The Riemann zeta function is a central object in mathematics, and its nontrivial zeros are the subject of one of the most famous unsolved problems in all of science: the Riemann Hypothesis, which predicts that all such zeros lie on a specific vertical line in the complex plane called the critical line. Even without proving this, mathematicians can try to show that a large fraction of zeros behave as expected, either by lying on the critical line or by being "simple" (meaning each zero occurs just once, not as a repeated root). This paper pushes those fractions to new records: more than 83.9% of zeros are distinct, more than 67.35% are both simple and on the critical line, and related statistics are also improved beyond previous best results.
The technical approach builds on a tool called Montgomery's pair correlation theorem, which describes how zeros relate to each other in a statistical sense. The key idea is an energy accounting argument: each zero of higher multiplicity uses up a disproportionately large share of a fixed total energy budget, so if multiple zeros existed in abundance, they would consume more energy than is available. The authors combine this with a large sieve inequality (a technique for bounding how many things can simultaneously be large) and a careful computer-assisted analysis of how groups of seven or eight consecutive zeros interact. The correction terms in this analysis are organized to cancel out neatly through a telescoping structure, making the bounds tight enough to set new records.
A notable feature of the work is its use of formal verification: the proofs are checked in a computer proof assistant called Lean 4, meaning a program has verified the logical steps with near-complete rigor. The one part Lean does not independently verify is the output of certain search programs, which are taken on trust as recorded results. The authors also describe the project as an experiment in AI-assisted mathematical research, suggesting that tools like large language models or automated reasoning systems played a supporting role in developing or checking the arguments. Whether that assistance proves to be a model for future mathematics is left as an open question.