Formation and consumption of single-ring aromatic hydrocarbons and their precursors in premixed acetylene, ethylene and benzene flames

View Article Online / Journal Homepage / Table of Contents for this issue

Formation and consumption of single-ring aromatic hydrocarbons and

their precursors in premixed acetylene, ethylene and benzene ﬂamesy

Henning Richter* and Jack B. Howard

Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge,

MA 02139-4307, USA. E-mail: richter@mit.edu

Received 6th November 2001, Accepted 12th March 2002

First published as an Advance Article on the web 30th April 2002

Kinetic modeling is becoming a powerful tool for the quantitative description of combustion processes covering

diﬀerent fuels and large ranges of temperature, pressure and equivalence ratio. In the present work, a reaction

mechanism which was developed initially for benzene oxidation, and included the formation of polycyclic

aromatic hydrocarbons was extended and tested for the combustion of acetylene and ethylene. Thermodynamic

and kinetic property data were updated. If available, data were taken from the recent literature. In addition,

density functional theory as well as ab initio computations on a CBS-Q and CBS-RAD level, partially published

previously, were carried out. Quantum Rice–Ramsperger–Kassel analysis was conducted in order to determine

pressure-dependent rate constants of chemically activated reactions. The model was developed and tested using

species concentration proﬁles reported in the literature from molecular beam mass-spectrometry measurements

in four unidimensional laminar premixed low-pressure ethylene, acetylene and benzene ﬂames at equivalence

ratios (f) of 0.75 and 1.9 (C H ), 2.4 (C H ) and 1.8 (C H ). Predictive capabilities of the model were found to

2

4

2

6

be at least fair and often good to excellent for the consumption of the reactants, the formation of the main

combustion products as well as the formation and depletion of major intermediates including radicals. Self-

combination of propargyl (C

formation pathway in both rich acetylene and ethylene ﬂames. In addition, reaction between vinylacetylene

) and vinyl radical (C

3 3

H ) followed by ring closure and rearrangement was the dominant benzene

(

C

4

H

4

2

H

3

) contributed to benzene formation in the f ¼ 1.9 ethylene ﬂame. Propargyl

formation and consumption pathways which involve reactions between acetylene, allene, propyne and singlet

and triplet methylene were assessed. Signiﬁcant overpredictions of phenoxy radicals indicate the necessity of

further investigation of the pressure and temperature dependence and the product distribution of phenyl

oxidation. The possible formation of benzoquinones, the ratio of the ortho and para isomers and their

degradation pathways are of particular interest.

In addition to the formation of undesired particulate matter,

combustion of hydrocarbons has long been used for large-scale

synthesis of carbon black and recently it has been shown sui-

1

. Introduction

1

0

Epidemiological studies have positively associated air pollu-

1

–3

11–15

tion with lung cancer and cardiopulmonary disease contri-

buting an estimated 6% of total mortality, half of which was

attributed to motorized traﬃc. Fine particles formed by

incomplete combustion, can be breathed more deeply into

the lungs and therefore are thought to pose a particularly

great risk to health. At least part of the hazardous health

eﬀects of atmospheric aerosols might be related to their asso-

ciation with polycyclic aromatic hydrocarbons (PAH), some

of which have been found to be mutagenic. A molecular bio-

logical pathway linking one PAH—benzo[a]pyrene—to human

lung cancer has been established.

The role of PAH in ﬂames as key intermediates in the incep-

tion and growth of soot which represents a major fraction of

atmospheric aerosols is widely accepted and the amounts

of diﬀerent PAH associated with diﬀerent aerosol size fractions

were measured. The minimization of the formation of soot in

combustion requires the control of chemical processes respon-

sible for the growth of larger and larger PAH as well as their

oxidation. Detailed kinetic modeling, the focus of the present

work, provides deeper insight into these processes and allows

the identiﬁcation of optimized operating conditions.

table for synthesis of fullerenes

other nanostructures.

and carbon nanotubes and

1

6–22

3

Optimization of combustion processes minimizing air pollu-

tion while maximizing energy release or yields of targeted

materials requires detailed assessment of the controlling chemi-

cal reaction pathways. Combustion involves complex competi-

tion between species formation and destruction which, without

oxidation, can lead to increasing large PAH and even soot par-

4

5

7,8

ticles. The amounts of PAH and soot in the ﬁnal combustion

products depend on conditions such as temperature, pressure,

fuel to oxygen ratio. Sooting tendencies of diﬀerent fuel types

increase from saturated aliphatics to unsaturated aliphatics

6

7

and to aromatics.

7

,8

The present work identiﬁes chemical reaction pathways

responsible for fuel consumption, e.g., bimolecular reactions

leading to more reactive radicals, unimolecular decay, or both.

Oxidation to CO, CO , H O and incompletely oxidized spe-

9

2

cies such as formaldehyde (CH O) and formation of inter-

2

mediates leading to PAH and soot were investigated.

Oxidation and breakdown can occur at nearly any stage of

combustion. For instance, consumption of initially formed

PAH and their precursors can suppress their contribution to

growth processes while acetylene formed in benzene destruc-

tion might play a signiﬁcant role in the formation and growth

of PAH and soot.

y Electronic supplementary information (ESI) available: Thermo-

dynamic and kinetic property data. See http://www.rsc.org/

suppdata/cp/b1/b110089k/

2

038

Phys. Chem. Chem. Phys., 2002, 4, 2038–2055

This journal is # The Owner Societies 2002

DOI: 10.1039/b110089k

View Article Online

A kinetic model which allows qualitative and quantitative

assessment of reactions responsible for fuel oxidation and for-

mation of PAH and their precursors has been developed and

applied to the combustion of acetylene, ethylene and benzene.

studied the formation of PAH and their precursors and com-

pared their model predictions to experimental results obtained

by other authors for acetylene and ethylene combustion at dif-

2

3,24,28,40

ferent conditions.

lene ﬂame of Westmoreland et al.,

in the present work. In a recent study, Pope and Miller

The low pressure premixed acety-

2

3,24

was also investigated

4

1

assessed benzene formation pathways in three fuel-rich low-

pressure premixed acetylene, ethylene and propene ﬂames.

Several researchers have developed kinetic models for fuel-

2

. Method

Experimental structures of unidimensional premixed acety-

lene,

4

2–44

2

3,24

25,26

27

and benzene low pressure ﬂames, con-

rich benzene combustion

and applied them to a premixed

ethylene

low-pressure benzene/oxygen/argon ﬂame, also studied in this

work, for which the structure was measured by Bittner and

sisting of temperature and mole fractions proﬁles of reactants,

products and intermediates were taken from the literature. The

comparison of experimental mole fractions to model predic-

tions allowed assessment of predictive capabilities of the

kinetic model. Main reaction pathways responsible for the for-

mation and consumption of key species were identiﬁed after

determination of the net rates of production, i.e., the contribu-

tion, of individual reactions.

Acetylene, ethylene and benzene are major components of

commonly used fuels, in addition acetylene and benzene are

thought to play key roles in the formation of PAH and soot.

For instance, the contribution of acetylene to PAH growth

involves its sequential addition to PAH radicals via the hydro-

2

7

Howard using MBMS. Recently, additional experimental

data for this ﬂame including concentration proﬁles of PAH

radicals became available from nozzle beam sampling with

radical scavenging and subsequent analysis by gas chromato-

4

5–47

graphy coupled to mass spectrometry (GC-MS).

The availability of experimental data is crucial for the devel-

opment of kinetic models. However, conclusions based on the

comparison between model predictions and experimental ﬂame

structures should also take into account possible uncertainties

in the experimental results. For instance, temperature mea-

surements, particularly in nearly-sooting conditions, are not

straight forward and the use of indirect calibration techniques

such as the estimation of ionization cross sections can induce a

signiﬁcant uncertainty.

2

8,29

gen-abstraction–acetylene-addition mechanism.

In the case

of non-aromatic fuels, the formation of the ﬁrst aromatic ring,

usually benzene, naphthalene or a derivative of these, is

thought to be crucial for the ‘‘kick-oﬀ’’ of the growth process.

Therefore, in this work, particular attention is paid to benzene

formation from aliphatic species as well as its consumption by

oxidation or subsequent PAH growth. Acetylene

5,26

lene ﬂames

of benzene formation while details of subsequent reaction

Kinetic model developed in present work

2

3,24

and ethy-

are particularly valuable for the investigation

The kinetic network developed in the present work was initi-

2

48

ally based on the study of Shandross et al. which focused

mainly on benzene oxidation. The mechanism was subse-

pathways, i.e., oxidation or PAH growth, can be assessed with

benzene as the initial fuel.

quently extended to PAH up to coronene (C24

and C₇₀fullerenes. Its predictions have been compared to

H

12) and C60

2

7

49

2

7

50

The experimental conditions of the four premixed unidimen-

sional ﬂat ﬂames investigated here are given in Table 1. Mole

fraction proﬁles of reactants, products and intermediate spe-

cies were determined for all four ﬂames by molecular beam

sampling coupled to mass spectrometry (MBMS). Thermocou-

ple measurements corrected for radiative heat loss gave tem-

peratures along the axis above the burner. The ability to

follow the evolution of radical concentrations makes MBMS

valuable for studying elementary chemical processes in hydro-

experimental data in nearly sooting and sooting premixed

benzene/oxygen/argon ﬂames. Thermodymamic data have

been revised using recent literature and density-functional the-

4

7

ory (DFT) calculations followed by vibrational analysis. The

potential energy surfaces of reactions of acetylene with phenyl

and 1-napththyl were explored by density-functional theory

and high-pressure-limit rate constants were deduced via transi-

tion state theory. Pressure- and temperature-dependent rate

constants which describe the formation of the diﬀerent pro-

ducts of these chemically-activated reactions were determined

by bimolecular QRRK (quantum Rice–Ramsperger–Kassel)

analysis using the modiﬁed strong collision approach. Details

of the techniques used in the present work for the determina-

tion of thermodynamic and kinetics properties were published

3

0

carbon combustion.

Detailed kinetic models are widely used in quantitative stu-

dies of combustion processes. Improved computational power

has permitted the development of increasingly complex reac-

3

1

tion networks. Experimental concentration proﬁles of stable

and radical species measured in fuel-rich acetylene/oxygen/

argon ﬂames were compared with model predictions.

5

1

recently. In addition, density functional theory coupled with

the use of isodesmic reactions and vibrational analysis served

for the determination of thermodynamic properties of a large

number of additional species included in the present mechan-

ism. For some selected species ab initio calculation on a

2

3,32

3

2,33

Routes leading to benzene have been suggested

formation of PAH described. Recently, Lindstedt and Ske-

and the

3

4

3

5

vis developed a detailed kinetic model including benzene for-

mation and oxidation and compared model predictions to

experimental data for six fuel-lean (normalized fuel/oxygen

or equivalence ratio f ¼ 0.12) to sooting (f ¼ 2.5) laminar,

premixed, low-pressure acetylene ﬂames.

5

2

CBS-Q level were performed. The complete set of standard

heats of formation, entropies and heat capacities used in the

present work is given in Table 2 (see Electronic Supplementary

5

3

Information (ESI)y) and is available on-line. In addition to

the above-mentioned reactions of acetylene with phenyl and

1-naphthyl, presssure-dependence of other chemically acti-

vated reactions was taken into account and rate constants

Kinetic schemes for fuel-rich ethylene combustion have been

developed and compared to experimental ﬂame structure data

3

6–38

39

measured at atmospheric pressure.

Wang and Frenklach

Table 1 Parameters of ﬂames investigated in the present work

ꢀ1

Flame

Fuel

Equivalence ratio (f)

Dilution/% of argon

Cold gas velocity/cm s

Pressure/Torr

Ref.

I

Acetylene

Ethylene

Benzene

2.4

0.75

1.9

1.8

5

50

20

30

20

23,24

25

26

II

56.91

50

30

62.5

50

III

IV

27,46

Phys. Chem. Chem. Phys., 2002, 4, 2038–2055

2039

View Article Online

a ﬂame depends on the local temperature and concentrations

of the reactants and products. Pressure has a signiﬁcant impact

on reactions with diﬀerent numbers of molecules of reactants

and products due to their pressure-depending equilibrium con-

stants and on unimolecular reactions whose rate coeﬃcients

increase with pressure.

Consumption of initial fuel: Acetylene

The analysis of the intial steps of acetylene consumption is

based on the modeling of the premixed low-pressure acety-

lene/oxygen/argon ﬂame (f ¼ 2.4, 5% argon, cold gas velo-

ꢀ

1

city v ¼ 50 cm s , 20 Torr) studied by Westmoreland et

2

3,24

al.

using MBMS (Flame I). Predicted mole fraction pro-

ﬁles of acetylene and oxygen are compared to experimental

data in Fig. 2. Key features such as a nearly constant acetylene

concentrations beyond 1.5 cm above the burner and the nearly

complete oxygen depletion in the postﬂame zone are predicted

correctly but the predicted consumptions of acetylene and oxy-

gen are slower than the data show. The acetylene mole fraction

remaining in the postﬂame zone is overpredicted by about

Fig. 1 Experimental temperature proﬁles used as input for kinetic

modeling: —: Flame I: fuel-rich acetylene/oxygen/argon ﬂame

ꢀ1

23,24

(

f ¼ 2.4, 5% argon, v ¼ 50 cm s , 20 Torr);

–ꢂꢂ–: Flame I:

increased temperature in the postﬂame zone (see text); ---: Flame II:

fuel-lean ethylene/oxygen/argon ﬂame (f ¼ 0.75, 56.91% argon,

ꢀ1

25

v ¼ 62.5 cm s , 30 Torr); ꢂ ꢂ ꢂ: Flame III: fuel-rich ethylene/oxy-

ꢀ

1

,

gen/argon ﬂame (f ¼ 1.9, 50.0% argon, v ¼ 62.5 cm

s

Torr); – –: Flame IV: nearly sooting benzene/oxygen/argon ﬂame

20

26

.

2

5%. Possible explanations might be uncertainties in the tem-

ꢀ1

27

(f ¼ 1.8, 30% argon, v ¼ 50 cm s , 20 Torr).

perature proﬁle used in the model calculation or possible per-

turbation of the ﬂame by the sampling probe. Temperature

measurement in near sooting or sooting ﬂames is particularly

challenging, e.g., due to possible deposits on the thermocouple

which can lead to a signiﬁcant decrease (up to several hundred

K) of the reading. The eﬀect of temperature on the predicted

acetylene proﬁle was tested. For instance, an increase of the

maximum temperature (at 0.8 cm from the burner) proﬁle by

for the low-pressure conditions pertinent to the present work

5

4,55

were determined by QRRK computations,

literature values were available. The mechanism developed in

if no suitable

5

3

the present study is given in Table 3y as well as on-line and

includes the sources of all kinetic data.

2

00 K to 2100 K and the maintain of this temperature in the

Testing of the model. All model calculations were performed

with the PREMIX ﬂame code, a part of the CHEMKIN soft-

postﬂame zone led to a good agreement of the model predic-

tions with the experimental acetylene mole fractions in the

burnt gas mixture (not shown). However, the latter tempera-

ture proﬁle is hypothetical and such large deviations from

measured data must be considered as unlikely.

Also, acetylene mole fractions predicted in the postﬂame

zone were found to be extremely sensitive to the thermody-

namic properties. A signiﬁcant improvement was also achieved

by using the thermodynamic properties of acetylene suggested

5

6–58

ware package.

Experimental temperature proﬁles, shown

in Fig. 1, were used for all computations. In addition to the

original experimental temperature proﬁle of the fuel-rich acet-

2

3,24

ylene ﬂame

(Flame I), a proﬁle modiﬁed in the postﬂame

zone is given and was used for the model predictions of this

ﬂame presented here. An increase of the temperature in the

postﬂame zone was based on inconsistencies in model predic-

tions of benzene and phenyl (see below) and possible soot

deposition on the thermocouple during the measurement.

Comparisons of model predictions with experimental mole

fraction proﬁles measured in the four investigated ﬂames

5

9

by Burcat and McBride instead of the NIST recommenda-

ꢁ

ꢀ1 60

tion for the heat of formation (DH_f¼ 54.19 kcal mol

)

in conjunction with the entropy and heat capacities determined

4

7,51

by means of density functional theory.

The diﬀerences

between the thermodynamic values suggested by Burcat and

(

Table 1) are discussed below. Major pathways of fuel con-

sumption and oxidation as well as of the formation of PAH

precursors were identiﬁed subsequent to the computation of

net rates of production of the species of interest using the post-

5

McBride and those used in our laboratory

9

47,51

are 0.16 kcal

ꢀ1

ꢁ

mol for the heat of formation (DH

f

) and ꢀ0.64, ꢀ0.77,

ꢀ

1

ꢀ1

K for the

) and heat capacities at 300, 400, 500, 600, 800,

5

8

ꢀ0.46, ꢀ0.31, ꢀ0.26, ꢀ0.16, 0.02, 0.20 cal mol

processor included in the CHEMKIN collection. Area

expansion of the unidimensional ﬂames was taken into account

in all computations using the corresponding expressions given

ꢁ

entropy (S

1

f

000, 1500 K, respectively. These discrepancies are close to

2

3,25–27

25

26

in the literature.

In the case of the lean and rich ethy-

lene ﬂames, area expansion was rescaled in order to be equal to

unity at the burner surface.

3

. Results

The initial consumption steps of the fuels acetylene, ethylene

and benzene, subsequent reaction steps leading to total oxida-

tion of the fuel, i.e., to CO, CO and H O, and competing

2 2

pathways responsible for PAH growth are discussed below.

The analysis of the rates production shows that the reactions

responsible for the consumption of the initial fuel, i.e., in the

present work acetylene, ethylene and benzene, depend signiﬁ-

cantly on the conditions of the combustion process, in particu-

lar temperature and pressure. Many reactions are at, or close

to, their equilibrium and reversibility was assumed for all reac-

tions included in the mechanism. Many PAH formation reac-

23,24

Fig. 2 Comparison between experimental mole fraction proﬁles

and model predictions in a fuel-rich acetylene/oxygen/argon ﬂame

3

4,39,47

ꢀ1

tions are tightly balanced, so that their contribution to

consumption or formation of a species at a given position in

(

f ¼ 2.4, 5% argon, v ¼ 50 cm s , 20 Torr); C

2

H

2

: S (experiment),

— (prediction); O : L (experiment), --- (prediction).

2

040

Phys. Chem. Chem. Phys., 2002, 4, 2038–2055

View Article Online

the limits of uncertainty imposed by the use of 14-parameter

ﬁts (‘‘NASA polynomia’’) for heats of formation, entropies

6

1

and heat capacities. The signiﬁcant impact of these discre-

pancies on the model prediction of acetylene emphasizes the

extreme importance of reliable thermodynamic data. In the

6

0

present work, the heat of formation recommended by NIST

and the entropy and heat capacities determined by means of

density functional theory followed by vibrational analysis

4

7,51

were used in order to maintain consistency of the thermody-

namic data set. Entropy and heat capacities of acetylene are

6

0

also included in the NIST database; the recommended values

are very similar to those used in present work. In addition, the

impact of the discussed variation of thermodynamic data of

acetylene on mole fraction proﬁles predicted by the kinetic

model has been assessed for other species such as HCO, only

minor eﬀects were observed.

The net consumption rate of acetylene, i.e., the sum of the

contributions of all reactions, increases rapidly close to the

burner and exhibits a sharp maximum at about 0.4 cm above

the burner. The net rate then decreases and reaches essentially

zero at about 2 cm from the burner. Simultaneously, the con-

tributions of all individual reactions also approach zero.

Therefore, the constant concentration of acetylene in the post-

ﬂame zone is not due to a balancing of acetylene formation

and destruction pathways but to a drastic slow down of all

reactions involving acetylene beyond the reaction zone. The

most important single reaction consuming acetylene is

2

3,24

Fig. 3 Comparison between experimental mole fraction proﬁles

and model predictions in a fuel-rich acetylene/oxygen/argon ﬂame

(

ꢀ1

f ¼ 2.4, 5% argon, v ¼ 50 cm s , 20 Torr); H: S (experiment), —

(

prediction); OH: L (experiment), --- (prediction); CH : ˘ (experi-

3

ment), ꢂ ꢂ ꢂ (prediction).

measure and are known with larger uncertainty. Accurate

modeling of radical formation and consumption is challenging

because of the enhanced sensitivity to thermodynamic and

kinetic properties and to the temperature proﬁle used. Mole

fraction proﬁles of H, OH, CH

puted for Flame I are compared to experimental data in Figs.

and 4. Abstraction of hydrogen atoms by H and OH is a key

3 2

, HCO, CO and CO com-

3

step in the formation of a large range of hydrocarbon radicals

while HCO is a key intermediate in the oxidation leading ulti-

mately to CO and CO . Methyl is the most abundant radical

C2H2 þ O Ð HCCO þ H:

The ketone radical HCCO is subsequently oxidized to CO, the

most important consumption reactions being

ð211Þ

2

3,24

identiﬁed in the rich acetylene ﬂame studied here.

Encouraging agreement between model predictions and experi-

mental mole fraction proﬁles can be seen in Figs. 3 and 4 for

OH, HCO and CH but a signiﬁcant overprediction of H, espe-

1

HCCO þ H Ð CH2 þ CO

ð165Þ

3

and

cially in the reaction zone, is observed.

HCCO þ O2 Ð 2CO þ OH:

Other signiﬁcant acetylene consumption pathways are, in

ð169Þ

Consumption of initial fuel: Ethylene

3

order of decreasing importance, oxidation to CH

2

and CO

25

26

A fuel-lean and a fuel-rich premixed low-pressure ethylene/

oxygen/argon ﬂame (Flames II and III, respectively; Table 1)

were studied and used for the determination of ethylene con-

sumption steps. The comparison of model predictions with

experimental data for both ﬂames allowed the reliability of

the present model to be assessed over a large range of equiva-

lence ratios. A major diﬀerence between these two ﬂames is the

1

via reaction (210)y and its reaction with CH to propargyl

2

3 3

(C H ) and H (reaction (321)). Both singlet and triplet methy-

1

lene ( CH₂and CH ) are easily oxidized to formaldehyde

3

2

(

formation via its recombination followed by rearrangement

2 2

CH O), CO and CO . Propargyl plays a key role in benzene

3

and will be discussed later in more detail.

Ethylene (C

next to acetylene, in the rich acetylene ﬂame studied here.

2 4

H ) is the second most abundant hydrocarbon,

fast and nearly complete conversion of ethylene to CO, CO

and H O in the fuel-lean ﬂame. Computed and experimental

2

3,24

2

Analysis of the rates of production showed that sequential

hydrogen addition to acetylene does not contribute signiﬁ-

cantly to ethylene formation. The major formation pathways

were found to be reactions

mole fraction proﬁles of ethylene and oxygen in Flame II are

given in Fig. 5 while proﬁles for the stable products, i.e.,

CH3 þ CH3 Ð C2H4 þ H2;

CH2 þ CH3 Ð C2H4 þ H

ð111Þ

ð235Þ

3

as well as unimolecular decay of n- and i-C

3

H

7

(reactions (436)

and (437)y). Ethylene consumption in this acetylene ﬂame

occurs mainly by hydrogen abstraction with H and OH form-

ing vinyl (C H ) (reactions (232) and (244)y).

2

3

Assessment of predictive capability. In the critical testing of

kinetic models for combustion chemistry by comparison of

model predicted species concentration proﬁles against experi-

mental data, proﬁles of radicals provide a more sensitive and

meaningful test than do the proﬁles of stable species. Radicals

are key intermediates in ﬂame chemistry and the ability to pre-

dict their concentrations at diﬀerent stages of combustion is

essential for the accurate prediction of the ﬂame properties

and behavior of interest. Compared to concentrations of stable

species, the concentrations of radicals are more diﬃcult to

23,24

Fig. 4 Comparison between experimental mole fraction proﬁles

and model predictions in a fuel-rich acetylene/oxygen/argon ﬂame

ꢀ1

(

f ¼ 2.4, 5% argon, v ¼ 50 cm s , 20 Torr); HCO: S (experiment, left

scale), — (prediction, left scale); CO: L (experiment, right scale), ---

prediction, right scale); CO

: ˘ (experiment, right scale), ꢂ ꢂ ꢂ (predic-

ght scale).

(

2

Phys. Chem. Chem. Phys., 2002, 4, 2038–2055

2041

View Article Online

26

Fig. 7 Comparison between experimental mole fraction proﬁles

and model predictions in a fuel-rich ethylene/oxygen/argon ﬂame

2

5

Fig. 5 Comparison between experimental mole fraction proﬁles

and model predictions in a fuel-lean ethylene/oxygen/argon ﬂame

ꢀ

1

(

—

f ¼ 1.9, 50.0% argon, v ¼ 62.5 cm s , 20 Torr); H: S (experiment),

ꢀ1

(

f ¼ 0.75, 56.91% argon, v ¼ 62.5 cm

s

,

30 Torr); ethylene

: L (experiment), --- (pre-

(prediction); OH: L (experiment), --- (prediction); O: ꢂ ꢂ ꢂ (predict-

(C

2

H

4

): S (experiment), — (prediction); O

2

ion).

diction).

2

6

2

5

Fig. 8 Comparison between experimental mole fraction proﬁles

and model predictions in a fuel-rich ethylene/oxygen/argon ﬂame

(

Fig. 6 Comparison between experimental mole fraction proﬁles

and model predictions in a fuel-lean ethylene/oxygen/argon ﬂame

(

ꢀ1

f ¼ 1.9, 50.0% argon, v ¼ 62.5 cm s , 20 Torr); CO: S (experiment),

(prediction); CO : L (experiment), --- (prediction); H O: ˘ (experi-

ment), ꢂ ꢂ ꢂ (prediction).

f ¼ 0.75, 56.91% argon, v ¼ 62.5 cm s , 30 Torr); CO: S (experi-

ment), — (prediction); CO : L (experiment), --- (prediction); H O:

(experiment), ꢂ ꢂ ꢂ (prediction).

—

2

˘

6

2

and a rate constant suggested by Tsang and Hampson was

2 2

CO, CO and H O are shown in Fig. 6. The insuﬃcient deple-

used for the channel leading to ethyl radicals, i.e.,

tion of CO predicted in the postﬂame zone is probably related

to a more than two-fold overpredition of hydrogen radicals, in

particular far from the burner, which shifts the equilibrium of

reaction

C2H4 þ H Ð C2H5

while the kinetic data derived by Knyazev et al. for

ð233Þ

ð232Þ

6

3

CO þ OH Ð CO2 þ H

ð39Þ

C2H4 þ H Ð C2H3 þ H2;

to the left and increases the CO concentrations. Another pos-

sible source of error might be an overestimation in the diﬃcult

ﬂame temperature measurement, discussed above. The use of

overestimated temperatures in the model computation for

Flame II could also explain the slightly too fast consumption

of ethylene and oxygen (Fig. 5), the overprediction of OH

based on experimental and ab initio computational results,

were used for hydrogen abstraction. Comparison of the main

reaction pathways responsible for ethylene consumption in a

fuel-lean and a fuel-rich ﬂame reveals signiﬁcant diﬀerences.

In the rich ethylene ﬂame, the reaction zone extends to ꢃ1.5

cm above the burner while in the lean ﬂame the net rate of pro-

duction of ethylene and the contribution of individual reac-

tions reach essentially zero at about 0.5 cm from the burner.

In the lean ﬂame, hydrogen addition leading to ethyl radi-

cals (reaction (233)) is the dominant consumption pathway

close to the burner and is responsible for a pronounced local

minimum of the net rate of production of ethylene at a height

above the burner of about 0.1 cm (Fig. 9). At increasing

heights above the burner, the contribution of reaction (233)

to ethylene consumption decreases dramatically and becomes

an ethylene forming pathway with a local maximum of its

net rate of formation at about 0.14 cm. This behaviour can

be explained by the relatively strong temperature dependence

of the rate constant of hydrogen abstraction (reaction (232))

favoring ethyl formation at low temperatures. Reactions of

ethylene with H, O and OH radicals, i.e., reactions (233),

(243)–(245) and (284)y result in an unambiguous net consump-

(

by up to 50%) and the ꢃ20% underprediction of H

Good to very good agreements of model predictions with

the corresponding experimental data were achieved in Flame

2

O (Fig. 6).

2

6

III, the fuel-rich ethylene ﬂame. Experimental and predicted

mole fraction proﬁles are shown in Figs. 7 and 8 for H, OH,

2 2

H O, CO and CO . In addition, the model prediction for O-

atoms is added in Fig. 7; experimental data for this species

in this ﬂame are not available, probably due to its low concen-

tration and mass overlap with methane in mass spectrometry.

The identity of the main ethylene consumption pathways

depends strongly on the local ﬂame conditions and therefore

on the height above the burner and the equivalence ratio.

Reaction of ethylene with hydrogen radicals can occur via

two diﬀerent channels: (i) hydrogen addition yielding ethyl

radicals and (ii) hydrogen abstraction giving vinyl radicals

(

C H

2 3

) and H

2

. Both reactions have been studied extensively

2

042 Phys. Chem. Chem. Phys., 2002, 4, 2038–2055

View Article Online

25

Fig. 10 Comparison between experimental mole fraction proﬁles

and model predictions in a fuel-lean ethylene/oxygen/argon ﬂame

ꢀ1

(

f ¼ 0.75, 56.91% argon, v ¼ 62.5 cm s , 30 Torr); vinyl (C

S (experiment), — (prediction); ketenyl (HCCO): L (experiment), ---

prediction).

2 3

H ):

Fig. 9 Rates of production of ethylene computed from model predic-

tions in a fuel-lean ethylene/oxygen/argon ﬂame (f ¼ 0.75, 56.91%

(

ꢀ

1

25

argon, v ¼ 62.5 cm s , 30 Torr) —: total; —: C

2

H

5

+ O

+ OH

+ O

+ OH Ð

2

Ð

C

2

H

4

+ HO

2

;

+ H O; —:

ꢂ ꢂ ꢂ:

C

2

H

4

+ H

+ O

+ OH Ð C

Ð

C

2

H

5

;

—:

+ HCO; ---:

+ H O; ---: C

C

2

H

5

2

C

2

H

4

Ð

CH

3

C

2

H

4

CH

2

CHO + H; ꢂ ꢂ ꢂ: C

+ CH O.

2

H

4

2

H

3

2

H

4

diction of the peak value is observed. The most important

vinyl consumption reactions are its oxidation with molecular

oxygen via reaction

3

2

tion of ethylene as shown in Fig. 9. The relative contributions

of the diﬀerent reactions depend on the relative concentrations

of H, O and OH radicals as well as on the local, temperature-

dependent, rate constants. For instance, in the lean ﬂame,

hydrogen abstraction by H radicals (reaction (232)) contri-

butes about 20 times less to ethylene consumption than the

corresponding reaction with OH. However, under fuel-rich

conditions reaction (232) is the dominant pathway, reﬂecting

the large concentration of H relative to that of OH.

Also in the rich ﬂame, formation of ethyl radicals via reac-

tion (233) leads to a local maximum of the ethylene consump-

tion rate close to the burner but in contrast to the lean case the

reverse reaction forms only insigniﬁcant amounts of ethylene

at higher heights above the burner. Vinyl formation via reac-

tions (232) and

C2H3 þ O2 Ð CH2O þ HCO

and unimolecular hydrogen-loss leading to acetylene, i.e., reac-

ð224Þ

tion

C2H3 Ð C2H2 þ H:

Pressure and temperature dependence of reaction (213) have

ð213Þ

6

been assessed by Knyazev and Slagle based on experimental

data and master equation modeling. Their suggested rate con-

stant was conﬁrmed in the present work by means of a QRRK

analysis of the reverse reaction and is used in the kinetic model

2

developed here. Also, hydrogen abstraction by H, O, O , OH

and CH (reactions (214), (216), (225), (226) and (229)y) con-

3

tributes signiﬁcantly to vinyl consumption and acetylene for-

mation. Other non-negligible vinyl-consumption pathways

involve its oxidation by O and OH:

C2H4 þ OH Ð C2H3 þ H2O

ð244Þ

is the dominant ethylene consumption channel. In addition,

C2H3 þ O Ð CH2CO þ H;

C2H3 þ O Ð CO þ CH3;

C2H3 þ O Ð HCO þ CH2

ð215Þ

ð217Þ

ð223Þ

ethylene oxidation by oxygen radicals, i.e., via reactions

1

C2H4 þ O Ð CH3 þ CHO

ð243Þ

ð284Þ

and

C2H3 þ OH Ð CH3CHO:

ð301Þ

C2H4 þ O Ð CH2CHO þ H;

contributes signiﬁcantly to its depletion. Another ethylene con-

sumption pathway, identiﬁed by analysis of the rates of pro-

duction, is its reaction with methine (CH) leading to allene

Acetylene. Acetylene formation and consumption are well

predicted in the rich ethylene ﬂame (Fig. 11) while the peak

mole fraction is by ꢃthreefold underpredicted in the lean ethy-

lene ﬂame. The analysis of the rates of production showed the

above discussed vinyl consumption reactions to be the major

acetylene formation pathways. In the lean ﬂame, acetylene is

completely depleted beyond about 0.8 cm from the burner.

In the rich ﬂame a signiﬁcant mole fraction of about 1%

remains in the postﬂame zone. The major acetylene consump-

tion pathway in both ﬂames is its oxidation with O to the kete-

nyl radical HCCO: (reaction (211)). Subsequently, HCCO is

consumed by reaction with H and O radicals: reactions (165)

and

(

responding rate constant suitable for low-pressure conditions

H

2

CCCH

2

) and hydrogen radicals (reaction (358)y). The cor-

5

4,55

was determined in the present work by a QRRK analysis

using the measurement of Thiesemann et al. for the entrance

6

4

6

5

channel and the expression derived by Tsang and Walker for

the unimolecular loss of hydrogen from the C H adduct.

3

5

2 3

Vinyl. Vinyl (C H ) is the major acetylene precursor in both

ethylene ﬂames. Hydrogen abstraction from ethylene by H and

OH (reactions (232) and (244)) is its dominant formation route

in all the ﬂames, including those of acetylene and benzene,

investigated in this work. Comparison of the predicted vinyl

mole fraction proﬁle with experimental data (Fig. 10) shows

good agreement of the peak value in the lean ethylene ﬂame

HCCO þ O Ð 2CO þ H:

ð166Þ

Species corresponding to the molecular mass of HCCO were

identiﬁed in both lean and rich ethylene ﬂames. Comparison

of the model predictions with the experimental mole fraction

proﬁles gave an excellent agreement in the lean ﬂame (Fig.

10) but an approximately four-fold overprediction was

observed in the rich ethylene ﬂame. Reaction pathways identi-

(

Flame II). However, the predicted breath of the proﬁle is only

about 0.4 cm compared to ꢃ0.7 cm for the experimental one.

Shape and peak location are in good agreement for all other

ﬂames (Flames I, III and IV) but a three to ﬁve-fold overpre-

Phys. Chem. Chem. Phys., 2002, 4, 2038–2055

2043

View Article Online

2

6

Fig. 11 Comparison between experimental mole fraction proﬁles

and model predictions in a fuel-rich ethylene/oxygen/argon ﬂame

Fig. 12 Comparison between experimental mole fraction proﬁles and

model predictions in a nearly sooting benzene/oxygen/argon ﬂame

ꢀ1

(

f ¼ 1.9, 50.0% argon, v ¼ 62.5 cm s , 20 Torr); C

2

H

2

: S (experi-

: L (experiment, right

: ˘ (experiment, right scale),

(f ¼ 1.8, 30% argon, v ¼ 50 cm s , 20 Torr); benzene: S (experi-

27

2

7

27

ment, left scale), — (prediction, left scale); CH

3

ment, left scale), — (prediction, left scale); O : L (experiment, left

2

scale), --- (prediction, right scale); CH

ꢂ ꢂ ꢂ (prediction, right scale).

4

scale), --- (prediction, left scale); C

phenyl: G (experiment, right scale), ꢂ ꢂ ꢂ (prediction, right scale).

6

H

5

: ˘ (experiment, right scale);

46

ﬁed as being responsible for acetylene consumption are quali-

tatively, and in a large extent quantitatively, identical to those

discussed above for a premixed acetylene/oxygen/argon

2

3,24

ﬂame.

The absence of a contribution of reaction

1

CH2 þ C2H2 Ð C3H3 þ H

ð321Þ

to acetylene consumption in the fuel-lean ethylene ﬂame is con-

sistent with propargyl (C ) being a precursor of benzene

which is only observed in trace quantities in this ﬂame.

3 3

H

Methyl. Methyl is usually the most abundant hydrocarbon

radical in ﬂames and it has been suggested to contribute to

2

7

PAH formation. In the present work, at least satisfactory

agreement between model predictions and experimental data

were achieved for methyl and methane in the rich ethylene

ﬂame (Fig. 11) and in the benzene ﬂame. However, the model

exhibited insuﬃcient methane depletion in the postﬂame zone

of the rich acetylene ﬂame and a signiﬁcant underprediction

for both methyl and methane in the lean ethylene ﬂame.

2

7

Fig. 13 Comparison between experimental mole fraction proﬁles

and model predictions in a nearly sooting benzene/oxygen/argon

ꢀ1

ﬂame (f ¼ 1.8, 30% argon, v ¼ 50 cm s , 20 Torr); H: S (experi-

ment), — (prediction); OH: L (experiment), --- (prediction); O:

˘

(experiment), ꢂ ꢂ ꢂ (prediction).

due to the overlap with methane in mass spectrometric detec-

tion.

Using the rate constant determined experimentally by

Consumption of initial fuel: Benzene

Benzene consumption by oxidative degradation and PAH

growth reactions was investigated here by kinetic modeling

of a premixed low-pressure nearly sooting benzene/oxygen/

6

7

Madronich and Felder over the temperature range from

90 to 1410 K at subatmospheric pressure, hydrogen abstrac-

tion via

7

2

7,46

argon ﬂame

(Flame IV) for which a large body of experi-

mental data including mole fraction proﬁles of radical inter-

mediates is available. The use of benzene ﬂame data in

kinetic modeling facilitates the study of major benzene con-

sumption reactions because the concentrations of the involved

species are considerably higher than in ﬂames of other fuels.

The analysis of the net rates of benzene consumption in the

premixed benzene shows the importance of the formation of

relatively unstable benzene derivatives in a ﬁrst reaction step

and which are consumed subsequently. H, O and OH radicals

are crucial for this initial activation and the relative contribu-

tion of the diﬀerent pathways is expected to depend signiﬁ-

cantly on the mole fractions of these species. The model

C H þ OH Ð C H þ H O

ð647Þ

6

5

2

was found to be the most important benzene consumption

pathway close to the burner in the benzene ﬂame studied here.

6

7

Madronich and Felder also showed that formation of phenol

and hydrogen radicals did not contribute signiﬁcantly to the

6 6

overall reaction rate of C H + OH. Nevertheless, the possibi-

lity of this pathway should not be entirely excluded at other

conditions than those investigated by these authors, e.g., high

pressure. Therefore, reaction (652)

C6H6 þ OH Ð C6H5OH þ H

ð652Þ

has been included in the present model. Its rate constant was

estimated at conditions relevant for this work in a QRRK

computation using input parameters similar to those sug-

gested by Shandross et al. Nevertheless, reaction (652) does

not contribute signiﬁcantly to benzene consumption. Reaction

of benzene with O atoms to give phenoxy and H, i.e.,

predictions for benzene and O agree well with the experimen-

2

tal data up to a distance of about 0.8 cm from the burner sur-

face while farther from the burner the predicted benzene and

2

7

5

4

8

O

(

2

depletion is slightly slower than in the experimental proﬁles

Fig. 12). However, consistent with the experiments, both ben-

zene and oxygen are entirely consumed in the postﬂame zone.

Predictions of mole fraction proﬁles of H, O and OH are given

in Fig. 13 and exhibit excellent agreement with experimental

data. Experimental data for atomic oxygen are only available

in the late part of the reaction zone and the postﬂame zone

C6H6 þ O Ð C6H5O þ H;

ð646Þ

is another important consumption pathway up to ꢃ0.7 cm

from the burner, i.e., nearly simultaneously with reaction

2

044

Phys. Chem. Chem. Phys., 2002, 4, 2038–2055

View Article Online

6

647). Ko et al. investigated C

8

(

suggested a rate constant over a temperature range from 300

6

H

6

+ O experimentally and

of its chemistry has proven to be challenging to model accu-

rately.

4

2–44,48,49

6

8

to 1450 K which was used here. Ko et al. also showed evi-

dence that the O atom addition mechanism, i.e., formation

of C H O + H, dominated under the conditions of their study.

6 5

Nevertheless, the possibility of a more important contribution

of hydrogen abstraction via reaction (645),

The contributions of individual reactions to the net phenyl

production rate in the premixed benzene ﬂame show that phe-

nyl is formed nearly entirely from benzene via the hydrogen

abstraction reactions (647) and (644), discussed above. The

reverse of unimolecular hydrogen loss from phenyl to give ben-

zyne,

C6H6 þ O Ð C6H5 þ OH

under low-pressure conditions can not be excluded taking into

ð645Þ

C6H5 Ð benzyne þ H;

ð636Þ

6

8

account that Ko et al. collected all experimental data at pres-

sures of not less than 240 mbar. Overestimation of phenoxy

formation via reaction (646) at low pressure might be a pos-

sible explanation for the overprediction of this species in a

is a minor phenyl formation pathway below ꢃ0.65 cm above

the burner and beyond 1.1 cm while the contribution of this

reaction to phenyl consumption exhibits a maxium at ꢃ0.8

cm. Unimolecular decay of phenyl was studied extensively by

Madden et al. in an ab initio MO study followed by a RRKM

treatment. Their rate constants for the formation of benzyne

4

8

72

premixed 22 Torr H

The contribution of hydrogen abstraction with H atoms, i.e.,

2

/O

2

/Ar ﬂame seeded with benzene.

6 4 6 4

and linear C H (C H , 1,5-hexadiyn-3-ene) were used in the

present work. Hydrogen loss to give linear C H , i.e.,

6 4

C6H6 þ H Ð C6H5 þ H2; ð644Þ

was shown to be similar to that of hydrogen abstraction with

OH but was shifted by ꢃ2 mm towards the postﬂame zone.

The kinetic data used in the present work for reaction (644),

and all similiar hydrogen abstraction reactions from PAH,

C6H5 Ð C6H4 þ H;

was found to be a major phenyl consumption pathway with a

maxium at ꢃ0.85 cm from the burner. Phenyl isomerization to

1,3-dien-6-yne radicals, i.e., ring opening,

ð637Þ

6 5 2

have been deduced from the C H + H rate constant com-

6

9

puted by Mebel et al., using the thermodynamic properties

given in Table 2.y The resulting rate constant is in excellent

agreement with the high temperature data determined in a

C6H5ðLÞ Ð C6H5;

ð631Þ

also identiﬁed on the potential energy surface by Madden et

7

0

14

72

al., is the most important phenyl consumption reaction in

shock tube study by Kiefer et al., i.e., 2.5 ꢄ 10 exp(ꢀ16 000

3

ꢀ1 ꢀ1

s .

0

00 cal/RT) cm mol

At the end of the reaction zone, unimolecular decay of ben-

zene to propargyl (C ) via the reverse of

the later part of the reaction zone, i.e., between 0.8 and 1.3

cm from the burner. Its production rates show that linear

C H (1,3-dien-6-yne radicals) decays quickly to linear C H

4

3

H

3

6

5

6

(

reaction (624)y). The peak mole fraction of linear C H was

6

5

C3H3 þ C3H3 Ð C6H6;

is the most important benzene consumption reaction. Its rate is

maximum at ꢃ1 cm above the burner. Propargyl recombina-

tion has been discussed as an important if not dominant

ð638Þ

predicted to be more than three orders of magnitude smaller

than that of phenyl. Model predictions compared with the

2

7

C

6

H

4

mole fraction proﬁle of species measured by MBMS

show good agreement in shape and location but the peak value

is overpredicted ꢃ2.5 fold if the C is linear or ꢃthreefold if

it is benzyne. The predicted 1,3,5-hexatriyne (triacetylene,

3

route for the formation of the ﬁrst aromatic ring. Phenyl + H

as well as benzene (and fulvene) have been suggested as possi-

ble products. In a recent shock tube study of pyrolytic reac-

6 4

H

C

6

H

2

) peak is ꢃ80% higher than the corresponding experimen-

7

1

27

tions of propargyl radicals, linear and/or aromatic C

species were identiﬁed as main products while the formation

rate of C + H was less than 10% of the total rate. Benzene

6

H

6

tal value. Analysis of production rates shows 1,3,5-hexa-

triyne to be formed by sequential hydrogen release and

abstraction beginning with linear C

H

6 5

6 4

H while

depletion by unimolecular decomposition to propargyl is con-

sistent with these observations, but quantitative assessment of

its contribution is not yet feasible due to the lack of a reliable

pressure-dependent rate constant.

C4H2 þ C2H Ð C6H2 þ H

was identiﬁed as major consumption pathway. A similar ana-

lysis in the acetylene ﬂame, shows reaction (613) to be a major

ð613Þ

Benzene consumption pathways identiﬁed by analysis of

benzene net production rates in the rich acetylene (Flame I)

are similar to the ﬁndings from the benzene ﬂame. Hydrogen

abstraction via reaction (647) is dominant, due to its low acti-

vation energy, up to ꢃ0.4 cm above the burner while reaction

1

release/abstraction from linear C H is also signiﬁcant.

Other, minor, phenyl consumption pathways in the later

part of the reaction zone of the benzene ﬂame are the reverse

of reactions

,3,5-hexatriyne formation pathway while sequential hydrogen

6

4

(

644) becomes dominant farther from the burner. Reaction

n-C4H3 þ C2H2 Ð phenyl

ð627Þ

with OH to give phenol, i.e., reaction (652), does not contri-

bute signiﬁcantly to benzene depletion while the reverse reac-

tion even leads to benzene formation close to the burner.

Benzene formation in the acetylene ﬂame results mainly from

propargyl recombination. However, the reverse reaction is a

minor consumption pathway at the end of the reaction zone

with a maximum rate at ꢃ6–7 mm above the burner. Details

of benzene formation chemistry in both acetylene and ethylene

ﬂames is discussed below.

The analysis of benzene net production rates showed that

phenyl and to some extent phenoxy radicals are key intermedi-

ates in benzene depletion. The chemistry of formation and sub-

sequently depletion of phenoxy and phenol is discussed below.

and

i-C4H3 þ C2H3 Ð phenyl þ H:

ð628Þ

In their theoretical study, Madden et al. also identiﬁed this n-

72

4 3

C H formation pathway on the potential energy surface of

phenyl decomposition. The low-pressure rate constant for

73

the forward reaction suggested by Westmoreland et al. from

a study of single-ring aromatics formation by chemically acti-

vated reactions was used in the present work. Reaction (628)

was taken from Pope and Miller. Recombination of phenyl

41

with H to benzene, i.e.,

C6H5 þ H Ð C6H6

ð634Þ

was found to be only a minor phenyl consumption pathway

and that in all three rich ﬂames investigated here. Depending

on the local conditions, reaction (634) contributed at certain

heights above the burner even to phenyl formation. Recombi-

Phenyl: A key intermediate in benzene depletion

Phenyl represents a bifurcation between further oxidation, i.e.,

decay to smaller species and growth to PAH. The complexity

Phys. Chem. Chem. Phys., 2002, 4, 2038–2055

2045

View Article Online

reaction with molecular oxygen, particularly the relative

contributions of pathways leading to phenoxy, ortho- and

eventually p-benzoquinone at diﬀerent pressures and tempera-

tures, are not completely understood.

The analysis of phenyl formation in the rich acetylene and

ethylene ﬂames, Flames I and III, conﬁrmed the above quali-

tative picture gained from Flame IV.

nation of phenyl and hydrogen to benzene competes with

hydrogen abstraction leading to benzyne and molecular

hydrogen:

C6H5 þ H Ð benzyne þ H2:

Low pressure rate constants determined for both channels by

ð635Þ

7

4

Mebel et al. in an ab initio MO/RRKM study were used

here. However, in all the ﬂames studied here, reaction (635)

was found not to contribute signiﬁcantly to phenyl decomposi-

tion.

Phenol and phenoxy. The above discussion of benzene and

phenyl consumption identiﬁes phenoxy radical (C

major intermediate. Analysis of the contributions of individual

reactions to C O production showed reactions (629) and

646) to be dominant in the benzene ﬂame. Unimolecular

6 5

H O) as a

In the early part of the reaction zone, phenyl consumption

occurred mainly by reaction with molecular oxygen,

6 5

H

(

C6H5 þ O2 Ð C6H5O þ O:

ð629Þ

decay to cyclopentadienyl and CO via

The rate constant of this reaction is pressure-dependent parti-

cularly at higher temperatures. Kinetics for reaction (629) were

C6H5O Ð C5H5 þ CO

was found to be the major consumption pathway using the rate

ð668Þ

5

determined by QRRK computations using input parameters

5

7

5

from the supporting information of DiNaro et al. Formation

of benzoquinone (C ) + H via phenyl oxidation has been

suggested from a shock tube study by Frank et al. Tan and

81

constant of Lin and Lin from a shock tube study by covering

6 4 2

H O

pressure and temperature ranges of 0.4–0.9 atm and between

1

7

6

000–1580 K. In the early part of the reaction zone, with a

maximum reaction rate at ꢃ5.5 mm from the burner, recombi-

4

Frank included this pathway in their reaction mechanism

used for modeling benzene oxidation in the same benzene

ﬂame studied here and observed a substantial reduction of

nation of phenoxy with H to give phenol, i.e.,

4

2,43,49

C6H5O þ H Ð C6H5OH;

ð667Þ

the phenyl overprediction reported by other authors.

However, relative to the experimental data, the maximum

contributes signiﬁcantly to phenoxy consumption while the

reverse reaction was found to be relatively unimportant in phe-

noxy formation between 7 and 11 mm.

7

6

was ꢃ2 mm closer to the burner. Frank et al. assumed the

major product to be p-benzoquinone rather than its ortho iso-

2

mer due to less stress in a transition state with a 1,4-O bridge.

Study of rates of production of phenol reveal reaction (667)

to be the dominant formation pathway while, in agreement

with the analysis of phenoxy formation and consumption, it

contributes to phenol consumption in the later part of the reac-

tion zone with a maximum reaction rate at ꢃ8 mm from the

burner. In this zone of the ﬂame, reaction (652) was found

to be the only signiﬁcant phenol forming reaction. Besides

hydrogen abstraction with OH,

7

Chai and Pfeﬀerle, using on-line mass spectrometry to study

benzene oxidation in a well-mixed reactor, detected trace

amounts of 108 u species which they attributed to benzoqui-

nones. The observation of no or only traces of benzoquinones

in another experimental and kinetic modeling study of benzene

7

8

oxidation led Alzueta et al. to conclude that o-benzoquinone

formation should be the most likely product channel. How-

ever, p-benzoquinone would be expected to build up to consid-

7

9

erable concentrations because of its thermal stability.

Similar to the suggestion of Alzueta et al.,

C6H5OH þ OH Ð C6H5O þ H2O;

unimolecular decomposition to cyclopentadiene (C

CO, i.e.,

ð680Þ

) and

7

8

H

5 6

C6H5 þ O2 Ð o-C6H4O2 þ H

ð684Þ

was included in the model developed in the present work and

was found to be the second most important phenyl consump-

tion pathway in the early part of the reaction zone. Fast decay

of o-benzoquinone to cylcopentadienone via

C6H5OH Ð C5H6 þ CO;

was the only important phenol consumption pathway identi-

ð671Þ

8

2

ﬁed. The rate constant deduced by Horn et al. in a high-tem-

perature shock tube experiment was used in the present work.

Comparison in Fig. 14 of the model prediction with the experi-

mental C H OH and C H proﬁles in Flame IV is given and

o-C6H4O2 Ð C5H4O þ CO

ð687Þ

6

5

6

led to the prediction of concentrations consistent with experi-

mental data on 108 u species, mass originally assigned to

shows satisfactory agreement. The slight shift of the predicted

2

7

C H O, but with a proﬁle shifted by ꢃ1.5 mm towards the

7

8

burner. The only pathway in the present model leading to p-

benzoquinone is reaction of atomic oxygen with phenoxy radi-

cals,

C6H5O þ O Ð p-C6H4O2 þ H:

ð686Þ

This reaction and a corresponding pathway to o-benzoqui-

none,

C6H5O þ O Ð o-C6H4O2 þ H

ð685Þ

8

were suggested by Lin and Mebel based on an ab initio MO

0

ꢀ

1

study. p-Benzoquinone is more stable by about 4 kcal mol

than its ortho-isomer and its consumption was found to be sig-

niﬁcantly slower. The model prediction of p-benzoquinone

gave a more than threefold higher peak concentration than

that of its ortho-isomer in Flame IV.

Fig. 14 Comparison between experimental mole fraction proﬁles and

model predictions in a nearly sooting benzene/oxygen/argon ﬂame

The comparison between model predictions and experimen-

tal data for phenyl in the benzene ﬂame, given in Fig. 12,

shows excellent agreement with two independent experimental

ꢀ1

(

f ¼ 1.8, 30% argon, v ¼ 50 cm s , 20 Torr) C

6

H

5

OH: S (experi-

27

46

ment, left scale); phenol: L (experiment, left scale), — (prediction,

left scale); C

scale).

27

5

H

6

: ˘ (experiment, right scale), --- (prediction, right

2

7,46

measurements.

However, details of phenyl depletion by

2

046 Phys. Chem. Chem. Phys., 2002, 4, 2038–2055

View Article Online

proﬁles towards the burner is consistent with that observed for

phenyl (Fig. 12) and might be attributable to imperfections in

the superposition of mole fraction proﬁles with the tempera-

ture proﬁle used in the computation.

for formation of cyclopentadienyl and CO

between phenoxy and O radicals was found in a ﬂow reactor

study by Buth et al. Mass spectrometric analysis of the reac-

tion products subsequent to molecular beam sampling and

electron impact ionization led to the detection of mass peaks

2

via reaction

8

7

Despite the good reproduction by the model of major inter-

mediates such as phenyl, phenol and cyclopentadiene, unre-

solved questions concerning phenoxy formation and

consumption persist. No experimental phenoxy proﬁle is avail-

assigned to p-benzoquinone (C

ever, as CO was only a minor component, representing less

than 15%, relative to C

quinone formation to be the major pathway of C

while a minor reaction channel would lead to C

Based on the overall rate constant for C

6 4 2 5 5 2

H O ), C H and CO . How-

2

8

7

6

H

4

O

2

, Buth et al. concluded benzo-

O + O

+ CO

O + O of

4

5

able for the ﬂames studied here but Hausmann et al. mea-

sured a phenoxy proﬁle with a peak mole fraction of

6 5

H

5

H

5

2

.

ꢀ6

1

ent concentration, identical to Flame IV. The model prediction

.5 ꢄ 10 in a benzene ﬂame under conditions, except for dilu-

6 5

H

1

4

3

ꢀ1 ꢀ1

87

1.68 ꢅ 0.35 ꢄ 10 cm mol

s

measured by Buth et al.

at room temperature and low pressure (1–5 mbar), Alzueta

ꢀ4

of a phenoxy peak value exceeding 4 ꢄ 10 for Flame IV

exceeds the Hausmann et al. value by more than two orders

of magnitude which cannot be attributed to the dilution and

associated temperature proﬁle diﬀerences. In addition, a signif-

icant overprediction of phenoxy radicals was also reported by

7

8

13

3

ꢀ1 ꢀ1

s to

both the formation of o- and p-benzoquinone (reactions

et al. assigned rate constants of 8.5 ꢄ 10 cm mol

1

3

ꢀ1 ꢀ1

s to

(685) and (686)) and k ¼ 1.0 ꢄ 10 cm mol

C6H5O þ O Ð C5H5 þ CO2:

4

8

ð670Þ

2 2

Shandross et al. for a rich H /O /Ar ﬂame with benzene

7

8

additive in which phenoxy concentration measured by MBMS

was close to the detection limit.

The rate constants suggested by Alzueta et al. are included in

the reaction mechanism developed in the present study. The

present analysis of the rates of production of phenoxy in the

benzene ﬂame showed the formation of o- and p-benzoquinone

via reaction with O radicals (reactions (685) and (686)) to be

minor and reaction (670) to be insigniﬁcant. Additional work

is needed to attain adequate understanding of phenoxy forma-

tion and consumption chemistry.

Reasons for the phenoxy overprediction might be an inade-

quate description of its formation, for instance the oxidation

of phenyl with O and the competion with other product chan-

2

nels such as ring opening or the formation of benzoquinones.

In recent years, several computational studies of potential

8

3

energy surfaces of phenyl + O

decomposition of phenylperoxy (C H OO) and phenoxy

2

and of the unimolecular

8

4

85

6

5

have been conducted. At lower temperatures the formation

of a dioxiranyl radical

Five-membered ring species: cyclopentadienyl, cyclopentadiene

and cyclopentadienone

Five-membered ring species are important intermediates in the

degradation of benzene, phenyl and their oxidation products,

in particular of phenoxy radicals. Recombination of cyclopen-

tadienyl radicals followed by rearrangement and hydrogen loss

with both O atoms bonded to the same carbon, followed by

further decomposition was identiﬁed to be lowest energy path-

way subsequent to the formation of the initital phenylperoxy

appears likely to play a signiﬁcant role in naphthalene forma-

36,37,47,49,88,89

tion under certain conditions.

formation and consumption of cyclic and aliphatic C species

Pathways for the

8

3

5

adduct. However, scission of the O–O bond to form phenoxy

and O radicals was found to be favorable at temperatures rele-

vant to the present work, and at temperatures greater than

about 1200 K direct formation of phenoxy and O without pas-

were analyzed in this study in the acetylene and benzene ﬂames

in which the corresponding experimental mole fraction proﬁles

are available.

8

3

sing through C

not identiﬁed in the investigation of potential energy surfaces

6 5

H OO was suggested. Benzoquinones were

Cyclopentadienyl. The rates of production of cyclopentadie-

8

3

nyl (C

(

5 5

H ) show unimolecular decomposition of phenoxy

reaction (668)) to be the dominant formation pathway in

of either phenyl + O

nylperoxy. The formation of this species via phenyl oxidation

2

or unimolecular decomposition of phe-

8

4

the benzene ﬂame. The only other signiﬁcant contribution to

cyclopentadienyl formation in this ﬂame results from hydrogen

abstraction from cyclopentadiene:

therefore remains questionable, while phenoxy can be consid-

ered a major product of the phenyl + O reaction. However,

a detailed investigation of the pressure and temperature depen-

dence of the rate constants describing diﬀerent product chan-

nels will be necessary for a quantitative assessment.

2

C5H6 þ H Ð C5H5 þ H2:

Recombination with hydrogen atoms, i.e.,

C5H5 þ H Ð C5H6;

ð590Þ

The potential energy surface including transition states of

thermal decomposition of phenoxy radicals was explored by

means of ab initio calculations. In qualitative agreement with

8

5

ð573Þ

8

1

Lin and Lin, cyclopentadienyl and CO were identiﬁed as ﬁnal

products. Bimolecular reaction of phenoxy with oxygen radi-

was identiﬁed as the major cyclopentadienyl depletion path-

way in the ﬁrst part of the reaction zone with a maximum con-

sumption rate at about 6 mm from the burner. Closer to the

end of the reaction zone, at about 8.5 mm, the reverse reaction

becomes a minor cyclopentadienyl formation route.

8

0

cals has been investigated thoroughly by Lin and Mebel in

an ab initio molecular orbital study and diﬀerent pathways

including formation of o- and p-benzoquinone, oxygen addi-

tion to a bridging position and to the terminal oxygen of the

Cyclopentadienyl oxidation to cyclopentadienone (C H O),

5

4

phenoxy radical with subsequent O

identiﬁed. Lin and Mebel also found a transition state allow-

ing for the insertion of a bridging oxygen into a C–C bond

2

elimination have been

i.e.,

8

0

C5H5 þ O Ð C5H4O þ H

ð581Þ

is a signiﬁcant contribution to C H depletion in the center of

6 5

forming the seven-membered ring structure C O(H )O but

5

concluded this pathway to be signiﬁcantly less favorable than

oxygen migration which yields a six-membered species with

oxygen atoms located at two adjacent carbons. Lin and

the reaction zone. The rate constant of reaction (581) was

determined by means of bimolecular QRRK using the input

55

90

parameters given by Zhong and Bozzelli. No pressure depen-

8

0

6 5

Mebel did not consider subsequent decay of C O(H )O to

dence of reaction (581) was observed. Unimolecular decompo-

sition of cyclopentadienyl is its most important consumption

pathway in the later part of reaction zone, particularly close

to the region of maximum temperature, a ﬁnding which is con-

8

cyclopentadienyl and CO as suggested by Carpenter based

6

2

on semiempirical PM3 computations but concluded that this

pathway deserved future study. Some experimental evidence

Phys. Chem. Chem. Phys., 2002, 4, 2038–2055

2047

View Article Online

sistent with the high activation energy of more than 100 kcal

mol of

ꢀ1

C5H5 Ð C3H3 þ C2H2:

ð574Þ

The potential energy surface of unimolecular cyclopentadienyl

reactions, including 1,2-hydrogen atom migration, ring open-

ing and decomposition, was studied thoroughly by Moskaleva

9

1

and Lin using ab initio MO calculations. Subsequently, they

determined kinetic properties for the diﬀerent reaction chan-

nels by means of multichannel RRKM theory for pressures

of 100 Torr, 1 atm and 10 atm covering a temperature range

from 1000 to 3000 K. Data given for 100 Torr are used in

the present work. Assessment of cyclopentadienyl production

rates identiﬁed hydrogen abstraction from cyclopentadienyl

to give cyclopentatriene via reactions

23,24

Fig. 16 Comparison between experimental mole fraction proﬁles

and model predictions in a fuel-rich acetylene/oxygen/argon ﬂame

C5H5 þ H Ð C5H4 þ H2

ð562Þ

ꢀ1

(

f ¼ 2.4, 5% argon, v ¼ 50 cm s , 20 Torr); cyclopentadienyl

C

5

H

5

): S (experiment, left scale), — (prediction, left scale); C

3 3

H :

and

˘

(experiment, right scale), ꢂ ꢂ ꢂ (prediction, right scale).

C5H5 þ OH Ð C5H4 þ H2O

ð563Þ

to be a minor cyclopentadienyl consumption route. Cyclopen-

tatriene is found to isomerize subsequently to 1,2-pentadiene-

experimental data were achieved for all of these species except

C H which is overpredicted by more than two orders of mag-

4

abstractions to give cyclopentatrienyl radicals (reactions

-yne (CH CCHCCH) (reaction (569)y). Also, hydrogen

2

5

2

nitude, probably due to the lack of consumption reactions.

Cyclic and acyclic C H and C H species were included in

(

553)–(555)y) are included in the model.

5

4

5

3

In the acetylene ﬂame, the reactions leading to cyclopenta-

the model but due to the nearly quantitative rearrangement

of cyclopentatriene to 1,2-pentadiene-4-yne (reaction (569)y),

predictions of the aliphatic species are signiﬁcantly larger than

those of their cyclic isomers. However, the current knowledge

dienyl radicals are qualitatively similar to those in the benzene

ﬂame but diﬀerences in the relative contributions of some reac-

tions are observed. For instance, hydrogen abstraction from

cylopentadiene by H and OH (reactions (590) and (593)y) con-

tributes more to cyclopentadienyl formation than does phe-

noxy decomposition (reaction (668)). Formation of

naphthalene via recombination of cyclopentadienyl radicals

of thermodynamic and kinetic properties of C

C H species and the related chemistry is scarce and further

5 2 5 3

H , C H and

5

4

developments might signiﬁcantly eﬀect the predicted relative

concentrations of cyclic and aliphatic species.

3

6,37,47,49,88,89

followed by rearrangement and hydrogen loss

played a role in cyclopentadienyl consumption only in the ben-

zene ﬂame.

Cyclopentadiene. Predicted cyclopentadiene mole fraction

proﬁles generally compare satisfactorily to experimental data.

However, a slight shift towards the burner in the benzene ﬂame

Comparison of predicted cyclopentadienyl mole fraction

proﬁles with experimental ones showed an overprediction of

ꢃ3.5 fold in the benzene ﬂame (Fig. 15) and an underpredic-

tion of ꢃ2-fold in the acetylene ﬂame (Fig. 16). Based on the

(

Fig. 14) and a ꢃ2-fold underprediction in the acetylene ﬂame

9

1

was observed. The net production rates of individual reactions

contributing to cyclopentadiene formation and consumption

indicate a close link to the cyclopentadienyl radical. Cyclopen-

tadiene formation and consumption were found to occur much

closer to the burner in the case of the acetylene ﬂame with

maxima and minima of the net production rate at about 0.8

mm and 2.5 mm compared to 6 mm and 8.5 mm in the benzene

ﬂame. In both ﬂames but especially in the acetylene one,

hydrogen addition to cyclopentadienyl, i.e., reaction (573) is

a major, and in ﬁrst part of the reaction zone even dominant,

formation pathway. The reverse direction of reaction (573) was

work of Moskaleva and Lin, rearrangement of cyclopenta-

dienyl to the acyclic

(

C

5

H

5

species 2-penten-4-ynyl

CH CHCHCCH) was included in the mechanism but in both

2

ﬂames its predicted mole fraction was more than two orders of

magnitude smaller than that of cyclopentadienyl. Mole frac-

tion proﬁles are also available for C H , C H , C H in the

5

2

5

3

5

4

5 3

acetylene ﬂame and for C H in the benzene ﬂame. Encoura-

ging agreements with less than four-fold diﬀerences from

9

2

studied experimentally by Roy et al. Forward and reverse

directions were expected to be signiﬁcantly pressure-depen-

dent, so the rate constant applied in the present work was

5

determined by bimolecular QRRK calculation using the

input parameters given by Zhong and Bozzelli. Unimolecular

9

0

8

2

decay of phenol, reaction (671), was found to be the major

cyclopentadiene formation pathway in the later part of the

reaction zone of the premixed benzene ﬂame with a maximum

net rate of reaction (671) at ꢃ7 mm from the burner and also

to play a signiﬁcant role in the acetylene ﬂame. Due to the

importance of direct cyclopentadiene formation via reaction

(

671) in the later part of the reaction zone, reaction (573) con-

tributes to cylopentadiene consumption at the end of the reac-

tion zone of the benzene ﬂame. Hydrogen abstraction from

cyclopentadiene by H, OH and O to give cyclopentadienyl

(reactions (590), (593) and (595)y) is the major C H consump-

Fig. 15 Comparison between experimental mole fraction proﬁles and

model predictions in a nearly sooting benzene/oxygen/argon ﬂame

ꢀ1

5

6

(

f ¼ 1.8, 30% argon, v ¼ 50 cm

s

,

20 Torr);

C H :

8 6

46

27

tion route. The kinetic properties of the most important single

S (experiment, left scale); phenylacetylene: L (experiment, left

9

2

27

reaction, reaction (590), were taken from Roy et al. who stu-

died the unimolecular decomposition of C behind reﬂected

scale), — (prediction, left scale); C

-

5

H

5

: ˘ (experiment, right scale),

-- (prediction, right scale).

5 6

H

2

048

Phys. Chem. Chem. Phys., 2002, 4, 2038–2055

View Article Online

shock waves. Bimolecular decomposition to give allyl (C

H CCHCH ) and C H , i.e.,

3

H

5

,

No unambiguous experimental identiﬁcation of cyclopenta-

dienone has been reported in the ﬂames studied in this work.

Comparison of the mole fraction proﬁle measured by MBMS

2

C5H6 þ H Ð C3H5 þ C2H2; ð592Þ

was identiﬁed as another major consumption pathway of

27

at 80 u in the benzene ﬂame with the cyclopentadienone pro-

ﬁle predicted by the present kinetic model showed similar

shape and peak location, but the predicted peak concentration

is ꢃ50-fold larger than the experimental value. Also, overlap

C

was found to be as important as that of reaction (590). Reac-

H

5 6

, particularly in the benzene ﬂame where its contribution

9

3

tion (592) was originally suggested by Dean, and the rate

6 8

with C H species such as 1,3- and 1,4-cyclohexadiene, which

9

constant given by Roy et al. was used in the present work.

2

5 4

have the same molecular mass as C H O cannot be excluded.

In addition to hydrogen-abstraction from the C-1 position

of cyclopentadiene, i.e., from a sp carbon atom, hydrogen-

There is clearly need for better understanding of formation

and/or depletion of cyclopentadienone.

3

2

abstraction from sp carbon atoms by H, OH, O and CH

3

was included in the model (reactions (591), (594), (596) and

consump-

tion pathway due to the signiﬁcantly higher C–H bond energy

Benzene formation

(

600)y) but was found to be only a negligible C

5 6

H

Literature review. Qualitative and quantitative understand-

ing of reaction pathways from aliphatic fuels to benzene and

its derivatives is essential for an adequate description of fuel-

rich combustion processes including formation of PAH and

soot as mentioned earlier. In addition to one-ring aromatics,

also species containing two rings, e.g., naphthalene, can also

be the ﬁrst aromatic compound formed in a combustion envir-

onment. As mentioned above, recombination of cyclopentadie-

3

compared to that involving the sp carbon atom.

Cyclopentadienone. The analysis of net rates of production

of cyclopentadienone (C

position of o-benzoquinone, i.e., reaction (687), suggested by

5 4

H O) showed unimolecular decom-

7

8

Alzueta et al., and cyclopentadienyl oxidation via reaction

581), discussed above, to be the dominant formation path-

ways in both the acetylene and benzene ﬂames. Other reaction

sequences such as association to C O,

3

6,37,47,49,88,89

(

nyl appears to form naphthalene.

However,

cyclopentadienyl radicals are formed via phenoxy decomposi-

tion and hence are strongly linked to six-membered ring

aromatics. Therefore, formation of naphthalene via cyclopen-

tadienyl recombination is expected to be determined essentially

by the eﬃciency of the formation of one-ring aromatic species,

particularly benzene and phenyl.

5 5

H

C5H5 þ O Ð C5H5O;

followed by hydrogen loss,

C5H5O Ð C5H4O þ H;

ð579Þ

ð582Þ

Mainly two classes of pathways have been considered for the

formation of one-ring aromatic species: (a) reactions of species

9

0

78

suggested by Zhong and Bozzelli and used by Alzueta et al.

have been implemented in the kinetic mechanism but did not

contribute signiﬁcantly to cyclopentadienone formation under

the conditions investigated here. Reaction with H to give n-

C H and CO (reaction (587)y) was a major cyclopentadienone

containing four carbon atoms (C

containing two-carbon atoms (C

between two molecules with three carbon atoms (C

4

-species) with compounds

-species) and (b) reactions

-species).

2

3

9

5

Cole et al. computed rates of benzene production from

mole fraction proﬁles determined by MBMS at diﬀerent

heights above the burner in a near-sooting premixed low pres-

sure 1,3-butadiene ﬂame. Based on kinetic data from the litera-

4

5

consumption pathway in both the acetylene and benzene

ﬂames while unimolecular decomposition to acetylene and CO,

9

5

C5H4O Ð 2C2H2 þ CO;

played a signiﬁcant role only in the benzene ﬂame. Both path-

ð589Þ

ture or from thermochemical estimations, Cole et al.

identiﬁed the reaction of the 1,3-butadienyl radical with acety-

lene followed by ring closure and hydrogen loss to be the

dominant pathway leading to benzene under the conditions

studied. Frenklach and Warnatz suggested in a modeling

study of a sooting low-pressure acetylene ﬂame, cyclization

7

ways were included in the model of Alzueta et al., with para-

8

meters for the latter reaction based on a computational study

3

4

9

4

94

of Wang and Brezinsky. Wang and Brezinsky explored the

potential energy surface of thermal cyclopentadienone decom-

position by means of ab initio MO calculations and subse-

quently determined kinetic parameters by RRKM theory for

6 5 6 5

of linear C H (n-C H ), the product of reaction between acet-

ylene and n-C H , to be the dominant pathway at the higher

4

3

ﬂame temperatures. Both

0

.1, 1, 10 and 100 atm. The potential energy surface showed

isomerization to bicyclo-C H O

5

4

n-C H þ C H Ð C H þ H

4

5

2

6

ð641Þ

and reaction (627) are included in the model developed in the

7

3

present work. Westmoreland et al. assessed benzene forma-

tion pathways in the same acetylene ﬂame investigated in the

present study.

2

3,24

They concluded that additions of n-C H

4 5

to be the inital step of cyclopentadienone decomposition, fol-

lowed by formation of a CO-substituted four-membered ring

species which decays to cyclobutadiene and CO. Similar to

and n-C to acetylene would be suﬃcient to account for

4

H

3

benzene formation in this ﬂame and suggested the correspond-

ing rate constants for diﬀerent temperatures and pressures

based on a bimolecular QRRK treatment for chemically acti-

vated reactions. A combined experimental/modeling study of

a rich premixed low pressure acetylene ﬂame led Bastin et

7

8

Alzueta et al., we assumed in the present work that cyclobu-

tadiene readily dissociates to acetylene and used the Wang and

9

4

Brezinsky 0.1 atm kinetic parameters which are pertinent to

the investigated conditions. Only reaction (589) was included

3

2

al.

to conclude that

C

4

H

5

+ C

2

H

2

! benzene + H and

in the ﬁnal version of the mechanism (Table 3y). C

bicyclo-C O followed by bicyclo-C

O Ð 2C

was found to be insigniﬁcant due to the potential energy of

5

H

4

O Ð

C H + C H ! phenyl can account for the formation of

4

3

2

5

H

4

5

H

4

H + CO

2 2

one-ring aromatic species. However, in the work of Bastin et

al. these reactions were treated as irreversible and n- and

3

2

ꢀ

1

bicyclo-C

5

H

4

O which is 50.6 kcal mol higher than that of

O. Another, minor, pathway of C O consumption

in the benzene ﬂame is its oxidation to vinylacetylene and

CO

i-isomers of both C H and C H were lumped. Miller and

4

3

4

5

9

4

33

Melius re-examined the ﬂame studied by Bastin et al. and

32

C

H

5 4

H

5 4

distinguished between n- and i-isomers for C H and C H

5

4

3

4

2

,

in their model. They included isomerization between the corre-

4 5

sponding isomers in their model with n-C and n-C H

4

H

3

C5H4O þ O Ð C4H4 þ CO2;

suggested by Alzueta et al.

ð588Þ

being thermodynamically signiﬁcantly less stable than their

isomers. Based on the resulting model predictions, Miller

7

8,79

Phys. Chem. Chem. Phys., 2002, 4, 2038–2055

2049

View Article Online

3

and Melius concluded that reactions (641) and (627) were not

suﬃcient to explain formation of the ﬁrst aromatic ring in acet-

ylene combustion under the conditions studied by Bastin et

3

2

al. even considering possible uncertainties in the thermody-

3

namic or kinetic property data. Miller and Melius investi-

gated in some detail the possible contribution of propargyl

(

with C

C

3

H

3

, H

2

CCCH) recombination and of reaction of propargyl

species and concluded that C + C

3

H

2

H

3 3

3

H

3

Ð

6 5

C H + H represents a potential pathway to one-aromatic ring

species in acetylene ﬂames. In the study of Miller and Melius,

3

the feasibility of this pathway was conﬁrmed by means of the

ab initio computational exploration of the potential energy sur-

faces between the diﬀerent initial recombination products and

benzene as well as phenyl + H. No prohibitive energy barrier

was found for the subsequent isomerization steps. This conclu-

2

3,24

Fig. 17 Comparison between experimental mole fraction proﬁles

and model predictions in a fuel-rich acetylene/oxygen/argon ﬂame

9

6

sion is consistent with the work of Alkemade and Homann

ꢀ1

(

f ¼ 2.4, 5% argon, v ¼ 50 cm s , 20 Torr); C

left scale), — (prediction, left scale); C : L (experiment, right scale),

-- (prediction, right scale).

6

H

6

: S (experiment,

who measured product distribution and the overall recombina-

tion rate constant in a temperature range of 623–673 K in a

ﬂow reactor operated at pressures between 300 and 600 kPa.

The relative product yield of benzene was up to 30% while lin-

ear C H species accounted for the remaining fraction. Recent

6

H

5

-

atomic hydrogen was measured on-line by optical absorption.

Formation of phenyl + H was found to represent less than 10%

of the overall reaction rate. Stable products comprised cyclic

6

3

5

36,37

modeling studies of premixed acetylene, ethylene

and

9

7

71

ethane ﬂames supported the dominant role of propargyl

recombination in the formation of the ﬁrst aromatic ring. Evi-

dence of contributions of additional pathways, namely, vinyl

and linear C

towards benzene in agreement with prior studies.

6 6

H species and the product distribution shifted

9

6,100

9

8

99

addition to vinylacetylene or 1,3-butadiene was recently

found in modeling of the premixed 1,3-butadiene ﬂame of Cole

Comparisons of model predictions for benzene and phenyl

to experimental mole fraction proﬁles in the acetylene ﬂame

are given in Fig. 17. Benzene is by ꢃtwo-fold overpredicted

while phenyl is underpredicted by ꢃ40%. The experimentally

observed increase of benzene in the postﬂame zone is overpre-

dicted (Fig. 17), both the benzene and phenyl predictions in

this zone are very sensitive to the temperature proﬁle and the

predicted benzene mole fraction in the postﬂame zone would

attain values exceeding the maximum in the reaction zone if

9

5

et al.

Despite complex sequences of rearrangement reactions iden-

3

tiﬁed by Miller and Melius and suggested by Alkemade and

3

9

6

Homann, most authors describe the formation of phenyl via

the global reaction C + C + H. However,

Ð C

included in their kinetic model ele-

mentary reaction steps forming diﬀerent C isomers fol-

lowed by isomerization to cyclic C H species, i.e., benzene

3

H

3

H

3 3

6 5

H

3

5,98

Lindstedt and Skevis

6 6

H

2

3,24

the originally published temperature proﬁle

was used. A

6

and fulvene. Referring to master equation calculations, Pope

temperature drop of more than 300 K between 1.2 cm and

3.8 cm above the burner in the original proﬁle is suspected

to be at least partially induced by soot deposits on the thermo-

couple used. Therefore, the temperature proﬁle was modiﬁed

in the postﬂame zone (Fig. 1) for all model calculations of

Flame I presented in this work.

4

1

and Miller accounted for fulvene and phenyl + H as product

channels of propargyl recombination with a nearly unity

branching ratio. In addition, isomerization between fulvene

and benzene was part of the reaction network they used for

modeling premixed acetylene, ethylene and propene ﬂames.

The formation of fulvene as product of propargyl recombina-

tion is consistent the potential energy surface determined by

The rates of production of benzene showed propargyl

recombination (reaction (638)) to be the dominant benzene

formation pathway in the acetylene ﬂame. However, the net

rate of this reaction shows its contribution to benzene deple-

tion in the later part of the reaction zone with a maximum

consumption rate at ꢃ0.7 cm from the burner before again

contributing to net benzene formation in the postﬂame zone.

Close to the burner, a minor contribution to benzene produc-

tion stems from reaction (652) with a maximum net rate at ꢃ2

mm from the burner and with phenol formed by addition of H

to phenoxy radicals (reaction (667)). Hydrogen abstraction,

i.e., reactions (644) and (647), was the major benzene con-

sumption route in Flame I while the importance of phenoxy

formation via reaction (646) was found to be secondary. The

picture of phenyl formation and consumption is similar to that

described above for the benzene ﬂame. The major diﬀerence is

the contribution of the isomerization of 1,3-dien-6-yne radicals

3

Miller and Melius but is in apparent contradiction with the

9

6

experimental ﬁndings of Alkemade and Homann as well as

a recent high-temperature shock tube study of propargyl reac-

7

1

tions under pyrolytic conditions where detection of fulvene

was not reported. However, in the pyrolysis of 1,5-hexadiyne,

a product of propargyl recombination, conducted by Stein

1

00

ꢀ3

et al. at atmospheric and very low (<10 Torr) pressures,

signiﬁcant amounts of fulvene were identiﬁed, in particular

at relative low temperatures and atmospheric pressure. A sig-

niﬁcant impact of the pressure was observed: at 1 atm, the ful-

vene/benzene ratio decreased steadily from 748 K to 833 K

(

the highest temperature reported for 1 atm) while at very

low pressure fulvene and benzene were formed simultaneously,

with a rate about 6 times faster for the latter species over the

whole range of investigated temperatures, i.e., up to 1073 K.

6 4 6 4

and hydrogen addition to linear C H (C H , 1,5-hexadiyn-3-

ene), i.e., the reverse of reactions (631) and (637), to phenyl

formation close to burner before becoming consumption path-

ways beyond about 3.5 mm from the burner surface. The ana-

lysis of the rates of production showed that in Flame I, both

Benzene formation in acetylene and ethylene ﬂames. In the

present work, the rate constant used for reaction (638) is

1

2

3

ꢀ1 ꢀ1

s

The direct formation of benzene instead of phe-

3

dies.

.0 ꢄ 10 cm mol

, consistent with several other stu-

3

6,37,97,98

linear C

(1,3,5-hexatriyne, triacetylene) resulting from reaction (613).

Sequential growth of polyynes such as C and C was

H and C H were formed sequentially from C H

6 5 6 4 6 2

nyl + H is based on a pyrolytic shock tube study of

7

propargyl reactions conducted by Scherer et al. Pyrolytic

1

H

6 2

4

2

7

reactions of propargyl radicals, produced by 3-iodo-propyne

(C H I) decomposition, behind reﬂected shock waves were stu-

3 3

previously considered as a route to soot formation but despite

its minor contribution to phenyl formation close to the burner,

died at 1000 to 1250 K. Residual gases were analyzed using gas

chromatographic techniques (GC-FID and GC-MS) and

no signiﬁcant role of C

matics could be identiﬁed in the present work.

6 2

H in the formation of one-ring aro-

2

050

Phys. Chem. Chem. Phys., 2002, 4, 2038–2055

View Article Online

ring aromatics in the acetylene and ethylene ﬂames. Addition

of vinyl to vinylacetylene contributes to some extent in the

ethylene combustion. However, because of remaining uncer-

tainties, particularly the thermodynamic properties of C

and C species, which might aﬀect the relative abundances

of the n- and i-isomers, the possibility of a signiﬁcant contribu-

tion of n-C or n-C to phenyl and benzene formation

4 3

H

4

H

5

4

H

3

4 5

H

cannot be deﬁnitively excluded. Also, additional work is

required to achieve suﬃcient quantitative understanding of

the formation of one-ring aromatic species via propargyl self-

combination under diﬀerent combustion conditions. A thor-

ough analysis, e.g., by means of a multi-well QRRK treat-

5

ment,

will be necessary to determine pressure and

temperature dependence of kinetics properties and product

distributions.

2

6

Fig. 18 Comparison between experimental mole fraction proﬁles

and model predictions in a fuel-rich ethylene/oxygen/argon ﬂame

(

ment, left scale), — (prediction, left scale); C

scale), --- (prediction, right scale).

In addition to reactions involving species containing two,

three or four carbon atoms, ﬁve-membered ring species might

also contribute to the formation of one-ring aromatic species.

A reaction sequence beginning with cyclopentadienyl and

methyl leading to fulvene and subsequently benzene has been

suggested based on exploration of the corresponding potential

ꢀ

1

f ¼ 1.9, 50.0% argon, v ¼ 62.5 cm s , 20 Torr); benzene: S (experi-

3 3

H : L (experiment, right

2

6

8

8,101

In the rich ethylene ﬂame (Flame III), the peak mole frac-

tion of benzene is ꢃ2-fold underpredicted and the computed

proﬁles have a signiﬁcantly narrower reaction zone than is

found experimentally (Fig. 18). Propargyl recombination

energy surface by means of ab initio computations.

zene formation pathway beginning with the reaction between

A ben-

3

8

ethylene and cyclopentadiene has been also suggested. For-

mation of one-ring aromatic species in pathways involving

ﬁve-membered ring species have not been investigated in the

present work, mostly due to the lack of available kinetics prop-

erty data, but deserve attention in future work.

(

reaction (638)) was identiﬁed as the dominant benzene forma-

tion pathway and the contribution of its reverse reaction to

benzene consumption was found to be signiﬁcantly more pro-

nounced than in Flame I. As in Flame I, hydrogen abstraction

by H and OH in reactions (644) and (647) was the other major

benzene consumption route. Besides propargyl recombination,

addition of vinyl to vinylacetylene, i.e.,

Propargyl, allene and propyne. Due to the importance of pro-

pargyl (H CCCH) for the formation of the ﬁrst aromatic ring

2

in the case of fuel-rich acetylene and ethylene combustion, its

formation and consumption were studied in some detail. Ther-

modynamic property values of propargyl and related species

such as C H , allene and propyne (Table 2y) were updated

C4H4 þ C2H3 Ð C6H6 þ H;

contributed to benzene formation in Flame III using the rate

ð639Þ

9

8

constant of Kubitza as cited by Lindstedt and Skevis. For-

mation and consumption pathways of phenyl in Flame III

are similar to those in Flame I, except for a contribution from

3

2

5

2,102

using ab initio computations on a CBS-Q level.

Its pro-

duction rates show that, in both the acetylene and ethylene

ﬂames, formation and consumption of propargyl were closely

4

1

reaction (628), suggested by Pope and Miller and which is

found to be a signiﬁcant phenyl formation pathway in

Flame III.

3 2 3 4

connected to C H as well as the two C H isomers, allene

(

agreement with Mebel et al., the structure HCCCH (propar-

CH CCH ) and propyne (methylacetylene, CH CCH). In

1

2

3

03

Reactions (641) and (627) are found to play no signiﬁcant

role in the formation of single-ring aromatics in both Flames

I and III; the reverse of reaction (627) even contributes to phe-

nyl depletion. However, meaningful assessment of relative con-

tributions of diﬀerent reactions requires adequate predictive

capability for all involved species. Propargyl predictions satis-

gylene) was assigned to the C H species. With propargylene

3

2

103

HCCCH) as product, Mebel et al. determined the bond dis-

(

sociation energy of propargyl (CH CCH) to be 94.1 kcal

2

ꢀ

1

mol in comparison to 96.3 kcal mol if vinylidenecarbene

ꢀ1

1

04

(

the triplet electronic state was found to have the lowest

H CCC) is formed. Similar to the results of Nguyen et al.,

2

factorily agree with experimental C

I and III and details of its formation and consumption are dis-

cussed below. The C and C isomers were undistinguish-

able with mass spectrometric analysis, so the MBMS measured

mole fraction proﬁles represent the sum of n- plus i-C and

. In the rich ethylene ﬂame, the predicted i-C

peak mole fractions are respectively ꢃ20- and 75-

and n-C . Comparison of the

sum of both isomers with the experimental proﬁles showed

ꢃ7- and 13-fold overpredictions for C and C , respec-

tively. The peak mole fraction of vinylacetylene (C H ) was

3 3

H proﬁles in both Flames

1

energy and was used in the present work. The heat of forma-

02

ꢀ

1

tion of 129.6 kcal mol , used in the present work (Table 2y),

H

4 3

4 5

H

ꢀ

1

is smaller by about 3 kcal mol

Nguyen et al. Eﬀects of potential uncertainties in thermody-

than that suggested by

1

04

4

H

3

namic property values are expected to be minimized by the

consistency of the computational technique used for all C

species.

n- plus i-C

and i-C

4

H

5

4 3

H

3 x

H

4

H

5

fold than those of n-C

4

H

3

4 5

H

The rates of production of propargyl in the premixed acety-

lene ﬂame showed H addition to C to be a major formation

pathway while hydrogen abstraction via

H

2

H

4 3

4 5

H

3

4

well predicted but the proﬁle was shifted towards the burner

by ꢃ1 mm relative to the experimental proﬁle.

C3H3 þ H Ð C3H2 þ H2

ð313Þ

In Flame I, shape and peak location of the computed viny-

lacetylene proﬁle are in good agreement with experiment but

the peak value is ꢃ3-fold underpredicted. The sum of n- and

i-C H is ꢃ3-fold overpredicted while the i-C H was ꢃ20-fold

was identiﬁed as the dominant propargyl consumption route.

H addition to C is included in the model as the reverse

of unimolecular dissociation of propargyl,

3

H

2

4

3

4

3

C3H3 Ð C3H2 þ H:

Unimolecular dissociation of propargyl was investigated theo-

ð312Þ

4 3

more abundant than n-C H . Unlike Flame III, the predicted

mole fraction of n-C H was only 4-fold smaller than that of

5

4

1

retically by Mebel et al. and the rate constant suggested by

03

i-C

4

H

5

and the sum of both was ꢃ10-fold underpredicted.

7

1

Scherer et al. was used in the present work. Formation of

Under the conditions of the present work and using a given

set of thermodynamic and kinetic property values, reaction

C H in the model occurs exclusively by hydrogen abstraction

3

2

(

638) is found to be the dominant pathway leading to single-

from propargyl by reaction with H and to some extent OH

Phys. Chem. Chem. Phys., 2002, 4, 2038–2055 2051

View Article Online

Hydrogen addition to propargyl, included as reverse reactions

of unimolecular decomposition of allene and propyne (reac-

tions (356) and (357)y), was found to play some role in propar-

gyl depletion. Due to the use of mass spectrometric analysis,

radicals (reactions (313) and (314)y). The reverse of reaction

312) was found to be the most important C H consumption

reaction, consistent with its importance in propargyl forma-

(

3

2

tion. In addition, oxidation with OH and O, i.e.,

no experimental distinction between C

ble. Model predictions for both allene and propyne are com-

pared in Fig. 19 to the proﬁle of the sum of all C species

3 4

H isomers was possi-

C3H2 þ O Ð C2H2 þ CO;

C3H2 þ OH Ð C2H2 þ HCO;

ð309Þ

ð310Þ

3 4

H

measured in Flame I. The predicted peak mole fraction of

and

4

cyclopropene is ꢃ10 -fold smaller than those of allene and pro-

pyne. The reverse reactions of unimolecular decomposition of

allene and propyne are major formation pathways of both spe-

cies, particularly close to the burner. In the later part of the

reaction zone, reactions

C3H2 þ OH Ð CHCHCHO

ð311Þ

contributed signiﬁcantly to C depletion. Mole fraction pro-

H

3 2

ﬁles computed for C and C H in Flame I are compared to

3 2

H

3 3

experimental data in Fig. 16 and Fig. 19. Overprediction of

both species in the postﬂame zone could indicate a need for

allene þ H Ð C2H2 þ CH3

CH3 þ C2H2 Ð propyne þ H

ð368Þ

ð383Þ

additional consumption pathways.

3 2

C H formation and

3 3

C H

consumption (and vice versa) are strongly coupled.

Reaction (321), with CH , and particularly at the end of

1

and

2

the reaction zone,

allene þ H Ð C3H5ðallyl; H2CCHCH2Þ;

are found to be major, even dominant, allene and propyne for-

mation pathways while they contribute net depletion of these

species close to the burner. Reaction

ð385Þ

3

CH2 þ C2H2 Ð C3H3 þ H

are major propargyl formation reactions. Based on the work of

ð322Þ

1

05

3

2 2 2

B o¨ hland et al., reaction of CH + C H is assumed to yield

directly propargyl and hydrogen atoms at the conditions perti-

nent to the present work. The rate constant determined by

these authors is used here. Addition of CH₂to C H is

C2H3 þ CH3 Ð C3H5 þ H

was identiﬁed to be mainly responsible for allyl (C H ,

ð395Þ

1

2

3 5

expected to form chemically activated cyclopropene followed

by isomerization to allene and propyne with subsequent hydro-

gen loss to give propargyl. In order to account for the pressure

H

2

CCHCH ) formation. Hydrogen abstraction reactions are

the major allene and propyne depletion routes.

The comparison of the model predictions with experimental

mole fraction proﬁles for C , C and C species in the

2

1

55

dependence of the CH

2

+ C

was performed and the resulting rate constants are included in

2

H

2

reaction, a QRRK analysis

3

H

2

H

3 3

3 4

H

rich ethylene ﬂame (Flame III) shows at least satisfactory

agreement (Figs. 18 and 20). In contrast to the acetylene ﬂame

(Flame I), allene was predicted to be more abundant than pro-

pyne. Similar to Flame I, propargyl depletion occurs by hydro-

gen abstraction leading to C

radicals which also consume C

formation was found to be signiﬁcantly diﬀerent from that in

Flame I. Reactions of CH and CH (reactions (321) and

2 2

(322)) were only secondary pathways despite the good agree-

ment of the predicted sum of these isomers with the experimen-

1

05

the model. The experimental rate constant of B o¨ hland et al.

for the entrance channel was used while kinetics parameters for

cyclopropene isomerization were taken from a combined

1

06

experimental and computational study of Karni et al.

The

H

3

2

and oxidation with O and OH

isomerization allene Ð propyne was found to proceed in a ser-

3 2

H

subsequently. Propargyl

1

06,107

ies of successive reactions via cyclopropene,

so the direct

1

3

reaction is not included. Kinetic parameters for unimolecular

hydrogen loss from allene and propyne were taken from the

7

1

shock tube study of Scherer et al. The QRRK results reveal

a signiﬁcant pressure dependence of the relative contributions

of the four product channels included, i.e., cyclopropene,

allene, propyne and propargyl + H. At pressures below 1

atm, propargyl + H is found to be the fastest pathway over

the whole temperature range investigated, i.e., from 300 to

tal CH proﬁle. Hydrogen abstractions by H and OH from

2

allene and propyne (reactions (379)–(382)y) are the dominant

propargyl formation pathways. Unimolecular hydrogen loss

from allene, reaction

allene Ð C3H3 þ H;

ð356Þ

2

100 K.

In addition, hydrogen abstraction from allene and propyne

is found to contribute to propargyl formation beyond ꢃ6.5

mm from the burner while the reverse reaction is a minor pro-

pargyl consumption pathway in the ﬁrst part of the reaction

by reaction with H and OH radicals, i.e., reactions (379)–

(

382)y, contributed to some extent to propargyl formation.

23,24

Fig. 19 Comparison between experimental mole fraction proﬁles

26

Fig. 20 Comparison between experimental mole fraction proﬁles

and model predictions in a fuel-rich ethylene/oxygen/argon ﬂame

and model predictions in a fuel-rich acetylene/oxygen/argon ﬂame

: S (experiment);

: ˘ (experi-

ꢀ

1

ꢀ1

(

f ¼ 2.4, 5% argon, v ¼ 50 cm s , 20 Torr); C

3

H

4

(f ¼ 1.9, 50.0% argon, v ¼ 62.5 cm s , 20 Torr); C

3

H

4

: S (experi-

allene: — (prediction); propyne: --- (prediction); C

ment), ꢂ ꢂ ꢂ (prediction).

3

H

2

ment); allene: — (prediction); propyne: --- (prediction); C

(experiment), ꢂ ꢂ ꢂ (prediction).

3

H

2

: ˘

2

052

Phys. Chem. Chem. Phys., 2002, 4, 2038–2055

View Article Online

zone. Major allene formation pathways are unimolecular pro-

penyl decay, i.e., the reverse of reaction (385) with allyl (C H ,

4. Conclusions

3

5

H

2

CCHCH

2

) formed by methyl addition to vinyl (reaction

Kinetic modeling is becoming an increasingly valuable tool for

combustion process optimization including minimization of

pollutants present in the exhaust gas. The meaningful applica-

tion of kinetic modeling requires suﬃcient predictive capability

under a large range of conditions. Therefore, reaction mechan-

isms need to be critically tested against experimental data from

well-deﬁned and easy to model ﬂames. Mole fraction proﬁles

measured in unidimensional laminar premixed low-pressure

ﬂames using MBMS are well suited for use in the development

and testing of kinetic models. Such data include not only reac-

tants and products but also radical intermediates. The data on

radicals allow particularly sensitive testing of the predictive

capability of kinetic models. Premixed ﬂames oﬀer a large

range of temperatures up to above 2000 K, pertinent to prac-

tical combustion devices and valuable for the assesssment of

reaction mechanisms. Measured temperature proﬁles are part

of the input necessary for modeling unidimensional premixed

ﬂames.

(

395)), as well as

CH þ C2H4 Ð C3H5:

ð384Þ

The rate constant for reaction (384) was determined by a

QRRK computation. Hydrogen abstractions (reactions (379)

and (380)y) leading to propargyl contribute signiﬁcantly to

allene depletion, as do also acetylene and methyl formation

via reaction (368) and isomerization to cyclopropene (reaction

(362)y). Cyclopropene, predicted only in trace quantities, is

found to isomerize nearly quantitatively to propyne (reaction

(

363)y), and this reaction is the dominant propyne formation

pathway. Unimolecular decomposition of propyne to propar-

gyl and H (reaction (357)y) contributes to its consumption

beyond about 8 mm from the burner but is a propyne forma-

tion pathway closer to the burner. Finally, H addition leading

to methyl and acetylene, the reverse of reaction (383), is found

to be a major propyne consumption pathway.

In the present work, published data on four premixed low

pressure ﬂames investigated by MBMS and thermocouple

pyrometry were used for the development and testing of a reac-

tion mechanism which describes the oxidation of the fuel as

well as the formation of single-ring aromatics. The fuels stu-

died are acetylene, ethylene, and benzene, thus including ali-

phatic and aromatic compounds pertinent to practical fuels.

Responsive to the interest in the formation of PAH and soot

in fuel-rich combustion, equivalence ratios of 2.4, 1.9 and 1.8

for ethylene, acetylene and benzene, respectively, were studied.

In addition, a lean (f ¼ 0.75) ethylene ﬂame was investigated

to assess the model under conditions of nearly complete fuel

oxidation.

The kinetic model developed in the present work takes into

account the recent literature. The thermodynamic and kinetic

property databases were updated. In cases where reliable data

were not available, density functional theory and in some case

CBS-Q and CBS-RAD computations including vibrational

analysis were conducted. QRRK analysis was used to deter-

mine pressure-dependent rate constants of chemically acti-

vated reactions.

Formation of PAH

Consumption of single-ring aromatics such as benzene and

phenyl and also cyclopentadienyl plays important roles in

molecular weight growth in fuel-rich combustion. Reaction

pathways leading to PAH containing up to six condensed rings

5

0

and C₆₀and C₇₀fullerenes in a benzene ﬂame (f ¼ 2.4) were

4

reported previously. A detailed analysis of the formation of

9

2

7,46

PAH with up to three aromatic rings in Flame IV

4

was pub-

7

46

lished recently. The data set of Benish will be used for the

further development of the reaction network presented in this

4

6

work, particularly its extension to larger species. Benish

detected and quantiﬁed radical species up to 201 u and stable

compounds up to 276 u using the nozzle beam sampling/radi-

cal scavenging method of Hausmann et al. with subsequent

analysis by GC-MS.

4

5

The reaction mechanism used in the present work (Table 3y)

includes the formation of PAH with up to three condensed

aromatic rings, i.e., phenanthrene and anthracene. Although

larger PAH are of general concern, in the ﬂames studied here

the concentrations of larger PAH not included in the model

are small enough relative to the concentrations of PAH that

are included, so that a signiﬁcant part of the error that would

be incurred by not including any PAH and their reactions with

smaller species is avoided. The peak mole fraction of phenyla-

cetylene, the largest species measured in the acetylene

Comparisons of model predictions with experimental mole

fraction proﬁles show in all four ﬂames the ability of the devel-

oped model to describe quantitatively the consumption of the

fuel and oxygen and the formation of major oxidation pro-

2

ducts such as CO, CO and water. H radical concentrations

are overpredicted in the acetylene and lean ethylene ﬂames

while good agreement is achieved in the rich ethylene and in

benzene ﬂames. Excellent agreement of model predictions with

experiment is observed for OH radicals, with a maximium

deviation of ꢃ40% overprediction in the lean ethylene ﬂame.

Discrepancies seen between experimental and computed mole

fractions, particularly for radical species, might be related

not only to imperfections of the model but also to uncertainties

of the experimental temperature proﬁle used as input for the

model computations. Formation and consumption of methyl,

the most abundant hydrocarbon radical, is reproduced cor-

rectly by the model but insuﬃcient depletion of methane is

observed in the postﬂame zone of the acetylene ﬂame. Vinyl

is ꢃ3-fold underpredicted in the acetylene ﬂame and up to 5-

fold overpredicted in the rich ethylene and benzene ﬂames.

This ﬁnding might indicate some lack of quantitative under-

standing of the initial phase of acetylene and ethylene con-

sumption, particularly the competition between addition and

abstraction reactions with hydrogen atoms. Also, uncertainties

on thermodynamic property data cannot be excluded.

2

3,24

ﬂame,

peak height and location is seen for both experimental pro-

is ꢃ3-fold overpredicted while good agreement of

2

7,46

ﬁles

determined by transition state theory and taking into account

in the benzene ﬂame (Fig. 15). Kinetics parameters

5

1

chemical activation are used in the description of phenylace-

tylene formation. Phenylacetylene mole fraction is overpre-

dicted in the postﬂame zone in both the acetylene and

benzene ﬂames, possibly indicating the need to include addi-

tional consumption pathways as those involved in PAH and

soot growth. Also, multiple-substituted phenylacetylene, e.g.,

1,2-, 1,3- or 1,4-diethynylbenzene, might be formed. Oxidation

of phenylacetylene by O radicals to give phenylcarbene

1

08

(

C

6

5

H CH), suggested by Eichholtz et al.

(reaction (740)y),

and subsequent reactions leading to benzyl (reaction (741)y)

and benzene (reactions (743) and (744)y) are included in the

present model. Thermodynamic properties of triplet phenyl-

carbene were determined using the CBS-RAD ab initio compu-

1

02,109

ꢁ

tational procedure.

A heat of formation (DH ) of 110.6

In general, the predictive capability of the model is found to

f

ꢀ

1

102

be at least satisfactory for most C

bons. However, C species are underpredicted in the acety-

lene ﬂame. Propargyl recombination is found to be the

1 2 3 4

, C , C and C hydrocar-

kcal mol was obtained. Finally, oxidation of the phenyla-

cetylene side chain by O radicals was found to have only a lim-

ited eﬀect on phenylacetylene proﬁles in the postﬂame zone.

4 5

H

Phys. Chem. Chem. Phys., 2002, 4, 2038–2055

2053

View Article Online

dominant benzene formation pathway in both acetylene and

rich ethylene ﬂames. The reaction between vinyl and vinylace-

tylene also contributes to benzene formation in ethylene com-

bustion. A detailed investigation of the kinetics of propargyl

self-combination including the temperature and pressure

dependence of the diﬀerent reaction products would be of

much interest for future work. The ratio between n- and i-iso-

7

8

9

B. S. Haynes and H. Gg. Wagner, Prog. Energy Combust. Sci.,

981, 7, 229–273.

H. Richter and J. B. Howard, Prog. Energy Combust. Sci., 2000,

6, 565–608.

J. O. Allen, N. M. Dookeran, K. A. Smith, A. F. Saroﬁm, K.

Taghizadeh and A. L. Laﬂeur, Environ. Sci. Technol., 1996, 30,

1023–1031.

1

2

10 J.-B. Donnet, R. C. Bansal and M.-J. Wang, Carbon Black:

Science and Technology, Dekker, New York, 2nd edn., 1993.

1

4 3 4 5

mers of C H and especially C H might be aﬀected by further

1

J. B. Howard, J. T. McKinnon, M. E. Johnson, Y. Makarovsy

and A. L. Laﬂeur, J. Phys. Chem., 1992, 96, 6657–6662.

J. B. Howard, Proc. Combust. Inst., 1992, 24, 933–946.

J. B. Howard, A. L. Laﬂeur, Y. Makarovsky, S. Mitra, C. J.

Pope and T. K. Yadav, Carbon, 1992, 30, 1183–1201.

improvements in thermodynamic property data; a change in

the contribution of these species to benzene formation could

result.

1

2

3

At least a qualitative understanding of benzene oxidation

has been achieved. Benzoquinones are shown to be possible

intermediates and their inclusion improved signiﬁcantly the

predictive capability of the model. However, the formation

of benzoquinones by oxidation of phenyl remains question-

able. In addition, the relative formation rates of o- and p-ben-

zoquinone and possible additional consumption pathways of

the latter isomer need to be addressed. Cyclopentadienyl radi-

cals are found to be formed in all ﬂames studied by unimole-

cular decay of phenoxy radicals. Unimolecular reaction to

propargyl and acetylene and oxidation to cyclopentadienone

are major cyclopentadienyl consumption pathways. A ꢃ4-fold

overprediction of cyclopentadienyl is observed in the benzene

ﬂame which might be related to uncertainties in phenoxy

14 H. Richter, A. Fonseca, S. C. Emberson, J.-M. Gilles, J. B.

Nagy, P. A. Thiry, R. Caudano and A. A. Lucas, J. Chim. Phys.,

1995, 92, 1272–1285.

H. Richter, A. J. Labrocca, W. J. Grieco, K. Taghizadeh, A. L.

1

5

Laﬂeur and J. B. Howard, J. Phys. Chem. B, 1997, 101, 1556–

560.

1

16 J. B. Howard, K. D. Chowdhury and J. B. Vander Sande, Nat-

ure, 1994, 370, 603.

7

1

K. D. Chowdhury, J. B. Howard and J. B. Vander Sande, J.

Mater. Res., 1996, 11, 341–347.

H. Richter, K. Hernadi, R. Caudano, A. Fonseca, H.-N.

Migeon, J. B. Nagy, S. Schneider, J. Vandooren and P. J. Van

Tiggelen, Carbon, 1996, 34, 427–429.

1

8

19 W. J. Grieco, J. B. Howard, L. C. Rainey and J. B. Vander

Sande, Carbon, 2000, 38, 597–614.

2

0

1

2

P. Hebgen, A. Goel, J. B. Howard, L. C. Rainey and J. B. Van-

der Sande, Proc. Combust. Inst., 2000, 28, 1397–1404.

R. L. Vander Wal and T. M. Ticich, Chem. Phys. Lett., 2001,

2

chemistry. Particularly, oxidation of phenyl by O and the

resulting product distribution at diﬀerent temperatures and

pressures need to be addressed in future work.

Finally, it can be concluded that at least satisfactory predic-

tive capability has been found for ethylene, acetylene and ben-

zene combustion at temperatures from near ambient up to

nearly 2100 K. QRRK analysis is found to be suitable for esti-

mating kinetics parameters of chemically activated reactions at

speciﬁc pressures. An atmospheric pressure version of the pre-

sent model is being developed and will be published later.

3

36, 24–32.

L. Yuan, K. Saito, W. Hu and Z. Chen, Chem. Phys. Lett., 2001,

346, 23–28.

23 P. R. Westmoreland, J. B. Howard and J. B. Longwell, Proc.

Combust. Inst., 1986, 21, 773–782.

2

3

4

5

6

7

8

9

0

P. R. Westmoreland, PhD Thesis, Massachusetts Institute of

Technology, Cambridge, MA, 1986.

A. Bhargava and P. R. Westmoreland, Combust. Flame, 1998,

1

15, 456–467.

A. Bhargava and P. R. Westmoreland, Combust. Flame, 1998,

13, 333–347.

J. D. Bittner and J. B. Howard, Proc. Combust. Inst., 1981, 18,

105–1116.

1

Acknowledgement

H. Bockhorn, F. Fetting and H. W. Wenz, Ber. Bunsen-Ges.

Phys. Chem., 1983, 87, 1067–1073.

M. Frenklach, D. W. Clary, W. C. Gardiner and S. E. Stein,

Proc. Combust. Inst., 1984, 20, 887–901.

U. Bonne, K. H. Homann and H. Gg. Wagner, Proc. Combust.

Inst., 1965, 10, 503–512.

This research was supported by the Chemical Sciences, Geos-

ciences and Biosciences Division, Oﬃce of Basic Energy

Sciences, Oﬃce of Energy Research, U.S. Department of

Energy under Grant DE-FGO2-84ER13282. We are grateful

to Dr R. Sumathi for providing thermodynamic properties

determined by CBS-Q and CBS-RAD computational techni-

ques, Prof. W. H. Green and Dr O. Mazyar for contributions

to the development of the thermodynamic and kinetic property

data sets, and Prof. J. W. Bozzelli for providing software and

valuable advice.

31 P. Lindstedt, Proc. Combust. Inst., 1998, 27, 269–285.

32 E. Bastin, J.-L. Delfau, M. Reuillon, C. Vovelle and J. Warnatz,

Proc. Combust. Inst., 1981, 22, 313–322.

3

4

J. A. Miller and C. F. Melius, Combust. Flame, 1992, 91, 21–39.

M. Frenklach and J. Warnatz, Combust. Sci. Technol., 1987, 51,

2

65–283.

R. P. Lindstedt and G. Skevis, Combust. Sci. Technol., 1997, 125,

3–137.

3

5

6

7

M. J. Castaldi, N. M. Marinov, C. F. Melius, J. Huang, S. M.

Senkan, W. J. Pitz and C. K. Westbrook, Proc. Combust. Inst.,

1996, 26, 693–702.

References

3

7

A. D’Anna and A. Violi, Proc. Combust. Inst., 1998, 27, 425–

433.

38 T. Faravelli, A. Goldaniga and E. Ranzi, Proc. Combust. Inst.,

1

D. W. Dockery, C. A. Pope, X. Xu, J. D. Spengler, J. H. Ware,

M. E. Fay, B. G. Ferris and F. E. Speizer, N. Engl. J. Med., 1993,

3

29, 1753–1759.

1998, 27, 1489–1495.

39 H. Wang and M. Frenklach, Combust. Flame, 1997, 110, 173–

221.

40 S. J. Harris, A. M. Weiner and R. J. Blint, Combust. Flame, 1988,

72, 91–109.

2

K. Siegmann and H. C. Siegmann, Molecular Precursor of Soot

and Quantiﬁcation of the Associated Health Risk, Current Pro-

blems in Condensed Matter, ed. Mor a´ n-L o´ pez, Plenum Press,

New York, 1998, pp. 143–160.

3

4

N. K u¨ nzli, R. Kaiser, S. Medina, M. Studnicka, O. Chanel, P.

Filliger, M. Herry, F. Horak, Jr., V. Puybonnieux-Texier, P.

Qu e´ nel, J. Schneider, R. Seethaler, J.-C. Vergnaud and H. Som-

mer, Lancet, 2000, 356, 795–801.

F. J. Miller, D. E. Gardner, J. A. Graham, R. E. Lee, Jr., W. E.

Wilson and J. D. Bachmann, J. Air Pollut. Control Assoc., 1979,

41 C. J. Pope and J. A. Miller, Proc. Combust. Inst., 2000, 28, 1519–

1527.

42 R. P. Lindstedt and G. Skevis, Combust. Flame, 1994, 99, 551–

561.

43 H.-Y. Zhang and J. T. McKinnon, Combust. Sci. Technol., 1995,

107, 261–300.

2

9, 610–615.

44 Y. Tan and P. Frank, Proc. Combust. Inst., 1996, 26, 677–684.

45 M. Hausmann, P. Hebgen and K.-H. Homann, Proc. Combust.

Inst., 1992, 24, 793–801.

46 T. G. Benish, , PhD thesis, Massachusetts Institute of Technol-

ogy, Cambridge, MA, 1999.

5

6

J. L. Durant, W. F. Busby, A. L. Laﬂeur, B. W. Penman and

C. L. Crespi, Mutat. Res., 1996, 371, 123–157.

M. F. Denissenko, A. Pao, M.-S. Tang and G. P. Pfeifer,

Science, 1996, 274, 430–432.

2

054

Phys. Chem. Chem. Phys., 2002, 4, 2038–2055

View Article Online

47

48

49

50

H. Richter, T. G. Benish, O. A. Mazyar, W. H. Green and J. B.

Howard, Proc. Combust. Inst., 2000, 28, 2609–2618.

R. A. Shandross, J. P. Longwell and J. B. Howard, Proc. Com-

bust. Inst., 1996, 26, 711–719.

H. Richter, W. J. Grieco and J. B. Howard, Combust. Flame,

1999, 119, 1–22.

W. J. Grieco, A. L. Laﬂeur, K. C. Swallow, H. Richter, K.

75 J. L. DiNaro, J. B. Howard, W. H. Green, J. W. Tester and J. W.

Bozzelli, J. Phys. Chem. A, 2000, 104, 10576–10586.

76 P. Frank, J. Herzler, T. Just and C. Wahl, Proc. Combust. Inst.,

1994, 25, 833–840.

77 Y. Chai and L. D. Pfeﬀerle, Fuel, 1998, 77, 313–320.

78 M. U. Alzueta, P. Glarborg and K. Dam-Johansen, Int. J. Chem.

Kinet., 2000, 32, 498–522.

Taghizadeh and J. B. Howard, Proc. Combust. Inst., 1998, 27,

1669–1675.

79 M. U. Alzueta, M. Oliva and P. Glarborg, Int. J. Chem. Kinet.,

1998, 30, 683–697.

5

1

2

H. Richter, O. Mazyar, R. Sumathy, W. H. Green, J. B. Howard

and J. W. Bozzelli, J. Phys. Chem. A, 2001, 105, 1561–1573.

R. Sumathi, H.-H. Carstensen and W. H. Green, J. Phys. Chem.

A, 2001, 105, 6910–6925.

http://web.mit.edu/richter/www/MITcomb.html.

A. M. Dean, J. W. Bozzelli and E. R. Ritter, Combust. Sci. Tech-

nol., 1991, 80, 63–85.

80 M. C. Lin and A. M. Mebel, J. Phys. Org. Chem., 1995, 8, 407–

420.

81 C.-Y. Lin and M. C. Lin, J. Phys. Chem., 1986, 90, 425–431.

82 C. Horn, K. Roy, P. Frank and T. Just, Proc. Combust. Inst.,

1998, 27, 321–328.

5

3

4

83 C. Barckholtz, M. J. Fadden and C. M. Hadad, J. Phys. Chem.

A, 1999, 103, 8108–8117.

5

6

A. Y. Chang, J. W. Bozzelli and A. M. Dean, Z. Phys. Chem.,

2000, 214, 1533–1568.

R. J. Kee, J. F. Grcar, M. D. Smooke and J. A. Miller, A FOR-

TRAN Program for Modeling Steady Laminar One-Dimensional

Premixed Flames, Sandia Report SAND85-8240, 1985.

R. J. Kee, F. M. Rupley and J. A. Miller, Chemkin-II: A Fortran

Chemical Kinetics Package for the Analysis of Gas-Phase Chemi-

cal Kinetics, Sandia Report SAND89-8009, Sandia National

Laboratory, Livermore, CA, 1989.

R. J. Kee, F. M. Rupley, J. A. Miller, M. E. Coltrin, J. F. Grcar,

E. Meeks, H. K. Moﬀat, A. E. Lutz, G. Dixon-Lewis, M. D.

Smooke, J. Warnatz, G. H. Evans, R. S. Larson, R. E. Mitchell,

L. R. Petzold, W. C. Reynolds, M. Caracotsios, W. E. Stewart,

P. Glarborg, C. Wang and O. Adigun, CHEMKIN Collection,

Release 3.6, Reaction Design, Inc., San Diego, CA, 2000.

A. Burcat and B. McBride, 1997 Ideal Gas Thermodynamic

Data for Combustion and Air-Pollution Use, Technion Aero-

space Engineering (TAE) Report # 804 June 1997, ftp://

ftp.technion.ac.il/pub/supported/aetdd/thermodynamics/.

NIST Chemistry WebBook, NIST Standard Reference Database

Number 69, February 2000, ed. W. G. Mallard and P. J.

Linstrom, National Institute of Standards and Technology,

Gaithersburg, MD 20899, http://webbook.nist.gov/.

84 M. J. Fadden, C. Barckholtz and C. M. Hadad, J. Phys. Chem.

A, 2000, 104, 3004–8011.

85 S. Olivella, A. Sol e´ and A. Garc ´ı a-Raso, J. Phys. Chem., 1995,

99, 10549–10556.

86 B. K. Carpenter, J. Am. Chem. Soc., 1993, 115, 9806–9807.

87 R. Buth, K. Hoyermann and J. Seeba, Proc. Combust. Inst.,

1994, 25, 841–849.

88 C. F. Melius, M. E. Colvin, N. M. Marinov, W. J. Pitz and S. M.

Senkan, Proc. Combust. Inst., 1996, 26, 685–692.

89 J. Filley and J. T. McKinnon, Combust. Flame, 2001, 124, 721–

723.

90 X. Zhong and J. W. Bozzelli, J. Phys. Chem. A, 1998, 102, 3537–

3555.

91 L. V. Moskaleva and M. C. Lin, J. Comput. Chem., 2000, 21,

415–425.

92 K. Roy, C. Horn, P. Frank, V. G. Slutsky and T. Just, Proc.

Combust. Inst., 1998, 27, 329–336.

93 A. M. Dean, J. Phys. Chem., 1990, 94, 1432–1439.

94 H. Wang and K. Brezinsky, J. Phys. Chem. A, 1998, 102, 1530–

1541.

95 J. A. Cole, J. D. Bittner, J. P. Longwell and J. B. Howard, Com-

bust. Flame, 1984, 56, 51–70.

96 U. Alkemade and K. H. Homann, Z. Phys. Chem., Neue Folge,

1989, 161, 19–34.

97 N. M. Marinov, W. J. Pitz, C. K. Westbrook, M. J. Castaldi and

S. M. Senkan, Combust. Sci. Technol., 1996, 116-117, 211–287.

98 R. P. Lindstedt and G. Skevis, Proc. Combust. Inst., 1996, 26,

703–709.

99 A. Goldaniga, T. Faravelli and E. Ranzi, Combust. Flame, 2000,

122, 350–358.

100 S. E. Stein, J. A. Walker, M. M. Suryan and A. Fahr, Proc. Com-

bust. Inst., 1990, 23, 85–90.

101 L. V. Moskaleva, A. M. Mebel and M. C. Lin, Proc. Combust.

Inst., 1996, 26, 521–526.

102 R. Sumathi, personal communication, Massachusetts Institute of

Technology, Cambridge, MA, 2001.

103 A. M. Mebel, W. M. Jackson, A. H. H. Chang and S. H. Lin, J.

Am. Chem. Soc., 1998, 120, 5751–5763.

5

7

8

5

6

9

0

1

R. J. Kee, F. M. Rupley and J. A. Miller, The Chemkin Thermo-

dynamic Data Base, Sandia Technical Report SAND87-8215B,

UC-4, Sandia National Laboratories, Livermore, CA, April

1

994.

W. Tsang and R. F. Hampson, J. Phys. Chem. Ref. Data, 1986,

5, 1087.

62

63

64

1

´

V. D. Knyazev, A. Bencsura, S. I. Stoliarov and I. R. Slagle, J.

Phys. Chem., 1996, 100, 11346–11354.

H. Thiesemann, E. P. Cliﬀord, C. A. Taatjes and S. J. Klippen-

stein, J. Phys. Chem. A, 2001, 105, 5393–5401.

W. Tsang and J. A. Walker, J. Phys. Chem., 1992, 96, 8378–8384.

V. D. Knyazev and I. R. Slagle, J. Phys. Chem., 1996, 100,

6

5

6

1

6899–16911.

S. Madronich and W. Felder, J. Phys. Chem., 1985, 89, 3556–

561.

T. Ko, G. Y. Adusei and A. Fontijn, J. Phys. Chem., 1991, 95,

745–8748.

67

68

69

70

71

72

73

74

3

104 T. L. Nguyen, A. M. Mebel and R. I. Kaiser, J. Phys. Chem. A,

2001, 105, 3284–3299.

105 T. B o¨ hland, F. Temps and H. Gg. Wagner, Proc. Combust. Inst.,

1986, 21, 841–850.

106 M. Karni, I. Oref, S. Barzilai-Gilboa and A. Lifshitz, J. Phys.

Chem., 1988, 92, 6924–6929.

107 T. Kakumoto, T. Ushirogouchi, K. Saito and A. Imamura, J.

Phys. Chem., 1987, 91, 183–189.

108 M. Eichholtz, A. Schneider, D.-V. Stucken, J.-T. Vollmer and

H. Gg. Wagner, Proc. Combust. Inst., 1994, 25, 859–866.

109 P. M. Mayer, C. J. Parkinson, D. M. Smith and L. Radom, J.

Chem. Phys., 1998, 108, 604–615.

8

A. M. Mebel, M. C. Lin, T. Yu and K. Morokuma, J. Phys.

Chem. A, 1997, 101, 3189–3196.

J. H. Kiefer, L. J. Mizerka, M. R. Patel and H.-C. Wei, J. Phys.

Chem., 1985, 89, 2013–2019.

S. Scherer, T. Just and P. Frank, Proc. Combust. Inst., 2000, 28,

1511–1518.

L. K. Madden, L. V. Moskaleva, S. Kristyan and M. C. Lin, J.

Phys. Chem. A, 1997, 101, 6790–6797.

P. R. Westmoreland, A. M. Dean, J. B. Howard and J. P. Long-

well, J. Phys. Chem., 1989, 93, 8171–8180.

A. M. Mebel, M. C. Lin, D. Chakraborty, J. Park, S. H. Lin and

Y. T. Lee, J. Chem. Phys., 2001, 114, 8421–8435.

Phys. Chem. Chem. Phys., 2002, 4, 2038–2055

2055

Article Doi

Formation and consumption of single-ring aromatic hydrocarbons and their precursors in premixed acetylene, ethylene and benzene flames

DOI: 10.1039/b110089k

Source and publish data:

Authors:

Article abstract of DOI:10.1039/b110089k

Full text of DOI:10.1039/b110089k

Products guided by the article

R&D Labs maybe for 124-38-9

Relevant to this article

Hot Product