Development of an Experimental Database and Kinetic Models for Surrogate Diesel Fuels

5/25/2018 Development of an Experimental Database and Kinetic Models for Surrogate Diesel Fuels

1/27

1

2007-01-02

Development of an Experimental Database and Kinetic Modefor Surrogate Diesel Fue

J.T. Farrell,*N.P. Cernansky,F.L. Dryer,D.G. Friend,C.A. Hergart,**C. K. Law

R.M. McDavid,**C.J. Mueller,A.K. Patel,**and H. Pitsc

Copyright 2007 SAE International. (Contribution, in part, of the National Institute of Standards and Technology. Not subject to copyright in the Unit

States).

ABSTRACT***

Computational fluid dynamic (CFD) simulations thatinclude realistic combustion/emissions chemistry hold

the promise of significantly shortening the developmenttime for advanced high-efficiency, low-emission engines.However, significant challenges must be overcome torealize this potential. This paper discusses thesechallenges in the context of diesel combustion andoutlines a technical program based on the use ofsurrogate fuels that sufficiently emulate the chemicalcomplexity inherent in conventional diesel fuel. Theessential components of such a program are discussedand include: (a) surrogate component selection; (b) theacquisition or estimation of requisite elementarychemical kinetic, thermochemical, and physical propertydata; (c) the development of accurate predictive

chemical kinetic models, together with the measurementof the necessary fundamental laboratory data to validatethese mechanisms; and (d) mechanism reduction toolsto render the coupled chemistry/flow calculationsfeasible. In parallel to these efforts, the need exists todevelop similarly robust models for fuel injection andspray processes involving multicomponent mixtures ofwide distillation character, as well as methodologies toinclude all of these high fidelity submodels incomputationally efficient CFD tools. Near- and longer-term research plans are proposed based on anapplication target of premixed diesel combustion. In thenear term, the recommended surrogate components

include n-decane, iso-octane, methylcyclohexane, andtoluene. For the longer term, n-hexadecane,heptamethylnonane, n-decylbenzene, and1-methylnaphthalene are proposed.

*ExxonMobil Research and Engineering

Drexel University

Princeton University

National Institute of Standards and Technology

**Caterpillar Inc.

Sandia National Laboratories

Stanford University

INTRODUCTION

Significant improvements in diesel engine efficiency emissions are required to meet future legislated targ

Conventional engine design approaches that rely prototype development are too time-consuming expensive to meet these challenges. The developmof predictive and efficient computational tools incorporate the relevant chemical kinetics, molectransport, and turbulent flow would representsignificant step forward in the ability to rapidly deshigh efficiency, low emission engines. Howesignificant gaps and uncertainties exist in our knowleof these fundamental processes and their coupling.

In recognition of this need, the National InstituteStandards and Technology (NIST) held a workshop

Combustion Simulation Databases for RTransportation Fuels in September 2003 [1]. The ovegoal was to develop the experimental databachemical kinetics, and computational methodolnecessary to simulate realistic combustion processwith fuels representative of actual commercial gasoldiesel, and aviation fuels. The present paper isoutgrowth of the original meeting and additioendeavors of a subsequently formed adhoc SurrogDiesel Fuel Working Group over the past three years.

Because of the complexity of diesel fuel composition reaction kinetics, studies focused on developing

molecular-level understanding of diesel combusoften employ surrogate fuels. Surrogate fuels simpler representations of fully blended fuels which comprised of selected species of known concentratiand that exhibit combustion characteristics similarthose of the real fuel. Experiments with surrogate fuare particularly useful when it is desirable to limit chemical and/or physical complexity of the fuelgenerate insight and understanding of underlyprocesses such as vaporization, mixing, ignition, pollutant formation.


2/27

2

Most surrogate studies, particularly those involving morethan one component and molecular weights that arecharacteristic of diesel fuel components, have beenlimited by the lack of well developed, validated, detailedchemical kinetic models of the multicomponentsurrogate mixture, and by the paucity of experimentalvalidation data needed to create accurate models. Ofequal significance, there presently exist few

dimensionally reduced versions of these kinetic modelsthat are compatible for use in multi-dimensional,computational fluid dynamic (CFD) engine design codes.

Formulating surrogates that match conventional dieselfuel combustion parameters requires matching both thechemical and physical properties of the fuel, and willlikely require more components than can be handled incurrent computational codes. Consequently, there arepresently practical reasons to limit the number of purecomponents used to represent full blend diesel chemicalkinetics. While the long-term goal of increasing thenumber of species considered should remain, near-termresearch should reflect a long-term desire tosimultaneously describe both physical and chemicalkinetic parameters of diesel fuel.

The goals of the present paper are to assess the currentstate of diesel surrogate fuel research and to outline aresearch program for the development of theexperimental database and modeling tools necessary toadequately emulate the chemical complexity of realdiesel fuel. The essential components of this researchprogram are: (a) the identification of a suitable smallnumber of fuel components representative of actual fuelcomposition that can be blended to mimic properties ofcommercial fuels, yet be sufficiently simple to permit

detailed kinetic modeling; (b) acquisition or estimation ofelementary chemical kinetic, thermochemical, andphysical property data needed for the surrogatemolecules and mixtures thereof; (c) the development ofaccurate, predictive chemical kinetic models, togetherwith the measurement of the necessary fundamentallaboratory data to validate these mechanisms; and (d)mechanism reduction tools to render feasible thecoupled chemistry/flow calculations. In parallel to effortsto develop more robust physical and chemical kineticmodels, there are needs to develop similarly robustmodels for fuel injection and spray processes involvingmulticomponent mixtures of wide distillation character, a

rigorous description of turbulence-chemistry interactionsin combustion, as well as methodologies to include all ofthese high-fidelity submodels in computationally efficientCFD tools. Ultimately, data acquired from wellcharacterized engines will be required for final validation.

The paper is organized as follows. First, the physicaland chemical properties of commercial diesel fuel aredescribed in order to elucidate the necessary propertytargets for realistic surrogates. Second, the key targetsdesired for matching a surrogate fuel to a real fuel arediscussed. Next, published work on surrogate diesel

components and fuels is reviewed to determine extent to which existing data on surrogates can be uas a foundation for this activity. We then proposeresearch plan for developing the necessary experimedatabase and kinetic models for diesel surrogate fuWe will distinguish the near-term (1-7 years) and longterm (7+ years) time frames.

OVERVIEW OF DIESEL FUEL COMPOSITION

Although a surrogate diesel fuel does not necessaneed to contain components that are representativethe molecules contained in diesel fuel to match the fuel behavior, it is reasonable to expect thatcompositional match may permit better agreement. Twe provide a brief overview of the compositioncommercial diesel fuel.

Commercial diesel fuels are complex blends of sevhundreds of individual species [2-4]. The primchemical and physical property ranges of No

American (U.S. and Canada) diesel fuels are givenTable 1 [5,6]. There is tremendous variability in diefuel properties not only internationally, but regionally locally as well, which reflects the nature of the crude the processes used at the refinery from which the originates.

Table 1. Chemical and physical properties of typicaNorth American diesel fuel [5,6]

Property Value

40-56

C10-C24

190-360

% normal, iso-paraffins 25-50

% cyclo-paraffins 20-40

% aromatics 15-40

Cetane Number

Carbon Number Range

Boiling Range (C)

Composition:

The diesel-range molecules produced from distillatiocrude oil (the first processing step) include paraffcyclo-paraffins, and aromatic molecules (see FigureThe paraffins include straight-chain (normal) branched (iso) paraffins. The iso-paraffins in diesel are very slightly branched, typically containing only or two methyl substitutions on a long (C10-C24) ch2-methyl analogues are the most common, followed

3-methyl, 4-methyl, etc. The cyclo-paraffins are prima1-ring cyclohexanes with multiple alkyl sidechains; 2-and larger cyclo-paraffins are usually present at levless than 5 % of the total fuel. Aromatics, whconstitute 20-40 % (30-35 % average) of commercialdiesel fuels [6], are primarily 1-ring analogues, i.e., albenzenes (15 %), with 5 % substituted 2-ring aroma(naphthalenes). Naphtho-aromatics and larger 3-cyclo-paraffins and aromatics can also be presendiesel, but the relative concentrations are small likely to decrease in the future.


3/27

3

Figure 1. Representative examples of the majorclasses of molecules contained in diesel fuel.Commercial fuels contain a very large number ofisomers of each class, and the specific C16 isomersshown are representative only. Adapted from Sato etal. [7].

Figure 2 shows a breakdown of molecular type vs.carbon number based on detailed analyticalcharacterization of three commercially available dieselfuels. The large compositional differences between

various market fuels is evident. The figure shows thatFuel B contains a significantly higher level ofcyclo-paraffins than Fuels A and C, and that Fuel Ccontains significantly more aromatics. These differencesunderscore an important point, i.e., that a surrogate fuelthat matches the combustion properties of a specific realdiesel fuel may not be broadly applicable to other marketfuels.

Diesel fuels from nonconventional resources areprojected to augment the petroleum-derived fuelsincreasingly in the future [8]. Biodiesel, i.e., fatty acidesters produced via transesterification with methanol or

ethanol of either vegetable oil or animal fat has beenused in Europe for several years, and is currently findingincreased use in the United States [9]. As biodiesel willbe available in relatively small quantities, it will beutilized primarily as a blending stock at levels up to 20 %.The molecular structure of biodiesel is much differentfrom that of petroleum-derived diesel fuel, with a morehomogeneous distribution. For example, soy-derivedbiodiesel typically consists of five methyl esters, withmethyl linoleate being the dominant component (~67 %)[10].

12

8

4

0

WeightPercent

30252015105

Carbon Number

n-paraffins

iso-paraffins

1-ring cycloparaffins


3+-ring cycloparaffins

1+-ring aromatics

2+-ring aromatics

3+-ring aromaticsC

16

12

8

4

0

WeightPercent

30252015105

n-paraffins

iso-paraffins




1+-ring aromaticsB

10

8

6

4

20

WeightPercent

30252015105

n-paraffins

iso-paraffins




1+-ring aromatics

2+-ring aromaticsA

Figure 2. Detailed hydrocarbon analysis of threcommercial diesel fuels as a function of molecutype and carbon number. Significant compositionvariability is evident.

Diesel fuels will also increasingly be derived fresources such as natural gas, oil sands, shale, cand petroleum coke. Commercial fuels produced fblending these materials will be required to mperformance, physical properties, and emissiregulations. The molecular structure distributions ove

and within the distillation curve associated with manthese materials are also different from thcharacteristic of conventional petroleum-derived diefuel. For example, fuels derived from tar sands mhave higher cyclo-paraffin content than conventiodiesel [11], and fuels produced from coal via presdirect coal liquefaction processes are predominacyclo-paraffinic [12]. Additionally, diesel-range fuproduced from natural gas, biomass, coal, petroleum coke via the Fischer-Tropsch (FT) procare comprised almost entirely of normal and paraffins. Like the iso-paraffins naturally present


4/27

4

petroleum-derived diesel, the iso-paraffins arepredominantly mono- and di-methyl paraffins, with themethyl substituent located on a carbon near the end ofthe molecule.

Since the vast majority of the commercial diesel fuel inthe near future will be petroleum-derived and will be thebaseline against which blends and/or complete

petroleum-derived fuel substitutions are evaluated, thispaper is constrained to discuss the petroleum derivedproblem in depth. However, it is expected that theprogram developed around petroleum-derived dieselfuels will be readily extendable to non-petroleum-derivedfuels.

TARGETS FOR SURROGATE DIESEL FUELS

Surrogate fuels are used to mimic the behavior of realdiesel fuel in various combustion devices, and thedefinition and complexity of the surrogate fuelformulation depends on the intended application. The

quantities used to compare the performance of asurrogate fuel to real diesel are often termed targets.We will distinguish here amongst three different types oftargets: property targets, development targets, andapplication targets.

The termproperty targetsrefers to fundamental physicaland chemical fuel properties. Common property targetssuch as density and hydrogen/carbon (H/C) ratio can bereadily matched with a single-component surrogate,though doing so obviously does not guarantee a goodmatch to diesel fuel combustion behavior. Otherproperties of the real fuel such as gross chemicalcomposition, phase behavior (e.g., vapor-liquidequilibrium and distillation characteristics), andmolecular transport properties may require many morecomponents to match well. Because the physical andchemical complexity of a surrogate fuel will be reducedcompared to that of a commercial diesel fuel, it will ingeneral not be possible (nor necessarily desirable) tomatch a wide range of properties such as viscosity,chemical composition (e.g., percent aromatics), surfacetension, etc. with a single surrogate formulation.

Development targets refers to kinetic and fluid dynamicprocesses that are important for validating surrogatemixture behavior, and that are typically evaluated in

devices with better controlled conditions than those inreal engines. Exemplar development targets include:kinetically-related phenomena (autoignition delay, burnrate, species evolution histories, emissions, etc.), withand without molecular-level transport coupling,measured in fundamental laboratory experimentaldevices; multi-component spray vaporization, dropletsize distribution, and liquid penetration lengthexperiments and model validations in combustionbombs; and elementary kinetic studies to define specificreaction pathways. As with property targets, it is likely

that several few-component surrogates will be requto match a broad range of development targets.

Application targetsrefers to results obtained from engexperiments. Exemplar application targets inclengine operating characteristics such as combusphasing and duration (timing and duration of cool flaand main heat release), combustion efficiency,

primary emissions (NOx, soot, CO, and unburhydrocarbons). In addition to steady-state behavdesirable application targets may include combusproperties during transient operation.

In many instances, developing surrogate fuels match application targets will be the ultimate reseagoal. However, it is important to recognize application targets for conventional diesel combuscannotbe used to assess the quality of a surrogate unless all relevant physical andchemical properties matched. This is because many processes, e.g., ignitsoot formation, etc., depend both on mixing kinetically controlled processes. The definition surrogates that match both physical and chemproperties of real fuel will likely require a larger numof components than necessary to match developmtargets, and thus constitutes a longer-term (7+ yeagoal.

In the interim, it is desirable to identify application targfor which the coupling of kinetic- and mixing-controprocesses is much reduced compared to thatconventional diesel combustion. In the present pawe have chosen premixed diesel combustion as appropriate relevant application target. Preliminexperimental results with premixed charge compress

ignition (PCCI) operation suggest that to first order, decoupling of kinetic and physical processes is a goapproximation.

GENERAL CONSIDERATIONS AND PREVIO

RESULTS FOR SURROGATE DIESEL FUELS

Depending on the application target chosen, a sincomponent surrogate may suffice. For examn-heptane is a gasoline-range molecule that has butilized heavily as a single component surrogate diesel fuel ignition. This frequent use reflects a cetnumber (CN) for n-heptane (~55) that is comparablethat for current European and Japanese diesel f

Additionally, detailed chemical-kinetic mechanisms low-, intermediate-, and high-temperature n-heptoxidation are available, e.g.,[13,14], and several modexist that have sufficiently reduced dimensiona(number of species and reactions) to enable their usCFD simulations.

Even in instances when n-heptane and a real diesel have comparable cetane numbers, the ignition behaof the two fuels likely will not match over a w


5/27

5

speed/load range. This is particularly true if the ignitionis dominated by mixing-controlled processes, since thesignificantly higher volatility of n-heptane will lead tolarge differences in liquid spray penetration andvaporization, and consequently, in local fuel/air ratios.Because of this, CFD simulations that incorporate n-heptane kinetics to predict real diesel combustionbehavior typically utilize physical properties for the real

diesel to provide realistic injection, break-up, collision,and evaporation behavior.

However, even under conditions where the fuel/airdistribution matches, the ignition behavior of n-heptaneand real diesel may differ. This is because the first stage(low temperature) heat release of n-heptane and realdiesel will likely exhibit a different dependence ontemperature and pressure, which will vary as the speedand load change. This difference reflects the fact thatthe oxidation and pyrolysis kinetics of n-heptane do notdescribe those of real diesel fuel. More specifically, thepotentially strong effects on ignition of aromatics, cyclo-paraffins, and iso-paraffins are not described well byn-heptane kinetics. This greatly reduced chemicalcomplexity also impacts the ability of n-heptane toreproduce the pollutant formation processes of realdiesel fuel. Moreover, the C/H ratio of the fuel departssignificantly from that of diesel, and thus local mixingphenomena cannot be represented accurately in termsof stoichiometries in diesel fuel applications. As aconsequence, while engine experiments with n-heptaneare capable of providing much insight, they may notaccurately reflect the combustion behavior of real dieselfuel. Similar limitations are expected to hold for anysingle component surrogate.

In recognition of this limitation, binary, ternary, andlarger blends have also been extensively employed assurrogates. Table 2 lists components that either havebeen extensively studied as surrogate fuel components,or are suitable for inclusion in future studies. Included inthe table are qualitative assessments (denoted by thelabels A, B, C, D and F, which are defined in the legendat the bottom) of how well (in relative terms, comparedto n-heptane) the kinetic and property information isknown for each component. Under the Understandingof Mechanism heading, separate assessments areprovided for the relative level of understanding of thelow- and intermediate-temperature vs.high temperature

kinetics. Similarly, qualitative evaluations are providedfor the thermophysical vs.transport properties under theProperty Information heading. It is important to notethat an A rating does not indicate that the detailedmechanisms that exist for these components have beenfully validated over all relevant conditions or areprecisely defined at the elementary kinetic level. Selectreferences for both mechanism development andexperimental validation data are also provided in thetable.

It is worth noting that many of the species includedTable 2 fall in the boiling range of gasoline or jet fuel.will be discussed in a later section, the inclusion of thspecies reflects the limited availability of validation dand kinetic models for diesel-range components, whis evident from the small number of entries in References columns. Of those studies that have udiesel-range molecules, several have been carried

with binary blends of the cetane reference sccomponents, i.e., n-hexadecane (or n-cetane, CN=1and either 1-methylnaphthalene (CN=0) or 2,2,4,4,6,heptamethylnonane (HMN, CN=15; also referred toiso-cetane or iso-hexadecane).

Similarly, a significant amount of work with binsurrogate fuels has also been carried out for jet applications (see [15] and references therein). Manythe components utilized fall in the front end of the dieboiling range. Of particular note, a coordinaindustry/academia European Union effort caComputational Fluid Dynamics for Combust(CFD4C) was initiated in 2000 with the goal developing surrogates for jet fuel combustion properand emissions [16]. Several binary surrogate fuinvolving n-decane and an aromatic species (toluen-propylbenzene, 1,2,4-trimethylbenzene) wevaluated, and a mixture of 80 vol % n-decane +vol % n-propylbenzene was defined as the fuel with best match to the application and design targets.

A diesel-boiling-range binary mixture of particrelevance is comprised of 70 vol % n-decane + 30 vo1-methylnaphthalene. This fuel was formulated as of the Integrated Diesel European Action (IDprogram [17], which was a concerted effort undertain the 1990s to specify a surrogate diesel fuel compriof C10and heavier compounds to facilitate comparisbetween experimental application targets and numersimulations. The IDEA fuel has a CN (55), nordensity (817 kg/m

3), and hydrogen-to-carbon ratio (

similar to those of European #2 diesel fuel [18,19].

Experiments in an optical direct injection (DI) dieengine showed that the ignition delay and heat-relerate of the IDEA fuel and the chosen reference diefuel closely matched. The IDEA fuel also showemissions trends similar to those of the real diesel fueengine output torque was varied, although NOxand semissions for the IDEA fuel were typically 3-7 % 10-30 % lower, respectively, [18,19]. The reasonaagreement in emissions levels is interesting given the liquid-phase penetration was considerably shorterthe IDEA fuel than for diesel, as determined fr


6/27

6

Fuels Low &

Intermediate

Temperatures

High

Temperatures

Thermo-

physical

Transport Mechanism Experimental

Straight-chain Alkanes

n-heptane A A A A- [81,91,112,180-

182,190]

[60,89,97,147,16

164,165,170,171

177,183-189]n-decane B A- A A [54-56,84-

86,155,156]

[55,56,156,54]

n-dodecane B B A A [63,157,190] [60,63,158,159]

n-hexadecane (n-cetane) C C B+ B [62,87,160] [62]

Branched-chain Alkanes

2,2,4-trimethyl-pentane (iso-octane) A- A B+ B [79,91,96] [89,91,96,97,147

161-168,170,171,1

2,2,4,4,6,8,8-heptamethylnonane (iso-cetane) C C B- C+ [63] [60,63,158,159]

Cycloalkanes

methylcyclohexane C C B+ B [92,169] [60,63,64,92,

93,158,159,170

ethyl/propyl/butyl cyclohexane D D B C [64]

decahydronaphthalene (decalin) D D B B- [63]

Single Ring Aromatics

toluene C C A B+ [94,96,98,172-176] [60,69,93,94,96-99,

172,177]

ethyl/propyl/butylbenzene C C B B [71] [69-71,100,178]

n-decylbenzene D D D D

Multi-ring Aromatics

tetralin D C B+ B-

1-methylnaphthalene C C B C [129,179] [60,63,158,159]

A B C D F

Detailedmechanism(s)

validated over wide

range

Mechanism(s)reported, but

with modest

discrepancies

or limitations

Mechanism(s)reported, but

with major

discrepancies

or limitations

No mechanismreported

Equation of state

available (density

to 0.3%)

Sufficient data

for model

(density to 3%)

Limited data

only

Extremely

limited/no

experimental data,

predictive model

feasible

No data or predict

model available

Correlations

available for

viscosity and

thermal

conductivity (5%)

Data available

for models (5-

10%)

Limited

viscosity and/or

thermal

conductivity

data

Extremely

limited/no

experimental data,

predictive model

feasible

No data or predict

model available

Table 2: Fuel Surrogate Components EvaluationUnderstanding of Mechanism Property Information Selected References

Thermo-physical Properties (Information on species as component

of mixture may be highly uncertain)

Transport Properties (Information on species as component of

mixture may be highly uncertain

Legend

Understanding of Mechanism


7/27

7

measurements conducted using a high-pressure, high-temperature spray chamber [20]. By comparison,n-decane, which was also tested in the engine,produced only ~ 10 % of the soot emissions of realdiesel. Development of the chemical-kinetic mechanismfor the IDEA fuel was discussed, though the finalmechanism has not been published in the open literature.Validation of the mechanism against development

targets has been limited by the lack of requisiteexperimental validation data. The efforts in Europe withthe IDEA fuel have been the most focused in definingand evaluating a surrogate fuel for the validation ofnumerical simulations of diesel combustion. Numerousother experimental investigations have employedsurrogate diesel fuels to match application targets forconventional diesel combustion [21-25]. However, noneof these studies compared surrogate fuel results withthose for real diesel using experimental or computationalapproaches that could isolate kinetic and/or physicaleffects. Even in the few studies where measured(surrogate and real diesel) and computed (surrogateonly) results have been compared, the selectionmethodology for the surrogate constituent compoundsand the measures taken to validate the chemical kineticmodels are not discussed, and the range of operatingconditions used is small. Thus, it is difficult to discern theextent to which the level of agreement reported reflectsa good match between the chemical kinetics of thesurrogate and real diesel fuel, or whether thecomparisons include compensating differences inchemical and physical property differences.

PHYSICAL PROPERTIES

Previous Diesel Spray Surrog ate Research

Many experimental studies on surrogate diesel fuelshave been conducted with constant-volume combustionvessels. Most of this work has been carried out toidentify a surrogate to simulate only the vaporization of#2 diesel fuel, not its ignition, combustion, or emissionscharacteristics. However, since fuel vaporization will beimportant to model in-cylinder fuel/air distribution indirect-injection premixed diesel operation, we brieflyreview the status of the development of diesel spraysurrogates.

Siebers has conducted a large parametric study using a

high-pressure common-rail fuel injector to determinehow the liquid length varies with ambient-gas conditions,injection parameters, and fuel volatility for n-hexadecane,HMN, and #2 diesel [26]. Measurements of liquid- andvapor-penetration rates and liquid-spray spreading anglehave also been reported for n-dodecane, n-heptane, andthe IDEA fuel [27]. The facility used in Reference 26 waslater used to compare liquid lengths of n-hexadecaneand HMN to those for #2 diesel, Fischer-Tropsch diesel,biodiesel, and several gasolines for ambient-gastemperatures from 700 to 1300 K and densities from 3.6to 59 kg/m

3at an injection pressure of 140 MPa [28]. A

follow-up study presented a model for predicting liquid length if the physical properties of the surrogcompound and certain fuel-injection parameters known [29]. The study provides criteria for a gvaporization surrogate, namely, that the liquid lendensity, and T90 (temperature at atmospheric pressfor 90 % distillation recovery) of the real and surrogfuels match. Based on these criteria, n-heptadec

was recommended as a good vaporization surrogate#2 diesel fuel [29]. It is clear, however, that this choiconly based on physical properties, and thus inappropriate for matching chemical kinetic targets.

Measurements of 1-methylnaphthalene fuel vaconcentration, droplet density, and mixture temperathave also been reported [30]. Results presentedReference 31 provide vapor- and liquid-fuel penetrameasurements over a range of ambient-gas injection conditions as well as comparisons to the resof Siebers [29]. Additionally, experimemeasurements of axial and radial penetration rates wall heat transfer have been presented for the IDEA and Swedish Class 1 diesel for single and split injecti[32]. However, the matched-density criterion mentioabove precludes 1-methylnaphthalene from beingsuitable vaporization surrogate for #2 diesel.

Thermophysica l Property Model ing and Exper ime

The development of computational models for processuch as fuel injection, atomization, vaporization, erequires detailed fundamental physics models as welaccurate physical properties for the fuel. In additionthe distillation curve, it is important to predict propersuch as the liquid density, liquid and vapor h

capacities, surface tension, and viscosity over a larange of temperatures and with pressures to more t5 MPa. Low-temperature properties may be imporfor cold-start models. A thermodynamically consisset of properties over the range of conditiencounteredpossibly including those of the pcomponents and of mixtures with intermediates airwould allow the most flexible possibilities for sysmodeling. However, representing such thermophysproperties of a complex mixture such as diesel fuel wa few-component surrogate is a major challenge, the variety of fuels encountered in the current and futmarketplace will require additional efforts to translate

knowledge learned from surrogates to an understandof more realistic engine situations.

More generally, CFD and computational combusmodeling require consistent property informationmaintain fidelity between the model and the systembe simulated. As noted earlier, the inclusion of baccurate property models and extensive kininformation within a combustion code is cleprohibitive with todays computational resources: number of coupled kinetic equations itself is unwieand when combined with accurate spatial- and ti


8/27

8

dependent density, heat capacity, viscosity, etc., fields,detailed numerical simulations are not feasible at thistime. In addition, studies have shown that ideal-gasapproximations or incompressible fluid approximationswithin CFD codes can lead to problems [33].

The first step in studying the properties of a surrogatemixture is an analysis of data available for each potential

constituent. With a potential slate of candidates, thereare a variety of sources of data that can be consulted.For instance, the AIChE DIPPR database providesrecommended data for many important compounds, andhas predictive capabilities [34]; the NIST ChemistryWebBook [35] provides many types of data on-line; theNIST REFPROP database [36] incorporates consistentsets of equations for more common fluids and theirmixtures; and the NIST ThermoDataEngine [37] canprovide evaluated properties based on a huge andgrowing collection of experimental data and a library ofpredictive models.

For the present paper, we have not compiled a completelist of property references. There are excellent standardreference correlations for the lower alkanes, and wenote that considerable information is available forn-decane: a validated equation of state from Lemmonand Span [38] gives saturated liquid densities with anuncertainty of 0.05 %, with compressed liquid densitiessomewhat higher and all thermodynamic propertiescalculable through a consistent Helmholtz energyformulation. A standard correlation for n-decaneviscosity was described by Huber et al. [39], and for n-decane thermal conductivity, an equation by Huber andPerkins [40] can be recommended. Thus n-decane canserve as a good starting point for alkane property

models, although the amount and reliability of theavailable data decreases as the carbon numberincreases. The thermophysical property models fornormal dodecane have also been well established, andextrapolations beyond this carbon numbercertainly ton-cetanewould be reasonable. Typically, theinformation on the transport properties for all of thehigher alkanes is less reliable than that for theequilibrium properties.

For the iso-paraffins, the sheer number of isomers asthe carbon number increases limits the possibility anddesirability of extensive experimental study for all

candidates. Although there is certainly propertyinformation available for some of the important iso-octane molecules, the largest molecule in this familyconsidered among the standard reference propertysurfaces within the NIST REFPROP database [36] isiso-pentane (2-methylbutane). The uncertainty in thedensity of iso-octane might be on the order of 3 % in therange of temperatures and pressures of interest. For thelarger species with only a few short side chains, such asthose typically encountered in real diesel fuels, modelsthat basically extrapolate from the normal alkanebackbone often give reasonable property results. When

the side chains are larger, or with multiple longer chathe ability to predict properties deteriorates, and baexperimental property studies on the fluid would warranted.

There are considerably fewer data and the models less reliable for both the cyclo-paraffins and the aromspecies. Evaluated models for cyclohexane have b

considered in the REFRPOP database. When methyl side chain is attached (methylcyclohexauncertainties in density of 3 % again might be expecToluene has been extensively studied and serves astandard in several applications, but data for the lararomatics and those with side chains are genersparse. The existing data, coupled with predicmodels, might allow estimates of properties at the 5level of uncertainty for density. As with the alkanboth data and the reliability of predictive models for transport properties are worse than those for equilibrium properties. New experimental data mayrequired for the larger cyclic and aromatic specimportant for surrogate diesel fuels.

Mixture property models generally require soinformation about the binary interactions in the systeFor many diesel-range systems, information on binary pairs is not available; for some of thepredictive models may be adequate. Such models appropriately developed and validated by soexperimental data for pairs chosen within and acrossclasses of interest (e.g., n-paraffins, iso-paraffins, cyparaffins, and aromatics). Sufficiently accurate quancalculations are not currently feasible for most of binary pairs under consideration; for low pressproperties, of primary importance here, the generatio

pair potentials from ab initio calculations could allodetermination of the lowest-order corrections to ideal-gas assumptions.

It has been recognized that it is difficult to match details of a real diesel distillation curve with a fcomponent surrogate. The introduction of newer dieblend-stocks (biodiesel, coal-derived diesel, etc., discussed above) will require surrogates to follow trends in the new distillation curves. Recent workadvanced distillation curve measurements [41] allowbetter connection between the distillation curve and other phase behavior properties, and it may be poss

to match specific features of the curve by varysurrogate composition.

Spray/Mul t icomponent Vapor izat ion Models

Various approaches to handling multi-componevaporation have been described. Lippert et al. promote an approach that employs a continuprobability density function to accurately capture entire range of fuel components. However, this will lto an inconsistency once the fuel evaporates if the


9/27

9

phase chemistry is subsequently represented by adiscrete set of components.

The vaporization of an ideal, miscible multicomponentdroplet is controlled by two factors, namely the relativevolatilities of the components and the liquid-phase massdiffusion. Earlier studies have assumed that thevaporization sequence is controlled solely by the

volatility differentials, and as such resembles thedistillation process. However, fundamental dropletvaporization studies show [43-45] that the excessivelysmall liquid-phase mass diffusion can limit the rates withwhich species can be transported to the surface andvaporized. In the limit of vanishing mass diffusivity thevaporization process would assume a so-called onion-skin mode, in which the fractional vaporization rate ofthe individual species is simply its mass fraction in theliquid [46]. This in turn implies that volatility differentialhas no influence on the relative rates of vaporization.The parameter that controls the behavior is avaporization Peclet number (Pe), which is the ratio of thedroplet d

2-law vaporization rate constant to the liquid-

phase mass diffusivity. Large Pe favors the onion-skinlimit, while small Pe favors the distillation limit [47]. Inrealistic situations the vaporization behavior is a mixedone [45]. Moreover, as a result of convective shear thatoccurs during the atomization process, the liquid phasewithin droplets is not quiescent. The resulting internalconvective motions within individual droplets significantlymodifies their effective Pe number.

The above fundamental understanding may also befurther modified in diesel applications by two additionalfactors: namely (a) the increase of the droplettemperature and the attainment of the thermodynamic

supercritical state by the droplet; and (b) the fact that thedroplet size is excessively small compared to diffusivetransport length scales. Both factors could promote thedistillation behavior. A number of papers have exploredsupercritical vaporization as a means to enhance mixingwith multi-component fuels, e.g., [48-52], but furtherwork is required to understand the details of supercriticalvaporization at engine conditions and the effects on in-cylinder mixture formation. Another issue of keyimportance is the mechanism of droplet breakup andcoalescence in dense sprays, which affects virtually allaspects of diesel combustion, e.g., spray penetration,fuel vapor distribution, ignition, etc.

As far as the development of gas-phase reactionmechanisms is concerned, such an uncertainty perhapsdoes not matter as long as the mechanisms aresufficiently comprehensive that they can describe fuelmixtures of any composition. On the other hand, care isneeded when interpreting experimental results affectedby both the vaporization and gas-phase reactionprocesses.

In summary, the key issue in the vaporization of fuelblends is the sequence in which components of different

volatilities and chemical identities vaporize, throeither droplets or ligaments. Current understandindicates that the mechanism is neither batch-distillatlike, controlled primarily by volatility differentials, onion-skin-like, controlled by liquid-phase mass diffus

A mixed-mode behavior, with the attendant complexitsolution, could be the controlling mechanism. On other hand, the attainment of super-criticality a

sufficiently early stage in vaporization could implcompletely different vaporization mechanism, althothere is little understanding of this phenomenon multicomponent fuels.

CHEMICAL KINETIC PROPERTIES

Previous K inetic Development Target Research

Experimental data from laminar premixed and diffusflames, flow reactors, shock tubes, and racompression machines (RCMs) constitute valuadevelopment targets for surrogate fuel mo

development and validation. A large number of studhave been reported for molecules relevant hydrocarbon transportation fuels, including those lisin Table 2. As noted earlier, many of the species listeTable 2 fall in the gasoline/jet boiling range, reflecthe limited availability of validation data and kinmodels for components relevant for diesel fuel. Howethe kinetics significant to the combustion of thsmaller components are important to describe kinetics of intermediates formed in the combustiondiesel species of larger molecular weight. Thus, in absence of suitable diesel-range molecules, thspecies may constitute a reasonable basis from whicformulate surrogate fuels to meet various near-tdevelopment targets. The discussion that folloincludes illustrative examples of the research performfor various fuel components, with particular focusdiesel boiling range molecules.

n-Paraffins: The kinetics of n-paraffins have bextensively studied, with most efforts concentratingspecies smaller than diesel fuel-range molecules.particular relevance for surrogate fuel development, oxidation of n-heptane has been studied in numerexperiments that emphasize chemical kinetics kinetics coupled with laminar convective/molecdiffusive transport. Much of this research has b

motivated by the use of n-heptane as one of gasoline primary reference fuels. Reference 53 contaa review of relevant experimental investigationsn-heptane useful for surrogate fuel development.

The detailed model development and validation effare substantially fewer for larger n-paraffins. n-Decan-dodecane, n-tetradecane, and n-hexadecane within the diesel boiling range and are more suitadiesel surrogate components. Of these, n-decane received the most attention in fundamental experime


10/27

10

and computational kinetic model development. Forexample, n-decane oxidation was explored in a

jet-stirred reactor at temperatures of 9001300 K, 0.1MPa pressure, and equivalence ratios of 0.52 [54] toobtain concentration profiles of intermediate species.High-temperature pyrolysis and oxidation data havebeen acquired at atmospheric pressure in a flow reactor,and shock tube ignition delays have been measured at

1.31.5 MPa [55]. Laminar burning velocities andburner-stabilized premixed species profiles are alsoavailable and have been used in modeling studies [56].Vovelle and coworkers have reported measurementsfrom fuel-rich, laminar premixed n-decane flames [57-59].The authors used extractive sampling to determinespecies profiles for both stable and radical intermediatesand products. At present, all of the above studiesconcentrate on the high-temperature pyrolysis andoxidative aspects of n-decane oxidation. Fewexperimental validation data presently exist for low andintermediate temperature characterization and modeling,which are important to diesel applications.

The amount of validation data and the number ofexperimental modeling efforts for n-paraffins larger thann-decane are sparse, and are mostly an outcome of jetfuel surrogate studies (see Reference 15). Intermediatesfor n-dodecane oxidation were quantified in a flowreactor at temperatures of 600800 K, 0.8 MPa pressure,and 0.25 equivalence ratio [60]. As part of a jet fuelsurrogate study, n-dodecane was also investigated in acounterflow diffusion flame, with oxygen molar fractionextinction limits and temperature profiles measured [61].The oxidation of n-hexadecane has been studied in a

jet-stirred reactor at 0.1 MPa pressure, temperatures of10001250 K, and equivalence ratios of 0.5, 1, and 1.5

[62]. Few data besides these have been reported.

In part, the shortage of data on diesel-range moleculesreflects the difficulty of carrying out experimentalmeasurements with fuels of such low volatility. Typically,experimental studies are limited to those for which verylow concentrations (~ less than a thousand ppm) ofhydrocarbon are required due to the difficulty ofvaporizing these heavy fuels and keeping the originalfuel and reaction intermediates from condensing in gassampling systems. This limitation affects all diesel-rangemolecules, not just n-paraffins, and constitutes a barrierto surrogate diesel fuel development. Moreover, high-

temperature operation leads to the potential of thermaldegradation of the fuel, which may be particularlyproblematic in experimental apparatuses that rely onpreheating to vaporize the fuel. Additionally,experimental studies of premixed systems that utilizeprevaporization and mixing with oxidizers (air or O2) maybe problematic with fuels such as long-chain n-paraffinsthat exhibit low-temperature kinetic activity andautoignition thresholds.

Iso-Paraffins: iso-octane is a gasoline primary referencefuel and by virtue of this has been widely studied (see

[53] and references therein). However, experimestudies of larger iso-paraffins are extremely limited. production during HMN oxidation has been studied pressurized flow reactor at temperatures of 6008000.8 MPa pressure, and 0.70 equivalence ratio [63]follow-up study quantified the intermediates producediso-hexadecane in that same facility at a leaequivalence ratio of 0.3 and mixed with n-dodecane

increase the radical pool [60]. Few data have breported for these diesel-range iso-paraffins over otranges of conditions, and essentially no data existother diesel-range iso-paraffins. This limitation in preflects the small number of specific iso-paraffin isomin the diesel distillation range that are availacommercially at high purity. This is particularly relevto experimental apparatuses such as engines require large quantities of fuel. Specialized, expenscustom synthesis of the components may be requiSimilar considerations also apply to diesel-range cyparaffins and aromatics.

Cyclo-paraffins: Virtually no experimental data hbeen reported for cyclo-paraffins in the diesel boirange, and even in the gasoline range, only a few dsets are available [53]. Data for alkyl cyclo-paraffins restricted to experiments in which alkyl cyclohexa(methylcyclohexane, 1,1-dimethyl-cyclohexane, dimethylcyclohexane, 1,3-dimethyl-cyclohexane, dimethylcyclohexane, ethylcyclohexane, vinylcyhexane, ethylidenecyclohexane, and ethynyl-cyhexane) were added to stabilized co-flow burner flamof methane and air [64]. Measurements of intermedspecies identified the aromatic intermediate specproduced by alkyl cyclo-paraffin oxidation. Compathe benzene production from the different cyclo-paraf

and several paraffin and olefin isomers that were astudied revealed that cyclo-paraffins may haveproportionately greater effect than non-cyclic parafon soot production because of oxidation routes direyielding aromatic species as intermediates. This findis consistent with experimental engine data fconventional diesel combustion experiments showthat cyclo-paraffins have an effect on soot formation tis intermediate between the effects of n-/iso-parafand aromatics [65].

Aromatics: There are scant data reported for dierange aromatics. It is clear, however, that reactions

the phenyl and benzyl radicals will play important roin the oxidation of these higher molecular weight spec[66]. For example, the oxidation of side chains on sinring aromatics typically occurs first, followed by furtoxidation of the remaining ring. Similarly, single-aromatic species are formed during the combustionnonaromatic species and constitute soot precursors [Consequently, a detailed understanding of diesel-raaromatics kinetics will require a detailed understandof gasoline-range aromatics reactions.


11/27

11

Extant experimental data for gasoline range aromaticsare discussed in [53]. The laminar burning velocities andMarkstein lengths of several C7-C9aromatics have beenreported at 450 K and 0.3 MPa [68]. The comparativeautoignition behavior of numerous alkylbenzenes at highpressure and low temperatures has been investigated ina rapid compression machine [69]. The data show thatwhile toluene, m-xylene, p-xylene and 1,3,5-

trimethylbenzene ignite only at temperatures greaterthan 900 K and pressures greater than 1.6 MPa,ethylbenzene, o-xylene, 1,2,3-trimethylbenzene, 1,2,4-trimethylbenzene, n-propylbenzene, 2-ethyltoluene, andn-butylbenzene ignite at much lower temperatures andpressures. The number of alkylated sites, their relativering locations, and their chain length all affect therelative importance of low- and high-temperature kineticson autoignition. Methylated aromatics with two or moreshort alkyl chains ignite at lower temperatures if alkylsubstitution is present in an ortho configuration. Alkylsubstitutions with groups longer than C3 also result inlow temperature activity.

A follow-up study measured autoignition delays andidentified intermediates from the oxidation of o-xylene,o-ethyltoluene, and n-butylbenzene at 600900 K,equivalence ratio of 1, and 1.41.9 MPa pressures [70].Identification of intermediates showed that in the low andintermediate temperature regimes, the long alkyl chainof n-butylbenzene undergoes oxygen addition andisomerization much as n-paraffins of similar chain lengthdo. Additionally, the oxidation of n-propylbenzene wasstudied in a jet-stirred reactor at 0.1 MPa pressure,temperatures of 9001250 K, and equivalence ratios of0.51.5 [71]. Other than these studies, few data arepresent in the literature on longer chain alkylated

benzenes with molecular weights in the diesel fuel range.

Fuel Blends: Numerous studies of binary and ternaryblends of n-heptane, iso-octane, and toluene have beenreported and are discussed in References 53 and 72.Dagaut conducted jet-stirred reactor studies of binaryblends of n-decane with either n-propylbenzene orn-propylcyclohexane, as well as ternary blends of thesethree components [54]. Their studies, carried out overthe intermediate temperature range (9001300 K),atmospheric pressure, and equivalence ratio range of0.52.0, indicated that the ternary mixture reproducedbest the species profiles of kerosene. The oxidation of

single and binary blends of n-dodecane,methylcyclohexane, decahydronaphthalene (decalin),iso-cetane, and 1-methylnaphthalene has beeninvestigated in the low temperature regime (600800 K)and at elevated pressures (0.81.2 MPa) in apressurized flow reactor [73,63,74]. Experimentalresults include CO/CO2 reactivity mapping and, in thelast two references, detailed profiles of intermediatespecies.

Additionally, ternary blends of n-hexadecane, decaand 1-methylnapthalene have been used to stignition quality and the effects of cetane improadditives in fundamental ignition bomb [75], and hpressure flow reactor [76] experiments overtemperature range from 500900 K at a pressure n1.25 MPa and a fixed reaction time of 1.8 s. In additignition delays and chemiluminescence during

autoignition process have been measured in shock tstudies for the previously discussed IDEA fuel several pressures, temperatures, and equivalence ra[77].

Previous Kinet ic Model ing Work

This section briefly overviews mechanism developmwork reported for surrogate diesel fuel components. Dto the shortage of diesel-range molecules for whdetailed models have been developed, and hierarchical nature of mechanism development, predictive capability of gasoline-range molecules is a

briefly assessed as appropriate. Recently repomodeling work is emphasized, while earlier referencan be found in the cited papers. Reference 78 canconsulted for a recent comprehensive review mechanism development for a wide range hydrocarbons.

n-Paraffins: A large number of kinetic models have bdeveloped for n-heptane [13,14,79-81]. As discussedPitz et al. [53], the kinetics for n-heptane are among best described of large-carbon-number species, thoseveral combustion parameters are still not wdescribed by existing models, in particular for HCcombustion with blends.

Lindstedt and Maurice reported an n-decane mechanthat satisfactorily reproduces several species proffrom the n-decane flame experiments of Vovelle coworkers [82]. Zeppieri et al. [83] and Bikas and Pet[84] have developed n-decane mechanisms basedexperimental measurements on shock-tube ignition, freactor, and jet-stirred reactor experiments. A unidetailed kinetic model for high molecular weight parafincluding n-decane, has also been developed temperatures of 6001200 K and at pressures of 0.MPa [85]. An n-decane mechanism for the IDEA was developed by Pitsch and Peters [86]. The h

temperature model for n-decane oxidation developedZeppieri et al. [83] has been further revised acompared against a much wider set of experimeparameters, including high-temperature pyrolysis oxidation data at atmospheric pressure in a flow reacand shock tube ignition delay at 1.31.5 MPa. Mrecently, Zhao et al. [56] made additional moexperimental comparisons to laminar burning measurements in a jet-wall stagnation flaconfiguration, burner-stabilized species profiles, stirred reactor data. Although fundamental experimedata in premixed and diffusion flame configurations


12/27

12

available, no model comparisons using n-decane exist,and additional model development work is required forthis molecule.

A model for n-hexadecane oxidation has beendeveloped by Ristori et al. [62] based on their jet-stirredreactor data. Concentration profiles of intermediatespecies (though not the parent molecule) compared well

with a detailed kinetic model consisting of 242 speciesand 1801 reactions. Results showed that at lowertemperatures (1000 K) and lean conditions,n-hexadecane reacts by H-atom abstraction via O andOH radicals, while at higher temperatures ( 1200 K)and stoichiometric and rich conditions, thermaldecomposition dominates. Fournet et al. [87] used theautomatic mechanism generation code EXGAS todevelop a high-temperature mechanism with 265species and 1787 reactions. Good agreement wasachieved through comparison of the Ristori et al. [62] jetstirred reactor data, though as with the model of Ristoriet al., the n-hexadecane profile was not reproduced well.Recently, Chaos et al. [88] have extended the modelingtechniques of Zeppieri et al. [83] for high-temperatureoxidation, comparing with these data as well as pyrolysisexperiments and high-temperature atmosphericpressure flow reactor data on n-heptane/n-hexadecanemixtures.

iso-Paraffins: iso-octane, like the other gasoline primaryreference fuel n-heptane, has been the subject of asignificant amount of mechanism development activities[89-91]. Conversely, little mechanism development hasbeen reported for diesel-range iso-paraffins. A lumpedmodel for HMN has been developed that reproduces theCO production in a flow reactor at temperatures between

600800 K [63], but additional experimental andmodeling work is required.

Cyclo-paraffins: As with iso-paraffins, no mechanismshave been reported for diesel-range cyclo-paraffins, andfew exist for gasoline-range cyclo-paraffins. The high-temperature oxidation of methylcyclohexane has beenmodeled by Orme et al. [92] through comparison to theirshock tube data and the flow reactor data from Zeppieriet al. [93]. As discussed by Pitz et al. [53], the generaltrends of the data are reproduced by this model, thoughseveral discrepancies have been identified that indicatethat further development is necessary. Mechanism

development and model validation efforts are alsounderway to encompass the low temperature aspects ofthe oxidation. These studies are critical to understandingthe competition between direct abstraction and ringrupture that will also be important for modeling diesel-range cyclo-paraffins.

Aromatics: Toluene is the aromatic species that hasreceived the most modeling attention. Dagaut et al. havedeveloped a detailed kinetic model for toluene oxidationconsisting of 120 species and 920 reactions [94] thatreproduces their jet-stirred reactor data well. Skokov et

al. found that this and other toluene mechanisms fareproduce experimental flow reactor measurements1.25 MPa and 960 K [95]. The mechanisms are allslow and do not produce the appropriate hot ignireactivity profiles. Furthermore Skokov et al. show the phenyl and benzyl production of radicalsinsufficient in terms of their subsequent reacchannels [95]. Additionally, Zhao et al. [96] describ

toluene model and n-heptane/iso-octane/toluene mothat reproduce the original high-pressure ignition-dedata [95] as well as shock-tube ignition-delay data fGauthier et al. [97]. Sivaramakrishnan et al. developed a mechanism to model their high-presstoluene shock tube data that showed good agreemwith their experimental results of intermediate specconcentrations, particularly at stoichiometric conditioThe aforementioned Dagaut model [94] underpredicthe consumption of toluene at the Sivaramakrishconditions. Additionally, the Dagaut model predictsignition delay times and OH concentration profiles fVasudevan et al. [99] well at very dilute conditions not as well at higher fuel concentrations. With exception of the toluene model developed by Zhao e[96], all other models predict significantly slower rateoxidation of pure toluene at lower temperatures (~ 921.25 MPa) [96]. All models show no low temperatnegative-temperature-coefficient activity for toluewhich is consistent with experimental observations.

Few mechanisms have been developed for aromalarger than C8. A detailed kinetic model with 124 specand 985 reactions has been developed n-propylbenzene and compared to jet-stirred readata acquired at 0.1 MPa pressure, temperatures9001250 K, and equivalence ratios of 0.51.5 [

Intermediate species were quantified and were in fagood agreement with the model. Furthermore, a detakinetic model for n-butylbenzene has been develowith 197 species and 1149 reactions [100]. The motook the n-butane mechanism from the n-heptane moof Curran et al. [13] and replaced a hydrogen atom wan unreactive aromatic ring. The model was compato experimental data of n-butylbenzene oxidized irapid compression machine at temperatures of 6840 K, 1.41.8 MPa pressures, and equivalence rati1. The model predicted the ignition delay times fairly wPitsch et al. [101] developed a mechanism 1-methylnaphthalene and combined it with the redu

mechanism for n-decane in a modeling study of IDEA fuel.

NOxand Soot: NOxemissions in engine simulations typically modeled with a variation of the extenZeldovich (thermal) mechanism, as this dominateshigh temperatures (~ 2400 K). However, several otpathways may contribute appreciably to NOx formatThese include prompt, N2O, NNH, and fuel-nitropathways [102,103]. Of these, the fuel-nitrogen pathwshould be a negligible source of diesel PCCI NOxsifuel nitrogen levels are low in market fuels and sho


13/27

13

be zero in surrogates derived from pure components. Ifthe market fuel includes 2-ethylhexyl nitrate (2-EHN) asa cetane improver, the associated kinetics can beincorporated into the surrogate mechanism via reactionsoutlined by Stein et al. [76]. It is clear, however, that thekinetics of 2-EHN + fuel reactions that promoteautoignition require further study.

To the extent that diesel PCCI combustion isflameless, the contributions of the prompt and NNHpathways might be expected to play a minimal role, asthey typically require high atom concentrations typical offlames. However, calculations by Dean and Bozzelli[103] indicate that the NNH pathway can be much moreimportant than Zeldovich near 1800 K, i.e., atemperature commensurate with diesel PCCI, thoughthe NNH mechanism appears to be most important indiffusion flames where NNH may form on the fuel-richside and react with O inside the flame sheet [102].Similarly, the N2O channel is important at high pressureand intermediate temperature [102] and has beensuggested to be as important as the thermal channel inPCCI combustion, in particular at high pressure [104].Further work is required to identify the importantpathways relevant for diesel PCCI operation and torefine the rate coefficients for the relevant pathways.

Historically, soot production has been modeled byphenomenological empirical models. While goodagreement has been demonstrated with these models[105-108], their predictive capability is limited outside therange over which they were developed. Elementarymodels have received increasing attention over the pasttwo decades [109]. The chemical reactions and physicalprocesses responsible for soot precursor and soot

formation are extremely complicated, and involve gas-phase reactions, nucleation, agglomeration, oxidation,and surface growth [110]. As such, detailed models maybe too complex to incorporate into CFD calculations.

Additionally, model inaccuracies will likely continue torequire calibration of empirical parameters byexperimental observations. Nevertheless, further effortsfocused on developing elementary soot models arerequired for predictive models, and this is an area whereconsiderable work needs to be devoted.

COMPUTATIONAL CONSIDERATIONS

It is clear that the computational effort involved inincorporating detailed mechanisms in combustionsimulations is overwhelmingly demanding under mostsituations. This problem will be exacerbated with thelarge mechanisms required for diesel-range fuels andmixtures, in particular if detailed soot kinetics areincorporated. This section addresses mechanismreduction and engine simulation considerations tofacilitate the development of surrogates for premixeddiesel combustion.

Reduced Kinet ic Mechanism Development

Reduced kinetic models developed for use in Ccalculations should be derived from validated detakinetic mechanisms, and should be constructed baon firm chemical and mathematical principles to ensthat the descriptive ability of the mechanism is compromised. The large computation time required

resolve the chemistry in reacting flows is due mostlytwo factors, namely the typically large number of specand reactions, and the computational stiffness due towide range of timescales present in chemically reacsystems in comparison to convective and diffustimescales. Mechanism reduction can therefore conducted in two steps. The first, skeletal reductinvolves eliminating unimportant species and reactioThe term unimportant refers to those species reactions that do not significantly affect the predictiothe targets of interest. The second step is basedtime-scale analysis and is discussed in more debelow.

Various methods have been developed for skelreduction. Prominent among them is the approadopted in the development of GRI-Mech [111]. Treduction involves performing a series of zedimensional computations over the range of initial taconditions representing those for which original minimized mechanisms should behave similarly. each such computation, individual reaction rates analyzed at each integration time step, and elementary reaction is retained in the reduced scheonly if certain criteria are satisfied. This procedprovides a clear, robust, and physically sensible mefor identifying and removing unimportant reactions f

large reaction schemes.

The utility and validity of this strategy have bdemonstrated for systems including n-heptane [1iso-octane, n-heptane/iso-octane mixtures, and gasolines [113], and a ternary mixture model n-heptane, iso-octane, and toluene [72]. Genalgorithm methodologies for mechanism reduction optimization against kinetic targets have also bdescribed [114]. A particularly notable recdevelopment is the application of directed relation gr(DRG) theory to minimize mechanisms, to similar hdegrees of accuracy but in very short computatio

times [115-118] (See Figure 3). By use of DRG, n-heptane mechanism consisting of 561 species been reduced to a skeletal mechanism consisting of species [117].

Of great importance for the development of futreduced chemical mechanisms are recent effortsautomatic mechanism reduction. Several methods hbeen proposed, for instance by Lovas et al. [119] andand Law [115]. These methods will greatly simplify reduction work of the anticipated large chemmechanisms for diesel surrogates and provide


14/27

14

flexibility that is required to generate application-specificreduced mechanisms. It is important to note, however,that minimization by use of any of the presenttechniques, including the automatic reduction techniquescited above, yields skeletal mechanisms of very similarsize (constrained by the accuracy of replication incomparison to target predictions with the originalmechanism).

Time-scale analysis is capable of further reducing thesize of the skeletal mechanism. It is based on theobservations that fast reaction entities, such as quasi-steady species (QSS) and partial equilibrium reactions,frequently exist and as such can be exploited to removeshort time scales as well as the number of variables inthe governing equations. The major approaches areintrinsic low-dimensional manifold (ILDM) [120] andcomputational singular perturbation (CSP) [121-123].Both methods perform eigenvalue decomposition for theJacobian matrix, and assume that the reaction ratecomponent vanishes along the direction of the basevectors associated with large negative eigenvalues. CSP,in addition, considers the time dependence of theJacobian matrix, which can be extremely useful inanalyzing controlling parameters in mechanism behavior[124]. Overall, the use of time-scale analysis techniquesreduces the computational time requirements byapproximately a factor of two.

A major benefit of the methods described above is thatthey can be automated and the corresponding accuracyquantified. However, these methods typically produceminimized mechanisms that nevertheless remain difficultto incorporate into most CFD simulations, and inparticular, if the goal is to parametrically investigate

design variations and parameters.

As a result, further dimensionally reduced mechanisms,even ones that encompass smaller envelopes of targetpredictions remain of interest. Hand reduction, i.e.,non-automated reduction achieved by reactionsubstitution or modification by individual researchers canbe invoked to further reduce the mechanismdimensionality. Along these lines, a detailed mechanismfor the auto-ignition and heat release of n-heptane hasbeen reported [125,126] that consists of 185 reactionsand 43 species. A reduced 18-step mechanism hasbeen provided [126]. The larger mechanism shows good

results compared with experimental auto-ignition delaytimes and species concentrations profiles fromlaboratory experiments.

A reduced n-heptane mechanism with a wider range ofapplication was derived by systematic reduction [127]from an updated and enhanced version of the detailedmechanism by Chevalier et al. [128]. The mechanismdescribes auto-ignition, the general heat release, sootprecursor, and NOxformation. This skeletal mechanismincludes approximately 100 species and has been

0

20

40

60

0 1000 2000 3000 400

Number of Reactions

ReductionTime,ms

0

20

40

60

0 200 400 600 800 100

Number of Species

Re

ductionTime,ms

(a)

(b)

iso-octane

n-heptane

iso-octane

n-heptane

iso-octane skeletal

n-heptane skeletal

iso-octane skeletal

n-heptane skeletal

ethylene

DME

DME

ethylene

Figure 3. Measured average reduction time Direct Relation Graph (DRG) for a single reactiostate as a linear function of (a) number of reactionsand (b) number of species, of the detailemechanism.

successfully applied in simulations of a high-press

combustion chamber and predictions of NOx and sformation in a direct injection diesel engine [129].

To facilitate efficient integration of the detailed chemiinto three-dimensional CFD calculations, a compn-heptane mechanism was developed by furtreducing the existing 42-species mechanism Golovitchev [14]. In this work [80], the researchers reduced the mechanism against 24 points represena range of pressure, temperature, and equivalence raThey then optimized the ignition delay with a micGenetic Algorithm against the ignition delay calculafrom the n-heptane mechanism by Curran et al. [13] o

the same P, T, and equivalence ratio range. Tresulting mechanism was further modified to improveaccuracy when used in KIVA/CHEMKIN to prediesel-like conditions. The resulting 29-species/reaction mechanism showed savings of 70 %computational time, while giving good agreemagainst the engine data taken with number 2 diesethe fuel.

In summary, several strategies have been identifiedderiving skeletal mechanisms with quantifiable accurabased on eliminating unimportant species and reactio


15/27

15

The subsequent use of time-scale analyses based onthe concepts of steady-state species and partialequilibrium reactions, or more systematically thecomputational singular perturbation (CSP), can furtherreduce the mechanism size and stiffness. Therequirement of iteration and possible non-convergencein solving the resulting algebraic equations can bemoderated through either tabulation such as In-Situ

Adaptive Tabulation (ISAT) or analytical solution [130]. Itis, however, important to emphasize that the sizes ofthese reduced mechanisms could still be fairly large ifchemical comprehensiveness is required, and thatdrastic, ad hoc reductions could lead to falsifiedchemistry. It is also important to note that accuracy inreduction could be further compromised through errorpropagation. Dynamic reduction performed locally maybe the ultimate solution.

Thermodynamic and Com putat ional Flu id Dynamic

Engine Simulat ions

Clearly, engine combustion simulations do not constitutean optimal test platform to evaluate chemicalmechanisms. In addition to the influence of the chemicalreaction mechanism, it is important to realize that thedegree to which a CFD simulation agrees with measureddata is also governed by the ability to describe a wholerange of subprocesses including turbulence, heattransfer, spray atomization, break-up, collision,evaporation, and wall impingement. The description ofthese phenomena, each taken individually, can have animpact on the final result equal to or greater than thechoice of reaction mechanism. As a result, the followingdiscussion focuses on the engine modelingconsiderations appropriate for premixed diesel

combustion via early injection, in which the spray hasbeen vaporized and the charge is premixed at the timeof ignition. These simulations are better suited for testingchemical mechanisms than simulations of conventional,mixing-controlled diesel combustion, since the impact ofthe spray and turbulence models can be regarded asrelatively small at the start of combustion.

Multizone models are the simplest methods that havebeen applied to premixed diesel combustion. Thesimplified fluid dynamics permits relatively large kineticmechanisms to be included. These models can bedivided into different classes: (1) Sequential fluid-

mechanic chemical kinetic methods, illustrated inFigure 4 [131-135], (2) coupled CFD multi-zonechemistry solvers [136,137], and (3) direct integrationwith detailed chemistry [138,139]. The first two methodsessentially decouple the numerical solution of thechemistry from that of the fluid dynamics. Thecomputational domain in the CFD-code is divided into asmaller number of zones, each characterized by e.g. atemperature [131-133,136], a temperature andequivalence ratio [134,135,137] or temperature,equivalence ratio, and EGR rate [139,140].

Figure 4. Schematic illustrating 3D sequentialcoupled multi-zone approach [135].

Whereas the majority of the referenced multizmodels have focused on either natural gas, propaneiso-octane, Hergart et al. [135] applied the sequenmulti-zone approach using the IDEA fuel in simulathe combustion process in a mid-range diesel engineuse of an improved n-decane mechanism, a relativgood match of the low-temperature heat release w

achieved. However, as a result of the significantly larmechanism (118 species, 1000 elementary reactiothe computational times increased markeFurthermore, there was some indication that the abto predict high-temperature combustion suffered aresult of adapting the mechanism to capture the ltemperature heat release.

Direct integration with detailed chemistry involves use of detailed kinetics to solve the chemistry of ecomputational cell in the CFD domain. In the appropresented by Singh et al. [141] for modeling combustion in a DI diesel engine operating in

different modes, each computational cell is treated aperfectly stirred reactor for which the detailed chemiis solved by a separate solver (CHEMKIN). A skelmechanism of n-heptane featuring 30 species andelementary reactions was used to represent dieSince the physical properties of n-heptane, such density and volatility, are very different from thoseactual diesel fuel, a higher hydrocarbon (n-tetradecawas used in modeling injection, break-up, collision, evaporation. Relatively good agreement was obtainecapturing the heat release and NOx emissions of ltemperature combustion, but the soot model was unato reproduce measured data.

All the multi-zone models discussed above rely on assumption that variations in the relevant quanti(temperature, equivalence ratio, and possibly EGR rare negligible within each zone. This assumption is likto be valid when port injection strategies and very edirect-injection cases are simulated. However, in presence of significant stratification, it is importanaccount for variances in the parameters provided to chemistry solver. In order to address the issuestratification, statistical approaches are required. Zhet al. [142] used a joint PDF of 40 chemical species a


16/27

16

mixture enthalpy to model HCCI combustion in agasoline engine. Their results demonstrate theimportance of accounting for turbulence-chemistryinteractions with increasing stratification. Otherexamples of approaches involving the use of a PDFcombined with detailed chemistry have been presented[127,129,143,19].

The assumption that fluctuations in the relevantquantities within each zone are negligibly small is astrong function of the size of the zones. Hence, cautionneeds to be exercised in choosing the number of zones.In cases of significant stratification, the previouslydescribed direct integration approach with detailedchemistry [141] has merit and has been shown to workwell. This approach could be viewed as an extreme kindof multi-zone approach, where each computation cellrepresents a separate zone. Transport between zones ishandled in the CFD code following the operator splittingtechnique.

In situations where one simply cannot neglect theturbulent fluctuations of temperature and species massfractions within a zone, even if this is chosen to be thecomputation cell, the previously mentioned PDF method[142] provides a good alternative. Unfortunately, themethod becomes rather computationally expensive asthe number of species in the reaction mechanismincreases.

Another approach involving a PDF, but using aconserved scalar coordinate transformation to reducethe computational cost, is the Representative InteractiveFlamelet (RIF) model described by Peters [144]. Theconcept relies on the assumption that chemical time

scales are much smaller than the turbulent ones and thethickness of the reaction zone is much thinner than thesmallest turbulent length scale as represented by theKolmogorov length. Unfortunately, these assumptionsare typically not valid for premixed, low-temperaturediesel combustion. However, recently Cook et al. [145]presented an enthalpy-based flamelet model that wasfound capable of predicting HCCI combustion in a rapidcompression machine.

Overall, it may be stated that the commonly usedapproaches to modeling turbulent reacting flows havethe capability of accommodating very detailed chemistry.

To date, most applications to diesel low temperaturecombustion have featured either n-heptane or a two-component model fuel as a diesel surrogate. In order topredict primarily soot emissions and CO with a higherdegree of accuracy, a more detailed diesel surrogatefuel will be required.

PROPOSED RESEARCH PLAN

The preceding sections have outlined the current stateof fuel surrogate development relevant to diesel

combustion. In this section, we outline a methodoland proposed research plan to develop the necessdatabase and kinetic mechanisms to develop diesurrogates. As discussed earlier, we restrict attention to premixed (but not necessahomogeneous) diesel combustion targets to minimcomplications due to interactions with the spray and corresponding impact on the local fuel/air/EGR mixtu

Surrogate Component Select ion

In general, a surrogate fuel should be blended with smallest number of components that provides desired agreement with chosen targets. For applicatargets requiring a description of fuel kinetics, a critconsideration is the level of kinetic understanding for individual surrogate molecules. Another imporconsideration is the level of computational complethat can be accommodated. Even with advancesmechanism reduction and computational efficiencthere is a strong incentive for minimizing the numbesurrogate components. Based on these consideratiand the desire for making rapid progress and develophighly accurate surrogate fuel models, a two-tier plasuggested for the kinetic-related targets:

Tier 1: Given that there will be a non-negligibly lalead time before the required kinetic database kinetic models for diesel-range surrogate componewill be developed, it is appropriate to consider surrogcompositions that can be utilized in the nearer term. reasonable to choose surrogate components in gasoline or jet fuel boiling range for which data available and mechanisms exist. We recommend these near-term surrogates be comprised of n-deca

iso-octane, methylcyclohexane, and/or toluene (Table 3). This recommendation is based on availability of kinetic models and data for all thcomponents, which, though incomplete, constitutereasonable starting point for further development.

Near-Term Longer-Term

n-decane n-hexadecane

iso-octane heptamethylnonane

methylcyclohexane n-decylbenzene

toluene 1-methylnaphthalene

Table 3: Recommended Diesel Surrogate Componen

for Kinetic Targets

Tier 2: Because of the large amount of work requireddevelop the database and mechanistic understandindiesel-range molecules, it is sensible even in Tier 2focus the efforts of the community on a small numbecandidates. We recommend that in the longer tesurrogates be comprised of n-hexadecaheptamethylnonane, n-decylbenzene, and 1-metnaphthalene. While few studies suitable for surrogdevelopment have been reported with these specthey provide the opportunity to blend surrogates w


17/27

17

wide variations in ignitability (CN) and composition,which will be required to match emission targets.Specifically, n-hexadecane and HMN are diesel primaryreference fuels, and provide the ability to vary CN over awide range. n-Decylbenzene provides the ability toadjust the aromaticity of the surrogate fuel, which altersthe ignitability and emissions behavior. While it is clearfrom examination of the entries for n-decylbenzene in

Table 2 that virtually no kinetic or property data exists forthis component, this same limitation applies to all alkylaromatic molecules in the ~ C16range. The choice of ann-alkyl aromatic is motivated by the use ofn-propylbenzene (also an unbranched alkyl aromatic) asa jet fuel surrogate component [15,16], though initialtests will be required to establish n-decylbenzene'ssuitability as a surrogate diesel fuel component.1-Methylnaphthalene will likely find the greatest utility asa component added in small concentrations to help meetsoot formation targets. While kinetic mechanisms forthese components are at very early stages ofdevelopment, they can be readily built upon those ofsmaller molecules for which accurate mechanisms exist.

An additional, key attribute of these components is theiravailability at high purity commercially. Although it wouldbe desirable to include a cycloparaffin in the longer termspecies list in Table 3, there are no C14-C16 range,single-ring alkyl cycloparaffins known to the authors tobe available commercially at high purity. Undercircumstances where a diesel-range (~ C10)

cycloparaffin is desired, either butylcyclohexane ordecalin may be suitable.

Physica l Property Development Targets

Various surrogate fuel formulations will be blended andtested against the real diesel fuel. As discussed earlier,the development of a suitable diesel PCCI vaporizationsurrogate may simply require matching the liquid length,density, and T90 [29]. Liquid length experiments inconstant volume vessels should be carried out toconfirm agreement and requisite refinements ifnecessary.

Thermophysical Property Determination: Accuratethermophysical property models are required that arevalid for equilibrium properties, phase behavior (e.g.vapor-liquid equilibrium and distillation characteristics),transport properties, and surface tension over the fullranges of temperature and pressure that might beexperienced. Robust equation of state models for few-component surrogate fuels would allow study ofvariations in composition, behavior of additives, andimportant comparisons with real fuels. Such models relyon careful study of the component species, and some

measurements on select binary pairs. For some of proposed components of the surrogate, the experimedatabase is already adequate, but for others, and some of the various families of molecules foundcurrent real diesels (and potential commercial fuefurther measurements and evaluations will be necess

In particular, the equilibrium properties for n-decane

toluene are well known, and our understanding of thproperties for iso-octane, methylcyclohexane, n-hexadecane may be adequate. Further evaluatiothe equilibrium properties of HMN, n-decylbenzene, 1-methylnatphthalene will be required, although current experimental data situation may be neadequate. Measurements of the density and sospeed over ranges of temperature and pressure for HMN, n-decylbenzene and 1-methylnaphtahalcomponents would be useful in the longer term. Amthe fluids being considered as surrogate componentsdiesel, standard reference correlations for the viscoand thermal conductivity are available only for n-decaFurther analysis and measurements for the otcomponents would be beneficial. Measurementstransport properties for n-decylbenzene 1-methylnaphthalene, in particular, are lacking, measurements of both viscosity and thermal conductof the saturation states and at high pressures shouldundertaken.

As the study of diesel surrogates progresses, it wilnecessary to (a) study distillation curves of surrogates and real fuels; (b) conduct transport propemeasurements and develop models for several of pure species; (c) conduct densimetry (i.e., PVT studand calorimetric experiments on a few of the p

components, as indicated above; and (d) perform somixture experiments in order to develop and validate appropriate mixture models. Current information mallow reasonable property estimates, but these are adequate for rigorous study and understanding.

Chemic al Kinetic Development Targets

Combustion phasing: One of the major applicatargets for diesel PCCI is the combustion phasing,the heat release timing, amount, and duration ducompression heating. There are several experimeapparatuses that may be appropriate to evaluate th

targets. For example, rapid compression machines shock tubes have been extensively used to measure first and second stage heat-release timings and amou[146,147]. These apparatuses have been routinoperated at 6 MPa pressure [97,98] and capabilitiesto 60 MPa have been demonstrated [98]. Experimenthigh pressures are particularly crucial as they are mrepresentative of diesel in-cylinder conditions, and hpressure ignition is poorly predicted by mmechanisms. Ignition delays should be acquired inand simulated EGR over a wide equivalence ratio ran

Additionally, flow reactor experiments have shown


18/27

18

the CO evolution during low temperature oxidation is agood measure of low temperature oxidation [148,149],and correspondingly, developing a surrogate that hasthe same CO profile as conventional diesel may beuseful. For these targets and the ones that follow, resultsfor individual components must be supplemented withthose for binary (and eventually ternary) blends over arange of conditions and varying composition to assist

model development and validation.

Spherically expanding and stagnation/opposed jet typeflames can also be used to provide laminar burningvelocities, Markstein lengths, and extinction limits. Whilenot directly relevant to diesel combustion, theseparameters are useful development targets formechanism validation. Measurements at high pressure(> 1.0 MPa) are desirable yet have been reported onlyfor very light fuels. Stagnation and opposed jet flamescan also be used to determine ignition and extinctionlimits for both non-premixed and premixed combustionmodes. Collectively, the flame measurements should becarried out at fuel/air ratios that span the flammabilityrange, at high pressure, and with high levels ofsimulated EGR.

Emissions: Hydrocarbon, CO, and NOx emissionsrepresent more challenging targets. Suitableapparatuses include turbulent flow reactors, combustionbombs, shock tubes, and engine-like systems. Rapidcompression machines with rapid sampling valvesshould be employed to provide instantaneousmeasurements of key intermediates prior to main ignition.Where possible, optical diagnostics should be used tocomplement the extractive measurements. However, thespecies amenable to optical detection and the pressur

Development of an Experimental Database and Kinetic Models for Surrogate Diesel Fuels

Documents

surrogate diesel fuels

conventional diesel

recommended surrogate

experimental database

prototype development

efficient computational

efficient cfd tools

robust models