Now
Energy-balance models.
The first stage is simple energy-balance models on an ordinary computer. They do not simulate plasma; they check the concept for consistency: whether there is enough energy for the exhaust, where the rest goes, and which parameters are critical. These models will serve as checkpoints for the more complex simulations.
Next stage · GPU
From single simulations to statistical ensembles.
A single run says nothing about robustness: we need hundreds of variants with scattered initial conditions. That takes GPUs.
| Task | Method | What it gives |
|---|---|---|
| Plasma-jet ensembles | Particle-in-cell and hydrodynamics on GPUs, for example WarpX | How spread in jet speed and timing affects liner symmetry |
| Neutron transport | Monte Carlo, for example OpenMC | How much neutron energy the liner, shield and coils absorb |
| Surrogate models | Neural networks trained on the ensemble data | Fast scans of operating modes instead of long runs for every point |
For this stage we are looking for compute partners: startup programs and research access to supercomputers.