The paper tackles a question about the prime factorization of whole numbers. Every positive integer can be broken down into prime factors, and researchers care especially about the largest and second-largest of these primes. For a given number n, call these P1(n) and P2(n). The question, posed by mathematicians Alladi and Johnson, was whether you could find a well-behaved function defined on the primes such that averaging it over P1(n) and P2(n) separately, across all integers up to some large bound, would give two different limiting values. In other words, could the largest and second-largest prime factors of a typical number behave differently in this averaged sense?
The paper proves the answer is no. If the average of your function evaluated at the largest prime factor converges to some value as you look at more and more integers, then the average evaluated at the second-largest prime factor must converge to exactly the same value. The two averages are "rigidly" linked. The proof works by reformulating both averages using a mathematical tool called the Dickman kernel, which is a well-known object in analytic number theory that describes how large prime factors are distributed among integers. Working on a logarithmic scale turns the averages into convolution integrals involving these kernels.
The key technical ingredient is a classical result from analysis called Wiener's Tauberian theorem, which roughly says that if a certain function has no zeros in its Fourier transform, then averaging operations involving it behave in a very controlled and predictable way. The authors show that the Dickman kernel relevant to the largest prime factor has this zero-free property in its Fourier transform. This guarantees that the weighted sums over primes converge in a suitable sense, and from there the convergence of the second average to the same limit follows as a consequence.