No abstract

It has been known that positive definiteness does not guarantee a bisequence to be a complex moment. However, it turns out that positive definite extendibility does (Theorems 1 and 22), and this is the main theme of this paper. The main tool is, generally understood, polar decomposition. To strengthen applicability of our approach we work out a criterion for positive definite extendibility in a fairly wide context (Theorems 9 and 29). All this enables us to prove characterizations of subnormality of unbounded operators having invariant domain (Theorems 37 and 39) and their further applications (Theorems 41 and 43) and a description of the complex moment problem on real algebraic curves (Theorems 52 and 56). The latter question is completed in the Appendix, in which we relate the complex moment problem to the two-dimensional real one, with emphasis on real algebraic sets. Academic PressIt is well known that there is a relationship between moment problems and Hilbert space operators, one of the most beautiful examples of this kind. In particular there is a link between the complex moment problem and cyclic subnormal operators (cf. [16], for instance), where any progress in one side impacts the other side. While the bounded case is pretty well understood, in the unbounded one a search for satisfactory solutions is still required. In this paper we provide a solution of the complex moment problem and, in a parallel way, a characterization of unbounded subnormal operators which are not necessarily cyclic. Though the latter case includes the former, we have decided to separate them, thus giving the possibility of a choice to readers with particular interests.Let us be more precise: A bisequence of complex numbers indexed by integer lattice points of the first quarter of the plane may be positive or positive definite. While positivity is necessary and sufficient for the bisequence to be a complex moment one, it requires information on how non-negative polynomials in two real variables look like and this is not article no. FU983284 432

No abstract

Abstract. Let A be a, not necessarily closed, linear relation in a Hilbert space H with a multivalued part mul A. An operator B in H with ran B ⊥ mul A * * is said to be an operator part of A when A = B b + ({0} × mul A), where the sum is componentwise (i.e. span of the graphs). This decomposition provides a counterpart and an extension for the notion of closability of (unbounded) operators to the setting of linear relations. Existence and uniqueness criteria for the existence of an operator part are established via the so-called canonical decomposition of A. In addition, conditions are developed for the decomposition to be orthogonal (components defined in orthogonal subspaces of the underlying space). Such orthogonal decompositions are shown to be valid for several classes of relations. The relation A is said to have a Cartesian decomposition if A = U + i V , where U and V are symmetric relations and the sum is operatorwise. The connection between a Cartesian decomposition of A and the real and imaginary parts of A is investigated.