PHYSICS OF FLUIDS VOLUME 13, NUMBER 11 NOVEMBER 2001 Effects of nonreacting solid particle and liquid droplet loading on an exothermic reacting mixing layer Richard S. Millera) Department of Mechanical Engineering, Clemson University, Clemson, South Carolina 29634-0921 共Received 14 September 2000; accepted 30 July 2001兲 Numerical simulations are conducted of two-dimensional 共2D兲 exothermic reacting mixing layers laden with either solid particles or evaporating liquid droplets. An irreversible reaction of the form fuel⫹r Oxidizer→共1⫹r兲 Products with exothermic Arrhenius kinetics is considered. The temporally developing mixing layers are formed by the merging of parallel flowing oxidizer and fuel streams, each uniformly laden with nonreacting particles or droplets. The gaseous phase is governed by the compressible form of the Navier–Stokes equations together with transport equations for the fuel, oxidizer, product, and evaporated vapor species concentrations. Particles and droplets are assumed smaller than the gas-phase length scales and are tracked individually in the Lagrangian reference frame. Complete ‘‘two-way’’ couplings of mass, momentum, and energy between phases are included in the formulation. The simulation parameters are chosen to study the effects of the mass loading ratio, particle Stokes number, vaporization, flow forcing, and reaction Zeldovich number on the flame evolutions. Quasi-one-dimensional simulations reveal that the asymptotic state of the laminar flames is independent of the particle or droplet loading. For forced 2D simulations, both particles and droplets are preferentially concentrated into the high-strain braid regions of the mixing layer. Cold solid particles entrained into the mixing zone cool the flame in the braid regions due to their finite thermal inertia. This results in flame suppression and, under certain conditions, local flame extinction in the braids. The potential for flame extinction is substantially enhanced by evaporating droplets through the latent heat, and also by the addition of nonreacting evaporated vapor which locally dilutes the reactant concentrations. In contrast, combustion proceeds robustly within vortex cores which have relatively dilute droplet distributions due to preferential concentration; particularly at moderate Stokes numbers. The extent of flame suppression and local extinction are increased with increasing reaction activation energy, dispersed phase mass loading, and also by decreasing particle or droplet Stokes number. © 2001 American Institute of Physics. 关DOI: 10.1063/1.1404137兴 I. INTRODUCTION structures which play a primary role in the turbulence dynamics.2,3 Later studies examined the role of these structures in chemically reacting mixing layers through theoretical,4 experimental,5–7 and numerical8,9 analyses. The vortical nature of this flow has a substantial impact on dispersed phase dynamics and dispersion. This occurs due to the ‘‘preferential concentration’’ mechanism,10,11 through which heavy particles migrate away from high-vorticity regions and towards high-strain regions of the mixing layer.12–20 Particle dispersion in mixing layers is maximized for ‘‘intermediate’’ Stokes numbers 共ratio of particle time constant to eddy turnover time scale兲 near unity.15,16 In contrast, very small Stokes number particles closely follow fluid motions, while very large Stokes number particles remain relatively unaffected by fluid motions due to their relatively large inertia. The vast majority of prior two-phase simulations have been restricted to solid particle laden flows. Generally, the particles are assumed to be governed by only the drag force and gravitational settling, under ‘‘one-way’’ coupling conditions. One-way coupling refers to cases with relatively small particle mass loading ratios for which the presence of the particles does not significantly affect the surrounding carrier gas flow.21–23 In this case only a limited number of particles Multiphase turbulent combustion occurs in a variety of natural and practical applications; including, liquid fueled combustion, solid propellant combustion, and fire suppression and control. An important subset of these reacting flows are those in which the dispersed phase is nonreacting; such as with fire suppressants. The scope of the present investigation is limited to gaseous nonpremixed turbulent flames laden with nonreacting discrete spherical solid particles or liquid droplets smaller than the smallest length scale of the 共corresponding single-phase兲 gas flow. Furthermore, only relatively small volumetric loadings (⬃10⫺3 ) are considered, for which direct particle–particle collisions may be neglected.1 The mixing layer formed by the merging of parallel flowing fluid streams remains one of the classic flow geometries for fundamental studies of turbulent mixing and combustion. Early investigations revealed that these turbulent shear flows are dominated by large scale organized vortical a兲 Telephone: 864-656-6248; [email protected] fax: 864-656-4435. 1070-6631/2001/13(11)/3303/18/$18.00 Electronic mail: 3303 © 2001 American Institute of Physics Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp 3304 Phys. Fluids, Vol. 13, No. 11, November 2001 (⬃103 →104 ) are tracked as dictated by the requirements for obtaining accurate statistics. Such calculations have been conducted for a variety of flows;11 including homogeneous turbulence,10,24 –28 mixing layers13,15,16,18,19,29 and channel flows.30–33 On the other hand, simulations of ‘‘two-way’’ coupled particle laden flows at larger mass loading ratios have received significantly less attention owing to the combined numerical difficulties of calculating the phase coupling terms and tracking sufficiently large numbers of particles (⬃105 →106 ) as required to achieve desired mass loadings. Two-way coupled solid particle simulations have typically been performed only for homogeneous turbulence.28,34 –39 One exception to this trend is the recent numerical investigation of the solid particle laden mixing layer by Meiburg et al.40 Their two-dimensional 共2D兲 simulations showed that two-way coupling effects are strongest for intermediate Stokes numbers. In contrast to solid particle laden flows, evaporating droplet laden flows must generally be described by a twoway coupled and compressible 共or variable density兲 formulation due to a vaporization source term in the gas-phase continuity equation. Due to these added complexities, the extent of simulations of liquid laden flows remain considerably less than for solid particle two-phase flows. Mashayek41,42 and Jaberi and Mashayek43 conducted simulations of both isotropic and homogeneous shear box turbulence with two-way coupled liquid droplets. Miller et al.44 – 46 extended two-way coupled computations to nonhomogeneous flows with simulations of a three-dimensional 共3D兲 planar mixing layer having one stream laden with evaporating hydrocarbon droplets. Under the conditions of these studies, the droplet laden stream reaches a state of ‘‘saturation’’ 共cessation of vaporization兲 due to the combined effects of gas cooling and growth of the evaporated vapor mass fraction. However, inside the mixing layer droplets mix with higher temperature fluid from the nonladen stream and evaporation proceeds to completion. In these regions, initially monodisperse droplet size distributions evolve to polydisperse distributions due to nonuniform rates of evaporation. Preferential concentration also has an important effect on the evolution of evaporating droplets which have timedependent Stokes numbers. At early times droplets with intermediate Stokes numbers accumulate in the braid regions of the layer for vaporization rates smaller than the eddy turnover time scale. Maximum vapor concentrations are therefore observed in the braid structures for both 2D and 3D mixing layers. The above citations have only addressed nonreacting flows. In contrast, numerical simulations of two-phase reacting flows have received markedly less attention. Miller and Bellan44 investigated an infinitely fast adiabatic reaction based on a passive conserved scalar mixture fraction in the context of analyzing the validity of the assumed particle distribution function 共PDF兲 method for two-phase combustion. Mashayek47 extended their earlier computations to simulate reacting homogeneous shear turbulence. In this case, fuel droplets are immersed in a carrier gas oxidizer and a simple chemical reaction of the form Fuel⫹Oxidizer → Products was considered. Simulation of two-phase react- Richard S. Miller ing non-homogeneous flows remains limited; however, Glaze and Frankel48 recently simulated one-way coupled solid particle dispersion in a reacting jet using the large eddy simulation 共LES兲 technique. The primary objective of the present paper is to use numerical simulation to investigate the complex couplings of turbulent flames with a nonreacting dispersed phase in the form of either solid particles or evaporating droplets. In particular, the effects of vortical rollup, reaction activation energy, dispersed phase mass loading, and particle or droplet size are considered. Only nonreacting particles and droplets are studied; both as a direct interest and as a prelude to future reacting droplet investigations. In this work 2D temporally developing mixing layers are considered as a model flow representative of the physical processes inherent in fully turbulent 3D reacting flows. The primary reason for this limitation is the exceedingly large computational cost of simulating the corresponding 3D flow. Miller and Bellan46 captured transition to turbulence in their simulations of a 3D temporally developing mixing layer using as many as 18⫻106 grid points and 5.7⫻106 evaporating droplets. However, Miller et al.49 were unable to reproduce the transition to turbulence for similar flow conditions due to the stabilizing effects of density stratification in their single-phase supercritical mixing layer. Preliminary tests performed for this study confirm that this situation is maintained for the present flow which contains stratification as well as exothermicity and particle loading. Therefore, rather than simulate low-Reynolds number 3D mixing layers with limited vortical development, we choose to examine 2D flows at relatively large Reynolds number with multiple vortical pairings. Extension of the present simulations to 3D domains under conditions conducive to transition would be prohibitively expensive 共e.g., extension of our cases SLD-HIGHc and EVAP-HIGHc described below to 3D would require 91⫻106 grid points and 447⫻106 particles or droplets兲. Despite the fact that many of the important physical processes found in 3D flows are also present in the 2D mixing layer 共including vortical structures and pairing, preferential concentration, particle–flow interaction, and local flame extinction in high-strain flow regions兲, the reader is cautioned that the results of the present investigation should not be quantitatively extrapolated to the more complicated 3D turbulent flows. The paper is organized as follows: The formulation, numerical approach, mixing layer configuration, and flow parameters are described in Sec. II. Results from the simulations are presented in Sec. III, which includes subsections addressing laminar flame simulations, flow visualizations, as well as effects of the Zeldovich number, the dispersed phase mass loading ratio and the initial Stokes number. Conclusions and final discussions are the subject of Sec. IV. II. FORMULATION AND APPROACH The governing equations describe the Lagrangian transport of discrete solid particles or evaporating droplets through a continuous, calorically perfect and chemically reacting carrier gas flow. Throughout the formulation, the subscripts F, O, P, and V refer to the fuel, oxidizer, product, and Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp Phys. Fluids, Vol. 13, No. 11, November 2001 Effects of nonreacting solid particle evaporated vapor species, respectively; whereas subscript G refers to the multispecies gas-phase mixture. For the dispersed phase, S denotes solid particles and L denotes liquid droplets. The compressible form of the governing equations for the gas-phase include mass, momentum and energy exchange between the gas and the nonreacting dispersed phase are ⫹ 关 u j 兴 ⫽S I , t x j 共1兲 共 ui兲⫹ 关 u i u j ⫹ P ␦ i j ⫺ i j 兴 ⫽S II,i , t x j 共2兲 冋 T ⫺u i i j 共 et兲⫹ 共 e t ⫹ P 兲 u j ⫺ t x j x j 册 ⫽S III ⫺ ˙ P ⌬H 0 , P⫽ 共3兲 兺␣ 关 R 0 /W ␣ 兴 T, 共4兲 where is the gas phase density, u i is the gas-phase velocity, P is the thermodynamic pressure, e t ⫽e⫹u i u i /2 is the total gas energy, i.e., kinetic energy plus internal energy 关e ⫽⌺ ␣ (Y ␣ C v , ␣ T⫹h V0 ) where the vapor reference enthalpy (h V0 ) is chosen to be the same for all gas phase species兴, and Y ␣ , R ␣ ⫽R 0 /W ␣ and C v , ␣ are the mass fraction, gas constant, and constant volume specific heat of species ␣, respectively 共the universal gas constant is R 0 , the molecular weight is W ␣ and C p, ␣ ⫽R ␣ ⫹C v , ␣ is the constant pressure heat capacity兲. Furthermore, i j is the Newtonian viscous stress tensor, is the viscosity, is the thermal conductivity, and ⌫ is the Fickian species diffusion coefficient. The right-hand side 共rhs兲 terms S I , S II,i , and S III describing the phase couplings of mass, momentum, and energy, respectively, are defined below. A single-step, irreversible and exothermic reaction of the form Fuel⫹r Oxidizer→共1⫹r兲 Products is considered through the solution of transport equations for the mass fractions (Y ␣ ) of the oxidizer ( ␣ ⫽O), fuel ( ␣ ⫽F), and product ( ␣ ⫽ P) 冋 册 Y ␣ Y ␣u j⫺ ⌫ ⫽ ˙␣, 共 Y ␣兲⫹ t x j x j 共5兲 and, in the case of liquid droplets, for the nonreacting evaporated vapor 冋 册 Y V Y Vu j⫺ ⌫ ⫽S I . 共 Y V兲⫹ t x j x j 共6兲 All species are assumed to diffuse at the same rate, with diffusivity ⌫. The chemical kinetics are based on the Arrhen˙ ␣ ) kinetics ius form of the reaction rate ( 冉 冊 冉 冊 ˙ O ⫽⫺ K R 共 T 兲 ˙ F ⫽⫺ K R 共 T 兲 WP Y Y , WF O F WP Y Y , WO O F 共7兲 共8兲 ˙ P ⫽⫹ 共 1⫹r 兲 K R 共 T 兲 冉 冊冉 冊 WP WF WP Y Y , WO O F 3305 共9兲 with temperature dependent reaction coefficient: K R (T) ⫽A 0 exp关⫺E0/(R0T)兴, where E 0 is the activation energy, and A 0 is the reaction rate constant 共with inverse units of time兲. Miller et al.50 compared eight different vaporization models with experimental data for single droplets. Based on their findings, a variant of the classical ‘‘D 2 law’’ is chosen to predict the droplet evolutions. The following model has previously been employed for calorically perfect species by Miller and Bellan45 in nonequilibrium form, and by Miller and Bellan46 in equilibrium form. Nonequilibrium effects were found to become significant only for very small droplet sizes 共⬍50 m兲 in high-temperature environments50 and are, therefore, not included in the present investigation. The modeled Lagrangian equations describing the transient position (X i ), velocity ( v i ), temperature (T p ), and mass (m p ) of a single particle 共or droplet兲 are dX i ⫽vi , dt 共10兲 dvi Fi ⫽ , dt mp 共11兲 dT p Q⫹ṁ p L V , ⫽ dt m pC L 共12兲 冉 冊冉 冊 dm p 1 ⫽ṁ p ⫽⫺m p dt p Sh ln关 1⫹B M 兴 , 3Sc G 共13兲 where F i is the modified Stokes drag force F i ⫽m p 冉 冊 f1 共 u i⫺ v i 兲, p 共14兲 Q is the heat flux to the surroundings Q⫽m p 冉 冊冉 f2 p 冊 NuC p,G 共 T⫺T p 兲 , 3PrG 共15兲 the subscript p denotes individual particle conditions, the particle time constant for Stokes flow is p ⫽ L D 2 /(18 ) 共substitute subscript L→S and dm p /dt⫽0 for solid particles兲, D is the particle diameter, C L is the heat capacity of the liquid, and the latent heat of evaporation is L V . Additionally, the gas mixture heat capacity is calculated using a mass averaging; C p,G ⫽⌺ ␣ Y ␣ C p, ␣ 共evaluated at particle locations兲. The gas-phase Prandtl and Schmidt numbers are Pr G ⫽ C p,G / and Sc G ⫽ /( ⌫), respectively. The evaporation rate is driven by the mass transfer number; B M ⫽(Y s f ⫺Y V )/(1⫺Y s f ) 共subscript sf denotes droplet surface conditions兲. The semiempirical Ranz–Marshall correlations are used for the Nusselt 共Nu兲 and Sherwood 共Sh兲 numbers, whereas f 1 is an empirical correction to Stokes drag accounting for finite droplet Reynolds numbers and evaporation.50 The function f 2 ⫽  /(e  ⫺1) is an analytical evaporative heat transfer correction, where the nondimensional evaporation parameter  ⫽⫺1.5 Pr G p ṁ p /m p is constant for droplets obeying the ‘‘D 2 law.’’ The vapor surface mass fraction is calculated directly from the surface molar fraction ( s f ) Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp 3306 Phys. Fluids, Vol. 13, No. 11, November 2001 Richard S. Miller which is obtained by equating the vapor and liquid fugacities at the surface 共i.e., s f P⫽ P sat兲, where the saturation pressure ( P sat) is provided by the Clausius–Clapeyron relation, yielding Ysf⫽ sf , s f ⫹ 共 1⫺ s f 兲 W G /W V sf⫽ LV 1 P atm 1 ⫺ exp P R V T B,L T p 再 冉 共16兲 冊冎 共17兲 , where P atm is atmospheric pressure, T B,L is the saturation temperature at P atm 共i.e., the normal boiling temperature兲. Finally, the latent heat is a linear function of temperature for calorically perfect species: L V ⫽h V0 ⫺(C L ⫺C p,V )T p . Based on the above formulation, the phase coupling terms are expressed as S I ⫽⫺ 兺␣ ⌬x 1 ⌬x 2 ⌬x 3⬘ 再 兺再 冋 S II,i ⫽⫺ S III ⫽⫺ 再 兺␣ ␣ w␣ 冎 共18兲 关 ṁ p 兴 ␣ , w␣ ⌬x 1 ⌬x 2 ⌬x 3⬘ 冎 关 F i ⫹ṁ p v i 兴 ␣ , 共19兲 冎册 冎 共20兲 w␣ ⌬x 1 ⌬x 2 ⌬x 3⬘ ⫻ v i F i ⫹Q⫹ṁ p 再 v iv i ⫹h V,s f 2 , ␣ where the summations are over local individual droplet contributions, h V,s f ⫽C p,V T p ⫹h V0 is the evaporated vapor enthalpy at the droplet surface, and the single droplet evaporation rate (ṁ p ), drag force (F i ) and heat transfer rate 共Q兲 are specified by Eqs. 共11兲–共13兲. The local summations are over all droplets residing within a local numerical discretization element 共⌬x 1 ⌬x 2 ; for 2D兲 and employ a geometrical weighting factor, w ␣ , used to distribute the individual droplet contributions to the four nearest neighbor surrounding grid points 共i.e., corners of the element ⌬x 2 兲. Note that the source terms are defined on a per unit volume basis since the particles are modeled as spherical entities, not as cross sections of infinitely long cylinders. In this case, the differential volume element in the above terms is ⌬x 1 ⌬x 2 ⌬x 3⬘ , where the characteristic length in the hypothetical x 3 direction (⌬x 3⬘ ) is defined below. These source terms are then minimally ‘‘smoothed’’ using a conservative operator in order to retain numerical stability of the Eulerian gas-phase equations.45 Potential limitations of the modeling of the source terms has been addressed previously in Ref. 45. A. Temporally developing mixing layer The flow geometry under consideration is the twodimensional 共2D兲 temporally developing mixing layer shown in Fig. 1. The streamwise (x 1 ) and cross stream (x 2 ) domain lengths are L 1 and L 2 , respectively. Stream 1 (x 2 ⬎0) is composed of pure oxidizer, whereas Stream 2 (x 2 ⬍0) is pure fuel. For two-phase simulations, both streams are uniformly laden with particles or droplets. Miller and Bellan45 found that the mass loading ratio 共ratio of liquid mass to FIG. 1. Schematic of the two-phase temporally developing mixing layer 共boxed region, above兲 including the initial flow profiles. The initial vorticity thickness ( ␦ ,0) is not shown to scale. gaseous mass in each stream兲 is the dominant parameter governing the flow modification in the droplet laden mixing layer. Therefore, we choose to fix the mass loading ratio to be the same in each stream; since the gas stream densities are not equal, the particle or droplet number density is not equal in the two streams. Boundary conditions are periodic in the streamwise direction, and nonreflecting outflow for the cross stream boundaries. The initial vorticity thickness is ␦ ,0 , where ␦ (t)⫽⌬U 0 / 具 u 1 / x 2 典 max ; the brackets 具 典 indicate averaging over the homogeneous x 1 direction and the mean velocity difference across the layer, ⌬U 0 ⫽U 1 ⫺U 2 , is calculated from a specified value of the convective Mach number (M c ). 51 The base flow mean velocity, mass fraction, temperature and number density are specified based on an error function profile; erf(1/2x 2 / ␦ ,0). 52 Initial mean profiles are illustrated in Fig. 1. The temperature is chosen to be constant and ‘‘low’’ in each stream at T⫽T 0 ; however, a temperature ‘‘spike’’ is added at the centerline to ignite the flame. The peak spiked temperature is denoted T F,0 . The fuel and oxidizer species are kept completely segregated in the initial profiles through the addition of a thin layer of product species. This prevents an overly fast initial reaction, as the flame can only begin to burn once diffusional mixing of reactants has commenced. In all cases, the evaporated vapor mass fraction, Y V , is initially zero everywhere. For ‘‘forced’’ simulations, perturbations are added to the velocity field in order to excite the growth of spanwise disturbances 共see below兲. Finally, we note that several recent investigations53–55 have noted that density stratification across the mixing layer can result in changes to both the ‘‘relaxed’’ laminar base flow profiles, as well as the most unstable forcing modes. Both effects were observed to significantly affect the development of the mixing layer. On the other hand, for the present reacting flow simulations the most appropriate base flow or forcing profiles are not easily determined due to the timedependent nature of the ensuing ignition and combustion, as well as to the presence of the dispersed phase. Therefore, the Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp Phys. Fluids, Vol. 13, No. 11, November 2001 approach used in this investigation is to fix the density stratification for all simulations, and to use the same error function and forcing used in constant density temporal mixing layer simulations.52 In order to minimize the effects of the initial conditions, the streamwise domain length is chosen to be eight times the smallest disturbance wavelength. As many as four primary vortex pairing events can occur during a simulation, thus allowing sufficient time for the mixing layer to develop in a more natural manner than at early times when the initial conditions are expected to have a stronger influence on the flow dynamics. B. Numerical approach The governing equations are solved numerically using fourth-order explicit Runge–Kutta temporal integration for all time derivatives. Spatial derivatives are disctretized using an explicit eighth order accurate central finite difference scheme in the streamwise direction (x 1 ), and by a fourthorder accurate tridiagonal compact finite difference in the cross stream direction (x 2 ). 56 The computational grid is uniformly spaced with constant ⌬x 1 ⫽⌬x 2 . A fourth order Lagrange interpolation procedure is used to obtain gas-phase variable values at droplet locations which, in general, do not reside at grid point locations. Transport equations are solved for every individual particle or droplet in the flow 共as opposed to stochastic entities representing many real particles兲. Nonreflecting outflow boundary conditions derived by Poinsot and Lele57 are employed for the x 2 domain boundaries. Droplets which pass through these boundaries and exit the domain are removed from the simulation. In the event that fluid enters the domain from the free stream, no new droplets are introduced. All simulations are conducted within a domain having L 1 ⫽8 1 ⫽58.32␦ ,0 and L 2 ⫽1.125L 1 , where 1 is the smallest forcing wavelength in the x 1 direction. Sinusoidal velocity perturbations are superimposed on the base flow profile having wavelengths L 1 , L 1 /2, L 1 /4, and L 1 /8. For convenience, the disturbances are generated as a spanwise vorticity distribution of the form 3 4 ⫽⫺⌺ m⫽1 f (x 2 ) 兩 A m sin(2x1 /(2m1))兩, where the cross stream weighting is given by f (x 2 )⫽exp(⫺x22/␦2 ,0), and the relative amplitudes of the harmonics are A 1 ⫽1, A 2 ⫽0.5, and A 3 ⫽A 4 ⫽0.35. The corresponding velocity disturbance distribution is then extracted via the solution of the appropriate Poisson equation. The nondimensional forcing * ) is characterized by the spanwise circulation amplitude (F 2D of the disturbance relative to the base flow circulation ( 1 ⌬U 0 ). The imposed disturbances instigate the development of eight initial vortices together with four pairing events prior to the nonphysical intervention of the domain * ⫽0) cases are boundaries. Both forced and nonforced (F 2D considered. The choice of most optimal forcing is particularly complicated for compressible, two-phase, stratified, and reacting mixing layers. A discussion of this subject pertaining to stratified layers may be found in Ref. 49. The approach chosen for this study is to fix the forcing mode and amplitude, and to simulate the mixing layer to relatively long times. Although long time influences of the initial forcing Effects of nonreacting solid particle 3307 TABLE I. Species property values used in the simulations. The subscripts denote the solid 共S兲 and liquid 共L兲 phases, and O, F, P, and V denote the oxidizer, fuel, product, and evaporated vapor species, respectively. All species have equal thermal conductivities and mass diffusivities with Pr⫽Sc⫽0.697. Property WO WF WP WV C p,O C p,F C p, P C p,V C S ,C L S ,L T B,L h 0V Value 28.97 kg共kg•mole兲⫺1 86.178 kg 共kg•mole兲⫺1 57.574 kg共kg•mole兲⫺1 142.0 kg共kg•mole兲⫺1 1043.8 J.kg⫺1 K⫺1 2251.5 J.kg⫺1 K⫺1 1947.7 J.kg⫺1 K⫺1 2394.5 J.kg⫺1 K⫺1 2520.5 J.kg⫺1 K⫺1 642.0 kg m⫺3 447.7 K 3.36⫻105 J.K⫺1 may exist, the present results are not used to investigate the asymptotic state of the two-phase mixing layer. The code is parallelized in both directions using the Message Passing Interface 共MPI兲 communication routines. Simulations were conducted on three different computers; a Silicon Graphics/Cray Origin2000, Hewlett Packard V2500, and a SUN HPC 6000, using from 8 to 48 CPUs. Computing times varied accordingly; e.g., the most intensive simulation 共EVAP-HIGHe, see below兲 was conducted using 48 CPUs of the V2500 and required approximately 6.9 seconds per iteration. The simulation time step (⌬t) is calculated at each temporal iteration from a specified value of the Courant number, C 共based on the convective plus acoustic velocities兲. Both single-phase and solid particle laden simulations have C⫽0.5, whereas evaporating droplet simulations employ C ⫽0.25 in order to ensure that the transient vaporization is well resolved temporally. C. Properties and nondimensional parameters Table I presents the properties of each of the species under consideration; fuel, oxidizer, product, evaporated vapor, as well as the solid or liquid species comprising the dispersed phase. The liquid and solid species have identical properties; the only distinction between these species is that the evaporation rate is nulled for solid particles 共i.e., ṁ p ⫽0兲. The species properties used in the investigation were chosen to model a typical 共though simplified兲 air– hydrocarbon diffusion flame. The oxidizer species has properties corresponding to those of air, the fuel species corresponds to hexane, the liquid 共or solid兲 and vapor species have properties of decane, and the product species properties are determined from mass and energy balances. All species are assumed to be calorically perfect. Several nondimensional parameters are introduced in order to simplify the characterization of the mixing layer and combustion. The flow Reynolds number is Re0 ⫽⌬U0␦,0 / . The particles or droplets are initially of uniform size specified by the initial Stokes number St0 ⫽ p,0⌬U 0 / ␦ ,0 . As mentioned previously, the initial mass loading ratio is prescribed to be the same for each stream. Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp 3308 Phys. Fluids, Vol. 13, No. 11, November 2001 Richard S. Miller TABLE II. Initialization parameters 共SP⫽single phase, SLD⫽solid particle, EVAP⫽evaporating droplet, LAM⫽laminar, LOW⫽low Ze, HIGH⫽high Ze兲. All cases have M c ⫽0.35, Re0⫽450, T 0 ⫽300 K, T F,0 ⫽565 K, Da(T F )⫽10, Ce⫽3.0, r⫽1, and P 0 ⫽1 atm. Two phase simulations are initialized with zero slip velocity ( v i ⫽u i ) and zero slip temperature 关T p ⫽T; except for EVAP cases in which T p ⫽min(T,TB,L)]. Case SP-LOW-LAM SP-HIGH-LAM SP-LOW SP-HIGH SLD-HIGH-LAM SLD-LOWa SLD-LOWb SLD-HIGHa SLD-HIGHb SLD-HIGHc SLD-HIGHd EVAP-HIGH-LAM EVAP-HIGHa EVAP-HIGHb EVAP-HIGHc EVAP-HIGHd N 1 ⫻N 2 Np M L0 St 0 Ze * F 2D ⌬x/D 0 p,0 /⌬t 10⫻576 10⫻576 512⫻576 512⫻576 10⫻576 512⫻576 512⫻576 512⫻576 512⫻576 512⫻576 512⫻576 10⫻576 512⫻576 512⫻576 512⫻576 512⫻576 0 0 0 0 3542 1.81⫻105 2.88⫻105 1.81⫻105 2.88⫻105 1.45⫻106 3.02⫻104 3542 1.81⫻105 2.88⫻105 1.45⫻106 3.02⫻104 ¯ ¯ ¯ ¯ 0.25 0.25 0.5 0.25 0.5 0.25 0.25 0.25 0.25 0.5 0.25 0.25 ¯ ¯ ¯ ¯ 2.0 2.0 2.0 2.0 2.0 0.25 12.0 2.0 2.0 2.0 0.25 12.0 9.21 15.35 9.21 15.35 15.35 9.21 9.21 15.35 15.35 15.35 15.35 15.35 15.35 15.35 15.35 15.35 0 0 0.1 0.1 0 0.1 0.1 0.1 0.1 0.1 0.1 0 0.1 0.1 0.1 0.1 ¯ ¯ ¯ ¯ 6.67 6.67 6.67 6.67 6.67 18.9 2.72 6.67 6.67 6.67 18.9 2.72 ¯ ¯ ¯ ¯ 90.5 90.5 90.5 90.5 90.5 11.3 543 181 181 181 22.6 1086 For constant density streams with uniform mass particles, the mass loading is defined as M L 0⫽ N p, ␣ m p ⬘ 共 ␣ 兲 /2 L 1 L 2 L 3, , 共21兲 for either individual stream 共N p, ␣ is the number of particles in stream ␣兲. In the above, a reference length in the x 3 direction must be specified to define a meaningful mass loading for a 2D flow 共the particles are spherical, not cross sections of infinitely long cylinders兲. Assuming that the particles are initially distributed at the ‘‘nodes’’of a hypothetical Cartesian grid in the 2D plane, then the corresponding isotropic 3D distribution would have the same particle separation in ⬘ ␣ ) is dethe third direction. The spanwise length scale L 3,( fined to be this hypothetical separation distance, which is determined by equating the assumed Cartesian particle grid area to the total area of either stream ⬘ 2共 ␣ 兲 ⫽ N p, ␣ L 3, L 1L 2 . 2 共22兲 For simplicity, we define the characteristic x 3 length scale for the phase coupling source terms, Eqs. 共18兲–共20兲, to be ⌬x 3⬘ ⬘ ⫹L 3,(2) ⬘ )/2 关this does not affect the physical ‘‘cor⫽(L 3,(1) rectness’’ of the approach, but does mean that the effective mass loading ratios are slightly altered from the definition in Eq. 共21兲兴. It is noted that the above defined mass loading parameter should not be quantitatively correlated to that of 3D flows. It does, however, provide a direct measure of the effect of increasing the total particle mass acting on the Eulerian gas-phase flow. The chemical reaction is also specified by nondimensional parameters. The Zeldovich number parameterizes the reaction activation energy: Ze⫽E 0 /(R 0 T 0 ), and the heat release parameter Ce specifies the heat of reaction ⌬H 0 :Ce ⫽⫺⌬H 0 /(C p,F T 0 ). With the above notation, the adiabatic flame temperature (T F ) is approximated by T F ⫽T 0 (1 ⫹Ce), where T 0 is the initial temperature of both streams in the mixing layer. Furthermore, the Damkohler number 共Da兲 is defined as the ratio of the characteristic time scale of the flow to the characteristic reaction time scale: Da(T) ⫽K R (T) ␦ ,0 /⌬U 0 . The Damkohler number is defined as a function of the temperature in order to simplify the following analyses. III. RESULTS The simulations conducted for this study are summarized in Table II which provides the designation, grid resolution, total number of particles (N p ), mass loading ratio, particle Stokes number, Zeldovich number, forcing amplitude, and relative initial particle diameters and particle time constants. The following conventions are used in naming each simulations: SP⫽single phase, SLD⫽solid particle laden, EVAP⫽evaporating droplet laden, LAM⫽laminar flow, LOW⫽low-activation energy, and HIGH⫽highactivation energy. The unforced ‘‘LAM’’ simulations are conducted with L 1 ⫽10⌬x 1 and correspond to quasi-onedimensional 共1D兲 laminar flames. The final two columns in Table II show that the particle sizes are all smaller than the grid spacing, and that the particle temporal evolutions are well resolved. In order to simplify the analysis, the remaining initialization parameters are fixed for all simulations. The flow Reynolds number is Re0⫽450, the convective Mach number is M c ⫽0.35, the pressure is initially constant and standard ( P 0 ⫽ P atm), the uniform temperature in both streams is T 0 ⫽300 K, and the spiked ignition temperature is chosen to be T F,0⫽565 K. For the reaction, the Damkohler number evaluated at the adiabatic flame temperature is in all cases Da(T F )⫽10, the reaction coefficient is r⫽1, and the heat release parameter is Ce⫽3 (T F ⫽1200 K). Finally, the particles or droplets are initially randomly dispersed throughout either stream, with specified size, zero slip velocity ( v i ⫽u i ), and zero slip temperature (T p ⫽T). However, droplet temperatures cannot exceed the boiling temperature; there- Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp Phys. Fluids, Vol. 13, No. 11, November 2001 FIG. 2. Gas-phase temperature dependence of 共a兲 the reaction Damkohler number, and 共b兲 the time to complete evaporation for a single droplet in an infinite quiescent environment. fore, for droplets T p ⫽min(T,TB,L). All simulations are conducted until a nondimensional time t * ⫽t⌬U 0 / ␦ ,0⫽160 at which point the final pairing has been nearly completed. A. Effect of combustion and droplet parameters A justification for the choices of several of the above parameters is provided in Fig. 2. Figure 2共a兲 depicts the reaction rate Damkohler number as a function of temperature for the two values of the Zeldovich number considered in this study. In both cases, Da(T F )⫽10 is specified; therefore, the reaction rate constant is also a function of Ze 共A 0 ⫽5.17⫻106 s⫺1 for Ze⫽9.21, and A 0 ⫽2.40⫻107 s⫺1 for Ze⫽15.35兲. In this manner, only the relative rate of the reaction at lower temperatures is affected by changes to the Zeldovich number. The Damkohler numbers at the initial spiked ignition temperature are Da(T F,0)⫽0.75, and Da(T F,0)⫽0.13 for Ze⫽9.21 and Ze⫽15.35, respectively. Therefore, LOW cases burn relatively vigorously early in the simulation, whereas HIGH cases will require a substantially longer time to commence vigorous combustion. Furthermore, the reaction rate is essentially negligible for both Zeldovich numbers at the ambient temperature T⫽300 K; thereby allowing for varying degrees of local flame extinction in the mixing layer simulations that follow. The droplet properties and time scales were chosen to mimic the behavior of typical species used in fire suppres- Effects of nonreacting solid particle 3309 FIG. 3. Mean cross stream profiles for laminar flame simulations at time t * ⫽160; 共a兲 temperature, and 共b兲 droplet number density. sion 共e.g., water; decane properties are used instead of actual water properties in order to simplify comparisons with a concurrent study of reacting droplets兲 in that evaporation occurs rapidly in flame regions, but very slowly under ambient conditions. The relative evaporation time (t E ) of single droplets is examined in Fig. 2共b兲. Droplets cannot be simulated to arbitrarily small sizes due to an eventual inadequate temporal resolution of the particle evolutions 共i.e., p will become too small relative to ⌬t兲. For the present investigation, droplets are assumed to be completely evaporated once their individual Stokes number reaches St⫽0.075, which was found to have an adequate temporal resolution. With this final Stokes number, all but 5.0⫻10⫺4 , 7.3⫻10⫺3 , and 1.6⫻10⫺1 of the initial mass is allowed to evaporate for droplets with St0⫽12, St0⫽2.0, and St0⫽0.25, respectively. The results in Fig. 2共b兲 correspond to the total time required to reach this size limit for single isolated droplets in infinite quiescent media of either pure fuel or oxidizer as a function of the ambient temperature. These curves are therefore solutions of only Eqs. 共12兲 and 共13兲 with u i ⫽ v i ⫽0. Each curve represents the results of 40 such single droplet simulations. Although only an approximation to what occurs in the more complicated 2D simulations, Fig. 2共b兲 provides an estimate of the droplet life time relative to the flow time scale for the range of gas temperatures that may be encountered by droplets in the flame. In all cases, the droplet life time is smaller than the total simulation time (t * ⫽160) for temperatures above T F,0 . In contrast, droplets in the free streams away from the combus- Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp 3310 Phys. Fluids, Vol. 13, No. 11, November 2001 Richard S. Miller FIG. 4. Temporal evolution of flow statistics for laminar flame simulations; 共a兲 momentum thickness, 共b兲 product thickness, 共c兲 maximum temperature, and 共d兲 maximum reaction rate. tion region encounter cooler fluid (T 0 ⫽300 K), and act essentially as nonevaporating particles at these temperatures. Note also that Fig. 2共b兲 shows that droplets evaporate faster in pure oxidizer than in pure fuel at high temperatures. This can result in a bias in the vaporization rates for droplets depending on which stream of the mixing layer they reside in. B. Laminar flame evolution Quasi-1D laminar flames are examined as an aid to understanding the 2D mixing layer flows. Simulations SPLOW-LAM, SP-HIGH-LAM, SLD-HIGH-LAM, and EVAPHIGH-LAM 共Table II兲 elucidate the effects of the Zeldovich number 共for SP flames兲, as well as either solid particle or liquid droplet loading (M L 0 ⫽0.25) on the laminar flame. All of these simulations are conducted with resolutions of 10⫻512 grid points. The complete set of 2D governing equations is solved for these flows; however, no forcing is used and the single-phase 共SP兲 cases are perfectly 1D solutions. On the other hand, minor x 1 variations of flow variables exist for the two-phase cases due to the phase coupling source terms in the Eulerian gas phase equations. Nevertheless, these spatial variations are negligible to the overall flow and the results are essentially 1D. Each of these flows are initialized in the same manner as the 2D mixing layers and are also allowed to evolve until a nondimensional time t * ⫽160. The final time mean temperature and mean number density profiles across the laminar flames are shown in Fig. 3. Note that the cross stream coordinate x 2 has been normalized by the instantaneous momentum thickness defined by46 ␦ m⫽ 1 关 具 u 1典 2⫺ 具 u 1典 1 兴 2 冕 ⫹L 2 /2 ⫺L 2 /2 兵 关 具 u 1 典 2 ⫺ 具 u 1 典 共 x 2 兲兴 ⫻ 关 具 u 1 典 共 x 2 兲 ⫺ 具 u 1 典 1 兴 其 dx 2 , 共23兲 where 具 u 1 典 ␣ is the mean momentum evaluated at the x 2 boundary of stream ␣. The particle number density 共n兲 is calculated in a similar manner to the phase coupling terms, n⫽ 兺 ␣ 兵 w ␣ 其 , and is interpreted as the number of particles per grid point. For evaporating droplets, any individual droplet which has reached the minimum size St⫽0.075 is removed from the simulation, and has no contribution to the summation. The results of Fig. 3 indicate that the long time laminar flames, which are nonsymmetric across the layer due to density and property stratification, are essentially independent of the Zeldovich number and the dispersed phase loading. At the observed laminar flame temperatures 关Fig. 3共a兲兴, the reaction rates are much faster than the characteristic flow time scale 关see Fig. 2共a兲兴. This means that all of the flames, regardless of the Zeldovich number, are diffusion controlled; i.e., limited by diffusion as opposed to being limited by Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp Phys. Fluids, Vol. 13, No. 11, November 2001 Effects of nonreacting solid particle 3311 FIG. 5. Temperature contours for simulation SP-LOW; 共a兲 t * ⫽40, 共b兲 t * ⫽80, 共c兲 t * ⫽120, and 共d兲 t * ⫽160. chemical kinetics. This accounts for the similarity in the SPLOW-LAM and the SP-HIGH-LAM flow profiles. On the other hand, the observed similarity in the particle and droplet laden flame profiles is attributed to the fact that the solid particles achieve an equilibrium temperature with the surrounding flame, whereas droplets are completely evaporated within the flame at long times. Once the liquid within the flame has evaporated, the gaseous vapor can diffuse away from the combustion zone, and the reaction continues in a similar manner to single-phase cases. This is clarified in Fig. 3共b兲 which shows a zero droplet number density in the approximate region ⫺2⭐x 2 / ␦ m ⭐3 due to the completion of vaporization. As a final note, we observe that the laminar flame temperatures 关Fig. 3共a兲兴 are substantially lower than the adiabatic flame temperature (T F ⫽1200 K) for all 1D simulations due to diffusional losses. Despite the observed long time similarities of the laminar flames, substantial differences in their evolutions are apparent at earlier times. Figure 4 presents the temporal development of the momentum thickness, product thickness ␦ P⫽ 冕 冕 ⫹L 2 /2 ⫺L 2 /2 L1 0 Y P dx 1 dx 2 , 共24兲 maximum temperature, and maximum reaction rate for the laminar flames. These results reveal that the diffusion limited asymptotic flame behavior occurs only after t * ⬇125. Prior to this time the flames develop quite differently, particularly as a function of the Zeldovich number. For LOW cases (Ze⫽9.21), the reaction rate is relatively fast immediately, as noted by the early and rapid rise of the maximum temperature and reaction rate 关Figs. 4共c兲 and 4共d兲兴. This is accompanied by an early development of linear momentum and product thickness growth rates for times t * ⬎50 关Figs. 4共a兲 and 4共b兲兴. On the other hand, for the HIGH cases (Ze⫽15.35), the reaction rate is relatively slow at the initial temperature (T F,0⫽565 K) and the maximum temperature decreases substantially until a sufficient mass of reactants have diffused into the flame region to ignite the flame at time t * ⬇50. Solid particles have a finite thermal inertia and, therefore, have the potential to delay ignition due to the energy lost to heating the particles. However, any changes depicted in the results of Fig. 4 due to solid particles are too small to draw conclusions in this regard. The characteristic time scale for particle heating is proportional to the particle Stokes number, in this case St⫽2.0, in units of the eddy turnover time 关see Eq. 共12兲兴. Evaporating droplets cause substantial additional losses of thermal energy from the flame due to latent heat effects 关note that the droplets cannot exceed the liquid boiling temperature, T B,L ⫽447.7 K, which results in large convective heat flux from the higher temperature flame; see Eq. 共12兲兴. However, the droplets reach complete evaporation relatively quickly at these temperatures 关see Fig. 2共b兲兴, and are rapidly destroyed within the reaction zone. Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp 3312 Phys. Fluids, Vol. 13, No. 11, November 2001 Richard S. Miller FIG. 6. Temperature contours for simulation SP-HIGH; 共a兲 t * ⫽40, 共b兲 t * ⫽80, 共c兲 t * ⫽120, and 共d兲 t * ⫽160. C. Mixing layer flow visualization As will be shown below, the qualitative and quantitative behaviors of the 2D forced mixing layer simulations can be markedly more complex than observed in the much simpler laminar flame simulations. The mixing layer analysis is begun in this section with a qualitative exposition of the flame development aided by flow visualizations. The typical evolution of a single-phase low activation energy flame is illustrated in Fig. 5 which depicts contours of the gas temperature for case SP-LOW at times t * ⫽40, 80, 120, and 160. Hereinafter, all gray scale contour plots indicate maximum domain values by the maximum labeled contour. At t * ⫽40 the flame is nonsymmetric about the centerline due to the density and property stratification across the mixing layer. At this time, the first pairings of the eight initial vortices have commenced. By time t * ⫽80 another set of pairings is occurring leaving two distinct spanwise vortices. The braid regions are highly strained, thereby greatly enhancing both thermal and mass diffusion of the hot product species into the surrounding cooler free steam fluid 共temperature maxima are within the vortices兲. However, the reaction is sufficiently strong at this Zeldovich number to prevent local flame extinction in the braids. In contrast, the evolution of the temperature contours is substantially altered at the higher Zeldovich number as observed in Fig. 6 for case SP-HIGH. In this case, the reaction rate is insufficient at lower temperatures to overcome the diffusional losses within the braid regions of the mixing layer. Clear evidence of local flame extinction is seen FIG. 7. Temperature contours for simulation SLD-LOWb; 共a兲 t * ⫽80, and 共b兲 t * ⫽160. Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp Phys. Fluids, Vol. 13, No. 11, November 2001 Effects of nonreacting solid particle 3313 FIG. 8. Contours for simulation SLD-HIGHb; 共a兲 temperature at time t * ⫽80, 共b兲 temperature at time t * ⫽160, 共c兲 particle number density at time t * ⫽80, and 共d兲 particle number density at time t * ⫽160. in the contours for times t * ⫽40 and 80. However, at later times the burning vortices eventually merge and re-ignite the extinguished braid fluid. The ‘‘thicker’’ nature of the final time flame in Fig. 6共d兲 as compared to Fig. 5共d兲 will be discussed below 共note that only a portion of the x 2 length of the domain is shown in these contour plots; the actual domain size is ⫺33⭐x 2 / ␦ ,0⭐⫹33兲. Analogous temperature contours for solid particle laden flames are presented in Figs. 7 and 8, corresponding to simulations SLD-LOWb and SLD-HIGHb 共with M L 0 ⫽0.5 and St0⫽2.0兲. The contours are depicted for times t * ⫽80 and 160. Contours of the particle number density 共n兲 are also provided for simulation SLD-HIGHb 关Figs. 8共c兲 and 8共d兲兴. As noted in previous investigations of two-way coupled solid particle17 and liquid droplet45,46 laden mixing layers, the presence of the dispersed phase suppresses the development of vortical structures through an effective dissipation of kinetic energy. A comparison of the vortices depicted in Figs. 7 and 8 with the single-phase results of Figs. 5 and 6 shows a qualitative consistency with this effect. Note that the more ‘‘jagged’’ nature of the contours in Figs. 7 and 8 is a physical effect associated with the spatial distribution of particles and the resulting coupling source terms in the Eulerian governing equations.45 The development of the two-phase flames is also qualitatively similar to the single-phase trends discussed above. However, for these flows the solid particles tend to preferentially concentrate in the braid regions 关Figs. 8共c兲 and 8共d兲兴. Particles entrained into the mixing layer from the free stream are therefore drawn into the braids with lower temperature than the local gas. The primary effect of this process is a further reduction to the flame temperature in the braids, and therefore, an additional tendency towards local extinction. This is evident for both Zeldovich numbers under consideration. For the lower activation energy 共Fig. 7兲, no complete extinction in the braids is observed; however, the braid flame remains substantially thinner and less developed than the corresponding single-phase flow 共Fig. 5兲. This effect is magnified for larger activation energy 共Fig. 8兲. In this case, local flame extinction in the braids is clearly evident at time t * ⫽80. Even at the final time, t * ⫽160 关Figs. 8共b兲 and 8共d兲兴, the presence of particles results in a very thin flame structure within the braids that was not observed for the single-phase flame. The extent of this local flame suppression and/or extinction will be quantified below. The last flow configuration for consideration in this section is the liquid droplet laden mixing layer. Contours of temperature, droplet number density and evaporated vapor mass fraction are provided in Fig. 9 for simulation EVAPHIGHb 共with M L 0 ⫽0.5 and St0⫽2.0兲 at times t * ⫽120 and 160. Differences in the behavior of the evaporating liquid laden flow are clearly evident; particularly in regards to extinction. At time t * ⫽120 only two distinct ‘‘flame balls’’ are Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp 3314 Phys. Fluids, Vol. 13, No. 11, November 2001 Richard S. Miller FIG. 9. Contours for simulation EVAP-HIGHb; 共a兲 temperature at time t * ⫽120, 共b兲 temperature at time t * ⫽160, 共c兲 droplet number density at time t * ⫽120, 共d兲 droplet number density at time t * ⫽160, 共e兲 evaporated vapor mass fraction at time t * ⫽120, and 共f兲 evaporated vapor mass fraction at time t * ⫽160. found, with a large extent of cold fluid between them 关Fig. 9共a兲兴. At this Zeldovich number (Ze⫽15.35), the reaction rate is too slow to overcome the cooling of the gas phase caused by droplets preferentially concentrating into the braids of the mixing layer 关Figs. 9共c兲 and 9共d兲兴. Furthermore, any droplets which cannot escape the burning vortices are completely evaporated in a relatively short time 关see Fig. 2共b兲兴. This results in the distinctly droplet devoid regions in the number density contours. Once the droplets in these vortex structures are evaporated, the flame development within these regions continues unimpeded by droplets. In addition, the earlier presence of droplets has already allowed a significant extent of diffusional mixing of reactants; therefore, the later developing flames burn rather robustly in the vortices. For example, the peak temperature at time t * ⫽160 is T MAX⫽1243 K 关Fig. 9共b兲兴 which is higher than the peak laminar flame temperature of T LAMINAR⬇900 K 关Fig. 3共a兲兴. Despite local extinction and preferential concentration in the braids, vortex pairing provides another route by which droplets are entrained into the burning vortices. Consider the final time 关Fig. 9共b兲兴, during which a thin but well defined layer of extinguished fluid resides between the two primary vortices. As with earlier vortex pairings 共not shown兲, the two distinct combustion zones will eventually merge into a single vortex of burning reactants, rapidly vaporizing any droplets which become engulfed within the combustion zone. An interesting feature of the evaporated vapor mass fraction contours in Figs. 9共e兲 and 9共f兲 is that the vapor resides in the vortices, and not in the braid regions where preferential Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp Phys. Fluids, Vol. 13, No. 11, November 2001 FIG. 10. Temporal evolution of 共a兲 the maximum temperature, and 共b兲 the extinction factor. concentration accumulates droplets. This is in contrast to the droplet laden 共nonreacting兲 mixing layer studied by Miller and Bellan,45,46 who observed a much more uniform distribution of vapor, with maximal concentrations within the braids. This is explained as follows. At very early times droplets residing within the initial preheat zone begin to evaporate, resulting in a thin layer of vapor mass fraction across the entire streamwise length of the layer. As the spanwise vortex rollup commences, the droplets accumulate in the braids and extinguish the flame due primarily to latent heat effects. As the braid fluid is cooled to near ambient temperatures, the vapor concentration gradient is enhanced by the straining of the braid structures, and the remaining vapor diffuses into the surrounding fluid. In contrast, droplets residing within the newly formed vortices rapidly evaporate, and more easily avoid being preferentially concentrated to the braids as they become smaller and more closely follow fluid motions. In this manner, the vapor frac- Effects of nonreacting solid particle 3315 FIG. 11. Temporal evolution of 共a兲 the momentum thickness, and 共b兲 the product thickness. tion inside the vortices is increased early in the simulation. The vapor concentration gradients within the vortices are also smaller than those found in the braids; resulting in a reduced rate of diffusional mixing of the vapor into the surrounding fluid. D. Quantitative mixing layer analysis In the previous section, qualitative flow visualizations were used to illustrate several differences between the behavior of the ‘‘turbulent’’ 2D mixing layer and laminar quasi-1D flames. In particular, the phenomena of local flame extinction and its relation to the preferential concentration of particles or droplets in high-strain regions was discussed. The analysis is now turned towards a quantitative investigation of the effects of particles or droplets on the reacting mixing layer. In order to aid in measuring the extent of local extinction during the following analysis, we introduce a new parameter, ␣ F ( ), denoted the ‘‘extinction factor:’’ ⫹L /2 L ␣ F共 兲 ⫽ 兰 ⫺L 2 /2兰 0 1 H 共 ⬘ 共 x兲 ⫺ 兲 H 共 Y O Y F 共 x兲 ⫺5.0⫻10⫺5 兲 dx 1 dx 2 2 ⫹L /2 L 兰 ⫺L 2 /2兰 0 1 H 共 Y O Y F 共 x兲 ⫺5.0⫻10⫺5 兲 dx 1 dx 2 , 共25兲 2 Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp 3316 Phys. Fluids, Vol. 13, No. 11, November 2001 FIG. 12. Temporal evolution of 共a兲 the momentum thickness, and 共b兲 the product thickness. where H is the Heaviside function, ⫽(T⫺T 0 )/(T F ⫺T 0 ) is the relative gas temperature between the ambient and the adiabatic flame temperature, and the product Y O Y F indicates the potential for chemical reactions to occur 关see Eqs. 共7兲– 共9兲兴. Therefore, ␣ F ( ) is the relative fraction of the domain with ‘‘premixed’’ reactants having a nondimensional temperature greater than . The extinction factor is bounded by 0⭐ ␣ F ( )⭐1, where ␣ F ⫽0 indicates that no portion of the domain with mixed reactants is hotter than , and ␣ F ⫽1 indicates that the entire portion of the domain with mixed reactants is hotter than . By choosing an appropriate value for within the range of essentially nonreacting temperatures 关such as ⫽0.1 (T⫽390 K); see Fig. 2共a兲兴, the fraction of extinguished reactants is approximately determined. In what follows, the extinction factor, as well as other statistical parameters, is used to quantify the effects of the Zeldovich number, mass loading ratio, and the initial Stokes number on the development of the reacting mixing layer. 1. Effect of the Zeldovich number The effects of the Zeldovich number on the reacting mixing layer are examined for three flow configurations: 共1兲 Single-phase flow 共cases SP-LOW and SP-HIGH兲, 共2兲 solid particle loading with M L 0 ⫽0.25 共cases SLD-LOWa and SLD-HIGHa兲, and 共3兲 solid particle loading with M L 0 ⫽0.5 共cases SLD-LOWb and SLD-HIGHb兲. Note that evaporating droplets are not included in this stage of the discussion in order to simplify the analysis. The temporal development of Richard S. Miller FIG. 13. Temporal evolution of 共a兲 the maximum temperature, and 共b兲 the extinction factor. the maximum temperature and the extinction factor ␣ F ( ⫽0.1) are presented in Fig. 10 for all three sets of the simulations. At early times, and relatively low temperatures, the peak temperature is observed to initially decrease for approximately the first 20 eddy turnover times for the HIGH Zeldovich number cases 共dashed curves兲. In contrast, the LOW Zeldovich number cases have a substantially larger reaction rate at the initial preheat temperature 关see Fig. 2共a兲兴, and the maximum temperature in these flows begins to increase much sooner. However, at long times the trends are reversed and the largest flame temperatures are found in HIGH Zeldovich number cases. This occurs because the initial delay in ignition for HIGH Ze mixing layers allows time for reactants to become relatively well mixed inside of the vortices. Once ignition does finally occur, the concentrations of mixed reactants are larger than can occur in the LOW cases, and a rapid and robust ignition at high-temperatures follows. Nevertheless, the extinction factors presented in Fig. 10共b兲 indicate that a significantly smaller portion of the mixed reactants are actually burning for HIGH Ze flows due to enhanced extinction in the braid regions. For the HIGH flows, only approximately 50% of the regions containing mixed reactants are sufficiently hot to burn robustly, even at the final simulation time. This same delay in ignition for larger Zeldovich number flows causes an enhancement in the long time ‘‘thickness’’ of the mixing layer which was observed previously in the flow visualizations 关see Figs. 5共d兲 and 6共d兲兴. Figure 11 quantifies Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp Phys. Fluids, Vol. 13, No. 11, November 2001 Effects of nonreacting solid particle 3317 FIG. 14. Particle and droplet number density contours at time t * ⫽160 for simulations; 共a兲 SLD-HIGHc, 共b兲 EVAP-HIGHc, 共c兲 SLD-HIGHd, and 共d兲 EVAP-HIGHd. this effect through the depiction of the momentum thickness and product thickness evolutions for the same simulations. For both single-phase and solid particle laden flows, the HIGH Ze cases exhibit a larger momentum thickness and product thickness at long times than the corresponding LOW cases. This feature can be attributed to the delay for ignition in the higher Zeldovich number flames. During early times, these flames behave more like nonreacting mixing layers in comparison to the lower Ze flows. McMurtry et al.9 conducted one of the first numerical studies of exothermic reacting mixing layers. Their results show that chemical heat release slows the development of large-scale structures and reduces the entrainment of reactants into the layer. Since the lower Zeldovich number flames are characterized by larger early heat release, these flames also exhibit more reduced growth and entrainment rates than the higher Ze flames. In contrast, ignition is delayed for high Zeldovich numbers, thereby allowing a faster early growth of the layer with significant mixing of reactants prior to ignition. This premixing of reactants allows for a more ‘‘explosive’’ growth of the product thickness once ignition finally occurs. A corresponding rapid increase in the momentum thickness also occurs corresponding to the delayed ignition due to thermal expansion effects.58 2. Effect of the mass loading ratio The results presented in Figs. 10 and 11 also elucidate the effect of the solid particle mass loading ratio for values M L 0 ⫽0, 0.25, and 0.5 at both Zeldovich numbers. In agreement with the previous nonreacting mixing layer results of Miller and Bellan,45,46 increasing liquid mass loading ratios result in reduced development of the momentum thickness of the pre-transitional layer 共note that Miller and Bellan46 observed a reverse in this trend for post-transitional flow; however, no similar conclusions can be drawn from the present 2D stratified and reacting flow results兲. Meiburg et al.40 observed a similar phenomena for the 2D solid particle laden mixing layer which they attributed to a reduction in the transport of vorticity from the braids to the vortex cores. Furthermore, the present results also show that both the maximum temperature and the extinction factor decrease with the addition of particles due to their finite thermal inertia and corresponding tendency to cool the surrounding fluid. Further studies of transitional mixing layers are required to determine the ultimate fate of the growth rate and peak temperature statistics in fully turbulent flows. The effects of the mass loading ratio are further investigated in Figs. 12 and 13 for both solid particles and evapo- Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp 3318 Phys. Fluids, Vol. 13, No. 11, November 2001 FIG. 15. Temporal evolution of 共a兲 the momentum thickness, and 共b兲 the product thickness. rating droplets. In both cases, mass loadings of M L 0 ⫽0, 0.25, and 0.5 are considered for a fixed Zeldovich number Ze⫽15.35 共HIGH兲. Evaporating droplets are also observed to reduce the momentum thickness and product thickness of the layer with increased liquid loading. In addition, the evaporating droplets substantially amplify the effects on the mixing layer in comparison to solid particles. For example, Fig. 13共a兲 reveals that the time required for the peak temperature to begin to grow 共ignition兲 is nearly twice as long for evaporating droplet laden flows due to their additional latent heat effects, as well as to the dilution of reactant concentrations by evaporated vapor. In addition, the extinction factor is approximately halved for the droplets 关Fig. 13共b兲兴. As noted previously, this is due to the droplets concentrating in the braid regions of the mixing layer and locally extinguishing the flame. 3. Effect of the Stokes number The last subject for investigation is the effect of the initial particle or droplet size, quantified in this case by the initial Stokes number, St0. To this point, only particles with ‘‘moderate’’ Stokes numbers St0⫽2.0 have been discussed; however, the simulations described in Table II also address ‘‘small’’ particles with St0⫽0.25, and ‘‘large’’ particles with St0⫽12.0. These are compared at fixed mass loading ratio (M L 0 ⫽0.25) and Zeldovich number 共Ze⫽15.35; HIGH兲. Contours of the final time (t * ⫽160) number density distri- Richard S. Miller FIG. 16. Temporal evolution of 共a兲 the maximum temperature, and 共b兲 the extinction factor. butions are presented for both the small and large particles and droplets in Fig. 14. As discussed in the introduction, it is well known that very small particles are able to closely follow fluid motions, and therefore, display relatively little preferential concentration. This is clearly evident in Fig. 14共a兲 which confirms that solid particles are indeed well distributed throughout both the vortex cores and the braid regions of the reacting mixing layer. On the other hand, the number density distribution for St0⫽0.25 evaporating droplets 关Fig. 14共b兲兴 is qualitatively very different. In this case, the droplets appear to exhibit very little preferential concentration in braid regions. However, the vortex regions are nearly perfectly devoid of droplets. Whereas the small droplets do indeed closely follow the fluid motions, the life time of small droplets in heated gas is very small compared to the characteristic flow time scales 关see Fig. 2共b兲兴. Therefore, any droplets following fluid motions into the flame regions are rapidly vaporized to completion and no longer contribute to the number density contours. In contrast to small particles, very large particles are also known to display reduced preferential concentration due to a reduced response to fluid motions. Figures 14共c兲 and 14共d兲 present the corresponding number density contours for cases SLD-HIGHd and EVAP-HIGHd which have St0⫽12. Although some reduction in the preferential concentration is observed in Fig. 14共c兲 for solid particles, the Stokes number is not large enough to show a dramatic decrease in comparison to St0⫽2.0 共simulation of larger St0 is not feasible at this Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp Phys. Fluids, Vol. 13, No. 11, November 2001 mass loading ratio since sufficiently large numbers of particles must be included to maintain relatively smooth phase coupling terms兲. Also, in contrast with the smaller evaporating droplets, some of the larger droplets are found inside of the combustion region in Fig. 14共d兲. Although these droplets do evaporate relatively quickly in the heated gas, their life times are much longer than for small droplets 关Fig. 2共b兲兴. Quantitative comparisons of the effect of the particle or droplet size are provided in Figs. 15 and 16. These figures show the momentum thickness, product thickness, maximum temperature, and extinction factor for both particles and droplets with St0⫽0.25, 2.0, and 12.0; all at fixed M L 0 ⫽0.25 and Ze⫽15.35. In agreement with previous nonreacting two-phase mixing layer results,45 changes to the initial Stokes number 共at fixed mass loading兲 do not significantly alter the growth rate of the layer as measured by the momentum thickness. However, the present results show that changes in the preferential concentration as a function of the Stokes number do result in a markedly increased product thickness for the largest particles and droplets 关Fig. 15共b兲兴. Since these particles do not closely follow fluid motions, fluid is constantly being swept past the droplets, and therefore, has less exposure time to be cooled by direct contact with particles or by liquid droplet latent heat effects. This allows for larger maximum temperatures 关Fig. 16共a兲兴 and a substantially reduced extent of local flame extinction 关Fig. 16共b兲兴 for large particles or droplets. For the liquid laden flames this effect is further augmented by the relatively longer life times of large droplets 关Fig. 2共b兲兴. IV. CONCLUSIONS Numerical simulations have been conducted of both solid particle and liquid droplet laden diffusion flames. The simulations were conducted for both quasi-1D laminar flames and for forced 2D mixing layers. A single-step irreversible and exothermic reaction of the form Fuel ⫹r Oxidizer→共1⫹r兲Products was considered. The particles 共or droplets兲 were assumed to be spherical, smaller than the gas phase length scales, and to obey a modified Stokes drag law. Droplet vaporization was assumed to be governed by the classical rapid mixing model in a form appropriate for calorically perfect species. Neither the particles or the droplets contribute to the reaction; droplets evaporate to a nonreacting vapor species different from either the fuel or oxidizer species. The primary parameters of interest to the study were the reaction Zeldovich number, the initial particle Stokes number and the dispersed phase mass loading ratio. The simulation results yielded both a qualitative and quantitative description of the two-phase reacting mixing layer. Laminar diffusion flames were observed to evolve to long time distributions which are essentially independent of the Zeldovich number, as well as the particle or droplet mass loading. Solid particles reach an equilibrium temperature with the flame at these times, whereas droplets are completely evaporated within the laminar flame region. In contrast, for 2D mixing layers the rollup of spanwise vortices creates high-strain braid regions in which local extinction of the reaction can occur for sufficiently large Zeldovich num- Effects of nonreacting solid particle 3319 bers. Local flame extinction can also be significantly enhanced by the presence of either solid particles or evaporating droplets. Solid particles are entrained into the mixing layer from the relatively cold free streams. Their finite thermal inertia, therefore, has the effect of further cooling the braid regions into which they are preferentially concentrated for moderate Stokes numbers. Evaporating droplets exhibit a more enhanced potential for local flame quenching due to the latent heat of vaporization, as well as through the diluting effects of the added nonreacting gaseous vapor species. In general, local flame extinction is maximized for large activation energies, large mass loading ratios, and by small particle or droplet sizes. These observations have direct consequences for practical flame suppressant delivery; suggesting that atomization of a large liquid mass to very small droplet sizes will optimize both local and global flame extinction through a more uniform delivery to both braids and vortices. ACKNOWLEDGMENTS This research was supported by the National Science Foundation through the Faculty Early Career Development Program; Grant No. CTS-9983762. Computational support was provided by the National Computational Science Alliance 共NCSA兲 under Grant No. CTS990040N and utilized the NCSA SGI/Cray Origin2000. Additional computational support was provided by the California Institute of Technology’s Center for Advanced Computing Research 共CACR兲 utilizing the Hewlett-Packard V2500, and by Clemson University’s Division of Computing and Information Technology utilizing a SUN HPC 6000. 1 C. T. Crowe, T. R. Troutt, and J. N. Chung, ‘‘Numerical models for twophase turbulent flows,’’ Annu. Rev. Fluid Mech. 28, 11 共1996兲. 2 G. L. Brown and A. Roshko, ‘‘On density effects and large scale structure in turbulent mixing layers,’’ J. Fluid Mech. 64, 775 共1974兲. 3 C. D. Winant and F. K. Browand, ‘‘Vortex pairing: The mechanisms of turbulent mixing layer growth at moderate Reynolds number,’’ J. Fluid Mech. 63, 237 共1974兲. 4 D. S. Shin and J. H. Ferziger, ‘‘Linear stability of the reacting mixing layer,’’ AIAA J. 29, 1634 共1991兲. 5 M. F. Miller, T. C. Island, B. Yip, C. T. Bowman, M. G. Mungal, and R. K. Hanson, ‘‘An experimental study of the structure of a compressible, reacting mixing layer,’’ AIAA Pap. 93-0354 共1993兲. 6 C. M. Coats and A. P. Richardson, ‘‘Nonpremixed combustion in turbulent mixing layers Part 1: Flame characteristics,’’ Combust. Flame 122, 253 共2000兲. 7 C. M. Coats, A. P. Richardson, and S. Wang, ‘‘Nonpremixed combustion in turbulent mixing layers Part 2: Mixing and flame stabilization,’’ Combust. Flame 122, 271 共2000兲. 8 J. J. Riley, R. W. Metcalfe, and S. A. Orszag, ‘‘Direct numerical simulations of chemical reacting turbulent mixing layers,’’ Phys. Fluids 29, 406 共1986兲. 9 P. A. McMurtry, J. J. Riley, and R. W. Metcalfe, ‘‘Effects of heat release on the large-scale structure in turbulent mixing layers,’’ J. Fluid Mech. 199, 297 共1989兲. 10 K. D. Squires and J. K. Eaton, ‘‘Preferential concentration of particles by turbulence,’’ Phys. Fluids A 3, 1169 共1991兲. 11 J. K. Eaton and J. R. Fessler, ‘‘Preferential concentration of particles by turbulence,’’ Int. J. Multiphase Flow 20, 169 共1994兲. 12 B. J. Lazaro and J. C. Lasheras, ‘‘Particle dispersion in a turbulent, plane, free shear layer,’’ Phys. Fluids A 1, 1035 共1989兲. 13 M. Samimy and S. K. Lele, ‘‘Motion of particles with inertia in a compressible free shear layer,’’ Phys. Fluids A 3, 1915 共1991兲. 14 F. Wen, N. Kamalu, J. N. Chung, C. T. Crowe, and T. R. Troutt, ‘‘Particle Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp 3320 Phys. Fluids, Vol. 13, No. 11, November 2001 dispersion by vortex structures in plane mixing layers,’’ J. Fluids Eng. 114, 657 共1992兲. 15 J. E. Martin and E. Meiburg, ‘‘The accumulation and dispersion of heavy particles in forced two-dimensional mixing layers. I. The fundamental and subharmonic cases,’’ Phys. Fluids 6, 1116 共1994兲. 16 N. Raju and E. Meiburg, ‘‘The accumulation and dispersion of heavy particles in forced two-dimensional mixing layers. Part 2: The effect of gravity,’’ Phys. Fluids 7, 1241 共1995兲. 17 K. T. Kiger and J. C. Lasheras, ‘‘Dissipation due to particle/turbulence interaction in a two-phase, turbulent, shear layer,’’ Phys. Fluids 9, 3005 共1997兲. 18 W. Ling, J. N. Chung, T. R. Troutt, and C. T. Crowe, ‘‘Direct numerical simulation of a three-dimensional temporal mixing layer with particle dispersion,’’ J. Fluid Mech. 358, 61 共1998兲. 19 Q. Wang, K. D. Squires, and L. P. Wang, ‘‘On the effects of nonuniform seeding on particle dispersion in two-dimensional mixing layers,’’ Phys. Fluids 10, 1700 共1998兲. 20 M. C. Soteriou and X. Yang, ‘‘Particle dispersion in variable density and viscosity shear flows,’’ Phys. Fluids 11, 1373 共1999兲. 21 R. A. Gore and C. T. Crowe, ‘‘Effect of particle size on modulating turbulent intensity,’’ Int. J. Multiphase Flow 15, 279 共1989兲. 22 S. Elghobashi, ‘‘On predicting particle-laden turbulent flows,’’ Appl. Sci. Res. 52, 309 共1994兲. 23 V. M. Kenning and C. T. Crowe, ‘‘On the effect of particles on carrier phase turbulence in gas-particle flows,’’ Int. J. Multiphase Flow 23, 403 共1997兲. 24 J. J. Riley and G. S. Patterson, ‘‘Diffusion experiments with numerically integrated isotropic turbulence,’’ Phys. Fluids 17, 292 共1974兲. 25 K. D. Squires and J. K. Eaton, ‘‘Measurements of particle dispersion obtained from direct numerical simulations of isotropic turbulence,’’ J. Fluid Mech. 226, 1 共1991兲. 26 S. Elghobashi and G. C. Truesdell, ‘‘Direct simulation of particle dispersion in decaying isotropic turbulence,’’ J. Fluid Mech. 242, 655 共1992兲. 27 L. P. Wang and M. R. Maxey, ‘‘Settling velocity and concentration distribution of heavy particles in isotropic turbulence,’’ J. Fluid Mech. 256, 27 共1993兲. 28 M. Boivin, O. Simonin, and K. D. Squires, ‘‘Direct numerical simulation of turbulence modulation by particles in isotropic turbulence,’’ J. Fluid Mech. 375, 235 共1998兲. 29 F. Wen, N. Kamalu, J. N. Chung, and C. T. Crowe, ‘‘Particle dispersion by vortex structures in plane mixing layers,’’ ASME J. Fluids Eng. 114, 657 共1992兲. 30 S. Pedinotti, G. Mariotti, and S. Banerjee, ‘‘Direct numerical simulation of particle behavior in the wall region of turbulent flows in horizontal channels,’’ Int. J. Multiphase Flow 18, 927 共1992兲. 31 H. Ounis, G. Ahmadi, and J. B. McLaughlin, ‘‘Brownian particle deposition in a directly simulated turbulent channel flow,’’ Phys. Fluids A 5, 1427 共1993兲. 32 M. Chen and J. B. McLaughlin, ‘‘A new correlation for the aerosol deposition rate in vertical ducts,’’ Colloid and Interface Sci. 169, 437 共1995兲. 33 D. W. I. Rouson, J. K. Eaton, and S. D. Abrahamson, ‘‘A direct numerical simulation of a particle-laden turbulent channel flow,’’ Technical Report, Department of Mechanical Engineering Report No. TSD-101, Stanford University, Stanford, CA, 1997. 34 K. D. Squires and J. K. Eaton, ‘‘Particle response and turbulence modification in isotropic turbulence,’’ Phys. Fluids A 2, 1191 共1990兲. 35 S. Elghobashi and G. C. Truesdell, ‘‘On the two-way interaction between homogeneous turbulence and dispersed solid particles. I: Turbulence modification,’’ Phys. Fluids A 5, 1790 共1993兲. 36 K. D. Squires and J. K. Eaton, ‘‘Effect of selective modification of turbulence on two-equation models for particle-laden turbulent flows,’’ J. Fluids Eng. 116, 778 共1994兲. Richard S. Miller 37 G. C. Truesdell and S. Elghobashi, ‘‘On the two-way interaction between homogeneous turbulence and dispersed solid particles. II: Particle dispersion,’’ Phys. Fluids 6, 1405 共1994兲. 38 M. R. Maxey, B. K. Patel, E. J. Chang, and L. P. Wang, ‘‘Simulations of dispersed turbulent multiphase flow,’’ Fluid Dyn. Res. 20, 143 共1997兲. 39 S. Sundaram and L. R. Collins, ‘‘A numerical study of the modulation of isotropic turbulence by suspended particles,’’ J. Fluid Mech. 379, 105 共1999兲. 40 E. Meiburg, E. Wallner, A. Pagella, A. Riaz, C. Hartel, and F. Necker, ‘‘Vorticity dynamics of dilute two-way coupled particle laden mixing layers,’’ J. Fluid Mech. 421, 185 共2000兲. 41 F. Mashayek, ‘‘Droplet-turbulence interactions in low Mach number homogeneous shear two-phase flows,’’ J. Fluid Mech. 367, 163 共1998兲. 42 F. Mashayek, ‘‘Direct numerical simulations of evaporating droplet dispersion in forced low Mach number turbulence,’’ Int. J. Heat Mass Transf. 41, 2601 共1998兲. 43 F. A. Jaberi and F. Mashayek, ‘‘Temperature decay in two-phase turbulent flows,’’ Int. J. Heat Mass Transf. 43, 993 共2000兲. 44 R. S. Miller and J. Bellan, ‘‘On the validity of the assumed pdf method for modeling binary mixing/reaction of evaporated vapor in gas/liquid-droplet turbulent shear flow,’’ Proceedings of the 27th Symposium 共International兲 on Combustion 共1998兲, pp. 1065–1072. 45 R. S. Miller and J. Bellan, ‘‘Direct numerical simulation of a confined three-dimensional gas mixing layer with one evaporating hydrocarbondroplet laden stream,’’ J. Fluid Mech. 384, 293 共1999兲. 46 R. S. Miller and J. Bellan, ‘‘Direct numerical simulation and subgrid analysis of a transitional droplet laden mixing layer,’’ Phys. Fluids 12, 650 共2000兲. 47 F. Mashayek, ‘‘Numerical investigation of reacting droplets in homogeneous shear turbulence,’’ J. Fluid Mech. 405, 1 共2000兲. 48 D. J. Glaze and S. H. Frankel, ‘‘Effect of dispersion characteristics on particle temperature in an idealized nonpremixed reacting jet,’’ Int. J. Multiphase Flow 26, 609 共2000兲. 49 R. S. Miller, K. G. Harstad, and J. Bellan, ‘‘Direct numerical simulation of supercritical fluid mixing layers applied to heptane-nitrogen,’’ J. Fluid Mech. 436, 1 共2001兲. 50 R. S. Miller, K. Harstad, and J. Bellan, ‘‘Evaluation of equilibrium and non-equilibrium evaporation models for many-droplet gas-liquid flow simulations,’’ Int. J. Multiphase Flow 24, 1025 共1998兲. 51 D. Papamoschou and A. Roshko, ‘‘The compressible turbulent mixing layer: An experimental study,’’ J. Fluid Mech. 197, 453 共1988兲. 52 R. D. Moser and M. M. Rogers, ‘‘Mixing transition and the cascade to small scales in a plane mixing layer,’’ Phys. Fluids A 3, 1128 共1991兲. 53 F. Kozusko, C. E. Grosch, T. L. Jackson, C. A. Kennedy, and T. B. Gatski, ‘‘The structure of variable property, compressible mixing layers in binary gas mixtures,’’ Phys. Fluids 8, 1945 共1996兲. 54 F. Kozusko, D. G. Lasseigne, C. E. Grosch, and T. L. Jackson, ‘‘The stability of compressible mixing layers in binary gases,’’ Phys. Fluids 8, 1954 共1996兲. 55 C. A. Kennedy and J. H. Chen, ‘‘Mean flow effects on the linear stability of compressible planar jets,’’ Phys. Fluids 10, 615 共1998兲. 56 C. A. Kennedy and M. H. Carpenter, ‘‘Several new numerical methods for compressible shear-layer simulations,’’ Appl. Numer. Math. 14, 397 共1994兲. 57 T. J. Poinsot and S. K. Lele, ‘‘Boundary conditions for direct numerical simulations of compressible viscous flows,’’ J. Comput. Phys. 101, 104 共1992兲. 58 R. S. Miller, C. K. Madnia, and P. Givi, ‘‘Structure of a turbulent reacting mixing layer,’’ Combust. Sci. Technol. 99, 1 共1994兲. Downloaded 02 Nov 2001 to 130.127.198.18. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp
© Copyright 2026 Paperzz