The spectral approximation of multiplication operators via asymptotic (structured) linear algebra (Q884418): Difference between revisions

Suppose \(\phi\) is a bounded function defined on a compact set \(K\) of \(\mathbb R^d\), \(d\geq 1\). Consider the multiplication operator \(M[\phi]\) defined on \(L_w^2(K)\) by \(M[\phi]h=\phi h\), \(h\in L_w^2(K)\), where \(w\) is a suitable weight function. The problem under consideration is to know how well the spectrum of \(M[\phi]\) is approximated by the eigenspectrum of its finite section, namely the \(N\times N\) matrix \(M_N[\phi]:=(\langle M[\phi]e_j, e_i\rangle)\) based on an orthonormal basis \(\{e_j\}\) of \(L_w^2(K)\). Using the classical Fourier complex exponential basis, the problem is reduced to the Toeplitz case. The author discusses how the case of a general real orthonormal basis on \(K=[-1, 1]\) can be reduced to the Fourier case and how the resulting Toeplitz case can be reduced to the so called circulant case in which that transform is the fast Fourier transform, and thus, the general case can be translated into a problem of (asymptotic) structured numerical linear algebra for wich an accurate solution can be determined with a low computational cost, namely, with \(O(N \log(N))\) complexity where \(N\) being the number of sections.