betaFeedback

WorksMathematics

Stability analysis of an extended quadrature method of moments for kinetic equations

Abstract

This paper performs a stability analysis of a class of moment closure systems derived with an extended quadrature method of moments (EQMOM) for the one-dimensional BGK equation. The class is characterized with a kernel function. A sufficient condition on the kernel is identified for the EQMOM-derived moment systems to be strictly hyperbolic. We also investigate the realizability of the moment method. Moreover, sufficient and necessary conditions are established for the two-node systems to be well-defined and strictly hyperbolic, and to preserve the dissipation property of the kinetic equation.

Introduction

By describing the evolution of problem-specific distribution functions, the kinetic models are founded with a solid basis in a wide range of complex interacting systems. For instance, the celebrated Boltzmann equation governs the distribution of the molecular velocity and is believed to better characterize the rarefied flows where their hydrodynamic counterparts, being Euler and/or Navier-Stokes equations, become less reliable [12]. However, solving the kinetic equation is challenging. For one thing, the binary collision operator of the Boltzmann equation causes quadratic costs while treating the velocity dependence. A widely-accepted measure is to apply the BGK operator which models the collision as a relaxation process towards the local equilibrium [2]. This model not only reduces the computational costs, but also has the desired conservation laws and an HH-theorem characterizing the dissipation properties due to collisions [12].

On the other hand, the high dimension of the phase space raises significant difficulties in computation, even for the BGK equation. Several deterministic methods have thus been developed [9, 18], including the discrete velocity model, the spectral methods, and various methods of moments, to remove the dependence upon the molecular velocity and deduce spatial-time models of macroscopic variables from the kinetic equation. This paper focuses on the method of moments. In the moment system, only the lower-order moments have clear physical interpretations after being related to the density, mean velocity and temperature (internal energy) of the system [12]. The higher-order moments may provide additional information beyond the classical hydrodynamic models.

All moment systems need to be closed, which is mostly done by reconstructing the distribution from the transported moments [16]. The well-known Grad’s 13-moment theory was established based on a linear expansion of the Maxwellian [11], but far away from equilibrium, this ansatz can lead to negative values of distributions. In contrast, the quadrature-based method of moments [17] was proposed with nonlinear reconstruction of the distribution function, and such a treatment seems to be suitable for non-equilibrium flows. However, approximating the distribution with a linear combination of multiple Dirac δ\delta-functions with unknown centers [17], the method (called QMOM) fails in the simulation of BGK equation due to the occurrence of singularity [10].

This highlights the importance of understanding the mathematical properties of the moment closure systems, which are usually first-order PDEs. For real-world physical models, the PDE is expected to be hyperbolic so that the system is robust against small perturbations of the initial data [22]. Indeed, the unphysical behaviors of the Grad’s theory and the QMOM approach can both be attributed to the lack of hyperbolicity [3, 5]. Let us mention some efforts to achieve hyperbolic regulation of the Grad’s theory [3, 23] and other moment closure systems [15]. Furthermore, a thorough stability analysis should as well account for the source term of the model, as the Boltzmann and BGK collisions are both featured with HH-theorems [12]. For the moment closure system, it is believed that the structural stability condition proposed in [25] for hyperbolic relaxation systems is a proper characterization of the dissipation property. The condition specifies how the source term should be coupled with the hyperbolic part in the vicinity of the equilibrium. Admitting such a structure, the resultant moment system is compatible with the classical theories [25]. Recently, it was shown that the structural stability condition is fulfilled by many moment closure systems, including the hyperbolic regularization models of rarefied gases [8, 28] and a series of hyperbolic shallow water moment models [14].

The objective of this paper is to investigate the stability properties of the extended QMOM (termed EQMOM) for the BGK equation. In EQMOM, the velocity distribution is reconstructed as a sum of multiple continuous kernel functions instead of the δ\delta-function [4, 27]. A list of kernels that can be used for the purpose of EQMOM is summarized in [20]. If the kernel is the Gaussian distribution (denoted Gaussian-EQMOM), it was found that the method is well-defined (namely, the unclosed terms can be uniquely determined), and the resultant moment system respects the structural stability condition [13]. However, little is known for other types of kernels, which may be suitable for different scenarios. For instance, the space plasmas follow kappa distributions with high energy tails deviated from a Maxwellian [19]. We also remark that the EQMOM with the beta function as the kernel is used to treat a simplified radiative transfer equation [1].

This paper deals with the EQMOM induced by a univariate kernel function (see (4)) and considers the spatial one-dimensional (1-D) BGK equation. Surprisingly, it is found that not all kernels share the nice properties of the Gaussian kernel. As the main result of this paper, we reveal the exact constraints on the kernel such that the two-node EQMOM (n=2n = 2) is well-defined, that the resultant moment system is strictly hyperbolic, and that the system properly respects the dissipation property (as judged by the structural stability condition). These constraints are inequalities of the moments of the kernel function, and are thus easy to check for specific kernels. Moreover, a sufficient condition is established for the nn-node EQMOM-derived moment system to be strictly hyperbolic. It is worth mentioning that, in our argument, the realizable moment set can be characterized for general n≥1n \ge1, which essentially contains the results in [6] (Proposition 3.1 for the two-node Gaussian-EQMOM) as a special case. This leads to a polynomial which is an answer to the question raised in [16] (footnote 16 on p.88). Furthermore, we present abundant examples of the kernel functions to which our theory can be applied in a straightforward manner.

The remainder of the paper is organized as follows. Our main results are presented in § 2. Section 3 is devoted to studying the realizability of the extended quadrature method of moments. Hyperbolicity of the resultant moment systems is analyzed in § 4. For two-node systems, the structural stability condition is verified in § 5. § 6 presents a number of specific kernel functions and investigates their behaviors numerically. The conclusions are given in § 7.

Preliminaries and main results

Consider the hypothetical 1-D BGK equation for the distribution f=f(t,x,ξ)f = f(t,x,\xi) with time t∈R+t \in\mathbb{R}_{+}, position x∈Rx \in\mathbb{R} and velocity ξ∈R\xi\in\mathbb{R}:

∂tf+ξ∂xf=1τ(feq−f),feq=feq(ρ,U,θ;ξ)=ρ2πθexp⁡(−(ξ−U)22θ).(1)\begin{aligned} \partial_t f + \xi\partial_x f &= \frac{1}{\tau}\left(f^{eq} - f\right), \\ f^{eq} = f^{eq}(\rho,U,\theta;\xi) &= \frac{\rho}{\sqrt{2\pi\theta}}\exp\left(-\frac{(\xi-U)^2}{2\theta}\right). \tag*{(1)} \end{aligned}

Here τ\tau is a relaxation time. In the local Maxwellian equilibrium feqf^{eq}, the density ρ\rho, mean velocity UU and temperature θ\theta are determined as the velocity moments of ff:

ρ=∫Rf dξ,ρU=∫Rξf dξ,ρθ+ρU2=∫Rξ2f dξ.\rho= \int_{\mathbb{R}} f\,\mathrm{d}\xi,\qquad\rho U = \int_{\mathbb{R}} \xi f\,\mathrm{d}\xi,\qquad\rho\theta+ \rho U^2 = \int_{\mathbb{R}} \xi^2 f\,\mathrm{d}\xi.

Define the jjth velocity moments of ff as

Mj=Mj(t,x)=∫Rξjf dξM_j=M_j(t,x)=\int_{\mathbb{R}}\xi^j f\,\mathrm{d}\xi

for j∈Nj\in\mathbb{N}. The evolution equation for MjM_j can be immediately derived from (1) as

∂tMj+∂xMj+1=1τ(ρΔjeq(U,θ)−Mj).(2)\partial_t M_j+\partial_x M_{j+1}=\frac{1}{\tau}\left(\rho\Delta_j^{eq}(U,\sqrt{\theta})-M_j\right). \tag*{(2)}

with

Δjeq(u,σ)=∫Rξj12πσexp⁡(−(ξ−u)22σ2) dξ.\Delta_j^{eq}(u,\sigma)=\int_{\mathbb{R}}\xi^j\frac{1}{\sqrt{2\pi}\sigma}\exp\left(-\frac{(\xi-u)^2}{2\sigma^2}\right)\,\mathrm{d}\xi.

Notice that the first nn equations for M0,…,Mn−1M_0,\ldots,M_{n-1} contain the term ∂xMn\partial_xM_n. Therefore, any finite truncation of the above equations leads to an unclosed system, and a closure procedure is required. In this paper, we are concerned with an extended quadrature method of moments (EQMOM) [4, 27].

Extended quadrature method of moments (EQMOM)

Let K=K(ξ)≥0\mathcal{K}=\mathcal{K}(\xi)\geq0 satisfy

mj:=∫RξjK(ξ) dξ<∞,∀j∈N,andm0=1.(3)\mathfrak{m}_j:=\int_{\mathbb{R}}\xi^j\mathcal{K}(\xi)\,\mathrm{d}\xi<\infty,\quad\forall j\in\mathbb{N},\quad\text{and}\quad\mathfrak{m}_0=1. \tag*{(3)}

In EQMOM, the distribution ff is approximated with the following ansatz

f(ξ)=∑i=1nwiσK(ξ−uiσ)(4)f(\xi)=\sum_{i=1}^{n}\frac{w_i}{\sigma}\mathcal{K}\left(\frac{\xi-u_i}{\sigma}\right) \tag*{(4)}

with the weights wiw_i, nodes uiu_i and ‘width’ σ>0\sigma>0 to be determined. To do this, the first (2n+1)(2n+1) lower-order moments are employed:

Mj=∑i=1nwiΔj(ui,σ),withΔj(u,σ):=∫Rξj1σK(ξ−uσ) dξ(5)M_j=\sum_{i=1}^{n}w_i\Delta_j(u_i,\sigma),\quad\text{with}\quad\Delta_j(u,\sigma):=\int_{\mathbb{R}}\xi^j\frac{1}{\sigma}\mathcal{K}\left(\frac{\xi-u}{\sigma}\right)\,\mathrm{d}\xi \tag*{(5)}

for j=0,…,2nj=0,\ldots,2n. This defines a map M:=(M0,…,M2n)T=M(W)\mathcal{M}:=(M_0,\ldots,M_{2n})^{T}=\mathcal{M}(W) for W=(w1,u1,…,wn,un,σ)T∈R2n+1W=(w_1,u_1,\ldots,w_n,u_n,\sigma)^{T}\in\mathbb{R}^{2n+1} with σ>0\sigma>0. Here the superscript ‘TT’ denotes the transpose of a vector or matrix.

Suppose M\mathcal{M} is injective on a certain domain W∈Ω⊂R2n+1W\in\Omega\subset\mathbb{R}^{2n+1}. Then for any M∈G:=M(Ω)M\in G:=\mathcal{M}(\Omega), there exists a unique W=M−1(M)W=\mathcal{M}^{-1}(M) solving (5). In this way, the EQMOM is well-defined and the next moment M2n+1M_{2n+1} can be evaluated as a function of MM:

M2n+1=M2n+1(M):=∑i=1nwiΔ2n+1(ui,σ).(6)M_{2n+1}=\mathcal{M}_{2n+1}(M):=\sum_{i=1}^{n}w_i\Delta_{2n+1}(u_i,\sigma). \tag*{(6)}

Consequently, the first (2n+1)(2n+1) equations in (2) are closed as a system of first-order PDEs:

∂tM+∂x(M1,…,M2n+1)T=1τ(ρΔeq(U,θ)−M).(7)\partial_tM+\partial_x(M_1,\ldots,M_{2n+1})^{T}=\frac{1}{\tau}\left(\rho\Delta^{eq}(U,\sqrt{\theta})-M\right). \tag*{(7)}

Here Δeq(U,θ)=(Δ0eq(U,θ),…,Δ2neq(U,θ))T∈R2n+1\Delta^{eq}(U,\sqrt{\theta})=(\Delta^{eq}_0(U,\sqrt{\theta}),\ldots,\Delta^{eq}_{2n}(U,\sqrt{\theta}))^{T}\in\mathbb{R}^{2n+1} with ρ=M0\rho=M_0, ρU=M1\rho U=M_1 and ρ(θ+U2)=M2\rho(\theta+U^2)=M_2.

The main goal of this paper is to investigate the injectivity of the map M\mathcal{M} in (5) for the general kernel K(ξ)\mathcal{K}(\xi), where the injectivity is closely related to the realizability of moments. Such analyses are useful to design efficient and robust algorithms to solve WW from (5). Moreover, we analyze hyperbolicity of the moment closure system (7) and its dissipation property inherited from the the HH-theorem of the kinetic equation.

Remark 2.1. For the sake of simplicity, throughout this paper we assume that the kernel function K(ξ)\mathcal{K}(\xi) is normalized, in the sense that m0=1\mathfrak{m}_0=1, m1=0\mathfrak{m}_1=0 and m2=1\mathfrak{m}_2=1. In fact, a general K(ξ)\mathcal{K}(\xi) can be normalized as K+(ξ)=hK(hξ+ξ0)\mathcal{K}^{+}(\xi)=h\mathcal{K}(h\xi+\xi_0) with ξ0=m1\xi_0=\mathfrak{m}_1 and h=m2−m12h=\sqrt{\mathfrak{m}_2-\mathfrak{m}_1^2} (note that m2≥m12\mathfrak{m}_2\geq\mathfrak{m}_1^2 due to the Cauchy-Schwartz inequality). Moreover, the map M2n+1(M)\mathcal{M}_{2n+1}(M) in (6) is the same as that derived from K(ξ)\mathcal{K}(\xi), so the moment closure system (7) is unchanged with the normalization.

Structural stability condition

For smooth solutions, the balance laws (7) can be written as

∂tM+A(M)∂xM=S(M):=1τ(ρΔeq(U,θ)−M)(8)\partial_t M + A(M)\partial_x M = S(M) := \frac{1}{\tau}\left(\rho\Delta^{eq}(U,\sqrt{\theta})-M\right) \tag*{(8)}

with coefficient matrix

References

  1. [1]G. W. Alldredge, R. Li, W. Li, Approximating the m-2 method by the extended quadrature method of moments for radiative transfer in slab geometry, Kinet. Relat. Models, 9 (2016) 237–249.DOI
  2. [2]P. L. Bhatnagar, E. P. Gross, 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.DOI
  3. [3]Z. Cai, Y. Fan, R. Li, Globally hyperbolic regularization of Grad’s moment system, Commun. Pur. Appl. Math., 67 (2013) 464–518.arxiv.org/abs/1203.0376
  4. [4]C. Chalons, R. O. Fox, M. Massot, A multi-Gaussian quadrature method of moments for gas-particle flows in a LES framework, in Proceedings of the Summer Program 2010, Center for Turbulence Research, Stanford University, Stanford, CA, (2010) 347–358.
  5. [5]C. Chalons, D. Fox, F. Lauret, M. Massot, Beyond pressureless gas dynamics: Quadrature-based velocity moment models, Commun. Math. Sci., 10 (2012) 1241–1272.arxiv.org/abs/1011.2974
  6. [6]C. Chalons, R. Fox, F. Lauret, M. Massot, A. Vié, Multivariate Gaussian extended quadrature method of moments for turbulent disperse multiphase flow, Multiscale Model. Simul. 15 (2017) 1553–1583.DOI
  7. [7]Y. Cheng, J.A. Rossmanith, A class of quadrature-based moment-closure methods with application to the Vlasov-Poisson-Fokker-Planck system in the high-field limit, J. Comput. Appl. Math., 262 (2014) 384–398.arxiv.org/abs/1212.4026
  8. [8]Y. Di, Y. Fan, R. Li, L. Zheng, Linear stability of hyperbolic moment models for Boltzmann equation, Numer. Math. Theor. Meth. Appl., 10 (2017) 255–277.arxiv.org/abs/1609.03669
  9. [9]G. Dimarco, L. Pareschi, Numerical methods for kinetic equations, Acta Numerica, 23 (2014) 369–520.DOI
  10. [10]R. Fox, A quadrature-based third-order moment method for dilute gas-particle flows, J. Comput. Phys. 227 (2008) 6313–6350.DOI
  11. [11]H. Grad, On the kinetic theory of rarefied gases, Comm. Pure Appl. Math., 2 (1949), 331–407.DOI
  12. [12]S. Harris, An Introduction to the Theory of the Boltzmann Equation, Dover Publications, New York, 2004.
  13. [13]Q. Huang, S.Q. Li, W.-A. Yong, Stability analysis of quadrature-based moment methods for kinetic equations, SIAM J. Appl. Math. 80 (2020) 206–231.arxiv.org/abs/1812.08320
  14. [14]Q. Huang, J. Koellermeier, W.-A. Yong, Equilibrium stability analysis of hyperbolic shallow water moment equations, Math. Meth. Appl. Sci. 45 (2022) 6459–6480.arxiv.org/abs/2011.08571
  15. [15]J. Koellermeier and M. Rominger, Analysis and numerical simulation of hyperbolic shallow water moment equations, Commun. Comp. Phys., 28 (2020), 1038–1084.DOI
  16. [16]D. L. Marchisio, R. O. Fox, Computational Models for Polydisperse Particulate and Multiphase Systems, Cambridge University Press, Cambridge, 2013.DOI
  17. [17]R. McGraw, Description of aerosol dynamics by the quadrature method of moments, Aerosol Sci. Technol., 27 (1997), 255–265.DOI
  18. [18]L. Mieussens, Discrete velocity model and implicit scheme for the BGK equation of rarefied gas dynamics, Math. Mod. Meth. Appl. Sci., 10 (2000) 1121–1149.DOI
  19. [19]V. Pierrard and M. Lazar, Kappa Distributions: Theory and Applications in Space Plasma, Sol. Phys., 267 (2010) 153–174.DOI
  20. [20]M. Pigou, J. Morchain, P. Fede, M. I. Pinet, G. Laronze, New developments of the Extended Quadrature Method of Moments to solve Population Balance Equations, J. Comput. Phys., 365 (2018) 243–268.DOI
  21. [21]K. Schmüdigen, The Moment Problem. Graduate Texts in Mathematics, vol 277, Springer, Cham, 2017.
  22. [22]D. Serre, Systems of Conservation Laws 1: Hyperbolicity, Entropy, Shock Waves, Cambridge University Press, Cambridge, 1999.
  23. [23]H. Struchtrup and M. Torrilhon, Regularization of Grad’s 13 moment equations: Derivation and linear analysis, Phys. Fluids, 15 (2003) 2668–2680.DOI
  24. [24]E. F. Toro, Riemann Solvers for Fluid Dynamics, 3rd edition, Springer, New York, 2009.
  25. [25]W.-A. Yong, Singular perturbations of first-order hyperbolic systems with stiff source terms, J. Differ. Equations, 155 (1999) 89–132.DOI
  26. [26]W.-A. Yong, Basic Aspects of Hyperbolic Relaxation Systems, in Advances in the Theory of Shock Waves, Birkhäuser Boston, Boston, MA, (2001) 259–305.DOI
  27. [27]C. Yuan, F. Laurent, R.O. Fox, An extended quadrature method of moments for population balance equations, J. Aerosol Sci., 51 (2012) 1–23.
  28. [28]W. Zhao, W.-A. Yong, L.-S. Luo, Stability analysis of a class of globally hyperbolic moment systems, Commun. Math. Sci., 15 (2017) 609–633.DOI

Paper details

vlnc:W_h6zrhehq
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.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents