A new arXiv preprint reports a numerical solver for a hybrid plasma equation that treats ions kinetically and electrons as a fluid. In a standing-shock test, relative deviations in mass, momentum and energy were reported at the order of 10^-12. In a separate weak-damping test, the corresponding conservation deviations were at the order of 10^-11. The standing-shock calculation also kept the average numerical rank very low.
The study asks whether a conservative, rank-adaptive method can solve the hybrid Vlasov-Fokker-Planck system while preserving its numerical structure. Its evidence comes from deterministic calculations covering several standard benchmark problems. The document is an arXiv version 1 preprint dated 20 August 2026.
How the calculation works
In the reported model, ions are handled through a kinetic description, while electrons are represented as a fluid. The configuration has one physical-space dimension and two cylindrically symmetric velocity dimensions, a setup known as 1D-2V. It assumes zero magnetic field and a single ion species.
The term low-rank refers to representing the evolving numerical state with a limited number of components instead of storing every part of the full calculation independently. The method can adjust that rank as the solution changes. In the heat benchmark, the reported solution usually stayed at rank 1, with occasional rank-2 values; the LBFP relaxation test began with a rank-2 state and temporarily increased in rank.
The framework combines RAIL for low-rank velocity updates with LoMaC, which performs conservative truncation. Both methods are extended to cylindrical velocity coordinates. The collision operator used in the LBFP test is discretized with the structure-preserving Chang-Cooper method.
For time integration, the kinetic-ion equation is advanced with a second-order, two-stage IMEX Runge-Kutta method. IMEX means that part of the update is treated implicitly and another part explicitly. The macroscopic first-stage system is solved with Newton's method, after which the moments are corrected with QCM before the kinetic update.
Four tests, different checks
The numerical evaluation covers four problems: a heat equation, homogeneous LBFP relaxation, a hybrid VFP standing shock and collisionless weak Landau damping. The reported comparisons examine temporal and velocity-mesh errors, conservation of mass, momentum and energy, equilibrium and entropy diagnostics, long-term shock behavior, solution rank and electric-field decay.
The heat-equation benchmark showed the reported temporal error behavior and machine-precision mass conservation. Its solution remained rank 1 for most of the calculation, with rank 2 appearing occasionally.
The homogeneous LBFP relaxation test also showed the expected temporal accuracy. The authors reported near-machine-precision conservation, along with near-machine-precision equilibrium and relative-entropy diagnostics. Relative entropy is a numerical measure used here to track the computed state against the equilibrium behavior being tested.
These findings are numerical comparisons rather than estimates from a sampled population. The paper reports no inferential statistical analysis or statistical uncertainty intervals; instead, it compares error norms with exact or reference solutions and uses conservation, equilibrium, entropy, rank and damping diagnostics.
The longer test was a standing shock
The standing-shock comparison provided the clearest long-run test of the conservative procedures. One calculation used the LoMaC and QCM steps, while a comparison omitted them. The latter run showed nonphysical drift in the shock, whereas the conservative calculation retained the reported shock profile.
In the conservative standing-shock run, relative deviations in mass, momentum and energy stayed at the order of 10^-12. The average rank across the spatial nodes remained very low, indicating that the calculation retained a compact numerical representation during the test.
A Mach 5 standing-shock refinement test showed near-second-order behavior in the tabulated velocity errors. The reported refinement orders were 2.18 and 2.29 for the perpendicular velocity direction, using the L1 and L∞ error measures, and 2.04 and 1.98 for the parallel direction.
What the results do not establish
In the collisionless weak Landau-damping test, the simulated electric-field decay agreed well with the stated theoretical ion damping rate of γLD = -0.04. Mass, momentum and energy were conserved to the order of 10^-11 in that calculation.
The results remain confined to the reported 1D-2V configuration, with zero magnetic field and one ion species. They therefore do not establish how the solver performs for a full three-dimensional, multi-species system with magnetic fields.
The authors identify nonlinear high-resolution limiters for projection-based rank-adaptive methods as an open issue. They also report a CFL time-step restriction on fixed structured grids, meaning that the grid and numerical update limit how large a time step can be.
The paper states that the source code generated and analyzed in the study is available from the authors' GitHub repository. Its acknowledgments report support from Swarthmore College, the DOE Office of Science ASCR and Triad National Security, LLC, with early discussions associated with an Air Force Summer Faculty Fellowship. The authors report no competing interests.
Taken together, the preprint reports a solver that maintained low numerical rank while matching expected temporal and mesh-refinement behavior in its selected tests. Its strongest demonstrations are the small conservation deviations in the standing-shock and weak-damping runs and the contrast between the conservative shock calculation and the run with those correction procedures omitted. Whether the same behavior extends beyond the stated model and numerical settings remains open.
Paper data and sources
Original title: An adaptive and conservative low-rank IMEX solver for the hybrid ion Vlasov-Fokker-Planck and fluid electron system
Authors: Joseph Nakao, Dylan T. Jacobs, William Taitano
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-20
DOI: Not available
Original paper · Full text