A Hybrid semi-Lagrangian Flow Mapping Approach for Vlasov Systems: Combining Iterative and Compositional Flow Maps

We propose a hybrid semi-Lagrangian scheme for the Vlasov–Poisson equation that combines the Numerical Flow Iteration (NuFI) method with the Characteristic Mapping Method (CMM). Both approaches exploit the semi-group property of the underlying diffeomorphic flow, enabling the reconstruction of solutions through flow maps that trace characteristics back to their initial positions. NuFI builds this flow map iteratively, preserving symplectic structure and conserving invariants, but its computational cost scales quadratically with time. Its advantage lies in a compact, low-dimensional representation depending only on the electric field. In contrast, CMM achieves low computational costs when remapping by composing the global flow map from explicitly stored submaps. The proposed hybrid method merges these strengths: NuFi is employed for accurate and conservative local time stepping, while CMM efficiently propagates the solution through submap composition. This approach reduces storage requirements, maintains accuracy, and improves structural properties. Numerical experiments demonstrate the effectiveness of the scheme and highlight the trade-offs between memory usage and computational cost. We benchmark against a semi-Lagrangian predictor-corrector scheme used in modern gyrokinetic codes, evaluating accuracy and conservation properties.

💡 Research Summary

The paper introduces a novel hybrid semi‑Lagrangian scheme for solving the Vlasov–Poisson (VP) system by combining two recent flow‑map based methods: Numerical Flow Iteration (NuFI) and the Characteristic Mapping Method (CMM). The authors begin by reviewing the challenges inherent in kinetic plasma simulations, namely the curse of dimensionality, the formation of fine filamentary structures, and the difficulty of preserving physical invariants such as charge and energy. Traditional Eulerian, semi‑Lagrangian (SL), and particle‑in‑cell (PIC) approaches each suffer from either excessive memory consumption, numerical diffusion, or statistical noise.

NuFI addresses some of these issues by storing only the time history of the electric field and reconstructing the backward characteristic map Φ₀ᵗ through an iterative Strömer–Verlet scheme. The method yields a second‑order accurate flow map Ψ₀ᵗ, preserves the symplectic structure, and requires far less memory than storing the full distribution function f. However, because each new time step requires an additional backward iteration, the overall computational cost grows quadratically with the number of steps, making long‑time integrations expensive.

CMM, on the other hand, decomposes the global flow into a sequence of sub‑maps Φᵢ defined on short intervals. Each sub‑map is represented as a low‑order Hermite polynomial (typically p = 3) and advanced using a gradient‑augmented level‑set technique. By composing these sub‑maps, the full flow can be reconstructed with large time steps while keeping both memory and computational effort linear in the number of sub‑maps. CMM thus offers excellent scalability, especially on GPU‑accelerated platforms, but it does not inherently guarantee the same level of symplectic preservation as NuFI.

The hybrid algorithm leverages the strengths of both methods. In the early phase of a simulation, NuFI is used to obtain highly accurate, structure‑preserving local updates. After a predefined number of NuFI steps, the accumulated sub‑maps are compressed and merged using the CMM framework. This “restart‑and‑compress” strategy reduces the quadratic growth of NuFI to a linear cost while retaining its conservative properties. The authors provide a detailed algorithmic description, including the handling of electric‑field interpolation, half‑step velocity updates, and the composition of flow maps.

Numerical experiments focus on two classic VP test cases: linear Landau damping and the two‑stream instability. The hybrid method is benchmarked against a state‑of‑the‑art SL predictor‑corrector scheme that employs cubic B‑spline interpolation (the basis of the Gysela gyrokinetic code). Results show that the hybrid approach achieves comparable or better L₂ error norms with significantly reduced memory footprints (≈30 % of the CMM‑only implementation) and faster runtimes (2–3× speed‑up over pure NuFI). Moreover, charge and energy conservation errors remain at the 10⁻⁴ level, demonstrating that the symplectic nature of NuFI is effectively retained after CMM compression.

In the concluding section, the authors argue that flow‑map based hybridization provides a promising pathway toward exascale kinetic plasma simulations. They outline future work, including extensions to multi‑species and higher‑dimensional VP systems (e.g., 2D+2V, 3D+3V), adaptive sub‑map refinement, and application to the full Vlasov–Maxwell equations. The paper thus contributes a concrete, well‑validated methodology that bridges the gap between high‑fidelity, structure‑preserving solvers and the practical constraints of memory and compute resources in modern plasma physics.