by Estêvão Gomes (@EstevaoMGomes)

This project is a Stellarator Coil Optimizer of alpha particles via differentiable JAX code and was developed as the research work for the New Talents in Physics Fellowship, awarded by the Calouste Gulbenkian Foundation.

The project was developed under the supervision of professor Rogério Jorge (@rogeriojorge).

In magnetic confinement fusion, one of the most promising approaches that allows steady state operation with no disruptions is the stellarator. A stellarator consists of electromagnetic coils that create a twisted magnetic field that needs to be optimized to confine a high-performing plasma. Such optimization is performed over a large set of parameters, typically of the order of several hundred or more. Furthermore, the target magnetic field is usually a fixed one, that has been previously obtained using another optimization based on the ideal MHD equations. With this work, we trace particles directly in the corresponding Biot-Savart magnetic fields stemming from a set of coils and optimize them to yield a small fraction of lost particles outside of the confinement region. Furthermore, to replace the need for hundreds of simulations per optimization step, we make use of automatic differentiation by implementing the guiding-center equations, magnetic field solver, and optimization routines in JAX. This allows us to streamline optimization efforts, and create a specialized, but very fast, numerical tool, to optimize force-free stellarator equilibria. As force-free equilibria are usually the first step in determining the viability of a device, such optimizations will be able to guide future designs based on ideal MHD equilibria.

After cloning the repository:

git clone https://github.com/uwplasma/ESSOS.git

The easiest way to run an example script is to create a conda environment. To install conda run on the terminal:

wget https://repo.anaconda.com/archive/Anaconda3-2024.06-1-Linux-x86_64.sh

Or install any other version avalaible on https://repo.anaconda.com/archive/. Then Run:

bash Anaconda3-2024.06-1-Linux-x86_64.sh -b -p ~/anaconda
rm Anaconda3-2024.06-1-Linux-x86_64.sh
echo 'export PATH="~/anaconda/bin:$PATH"' >> ~/.bashrc 

To refresh, run

source .bashrc

Then you can create a conda environment with

conda create --name myenv --file dependencies-file.yml python=3.12.2

where "myenv" is the name you want your environment to have and "dependencies-file.yml" can be dependencies-gpu.yml or dependencies-cpu.yml, whether you want to run the scripts on CPU or GPU.