Diffusion For Fusion

Designing Stellarators with Generative AI

Stellarators are a prospective class of fusion-based power plants that confine a hot plasma with three-dimensional magnetic fields. We show that stellarators can be designed using Denoising Diffusion models, a class of generative models commonly used for image generation.

Appears in the NeurIPS AI4Science Workshop, 2025 ›

STAR_Lite

A Laboratory-Scale Stellarator for Fusion Energy at Hampton University

We introduce the STAR_Lite Stellarator, a laboratory-scale device for studying non-resonant divertors. STAR_Lite is under construction at Hampton University.

Submitted to Journal of Plasma Physics ›

Fast Ion Confinement

Direct Optimization of Fast-Ion Confinement in Stellarators

Keeping fast ions confined inside a stellarator is essential for producing fusion energy, but simulating their trajectories is expensive. We directly optimize stellarator designs by simulating the actual fast ion trajectories, and show that this approach successfully produces configurations with significantly fewer ion losses.

Appears in the Journal of Plasma Physics ›

Variational Inference

Scaling Gaussian Processes with Derivatives

Augmenting Gaussian Processes (GP) with derivative information can significantly enhance their predictive capacity. However, there are serious scalability issues associated with training and inference when using GPs with derivatives, particularly in high dimensions. We developed a Variational Inference technique for training GPs with derivatives that learns a reduced GP model, allowing for scalable training and inference at a minimal loss in accuracy.

Appears in NeurIPS 2021 ›

Multiobjective Optimization

Understanding trade-offs in stellarator design

Every stellarator design involves trade-offs: better particle confinement might require complex coils, while cutting costs can hurt energy production. Understanding these trade-offs is key to building devices that satisfy physical, engineering, and financial requirements all at once. We apply multi-objective optimization to map out these trade-offs and reveal how fundamental design choices influence the overall design.

Appears in the Journal of Plasma Physics›

Constrained Optimization

Modeling Approaches for Addressing Simple Unrelaxable Constraints with Unconstrained Optimization Methods

When optimizing chemical concentrations in a chemical system or probabilities of transmission in an epidemiological simulation, negative concentration levels and negative probabilities, respectively, are readily modeled as belonging to regions that an optimization algorithm should never probe. These constraints are known as unrelaxable constraints, constraints which must be satisfied in order to produce meaningful output of the objective. Solving optimization problems with unrelaxable constraints is difficult, and traditionally requires specialized algorithmic solutions. We have developed modeling approaches which open the doors to "off-the-shelf" optimization routines, such as BFGS, to solve optimization problems with simple unrelaxable constraints.

Appears in Optimization Letters ›

Fusion Energy

Stochastic Optimization of Stellarator Coil Designs

The Stellarator is a state-of-the-art nuclear fusion device designed to generate sustainable energy. We are developing a stochastic optimization framework within FOCUS, an optimization toolbox for designing the magnetic coils for Stellarators. Our accelerated stochastic optimization methods allow for rapid development of coils with low construction tolerances, with the goal of saving millions of dollars in engineering costs.

Appears in the Journal of Plasma Physics ›

Particle Accelerators

Derivative-Free Optimization of a Rapid-Cycling Synchrotron

Fermilab houses a proton accelerator to be used as a part of Deep Underground Neutrino Experiment (DUNE), an international, multi-decadal physics program for leading-edge neutrino science and proton decay studies. However by the inception of DUNE in 2032, the FermiLab proton complex must undergo a substantial upgrade in order to meet the particle accelerator beam power requirements. We develop and solv a derivative-free optimization model which seeks these high power accelerator configurations.

Appears in Optimization and Engineering ›

Microstructures

Fluctuating Hydrodynamics Methods for the Drift-Diffusion Dynamics of Microstructures

As a model for protein motion in lipid-bilayer membranes (cell membranes), we develop fluctuating hydrodynamics approaches for capturing the fluid and microstructure interactions in spherical fluid interfaces. We apply the model model to study the drift-diffusion dynamics of microstructures compare with and without hydrodynamic coupling within the curved fluid interface.

Appears in the Journal of Computational Physics ›