A note on the convexity of the sum of subpermanents (Q1923194)

Let \(\Omega_n\) denote the set of \(n \times n\) doubly stochastic matrices and \(\sigma_k(A)\) the sum of all subpermanents of order \(k\) of the matrix \(A\). The author proves that \(\sigma_2 (A)\) and \(\sigma_3 (A)\) are convex on \(\Omega_n\) for \(n \geq 2\) and \(n \geq 4\), respectively. In addition, as a generalization of \textit{E. T. H. Wang}'s conjecture [Linear Multilinear Algebra 5, 145-148 (1977; Zbl 0367.15007)], the author proposes the following conjecture: For every \(k \geq 3\) there exists \(n_k \geq k+1\) such that the inequality \[ \sigma_k \bigl(\alpha J_n + (1-\alpha) A\bigr) \leq\alpha \sigma_k (J_n) + (1-\alpha) \sigma_k \] holds for all \(\alpha \in [0,1]\) and all \(A\in \Omega_n\) with \(n \geq n_k\), where \(J_n= ({1 \over n})^n_{i,j=1} \in \Omega_n\). It is shown that this conjecture is true for \(k\geq 4\) with \(n_3 = 4\) and \(n_4 = 6\).