numerical and experimental studies of ethanol flames

Proceedings

Proceedings of the Combustion Institute 31 (2007) 1149–1156

www.elsevier.com/locate/proci

of the

CombustionInstitute

Numerical and experimental studies of ethanol flames

Priyank Saxena *, Forman A. Williams

Center for Energy Research, Department of Mechanical and Aerospace Engineering, University of California,

San Diego, La Jolla, CA 92093, USA

Abstract

Ethanol combustion is investigated on the basis of a new chemical-kinetic mechanism consisting of 192elementary steps among 36 species, augmented by 53 additional steps and 14 additional species to addressthe formation of oxides of nitrogen and 43 steps and 7 species to address formation of compounds involv-ing three carbon atoms. The mechanism is tested against shock-tube autoignition-delay data, laminar burn-ing velocities, counterflow diffusion-flame extinction and measurements of structures of counterflowpartially premixed and diffusion flames, the last of these newly completed and reported here for the firsttime. These measurements, on ethanol–air flames at a strain rate of 100 s�1, employing prevaporized eth-anol with a mole fraction of 0.3 in a nitrogen carrier stream, were made for the pure diffusion flame and fora partially premixed flame with a fuel-side equivalence ratio of 2.3 and involved thermocouple measure-ments of temperature profiles and determination of concentration profiles of C2H5OH, CO, CO2, H2,H2O, O2, N2, CH4, C2H6 and C2H2 + C2H4 by gas chromatographic analysis of samples withdrawnthrough fine quartz probes. Computational investigations also were made of profiles of oxides of nitrogenand other potential pollutants in similar partially premixed flames of ethanol and other fuels for compar-ison purposes. The computational results are in reasonable agreement with experiment and perform as wellas or better than predictions of other, generally much larger, mechanisms available in the literature. Fur-ther research is, however, warranted for providing additional and more stringent tests of the mechanismand its predictions.� 2006 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Keywords: Ethanol; Reaction mechanism; Partially premixed flame; Nonpremixed flame; Pollutants

1. Introduction

Although ethanol is attractive for use as apossible low-pollution source of renewable ener-gy, significant uncertainties remain in its com-bustion chemistry. Despite the availability of anumber of detailed chemical-kinetic mechanismsfor ethanol combustion [1–6], there are differenc-es in predictions of these mechanisms, and most

1540-7489/$ - see front matter � 2006 The Combustion Institdoi:10.1016/j.proci.2006.08.097

* Corresponding author. Fax: +1 858 534 5354.E-mail address: [email protected] (P. Saxena).

of them have not been tested entirely thoroughlyagainst experimental combustion data. A needtherefore exists to investigate possibilities ofimproving the mechanisms and to attempt tovalidate them against measurements made inflames. The present paper addresses this needby suggesting a somewhat different detailedmechanism and by comparing predictions withnew measurements of structures of partially pre-mixed and diffusion flames in a counterflowgeometry. The paper also investigates computa-tionally profiles of concentrations of pollutantsin these and related flames.

ute. Published by Elsevier Inc. All rights reserved.

mailto:[email protected]

Fig. 1. Reaction-path diagram of the ethanol partiallypremixed flame listed in Table 2.

1150 P. Saxena, F.A. Williams / Proceedings of the Combustion Institute 31 (2007) 1149–1156

2. The reaction mechanism

The chemical-kinetic mechanism is an augmen-tation of a mechanism developed for the combus-tion of hydrogen [7], carbon monoxide [7],methane [8,9], ethane [10], ethylene [11], acetylene[12,13], propane [14], propene [14], propyne [14],allene [14] and methanol [15,16]. The mechanismis relatively short for a detailed mechanism, sim-plifications having been achieved by restrictingattention to temperatures above about 1000 K,pressure below about 100 bar, equivalence ratiosless than about 3 in premixed systems and poten-tial-flow strain rates greater than about 50 s�1 innonpremixed or partially premixed systems. Theserestrictions preclude addressing soot formationand cool flames, for example. The mechanism isextended here to ethanol by adding many of thesteps and rate parameters of Li [6].

Specifically, 33 reactions are added thatinvolve C2H5OH or one of the three isomers pro-duced by abstraction of an H atom from it,CH3CHOH, CH2CH2OH and CH3CH2O, and22 reactions are added that involve acetaldehydeor one of the two isomers produced by abstractionof H from it, CH2CHO and CH3CO. Thus, intotal, 55 new reactions and 6 new species are add-ed. Consistent with our earlier practice and high-temperature objective, the peroxide HOC2H4O2

and reactions related to it, which are included byLi [6], are not part of the present mechanism.All rate parameters for the newly added reactionsare those of Li [6], except for the two initiationsteps for which Li gives values only at particularpressures and for which, seeking simplicity withinbounds of present uncertainty, we constructed aTroe-type falloff fit with Fc = 0.5 and with zerotemperature exponents at both low and high pres-sures, resulting in low-pressure and high-pressurespecific reaction-rate constants given in Table 1.Because of falloff complexities, the low-pressurerate parameters may not apply below about10�2 bar, and rates calculated at the highest tem-peratures likely are too high in Table 1. Theresulting mechanism, available on the web [17],incorporates some revisions of rates given in pre-vious publications and involves 288 elementarysteps among 57 chemical species where the 53

Table 1New ethanol rate parameters

Step

C2H5OH + Ma fi CH3 + CH2OH + M

C2H5OH + Ma fi C2H4 + H2O + M

a Chaperon efficiencies are 2.0 for H2, 6.0 for H2O, 1.5 for COall other species; Troe falloff with Fc = 0.5; units of activation

steps among 14 species needed to address NOx

formation and the 43 steps among 7 species need-ed to address formation of compounds involvingthree carbon atoms are appended. All revisionsother than those indicated above have been dis-cussed in recent publications [7,14].

Figure 1 shows a representative reaction-pathdiagram for this mechanism. The conditionsselected for construction of this figure correspondto the partially premixed case of Table 2, and thefigure tracks the carbon history, the arrows thatshow the main pathways being labeled with theagents and the percentage of their contributions.These percentages are obtained by integratingconsumption rates over the entire field. Indica-tions of the fates of some of the minor species,such as C2H, C3H4, C3H6 and C3H8, are omittedfor clarity because they are present in very smallquantities and do not affect the rest of the chemis-try significantly.

Direct fuel decomposition by the second steplisted in Table 1 seems to be more important forthese ethanol flames than is direct decomposition

k0 = 3.0 · 1016 e(�242.4/RT) cm3/mol sk1 = 5.0 · 1015 e(�342.8/RT) s�1

k0 = 1.0 · 1017 e(�225.7/RT) cm3/mol sk1 = 8.0 · 1013 e(�271.7/RT) s�1

, 2.0 for CO2, 2.0 for CH4, 0.7 for Ar and He and 1.0 forenergies are kJ/mol.

Table 2Experimental conditions and boundary conditions usedin the numerical calculations

Non premixed Partially premixed

Fuel streamX C2H5OH 0.3 0.1385X N2

0.7 0.6803X O2

0 0.1812VF 29.8 cm/s 30.22 cm/sTF 340 K 327 K

Oxidizer streamX N2

0.79 0.79X O2

0.21 0.21VOx 30 cm/s 30 cm/sTOx 298 K 298 KTad 2070 K 2240 K

Symbols: X, mole fraction; V, duct exit velocity; T,temperature; subscripts: F, fuel stream; Ox, oxidizerstream; ad, adiabatic flame.

P. Saxena, F.A. Williams / Proceedings of the Combustion Institute 31 (2007) 1149–1156 1151

in corresponding flames of many other fuels; thispath contributes more than 20% of the fuelremoval in Fig. 1 and nearly 50% in the corre-sponding diffusion flame for the conditions ofTable 2. Ethylene concentrations will be high inthese flames because it is produced not only bythis direct decomposition but also by decomposi-tion of one of the hydroxy ethyl radicals formedby H abstraction, seen at the upper left ofFig. 1; (it will be even higher in the diffusion flamethan in the partially premixed flame since about70% of the fuel is calculated to go to C2H4 inthe diffusion flame). The other hydroxy ethyl rad-ical, as well as the ethoxy radical, at the upperright of Fig. 1, are seen to lead instead to acetal-dehyde, the peak concentrations of which will bequite appreciable, computationally about twicethat of formaldehyde (which arises from oxygenattack on vinyl and hydroxyl attack on ketene inthese particular flames, in addition to the familiaroxygen attack on methyl). The acetaldehyde con-centration nevertheless is still less than half that ofethylene. Acetaldehyde is seen in Fig. 1 to play amajor role in the formation of CH3 radicals,directly by CH3CHO + M fi CH3 + HCO + M,or indirectly, either through the set of reactionsCH3CHO + H fi CH2CHO + H2, CH2CHO fiCH2CO + H and CH2CO + H fi CH3 + CO orby radical attack forming CH3CO, which in turndecomposes to CH3 and CO. Ethoxy also appre-ciably contributes to formation of CH3 directly(the path at the far right), since nearly 40%of it decomposes by CH3CH2O + M fi CH3 +CH2O + M, the other 60% of passing throughacetaldehyde. Ketene is produced from acetalde-hyde via CH2CHO and also from the ethylenepath at the left, through vinyl and acetylene,which thereby provides an additional, relativelysmaller contribution to methyl. In this mecha-nism, both methane and ethane come only from

CH3, at the bottom of the figure. For simplicity,the fate of HCCO, namely forming CO andCO2, is not shown in the figure.

The paths shown in Fig. 1 are not quantitative-ly representative of other conditions or other com-bustion processes. For example, although the firstentry in Table 1 plays no significant role here, it isimportant in autoignition. The general ideas,however, such as the observation that ethyleneand acetaldehyde are very important stable inter-mediates in ethanol combustion, extend to all pro-cesses considered here.

3. Autoignition, laminar burning velocities anddiffusion-flame extinction

It is of interest to test predictions of the mech-anism against experimental data on combustionprocesses that are available in the literature. Thecomputer program CHEMKIN 3.7 [18] mainlywas employed for this purpose, although theFlameMaster [19] program also was often usedto make sure that the predictions of the two differ-ent programs were the same. The autoignitioncomputations were performed for homogeneous,adiabatic, isochoric conditions, with the maxi-mum rate of increase of temperature employedas the criterion to define the ignition time. Forthe cases computed this was found to correspondclosely to the maximum rate of increase of pres-sure and is a very good approximation to theshock-tube conditions to which the calculationsare applied; in these cases the dependence of theignition time on the ignition criterion is weak,and the results would be not too different ifisobaric rather than isochoric conditions hadbeen selected for the computation. The burning-velocity computations included multicomponenttransport, the Soret effect and radiant energyloss from CO2 and H2O bands, while the diffu-sion-flame computations included these as wellas an alternative mixture-averaged transportapproximation.

There are various sources of shock-tube igni-tion-delay data [20–22], and Figs. 2 and 3 showrepresentative comparisons. The agreements inFig. 2, where the experiments were performedin 90% Ar at different equivalence ratios u andpressures P, are quite good. It is seen that, underthese conditions, there is practically no depen-dence of the ignition delay on u, and there is asmall decrease in delay with increasing P, bothexperimentally and computationally. For theleaner conditions at higher pressures shown inFig. 3, where the dilution was 92% Ar [21], theexperimental pressure dependence of the ignitiontime is slightly greater than predicted, and theexperimental slopes of the curves are somewhatless than predicted. These differences, which areexhibited to an even greater extent by other

10

100

1000

0.55 0.6 0.65 0.7 0.75 0.8

1000/ T (K -1)

( emi

T yaleD

noiti

ngI

)s

P=1 bar, =0.5

P=1 bar, =1.0

P=1 bar, =2.0

P=2 bar, =2.0

Fig. 2. Comparisons of measured [20] and calculatedautoignition delay times for u = 0.5, 1.0 and 2.0 atP = 1 bar and for u = 2.0 at P = 2 bar, with 90% Ardilution; the experimental ignition criterion was firstvisible light emission.

10

100

1000

10000

0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95

1000/ T (K -1)

( emi

T yaleD

noiti

ngI

)s

P=2 barP=3.4 barP=4.6 barPrediction

Fig. 3. Comparisons of measured [21] and calculatedautoignition delay times for autoignition delay times foru = 0.25 at P = 2 bar, P = 3.5 bar and P = 4.6 bar, with92% Ar dilution; the experimental ignition criteria werepressure rise, OH emission and the chemiluminescentreaction between CO and O.

10

20

30

40

50

60

70

80

90

100

0.5 0.7 0.9 1.1 1.3 1.5 1.7Equivalence Ratio

)s/mc( seitic

oleV

gni

nru

B rani

maL

298 K363 K428 K453 KPrediction

Fig. 4. Comparison of measured [23] and calculatedlaminar burning velocities for ethanol–air mixtures atinitial temperatures of 298, 363, 428, 453 K atP = 1 atm.


0

50

100

150

200

250

300

350

400

450

500

0.1 0.15 0.2 0.25 0.3 0.35 0.4Fuel Mass Fraction in Fuel Stream

s( n

oitcnitx

E ta etaR

niartS e

diS-ri

A1-)

ExperimentsMulti-ComponentMixture-Average

Fig. 5. Comparison of measured [24] and calculateddiffusion-flame extinction strain rates at P = 1 atm forfuel-stream temperature TF = 323 ± 10 K and air tem-perature TOx = 298 K; distance between the exit ductswas 10 mm, and the strain rate was calculated frommeasured flow rates using a plug-flow formula [25].

mechanisms, also extend to other data inthis pressure range at richer conditions [21,22]and may reflect a combination of inadequaciesin the mechanisms and experimental difficultiesat the lower temperatures. For purposes ofcomparison, the prediction of the mechanism ofLi [6] at 2 bar is shown as light dashed curvesin Figs. 2 and 3; other mechanisms are in pooreragreement.

Figure 4 compares predictions of the presentmechanism with measurements of laminarburning velocities at various initial temperatures[23]. Agreements are seen to be good, exceptbetween equivalence ratios of 0.7 and 0.9, whereburning velocities are overpredicted by up to20%. The predictions in Fig. 4 agree with thoseof Li [6] and are in better agreement with thedata than are the predictions of a mechanism ofMarinov [5].

Measurements have been made of counterflowdiffusion-flame extinction at normal atmosphericpressure for fuel streams of prevaporized ethanolwith mass fraction in nitrogen varying between0.18 and 0.39, flowing against air at room temper-ature [24]. Figure 5 compares this data with pre-dictions of the present mechanism based on twodifferent transport-property descriptions, multi-component and mixture-averaged formulationsas implemented in CHEMKIN 3.7. There areknown deficiencies of transport-model implemen-tation in CHEMKIN 3.7 [26,27], and it is unclearwhich of the two is better. It is seen from Fig. 5that, although the mixture-averaged computationproduces excellent agreement with experiment, themulticomponent description, which might bethought to be better, exhibits noticeable differenc-es. At present it is uncertain whether the differenc-es between computation and experiment reflect


transport uncertainties, experimental departuresfrom plug-flow boundary conditions or deficien-cies in the mechanism. In any event, the sametypes of differences are shared by the other mech-anisms in the literature.

0

0.05

0.1

0.15

0.2

0.25

0.3

0 2 4 6 8 10 12

Distance (mm)

noitcar

F elo

M

C2H5OHO2N2/3H2OCOCO2H2Prediction

C2H5OHO2N2 /3H2OCOCO2H2Prediction

300

800

1300

1800

2300

0 2 4 6 8 10 12

Distance (mm)

)K( er

utarep

meT

0

0.003

0.006

0.009

0.012

0.015

0.018

noitcar

F elo

M

TemperatCH4C2H2+C2H4C2H6PredictionNOx

TCH4C2H2+C2H4C2H6PredictionNOx*100

b

a

Fig. 6. Ethanol partially premixed flame structures: (aconcentration profiles for C2H5OH, N2, O2, H2O, COCO2 and H2, (b) temperature profile and concentrationprofiles for CH4, C2H2 + C2H4, C2H6 and NOx.

4. Counterflow flame structures

Experiments on flame structures were per-formed using a counterflow burner described pre-viously [28]. In this burner, fine wire meshes at theduct exits promote plug-flow boundary conditionsthere. The opposing ducts, 23.1 mm in inner diam-eter with shielding annular nitrogen curtains, wereseparated by 12 mm in these experiments. Airflows through the top duct and fuel through thebottom duct at flow rates adjusted according toa momentum balance to maintain the stagnationplane halfway between the duct exits. An insulat-ed vaporizer, temperature-controlled to providethe desired ethanol mole fraction in nitrogen, gen-erates the fuel vapor which flows through a heatedline to the lower duct, first having been mixed withoxygen to reconstitute air for the partially pre-mixed experiments. Computer-controlled flowme-ters adjust the flow rates, which in theseexperiments provided an air-side strain rate of100 s�1 according to the plug-flow formula [25].

Temperature profiles were measured by anuncoated Pt–Pt13%Rh (Type R) thermocouple,except in the high-temperature region of the par-tially premixed flame. The maximum temperatureexperienced in the partially premixed flame wasmuch higher than that in the diffusion flame andsurpassed the melting point of Pt, necessitatinguse of an uncoated Pt-6%Rh vs. Pt-30%Rh (TypeB) thermocouple for near-flame-temperature mea-surements in the partially premixed flame. Stan-dard radiation corrections were made [29], butcatalytic effects were neglected, leading to temper-atures estimated to be about 120 K too high at thehighest temperatures, based on experimentalobservations and the ideas in the literature [29].

Following best established practice [28] gassamples were collected 5 mm off axis by a quartzmicroprobe, tip 88 lm inner diameter and168 lm outer diameter, contoured for rapid cool-ing to quench reactions and inserted radially tominimize flow disturbances. The location of thesampling probe and the thermocouple in the flowfield was determined using a digital camera with apixel size corresponding to a distance in the flowfield of approximately 17 lm. The samples fromthe quartz probe flowed steadily, through linesheated to prevent condensation, to a SRI 8610Cgas chromatograph, equipped with a 4.5 ft mole-sieve (80/100 mesh) for separating H2, O2 plusAr, N2, CH4, and CO and a 12 ft Porapaq Q col-umn for separating CO2, C2H6, C2H2 + C2H4,and C2H5OH. Temperature programming and

valve switching were employed to optimize theseparation performance of both columns. Thespecies eluting from the column were sensed bya thermal-conductivity detector (TCD) and aflame-ionization detector (FID). The chromato-grams were analyzed by in-house software, abso-lute mole fractions being determined bycomparing with runs of known samples. Since eth-ylene and acetylene appear as a single peak, theirsum was reported using the calibration for ethyl-ene as in earlier work [30]. Since argon elutedtogether with oxygen from the columns, a correc-tion was made for this [31]. Recently developedmethods [31] were employed to overcome difficul-ties in measuring water concentrations. Theexpected accuracy for the maximum concentra-tions is better than ±10% for species that can beclearly identified on either the FID or TCD.Hydrogen gives a very small signal on the TCD,so its concentration is estimated to be accurateto within ±25%, while the expected accuracy forwater is estimated to be ±20%.

Figures 6 and 7 show the results of these flame-structure measurements. In these figures, theexperimental results are compared with predic-tions of the chemical-kinetic mechanism, whichare shown by the solid curves. These computa-tions included multicomponent diffusion, theSoret effect and radiative heat loss from carbon

),

300

900

1500

2100

0 3 6 9 12

Distance (mm)

)K( er

utarep

meT

0

0.005

0.01

0.015

0.02

0.025

0.03

noitcar

F elo

M

TCH4C2H2+C2H4C2H6PredictionNOx

TCH4C2H2 +C2H4C2H6PredictionNox*200

0

0.1

0.2

0.3

0 3 6 9 12

Distance (mm)

noitcar

F elo

M



a

b

Fig. 7. Ethanol nonpremixed flame structures: (a) con-centration profiles for C2H5OH, N2, O2, H2O, CO, CO2

and H2, (b) temperature profile and concentrationprofiles for CH4, C2H2 + C2H4, C2H6 and NOx.


dioxide and water bands, with buoyancy neglectedand plug-flow boundary conditions. Figures 6aand 7a give the concentrations of the majorspecies in the partially premixed flame and in thediffusion flame, respectively. The excellent agree-ments between the experimental and computa-tional results in these figures are mutuallysupportive and help to verify the ability to mea-sure H2O with good accuracy. The minimum ofthe N2 mole fraction somewhat on the air sideof stoichiometry has also been seen for other fuelsand is associated with differential diffusion effectsand changes in the numbers of moles, since N2

is not consumed chemically. The disappearanceof the reactants at the premixed-flame and diffu-sion-flame sheets is evident in these figures, andthe H2, CO, H2O and CO2 profiles also are all justas would be expected for these conditions. Thecomputational results in these figures are indistin-guishable from those obtained when the mecha-nism of Li [6] is employed, but some of theearlier mechanisms exhibit different results andpoorer agreement with experiment.

Figures 6b and 7b show the profiles of temper-ature and of mole fractions of measured stableintermediates. The measured temperature profilesagree with computations within experimentalerror. The catalytic effect of the high-temperature

thermocouple, not accounted for in data reduc-tion, is seen between the premixed flame and thediffusion flame in Fig. 6b where the measured val-ues are too high by roughly 100 K; the displace-ment of the temperature profile on the air side inFig. 7b is within experimental positioning uncer-tainty. Because of the more stringent spatial reso-lution requirements in the diffusion flame,observed differences in profile shapes in Fig. 7cannot definitively be attributed to deficiencies inthe computational results.

The agreements of the profiles of methane andof ethylene plus acetylene (computationally about80% ethylene) are good in both of these figures,thereby lending experimental support to the pres-ent mechanism. In contrast, the mechanism of Li[6] produces poorer agreement, giving slightlyhigh predictions of C2H2 + C2H4 peaks andunderpredicting the height of the CH4 peaks bynearly a factor of two, the latter illustrated bythe light dashed curve in Fig. 6b. It may be notedthat in the diffusion flame the peak CH4 concen-tration is only about 60% that in the partially pre-mixed flame, while the peak C2H4 concentration is25% greater, illustrating the proportionally largercontributions of the paths on the left-hand side ofFig. 1 for the diffusion flame. The mechanismunderpredicts ethane concentrations somewhat,by about 30% for the partially premixed flameand nearly 50% for the diffusion flame; the under-predictions by other mechanisms [5,6] are compa-rable to this. This difference can be removed forthe present mechanism by decreasing the rate ofthe step C2H6 + H fi C2H5 + H2 by a factor oftwo (within uncertainties) or by adding a stepC2H5OH + H fi C2H6 + OH at a rate somewhatless than that of other H attacks on the fuel; thelatter of these has not been considered in studies[32] of this system and may not be likely.

5. Concentration profiles related to pollutants

Figures 6 and 7b also show computed NOx

profiles in these flames. Here NOx is defined asthe sum of NO and NO2, but in all cases NO dom-inates in the flames. In contrast to hydrocarbonflames, the thermal mechanism is much moreimportant than the prompt mechanism for alco-hol flames in these counterflow, atmospheric-pres-sure experiments because the CH concentrationsare smaller. They are entirely negligible in metha-nol flames and, for these ethanol flames, less thanhalf the values calculated for hydrocarbon flames.In Fig. 6b the peak NOx concentration, about40 ppm, is nearly twice that in Fig. 7b, mainlybecause the peak flame temperature of the partial-ly premixed flame is higher in these two experi-ments. In general, peak NO concentrations inethanol flames were calculated to be comparablewith those in methanol flames and about half

0

10

20

30

40

50

60

-10 -5 0 5 10 Distance (mm)

)m

pp(

noitcarf el

om

ON

NOPrediction

Fig. 8. Comparison of measured and calculated NOconcentration profiles for a partially premixed ethanol–air flame with fuel-side equivalence ratio u = 2.3 andplug-flow strain rate 53 s�1.


those in hydrocarbon flames. Although NO pro-files were not measured in these experiments, theyhave been measured earlier in our laboratory byT. Hiraiwa under quite similar conditions andare shown in Fig. 8 along with a curve giving com-puted results for those conditions. The closeagreement in Fig. 8 lends credence to the compu-tational NOx results.

For these types of counterflow flames, extend-ing to partial-premixing equivalence ratios of 1.5,peak methane concentrations in ethanol flamesare calculated to be about twice those in methanolor hydrocarbon flames, while peak total concen-trations of C2 species (sum of C2H2, C2H4 andC2H6) in ethanol flames are calculated to be com-parable with those of most hydrocarbon flames,more than twice those in methane flames andnearly one hundred times those in methanolflames. This last result may help to account forthe relatively high sooting tendency of ethanol(and higher alcohols); ethylene has been seen tobe a very significant decomposition product ofethanol in these flames. Alcohol diffusion and par-tially premixed flames also exhibit higher maxi-mum CO concentrations than hydrocarbonflames. These observations, along with the previ-ously noted high acetaldehyde and formaldehydeconcentrations in the ethanol flames studied here,suggest that pollutant-emission concerns will bedifferent for ethanol than they are for methanolor for hydrocarbons.

6. Conclusions

Although this work has shown that agreementsbetween experiments and predictions for near-atmospheric pressures are reasonably good,more work should be done in the chemical-kineticmodel in evaluating the performance of the

mechanism. This would necessitate additionalexperimental, chemical-kinetic and computationalwork. It would be of interest to measure addition-al species in these flames at normal atmosphericpressure and also to perform experiments andmake comparisons at higher pressures, since manyof the applications of interest are at elevated pres-sures. The current study is a step forward in theeffort to minimize the uncertainties in the predic-tions of ethanol combustion and to identify theassociated chemical-kinetic channels. Similar tomethanol, ethanol is expected to reduce NOx

emissions, but it may not be comparably effectivein soot reduction and may exacerbate concernsabout aldehydes. This study may contributetowards enhancement of understanding of ethanolcombustion and assessment of its utility as a fuel.

References

[1] M.P. Dunphy, P.M. Patterson, J.M. Simmie, Jour-nal of the Chemical Society Faraday Transactions 87(1991) 2549–2559.

[2] A.A. Borisov, V.M. Zamanskii, A.A. Konnov,V.V. Lisyanskii, S.A. Rusakov, G.I. Skachkov,Soviet Journal of Chemical Physics 9 (1992) 2527–2537.

[3] P. Dagaut, J.C. Boettner, M. Cathonnet, Journal deChimie Physique 89 (1992) 867–884.

[4] T.S. Norton, F.L. Dryer, International Journal ofChemical Kinetics 24 (1992) 319–344.

[5] N.M. Marinov, International Journal of ChemicalKinetics 31 (1999) 183–220.

[6] J. Li, Experimental and Numerical Studies of Etha-nol Chemical Kinetics, PhD thesis, Princeton Uni-versity, 2004.

[7] P. Saxena, F.A. Williams, Combust. Flame 145(2006) 316–323.

[8] S.C. Li, F.A. Williams, Combust. Flame 118 (1999)399–414.

[9] T. Shimizu, F.A. Williams, A. Frassoldati, Journalof Propulsion and Power 21 (2005) 1019–1028.

[10] M.M.Y. Waly, S.C. Li, F.A. Williams, Journal ofEngineering of Gas Turbine and Power 122 (2000)651–658.

[11] B. Varatharajan, F.A. Williams, Journal of Propul-sion and Power 18 (2002) 344–351.

[12] M.M.Y. Waly, S.C. Li, F.A. Williams, Proc.Combust. Inst. 28 (2000) 2005–2012.

[13] B. Varatharajan, F.A. Williams, Combust. Flame124 (2001) 624–645.

[14] M.V. Petrova, F.A. Williams, Combust. Flame 144(2006) 526–544.

[15] S.C. Li, F.A. Williams, Proc. of the Combust. Inst.26 (1996) 1017–1024.

[16] S.C. Li, F.A. Williams, Proc. Combust. Inst. 27(1998) 485–493.

[17] Available at: http://maemail.ucsd.edu/combustion/cermech/.

[18] R.J. Kee, F.M. Rupley, J.A. Miller, et al., 2003,Chemkin Collection, Release 3.7.1, Reaction DesignInc., San Diego, CA.

[19] H. Pitsch, Entwicklung eines Program-pakets zurBerechnung eindimensionaler Flammen am Beispiel

http://maemail.ucsd.edu/combustion/cermech/

http://maemail.ucsd.edu/combustion/cermech/


einer Gegenstvomdiffusionflamme, MS thesis,RWTH Aachen, Germany, 1993.

[20] K. Natarajan, K.A. Bhaskaran, Thirteenth Interna-tional Symposium on Shock Waves, 1982, pp. 834–842.

[21] M.P. Dunphy, J.M. Simmie, Journal of the Chem-ical Society, Faraday Transactions 87 (1991) 1691–1696.

[22] H.J. Curran, M.P. Dunphy, J.M. Simmie, C.K.Westbrook, W.J. Pitz, Proceedings of the Combus-tion Institute 24 (1992) 769–776.

[23] F.N. Egolfopoulos, D.X. Du, C.K. Law, Proceed-ings of the Combustion Institute 24 (1992) 833–841.

[24] R. Seiser, K. Seshadri, Western States Section, TheCombustion Institute, 2002 Spring Meeting, Univer-sity of California at San Diego, La Jolla, CA 92093-0411, March 25–26, 2002, Paper # 02S-27.

[25] K. Seshadri, F.A. Williams, International Journal ofHeat and Mass Transfer 21 (1978) 251–253.

[26] P. Paul, J. Warnatz, Proceedings of the CombustionInstitute 27 (1998) 495–504.

[27] Y. Dong, A.T. Holley, M.G. Andac, F.N. Egolfo-poulos, S.G. Davis, P. Middha, H. Wang, WesternState Section/The Combustion Institute (2004) WSS/CI 04S-19.

[28] R. Seiser, Nonpremixed Combustion of LiquidHydrocarbon Fuels, PhD thesis, Technical Univer-sity of Graz, 2000.

[29] C.R. Shaddix, Practical Aspects of CorrectingThermocouple Measurements for Radiation Loss,Western State Section/The Combustion Institute,University of Washington, Seattle, WA, October26–27, 1998, WSS/CI 98F-24.

[30] R. Seiser, L. Truett, D. Trees, K. Seshadri, Pro-ceedings of the Combustion Institute 27 (1998) 649–657.

[31] K. Tanoue, R. Seiser, K. Seshadri, Proceedings ofthe Third Joint Meeting of the US sections of TheCombustion Institute, The University of Illinois atChicago, Chicago, March 16–19, 2003, Paper A31.

[32] J. Park, Z.F. Xu, M.C. Lin, Journal of ChemicalPhysics 118 (2003) 9990–9996.

Comment

Eva Gutheil, Heidelberg University, Germany. As yougo from the non-premixed flame to partially premixedflame, a second green premixed flame zone appears ata certain degree of premixedness. How does this flameaffect the NO profile?

Reply. As the degree of partial premixing increases ata constant strain rate, the NO profile broadens, but thepeak remains at the diffusion flame, where the tempera-ture is highest. The variation of the peak NO concentra-tion with the degree of partial premixing depends,however, on the degree of dilution of the fuel stream,as a consequence of the competition between the promptand thermal pathways. Of our two experiments, the dif-fusion-flame experiment involved nitrogen-diluted fuelagainst air and had a lower flame temperature, resultingin a calculated peak NO concentration of 25 ppm, al-most all from the prompt mechanism. The partially pre-

mixed flame, on the other hand, involved a fuel–airmixture against air, resulting in a higher flame tempera-ture and a peak NO concentration of 40 ppm, about30% of which was attributable to the thermal mecha-nism. For the latter case, computations indicated veryslight increases in peak NO concentrations with increas-ing premixing until a fuel-side equivalence ratio of about10 was reached, beyond which, at lower equivalence ra-tios, there was a sharp drop in the peak NO concentra-tions with increasing premixing through reduction in theprompt pathways. If oxygen were added to the fuelstream in our diffusion-flame experiment, on the otherhand, computations predict significantly increased peakNO concentrations partially through onset of the ther-mal mechanism (which occurs at about an equivalenceratio of 5) as a consequence of the increased flame tem-perature, followed again by a sharp decrease at high pre-mixing through suppressing of the prompt mechanism.

numerical and experimental studies of ethanol flames

Documents