The paper studies a classical problem in additive number theory: how often can a positive integer be written as the sum of a "structured" piece and a more "flexible" piece? The structured piece here is a product of fixed bases raised to varying powers, like 2 raised to some power times 3 raised to some other power. The flexible piece comes from a set of integers with known density properties, such as the prime numbers or numbers expressible as a sum of two squares. This setup generalizes a famous 1934 result by Romanoff, who showed that a positive proportion of all integers can be written as a prime plus a power of 2.
The main technical contribution is a lower bound on how many integers up to a large threshold admit many such representations. The count of representations depends on the number of bases used and on something called the "multiplicative rank," which roughly measures how independent the bases are from one another. For example, if you use bases 2 and 4, they are not multiplicatively independent because 4 is just a power of 2, and this redundancy affects the count. Crucially, the authors handle this situation without requiring the bases to be independent or coprime, which are common simplifying assumptions in earlier work.
When the flexible set is the primes, the authors show that a positive proportion of integers up to a large number have at least on the order of the logarithm raised to the power of one less than the number of bases in representations of the desired form. When the flexible set is sums of two squares, the exponent in the logarithm is slightly larger due to the higher density of that set. The results unify and extend previous work on Romanoff-type theorems and provide a flexible framework that accommodates many different choices of bases and additive sets simultaneously.