$Γ$-convergence of a diffeomorphism-natural MDL functional to Einstein-Hilbert with Gibbons-Hawking-York boundary term
Marko Lela
TL;DR
The paper proves a Γ-convergence result showing that a diffeomorphism-natural discrete MDL functional converges to the Einstein–Hilbert action with the Gibbons–Hawking–York boundary term on boundary-fitted, shape-regular meshes. It identifies interior and boundary Carathéodory densities as $f_{in}=\alpha_0+\alpha_1 R$ and $f_{bdry}=\beta_1 K$, and proves the liminf/limsup inequalities via a recovery sequence based on reflected Fermi smoothing, yielding $F(g)=c_0\int_M dV_g + c_1\int_M R_g dV_g + c_2\int_{\partial M} K_g dS_g$ with $c_0=\alpha_0$, $c_1=\alpha_1$, and $c_2=\beta_1$. A boundary first-layer analysis shows leading $O(h^{d-1})$ contributions per boundary cell and a global $O(h)$ boundary remainder, while the interior remainder is $O(h^2)$. The results are established under equi-coercivity, bounded geometry, and boundary regularity, and Appendix F provides a reproducible protocol for rate checks and calibration of the constants. This work situates EH+GHY as the natural continuum limit of a data-efficient, diffeomorphism-natural discretization, advancing the MIS program's aim to derive continuum physics from information-theoretic discretizations.
Abstract
We prove a \(Γ\)-convergence result for a diffeomorphism-natural discrete MDL-type functional to the Einstein-Hilbert action with the Gibbons-Hawking-York boundary term. On boundary-fitted, shape-regular meshes we establish interior and boundary blow-ups, identify the Carathéodory densities \(f_{\mathrm{in}}=α_0+α_1 R\) and \(f_{\mathrm{bdry}}=β_1 K\), and obtain the \(\liminf/\limsup\) bounds via a recovery sequence based on reflected Fermi smoothing. A boundary first-layer asymptotics shows that boundary cells contribute at order \(h^{d-1}\), yielding a global \(O(h)\) boundary remainder, while the interior remainder is \(O(h^2)\). The paper is foundational; Appendix~E specifies a reproducible protocol for rate checks and calibration of \(α_0,α_1,β_1\).
