The structure of logarithmically averaged correlations of multiplicative functions, with applications to the Chowla and Elliott conjectures (Q2328043)

The Chowla conjecture concerning correlations of the Liouville function \(\lambda\) asserts that for any \(k\geq 0\) and any distinct integers \(h_0,\dots,h_k\), as \(x\to\infty\) one has \[ \mathbb{E}_{n\leq x}\lambda(n+h_0)\cdots\lambda(n+h_k)=o(1), \] where \[ \mathbb{E}_{n\in A}f(n)=\frac{1}{|A|}\,\sum_{n\in A}f(n). \] We adopt the convention that \(\lambda(n)=0\) when \(n\leq 0\). The paper under review is concerned with the structure of multiple correlations between bounded multiplicative functions, with application to the logarithmically averaged Chowla conjecture, for which the authors consider an averaged form in which the above unweighted averages are replaced by logarithmic averages \[ \mathbb{E}^{\log}_{a\in A} f(a)=\frac{\sum_{a\in A}\frac{f(a)}{a}}{\sum_{a\in A}\frac{1}{a}}. \] The logarithmically averaged Chowla conjecture then asserts that, for any \(k\geq 0\) and any distinct integers \(h_0,\dots,h_k\), as \(x\to\infty\) one has \[ \mathbb{E}_{n\leq x}^{\log}\lambda(n+h_0)\cdots\lambda(n+h_k)=o(1). \] Among several results toward to this conjecture, the authors prove it for any even \(k\geq 0\). Then, they consider more general multiplicative functions with application to a logarithmically averaged version of the so-called Elliott conjecture. The authors let \(g_0,\dots,g_k:\mathbb{N}\to\mathbb{D}\) be \(1\)-bounded multiplicative functions, and let \(h_0,\dots,h_k\in\mathbb{Z}\) be shifts, and consider correlation sequences \(f:\mathbb{N}\to\mathbb{Z}\) of the form \[ f(a):=\widetilde{\lim_{m\to\infty}}\mathbb{E}^{\log}_{\frac{x_m}{\omega_m}\leq n\leq x_m}\,g_0(n+ah_0)\cdots g_k(n+ah_k), \] where \(1\leq \omega_m\leq x_m\) are numbers going to infinity as \(m\to\infty\) and \(\widetilde{\lim}\) is a generalized limit functional extending the usual limit functional. In their main theorem, the authors show that \begin{itemize} \item[(i)] these sequences \(f\) are the uniform limit of periodic sequences \(f_i\), \item[(ii)] if \(g_0\cdots g_k\) does not weakly pretend to be any Dirichlet character, then \(f\) must vanish identically, \item[(iii)] if the multiplicative function \(g_0\cdots g_k\) ``weakly pretends'' to be a Dirichlet character \(\chi\), the periodic functions \(f_i\) can be chosen to be \(\chi\)-isotypic in the sense that \(f_i(ab)=f_i(a)\chi(b)\) whenever \(b\) is coprime to the periods of \(f_i\) and \(\chi\). \end{itemize} The authors apply their main theorem to obtain several new cases of the logarithmically averaged Elliott conjecture. Also, they give a number of applications of these special cases, including the conjectured logarithmic density of all sign patterns of the Liouville function of length up to three and of the Möbius function of length up to four. Among several tools for proving the above results, the authors apply the Furstenberg correspondence principle and the entropy decrement argument, and as a corollary of the countably additive probability theory interpretation, they interpret the correlation sequence \(f(a)\) in the language of ergodic theory. Finally, in the Appendix of the paper they generalize the main result to cover more general correlation sequences of the form \[ f(a):=\widetilde{\lim_{m\to\infty}}\mathbb{E}^{\log}_{\frac{x_m}{\omega_m}\leq n\leq x_m}\,g_0(q_0n+ah_0)\cdots g_k(q_kn+ah_k), \] for fixed \(q_0,\dots,q_k\).