Combustion and Flame 145 (2006) 160–180www.elsevier.com/locate/combustflame

Aerodynamics inside a rapid compression machine

Gaurav Mittal, Chih-Jen Sung ∗

Department of Mechanical and Aerospace Engineering, Case Western Reserve University, Cleveland, OH 44106, USA

Received 9 May 2005; received in revised form 11 October 2005; accepted 22 October 2005

Available online 15 December 2005

Abstract

The aerodynamics inside a rapid compression machine after the end of compression is investigated using planarlaser-induced fluorescence (PLIF) of acetone. To study the effect of reaction chamber configuration on the resultingaerodynamics and temperature field, experiments are conducted and compared using a creviced piston and a flatpiston under varying conditions. Results show that the flat piston design leads to significant mixing of the coldvortex with the hot core region, which causes alternate hot and cold regions inside the combustion chamber.At higher pressures, the effect of the vortex is reduced. The creviced piston head configuration is demonstratedto result in drastic reduction of the effect of the vortex. Experimental conditions are also simulated using theStar-CD computational fluid dynamics package. Computed results closely match with experimental observation.Numerical results indicate that with a flat piston design, gas velocity after compression is very high and the coreregion shrinks quickly due to rapid entrainment of cold gases. Whereas, for a creviced piston head design, gasvelocity after compression is significantly lower and the core region remains unaffected for a long duration. As aconsequence, for the flat piston, adiabatic core assumption can significantly overpredict the maximum temperatureafter the end of compression. For the creviced piston, the adiabatic core assumption is found to be valid even upto 100 ms after compression. This work therefore experimentally and numerically substantiates the importance ofpiston head design for achieving a homogeneous core region inside a rapid compression machine. 2005 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Keywords: Rapid compression machine; PLIF of acetone; Adiabatic core hypothesis; Computational fluid dynamics;Autoignition

1. Introduction

In a rapid compression machine (RCM) study, pri-mary experimental data consist of the pressure traceof a given reacting mixture. A typical pressure tracebefore autoignition taking place shows a rapid rise inpressure during the compression stroke, which is ofthe order of 15–40 ms, followed by a gradual decreasein pressure due to heat loss from a constant-volume

* Corresponding author.E-mail address: cjs15@po.cwru.edu (C.-J. Sung).

0010-2180/$ – see front matter 2005 The Combustion Institute.doi:10.1016/j.combustflame.2005.10.019

chamber at the end of compression. Although in prin-ciple RCM simulates a single compression event,complex aerodynamic features can affect the state ofthe reacting core in the reaction chamber. Previousstudies (e.g., [1–3]) have shown that the motion of thepiston creates a roll-up vortex, which results in mix-ing of the cold gas pockets from the boundary layerwith the hot gases in the core region. Such undesiredmixing leads to difficulties in accurately characteriz-ing the state of the reacting mixture.

In modeling the RCM experiments, for simplicityit is often assumed that the aerodynamic effects donot play any significant role at the short time scales

Published by Elsevier Inc. All rights reserved.

G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180 161

encountered in the RCM. It is further assumed thatthe core gas, away from the thermal boundary layer,is compressed adiabatically. Thus, temperature evolu-tion is determined from the measured pressure pro-file by assuming the adiabatic core hypothesis. Basedon this hypothesis, the core temperature at any in-stant during compression, T (t), can be determinedfrom the experimentally measured pressure, P(t), ac-

cording to the relation∫ T (t)T0

γγ−1

dTT

= ln[P(t)/P0],where P0 is the initial pressure, T0 the initial tem-perature, and γ the specific heat ratio. Furthermore,to model the RCM data, the computed pressure vari-ation needs to match with the experimental pressuretrace by including an empirically determined heattransfer coefficient or volume expansion with a zero-dimensional model [4–6]. However, substantial dis-crepancies have been observed between data takenfrom different rapid compression machines even un-der similar conditions of temperature and pressure[7,8]. These discrepancies are attributed partly to thedifferent heat loss characteristics after the end of thecompression stroke and partly to the difference inaerodynamics between various machines.

The effect of aerodynamics is particularly morecomplicated because it does not show up in the pres-sure trace and it may lead to significant temperaturegradients and ultimately to the failure of the adiabaticcore hypothesis. If the aerodynamic effect becomessignificant and the adiabatic core hypothesis fails,there is no easy way to determine the temperatureinside the reaction chamber. As a result, unambigu-ous determination of the state of the reacting mixtureand systematic characterization of the resulting aero-dynamic field inside an RCM are important for ob-taining reliable kinetic data from RCMs and bridgingthe gap between data taken from different machines.

Several studies have contributed to the understand-ing of the aerodynamics and temperature field insidean RCM by computational fluid dynamics (CFD) cal-culations and experimental measurements. Griffiths etal. [1] numerically showed that the hot core regiongenerated at the end of compression is virtually adia-batic and spans approximately 70% of the volume ofthe combustion chamber at the end of compression.Griffiths et al. [1] also observed differences betweencomputational results using spatially uniform condi-tions and CFD simulation. Differences between twosets of computations were attributed to the effect oftemperature gradient that was accounted for in CFDanalysis [1]. In the recent study of Clarkson et al. [2],the temperature field was imaged by Rayleigh scatter-ing and laser-induced fluorescence (LIF) of acetone.Acetone-LIF was found to nicely characterize thetemperature variations in the RCM, whereas Rayleighscattering was relatively less sensitive [2]. It was ex-perimentally observed that the roll-up vortex had pen-

etrated the center of the combustion chamber at theend of compression [2]. The temperature differencebetween the hot gases and the roll-up vortex was esti-mated to be 50 K [2]. Furthermore, LIF of acetonegenerated by decomposition of di-t-butyl peroxideunambiguously showed the temperature stratificationat the center of the reaction chamber [2].

Griffiths et al. [9] investigated temperature andconcentration fields in a rapid compression machineby using a number of experimental techniques, in-cluding Schlieren photography, planar laser-inducedfluorescence (PLIF) of acetone, PLIF of formalde-hyde, and chemiluminescence imaging. In order toillustrate the interaction of chemistry with the temper-ature field in the RCM, Griffiths et al. [9] contrastedthe combustion behavior of di-t-butyl peroxide withthat of n-pentane. The overall reaction of the former ischaracteristic of thermal ignition, while the combus-tion of the latter was investigated in the compressedtemperature range exhibiting a negative temperaturedependence of the overall reaction rate. With imag-ing being taken up to 10 ms after the end of com-pression, results showed that the di-t-butyl peroxidereaction proceeded faster in the zone of peak temper-ature [9]. Somewhat similar behavior was observedfor n-pentane combustion when the compressed tem-perature was at the lower end of the negative tempera-ture dependence range [9]. By contrast, at compressedtemperatures close to the upper end of the nega-tive temperature dependence region, the reaction inthe cooler zone developed faster and the temperatureinhomogeneity inside the reaction chamber rapidlysmoothed out [9]. However, Griffiths et al. [9] pointedout that spatial inhomogeneity of concentrations ofintermediates can be there, which would affect theeventual evolution of spontaneous ignition.

Griffiths et al. [10] used chemiluminescence imag-ing along with filtered Rayleigh scattering in a rapidcompression machine to characterize the transitionfrom nonknocking to knocking reaction and the evo-lution of the spatial development of the reactivity.Results from filtered Rayleigh scattering gave evi-dence of a cooler core region at compressed tem-peratures below the negative temperature depen-dence region [10]. In contrast, Rayleigh scatteringdid not show a cooler core at compressed tempera-tures within the region of negative temperature de-pendence. Hence, the effect of the negative tempera-ture dependence of reaction rate is to smooth out thetemperature inhomogeneity inside the reaction cham-ber [10].

Lee and Hochgreb [3] theoretically demonstratedthat the roll-up vortex due to the piston motion can becounteracted by deliberately machining a crevice intothe side of the piston. Creviced piston allowed moreaccurate predictions of the reacting temperature from

162 G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180

pressure–time records. Predictions of a simple heattransfer model in conjunction with a creviced pistonwere found to agree well with experimental pressurehistory [3].

Desgroux et al. [11,12] made direct measurementsof temperature in an RCM using thermocouple andsingle point Rayleigh scattering. These measurementsconfirmed the existence of an adiabatic core gas atthe end of compression. Rayleigh scattering measure-ments were made with a temperature accuracy of 3–4%, but after the end of compression the local temper-ature exhibited a standard deviation of 10–15% [11].Desgroux et al. [12] conducted fine-wire thermocou-ple measurements for nonreactive and reactive mix-tures at different radial locations. They observed auniform temperature field for a few milliseconds af-ter the end of compression and subsequent develop-ment of temperature inhomogeneity due to heat lossto the wall and gas recirculation [12]. For nonreactivemixtures the temperature inhomogeneity persisted af-ter compression, whereas for reactive iso-octane mix-tures a temperature-leveling effect, ascribed to thenegative temperature coefficient of reaction rate, wasobserved [12].

Donovan et al. [13] took direct thermocouple mea-surement in a free-piston rapid compression facility toassess the adiabatic core hypothesis. Measurementswere corrected for slower response of the thermocou-ple. Donovan et al. [13] asserted that experimentalmeasurements at the end of the compression strokejustified the adiabatic core gas assumption. However,the thermocouple measurements failed to give a truerepresentation of temperature after the end of com-pression because of the failure of the model used tocorrect the time response of the thermocouple [13].

Recently, Würmel and Simmie [14] conductedCFD studies of a twin-piston rapid compression ma-chine using Star-CD. From the CFD simulations, thepiston head crevices were optimized for the twin-piston RCM in terms of volume, location, and the di-mension and geometry of the channel connecting thecrevice and the chamber. The channel connecting thecrevice and the chamber was optimized and the idealgeometry was found to be rectangular [14]. An an-gled channel design was seen to help somewhat in thecooling of the trapped gas in the piston head crevice.Würmel and Simmie [14] also observed strong depen-dence of the crevice performance on the test gas used.Specifically, although an optimal crevice was identi-fied for test gases such as nitrogen, oxygen, and argon,it was not possible when using helium, due to the as-sociated high heat loss. Instead of helium, the use ofxenon as a bath gas for RCM experiments was recom-mended [14].

The present investigation aims to extend the previ-ous efforts by experimentally and numerically study-

ing the effect of the piston head geometry on aerody-namics and temperature field inside an RCM. It hasimplications for kinetic modeling in the sense thathow well the prediction of the adiabatic core hypoth-esis matches with the actual temperature in the reac-tion chamber for a long time duration after the end ofcompression. Requirement of experimental data, par-ticularly under the conditions of homogeneous chargecompression ignition (HCCI) operation, requires theability to sustain unambiguous reaction conditions forlonger times. This is so because the relevant HCCIconditions are extremely lean or highly diluted, whichsignificantly increases the ignition delay times. There-fore, it is particularly important to characterize thestate of the aerodynamics and the resultant tempera-ture field for long time duration after the end of com-pression stroke.

In the present work, temperature field inside anRCM is studied using planar laser-induced fluores-cence (PLIF) of acetone. Laser diagnostic techniquein an RCM offers many advantages in comparison tothe thermocouple measurements. Thermocouple mea-surements suffers from the drawback that one cannotreadily obtain information about instantaneous spa-tial variation of temperature. Moreover, there will beaerodynamic disturbances if the thermocouple probeis not sufficiently fine [12]. Additionally, correctionof the measured temperature due to the slow time re-sponse of the thermocouple requires knowledge of thevelocity field inside the chamber. The validity of themodel used to determine the velocity field and cor-rect for time response may be under question. Byresorting to a longer compression time, as conductedin [12], while the effect of thermal inertia of the ther-mocouple can be minimized, aerodynamics at longcompression times may not be a true representationof what happens at short compression times. In con-trast, temperature mapping using the laser techniquesis practically nonintrusive and is capable of giving in-stantaneous spatially resolved temperature field.

Both a flat piston and a creviced piston are em-ployed and compared in the present study. These ex-periments will demonstrate whether the uniformity ofthe temperature field is indeed improved by using acreviced piston, as suggested by the earlier theoret-ical study [3]. The present investigation is particu-larly useful because experiments are conducted on thesame RCM. The mapping of the temperature field iscarried out for two different piston head configura-tions under varying operating conditions for a longtime duration after the compression. In order to in-vestigate the effect of the piston geometry on ignitiondelay, experiments are also conducted for autoigni-tion of iso-octane using both piston head configura-tions. In addition, numerical simulations are carriedout using Star-CD CFD package. Computed results

G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180 163

further provide insights into the nature of the aerody-namics inside an RCM. Based on these experimentaland computational studies, conditions under whichthe adiabatic core hypothesis can be satisfactorily ap-plied are delineated.

In the following sections we will first highlightthe features of the RCM employed herein. Experi-mental specifics for the ignition delay measurementsand PLIF of acetone are then described, followed byexperimental and numerical results on the characteri-zation of aerodynamics inside the RCM.

2. Experimental

2.1. Rapid compression machine

Fig. 1 shows the schematic of the present RCMsystem that consists of a driver piston, reactor piston,hydraulic motion control chamber, and a driving airtank. The driver cylinder has a bore of 5 in. (12.7 cm)and that of the reactor cylinder is 2 in. (5.08 cm).The machine is pneumatically driven and hydrauli-cally stopped. Stroke can be varied between 7 and10 in. (17.78 and 25.4 cm) by adjusting the spacerson the hydraulic cylinder. Clearance is also adjustableand can be varied by using split shims between thehydraulic cylinder head and the reactor cylinder. Fur-

ther details of the present rapid compression machinecan be found in [15]. The cylindrical reaction cham-ber is equipped with the sensing devices for pressureand temperature, gas inlet/outlet ports for preparingthe reactant mixture, and quartz windows for opticalaccess. Dynamic pressure during compression is mea-sured using Kistler 6125B transducer with a 5010Bcharge amplifier.

In order to study the effect of the piston head de-sign on aerodynamics in an RCM, experiments areconducted for two different piston head configura-tions: a creviced piston and a simulated flat piston.An enlarged view of the configuration of the crevicedpiston head is also shown in Fig. 1. The location ofthe crevice is on the cylindrical surface of the pistonand the piston face is kept flat. On the other hand, theterm “flat piston” indicates a piston that does not havea crevice along its cylindrical periphery and hence isflat on the face and along the piston circumference. Inthe present study, the flat piston head is simulated byfilling the crevice volume with a pack of o-rings. Thisfilling results in approximately 90% reduction in thecrevice volume and gives an almost flat piston.

Two sets of experiments are conducted and com-pared using both piston head configurations. First,ignition delays for stoichiometric iso-octane/oxygen/inert gas mixtures are measured in the temperaturerange of 684 to 878 K, from which the effect of piston

Fig. 1. Schematic of the rapid compression machine (RCM), the creviced piston head configuration, and the acetone-PLIF setup.Dimensions of the creviced piston in mm: A = 0.50, B = 4.0, C = 0.15, D = 20.0, and E = 1.50.

164 G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180

head configuration on autoignition is demonstrated.Second, PLIF imaging of acetone for temperaturefield mapping is carried out under conditions of rel-atively low pressure (approximately 12 bar) and highpressure (approximately 39.5 bar). These experimentsalso provide insights into the effect of pressure on theresulting temperature field inside the rapid compres-sion machine.

2.2. PLIF of acetone

For the characterization of the temperature field,laser-induced fluorescence of acetone tracer in nitro-gen is studied from the end of the compression stroke.The mixture consists of 1 to 2% (by concentration)acetone and remaining nitrogen. The schematic of thePLIF imaging setup is also depicted in Fig. 1. An ap-proximately 5-mm-wide laser sheet with a waist of50 µm is made to traverse the central plane of thecombustion chamber, which is equidistant from thereactor piston face and the end of the chamber. Thecombustion chamber is equipped with two 0.66-in.(1.67-cm) diameter optically flat quartz windows forthe traversal of the laser sheet. The end of the cham-ber is fitted with a 2-in. (5.08-cm) diameter and 1.7-in. (4.318-cm) thick quartz window, which providesfull view of the combustion chamber for fluorescenceimaging. An intensified CCD camera (Princeton In-struments, PI-MAX, 1024 × 256 pixels) with a UVlens (Nikon, f = 105 mm) is placed perpendicularto the laser sheet in order to record the fluorescenceimages. A WG305 (Schott) filter is used to filter scat-tered light from the fluorescence signal. A frequency-doubled Continuum Nd:YAG system is used in con-junction with a dye laser filled with Rhodamine 590solution. The resulting laser energy at wavelength of279 nm is around 8 mJ/pulse.

The synchronization of the laser firing with ma-chine is achieved in the following manner. The laseris continuously fired at a repetition rate of 10 Hz. Thesignal generated from the start of the ICCD cameraacquisition is used to actuate a relay after a speci-fied time delay. Actuation of the relay fires the RCM.A subsequent laser pulse, which occurs after the endof compression, provides the pumping source for theinduced fluorescence. The timings of laser pulses andpressure trace of RCM are simultaneously recordedusing a data acquisition system, which allows accu-rate determination of the timing of laser pulse relativeto the end of compression. After every experiment,the laser optics is realigned to correct for any minormovement of the machine as a result of firing. Sincethe repetition rate of the laser system is 10 Hz, PLIFimages are single-shot measurements.

Fig. 2. Ignition delay time versus the adiabatic core tem-perature at TDC for stoichiometric iso-octane/oxygen/inertmixtures. Composition: iC8H18/O2/inert = 1/12.5/47. Ini-tial conditions: P0 = 331 Torr and T0 = 297 K. The adi-abatic core temperature at TDC is varied by changing thecomposition of the inert gases.

3. Experimental results

3.1. Ignition delay

Experiments are generally conducted for a strokeof 10 in. (25.4 cm) and clearance of 0.525 in.(1.33 cm). Time for the compression stroke is ap-proximately 30 ms, and the geometric compressionratio when using a creviced piston is 15.1. Fig. 2shows a plot of ignition delay, measured in the presentRCM, versus the adiabatic core temperature at thetop dead center (TDC), Tc, for stoichiometric iso-octane/oxygen/inert gas mixtures. Based on the ex-perimental pressure trace, Tc is calculated according

to∫ TcT0

γγ−1

dTT

= ln[Pc/P0], where Pc is the com-pressed pressure at TDC. Experiments are conductedusing a flat piston and a creviced piston. For a givenpiston head configuration, the adiabatic core temper-ature is varied by changing the composition of theinert gases (argon and nitrogen), while keeping a fixedcompression ratio. Results of the present work arealso compared with the experimental data of Minettiet al. [16] under similar conditions of pressure, com-position, and equivalence ratio. For the creviced pis-ton, the initial pressure and the initial temperature arekept fixed at 331 Torr and 297 K, respectively. Forthe flat piston, clearance is increased to compensatefor the absence of the crevice so that the conditionsat TDC are identical to those for the creviced pistonwhen same initial conditions of pressure, temperature,and composition are used. This results in adiabaticcore temperature at TDC ranging from 684 to 878 Kand pressure at TDC varying from 13.3 to 16.25 bar.

In spite of similar conditions, Fig. 2 demonstratessubstantial discrepancies among different sets of ex-perimental data. This comparison shows that not onlythe data from different RCMs under similar test con-

G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180 165

ditions may be different, but also for the same RCMthere may be significant differences depending uponthe configuration of the piston head and the final re-acting volume. Furthermore, it is seen from Fig. 2 thatwhen a flat piston is used, ignition delays are sig-nificantly reduced as compared to those obtained byusing a creviced piston. It is therefore expected thatchanging from a creviced piston head to a flat pistonhead significantly affects the heat loss and the result-ing aerodynamics inside the RCM, and the effect re-flects in the form of considerable change in ignitiondelay. This example is presented here to highlight theimportance of the piston head configuration effect onignition delay.

3.2. Acetone fluorescence

The acetone fluorescence signal can be highly sen-sitive to temperature. At 279 nm excitation, fluores-cence signal reduces by 42% as temperature increasesfrom 600 to 800 K [17,18]. Fig. 3 shows a represen-tative single-shot PLIF image of a homogeneous ace-tone/nitrogen mixture at room temperature of 297 Kand pressure of 44 bar, along with the associated pho-ton count integrated along the width of the laser sheet.For this experiment, a mixture of acetone and nitrogenat an initial pressure of 2.91 bar is gradually com-pressed by slowly moving the reactor piston. Suchcompression is nearly isothermal because the com-pressed gases are allowed to cool down to the roomtemperature. After the desired compressed pressureis reached, the fluorescence measurement is subse-quently taken. Laser sheet enters from the right sideof the image. The location of cylinder wall is alsoshown in the figure. Decay in the fluorescence signal

Fig. 3. A representative single-shot acetone-PLIF image andthe associated photon counts integrated along the width ofthe laser sheet. Chamber conditions: pressure = 44 bar andtemperature = 297 K. Laser sheet enters from the right sideof the image.

from right to left is because of strong absorption. Ac-cording to the Beer–Lambert law, in a homogeneousfield with uniform temperature and concentration, sig-nal decays exponentially due to absorption of the laserintensity. By correcting for absorption and incorporat-ing for temperature sensitivity of fluorescence signal,it is possible to determine the resultant temperaturefield from fluorescence signal. Parameters for tem-perature sensitivity of acetone fluorescence are takenfrom Thurber et al. [17,18].

Using the fluorescence intensity shown in Fig. 3,Fig. 4 demonstrates the procedure for deducing thetemperature distribution. In Fig. 4, radial directionequal to 0 corresponds to the central axis of the cylin-drical chamber, while 2.54 and −2.54 cm representthe cylinder wall on either side. Moreover, the dot-ted line is the actual fluorescence signal, while thedashed line represents the exponential decay in thefluorescence intensity, as calculated using the Beer–Lambert law and the known temperature of 297 K.In addition, the deduced temperature distribution isdenoted by the solid line. It is seen from Fig. 4 thatin a uniform temperature field, signal follows the ex-ponential decay feature and the deduced temperaturenicely predicts the temperature distribution. In Fig. 4,the deduced temperature is noted to be within ±4 K,which is within 1.35% of the actual temperature. Thisgives an estimate of the uncertainty of the deducedtemperature due to the noise in the fluorescence sig-nal.

When there exists temperature nonuniformity inthe chamber, knowledge of the temperature at oneanchor point is required to deduce the temperaturedistribution from the fluorescence signal. Relative tothis anchor point, temperature along the radial direc-tion can subsequently be deduced by knowing thetemperature dependence of the fluorescence intensity.

Fig. 4. Temperatures deduced from the fluorescence signal.Chamber conditions: pressure = 44 bar and temperature =297 K. Dotted line: raw fluorescence intensity. Dashed line:calculated fluorescence intensity using the Beer–Lambertlaw. Solid line: deduced temperature distribution.

166 G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180

However, such an anchor temperature is difficult to bemeasured during the actual RCM experiment. In thepresent work, the maximum temperature in the cham-ber at any time instant is taken as the adiabatic coretemperature, calculated from the pressure trace based

on the expression∫ T (t)T0

γγ−1

dTT

= ln[P(t)/P0].The procedure to deduce the temperature distribu-

tion from the integrated fluorescence intensity profileis as follows. An anchor point in the fluorescence in-tensity profile is first arbitrarily chosen at some pixelinside the chamber and given a temperature valueequal to the adiabatic core temperature derived fromthe experimental pressure trace at that instant. Sub-sequently, the temperatures at other radial locationsare determined by marching radially in both direc-tions toward the wall from this anchor point. Whilemarching radially, correction for absorption is con-ducted according to the Beer–Lambert law. Accord-ing to this law, attenuation in the laser intensity (I )as the laser traverses a length dx is expressed asI/I0 = exp(−σndx), where I0 is the incident laserintensity, σ is the absorption coefficient, and n repre-sents the acetone number density. At each marchingstep, after correcting for the absorption, the temper-ature is calculated from the knowledge of the tem-perature dependence of the fluorescence intensity, forwhich data is taken from Thurber et al. [17,18]. Bymarching radially from the first chosen anchor point,the temperature distribution along the entire domainof the chamber is therefore deduced. If the result-ing maximum temperature in the chamber is higherthan the adiabatic core temperature, the temperatureof the anchor point is reduced for the next iteration.The iteration procedure continues until the maximumdeduced temperature in the chamber equals the adia-batic core temperature.

It has to be pointed out that the assumption of themaximum temperature being equal to the adiabaticcore temperature does not affect the pattern of the de-duced temperature in any way. If the actual maximumtemperature is somewhat lower than the adiabatic coretemperature, the entire temperature profile would shiftdownward, without affecting the pattern of the tem-perature distribution.

3.3. Flat piston

3.3.1. Low compressed gas pressureFig. 5 shows the experimental results conducted at

an initial pressure of 236 Torr and an initial temper-ature of 297 K using the simulated flat piston. Themixture consists of 2% acetone in nitrogen. At theend of compression, the gas pressure and tempera-ture are 12.5 bar and 780 K, respectively. Again, theraw fluorescence signal is shown as the dotted lineand the deduced temperature is denoted as the solid

line. Exponential decay of the fluorescence intensity,as may be observed in a uniform temperature field, isshown as the dashed line. Time 0 is taken as the endof the compression stroke. The fluorescence signalat 1 ms after compression shows increased intensityin the central portion of the chamber. Any deviationof fluorescence signal from exponential decay is in-dicative of temperature inhomogeneity. In addition,any hump corresponds to reduced temperature, whileany depression indicates an increase in the tempera-ture. At 1 ms, an increased fluorescence intensity inthe central portion is due to the effect of the roll-upvortex, which brings in cold gases from the bound-ary layer to the center of the chamber. In the tem-perature field, this increase in fluorescence intensitycorresponds to approximately 100 K reduction in thetemperature. At 6 ms postcompression, as we go fromthe wall to the centerline, there are alternating coldand hot regions. Apart from the temperature reductionof approximately 50 K in the central regime, there isanother zone of low temperature near the wall. Thereason for this type of temperature nonuniformity willbe discussed later on along with the discussion ofcomputational results.

At time steps of 8 and 18 ms, the width of the cen-tral depression zone increases due to thermal trans-port, whereas the temperature difference associatedwith this depression decreases. In addition, the ef-fect of the side depression becomes more pronounced.Similar features are observed in the temperature fieldat subsequent time steps. Even at a long time scaleof 129 ms, sharp temperature gradients exist in thechamber. Furthermore, these PLIF results are consis-tent with those of previous studies [2,9].

Since only one PLIF measurement can be im-aged during every run of the RCM, aerodynamic fea-tures captured at the specified time instance may varyslightly in different runs. For instance, at 19 ms inFig. 5, instead of depression in the temperature dis-tribution at the center, a peak in the center and twoadjacent zones of low temperature are observed. Thiswould suggest that the cold gases flowing across thepiston head did not reach the center of the chamberfrom either direction. Although slightly different pat-tern may be observed in different runs even at thesame reference time after compression, it does not al-ter the conclusion that the use of the flat piston leadsto substantial temperature nonuniformity due to theeffect of the roll-up vortex. In general, PLIF experi-ments are found to give fairly repeatable temperaturedistribution.

3.3.2. High compressed gas pressureFig. 6 illustrates another set of experimental re-

sults using the simulated flat piston, with an initialpressure of 700 Torr and an initial temperature of

G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180 167

Fig. 5. PLIF intensities and the deduced temperature distributions at varying times after compression for a simulated flat pistonhead. Gas composition: 2% acetone in nitrogen. Conditions at TDC: pressure = 12.5 bar and temperature = 780 K. Dotted line:raw fluorescence intensity. Dashed line: calculated fluorescence intensity in a uniform temperature field using the Beer–Lambertlaw. Solid line: deduced temperature distribution.

297 K. The molar concentration of acetone is 1% innitrogen. This results in the compressed gas pressureof 39.5 bar and compressed temperature of 815 K atTDC. The features of the temperature field are simi-lar to those obtained at the low compressed pressure.Specifically, there exists temperature depression inthe center at the end of compression and side depres-

sions appear at the subsequent time steps. In contrastto the measurement at low compressed pressure, athigh compressed pressure the central depression inthe temperature distribution is smaller, being around65 K. At the subsequent time steps, the central tem-perature depression at higher compressed pressure isgenerally less than that at lower compressed pressure.

168 G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180

Fig. 6. PLIF intensities and the deduced temperature distributions at varying times after compression for a simulated flat pistonhead. Gas composition: 1% acetone in nitrogen. Conditions at TDC: pressure = 39.5 bar and temperature = 815 K. Dottedline: raw fluorescence intensity. Dashed line: calculated fluorescence intensity in a uniform temperature field by using theBeer–Lambert law. Solid line: deduced temperature distribution.

3.4. Creviced piston

3.4.1. Low compressed gas pressureUsing a mixture consisting of 1% acetone in ni-

trogen, a series of experiments with the creviced headpiston are conducted. Initial conditions of 275 Torrand 297 K yield compressed gas conditions at TDCof 11.95 bar and 760 K. PLIF results and the de-

duced temperature distribution at varying time stepsare shown in Fig. 7.

At 4.1 ms postcompression, the effect of the vor-tex has not reached the center of the chamber and thedepression in temperature is observed to be approx-imately 35 K. At 6.2 ms, the central depression cor-responds to approximately 30 K, as shown in Fig. 7.However, at 12 ms the effect of the vortex has grown

G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180 169

Fig. 7. PLIF intensities and the deduced temperature distributions at varying times after compression for a creviced piston head.Gas composition: 1% acetone in nitrogen. Conditions at TDC: pressure = 11.95 bar and temperature = 760 K. Dotted line: rawfluorescence intensity. Dashed line: calculated fluorescence intensity in a uniform temperature field by using the Beer–Lambertlaw. Solid line: deduced temperature distribution.

and the central depression in temperature is approxi-mately 80 K.

In comparison with the case of the flat piston atlow pressure, the effect of the vortex is reduced, al-though it is not eliminated. Fig. 7 also demonstratesthat the development of temperature inhomogeneity isrelatively gradual for the case of the creviced piston.

3.4.2. High compressed gas pressureFig. 8 shows the experimental results for the mix-

ture consisting of 1% acetone in nitrogen, with theinitial conditions of 870 Torr and 297 K. At the end ofcompression, the compressed gas conditions at TDCare 39.5 bar and 770 K. It is seen from Fig. 8 thatthere is no evidence of any temperature inhomogene-

170 G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180

Fig. 8. PLIF intensities and the deduced temperature distributions at varying times after compression for a creviced piston head.Gas composition: 1% acetone in nitrogen. Conditions at TDC: pressure = 39.5 bar and temperature = 770 K. Dotted line: rawfluorescence intensity. Dashed line: calculated fluorescence intensity in a uniform temperature field by using the Beer–Lambertlaw. Solid line: deduced temperature distribution.

ity at 2 ms postcompression. The temperature in thechamber is quite uniform and the fluorescence sig-nal shows exponential decay. At the subsequent timesteps, even up to 114 ms, there is no effect due to thevortex. Eventually, the effect of the vortex becomesnoticeable. At 200 ms, the depression due to the vor-tex is approximately 40 K, while at 390 ms the centraldepression corresponds to approximately 100 K.

These results clearly show the effects of the pres-sure and piston head design on the aerodynamicswithin an RCM. In comparison to the results usingthe creviced piston at low pressure (Fig. 7), at highpressure the effect of the vortex roll-up is significantlyreduced. While at low pressure the effect of the vortexappears from the end of compression, at high pres-sure such an effect becomes noticeable only at time

G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180 171

greater than 114 ms postcompression. Furthermore, incomparison to the results using the flat piston at highpressure, the creviced piston at high pressure yieldsa much higher degree of homogeneity inside the testchamber for long time duration.

4. Numerical analysis

Star-CD CFD package is used to simulate the ex-perimental conditions and provide insights into thedetailed evolution of the aerodynamic field inside anRCM. As with the experimental tests, simulation isconducted for conditions of low, intermediate, andhigh compressed gas pressures, and for the flat as wellas the creviced piston head configurations.

Because of the cylindrically symmetric geometryof the combustion chamber, numerical simulation iscarried out on an axisymmetric grid distribution. Inthe actual physical situation in the RCM, flow may de-viate from axisymmetric behavior and exhibit asym-metrical patterns. However, axisymmetric configura-tion is nevertheless chosen because it considerablyreduces the computational time while capturing theessentials of the underlying fluid dynamics. Compu-tations are performed from the beginning of the com-pression stroke with a compression time of 30 ms.Initially, the test gas, nitrogen, at rest is specified withuniform temperature and pressure. A constant temper-ature of 297 K and no slip condition are specified atthe cylinder wall boundary and piston surface.

In the simulation, the piston starts from rest andits motion is given in a manner similar to the pis-ton motion in an engine, by specifying dimensions forthe crank radius and the connecting rod length. Thisspecification of velocity profile is consistent with theoperation of the present RCM, in which the piston ac-celerates from rest initially, eventually decelerates ata constant rate, and comes to a stop due to the hy-draulic stopping mechanism. Furthermore, by varyingthe crank radius and the length of the connecting rodin the CFD calculations, it is observed that as long asthe compression time is kept constant, the resultingflow field is insensitive to reasonable changes in thevelocity profile. This observation is consistent withthe finding of Würmel and Simmie [14]. At top deadcenter, the piston comes to rest and remains therefor subsequent time steps. A time step of 55.55 µsis taken, as further reduction of the time step by afactor of two results in identical solution. During thecompression stroke, the vertex filling methodology ofStar-CD [19] is chosen for the motion of the grid. Inthis method, the same number of grids is maintainedbetween the cylinder head and the piston head, but thegrids become compressed as the piston moves. Afterthe end of the compression stroke, the computational

Fig. 9. Computational grid distribution for a creviced pistonat the end of compression. All length dimensions are in mm.

grids remain the same for the post compression pe-riod.

The computational grid distribution employed forthe creviced piston configuration at the end of com-pression is shown in Fig. 9. The main reaction cham-ber consists of 100 grids in the radial direction and140 grids in the axial direction. In the radial direc-tion, the finer grids are used near the boundary. In theaxial direction, an accordion distribution is used, withthe finer grids near the piston and cylinder head andthe coarser grids in between. In order to ascertain thequality of the grid distribution, selective cases are cal-culated on a finer grid distribution of 150 × 200 in theradial and axial directions, respectively. The use ofthe finer grids results in an identical flow pattern andless than 0.25% change in predicted temperature. Asa result, all the computed results shown herein are ob-tained on a 100 × 140 grid distribution.

In addition, calculations are performed and com-pared for both laminar and turbulent flow conditions.Since it cannot be easily ascertained a priori whetherthe postcompression flow field inside an RCM is lam-inar or turbulent, it is important to perform calcula-tions using both laminar and turbulent models, andassess their predictions by comparing with the exper-imental results. Various turbulent models are consid-ered. A standard version of the k–ε model is found toyield extremely high turbulent mixing. On the otherhand, calculations using a RNG model yield relativelyless turbulent mixing and are found to be closer to theexperimental observation in comparison to the resultsobtained using the k–ε model. Furthermore, computa-tional results show that the features of aerodynamicsare better predicted by laminar calculations, while theactual aerodynamic pattern is expected to be some-

172 G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180

Fig. 10. STAR-CD simulation of temperature and velocity fields at varying times after compression for a flat piston at a pressureof 12.8 bar at TDC under laminar condition.

where in between the laminar and turbulent calcula-tions.

5. Computational results

Using the flat piston, Fig. 10 shows the computedfields of velocity and temperature for various time

steps after the end of compression under laminar con-ditions. This is the case with a relatively low com-pressed pressure of 12.8 bar at TDC. At the end ofcompression, i.e., at 0 ms, maximum velocity in thechamber is 12.33 m/s. There is a big vortex span-ning almost the entire chamber, which shears coldgases from the boundary and brings them inside. As

G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180 173

Fig. 11. STAR-CD simulation of temperature and velocity fields at varying times after compression for a flat piston at a pressureof 13.14 bar at TDC under turbulent condition.

such, the regime near the centerline is greatly af-fected by the flow of the cold gases from the bound-ary. Along with the main vortex, the computed re-sults show the formation of corner vortices near thecylinder wall, the piston head, and the centerline.At 20 ms after compression, the maximum veloc-ity has reduced to 3.27 m/s, but the velocity patternis markedly different from that at the end of com-pression. The corner vortex on the piston head alsobecomes significantly large. At the subsequent timesteps, the pattern of the flow field remains almostidentical and the velocity in the chamber continuesto decay. At 60 ms, the maximum velocity is reducedto 1.19 m/s.

As a comparison, Fig. 11 shows the computed re-sults using the RNG turbulent model under similarconditions of Fig. 10. The compressed pressure atTDC is now 13.14 bar. For this turbulent simulation,the maximum velocities at the end of compression(0 ms) and at 10 ms postcompression are respectively13.38 and 5.48 m/s, which are not much differentfrom those of the laminar simulation. However, theturbulent calculations do not exhibit any corner vor-tex, as seen in the laminar model. This is expectedas turbulence suppresses the onset of flow separation.Fig. 11 also demonstrates that the nature of velocityfield remains identical for all the time steps calcu-lated. In the temperature field, enhanced mixing and

174 G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180

shallow temperature gradient are shown in Fig. 11, ascompared to the laminar results.

When comparing PLIF data with the computa-tional results, it can be clearly seen that the simulationusing the laminar model agrees well with the fea-tures of the aerodynamics observed in experiments,whereas the turbulent calculations fail to do so. Theexistence of sharp temperature gradients in the experi-mentally deduced temperature profile is also shown inthe laminar flow calculations. It should be mentionedthat although the turbulent calculations may not com-pletely capture the qualitative trend observed in thePLIF experiments, the comparison of Figs. 10 and 11provide insights into the effect of turbulence on theaerodynamics inside an RCM.

As with the experimental results, the effect of thevortex on the temperature distribution near the cen-tral portion of the chamber at the end of compressionis also noted in the numerical simulation. However,depression in the temperature predicted by the sim-ulation is greater (200 K) than that obtained by theexperimental data (100 K). This quantitative differ-ence can be attributed to a number of reasons as fol-lows. The piston used in the experiment is a simulatedflat piston with approximately 10% of the crevice vol-ume and not a truly flat piston as employed in thesimulation. Therefore, relatively smaller effect of thevortex in the experiment than what would occur fora truly flat piston for which simulation is calculatedis expected. In addition, simulation is conducted forconditions of constant wall temperature. But in theactual experiment, it is expected that the wall temper-ature increases by a few kelvin, resulting in smallertemperature nonuniformity.

Furthermore, there exist alternating hot and coldregions in the experimental results using the flat pis-ton (cf. Figs. 5 and 6). Moving from the wall bound-ary toward the center, there is a zone of high tem-perature. After this zone, there is a low temperatureregime, followed by another zone of high tempera-ture. Finally, the center of the chamber is at low tem-perature. Such a temperature field can be producedonly by multiple vortices, rotating in different direc-tions. Similar features can be observed in the com-puted temperature and velocity fields at 20 ms for thelaminar calculations, as shown in Fig. 10b. There arealternating hot and cold regions and the velocity fieldexhibits primarily two vortices.

Based on the laminar calculations, Fig. 12 showsthe computed temperature and velocity fields for thecreviced piston at a relatively low pressure of 11.9 barat TDC. For the ease of representation, only themain chamber volume, apart from the crevice zone, isshown in Fig. 12. At the end of compression (0 ms),there is small temperature nonuniformity near the pis-ton head. The maximum predicted gas velocity at this

time is 2.29 m/s, which is much smaller than the gasvelocity for the flat piston. The effect of the vortex isseen to gradually grow at the subsequent time steps.Even at 80 ms post compression, unlike the flat pistoncase, significant part of the chamber is unaffected bythe vortex, as shown in Fig. 12e.

Fig. 13 shows the simulated temperature fields forthe flat piston at a higher pressure of 40.7 bar at TDC,using the laminar model. Patterns of the velocity andtemperature fields are identical to those calculated atlow pressures. However, at higher pressures, the effectof the vortex penetrates at a much slower rate. Forinstance, the temperature distribution at 40 ms under ahigher compressed pressure (Fig. 13c) is very similarto that at 20 ms under a lower compressed pressure(Fig. 10b).

Compared with Fig. 13, Fig. 14 shows the laminarsimulation for the creviced piston at a similar com-pressed pressure of 39.9 bar at TDC. At the end ofcompression, apart from minor nonuniformity at thecenter, the effect of the vortex is practically nonex-istent. At the postcompression time of 100 ms, thevortex effect eventually reaches the central plane ofthe chamber. Similar features are observed in the ex-perimental results. However, the laminar calculationspredict an earlier onset time of the temperature inho-mogeneity, as the experimental results do not observeany apparent vortex effect until 114 ms post compres-sion.

In addition, the effect of pressure on the vortex for-mation for the creviced piston is consistent in both theexperimental and computational results. Increase inpressure reduces the effect of the vortex. For a flat pis-ton, although such a pressure effect cannot be clearlyobserved in the experimental data, it is demonstratedin the computational results. The effect of pressureon the vortex roll-up and the induced temperature in-homogeneity can be attributed to the change in thethermal diffusivity. At a higher compressed gas pres-sure, the mixture thermal diffusivity is lower, therebyleading to a thinner thermal boundary layer. Sincetemperature inhomogeneity is induced by the shear-ing of cold gases from the boundary to the interior ofthe cylindrical chamber, a thinner boundary layer un-der high compressed gas pressures will slow down thedevelopment of the temperature inhomogeneity.

6. Assessment of the adiabatic core hypothesis

The adiabatic core hypothesis is based on the as-sumption that the effect of heat loss due to the cham-ber wall does not affect the temperature of the core re-gion at the short time scales encountered in an RCM.From the present experimental results using both flat

G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180 175

Fig. 12. STAR-CD simulation of temperature and velocity fields at varying times after compression for a creviced piston at apressure of 11.9 bar at TDC under laminar condition.

and creviced piston heads, the validity of the adiabaticcore hypothesis can be clearly identified.

For the case of flat piston at a low pressure, atthe end of compression (Fig. 5a) there is large tem-

perature gradient at the center of the chamber, whilethe rest of the chamber is not much affected. How-ever, at the post compression time of 6 ms (Fig. 5b),the temperature gradient is present across the entire

176 G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180

Fig. 13. STAR-CD simulation of temperature field at varying times after compression for a flat piston at a pressure of 40.7 barat TDC under laminar condition.

domain of the chamber and the effect of rolled-upcold gases has spread everywhere. Therefore, at 0 msthere is a possibility that the adiabatic core assump-tion may hold, but the actual temperature quickly be-gins to deviate from the adiabatic core temperature.From 6 ms onward, the failure of the adiabatic corehypothesis is expected. For the flat piston at a highcompressed pressure, the temperature gradient acrossthe entire chamber is observed at 15.8 ms postcom-pression (Fig. 6c).

For the case of creviced piston at a high pressure,even at 114 ms post compression (Fig. 8f), the ef-fect of the vortex is not observed. This suggests thevalidity of the adiabatic core assumption up to thistime when using the creviced piston at high pressures.However, at low pressures, the validity of the adia-batic core hypothesis may not be as good.

Similar features regarding the validity of adia-batic core hypothesis are also observed in the com-putational results. Fig. 15 plots the time evolutionof the pressure and the maximum temperature, ob-tained from the simulations employing different con-ditions and piston head configurations. In addition,the temperature profile calculated by using the adi-

abatic core hypothesis is also shown and compared.In Fig. 15, the temperature based on the adiabaticcore hypothesis is deduced from the correspondingsimulated pressure trace. At the end of compression(0 ms), the core temperature predicted by the hypoth-esis exactly matches with the maximum temperatureof the CFD analysis. For the flat piston, however,the computed maximum temperature using the lam-inar model, shown in Figs. 15a and 15b, quickly be-gins to significantly deviate from the temperature ofthe adiabatic core assumption as the postcompres-sion time proceeds. When the turbulent simulationis used, Fig. 15c shows that a larger overpredictionresulting from the adiabatic core assumption is ob-served. In contrast, for the creviced head piston theadiabatic core hypothesis accurately predicts the max-imum temperature for a long period of postcompres-sion time, as shown in Figs. 15d, 15e, and 15f.

Fig. 16 compares the extent of the core regionfrom the end of compression simulated using two dif-ferent piston head configurations, for both laminarand turbulent calculations. Here the core region is de-fined as the ratio of the volume of the chamber thatis within 5% of the maximum instantaneous temper-

G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180 177

Fig. 14. STAR-CD simulation of temperature field at varying times after compression for a creviced piston at a pressure of39.9 bar at TDC under laminar condition.

ature to the total volume of the chamber at the endof compression. For the case of creviced piston, onlythe main cylinder volume, apart from the crevice, isconsidered for the core region calculation. Note thatthe core region may not be the geometric center of thechamber, but the region of the maximum temperature.

For the flat piston at low pressure of 12.8 bar atTDC, while at the end of compression (time = 0 ms)the core region encompasses approximately 70% ofthe volume of the chamber, it rapidly reduces to35% at 20 ms and continues to reduce as the post-compression time proceeds. The present simulationalso demonstrates that although relatively large gasvolume is at adiabatic core temperature at the end ofcompression, the heat transport rates from this coreregion are not low. The failure of the adiabatic coreassumption after the end of compression results fromthe high gas velocity field in a small confined cham-ber. This can be understood from the time evolutionof the maximum gas velocity from the end of com-pression, as shown in Fig. 17. Significantly highervelocities with the flat head piston in comparison tothe creviced piston are seen in Fig. 17. High gas ve-

locities at the end of compression quickly bring theeffect of wall into the core region.

For the flat piston at a high pressure of 40.7 barat TDC, Fig. 16a shows that the extent of the core re-gion is relatively larger than that at a low compressedpressure of 12.8 bar. In contrast, with the crevicedpiston, the extent of core is significantly increased.Fig. 16b further demonstrates the effect of the turbu-lence on the extent of the core region. In general, theeffect of turbulent mixing lowers the temperature ofthe core region while increases its spatial extent. Itshould be noted that under both laminar and turbu-lent conditions the primary reason for the failure ofthe adiabatic core hypothesis is due to the mixing ofcold gases sheared from the wall boundary with thecore region, rather than the heat loss from cold wallitself.

Fig. 17 also illustrates that the use of a crevicedpiston head substantially reduces gas velocity at theend of compression. As a result, the temperaturepredicted by the adiabatic core hypothesis matchesclosely with the actual maximum temperature for longtime interval, as shown in Fig. 15. Additionally, the

178 G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180

Fig. 15. Assessment of the adiabatic core hypothesis based on computational results. Pressure at TDC is indicated in the figure.Dashed line, solid line, and line with symbols respectively represent pressure, calculated temperature using the adiabatic corehypothesis, and actual maximum temperature in the chamber.

extent of the core region is greatly improved, as seenin Fig. 16. Since the determination of the core tem-perature in an RCM is typically indirect, the presentnumerical experiments therefore demonstrate that it isextremely important to have proper operating condi-tions and geometric configuration of the test chamberso that the adiabatic core hypothesis is valid.

Although various RCMs have different operatingconditions of clearance, stroke, etc., similar featuresof aerodynamics as observed in this study are ex-pected. Depending upon the extent of the vortex ef-fect, various RCMs would vary in terms of the validityof the adiabatic core hypothesis after the end of com-pression, although such a hypothesis is expected to beaccurate at TDC. This would explain the discrepan-cies in the reported experimental data from differentRCMs, even under similar compressed conditions.

7. Effect of the piston head configuration onignition delay

It has been shown in Fig. 2 that under similarconditions the use of different piston head configu-rations can lead to different ignition delay. In order toaddress the reasons, additional simulations and exper-iments are carried out based on the same compressedgas temperature and pressure for a flat and a crevicedpiston head configurations. To achieve the same con-dition at the end of compression, the clearance isincreased to account for the absence of the crevicevolume in the flat piston case.

Fig. 18 shows the numerical comparison using ni-trogen. Although pressure and temperature at TDCare the same for both piston head configurations, therate of pressure drop for the creviced piston is higherthan that for the flat piston. This result is attributed to

G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180 179

Fig. 16. Extent of the core region, calculated from com-putational results, as a function of postcompression time.(a) Effect of pressure and piston head configuration usingthe laminar model. (b) Effect of turbulence. Pressure at TDCis indicated in the figure. Solid line: laminar calculation.Dashed line: turbulent calculation.

Fig. 17. Time variation of the maximum velocity in the cylin-der from the end of compression for laminar calculation. Thepressure at TDC is indicated in the figure.

the larger surface area to volume ratio for the crevicedpiston case. Crevice is a zone of small volume butlarge surface area, which accelerates the rate of heatloss to the wall. Due to the higher postcompressionpressure in the case of the flat piston, the core temper-ature calculated by the adiabatic core hypothesis isalso higher. However, it is seen from Fig. 18 that thecomputed maximum temperature for the flat head pis-ton case can become lower than that for the crevicedhead piston case. Consistent with the computational

Fig. 18. Computational results on the effect of piston headconfiguration under the same conditions of temperature andpressure at TDC. Cases of a flat and a creviced piston headconfigurations with nitrogen as the test gas are shown.

Fig. 19. Comparison of measured pressure profiles foriso-octane autoignition experiments using a flat and acreviced piston head configurations. Composition: iC8H18/O2/inert = 1/12.5/47. The adiabatic core temperature atTDC is indicated in the figure. Solid line: flat piston. Dashedline: creviced piston.

results shown in Figs. 15–17, for the flat piston casethe adiabatic core assumption is not valid, and thesignificant effect of the roll-up vortex decreases thecore temperature, despite of higher postcompressionpressure. The resulting lower core temperature thanthat determined based on the adiabatic core hypoth-esis and the higher postcompression pressure for theuse of the flat piston will in turn affect the autoigni-tion delay.

Fig. 19 compares the experimental pressure tracesusing stoichiometric iso-octane/O2/inert gas mixturesfor both piston head configurations under similarcomposition, compressed gas temperature, and TDCpressure. Results for three different adiabatic coretemperatures, namely 684, 745, and 878 K at TDC,are compared in Fig. 19. Both the single-stage (878 K)and the two-stage (684 and 745 K) ignition phenom-ena are clearly shown. Although the compressed pres-sures at TDC match quite well for both pistons, it isseen in Fig. 19 that the postcompression pressure for

180 G. Mittal, C.-J. Sung / Combustion and Flame 145 (2006) 160–180

the flat piston is higher, as demonstrated by the com-putational results. The combined effect of the lowercore temperature than that calculated using the adia-batic core hypothesis and the higher postcompressionpressure for the flat piston case alters the ignition de-lay as depicted in Fig. 19.

8. Conclusions

The purpose of this work is to investigate the ef-fects of aerodynamics on the measured propertieswithin a rapid compression machine. Such an un-derstanding is important in order to unambiguouslycharacterize the state of the reacting mixture and toexplain the reasons for the mismatch of ignition delaydata obtained from various RCMs. Experimental andcomputational studies are conducted for a flat and acreviced piston head configurations. Experimental re-sults using PLIF of acetone demonstrate the suitabil-ity of a creviced piston to provide a homogeneous re-acting core. Computational results under laminar con-ditions closely reproduce the features of the aerody-namics that are observed in the experiment. For a flatpiston, although the adiabatic core assumption accu-rately predicts the temperature at the end of compres-sion, it significantly overpredicts the postcompressiontemperature as the effect of fast-moving cold gasesquickly spreads all over the chamber. A creviced pis-ton head is found to significantly increase the extentof the core region. Therefore, when using a properlydesigned creviced piston, the postcompression tem-perature can be accurately predicted based on the adi-abatic core hypothesis. It is also observed that the ef-fect of the roll-up vortex is reduced as the compressedpressure is increased.

The present results demonstrate that using acreviced piston design, the actual temperature in anRCM can be deduced from the pressure trace. Fromthe kinetic modeling point of view, unambiguous de-termination of the temperature is the prime require-ment, as the heat loss associated with an RCM canbe easily accounted for by using an empirical rela-tion. This work suggests that when a properly de-signed creviced piston is used, a zero-dimensionalmodel should satisfactorily model the experimentalresults using an RCM. However, when a flat pistonis employed, the use of a zero-dimensional model

to simulate the experimental data is deemed inade-quate.

Acknowledgment

This work was supported by the National ScienceFoundation under Grant No. 0133161.

References

[1] J.F. Griffiths, Q. Jiao, A. Schreiber, J. Meyer, K.F.Knoche, W. Kardylewski, Combust. Flame 93 (1993)303–315.

[2] J. Clarkson, J.F. Griffiths, J.P. Macnamara, B.J.Whitaker, Combust. Flame 125 (2001) 1162–1175.

[3] D. Lee, S. Hochgreb, Combust. Flame 114 (1998) 531–545.

[4] C.K. Westbrook, W.J. Pitz, J.E. Boercker, H.J. Curran,J.F. Griffiths, C. Mohamed, M. Ribaucour, Proc. Com-bust. Inst. 29 (2002) 1311–1318.

[5] S. Tanaka, F. Ayala, J.C. Keck, Combust. Flame 133(2003) 467–481.

[6] M. Ribaucour, R. Minetti, L.R. Sochet, H.J. Curran,W.J. Pitz, C.K. Westbrook, Proc. Combust. Inst. 28(2000) 1671–1678.

[7] R. Minetti, M. Carlier, M. Ribaucour, E. Therssen, L.R.Sochet, Proc. Combust. Inst. 26 (1996) 747–753.

[8] E.J. Silke, H.J. Curran, J.M. Simmie, Proc. Combust.Inst. 30 (2005) 2639–2647.

[9] J.F. Griffiths, J.P. MacNamara, C. Mohamed, B.J.Whitaker, J. Pan, C.G.W. Sheppard, Faraday Dis-cuss. 119 (2001) 287–303.

[10] J.F. Griffiths, J.P. MacNamara, C.G.W. Sheppard, D.A.Turton, B.J. Whitaker, Fuel 81 (2002) 2219–2225.

[11] P. Desgroux, L. Gasnot, L.R. Sochet, Appl. Phys. B 61(1995) 69–72.

[12] P. Desgroux, R. Minetti, L.R. Sochet, Combust. Sci.Technol. 113 (1996) 193–203.

[13] M.T. Donovan, X. He, B.T. Zigler, T.R. Palmer, M.S.Wooldridge, A. Atreya, Combust. Flame 137 (2004)351–365.

[14] J. Würmel, J.M. Simmie, Combust. Flame 141 (2005)417–430.

[15] G. Mittal, C.J. Sung, Combust. Sci. Technol., in press.[16] R. Minetti, M. Ribaucour, M. Carlier, L.R. Sochet,

Combust. Sci. Technol. 113–114 (1996) 179–191.[17] M.C. Thurber, F. Grisch, R.K. Hanson, Opt. Lett. 22

(1997) 251–253.[18] M.C. Thurber, F. Grisch, B.J. Kirby, M. Votmeier, R.K.

Hanson, Appl. Opt. 37 (1998) 4963–4978.[19] Star-CD, Version 3.20 User Guide.