Degree complexity of birational maps related to matrix inversion: symmetric case (Q415470)

Let \(\mathcal M_q\) be the space of \(q\times q\) complex matrices and \(K : \mathbb P (\mathcal M_q)\to \mathbb P (\mathcal M_q)\) the birational map defined by \(K= I\circ J\), where \(J((x_{i, j})_{1\leq i,j\leq q})= (x_{i,j}^{-1})_{1\leq i,j\leq q}\) and \(I\) is the map that sends a matrix to its inverse. The map \(K\) is of interest in statistical mechanics. The degree complexity of \(K\) is the exponential rate of growth of the degrees of the iterates of \(K\): \(\delta(K)= \lim_{n\to \infty} (\mathrm{deg} K^n)^{1/n}\). The main result of the paper is a determination of the degree complexity of \(K\) when restricted to the \(K\)-invariant subspace of symmetric matrices \(\mathcal S_q\). The author verifies a conjecture made in [\textit{J.-C. Anglès d'Auriac, J. M. Maillard, J.M.} and \textit{C. M. Viallet}, J. Phys. A, Math. Gen., No 39, 3641--3654 (2006; Zbl 1086.14504)] that predicts that the degree complexity \(\delta(K)\) equals to \(\delta(K| \mathcal S_q)\). Further, the author proves that \(\delta(K)\) is the largest modulus of the roots of the polynomial \(\lambda^2-(q^2-4q+2)\lambda+1\). The proof relies on blowing up suitable subvarieties to obtain a model \(Z\to \mathbb{P}\mathcal (S_q)\) such that \(\delta(K)\) is the spectral radius of the pullback of the lifted map \(K_Z^*: \mathrm{Pic } Z\to \mathrm{Pic } Z\). Unlike previous known cases, the proof does not rely on showing that \(K_Z\) is algebraically stable, but on a careful study of the action of iterates \(K_Z^n\) on exceptional surfaces.