``Extrinsic'' and ``intrinsic'' data in quantum measurements: asymptotic convex decomposition of positive operator valued measures (Q1424236)

Consider a positive operator valued measure (POVM) \(a\) with values in a finite-dimensional Hilbert space \(H\) and defined on a discrete finite set. The main result of the paper shows that the tensor power of \(a\) can be asymptotically decomposed into a convex mixture of POVM's having a small number of sequences on which they are supported. This number is estimated in terms of the quantum mutual information between a sender and a receiver. The asymptotic optimality of the decomposition is derived. The main theorem is then applied to the asymptotic decomposition of completely positive trace preserving map into maps of this kind which have Kraus decomposition with a small number of contributing operators. Further applications (simpler proof of the Holevo Bound Theorem, etc.) as well as open problems are widely discussed.