Smooth singular value decomposition on the symplectic group and Lyapunov exponents approximation (Q876738): Difference between revisions

Let \(X\in C^k\) be a matrix-valued function in the symplectic group \(\mathcal{S}\). The latter means that \(X^T(t)JX(t)=J=(J_{i,j})_{i,j=1,2}\), where \(J_{i,j}\in {\mathbb{R}}^{n\times n}\), \(J_{1,1}=J_{2,2}=0\), \(-J_{2,1}=J_{1,2}=I_n\). In addition, \(X\) is a fundamental matrix solution of a Hamiltonian system \(\dot{X}=C(t)X\), where \(C^T(t)J+JC(t)=0\). The main contribution of the paper consists in presenting a constructive argument to establish the existence of a smooth singular value decomposition (SVD) in \(\mathcal{S}\) for \(X\). The theory follows from the explicit structure of the polar factorization \(X=QP\) of \(X\), where \(Q\) is orthogonal and \(P\) is positive definite (\(P>0\)) and dichotomic (eigenvalues different from \(1\) for all \(t\)). It is shown how to construct the SVD of \(X\) from the eigendecomposition of \(P\). This constructive approach yields a new algorithm to compute the SVD of \(X\), which can be used to approximate the Lyapunov exponents of Hamiltonian systems. Some examples are given to illustrate the power of this method.