Speaker: Alan Kaptanoglu

Title: Sparse regression and local basis functions in stellarator optimization problems  

Abstract: "Electromagnetic coils can be designed by using volume elements (voxels) of electric current, constrained so the current is divergence-free. I will discuss how this voxel method can be used to design electromagnetic coils for a class of plasma experiments known as stellarators. We produce topologically-exotic coils for several new stellarator designs and show that these solutions can be interpolated into a filamentary representation and then further optimized. Very similar optimization problems can be formulated for designing permanent magnet arrays for magnetostatic problems. We explore a class of greedy algorithms and use them to improve the original PM4Stell project for designing permanent magnet arrays for a quadrant of a 0.5 Tesla scaled NCSX stellarator. Unlike previous methods, our new greedy algorithms can run extremely fast, require no post-processing steps, and match the errors of the previous solution found for PM4Stell, but use 30% fewer magnets. Lastly, I will discuss recent work using local basis functions for calculating 3D MHD equilibria."

Bio: "Alan Kaptanoglu, Assistant Professor of Mathematics at the Courant Institute at NYU, works at the intersection of numerical methods, machine learning, and fusion energy. He received a B.S. in Physics from Stanford University in 2016 and a Ph.D. in Physics from the University of Washington in 2021. Most recently, he was a Postdoctoral Researcher at the University of Maryland, working on stellarator optimization problems."

This talk will be offered in a hybrid format. If you wish to participate remotely, please send an email to [email protected].