A divisor set is a collection of positive integers that covers all integers from 1 to n in the sense that every integer in that range divides at least one number in the set. Researchers sometimes want these covering sets to be structured, and one natural structure is an arithmetic progression, meaning the numbers in the set are evenly spaced like 3, 7, 11, 15, and so on. Umans and Wang conjectured that you could build such a covering arithmetic progression where both the number of terms and the size of the terms stay relatively small, with specific bounds controlled by two parameters called alpha and beta.
This paper disproves that conjecture for a wide range of those parameters. The authors prove that any arithmetic progression that covers all integers up to n must either be very long or contain very large numbers. Specifically, if the numbers in the progression are not astronomically large, then the progression itself must have far more terms than Umans and Wang believed was necessary. The key forbidden point in their conjecture was when both alpha and beta equal one third, and the paper shows the conjecture fails there even if you allow a little extra slack in the bounds.
The proof works by focusing on a clever class of numbers called semiprimes, which are products of exactly two primes. By restricting attention to primes in a carefully chosen range near the square root of n, the authors translate the divisibility problem into a question about how points and lines meet in a finite geometric structure. A classical and elementary counting argument about such incidence structures then forces the arithmetic progression to be longer than the conjecture allows. Importantly, the result is specific to one-dimensional arithmetic progressions and does not rule out the broader, higher-dimensional version of the original conjecture.