On the number of real eigenvalues of a product of truncated orthogonal random matrices (Q2076654)

Let \(O\) be a Haar-distributed real orthogonal matrix of size \((N+L) \times (N+L)\) and let \(\tilde{O}\) denote the upper-left \(N \times N\) submatrix of \(O\). Consider independent copies \(\tilde{O}_1, \tilde{O}_2,\dots, \tilde{O}_m\) of \(\tilde{O}\). In this article, the authors study the real eigenvalues of the product matrix \(X^{(m)}:=\tilde{O}_1 \tilde{O}_2 \cdots \tilde{O}_m\). Suppose that \(L=L_N\) is a sequence of positive integers such that \(L_N/N\) converges to a positive finite number as \(N \to \infty\). The main theorem states that for the total number \(N^{(m)}_\mathbb{R}\) of real eigenvalues of \(X^{(m)}\) one has \[ \mathbb{E} (N_\mathbb{R}^{(m)})= \sqrt{\frac{2mL_N}{\pi}} \operatorname{arctanh} \left(\frac{N}{N+L_N} \right) +O(1), \qquad N \to \infty. \] Furthermore, the limiting real eigenvalue distribution of \(X^{(m)}\) is determined. These results prove conjectures suggested by \textit{P. J. Forrester} et al. [Exp. Math. 29, No. 3, 276--290 (2020; Zbl 1466.15035)]. In their proof, they apply the fact that the real eigenvalues of \(X^{(m)}\) forms a Pfaffian point process.