Influence-free states on compound quantum systems

TL;DR

This paper explores the structure of influence-free states in compound quantum systems, revealing their mathematical properties, limitations in combining states, and connections to Bayesian probability theory.

Contribution

It characterizes influence-free states using positive linear maps, analyzes extremal states, and links quantum influence-free states to Bayesian probability frameworks.

Findings

Extremal influence-free states are positive and PTP states.

Influence-free states cannot always be combined as tensor products.

A Bayesian interpretation of quantum state updates is established.

Abstract

Let Alice and Bob be able to make local quantum measurements and communicate classically. The set of mathematically consistent joint probability assignments (``states'') for such measurements is properly larger than the set of quantum-mechanical mixed states for the Alice-Bob system. It is canonically isomorphic to the set of positive (not necessarily completely positive) linear maps Phi from the bounded linear operators on Alice's space to those on Bob's, for which Tr Phi(I)=1. We review the fact that allowing classical communication is equivalent to enforcing ``no-instantaneous-signalling'' (``no--influence'') in the direction opposite the communication. We establish that in the subclass of ``decomposable'' states, i.e. convex combinations of positive states with "PTP" ones whose partial transpose is positive, the extremal states are just the extremal positive and extremal PTP states. We show that two such states, shared by the same pair of parties, cannot necessarily be combined as independent states (their tensor product) if the full set of quantum operations is allowed locally to each party. We use a framework of ``test spaces'' and states on these, suited for exhibiting the analogies and deviations of empirical probabilistic theories from classical probability theory. This leads to a deeper understanding of analogies between quantum mechanics and Bayesian probability theory. The existence of a ``most Bayesian'' quantum rule for updating states after measurement, and its association with the situation when information on one system is gained by measuring another, is a case of a general proposition holding for test spaces combined subject to the no-signalling requirement.