Discrete Velocity Half-Space Problems

• Discrete velocity half-space problems are boundary value problems in kinetic theory defined on semi-infinite domains with a finite set of velocities, modeling interface and coupling phenomena.
• The analysis employs a damping modification and even-odd decomposition to achieve unique solvability and exponential decay, enhancing stability in numerical simulations.
• Galerkin and spectral discretization methods yield rapid convergence and enable precise extraction of macroscopic parameters, such as extrapolation lengths in networked domains.

Discrete velocity half-space problems are boundary value problems for kinetic equations defined on semi-infinite domains, with a finite set of discrete velocities. These problems serve as model settings for kinetic-fluid coupling and as local descriptions of interface and boundary layers in multiscale transport phenomena, especially where the kinetic regime transitions to macroscopic models, such as hydrodynamic or wave equations. Discrete velocity half-space formulations are foundational for developing accurate numerical methods and rigorous coupling conditions in kinetic theory and network dynamics.

1. Mathematical Formulation of Discrete Velocity Half-Space Problems

In the discrete velocity context, the velocity variable $v$ takes values in a finite set $\{v^{(i)}\}_{i=1}^M$. The standard stationary (time-independent) half-space equation is: $$v_1\,\partial_x f(x,v) + L\,f(x,v) = 0, \qquad x > 0, \; v \in \{v^{(i)}\}$$ where $f(x,v)$ is the kinetic density, $L$ is an $M\times M$ self-adjoint collision matrix (non-negative with finite-dimensional null space), and the domain is $x > 0$. The problem is closed with incoming boundary data at $x=0$ for all $v_1 > 0$,

$f(0,v) = \phi(v), \quad v_1 > 0$

and the requirement that $f(x,\cdot)\to f_\infty$, an equilibrium in $\text{Null}\,L$, as $x\to\infty$. This setting emerges naturally in the description of boundary/interface layers for kinetic models, with $L$ typically derived from linearized BGK, radiative transfer, or linearized Boltzmann-type operators (Li et al., 2014, Borsche et al., 4 Dec 2025).

A variant appears in networked domains, where discrete velocity BGK models on graph edges with velocities $\{v_i\}$ are studied. The kinetic equation in the Knudsen regime is

$$v_i\,\frac{d\varphi_i}{dx} = -\big(\varphi_i - M_i[\varphi]\big), \qquad x \in [0,\infty), \; i=1,\dots,2N$$

where $M_i[\varphi]$ is a Maxwellian moment projection, $\rho$ and $q$ the zeroth and first moments. For networks, additional matching and conservation conditions arise at graph nodes (Borsche et al., 4 Dec 2025).

2. Analytical Approach: Damping, Parity Decomposition, and Well-Posedness

Direct analysis of the undamped half-space problem lacks coercivity due to the presence of collision invariants (null-modes of $L$), which renders well-posedness proofs and numerical stability challenging. The methodology introduced by Qin–Lu–Sun establishes a three-step analytic foundation (Li et al., 2014):

1. Damping: Augmenting the collision operator $L$ to $L_d$ by terms that penalize all null-space components, using an $L$-orthonormal decomposition of $\text{Null}\,L$ into subspaces sorted by sign of $v_1$-moment. Explicitly,

$$L_d f = Lf + \alpha \sum_{\ell}(v_1 X_\ell)(v_1 X_\ell,f) + \cdots$$

where $\alpha > 0$ small.

1. Even-Odd (Parity) Decomposition: Introducing $f^\pm(v) = \frac12(f(v) \pm f(-v))$ to separate parity sectors, facilitating the identification of decaying and non-decaying modes across $x$.
2. Inf-Sup and Weak Formulation: The resulting damped problem is posed on a Hilbert space $\Gamma$ of functions with square-integrable parity and discrete derivatives, with the weak form ensuring boundedness and unique solvability by a Babuška–Lax–Milgram argument. The solution to the damped system decays exponentially as $x \to \infty$.

This framework provides a rigorous path to well-posedness, overcoming the spectral degeneracy inherent in the kinetic half-space setting.

3. Galerkin and Spectral Discretization Methods

Numerical resolution of discrete velocity half-space problems is based on Galerkin-spectral approaches that leverage parity-adapted basis expansions in velocity:

• Galerkin Method: Select a finite-dimensional subspace $\Gamma_N$ of parity-adapted basis functions (e.g., half-range Hermite or Legendre polynomials). The solution is expanded as $f_N(x,v) = \sum_k a_k(x)\psi_k(v)$.
• System Reduction: Projecting the weak form onto $\Gamma_N$, one obtains a system of ODEs in $x$ for the modal coefficients $(a_k(x))$. Boundary constraints at $x=0$ are enforced by upwind trace terms and by imposing orthogonality to positive generalized eigenmodes of the transport-collision ODE system.
• Spectral Implementation: For bounded velocity models (using Gauss–Legendre nodes and orthogonal polynomials), modal decompositions yield block systems for discrete moments, which are solved via eigen-decomposition and Gaussian elimination. The full coupling at network nodes reduces to block-Toeplitz systems, streamlining the extraction of macroscopic invariants and coupling coefficients (Borsche et al., 4 Dec 2025).

The table below summarizes key aspects of the two main spectral approaches:

Feature	Qin–Lu–Sun (Li et al., 2014)	Borsche–Damm–Klar–Zhou (Borsche et al., 4 Dec 2025)
Basis in $v$	Half-range Hermite/Legendre	Legendre polynomials, Gauss nodes
Damping	Explicit construction $L_d$	Not required, but moments are stable
Core unknowns	Parity-adapted coefficients	Moment vectors $G = S\varphi$
Node coupling (nets)	N/A	Block-Toeplitz / algebraic invariance

In both settings, quasi-optimality is established: the discrete numerical error is controlled by best approximation error in the chosen modal space.

4. Extraction of Macroscopic Coupling: Extrapolation Lengths and Network Applications

In networked kinetic models, the interface (half-space) problem yields crucial macroscopic parameters—most notably, the extrapolation length $\delta$—that enter the coupling conditions for macroscopic wave equations. The half-space solution near a node, analyzed in moment variables, leads generically to a linear invariant

$$\rho_0^i(0) + \delta q_0^i(0) = \text{constant across edges}\,,$$

together with total flux conservation $\sum_i q_0^i(0) = 0$ and prescribed outgoing Riemann data. The coefficient $\delta$ encapsulates the effect of the kinetic boundary/interface layer and is determined numerically via the spectral decomposition.

For the discrete velocity BGK model on a three-edge node with bounded velocities in $[-1,1]$, one obtains $\delta(\infty) = 0.7307$ as $N \to \infty$. This quantity is essential for accurate macroscopic closures and for ensuring the equivalence between kinetic and effective macroscopic descriptions in the hydrodynamic limit (Borsche et al., 4 Dec 2025).

5. Numerical Performance, Accuracy, and Implementation Considerations

The Galerkin-spectral methods for discrete velocity half-space problems exhibit rapid convergence. Typically, 10–20 half-range modes suffice to reach $10^{-6}$ accuracy in representative benchmarks (Li et al., 2014). The main practical considerations include:

• Basis Selection: Maxwellian kernels favor Hermite basis; transport kernels prefer Legendre basis on $[0,1]$ or $[-1,1]$.
• Quadrature: Gaussian quadrature of high order is used to compute moments and matrix entries, avoiding aliasing.
• Eigenvalue Problems: Generalized eigenproblems identifying decaying modes can be solved offline and reused across multiple boundary data sets.
• Oscillation Control: Gibbs phenomena due to discontinuity at $v_1=0$ are moderated by spectral/cosine filters or post-processing via Gegenbauer reprojection.
• Network Nodes: In the network context, the spectral machinery is essential in extracting the macroscopic coupling coefficients efficiently and reliably (Borsche et al., 4 Dec 2025).

6. Significance and Broader Applications

Discrete velocity half-space problems are central in multiscale modeling at interfaces and boundaries where kinetic and macroscopic models are coupled, such as in rarefied gas dynamics, radiative transfer, and gas network simulations. Their analysis and numerical resolution provide the foundation for:

• Rigorous derivation of macroscopic coupling conditions in kinetic-fluid and kinetic-wave systems, especially via asymptotic analysis in the small Knudsen number limit (Borsche et al., 4 Dec 2025).
• High-accuracy simulation of boundary layers, with quasi-optimal error control.
• Transfer of kinetic interface information to effective macroscopic models, including computation of extrapolation lengths and invariant relations.

Numerical experiments on both single half-space and full network configurations confirm that these spectral methods achieve spectral convergence rates for the kinetic-to-macroscopic interface coupling and robustly reproduce macroscopic limits up to $O(\varepsilon)$, where $\varepsilon$ is the Knudsen number (Borsche et al., 4 Dec 2025).