arXiv · 2611.14650
Ernst-Potential Driven Recovery of Kerr with Neural Networks
Abstract
We use artificial neural networks to recover the Kerr solution by directly minimizing a regularized Ernst sigma-model action. Reduction along the axial Killing direction avoids the divergence at the ergosphere. Compactified coordinates cover the entire black-hole exterior, while the network parametrization enforces the asymptotic data and axis regularity. For dimensionless spins $0.1 \le a/M \le 0.99$, the median maximum relative error in the recovered metric component $g_{ϕϕ}$ remains below $7.2\times10^{-4}$. When only the horizon rod length and angular momentum are prescribed, the networks also predict the mass with median relative errors below $1.8\times10^{-3}$. These results demonstrate the practicality of direct variational training for rotating black-hole geometries.
Explore related subjects
Keep this discovery
Explore connections, maps & timelines
Veselin G. Filev, Stoytcho S. Yazadjiev. 2026-11-10. Ernst-Potential Driven Recovery of Kerr with Neural Networks. https://arxiv.science/abs/2611.14650
Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite arXiv Science by Indiciae: BibTeX · RIS.