Identification of minimum-phase-preserving operators on the half-line (Q2898436)

The authors address the inverse problem of identifying a minimum phase preserving operator from its output. It is motivated by considerations in seismic imaging, where seismic traces are widely assumed to be minimum phase functions, or translates of minimum phase functions. The results obtained here show that, if the minimum phase hypothesis is valid for seismic data, then current techniques such as the Wiener deconvolution are inadequate in the most general setting. A non-stationary operator representing the process of an input wavelet travelling through an interval of rock, if it preserves minimum phase, cannot be determined from a single test function. Therefor,e an input wavelet cannot be computed from a single seismic trace -- even with non-stationary methods; at least two traces, arising from two different source functions are required to determine the operator, and thus to invert it. This offers the exciting possibility that a re-evaluation of the seismic deconvolution techniques can lead to substantial improvement.NEWLINENEWLINEThe main results of this article solve the inverse problem concerning linear operators on the half line that preserve the class of minimum phase or translated minimum phase-signals. They show that every such operator is conjugate via the Fourier-Laplace transform to a product-composition operator on Hardy space, that precisely two test functions are required to reconstruct such an operator, and that there is a simple reconstruction formula in terms of the exponentially damped linear test functions \(e^{-t}(1 - t)\) and \(e^{-t}t\). The results obtained are heavily based on recent developments in the theory of stable polynomials.