Pith. sign in

REVIEW 3 major objections 5 minor 34 references

Kinetic Optimization of Magnetic Mirror Confinement: Beyond Classical Loss-Cone Theory

T0 review · 3 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The optimal shape of a magnetic mirror depends on the kinetic model: electron-only plasmas favor a centrally peaked field, while electron-ion plasmas recover the classical boundary-peaked mirror.

desk verdict Worth refereeing: internally coherent, robust optimization, public code — but the topology switch is only demonstrated at mi/me=25, and the claim that 25 captures the species separation is asserted, not tested. read the letter →

arxiv 2607.26479 v1 pith:PM4Z7P4B submitted 2026-07-29 physics.plasm-ph math.OC

Lukas Einkemmer , Martin Guerra , Qin Li , Leonardo Zepeda-Núñez This is my paper · ORCID
classification physics.plasm-phmath.OC
keywords magneticmirrordrift-kineticVlasov-Poissonlossconeself-consistentelectricfieldPDE-constrainedoptimizationautomaticdifferentiationplasmaconfinement
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper sets out to show that optimal magnetic mirror design cannot be decided by classical loss-cone arguments alone; the self-consistent electric field is a first-class confinement mechanism, and the best field shape depends on which kinetic model governs the plasma. Using a reduced drift-kinetic–Poisson model, the optimization discovers that an electron-only system favors a centrally peaked double-well magnetic profile, whereas a coupled electron-ion system recovers the familiar boundary-peaked mirror. A sympathetic reader cares because this indicates that shape optimization of mirrors should be performed through kinetic simulation and that electrostatic trapping is an exploitable design knob, not a small correction.

What carries the argument

The load-bearing object is the coupled 1D1V drift-kinetic Vlasov–Poisson system (2.1)–(2.2), parameterized by the conserved magnetic moment μ. The magnetic mirror force is -μ ∂_z |B|; the self-consistent electrostatic potential φ is obtained from a 1D flux-tube Poisson equation with homogeneous Dirichlet boundary conditions, producing an electric field E = -∂_z φ that adds a species-dependent acceleration. The optimization parameterizes the magnetic profile |B(z)| as a neural network with a fixed mirror ratio R_m = 10, and maximizes the total retained longitudinal density at a final time using reverse-mode automatic differentiation through the discretized solver. The Poisson coefficient prop

What would settle it

Repeat the optimization with mobile ions at the physical mass ratio (m_i/m_e ≈ 1836) and a time horizon that extends past the ambipolar phase; if the centrally peaked electron-only optimum disappears, it is an artifact of the frozen-ion assumption. Alternatively, measure in an experiment or kinetic simulation with effectively immobile ions whether a centrally peaked field with mirror ratio 10 retains more electrons than the classical boundary-peaked field of the same ratio over the escape time t≈π/2.

Watch

Extended reading notes

Core claim

Within the reduced 1D1V drift-kinetic–Poisson model, the optimizing magnetic field configuration depends qualitatively on the kinetic model: an electron-only model with a stationary neutralizing ion background favors an unconventional centrally peaked double-well field, while the fully coupled electron-ion model (mass ratio 25) recovers the classical boundary-peaked mirror. The self-consistent electric field generated through the Poisson equation is the second confinement mechanism: it traps particles whose magnetic moment is nearly zero—particles the classical loss cone predicts will escape—and its strength relative to the mirror force differs between the two regimes. The paper also documen

Load-bearing premise

The central conclusion rests on the reduced 1D1V drift-kinetic–Poisson model, with stationary ions in the electron-only case, a reduced mass ratio of 25 in the coupled case, and a finite-time retention objective; if these simplifications do not capture real mirror dynamics, the found model-dependence of the optimal shape could be a modeling artifact rather than a physical design principle.

Editorial extensions

If this is right

  • Two magnetic fields with the same mirror ratio can confine very differently once the self-consistent electric field is included, so mirror-ratio heuristics alone are insufficient.
  • The self-consistent electric field acts as a second confinement barrier, especially for low-magnetic-moment electrons, and raises retained mass from about 94% (linear loss-cone prediction) to above 97% in the electron-only model.
  • In an electron-only regime with stationary ions, the optimized field is a centrally peaked double-well profile rather than the classical single-well mirror, and this preference is consistent across 100 random initializations.
  • In the fully coupled electron-ion system, the optimizer recovers the classical boundary-peaked mirror, indicating that ion dynamics drive the topology back toward the conventional design.
  • End-to-end gradient-based optimization through a kinetic solver is a viable method for finding non-heuristic mirror configurations.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Interpreted as a fast-timescale solution, the electron-only central peak suggests a two-phase operating scheme: start with a centrally peaked field to build an electrostatic barrier for electrons, then transition toward the boundary-peaked profile as ions become mobile. The paper does not draw this control implication; it follows from the separation of electron and ion confinement timescales the p
  • The finite-time objective used for optimization may itself be a design lever: if the horizon is short, the electron-only central peak looks optimal; if extended through the field-reversal phase, the classical shape may regain dominance. A testable extension is to optimize with a time-averaged retention objective.
  • Because the model is one-dimensional along the field line, the centrally peaked double-well topology is a clean but possibly fragile prediction; extending the drift-kinetic setup to 2D or including finite Larmor radius effects could alter the electrostatic barrier that creates the central preference.
  • The four-phase ambipolar dynamics implies that 'confinement' is not a single number; mirror designs should be assessed by their retention curve over time, and optimization targets should reflect the intended phase of operation. This follows from the paper's own results but is not stated as a design conclusion.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper formulates magnetic mirror design as a PDE-constrained optimization problem governed by a reduced 1D1V drift-kinetic–Poisson model parameterized by the magnetic moment. After deriving the classical loss-cone trapped fraction for a Maxwellian, forward simulations show that the self-consistent electric field adds a confinement mechanism beyond the magnetic mirror force, leading to four-phase escape dynamics in the coupled electron–ion system. Neural-network parameterizations of the magnetic field are then optimized under a fixed mirror ratio R_m=10. The authors report that the electron-only model with stationary ions optimizes to a centrally peaked double-well field, whereas the fully coupled electron–ion model at m_i/m_e=25 recovers a boundary-peaked classical mirror configuration. The paper interprets this as a qualitative model-dependence of optimal magnetic mirror design and argues that optimization must be based on self-consistent kinetic dynamics rather than loss-cone heuristics.

Significance. If the central claim holds, the paper identifies a physically meaningful and non-obvious effect: self-consistent electrostatic trapping can invert the classical loss-cone design heuristic, and the optimal magnetic topology depends on the kinetic model. The paper has clear strengths: the linear simulation reproduces the analytic loss-cone fraction (~94%) and the earliest escape time; the optimization is repeated over 100 random initializations with narrow 95% confidence bands; the loss-cone formula is derived independently of the optimization; and public code is provided. These features make the internal reasoning reproducible and internally consistent. However, the headline qualitative conclusion is currently tied to a reduced mass ratio and a finite-time objective, and the numerical verification lacks grid-convergence evidence. The result is significant if it survives those tests, but that is not yet demonstrated.

major comments (3)
  1. [§4.2.2, Table 1] The central dichotomy compares the electron-only model with stationary ions (§4.2.1) to a fully coupled model at m_i/m_e=25 (§4.2.2). These two settings differ simultaneously in ion mobility and in mass ratio, so the observed topology change cannot be cleanly attributed to the presence of ion dynamics per se. The text asserts that m_i/m_e=25 is “sufficient to capture the distinct macroscopic timescales,” but no numerical evidence is given. At the physical proton–electron ratio (~1836), ions are almost immobile on electron confinement timescales, and the coupled model should approach the frozen-ion electron-only limit; if it does, the centrally peaked optimum may re-emerge. This directly bears on the abstract’s claim that the optimized configuration depends qualitatively on the underlying kinetic model. I request either simulations at several increased mass ratios (e.g., 100, 400, or as l
  2. [§2.3, Eq. (2.5); §3.3 and §4.2] The optimization objective is the retained mass at a single finite horizon T, with T=25 for the electron-only case and T=80 for the coupled case. The topology switch may therefore be an artifact of comparing different horizons: in the coupled case electrons have already undergone two escape phases well before T=80, so the joint objective may be dominated by ion retention; conversely, an electron-only run extended to T=80 might still prefer a centrally peaked field. The paper does not study the sensitivity of the optimized profile to T, nor does it use a steady-state or time-averaged objective. Since the abstract claims a qualitative model-dependence, the comparison should be made at comparable or systematically varied horizons.
  3. [§4.1, Appendix A, Table 1] The numerical scheme is described in detail, but the paper does not report grid sizes (N_z, N_v, N_μ), the time step Δt, or any grid-convergence study. Because gradients are computed through the fully discretized solver, the optimizer could exploit numerical diffusion or discretization artifacts rather than physical confinement mechanisms. The 100-initialization robustness addresses optimizer variability but not discretization error. I ask for a convergence study showing that the optimal topology (centrally peaked vs. boundary peaked) is unchanged under spatial, velocity, and time-step refinement, and for the actual resolution used in Table 1 to be stated.
minor comments (5)
  1. [Eq. (2.2)] The integral sign is rendered as an “x” in the displayed Poisson equation, making the equation hard to read. The notation should be corrected to a proper integral over v and μ.
  2. [Related Work, Introduction] The last paragraph of §1.1 says the code enables optimization “within a three-dimensional drift-kinetic model,” but the model is 1D1V parameterized by μ (three-dimensional phase space but one spatial dimension). This wording is misleading and should be corrected.
  3. [Table 1 and Appendix A] Table 1 omits the grid resolution and time step, while Algorithm 2 refers to an “isotropic grid” although z, v, and μ have different domains and units. Please state the exact resolutions used for all experiments.
  4. [Figures 8 and 11] The optimized profiles are shown with confidence bands in Appendix C, but the main-text figures would benefit from overlaying the band width or stating the standard deviation at key z locations, so the reader can see how much the central peak varies across the 100 runs.
  5. [§3.3] The four-phase interpretation is descriptive and plausible, but it would be strengthened by a quantitative indicator (e.g., sign of net charge or time of field reversal) plotted alongside the retained mass, rather than inferred from the mass curves alone.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central claims are outputs of forward simulation and PDE-constrained optimization, with the loss-cone formula derived independently and no parameter fitted to reproduce the reported effects.

full rationale

The paper's derivation chain is self-contained. The drift-kinetic-Poisson model (2.1)-(2.2) is stated as the governing equations, the objective (2.5) is a finite-time retention functional, and the magnetic field is parameterized by a neural network with fixed mirror ratio and symmetry constraints. The classical loss-cone prediction (3.3) is derived independently from conservation of magnetic moment and kinetic energy and is used only as a benchmark; it is not fitted to the simulations. The electrostatic confinement effect and the qualitative difference between the electron-only and coupled optimal topologies are outputs of the numerical solver, not assumptions built into the objective or the admissible set. The admissible set contains both centrally-peaked and boundary-peaked profiles, so the optimizer's choice is not forced by construction. Self-citations appear only in the related-work survey and in standard numerical-scheme references; no load-bearing step is justified solely by a self-citation. The acknowledged reduced mass ratio (mi/me=25) and the frozen-ion approximation are modeling and validity concerns, not circularity, because the reported predictions are not defined in terms of those assumptions. Overall, no step reduces to its own input by construction.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

All claims are conditioned on the reduced kinetic model; the model reductions, the frozen-ion approximation, the reduced mass ratio, and the final-time objective are the main premises. No new physical entities are introduced, and no external benchmark validates the reduced model against full 3D kinetic or experimental data.

free parameters (5)
  • Reduced ion-to-electron mass ratio m_i/m_e = 25 = 25
    Chosen to reduce computational cost (Section 3.1, Table 1). The multi-species optimum and the two-stage leakage claim depend on this ratio being representative of the physical 1836 ratio.
  • Simulation horizon T = 25 (single-species), 80 (multi-species)
    The objective (2.5) maximizes retained mass at the final time; optimal topologies may be finite-horizon artifacts. Chosen by hand in Table 1.
  • Mirror ratio R_m = 10 = 10
    Fixed design constraint in Table 1 and Section 4.1; all compared profiles have the same mirror ratio, so optimization can only reshape geometry. Chosen by hand.
  • Initial Gaussian width σ_z and center z_c = Not stated numerically in text
    Initial condition parameters in (3.1) are chosen by hand; confinement optimization may depend on them, but their values are not reported in Table 1.
  • Velocity domain v0 = 4 and μ cutoff tolerance = v0 = 4; e^{-μ0 Bmax} = 10^{-12}
    Computational domain choices in Table 1 that truncate phase space and affect escape-time estimates.
assumptions (6)
  • domain assumption Adiabatic invariance of μ and guiding-center reduction to 1D1V drift-kinetic model are valid
    Sections 2.1-2.2. The entire optimization is performed in this reduced model; violations (strong gradients, non-paraxial geometry, finite Larmor radius) would change the optimum.
  • domain assumption The 1D flux-tube Poisson equation (2.2) with homogeneous Dirichlet boundary conditions correctly captures ambipolar fields
    Section 2.2. The electrostatic barrier mechanism and optimized profiles depend on this specific model; boundary conditions at z = ±L_z are a modeling choice.
  • domain assumption Electron-only limit with stationary neutralizing ions isolates fast electron confinement
    Section 3.1. The centrally-peaked optimum is found only in this limit; if ions cannot be treated as stationary over the relevant time, the result is an artifact.
  • domain assumption Reduced mass ratio m_i/m_e = 25 preserves the qualitative two-timescale separation
    Section 3.1 and Table 1. The multi-species conclusion (recovery of boundary-peaked mirror) may depend on this; not validated against m_i/m_e = 1836.
  • domain assumption Final-time retained mass (2.5) is a valid confinement objective
    Section 2.3. Maximizing mass at T may favor delaying losses rather than true steady confinement; the centrally-peaked profile is a finite-horizon optimum.
  • standard math The numerical discretization is accurate and converged for the optimization
    Appendix A. No convergence study or grid-resolution sensitivity is reported; optimization may exploit numerical diffusion or artifacts of the PCHIP semi-Lagrangian scheme.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Kinetic Optimization of Magnetic Mirror Confinement: Beyond Classical Loss-Cone Theory." pith.science (2026). https://pith.science/paper/PM4Z7P4B

@misc{pith2026260726479,
  author       = {Pith},
  title        = {Pith review of: Kinetic Optimization of Magnetic Mirror Confinement: Beyond Classical Loss-Cone Theory},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PM4Z7P4B}},
  note         = {Machine review of arXiv:2607.26479}
}
read the original abstract

Magnetic mirrors are among the conceptually simplest plasma confinement configurations and remain promising candidates for thermonuclear fusion. Their design requires shaping an externally applied magnetic field to confine plasma within an open-ended cylindrical device. In contrast to toroidally closed devices such as tokamaks and stellarators, confinement in magnetic mirrors depends intrinsically on kinetic mechanisms, particularly velocity-space trapping and particle loss through the open ends. We formulate magnetic mirror design as a PDE-constrained optimization problem governed by a reduced multispecies drift-kinetic-Poisson model. The resulting optimization reveals two physical effects not captured by the classical loss-cone argument. First, the self-consistent electric field generated through Poisson coupling acts as a secondary confinement barrier and substantially alters particle retention in the nonlinear regime. Second, the optimized magnetic-field configuration depends qualitatively on the underlying kinetic model: an electron-only model favors an unconventional centrally peaked field, whereas the fully coupled electron-ion model recovers the classical boundary-peaked mirror configuration. These results demonstrate that optimal magnetic mirror design cannot be determined solely from loss-cone considerations, but must account for the self-consistent nonlinear kinetic dynamics of the plasma.

Figures

Figures reproduced from arXiv: 2607.26479 by the authors.

Figure 1
Figure 1. Initial setup for |B(z)| (left) and fs(0, z, v, µ) (right). We first consider a simplified electron-only regime where the ion distribution is treated as an stationary neutralizing background so fi does not evolve. This approximation is valid on timescales that are short enough such that the heavy ions do not have enough time to respond to the fast electrons. Therefore, this approximation isolates the fast electron d… view at source ↗
Figure 2
Figure 2. Electron density fe(T = 25) for simulating (2.1) with a stationary ion background without the presence of E (top) and with the presence of E (bottom). Clearly, the electric field provides another layer of confinement [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Physical quantities computed from the simulation of the electron drift-kinetic equation with a [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Final particle density fs(T = 80) (s ∈ {e, i}) of (2.1) without the presence of E (top 2 rows) and with the presence of E (bottom 2 rows). • if Rm = 1 (no magnetic mirror), then Ftrapped = 0, meaning all particles escape; • if Rm → ∞ (perfect mirror), then Ftrapped → 1…
Figure 5
Figure 5. Figure 5: Physical quantities computed from the simulation of the coupled electron-ion drift-kinetic equa [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Retained electron mass as a function of time for simulations with and without the self-consistent [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Retained particle mass for the fully coupled electron-ion system. Left: electron mass evolution. [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]
Figure 8
Figure 8. Figure 8: Optimization for the electron drift-kinetic equation with stationary ions. Top-left: initial and [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 9
Figure 9. Figure 9: Initial phase-space distributions generated from the optimized double well magnetic fields for the [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 10
Figure 10. Figure 10: Electron phase-space distributions at final time [PITH_FULL_IMAGE:figures/full_fig_p016_10.png]
Figure 11
Figure 11. Figure 11: Optimization results in the fully coupled electron-ion system. Top-left: initial and optimized [PITH_FULL_IMAGE:figures/full_fig_p017_11.png]
Figure 12
Figure 12. Figure 12: Initial phase-space distributions for the optimized multi-species configuration. [PITH_FULL_IMAGE:figures/full_fig_p018_12.png]
Figure 13
Figure 13. Figure 13: Phase-space distributions at T = 80 for the optimized multi-species configuration. [7] John R. Cary and Alain J. Brizard. “Hamiltonian theory of guiding-center motion”. In: Review of Modern Physics 81 (2 2009), pp. 693–738. doi: 10.1103/RevModPhys.81.693. url: https:/…
Figure 14
Figure 14. Figure 14: Architecture of the MLP used to define |B(z; θ)|. C Robustness of the optimization routine We run Algorithm 1 for 100 different initializations and plot the mean of the different optimal magnetic profiles and a 95% confidence interval for both, the single-species and …
Figure 15
Figure 15. Figure 15: Mean of optimal magnetic profiles |B(z; θ ∗ )| for different initializations θ0 and shaded area representing a 95% confidence interval. From left to right: single-species and multi-species. 24 [PITH_FULL_IMAGE:figures/full_fig_p024_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 8 canonical work pages

  1. [1]

    Accelerated methods for direct computation of fusion alpha particle losses within, stellarator optimization

    Christopher G Albert, Sergei V Kasilov, and Winfried Kernbichler. “Accelerated methods for direct computation of fusion alpha particle losses within, stellarator optimization”. In:Journal of Plasma Physics86.2 (2020), p. 815860201

    2020

  2. [2]

    Instantaneous control strategies for magnetically confined fusion plasma

    Giacomo Albi, Giacomo Dimarco, Federica Ferrarese, and Lorenzo Pareschi. “Instantaneous control strategies for magnetically confined fusion plasma”. In:Journal of Computational Physics527 (2025), p. 113804.issn: 0021-9991.doi:10.1016/j.jcp.2025.113804.url:https://www.sciencedirect. com/science/article/pii/S0021999125000877

    arXiv 2025

  3. [3]

    Robust feedback control of collisional plasma dynamics in presence of uncertainties

    Giacomo Albi, Giacomo Dimarco, Federica Ferrarese, and Lorenzo Pareschi. “Robust feedback control of collisional plasma dynamics in presence of uncertainties”. In:Journal of Computational Physics (2025), p. 114512

    2025

  4. [4]

    On the stabilization of a kinetic model by feedback-like control fields in a Monte Carlo framework

    Jan Bartsch and Alfio Borzi. “On the stabilization of a kinetic model by feedback-like control fields in a Monte Carlo framework”. In:Kinetic and Related Models17.6 (2024), pp. 892–913.issn: 1937-5093. doi:10.3934/krm.2024005

    work page doi:10.3934/krm.2024005 2024

  5. [5]

    Controlling a Vlasov–Poisson Plasma by a Particle-in-Cell Method Based on a Monte Carlo Framework

    Jan Bartsch, Patrik Knopf, Stefania Scheurer, and J¨ org Weber. “Controlling a Vlasov–Poisson Plasma by a Particle-in-Cell Method Based on a Monte Carlo Framework”. In:SIAM Journal on Control and Optimization62.4 (2024), pp. 1977–2011.doi:10.1137/23M1563852. eprint:https://doi.org/10. 1137/23M1563852.url:https://doi.org/10.1137/23M1563852

    doi:10.1137/23m1563852 2024

  6. [6]

    Foundations of nonlinear gyrokinetic theory

    A. J. Brizard and T. S. Hahm. “Foundations of nonlinear gyrokinetic theory”. In:Review of Modern Physics79 (2 2007), pp. 421–468.doi:10.1103/RevModPhys.79.421.url:https://link.aps.org/ doi/10.1103/RevModPhys.79.421. 18 (a)f s =f e (b)f s =f i Figure 13: Phase-space distributions atT= 80 for the optimized multi-species configuration

    work page doi:10.1103/revmodphys.79.421.url:https://link.aps.org/ 2007

  7. [7]

    Hamiltonian theory of guiding-center motion

    John R. Cary and Alain J. Brizard. “Hamiltonian theory of guiding-center motion”. In:Review of Modern Physics81 (2 2009), pp. 693–738.doi:10.1103/RevModPhys.81.693.url:https://link. aps.org/doi/10.1103/RevModPhys.81.693

    work page doi:10.1103/revmodphys.81.693.url:https://link 2009

  8. [8]

    New York, NY: Springer, 1984

    Francis F Chen.Introduction to Plasma Physics and Controlled Fusion. New York, NY: Springer, 1984. isbn: 9783319223087.doi:10.1007/978-3-319-22309-4

    doi:10.1007/978-3-319-22309-4 1984

Show all 34 references
  1. [9]

    Semi-Lagrangian SA V method for Vlasov-Maxwell equations

    Nicolas Crouseilles, Hongtao Liu, and Yukun Yue. “Semi-Lagrangian SA V method for Vlasov-Maxwell equations”. In:Journal of Computational Physics549 (2026), p. 114606.issn: 0021-9991.doi:10. 1016 / j . jcp . 2025 . 114606.url:https : / / www . sciencedirect . com / science / ar...

    2026

  2. [10]

    Conservative semi-Lagrangian schemes for Vlasov equations

    Nicolas Crouseilles, Michel Mehrenberger, and Eric Sonnendr¨ ucker. “Conservative semi-Lagrangian schemes for Vlasov equations”. In:Journal of Computational Physics229.6 (2010), pp. 1927–1953. issn: 0021-9991.doi:10 . 1016 / j . jcp . 2009 . 11 . 007.url:https : / / www . scie...

    2010

  3. [11]

    Control of kinetic plasma instabilities by laser fields

    Nicolas Crouseilles et al. “Control of kinetic plasma instabilities by laser fields”. In:arXiv:2509.23080, to appear in Physics of Plasmas(2025)

    2025

  4. [12]

    Magnetic control of tokamak plasmas through deep reinforcement learning

    Jonas Degrave et al. “Magnetic control of tokamak plasmas through deep reinforcement learning”. In: Nature602 (2022), pp. 414–419.doi:10.1038/s41586-021-04301-9

    2022 doi

  5. [13]

    Stabilization of beam heated plasmas by beam modulation

    Lukas Einkemmer. “Stabilization of beam heated plasmas by beam modulation”. In:Physics of Plasmas 31.12 (2024)

    2024

  6. [14]

    Control of instability in a Vlasov- Poisson system through an external electric field

    Lukas Einkemmer, Qin Li, Cl´ ement Mouhot, and Yukun Yue. “Control of instability in a Vlasov- Poisson system through an external electric field”. In:Journal of Computational Physics530 (2025), p. 113904.issn: 0021-9991.doi:10.1016/j.jcp.2025.113904.url:https://www.sciencedire...

    2025

  7. [15]

    Suppressing instability in a Vlasov–Poisson system by an external electric field through constrained optimization

    Lukas Einkemmer, Qin Li, Li Wang, and Yang Yunan. “Suppressing instability in a Vlasov–Poisson system by an external electric field through constrained optimization”. In:Journal of Computational Physics498 (2024), p. 112662.issn: 0021-9991.doi:10.1016/j.jcp.2023.112662.url:htt...

    2024

  8. [16]

    A semi-Lagrangian discontinuous Galerkin method for drift-kinetic simulations on GPUs

    Lukas Einkemmer and Alexander Moriggl. “A semi-Lagrangian discontinuous Galerkin method for drift-kinetic simulations on GPUs”. In:SIAM Journal on Scientific Computing46.2 (2024), B33–B55

    2024

  9. [17]

    Kinetic scrape off layer simulations with semi-Lagrangian discontinuous Galerkin schemes

    Lukas Einkemmer and Alexander Moriggl. “Kinetic scrape off layer simulations with semi-Lagrangian discontinuous Galerkin schemes”. In:Computer Physics Communications(2025), p. 109775

    2025

  10. [18]

    Convergence Analysis of a Discontinuous Galerkin/S- trang Splitting Approximation for the Vlasov–Poisson Equations

    Lukas Einkemmer and Alexander Ostermann. “Convergence Analysis of a Discontinuous Galerkin/S- trang Splitting Approximation for the Vlasov–Poisson Equations”. In:SIAM Journal on Numerical Analysis52.2 (2014), pp. 757–778.doi:10.1137/120898620

    2014 doi

  11. [19]

    Conservative Numerical Schemes for the Vlasov Equation

    Francis Filbet, Eric Sonnendr¨ ucker, and Pierre Bertrand. “Conservative Numerical Schemes for the Vlasov Equation”. In:Journal of Computational Physics172.1 (2001), pp. 166–187.issn: 0021-9991. doi:10.1006/jcph.2001.6818.url:https://www.sciencedirect.com/science/article/pii/ ...

    2001

  12. [20]

    What metric to optimize for sup- pressing instability in a Vlasov-Poisson system?

    Martin Guerra, Qin Li, Yukun Yue, and Leonardo Zepeda-N´ u˜ nez. “What metric to optimize for sup- pressing instability in a Vlasov-Poisson system?” In:Journal of Computational Physics564 (2026), p. 115169.issn: 0021-9991.doi:10.1016/j.jcp.2026.115169.url:https://www.sciencedi...

    2026

  13. [21]

    Hazeltine and James D

    Richard D. Hazeltine and James D. Meiss.Plasma Confinement. Courier Corporation, 2003.isbn: 9780486432427

    2003

  14. [22]

    An implicit, conservative electrostatic particle-in-cell algorithm for paraxial magnetic nozzles

    Pedro Jim´ enez, Luis Chac´ on, and Mario Merino. “An implicit, conservative electrostatic particle-in-cell algorithm for paraxial magnetic nozzles”. In:Journal of Computational Physics502 (2024), p. 112826. issn: 0021-9991.doi:10 . 1016 / j . jcp . 2024 . 112826.url:https : /...

    2024

  15. [23]

    A. S. Joglekar et al.Differentiable Programming for Plasma Physics: From Diagnostics to Discovery and Design. 2026. arXiv:2603.11231 [physics.plasm-ph].url:https://arxiv.org/abs/2603.11231

    2026

  16. [24]

    Loss-cone stabilization in rotating mirrors: thresholds and thermodynamics

    E.J. Kolmes, I.E. Ochs, and N.J. Fisch. “Loss-cone stabilization in rotating mirrors: thresholds and thermodynamics”. In:Journal of Plasma Physics90.2 (2024).doi:10.1017/S0022377824000205

    2024 doi

  17. [25]

    Monte Carlo Gradient in Optimization Constrained by Radiative Transport Equation

    Qin Li, Li Wang, and Yunan Yang. “Monte Carlo Gradient in Optimization Constrained by Radiative Transport Equation”. In:SIAM Journal on Numerical Analysis61.6 (2023), pp. 2744–2774.doi:10. 1137/22M1524515. eprint:https://doi.org/10.1137/22M1524515.url:https://doi.org/10. 1137/...

    2023 doi

  18. [26]

    Dynamical feedback control with operator learning for the Vlasov-Poisson system

    Jingcheng Lu, Li Wang, and Jeff Calder. “Dynamical feedback control with operator learning for the Vlasov-Poisson system”. In:arXiv preprint arXiv:2509.23063(2025)

    2025

  19. [27]

    Collisional losses of electrons from an adiabatic trap in a plasma with a positive potential

    V.P. Pastukhov. “Collisional losses of electrons from an adiabatic trap in a plasma with a positive potential”. In:Nuclear Fusion14.1 (1974), pp. 3–6.doi:10.1088/0029-5515/14/1/001

    1974 doi

  20. [28]

    The magnetic mirror approach to fusion

    R.F. Post. “The magnetic mirror approach to fusion”. In:Nuclear Fusion27.10 (1987), p. 1579.doi: 10.1088/0029- 5515/27/10/001.url:https://iopscience.iop.org/article/10.1088/0029- 5515/27/10/001. [29]Realta Fusion.https://realtafusion.com/. Accessed: 2026-05-21

    1987 doi

  21. [30]

    The semi-Lagrangian method for the numerical resolution of the Vlasov equation

    Eric Sonnendr¨ ucker, Jean Roche, Pierre Bertrand, and Alain Ghizzo. “The semi-Lagrangian method for the numerical resolution of the Vlasov equation”. In:Journal of Computational Physics149.2 (1999), pp. 201–220.doi:10.1006/jcph.1998.6148.url:https://www.sciencedirect.com/scie...

    1999

  22. [31]

    Numerical Analysis of Fluid Estimation for Source Terms in Neutral Particle Simulations

    Zhirui Tang, Emil Lovbak, Julian Koellermeier, and Giovanni Samaey. “Numerical Analysis of Fluid Estimation for Source Terms in Neutral Particle Simulations”. In:Contributions to Plasma Physics n/a.n/a (), e70129.doi:https://doi.org/10.1002/ctpp.70129. eprint:https://onlinelib...

    doi

  23. [32]

    Drift-kinetic PIC simulations of plasma flow and energy transport in the magnetic mirror configuration

    Mikhail Tyushev et al. “Drift-kinetic PIC simulations of plasma flow and energy transport in the magnetic mirror configuration”. In:Physics of Plasmas32.3 (2025)

    2025

  24. [33]

    Complete quasilinear model for the acceleration-driven lower hybrid drift instability and a computational assessment of its validity

    GV Vogman and JH Hammer. “Complete quasilinear model for the acceleration-driven lower hybrid drift instability and a computational assessment of its validity”. In:Physical Review E110.2 (2024), p. 025201

    2024

  25. [34]

    Xishuo Wei et al.Low-dimensional geometry learning for turbulence prediction in optimized stellarators

  26. [2026]

    A Numerical Scheme In this section, we explain in detail the numerical scheme employed to solve (2.1)

    arXiv:2603.17366 [physics.plasm-ph].url:https://arxiv.org/abs/2603.17366. A Numerical Scheme In this section, we explain in detail the numerical scheme employed to solve (2.1). We recall that our phase- space is (z, v(s), µ) for speciess∈ {e, i}which gets discretized into an i...

    arXiv

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.