A ``supernormal'' partition statistic (Q2167492)

The authors study a statistic on integer partitions called the ``supernorm.'' Previously, the ``norm'' \(N(\lambda)\) of a partition \(\lambda\) was introduced by the third author, namely the product of all of its parts (with the empty partition having norm 1). Observe, then, that the set of partitions into primes has norms precisely covering the positive integers. Previous papers of the third author and collaborators have developed a theory in which numerous results of classical multiplicative nuumber theory are the special case of theorems about the partition supernorm, restricted to this set of partitions into primes. In this paper, the supernorm is defined instead, for the partition \(\lambda = \lambda_1 + \dots + \lambda_r\), by \(\widehat{N}(\lambda) = p_{\lambda_1} \dots p_{\lambda_r}\), with \(p_i\) denoting the \(i\)-th prime. The authors produce theorems relating the length, size, norm, and supernorm of partitions, producing for instance the inequalities \begin{align*} N(\lambda) &\leq \widehat{N}(\lambda) \leq N(\lambda)^{\frac{\log 3}{\log 2}} \\ p_n &\leq \widehat{N}(\lambda) \leq 2^n. \end{align*} The supernorm maps the monoid of all partitions with multiset union as the multiplication to the monoid \((\mathbb{N}, \cdot)\). The authors now explore the relation between partition theory with the supernorm and classical number theory, offering a ``dictionary'' of correspondences, e.g. partitions with no 1s correspond to integers with no 2s as prime factors, i.e. odd integers, and the length of \(\lambda\) corresponds to the number of prime factors of an integer. For example, denote by \(\mu(n)\) the classical Möbius function, \(p_{\min}(n)\) the smallest prime factor of an integer \(n\), and the index of the \(i\)-th prime \(p_i\) by \(\operatorname{idx}(p_i) = i\). Then the authors show: For \(d(S)\) the arithmetic density of a set \(S \subseteq \mathbb{N}\), \[ \sum_{\substack{n \geq 2 \\ \operatorname{idx}(p_{\min}(n)) \in S}} \frac{-\mu(n)}{n} = d(S). \] A more general version is given as Conjecture 4.1: for \(f(n)\) an arithmetic function, if the limit of the average value \(f_\infty = \lim_{N \rightarrow \infty} \sum_{k=1}^N f(k)\) exists, then does it hold that \[ \lim_{s \rightarrow 1^+} \frac{\sum_{p \in \mathbb{P}} f(\operatorname{idx}(p)) p^{-s}}{\sum_{p \in \mathbb{P}} p^{-s}} = f_\infty? \] To what extent this program can continue to generalize classical number theory, and what depth of new theorems can be unearthed from the attempt, remains to be seen. It would perhaps be most interesting if some open question in classical number theory could be resolved thanks to analysis from the more general viewpoint.