New criteria to identify spectrum (Q2483778)

The components of a probability measure can be identified via a transform of the measure. Two of these are well-known, viz., the Fourier transform and the Borel transform. In this paper, the authors address the question of identifying the components using a more general transform. They give results using a general approximate identity, and an associated continuous wavelet transform. In fact, the authors give some new criteria for identifying the components of a probability measure in its Lebesgue decomposition. This gives new criteria to identify spectral types of self-adjoint operators on Hilbert spaces. The connection between an approximate identity and the continuous wavelet transform was discussed in [\textit{M. Holschneider}, ``Wavelets. An analysis tool'' (Oxford Mathematical Monographs) (The Clarendon Press/New York and Oxford University Press) (1995; Zbl 0874.42020)], while wavelet coefficients of fractal measures were studied in [\textit{R. S. Strichartz}, J. Geom. Anal. 1, No.~3, 269--289 (1991; Zbl 0734.28011)]. In the theory of self-adjoint operators, a finer decomposition of spectra with respect to Hausdorff measures was first used in [\textit{Y. Last}, J. Funct. Anal. 142, No.~2, 406--445 (1996; Zbl 0905.47059)] and general criteria for recovering a measure from its Borel transform were given in [\textit{B. Simon}, Proc. Am. Math. Soc. 123, No.~2, 3749--3755 (1995; Zbl 0853.28001)].