Derivation of the rules of quantum mechanics from information-theoretic axioms (Q421253)

The author proposes a set of reasonable axioms under which observed transition probabilities come from the standard Hilbert-space quantum model. Specifically, the author assumes that there is a set \(S\) of states, and that for each state \(s\in S\), we have a filter that transforms a general state \(s'\) into a state \(s\) with a certain probability \(s(s')\in [0,1]\); a good example is different polarizers. In quantum mechanics, each state \(s\) is described by a unit vector \(\hat s\) in a Hilbert space so that \(s(s')=|{\hat s}^*\cdot {\hat s}'|^2\). The axioms describe when the matrix \(s(s')\) of transition probabilities can be thus represented. Axiom~I states that the transition probability is symmetric (\(s(s')=s'(s)\)) and that \(s(s')=1\) if and only if \(s=s'\). Axiom II states, in effect, that the set \(S\) is compact (a reader should be warned that its formulation is not clear, but it is clarified in the following proposition). The most informative Axiom III states that for every \textit{frame} \(F\), i.e., for every maximal set whose elements are \textit{mutually orthogonal} (i.e., for which \(f(f')=0\) for all \(f\neq f'\) from \(F\)), and for every state \(s\), we have \(\sum_{f\in F} f(s)=1\). From the physical viewpoint, this axiom enables us to interpret the values \(f(s)\geq 0\) corresponding to different elements \(f\in F\) as the probabilities of observing \(f\) in the corresponding measurement process. Axioms I--III lead to the desired algebraic structure for \(s(s')\) in terms of the corresponding unit vectors, but this structure includes both the quantum case and the classical (hidden variables) case. To specify the quantum case, the author introduces an additional Axiom IV, according to which for every two states \(a,b\in S\), there exist mutually orthogonal states \(c\) and \(c'\) for which the function \(s\to (a(s)+b(s))/2\) can be represented as \(\lambda\cdot c(s)+(1-\lambda)\cdot c'(s)\) for some \(\lambda\in [0,1]\). The authors show that from the physical viewpoint, Axiom IV equivalent to a phenomenon first observed by von Neumann and Turing: that in quantum mechanics (in contrast to the hidden variable theories), the entropy of the measurement result can be decreased if we first perform an appropriate intermediate measurement. The proposed axiomatization is based solely on the observational data, and it works even in the 2-D case, when the usual axiomatizations of Hilbert space -- e.g., the ones based on quantum logic and Gleason's theorem -- do not work.