Advancing Magnetized Target Fusion by Solving an Inverse Problem with COMSOL Multiphysics

General Fusion's Magnetized Target Fusion approach involves compressing a spherical tokamak plasma to achieve fusion conditions. This presentation will detail how the COMSOL Multiphysics® software was used in the LM26 fusion demonstration, which has been in operation since February 2025. Initially, COMSOL® was used to model the magnetomechanical compression of small-scale lithium rings and cylinders. These 2D axisymmetric models, which coupled the nonlinear solid mechanics, magnetic field, and heat transfer modules, were validated against high-speed imagery and laser diagnostics from those experiments. The validated models were then instrumental in defining the LM26 compressor design and operating conditions.

A central challenge is that plasma equilibrium characteristics and lithium liner model parameters need to be adjusted during a compression shot. While some material testing was conducted on lithium samples in tension and compression, the investigated range was insufficient to cover all the experimental conditions of LM26. To overcome this, a Bayesian inference reconstruction process was used to solve an inverse problem. This process began by recreating the lithium liner's compression sequence using a parametric sweep of COMSOL Multiphysics models, which were constrained by experimental structured light reconstruction (SLR) and photon doppler velocimetry (PDV) measurements in LM26. This method enabled General Fusion to provide precise magnetic flux boundary conditions to internal Grad–Shafranov magnetohydrodynamic (MHD) solvers. The MHD solvers, in turn, were used to reconstruct the plasma equilibrium and determine the plasma density profiles needed to calculate its temperature. This work is critical, as General Fusion aims to have LM26 reach a plasma temperature of 1 keV, with a future goal of 10 keV.

*Please see www.comsol.com/privacy for COMSOL's Privacy Policy. Contact COMSOL at www.comsol.com/contact for more information.