betaFeedback

WorksPhysics and Astronomy

Linear regularized 13-moment equations with Onsager boundary conditions for general gas molecules

Abstract

We develop the steady-state regularized 13-moment equations in the linear regime for rarefied gas dynamics with general collision models. For small Knudsen numbers, the model is accurate up to the super-Burnett order, and the resulting system of moment equations is shown to have a symmetric structure. We also propose Onsager boundary conditions for the moment equations that guarantees the stability of the equations. The validity of our model is verified by benchmark examples for the one-dimensional channel flows.

Introduction

The modeling and computation of rarefied gas dynamics has been a classical research topic in the history of fluid mechanics. On one hand, significant progress has been made in the development of efficient solvers for the Boltzmann equation [11, 14, 17, 29, 20]; On the other hand, many researchers still try to avoid the high computational cost and find extensions of classical fluid models such as Euler equations and Navier-Stokes equations and hope that the new models are capable of describing the motion of moderately rarefied gases. The moment method introduced by Grad [13] is one of the important approaches in this direction. While Grad’s original method suffers from a number of deficiencies such as loss of hyperbolicity and convergence [19, 5, 8], many new ideas have been proposed in the recent years to improve its robustness and make moment methods more widely applicable [22, 7, 3, 24, 12, 21].

In this work, we will study the regularized version of Grad’s 13-moment equations, which are originally proposed in [26] for Maxwell molecules and extended to general gases in [27, 7]. In the literature, this model is called R13 equations for short. The R13 equations for Maxwell molecules have been verified for a variety of problems [30, 16, 33, 9, 10]. Recently, attentions have been drawn to the study of Onsager boundary conditions, which guarantees the stability of simulations of boundary value problems [1, 23]. Here we will consider the formulation of Onsager boundary conditions for general R13 equations with arbitrary elastic collision models. The linearized Boltzmann equation will be taken as the base model, so that rigorous theory of stabilization can be established for the moment equations.

In the linearized setting, one typical form of the moment equations is as follows:

A0∂u∂t+∑j=13Aj∂u∂xj=Lu,(1)A_0\frac{\partial u}{\partial t}+\sum_{j=1}^{3}A_j\frac{\partial u}{\partial x_j}=Lu, \tag*{(1)}

where A0A_0 is symmetric positive semidefinite, and LL is symmetric negative semidefinite. The matrices AjA_j, j=1,2,3j=1,2,3 are symmetric. Here we allow A0A_0 to have zero eigenvalues so that the form (1) can also cover parabolic equations such as the R13 equations. For problems with unbounded domains or periodic boundary conditions, one can show that

ddt∫uTA0u dx=∫uTLu dx≤0,\frac{d}{dt}\int u^T A_0u\,\mathrm{d}x=\int u^T Lu\,\mathrm{d}x\leq0,

indicating the L2L^2 stability. For problems on bounded domains, additional conditions on the boundary conditions are suggested in [23, 35, 4] to preserve the L2L^2 stability. Assume that n=(n1,n2,n3)Tn=(n_1,n_2,n_3)^T is the

Key words and phrases. Regularized 13-moment equations, super-Burnett order, Onsager boundary conditions.

Zhenning Cai’s work was supported by the Academic Research Fund of the Ministry of Education of Singapore under Grant No. A-0004592-00-00. outer unit normal vector on the boundary point. The conditions are based on the following structures of the matrices AjA_j:

P(n1A1+n2A2+n3A3)PT=(0AoeAeo0),(2)\mathbf{P}(n_1\mathbf{A}_1+n_2\mathbf{A}_2+n_3\mathbf{A}_3)\mathbf{P}^{\mathsf{T}}= \begin{pmatrix} 0 & \mathbf{A}_{\mathrm{oe}}\\ \mathbf{A}_{\mathrm{eo}} & 0 \end{pmatrix}, \tag*{(2)}

where P\mathbf{P} is an orthogonal matrix converting the moments u\mathbf{u} to another set of moments, each of which is either odd or even in the normal direction, and we choose the permutation such that

Pu=(uoddueven).\mathbf{P}\mathbf{u}= \begin{pmatrix} \mathbf{u}_{\mathrm{odd}}\\ \mathbf{u}_{\mathrm{even}} \end{pmatrix}.

The moments in uodd\mathbf{u}_{\mathrm{odd}} include quantities that changes sign when the frame of reference changes by flipping the normal vector n\mathbf{n}, and the moments in ueven\mathbf{u}_{\mathrm{even}} remain unchanged under this transformation. By the symmetry of Aj\mathbf{A}_j, we have Aoe=AeoT\mathbf{A}_{\mathrm{oe}}=\mathbf{A}_{\mathrm{eo}}^{\mathsf{T}}. The structure (2) comes from the fact that the normal flux of an odd moment is an even moment, and the normal flux of an even moment is an odd moment. With this structure, if Aoe\mathbf{A}_{\mathrm{oe}} has full row rank, the L2L^2 stable boundary conditions have the form

uodd=QAoe(ueven−gext),(3)\mathbf{u}_{\mathrm{odd}}=\mathbf{Q}\mathbf{A}_{\mathrm{oe}}(\mathbf{u}_{\mathrm{even}}-\mathbf{g}_{\mathrm{ext}}), \tag*{(3)}

where Q\mathbf{Q} is a positive semidefinite matrix, gext\mathbf{g}_{\mathrm{ext}} refers to the external source coming from the boundary of the domain. Such boundary conditions are known as Onsager boundary conditions. The particular form of boundary conditions that has odd moments on the left-hand side originates from Grad’s work [13], which ensures the continuity of boundary conditions with respect to the accommodation coefficient. However, for a specific moment system, choosing uodd\mathbf{u}_{\mathrm{odd}} to be all the odd moments may result in a rank-deficient Aoe\mathbf{A}_{\mathrm{oe}}, so that (3) will provide too many boundary conditions. In this work, we will encounter such a situation during our derivation, requiring us to adjust the selections of uodd\mathbf{u}_{\mathrm{odd}} and ueven\mathbf{u}_{\mathrm{even}} to restore the surjective property of Aoe\mathbf{A}_{\mathrm{oe}}. The structure of Onsager boundary conditions is useful not only for time-dependent problems. For steady-state problems (time derivative removed in (1)), Onsager boundary conditions can provide a symmetric weak form, which helps develop the theory of well-posedness and the finite element methods [32].

For the regularized 13-moment equations for Maxwell molecules, the boundary conditions with this particular structure have been devised in [35, 32]. However, the technique to derive the Onsager boundary conditions cannot be directly generalized to the R13 equations for more general molecules. In this work, we will reconsider the derivation of the linearized steady-state R13 equations for general collision models, from which we will show clearly how the structure (2) is built into the derivation of moment equations, and thus the boundary conditions in the form (3) can be naturally obtained.

In the following section, we will review the moment method for the linear Boltzmann equation and the asymptotic expansions of the moments when the Knudsen number is small. Our main results are presented in Section 3, where the explicit forms of the linear R13 equations and the Onsager boundary conditions are provided. The derivation of the R13 equations and boundary conditions are respectively given in Section 4 and 5. In Section 6, we verify the accuracy of our model by one-dimensional channel problems. A brief conclusion is given in Section 7.

Review of the moment equations and asymptotic properties of moments

We consider the steady-state linear Boltzmann equation

ξj∂f∂xj=1KnL[f](4)\xi_j\frac{\partial f}{\partial x_j}=\frac{1}{\mathrm{Kn}}\mathcal{L}[f] \tag*{(4)}

where f(x,ξ)f(\mathbf{x},\boldsymbol{\xi}) is the distribution function, x=(x1,x2,x3)\mathbf{x}=(x_1,x_2,x_3) denotes the position and ξ=(ξ1,ξ2,ξ3)\boldsymbol{\xi}=(\xi_1,\xi_2,\xi_3) stands for the velocity of the gas molecules. We apply Einstein’s summation convention throughout this work, meaning that when an index appears twice in the same term, the expression represents the sum of this term with this index running from 1 to 3. For example, in (4),

ξj∂f∂xj=ξ1∂f∂x1+ξ2∂f∂x2+ξ3∂f∂x3.\xi_j\frac{\partial f}{\partial x_j}=\xi_1\frac{\partial f}{\partial x_1}+\xi_2\frac{\partial f}{\partial x_2}+\xi_3\frac{\partial f}{\partial x_3}.

For indices whose ranges are not from 1 to 3, the summation symbol will be written explicitly. On the right-hand side of (4), the constant Kn is the Knudsen number characterizing how rarefied the gas is, and L[f]\mathcal{L}[f] is the linearized Boltzmann collision operator. Below we will introduce the general moment equations for the linearized Boltzmann equation and the asymptotic expansion of the moments in the case of a small Knudsen number.

Series expansion and moment equations

2.1. Series expansion and moment equations. Following [34], we expand the distribution function into an infinite series:

f(x,ξ)=∑l=0+∞∑m=0+∞(2l+1)!!l!wi1⋯ilm(x)ψi1⋯ilm(ξ).(5)f(x,\xi)=\sum_{l=0}^{+\infty}\sum_{m=0}^{+\infty}\frac{(2l+1)!!}{l!}w_{i_1\cdots i_l}^{m}(x)\psi_{i_1\cdots i_l}^{m}(\xi). \tag*{(5)}

Here, the basis functions ψi1,…,ilm\psi_{i_1,\ldots,i_l}^{m} are defined by

ψi1,…,ilm(ξ)=1(2π)3/2L‾m(l+1/2)(∣ξ∣22)ξ⟨i1⋯il⟩exp⁡(−∣ξ∣22),(6)\psi_{i_1,\ldots,i_l}^{m}(\xi)=\frac{1}{(2\pi)^{3/2}}\overline{L}_{m}^{(l+1/2)}\left(\frac{|\xi|^2}{2}\right)\xi_{\langle i_1\cdots i_l\rangle}\exp\left(-\frac{|\xi|^2}{2}\right), \tag*{(6)}

where ξ⟨i1⋯ξl⟩\xi_{\langle i_1\cdots\xi_l\rangle} is the trace-free part of the tensor ξi1⋯ξil\xi_{i_1}\cdots\xi_{i_l} (see [28 Appendix A]), and L‾n(l+1/2)\overline{L}_{n}^{(l+1/2)} is the normalized Laguerre polynomial

L‾n(l+1/2)(x)=π2l+1n!Γ(n+l+3/2)x−(l+1/2)(ddx−1)nxn+l+1/2.(7)\overline{L}_{n}^{(l+1/2)}(x)=\sqrt{\frac{\sqrt{\pi}}{2^{l+1}n!\Gamma(n+l+3/2)}}x^{-(l+1/2)}\left(\frac{\mathrm{d}}{\mathrm{d}x}-1\right)^{n}x^{n+l+1/2}. \tag*{(7)}

In literature, people usually define the Maxwellian

fM(ξ)=1(2π)3/2exp⁡(−∣ξ∣22),f_M(\xi)=\frac{1}{(2\pi)^{3/2}}\exp\left(-\frac{|\xi|^2}{2}\right),

so that ψi1,…,ilm(ξ)=L‾m(l+1/2)(∣ξ∣2/2)ξ⟨i1⋯il⟩fM(ξ)\psi_{i_1,\ldots,i_l}^{m}(\xi)=\overline{L}_{m}^{(l+1/2)}(|\xi|^2/2)\xi_{\langle i_1\cdots i_l\rangle}f_M(\xi). Compared with the classical series expansion by Grad [13], the Maxwellian fMf_M is a global equilibrium state due to our linearized setting. This expansion requires us to assume that the distribution function is defined in the following Hilbert space:

L2(R3,[fM(ξ)]−1 dξ):={f∣∫R3[f(ξ)]2fM(ξ) dξ<+∞},L^2(\mathbb{R}^3,[f_M(\xi)]^{-1}\,\mathrm{d}\xi):=\left\{f\mid\int_{\mathbb{R}^3}\frac{[f(\xi)]^2}{f_M(\xi)}\,\mathrm{d}\xi<+\infty\right\},

so that all the moments of the distribution function can be properly defined. The inner product of this Hilbert space is

⟨g1,g2⟩:=∫R3g1(ξ)g2(ξ)fM(ξ) dξ,\langle g_1,g_2\rangle:=\int_{\mathbb{R}^3}\frac{g_1(\xi)g_2(\xi)}{f_M(\xi)}\,\mathrm{d}\xi,

so that we can express the coefficients wi1⋯ilmw_{i_1\cdots i_l}^{m} as the moments of the distribution function:

wi1⋯iln(x)=⟨f,ψi1⋯ilm⟩.(8)w_{i_1\cdots i_l}^{n}(x)=\langle f,\psi_{i_1\cdots i_l}^{m}\rangle. \tag*{(8)}

In particular, we would like to highlight the relationship between these coefficients and the quantities in Grad’s 13-moment equations:

References

  1. [1]Alexander Felix Beckmann, Anirudh Singh Rana, Manuel Torrilhon, and Henning Struchtrup. Evaporation boundary conditions for the linear r13 equations based on the Onsager theory. Entropy, 20(9):680, 2018.DOI
  2. [2]G. A. Bird. Molecular Gas Dynamics and the Direct Simulation of Gas Flows. Oxford: Clarendon Press, 1994.
  3. [3]Niclas Böhmer and Manuel Torrilhon. Entropic quadrature for moment approximations of the Boltzmann-BGK equation. Journal of Computational Physics, 401:108992, 2020.DOI
  4. [4]Jonas Bünger, Edilbert Christruraj, Andrea Hanke, and Manuel Torrilhon. Structured derivation of moment equations and stable boundary conditions with an introduction to symmetric, trace-free tensors. Kinetic and Related Models, 16(3):458–494, 2023.DOI
  5. [5]Z. Cai, Y. Fan, and R. Li. On hyperbolicity of 13-moment system. Kinet. Relat. Mod., 7(3):415–432, 2014.arxiv.org/abs/1401.7523
  6. [6]Z. Cai and M. Torrilhon. Approximation of the linearized Boltzmann collision operator for hard-sphere and inverse-power-law models. J. Comput. Phys., 295:617–643, 2015.DOI
  7. [7]Z. Cai and Y. Wang. Regularized 13-moment equations for inverse power law models. J. Fluid Mech., 894:A12, 2020.DOI
  8. [8]Zhenning Cai and Manuel Torrilhon. On the Holway-Weiss debate: Convergence of the Grad-moment-expansion in kinetic gas theory. Physics of Fluids, 31(12):126105, 2019.
  9. [9]Rory Claydon, Abhay Shrestha, Anirudh S. Rana, James E. Sprittles, and Duncan A. Lockerby. Fundamental solutions to the regularised 13-moment equations: efficient computation of three-dimensional kinetic effects. Journal of Fluid Mechanics, 833:R4, 2017.DOI
  10. [10]Thomas C. De Fraja, Anirudh S. Rana, Ryan Enright, Laura J. Cooper, Duncan A. Lockerby, and James E. Sprittles. Efficient moment method for modeling nanoporous evaporation. Phys. Rev. Fluids, 7(2):024201, 2022.DOI
  11. [11]Giacomo Dimarco, Raphaël Loubère, Jacek Narski, and Thomas Rey. An efficient numerical method for solving the Boltzmann equation in multidimensions. Journal of Computational Physics, 353:46–81, 2018.arxiv.org/abs/1608.08009
  12. [12]Rodney O. Fox and Frédérique Laurent. Hyperbolic quadrature method of moments for the one-dimensional kinetic equation. SIAM Journal on Applied Mathematics, 82(2):750–771, 2022.
  13. [13]H. Grad. On the kinetic theory of rarefied gases. Comm. Pure Appl. Math., 2(4):331–407, 1949.DOI
  14. [14]J. Hu and K. Qi. A fast Fourier spectral method for the homogeneous Boltzmann equation with non-cutoff collision kernels. J. Comput. Phys., 423:109806, 2020.DOI
  15. [15]Z. Hu, S. Yang, and Z. Cai. Flows between two parallel plates: Analytical solutions of regularized 13-moment equations for inverse-power-law models. Phys. Fluids, 32(12):122007, 2020.DOI
  16. [16]I. E. Ivanov, I. A. Kryukov, and M. Yu. Timokhin. Application of moment equations to the mathematical simulation of gas microflows. Comput. Math. Math. Phys., 53:1534–1550, 2013.DOI
  17. [17]Chang Liu, Yajun Zhu, and Kun Xu. Unified gas-kinetic wave-particle methods I: Continuum and rarefied gas flow. Journal of Computational Physics, 401:108977, 2020.arxiv.org/abs/1811.07141
  18. [18]J. C. Maxwell. On stresses in rarefied gases arising from inequalities of temperature. Proc. R. Soc. Lond., 27(185–189):304–308, 1878.
  19. [19]I. Müller and T. Ruggeri. Rational Extended Thermodynamics, Second Edition, volume 37 of Springer tracts in natural philosophy. Springer-Verlag, New York, 1998.
  20. [20]Lorenzo Pareschi and Thomas Rey. Moment preserving Fourier-Galerkin spectral methods and application to the Boltzmann equation. SIAM Journal on Numerical Analysis, 60(6):3216–3240, 2022.arxiv.org/abs/2105.13158
  21. [21]Teddy Pichard. A moment closure based on a projection on the boundary of the realizability domain: Extension and analysis. Kinetic and Related Models, 15(5):793–822, 2022.DOI
  22. [22]A. S. Rana, V. K. Gupta, and H. Struchtrup. Coupled constitutive relations: a second law based higher-order closure for hydrodynamics. Proc. Roy. Soc. A, 474:20180323, 2018.DOI
  23. [23]N. Sarna and M. Torrilhon. On stable wall boundary conditions for the Hermite discretization of the linearised Boltzmann equation. J. Stat. Phys., 170:101–126, 2018.DOI
  24. [24]Neeraj Sarna. A positive and stable L2-minimization based moment method for the boltzmann equation of gas dynamics. Journal of Computational Physics, 440:110428, 2021.DOI
  25. [25]H. Struchtrup. Stable transport equations for rarefied gases at high orders in the Knudsen number. Phys. Fluids, 16(11):3921–3934, 2004.DOI
  26. [26]H. Struchtrup and M. Torrilhon. Regularization of Grad’s 13 moment equations: Derivation and linear analysis. Phys. Fluids, 15(9):2668–2680, 2003.DOI
  27. [27]H. Struchtrup and M. Torrilhon. Regularized 13 moment equations for hard sphere molecules: Linear bulk equations. Phys. Fluids, 25:052001, 2013.DOI
  28. [28]Henning Struchtrup. Macroscopic transport equations for rarefied gas flows. In Macroscopic transport equations for rarefied gas flows, pages 145–160. Springer, 2005.
  29. [29]Wei Su, Lianhua Zhu, and Lei Wu. Fast convergence and asymptotic preserving of the general synthetic iterative scheme. SIAM Journal on Scientific Computing, 42(6):B1517–B1540, 2020.arxiv.org/abs/2003.09958
  30. [30]Peyman Taheri and Henning Struchtrup. Effects of gas reflection in microflows between coaxial cylinders. Phys. Rev. E, 80(6):066317, 2009.DOI
  31. [31]Peyman Taheri, Manuel Torrilhon, and Henning Struchtrup. Couette and Poiseuille microflows: Analytical solutions for regularized 13-moment equations. Phys. Fluids, 21(1):017102, 2009.DOI
  32. [32]Lambert Theisen and Manuel Torrilhon. FenicsR13: A tensorial mixed finite element solver for the linear r13 equations using the FEniCS computing platform. ACM Trans. Math. Softw., 47(2), 2021.DOI
  33. [33]M. Yu. Timokhin, H. Struchtrup, A. A. Kochkhanhik, and Ye. A. Bondar. Different variants of R13 moment equations applied to the shock-wave structure. Physics of Fluids, 29(3):037105, 2017.
  34. [34]M. Torrilhon. Convergence study of moment approximations for boundary value problems of the Boltzmann-BGK equation. Comm. Comput. Phys., 18(3):529–557, 2018.
  35. [35]Manuel Torrilhon and Neeraj Sarna. Hierarchical boltzmann simulations and model error estimation. Journal of Computational Physics, 342:66–84, 2017.DOI

Paper details

vlnc:W_kc7yqhab
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents