Expansion, divisibility and parity
Harald Andrés Helfgott, Maksym Radziwiłł
TL;DR
The paper develops a framework to study correlations of multiplicative functions, centering on a prime-divisibility graph and a local-expansion phenomenon. It proves a strong local Ramanujan-type bound for an adjacency-like operator A restricted to a large subset X, implying rapid mixing and cancellation for walks on the graph. This machinery yields improved results for parity-related averages of the Liouville function, including Chowla-type cancellations at almost all scales and refined short-interval correlations, extending beyond Tao’s entropy-based approach. The methods combine trace techniques, a sophisticated composite-modulus sieve via Rota’s cross-cut theorem, and a multi-dimensional Kubilius model, enabling simultaneous control over many primes and diverse divisibility constraints, with clear structural and graph-theoretic interpretations of the combinatorics of walks.
Abstract
Let $\mathbf{P} \subset [H_0,H]$ be a set of primes, where $\log H_0 \geq (\log H)^{2/3 + ε}$. Let $\mathscr{L} = \sum_{p \in \mathbf{P}} 1/p$. Let $N$ be such that $\log H \leq (\log N)^{1/2-ε}$. We show there exists a subset $\mathscr{X} \subset (N, 2N]$ of density close to $1$ such that all the eigenvalues of the linear operator $$(A_{|\mathscr{X}} f)(n) = \sum_{\substack{p \in \mathbf{P} : p | n \\ n, n \pm p \in \mathscr{X}}} f(n \pm p) \; - \sum_{\substack{p \in\mathbf{P} \\ n, n \pm p \in \mathscr{X}}} \frac{f(n \pm p)}{p}$$ are $O(\sqrt{\mathscr{L}})$. This bound is optimal up to a constant factor. In other words, we prove that a graph describing divisibility by primes is a strong local expander almost everywhere, and indeed within a constant factor of being "locally Ramanujan" (a.e.). Specializing to $f(n) = λ(n)$ with $λ(n)$ the Liouville function, and using an estimate by Matomäki, Radziwiłł and Tao on the average of $λ(n)$ in short intervals, we derive that \[\frac{1}{\log x} \sum_{n\leq x} \frac{λ(n) λ(n+1)}{n} = O\Big(\frac{1}{\sqrt{\log \log x}}\Big),\] improving on a result of Tao's. We also prove that $\sum_{N<n\leq 2 N} λ(n) λ(n+1)=o(N)$ at almost all scales with a similar error term, improving on a result by Tao and Teräväinen. (Tao and Tao-Teräväinen followed a different approach, based on entropy, not expansion; significantly, we can take a much larger value of $H$, and thus consider many more primes.) We can also prove sharper results with ease. For instance: let $S_{N,k}$ the set of all $N<n\leq 2N$ such that $Ω(n) = k$. Then, for any fixed value of $k$ with $k = \log \log N + O(\sqrt{\log \log N})$ (that is, any "popular" value of $k$) the average of $λ(n+1)$ over $S_{N,k}$ is $o(1)$ at almost all scales.