Integral representations for generalized difference kernels having a finite number of negative squares (Q1030500)

The main object under consideration in the present paper is a bounded operator on the Lebesgue space \(L_m^2(0,l)\) of \(m\)-dimensional vector-valued functions, given by \[ (Sf)(x)=\frac{d}{dx}\,\int_0^l\,s(x-t)f(t)\,dt. \] In this case, \(s(x-t)\) is called a generalized difference kernel. The operator \(S\) with an additional assumption \(S\geq 0\) has been studied in detail by the second author, and the integral representation for the kernel has been obtained. In the present paper, the authors adapt some ideas from the abstract interpolation problem to the indefinite case, that is, \(S\geq0\) is replaced by the requirement that the negative spectrum of \(S\) consists of finitely many eigenvalues of finite total multiplicity \(\kappa_S\). The main result concerns the integral representation of the generalized difference kernel in the indefinite case. As a consequence, the authors prove the extension theorem for the kernel to arbitrarily large intervals with a bound on the number of negative eigenvalues.