Source-linked AI summary
Spectral - Lagrangian methods for Collisional Models of Non - Equilibrium Statistical States
Irene M. Gamba, Sri Harsha Tharkabhushanam
TL;DR
Stochastic Boltzmann simulation can make evolving non-equilibrium distributions costly or unreliable, while existing deterministic approaches have limitations for non-conservative interactions. This paper introduces a Fourier-based spectral solver with Lagrangian moment constraints, and reports close agreement with available analytical results, including an explicit self-similar benchmark.
Problem
Stochastic simulation can produce fluctuations that make evolving distributions costly or unreliable in non-stationary or non-equilibrium states.
Source: §1 · p. 2Method
The solver evaluates the collision operator in Fourier space and enforces required moment conservation through Lagrangian constraints.
Source: §1 · p. 4Results
Computed solutions are reported to be very close to analytical results, including an explicit self-similar solution for a Maxwell-type slow-down model.
Source: §1 · p. 5, §6 · p. 36Takeaways & Limitations
The method computes the distribution function itself rather than only averages and is reported to handle elastic and inelastic VHP interactions without additional modification.
Source: §6 · pp. 36–37Convergence and error results for the Fourier Transform Lagrangian method are still being developed locally in time.
Source: §1 · p. 5
Abstract
from arXiv · showhide
We propose a new spectral Lagrangian based deterministic solver for the non-linear Boltzmann Transport Equation for Variable Hard Potential (VHP) collision kernels with conservative or non-conservative binary interactions. The method is based on symmetries of the Fourier transform of the collision integral, where the complexity in its computing is reduced to a separate integral over the unit sphere $S^2$. In addition, the conservation of moments is enforced by Lagrangian constraints. The resulting scheme, implemented in free space is very versatile and adjusts in a very simple manner, to several cases that involve energy dissipation due to local micro-reversibility (inelastic interactions) or elastic model of slowing down process. Our simulations are benchmarked with the available exact self-similar solutions, exact moment equations and analytical estimates for homogeneous Boltzmann equation for both elastic and inelastic VHP interactions. Benchmarking of the simulations involves the selection of a time self-similar rescaling of the numerical distribution function which is performed using the continuous spectrum of the equation for Maxwell molecules as studied first in \cite{asympGranular} and generalized to a wide range of related models in \cite{BCG}. The method also produces accurate results in the case of inelastic diffusive Boltzmann equations for hard-spheres (inelastic collisions under thermal bath), where overpopulated non-Gaussian exponential tails have been conjectured in computations by stochastic methods in \cite{vanNoi-Ernst, ernst-brito, MSS, GRW04} and rigourously proven in \cite{diffGran} and \cite{highEnergyTails}.
1 Introduction
The introduction motivates deterministic Boltzmann solvers by the limits of stochastic simulation and the restricted scope of existing deterministic approaches. It presents a Fourier-based, Lagrangian-constrained method for elastic and inelastic VHP interactions and benchmarks it against analytical behavior.
1 Introduction
- DSMC has statistical fluctuations that can make evolving probability distributions costly or unreliable to resolve in non-stationary or non-equilibrium states.The introduction notes that DSMC can converge for classical monoatomic rarefied gases, but accurate high-velocity tails remain an active concern.Source: §1 · p. 2
- Earlier deterministic spectral solvers were largely restricted to elastic conservative interactions, while some lacked momentum and energy conservation and could produce negative distributions.The authors identify loss of positivity as a shortcoming that their proposed method does not remove.Source: §1 · pp. 3–4
- The method combines Fourier-transform evaluation of the collision operator with Lagrange-multiplier constraints, supporting elastic, inelastic, and energy-dissipative VHP models.The required moments are enforced through constrained optimization rather than a periodic representation of the distribution.Source: §1 · p. 4
- For general interactions, collision-integral evaluation costs O(N^(2d) log(N) + N^(2d)); symmetry can reduce this to O(N^d log(N)) in specified cases.The reduced order applies when the differential cross section is independent of scattering angle, including hard spheres in d+2 dimensions or Maxwell models in 2D.Source: §1 · pp. 4–5
- Computed dynamically scaled solutions are compared with an explicit self-similar solution for a Maxwell-type slow-down model.That model's analytical behavior includes a Fourier-space formula for particular parameters and power-law high-energy tails.Source: §1 · p. 5
- Convergence and error results for the Fourier Transform Lagrangian method are still being developed locally in time.Optimal complexity for free-space approximation using a non-equispaced FFT is expected, rather than established here.Source: §1 · p. 5
2 Preliminaries
The preliminaries define the collision models and conservation laws, describe heating and cooling settings, and derive a Fourier-space collision representation that separates angular integration from the remaining computation.
2 Preliminaries
- VHP interactions have collision frequency proportional to relative speed raised to a power between zero and one; Maxwell molecules and hard spheres correspond to powers zero and one.Elastic collisions conserve density, momentum, and energy, whereas inelastic collisions conserve density and momentum.Source: §2 · p. 7
2.1 Boltzmann collisional models with heating sources
- Heating-source models add an operator G(f) to the homogeneous Boltzmann equation and can admit stationary states with finite energy.The section considers random heating, dynamically rescaled models, and a thermostat model of slowing down.Source: §2.1 · pp. 9–10
- In the equal-mass Maxwell-type thermostat model, elastic light-particle collisions do not conserve total energy, and cooling-background asymptotics can produce power-law tails and loss of moments over time.The heavy particles form the equilibrium background for the modeled light-particle distribution.Source: §2.1 · pp. 11–12
2.2 Collision Integral Representation
- The Fourier transform of the collision integral isolates dependence on the scattering direction into a separate unit-sphere integral for anisotropic interactions.For isotropic interactions, that angular integral has a closed form.Source: §2.2 · pp. 12–13
- The Fourier representation yields O(N^(2d)) operations, with symmetry reducing the cost to O(N^d log(N)) in particular cases.The stated reduction follows from exploiting symmetric collision kernels.Source: §2.2 · p. 14
3 Numerical Method
The numerical method discretizes velocity and Fourier space, advances the distribution in time, and corrects discretization-induced moment errors through constrained optimization.
3.1 Discretization of the Collision Integral
- The collision-integral discretization assumes a sufficiently large bounded velocity domain, uniform velocity and Fourier grids, and grid spacings satisfying h_v h_ζ = 2π/N.The support assumption is used to avoid mass loss, with the distribution normalized to zero mean velocity.Source: §3.1 · p. 14
3.2 Time Discretization
- Time evolution uses a second-order Runge-Kutta or forward Euler step, with the timestep chosen to correspond dimensionally to 0.1 times the interval between collisions.The interval depends on collision frequency.Source: §3.2 · pp. 14–15
3.3 Conservation Properties - Lagrange Multipliers
- Because discretization can violate the collision operator's conservation laws, the method corrects the computed distribution by constrained Lagrange-multiplier optimization.The constraints are density for the linear operator, density and momentum for inelastic collisions, and density, momentum, and energy for the elastic Boltzmann operator.Source: §3.3 · pp. 15, 17, 18
- The correction can enforce higher-order moments, and the chosen constraints depend on which observables the collision model conserves.The method also uses FFTs and is stated to cover elastic and inelastic collisions for 1/2 ≤ β ≤ 1.Source: §3.3 · pp. 15, 17, 18
4 Self-Similar asymptotics for a general elastic or inelastic BTE of Maxwell type or the cold thermostat problem - power law tails
The section uses self-similar asymptotics as benchmarks for Maxwell-type Boltzmann models, including thermostat and inelastic cases. It describes the limiting distributions and relates their moment bounds to the spectral function μ(p).
4 Self-Similar asymptotics for a general elastic or inelastic BTE of Maxwell type or the cold thermostat problem - power law tails
- The section presents self-similar asymptotics as a benchmark in properly rescaled time for Maxwell-type elastic or inelastic models and the cold-thermostat mixture problem.The spectral function μ(p) is given for the homogeneous elastic or inelastic Maxwell model and the corresponding mixture problem.Source: §4 · p. 18, §4.2 · p. 24
4.1 Self-Similar Solution for a non-negative Thermostat Temperature
- For positive background temperature T, the self-similar state approaches a rescaled Maxwellian with that background temperature.The Fourier-space hot-thermostat solution is obtained from the zero-temperature solution by multiplication by e^-Tx.Source: §4.1 · pp. 19–20
- As the thermostat temperature tends to zero, the distribution develops power-like large-velocity tails and an integral singularity at zero, while retaining finite mass and kinetic temperature.The text characterizes this as overpopulated tails relative to a Maxwellian and infinitely many particles at zero energy.Source: §4.1 · p. 20
- The solver captures the self-similar states with spectral accuracy, and these states are attractors for a large class of initial states.The section presents these simulations as benchmarks for the proposed scheme.Source: §4.1 · p. 20
4.2 Self-Similar asymptotics for a general problem
- For the general Maxwell-type problem, the spectral function determines which moments of the self-similar limit are bounded; when p = 1, the maximal bounded moment order depends on the root μ(p*) = μ(1).Moments of order q ≤ p are bounded for 0 ≤ p ≤ 1; in the finite-energy case, boundedness above order 1 is controlled by p*.Source: §4.2 · p. 23
5 Numerical Results
The numerical section tests the solver across elastic, inelastic, diffusive, and slowing-down models, comparing its results with available analytical moments and solutions. It reports agreement in several benchmark cases alongside behaviors and numerical caveats specific to the models.
5.1 Maxwell type of Elastic Collisions
- For elastic Maxwell molecules, the solver is compared with exact moment evolution, and larger N gives a smaller conservation correction at final time.The computed distribution is also plotted evolving toward a Maxwellian.Source: §5.1 · p. 25
5.2 Maxwell type of Elastic collisions - Bobylev-Krook-Wu (BKW) Solution
- The BKW test compares numerical approximations with the exact solution across discretizations and time steps.Source: §5.2 · p. 27
5.3 Hard-Sphere Elastic Collisions
- Hard-sphere elastic moments evolve toward equilibrium somewhat faster than in the Maxwell-type case, although explicit moment formulas and a solution expression are unavailable.Source: §5.3 · p. 28
5.4 Inelastic Collisions
- For inelastic Maxwell collisions, computed kinetic energy is reported accurate against its analytical evolution even though energy is not imposed as a conservation constraint.The conservation routine constrains density and momentum only.Source: §5.4 · p. 30
5.5 Inelastic Collisions with Diffusion Term
- With diffusion, inelastic Maxwell collisions approach a positive equilibrium temperature; simulations compare Maxwell temperatures with the analytical expression and show hard-sphere asymptotics.Source: §5.5 · p. 31
5.6 Maxwell type of Elastic Collisions - Slow down process problem
- At positive thermostat temperature, the slowing-down model converges to a Maxwellian background and agrees with the analytical self-similar distribution; the zero-temperature power-tail regime occurs only at T = 0.Setting T = 0 directly caused instability, so the simulations instead used a decaying temperature.Source: §5.6 · p. 33
- The authors expect optimal complexity from a non-equispaced FFT for the rescaled Fourier modes, but defer that analysis to future work.Source: §5.6 · p. 33
- For decaying thermostat temperature, the computed distribution develops power-like high-energy tails and deviates substantially from a Maxwellian.Source: §5.6 · p. 34, Figure 13 · p. 35
6 Conclusions and Future Work
The method handles elastic and inelastic variable hard potential interactions and agrees closely with available analytical results. The authors also identify extensions to non-uniform grids and space-inhomogeneous problems as future work.
6 Conclusions and Future Work
- The method works for elastic and inelastic variable hard potential interactions, with computed results close to available analytical benchmarks.The comparisons include elastic moment equations, the BKW self-similar solution, and attracting Bobylev-Cercignani-Gamba self-similar solutions for elastic slowing down.Source: §6 · p. 36
- The collision integral is evaluated using a Fast Fourier Transform, avoiding unit-sphere integration in favor of integration on a regular velocity grid.The method uses a uniform velocity grid; a non-uniform grid is possible, but requires a non-uniform FFT.Source: §6 · p. 36
- Figures 14 and 15 show m_q(t) for T = e^-2t/3.Source: Figure 14 · p. 37, Figure 15 · p. 38
- The scheme uses a second-order Runge-Kutta time integrator and can compute the distribution function itself, rather than only averages.A trapezoidal rule is used for integration, with Gaussian quadrature suggested as an alternative for improved accuracy.Source: §6 · p. 37
- Space-inhomogeneous implementations using splitting of advection and collision components are in development, with practical 1- and 2-D problems proposed as next steps.Examples include shock-tube phenomena and Rayleigh-Benard or Couette flow.Source: §6 · p. 37