We are a computational research group in the Department of Computational Mathematics, Science and Engineering (CMSE) at Michigan State University, led by Prof. Michael S. Murillo. Our work sits at the intersection of physics, applied mathematics, and data science, with a shared focus on understanding the collective dynamics of interacting many-body systems.
Plasma Physics and Kinetic Theory
Our core research area is the physics of non-ideal and strongly coupled plasmas — systems where particle interactions are so strong that standard approximations break down. We develop and apply kinetic theory, molecular dynamics, and hydrodynamic models to understand transport properties, thermodynamic structure, and non-equilibrium dynamics in plasmas ranging from ultracold neutral plasmas to warm dense matter and inertially confined fusion environments. Recent work includes tensor network methods for high-dimensional kinetic equations, quantum statistical potentials for dense plasmas, and multispecies BGK kinetic models for charged-particle transport. Our open-source molecular dynamics package Sarkas is freely available for the plasma physics community.
Machine Learning and Ensemble Modeling
We develop and apply machine learning methods to accelerate and improve scientific modeling across physical and biological systems. A central focus is on hybrid approaches that couple physics-based models with machine learning to achieve accuracy and generalizability that neither achieves alone. Recent highlights include an ensemble modeling system — coupling five ecosystem models with four machine learning algorithms — that substantially improves predictions of nitrous oxide emissions from US croplands, published in PNAS (2026). We also develop surrogate models and multi-fidelity methods for plasma transport data, and apply data-driven approaches to electrical conductivity and equation-of-state problems in dense plasmas.
Agent-Based and Sociophysical Modeling
We use the mathematical tools of statistical mechanics — generalized Langevin equations, memory kernels, network theory — to study collective behavior in social and sociophysical systems. Current work develops an exact mapping from linear agent-based models to generalized Langevin equations, revealing how unobserved environmental agents create memory effects and how network topology shapes collective dynamics. We apply this framework to understanding covert influence operations and opinion dynamics. This thread connects directly to our plasma physics work: the same mathematical structures that describe strongly coupled charges also describe interacting agents.
Join Us
We have openings for motivated Ph.D. students with backgrounds in physics, applied mathematics, computational science, or related fields. If you are interested in working at the frontier of plasma physics, scientific machine learning, or computational social science, please reach out to Prof. Murillo directly.