ICNSP 2025 Summary

The Totality of ICNSP 2025: Charting the Computational Frontier of Fusion Plasma

The pursuit of practical fusion energy stands as one of humanity's greatest scientific and engineering challenges. At General Physics, we understand that unlocking this potential hinges on our ability to model and predict the complex, non-linear dynamics of plasma. This commitment to computational fidelity was powerfully reaffirmed at the recent International Conference on Numerical Simulation of Plasmas (ICNSP 2025).

Held from November 12–14, 2025, on the vibrant campus of the University of California, Irvine, this meeting served as an essential follow-up to the critical 2024 Theory of Fusion Plasmas Conference in Varenna. The discussions focused intensely on the cutting-edge of numerical simulation methods, plasma physics, and advances in computational models, offering a panoramic view of the field's current trajectory.

Bridging Scales: Innovations in Kinetic and Fluid Dynamics

The conference showcased tremendous progress in refining the computational methods that capture plasma behavior across vast spatial and temporal scales.

A major thematic thread centered on Kinetic Theory and Vlasov Dynamics. Nanyi Zheng (University of Delaware) presented a significant step forward with a semi-Lagrangian adaptive rank method, leveraging tensor products and matrix mechanics to tackle high-dimensional Vlasov dynamics with unprecedented efficiency. This push for structure-preserving accuracy was echoed by Jingwei Hu, who discussed similar methods for the challenging Vlaslov-Maxwell-Landau equation in collisional plasmas. Further exploring the kinetic realm, Uri Shumlak (University of Washington) detailed the physics of multi-species ring plasmas, while the quantum-level complexity was addressed by Andrew Christlieb and Sining Gong (Michigan State), focusing on adaptive-rank solutions for quantum Boltzmann models and the intricacies of Wigner Poisson and Fermi-Dirac quantum fluid models.

Simultaneously, the development of robust Magnetohydrodynamic (MHD) and Fluid models continues to evolve. Byung Kyu Na (Max Planck Institute) introduced STRUPHY, a Python resource for structure-preserving hybrid MHD simulations. Addressing turbulence, Micol Bassanini (EPFL) showcased advanced IMEX integration for tokamak edge turbulence, while Haotian Chen (Peking University) presented work on zonal flow dynamics and geometric effects in turbulent transport. Even feedback systems were scrutinized, with Gijs Derks (Dutch Institute for Fundamental Energy Research) detailing characteristic timescales for feedback systems using TCM and SOLDECS methodologies.

The Computational Engine: HPC, AI, and Next-Gen Algorithms

The true power of modern plasma physics lies in its synergy with high-performance computing (HPC) and artificial intelligence (AI).

William Tang (Princeton) offered a forward-looking review of national lab efforts, emphasizing HPC, scaling laws, Verification, Validation, and Uncertainty Quantification (VVUQ), and the emergence of digital twin simulations and "Fusion GPT." This drive for computational muscle was evident in the development of particle-in-cell (PIC) methods: Justin Angus (Lawrence Livermore) highlighted the powerful WARPX code for high-performance collisional impact PIC, complemented by Dan Barnes (Tokamak Energy) with new Darwin energy-conserving PIC algorithms for WARPX. Jean-Luc Vay (Lawrence Berkeley) contributed to the algorithmic advancement by discussing energy-preserving coupling of explicit PIC with Monte Carlo collisions.

Furthermore, several speakers demonstrated novel ways to make simulations more tractable:

* Golo Wimmer (Los Alamos) introduced efficient finite-element constraint solvers for MHD.

* Luis Chacon (Los Alamos) tackled the sheer "tyranny of scales" in simulations, particularly in the context of Hall MHD.

* Peiyi Chen (University of Wisconsin) demonstrated the use of neural networks to enforce the Penrose condition in magnetized plasma instability analysis.

* Kentaro Hara (Stanford University) explored sequential data assimilation for plasma dynamics, a crucial technique for combining simulation with experimental data.

Synthesis and Future Directions

The conference served as a vital forum for comparison and collective advancement. Xueqiao Xu (Lawrence Livermore) led a key discussion focusing on the comparison and deployment of major simulation tools, including BOUT++, GEM, SOLPS, and VPIC. In the same vein, Ayumi Takano & Tomohiko Watanabe (Nagoya University) discussed challenges in reconciling MHD codes (BOUT, JOREK, Nimrod) with complex toroidal plasma symmetry issues.

Finally, the analytical methods for interpreting complex simulation output are becoming increasingly sophisticated. Alexandra Dudkovskaia (University of York) presented an elegant application of dynamic-mode decomposition and trikinetic eigenmode analysis to electromagnetic fusion plasmas.

The breadth and depth of ICNSP 2025—from the foundational mathematics of Logan Tate Meredith's kinetic theory to the massive computational challenges discussed by Xiahuo Wei on turbulence transport—demonstrated a vibrant, rapidly evolving field. Though the interesting talks by Yao Zhao (Shanghai Jiao Tong) and Ion Farcas (Virginia Tech) await separate detailed publication, the collective momentum from UC Irvine is clear.

The fusion community is not just iterating on existing models; it is fundamentally transforming how we approach the physics problem, driven by the revolutionary potential of modern computational science. General Physics proudly stands ready to leverage these new models and algorithms as we continue to support the mission of making fusion energy a reality.