+ All Categories
Home > Documents > High Pressure Studies of Propene Combustion

High Pressure Studies of Propene Combustion

Date post: 20-Nov-2021
Category:
Upload: others
View: 1 times
Download: 0 times
Share this document with a friend
12
Paper # 070RK-0172 Topic: Reaction Kinetics 8 th U. S. National Combustion Meeting Organized by the Western States Section of the Combustion Institute and hosted by the University of Utah May 19-22, 2013 High Pressure Studies of Propene Combustion Jeffrey Santner, 1 Francis M. Haas, 1 Xiaobo Shen, 1,2 Yiguang Ju, 1 Frederick L. Dryer, 1 1 Department of Mechanical and Aerospace Engineering, Princeton University, Engineering Quad, Olden Street, Princeton, NJ 08544 2 State Key Laboratory of Fire Science, University of Science and Technology of China, Hefei 230026, PR China Oxidation of propene, an important stable intermediate formed in the combustion of C 3 and larger hydrocarbons and oxygenated biofuels, was studied in both a high pressure laminar flow reactor at ~800 K and in high pressure rich and lean premixed laminar flames. These experimental platforms provide insight to propene oxidation kinetics for a broad range of temperature and fuel loading conditions at high pressure, where few experimental data are available. Such high pressure conditions shift the chemistry to regimes of higher kinetic uncertainty and sensitivity. Lean premixed laminar burning rates measured up to 20 atm in a nearly constant pressure chamber agree with model predictions, though model predictions are significantly slower for rich conditions. Rich flame predictions are particularly sensitive to reactions of the fuel and its fragments that control the radical pool, such as CH 3 +H(+M)=CH 4 (+M), C 3 H 6 +H=aC 3 H 5 +H 2 , aC 3 H 4 +H=aC 3 H 5 , and reactions of other C 1 and C 2 fragments. Flow reactor measurements at 15 atm and ~800 K reveal slow reaction with none observed at 770 K, both observations being in contrast with predictions using models found in the literature. Predictions of flow reactor species evolution profiles are sensitive to initiation reactions involving long-lived resonantly-stable allyl (aC 3 H 5 ) radical self- recombination. Inclusion of this reaction submodel greatly increases predicted induction time and improves model predictions of species gradients. 1. Introduction Propene (C 3 H 6 ) is an important intermediate alkene formed in the combustion of C 3 and larger hydrocarbons and oxygenated biofuels/additives. For example, at high temperatures, n-alkane decomposition typically proceeds via H-atom abstraction, followed by β-scission, producing an alkyl radical and an alkene. This alkyl radical can further decompose
Transcript
Page 1: High Pressure Studies of Propene Combustion

Paper # 070RK-0172 Topic: Reaction Kinetics

8th U. S. National Combustion Meeting

Organized by the Western States Section of the Combustion Institute

and hosted by the University of Utah

May 19-22, 2013

High Pressure Studies of Propene Combustion

Jeffrey Santner,1 Francis M. Haas,1 Xiaobo Shen,1,2 Yiguang Ju,1 Frederick L. Dryer,1

1Department of Mechanical and Aerospace Engineering, Princeton University, Engineering Quad,

Olden Street, Princeton, NJ 08544 2State Key Laboratory of Fire Science, University of Science and Technology of China, Hefei

230026, PR China

Oxidation of propene, an important stable intermediate formed in the combustion of C3 and larger hydrocarbons

and oxygenated biofuels, was studied in both a high pressure laminar flow reactor at ~800 K and in high pressure rich and

lean premixed laminar flames. These experimental platforms provide insight to propene oxidation kinetics for a broad range

of temperature and fuel loading conditions at high pressure, where few experimental data are available. Such high pressure

conditions shift the chemistry to regimes of higher kinetic uncertainty and sensitivity.

Lean premixed laminar burning rates measured up to 20 atm in a nearly constant pressure chamber agree with

model predictions, though model predictions are significantly slower for rich conditions. Rich flame predictions are

particularly sensitive to reactions of the fuel and its fragments that control the radical pool, such as CH3+H(+M)=CH4(+M),

C3H6+H=aC3H5+H2, aC3H4+H=aC3H5, and reactions of other C1 and C2 fragments.

Flow reactor measurements at 15 atm and ~800 K reveal slow reaction with none observed at 770 K, both

observations being in contrast with predictions using models found in the literature. Predictions of flow reactor species

evolution profiles are sensitive to initiation reactions involving long-lived resonantly-stable allyl (aC3H5) radical self-

recombination. Inclusion of this reaction submodel greatly increases predicted induction time and improves model

predictions of species gradients.

1. Introduction

Propene (C3H6) is an important intermediate alkene formed in the combustion of C3 and larger hydrocarbons and

oxygenated biofuels/additives. For example, at high temperatures, n-alkane decomposition typically proceeds via H-atom

abstraction, followed by β-scission, producing an alkyl radical and an alkene. This alkyl radical can further decompose

Page 2: High Pressure Studies of Propene Combustion

2

via β-scission to form a smaller alkyl radical and an alkene, or H-atom abstraction from this parent radical can produce

an alkene. From this well-known alkane “unzipping” and similar mechanisms for other fuel molecular classes, large

quantities of low carbon number alkenes are formed during hydrocarbon combustion. A recent flow reactor study [1]

further suggests that at intermediate combustion temperatures, alkenes may also be produced from large n-alkane

oxidation due to HO2 elimination from alkylperoxy (RO2) radicals. This pathway remains poorly understood. Among

small alkenes, propene is of particular interest because it is the simplest alkene containing the resonantly-stabilized allyl

(CH2=CH-ĊH2 ↔ ĊH2-CH=CH2, aC3H5) group. Thus, aside from its own inherent merits, study of propene chemistry

can be used to study chemistry of the allyl group, which is also present in larger fuel molecules such as those found in

biodiesel.

Few works have considered the oxidation mechanism of propene at higher pressures relevant to applied combustors,

which typically operate at elevated pressure for reasons of increased energy conversion efficiency. Across the

combustion temperature range, higher pressures emphasize radical recombination and HO2 reactions compared to similar

conditions at lower pressure. Propene flames have been studied extensively by the Bielefeld group at 50 mbar (e.g., [2,

3]), while other flame studies have included pressures from 1 atm [4] to 5 atm [5]. However, at these lower pressure

conditions, the HO2-driven kinetics typical of high pressure systems are much less important. Conditions at intermediate

combustion temperatures are characterized by radical addition across the double bond as well as allyl oxidation and

recombination chemistry. These families of reactions have relatively large rate coefficient uncertainties. While studies of

propene oxidation in flow reactors and jet stirred reactors range from atmospheric pressure [4, 6] to 8 atm [7], they

typically investigate higher temperature regimes where HO2 and O2 addition pathways are not emphasized. The study of

Hori et al. [6] considers the intermediate temperature oxidation of propene, but interpretation of its results is confounded

by chemistry of added NOx species. The present study expands on the work of Zheng et al. [8], which investigates

propene oxidation at both intermediate temperature and high pressure; herein higher pressures, lower temperatures, and

broader range of equivalence ratio are considered.

Through both experiments and modeling, the present study investigates the kinetics of propene oxidation at high

pressure, where observables appear to be sensitized to elementary reactions involving HO2, oxygen addition, and radical

recombination. This regime is studied experimentally using two platforms to access both high and intermediate

combustion temperatures. First, burning rates are measured in lean and rich flames from 1 to 20 atm. Second, the major

stable species involved in propene oxidation are measured in a flow reactor at 15 atm and 800 K over a range of

equivalence ratios. Several flow reactor conditions for which no measurable extent of reaction occurred are additionally

discussed. Present experimental results for this high pressure regime provide useful constraint for elucidating the

oxidation of propene and other allylic systems.

2. Methods

2.1 High Pressure Flame Experiments

Burning rates were measured using the outwardly propagating spherical flame method in a 10 cm diameter

cylindrical chamber with a concentric pressure release chamber and two windows. For details on the device, see [9].

Page 3: High Pressure Studies of Propene Combustion

3

Mixtures were created from propene (>99%), oxygen

(99.5%), helium (99.995%), nitrogen (99.99%), and

synthetic air using the partial pressure method. After

allowing the mixture ten minutes to become quiescent,

it is centrally ignited by a spark. High speed (8000 fps

and 15000 fps) schlieren imaging is utilized to image

the flame propagation up to a radius of 3 cm. The

combustion pressure rise is released to the outer

chamber after the flame front has passed the edge of the

viewing window.

An edge detection program and circle fitting

algorithm are used to determine the flame radius from

each image. The stretched propagation speed sb and

stretch rate κ are extracted from the radius time history

and corrected for compression-induced flow effects as discussed in [10]. The unstretched flame propagation speed sb,0 is

then calculated through linear extrapolation to zero stretch using the linear stretch relation (sb = sb,0 - κLb). Extrapolation

endpoints are determined iteratively by locating the range where the residuals from a linear fit of the strain-speed data

are below a threshold value and using this range to compute a new linear fit. This process is repeated while decreasing

the threshold until stable endpoints are found. The extrapolated burning velocity is multiplied by the calculated burned

gas density [11] to give the mass burning rate. No data analyses were performed for flames that were observed to be

wrinkled due to cellular or spiraling instabilities, affected by buoyancy, or influenced by transient or non-linear response

of the flame speed to stretch rate.

All experiments were performed at room temperature (measured as 298 ± 1 K) with pressures from 1 to 20 atm.

Atmospheric pressure experiments in air were performed to compare with published measurements and appear to fall

among the scatter of reliable literature values [4, 5]

(Fig. 1). Burning rates were then measured at high

pressures with calculated, fixed flame

temperatures of 2000 K in N2/He dilution for

equivalence ratios of 0.8 and 1.3. The N2/He ratio

was adjusted to achieve stable flames and remove

ignition difficulties. For ϕ=0.8, the N2/He ratio was

2, and for φ=1.3, the ratio was 1. Experiments

were repeated at some conditions to indicate

experimental repeatability.

Figure 1: Present measurements compared with literature measurements [4,5] for propene flame speed in air at 1 atmosphere.

0

5

10

15

20

25

30

35

40

45

50

0.60 0.80 1.00 1.20 1.40 1.60 1.80

Lam

inar

flam

e spe

ed (c

m/s

)

Equivalence Ratio

PresentJomaas et al.Davis et al. (nonlinear)Davis et al. (linear)USC Mech IIAramco Mech

C3H6/Air1 atm

Figure 2: Schematic of the HPLFR facility.

Page 4: High Pressure Studies of Propene Combustion

4

2.2 High Pressure Laminar Flow Reactor (HPLFR)

The HPLFR is a new flow reactor facility developed

to measure both fundamental chemical rate coefficients

as well as global oxidation features for species of

interest to combustion. The design of the HPLFR

includes features of both the VPFR (e.g., [1, 12]) as well

as the Technical University of Denmark (DTU) laminar

flow reactor [13]. The HPLFR accesses relatively high

pressures (≤ 30 atm) and temperatures from ~500-1000

K.

Figure 2 presents a general schematic of the HPLFR

with major subsystems indicated. A PID-thermostatted

three-zone tube furnace encloses a 1.5 in. OD pressure

shell, which in turn encloses a reactor duct. This duct is

fed by a steady flow of premixed, preheated gaseous reactants from the Feed/Calibration System. Under conditions

favoring reaction, this premixed gas feed converts into products in the duct and subsequently exhausts from the reactor.

A back pressure regulator at the exhaust controls duct pressure, and some pressurized exhaust bleeds into the annular

space between duct and pressure shell to maintain pressure equilibrium across the duct wall. A hot water-cooled,

convection quench probe with integrated thermocouple extracts a small, quenched, continuous sample flow from a fixed

axial coordinate in the duct test section. This flow passes through heated transfer lines to a pressure-regulated online

Inficon 3000 micro gas chromatograph analytical system, which permits identification and quantification of stable

species of interest. A screw drive translates the probe axially through the duct, enabling sample collection along the duct

axis. A simple velocity-axial displacement relationship determines relative residence time in the test section under plug

flow and additional idealized flow reactor assumptions.

Though flow is laminar as characterized by duct Reynolds

number, measured reaction occurs in the nearly-plug flow

entry length following a sudden expansion in the duct,

minimizing flowfield aberrations arising from the steady-

state parabolic profile usually associated with laminar

flows.

Quenched gas samples from the probe are transferred

through a heated (393 K) Teflon line to a gas

chromatograph (GC) using three columns (backflushed

5Å molecular sieve, PLOTQ, and OV1 using helium or

argon carrier gases) and a thermal conductivity detectors

(TCDs) on each column. Area responses and retention

times are determined from dilutions of calibration

Figure 3: Mass burning rate of lean propene flames from 5-20 atm compared to predictions from USC Mech [16] and Aramco Mech [15].

0

0.05

0.1

0.15

0.2

0.25

0 5 10 15 20

Mas

s bur

ning

rate

(g c

m-2

s-1)

Pressure (atm)

Experiments

USC Mech

Aramco Mech

Propene/O2/N2/HeN2/He=1Tf=2000 Kϕ=0.8

Figure 4: Mass burning rate of rich propene flames from 3-20 atm compared to predictions from USC Mech [16] and Aramco Mech [15].

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0 5 10 15 20

Mas

s bur

ning

rate

(g c

m-2

s-1)

Pressure (atm)

ExperimentsUSC Mech IIAramco Mech

Propene/O2/N2/HeN2/He=1Tf=2000 Kϕ=1.3

Page 5: High Pressure Studies of Propene Combustion

5

standards.

Further discussion of the HPLFR facility is to appear

in the thesis of Haas [14]. However, it is worth noting

that this apparatus has been previously validated by

favorable comparisons of its measurements to relatively

well-established rate coefficients for H+O2+M→ HO2+M

for M = (N2, Ar) and H+NO2→NO+OH.

2.3 Modeling

Calculations used two kinetic models - Aramco

Mech [15] and USC-Mech II [16]. Flame simulations

were performed using the PREMIX code [11] utilizing

multicomponent transport properties and Soret diffusion,

with both gradient and curvature limits set to 0.05.

SENKIN from the CHEMKIN II package [17] was used

to simulate the flow reactor experiment at constant pressure conditions, under the adiabatic assumption for VPFR

(reactivity) experiments, and the isothermal assumption for HPLFR (speciation) experiments further discussed below.

3. Results and Discussion

3.1 Flame speeds

Laminar flame speeds of propene in air at 1 atm, discussed above in Section 2.1, are shown in Fig. 1. Model

predictions appear to be significantly slower than the present experiments (and others) for the rich flames, although

Aramco Mech predictions for the rich side agree well with

the Jomaas et al. [5] measurements. Mass burning rates for

lean and rich propene flames from 3-20 atm are shown in

Figs. 3-4. The results are similar to those at 1 atm (Fig. 1)

- the model predictions agree with experimental

measurements for lean flames, but predictions are slower

than measurements for rich flames, by up to 35%.

3.2 Flow reactor measurements

Results of a reactivity profile experiment [8] from the

Princeton Variable Pressure Flow Reactor (VPFR) are

shown in Fig. 5 at 12.5 atm and stoichiometric conditions

for a residence time of 1.8 seconds. This experiment

measures stable C3H6 and H2O evolution as a function of

Figure 5: Reactivity comparison of VPFR experiments [8] at 12.5 atm, with an equivalence ratio of 1 and residence time of 1.8 seconds.

0

500

1000

1500

2000

2500

3000

3500

4000

500 600 700 800 900 1000

Mol

e Fra

ctio

n (p

pm)

Temperature (K)

Aramco MechUSC-Mech IIUpdated ModelC3H6H2O

Figure 6: Flow reactor speciation at 15 atm, 800 K, with an equivalence ratio of 0.35. Oxygen profiles have been scaled stoichiometrically with initial C3H6 mole fraction.

0

500

1000

1500

2000

2500

3000

3500

4000

4500

0 1 2 3

Mol

e fra

ctio

n (p

pm)

Time (s)

stoic. scaled O2C3H6CO/2CO2Aramco MechUSC-Mech IIUpdated Model

15 atm800 Kϕ=0.35

Page 6: High Pressure Studies of Propene Combustion

6

varying initial temperature at fixed residence time. The

primary purpose of including these experimental results

is to show that there is no low temperature chemistry or

negative temperature coefficient behavior that often

results from oxidation of larger monoalkenes and alkanes

(e.g., [1]) under similar conditions. However, propene

slowly reacts at high pressures and intermediate

temperatures, evident in the slope of the measured

propene and water profiles below 800 K. Modeling and

model interpretation of this type of reactivity profile

requires further considerations, beyond the scope of the

present work (e.g., [1, 12]); however, the lower

temperature reactivity (relative to measurements) of both

the USC Mech II and Aramco Mech model predictions

suggests that the predicted global reactivity of these models is too fast. The present speciation experiments, which

measure stable species evolution as a function of time for fixed temperature, further investigate propene oxidation

behavior in this intermediate temperature, high pressure regime.

Species profiles measured in the flow reactor for various equivalence ratios are shown in Fig. 6-9. All tests were

conducted at 15.0 ± 0.1 atm and 800 ± 5 K, with less than 20 K measured temperature rise due to reaction for the lean

experiments, and less than 5 K temperature rise for the rich and stoichiometric cases. For this reason, the experiments are

modeled as isothermal. Fuel concentration ranged from 4000 ppm to 6240 ppm. In addition to the species reported in the

figures, the GC was calibrated to measure H2, CH4, C2H6, aC3H4 , and pC3H4, but the mole fractions of these species

were below detection/quantification limits of ~ tens of ppm in all experiments. Water and formaldehyde have been

removed from the lean measurements (Figs 6-7) due to sample condensation observed during the experiments. The GC

Figure 7: Flow reactor speciation at 15 atm, 800 K, with an equivalence ratio of 0.5. Oxygen profiles have been scaled stoichiometrically.

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

0 1 2 3

Mol

e fra

ctio

n (p

pm)

Time (s)

stoic. scaled O2C3H6CO/2CO2

15 atm800 Kϕ=0.5

Aramco MechUSC-Mech IIUpdated Model

Figure 8: Flow reactor speciation at 15 atm, 800 K, with an equivalence ratio of 1.0. Note that CH2O measurements are reported in arbitrary units and have been scaled to match the approximate intersection of the three models represented. The same scale factor is used in Fig. 9. Oxygen profiles have been scaled stoichiometrically. Models have been time shifted to match the highest experimental H2O mole fraction measurement, as discussed in the text.

0

20

40

60

80

100

120

140

160

180

200

0 1 2 3

Mol

e fra

ctio

n (p

pm)

Time (s)

COCO2CH2O - A.U.

15 atm800 Kϕ=1.0

Aramco MechUSC-Mech IIUpdated Model

0

50

100

150

200

250

300

350

400

0

1000

2000

3000

4000

5000

0 1 2 3

H2O

mol

e fr

actio

n (p

pm)

Mol

e fra

ctio

n (p

pm)

Time (s)

Scaled O2C3H6H2O

15 atm800 Kϕ=1.0

Aramco MechUSC-Mech IIUpdated Model

Page 7: High Pressure Studies of Propene Combustion

7

peak for CH2O was identified using formalin solution,

but the mole fraction could not be accurately quantified -

formaldehyde results are therefore displayed in arbitrary

units using the same scaling factor for Figs. 8 and 9.

The lean modeling results (Figs. 6-7) have been time

shifted to coincide with points in the early portion of the

propene profile according to discussion in the

Supplementary Material of [12]. Both models predict

slightly faster reaction than measured, with the USC-

Mech [16] performing slightly better than Aramco Mech

[15]. The stoichiometric and rich results (Figs. 8-9) have

been time shifted to coincide with the highest measured

water concentration, as the measured water profile has

the highest gradient relative to its measurement

uncertainty. There is significant disagreement between

the models and experiment consistent with the steeper

oxidation gradients and (lack of) induction behavior

noted for the lean and Zheng [8] cases. An updated

model (discussed in section 3.3) appears to improve

predictions against experiment, especially for the more

sensitive profiles of H2O, CH2O (gradient shape), and

CO. These profiles provide sufficient partial constraint to

underscore points made subsequently about allyl self-

recombination chemistry, though in general, it is

desirable to capture more of the fuel oxidation gradient than exhibited by the stoichiometric and rich data provided

herein

In addition to the speciation profiles shown, measurements were also attempted at lower pressure and temperature

conditions at an equivalence ratio of ~0.35. At 770 K and pressures of 8, 12, and 15 atm, no quantifiable fuel loss or

product formation was observed at residence times of 1.2 s (8 atm), 1.8 s (12 atm) and 2.3 s (15 atm), although both

models predict significant reaction at these conditions.

3.3 Kinetic analysis and modeling

Further kinetic analysis for the flame conditions is performed using USC Mech II [16], though Aramco Mech

performs similarly. The model performs well for lean flames, but predictions are too slow for rich flames. A-factor

sensitivity analysis (Figure 10) conducted for conditions represented in Figs. 3 and 4 reveals that the burning rate is most

sensitive to H+O2=OH+O, however the USC Mech II rate coefficient is within 4% of the recent, low uncertainty

recommendation of Hong et al. [18] from 800 - 2300 K for this reaction, suggesting it is unlikely to be the cause of much

Figure 9: Flow reactor speciation at 15 atm, 800 K, with an equivalence ratio of 1.25. Caveats as noted in Fig. 8.

0

50

100

150

200

250

300

350

400

0

1000

2000

3000

4000

5000

6000

7000

0 1 2 3

H2O

mol

e fr

actio

n (p

pm)

Mol

e fra

ctio

n (p

pm)

Time (s)

Scaled O2C3H6H2OAramco MechUSC-Mech IIUpdated Model

15 atm800 Kϕ=1.25

0

20

40

60

80

100

120

140

160

0 1 2 3

Mol

e fra

ctio

n (p

pm)

Time (s)

COCO2CH2O - A.U.Aramco MechUSC-Mech IIUpdated Model

15 atm800 Kϕ=1.25

Page 8: High Pressure Studies of Propene Combustion

8

of the discrepancy between predictions and

experiment. Moreover, adjustment of this rate

coefficient to improve prediction of rich burning

rates would tend to deteriorate predictions against

lean cases.

The next most sensitive reaction is

CH3+H(+M)=CH4(+M). The high (sensitivity ×

uncertainty) for this reaction at rich conditions might

appear to explain much of the model prediction

discrepancy (Fig. 4). To bring model predictions

closer to experimental results, this rate needs

reduction. The rate coefficient used in USC Mech II

is slightly lower than recent calculations of Jasper &

Miller [19], and though it is based on GRI Mech

[20], it has been optimized to be up to 40% slower

than the (GRI-optimized) value. Uncertainties

associated with this reaction rate coefficient merit yet

further consideration, beyond the scope of the

present work, towards improving propene burning rate predictions.

Model prediction discrepancies may also be caused by a variety of reactions that affect rich flames more strongly

than lean flames, such as C3H6+H=aC3H5+H2, aC3H4+H=aC3H5, and many reactions of C1 and C2 fuel fragments. As

expected, elevated pressure also causes flame predictions to become sensitive to reactions involving HO2, such as

H+O2(+M)=HO2(+M), and aC3H5+HO2=OH+C2H3+CH2O. However, these reactions are no more sensitive for rich

conditions than for lean conditions, so while HO2 chemistry remains highly uncertain, it is not likely to be the cause of

the present model prediction discrepancies for rich flames.

Further modeling work for flow reactor conditions is performed using Aramco Mech [15]. Though both models

perform similarly, Aramco Mech was developed with additional focus on low to intermediate temperature chemistry,

where we currently find some deficiencies in modeled propene oxidation pathways. Time-integrated path flux analysis

(Fig. 11) reveals that, for the rich flow reactor conditions, the major channel for complete propene destruction is through

allyl (aC3H5), which reacts with HO2 to form C3H5O+OH. The lack of significant predicted reaction of allyl with O2,

which is present in much higher concentrations than any radical species, merits further investigation, though apparent

absence of these channels may be due in part to the timescale selected for path flux analysis of Fig. 11.

Figure 10: A-factor sensitivity of the burning rate to elementary reaction rates for flame conditions representative of those presently investigated.

-0.4 -0.2 0 0.2 0.4 0.6A-factor sensitivity

Rich, 20 atmRich, 1 atmLean, 20 atmLean, 1atm

Sensitivity of propene flame speedUSC-Mech IITf ~ 2000 K

H+O2=O+OH

CH3+H(+M)=CH4(+M)

CO+OH=CO2+H

H+O2(+M)=HO2(+M)

C3H6+H=aC3H5+H2

C3H6+OH=aC3H5+H2O

aC3H5+HO2=OH+C2H3+CH2O

aC3H5+H(+M)=C3H6(+M)

HCO+M=CO+H+M

H+OH+M=H2O+M

Page 9: High Pressure Studies of Propene Combustion

9

In further consideration of this observation, this study replaces the four allyl+O2 pathways present in the original

model with the rate coefficient calculations from Lee & Bozzelli [21, 22]. Also added to the Aramco Mech model is their

pressure-dependent oxygen addition pathway, which as shown in Fig. 12, is ~1000 times faster than each of the other

four allyl+O2 pathways [21, 22]. Decomposition reactions and thermochemistry for the adduct (aC3H5O2) have been

taken from a Lawrence Livermore National Laboratory model [6]. Lee describes yet additional channels for C3H5O2

adducts (9 total pathways), but reaction rates are slow relative to the major addition pathway (Fig. 12.). However, due to

the shallow energy well of the adduct, the major pathway for this reaction is to proceed back to the reactants, allyl and O2

[23]. This entire updated reaction submechanism is shown to have minimal effect on the modeling results (Fig. 13),

though it does provide more consistent treatment of the allyl+O2 reaction system through inclusion of the relatively faster

O2 addition channel.

Because allyl is not readily consumed by O2, allyl radicals are long-lived, and the allyl self-recombination reaction is

expected to be important. The allyl recombination submodel from a LLNL model [6] has also been added to the original

Aramco Mech. This submodel includes aC3H5 recombination to form C6H10 (1,5-hexadiene) from Tsang [24] and two

reactions for the destruction of C6H10. Onset of predicted propene oxidation is extremely sensitive to the highly uncertain

rate of allyl recombination, as the LLNL-Tsang submodel greatly slows predicted induction, moving the time for full

Figure 11: Path flux for complete propene destruction under flow reactor conditions of φ=1.25, 15 atm, and 800 K using Aramco Mech [15] and the updated model. Numbers refer to the percentage of each species being consumed in each pathway. Numbers in parentheses refer to flux using the updated model. Where only one number is present, the updated model is equivalent to Aramco Mech. Note that flux analysis results depend on choice of time integration interval and so specific results for the induction period, half-fuel consumption point, etc., may vary significantly using otherwise equal kinetic/physical model inputs.

Aramco Mech(Updated model)phi=1.25, 15 atm, 800 K

C3H6

aC3H5

+OH 34 (35)+HO2 4 (5)

C3H6OH

+OH 10

iC3H7

+H 26 (24)

C3H5O

+HO2 82 (73)+CH3O2 13 (12)

C2H3CHO

-H 87+O2 7

C2H3CO

+HO2 59 (58)+H 18

+OH 9 (10)+CH3O2 5

C2H3+CO

HOC3H6O2

+O2

CH3CHO+CH2O+OH

CH3CO

+OH 40+HO2 23 (24)

CH3+HOCHO

+OH 18

CH3+CO

iC3H7O2

+O2 99 (98)

C3H6OOH2-1

33

CH2O+HCO

+O2 65 CH2CHO+O2 26

CH2O+CO+OH

+O2 32

nC3H7

+H 6

CH3+C2H4

66

O2CH2CHO

+O2 67

C6H10

+aC3H5 0 (13)

C6H9

+OH

C2H3CHO+aC3H5

+OH

+HO2

HCO +OH 53+HO2 9

HOCHO+OH, -H 19

CO

+O2 91 (92)

CO2

+OH 94 (93)

Page 10: High Pressure Studies of Propene Combustion

10

oxidation from ~1 second to ~8 seconds for the

ϕ=0.35 condition and to ~20 seconds for the ϕ=1.25

condition (see Fig. 13). However, updating the allyl

recombination rate to the significantly more recent

recommendation of Matsugi et al. [25], which is a

factor of 5x105 slower at 800 K, gives more

reasonable results for full oxidation time - ~2

seconds for ϕ=0.35 and ~7 seconds for ϕ=1.2 5.

Inclusion of the allyl recombination pathway causes

strong induction inhibition because it terminates two

allyl radicals into the stable 1,5-hexadiene (C6H10)

intermediate.

Though allyl self-recombination chemistry is

presently missing from both the Aramco Mech and

USC Mech II kinetic models, this chemistry should

be included as an important, sensitive consideration

in prediction of high pressure, lower temperature

ignition and speciation. Absence of allyl self-reaction

chemistry at these conditions suggests compensatory tuning of other rate coefficients and attendant implied uncertainties

in the present modeling work. Treatment of the C6 species produced is admittedly a complicating model development

factor, but the present approach borrowed from LLNL [6] may provide a satisfactory approximation.

The updated model (model 4 in Fig. 13), consisting of the original Aramco Mech with both allyl+O2 and allyl+allyl

chemistry submodel revisions previously described, provides slightly improved performance for the lean conditions as

compared to Aramco Mech (Figs. 6-7). In addition to exhibiting physically plausible time shifts discussed above, the

Figure 12: Reaction rates of the 9 allyl+O2 paths from Lee & Bozzelli [21,22]. Solid lines indicate pathways that are added to/updated in the present updates to the Aramco Mech model.

1.E-061.E-041.E-021.E+001.E+021.E+041.E+061.E+081.E+101.E+12

0.5 1 1.5 2 2.5 3

Bim

olec

ular

rate

coe

ffic

ient

(cm

3m

ol-1

s-1)

1000/T (K-1)

C3H5-A+O2(+M)<=>C3H5O2(+M)C3H5-A+O2<=>C2H3CHO+OH C3H5-A+O2<=>C3H4-A+HO2 C3H5-A+O2<=>C2H2+CH2O+OH C3H5-A+O2<=>CH2CHO+CH2OC3H5-A+O2<=>C3H4OOHC3H5-A+O2<=> YC=CCO + OH C3H5-A+O2<=> YCC.COO C3H5-A+O2<=> C.H2 YCCOO

15 atm

800 K

Figure 13: Propene mole fraction simulations for lean and rich conditions using (1) the unaltered Aramco Mech [15], with (2) updated allyl+O2 chemistry [21,22], (3) updated allyl+O2 chemistry and the Tsang allyl recombination rate [24], and (4) updated allyl+O2 chemistry with the Matsugi allyl self-recombination rate [25]. For models 3 and 4, C6H10 chemistry is taken from LLNL[6].

0

0.0005

0.001

0.0015

0.002

0.0025

0.003

0.0035

0.004

0.0045

0 2 4 6 8

Prop

ene m

ole

frac

tion

Time (s)

1234

15 atm800 Kϕ=0.35

0

0.001

0.002

0.003

0.004

0.005

0.006

0.007

0 5 10 15 20

Prop

ene m

ole

frac

tion

Time (s)

1234

15 atm800 K

ϕ=1.25

Page 11: High Pressure Studies of Propene Combustion

11

updated model better predicts the slow oxidation “shoulder” feature of the temporal fuel oxidation profile before ~1.75

and ~2.0 seconds in the φ = 0.35/Fig. 6 and φ = 0.5/Fig. 7 cases, respectively. This change is largely attributable to the

reaction pathways influenced by adding allyl self-reaction. Predicted gradients for all three models considered remain

higher than measurements, similar to comparisons of predictions for the higher temperature, lower pressure species

profiles of Zheng et al. [8] (not shown here). Further scrutiny of propene and other molecular fragment chemistry is

clearly warranted. It is notable, however, that the USC Mech II and Aramco Mech models developed without

consideration of the present flow reactor data capture the primary features of the experimentally-measured species

evolution profiles.

Unlike the lean cases, model comparison with experiment is somewhat more difficult for the stoichiometric (Fig. 8)

and rich (Fig. 9) cases. Experimental measurements lie primarily in the early phases of oxidative reaction, and additional

measurements at extended residence times are desirable for elucidating the major features of the fuel and oxygen

consumption. However, the updated model does give improved predictions of the H2O, scaled CH2O, and (limited) CO

profiles.

Comparisons can also be made regarding what was not detected. Models predict up to 100-150 ppm of methane

evolved for the lean conditions, however methane was not experimentally observed. In the lean mixtures, methane is

formed from methyl, so this overprediction could result from failure of the model to predict methyl production, though

the assumed branching ratio between CH3+HO2 or CH3O2 reaction paths may also result in the disparity.

4. Conclusions

Premixed mass burning rates were measured from 3 to 20 atm under lean and rich conditions using the outwardly

propagating flame technique. Similar to present and literature measurements in air at 1 atm, model predictions from USC

Mech II [16] and Aramco Mech [15] agree with lean measurements, but predictions are slower than measurements for

rich conditions. Discrepancies are likely due to uncertainties in the rate parameters for CH3+H(+M)=CH4(+M),

C3H6+H=aC3H5+H2, aC3H4+H=aC3H5, and reactions of other C1 and C2 fragments.

Speciation profiles were measured at 800 K and 15 atm over a range of equivalence ratios in a nearly isothermal flow

reactor. After induction periods of significantly varying length, predicted temporal species oxidation gradients are higher

than measurements, particularly for stoichiometric and rich conditions. The allyl+O2 pathways in Aramco Mech [15]

have been updated, including a missing O2 addition pathway. Despite the present model modifications increasing the

overall rate of allyl+O2 by nearly three orders of magnitude, this entire submodel has negligible bearing on present

predictions at the flow reactor conditions considered. However, the preliminary allyl self-reaction submodel added to

Aramco Mech greatly increases predicted induction time and yields improved predicted species gradients when

compared to measurements. Kinetic model predictions of propene induction are found to be extremely sensitive to allyl

recombination, and further study on this pathway is needed in order to accurately predict oxidation and, particularly,

ignition of this species and other allylic compounds.

Page 12: High Pressure Studies of Propene Combustion

12

Acknowledgements

We thank Professor Henry Curran and the NUI Galway CCC for providing the unpublished “Aramco Mech” kinetic

model [15] used in the present work. This research was supported as part of the Combustion Energy Frontiers Research

Center, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Office of

Basic Energy Sciences under Award Number DE-SC0001198.

References

1. S. Jahangirian, S. Dooley, F.M. Haas, F.L. Dryer, Combust. Flame 159 (1) (2012) 30-43. 2. B. Atakan, A.T. Hartlieb, J. Brand, K. Kohse-Höinghaus, Proc. Combust. Inst. 27 (1) (1998) 435-444. 3. A. Lamprecht, B. Atakan, K. Kohse-Höinghaus, Combust. Flame 122 (4) (2000) 483-491. 4. S.G. Davis, C.K. Law, H. Wang, Combust. Flame 119 (4) (1999) 375-399. 5. G. Jomaas, X.L. Zheng, D.L. Zhu, C.K. Law, Proc. Combust. Inst. 30 (1) (2005) 193-200. 6. M. Hori, N. Matsunaga, N. Marinov, W.J. Pitz, C.K. Westbrook, Proc. Combust. Inst. 27 (1) (1998) 389-396. 7. P. Dagaut, M. Cathonnet, J.C. Boettner, J. Phys. Chem. 92 (3) (1988) 661-671. 8. L. Zheng, A. Kazakov, F.L. Dryer, in: Third joint meeting of the U.S. sections of the combustion institute, Chicago, Il, 2003. 9. X. Qin, Y. Ju, Proc. Combust. Inst. 30 (1) (2005) 233-240. 10. M.P. Burke, Z. Chen, Y. Ju, F.L. Dryer, Combust. Flame 156 (4) (2009) 771-779. 11. R.J. Kee, J.F. Grcar, M.D. Smooke, J.A. Miller, A Fortran Program for Modeling Steady Laminar One-Dimensional Premixed Flames. 12. Z. Zhao, M. Chaos, A. Kazakov, F.L. Dryer, Int. J. Chem. Kinet. 40 (1) (2008) 1-18. 13. C.L. Rasmussen, J. Hansen, P. Marshall, P. Glarborg, International Journal of Chemical Kinetics 40 (8) (2008) 454-480. 14. F.M. Haas. Studies of Small Molecule Reactions Foundational to Combustion Chemistry, Including Experimental Measurements from a Novel High Pressure Flow Reactor. Ph.D. Thesis, Princeton University, 2013. 15. W.K. Metcalfe, S. Burke, H.J. Curran, Combust. Flame (Submitted for Review). 16. H. Wang, X. You, A.v. Joshi, S.G. Davis, A. Laskin, F.N. Egolfopoulos, C.K. Law, USC Mech Version II. High-Temperature Combustion Reaction Model of H2/CO/C1-C4 Compounds. http://ignis.usc.edu/USC_Mech_II.htm 17. R.J. Kee, F.M. Rupley, J.A. Miller, Technical Report SAND89-8009, Sandia National Laboratories, Albuquerque, NM, 1989. 18. Z. Hong, D.F. Davidson, E.A. Barbour, R.K. Hanson, Proc. Combust. Inst. 33 (1) (2011) 309-316. 19. A.W. Jasper, J.A. Miller, J. Phys. Chem. A 115 (24) (2011) 6438-6455. 20. G.P. Smith, D.M. Golden, M. Frenklach, N.W. Moriarty, B. Eiteneer, M. Goldenberg, C.T. Bowman, R.K. Hanson, S. Song, W.C. Gardiner, V.V. Lissianski, Z. Qin GRI-Mech 3.0. http://www.me.berkeley.edu/gri_mech/ 21. J. Lee. Thermochemistry and kinetic analysis on radicals of acetaldehyde + O2, allyl radical + O2 and diethyl and chlorodiethyl sulfides. New Jersey Institute of Technology, 2003. 22. J. Lee, J.W. Bozzelli, Proc. Combust. Inst. 30 (2005) 1015-1022. 23. J.W. Bozzelli, A.M. Dean, J. Phys. Chem. 97 (17) (1993) 4427-4441. 24. W. Tsang, J. Phys. Chem. Ref. Data 20 (2) (1991) 221-274. 25. A. Matsugi, K. Suma, A. Miyoshi, J. Phys. Chem. A 115 (26) (2011) 7610-7624.


Recommended