A deterministic-stochastic method for computing the Boltzmann collision integral in $\mathcal{O}(MN)$ operations

Alexander Alekseenko; Truong Nguyen; Aihua Wood

doi:10.3934/krm.2018047

Article Contents

2018, Volume 11, Issue 5: 1211-1234. Doi: 10.3934/krm.2018047 open access

This issue Previous Article On global solutions to the Vlasov-Poisson system with radiation damping Next Article Stability of traveling waves for nonlocal time-delayed reaction-diffusion equations

A deterministic-stochastic method for computing the Boltzmann collision integral in $\mathcal{O}(MN)$ operations

1.
Department of Mathematics, California State University Northridge, Northridge, CA 91330, USA
2.
Department of Mathematics & Statistics, Wright State University, Dayton, OH 45435, USA
3.
Department of Mathematics & Statistics, Air Force Institute of Technology, WPAFB, OH 45433, USA

* Corresponding author: alexander.alekseenko@csun.edu
* Corresponding author: alexander.alekseenko@csun.edu

Received: February 29, 2016

Revised: August 31, 2017

Published: May 2018

The first author was supported by NSF grant DMS-1620497 and HPTi PETTT grant PP-SASKY06-001. The third author was supported by AFOSR Grant No. F4FGA04296J00.

Abstract Full Text(HTML) Figure(2) / Table(5) Related Papers Cited by

Abstract

We developed and implemented a numerical algorithm for evaluating the Boltzmann collision integral with $O(MN)$ operations, where $N$ is the number of the discrete velocity points and $M <N$. At the base of the algorithm are nodal-discontinuous Galerkin discretizations of the collision operator on uniform grids and a bilinear convolution form of the Galerkin projection of the collision operator. Efficiency of the algorithm is achieved by applying singular value decomposition compression of the discrete collision kernel and by approximating the kinetic solution by a sum of Maxwellian streams using a stochastic likelihood maximization algorithm. Accuracy of the method is established on solutions to the problem of spatially homogeneous relaxation.

Keywords:

Kinetic equations,
collision integral,
nodal-discontinuous Galerkin discretizations,
convolution formulation,
dynamics of non-continuum gas.

Mathematics Subject Classification: Primary: 76P05, 76M10; Secondary: 65M60.

Citation:

FullText(HTML)

Figure 1. Magnitudes of eigenvalues of $A_{\alpha\beta}$ for the cases $s_u = s_v = s_w = 1$, $N = 15^3$ (a) and $s = 1$, $N = 33^3$ (b)

Download: Full-size image PowerPoint slide

Figure 2. Directional temperatures in the solutions to the problem of spatially homogeneous relaxation for the initial data corresponding to Mach 1.55 (a), (b) and Mach 3.0 (c), (d) shock waves. The solutions shown in Figures (a) and (c) are obtained by using the truncated kernel method for evaluating the collision integral. The solid line corresponds to the solution using the full Boltzmann integral. The new model looses accuracy at about three mean free times for both Mach 1.55 and Mach 3.0 solutions. The solutions shown in Figures (b) and (d) are obtained by using the truncated kernel model for about three mean free times and by using the ellipsoidal-statistical Bhatnagar-Gross-Krook model after

Download: Full-size image PowerPoint slide

Table 1. Convergence of errors in the approximation of $\mathcal{A}(f, f)$ by $\hat{\mathcal{A}}_M(f, f)$ for the case $N = 15^3$ with respect to the number $M$ of retained eigenvectors: $\vert\lambda_{M}\vert$ is the magnitude of the $M^{\mathrm{th}}$ eigenvalue; $\varepsilon_1$ is the absolute error between $\mathcal{A}(f, f)$ and $\hat{\mathcal{A}}_M(f, f)$ when $f(t, \vec{x}, \vec{v})$ is a single Maxwellian stream with density $n = 1$, bulk velocity $\bar{\vec{u}} = (0, 0, 0)$ and temperature $T = 0.59$; $\varepsilon_2$ is the relative error between $\mathcal{A}(f, f)$ and $\hat{\mathcal{A}}_M(f, f)$ when $f(t, \vec{x}, \vec{v})$ is a sum of two Maxwellian streams with densities $n_1 = n_2 = 1$, bulk velocities $\bar{\vec{u}}_1 = (0.4, 0, 0)$ and $\bar{\vec{u}}_2 = (-0.4, 0, 0)$ and temperatures $T_1 = T_2 = 0.89$; $\varepsilon_3$ is the relative error between $\mathcal{A}(f, f)$ and $\hat{\mathcal{A}}_M(f, f)$ when $f(t, \vec{x}, \vec{v})$ is a sum of two Maxwellian streams with densities $n_1 = n_2 = 1$, bulk velocities $\bar{\vec{u}}_1 = (0.4, 0, 0)$ and $\bar{\vec{u}}_2 = (-0.4, 0, 0)$ and temperatures $T_1 = T_2 = 0.31$

$M$	$\vert\lambda_{M}\vert$	$\varepsilon_{1}$	$\varepsilon_{2}$	$\varepsilon_{3}$
10	4.1E+0	1.7E-1	3.1E-1	2.1E-1
80	1.0E+0	3.7E-2	5.8E-1	5.3E-2
160	5.7E-1	1.6E-2	2.9E-2	2.0E-2
320	3.6E-1	1.5E-2	2.5E-2	1.8E-2
640	2.2E-1	3.3E-3	8.4E-3	4.4E-3
960	1.1E-1	1.3E-3	5.7E-3	2.2E-3
1120	5.8E-2	2.3E-3	5.9E-3	2.3E-3
1280	4.1E-3	1.7E-5	2.1E-3	3.1E-5
1400	4.6E-5	2.2E-7	2.0E-3	3.9E-7

| Show Table

DownLoad: CSV

Table 2. Convergence of errors in the approximation of $\mathcal{A}(f,f)$ by $\hat{\mathcal{A}}_M(f,f)$ for the case $N=33^3$ with respect to the number $M$ of retained eigenvectors: $\vert\lambda_{M}\vert$ is the magnitude of the $M^{\mathrm{th}}$ eigenvalue; $\varepsilon_1$ is the absolute error between $\mathcal{A}(f,f)$ and $\hat{\mathcal{A}}_M(f,f)$ when $f(t,\vec{x},\vec{v})$ is a single Maxwellian stream with density $n=1$, bulk velocity $\bar{\vec{u}}=(0,0,0)$ and temperature $T=0.89$. $\varepsilon_2$ is the relative error between values of $\mathcal{A}(f,f)$ and values $\hat{\mathcal{A}}_M(f,f)$ when $f(t,\vec{x},\vec{v})$ is a sum of two Maxwellian streams with densities $n_1=n_2=1$, bulk velocities $\bar{\vec{u}}_1=(0.\overline{36},0,0)$ and $\bar{\vec{u}}_2=(-0.\overline{36},0,0)$ and temperatures $T_1=T_2=0.89$. The quantity $\varepsilon_3$ is the relative error between values of $\mathcal{A}(f,f)$ and values $\hat{\mathcal{A}}_M(f,f)$ when $f(t,\vec{x},\vec{v})$ is a sum of two Maxwellian streams with densities $n_1=n_2=1$, bulk velocities $\bar{\vec{u}}_1=(0.\overline{36},0,0)$ and $\bar{\vec{u}}_2=(-0.\overline{36},0,0)$ and temperatures $T_1=T_2=0.31$

$M$	$\vert\lambda_{M}\vert$	$\varepsilon_1$	$\varepsilon_2$	$\varepsilon_3$
10	5.0E+0	1.0E-1	5.2E-1	3.7E-1
100	1.7E+0	4.1E-2	2.0E-1	3.5E-1
500	5.4E-1	5.0E-3	2.3E-2	2.5E-2
1000	2.9E-1	3.3E-3	1.6E-2	1.5E-2
2000	1.7E-1	1.6E-3	7.9E-3	4.7E-3
2900	1.3E-1	1.5E-3	7.5E-3	3.9E-3
4800	9.6E-2	9.2E-4	4.6E-3	2.3E-3
9900	4.3E-2	1.6E-4	3.5E-3	5.3E-4
14400	2.8E-12	2.8E-5	3.5E-3	9.2E-7

| Show Table

DownLoad: CSV

Table 3. CPU time in seconds for computing one time step of the spatially homogeneous solution on a 2.4 GHz processor. Results of simulations for the $N = 15^3$ and $N = 33^3$ grids are presented. $M$ is the number of eigenvectors used in (17). The bottom row gives the time for computing one time step in the method of [3]. Speedup is the ratio of the time used in the new method and the time used by the algorithm of [3]

$n=15$, $s_u=s_v=s_w=1$			$n=33$, $s_u=s_v=s_w=1$
$M$	CPU Time, s	Speedup	$M$	CPU Time, s	Speedup
160	2.0	16.4	4000	444.1	40.0
320	3.8	8.7	6000	668.10	26.6
480	5.4	6.1	8000	890.2	19.9
640	7.0	4.8	10000	1113.4	15.9
960	10.3	3.2	12000	1333.6	13.3
1400	16.0	2.1	14000	1559.4	11.4
DG Boltzmann CPU Time: 33.18 s			DG Boltzmann CPU Time: 17758 s

| Show Table

DownLoad: CSV

Table 4. Results of approximating the solution to the problem of spatially homogeneous relaxation [3] by a sum of two Maxwellian distributions at different stages of the relaxation process. The values of the time are normalized to the mean free time. The initial data for the problem is a sum of two Maxwellians with dimensionless densities, bulk velocities and temperatures, $n_1 = 1.609$, $\bar{\vec{u}}_1 = (0.775, 0, 0)$, $T_1 = 0.3$ and $n_2 = 2.863$, $\bar{\vec{u}}_2 = (0.436, 0, 0)$, $T_2$ = 0.464, respectively. Columns $n_i$, $T_i$, $\bar{u}_i$, $\bar{v}_i$, and $\bar{w}_i$, $i = 1, 2$ show the density, temperature, and the components of the bulk velocity of the approximating Maxwellians, respectively; $\varepsilon_{L^1}$ is the $L^1$ norm error given by (21)

$t/\tau$	$n_1$	$T_1$	$\bar{u}_{1}$	$\bar{v}_{1}$	$\bar{w}_{1}$	$n_2$	$T_2$	$\bar{u}_{2}$	$\bar{v}_{2}$	$\bar{w}_{2}$	$\varepsilon_{L_1}$
0.0	1.54	0.30	0.78	1.1E-2	3.6E-4	2.94	0.46	0.44	-7.4E-3	-2.9E-3	8.5E-3
0.04	1.66	0.30	0.77	1.9E-3	-7.4E-3	2.81	0.47	0.43	-2.5E-4	9.7E-4	8.3E-3
0.45	3.31	0.46	0.48	5.8E-4	1.2E-3	1.16	0.29	0.78	6.3E-3	4.7E-4	7.8E-3
0.87	1.19	0.30	0.76	1.9E-3	1.0E-2	3.28	0.45	0.48	3.3E-4	-2.0E-3	5.8E-3
1.7	1.30	0.33	0.69	6.6E-4	1.5E-2	3.17	0.46	0.50	6.6E-4	-8.9E-3	7.8E-3
2.9	2.93	0.45	0.51	-3.7E-3	7.8E-3	1.54	0.37	0.65	4.5E-3	-9.3E-3	7.1E-3
3.8	1.29	0.36	0.65	-1.6E-2	-1.2E-2	3.18	0.44	0.52	9.0E-3	5.2E-3	6.9E-3
4.2	3.53	0.44	0.53	-4.0E-3	-3.2E-3	0.94	0.36	0.66	1.2E-2	1.3E-2	5.9E-3
4.5	2.21	0.44	0.52	-1.2E-2	-2.7E-2	2.26	0.40	0.60	1.3E-2	2.8E-2	7.0E-3
6.2	2.34	0.42	0.53	2.6E-2	-8.2E-3	2.13	0.42	0.59	-2.7E-2	9.9E-3	6.1E-3
7.9	1.58	0.42	0.53	4.4E-2	-4.3E-2	2.89	0.42	0.57	-2.4E-2	2.3E-2	4.6E-3

| Show Table

DownLoad: CSV

Table 5. Results of approximating the solution to the problem of spatially homogeneous relaxation by a sum of three Maxwellian distributions at different stages of the relaxation process. The initial data for the problem is a sum of two Maxwellians with dimensionless densities, bulk velocities and temperatures, $n_1 = 1.609$, $\bar{\vec{u}}_1 = (0.775, 0, 0)$, $T_1 = 0.3$ and $n_2 = 2.863$, $\bar{\vec{u}}_2 = (0.436, 0, 0)$, $T_2$ = 0.464, respectively

$t/\tau$	$n_1$	$T_1$	$\bar{u}_1$	$\bar{v}_1$	$n_2$	$T_2$	$\bar{u}_2$	$\bar{v}_2$	$n_3$	$T_3$	$\bar{u}_3$	$\bar{v}_3$	$\varepsilon_{L_1}$
0.0	1.35	0.48	0.38	-2.8E-2	1.51	0.44	0.48	3.0E-3	1.61	0.30	0.78	-5.9E-3	6.6E-3
.04	1.97	0.43	0.51	1.0E-2	1.17	0.48	0.38	-1.4E-2	1.33	0.29	0.79	-9.6E-3	6.2E-3
.45	1.86	0.33	0.73	3.6E-3	0.68	0.47	0.45	-4.9E-2	1.93	0.47	0.43	1.3E-2	7.9E-3
.87	0.89	0.46	0.40	-9.0E-2	2.47	0.44	0.53	3.9E-2	1.11	0.31	0.75	-1.1E-2	8.9E-3
1.7	0.92	0.29	0.73	5.5E-3	0.68	0.44	0.52	1.0E-1	2.87	0.45	0.51	-2.7E-2	7.0E-3
2.9	1.59	0.42	0.58	1.4E-2	0.92	0.34	0.67	-1.9E-2	1.96	0.45	0.49	5.7E-2	7.0E-3
3.8	2.01	0.39	0.62	-3.0E-3	1.03	0.43	0.52	5.2E-2	1.43	0.45	0.50	-3.0E-2	6.2E-3
4.2	1.64	0.43	0.54	-2.5E-2	1.03	0.38	0.64	-2.7E-2	1.80	0.43	0.52	-6.7E-3	7.6E-3
4.5	1.50	0.45	0.49	-4.1E-2	0.84	0.37	0.65	-4.0E-2	2.13	0.42	0.57	4.2E-2	5.4E-3
6.2	1.90	0.41	0.56	-2.3E-2	1.81	0.44	0.53	4.3E-2	0.76	0.39	0.63	-3.0E-2	8.0E-3
7.9	1.34	0.42	0.60	2.9E-4	2.27	0.42	0.57	-2.3E-2	0.84	0.41	0.47	6.1E-2	5.9E-3

| Show Table

DownLoad: CSV

Related Papers

Cited by

References

[1]	A. Alekseenko and C. Euler, A Bhatnagar-Gross-Krook kinetic model with velocity-dependent collision frequency and corrected relaxation of moments, Continuum Mechanics and Thermodynamics, 28 (2016), 751-763. doi: 10.1007/s00161-014-0407-0.
[2]	A. Alekseenko and E. Josyula, Deterministic solution of the Boltzmann equation using a discontinuous Galerkin velocity discretization, in 28th International Symposium on Rarefied Gas Dynamics, 9-13 July 2012, Zaragoza, Spain, AIP Conference Proceedings, American Institute of Physics, 2012, 8pp.
[3]	A. Alekseenko and E. Josyula, Deterministic solution of the spatially homogeneous Boltzmann equation using discontinuous Galerkin discretizations in the velocity space, emphJournal of Computational Physics, 272 (2014), 170–188, URL http://www.sciencedirect.com/science/article/pii/S0021999114002186. doi: 10.1016/j.jcp.2014.03.031.
[4]	L. Andallah and H. Babovsky, A discrete Boltzmann equation based on a cub-octahedron in $\mathbb{R}^3$, SIAM Journal on Scientific Computing, 31 (2009), 799-825. doi: 10.1137/060673850.
[5]	E. Anderson, Z. Bai, C. Bischof, S. Blackford, J. Demmel, J. Dongarra, J. Du Croz, A. Greenbaum, S. Hammarling, A. McKenney and D. Sorensen, LAPACK Users' Guide, 3rd edition, Society for Industrial and Applied Mathematics, Philadelphia, PA, 1999.
[6]	V. V. Aristov, Direct Methods for Solving the Boltzmann Equation and Study of Nonequilibrium Flows, Fluid Mechanics and Its Applications, Kluwer Academic Publishers, 2001. doi: 10.1007/978-94-010-0866-2.
[7]	V. V. Aristov and S. A. Zabelok, A deterministic method for the solution of the Boltzmann equation with parallel computations, Zhurnal Vychislitel'noi Tekhniki i Matematicheskoi Physiki, 42 (2002), 425-437.
[8]	H. Babovsky, Kinetic models on orthogonal groups and the simulation of the Boltzmann equation, AIP Conference Proceedings, 1084 (2008), 415–420, URL http://scitation.aip.org/content/aip/proceeding/aipcp/10.1063/1.3076513.
[9]	P. L. Bhatnagar, E. P. Gross and M. Krook, A model for collision processes in gases. I. Small amplitude processes in charged and neutral one-component systems, Phys. Rev., 94 (1954), 511-525.
[10]	G. A. Bird, Molecular Gas Dynamics and the Direct Simulation of Gas Flows, Oxford Engineering Science Series, Oxford University Press, New York, USA, 1995.
[11]	C. M. Bishop, Neural Networks for Pattern Recognition, Advanced Texts in Econometrics, Clarendon Press, 1995.
[12]	A. V. Bobylev and S. Rjasanow, Difference scheme for the Boltzmann equation based on the fast Fourier transform., European Journal of Mechanics -B/Fluids, 16 (1997), 293-306.
[13]	A. V. Bobylev and S. Rjasanow, Fast deterministic method of solving the Boltzmann equation for hard spheres, European Journal of Mechanics -B/Fluids, 18 (1999), 869–887, URL http://www.sciencedirect.com/science/article/pii/S0997754699001211. doi: 10.1016/S0997-7546(99)00121-1.
[14]	I. D. Boyd, Vectorization of a Monte Carlo simulation scheme for nonequilibrium gas dynamics, Journal of Computational Physics, 96 (1991), 411–427, URL http://www.sciencedirect.com/science/article/pii/002199919190243E.
[15]	J. M. Burt, E. Josyula and I. D. Boyd, Novel Cartesian implementation of the direct simulation Monte Carlo method, Journal of Thermophysics and Heat Transfer, 26 (2012), 258-270.
[16]	L. Devroye, General principles in random variate generation, in Non-Uniform Random Variate Generation, Springer New York, 1986, 27–82. doi: 10.1007/978-1-4613-8643-8_2.
[17]	G. Dimarco and L. Pareschi, Numerical methods for kinetic equations, Acta Numerica, 23 (2014), 369-520. doi: 10.1017/S0962492914000063.
[18]	I. D. Dinov, Expectation maximization and mixture modeling tutorial, in Statistics Online Computational Resource, UCLA: Statistics Online Computational Resource, 2008, URL http://escholarship.org/uc/item/1rb70972.
[19]	F. Filbet and C. Mouhot, Analysis of spectral methods for the homogeneous Boltzmann equation, Transactions of the American Mathematical Society, 363 (2011), 1947–1980, URL http://www.jstor.org/stable/41104652. doi: 10.1090/S0002-9947-2010-05303-6.
[20]	F. Filbet, C. Mouhot and L. Pareschi, Solving the Boltzmann equation in ${N}\log_2{N}$, SIAM Journal on Scientific Computing, 28 (2006), 1029-1053. doi: 10.1137/050625175.
[21]	F. Filbet, L. Pareschi and T. Rey, On steady-state preserving spectral methods for homogeneous Boltzmann equations, Comptes Rendus Mathematique, 353 (2015), 309–314, URL http://www.sciencedirect.com/science/article/pii/S1631073X15000412. doi: 10.1016/j.crma.2015.01.015.
[22]	E. Fonn, P. Grohs and R. Hiptmair, Hyperbolic cross approximation for the spatially homogeneous Boltzmann equation, IMA Journal of Numerical Analysis, 35 (2015), 1533-1567. doi: 10.1093/imanum/dru042.
[23]	R. O. Fox and P. Vedula, Quadrature-based moment model for moderately dense polydisperse gas-particle flows, Industrial and Engineering Chemistry Research, 49 (2010), 5174-5187. doi: 10.1021/ie9013138.
[24]	I. M. Gamba and S. H. Tharkabhushanam, Spectral-Lagrangian methods for collisional models of non-equilibrium statistical states, J. Comput. Phys., 228 (2009), 2012-2036. doi: 10.1016/j.jcp.2008.09.033.
[25]	I. M. Gamba and S. H. Tharkabhushanam, Shock and boundary structure formation by spectral-lagrangian methods for the inhomogeneous Boltzmann transport equation, Journal of Computational Mathematics, 28 (2010), 430-460. doi: 10.4208/jcm.1003-m0011.
[26]	I. M. Gamba and C. Zhang, A conservative discontinuous Galerkin scheme with $O(n^2)$ operations in computing Boltzmann collision weight matrix, in 29th International Symposium on Rarefied Gas Dynamics, July 2014, China, AIP Conference Proceedings, American Institute of Physics, 2014, 8pp.
[27]	B. I. Green and P. Vedula, Validation of a collisional lattice Boltzmann method, in 20th AIAA Computational Fluid Dynamics Conference, 27-30 June 2011, Honolulu Hawaii, AIP Conference Proceedings, American Institute of Physics, 2011, 14pp.
[28]	L. H. Holway, New statistical models for kinetic theory: Methods of construction, Phys. Fluids, 9 (1966), 1658-1673. doi: 10.1063/1.1761920.
[29]	M. Ivanov, A. Kashkovsky, S. Gimelshein, G. Markelov, A. Alexeenko, Y. Bondar, G. Zhukova and S. Nikiforov, SMILE system for 2D/3D DSMC computations, in 25th International Symposium on Rarefied Gas Dynamics, 21–28 July 2006, St. Petersburg, Russia, AIP Conference Proceedings, Publishing House of the Siberian Branch of the Russian Academy of Sciences, Novosibirsk, Russia, 2007, 8pp.
[30]	D. Kalman, A singularly valuable decomposition: The SVD of a matrix, College Math Journal, 27 (1996), 2-23.
[31]	R. Kirsch and S. Rjasanow, A weak formulation of the Boltzmann equation based on the Fourier transform, Journal of Statistical Physics, 129 (2007), 483-492. doi: 10.1007/s10955-007-9374-1.
[32]	Y. Y. Kloss, F. G. Tcheremissine and P. V. Shuvalov, Solution of the Boltzmann equation for unsteady flows with shock waves in narrow channels, Computational Mathematics and Mathematical Physics, 50 (2010), 1093-1103. doi: 10.1134/S096554251006014X.
[33]	R. Larsen, PROPACK: Computing the singular value decomposition of large and sparse or structured matrices. Computer software, 2005.
[34]	C. D. Levermore, Moment closure hierarchies for kinetic theories, Journal of Statistical Physics, 83 (1996), 1021-1065. doi: 10.1007/BF02179552.
[35]	A. Majorana, A numerical model of the Boltzmann equation related to the discontinuous Galerkin method, Kinetic and Related Models, 4 (2011), 139-151. doi: 10.3934/krm.2011.4.139.
[36]	C. Mouhot and L. Pareschi, Fast algorithms for computing the Boltzmann collision operator, Mathematics of Computation, 75 (2006), 1833–1852, URL http://www.jstor.org/stable/4100126. doi: 10.1090/S0025-5718-06-01874-6.
[37]	A. Munafò, J. R. Haack, I. M. Gamba and T. E. Magin, A spectral-lagrangian Boltzmann solver for a multi-energy level gas, Journal of Computational Physics, 264 (2014), 152–176, URL http://www.sciencedirect.com/science/article/pii/S0021999114000631. doi: 10.1016/j.jcp.2014.01.036.
[38]	A. Narayan and A. Klöckner, Deterministic numerical schemes for the Boltzmann equation, arXiv: 0911.3589.
[39]	V. A. Panferov and A. G. Heintz, A new consistent discrete-velocity model for the Boltzmann equation, Mathematical Methods in the Applied Sciences, 25 (2002), 571-593. doi: 10.1002/mma.303.
[40]	L. Pareschi and B. Perthame, A Fourier spectral method for homogeneous Boltzmann equations, Transport Theory and Statistical Physics, 25 (1996), 369-382. doi: 10.1080/00411459608220707.
[41]	C. R. Schrock and A. W. Wood, Convergence of a distributional Monte Carlo method for the Boltzmann equation, Advances in Applied Mathematics and Mechanics, 4 (2012), 102-121. doi: 10.4208/aamm.10-m11113.
[42]	C. R. Schrock and A. W. Wood, Distributional Monte Carlo solution technique for rarefied gasdynamics, Journal of Thermophysics and Heat Transfer, 26 (2012), 185-189.
[43]	N. Selden, C. Ngalande, N. Gimelshein, S. Gimelshein and A. Ketsdever, Origins of radiometric forces on a circular vane with a temperature gradient, Journal of Fluid Mechanics, 634 (2009), 419–431, URL http://journals.cambridge.org/article_S0022112009007976. doi: 10.1017/S0022112009007976.
[44]	E. M. Shakhov, Approximate kinetic equations in rarefied gas theory, Fluid Dynamics, 3 (1968), 112-115. doi: 10.1007/BF01016254.
[45]	E. M. Shakhov, Generalization of the Krook kinetic relaxation equation, Fluid Dynamics, 3 (1968), 95-96. doi: 10.1007/BF01029546.
[46]	K. Stephani, D. Goldstein and P. Varghese, Generation of a hybrid DSMC/CFD solution for gas mixtures with internal degrees of freedom, in 50th AIAA Aerospace Sciences Meeting including the New Horizons Forum and Aerospace Exposition, Aerospace Sciences Meetings, American Institute of Aeronautics and Astronautics, 2012, p648. doi: 10.2514/6.2012-648.
[47]	H. Struchtrup, Macroscopic Tansport Equations for Rarefied Gas Flows. Approximation Methods in Kinetic Theory, Interaction of Mechanics and Mathematics Series, Springer, Heidelberg, 2005.
[48]	S. Succi, The Lattice Boltzmann Equation: For Fluid Dynamics and Beyond, Numerical Mathematics and Scientific Computation, Clarendon Press, Oxford, 2013, URL https://books.google.com/books?id=OC0Sj_xgnhAC.
[49]	F. G. Tcheremissine, Solution to the Boltzmann kinetic equation for high-speed flows, Computational Mathematics and Mathematical Physics, 46 (2006), 315-329.
[50]	F. G. Tcheremissine, Method for solving the Boltzmann kinetic equation for polyatomic gases, Computational Mathematics and Mathematical Physics, 52 (2012), 252-268. doi: 10.1134/S0965542512020054.
[51]	V. A. Titarev, Efficient deterministic modelling of three-dimensional rarefied gas flows, Communications in Computational Physics, 12 (2012), 162-192. doi: 10.4208/cicp.220111.140711a.
[52]	L. Wu, C. White, T. J. Scanlon, J. M. Reese and Y. Zhang, Deterministic numerical solutions of the Boltzmann equation using the fast spectral method, Journal of Computational Physics, 250 (2013), 27–52, URL http://www.sciencedirect.com/science/article/pii/S0021999113003276.
[53]	L. Wu, J. Zhang, J. M. Reese and Y. Zhang, A fast spectral method for the Boltzmann equation for monatomic gas mixtures, Journal of Computational Physics, 298 (2015), 602– 621, URL http://www.sciencedirect.com/science/article/pii/S0021999115004167. doi: 10.1016/j.jcp.2015.06.019.