arXiv:2607.02861
Learning Effective Bulk Fields from Fermionic Spectra: Application to an Extended Power-Law Liquid Model for Cuprate Strange Metals
Hong-Zhi Xiao, Zhan-Zhi He, Zhuo-Yu Xian, Shao-Feng Wu
Abstract
We develop a Neural-ODE framework to infer effective bulk metric functions and the charge-weighted gauge potential from boundary fermionic spectra within a static, planar black-hole ansatz. Controlled benchmarks recover the known bulk profiles of the Einstein--Maxwell and Gubser--Rocha models. We then use normalized interaction self-energy targets generated from an extended power-law liquid model with parameters calibrated by angle-resolved photoemission measurements to test the proposed geometric explanation of cuprate strange-metal spectra in a semi-holographic setting. For this application, we use a massless probe in backgrounds with conformal-to-$\mathrm{AdS}_2\times\mathbb{R}^2$ asymptotics and derive a spectral extraction formula from the conserved radial Dirac flux. We identify coordinate/Weyl redundancy and a degeneracy in the scaled Hawking temperature of the normalized inverse problem. The conformal factor remains undetermined, so bulk entropy and the associated contribution to electronic specific heat require independent input. Low-temperature targets are fitted accurately across the studied doping range. In the high-temperature test, however, the present ansatz and optimization search have not yielded a refined geometry that accurately reproduces the thermal spectral shape while preserving frequency symmetry. These results may guide the further development of semi-holographic models for cuprates, and the framework may also find applications in other strongly correlated electron systems.