Learning the Emergent Bulk Geometry of the Three Dimensional Ising Model

Abstract

We use Monte Carlo data for the three-dimensional Ising model at the critical temperature to predict an emergent radial geometry of the bulk using machine learning. Specifically, we employ a differentiable solver for a probe field on an asymptotically $\mathrm{AdS}_4$ background, trained to reproduce the measured momentum-space two-point functions of the boundary operators ($σ$ and $ε$). We find that a single metric cannot simultaneously serve both operators. Fitting each channel separately improves its own residual by a factor of three to five, but degrades the other by up to two orders of magnitude. We quantify this tension between the two operators and show that it survives changes to the momentum window and the lattice size. Our results numerically demonstrate that the bulk geometry of a theory with an $\mathcal{O}(1)$ central charge, such as the 3D Ising model, is not captured by a single classical metric.