On the energy image density conjecture of Bouleau and Hirsch
Sylvester Eriksson-Bique, Mathav Murugan
TL;DR
The paper resolves the energy image density conjecture of Bouleau and Hirsch by developing two robust proofs—one via De Philippis–Rindler normal currents and a second via a lower semicontinuity framework enriched by Alberti–Marchese decomposability bundles and cone-null sets—applying them across regular, local $p$-Dirichlet spaces, $p$-Dirichlet structures, and Sobolev-type fractal energies. Central to the approach is the carré du champ matrix $\\gamma$, whose invertibility or $p$-independence guarantees that pushforwards of energy measures are absolutely continuous with respect to Lebesgue measure; this yields the $EID$ property in broad settings. The results yield concrete applications including the finiteness of martingale dimension under sub-Gaussian heat kernel bounds and a new proof of Cheeger’s conjecture for PI spaces, demonstrating how $EID$ bridges stochastic analysis, geometric measure theory, and metric-space analysis. Overall, the work unifies several nonlinear and linear energy frameworks under a common criterion and opens avenues for dimension theory and regularity in diverse energy settings, including fractals and metric-measure spaces.
Abstract
We affirmatively resolve the energy image density conjecture of Bouleau and Hirsch (1986). Beyond the original framework of Dirichlet structures, we establish the energy image density property in several related settings. In particular, we formulate a version of the property that encompasses strongly local, regular Dirichlet forms, Sobolev spaces defined via upper gradients, and self-similar energies on fractals, thereby unifying these under a single framework. As applications, we prove the finiteness of the martingale dimension for diffusions satisfying sub-Gaussian heat kernel bounds, and we obtain a new proof of a conjecture of Cheeger concerning the Hausdorff dimension of the images of differentiability charts in PI spaces. The proof of the energy image density property is based on a structure theorem for measures and normal currents in $\mathbb{R}^n$ due to De Philippis--Rindler, together with the notions of decomposability bundles due to Alberti--Marchese and cone null sets due to Alberti--Csörnyei--Preiss and Bate.
