A central limit theorem for integer partitions (Q5961997): Difference between revisions

Let \(\Lambda = (\Lambda_k)_{k\geq 1}\) be a non-decreasing sequence of positive integers. A \(\Lambda\)-partition of \(n\) is a representation of \(n\) of the form \(n = \sum_{j=1}^s \Lambda_{i_j} = n\) with \(i_1 < i_2 < \cdots < i_s\). The main result of this paper is a central limit theorem for the number of summands \(s\) of \(\Lambda\)-partitions of \(n\), provided that the Dirichlet series \(D(s) = \sum_{k\geq 1} \Lambda_k^{-s}\) satisfies certain analyticity properties. Usually it is assumed that \(D(s)\) has a simple pole at the abscissa of convergence \(\sigma_c\) and that \(D(s)(s-\sigma_c)\) can be continued analytically to \(\text{Re}(s) > \sigma_c - \eta\) (for some \(\eta> 0\)) -- this is the so-called Meinardus condition. In this case a central limit theorem was already known by \textit{H.-K. Hwang} [J. Comb. Theory, Ser. A 96, No. 1, 89--126 (2001; Zbl 1029.60013)]. The authors extend this property to situations where \(D(s)\) has infinitely poles \(s_m = \sigma_c + 2\pi i m \omega\), \(m\in \mathbb{Z}\), on the line \(\text{Re}(s) = \sigma_c\). This assumption is (usually) satisfied for certain \textit{automatic sequences} like missing digit sequences and gives rise to periodic terms in the asymptotic expansions. The proof is technical and quite involved. In particular it uses techniques from analytic number theory like the Mellin transform.