The paper tackles a problem in number theory involving special sums of fractions. The central objects are "central binomial coefficients," which are numbers that appear in the middle of Pascal's triangle, and "harmonic numbers," which are partial sums of the series 1 + 1/2 + 1/3 + 1/4 + so on. Mathematicians have long noticed that certain sums built from these objects, when evaluated at a prime number p, tend to satisfy surprising divisibility properties called congruences. These congruences say that two complicated expressions are equal when you only care about remainders after dividing by a power of a prime, much like how a clock keeps time modulo 12.
The specific result proven here settles a conjecture made by the prolific Chinese mathematician Zhi-Wei Sun. The conjecture concerns a sum where you add up terms of the form 1 divided by (k to the fourth power times the central binomial coefficient for 2k), as k runs from 1 up to p minus 1. Sun predicted that this sum, modulo p squared, equals a clean expression involving a harmonic number and a Bernoulli number (another classical sequence of numbers with deep connections to prime numbers). Proving this kind of statement is hard because it requires tracking extremely precise arithmetic behavior of these sums as the number of terms grows with the prime. As a bonus, the authors also confirm a related conjecture by two other researchers involving a more complicated sum with generalized harmonic numbers.
The proof brings together several distinct mathematical tools. The authors work in the world of "p-adic" arithmetic, where numbers are measured by how divisible they are by a prime rather than by their usual size. They combine known identities for finite sums, careful manipulation of harmonic number congruences, and a technique from complex analysis called the residue theorem, which is normally used to evaluate integrals but here serves to extract key coefficients from generating functions. The blend of number theory and complex analysis is what makes the approach notable.