AIAA SciTech Forum 10.2514/6.

2020-0872
6-10 January 2020, Orlando, FL
AIAA Scitech 2020 Forum

An OpenFOAM-based fully compressible reacting flow solver

with detailed transport and chemistry for high-speed
combustion simulations

Dezhi Zhou∗ , Shufan Zou† , Suo Yang‡

Department of Mechanical Engineering, University of Minnesota - Twin Cities, Minneapolis, MN 55455

In this paper, a fully compressible reacting flow solver with detailed species transport and
chemistry for high-speed combustion simulations is developed. This new solver, which is based
on the open source platform OpenFOAM, is validated with one-dimensional (1D) reactive shock
Downloaded by UNIVERSITY OF TEXAS AT AUSTIN on January 9, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-0872

tube, two dimensional (2D) cold jet in cross flow (JICF) with detailed chemical kinetics and
transport properties. In addition, the new solver is also tested with the method of tabulated
dynamic adaptive chemistry (TDAC) for a low-Mach flame (Sandia Flame D). The results show
that this solver is robust enough to capture the shock/reaction wave, able to predict the pressure
and temperature variation induced by shocks in JICF with Ma = 3.5 and capture the flame
structure in flame with Ma = 0.1 − 0.2. The TDAC method implemented with this new solver
is also tested in the Sandia Flame D case and indicates good accuracy.

I. Introduction
ocket engines have been successfully applied to many supersonic flight vehicles (e.g X-15 reaching Mach 6.7 in
R 1967). One of the biggest constraints of rocket engine is its limited ranges due to the heavy carry-on oxidizer. Air
breathing engines which ingest air as oxidizer are thus an attractive solution. However, capturing atmospheric air and
burning fuels with air at speeds of order 6000-10000 km/h is extremely difficult. In ramjet engines, the air is decelerated
so that it burns with the fuel under subsonic conditions. At even higher speeds (e.g., Mach number greater than 5), such
deceleration of the airflow to subsonic velocities leads to large energy losses and thermal load. To this end, burning
fuels in supersonic airflow stimulates the development of supersonic combustion ramjets (scramjets). Due to the limited
and expensive experimental facility for supersonic combustion test, scramjet engine simulation is an economical way to
obtain insights into supersonic combustion [1].
Scramjet engine simulations encounter many challenges, such as modeling the high-speed flow-turbulence-chemistry
interaction [2]. To simulate these effects accurately, highly refined mesh size as well as finite rate chemistry (FRC) are
typically needed, which could easily cause prohibitive computation cost. Even though the low-manifold models such as
flamelet/progress-variable (FPV) model is used in some supersonic combustion [3], it is pointed out in [4] that most of
these low-manifold models need complicated corrections when applied to fully compressible flows and often mispredict
key physical quantities. As a result, mechanism reduction has to be employed for FRC simplifications.
The most commonly seen simplification methodology in combustion simulations is a priori chemical mechanism
reduction, such as the methods of Direct Relation Graph (DRG) [5], DRG with Error Propagation (DRGEP) [6], Path
Flux Analysis [7], and Global Pathway Selection (GPS) [8]. The conventional way for mechanism reduction is to
generate the mechanisms under a certain range of operating conditions a priori and subsequently implement these
reduced mechanisms into computational fluid dynamics (CFD) combustion simulations. In recent years, the concept of
dynamic adaptive chemistry (DAC) scheme which reduces the mechanism on-the-fly at each time step and at each local
computational cell has been extensively studied in reactive flow simulations [9, 10]. In this way, the number of species
could be reduced to smaller values than the conventional calculations, but the on-the-fly mechanism reduction process
itself could incur CPU cost. The tabulated chemistry techniques further reduce the effort of mechanism reduction
process as well as the chemistry integration process. Examples of combining DAC and tabulated chemistry can be
found in [11–13].
Before applying these tabulated chemistry and dynamic adaptive chemistry techniques into high-speed combustion
simulations, a robust multi-dimensional fully-compressible CFD solver with the capability to deal with complex
∗ Post-Doctoral Associate, Member AIAA.
† Ph.D. Student.
‡ Richard & Barbara Nelson Assistant Professor, [email protected] (Corresponding Author), Member AIAA.

Copyright © 2020 by Suo Yang. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
 geometry, unstructured grid, detailed chemistry and species transport is needed. In this study, a new density-based
fully-compressible solver based on OpenFOAM (OF) is developed. The solver is validated with 1D reacting shock
tube, jet in cross flow (JICF) with detailed chemistry and mixture averaged transport models. Finally, tabulated DAC
(TDAC) [11] is coupled in our new solver in a 2D Sandia Flame D simulation. All the results show that the current
solver is able to capture the shock wave, flame, and their interaction very well. The Sandia Flame D results show that
TDAC application in our new solver predicts the flame structure and species distribution with good agreement with the
experimental data.

II. Methodology

A. Governing Equations
The equation of mass conservation, momentum (neglecting body force), energy, together with species transport can
Downloaded by UNIVERSITY OF TEXAS AT AUSTIN on January 9, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-0872

be written as:
∂ρ
+ ∇ · (ρu) = 0 (1)
∂t
∂ ρu
+ ∇ · (ρu ⊗ u) = −∇p + ∇ · τ (2)
∂t
∂ ρE
+ ∇ · (ρE u) = −∇ · (up) + ∇ · (u · τ) + ∇ · q (3)
∂t
∂ ρYs
+ ∇ · (ρYs u) = −∇ · js + ρωÛ s , s = 1, ..., ns (4)
∂t
where ρ is the mixture density, p is the pressure, u is the velocity, τ = τi j g i g j is the viscous stress tensor, q is the
heat flux, E = e + 21 u · u is the total energy. In Eq. (4), Ys , js and ωÛ s are the mass fraction, diffusion mass flux and
production rate of specie s, while ns is the total number of species. To better capture the heat formation in reacting
compressible flow, here we use the energy equation in the expression of enthalpy, which can be obtained by combine Eq.
(1), (2) and (3) together and can be written as:

∂ ρh ∂ ρK ∂p
+ ∇ · (ρhu) + + ∇ · (ρK u) − = ∇ · (τ · u) + ∇ · q + ρr (5)
∂t ∂t ∂t
where h = e + ρp is the total enthalpy, r is the heat source (chemistry or radiation). The pressure is calculated by the
equation-of-state of perfect gas
Õ Ys
p = ρRT (6)
s
Ws
where Ws is the molar mass of specie s, and R is the universal gas constant. The heat flux can be computed as the
combination of heat transfer and convection
Õ
q = −λ∇T − hs j s (7)
s

where λ is the mixture thermal conductivity, and hs are the enthalpy of specie s.The diffusion mass flux can be expressed
as  
∇Xs
js = −ρYs Ds + uc (8)
Xs
where uc is the correction velocity required to enforce a zero net diffusion flux,Xs is mole fraction and Ds is the
diffusivity of specie s. Ds is calculated with the mixture average transport model [14], where the transport variables
(.e.g., viscosity and thermal conductivity) are computed with coupled Cantera package [15].

2
 B. Numerics and algorithm
The framework of second-order central-upwind scheme for finite volume method is implemented in our new solver
(named as rhoCentralSpeciesFoam), following the implementation of rhoCentralFoam [16]. However, as we changed
the expression of energy equation and introduce reactions, some proper adjustment are needed.
As for the convective terms at the expression of ∇ · (uΨ), the discretization in finite volume could be written as
∭ ∯ Õ Õ
∇ · (uΨ)dV = φf Ψf ≈ αφ f + Ψ f + + (1 − α)Φ f − Ψ f − + ω f Ψ f − − Ψ f +


dS · uΨ ≈ (9)
V s f f

where the third term is strictly only required when the convective term is a part of a material derivative. S f is the vector
in the normal direction of surface f with modulus of the area of surface f ,φ f = S f · u f is the volumetric flux, subscript
f indicates the surface of a cell, volumetric fluxes associated with the local speed of propagation can be given by
Downloaded by UNIVERSITY OF TEXAS AT AUSTIN on January 9, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-0872

ψ f + = max c f + S f  + φ f +, c f − S f  + φ f −, 0
   

(10)
ψ f − = max c f + S f  − φ f +, c f − S f  − φ f −, 0
   

and parameter α and ω satisfy

ψf +
α= ψ f + +ψ f −
(11)
ω f = α(1 − α)(ψ f + + ψ f − )
Similarly, the gradient terms can be discretized as
∭ ∯ Õ Õ
∇ΨdV = αS f Ψ f + + (1 − α)S f Ψ f −

dSΨ ≈ Sf Ψf ≈ (12)
y s f f

As for the Laplacian terms with diffusion coefficient Γ, they can be discretized like
"  2  2 ! #
∭ ∮ Õ Õ S f  S f 
∇ · (Γ∇Ψ)dV = dS · (Γ∇Ψ) ≈ ΨN − Ψp + Sf −


Γ f S f · (∇Ψ) f ≈ Γf d · (∇Ψ) f
v s f f
Sf · d Sf · d
(13)
A second-order implicit Euler backward scheme is used for time marching, resulting in a second-order accurate numerical
framework.
As the the total energy equation(Eq. 3) is moved to total enthalpy equation (Eq. 5), Eq.(5) is not closed as an
additional time derivative of pressure appears. The closure of this term requires the information of what Eq. 5 is solved
for, because in the current rhoCentralSpeciesFoam density-based solver, the pressure is a function of temperature (see Eq.
6). It is then seen that Eq. 5 and 6 need to be solved together. In this study, we introduce an p-T prediction-correction
iteration to appropriately solve these two equations. Basically, the predicted pressure (initially, the old pressure at the
previous time step is used as the predicted pressure) is used to evaluate the time derivative of pressure term and then
Eq. 5 is solved. The resulting predicted enthalpy is then employed for the temperature prediction. Subsequently, the
pressure is corrected based on the equation of state (Eq. 6), using the predicted temperature. This loop is iterated until
the pressure converges. A relative error tolerance 1.0e−6 is used for the pressure convergence.

C. TDAC
Tabulated dynamic adaptive chemistry (TDAC) [11] is used in this study to accelerate the finite rate chemistry
integration. TDAC is an algorithm that combines in situ adaptive tabulation (ISAT) [17] and dynamic adaptive chemistry
(DAC) [9]. In conventional DAC, the mechanism reduction process is conducted at each time step and each grid
point to generate an optimal reduced mechanism considering the local and current thermodynamic conditions and
species composition. However, as pointed out in [12, 13, 18], the mechanism reduction process could incur expensive
computation, especially with large amount of grid points. In TDAC, this issue is addressed by the ISAT technique,
which only requires mechanism reduction when the current and local state is not within the accuracy region of the
tabulated chemistry. In ISAT, this accuracy region is quantified and the tabulation is dynamically increased.
Another important feature of TDAC is the reduction on species transport equations. When reduced mechanism is
produced and unimportant species are eliminated, only the important species transport equations are solved, which
further significantly reduce the computational cost. For more details on TDAC, interested readers could refer to [11].

3
 III. Results

A. Multi-species reactive shock tube

The current solver is firstly validated with a multi-species reactive shock tube problem, which is aimed at testing
the competition between convection and reaction terms. Moreover, this problem could assess the robustness of the
numerical schemes in rhoCentralSpeciesFoam in a reactive mixture with strong pressure variations. This problem was
studied firstly by Oran et al. [19] and then used in other investigations [20–22] as a benchmark case for their newly
developed numerical schemes. The shock tube is initially filled with reactive mixture and subject to a shock wave. Due
to the closed and reflecting wall at the shock tube boundary, the reactive mixture is compressed and a reaction wave is
formed after ignition delay. The reaction wave catches up the reflecting shock wave and finally merges with the shock
wave structure.
In the simulation of this problem, a 12 cm 1D computational domain is set up with 400 grid points, and filled with
Downloaded by UNIVERSITY OF TEXAS AT AUSTIN on January 9, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-0872

premixed reactive mixture (2/1/7 molar ratio of H2 /O2 /Ar). The left side applied wall boundary conditions while at
x=12 cm, a non-reflecting boundary condition is used. Initially, from x=0 cm to x=6 cm, the pressure, velocity and
density of the mixture is 7173 Pa, 0 m/s and 0.072 kg/m3 while from x=6 cm to x=12 cm, the corresponding values
are 35594 Pa, -487.34 m/s and 0.18075 kg/m3 . The chemical kinetics is considered by a 8 species and 18 reaction
steps mechanism, which is the same as used in Ref. [22] because we will use Ref. [22] as a benchmark to validate our
results. It is noted that the current reactive shock tube case is inviscid and thus the transport equation as described in
Section II.A are restricted to their Euler form.
The temperature profiles at time = 170 µs, 190 µs and 230 µs are displayed and compared with the results in [22], as
shown in Fig. 1. To better visualize the interaction between the reactive wave and shock wave structure, the temperature
contours are also demonstrated in Fig. 2. As observed, our solver perfectly matches the benchmark results at different
times and well captures the shock wave propagation, reaction wave initiation and interaction/merge of shock and reaction
wave.

Fig. 1 Temperature profiles as a function of distance to the left boundary wall at three diferent time (170
µs, 190 µs and 230 µs). Symbols are results of the benchmark case [22]. Lines are results of our solver
rhoCentralSpeciesFoam.

B. 2-D jet in supersonic cross flow

Transverse injection, or jet in cross flow (JICF) is one of the fundamental interests of the research for scramjet engines
which suggests a potential strategy of the fuel injection. In this case, a cold flow high Mach JICF with nitrogen-injection
is simulated to test the numerical schemes and species diffusion. This case corresponds to the experiment by Aso
et al. [23]. As shown in 3, the geometry consists of a long flat plain with a slot.The distance from the leading edge
to the nozzle is L=330 mm, which is the same to the experiment configuration. The air flows into the domain with
a total pressure p0 = 1.2 M Pa, Mach number Ma = 3.75, total temperature T0 = 299K. And the corresponding

4
Downloaded by UNIVERSITY OF TEXAS AT AUSTIN on January 9, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-0872

Fig. 2 Temperature contours at three diferent time (170 µs, 190 µs and 230 µs), calculated with our solver
rhoCentralSpeciesFoam.

parameters of the nitrogen injection is p0i n j = 0.31p0 and T0i n j = 299 K. By using a converging sonic throat bonded to
the slot, the Mach number of injection is kept as Ma = 1.The simulation is performed under a non-uniform structural
mesh with 692 × 175 grids, with first ∆z = 0.005(mm) ≈ 200y + . The temperature field near the injection is shown in

Fig. 3 Geometry of the non-reactive transverse injection case

Fig.5 and a boundary layer and before the injector shock induced high temperature regions are clearly observed. The
non-dimensionalized wall pressure p∗ = pwp∞all is plotted in Fig. 4, which indicates a considerable agreement with the
experiment data.

C. 2-D supersonic cavity flow

In this part, a 2-D Ma = 3 cavity flow is simulated to show the ability for high speed cavity flow. The geometry in
this case is shown in Fig. 6, which contains a cavity with the size (44.5mm × 8.9mm). The free stream condition in this
cavity flow is total temperature T0 = 300K, total pressure p0 = 690k Pa and Mach number Ma = 3, the same as the
experimental case LD5-O1-16 of Gruber et al [24].
The computational domain is split into two blocks, where the main flow region is discretized to 1312 × 102 while
the cavity region contains 300 × 60 cells. Both Reynolds averaged Navier Stokes (RANS) and large eddy simulations
(LES) are used for this case. The pressure on the wall in the cavity is shown in Fig.7 which suggest not only the RANS
results but LES results fits the experiment data well. In addition, we treat air as a mixture of 79% nitrogen and 21%
oxygen in this case.

5
 4
present
Aso et al.
3

0
Downloaded by UNIVERSITY OF TEXAS AT AUSTIN on January 9, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-0872

290 300 310 320 330 340 350 360

x(mm)

Fig. 4 Pressure on the wall near the injection, compared to experiment data by Aso et al. [23].

Fig. 5 Temperature field near the injection

D. Sandia Flame D
Although the density-based solver developed in this study is fully-compressible, we will show in this section that it is
still able to well capture the flow and flame field at Ma = 0.1 − 0.2. The Sandia Flame D piloted partially premixed flame
[25] (the maximum Mach number in this flame is about 0.17) is simulated with the TDAC method. Sandia Flame D
documentation provides detailed experimental data and the flames are widely validated with different CFD frameworks
[26, 27]. Fully compressible solvers is firstly used for Sandia Flame D by Yang et al. [28] with preconditioning scheme.
The flow conditions in Sandia Flame D is presented in Table 1.
In this study, the current solver is used for the low-speed Sandia Flame D with purpose of testifying the capability
of our solve in the low Mach number range. A 2D wedge computational domain is used in this simulation with
the Reynolds Averaged transport equations. The Reynolds stress term is closed with the standard k −  model [29].
Turbulence-chemistry interaction is handles by the EDC model [30]. The chemistry source term integration is achieved
by solving a series of ODE equations, where tabulated dynamic adaptive chemistry (TDAC) is used to accelerate the
computation. GRI3.0 mechanism [31] is used for the chemical kinetics.
Before the comparison with experimental data, we plot the statistically stationary temperature results calculated by
the existing OF solver (i.e., reactingFoam) and our rhoCentralSpeciesFoam solver. There is no obvious discrepancy

Fig. 6 geometry of 2-D cavity flow case

6
Downloaded by UNIVERSITY OF TEXAS AT AUSTIN on January 9, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-0872

Fig. 7 pressure distribution on the wall in the cavity

Table 1 Flow conditions in Sandia Flame D.

Inner diameter Outer diameter Bulk velocity Temperature

Composition
(mm) (mm) (m/s) (K)
25% CH4 /75% air
Fuel jet 7.2 7.7 49.6 294
(by volume)
CH4 /air equilibrium
Piloted flame 7.7 18.2 11.4 1880
mixture (φ = 0.77)
Co-flow Air 18.9 N.A. 0.9 291

observed for the temperature distribution as seen in Fig. 8.

To further validate our solver, the radial temperature and OH mass fraction distribution calculated with the current
solver is compared with the experimental data, as plotted in Fig. 9. It is then seen that the solver developed for
fully-compressible reacting flow with TDAC is also able to capture the flame structure and speciation data in the
low-speed Sandia Flame D case. Note that the current simulation is only a RANS simulation, which might be the
reason of discrepancy between the predicted and measured data, especially at x/d = 15. Future work employing LES
framework in the current solver is desirable.

IV. Conclusion
In this study, a OpenFoam-based fully compressible solver rhoCentralSpeciesFoam for reacting flow is developed.
This solver is validated with a reactive shock tube case firstly and it is found that rhoCentralSpeciesFoam is able to
capture the shock wave, reaction wave and their interaction very well, indicating the robustness of the numerical schemes
with the presence of strong pressure variation and reactions. In addition, a cold JICF with Ma = 3.5 is simulated and
the wall pressure near the nitrogen injection agrees well with the experimental data. With respect to relatively low
Mach flow, a piloted partially flame (Sandia Flame D) is modelled, with the method of TDAC. The comparison of the
predicted flame structure between the existing low speed OF solver (i.e., reactingFoam) and by our solver show nearly
no difference. In addition, the computed radial temperature and OH distributions match well with the experimental data.
In sum, the newly developed fully-compressible reacting flow solver is robust for high Mach number reacting flow.
Within the range of Ma = 0.1 - 0.2, this solver is still able to give reliable results. This solver employs detailed finite rate
chemistry and detailed transport, with the ease to implement mechanism reduction and chemistry tabulation methods
(e.g., TDAC). Future work should validate the current solver within the LES framework for more complex geometry and
highly turbulent flames.

7
Downloaded by UNIVERSITY OF TEXAS AT AUSTIN on January 9, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-0872

Fig. 8 Statistically stationary temperature distribution, calculated with the (a) low Mach reacting flow solver
reactingFoam, (b) our new full-compressible solver rhoCentralSpeciesFoam.

Fig. 9 Radial OH mass fraction and temperature distribution at x/d = 15 and x/d = 45. Symbols are
experimental data while lines are simulation data.

Acknowledgments
S. Yang gratefully acknowledges the faculty start-up funding from the University of Minnesota – Twin Cities. The
authors gratefully acknowledge Prof. Graham V. Candler and the Minnesota Supercomputing Institute (MSI) for the

8
 computational resources.

References
[1] Urzay, J., “Supersonic combustion in air-breathing propulsion systems for hypersonic flight,” Annual Review of Fluid Mechanics,
Vol. 50, 2018, pp. 593–627.

[2] Gonzalez-Juez, E. D., Kerstein, A. R., Ranjan, R., and Menon, S., “Advances and challenges in modeling high-speed turbulent
combustion in propulsion systems,” Progress in Energy and Combustion Science, Vol. 60, 2017, pp. 26–67.

[3] Cymbalist, N., Candler, G. V., and Dimotakis, P., “Application of the Evolution-Variable Manifold Approach to Cavity-Stabilized
Ethylene Combustion,” 46th AIAA Fluid Dynamics Conference, 2016, p. 3481.

[4] Mrema, H. F., and Candler, G. V., “Large Eddy Simulation of Supersonic Combustion using the Flamelet/Progress-Variable
Downloaded by UNIVERSITY OF TEXAS AT AUSTIN on January 9, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-0872

Approach and the Evolution-Variable Manifold Approach,” AIAA Scitech 2019 Forum, 2019, p. 1446.

[5] Lu, T., and Law, C. K., “A directed relation graph method for mechanism reduction,” Proceedings of the Combustion Institute,
Vol. 30, No. 1, 2005, pp. 1333–1341.

[6] Pepiot-Desjardins, P., and Pitsch, H., “An efficient error-propagation-based reduction method for large chemical kinetic
mechanisms,” Combustion and Flame, Vol. 154, No. 1-2, 2008, pp. 67–81.

[7] Sun, W., Chen, Z., Gou, X., and Ju, Y., “A path flux analysis method for the reduction of detailed chemical kinetic mechanisms,”
Combustion and Flame, Vol. 157, No. 7, 2010, pp. 1298–1307.

[8] Gao, X., Yang, S., and Sun, W., “A global pathway selection algorithm for the reduction of detailed chemical kinetic mechanisms,”
Combustion and Flame, Vol. 167, 2016, pp. 238–247.

[9] Liang, L., Stevens, J. G., and Farrell, J. T., “A dynamic adaptive chemistry scheme for reactive flow computations,” Proceedings
of the Combustion Institute, Vol. 32, No. 1, 2009, pp. 527–534.

[10] Yang, S., Ranjan, R., Yang, V., Menon, S., and Sun, W., “Parallel on-the-fly adaptive kinetics in direct numerical simulation of
turbulent premixed flame,” Proceedings of the Combustion Institute, Vol. 36, No. 2, 2017, pp. 2025–2032.

[11] Contino, F., Jeanmart, H., Lucchini, T., and D’Errico, G., “Coupling of in situ adaptive tabulation and dynamic adaptive
chemistry: An effective method for solving combustion in engine simulations,” Proceedings of the Combustion Institute, Vol. 33,
No. 2, 2011, pp. 3057–3064.

[12] Yang, S., Ranjan, R., Yang, V., Sun, W., and Menon, S., “Sensitivity of predictions to chemical kinetics models in a temporally
evolving turbulent non-premixed flame,” Combustion and Flame, Vol. 183, 2017, pp. 224–241.

[13] Zhou, D., Tay, K. L., Li, H., and Yang, W., “Computational acceleration of multi-dimensional reactive flow modelling using
diesel/biodiesel/jet-fuel surrogate mechanisms via a clustered dynamic adaptive chemistry method,” Combustion and Flame,
Vol. 196, 2018, pp. 197–209.

[14] Bird, R. B., Stewart, W. E., and Lightfoot, E. N., Transport phenomena, John Wiley & Sons, 2007.

[15] Goodwin, D. G., Speth, R. L., Moffat, H. K., and Weber, B. W., “Cantera: An Object-oriented Software Toolkit for Chemical
Kinetics, Thermodynamics, and Transport Processes,” https://www.cantera.org, 2018. doi:10.5281/zenodo.1174508,
version 2.4.0.

[16] Greenshields, C. J., Weller, H. G., Gasparini, L., and Reese, J. M., “Implementation of semi-discrete, non-staggered central
schemes in a colocated, polyhedral, finite volume framework, for high-speed viscous flows,” International journal for numerical
methods in fluids, Vol. 63, No. 1, 2010, pp. 1–21.

[17] Pope, S. B., “Computationally efficient implementation of combustion chemistry using in situ adaptive tabulation,” 1997.

[18] Sun, W., Gou, X., El-Asrag, H. A., Chen, Z., and Ju, Y., “Multi-timescale and correlated dynamic adaptive chemistry modeling
of ignition and flame propagation using a real jet fuel surrogate model,” Combustion and Flame, Vol. 162, No. 4, 2015, pp.
1530–1539.

[19] Oran, E., Young, T., Boris, J., and Cohen, A., “Weak and strong ignition. I. Numerical simulations of shock tube experiments,”
Combustion and Flame, Vol. 48, 1982, pp. 135–148.

9
 [20] Fedkiw, R. P., Merriman, B., and Osher, S., “High accuracy numerical methods for thermally perfect gas flows with chemistry,”
Journal of Computational Physics, Vol. 132, No. 2, 1997, pp. 175–190.

[21] Deiterding, R., “A parallel adaptive method for simulating shock-induced combustion with detailed chemical kinetics in complex
domains,” Computers & Structures, Vol. 87, No. 11-12, 2009, pp. 769–783.

[22] Ferrer, P. J. M., Buttay, R., Lehnasch, G., and Mura, A., “A detailed verification procedure for compressible reactive
multicomponent Navier–Stokes solvers,” Computers & Fluids, Vol. 89, 2014, pp. 88–110.

[23] Aso, S., KAWAI, M., and ANDO, Y., “Experimental study on mixing phenomena in supersonic flows with slotinjection,” 29th
Aerospace sciences meeting, 1991, p. 16.

[24] Gruber, M., Baurle, R., Mathur, T., and Hsu, K.-Y., “Fundamental studies of cavity-based flameholder concepts for supersonic
combustors,” Journal of Propulsion and power, Vol. 17, No. 1, 2001, pp. 146–153.
Downloaded by UNIVERSITY OF TEXAS AT AUSTIN on January 9, 2020 | http://arc.aiaa.org | DOI: 10.2514/6.2020-0872

[25] Barlow, R., and Frank, J., “Effects of turbulence on species mass fractions in methane/air jet flames,” Symposium (International)
on Combustion, Vol. 27, Elsevier, 1998, pp. 1087–1095.

[26] Pitsch, H., and Steiner, H., “Large-eddy simulation of a turbulent piloted methane/air diffusion flame (Sandia flame D),” Physics
of fluids, Vol. 12, No. 10, 2000, pp. 2541–2554.

[27] Jones, W., and Prasad, V., “Large Eddy Simulation of the Sandia Flame Series (D–F) using the Eulerian stochastic field method,”
Combustion and Flame, Vol. 157, No. 9, 2010, pp. 1621–1636.

[28] Yang, S., Wang, X., Huo, H., Sun, W., and Yang, V., “An efficient finite-rate chemistry model for a preconditioned compressible
flow solver and its comparison with the flamelet/progress-variable model,” Combustion and Flame, Vol. 210, 2019, pp. 172–182.

[29] Launder, B. E., and Spalding, D. B., “The numerical computation of turbulent flows,” Numerical prediction of flow, heat
transfer, turbulence and combustion, Elsevier, 1983, pp. 96–116.

[30] Magnussen, B. F., and Hjertager, B. H., “On mathematical modeling of turbulent combustion with special emphasis on soot
formation and combustion,” Symposium (international) on Combustion, Vol. 16, Elsevier, 1977, pp. 719–729.

[31] Smith, G. P., Golden, D. M., Frenklach, M., Moriarty, N. W., B., M. E., Goldenberg, C. T., B., Hanson, R. K., Song, S., Jr., W.
C. G., Lissianski, V. V., , and Qin, Z., “GRI-MECH 3.0,” http://www.me.berkeley.edu/gri_mech/, 1999.

10