The Dirichlet spectrum is a set of numbers that arises in a classical problem in number theory: how well can real numbers be approximated by fractions? For any real number, you can measure the quality of its rational approximations in a precise way, producing a single numerical score called its Dirichlet value. The collection of all possible such scores forms the Dirichlet spectrum. Mathematicians have known for some time that this spectrum has an interesting structure: beyond a certain threshold value, every number belongs to the spectrum, forming what is called a "ray." Below that threshold, the spectrum has gaps and isolated points scattered in a complicated pattern. Understanding exactly where this threshold sits and what the spectrum looks like near it is the central challenge.
This paper makes progress on several fronts. The authors narrow down the location of the threshold value, proving it lies between roughly 5.96 and 18.48, improving on previous estimates. They identify a concrete interval of numbers that are completely absent from the spectrum, which gives the improved lower bound. Finding this gap required a computer-assisted search through a finite but large collection of objects called continued-fraction words, which are the main tool for encoding how well numbers can be approximated by fractions. The upper bound comes from constructing explicit real numbers whose Dirichlet values are demonstrably close to the threshold.
The paper also resolves questions about special isolated points in the spectrum, previously identified by a mathematician named Ivanov. The authors prove rigorously that all of these points are genuinely isolated, meaning no other spectrum values cluster arbitrarily close to them. They also work out precisely how the gaps surrounding each of these special points behave as you move along the spectrum, and they fully characterize which real numbers share the same Dirichlet value at each of these isolated points. Taken together, these results give a clearer and more detailed picture of one of the more intricate objects in classical approximation theory.