Eigenvalue Problem for Infinite Matrices and Its Applications

点击翻译按钮获取中文摘要

We consider an infinite complex symmetric (not necessarily Hermitian) tridiagonal matrix T whose diagonal elements diverge to * in modulus and whose off-diagonal elements are bounded. We regard T as a linear operator mapping maximal domain in the Hilbert space l^2 into l^2. Assuming the existence of T^<-1> we consider the problem of approximating a given simple eigenvalu lambda of T by an eigenvalue lambda_n of T_nthe n-th order principal submatrix of T.Let T=[x^<(1)>, x^<(2)>, ...]^T be an eigenvector corresponding to lambda. Assuming X^T X * 0 and f_<n+1>x^<(n+1)>/x^<(n)>*0 as n**, we will show that there exists a sequence {lambda_n} of T_n such that lambda-lambda_n=f_<n+1>^<(n)>x^<(n+1)>[1+o(1)]/(X^TX)*0, where f_<n+1> represents the (n, n+1) element of T.Application to the following problems is included : (a) solve J_<nu>(z)=0 for nu, given z * 0 and (b) compute the eigenvalues of the Mathieu equation. Fortunately, the existence of T^<-1> need not be verified for these examples since we may show that T+alphaI with alpha taken appropriately has an inverse.