A Poisson-Voronoi tessellation is a way of dividing space into random cells. You scatter points randomly throughout space according to a Poisson process, then assign every location to whichever seed point is nearest. The resulting patchwork of regions is called a Voronoi tessellation, and studying the typical cell, meaning a randomly chosen cell, is a classical problem in stochastic geometry. One quantity of particular interest is the volume of that typical cell, which turns out to be surprisingly difficult to characterize exactly in higher dimensions.
The paper's central contribution is a clean mathematical decomposition of the typical cell volume into two independent pieces, conditioned on knowing how many faces the cell has. One piece captures the overall size of the cell and follows a Gamma distribution, which is a well-understood and tractable probability law. The other piece captures only the shape of the cell, stripped of its size. Because these two pieces are independent given the face count, many useful formulas fall out directly: exact expressions for moments and transforms of the volume, bounds on how the cell relates to a companion geometric object called the Voronoi flower, and a precise accounting of where randomness comes from. Specifically, two separate sources of variability are identified: shape fluctuations when the number of faces is held fixed, and additional randomness from the face count itself varying.
The paper also makes progress on understanding the probability that a cell is extremely small, which is a delicate question tied to rare geometric configurations. The authors reduce this lower-tail problem to two more manageable mathematical conditions and derive the leading-order asymptotic behavior for small volumes given a fixed face count. They verify their framework exactly in one dimension, work out explicit formulas in two dimensions, and support the theoretical findings with simulations in two, three, and four dimensions, giving confidence that the results extend usefully across different spatial settings.