arXiv subjects

Kyungtak Lim

Publications and source records attributed to Kyungtak Lim.

8 recordsLinked to original sources

arXiv2026

Surrogate modeling of drift-reduced Braginskii turbulence with resistivity-conditioned Koopman neural operators

Machine-learning-driven surrogate operators are developed for three-dimensional, nonlinear, flux-driven simulations of boundary plasma turbulence based on the two-fluid drift-reduced Braginskii model. Resistivity-conditioned Koopman neural operators (KNOs) are trained on Global Braginskii Solver (GBS) simulations, spanning low- to high-resistivity regimes. Separate fieldwise models are constructed for plasma density, electron temperature, electric potential, and vorticity. Evaluation at a held-out resistivity shows that the surrogates reproduce key short-horizon statistical features, including strong one-step agreement, spectral trends, and reduced pressure-gradient diagnostics. Field-dependent limitations remain, with vorticity showing the largest discrepancies and autoregressive rollout progressively departing from the reference simulation. The results demonstrate that resistivity-conditioned fieldwise neural operators provide useful fast emulators for selected boundary-plasma turbulence diagnostics, while stable long-horizon dynamical closure remains unresolved.

physics.plasm-ph↗

arXiv2026

Plasma GraphRAG: Physics-Grounded Parameter Selection for Gyrokinetic Simulations

Accurate parameter selection is fundamental to gyrokinetic plasma simulations, yet current practices rely heavily on manual literature reviews, leading to inefficiencies and inconsistencies. We introduce Plasma GraphRAG, a novel framework that integrates Graph Retrieval-Augmented Generation (GraphRAG) with large language models (LLMs) for automated, physics-grounded parameter range identification. By constructing a domain-specific knowledge graph from curated plasma literature and enabling structured retrieval over graph-anchored entities and relations, Plasma GraphRAG enables LLMs to generate accurate, context-aware recommendations. Extensive evaluations across five metrics, comprehensiveness, diversity, grounding, hallucination, and empowerment, demonstrate that Plasma GraphRAG outperforms vanilla RAG by over $10\%$ in overall quality and reduces hallucination rates by up to $25\%$. {Beyond enhancing simulation reliability, Plasma GraphRAG offers a methodology for accelerating scientific discovery across complex, data-rich domains.

physics.plasm-ph↗

arXiv2026

Machine learning prediction of plasma behavior from discharge configurations on WEST

Accurately predicting plasma behavior based on discharge configurations is essential for the safe and efficient operation of tokamak experiments. While physics-based integrated modeling codes provide valuable insights, their high computational cost limits their applicability for fast scenario design and control optimization. In this study, we propose a transformer-based machine learning model to predict key global plasma parameters on the WEST tokamak, including the normalized beta ($β_{n}$), toroidal beta ($β_{t}$), poloidal beta ($β_{p}$), plasma stored energy ($W_{\mathrm{mhd}}$), safety factor at the magnetic axis ($q_{0}$), and safety factor at the 95% flux surface ($q_{95}$). The model uses only signals that can be defined before the discharge, such as magnetic coil currents, auxiliary heating power, plasma current reference, and line-averaged plasma density. Trained on 550 discharges from the WEST campaigns, the model demonstrates an average mean square error (MSE) loss of 0.026, an average coefficient of determination $R^{2}$ of 0.94, and achieves inference times on the order of 0.1 seconds. These results highlight the potential of data-driven surrogate models for assisting in discharge planning, scenario evaluation, and real-time control of tokamak plasmas.

physics.plasm-ph↗

arXiv2025

Reconstructing High-fidelity Plasma Turbulence with Data-driven Tuning of Diffusion in Low Resolution Grids

Developing physically consistent closure models is a longstanding challenge in simulating plasma turbulence, even in minimal systems such as the two-field Hasegawa-Wakatani (HW) model, which captures essential features of drift-wave turbulence with a reduced set of variables. In this work, we leverage theoretical insights from Direct Interaction Approximation (DIA) to construct a six-term closure structure that captures the dominant turbulent transport processes, including both diffusion and hyper-diffusion. While the mathematical form of the closure is fully prescribed by DIA, the corresponding transport coefficients are learned from data using physics-informed neural networks (PINNs). The resulting Extended HW model with Closure (EHW-C) model reveals several nontrivial features of plasma turbulence: notably, some inferred coefficients become negative in certain regimes, indicating inverse transport, a phenomenon absent in conventional closure models. Moreover, the EHW-C model accurately reproduces the spectral and flux characteristics of high-resolution Direct Numerical Simulations (DNS), while requiring only one-eighth the spatial resolution per direction, yielding a tenfold speed-up. This work demonstrates how theory-guided machine learning can both enhance computational efficiency and uncover emergent transport mechanisms in strongly nonlinear plasma systems.

physics.plasm-ph↗

arXiv2025

A high-fidelity surrogate model for the ion temperature gradient (ITG) instability using a small expensive simulation dataset

One of the main challenges in building high-fidelity surrogate models of tokamak turbulence is the substantial demand for high-quality data. Typically, producing high-quality data involves simulating complex physical processes, which requires extensive computing resources. In this work, we propose a fine tuning-based approach to develop the surrogate model that reduces the amount of high-quality data required by 80\%. We demonstrate the effectiveness of this approach by constructing a proof-of-principle ITG surrogate model using datasets generated from two gyrokinetic codes, GKW and GX. GX needs in terms of computing resources are much lighter than GKW. Remarkably, the surrogate models' performance remain nearly the same whether trained on 798 GKW results alone or 159 GKW results plus an additional 11979 GX results. These encouraging outcomes indicate that fine tuning methods can significantly decrease the high-quality data needed to develop the simulation-driven surrogate model. Moreover, the approach presented here has the potential to facilitate surrogate model development for heavy codes and may ultimately pave the way for digital twin systems of tokamaks.

physics.plasm-ph↗

arXiv2024

Effect of negative triangularity on SOL plasma turbulence in double-null L-mode plasmas

The effects of negative triangularity (NT) on boundary plasma turbulence in double-null (DN) configurations are investigated using global, nonlinear, three-dimensional, flux-driven two-fluid simulations. NT plasmas exhibit suppressed interchange-driven instabilities, resulting in enhanced confinement and lower fluctuation levels compared to positive triangularity (PT) plasmas. This reduction in interchange instability is associated with the weakening of curvature effects in the unfavorable region, caused by the stretching of magnetic field lines at the outer midplane. The magnetic disconnection between the turbulent low-field side (LFS) and the quiescent high-field side (HFS) results in most of the heat flux reaching the DN outer targets. In NT plasmas, the power load on the outer target is reduced, while it increases on the inner target, indicating a reduced in-out power asymmetry compared PT plasmas. Furthermore, the analysis of power load asymmetry between the upper and lower targets shows that the up-down power asymmetry is mitigated in NT plasmas, mainly due to the reduced total power crossing the separatrix. The reduction of interchange instabilities in NT plasmas also affects the blob dynamics. A three-dimensional blob analysis reveals that NT plasmas feature smaller blob sizes and slower propagation velocities. Finally, an analytical scaling law for blob size and velocity that includes plasma shaping effects is derived based on the two-region model and is found to qualitatively capture the trends observed in nonlinear simulations.

physics.plasm-ph↗

arXiv2023

Self-consistent gyrokinetic modelling of turbulent and neoclassical tungsten transport in toroidally rotating plasmas

The effect of toroidal rotation on both turbulent and neoclassical transport of tungsten (W) in tokamaks is investigated using the flux-driven, global, nonlinear 5D gyrokinetic code GYSELA. Nonlinear simulations are carried out with different levels of momentum injection that drive W to the supersonic regime, while the toroidal velocity of the main ions remains in the subsonic regime. The numerical simulations demonstrate that toroidal rotation induces centrifugal forces that cause W to accumulate in the outboard region, generating an in-out poloidal asymmetry. This asymmetry enhances neoclassical inward convection, which can lead to central accumulation of W in cases of strong plasma rotation. The core accumulation of W is mainly driven by inward neoclassical convection. However, as momentum injection continues, roto-diffusion, proportional to the radial gradient of the toroidal velocity, becomes significant and generate outward turbulent flux in the case of ion temperature gradient (ITG) turbulence. Overall, the numerical results from nonlinear GYSELA simulations are in qualitative agreement with the theoretical predictions for impurity transport, as well as experimental observations.

physics.plasm-ph↗

arXiv2023

Effect of triangularity on plasma turbulence and the SOL-width scaling in L-mode diverted tokamak configurations

The effect of triangularity on tokamak boundary plasma turbulence is investigated by using global, flux-driven, three-dimensional, two-fluid simulations. The simulations show that negative triangularity stabilizes boundary plasma turbulence, and linear investigations reveal that this is due to a reduction of the magnetic curvature drive of interchange instabilities, such as the resistive ballooning mode. As a consequence, the pressure decay length $L_p$, related to the SOL power fall-off length $λ_q$, is found to be affected by triangularity. Leveraging considerations on the effect of triangularity on the linear growth rate and nonlinear evolution of the resistive ballooning mode, the analytical theory-based scaling law for $L_p$ in L-mode plasmas, derived by Giacomin \textit{et al.} [{Nucl. Fusion}, \href{https://doi.org/10.1088/1741-4326/abf8f6}{\textbf{61} 076002} (2021)], is extended to include the effect of triangularity. The scaling is in agreement with nonlinear simulations and a multi-machine experimental database, which include recent TCV discharges dedicated to the study of the effect of triangularity in L-mode diverted discharges. Overall, the present results highlight that negative triangularity narrows the $L_p$ and considering the effect of triangularity is important for a reliable extrapolation of $λ_q$ from present experiments to larger devices.

physics.plasm-ph↗