Matrix formulation for infinite-rank operators (Q2912205)

Let \(X\) be a Banach space of dimension \(d\leq \infty\). Then a linear operator \(S\) on \(X\) of finite rank \(n\) can be written as \(Sx=\sum_{j=1}^n f_j(x)x_j\) with \(x_j\in X\) and \(f_j\) linear functionals on \(X\). So we can think of it as a matrix \(S\in \mathbb{C}^{d\times d}\) that can be written as \(S=LK\) with analysis operator \(K\in\mathbb{C}^{n\times d}\) whose rows represent the \(f_j\) and synthesis operator \(L:\mathbb{C}^{d\times n}\to X\) whose columns represent the \(x_j\). This operator \(S\) on \(X\) can be associated with a matrix \(A\in\mathbb{C}^{n\times n}\) by \(A=KL\), hence satisfying \(AK=KS\). Let \(U\) and \(V\) be other operators on \(X\) of infinite rank that satisfy some invariance properties, namely \(CK=KU\) and \(LD=VL\) for some matrices \(C,D\in\mathbb{C}^{n\times n}\). This paper relates the spectral properties of the operator \(T=S+U\) (or \(T=S+V\)) on \(X\) and its \(n\)-dimensional counterpart, the matrix \(B=A+C\) (or \(B=A+D\)). This allows to approximate a large \(d\)-dimensional problem for a low rank \(n\) operator (\(S\)) with particular high rank perturbations (\(U\) or \(V\)) by a low \(n\)-dimensional matrix problem (\(B\)).