The Dirichlet spectrum of a domain is the collection of special frequencies at which sound (or heat, or quantum wavefunctions) can resonate inside that domain with the boundary held fixed at zero. These eigenvalues form an infinite sequence of positive numbers that grow without bound. A natural question is whether the gaps between consecutive eigenvalues can stay bounded away from zero forever, meaning there is some minimum spacing that no two eigenvalues ever violate. In two dimensions this question is especially delicate because the classical Weyl law, which describes how eigenvalues grow on average, neither forces gaps to stay large nor forces them to shrink to zero.
The paper proves that uniform separation of eigenvalues is actually a rare, exceptional property among planar domains. The authors work with a carefully defined collection of rough domains inside a fixed bounded region, equipped with a natural notion of closeness based on how their complements look geometrically. They show this collection is large and well-behaved in a precise topological sense, specifically that it satisfies the Baire category theorem, which allows them to distinguish between typical and exceptional properties. They also transfer a classical result about generic simplicity of spectra into this rough-domain setting, meaning that for a typical domain in their collection, no two eigenvalues coincide.
The main result combines both of these threads: a typical domain has a simple spectrum, meaning all eigenvalues are distinct, yet the gaps between consecutive eigenvalues still shrink toward zero somewhere in the sequence. The proof works by showing that any domain can be slightly modified, through a careful local surgery, to introduce a pair of consecutive eigenvalues that are extremely close together without being equal. Because this surgery can always be performed and the resulting domains are dense in the collection, the property of having gaps that approach zero is not just common but generic, holding for all domains outside a topologically negligible set.