The paper investigates a classical question in number theory: how often do prime numbers appear in pairs or in evenly spaced triplets, when the gaps between them are constrained in a specific way? The constraint is that the difference between primes must be a number whose digits, written in some base, all come from a restricted set of allowed digits. For example, in base 10 you might only allow differences whose digits are drawn from the set containing 1, 2, and 5. The central goal is to count how many such prime pairs or arithmetic progressions exist and to find precise formulas describing their abundance.
To tackle this, the authors use tools from Fourier analysis, which is a way of breaking down complicated counting problems into oscillating wave-like components. They identify a key condition, called "digit non-resonance," which roughly ensures that the restricted digit set does not accidentally align in a way that would interfere with the distribution of primes. When this condition holds, they prove that the count of prime pairs and three-term progressions follows clean asymptotic formulas, meaning they can describe the long-run behavior of these counts very precisely, including the leading constant that governs the overall size.
The most technically demanding part of the work is actually verifying that the non-resonance condition is satisfied in practice for natural choices of digit sets. The authors handle this using a "transfer operator" approach, which is a technique borrowed from dynamical systems that tracks how information propagates through repeated digit-by-digit constructions. By carefully analyzing this operator, they confirm that the required condition holds across a range of explicit examples, making the abstract theoretical results concrete and applicable.