Plasma simulations with adaptive mesh and algorithm refinement
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.
Kai: We're looking at the paper: "Plasma simulations with adaptive mesh and algorithm refinement".
Mira: Here's the gist of the paper. Plasma simulations with adaptive mesh and algorithm refinement describes PAMAR, which extends Adaptive Mesh and Algorithm Refinement to weakly ionized plasmas by making each species an independent adaptive-mesh solver with its own phasespace dimensionality and refinement criteria, coupled through a field solver in one executable The paper demonstrates this capability for a planar capacitively coupled discharge computed with 1d1v two-stream kinetic solvers and for a floating sphere in a weakly collisional plasma computed with 1d2v octree solvers.
Kai: PAMAR combines adaptive Cartesian meshes with automatic selection of kinetic or fluid solvers across a computational domain It was introduced for rarefied gas dynamics in the Unified Flow Solver and extended to gas mixtures, particle and lattice-Boltzmann solvers, computational electromagnetics, multi-phase flows, and heterogeneous CPU–GPU architectures.
Mira: The central idea is that each plasma species is described by an independent adaptive-mesh solver with its own phase-space dimensionality, refinement criteria, and time step The physical model for electrons and ions is a build-time switch between a fully kinetic, a reduced-kinetic, and a fluid solver behind one fixed interface.
Kai: Transport Model Hierarchy A hierarchy of transport models describes electrons and ions in a weakly ionized plasma This hierarchy includes kinetic, angular moments with closures (an intermediate kinetic level), and fluid (velocity moments).
Paper summary: Mira: To summarize what they did here, PAMAR describes a plasma extension of their adaptive mesh and algorithm refinement method that allows electrons, ions, neutral species, and electromagnetics to have independent adaptive mesh solvers with their own phase-space dimensionality, refinement criteria, and time steps per solver.
Kai: It claims this is possible even in weakly ionized plasmas where the mass, velocity, and time scale differences between species are huge compared to gas dynamics.
Lev: They show two demonstrations: a transient simulation of a planar capacitively coupled plasma using one-dimensional two-stream kinetic solvers for electrons and ions on independent adaptive quadtrees, and another demonstration with a steady spherical plasma problem using one-dimensional two-velocity mu octree solvers for both species.
Kai: The results they got from these simulations are noise-free kinetic models of a capacitive discharge around a spherical probe, and they also resolve the three electron groups and the different time scales of electron and ion kinetics without any statistical noise.
Mira: They use this framework to show that when you look at specific plasma problems like planar discharges, they can resolve dynamic properties without the statistical noise you often get from particle-in-cell simulations.
Lev: The way they handle solver selection is a hybrid approach: the ion solver keeps the kinetic solver where its own solution shows temperature above one point five times the gas temperature and uses a fluid model elsewhere, while the electron hybrid keeps the two dimensional octree only for escaping electrons in the loss-cone layer and uses a Fokker–Planck model in position and total energy elsewhere.
Kai: That sounds like they are trying to be efficient by only running the most complex solver when it's actually needed, which is smart when you're dealing with computational costs.
Mira: They also mention that for the spherical results, the adaptive ion octree resolves cold ions at a tenth of the uniform mesh cost, and graded meshes improve absorbed flux, collected currents, and density profiles.
Lev: The paper notes that their per-species clocks help accelerate the approach to steady state in both demonstrations.
Kai: It seems like they’ve managed to get some high-resolution kinetic physics in a way that’s computationally tractable for these complex plasmas.
Mira: The paper itself is called "Plasma simulations with adaptive mesh and algorithm refinement," which points to how this methodology extends past its original use in rarefied gas dynamics into the realm of plasma physics.
Conclusion: Kai: So, thinking about the title, "Plasma simulations with adaptive mesh and algorithm refinement," it really tells you that this isn't just a minor tweak; they’re fundamentally changing how we model different physical scales in plasmas.
Mira: I think what matters is that they showed how to apply this complex adaptive meshing to weakly ionized plasmas where the species are so different in mass and time scale, which was a big challenge for their original work.
Lev: The implications are pretty deep because they’ve resolved things like the formation of electron groups and the two time averages of the potential that govern electrons and ions in planar discharges without statistical noise from PIC simulations.
Kai: And for the spherical case, it shows that adaptive ion octrees can resolve cold ions at a fraction of a uniform mesh cost while improving flux profiles.
Mira: The paper suggests that using angular-moment levels is the right tool for efficiently describing electrons in gas-discharge plasmas, which is something they found effective.
Lev: A major limitation they point out is that their GPU path is constrained by the current backend of Basilisk, which only supports Cartesian grids and not the adaptive quadtrees and octrees used by these kinetic solvers.
Kai: And another big one mentioned is that replacing explicit time advance with an implicit scheme for electrons coupled to Poisson's equation would let the coupled system advance on the ion or RF time scale, which they suggest would reduce the cost by several orders of magnitude.
Mira: So, while this method gives them great results for specific geometries like planar discharges and spheres, the current computational path is limited by hardware capabilities and explicit time stepping.
Lev: That makes sense; I'm always looking at how to take these theoretical advancements and make them actually runnable on real hardware with minimal overhead.
University of Alabama in Huntsville · CFD Research Corporation
physics.comp-ph, physics.plasm-ph
Submitted: 2026-11-10
Updated: 2026-11-10
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 79/100
The gist: Plasma simulations with adaptive mesh and algorithm refinement describes PAMAR, which extends Adaptive Mesh and Algorithm Refinement to weakly ionized plasmas by making each species an independent…
Key concepts
- PAMAR
- A simulation framework that adapts the mesh and algorithm independently for each plasma species. Each species gets its own solver based on its phase-space needs, connected by a field solver, allowing for highly detailed modeling of complex plasmas.
- Species-Specific Solvers
- The core idea is treating electrons and ions separately. Each species uses an adaptive mesh and refinement criteria tailored to its specific physics (e.g., kinetic vs. fluid). This allows the simulation to use the most appropriate mathematical model for each particle type.
- Transport Model Hierarchy
- A layered approach describing plasma transport, starting from kinetic models for electrons and ions, moving through angular moments, and ending with fluid equations at larger scales. This hierarchy ensures efficiency by using simpler models when spatial or temporal scales are large enough.
- Hybrid Solvers
- The framework uses different solvers based on local conditions. For example, the ion solver switches between kinetic and fluid models depending on temperature, while the electron solver uses a kinetic octree for escaping electrons and a Fokker-Planck model elsewhere.
Terminology
Summary
Plasma simulations with adaptive mesh and algorithm refinement describes PAMAR, which extends Adaptive Mesh and Algorithm Refinement to weakly ionized plasmas by making each species an independent adaptive-mesh solver with its own phasespace dimensionality and refinement criteria, coupled through a field solver in one executable The paper demonstrates this capability for a planar capacitively coupled discharge computed with 1d1v two-stream kinetic solvers and for a floating sphere in a weakly collisional plasma computed with 1d2v octree solvers. p. 0
Methodology of PAMAR
PAMAR combines adaptive Cartesian meshes with automatic selection of kinetic or fluid solvers across a computational domain It was introduced for rarefied gas dynamics in the Unified Flow Solver and extended to gas mixtures, particle and lattice-Boltzmann solvers, computational electromagnetics, multi-phase flows, and heterogeneous CPU–GPU architectures. The central idea is that each plasma species is described by an independent adaptive-mesh solver with its own phase-space dimensionality, refinement criteria, and time step The physical model for electrons and ions is a build-time switch between a fully kinetic, a reduced-kinetic, and a fluid solver behind one fixed interface. p. 2
Transport Model Hierarchy
A hierarchy of transport models describes electrons and ions in a weakly ionized plasma This hierarchy includes kinetic, angular moments with closures (an intermediate kinetic level), and fluid (velocity moments). The energy moments reduce the distribution further to fluid equations when the spatial and temporal scales exceed the energy-relaxation length For electrons, angular-moment level lies between the Boltzmann and fluid models. p. 4
Demonstrations of Results
The paper presents two demonstrations: a transient simulation of a planar capacitively coupled plasma (CCP) in argon using 1d1v two-stream kinetic solvers for electrons and ions on independent adaptive quadtrees The second demonstration is a steady spherical plasma problem using 1d2v (r, v, μ) octree solvers for both species These simulations resolve the three electron groups and the different time scales of electron and ion kinetics without statistical noise.
Solver Selection and Hybridization
The framework is built on Basilisk [23], a free software system for solving partial differential equations on adaptive Cartesian meshes The hybrid ion solver keeps the kinetic solver where its own solution shows the ion temperature above 1.5 times the gas temperature and uses the fluid model elsewhere. p. 0 A hybrid electron solver keeps the 1d2v octree only for the escaping electrons in the loss-cone layer and uses a Fokker–Planck model in position and total energy elsewhere. p. 0
Conclusions
The CCP results resolve, without the statistical noise of PIC and with RF-phase resolution, the dynamic properties of CCP: the formation of electron groups, the two time-averages of the potential that govern electrons and ions, the secondary peaks of the ion energy distribution, and the transition at 5 MHz to a regime in which the sheath oscillation spans the gap. The spherical results show that the adaptive ion octree resolves the cold ions at a tenth of the uniform-mesh cost, and that graded meshes improve absorbed flux, collected currents, and density profiles. p. 0 p. 0 The per-species clocks accelerate the approach to the steady state. p. 3 p. 3 The fluid ions over-collect by 50–70% in spherical geometry because they carry no angular momentum In the planar discharge, where ions are born at rest and carry no angular momentum, their error comes from the isothermal closure at the sheath edge and the per-cell selection in Section 3.7 confines them to the bulk, where the closure holds. p. 1 p. 1 The hybrid carries a slightly longer tail beyond 170 eV, four orders of magnitude below the peak. p. 6 p. 6 The kinetic solver would then serve as the benchmark for calibration and validation The angular-moment level is the right tool for efficiently describing electrons in gas-discharge plasmas.
Future Work
Future work will focus on parallelizing the species solvers, coupling an implicit electron solver to Poisson's equation, developing a split phase-space formulation for multi-dimensional and highly transient problems, and coupling independent solver instances across material regions for multi-phase plasmas. The GPU path of PAMAR is constrained by Basilisk's current GPU backend, which supports only Cartesian grids and not the adaptive quadtrees and octrees used by the kinetic solvers Replacing explicit time advance with an implicit scheme for electrons coupled to Poisson's solver would let the coupled system advance on the ion or RF time scale, reducing the cost by several orders of magnitude. The hybrid ion solver of Section 3. p. 5 p. 57 also shows where the cost of a transient CCP run goes. p. 0 p. 0 The electron time step remains the main limitation, and the implicit electron solver discussed above is the main target. p. 2 p. 2
Acknowledgments
This work was supported by the NSF EPSCoR FTPP project (Cooperative Agreement OIA-2148653). The authors used Claude Code (Anthropic) to assist with code development, data analysis, and manuscript editing
References
[5] V. I. Kolobov, R. R. Arslanbekov, V. V. Aristov, A. A. Frolova, S. A. Zabelok, Unified solver for rarefied and continuum flows with adaptive mesh and algorithm refinement [6] V. p. 0 I. Kolobov, R. R. Arslanbekov, Towards adaptive kinetic-fluid simulations of weakly ionized plasmas [7] V. p. 1 I. Kolobov, R. R. Arslanbekov, S. Bayyuk, V. V. Aristov, A. A. Frolova, S. A. Zabelok, F G Tcheremissine Unified kinetic approach for simulation of gas flows in rarefied and continuum regimes Final Report AFRL-RBWP-TR-2008-3031 Air Force Research Laboratory Wright-Patterson AFB OH 2007] [8] R. p. 0 R. Arslanbekov V I Kolobov A A Frolova Kinetic solvers with adaptive mesh in phase space Phys Rev E 88 2013 063301 [9] V I Kolobov R R Arslanbekov A A Frolova Solving kinetic equations with adaptive mesh in phase space for rarefied gas dynamics and plasma physics AIP Conf Proc. p. 0 1628 2014 952–959 [10] V I Kolobov R R Arslanbekov D Levko Boltzmann-Fokker-Planck kinetic solver with adaptive mesh in phase space AIP Conf Proc. p. 1 p. 1 2132 2019 060011 [11] V I Kolobov R R Arslanbekov D Levko Kinetic solvers with adaptive mesh in phase space for lowtemperature plasmas J Phys: Conf. p. 0 p. 0 Ser. 1225 2019 012016 [37] N L Aleksandrov A M Konchakov A P Napartovich A N Starostin in B M Smirnov Ed Plasma Chemistry vol. p. 2 p. 2 11 Energoatomizdat Moscow 1984 p. 3 (in Russian) [48] V I Kolobov R R Arslanbekov P Baruval Spherical plasma simulations with adaptive mesh and algorithm refinement in preparation companion paper [48] [49] J G Laframboise Theory of spherical and cylindrical Langmuir probes in a collisionless Maxwellian plasma at rest UTIAS Report No. p. 0 p. 0 100 University of Toronto 1966 [50] S K Thagunna V I Kolobov G P Zank Self-pulsing of dielectric barrier discharges at low driving frequencies Phys. Plasmas 31 2024 053507 [51] V A Godyak R B Piejak Abnormally low electron energy and heating-mode transition in a low-pressure argon rf discharge at 13. p. 4 p. 456 MHz Phys Rev Lett. p. 5 p. 5 65 1990 996–999 [52] J Taillet Resonance-sustained radio frequency discharges Am J Phys. p. 3 p. 3 37 1969 423–441 [53] T Mussenbrock R P Brinkmann Nonlinear electron resonance heating in capacitive radio frequency discharges Appl Phys Lett. p. 6 p. 6 88 2006 151503 [54] U Czarnetzki T Mussenbrock R P Brinkmann Nonlinear electron resonance heating in capacitive radio frequency discharges Phys. p. 6 p. 6 Plasmas 13 2006 123503 [57] Ya L Al'pert A V Gurevich L P Pitaevskii Space Physics with Artificial Satellites Consultants Bureau New York 1965 [58] A V Filippov D Tsendin Electron dynamics in low pressure capacitively coupled radio frequency discharges Plasma Sources Sci. Technol. 4 1995 200–211 [63] V I Kolobov R R Arslanbekov D Levko Electron groups in solar wind and gas discharge plasmas J Phys: Conf. p. 3 p. 3 Ser. 1623 2020 012006 [64] S B Harris J H Paiste T J Holdsworth R R Arslanbekov R P Camata Laser-generated plasmas in length scales relevant for thin film growth and processing simulation and experiment J Phys D: Appl. Phys. 53 2020 015203
Pruning redundant reasoning tokens
The paper describes PAMAR, which extends the Adaptive Mesh and Algorithm Refinement methodology to weakly ionized plasma by making each species an independent adaptive-mesh solver with its own phasespace dimensionality and refinement criteria, coupled through a field solver in one executable.
How the training signal is built
The PAMAR solvers currently run sequentially in a single process, but intra-solver thread parallelism with OpenMP across phase-space grid cells and inter-solver concurrency by advancing the solvers simultaneously between field solves already permits this. The load must be rebalanced after every adaptation, and a kinetic cell costs orders of magnitude more than a fluid one. p. 2
Where the gains come from
The hybrid ion solver keeps the kinetic solver where its own solution shows the ion temperature above 1.5 times the gas temperature and uses the fluid model elsewhere The hybrid electron solver keeps the 1d2v octree only for the escaping electrons in the loss-cone layer and uses a Fokker–Planck model in position and total energy elsewhere The CCP results resolve, without the statistical noise of PIC and with RF-phase resolution, the dynamic properties of CCP: the formation of electron groups, the two time-averages of the potential that govern electrons and ions, the secondary peaks of the ion energy distribution, and the transition at 5 MHz to a regime in which the sheath oscillation spans the gap. p. 0 The spherical results show that the adaptive ion octree resolves the cold ions at a tenth of the uniform-mesh cost, and that graded meshes improve absorbed flux, collected currents, and density profiles. The per-species clocks accelerate the approach to the steady state. The fluid ions over-collect by 50–70% in spherical geometry because they carry no angular momentum In the planar discharge, where ions are born at rest and carry no angular momentum, their error comes from the isothermal closure at the sheath edge and the per-cell selection in Section 3.7 confines them to the bulk, where the closure holds. The hybrid carries a slightly longer tail beyond 170 eV, four orders of magnitude below the peak. The kinetic solver would then serve as the benchmark for calibration and validation The angular-moment level is the right tool for efficiently describing electrons in gas-discharge plasmas.
Future work will focus on parallelizing the species solvers, coupling an implicit electron solver to Poisson's equation, developing a split phase-space formulation for multi-dimensional and highly transient problems, and coupling independent solver instances across material regions for multi-phase plasmas. The GPU path of PAMAR is constrained by Basilisk's current GPU backend, which supports only Cartesian grids and not the adaptive quadtrees and octrees used by the kinetic solvers. Replacing explicit time advance with an implicit scheme for electrons coupled to Poisson's solver would let the coupled system advance on the ion or RF time scale, reducing the cost by several orders of magnitude. The hybrid ion solver of Section 3.7 also shows where the cost of a transient CCP run goes. The electron time step remains the main limitation, and the implicit electron solver discussed above is the main target.
This work was supported by the NSF EPSCoR FTPP project (Cooperative Agreement OIA-2148653) The authors used Claude Code (Anthropic) to assist with code development, data analysis, and manuscript editing.
[5] V. I Kolobov R R Arslanbekov V V Aristov A A Frolova S A Zabelok Unified solver for rarefied and continuum flows with adaptive mesh and algorithm refinement J Comput Phys 223 2007 589–608 [6] V I Kolobov R R Arslanbekov Towards adaptive kinetic-fluid simulations of weakly ionized plasmas J Comput Phys 231 2012 839–869 [7] V I Kolobov R R Arslanbekov S Bayyuk V V Aristov A A Frolova S A Zabelok F G Tcheremissine Unified kinetic approach for simulation of gas flows in rarefied and continuum regimes Final Report AFRL-RBWP-TR-2008-3031 Air Force Research Laboratory Wright-Patterson AFB OH 2007 [8] R R Arslanbekov V I Kolobov A A Frolova Kinetic solvers with adaptive mesh in phase space Phys Rev E 88 2013 063301 [9] V I Kolobov R R Arslanbekov A A Frolova Solving kinetic equations with adaptive mesh in phase space for rarefied gas dynamics and plasma physics AIP Conf Proc. p. 0 1628 2014 952–959 [10] V I Kolobov R R Arslanbekov D Levko Boltzmann-Fokker-Planck kinetic solver with adaptive mesh in phase space AIP Conf Proc. 2132 2019 060011 [11] V I Kolobov R R Arslanbekov D Levko Kinetic solvers with adaptive mesh in phase space for lowtemperature plasmas J Phys: Conf. Ser. 1225 2019 012016 [37] N L Aleksandrov A M Konchakov A P Napartovich A N Starostin in B M Smirnov Ed Plasma Chemistry vol. 11 Energoatomizdat Moscow 1984 p. 3 (in Russian) [48] V I Kolobov R R Arslanbekov P Baruval Spherical plasma simulations with adaptive mesh and algorithm refinement in preparation companion paper [48] [49] J G Laframboise Theory of spherical and cylindrical Langmuir probes in a collisionless Maxwellian plasma at rest UTIAS Report No. 100 University of Toronto 1966 [50] S K Thagunna V I Kolobov G P Zank Self-pulsing of dielectric barrier discharges at low driving frequencies Phys. Plasmas 31 2024 053507 [51] V A Godyak R B Piejak Abnormally low electron energy and heating-mode transition in a low-pressure argon rf discharge at 13.56 MHz Phys Rev Lett. 65 1990 996–999 [52] J Taillet Resonance-sustained radio frequency discharges Am J Phys. 37 1969 423–441 [53] T Mussenbrock R P Brinkmann Nonlinear electron resonance heating in capacitive radio frequency discharges Appl Phys Lett. 88 2006 151503 [54] U Czarnetzki T Mussenbrock R P Brinkmann Nonlinear electron resonance heating in capacitive radio frequency discharges Phys. Plasmas 13 2006 123503 [57] Ya L Al'pert A V Gurevich L P Pitaevskii Space Physics with Artificial Satellites Consultants Bureau New York 1965 [58] A V Filippov D Tsendin Electron dynamics in low pressure capacitively coupled radio frequency discharges Plasma Sources Sci. Technol. 4 1995 200–211 [63] V I Kolobov R R Arslanbekov D Levko Electron groups in solar wind and gas discharge plasmas J Phys: Conf. Ser. 1623 2020 012006
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Adaptive Mesh and Algorithm Refinement (AMAR) extension allows for separate adaptive-mesh solvers for electrons, ions, and fields in one executable with independent phase-space dimensionality, refinement criteria, and time steps. This enables noise-free kinetic models of capacitive discharges by balancing cost and detail between particle-in-cell (PIC) simulations and fast CCP models.
-
Per species selection of the ion solver allows for a hybrid ion solver that keeps the kinetic solver where its own solution shows the ion temperature above 1.5 times the gas temperature and uses the fluid model elsewhere. This mechanism confines ions to regions where their description is accurate, reducing bulk deficit in fluid runs by confining them to where the closure holds.
-
Per-cell selection of electron solver allows for a hybrid electron solver that keeps the one d two v octree only for escaping electrons in the loss-cone layer and uses a Fokker–Planck model in position and total energy elsewhere. This reduces computational cost by 1.6 to 5.3 times compared to running the all-kinetic solution across the entire domain.
-
The PAMAR architecture can resolve three electron groups, the two time-averages of the potential that govern electrons and ions, secondary peaks of the ion energy distribution, and transition regimes in frequency dependence for capacitively coupled plasmas without statistical noise from PIC.
-
The system can be parallelized using intra-solver thread parallelism across phase-space grid cells and inter-solver concurrency by advancing solvers simultaneously between field solves. This allows for distributed memory MPI scaling if moment-integration functions include global reductions to handle domain decomposition transparently via Basilisk's infrastructure.
Related papers
- Gradient-Based Inverse Design of Free-Energy Landscapes with Diffusion Models
- Mixture of Polyconvex Neural Potentials for Parametric Hyperelasticity: Towards Foundation Material Models
- Development and large-scale benchmarks of a protein--ligand absolute binding free energy toolkit
- Spin disorder competing with positional symmetry breaking governs the metal-insulator behavior in oxide paramagnets
- Investigation of the Asymptotic Properties of Active Impedance in Large Finite Array Antennas
- Automatic generation of exchange-correlation response kernels