An Information-Theoretic Reconstruction of Curvature

Abstract

We develop an intrinsic information-theoretic approach for recovering Riemannian curvature from the small-time behaviour of heat diffusion. Given a point and a two-plane in the tangent space, we compare the heat mass transported along that plane with its Euclidean counterpart using the relative entropy of finite measures. We show that the leading small-time distortion of this directional entropy encodes precisely the local curvature of the manifold. In particular, the planar information imbalance determines both the scalar curvature and the sectional curvature at a point, and assembling these directional values produces a bilinear tensor that coincides exactly with the classical Riemannian curvature operator. The method is entirely analytic and avoids Jacobi fields, curvature identities, or variational formulas. Curvature appears solely through the behaviour of heat flow under the exponential map, providing a new viewpoint in which curvature is realized as an infinitesimal information defect of diffusion. This perspective suggests further connections between geometric analysis and information theory and offers a principled analytic mechanism for detecting and reconstructing curvature using only heat diffusion data.