us

Available online 12 August 2010

Keywords:Multi-component fuelReduced mechanismAuto-ignition

meuto

stants of selected reactions. Using a similar approach to that used to develop the reduced PRF mechanism,

ns, aromatics, and cycloalkanes, with a wide range of molecularweights. In addition, the exact composition of these fuels variesbased on the source, distributor, and intended region and seasonof use. Because of the limitations of computational cost to deal witha large number of components inmodeling themixture preparationand subsequent combustionprocesses,most previous studiesmodelthe ignition/combustion characteristics of automotive fuels by rep-

species that exhibits common combustion characteristics of thefuel class. For example, n-heptane is widely used as a surrogatefor n-alkanes and iso-octane represents iso-alkanes.

Comprehensive detailed kinetic mechanisms have been com-piled for certain fuels. For example, Curran et al. [1,2] developedmechanisms to study the oxidation of n-heptane and iso-octane.Fisher et al. [3] developed a detailed kinetic reaction mechanismthat contained 264 species and 1219 reactions for methyl butano-ate, which is regarded as a simple but representative surrogate formethyl esters (biofuels). A detailed mechanism for toluene (thesmallest alkyl benzene) was developed by Bounaceur et al. [4],

* Corresponding author. Address: Engine Research Center, 1500 EngineeringDrive, ERB #1016A, Madison, WI 53706, United States. Fax: +1 608 262 6707.

Combustion and Flame 158 (2011) 6990

Contents lists availab

n

eviE-mail address: yra@wisc.edu (Y. Ra).As more attention is paid to methods for the reduction ofpollutant emissions, detailed information about the combustionchemistry in engines is becoming critical for improving engine per-formance. In particular, in HCCI/PCCI engine research, where com-bustion is a chemical-reaction-governed process, the importance ofaccurate chemical kinetic mechanisms and their computationalefciency is emphasized.

Practical engine fuels consist ofmulti-componentswith differentclasses of chemical structures. For instance, gasoline and diesel con-sist ofmany different classes of hydrocarbons, such as parafns, ole-

tion. However, as more detailed information about the combustionchemistry of the major components of the fuels and their effectson engine emissions is required for improving engine design, moreattention is being paid to the inclusion of more realistic multi-com-ponent compositions in simulations.

Ideally, each fuel component requires its own set of combustionspecies and reactions. However, the resulting reaction mechanismswould be computationally prohibitive, even with current state-of-the-art computational facilities. Thus, it is usual to representcomplex fuels in a class with a single, relatively simple, surrogateHCCISpray combustionInternal combustion engines

1. Introduction0010-2180/$ - see front matter 2010 The Combustdoi:10.1016/j.combustame.2010.07.019reduced mechanisms for the oxidation of n-tetradecane, toluene, cyclohexane, dimethyl ether (DME),ethanol, and methyl butanoate (MB) were built and combined with the PRF mechanism to form amulti-surrogate fuel chemistry (MultiChem) mechanism. The nal version of the MultiChem mechanismconsists of 113 species and 487 reactions. Validation of the present MultiChem mechanism was per-formed with ignition delay time measurements from shock tube tests and predictions by comprehensivemechanisms available in the literature.A combustion model was developed to simulate engine combustion with multi-component fuels using

the present MultiChem mechanism, and the model was applied to simulate HCCI and DI engine combus-tion. The results show that the present multi-component combustion model gives reliable performancefor combustion predictions, as well as computational efciency improvements through the use ofreduced mechanism for multi-dimensional CFD simulations.

2010 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

resenting them as blends of just two pure hydrocarbons (primaryreference fuel), i.e., iso-octane and n-heptane, as a rst approxima-Received in revised form 16 November 2009Accepted 22 July 2010

tion engines. Starting from an existing reduced mechanism for primary reference fuel (PRF) oxidation,further improvement was made by including additional reactions and by optimizing reaction rate con-A combustion model for IC engine combwith multi-component fuels

Youngchul Ra *, Rolf D. ReitzEngine Research Center, University of Wisconsin-Madison, United States

a r t i c l e i n f o

Article history:Received 3 September 2009

a b s t r a c t

Reduced chemical kinetictypical multi-component a

Combustio

journal homepage: www.elsion Institute. Published by Elseviertion simulations

chanisms for the oxidation of representative surrogate components of amotive fuel have been developed and applied to model internal combus-

le at ScienceDirect

and Flame

er .com/locate /combustflameInc. All rights reserved.

n anand includes 148 species and 1013 reactions. Silke et al. [5] pro-posed a comprehensive reaction mechanism with 1081 speciesand 4269 reactions for oxidation of cyclohexane, which is the sim-plest representative cycloalkane. For the oxidation processes ofC10C16 alkanes, Westbrook et al. [6] recently developed a complexreaction mechanism that consists of 2116 species and 8130reactions.

These comprehensive mechanisms are very useful to give de-tailed information about reaction pathways. However, with reac-tion mechanisms that consist of several hundred species andseveral thousand reactions, it is still much too costly to use a de-tailed chemical kinetic mechanism directly in engine combustionstudies using multi-dimensional CFD codes. It is necessary to de-velop chemical reaction mechanisms that retain the essential fea-tures of the fuel chemistry predicted by comprehensive reactionmechanisms, but with much improved computational efciencyin terms of memory usage and CPU time. The extent of comprehen-siveness required depends on the computational resources avail-able and the type of information desired from the simulation.

Signicant effort has been made to develop reduced mecha-nisms for the oxidation chemistry of surrogates of conventional en-gine fuels. For example, Seshadri et al. [7] developed a skeletalmechanism with 23 species and 34 reactions for n-heptane oxida-tion and applied it to study the structure of premixed n-heptaneames. Peters et al. [8] developed a skeletal mechanism with 35species and 56 reactions for n-heptane oxidation and further usedit to derive global mechanisms for different temperature regimesby applying steady-state assumptions to many of the intermediatespecies. Hasse et al. [9] employed a reduced mechanism (29 spe-cies and 48 reactions) that includes reaction pathways involvingintermediate hydrocarbon/oxygenate species for iso-octane oxida-tion in their study of the quenching of ames near cold walls.

Brakora et al. [10] developed a skeletal chemical reactionmechanism (41 species and 150 reactions) from the detailedmethyl butanoate oxidation mechanism and combined it with areduced n-heptane mechanism to give a reduced mechanism forbiodiesel. Recently, Ra and Reitz [11] developed a reduced mecha-nism for PRF oxidation that consists of 41 species and 130 reac-tions. The mechanism was validated with measurements fromshock tube tests, and HCCI and direct injection engine experimentsavailable in the literature, which demonstrated the mechanismsreliable performance for combustion predictions, as well as com-putational efciency improvements for multi-dimensional CFDsimulations.

A number of researchers have modeled the chemistry of multi-surrogate fuelmixtures. Dagaut andCathonnet [12] reviewed recentresearch on chemical kinetic modeling of kerosene combustion, aswell as the relevant experimental studies. They surveyed a largebody of experiments and kinetic modeling of kerosene combustion,showing how surrogate fuelmixtures, including an n-decane kineticmechanism with aromatic and other species, can predict sootingbehavior and overall heat release.

Buda et al. [13] discussed the importance of unied kineticmechanisms for alkanes over both the low and high temperatureregimes. Similarly, Ranzi et al. [14] proposed a lumped approachto generate mechanisms for n-decane, n-dodecane and n-hexadec-ane, and to reduce the size of each mechanism. They demonstratedthat heavy n-alkanes displayed the same kinetic behavior in bothhigh- and low-temperature regions and an overall kinetic schemecould be successfully extended directly.

Recently, the development of surrogate mixtures that representgasoline and diesel combustion behavior was reviewed by Pitzet al. [15] and Farrell et al. [16], respectively. Pitz et al. described

70 Y. Ra, R.D. Reitz / Combustiothe general composition of gasoline in terms of fuel classes, viz,dominant component(s) in n-alkanes, iso-alkanes, naphthenes,olens, aromatics, and oxygenates, which would be desirablesurrogate components for gasoline. They further discussed the rel-ative importance and availability of mechanisms for the majorcomponents in each class in terms of relevance to practical sys-tems, understanding of the mechanisms, and property information.Farrell et al. [16] suggested that lower molecular weight surrogateswould provide the most suitable basis for creating surrogate dieselfuels in the near term. The recommended surrogate componentsinclude n-decane, iso-octane, methylcyclohexane, and toluene.They also indicated the importance of liquid fuel vaporization pro-cesses in modeling the in-cylinder fuel/air/EGR mixture distribu-tion in diesel engine combustion, emphasizing the lack ofunderstanding of the implications of supercriticality with multi-component fuels. In addition, they stated that it is necessary to de-velop methods to generate substantially smaller mechanisms thanthose obtained from currently available reduction techniques,without sacricing chemical comprehensiveness.

More recently, Puduppakkam et al. [17] reported a surrogate-blending technique and proposed a surrogate-fuel blend to repre-sent a real gasoline fuel. In their model, detailed chemical-kineticsmechanisms for each component of the surrogate fuel were assem-bled and validatedwith kinetics experiments available in the litera-ture. They further tested the model with HCCI engine experiments.

In the present study, a reduced chemical kinetic mechanism forthe oxidation of representative surrogate components of typicalmulti-component automotive fuels is developed. Starting from anexisting reduced mechanism for primary reference fuel (PRF) oxi-dation, further improvement is made to the performance of heatrelease and emissions predictions by including additional reactionsand by optimizing reaction rate constants of selected reactions.Using a similar approach to that used for development of thereduced PRF mechanism, reduced mechanisms for oxidation ofn-tetradecane, toluene, dimethylether (DME), ethanol, and cyclo-hexane are built and combined with the PRF mechanism and anexisting reduced mechanism for methyl butanoate to form a mul-ti-surrogate fuel chemistry (MultiChem) mechanism, which will bedescribed in the following section. Validation of the present Multi-Chem mechanism is performed with ignition delay time measure-ments from shock tube tests and predictions by comprehensivemechanisms available in the literature.

A combustion model is also proposed to simulate engine com-bustion with multi-component fuels using the present MultiChemmechanism. In order to model the evaporation of multi-componentgasoline and diesel sprays in engines, a discrete multi-component(DMC) vaporization model developed by Ra and Reitz [18] andtested for evaporation processes of single and multi-componentfuel droplets and sprays in a constant volume chamber and enginecombustion [19,20], is employed. In the vaporization model, un-steady vaporization of multi-component fuel droplets and sprayswas considered for both normal and ash-boiling evaporation con-ditions. In addition, a characteristic diffusion length was dened todetermine the amount of vaporized fuel as a function of time in or-der to treat surface phase change under trans-critical conditions.The physical properties of multi-component fuels were modeledusing various correlations available in literature [21], as a functionof temperature, composition and, if necessary, pressure.

Application of the model to simulate HCCI and DI engine com-bustion is also presented.

2. Mechanism reduction methodology

2.1. Reduction method

The basic aim of mechanism reduction is to identify unimpor-

d Flame 158 (2011) 6990tant species and reaction pathways in order to reduce the complex-ity of chemistry mechanisms and the number of variables, but stillto maintain the important features of full schemes.

employed.In the method, the reduction was performed using a combina-

n antion of chemical lumping, graphical reaction ow analysis andelimination methods. It can be summarized in ve steps: (1)identication of the most essential species and reaction steps ofthe base mechanism using graphical reaction ow analysis, (2)elimination of unimportant species and reactions from the basemechanism to obtain a new reduced mechanism, (3) tests ofthe new reduced mechanism at appropriate operating conditionsand comparison of the proles of species concentrations of se-lected species with reference data for validation, (4) modicationof the reduced mechanism by adding/removing species and rele-vant reaction steps to/from the reduced mechanism or in order toimprove the prediction of species concentration histories, and (5)optimization of the reaction rate constants of selected reactionsto give the best agreement of ignition delay times with experi-mental measurements. In the nal stage of the procedure, the re-sults of sensitivity analysis are used to minimize the adjustmentof reaction rate constants from those of the base mechanisms. Formore details of the reduction methodology, readers are referredto Ref. [11].

2.2. Optimization of reaction rate constants

Once the reaction pathways of a reduced mechanism are estab-lished based on the reduction methods described above, the reac-tion rate constants of selected reaction steps can be optimized suchthat predicted ignition delay curves match experimental data orthose of reference detailed mechanisms. In the present study, thereactions to be optimized in the reduced mechanism for each sur-rogate fuel were selected from the ignition delay curve sensitivityanalysis. While conventional sensitivity analysis [34] assesses thetemporal sensitivity of temperature and species concentrations toindividual reaction steps during a single ignition process, the pres-ent approach enables an estimate of the global contribution ofindividual reaction steps to the ignition process over a wide rangeof temperature variation, such as that found during the compres-sion stroke of HCCI engines. For the purpose of the study, the igni-tion delay time in a constant volume ignition process is dened asthe time when a 400 K increase in mixture temperature above theinitial temperature is reached. Among the six parameter variationsexplored by Ra and Reitz [11], two variations of pre-exponentialVarious methods have been suggested to determine the impor-tance of species and reaction pathways in a mechanism. Sensitivityanalysis [2225], a reaction elimination method [26], the methodof detailed reduction [27], and chemical lumping techniques [28]are among the reduction methods that have been quite widely ex-plored. Each method has its own disadvantages, as well as advan-tages. Automation of reduction procedures has drawn muchinterest as well. Soyhan et al. [29] developed an automatic reduc-tion technique and applied it to reduce a natural gas mechanism,Montgomery et al. [30] have used the Computer Aided ReductionMethod (CARM) to automate the mechanism reduction process,and more recently, the direct relation graph method [31] hasbeen suggested and used to help automation of reductionprocedures.

Patel et al. [32], Brakora et al. [10] and Ra and Reitz [11] used acombination of chemical lumping, graphical reaction ow analysis[33] and elimination methods to generate reduced mechanisms forthe oxidation of n-heptane, methyl butanoate, and primary refer-ence fuel, respectively, from more comprehensive reaction mecha-nisms. In the present study, the method used by Ra and Reitz is

Y. Ra, R.D. Reitz / Combustiofactor, i.e., k1 = 2 and k2 = 0.5 times the base line value, were inves-tigated in the present study, since this is sufcient to demonstratesensitivities.2.2.1. Sensitivity coefcientsAn ignition delay sensitivity coefcient was calculated by

changing the pre-exponential factor of the baseline value by thefactors of k1 and k2, as A = k Abase. The equation to obtain the igni-tion delay sensitivity is

SigT 100 log10tk1 log10tk2 log10tbaselog10k1=k21

where Sig(T) is the ignition delay sensitivity coefcient in percentileat initial temperature T, tk1 and tk2 are the ignition delay times ob-tained using the factors of k1 and k2, respectively, and tbase is theignition delay time of the baseline case corresponding to initialtemperature T. If Sig is positive the ignition delay time shortens asthe reaction rate (pre-exponential factor) increases.

The sensitivity of the gradient of ignition delay time to the var-iation of reaction rate of a selected reaction was also monitored.Based on the argument of Westbrook [35], the ignition delay timeof the system can be reasonably formulated as tig ATeE

=T , whereA is a constant and E is the system activation energy. Therefore,the gradient of the logarithmic ignition delay times with respectto the initial temperature is given as d log tigdT 1T E

T2. This indicates

that gradient of the logarithmic ignition delay times is inherentlya function of system temperature and activation energy and theinuence of the system activation energy becomes more apparentas the system temperature decreases. Thus, at a given system tem-perature, the change of the temperature gradient of the logarith-mic ignition delay times with respect to the variation of reactionrate of an individual reaction reects the effect of the individualreaction on the system activation energy change.

Using the same change of pre-exponential factor as in the igni-tion delay sensitivity, the gradient sensitivity coefcient, Sgr(T), iscalculated as

SgrT dlog10tk1

dT dlog10tk2

dT

log10k1=k2 100 2

The unit of the gradient sensitivity is (s/K). Note that the varia-tion of the ignition delay curve shape at a given temperature is ob-tained as an absolute change of the gradient of the ignition delaycurves on the tigT plot in a piecewise interval containing the tem-perature point of interest. Graphically, if Sgr is positive the ignitiondelay curve tends to rotate in counter-clockwise direction on thetigT plot. The higher the absolute value of the gradient sensitivity,the more the ignition delay curve rotates.

In the present study, only newly added reactions were consid-ered in the sensitivity analysis, since the C0C3 stem reactions inthe base PRFmechanismhave been successfully validated in numer-ical models for engine studies [32,36,37] and were held in commonin building the MultiChemmechanism. An example of the sensitiv-ity method is given here for the tetradecane mechanism.

2.2.2. Sensitivity analysis of n-tetradecane auto-ignitionFigure 1 shows changes of the ignition delay curves correspond-

ing to variations of the pre-exponential factors (A = kAbase, k = 2.0and 0.5) for reactions that have signicant effects on the oxidationprocess among those listed in Table 1. Also, sensitivity coefcientplots are presented for each reaction case considered. FromFig. 1a, for example, it can be seen that reaction class II affectsdominantly the ignition delay times in low-to-NTC temperature re-gimes only (T < 1000 K). This is expected since the activation bar-rier of the reaction used in the mechanism is substantially low(35 cal/mol-K) and the OH radicals govern the ignition process.The ignition delay sensitivity coefcients are large positive values

d Flame 158 (2011) 6990 71in these temperature regimes, as shown in Fig. 1b, which indicatesthat the ignition delay curve transitionally shifts down signi-cantly by increasing the reaction rate of the reaction class II of

y se

nsiti

vity

[%]

n an1

10 d

elay

[ms]

baselinek=2k=0.5

72 Y. Ra, R.D. Reitz / Combustiothe reduced n-tetradecane mechanism. The variation of the igni-tion delay curve in terms of the local slope of the curve is well pre-sented by the corresponding variation of gradient sensitivitycoefcients, changing from negative (clockwise rotation) to posi-tive (counter-clockwise rotation) values.

As mentioned above, among the hydrogen atom abstractionreactions, abstraction by OH radicals (reaction class II) has a stronginuence on ignition delay times in low-T and the NTC regimes ofn-tetradecane oxidation. Reaction classes V and X, shown in Fig. 1e,f, k and l, mainly affect the temperature range of the NTC regime,

0.01

0.1

initial temperature [K]

igni

tion

igni

tion

dela

0.01

0.1

1

10

initial temperature [K]

igni

tion

dela

y [m

s]

baselinek=2k=0.5

igni

tion

dela

y se

nsiti

vity

[%]

0.01

0.1

1

10

initial temperature [K]

igni

tion

dela

y [m

s]

baselinek=2k=0.5

600 700 800 900 1000 1100 1200 1300 1400

600 700 800 900 1000 1100 1200 1300 1400

600 700 800 900 1000 1100 1200 1300 1400

igni

tion

dela

y se

nsiti

vity

[%]

(a)

(c)

(e)Fig. 1. Ignition sensitivity to variations of reaction rate constants of selected reactionsreaction class II, (c and d) reaction class III, (e and f) reaction class V, (g and h) reaction clain Table 1.10

15

20

25

102030

405060

sitiv

ity c

oeffi

cien

tignition delay sensitivitygradient sensitivity

d Flame 158 (2011) 6990while reaction classes III and IX inuence both the NTC and hightemperature regimes, signicantly. It is notable that reaction clas-ses V and X have the opposite effects in terms of the shifting direc-tion of the ignition delay curves. Enhancing the b-scission reaction(class X) of n-tetradecyl radicals tends to inhibit the low-tempera-ture branching that forms two OH radicals via ketoperoxide forma-tion/decomposition reactions.

The characteristic effects of n-tetradecane oxidation reactionsin most of the reaction classes shown in Table 1 are very similarto those reported for n-heptane [11], which indicates that a

-5

0

5

initial temperature [K]

-40-30-20

-100

grad

ient

sen

-5

0

5

10

15

20

25

initial temperature [%]

-40-30-20-100102030405060

grad

ient

sen

sitiv

ity c

oeffi

cien

t

ignition delay sensitivitygradient sensitivity

-5

0

5

10

15

20

25

650 750 850 950 1050 1150 1250 1350 1450

650 750 850 950 1050 1150 1250 1350 1450

650 750 850 950 1050 1150 1250 1350 1450initial temperature [K]

-40-30-20-100102030405060

grad

ient

sen

sitiv

ity c

oeffi

cien

t

ignition delay sensitivitygradient sensitivity

(b)

(d)

(f)of n-tetradecane oxidation. Stoichiometic n-tetradecane in air. P = 40 bar. (a and b)ss VI, (i and j) reaction class VII, (k and l) reaction class IX, (m and n) reaction class X

11

2

2

n se

nsiti

vity

[%]

n an1

10

ion

dela

y [m

s]baselinek=2k=0.5

Y. Ra, R.D. Reitz / Combustioreduced mechanism can be adjusted for different fuels with similarchemical structures based on ignition delay curve sensitivityinformation. However, contrary behavior is seen for classes IVand V. Hydrogen atom abstraction by oxygen molecules (reactionclass IV) has negligible inuence in the present n-tetradecaneoxidation, while it has a signicant effect on the depth of theNTC regime in the n-heptane oxidation branch.

This is attributed to the greater complexity of the n-tetradecanereaction pathways in the present reaction mechanism. Note thatn-tetradecane is assumed to also decompose to n-heptyl radicalsin the present reaction mechanism. Thus, the reaction class X leads

0.01

0.1

initial temperature [K]

igni

t

igni

tio

0.01

0.1

1

10

initial temperature [K]

igni

tion

dela

y [m

s]

baselinek=2k=0.5

igni

tion

dela

y se

nsiti

vity

[%]

0.01

0.1

1

10

initial temperature [K]

igni

tion

dela

y [m

s]

baselinek=2k=0.5

600 700 800 900 1000 1100 1200 1300 1400

600 700 800 900 1000 1100 1200 1300 1400

600 700 800 900 1000 1100 1200 1300 1400

igni

tion

dela

y se

nsiti

vity

[%]

(g)

(i)

(k)Fig. 1 (cont0

5

0

5

0102030405060

nsiti

vity

coe

ffici

ent

ignition delay sensitivitygradient sensitivity

d Flame 158 (2011) 6990 73to another important level of alkyl ketoperoxide formation/decom-position reactions, which is C7-ketoperoxide in the present reac-tion mechanism. This makes the effects of the reactions in thereaction classes shown in Table 1 become more complex than inthe mechanism for lower C-number species oxidation.

It should be noted that the sensitivity coefcients results of thepresent ignition delay curves of the reduced mechanism may notbe necessarily indicative of those of detailed mechanisms, becausethe reduced mechanism might not reect some possible effects ofomitted reaction pathways. Nevertheless, the sensitivity results ofthe reactions selected in the present study are believed to give

-5

0

5

initial temperature [K]

-40-30-20-10

grad

ient

se

-5

0

5

10

15

20

25

initial temperature [K]

-40-30-20-100102030405060

grad

ient

sen

sitiv

ity c

oeffi

cien

t

ignition delay sensitivitygradient sensitivity

-5

0

5

10

15

20

25

650 750 850 950 1050 1150 1250 1350 1450

650 750 850 950 1050 1150 1250 1350 1450

650 750 850 950 1050 1150 1250 1350 1450initial temperature [K]

-40-30-20-100102030405060

grad

ient

sen

sitiv

ity c

oeffi

cien

t

ignition delay sensitivitygradient sensitivity

(h)

(j)

(l)inued)

110io

n de

lay

[ms]

baselinek=2k=0.5

-5

0

5

elay

sen

sitiv

ity [%

]

01020304050

ensi

tivity

coe

ffici

entignition delay sensitivity

gradient sensitivity

cont

Table 1S1, S

74 Y. Ra, R.D. Reitz / Combustion and Flame 158 (2011) 6990Reaction classes modeled for low temperature oxidation of n-tetradecane. (species).

No. Reaction

_0.01

0.1

initial temperature [K]

igni

t

600 700 800 900 1000 1100 1200 1300 1400

(m)Fig. 1 (important information about how the performance of a reducedmechanism can be optimized with minimal adjustment of reactionrate constants of the mechanism.

3. MultiChem mechanism formulation

Starting from an existing reduced mechanism for primary refer-ence fuel (PRF) oxidation [11], further improvement was made byincluding additional reactions and by optimizing reaction rate con-stants of selected reactions. Using a similar approach to that usedfor development of the reduced PRF mechanism [11], reducedmechanisms for the oxidation of n-tetradecane, toluene, cyclohex-ane, dimethyl ether (DME), ethanol, and methyl butanoate (MB)were built and combined with the PRF mechanism to form thepresent multi-surrogate fuel chemistry (MultiChem) mechanism.

Table 2Detailed mechanisms selected to be the base mechanisms for the reduced mechanism for

Surrogate component Fuel family represented Base detailed mechanis

n-Tetradecane Heavy alkanes Westbrook et al. [6]Toluene Aromatics Bounaceur et al. [4] andCyclohexane Napthenes Silke et al. [5]Dimethyl ether Ethers Curran et al. [58,59]Ethanol Alcohols Marinov [55]Methyl butanoate Esters Fisher et al. [3]

I RH H R H2OII RH OH _R H2OIII RH HO2 _R H2O2IV RH O2 _R HO2V _R O2 R _OOVI _ROO O2 R-keto OHVII R-keto CH2O R0CO OHVIII R0CO R0 COIX _R S1 S2 S3X _R Y1 Y2XI R0 O=OH=CH3 R00 OH=H2O=CH4 HXII R0 X1 X2 X3XIII R00 Z1 Z2 Z3-20

-15

-10

650 750 850 950 1050 1150 1250 1350 1450initial temperature [K]

igni

tion

d

-50-40-30-20-10

grad

ient

s

(n)inued)

2, S3, X1, X2, X3, Y1,Y2, Z1, Z2, and Z3 are generic intermediate hydrocarbon

No. in Table A1 Class # in Ref. [1]As base mechanisms to reduce for each surrogate component,detailed mechanisms were chosen, as shown in Table 2. In this sec-tion, the mechanism formulation of each surrogate component se-lected in this study is briey mentioned. Detailed descriptions ofmechanisms for each MultiChem surrogate component are pro-vided in supplementary material [78].

It is notable that the present reduced mechanism does not sharethe full small molecule kinetics and themochemistry with the de-tailed mechanisms of interest, although the PRF base mechanismshares important major reactions and reaction pathways for smallmolecules with the detailed mechanisms and has been well vali-dated against detailed mechanisms as well as experimental datawith concentration proles of selected species. This may lead topotential discrepancy of performance between the reduced andthe detailed mechanisms. Also, the effects of the kinetics of smallmolecules missing in the reduced mechanism may depend on the

surrogate fuel components considered in the present study.

m No of species and reactions (species, reaction)

Detailed mechanism Present mechanism

(1668, 6449) (47, 181)Andrae et al. [42] (148, 1036) (1083, 4635) (50, 255)

(1081, 4269) (67, 287)(79, 351) (27, 96)(57, 383) (27, 117)(264, 1219) (42, 178)

21 2

22 2

23 2

24 2

25 10

26 12, 22, 23

27 2428 Low T branch + lower level ketoperoxides29, 30 Low T branch + lower level ketoperoxides

31 Low T branch + lower level ketoperoxides

35, 36, 37 Lower level H-abstraction38 Low T branch + lower level ketoperoxides40, 41, 42 Lower level alkyl radical decomposition

reaction conditions. Since the present reduced mechanism mainlyfocuses on low temperature regime, discrepancies of predicted re-sults between the reduced and detailed mechanisms may becomelarger at high temperature conditions. The authors leave the possi-ble further improvement of the present reduced mechanism in thehigh temperature regime for future study.

3.1. Primary reference fuel (PRF) mechanism

In order to make a more robust common pathway of low-Chydrocarbon oxidation, which is important in extending the rangeof fuel components included in the reduced mechanism to othersurrogate fuels, certain reaction pathways involving C3 intermedi-ate species in the PRF mechanism were updated to improve theprediction of UHC emissions speciation in HCCI operation undervarious equivalence ratios, especially for rich mixtures. In the ori-ginal PRF mechanism [11], less attention was paid to the speciationof UHC emissions of HCCI operation at rich conditions and theunderprediction of C2H2 and over-prediction of CH4 were foundto inuence signicantly the soot modeling that employs C2H2 asan inception species. By improving the prediction of gas phasereactions related to formation of C2H2, the reduced mechanismcould capture emissions characteristics much better without anycost of performance of ignition timing prediction.

Four more reactions that involve propene (C3H6) and active rad-icals were added and ve more reactions were added to supple-ment propenyl (C3H5) reaction pathways. In addition, ve morereactions of C3H4 (methylacetylene or propadiene) oxidation wereadded. Ethane oxidation pathways were improved, as well, by add-ing two reactions to consider the ethylperoxide formation path-

ways and two H-atom abstraction reactions by H and CH3radicals. Ethylene oxidation was supplemented with the additionof two H-atom abstraction reactions by H and O2 molecules andone O-atom addition reaction to form CH3 and HCO. In addition,a pressure-dependent decomposition reaction of methoxy (CH3O)radicals to form formaldehyde and H-atom, and a reaction ofCH2CO with OH to form CH3O + CO were added to improve the per-formance of the reduced mechanism. Finally, the reaction rate con-stants of several CH4 oxidation reactions were updated with thosein a more recent detailed mechanism [38].

3.2. n-Tetradecane mechanism

Due to its relative abundance in the composition and the close-ness of the physical properties to those of typical diesel fuels [39],it is reasonable to select n-tetradecane as one of the major surro-gate components to represent the chemical kinetics as well asthe physical properties of diesel-like fuels. With the n-tetradecanereaction branch (1668 species, 6449 reactions) in the parent com-prehensive mechanism by Westbrook et al. [6,40] being used as abase mechanism (see Table 2), the basic reactions pathways ofthe reduced n-tetradecane mechanism were formulated accordingto the generic form of the reaction equations that were developedfor n-heptane and iso-octane oxidation. As part of the mechanism,twelve major reaction pathways were used to describe the low-temperature reaction branches from n-tetradecane to intermediatehydrocarbon species that contain up to three carbons. Basically,one chemical reaction equation was assigned to each step of thegeneric form of the reaction equations, as shown in Table 1, except

Y. Ra, R.D. Reitz / Combustion and Flame 158 (2011) 6990 75C14H30

C14H29 C14H29O2

C14H28OOH

C12H25CO

C14ket

C12H23

O2C14H28OOH

C12H25

C3H4, C2H4, C2H3, CH3CHO,

C4H7

C3H6, C2H4, C7H15

C3H6, CH3

C3H5, C2H3, C7H15

C5H10

C4H9

C4H9

C3H7, C3H5, C2H5, CH3CHO, CH3CO, CH2O, HCO, CH3

C7H15 n-heptane branch

CH2O

C14H30

C14H29 C14H29O2

C14H28OOH

C12H25CO

C14ket

C12H23

O2C14H28OOH

C12H25

C3H4, C2H4, C2H3, CH3CHO,

C4H7

C3H6, C2H4, C7H15

C3H6, CH3

C3H5, C2H3, C7H15

C5H10

C4H9

C4H9

C3H7, C3H5, C2H5, CH3CHO, CH3CO, CH2O, HCO, CH3

C7H15 n-heptane branch

CH2OCH3O, CH2O, CO CH3O, CH2O, CO

Fig. 2. Major reaction pathways of tetradecane oxidation.for reaction class IX and XIII. Two and three reactions were as-signed to reaction classes IX and XIII, respectively, to ensure vari-ous decomposition species from the radical. In order to make thereaction pathway between C12-species and C3-species, reducedsub-mechanisms for oxidation of n-pentene (C5H10) and butenyl

Toluene

Benzyl

C6H5CHO

C6H5

C6H6

C5H5

C5H4OH

OC6H4O C6H5O C6H4OH

C5H4O

C6H5OH

C5H5O

C4H6

C4H5 C4H4

C4H3

C2H2, HCO

C2H2, C2H3

C2H2, C2H3

C2H2

C5H3O C2H2, HCCO

C2H2, CH2CO, HCCO

C2H2, C2H3, C2H4, HCO

C4H4, C2H3

OC6H4CH3

Toluene

Benzyl

C6H5CHO

C6H5

C6H6

C5H5

C5H4OH

OC6H4O C6H5O C6H4OH

C5H4O

C6H5OH

C5H5O

C4H6

C4H5 C4H4

C4H3

C2H2, HCO

C2H2, C2H3

C2H2, C2H3

C2H2

C5H3O C2H2, HCCO

C2H2, CH2CO, HCCO

C2H2, C2H3, C2H4, HCO

C4H4, C2H3

OC6H4CH3(b)Fig. 3. Major reaction pathways of toluene oxidation.

(C4H7) were included in the present mechanism, as shown in Fig. 2.Overall, a total of 11 species and 35 reactions were added to thebase PRF mechanism to generate the n-tetradecane oxidationbranch in the present reaction mechanism. The detailed reactionsand their rate constants are listed in Table A1 [78].

3.3. Toluene mechanism

Toluene is the most abundant aromatic hydrocarbon in gaso-lines (up to 10%), with a research octane number of 120 [39].Although it was reported by Roubaud et al. [41] that the type ofaromatic present in practical engine fuels inuences auto-ignitionat low temperatures, as a rst approximation the simple chemicalstructure and oxidation reaction pathways of toluene are regardedto be well representative of the characteristics of aromatic fuels.Based on the mechanisms by Bounaceur et al. [4] and Andraeet al. [42], the major reaction pathways which include primary tol-uene oxidation pathways and subsequent secondary reactionswere modeled in the present study. The primary reaction pathwaysare simplied as two reaction ows from toluene: (1) toluene tobenzaldehyde, and (2) toluene to benzene. The secondary reactionpathways are involved with phenyl radicals and benzene to givelower C-number hydrocarbon species and radicals. In the present

study, the reactions for C3C6 hydrocarbons such as C5H5, C4H5,C4H4, C4H3, etc., were adopted from the benzene oxidation mecha-nism of Bounaceur et al. [4], and combined with the C0C3 hydro-carbon reactions of the base PRF reaction mechanism [11]. A totalof 20 species and 121 reactions were added to the base PRFmechanism to account for aromatics in the present MultiChemmechanism, as shown in Fig. 3, and the reactions and their rateconstants are listed in Table A1 [78]. Note that the number ofspecies and reactions added to the base mechanism is relativelylarge compared to those for other components considered in thepresent study. This is due to the fact that toluene has a ring struc-ture while other components mainly have straight chain struc-tures. Due to this characteristic structure, the toluene oxidationprocess involves ring-opening stages. This leads to substantiallydifferent reaction pathways and intermediate species from thoseof other components.

3.4. Cyclo-hexane mechanism

Cycloalkanes are an important chemical class of hydrocarbonsthat make up a signicant portion in diesel (up to 35%), jet (upto 20%), and gasoline (up to 10%) fuels [15,16]. As a representa-tive surrogate for cycloalkanes, cyclohexane was chosen, since not

Table 3Cyclohexane ring-opening reactions modeled in the present reaction mechanism.

Ring-opening reaction Combined reactions Reaction in the present mechanism

cyOC6H9O = OC5H8CHO NA cyOC6H9O = OC5H8CHOcyC6H9O = C5H8CHO C2H2 + C3H6CHO = C5H8CHO cyC6H9O = C2H2 + C3H6CHOHOOC5H8CHO + OH = cyO2C6H10OOH cyC6H11O2 = cyC6H10OOH cyC6H11O2 + O2 = HOOC5H8CHO + OH

cyC6H10OOH + O2 = cyO2C6H10OOHC5H10CHO = cyC6H11O C5H10CHO = C2H4 + C3H6CHO cyC6H11O = C2H4 + C3H6CHOC6H11 = cyC6H11 NA C6H11 = cyC6H11C6H10OOH = cyC6H10OOH cyC6H11O2 = cyC6H10OOH cyC6H11O2 = C4H6OOH + C2H4

C4H6OOH + C2H4 = C6H10OOH

C

cy

OC

-

C

cy

OC

-

76 Y. Ra, R.D. Reitz / Combustion and Flame 158 (2011) 6990chx

cyC6H11

cyC6H11O

C2H4, C3H6

cyC6H11O2

HOOC6H9OOH

HOOC5H8CHO

OC5H8CHO

C2H3CHO, C2H4, HCO

CHOC4H7CHO

C5H10CHO

C3H6CHO

- C2H4

C2H3CHO, C3H6, C2H4, CH2CHO, CH3

C6H11C2H4, C4H7, C2H5, C4H6

chx

cyC6H11

cyC6H11O

C2H4, C3H6

cyC6H11O2

HOOC6H9OOH

HOOC5H8CHO

OC5H8CHO

C2H3CHO, C2H4, HCO

CHOC4H7CHO

C5H10CHO

C3H6CHO

- C2H4

C2H3CHO, C3H6, C2H4, CH2CHO, CH3

C6H11C2H4, C4H7, C2H5, C4H6Fig. 4. Major reaction pathwaysHOCO, C2H4

hexe

C6H6cyC6H10OOH

cyO2C6H10OOH

cyO2C6H10O

cyOC6H9O

OC5H8CHO

C3H6CHO3H4CHO

- CH2COC2H4

C2H3CHO, C3H6,

cyC6H9OOH

cyC6H9O

cyC6H9O

C5H8CHO- C2H2

cyC6H9 cyC6H8

C2H4, C2H2

HOCO, C2H4

hexe

C6H6cyC6H10OOH

cyO2C6H10OOH

cyO2C6H10O

cyOC6H9O

OC5H8CHO

C3H6CHO3H4CHO

- CH2COC2H4

C2H3CHO, C3H6,

cyC6H9OOH

cyC6H9O

cyC6H9O

C5H8CHO- C2H2

cyC6H9 cyC6H8

C2H4, C2H2C2H4, CH2CHO, CH3C2H4, CH2CHO, CH3

of cyclohexane oxidation.

radicals, followed by successive additions of oxygen leading to

important characteristic oxidation processes of alcohols. Therefore,

ployed in the MultiChem mechanism, as shown in Fig. 6 and listedin Table A1 [78].

3.7. Methyl butanoate (MB) mechanism

It is believed that the overall oxidation characteristics of biodie-sel can be reasonably represented by those of simpler surrogatefuels that are in the same ester class [60,61]. One of the simplestfuels that has been used as a biodiesel surrogate for kinetic model-ing is methyl butanoate (MB, CH3CH2CH2(C@O)OCH3), which con-tains the essential chemical structure of the long-chain fatty acidsand a shorter, but similar, alkyl chain.

Brakora et al. [10] developed a skeletal chemical reactionmechanism (41 species and 150 reactions) and validated it formulti-dimensional engine combustion simulations. The reduced

n and Flame 158 (2011) 6990 77ethanol was selected as a surrogate fuel to represent the alcohols.A detailed reaction mechanism of Marinov [55] with 57 species

and 383 reactions was used as the base ethanol mechanism. Froma ux analysis four major reaction pathways were identied. Themajor reaction pathways include dominant reactions at high andlow temperatures. With further eliminating relatively unimportantreaction pathways from the reduced mechanism at no signicantexpense of computational accuracy, four species (ethanol,CH3CHOH, CH3CH2O, CH3CHO) and 31 reactions were newly com-bined with the base PRF mechanism to generate the MultiChemmechanism, as shown in Fig. 5 and listed in Table A1 [78].

3.6. Dimethyl ether (DME) mechanism

The CO bond energy (362 kJ/mol [56]) is smaller than that ofthe CH bond (407 kJ/mol [56]) and the distortion of the CObonds in the DME molecule weakens the bonding strength. Thus,the CO bond breaks easier than the CH bond. This enables thechain reaction to start at relatively low temperatures, which leadsto the low auto-ignition temperature. In addition, its simple chem-ical structure that has no CC bond to act as a seed for polymeriza-tion, which leads to polyaromatic hydrocarbons, and its highoxygen content enable substantial reduction of particulate matterat a slight expense of NOx [57].

In the present study, a comprehensive reaction mechanism ofCurran et al. [58,59] was used as the base DME mechanism. In or-chain-branching pathways of the NTC region.Among the ring-opening reactions modeled in the comprehen-

sive mechanism, a total of six ring-opening steps were consideredin the present reaction mechanism, as listed in Table 3. In the table,the reactions from the comprehensive mechanism that were com-bined are listed, as well. Some of the ring-opening reactions werecombined with other reactions to make reduced steps in the pres-ent reaction mechanism. A total of 17 species and 53 reactions todescribe cyclohexane oxidation were added to the base PRF mech-anism to form the present MultiChem mechanism, as shown inFig. 4 and listed in Table A1 [78].

3.5. Ethanol mechanism

Important intermediates of the oxidation process of alcoholshave been identied as acetaldehyde, acetic acid, formaldehyde,and formic acid by many researchers [52,53]. Among them, aceticacid and formic acid are attributed to minor reaction pathways[54]. Much of the chemistry for alcohol compounds should be rel-evant to ethanol [54], where reaction pathways include most of theonly there are many experimental data [4348] and detailedkinetic mechanisms [5,45,4951] available in the literature, butalso it is believed that cyclohexane, as the simplest cycloalkane,can represent the major reaction characteristics of cycloalkane oxi-dation processes [5].

In the present study, the comprehensive mechanism (1081 spe-cies, 4269 reactions) by Silke et al. [5] was used as the parentmechanism. The general features of cyclohexane oxidation are thatcyclohexane oxidizes at high temperatures mainly via unimolecu-lar decomposition leading to linear products, H-atom abstractionleading to both dehydrogenation and formation of benzene, andb-scission reactions that break the cyclic ring. At lower tempera-tures, the dominant reaction path for cyclohexane is H-atomabstraction from the parent fuel molecule by OH, HO2, and other

Y. Ra, R.D. Reitz / Combustioder to model DMEs primary reaction pathways, a total of eightadditional species and 36 additional reactions were newly coupledwith the stem reactions for the low-C-number hydrocarbons em-mechanism was generated from the detailed methyl butanoateoxidation mechanism (264 species and 1219 reactions) of Fisheret al. [3]. Adopting the reduced MB mechanism by Brakora et al.with minor modication of pathways in the present study, a totalof nine species and 44 reactions were newly added to the presentMultiChem mechanism to describe the MB oxidation pathways, aslisted in Table A1 [78].

3.8. Co-oxidation reactions

There is no general information available about the importanceof co-oxidation reactions of fuel components in the combustionchemistry of a blended or multi-component fuel. Several research-ers [6265] found that co-oxidation reactions between the individ-ual fuel components were important and the range of co-oxidationreactions depended on the variety of species used in the mecha-nism. However, it has also been suggested that interactions be-tween the fuel components are unimportant and it is possible todevelop a reduced mechanism for mixtures by combining the re-duced mechanisms of the single components [64].

In the present study, several co-oxidation reactions between thesurrogate components were considered in order to take the possi-ble contribution of those reactions into account without any signif-icant additional computational cost (no additional number ofspecies). Since only dominant reaction pathways described withsingle lumped species were used to model the branching reactionsfor each surrogate fuel considered in the present study, only co-oxidation reactions that involve the radicals produced from oneof the fuels through H-atom abstraction (e.g., alkyl, benzyl, andcyclohexyl radicals) and the other fuels, R1 + R2HM R1H + R

2, were

included in the reduced mechanism. According to the argument byAndrae et al. [62], co-oxidation reactions between similar fuels

C2H5OH

HOC2H4O2

C2H4OHCH3CH2OCH3CHOH

CH3CHO

CH2CHO

CH2OHC2H4C2H5

CH3CHO

C2H5OH

HOC2H4O2

C2H4OHCH3CH2OCH3CHOH

CH3CHO

CH2CHO

CH2OHC2H4C2H5

CH3CHO

CH2OHC2H4C2H5

CH3CHOCH2O, CH2CO, CH3CH2O, CH2CO, CH3

Fig. 5. Major reaction pathways of ethanol oxidation.

the consistency of the thermodynamic properties between the re-

CH3OCH2

CH3OCH2O

CH3OCHO

CH3OCO

CH3OCH3

CH3OCH2O2

CH3OCH2OOH

CH3O, CH2O

CH2OCH2OOH

O2CH2OCH2OOH

HO2CH2OCHO

OCH2OCHO

HOCH2OCO

HOCH2O

HCO2, CH2O

CH3O, CH3

HCO2, CH3

CH3OCH2

CH3OCH2O

CH3OCHO

CH3OCO

CH3OCH3

CH3OCH2O2

CH3OCH2OOH

CH3O, CH2O

CH2OCH2OOH

O2CH2OCH2OOH

HO2CH2OCHO

OCH2OCHO

HOCH2OCO

HOCH2O

HCO2, CH2O

CH3O, CH3

HCO2, CH3

Fig. 6. Major reaction pathways of d

78 Y. Ra, R.D. Reitz / Combustion and Flame 158 (2011) 6990only need be considered, and co-oxidation reactions between fuelsthat are substantially different can be neglected. Other co-oxida-tion reactions such as those between fuels and alkylperoxy radicalsare not included in the present work, but may be added later to en-hance the completeness of the co-oxidation pathways. The rateconstants for the co-oxidation reactions were set to those usedfor the PRF mechanism by Ra and Reitz [11] and listed in TableA1 [78].

With reactions pathways for the surrogate components consid-ered in the present study combined together, the nal version ofthe MultiChem mechanism consists of 113 species and 487 reac-CH3O, CH3, CO2, COCH3O, CH3, CO2, COtions, as shown in Table A1 [78].

3.9. Thermodynamic properties

The thermodynamic properties of species considered in thepresent reaction mechanism are based on those used in the parentmechanisms. When a generic species was selected to represent its

0.01

0.1

1

10

1000/T [1/K]

igni

tion

dela

y [m

s]

PRFMultiChemExp, Fieweger et al. (1997)

0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5

(a)Fig. 7. Comparison of ignition delay time predictions between the present MultiChem minitial pressure was 40 bar. Shock tube tests are from Ref. [75]. (a) Stoichiometric iso-ocduced and parent mechanisms was maintained as much as possi-ble. Similarly, transport properties of species can be set, althoughno transport properties were used in the present study since vali-dation was performed for auto-ignition only.

4. Experiments for comparisonisomers the thermodynamic data of the species that existed mostin the oxidation process were used for its properties. In this way,

HCO2H

HCO2

H, CO3

HCO2H

HCO2

H, CO3

imethy ether (DME) oxidation.4.1. Ignition delay time experiments

Ignition delay predictions using the present reaction mecha-nism were validated using experimental data available in theliterature. For the n-tetradecane mechanism, experimental mea-surements of ignition delays at pressures close to engine combus-tion conditions are very rare. Shen et al. [66] recently reportedignition delay measurements of n-tetradecane/air mixtures in

0.01

0.1

1

10

100

0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.51000/T [1/K]

igni

tion

dela

y [m

s]

PRFMultiChemExp, Fieweger et al. (1997)

(b)echanism and the reduced PRF mechanism [11] for various initial temperatures. Thetane/air mixtures, (b) stoichiometric n-heptane/air mixtures.

n anY. Ra, R.D. Reitz / Combustioshock tube tests at intermediate-to-high temperatures andelevated pressures for two different equivalence ratios. This exper-

0

20

40

60

80

100

120

140

160

-30 -20 -10 0 10 20 30crank angle [deg atdc]

pres

sure

[bar

]

0

100

200

300

400

500

600

700

800H

RR

[J/d

eg-g

]P, mechanism of Ref [11]P, present mechanismHRR, mechanism of Ref [11]HRR, present mechanism

(a)Fig. 8. Improvement of PRF mechanism in the MultiChem mechanism. (a) Pressure andcombustion condition (O2 = 8.58%, /overall 0.68, engine speed = 2000 rev/min), (b)(EGR = 50%) among the LLNL detailed mechanism [38], the PRF mechanism by Ra and R

0.01

0.1

1

10

0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5

igni

tion

dela

y [m

s]

1000/T [1/K]

present mechanism, phi=1.0Westbrook et al., phi=1.0Exp, Shen et al., phi=1.0present mechanism, phi=0.5Westbrook et al., phi=0.5Exp, Shen et al., phi=0.5

(a)Fig. 9. Comparison of prediction of ignition delay times of n-tetradecane oxidation at conand prediction by detailed mechanisms [6]. (a) Ignition delay times of n-tetradecane/airstoichiometric n-heptane/air and n-tetradecane/air mixtures at 40 bar initial pressure.d Flame 158 (2011) 6990 79imental data was used for validation of the n-tetradecane oxidationpredicted by the present reaction mechanism.

0.00.20.40.60.81.01.21.41.61.82.0

equi

vale

nce

ratio

T=1500 K

T=1200 K

CO > 88 g/kg-f

CO < 88 g/kg-f

UHC > 20 g/kg-f

UHC < 20 g/kg-f

0.00.20.40.60.81.01.21.41.61.82.0

equi

vale

nce

ratio

T=1500 K

T=1200 K

CO > 88 g/kg-f

CO < 88 g/kg-f

UHC > 20 g/kg-f

UHC < 20 g/kg-f

0.00.20.40.60.81.01.21.41.61.82.0

-20 0 20 40 60 80 100crank angle [deg atdc]

equi

vale

nce

ratio

T=1500 K

T=1200 K

UHC > 20 g/kg-f

CO > 88 g/kg-f

UHC < 20 g/kg-f

CO < 88 g/kg-f

Mechanism of Ra and Reitz [11]

Present mechanism

LLNL mechanism [36]

(b)HRR proles of HCCI combustion of PRF0 under a medium load low temperature

Comparison of emissions and gas temperature predictions of HCCI combustioneitz [11], and the PRF mechanism in the present study.

0.01

0.1

1

10

0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5

igni

tion

dela

y [m

s]

1000/T [1/K]

present mechanism, n-tetradecanepresent mechanism, n-heptane

(b)stant volume between the present reaction mechanism and shock tube test data [66]mixtures at 40 bar initial pressure, (b) comparison of ignition delay times between

For the toluene mechanism, three shock tube experimental re-searches were employed; experiments by Bounaceur et al. [4] forhigh temperature toluene/O2/Ar mixtures, by Davidson et al. [67]for toluene/air mixture at intermediate-to-high temperatures,and by Herzler et al. [68] for n-heptane/toluene/air mixtures atlow-to-intermediate temperatures.

In order to validate the performance of the reduced cyclo-hex-ane mechanism, rapid compression machine (RCM) data by Lem-aire et al. [45] and shock tube test data by Daley et al. [43] wereused. The RCM data, which was originally obtained in pressurerange of 1114 atm, was scaled to 12.5 bar using the pressure scal-ing proposed by Daley et al., ignition delay, sP1.1, in order to becombined with the shock tube data by Daley et al.

For validation of the DME oxidation reactions of the presentreaction mechanism, shock tube tests by Pfahl et al. [69] wereused. In the experiments, ignition delay times of stoichiometricDME/air mixtures were measured at two different initial pressures,one of which is close to engine conditions.

For the MB mechanism, shock tube tests by Dooley et al. [61]were used for validation. In the experiments, high temperatureMB/O2/Ar mixtures with various equivalence ratios were testedat two different reected shock pressures. In the present studythe measurements at 4 atm reected shock pressure were used.

For validation of the ethanol mechanism as an individual compo-nent, predictions using the corresponding detailed mechanismswere compared with those by the present reaction mechanism.

Auto-ignition measurements in shock tube tests by Gauthieret al. [70] were used to validate the performance of the presentreaction mechanism for ignition predictions of gasoline surrogatemixtures.

They tested two ternary surrogate mixtures consisting ofiso-octane, toluene, and n-heptane; Surrogate-A (63/20/17%) andsurrogate B (69/14/17%) by liquid volume, as well as RD387 gaso-line (see Table D1 [78]). The measured ignition delay times ofgasoline/air and ternary surrogate/air mixtures for various temper-atures, equivalence ratios and pressures were compared with thepredicted values using the same ternary surrogate fuels proposedby Gauthier et al.

Also, based on the measured composition of RD387 gasoline[70], a 4-component surrogate model is proposed in the presentstudy and its ignition delay predictions were compared with themeasured data. Ten major components of the gasoline, i.e., cyclo-pentane, toluene, ethylbenzene, m-xylene, n-pentane, n-heptane,isopentane, 3-methylhexane, 2-methylhexane and iso-octane aregrouped into four surrogate components (cyclohexane, toluene,n-heptane and iso-octane) according to their chemical classes.

0.001

0.01

0.1

1

10

0.55 0.6 0.65 0.7 0.75 0.8

igni

tion

dela

y [m

s]

Exp, Bounaceur et al.mechanism of Bounaceur et al.mechanism of Andrae et al.present mechanism

0.001

0.01

0.1

1

10

0.6 0.65 0.7 0.75 0.8

igni

tion

dela

y [m

s]

Exp, Bounaceur et al.mechanism of Bounaceur et al.mechanism of Andrae et al.present mechanism

tant

80 Y. Ra, R.D. Reitz / Combustion and Flame 158 (2011) 69901000/T [1/K]

0.01

0.1

1

10

100

1000

0.7 0.8 0.9 1 1.1 1.2

igni

tion

dela

y [m

s]

1000/T [1/K]

present mechanismmechanism of Bounaceur et al.mechanism of Andrae et al.Exp, Davidson et al.

(a)

(c)Fig. 10. Comparison of prediction of ignition delay times of toluene oxidation at cons

prediction by detailed mechanisms [4,42]. (a) Ignition delays of toluene/O2/Ar mixturesO2/Ar mixtures at equivalence ratio of 0.5 and pressures of 79.5 bar, (c) ignition delays otoluene/air mixtures at 40 bar.1000/T [1/K]

0.01

0.1

1

10

100

1000

0.7 0.8 0.9 1 1.1 1.2 1.3

igni

tion

dela

y [m

s]

1000/T [1/K]

present mechanismBounaceur et al.Andrae et al.

(b)

(d)volume between the present reaction mechanism and experimental data [4,67] and

at equivalence ratio of 1.0 and pressures of 79.5 bar, (b) ignition delays of toluene/f stoichiometric toluene/air mixtures at 50 bar, (d) ignition delays of stoichiometric

For detailed information about the composition of the gasoline andmodel surrogate fuel, refer to Ref. [70] and Table D1 [78]. Since thegrouping method employed in this study does not adjust the pro-portion of individual surrogate components, the average propertiesof the surrogate mixtures may be different from the target mix-tures. For example, the average molecule weight and H/C ratio ofthe 4-component surrogate mixtures are 96.6 and 1.95, respec-tively, while those of the RD387 (assuming the major componentsrepresent the entire composition) are 86.5 and 1.84, respectively.However, this approximation was considered to be reasonable.

4.2. HCCI engine experiments

In order to test the present mechanism in homogeneous chargecompression ignition engine simulations, experiments by Pud-uppakkam et al. [17] were employed. Among the fuels with whichthey performed engine tests, the 4-component gasoline surrogatefuel was considered for validation of the reduced mechanisms ofthe present study. Also, 5-component model-fuel blend was de-rived from their 7-component model-fuel blend since some ofthe fuel components considered in Ref. [17] were not modeled inthe present reaction mechanism and tested for HCCI combustion.In this case, isohexane was merged into iso-octane and styrene intotoluene. Methycyclohexane was modeled using cyclohexane.

1-Pentene was modeled using a submechanism of n-tetradecaneoxidation in the present study. The details of model compositionof the fuels is listed in Table D2 [78], along with the compositionproposed by Puduppakkam et al. [17]. Details of the engine aregiven in Table D3 [78].

4.3. DI diesel engine experiment

Finally, in order to demonstrate the performance of the reducedmechanism for spray combustion applications, low temperaturecombustion (LTC) compression-ignition experimental results froma direct-injection, light-duty diesel engine were used [71]. In theexperiments, the distillation and fuel composition in classes weremeasured, as well as proles of cylinder gas pressure forcombusting cases. Detailed specications of the engine are givenin Ref. [72]. The fuel specications and operating conditions ofthe engine experiment simulated in this study are given in TableD4 [78].

5. Results and discussion

The performance of the present MultiChem mechanism wastested by comparing the predicted results with experimental data

0.01

0.1

1

10

100

0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5

igni

tion

dela

y [m

s]

EXP, Herzler et al., 10 barpresent mechanism, 10 barEXP, Herzler et al., 30 barpresent mechanism, 30 barEXP, Herzler et al., 50 barpresent mechanism, 50 bar 0.1

1

10

100

0.8 0.9 1 1.1 1.2 1.3

igni

tion

dela

y [m

s]

present mechanism, 10 barpresent mechanism, 30 barpresent mechanism, 50 barEXP, Herzler et al., 10 barEXP, Herzler et al., 30 barEXP, Herzler et al., 50 bar

mix.3.

Silke et al., phi=1.0

Y. Ra, R.D. Reitz / Combustion and Flame 158 (2011) 6990 811000/T [1/K]

(a)Fig. 11. Comparison of prediction of ignition delay times of n-heptane/toluene/airn-heptane and 65% toluene by volume. Ignition delays (a) at / = 1.0 and (b) at / = 0

0.01

0.1

1

10

100

igni

tion

del

ay [m

s]

1000/T [1/K]

present mechanism, phi=1.0Silke et al., phi=1.0experiment

0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6

(a)

Fig. 12. Comparison of prediction of ignition delay times of cyclohexane/air mixtures at c12.5 bar. (a) Comparison with experiments by Lemaire et al. [45] and Daley et al. [43], (b)by Silke et al. [5] for three equivalence ratios (0.5, 1.0, and 2.0).0.010.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6

1000/T [1/K]

Silke et al., phi=2.0

(b)0.1

1

10

100

1000

ign

itio

n d

ela

y [m

s]

present mechanism, phi=0.5present mechanism, phi=1.0present mechanism, phi=2.0Silke et al., phi=0.51000/T [1/K]

(b)tures with experimental data. Experiments by Herzler et al. [68]. Mixtures of 35%onstant volume with experimental data and a detailed mechanism. Initial pressure iscomparison between the present reaction mechanism and the detailed mechanism

are in much better agreement with those by the detailed mecha-nism [38] than the previous mechanism [11].

5.1.2. n-Tetradecane auto-ignitionIgnition delay times of n-tetradecane/air mixtures at constant

volume were simulated with the present MultiChem mechanismfor two different equivalence ratios (/ = 1.0 and 0.5) at an initialpressure of 40 bar. The experimental measurements by Shenet al. [66] were used for comparison. Figure 9a shows comparisonof predicted ignition delays using the present mechanism with theshock tube test data and predictions obtained using the detailedmechanisms of Westbrook et al. [6]. The present predicted ignitiondelays are in good agreement with experimental data for bothequivalence ratios except for one stoichiometric mixture datapoint at low temperature (870 K). However, for this condition,which falls in the NTC regime for other n-alkane fuels, the mea-

n and Flame 158 (2011) 6990and predictions using detailed mechanisms. Shock tube tests andengine experimental data available in the literature were em-ployed in the validation, as described above. For comparison withshock tube ignition delay data, the calculations were performedwith the SENKIN code [34] under constant volume conditions. Also,the predicted ignition delay times were compared with those ob-tained using other mechanisms available in the literature. The de-tailed mechanisms for tetradecane oxidation of Westbrook et al.[6], toluene oxidation of Bounaceur et al. [4] and Andrae et al.[42], cyclohexane oxidation of Silke et al. [5], ethanol oxidationof Marinov [55], DME oxidation of Fischer et al. [59] and Curranet al. [58], and MB oxidation of Fisher et al. [3] were chosen forthe comparisons.

In order to conrm the repeatability of predictions for PRF oxi-dation, a comparison of ignition delay prediction between the Mul-tiChem mechanism and the previous reduced PRF mechanism wasalso performed. Simulations of engine combustion were performedusing the multi-dimensional KIVA-3 V code [73] coupled with theCHEMKIN library [74] and with physical sub-models developedat the Engine Research Center of the University of Wisconsin.

5.1. Validation of the reduced mechanism

Since the major interest of the present study was to develop areduced multi-component chemistry mechanism to be used forauto-ignition processes of multi-component fuels, the mechanismvalidity was gauged by the ability to model auto-ignition of indi-vidual fuel components and multi-component sprays.

When no experimental data to compare and validate thereduced mechanism was available, the reduced mechanism wasvalidated with the parent detailed mechanism.

5.1.1. PRF auto-ignitionIgnition delay predictions using the present MultiChem mecha-

nism were validated with those obtained by the reduced PRFmechanism [11] for stoichiometric mixtures of n-heptane andiso-octane under various initial temperatures, as shown in Fig. 7.The simulations were performed at a pressure of 40 bar. For com-parison, shock tube experimental data of Fieweger et al. [75] is in-cluded. As can be seen, the predictions by the present MultiChemmechanism are in excellent agreement with those by the reducedPRF mechanism, which conrms that the PRF oxidation branchesare well represented in the new reduced mechanism.

Further improvement was made to the performance of heat re-lease and species predictions by including additional reactions andby optimizing reaction rate constants of selected reactions. Heatrelease rate predictions of HCCI combustion of gas mixtures highlydiluted with exhaust gas recirculation (O2 = 8.58% at intake valveclosure, /overall 0.68) were signicantly improved, as shown inFig. 8a. The PRF mechanism of Ra and Reitz [11] tends to show dou-ble-humps in heat release rate proles due to excessive termina-tion of active radicals before radical branching reactions (e.g.,H2O2(+m) = OH + OH(+m)) take place. It is clearly seen in the gurethat the double-hump behavior was eliminated by the presentmechanism and the pressure prole becomes smoother duringthe main ignition period.

In addition, the UHC and CO emissions predictions were im-proved in the present mechanism. In Fig. 8b, emissions and gastemperature predictions are compared among the LLNL detailedmechanism, the PRF mechanism by Ra and Reitz [11], and thePRF mechanism in the present study for HCCI engine operation atvarious mixture equivalence ratios. Contours of crank angle atwhich, for each equivalence ratio, gas temperatures become 1200

82 Y. Ra, R.D. Reitz / Combustioand 1500 K, and UHC and CO emissions become 20 and 88 g/kg f,respectively, are plotted. The contours predicted by the presentPRF mechanism, especially the UHC contour for rich mixtures,sured ignition delay time is much longer than correspondingvalues for other n-alkanes, especially n-heptane (see Fig. 9a, referto [11]). Moreover, the ignition delays of n-tetradecane arereasonably expected to be faster than those of n-heptane, basedon the reported CN numbers of the two fuels (CNnC7H16 = 52 vs.CNnC14H30 = 93, [76]). More experimental data of n-tetradecaneignition delay at low temperatures would be desirable for bettervalidation of reaction mechanisms. It is seen from the gure thatthe detailed mechanism tends to overpredict the measuredignition delay over the entire temperature range considered forboth equivalence ratios. The ignition delay times of stoichiometricn-tetradecane/air mixtures are compared in Fig. 9b with those ofstoichiometric n-heptane/air mixtures using the present reactionmechanism. The n-tetradecane ignition delays are predicted to beslightly faster than those of n-heptane, which is consistent withthe results reported in Westbrook et al. [6] and the CN expectation.

5.1.3. Toluene auto-ignitionThe toluene oxidation branch of the present reaction mecha-

nism was validated with various experiments and predictions bydetailed mechanisms. The shock tube experimental data by Boun-aceur et al. [4] and Davidson et al. [67] were compared with pre-dictions for toluene/O2/Ar mixtures at high temperatures andtoluene/air mixture at intermediate-to-high temperatures, respec-tively. In order to check the prediction performance for n-heptane/toluene/air mixtures at low-to-intermediate temperatures, theexperiments by Herzler et al. [68] were used. In addition, thepredictions obtained using detailed mechanisms available in theliterature were compared with those by the present reactionmechanism.

0.01

0.1

1

10

100

1000

0.7 0.8 0.9 1 1.1 1.2 1.3

igni

tion

dela

y [m

s]

1000/T [1/K]

Marinov, phi=2.0present mechanism, phi=2.0Marinov, phi=1.0present mechanism, phi=1.0Marinov, phi=0.5present mechanism, phi=0.5Fig. 13. Comparison of prediction of ignition delay times of ethanol oxidation withair at constant volume between the present reaction mechanism and a compre-hensive mechanism [55]. Initial pressure is 40 bar.

0.001

0.01

0.1

1

10

100

1000

igni

tion

dela

y [m

s]

10000/T [1/K]

present mechanism, phi=1.5EXP, phi=1.5present mechanism, phi=1.0EXP, phi=1.0present mechanism, phi=0.5EXP, phi=0.5present mechanism, phi=0.25EXP, phi=0.25

0.001

0.01

0.1

1

10

100

5.5 6 6.5 7 7.5 8 8.5 9 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5

igni

tion

dela

y [m

s]

1000/T [1/K]

Fisher et al., phi=1.0present mechanism, phi=1.0Fisher et al., phi=0.5present mechanism, phi=0.5

(a) (b)Fig. 15. Comparison of ignition delay time predictions between the present reaction mechanism and shock tube experiments by Dooley et al. [61] and predictions with adetailed mechanism by Fisher et al. [3]. (a) Comparison with shock tube data was for MB/O2/Ar mixtures at an initial pressure of 4 atm, (b) comparison with detailedmechanism for MB/air mixtures at an initial pressure of 40 bar.

0.01

0.1

1

10

igni

tion

dela

y [m

s]

1000/T [1/K]

present mechanism, 55 barexperiment, 55 barpresent mechanism, 20 barexperiment, 20 bar

0.01

0.1

1

10

0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2

igni

tion

dela

y [m

s]

1000/T [1/K]

present mechanism, 55 barexperiment, 55 barpresent mechanism, 20 barexperiment, 20 bar

(a) (b)Fig. 16. Comparison of predicted ignition delay times with shock tube test data for gasoline surrogate compositions of Gauthier et al. [70]. Stoichiometric mixtures of the fuelin air were considered. (a) Surrogate-A: 63% iso-octane + 20% toluene + 17% n-heptane, (b) Surrogate-B: 69% iso-octane + 14% toluene + 17% n-heptane.

0.01

0.1

1

10

0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6

igni

tion

dela

y [m

s]

1000/T [1/K]

present mechanism present mechanismdetail mechanism detail mechanismexperiment experiment

13 bar

40 bar

0.01

0.1

1

10

0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5

igni

tion

dela

y [m

s]

1000/T [1/K]

present mechanism, phi=0.5present mechanism, phi=1.0present mechanism, phi=2.0detail mechanism, phi=0.5detail mechanism, phi=1.0detail mechanism, phi=2.0

(a) (b)Fig. 14. Comparison of prediction of ignition delay times of DME oxidation with air at constant volume between the present reaction mechanism and a comprehensivemechanism [58,59]. Initial pressure is 40 bar.

Y. Ra, R.D. Reitz / Combustion and Flame 158 (2011) 6990 83

Ignition delay times of toluene/O2/Ar mixtures were calculatedat an initial pressure of 50 bar for two different equivalence ratios(0.5 and 1.0). In both cases, the toluene mole fraction was xed at1.25%. In Fig. 10a and b, predicted ignition delays under high initialtemperature conditions are compared with experimental data andpredictions by detailed mechanisms of Bounaceur et al. [4] andAndrae et al. [42]. The present reaction mechanism slightly over-predicts the ignition delay times compared to the experimentaldata. However, it is seen that the predicted ignition delays by thepresent reaction mechanism are in best agreement with the mea-sured data and capture the trend of ignition delays excellentlycompared to the two detailed mechanisms.

Ignition delay times of toluene/air mixtures predicted by thepresent mechanism were also compared with experimental datafor stoichiometric mixtures at 50 bar initial pressure and interme-diate temperatures, as shown in Fig. 10c. Overall, the present reac-tion mechanism predicts ignition delay times and their trends withrespect to initial temperature variation reasonably well, although ittends to overpredict ignition delays as the initial mixture temper-ature is decreased. The detailed mechanism by Andrae et al. [42]captures the ignition delay times quite well, while the mechanismby Bounaceur et al. tends to signicantly overpredict.

In Fig. 10d, the performance of the three mechanisms are com-pared for stoichiometric toluene/air mixtures at an initial pressureof 40 bar over low-to-high temperature ranges. Compared to theignition delay times of n-alkanes under similar conditions, toluene

ignition delay times are predicted to be much longer. It is notablethat no NTC behavior is seen in toluene oxidation over the entiretemperature range considered. The predictions by the presentreaction mechanism are in good agreement with those by Andraeet al. [42].

In order to validate the performance of the present reactionmechanism in predicting ignition delay times of n-heptane/tolueneblended mixtures, auto-ignition of a mixture of 35% n-heptane and65% toluene in air was simulated. In Fig. 11a stoichiometric igni-tion delay times are compared with shock tube experimental datafor various initial pressures. Ignition delay times at 10 and 30 barare reasonably well predicted by the present reaction mechanism,but the 50 bar case tends to be underpredicted. Due to the effectsof the n-heptane oxidation pathways, a clear NTC behavior is seenin all three pressure cases. The present reaction mechanism alsoperforms well for mixtures with equivalence ratio of 0.3, as shownin Fig. 11b. Overall, the experimental ignition delay times are wellcaptured by the present reaction mechanism, although the ignitiondelay times in the lowest initial pressure case tend to be underpre-dicted at high temperatures.

5.1.4. Cyclohexane auto-ignitionThe performance of the present reaction mechanism was vali-

dated with ignition delay measurements in a rapid compressionmachine by Lemaire et al. [45] and in shock tube tests by Daleyet al. [43]. Ignition delay times of cyclohexane/air mixtures were

0.1

1

10

nitio

n de

lay

[ms]

Sur-ASur-B 0.1

1

10

ition

del

ay [m

s]

present model, phi=0.5present model, phi=1.0

0.9

84 Y. Ra, R.D. Reitz / Combustion and Flame 158 (2011) 69900.01

1000/T [1/K]

ig

present modelisooctaneexperiment

0.01

0.1

1

10

0.75 0.8 0.85

0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2

igni

tion

dela

y [m

s]

(a)Fig. 17. Comparison of predicted ignition delay times with the shock tube test data forstoichiometric gasoline/air mixtures, (b) effect of equivalence ratio variation predicted usthe present 4-component model.0.01

ign

1000/T [1/K]

present model, phi=1.6experiment, phi=0.5experiment, phi=1.0experiment, phi=1.6

0.95 1 1.05 1.1 1.15 1.2

0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2

1000/T [1/K]

present model, phi=0.5present model, phi=1.0present model, phi=1.6experiment, phi=0.5experiment, phi=1.0experiment, phi=1.6

(b)

(c)

gasoline by Gauthier et al. [70]. Initial pressure 55 bar. (a) Ignition delay times ofing Surrogate-A composition, (c) effect of equivalence ratio variation predicted using

20

30

40

50

60

70

80experiment, 453 Kpresent model, 453 KPuduppakkam et al. model, 453 Kexperiment, 433 Kpresent model, 433 KPuduppakkam et al. model, 433 K

20

30

40

50

60

70

80

pres

sure

[bar

]

experiment, 453 Kpresent model, 453 KPuduppakkam et al. model, 453 Kexperiment, 433 Kpresent model, 433 KPuduppakkam et al. model, 433 K

(a)

n an20

30

40

50

60

70

80experiment, 437 Kpresent model, 437 KPuduppakkam et al. model, 437 Kexperiment, 430 Kpresent model, 430 KPuduppakkam et al. model, 430 K

20

30

40

50

60

70

80

pres

sure

[bar

]

experiment, 437 Kpresent model, 437 KPuduppakkam et al. model, 437 Kexperiment, 430 Kpresent model, 430 KPuduppakkam et al. model, 430 KY. Ra, R.D. Reitz / Combustiocalculated at an initial pressure of 12.5 bar. For comparison, igni-tion delay calculations using the detailed mechanism by Silkeet al. [5] were performed for the same conditions. In Fig. 12a, pre-dictions and measured data are compared for stoichiometric cyclo-hexane/air mixtures. Note that the RCM data, which was originallyobtained in the pressure range of 1114 atm, was scaled to12.5 bar to be combined with the shock tube data. The ignition de-lays by the present reaction mechanism are in good agreementwith the measurements except that it tends to underpredict thedelays for very low temperatures, while the detailed mechanismoverpredicts ignition delays over the low-to-intermediate temper-ature range. The NTC regime temperatures of cyclohexane oxida-tion at initial pressure of 12.5 bar are much lower than those ofn-heptane and n-tetradecane at 40 bar (compare Fig. 9). It wasfound that the range of temperature corresponding to the NTCbehavior in cyclohexane oxidation moves to higher temperaturesas the initial pressure increases (not shown).

The effect of equivalence ratio variation was tested for the sameinitial pressure. In Fig. 12b, the predicted ignition delay times atthree equivalence ratios (0.5, 1.0 and 2.0) are compared with thoseby the detailed mechanism. For all three equivalence ratio cases,the detailed mechanism tends to predict longer ignition delays, ex-cept for temperatures higher than around 1100 K. Overall, the ef-fect of the equivalence ratio variation is seen to be similar inboth mechanisms.

5.1.5. Ethanol auto-ignitionThe performance of the present MultiChem mechanism for the

auto-ignition of ethanol was validated by comparing the ignition

10-30 -20 -10 0 10 20 30

10-30 -20 -10 0 10 20 30

crank angle [deg atdc]

(b)Fig. 18. Comparison of predicted pressure proles in a gasoline HCCI engine with experimthe predictions using the present model, (b) Blend-1 for initial intake temperatures of 4d Flame 158 (2011) 6990 85delay times at constant volume for various initial temperatureand equivalence ratios with those obtained using a comprehensivemechanism [55], as shown in Fig. 13. The ignition delay times ofethanol at relatively low temperatures are substantially longer thatthose of typical aliphatic components in gasoline and diesel fuels,as can be expected from its typical CN number (10).

Over the entire temperature range considered, it is seen that thepredictions by the present reaction mechanism are in good agree-

10-30 -20 -10 0 10 20 30

10-30 -20 -10 0 10 20 30

crank angle [deg atdc]

(c)ents and predictions by Puduppakkam et al. [17]. (a) 2-D computational grid used in37 and 430 K, (c) Blend-2 for intake temperatures of 453 and 433 K.

Fig. 19. Computational grid used in the engine simulations. 1/7th sector grid with10,470 cells for DI diesel combustion simulation.

ment with those by the comprehensive mechanism, although thereis slight discrepancy for lean mixtures at high initial temperatures.

5.1.6. DME auto-ignitionThe performance of the MultiChem mechanism for the auto-

ignition of DME was validated with the shock tube tests of Pfahlet al. [69]. In the experiments, ignition delay times of stoichiome-tric DME/air mixtures were measured at two different initial pres-sures (13 and 40 bar). Simulations of the ignition process atconstant volume were performed for the two initial pressure con-ditions with the present reaction mechanism. For the comparisonbetween the reduced and the detailed mechanism [58,59], ignitiondelay times at 40 bar initial pressure were calculated for variousequivalence ratios.

Figure 14a shows comparisons of ignition delay with respect toinitial temperature between measurements and predictions. For

In Fig. 15, ignition delay times of MB/O2/Ar mixtures at an initialpressure of 4 atm are compared for various equivalence ratios be-tween predictions by the present mechanism and experimentaldata. For stoichiometric and rich mixtures, the present mechanismpredicts the experimental ignition delay times reasonably well,while it tends to underpredict as the mixtures become leaner,especially at lower temperatures. However, when the performanceof the present mechanism is compared with that of the compre-hensive mechanism, the predictions by both mechanisms are ingood agreement over the entire temperature range considered, asshown in Fig. 15b. Note that the MB oxidation mechanism wasoptimized to match the predictions by the comprehensive mecha-nism when it rst was developed by Brakora et al. [10]. Therefore,further improvement of MB mechanism to match the experimentalmeasurements would be desirable.

ion

BRBRBRBRBRBRBR

86 Y. Ra, R.D. Reitz / Combustion and Flame 158 (2011) 699040 bar initial pressure, ignition delay times predicted by the pres-ent reaction mechanism are in good agreement with the experi-mental data, while the detailed mechanism slightly overpredictsignition delays in the low temperature range. On the contrary,the present reaction mechanism tends to overpredict ignition de-lay times at low temperatures, while it tends to underpredict atintermediate-to-high temperatures. It would be necessary to fur-ther optimize the pressure dependency of the reduced mechanismto improve the performance of the reduced mechanism.

The performance of the reduced mechanism in terms of equiv-alence ratio variation is compared with that of the detailed mech-anism in Fig. 14b. Three equivalence ratios were considered. As theequivalence ratio increases, the ignition delay times shorten inboth mechanisms. However, the variation of ignition delay timeswith equivalence ratio in the detailed mechanism case is smallerthan that of the present reaction mechanism at low temperaturesup to 850 K, while it becomes greater for temperatures higher than900 K. It is interesting that the ignition delay behavior predicted bythe reduced mechanism for equivalence ratio 2.0 mixtures showsthe NTC behavior clearly, while the detailed mechanism showsonly a gradual decrease of ignition delay times with increasinginitial mixture temperature.

5.1.7. MB auto-ignitionThe performance of the present reaction mechanism in ignition

delay predictions of MB was validated with shock tube experimen-tal data by Dooley et al. [61] and ignition delay times calculatedusing the detailed mechanism by Fisher et al. [3], which was thebase mechanism for the MB oxidation branch in the present reac-tion mechanism. For detailed composition information of the mix-tures, see Ref. [61].

Table 4Reaction rate constants of modied reactions for DI spray combust

Reaction no. Reaction

14 nC7H16 + HO2 = C7H15 2 + H2O2

17 C7H15O2 + O2 = C7KET12 + OH

18 C7KET12 = C5H11CO + CH2O + OH

20 C7H15-2 = C2H5 + C2H4 + C3H6

23 C14H30 + HO2 = C14H29 + H2O2

25 C14H29 + O2 = C14H29O2

26 C14H29O2 + O2 = C14KET12 + OH27 C14KET12 = C12H25CO + CH2O + OH BR5.1.8. Gasoline auto-ignitionWith the individual surrogate component mechanisms having

been validated against ignition delay measurements and predic-tions of detailed mechanisms, the present reaction mechanismwas further validated with experimental ignition delay measure-ments with known compositions of multi-component fuels. Igni-tion delay times at constant volume condition were calculatedfor the two iso-octane/toluene/n-heptane compositions proposedby Gauthier et al. [70] as a gasoline surrogate. In Fig. 16 the pre-dicted ignition delay times at two different initial pressures arecompared with the experimental data of Gauthier et al. The igni-tion delay times predicted by the present reaction mechanismare in excellent agreement with the measured values for both sur-rogate compositions.

In order tomodel the auto-ignition ofmulti-component gasoline,predictions by thepresent reactionmechanismwere also performedusing three other surrogate composition models. A new 4-compo-nent composition including cyclohexane to account for naphtheneson gasoline auto-ignition behavior was tested, along with the twosurrogate compositions suggested byGauthier et al. [70], as summa-rized in Table D1 [78].

The performance of the three composition models are comparedwith measured data in Fig. 17a. Ignition delay times of 100% iso-oc-tane are plotted for comparison, as well. All calculations were donefor stoichiometricmixtures at 55 bar initial pressure. Overall, Surro-gate-A gives best agreement with measurements among the surro-gate models. As was expected, 100% iso-octane showed the longestignition delay times, especially at lower temperatures. The present4-componentmodel tends to slightly underpredict, which indicatesthat adjustment of the composition of the fuel mixture may be nec-essary for better agreement. However, the present 4-component

simulations (both base and revised values are shown).

A n Ea (cal/mole)

ase 2.44E + 14 0.0 16950.0evised 4.88E + 14 0.0 16950.0ase 1.54E + 14 0.0 18232.7evised 2.30E + 14 0.0 18232.7ase 9.01E + 14 0.0 41100.0evised 1.80E + 15 0.0 41100.0ase 2.54E + 15 0.0 34600.0evised 3.80E + 15 0.0 34600.0ase 1.01E + 14 0.0 17690.0evised 1.51E + 14 0.0 17690.0ase 3.77E + 12 0.0 0.0evised 2.64E + 12 0.0 0.0ase 1.40E + 14 0.0 19232.7evised 2.10E + 14 0.0 19232.7

ase 6.00E + 16 0.0 39000.0evised 8.00E + 16 0.0 39000.0

model adequately captures the effects of equivalence ratio variation,as shown in Fig. 17c. While the present 4-component model stilltends to slightly underpredict the ignition delay times comparedto thoseof the Surrogate-Amodel (Fig. 17b), the trendof equivalenceratio variation is better captured by the present 4-component mod-el, especially at higher temperatures. Therefore, the 4-componentmodel can reasonably be expected to provide information aboutthe effects of naphthenes on the auto-ignition of gasoline fuels bet-ter than the three-component Surrogate-A and -B models.

5.2. Validation of MultiChem mechanism with engine experiments

5.2.1. Gasoline HCCI combustionThe present reduced MultiChem mechanism was also tested

against homogeneous charge compression ignition engine experi-ments by Puduppakkam et al. [17]. Two fuels, Blends 1 and 2 aslisted in Table D2 [78] were considered with two different initial

performed using the KIVA code [73] coupled with the CHEMKIN

[72] was simulated using a CFD computational grid with 10,470cells in a 1/7th sector of the cylinder, as shown in Fig. 19. The en-gine was operated in the low temperature combustion regime withintake oxygen fraction of 9.5%. Table D4 [78] provides the details of

ep

0.01

0.1

1

10

0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.51000/T [1/K]

igni

tion

dela

y [m

s]

toluenediesel

(b)

Table 5Model composition of surrogate diesel fuels for DI diesel engine tests.

8-Component spray model Component in chemistry model

Component Mass fraction (%)

Cyclohexane 3.7 CyclohexaneDecalin 6.4Toluene 5.0 Toluenen-Decane 5.6 n-Heptanen-Dodecane 20.9n-Tetradecane 25.9 n-Tetradecanen-Hexadecane 16.6n-Octadecane 15.8

Y. Ra, R.D. Reitz / Combustion and Flame 158 (2011) 6990 87400

450

500

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1evaporated fract