Numerical thermalization in n-D particle-in-cell simulations
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: "Numerical thermalization in n-D particle-in-cell simulations".
Mira: Here's the gist of the paper. The particle-in-cell (PIC) simulation method can suffer from artificially high collisionality due to the underresolution of particle number, and this paper presents a theory to predict numerical thermalization time so PIC users can develop constraints addressing numerical thermalization a priori The result is a means to predict the timescale of self-consistent Coulomb interactions in the PIC simulation and thus characterize the relevance and implications of numerical thermalization as a function of grid spacing and macroparticle weight It is determined that reaching the physical thermalization time, let alone approximating the collisionless Vlasov limit, may often be intractable in 3D for macroparticle sizes that resolve the Debye length Numerical Thermalization Theory and Scaling The kinetic theory developed by Okuda, Birdsall, and Langdon accurately predicts the self-consistent Coulomb collision time in 1D, 2D and 3D PIC simulations The primary influences on the collision time are the macroparticle shape, macroparticle weight, and the physical plasma parameter ND The finite shape weakens and smooths the force, whereas the macroparticle weight amplifies it The key numerical parameters influencing numerical thermalization are the shape function S(r), and the macroparticle weight W ≡ ND/N M D The kinetic theory is shown to accurately predict the self-consistent Coulomb collision time in 1D, 2D and 3D PIC simulations Constraints on Simulation Design The quality of the trade-off between shape and weight relies on several assumptions 1.
Kai: The weight must be large enough given the physical ND that the simulation is feasible to run 2.
Mira: The smallest scale phenomena of interest must be resolved by the shape width 3.
Kai: The resulting numerical thermalization time must be acceptable compared to the physical thermalization time, the simulation (or residence) time, or other characteristic timescales, such as electron-neutral collisions Validation through PIC Simulations The collision time is validated between electrostatic PIC simulation and kinetic theory in order to describe how particles exchange momentum and thermalize when the characteristic width of the macroparticle shape is varied The exponential decay time of Z(t) is equivalent to the overall numerical Coulomb collision time Simulations with large macroparticle weights suffer from rapid numerical thermalization unless they smooth particle density gradients and forces over many Debye lengths In the filtered case, there is no observed heating and a good agreement with the theory is reached in both large and small macroparticle radius limits Kinetic Framework for PIC Forces The PIC pair interaction force can be encapsulated into the macroparticle pair force In a momentum-conserving PIC plasma, the kernel becomes an operator modulated by the Fourier transform of the shape function, S(k,∆), a gradient discretization tensor D (k,∆), and an optional filter ϕ (k,∆) The force kernel in this case becomes KMC (k,∆) = −ik/k squared Dϕ squared S ΠΠ∗ S This kernel represents the “cloud” plasma model of Okuda Estimation of Thermalization Times The relationship between the velocity-dependent FokkerPlanck coefficients and the corresponding transport coefficients is provided by velocity moments taken over the plasma kinetic equation in the usual way The collision rate νss′ associated with this drift processes is given by Eq.
Paper summary: Kai: So, this paper, "Numerical thermalization in n-D particle-in-cell simulations," it looks at how we get fake collisions when we use too many particles or a heavy weight in particle-in-cell simulations Mira right? The main thesis is that PIC can suffer from artificially high collisionality because of underresolution, which means using a large macroparticle weight Lev yeah, like you're saying the number of particles isn't high enough to really smooth things out Kai and what they claim is a way to predict this numerical thermalization time so users can set constraints beforehand Mira that helps them control how much fake collision happens in their simulations.
Kai: It’s about predicting the timescale of self-consistent Coulomb interactions in these PIC simulations, and it characterizes how important numerical thermalization is depending on the grid spacing and the macroparticle weight Lev so it's a way to predict that timescale for you.
Mira: The paper does this by comparing the collision time calculated from PIC simulations—based on the decay rate of the velocity autocorrelation function—directly against what they got from kinetic theory, which is developed by Okuda, Birdsall, and Langdon Kai to see how accurate it is across different conditions.
Lev: And they found that this kinetic theory accurately predicts the simulated collision time for 1D, 2D, and three dee PIC simulations when you vary the grid spacing, the plasma conditions, and the simulation dimensionality Mira which means the theory holds up across those dimensions <ref:2606.25528#pg1>.
Kai: It seems like it’s about finding a way to tell when we’re hitting a wall where reaching physical thermalization time, or even just approximating that collisionless Vlasov limit, might be impossible in three dimensions for macroparticles that resolve the Debye length Lev which is important because in three dee, things can get really tricky <ref:2606.25528#pg1>.
Mira: And the paper sets up some constraints for simulation design based on this, like you need a weight large enough given the physical plasma parameter ND so it's even feasible to run Kai and you also need the smallest scales you care about to be resolved by the shape width Lev plus that resulting numerical thermalization time has to be acceptable compared to your physical thermalization time or other timescales like electron-neutral collisions Mira.
Lev: From my side, what this means for real hardware is that we need to think about how those constraints affect the actual runtime and overhead of running these simulations on real machines Kai because if the numerical thermalization time is too short, you're just wasting compute cycles on fake interactions.
Kai: So they validated this collision time by comparing it to kinetic theory by varying the characteristic width of the macroparticle shape and seeing how the exponential decay time of Z t matches up Lev and they showed that simulations with large macroparticle weights actually suffer from rapid numerical thermalization unless you smooth those particle density gradients and forces over many Debye lengths Mira.
Paper summary: Mira: And in a filtered case, they found there's no observed heating, which gives them good agreement with the theory whether you're looking at large or small macroparticle radius limits Kai which suggests the shape smoothing plays a big role.
Lev: The kinetic framework they use to describe the PIC pair interaction force is encapsulating it into a macroparticle pair force kernel, KMC, which is an operator modulated by the Fourier transform of the shape function S of k and delta D of k and delta, and an optional filter phi Kai so it’s like modeling the "cloud" plasma model that Okuda described.
Mira: And they estimate these thermalization times by relating them to velocity-dependent Fokker-Planck coefficients, which are derived from moments taken over the plasma kinetic equation in a usual way Lev and then they insert the slow speed approximation into an equation to give you the collision rate nu ss prime Kai.
Lev: That’s where it gets concrete, because that equation gives you the collision rate associated with those drift processes, which is what really drives how fast things thermalize in practice.
Kai: So for someone just listening on a podcast, this paper is about putting a number on how good your particle-in-cell simulation actually is when you use heavy weights and grid spacing Mira it tells you that if your numerical collision time doesn't match the physical one, your results are probably skewed.
Mira: And the authors show the relationship between the velocity moments and those transport coefficients by taking them over the plasma kinetic equation, which sets up how you calculate those rates Lev and they provide this formula for nu ss prime using Eq. forty when you use that slow speed approximation Kai <ref:2606.25528#pg2>.
Lev: It’s a practical tool then, because it’s not just theory; it helps define the boundary conditions for your simulations so you don't waste time running things that won't converge to the right physics Mira.
Kai: So, to wrap up this part of "Numerical thermalization in n-D particle-in-cell simulations," we see how the collision time scales with macroparticle weight and shape, and how it compares to the physical thermalization time Lev which gives us a clear picture of when numerical effects become too much.
Mira: It really frames simulation design by showing that the quality of the trade-off between shape and weight depends on those three assumptions we talked about earlier, like feasibility, resolution, and acceptable thermalization time Kai.
Lev: The implications for me is that when I'm building error correction codes for real hardware, I need to know this timescale because if it's too fast numerically, the overhead of the simulation itself will dominate Mira.
Kai: So ultimately, the authors show that reaching physical thermalization time in three dimensions for macroparticles resolving the Debye length is often intractable, which is a limitation we have to respect Lev.
Mira: The paper’s title and authors suggest they are tackling a fundamental problem in how we model collisions in these complex simulations Kai by providing a predictive tool that links the numerical setup to physical reality.
Conclusion: Kai: So, this paper is about how we get these fake collisions in particle-in-cell simulations when we use too many particles or heavy weights Mira
Mira: Right, specifically addressing that artificial collisionality problem caused by underresolution Lev
Kai: They develop a theory to predict the numerical thermalization time so users can set constraints before they even run the simulation Mira
Lev: It’s about predicting how long it takes for these self-consistent Coulomb interactions to happen in three dimensions, based on things like grid spacing and macroparticle weight Kai
Mira: And they show that reaching the actual physical thermalization time is often impossible in three dimensions if you're using macroparticles that resolve the Debye length Lev
Kai: So it sets a boundary for what we can realistically expect from these simulations Mira
Lev: It gives us a way to see where our numerical approximations actually break down compared to physics Kai
Mira: They validate this by checking their theory against actual PIC simulations when they change the shape of the macroparticles Lev
Kai: And they found that if you use large macroparticle weights, things thermalize very fast unless you smooth out those density gradients over many Debye lengths Mira
Lev: That means for real hardware, we need to account for this numerical speed because if it’s too fast the simulation overhead just dominates Kai
Mira: Ultimately, this work frames how we should design these simulations by linking the particle setup directly to physical reality Lev
Kai: So, what does this mean when we look at the title and who wrote it?
Department of Nuclear Engineering and Radiological Sciences, University of Michigan · Sandia National Laboratories
physics.plasm-ph
Submitted: 2026-06-24
Updated: 2026-06-24
Comments: 17 pages, 5 figures, 1 table
Journal ref: Phys. Plasmas 33, 073906 (2026)
DOI: 10.1063/5.0322906
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 83/100
The gist: The particle-in-cell (PIC) simulation method can suffer from artificially high collisionality due to the underresolution of particle number…
Key concepts
- Numerical Thermalization Time
- This is the timescale at which particles in a PIC simulation reach a state where their behavior mimics physical collisions. The paper provides a method to predict this time based on simulation parameters, helping users understand when their simulations are accurate.
- Macroparticle Weight (W)
- Weight is defined as the ratio of the plasma density to the particle number density multiplied by the Debye length (ND/NMD). It influences how strongly macroparticles interact with each other and affects the resulting collision time in 1D, 2D, and 3D simulations.
- Shape Function S(r)
- The shape function describes the physical geometry of a macroparticle. Its finite shape determines how forces are weakened or smoothed as particles approach each other, which is a key factor in the collision time calculation.
Terminology
Summary
The particle-in-cell (PIC) simulation method can suffer from artificially high collisionality due to the underresolution of particle number, and this paper presents a theory to predict numerical thermalization time so PIC users can develop constraints addressing numerical thermalization a priori The result is a means to predict the timescale of self-consistent Coulomb interactions in the PIC simulation and thus characterize the relevance and implications of numerical thermalization as a function of grid spacing and macroparticle weight It is determined that reaching the physical thermalization time, let alone approximating the collisionless Vlasov limit, may often be intractable in 3D for macroparticle sizes that resolve the Debye length
Numerical Thermalization Theory and Scaling
The kinetic theory developed by Okuda, Birdsall, and Langdon accurately predicts the self-consistent Coulomb collision time in 1D, 2D and 3D PIC simulations The primary influences on the collision time are the macroparticle shape, macroparticle weight, and the physical plasma parameter ND The finite shape weakens and smooths the force, whereas the macroparticle weight amplifies it The key numerical parameters influencing numerical thermalization are the shape function S(r), and the macroparticle weight W ≡ ND/N M D The kinetic theory is shown to accurately predict the self-consistent Coulomb collision time in 1D, 2D and 3D PIC simulations
Constraints on Simulation Design
The quality of the trade-off between shape and weight relies on several assumptions
-
The weight must be large enough given the physical ND that the simulation is feasible to run
-
The smallest scale phenomena of interest must be resolved by the shape width
-
The resulting numerical thermalization time must be acceptable compared to the physical thermalization time, the simulation (or residence) time, or other characteristic timescales, such as electron-neutral collisions
Validation through PIC Simulations
The collision time is validated between electrostatic PIC simulation and kinetic theory in order to describe how particles exchange momentum and thermalize when the characteristic width of the macroparticle shape is varied The exponential decay time of Z(t) is equivalent to the overall numerical Coulomb collision time Simulations with large macroparticle weights suffer from rapid numerical thermalization unless they smooth particle density gradients and forces over many Debye lengths In the filtered case, there is no observed heating and a good agreement with the theory is reached in both large and small macroparticle radius limits
Kinetic Framework for PIC Forces
The PIC pair interaction force can be encapsulated into the macroparticle pair force In a momentum-conserving PIC plasma, the kernel becomes an operator modulated by the Fourier transform of the shape function, S(k,∆), a gradient discretization tensor D (k,∆), and an optional filter ϕ (k,∆) The force kernel in this case becomes KMC (k,∆) = −ik/k squared Dϕ squared S ΠΠ∗ S This kernel represents the “cloud” plasma model of Okuda
Estimation of Thermalization Times
The relationship between the velocity-dependent FokkerPlanck coefficients and the corresponding transport coefficients is provided by velocity moments taken over the plasma kinetic equation in the usual way The collision rate νss′ associated with this drift processes is given by Eq. (40) Inserting the slow speed approximation into Eq. p. 9 For distributions that are in a linear relaxation
regime, collision time models for each the dynamic, static,
and slow speed drag integrals can be evaluated to estimate numerical thermalization with varying ease. Inserting the slow
speed approximation into Eq.Where the paper says it
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
-
AI systems can be constrained to develop simulations where numerical thermalization time is far below physical collision time so that
the physical collision rate can be added in using Monte Carlo.
-
AI systems can use kinetic theory, such as the Lenard-Balescu like equation, to predict the
numerical thermalization time in PIC simulations with a variety of conditions, dimensions, and interpolation/solver schemes.
-
AI systems can incorporate constraints based on the comparison between "the numerical thermalization rate (on the order of tau num ss') and both the physical Coulomb collision time tau phys ss' for each species combination, s,s', as well as to other characteristic timescales of the simulation."
-
AI systems can be designed to account for
shape effects that nominally reduce collisions are mediated through the inter-particle force,
allowing them to determine if numerical thermalization is an issue based on comparing macroparticle size and other relevant length scales. -
AI systems can use the calculated drag models, such as the
dynamic screening expression for the Coulomb collision frequency
(Eq. 43), to estimate numerical thermalization with varying ease across different macroparticle radii and dimensions. -
AI systems can be used to evaluate
the permissible extent of such scaling is limited by numerical thermalization because the PIC Coulomb collision time is determined by N M D regardless of the ND in the hypothetically smaller system.
-
AI systems can be trained to recognize when
numerical thermalization is not a problematic constraint here because reaching the physical ND ≈ 2 is feasible and the shape effects extend the collision time well beyond tau phys omega p.
Abstract
The particle-in-cell (PIC) simulation method is often understood to solve the collisionless Vlasov equation due to the finite shape of its macroparticles. In reality, it can suffer from artificially high collisionality due to the underresolution of particle number; i.e., the use of a large macroparticle weight. The degree to which particle shape effects compensate for a large macroparticle weight in 1D, 2D, and 3D is presented. The collision time is calculated from PIC simulations based on the decay rate of the velocity autocorrelation function and compared directly with the kinetic theory of Okuda, Birdsall, and Langdon. The theory is found to accurately predict the simulated collision time with varied grid spacings, plasma conditions, and simulation dimensionalities. The result is a means to predict the timescale of self-consistent Coulomb interactions in the PIC simulation and thus characterize the relevance and implications of numerical thermalization as a function of grid spacing and macroparticle weight. It is determined that reaching the physical thermalization time, let alone approximating the collisionless Vlasov limit, may often be intractable in 3D for macroparticle sizes that resolve the Debye length.
Related papers
- Implementation and verification of a new free boundary module for Resistive Wall Modes simulations with the nonlinear 3D full MHD code SpeCyl
- System Size Dependence of Collisionless Reconnection Rate
- SPECTRE: A robust solver for 3D equilibria with arbitrary topology
- High-quality electron beam generation from laser wakefield accelerators for driving compact free electron lasers
- Deflation Techniques for Stellarator Equilibrium and Optimization
- Rapid Vacuum Response Calculations for Non-axisymmetric Plasma Geometries