Speaker
Description
We discuss numerical simulations of relativistic spin hydrodynamics solved by physics-informed neural networks (PINNs) [1]. We take advantage of PINNs to explicitly impose the total angular momentum, i.e., the sum of orbital and spin angular momentum, accurately conserved by including the conservation law directly in the loss function as a training target. This enables controlled numerical studies of the spin-orbit conversion driven by the rotational viscous effect. We study both orbital-to-spin and spin-to-orbital conversion, analogous to the Barnett and Einstein–de Haas effects, respectively. Our results provide a numerical demonstration of spin–orbit conversion in fully nonlinear relativistic spin hydrodynamics.
[1] Hidefumi Matsuda, Koichi Hattori, Koichi Murase, arXiv: 2512.17971
