Distribution of neighboring values of the Liouville and M\"obius functions

Abstract

Let $\lambda(n)$ and $\mu(n)$ denote the Liouville function and the M\"obius function, respectively. In this study, relationships between the values of $\lambda(n)$ and $\lambda(n+h)$ up to $n\leq10^8$ for $1\leq h\leq1,000$ are explored. Chowla's conjecture predicts that the conditional expectation of $\lambda(n+h)$ given $\lambda(n)=1$ for $1\leq n\leq X$ converges to the conditional expectation of $\lambda(n+h)$ given $\lambda(n)=-1$ for $1\leq n\leq X$ as $X\rightarrow\infty$. However, for finite $X$, these conditional expectations are different. The observed difference, together with the significant difference in $\chi^2$ tests of independence, reveals hidden additive properties among the values of the Liouville function. Similarly, such additive structures for $\mu(n)$ for square-free $n$'s are identified. These findings pave the way for developing possible, and hopefully efficient, additive algorithms for these functions. The potential existence of fast, additive algorithms for $\lambda(n)$ and $\mu(n)$ may eventually provide scientific evidence supporting the belief that prime factorization of large integers should not be too difficult. For $1\leq h\leq1,000$, the study also tested the convergence speeds of Chowla's conjecture and found no relation on $h$.