On the statistical nature of Betti numbers and Euler characteristic of smooth random fields

Abstract

We represent excursion sets of smooth random fields as unions of a topological basis consisting of a sequence of simply and multiply connected compact subsets of the underlying manifold. The associated coefficients, which are non-negative discrete random variables, reflect the randomness of the field. Betti numbers of the excursion sets can be expressed as summations over the coefficients, and the Euler characteristic and the sum of Betti numbers can also be expressed as their (alternating) sum. This enables understanding their statistical properties as sums (or differences) of discrete random variables. We examine the conditions under which each topological statistic can be asymptotically Gaussian as the size of the manifold and the resolution increase. The coefficients of the basis elements are then modeled as Binomial variables, and the statistical natures of Betti numbers, Euler character and sum of Betti numbers follow from this fundamental property. We test the validity of the modeling using numerical calculations, and identify threshold regimes where the topological statistics can be approximated as Gaussian variables. The new representation of excursion sets thus maps the properties of topological statistics to combinatorial structures, thereby providing mathematical clarity on their use for physical inference, particularly in cosmology.