From molecular dynamics to kinetic models: data-driven generalized collision operators in 1D3V plasmas

Yue Zhao, Guosheng Fu, Huan Lei · 2026

Read From molecular dynamics to kinetic models: data-driven generalized collision operators in 1D3V plasmas by Yue Zhao, Guosheng Fu, Huan Lei (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-27828.txt.

Introduction

Zhao and co-authors learn plasma collision rules straight from molecular dynamics. Standard Landau operators miss much of real plasma behavior. Their learned operator is anisotropic and non-stationary. It captures uneven energy transfer from many-body effects. A low-rank tensor form keeps evaluation cheap. Conservation laws hold strictly in the kinetic equation. A second-order scheme conserves mass and total energy fully. Transport coefficients match reference data well.

Catalog record

CREATORYue Zhao, Guosheng Fu, Huan Lei
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE61705
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.27828

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

From molecular dynamics to kinetic models: data-driven generalized collision operators in 1D3V plasmas Yue Zhao, Guosheng Fu, Huan Lei Abstract We present a data-driven approach for constructing generalized collisional kinetic models for inhomogeneous plasmas in one-dimensional physical space and three-dimensional velocity space (1D–3V). The collision operator is directly learned from micro-scale molecular dynamics (MD) and accurately accounts for the unresolved particle interactions over a broad range of plasma conditions. Unlike the standard Landau operator, the present operator takes an anisotropic, non-stationary form that captures the heterogeneous collisional energy transfer arising from the many-body interactions, which is crucial for plasma kinetics beyond the weakly coupled regime. Efficient numerical evaluation is achieved through a low-rank tensor representation with 𝒪 ⁡ ( N ​ log ⁡ N ) \mathcal{O}(N\log N) computational complexity. The constructed kinetic equation strictly preserves conservation laws and physical constraints and therefore, enables us to develop an explicit second-order, energy-conserving scheme that ensures fully discrete conservation of mass and total energy. Numerical results demonstrate that the present model accurately predicts both transport coefficients and several 1D-3V kinetic processes compared with MD simulations across a broad range of densities and temperatures in spatially inhomogeneous settings. This work provides a systematic pathway for bridging micro-scale MD and inhomogeneous plasma kinetic descriptions where empirical models show limitation. 1 Introduction Collisional kinetic theory provides a fundamental framework for modeling the evolution of particle distribution functions in plasmas and rarefied gases, bridging microscopic particle interactions and macroscopic transport phenomena. In spatially inhomogeneous settings, the kinetic description typically consists of a Vlasov-type transport term coupled with a collision operator that accounts for unresolved particle interactions beyond the mean-field approximation. Among various models, the Landau collision operator [ 23 ] has been widely adopted for plasmas with long-range Coulomb interactions, owing to its rigorous conservation properties and consistency with the H-theorem. It can be formally derived from the Boltzmann collision operator [ 4 , 40 ] in the grazing collision limit or from the Balescu–Lenard equation [ 27 , 3 ] under appropriate dielectric screening assumptions. In practical plasma simulations, spatial inhomogeneity plays a crucial role in determining transport processes, wave–particle interactions, and relaxation dynamics. As a result, kinetic equations posed in one-dimensional physical space and three-dimensional velocity space (1D–3V) have become a standard modeling paradigm in plasma physics [ 11 ] . Such models retain the essential velocity-space physics while allowing for nontrivial spatial transport, and have been extensively used in applications ranging from fusion plasmas to space and astrophysical systems. However, incorporating collisions in 1D–3V kinetic models significantly increases both analytical and computational complexity, as the collision operator must be evaluated locally in physical space while preserving its intrinsic high-dimensional structure in velocity space. The numerical solution of collisional kinetic equations in 1D–3V phase space presents substantial challenges. Even for the classical Landau operator, the collision term involves a nonlinear, three-dimensional integro-differential operator evaluated at every spatial grid point, leading to extremely high computational cost. Considerable efforts have been devoted to developing efficient numerical methods for collisional kinetic equations in spatially inhomogeneous settings. Early approaches include finite-difference and finite-volume discretizations combined with conservative formulations to preserve mass, momentum, and energy [ 9 , 24 ] . Entropy-based and structure-preserving schemes have been further developed to ensure discrete consistency with the H-theorem [ 5 , 13 ] . To alleviate the computational burden, various fast algorithms have been proposed for evaluating the Landau collision operator, including Fourier spectral methods that exploit the convolution structure of the isotropic Landau kernel [ 32 , 16 , 41 ] , multigrid and multipole-based approaches [ 6 , 25 ] , Hermite spectral based methods [ 39 , 30 , 29 ] , as well as formulations based on Rosenbluth potentials [ 34 ] coupled with Poisson solvers [ 8 , 37 ] . Particle-based and hybrid methods have also been explored to improve scalability and robustness in high-dimensional settings [ 38 , 7 , 2 , 18 ] . To address the stiffness induced by disparate temporal scales between transport and collisions, implicit and asymptotic-preserving schemes have been proposed [ 14 , 8 , 26 , 15 , 21 ] . Despite these advances, the evaluation of the collision operator remains a dominant computational bottleneck in large-scale 1D–3V kinetic simulations. Beyond computational considerations, a more fundamental limitation arises from the modeling assumptions underlying the Landau collision operator. The Landau formulation is valid in the weakly coupled regime, where binary interactions are dominated by small-angle scattering and particle correlations are negligible. However, in many physically relevant scenarios, such as dense laboratory plasmas, warm dense matter, and inertial confinement fusion, particle correlations and collective effects become increasingly important. In these regimes, the Landau operator may fail to accurately capture collisional relaxation and energy transfer processes, even if it is evaluated with high numerical fidelity. This modeling deficiency poses a significant challenge for spatially inhomogeneous kinetic simulations, where local thermodynamic conditions can vary strongly in space and time. To overcome these limitations, data-driven approaches have recently emerged as a promising pathway for constructing effective collision models informed by microscopic particle dynamics. Molecular dynamics (MD) simulations provide a first-principles description of interparticle interactions beyond weak-coupling assumptions, offering a natural foundation for learning generalized collisional behavior [ 35 , 1 ] . In our previous work [ 42 ] , we introduced a data-driven collision operator for one-component plasmas that generalizes the classical Landau formulation beyond the weakly coupled regime. The operator is characterized by a non-stationary and symmetry-breaking collision kernel that captures correlation-induced anisotropic collisional effects absent from the classical Landau formulation. This generalized collision kernel was shown to accurately model moderately coupled plasma dynamics in spatially homogeneous (0D–3V) settings. In a subsequent study [ 43 ] , we further developed a fast spectral separation representation of the generalized collision kernel, which exploits a low-rank tensor decomposition to recover a convolution structure and enables efficient evaluation with \mathcal{O}(N\log N) computational complexity. However, both the generalized collision model and its spectral separation formulation were restricted to spatially homogeneous 0D–3V systems, and show limitations to model spatially inhomogeneous collisional kinetics. In this work, we aim to extend the data-driven collision framework to spatially inhomogeneous 1D–3V kinetic systems within a general metriplectic formulation [ 31 , 17 ] . In contrast to the homogeneous case, the 1D–3V Vlasov–collision system involves nontrivial coupling between the spatial transport under a self-consistent electric field and the collisional process, which requires a careful balance between Hamiltonian transport and dissipative dynamics at both continuum and discrete levels. From the metriplectic perspective, this extension introduces additional structural challenges: the collision operator must be consistently coupled with the Vlasov dynamics while preserving conservation laws, entropy dissipation, and compatibility with spatial inhomogeneity. In particular, the collision process further depends on the local plasma conditions that introduce spatially heterogeneous dissipative mechanisms. To capture this effect, we construct a generalized collisional kernel directly learned from the micro-scale MD simulations that explicitly captures the dependence on the local macroscopic moments of the distribution function. The developed model provides a consistent 1D–3V kinetic description of collisional plasma dynamics over a broad range of physical regimes. Moreover, the constructed model strictly preserves physical constraints and frame-indifference symmetry conditions, and

Read the complete text