TITLE: Frequency of large values of the Mertens function
QUESTION [5 upvotes]: It is known that with $M(x) = \sum_{n\le x}\mu(n)$, there are infinitely many $x$ s.t. $|M(x)|\ge x^{\frac{1}{2} - \varepsilon}$ (see Chapter 15 of Montgomery-Vaughan, for example). Is there any way to make this result effective in the following sense: show that for large $X$, there exists some $x\in [X, f(X)]$ s.t. $|M(x)|\ge x^{\frac{1}{2} - \varepsilon}$ (or $x^{\sigma - \varepsilon}$ if $\zeta(\sigma + it) = 0$ for some $\sigma > 1/2$), where $f(X)$ is some explicit function in $X$? (Potentially naive) heuristics suggest one can take $f(X) = X + X^{2\sigma}$.

REPLY [3 votes]: It was proved by Kaczorowski and Pintz (Acta Math. Hungar. 48 (1986), 173-185, doi: 10.1007/BF01949062) that there exists $x\in[X,X^{1+o(1)}]$ such that $M(x)\geq x^{\sigma-\varepsilon}$, and there also exists $x\in[X,X^{1+o(1)}]$ such that $M(x)\leq -x^{\sigma-\varepsilon}$. See Corollary 1 in their paper.