3, C4H5 are predicted in good agreement with experimental results.

The maximum concentrations of C2H4 and further C3 and C4 species are also predicted within the error margins given as 30% for major species (mole fraction >10−3) and a factor of three for radicals and minor species (mole fraction<10−3) by Atakan et al.

(see Table 2).

The mass flow analysis of the reaction pathways from the fuel to benzene shows that C2H2 is mainly consumed by the reactionsC2H2 + O → 3CH2 + COC2H2 + O → HCCO + H.C3H3 is predominantly formed by the reactions of 3CH2 and 1CH2 with the fuel where 3CH2 and 1CH2 are the triplet and singlet methylenes respectively.C2H2 + 1CH2 → C3H3 + HC2H2 + 3CH2 → C3H3 + HThe rate coefficients of these C2H2/CH2 reactions and other pathways (R196,197,209,210) to C3H4/C3H4P (allene/propyne) are fitted to give optimum results for C3H3 and C3H4 concentrations profiles for all three flames.

The overall rate coefficients lie within the error range of the reported values.10,20

The importance of these reactions is also emphasized by Pope and Miller13 (see also notes added in proof).41

The C4H5 radical is mainly produced via the sequence of reactionsC3H3 + CH3 → C4H6C4H6 + (H,OH) → i-C4H5, n-C4H5 + (H2,H2O)This holds for all studied flames.

The potential benzene precursor n-C4H3 is produced in all flames via the reactionsC4H2 + H → n-C4H3C4H4 + H → n-C4H3 + H2As a result of reaction flow analysis benzene is almost exclusively formed by the propargyl combinationC3H3 + C3H3 → A1while R449 is the main consuming path boosted by the high benzene concentrations.

This changes when propargyl combination is omitted in the kinetic model.

Another flow analysis revealed that benzene in this case is formed exclusively by n-C4H5 + C2H2 → A1 + H

A reaction sensitivity analysis (Fig. 9) at the position of the maximum concentration of benzene shows that the most sensitive reactions are the propargyl combination and radical/atom + fuel reactions on the path to C3H3.

This applies for all three flames invesigated here.

Considering the acetylene flame in Wang and Frenklach’s study reaction (R449) was the major source for benzene.

Its concentration profiles could be reasonably well predicted even if the propargyl combination was excluded from the kinetic model.

Here reaction (R449) is still among the most sensitive reactions.

But the computed A1 concentration profiles reach just 20% of the experimental value when reaction (R383) is neglected in the kinetic scheme.

In this simulation the n-C4H5 concentration was at the upper limit of the stated experimental error and the predicted A1 concentration was compared to the lower experimental limit.

Thus benzene formation cannot be described without a predominant contribution from reaction (R383).

The A1 concentration profile was also simulated for the acetylene flame studied by Wang and Frenklach.

Here the concentration of A1 at the peak was reduced only by 30% when propargyl combination with k383 = 1 × 1011 cm3 mol−1 s−1 is omitted (rate coefficient for 20 torr in Wang and Frenklach’s model).

The maximum benzene mole fraction is predicted 30% lower than with Wang and Frenklach’s model but within the experimental error.

Our simulation likewise showed that the benzene concentration is overpredicted in the post flame zone because of the persistent production via propargyl combination.

As already pointed out by Wang and Frenklach the formation and destruction of benzene depends strongly on temperature and thus on the thermodynamic properties of the species involved.

This is a critical point when simulating benzene concentration profiles and further kinetic studies are necessary for solving this problem.

Fig. 10 illustrates the reaction paths from the fuel to benzene identified by reaction and mass flow analysis.

The scheme is comparatively straightforward as the main benzene precursors are formed in one or two steps from the fuel molecule.

Flames 2 and 3 with different fuels show significantly more complex pathways but the route from acetylene is still an important route.

Flame 2: Propene

A very good match between the predicted and the experimental concentrations is found for the major species H2, H2O, CO, CO2 and is displayed in Fig. 11.

In case of the propene flame the CO2 concentration profile is predicted within the experimental error of 30%.

This is due to the addition of the HCCO decomposition reaction (R103) discussed above which lowered the maximum CO2 concentration again by 30%.

In Figs. 12 and 13 the mole fraction profiles of CH3, C2H2, C2H4, various C3/C4 and A1 species are displayed.

Profiles for species with mole fractions >10−3 are predicted in excellent agreement with the experimental data.

This holds especially for intermediate species on the reaction paths leading to benzene (see Fig. 14).

Even for radicals like C3H3, C4H3 and C4H5 maximum concentrations and profiles are simulated very close to the experimental values.

The new compilation of the C3 chemistry was essential to achieve these results.

It comprises large parts of Warnatz’s C3-chemistry but some modifications of key reactions were carried out which are very sensitive to the mass flow in propene flames.

The C3H4 isomer propyne, C3H4P, is central for simulating propene oxidation.

Most of its reactions are formulated analogously to those of allene.

The isomerization of allene and propyne explains the relative high C3H4 mole fraction.

Especially these modifications enabled a simulation of the C3H4 concentration profile.

The experimental values are reproduced very well as seen in Fig. 9(d).

The C3H4 (allene/propyne) decomposition by C–H-bond breaking was described via the the backward reactionsC3H3 + H → C3H4C3H3 + H → C3H4PThe kinetic data are taken from Wang and Frenklach.7

Some rate coefficients for H-atom abstraction reactions likeC3H5 + H → C3H4 + H2were slightly changed within the experimental error of the original reported values in order to predict simultaneously flame speeds for the same fuel.

The importance of the reactions of C2H2 with CH2 for predicting benzene concentration profiles of all flames was discussed above.

They are a vital part of the C3 chemistry derived here.

The experimental flame structure analysis shows an earlier formation of species C6Hx,x>6 compared to benzene (A1).

A direct simulation with the Wang–Frenklach C6-chemistry resulted in C6H8 concentrations which are almost two orders of magnitude too low.

Even benzene concentrations were predicted one order of magnitude too low.

Therefore additional chemistry was considered.

Obvious candidates for early C6Hx,x>6 formation are reactions of the species C3H6 and C3H5, being abundant early in the flame.

As already suggested by Atakan et al12. the allyl combinationC3H5 + C3H5 → C6H10and the reactionC3H6 + C3H3 → C6H8 + Hwere introduced, the latter analogously to C2H3 + C2H3.

This elementary reaction was studied by Knyazev et al39. in detail.

Adding the set of reactions (R384)–(R394), which is rather small, including some global reaction steps, allows the prediction of early C6Hx>7 formation and of accurate C3H3/A1 concentrations.

This opens a new pathway to C3H3 and to benzene which is particular to the fuel propene as will be shown below.

It should be noted that the considered allyl combination allowed a reasonable simulation of C6H10 concentrations but does not significantly contribute to benzene formation.

Further kinetic studies are needed to fully understand the mechanism of early C6Hx formation.

Simulations were also performed with reaction (R*) added to the mechanism:13C3H3 + C3H5 → A1 + H + HThough benzene concentrations in the propene flame could be predicted well even when reactions for C6H8 build up, reaction (R384)–(R394), were omitted, there arise several drawbacks.

The C6H8 maximum concentration is predicted much too low when reaction (R*) is added and the latter reactions are excluded.

The maximum position of C3H3 is shifted to higher burner distances matching the experimental data less accurately.

In Flame 1 and 3 reaction (R*) does not significantly contribute to benzene formation and the simultaneous prediction of benzene concentration profiles for all flames is no longer possible without major changes to the kinetic model.

In principle reaction (R*) is a possible source of benzene in propene flames but it is not included in the mechanism for the reasons given above.

The mass flow analysis shows that C3H6 is mainly consumed by H-atom abstractionC3H6 + H → C3H5 + H2and H-atom additionC3H6 + H → n-C3H7In the propene flame C3H3 is again formed to a high extent by the reactions of 3CH2 and 1CH2 with acetylene.

Another source is H-atom abstraction from propyne, an intermediate in the stepwise abstraction route from propene to propargyl.C3H4P + (H,OH) → C3H3 + (H2,H2O)A third important source of C3H3 is connected to the early formation of C6H8 as discussed above where C3H3 is formed by the reverse of reaction (R391).C6H8 + H → C3H6 + C3H3Benzene is formed predominantly by propargyl combination.

A minor contribution is linked to the chemistry of C6H8, being formed early in the flame.

This additional reaction sequence is via C6H8 formation, subsequent H-atom abstraction and n-C6H7 decompostionC6H8 + (H,OH) → n-C6H7 + (H2,H2O)n-C6H7 → A1 + HThe reaction sensitivity analysis (Fig. 13) at the position of the maximum concentration of benzene shows that the most sensitive reactions are the first and the last step of the pathway to benzene formation, i.e. propene H-atom abstraction und propargyl combination:C3H6 + H → C3H5 + H2C3H3 + C3H3 → A1Other sensitive reactions are related to reaction paths forming the intermediate C3H3.

The comparable high sensitivity of reactions (R258R) and (R383),C4H6 → C3H3 + CH3C3H3 + 3CH2 → C4H4 + Hshows that matching concentration profiles of C3 and C4-species simultaneously is essential (see Figs. 12(c–e) and 13).

In Fig. 14 the results of the mass flow and reaction sensitivity analysis are illustrated by a reaction flow diagram.

In case of propene the propagyl radical cannot be formed directly from the fuel.

Compared with the case of acetylene this leads to a more complex mass flow to benzene with a higher number of reaction steps involved.

In particular the route via C2H2, the main route of the acetylene flame, consists of six steps but is still part of a main reaction path.

The reaction paths via successive H-atom abstraction and the early formed C6H8 are characteristic for a propene flame.

Reaction (R391R) in particular is a source of early formed propargyl and accounts for the high benzene concentrations compared to the acetylene flame.

Flame 3: Acetylene/propene (1:1)

Since the mechanism derived here is developed and optimized for the oxidation of pure fuels its application to a blended fuel is a critical test.

No specific modifications to the kinetic scheme were carried out to predict the structure of a acetylene/propene flame.

In a recent flame study of C2H2/C3H6–O2 in Ar the concentration profiles of some intermediates are reported, although major species like H2 or CO are not included.35

Yet predictions of profiles and maximum concentrations of C1–C4 key species taken into account agree reasonably well with the experimental data and are displayed in Figs. 16–18 or listed in Table 2.

It can be seen that calculated concentrations of the fuel C2H2 and the radicals CH3 and C3H3 reach an almost constant level not detected in the experiment.

When predicting concentration profiles for the pure flames this concentration evolution is not observed but too much C3H3 is predicted at high burner distances in the propene flame (Fig. 12).

In case of the acetylene/propene flame one of the fuel components (acetylene) is formed to a large extent from the degradation of the other component (propene) and is thus present in very high concentrations in the late flame and post flame zone (see Fig. 18).

This fuel to fuel interaction is probably not fully considered by the model and likely to cause the discrepancies seen in the C2H2, CH3 and C3H3 mole fraction profiles.

However, the results show that the model developed for pure fuels can be applied to a fuel mixture without substantial drawbacks.

Analysis of the reaction flow in the flame with the blended fuel yields the same reaction pathways from the fuel to benzene as seen for the pure fuel components, but relative distribution of different pathways are partially changed.

The reaction path to benzene via early formed C6H8 dicussed above is no longer essential to predict benzene concentrations, a match with experimental results could be achieved even when this channel is omitted.

This is due to the the high contribution of C2H2 reactions with methylene to benzene formation.

Acetylene is present in high concentrations in the flame zone because of the permanent C2H2 production from C3H6 degradation which is discussed above.

Fig. 19 shows the results of the reaction sensitivity analysis at the maximum concentration of benzene.

The most sensitive reactions are those discussed for the pure flames.

Here the highest relative sensitivity is found for propargyl combination and not for a fuel radical reaction because the fuel concentrations are diluted in the blended fuel (see Figs. 9 and 15).

The high concentration of C2H2 in the flame zone not only boosts benzene formation via C3H3 but also via reaction (R449), n-C4H5 + C2H2, although only half of the fuel comprises acetylene.

Therefore the relative sensitivity of reaction (R449) has the same value as in Flame 1.

The flow diagram for the acetylene/propene flame (Fig. 20) is very similar to the propene flow diagram.

Here acetylene is both fuel and intermediate.

The path via n-C4H5 is contributing as much to benzene formation as in case of the acetylene flame.

The minor route via C6H7 of the propene flame is neglegible in the acetylene/propene flame.


The mechanism presented here is capable to predict flame speeds, especially under fuel rich conditions, and ignition delay times for a large range of C1–C4 hydrocarbon fuels in good agreement with experimental data.

Cross-checking of this very broad range of targets is used to identify associated sets of sensitive reactions and to optimize the model.

This constitutes an important precondition for its application to flame structure studies, simulating concentration profiles of major species and key intermediates in fuel rich acetylene and propene flames in good agreement with the experiments.

The chemical structure analysis for fuel-rich C2–C3 laminar low pressure premixed flames shows fuel-dependent reaction pathways to benzene.

The propargyl radical is a key species for every reaction path found but its formation is fuel specific.

Benzene is produced in all flames predominantly by propargyl combination.

Acetylene addition to n-C4H5 is a secondary benzene source in the acetylene and the mixed acetylene–propene flame.

In the propene flame the C6H7 decomposition is a minor secondary source of benzene.

The experimentally found early formation of C6H8 in the propene flame is considered in the model.

Benzene formation via C3H5 + C3H3 is identified as a possible benzene source in propene flames.