Log-concavity of infinite product generating functions (Q2164877): Difference between revisions

A sequence of non-negative integers \((a_n)\) is said to be log-concave at \(n\) if \(a_n^2-a_{n-1}a_{n+1}\geq 0\). Define \(p_d(n)\) to be the coefficients of the generating function \[ \sum_{n\geq 0}p_d(n)q^n=\prod_{n\geq 1}(1-q^n)^{-n^{d-1}}. \] It is known by results of \textit{J.-L. Nicolas} [Ann. Inst. Fourier 28, No. 4, 1--16 (1978; Zbl 0363.10027)] and \textit{S. DeSalvo} and \textit{I. Pak} [Ramanujan J. 38, No. 1, 61--73 (2015; Zbl 1322.05018)] that the coefficients \(p_1(n)\) are log-concave for \(n>25\), and by recent results of \textit{K. Ono} et al. [Adv. Math. 409 B, Article ID 108692, 31 p. (2022; Zbl 1504.11110)] that the coefficients \(p_2(n)\) are log-concave for \(n>11\). The authors extend these studied cases to a general \(d\) and prove that \(p_d(n)\) is almost log-concave for \(3|n\) and almost strictly log-concave otherwise. The authors prove their results using two easy combinatorial lemmas and a formula for \(p_d(n)\) in terms of the sigma function \(\sigma_d(n)\) due to \textit{B. Kostant} [Invent. Math. 158, No. 1, 181--226 (2004; Zbl 1076.17002)]. The proofs are easy to follow and written out in details which makes the paper quite readable.