Further extensions of Hartfiel's determinant inequality to multiple matrices (Q825557)

As a refinement of a result by \textit{E. V. Haynsworth} [Proc. Am. Math. Soc. 24, 512--516 (1970; Zbl 0198.05202)], \textit{D. J. Hartfiel} [Proc. Am. Math. Soc. 41, 463--465 (1973; Zbl 0282.15011)] proved the following determinant inequality: Theorem. Suppose \(A, B\) are \(n\times n\) positive definite matrices. Let \(A_k\) and \(B_k\), \(k = 1, \ldots, n-1\), denote the \(k\)-th leading principal submatrices of \(A\) and \(B\), respectively. Then \[ \det(A+B) \ge\left(1+\sum_{k=1}^{n-1}\frac{\det B_k}{\det A_k}\right) \det A +\left(1+\sum_{k=1}^{n-1}\frac{\det A_k}{\det B_k}\right)\det B +(2^n-2n)\sqrt{\det AB}. \] Recently, some authors extended the above result to a larger class of matrices, namely the matrices whose numerical range is contained in a sector \[ S_\theta=\{z\in \mathbb{C}: \Re z>0, |\Im z| \le (\Re z) \tan \theta\}.\] The main result of this paper contains the following inequality. Theorem. Let \(A_j\) be \(n\times n\), \(j=1, \ldots, m\), such that \(W(A_j) \subset S_\theta\), and let \(A_{jk}\), \(k=1, \ldots, n-1\), denote the \(k\)-th leading principal submatrix of \(A_{j}\). Then it holds \[ \left|\det\left(\sum_{j=1}^mA_j\right)\right|\ge \sum_{1\le i<j\le m}\prod_{k=1}^n\left(\frac{|\det A_{ik}|}{|\det A_{i(k-1)}|}+\frac{|\det A_{jk}|}{|\det A_{j(k-1)}|}\right)\cos^k\theta -(m-2)\sum_{j=1}^m|\det A_j|, \] where by convention \(\det A_{j0}=1\) for all \(j\). This result generalizes Theorem 2.8 in [\textit{Y. Zheng} et al., Appl. Math. Comput. 369, Article ID 124827, 5 p. (2020; Zbl 1433.15024)].